A Tale of Two Tissues: AtGH9C1 Is an Endo-β-1,4-Glucanase Involved in Root Hair and Endosperm Development in Arabidopsis

Arabidopsis AtGH9C1 is an endo-β-1,4-glucanase possessing a carbohydrate-binding domain (CBM49). Analysis of AtGH9C1 expression by promoter-reporter GUS, RT-PCR, public transcriptome databases and GFP protein tagging demonstrated a high and selective expression of AtGH9C1 in root hairs and in the endosperm. Expression in root hair cells started prior to bulge formation and continued during hair elongation. AtGH9C1 expression increased with treatments that increase density (ACC) or length (sucrose) of root hairs. Expression in the endosperm extended sequentially to the micropylar, peripheral and chalazal compartments. A mutant with reduced AtGH9C1 expression had a delay in germination and a marked reduction in root hair presence. Complementation of the mutant partially improved both germination and root hair density. Experiments with ectopically expressed AtGH9C1-GFP with and without the CBM49, demonstrated that both forms of the protein are secreted and that CBM49 targets the protein to specific regions of the cell wall, but what makes these regions special is still unknown. The amino acid alignment of angiosperm GH9 genes with C-terminal extensions illustrate that AtGH9C1 belongs to a different clade than its tomato homolog, S1GH9C1. The latter has a CBM49 that was shown to bind crystalline cellulose. We suggest that AtGH9C1 is associated with the weakening of the cell wall during formation and growth of the root hair as well as with the sequential anterior-posterior breakdown of the endosperm cell wall that provides space for the growing embryo. Thus, is likely that the CBM49 of AtGH9C1 recognizes a form of cellulose or glucan polymer that is prevalent in the wall of these specialized tissues and that is different than the one recognized by S1GH9C1.


Introduction
The role of hydrolytic (b-1, 4) glucanases in plant cell wall loosening is not clear albeit these enzymes were formerly envisioned as the prime movers in the wall expansion process [1,2]. All plants have genes that encode endo (b-1, 4) -glucanases, EC 3.2.1.4, that catalyze the breaking of b-1, 4 glycosidic bonds in the interior of longer b-1, 4 glucose chains. While it is not known if they can break the glycosidic bonds of the crystalline cellulose matrix [3,4], it is recognized that this activity could contribute to the loosening of the cell wall to make it more pliable for expansion, [3,5]. The family of (b-1, 4) glucanases belongs to the Glycosyl Hydrolase family 9 or GH9 (CAZy; http://www.cazy.org). This family is a member of the clan CL0059 (http://pfam.sanger.ac. uk/clan/CL0059), enzymes that have a three dimensional structure composed of 6 helical hairpins.
The GH9 family in plants is extended, divergent and is subdivided into classes A, B and C, all having the conserved catalytic module (,450AA) composed of two signatures that distinguish the GH9 endoglucanases from other hydrolases [6,7]. The GH9 class C is of special interest because it comprises putative secreted proteins of ,600AA containing an amino acid extension at the C-terminus (,100 -amino acids long) that constitute a carbohydrate binding module (CBM) [8]. A CBM is a short amino-acid sequence that folds into a discrete threedimensional structure forming a putative carbohydrate binding cleft. Most prokaryote enzymes that hydrolyze insoluble polysaccharides (i.e. cellulases, amylases, and chitinases) have a CBM in addition to a catalytic module to facilitate hydrolysis. The CBM in bacteria cellulases is important [9], because it helps position the enzyme's catalytic site on the microfibrils for efficient breakage of the endoglycosidic bonds [10,11]. In some cases, associating a different CBM to the same catalytic module conferred binding specificity to different cellulose forms (amorphous vs. crystalline), [12]. Crystalline cellulose is found at the core of all cellulose microfibrils, but at the surface of the microfibril there are different percentage of amorphous cellulose depending on the size of the microfibrils [13]. The thicker microfibrils found in secondary walls have lower surface to volume ratio and only 6,7% amorphous cellulose, while the smaller microfibrils found in primary cell wall have 33-80% amorphous cellulose [13,14].
The CBM of plant GH9 class C constitutes a distinctive family, which has been given the designation of CBM49 (CAZy). The CBM49 of S1GH9C1, (SlCel9C1), an endo (b-1, 4) glucanase from Solanum lycopersicum, was characterized first and shown to bind crystalline cellulose in vitro [15]. Thus, it was suggested that S1GH9C1, might facilitate cellulose degradation during tomato ripening [15]. However, the physiological role of S1GH9C1, and for that matter the role of all plant (GH9) class C enzymes, remains unknown.
We selected A. thaliana to analyze the role of one GH9 gene containing a CBM49 referred to as AtGH9C1. This gene is one of three GH9 class C enzymes in Arabidopsis and its role is more intriguing because it is the most dissimilar of the three [8]. The other two, AtGH9C2 and AtGH9C3 have high sequence identity and constitute a duplicated gene pair positioned within duplicated DNA segments of the genome [16,17,18].
Here we analyze the spatial-temporal distribution of AtGH9C1 message, determine the localization of AtGH9C1-GFP-fusion protein with and without the CBM49, and analyze a mutant with reduced AtGH9C1 expression. Our results demonstrate that AtGH9C1 is expressed primarily in root hair and endosperm cells and agree with results from public transcriptomes from these tissues. We established that the AtGH9C1 is a secreted protein and that the CBM confers localization specificity within the cell wall. A dendogram of all known GH9 class C genes in angiosperms demonstrates that AtGH9C1 and S1GH9C1 belong to different clades. Other genes that cluster with AtGH9C1 are also expressed preferentially in root hairs. The walls of hair cells, like the walls of the cellularized endosperm, are special, unlignified and short-lived. We suggest that AtGH9C1, and other members of the same clade, hydrolyze b-1, 4 glucan chains in a cell wall context that is different than the one recognized by S1GH9C1, and that AtGH9C1 plays a role in the development of the root hair and of the endosperm in Arabidopsis.

AtGH9C1 is Expressed in Roots and Seeds of Arabidopsis
AtGH9C1 is located on the upper arm of chromosome 1, (At1g48930) and encodes for a polypeptide 627 amino acids long, (69 kD, pI 9.18), that is moderately glycosylated. This gene is composed of 8 exons, of which the first one carries a predicted signal peptide [e.g. 19] and the last one encodes the entire CBM49 that is tethered to the catalytic core by a short polypeptide linker encoded in part by exon 7. The remaining exons encode the catalytic module.
To gain a first insight into where AtGH9C1 is expressed, RT-PCR was performed using a PCR primer pair based on the 6th and last AtGH9C1 exons and RNA templates from major Arabidopsis plant organs. Results indicated that AtGH9C1 is expressed abundantly in young siliques, and in roots (Fig. 1A). A band of low intensity was also observed in young shoots but bands were undetected in flowers, leaves and stem. Comparison of RT-PCR band intensity at 25 PCR cycles, (Fig.1B), also revealed that expression was absent in flowers, low in very young siliques (S1, ,2 mm long), peaks at S3, (S3,,5 mm long), and was absent at S5 as siliques reach their maximal length (.1 cm long).

Spatial-temporal Expression of AtGH9C1
To examine expression of AtGH9C1 with spatio-temporal resolution, its putative promoter (sequence of 1,600 bp preceding AtGH9C1 coding region) was fused to the GUS reporter gene (Escherichia coli ß-glucuronidase) to generate AtGH9C1::GUS and used to transform A. thaliana Columbia, Lansberg and Shahadara ecotypes. Independent transgenic lines from each ecotype showed GUS activity in roots and young siliques, (Fig.2). Shahadara transgenic seeds were selected to monitor GUS staining at daily intervals after cold stratification, and transfer to continuous light. GUS activity was negative at 16-24 hours, when some seeds had a ruptured testa (seed coat) and some had an emerging radicle, ( Fig. 2a and b). Between 30-72 hours, when elongation of the radicle continues, GUS activity was strong in all the epidermal cells of the collet, i.e., the root-shoot junction, (Fig. 2c and 3d). After 5 days, when cotyledons have emerged and the primary root is rapidly elongating, the strong GUS activity continued in the collet and extended to the differentiation zone of the main root (Fig. 2e). The blue staining was very strong only in the files of hairforming cells (trichoblast), indicating low or no GUS activity in the atrichoblasts. Note that GUS staining was detected in the hair cell before the bulging and emergence of the root hair (Fig. 2f), and that the hair emerged closer to the bottom cross wall of the trichoblast (Fig. 2g), as described by others. Seedlings growing in media supplemented with 1% sucrose have longer root hairs and strong GUS staining in the hairs (Fig 2h). GUS activity was absent from the epidermal cells of the root cap and the meristematic region of the root tip.
GUS activity was also detected within the seeds of young siliques (silique # 2-3 from the top), in the compartments of the seed endosperm, particularly at the micropylar endosperm, (MCE), and the peripheral endosperm (PEN), (Fig. 2i), while older siliques displayed no stain. Note that even the embryo proper shows a weak blue coloration (Fig. 2j), revealed through the seed coat.
To substantiate our data, we searched Genevestigator, https:// www.genevestigator.com for AtGH9C1 gene expression in publically available transcriptomes. We confirmed that the AtGH9C1 message was detected most abundantly at the root zone where the root hairs develop. We also confirmed the expression of AtGH9C1 in developing seeds noticing that AtGH9C1 message is present in the three domains of the seed endosperm i.e. the MCE, the PEN and the chalazal (CZE).

AtGH9C1 Expression is Comparable to Root Hair Marker Genes
Next, we expanded the Genevestigator search to compare AtGH9C1 expression with cell-wall genes known to regulate root hair morphogenesis. A heat map was developed with this public data ( Fig 3A) to illustrate that the abundance of AtGH9C1 message, Detection of AtGH9C1 and Actin 11 messages by RT-PCR, using as template total RNA isolated from flowers (F), young siliques (S1 and S3), roots (R), shoots (Sh), expanded leaves (L) and stem (St); (B) RT-PCR comparing flower (F) with three stages of silique development (S1, ,2 mm long), S3, (,5mm long), S5, siliques reaching their maximal length (.1 cm long). doi:10.1371/journal.pone.0049363.g001 ATGH9C1 in Root Hair and Endosperm Development PLOS ONE | www.plosone.org (column 1) in the root hair zone of the main and the lateral root, is comparable to that of root-hair marker genes, including endoxyloglucan transglycosylases AtXTH14/XTR9, [4,20], (column 2), expansins, AtEXP7 [21] (column 3), AtEXP18, (column 4), structural cell wall proteins, LRX1 [22], (column 5) and RHS12, a pectin esterase, (column 6). Similarly, and comparable to those root hair marker genes, AtGH9C1 expression is lower in the stem, in adult leaves and in other root cell types including the endodermis, the cortex, atrichoblast and the stele.

AtGH9C1 Expression in the Endosperm Follows an Anterior-posterior Polar Direction
We also searched the publically available transcriptomes of Arabidopsis endosperm before [23], and after cellularization (http://estdb.biology.ucla.edu/genechip. AtGH9C1 expression was found in the proliferating syncytial endosperm, in the group of genes expressed preferentially in the endosperm, (''endosperm preferred'', EP, with a.2-fold differential expression in the endosperm sample compared with the silique sample) [23]. A heat map (Fig 3B) illustrates the progression of AtGH9C1 expression level in the three domains of the endosperm and in the embryo proper through 5 stages of seed development (from preglobular to mature green embryo stage). AtGH9C1 message is present in the MCE at low level at the globular developmental stage, and persists in the MCE at the heart developmental stage. AtGH9C1 expression then extends into the PEN at the linear cotyledon stage, reaching here the highest level of message accumulation, and finally expands into the CZE at the green cotyledon stage. Thus, AtGH9C1 expression seems to follow the sequential polar antero-posterior organization of endosperm cellularization and endosperm dissolution. There is also expression of AtGH9C1 in the embryo proper at the heart and linear cotyledon stages [24]. Interestingly, and consistent with our analysis of GUS transgenic seeds during imbibition and germination, the transcriptomes data of seed germination indicated that AtGH9C1 message level is very low and unchanged all through seed imbibition, rupture of the endosperm, breakage of the testa, and radicle emergence ( [25] Supplemental Dataset 1).

AtGH9C1 Expression Responds Like a Root Hair Marker Gene
If AtGH9C1 expression is localized at the root hairs, GUS activity in the whole root of the transgenic lines should quantitatively correlate with root length, since longer roots will have more root hairs. Thus, seedlings of the transgenic line were exposed to treatments that generate either shorter roots (increasing concentration of mannitol) or longer roots (1% or 2% sucrose, [26]), over an equal growth period (7 days post illumination). The correlation between root length and GUS activity was confirmed in the mannitol treatment where both measurements were negatively impacted by the increasing osmotic stress imposed by mannitol ( Fig 4A). Moreover, each of the measurements was reduced in about equal proportion, e.g. a linear regression between the two variables extrapolated to around zero. However, when transgenic seeds were germinated and maintained in control media supplemented with 1% or 2% sucrose GUS activity increased to a much greater degree than root length (Fig. 4A). In addition, seedlings grown in sucrose concentration had longer root hairs compared with seedlings in control or mannitol media (Fig. 4B). Thus, assuming that those treatments do not affect the number of root hair forming cells, the data suggests that AtGH9C1 contributes not only to the formation of the root hair but also to the lengthening of the hair.
Another treatment known to affect root growth and root hair density is the plant hormone ethylene [27]. Thus, seeds of transgenic AtGH9C1::GUS (Columbia) were germinated in agar plates supplemented with 10 mM 1-aminocyclopropane-1-carboxylate (ACC), a substrate for ethylene synthesis, in both light and dark conditions. At 72 h post-germination in continuous light, seedling roots displayed an increase in root hair abundance as expected, and a strong blue coloration throughout the root (Fig. 5A). At 48 h post-germination in darkness, the transgenic seedling showed the triple response and a strong blue coloration in the root (5B-right). This contrasted with the very weak blue coloration observed in the root of the etiolated ethylene insensitive mutant (ers-2), also carrying the same AtGH9C1::GUS construct and germinated in darkness under the same conditions ( Fig. 5Bleft).
To further analyze the regulation of AtGH9C1 expression by ethylene, transgenic AtGH9C1::GUS seeds were germinated in agar media supplemented with 10 mM ACC and kept in darkness for 3 days. After the treatment the whole seedling was used for GUS activity determination. Results showed a 5-fold increase in GUS activity in etiolated seedlings grown in the presence of ACC compared to those grown in control media ( Table 1). The effect of 20 mM AVG, an inhibitor of ethylene synthesis, and the effect of both AVG and ACC, to test for inhibition reversal, was also analyzed in 3 day old transgenic seedlings, kept in continuous light. The activity from seedlings in AVG was around half the activity in ACC. The activity of seedlings exposed to both chemicals was less than the activity in presence of ACC alone. This indicated that simply adding ACC could not fully reverse the inhibition caused by AVG, as noted by others, [21]. Finally, the effect of ACC was dose dependent as shown by the analysis of GUS activity in 7 day old seedlings exposed to 0.1 mM, 1 mM or 10 mM ACC and kept in continuous light. Both shoot and root were analyzed separately and both showed a concomitant increase in GUS activity compared to seedlings grown on control plates (no ACC) and the activity was always greater in the root than in the shoot tissue (Table 1). Overall, these results indicate that AtGH9C1 expression is regulated by ethylene, which increases root hair density (Fig 5A), [28].    The expression of AtGH9C1 and of the same set of root-hair marker genes mentioned above was searched in public root transcriptomes that used treatments or genetic backgrounds that promote or reduce root hair development. A heat map ( Figure S1), shows that different treatments that promote root hair growth over control treatment, such as iron [29], phosphorous [30,31] and water deficiency, changed the expression of AtGH9C1and of the marker genes in similar ways. Particularly, all increased around two fold with phosphorous and water deficiency (red panels). In contrast, in mutant lines, that develop shorter root hairs, like rsl4 [32] and rhd2-2 [33], the expression of these genes decreased compared to wild type controls (green panels). Moreover, in a line that over expressed RSL4 (35S::RSL4) [32], and resulted in longer root hairs, AtGH9C1 and the marker genes increased gene expression (GEO accession number GSE19530).

Cell Wall Localization of AtGH9C1-GFP with and without CBM49 before and after Plasmolysis
The deduced amino acid sequence of AtGH9C1 predicted the presence of a signal peptide at the N-terminus suggesting that the nascent polypeptide is imported into the endoplasmic reticulum and then secreted outside the cell. To elucidate experimentally the cellular localization of this protein and the role of the CBM in localization we generated fusions between the GFP and AtGH9C1 (with and without the CBM). The fusion protein was expressed using the 2X 35S promoter to guarantee a strong expression. Both, C-terminal AtGH9C1-GFP:6CBM, (Fig.6) and N-terminal (GFP-AtGH9C1:6CBM) ( Figure S2) fusions were generated to test for possible interference from the GFP. Each construct was infiltrated into N. benthamiana leaves to induce transient expression and then the localization of the fluorescence protein (with and without the CBM) was analyzed before and after plasmolysis by confocal laser microscopy.
There was a clear difference in the distribution of fluorescence from AtGH9C1-GFP fusion with the CBM (Fig.6A) vs. without the CBM (Fig 6B). With the CBM, the fluorescence contours only some segments of the cell wall surface (Fig.6A), The discrete fluorescence segments were positioned in some neck and lobe regions of the epidermal cells and the intensity of the green fluorescence was high. In contrast without the CBM the fluorescence was diffused and dispersed throughout the epidermal cell wall (Fig. 6B).
When N. benthamiana leaves were infiltrated as above and subsequently incubated in 1M sucrose (instead of water), the localization of fluorescence from AtGH9C1-GFP with the CBM showed no signs of alteration despite the retraction of the plasma membrane (Fig. 6C). In contrast, the localization of fluorescence from AtGH9C1-GFP lacking the CBM became diffused accumulating in extracellular regions as blotches of random sizes (Fig. 6D). No fluorescence was observed in the cytosol in either case. This contrasted with the distribution of fluorescence from the 35S:GFP alone, (Fig. 6E) which appears inside and outlining each cell very strongly. Thus, secretion AtGH9C1 to the cell wall was confirmed for both AtGH9C1-GFP with and without the CBM and the localization was altered when lacking the CBM.
Although the fluorescence was weaker, the same pattern of localization was observed for the N-terminal fusion with and without the CBM, where the GFP was inserted between the signal peptide and the rest of AtGH9C1 ( Figure S2). This infers that GFP does not interfere with the secretion of AtGH9C1 in either the Cor N-terminal fusions. Because of its more intense fluorescence, we used the C-terminal (AtGH9C1-GFP: 6CBM) fusions for further studies.

Stable Over-expression Expression of AtGH9C1-GFP with and without CBM and Ectopic Localization AtGH9C1-GFP Protein in Arabidopsis Thaliana
To analyze the possible consequences of over expressing AtGH9C1-GFP with and without the CBM, we transformed wild type Arabidopsis with the C-terminal (AtGH9C1-GFP: 6CBM) fusion constructs. Although GFP was observed in the transformed lines, there were no obvious phenotypic difference among the over-expressor lines (T3 generation) and wild type. The studies included analysis of germination efficiency, and growth of root and hypocotyl, at 5 day-post germination (in light or continuous dark). Nevertheless, the images of GFP fluorescence of these lines illustrated again a different spatial distribution between AtGH9C1-GFP with and without the CBM ( Figure S3). Most noticeable was the accumulation of GFP fluorescence in association with the stomata cavity only if the CBM is present.

AtGH9C1-T-DNA Insertion Mutants: Phenotypic Characterization
To gain further insight into AtGH9C1 gene function, we obtained from the SALK collection three lines with a T-DNA insertion in AtGH9C1 gene [34]. The SALK 006684 and SALK 006767 lines were indicated to have a T-DNA insertion at intron number two, while the line SALK 051517 in exon number eight (Fig.7A). Using gene-specific primers and a T-DNA left border primer, a homozygous mutant for each line was identified and the position of insertion confirmed by sequencing the DNA segment amplified from the T-DNA left border primer. Our sequence determined that the T-DNA insertion is intronic in SALK 006684 and SALK 006767, but indicated that in the SALK 051517 line the insertion was also intronic and toward the end of intron 6.
Expression of AtGH9C1 in SALK 006684 and 051517 was further analyzed by RT-PCR using total RNA from green siliques and a primer pair based on exon seven (Fw) and on the 39UTR (Rv). These primers amplify a 500 bp segment from the In general, growing in soil, both mutant lines germinated, produced normal plants and abundant seeds, yet SALK 006684 plants consistently displayed a reduction in rosette size compared to wild type and SALK 051517 line (Fig.7C). Therefore, growth of SALK 006684 was further characterized on 1/2MS agar plates always comparing with wild type. Seeds of SALK 006684 showed a delay in germination of 12h-24h, and the delay became even more evident (36h-48) when the agar media was supplemented with hyperosmotic 4% Mannitol, (Fig. 8A). SALK 006684 and wild type showed no apparent differences in root hair abundance or root hair length at the collet or at the main root, 72 h post germination. However, once the root began to grow down the agar plate, the main root of the mutant showed a noticeably lower root hair abundance compared to wild type (Fig 8B-C). The average root hair abundance for wild type, few millimeters above the zone of differentiation was , 9 hairs/mm while for the mutants was , 0.2 hairs/mm (Fig. 9). This suggests that the root hair phenotype becomes visible when a low activity threshold for AtGH9C1 is reached. Other studies comparing wild type with SALK 006684 included a comparison of root length growth in control media, over a 7-day period post-light incubation, ( Figure S4A). The time course demonstrated consistently that the wild type roots were longer than those of the mutant, and that the absolute differences between them remain about the same over the 7-day period of growth. Therefore, the difference in root length observed was a reflection of the delay in germination and not due to a reduced rate of root growth. A comparison in a media supplemented with sucrose to stimulate root hair growth ( Figure  S4B) demonstrated clearly the root hair abundance difference between the two lines. Note that the mature SALK006684 plants ( Figure S4C, left) have small stature compared to wild type ( Figure S4C, right), and that some siliques from the main stem were shorter and defective ( Figure S4D). Results of equivalent experiments performed with SALK 051517 had similar tendencies but differences were smaller.
Finally, a feature common to both mutant lines was that only 1% of the seeds germinated compared to 50% of wild type seeds after four years of storage in a dry cabinet. These results suggested that both mutants' seeds lost viability faster than those of the wild type and that AtGH9C1 expression in the endosperm may contribute to seed longevity during storage.

Mutant Complementation Test
Subsequently, we complemented the mutants with the wild type gene to examine whether the reduction of AtGH9C1 expression in the mutants contributed to the delay in seed germination and to the reduction in root hair abundance and length. Complementation was accomplished by transformation of the homozygous SALK 006684 and SALK 051517 lines with a construct carrying  the full-length AtGH9C1-GFP driven by the double 35S promoter (referred to as Compl-SALK 006684 and Compl-SALK 051517). The selected complemented lines used for phenotypic characterization were hygromycin resistant, GFP positive, and either homozygous or heterozygous dominant because gene expression under the 2X 35S promoter confers a dominant character [35]. Three phenotypic characters, % germination and root hair abundance and length were analyzed. In each study, the comparison included the wild type, the mutant and the complemented mutant.
Two independent trials were conducted to measure percent (%) germination of seeds at 24 h post-light incubation, (Table 2). In both trials, the % germination of SALK 006684 improved significantly with complementation, (Tukey HSD test, p = 0.0520.01, respectively), but only in trial two did reach the % germination level of the wild type. For SALK 051517, the % germination was not different compared to the wild type, and although the % germination of Compl-SALK 051517 was slightly higher than the wild type, the difference was not statistically significant ( Table 2). We concluded that the complemented lines were only partially rescued by the fusion construct with the GFP and that the fusion enzyme may have reduced enzyme activity. Nevertheless, they also supported the hypothesis that AtGH9C1 plays a role preparing the seed for germination during the period of high expression in the endosperm.
Analysis of root hair abundance and length, were conducted in 7 day-old seedlings growing on control agar plates (Fig. 9). While we could find ''patches'' with hairs on the roots Compl-SALK 006684 (Fig. 9b Inset), it was not easy to verify the increase by visual observation alone. Therefore, we systematically counted the root hairs on pictures of consecutive mm of root segments (see details in Methods) to quantify an average density. Density was quite variable between roots and within a root, with a range (min to max) of 7-11.6 hairs mm 21 for WT and 0.3-2.8 hairs mm 21 for Compl-SALK 006684 (Fig. 9). Nevertheless, the complemented mutant had a significantly greater average root hair density than the mutant (t-test, n = 14, p = 0.013). As mentioned above, complementation did not totally rescue the WT phenotype, in terms of root hair abundance and length (Fig. 9c inset). The average (6SD) root hair length for wild type was 0.4860.22 mm while for the Compl-SALK 006684 was 0.2160.1 mm.

Phylogeny of CBM49 from Higher Plants and Lower Photosynthetic Species
All sequenced genomes of higher plants and early photosynthetic species such as Physcomitrella, Selaginella and the unicellular algae Chlamydomonas contain GH9 genes coding for endo-b-1,4glucanase possessing a C-terminal extension. The specific protein sequence accessions numbers and locus identification for each gene are presented in Table S1. The phylogenetic tree, using Chlamydomonas as a root, shows that the proteins from lower photosynthetic species form a separate clade (Clade 3) that is distinct from the two other clades (Clade 1 and 2) formed by genes from higher plants (Fig. 10). Each clade from higher plants includes members of the monocot and dicot plants. S1GH9C1, the tomato GH9 with a CBM49 at the C-terminal extension that was shown to bind crystalline cellulose [15] is placed in Clade 2. The Arabidopsis AtGH9C1 and its paralogue in A. lyrata also possessing a CBM49 at the C-terminal extension are both placed in Clade 1. Instead, AtGH9C2 and AtGH9C3, and corresponding paralogues of A. lyrata are all placed in Clade 2, of which several are expressed in fruit [15,36,37].
It's apparent then, that GH9 enzymes with C-terminal extensions from Clade 1 are distinct from those in Clade 2 and this distinction (based on percent amino acid identity) may represent differences in protein structure and possibly functional differences. The difference in amino acid composition between members of Clade 1 and Clade 2 is illustrated in the amino acid alignment of dicot GH9 with C-terminal extensions ( Figure S5). The C-terminal extension of members of Clade1 are enriched in glutamine (,9%) while, members of Clade 2 are enriched in serine (,20%).

Expression Profile of Other GH9C-type Enzymes from Clade 1
To determine if other enzymes from Clade 1 have same functions as AtGH9C1 we searched for their expression profile in public microarray databases. The Medicago truncatula Gene Expression Atlas indicates that MtrGH9C1 of Clade 1 (http:// mtgea.noble.org/v2/probeset.php?id = Mtr.39531.1. S1_at&submit = Go) is exclusively expressed in root hair cells. Also, the expression of the two soybean members of Clade1 (Glyma11g10760 and Glyma12g03050) had been analyzed along segments of the main root by quantitative RT-PCR using primers that amplify both genes. The highest expression of these genes occurs in the root segment 2-7 mm behind the root meristem where cell elongation occurs and root hair initiation begins. At each side of this segment the expression is lower and is undetected in leaves (Dr. Tucker, unpublished results). Unfortunately, probes for these two soybean genes are not included in the GeneChip microarray of microdissected soybean seeds (http:// seedgenenetwork.net/). Efforts to have global transcriptome profiling of every tissue in Poplar (http://www.popgenie.org) and Sorghum [38] are in progress. From the data already available, we found that the corresponding GH9 members of Clade 1 from these two plants (POPTR_0007s07790.1 and Sb06g032760) are both preferentially expressed in the root over shoots with some expression in young leaves and nodes in Poplar.

Discussion
AtGH9C1 is one of the three endo (b-1, 4) glucanase (GH9) genes of A. thaliana possessing both a catalytic module and a carbohydrate-binding module, family 49 (CBM49), at the Cterminus. Here we demonstrated that AtGH9C1 is expressed primarily in root hairs and endosperm cells. Both cell types have

A Role for ATGHC1 in Root Hair Development
The tubular outgrowth of root hairs from trichoblasts involves a polarized secretion of enzymes and addition of new cell wall polysaccharides at the growing apex [39,40]. These molecules bring about the concurrent localized loosening and reinforcement of the root hair cell wall [41]. Several proteins contribute to cell wall loosening in Arabidopsis root hairs, including expansins [21] and xyloglucan endo transglycosylase-hydrolases (XTHs) [20,42,43,44,45]. Our data demonstrate that AtGH9C1 constitutes another molecular player in Arabidopsis root hair growth and is expressed before the initiation of the bulge at the trichoblasts, and contributes to the elongation of the hair cell (tip growth) in coordination with other cell wall enzymes.
The specific expression of AtGH9C1 in the root hair was clearly demonstrated by our transgenic lines carrying the promoter AtGH9C1::GUS construct and by published root microarray experiments that analyzed the differentiation zone where root hairs develop. GUS activity was observed in the trichoblast committed to the development of hairs, both before hair initiation and after the hair formed. This activity was increased by treatments that increased root hair density (ACC), increase root length and root hair length (sucrose) and decreased by treatments that reduced root length (mannitol).
A role of AtGH9C1 expression in root hair development was further supported by published microarray experiments that included the concomitant response of AtGH9C1 expression to treatments that, a) increased root hair length (P and Fe deficiency); or b) mutants with short or defective root hairs (rhd2, rsl4) [33,46]. Specifically, in the rhd2-1 mutant, the expression of AtGH9C1 is reduced 58% compared to wild type, which is comparable to the reduction observed for the root-hair specific expansins [33]. Finally, a role of AtGH9C1 expression in root hair development was also supported by the root hair phenotype of the homozygous SALK 006684 mutant, which became visible when a low activity threshold for AtGH9C1 was reached.
AtGH9C1 responds to ethylene signaling and its transcriptional regulation includes several transcription factors that control root hair growth and elongation such as RSL4, a basic helix-loop-helix (bHLH) [32]. By searching the transcriptome of RSL4, we found that the expression of AtGH9C1 followed a pattern of gene expression consistent with the regulation of root hair formation by RSL4. More recently, the transcriptome data of upbeat [46] become available and showed that AtGH9C1 expression was 15 times reduced compared to wild type in the root elongation zone. UPBEAT1 is a transcription factor that regulates the balance between the zones of cellular proliferation and cell elongation [46]. Since our AtGH9C1::GUS transgenic lines showed strong reporter activity in the root-hair cell before bulge formation, it suggest that AtGH9C1 expression is de-repressed by UPBEAT1 once the cell has reached the transition zone between cell elongation and root hair differentiation.

A Role for AtGHC1 in Endosperm Development
In addition to expression in root hair, RT-PCR demonstrated AtGH9C1 expression in young siliques while our GUS transgenic established a localization of the expression to the seed endosperm. But the connection of this expression with the development of the seed endosperm was revealed by reference to the public transcriptome series of Arabidopsis seed development (before and after endosperm cellularization). The endosperm is an specialized tissue that provides the essential nutrients for the development of the embryo and the carbon needed for the synthesis of wall polysaccharides of the new embryo cells [47]. Some of this carbon is derived from catabolic reactions occurring in the cell walls of the mature endosperm, which in many seeds functions as a carbohydrate-storing compartment. However, in Arabidopsis, the cell walls of the endosperm are transient and thin. They contain limited amounts of cellulose, high amounts of hemicellulosic polysaccharides and they are totally dismantled to give room to the growing embryo [48]. The breakdown of the endosperm compartments occurs after cellularization, following a polar anterior-posterior sequence. It starts in the MCE (late heart stage) even before the PEN and the CZE are cellularised. Once the PEN cellularizes, it persists only for a few days, degenerating when the embryo progresses from the linear cotyledon to the green cotyledon stage [49]. Similarly, the CZE, which is the last region to cellularize, degenerates partially as the embryo matures to the green stage [50,51]. AtGH9C1 expression is found in all three endosperm compartments and seems to follow the progression of cellularization and dissolution in each compartment. The characterization of SALK 006684 demonstrated that reducing AtGH9C1 activity contributes to delayed seed germination and reduced seed viability. Thus, AtGH9C1 Figure 10. Phylogeny and clade organization of proteins endo-b 1,4 glucanases with a C-terminal extension. A rooted consensus tree was constructed from the deduced amino acid sequences of angiosperms and lower plant endo-b 1,4 glucanases with a C-terminal extension. Protein sequences were aligned using CLUSTAL, which was subsequently input into Phylip (see further details in Materials and Methods) to perform maximum likelihood phylogeny reconstruction. Arrows mark AtGH9C1 in CLADE 1 and S1GH9C1, in CLADE 2. The different proteins in the tree are identified by species followed by a letter suffix to distinguish multiple members from the same species. For Arabidopsis the suffix distinguishes the three members of GH9 class C. The branches of the tree are color coded to denote the monocot and dicot members of CLADE 1 and CLADE 2. The specific protein sequence accessions numbers and locus identification for each gene (http://www.ncbi.nlm.nih.gov/) are presented in Table S1. doi:10.1371/journal.pone.0049363.g010 could be involved in the process of gradually breaking down the endosperm cell walls as the embryo grows. Moreover, since AtGH9C1 begins to be expressed prior to, or at, the onset of MCE cellularization, it is likely that AtGH9C1 activity is present even before the wall is fully formed. Thus we speculate that AtGH9C1may contribute to maintain a thin and easily degradable wall of the endosperm cells.

Different Localization of AtGH9C1-GFP with and without the CBM
The deduced amino acid sequence of AtGH9C1 predicted the presence of a signal peptide at the N-terminus. Results after plasmolysis of N. benthamiana leaves, transiently expressing AtGH9C1-GFP with and without the CBM, confirmed that AtGH9C1 protein is secreted and demonstrated that the CBM immobilized the protein to specific regions of the cell wall. Transgenic Arabidopsis constitutively expressing AtGH9C1-GFP also displayed a different localization of the fusion protein when the CBM was absent.

Two Clades of CBM 49
Phylogenetic analysis, based on amino acid sequence similarity to AtGH9C1, showed that GH9 proteins with CBM49 modules exist in a wide range of photosynthetic organisms from lower species such as algae, ferns and mosses to higher plants. This suggests an important and conserved role for the CBM49 of GH9 enzymes. The tree shows two different Clades (Clade 1 and 2) for the angiosperms while the lower photosynthetic species form a distinct Clade from higher plants (Clade 3). The two Clades of higher plants indicate two subtypes of GH9 proteins with CBM49 due to some differences in amino acid composition. The biological significance of these subtypes is unknown but they could represent differences in specificity or possibly functional differences, among them. Their broad occurrence in angiosperm suggests that they play a role in plant development perhaps because they confer the enzyme's ability to recognize distinct forms of a substrate common to all cell walls. The CBM49 characterized in the tomato cellulase (SlGH9C1) by Urbanowicz et al. [15], binds crystalline cellulose and belongs to Clade2, while AtGH9C1 belongs to Clade1. Crystalline cellulose is found in all plant cell walls. Apart from this ordered physical form, cellulose is also found in more disordered, amorphous forms, at the surface of small cellulose microfibrils and when forming complex structures with other wall polysaccharides [13]. Cellulases containing cellulose-binding modules with binding specificities for different cellulose forms exsit in microbes and fungi, [10,52,53]. Thus, if it is assumed that the CBM49 in both clades bind cellulose, we speculate that the distinction between clades might reflect their ability to recognize different cellulose forms. AtGH9C1 is expressed in root hairs and endosperm cells. Other genes that cluster with AtGH9C1 are also expressed preferentially in root hair cells. The walls of hair cells, like the walls of the cellularized endosperm, are special, unlignified and short-lived. The cellulose of these walls is probably like the cellulose found in the primary wall that is expanding and consists of smaller microfibrils that could bear between 33-80% of amorphous cellulose [13,14].
The binding of the CBM to cellulose depends on the prevalence and location of the planar side chains of aromatic residues at the protein-ligand interface. This allows hydrophobic stacking with the planar crystal cellulose [10,15]. In addition in the C. fimi Cel9B CBM 4-1, some glutamine and asparagine have been shown to contribute to the affinity of towards soluble sugars and soluble cellulose derivatives [54]. The CBM49 of members of Clade1 and Clade2, have the same number and position for conserved tryptophans and tyrosines, but members of Clade1 are enriched in glutamine (,9%).
So the question remains as to which cell-wall substrate AtGH9C1 hydrolyzes, to which ligand the CBM binds and what is the distinctive aspect of this substrate where AtGH9C1 is active. When transiently expressed, AtGH9C1 is localized in some lobeneck regions of the pavement cells where a restricted outgrowth is coordinated and compensatory to the expansion of the adjacent cell [55,56]. So our results indicate that the substrate of AtGH9C1 is either localized in selected regions of the wall, or it is localized in all regions but is being modified only in specific areas that the CBM49 of AtGH9C1 recognizes. If these localized regions are expanding, this substrate is either amorphous cellulose or a glucan polymer that AtGH9C1 can modify to facilitate cell-wall weakening. Alternatively, as proposed by Shpigel et al. [57], the binding of the CBM49 to the glucan chains may inhibit crystallization of microfibrils and thus weakening the wall to facilitate expansion. However, we cannot rule out that AtGH9C1 could bind to a different type of polysaccharide such as callose or hemicelluloses.

Conclusions
AtGH9C1 is a secreted endo b-1,4 glucanase that has, in addition to the canonical cellulase catalytic site, a carbohydrate binding domain (CBM49). The CBM49 associated with this gene targets the protein to some regions of the wall and when the CBM is missing, the protein is diffusely distributed throughout the wall. Whatever the substrate may be, AtGH9C1 plays a role in root hair and endosperm development in Arabidopsis. We suggest that AtGH9C1 recognizes a form of cellulose or glucan polymer that is prevalent in the wall of root hair, in the endosperm and in the growing regions of the cell wall and that is different than the one recognized by S1GH9C1.

Plant Material and Growth Conditions
Seeds of Arabidopsis ecotypes Columbia, Lansberg or Shahadara (as indicated) were germinated in half-strength MS [58] basal salts (Sigma, St. Louis, MO), 0.5 g/L 2-[N-Morpholino] ethanesulfonic acid (Sigma), pH 5.7, supplemented with 1X Gamborg's vitamins and solidified with 1% (w/v) plant tissue culture agar (Type E, Sigma). Seeds were surface-sterilized, cold treated for 4 days at 4uC and then placed vertically in a growth chamber (Conviron) at constant 20uC under 16-h light/8-h-dark regime. For germination in soil, seeds were dispersed over wet soil (Metro-Mix 300, Scotts Company), cold treated for 4 days at 4uC and grown to maturity in a growth chamber at 20uC under 16-h light/8-h-dark cycle. Nicotiana benthamiana seeds were planted in the same manner but cold stratification was not performed.

GUS Promoter-reporter
A chimeric construct was generated by fusing the putative AtGH9C1 promoter in frame with the GUS gene. The AtGH9C1 promoter was derived from Arabidopsis genomic DNA amplified with a forward primer that contains a BamHI restriction site at the 5' end and a reverse primer with a SmaI restriction site at the 59end. The PCR product contained 1,600-bp of the putative promoter plus the 59 UTR of the AtGH9C1 gene. This fragment was cloned first into PCRII vector (Invitrogen) and then cut at the SmaI and BamHI restriction sites. The fragment was purified and fused upstream of the GUS gene in the pBI101.3 plant transformation vector opened at its BamHI/SmaI site. A plasmid containing the promoter fragment was selected based on restriction size digestion and confirmed by DNA sequence. The construct also contains the NPTII gene that confers kanamycin resistance and was delivered to Columbia, Lansberg and Shahadara wild-type plants via Agrobacterium transformation by the floral dip method [59]. Seeds derived from transformed plants were selected in germination media containing kanamycin. The AtGH9C1/GUS construct was introduced also in the ethylene insensitive ers-2-2 mutant by crossing it with the AtGH9C1/GUS transgenic in Columbia background. Seeds from F2 generation were screen in half-strength MS agar plates containing 10 uM ACC in the dark and seedlings that showed the tall hypocotyl phenotype were selected and assayed for GUS activity in the presence of 10 uM ACC. The primers used to clone t h e p r o m o t e r w e r e F w -

Histochemical GUS Analysis
Transgenic seedlings and siliques were collected in 90% acetone, incubated on ice for a few minutes, rinsed with 50 mM NaPO 4 pH 7.2, 0.5 mM K 3 Fe(CN) 6 and 0.5 mM K 4 Fe (CN) 6 , and then placed in staining solution 2 mM X-Glu; (5-bromo-4chloro-3-indolyl-b-D-glucopyranoside from Sigma), dissolved in rinse solution. They were vacuum infiltrated for 10 min and incubated at 37uC for 2 h or overnight. Seedlings were examined using a microscope and photographed with a Digital Nikon 990 Camera.

GUS Spectrophotometric Assay
For quantification of GUS activity, a fluorogenic substrate, 1 mM MUG (4-methyl-umbelliferyl-b-D-glucuronide), was used in the assay according to the method of Jefferson et al. [60]. Root and shoot tissues were collected and extracted to analyze GUS activity from transgenic AtGH9C1-GUS seedlings growing vertically in 1/2MS agar plates for 7 days. For the analysis of the effect of mannitol, sucrose or ethylene, the half-strength MS medium was supplemented with either 1%, 2%, 4% mannitol, 1%, 2% sucrose, 0.1 mM, 1 mM, and 10 mM of 1 aminocyclopropane-1 carboxylic acid (ACC, Sigma), 20 mM AVG (aminoethoxyvinylglycine, Sigma) made from a stock (1,000X) prepared in water. Tissues were ground in reaction buffer 2:1 (v/w) and centrifuged for 5 min at maximal speed in a microfuge to clear the supernatant. An aliquot of crude extract was mixed with an equal volume of reaction buffer containing 2 mM MUG and incubated at 37uC. Aliquots were taken at 15 min intervals and the reaction was stopped with 0.2M Na 2 CO 3 . Fluorescence was determined using a one-channel fluorometer (Turner Design Picofluor) with an excitation range of 365-395 nm, emission wavelength greater than 430 nm and calibrated with known concentrations of 4-methylumbelliferone (MU). The fitted linear slope (6 standard error) of MU produced vs incubation time was used as relative measure of GUS content for comparison with control treatments.

RNA Isolation and RT-PCR Analysis
For tissue specificity studies, total RNA was isolated from 200 mg of different plant tissues (buds, green siliques, rosette leaves, stems and roots) as described in the RNA Plant Mini Kit (Ambion). A two-step RT-PCR was performed using the Retroscript kit (Ambion, Austin, Texas). Total root RNA was pretreated with DNAse prior to reverse transcription (Ambion DNA-free kit). First strand cDNA was prepared with oligo dT primers and used as template for PCR reactions. RT-PCR was normalized using Actin-11 primers. Primers for AtGH9C1 were: RV-59GA-CAAAAAAGCATTCTCTACCAAGG3' based on the 39UTR; Fw-59AAACCTTATGAAACTACAAAA-CCAG3' based on the 7 th exon. Primers for Actin-11 were, Fw-59-ATGGCAGATGGT-GAAGACATTCAG-39; Rv-59GAAGCACTTCCTGTGGAC-TATTGA-39. Each PCR reaction was run for 25 or 30 cycles with a denaturation step for 45 sec at 94uC, and an extension for 70 seconds at 72uC, Annealing was for 30 seconds at optimal temperature for each primer pair: AtGH9C1 48uC; Actin-11, 60uC.

N-terminal and C-terminal GFP Fusion of AtGH9C1 with and without CBM
For N-terminal GFP fusions the signal peptide was fused upstream of the GFP followed downstream by the rest of the AtGH9C1cDNA (with and without the CBM). For the C-terminal GFP fusions, the AtGH9C1cDNA (with and without the CBM) was fused upstream of the GFP after removing the stop codon and the 39 UTR from the AtGH9C1cDNA. To prepare the C-terminal-GFP fusion, the pMDC83 vector was used. DNA segments of AtGH9C1 cDNA, with and without the CBM were amplified by PCR and cloned upstream of the GFP. The Fw primer was based on the 59UTR of AtGH9C1. The Rv primer including the CBM was based on AtGH9C1 last exon, excluding the stop codon (UAG). The Rv primer for the segment without the CBM was based on the last 20 bases of exon six. Forward and reverse primers were designed with terminal restriction sites, BamHI and KpnI, respectively, for subsequent cloning into pMDC83.
To prepare the N-terminal-GFP fusion, the pMDC45 vector was used. A stitching PCR approach was required to fuse the predicted signal peptide (SP), (MRKFGGSLFGVSLLLSVLLAAATAAÂ EYYN) of AtGH9C1 (SignalP 3.0) upstream to the GFP gene and then fuse, downstream of the GFP, the rest of the AtGH9C1cDNA (with and without the CBM). This would allow for the secretion of the fusion protein to the cell wall. The ' symbol represent the site for the endopeptidase.
The GFP gene was amplified from pMDC45 using Fw and Rv primers. The Rv primer to stitch the AtGH9C1 signal peptide to the N-terminus of GFP, and the Fw primer to stitch the C-terminus of the GFP to the rest of AtGH9C1cDNA were both 60 bp ULTRAMERS TM . The Fw and Rv primers to stitch all pieces together to be cloned into pMDC45 included, a KpnI and a PacI restriction sites, respectively. The Rv primer designed to amplify AtGH9C1 without the CBM included the sequence preceding the C-terminal extension, a stop codon and a Pac I restriction site. All primers were designed using Integrated DNA technologies (IDT) PrimerQuestSM and they are listed in Table S2 indicating the pair that was used for each specific reaction.
(NEB), 250 mM dNTP mixture, 240 pM forward primer, 240 pM reverse primer, 2.5 mM MgCl 2 , 2.5U Taq Polymerase (NEB) & approximately 50ng DNA. The reactions were run for 20 cycles with denaturation at 94uC for 30 seconds, annealing at Tm appropriate for primers for 30 seconds and elongation at 72uC for 3 minutes. Following amplification, gel electrophoresis and extraction from the gel, the three fragments: SP-GFP, GFP and GFP-AtGH9C1, were used as templates in a modified PCR reaction that included 50ng of each fragment, 24pM of the terminal primers and 2.5U Taq Polymerase as above. The PCR stitching fragment was confirmed on a 0.6% gel, isolated and subcloned into a pSTblue-1 vector (Novagen) blue according to the manufacturer's instruction. The clones containing the stitching fragments (SP-GFP and GFP-AtGH9C1) were sequenced to confirm that were fused in frame to yield the expected amino acid sequence. Sequencing was performed on a Biosystems DNA Sequencer model 3730 (UMBI DNA sequencing facility) using a T7 promoter primer (1 pM/ml). See the list of primers used in Table S2.

Ligation into pMDC45 or pMDC83
The pSTblue-1 and plant transformation vectors were restricted with the appropriate restriction enzymes: (KpnI+PacI) for pMDC45 and (BamHI+KpnI) for pMDC83. The reactions were run on an agarose gel and the needed bands were excised and purified. The appropriate fragments were ligated over night at 16uC in a 10 ml final volume for each reaction. The reaction mixture included 1/10th volume vector and 3/10th volume insert, T4 DNA ligase (2000 cohesive end units; NEB) and 1x ligase buffer (NEB). Each ligation reaction was used to transform competent high efficiency DH10B E.coli cells (NEB). Plasmids from individual colonies were isolated and restricted with KpnI+EcoRI for N-terminal-GFP fusion clones or with BamHI and KpnI, for C-terminal-GFP fusion clones.

Generating Agrobacterium Competent Cells
N-terminal-GFP and C-terminal-GFP fusion plasmids were each introduced into competent GV3101 Agrobacterium tumafaciens, plated on LB agar (1.5% w/v Difco bacto agar) containing kanamycin (50 mg/ml), gentamycin (25 mg/ml) and rifampicin (50 mg/ml) and incubated at 30uC. Individual colonies were screened by colony PCR, in which after touching and boiling for 5 minutes, the supernatant (2 ml) was used as template for PCR reaction with the appropriate primers.

Plant Transient and Stable Transformation with AtGH9C1-GFP Fusion Constructs
Nicotiana benthamiana plants (4-5 week old) were used to examine transient expression of the GFP constructs in leaf epidermal cells [61]. The GV3101 strains carrying the appropriate clones were grown on Luria broth (LB) overnight at 30uC, with the appropriate antibiotics. Three ml of each culture were centrifuged at 6000 rpm for 15 minutes and the pelleted cells resuspended in infiltration medium (10 mM CaCl2, 10 mM MgCl2, pH 5.7) to an O.D. of 0.8. Then, acetosyringone (39, 59-Dimethoxy-4'hydroxyacetophenone) in DMSO (Phytotechnology Laboratories) was added to the suspension at 70 mM final concentration and left at room temperature for 3 hours without shaking. The abaxial side of 1.5-2 inch leaves was infiltrated using a 1 ml syringe (BD integra) supporting the adaxial side of the leaf with the paper card. The infiltration medium bled into the leaf forming a ,0.25 inch diameter blotch. The plant was placed under natural sunlight for 48 hours with watering and infiltrated areas were observed with a Zeiss LSM-510 confocal microscope.
For stable A. thaliana transformation, GV3101 strains carrying the GFP fusion clones were grown in culture, and used as recommended for floral dip transformation [59]. Seeds were screened on K MS media agar plates with hygromycin (40 mg/ ml). At least 5 independent transformant lines were collected for each and F2 generation seeds were used for further studies. Transformation was confirmed in seedlings by surveying GFP florescence using an EPI fluorescence micrscope.

Root Length Measurements
Seedlings were grown in control half-strength MS media using large petri dishes (150 mm 615 mm), kept vertically for up to 10-12 days. The dish base was divided into two half so that the two Arabidopsis lines to be compared were growing in the same dish. Photographs of the plates that included a ruler were taken to determine mean root length (6SE, N = 30) using ImageJ software.

Root Hair Density and Length Measurements
Root hair density was determined on the main root using photographs of the main roots at 40X magnification. Efforts were made to maintain the roots growing over the surface of the agar, to keep as much as possible the roots separated from each other and maintain constant humidity while taking the pictures to avoid collapsing the root hairs. The root hairs longer than 0.05 mm were counted in 1 mm consecutive intervals starting at 3-5 mm from the tip to exclude the zone of elongation and the zone of root hair formation containing root hairs at different developmental stages. The root length surveyed was 3-4 mm for wild type and up to 6-8 mm for the mutant and the complemented line. The analysis included around 8 roots per line. Differences in mean root hair density between the mutant and the complemented line were tested for significance using ANOVA. Although the root hair length and density changes along older segments of root where fewer root hairs are visible, the newer segments, if the root is growing outside the agar media, have turgid and abundant root hairs that can be easily detected with a microscope.

Confocal Laser Scanning Microscopy (Zeiss LSM-510)
LSM-510 image browser software was used under expert mode. The argon laser was used at 50% power. The Plan-Neofluar 10x/ 0.3 objective lens was used to visualize the image at 10x magnification under transmitted light. After the sample was focused the transmitted light was turned off and the blue excitation light with a GFP filter was turned on. Once the desired image was observed, image scanning ('SCAN' icon) was performed with the argon laser (488 nm). The 'FITC dual with Rhodamine' single track configuration ('CONFIG' icon) was used to scan the GFP signal (BP505-550 emission filter) as well as the red chlorophyll autofluorescence (LP 560 emission filter). The transmitted light image was scanned simultaneously through a different channel. A 'Single Scan' with line averaging of 8 was used to obtain the final image. A C-Apochromat 63x/1.2 W corr water objective lens (Zeiss) was used to image samples at 63x magnification.

Phylogeny of GH9 with C-terminal Extensions
The amino acid sequence of AtGH9C1was used as queries to BLAST search 14 fully sequenced plant genomes plus genes from partially sequenced plants described in the Carbohydrate Activating Enzyme database (Phytozome 4.0 and CAZy protein database). A percent identity multiple sequence alignment was generated using CLUSTALW [62] at http://www.ebi.ac.uk/ Tools/clustalw2/index.html. Seqboot was used for a 1000 data set bootstrap resampling of the CLUSTALW data. The Consense file was used to draw a rectangular phylogram in Dendroscope [63]. Protpars was used to obtain the most likely unrooted phylogenies using the parsimony method. Consense was used to obtain a rooted consensus tree from the Protpars multiple phylogeny output (1000 datasets) and input into Phylip to construct a rooted consensus tree (http://www.ebi.ac.uk/Tools/clustalw2/index.html). Chlamydomonas was used as a root because it is considered the most distant photosynthetic organism to land plants. Figure S1 Gene Expression in response to treatments that affect root hair. Heat map (based on public root transcriptomes) showing the up-fold changes (red) in gene expression of AtGH9C1 and of other root hair genes (columns) in the root hair zone [65,66], after treatments (rows) that promote root hair development like iron and phosphorous deficiency (-Iron, -P) and salt (+Salt). Genes expression fold decrease (green) in the mutants, rhd2 and rsl4, where the mutation cause a reduction in root hair presence. The expression of AtGH9C1 and of the other root hair marker genes is restored in the RSL4 line, which complements rsl4 and cause an increase in root hair length.