TRICHOME BIREFRINGENCE and its homolog At5g01360 encode plant-specific DUF231 proteins required for cellulose biosynthesis in Arabidopsis

The Arabidopsis thaliana trichome birefringence ( tbr ) mutant has severely reduced crystalline cellulose in trichomes, but the molecular nature of TBR was unknown. We determined TBR to belong to the plant-specific DUF231 domain gene family comprising 46 members of unknown function in Arabidopsis. The genes harbor another plant-specific domain, called TBL domain, which contains a conserved GDSL motif known from some esterases/lipases. TBR and TBR-like 3 ( TBL3 ) are transcriptionally coordinated with primary and secondary CESA genes, respectively. The tbr and tbl3 mutants hold lower levels of crystalline cellulose and have altered pectin composition in trichomes and stems, respectively, tissues generally thought to contain mainly secondary wall crystalline cellulose. In contrast, primary wall cellulose levels remain unchanged in both mutants as measured in etiolated tbr and tbl3 hypocotyls, while the amount of esterified pectins is reduced and pectin methylesterase activity is increased in this tissue. Furthermore, etiolated tbr hypocotyls have reduced length with swollen epidermal cells, a phenotype characteristic for primary cesa mutants or wild-type treated with cellulose synthesis inhibitors. Taken together, we show that two TBL genes contribute to the synthesis and deposition of secondary wall cellulose presumably by influencing the esterification state of pectic polymers. Transcriptional profiling mature Arabidopsis trichomes reveals that NOECK encodes the MIXTA-like transcriptional regulator MYB106.


INTRODUCTION
As the major component of the plant cell wall, cellulose has diverse functions. In the primary wall, cellulose is important for the production of the cell plate during cell division, and also for anisotropic cell expansion, and for turgor pressure distribution (Shedletzky et al., 1992). Thus, the cellulose microfibrils largely determine cell shape and patterns of development (Carpita and McCann, 2000). Once plant cells have stopped expanding, some cell types deposit a secondary cell wall (Taylor et al., 2001) that is mainly comprised of highly aligned, crystalline cellulose microfibrils, non-cellulosic polysaccharides, such as xylans and mannans (Brown et al., 2005;Carpita and McCann, 2000;Ebringerova and Heinze, 2000), and lignin. These polymers provide a framework, which provide further strength to cells which have to sustain enhanced mechanical stress. Secondary cell walls are the major components in wood and plant fibers, underlining their economical importance (Brown et al., 2005). Cell walls are also important for protecting cells against pathogens, dehydration or other environmental factors (Braam et al., 1999;Jones and Takemoto, 2004;Vorwerk et al., 2004).
More than 1000 genes in the Arabidopsis genome are estimated to encode cell-wall related proteins, but the specific biological contexts and the biochemical functions of most of these proteins are largely unknown (Carpita et al., 2001;Somerville et al., 2004). The first plant cellulose synthase (CESA) genes were identified in cotton through sequence homology to conserved regions of bacterial cellulose synthases, and high expression levels coinciding with high rates of cellulose synthesis (Pear et al., 1996). A family of CESA-related genes was rapidly identified in Arabidopsis, and their function as cellulose synthases was subsequently corroborated by classical genetic approaches (Arioli et al., 1998;Taylor et al., 1999;Fagard et al., 2000; 6 fragile fiber (fra) mutants, in which interfascicular fibers exhibit reduced mechanical strength, resulted in the identification of mutant alleles for the secondary CESA genes (fra5, fra6; Zhong et al., 2003), but also yielded novel components, some of which presumably are indirectly required for secondary wall synthesis. These include the kinesin-like protein FRA1, which is essential for oriented deposition of cellulose microfibrils and cell wall strength (Zhong et al., 2002), the katanin-like protein FRA2 involved in regulating microtubule disassembly by severing microtubules (Burk et al., 2001), a type II inositol polyphosphate 5-phosphatase (Zhong et al., 2004) and the GTP-binding protein RHD3 (Wang et al., 1997, Hu et al., 2003, both required for actin organization in fiber cells. Identification of FRA8 as a putative xylan glucuronyltransferase (Zhong et al., 2005) from the glycosyl transferase (GT) 47 family of carbohydrate-active enzymes (cf. www.cazy.org) points to the importance of acidic xylan (i.e. glucuronoxylan) modifications for normal cellulose deposition during secondary wall formation in Arabidopsis (Reis and Vian;Peña et al., 2007)..
Co-expression analyses of microarray data with CESA genes as baits have been used to identify genes associated with cellulose synthesis (Persson et al., 2005;Brown et al., 2005).
Mutations in several such genes display cell wall phenotypes characteristic of cellulose deficiency, for example the COBRA-like gene IRX6 that contains reduced levels of secondary wall cellulose (Brown et al., 2005). Likewise, the irregular xylem 8 and 9 mutants (irx8, irx9) display slighty reduced levels in stem secondary wall cellulose (Brown et al., 2005), but appear to be associated with the synthesis of xylans. IRX8 (At5g54690) and IRX9 (At2g37090) encode putative glycosyl transferase genes from the GT8 and GT43 families, respectively (cf. www.cazy.org), and were found to be involved in glucuroxylan (GX) synthesis (Bauer et al., 2006;Peña et al., 2007;Persson et al., 2007), suggesting a requirement for normal hemicellulosic polysaccharide synthesis in order for normal secondary wall synthesis and cellulose deposition to occur.
Another important component thought to be required for secondary wall cellulose synthesis is the gene that controls a trait referred to as TRICHOME BIREFRINGENCE (TBR) (Potikha and Delmer, 1995). The highly-ordered cellulose found in the cell walls of Arabidopsis trichomes displays strong birefringence under polarized light, whereas the Arabidopsis tbr mutant displays no such birefringence (cf. Fig. 1), and the cellulose content in tbr mutant trichomes is strongly reduced (Potikha and Delmer, 1995). In this work we report the identification of the gene responsible for the TBR trait, and a further characterization of the Arabidopsis tbr mutant.

7
We show that TBR belongs to a plant-specific, poorly described gene family (TBR-like, TBL) with 46 members in Arabidopsis. TBR and other gene family members are strongly co-expressed with primary-and secondary CESA genes, respectively. We further provide evidence that TBR and an additional member of the family influences secondary wall cellulose deposition. Our results also suggest involvement of the latter genes in pectin modification, more specifically methylesterification that may affect the deposition of secondary walls.

Novel phenotypes of the tbr mutant
The tbr mutant of Arabidopsis was previously characterized as lacking leaf and stem trichome birefringence (Potikha and Delmer, 1995;Fig. 1) characteristic of plant cells which contain highly ordered cellulose in their secondary walls. Consistently, cellulose levels were decreased in tbr trichomes by >80%. In addition, the cellulose content in the leaf vasculature was decreased with ~30%, whereas cellulose contents in stems, roots or callus tissue were unaffected (Potikha and Delmer, 1995). Other tbr phenotypes reported by Potikha and Delmer (1995) included reduced trichome density, altered trichome shape and surface appearance, lack of trichome papillae and basal cells, altered stomata shape, and altered patterns of callose deposition.
An additional cell wall-related phenotype was detected when we grew tbr mutant seedlings in the dark. Mutant seedlings frequently displayed a marked reduction in hypocotyl length ( Fig. 2A), reminiscent of etiolated primary cesa mutant seedlings, e.g. prc1-8, ixr1-2 or rsw1-10 (Fagard et al., 2000;Mouille et al., 2003), and dark-grown wild-type seedlings treated with cellulose synthesis inhibitors, like thaxtomin A (Scheible et al., 2003;Bischoff et al., 2009) or DCB (Robert et al., 2004). In addition, young etiolated tbr mutant seedlings occasionally showed slight isotropic cell expansion symptoms in the upper part of the hypocotyl (Fig. 2B, C), a phenotype that was reported previously for thaxtomin A-treated seedlings (Scheible et al., 2003). Such symptom was never observed in wild-type seedlings (Fig. 2D). These results point towards a function of TBR in primary cell wall synthesis in etiolated seedlings. To assess whether this is due to reduced levels of primary wall cellulose levels we measured the amount of crystalline cellulose in tbr and wild-type hypocotyls (Fig. 2E). We did not detect any differences in cellulose levels between the tbr and the wild-type control, indicating that alterations in other cell wall polymers may be the cause of the reduced hypocotyl elongation.
8 Greenhouse-grown tbr mutants (i.e. progeny with five backcrosses to Col-0 wild type) also repeatedly displayed a marked growth variation that is unrelated to variation in germination time. The size of tbr mutants ranged between wild type-like and considerably dwarfed, with the frequency of the latter being less than 10% (supplemental Fig. S1). Additional visible phenotypes of tbr mutants included a ~40% reduced stem thickness at the base (0.82±0.12 mm in tbr vs. 1.47±0.21 mm in wild type), and reduced leaf size (supplemental Fig. S1).

Cell wall composition of tbr trichomes
To investigate the effects on cell wall composition of the TBR gene product we isolated mature trichomes from tbr and wild-type rosette leaves. Cellulose measurements in the trichome preparations revealed a ~70% decrease in tbr relative to wild type ( Fig. 3) thus confirming previous results (Potikha and Delmer, 1995). We also analyzed the non-cellulosic sugar composition and uronic acid content in trichomes by HPAEC (Fig. 3). Compared to wild type, tbr trichomes displayed a 25% increase of the pectic component galacturonic acid. Similarly, another prominent pectic component, rhamnose was increased by ~15%. In contrast, the relative galactose-and arabinose-contents were reduced by ~15%, while glucuronic acid, a minor but for biomineralization processes important compound found in plant cell walls, was unaltered.
Neutral sugars, i.e. xylose and fucose, apparent in the side chains of hemicelluloses such as xyloglucans, were increased by around 20 % (Fig. 3). A clear separation of xylose and mannose could not be achieved by the HPAEC method used, but additional data obtained from analysis of alditol acetates by GC-MS (not shown), suggested that the xylose content was increased in tbr mutant trichomes, whereas the mannose content was unchanged.

Identification of TBR
The morphological changes and the reduction in cellulose in tbr mutant trichomes are robust and transmitted to the progeny in a pattern typical for recessive mutants (Potikha and Delmer, 1995).
Since the genetic lesion in the EMS-induced tbr mutant was unknown, we identified the TBR gene by map-based cloning (Lukowitz et al., 2000), and cosmid complementation (  Table S1). This led to the identification of a complementing clone B and an overlapping non-complementing clone C (Figs. 4 and S2). End-sequencing of the Arabidopsis genomic sequence integrated in clone B and C revealed that the only TBR candidate gene was AT5G06700. This gene is completely comprised 9 in B whereas the first two thirds of the gene are absent in C. Sequencing of AT5G06700 from tbr mutant DNA subsequently revealed a G A transition in the third exon of the annotated coding region, resulting in replacement of Gly427 by Glu in the predicted protein (Fig. 4A). A CAPS marker was developed based on the single base change found in AT5G06700. Co-segregation of strong trichome birefringence and the heterozygous CAPS genotype was found in the progeny of tbr mutants transformed with clone B, as expected (supplemental Fig. S2D).
To provide further and independent confirmation of the identity of TBR, we investigated additional reduction/loss of function alleles for AT5G06700. To this end, we first tested several T-DNA insertion lines (SALK_134006, SALK_134014, SALK_058509, SAIL_707_D07) but were unable to (i) identify homozygous mutants by PCR and (ii) detect plants with reduced or lacking trichome birefringence. Therefore we next produced an RNA interference (RNAi) construct to AT5G06700, and introduced it into wild type. Many of the resulting RNAi lines displayed considerable or complete loss of trichome birefringence (Fig. 4B). In addition, complementation of the tbr mutant was achieved by expression of the annotated AT5G06700 coding sequence under control of the CaMV 35S promoter (supplemental Fig. S3).

TBR and its homologs are plant-specific DUF231 proteins
According to the annotation provided by TAIR, TBR spans ~2.173 kb (from start to stop codon, Fig. 4A), contains five exons and encodes a 2.207 kb transcript that yields a predicted 608 amino acid 67.9 kD protein with an isoelectric point of 9.12. BLASTP and TBLASTN searches with the TBR protein sequence revealed that TBR is plant-specific, and has 45 homologs (e-values <10 -29 ) in the Arabidopsis genome. We aligned the protein sequences (Table S2), and created an unrooted phylogenetic tree (Fig. 5). The proteins cluster into three major branches consistent with previous analyses (Xin et al., 2007), with TBR (AT5G06700) being part of the smallest clade  Table S2). Three additional Arabidopsis TBLs have Ala and the residual 16 an aromatic amino acid (Y, W and F) in this position.
Furthermore, the 46 proteins contain another plant-specific region, typically 87-89 residues long, that has not been recognized previously and which we termed TBL domain (supplemental Fig. S4). Thirteen amino acids in this domain, including five unequally spaced cysteine residues, two tryptophans and a glycine-aspartate-serine ( Table S2). The GDS signature is followed by a Leu in over half of the proteins (24 proteins), or by a similar non-aromatic, aliphatic amino acid like Ile and Val or Met, which in turn is preceded by several hydrophobic amino acids. GDS in the same amino acid context (i.e. GDSL) has previously been found to be a conserved motif in some esterases/lipases (Upton and Buckley, 1995;Akoh et al., 2004).
There is a remarkable stretch of highly conserved amino acids towards the end of the DUF231 domain including a DxxH motif (Supplemental Fig. S4B; Supplemental Table 2  To investigate where the TBR gene is active we created promoter GUS gene reporter lines. When expressed under the control of 1.6 kb sequence upstream of the annotated start codon (which includes 140 bp annotated 5'-UTR), GUS activity in young plantlets was prominent in leaf and stem trichomes (Figs. 6A to C). However, similar GUS constructs for the close TBR homologs At3g12060 (TBL1) and At1g60790 (TBL2) showed that these genes were not active in trichomes (Figs. 6D, and E). Interestingly, the TBL1 coding sequence expressed under control of the TBR promoter complemented the trichome birefringence phenotype of tbr mutants (data not shown).
These results suggest that the TBL1 is functionally equivalent to TBR, but may work in different tissues or cell-types compared to TBR.
In 3-weeks-old TBR:GUS gene reporter plants GUS activity was mainly associated with the leaf vasculature, and was also present in younger expanding/maturating rosette leaves ( Fig.   6F), as well as rapidly growing parts of the root, e.g. lateral root tips (data not shown). In 4weeks-old plants the signal persisted in the vasculature and trichomes, and was also strong in rapidly expanding, fortifying inflorescence stems (Fig. 6G), where primary and secondary wall celluloses are deposited. With increasing age and organ maturation GUS activity continuously decreased ( Fig. 6H) until it was hardly detectable in 6-weeks old plants (Fig. 6I). This expression pattern is in agreement with the developmental AtGenExpress ATH1 genechip data (Schmid et al., 2005; for visualization see http://www.bar.utoronto.ca/efp), and is also consistent with the one expected for a component required during cellulose deposition.

Co-expression of TBR and TBLs with CESA and other cell wall genes
The strong reduction in crystalline cellulose in tbr trichomes (Potikha and Delmer, 1995) led us to investigate whether the gene is co-expressed with cellulose synthase (CESA) genes. Analyses using the GeneCAT co-expression tool (http://genecat.mpg. de;Mutwil et al., 2008) revealed that CESA5, CESA6, CESA3 rank at positions 2, 6 and 12 on the list of genes co-expressed with TBR (supplemental Table S3). We also found the endochitinase ELP/POM/CTL1 (At1g05850) and the GPI-anchored protein encoding COBRA gene (At5g60920) on this list. These genes are wellknown components for primary wall cellulose production (Hauser et al., 1995;Schindelman et al., 2001;Zhong et al., 2002). Similar results were also obtained using other co-expression tools,  To see if also other members of the family may be transcriptionally coordinated with cellulose related processes we extended the analysis including the additional 35 DUF231 family members represented on ATH1 genechips. Interestingly, the TBL gene At5g01360 was tightly coexpressed with the secondary CESA genes, and conversely the analysis of genes co-expressed with At5g01360 in GeneCAT revealed many of the known or suspected genes important for secondary wall cellulose synthesis (Brown et al., 2005), including all three secondary CESA genes, IRX8, IRX9, IRX12 and COBL4 (supplemental Table S3).

Isolation and phenotypic characterization of tbl3 mutants
To investigate the potential involvement of At5g01360/TBL3 in cellulose / cell wall synthesis, we Growth analysis revealed that the homozygous mutant progeny displayed a variety of growth phenotypes, ranging from mild to severe reductions in inflorescence stem elongation (supplemental Fig. S6), whereas rosette leaf or root growth was only mildly affected (supplemental Fig. S6; data not shown). The variation in growth and stem size is similar to what was observed for the tbr mutants (supplemental Fig. S1). However, TBR gene expression was not changed in the tbl3 mutants (data not shown). The tbl3 mutants also frequently displayed 13 significantly (10-20%) reduced stem diameter (data not shown) as found for tbr mutants and for other secondary wall mutants (Turner & Somerville, 1997;Taylor et al., 2001). To corroborate the mutant phenotype we also produced gene-specific RNAi lines for At5g01360, and found the tbl3 growth phenotypes recapitulated in these (data not shown). These results suggest that TBL3 is required for normal stem development in Arabidopsis.
Biochemical analysis of tbl3 T-DNA mutant stems revealed a 15-20% reduction in cellulose content independent of stem age (Fig. 7A). These data were reproduced using the tbl3 RNAi lines (Fig. 7A). We also detected changes in the non-cellulosic carbohydrate composition in stem material sampled 20 days after bolting (Fig. 7B). In agreement with previous studies, wild-type stems showed a high xylose content characteristic of secondary cell walls (Turner and Somerville, 1997;Brown et al., 2005). Cell wall material from tbl3 mutant stems displayed a 20% relative reduction in xylose content, and a strong relative increase in arabinose, galactose, rhamnose and fucose, as well as uronic acids (Fig. 7B). These changes resemble the sugar compositions measured in stems of irx7 or irx8 mutants, and of mutants in other genes coexpressed with secondary CESAs (Brown et al., 2005;Persson et al., 2007). Finally, stem sections were cut from the base of the mature inflorescence stem to investigate xylem morphology of tbl3 mutant and wild type. The xylem vessels of tbl3 mutants were indistinguishable from wild-type vessels displaying open xylem elements with relatively round shape (data not shown). These data suggest that while the constituents for several important secondary wall polymers were decreased in tbl3 the integrity of the wall still is sufficient for normal xylem morphology.

Additional hypocotyl phenotypes of tbr and tbl3
To determine possible structural changes in the cell walls of tbr hypocotyls, we performed Fourier Transform Infrared (FTIR) microspectroscopy (Fig. 8A). This analysis revealed, according to Student t-test, significant changes at wave number 1168 cm -1 and also at wave number 1774 cm -1 , which most likely correspond to decrease of ester linkages presumably of methyl-esterified pectins (Mouille et al., 2003;Sene et al., 1994). The peak at 1546 cm -1 may represent changes in not further specified cell wall proteins. These results indicate that the tbr hypocotyls are not holding lower levels of cellulose (Fig. 2E) observation, a significantly elevated pectin methylesterase (PME) activity was detected in tbr protein extracts derived from etiolated seedlings (Fig. 8B).
In contrast to tbr (Fig. 2), etiolated tbl3 seedlings did not show a significant reduction in hypocotyl length (data not shown). The level of crystalline cellulose in these hypocotyls tended to be slightly increased on a dry weight basis, and FTIR analysis revealed similar changes in cell wall structures of etiolated tbl3 hypocotyls as seen for tbr (Fig. 8A). We found significant changes at wave numbers 852, 917, 933, 1191 cm -1 , and at wave numbers 1762, 1731 and 1712 cm -1 (Fig. 8A), indicating structural changes of cellulose and esterified pectins in tbl3 compared to wild-type seedlings. In addition and similar to tbr, PME activity was significantly increased in tbl3 (Fig. 8B). These data suggest that the tbr and tbl3 seedling hypocotyls hold lower levels of methylesterified pectins due to enhanced pectin methylesterase activities.

DISCUSSION
In this study we applied forward and reverse genetic approaches to identify two members of the undescribed DUF231 gene family, i.e. TBR (At5g06700) and TBL3 (At5g01360), as novel components contributing to secondary wall cellulose synthesis in higher plants. The tbr mutant was previously reported as a mutant that lacks crystalline cellulose in trichomes, and to lesser extent also in the vasculature (Potikha and Delmer, 1995). The involvement of TBR in cellulose synthesis is underpinned by the expression pattern of the gene. TBR displays extraordinary coexpression with primary CESA genes, such as CESA3, CESA5 or CES6, while co-expression with secondary CESA genes is inconspicuous. Several genes co-expressed with the primary and secondary CESA genes were previously shown to be associated with primary and secondary wall cellulose production, respectively (see introduction and results). The expression patterns obtained from the microarray datasets were mirrored in the TBR promoter::GUS staining patterns, corroborating that TBR is similarly expressed as the primary wall CESA genes. Considering the substantial reduction in trichome associated crystalline cellulose levels in tbr, it is perhaps surprising that no decrease in primary wall cellulose levels were observed. However, it is possible that other TBR-related gene products may functionally compensate for the loss of TBR in these tissues. In support of such scenario, TBL1 could rescue the trichome birefringence phenotype when expressed under the TBR promoter. We showed that the reduction in cellulose content in tbr trichomes was accompanied by an increased amount of pectic compounds, i.e. galacturonic acids, whereas the relative galactose and arabinose contents were reduced by ~15%. These results imply increased amounts of HG and decreased abundance of the galactose-and arabinose-containing side chains of the pectin rhamnogalacturonan I in the mutant. The amount of glucuronic acid remained stable in trichomes, in turn. This result confirms that the changes in uronic acids are associated with pectic structures, and rules out the idea that loss of trichome birefringence is a result of reduced uronicacid-dependent bio-mineralization/crystallization. The amount of xylose-mannose was increased indicating elevated amounts of xyloglucans. Since the effects on pectin/xyloglucan content, and/or modification, was significant we propose that tbr directly interferes with cellulose biosynthesis in trichomes, and directly or indirectly influences pectin composition in this organ.
In support of this, several mutants deficient in cellulose production display increased levels of pectins, most likely homogalacturonans (His et al., 2001).
tbr hypocotyls contained wild-type levels of primary wall cellulose, less methylesterified pectins, and hold modified cell wall structures as assessed by FTIR analysis. In addition, results of the PME assay confirm that the lower degree of pectin methylesterification comes along with a higher methylesterase activity in tbr compared to wild-type seedlings. Many primary CW mutants e.g. prc1-1, rsw1-10, kor1-1, kob1-1 have less crystalline cellulose and short hypocotyls (Desnos et al., 1996, Arioli et al., 1998, Nicol et al., 1998, Fagard et al., 2000, Pagant et al., 2002. However, dwarfed hypocotyls are not necessarily linked to the level of crystalline cellulose, but can also result from altered levels of methylesterified pectins (Derbyshire et al., 2007). In addition, it is interesting to note that the tbr hypocotyl phenotypes described partially resemble the mutant quasimodo2 (qua2), affecting a putative pectin methyltransferase. qua2 has higher levels of crystalline cellulose, a shorter hypocotyl as well as altered pectin composition (Mouille et al., 2003(Mouille et al., , 2007. Our results suggest that TBR, besides its role in secondary wall cellulose deposition in trichomes, is involved in pectin modifications in etiolated hypocotyls, a tissue mainly containing primary wall cellulose. Alternatively, the changes described could reflect a feedback loop from secondary to primary cell wall biosynthesis as has been described for mur10 (Bosca et al., 2006 (Marks et al., 2008). This may suggest that the "secondary wall" crystalline cellulose deposited in trichomes is a product of the primary CESA complex.
Alternatively, the secondary CESA complex may be recruited from other cell types, such as neighboring trichome companion or basal cells. However, both interpretations seem to be rejected by the observation that irx3 (cesa7) mutants as well as primary cesa mutants (cesa3, cesa2, cesa5, cesa6, cesa2/cesa5 and cesa2/cesa6) display wild-type like trichome birefringence (V. Bischoff, S. Nita and W. Scheible, unpublished data). Another possibility would be that cellulose synthase-like (CSL) proteins synthesize cellulose in trichomes, similar to the situation described for pollen tubes of Nicotiana alata (Doblin et al., 2001).
The second TBL/DUF231 domain gene family member that we functionally implicated in secondary cell wall and cellulose synthesis is TBL3 (AT5G01360). This gene was strongly coexpressed with CESA4, CESA7, CESA8, and analysis of a T-DNA insertion null allele and RNA interference lines confirmed its function in secondary cell wall cellulose deposition. The T-DNA mutants have shorter and weaker stems at high frequency, contained significantly less cellulose in the inflorescence stems and displayed a cell wall monosaccharide pattern that highly resembles those of several irx mutants (Brown et al., 2005). tbl3 stems appear to be enriched in pectin as judged from increased levels of uronic acids, arabinose, galactose and rhamnose. While xylose levels decreased by ~20%, this neutral sugar pattern is reminiscent of those found in irx7, irx8 or irx9 mutants (Brown et al., 2005). The irregular xylem structure is a known criteria for mutants involved in secondary wall cellulose biosynthesis (Turner and Somerville, 1997). However, the reduction of cellulose and/or the structural changes in the secondary cell wall of tbl3 were not sufficient to yield an irregular xylem phenotype with collapsed xylem vessels (data not shown).
This is in agreement with several other mutants notably involved in secondary wall cellulose biosynthesis. Some of the latter showed normal or modestly disturbed xylem vessels (Brown et al., 2005;Persson et al., 2005;Brown et al., 2009) specific tissues and also has an influence on the pectic structures in tissue containing mainly primary cell wall. A role of AT5G01360 in secondary cell wall biosynthesis was previously suggested as it turned up as member of a core xylem-specific gene set resulting from global comparative transcriptome analysis (Ko et al., 2006).
What might be the biochemical functions of the plant-specific DUF231 domain genes?
One report (Yoshida et al., 2001) describes YLS7 (AT5G51640) as highly expressed in senescent leaves, but beyond this the gene remained anonymous. Another member of the gene family (ESK1; AT3G55990) was found to act as a negative regulator of cold acclimation, as mutations in the ESK1 gene provide strong freezing tolerance (Xin et al., 2007). In a third report, it was noted that a null mutation in the PMR5 gene (i.e. AT5G58600) rendered Arabidopsis resistant to powdery mildew species, and based on cell wall and FTIR analyses pmr5 cell walls were enriched in pectin and displayed a reduced degree of pectin methylesterifiation relative to wildtype cell walls, suggesting that the gene affects pectin composition (Vogel et al., 2004). This observation is in line with our results, as we also noted changes in pectin composition in various tissues of tbr and tbl3, respectively. Although the alterations in pectin composition in tbr trichomes and tbl3 stems seem to be due to compensatory effects following the reduction in cellulose levels, our results from etiolated tbr and tbl3 hypocotyls suggest a more direct impact of these gene products on pectin modification. FTIR analysis revealed less esterified pectins and this could be confirmed by an elevated PME activity. These results indicate that at least some DUF231 gene family members, including PMR5, TBR and TBL3, are required to maintain higher pectin methylesterification states, by e.g inhibiting PME activity, rather than being PMEs by themselves, as might be deduced from the conserved GDSL and DxxH motifs known from some esterases (cf. results).
Whatever the biochemical function of the proteins, loss of TBR or TBL3 appears to increase PME activity, reduce pectin esterification and to decrease cellulose deposition in trichomes and stems, respectively. There is existing evidence indicating that correct modifications of non-cellulosic cell wall polysaccharides are important for growth of cells undergoing secondary wall thickening, the synthesis of secondary wall cellulose and secondary wall integrity. For example, the dwarfed Arabidopsis irx8/gaut12 mutant is deficient in GX and HG (Persson et al., 2007), has alterations of the glycosyl sequence at the GX reducing end (Peña et al., 2007) and has a significant reduction in secondary cell wall thickness and cellulose content (Persson et al., 2007). The degree of pectin methylesterification can limit cell growth and hypocotyl elongation in Arabidopsis, as shown by ectopic expression of a fungal pectin methylesterase (Derbyshire et al., 2007). Similarly, pectin methylesterase prevents correct growth of developing wood cells/fibres in poplar (Siedlecka et al., 2008), suggesting that proper pectin esterification is likely to be essential for xylem development and lignification (Pelloux et al., 2007). Furthermore, there is evidence for potent binding of pectins, which are enriched in neutral side chains, to cellulose (Zykwinska et al., 2005), thus leading to the assumption that pectincellulose interactions are significant for cell wall assembly and normal cellulose deposition during primary and secondary cell wall formation. In tobacco leaf explants oligogalacturonides derived from esterified HG were shown to stimulate cellulose deposition and cell wall thickening (Altamura et al., 1998), and such oligogalacturonides were also shown to elicit production of hydrogen peroxide (H 2 O 2 ) (Legendre et al., 1993;Svalheim and Robertson, 1993), probably via induction of small Rac-type GTPases which activate H 2 O 2 -producing plasma-membrane NADPH oxidases (Keller et al., 1998;Potikha et al., 1999;Wong et al., 2007). In this respect, it should be noted that the plasma-membrane-localized small GTPase AtRAC2/ROP7 (At5g45970) is specifically induced during later stages of xylem differentiation in Arabidopsis (Brembu et al., 2005), and strong co-expression with TBL3 was detected in our study (cf. Supplemental Table   S3). Constant production of low levels of H 2 O 2 stimulates the onset of secondary wall cellulose synthesis and secondary wall differentiation as shown in cotton fiber cells (Potikha et al., 1999;Karlsson et al., 2005;Hovav et al., 2008) which from the botanical point of view represent seed trichomes. It is therefore tempting to speculate that (i) TBR and/or TBL3 function in maintaining HG esterification, as suggested by our results, and that (ii) this is required for triggering secondary wall cellulose synthesis in a manner similar to the suite of events outlined above. The esterification state of pectic polymers might therefore influence the synthesis and deposition of cellulose in plants.

Plant materials and growth conditions.
Plants were grown in environmental chambers (

Positional cloning of the TBR locus
See Protocol S1

Cell wall analyses and pectin methylesterase activity assay
See Protocols S2 and S3

Analysis of trichome birefringence.
Trichome birefringence of old rosette leaves was analyzed by incubation in methanol for 20 min, followed by discoloration for 1 h in boiling 85% lactic acid, and rinsing with water (3 times). The leaves were then observed under a stereomicroscope (Leica MZ 12.5, Germany) equipped with a polarizer.

Fourier-transform infrared microspectrometry.
Four-day-old seedlings were squashed between two barium fluoride windows and rinsed abundantly with distilled water for 2 min, before drying at 37 °C for 20 min. For each mutant, twenty spectra were collected from individual hypocotyls of seedlings from four independent cultures (five seedlings from each culture), as described by Mouille et al. (2003). Normalization of the data and statistical analyses were performed as described by Mouille et al. (2003).
Normalization of the data set and statistical analyses were performed using the statistical language R V2.6. (R Development Core Team, 2006). To normalize the spectra, the baseline, estimated using a linear regression involving 10 points at each end of the spectrum, was subtracted from each absorbance value and the area was set to one by dividing each absorbance 20 value by the sum of all absorbance values. To determine the difference of the composition and the structure between mutants and wild-type, Student's t-test was performed.

Promoter-β-glucuronidase constructs and GUS staining.
To obtain a TBR promoter--glucuronidase (GUS) gene fusion construct, a PCR product was amplified from genomic DNA of Arabidopsis wild-type Col-0 using PfuTurbo polymerase  spanning the insertion site (cf. Fig. 7A) was performed on wild-type and tbl3 cDNA to confirm the absence of a PCR product when using the mutant template (cf. Fig. 7B). Quantitative realtime PCR was used to analyze the expression level of At5g01360, using a Applied Biosystems HT7900 sequence detection system, SYBR green reagent (Applied Biosystems) and primers 5' -ATCGCATCGACGCTCACAC-3' and 5' -TCAGCGGTTAGGATCTTGCC-3'. For further details and subsequent data normalization procedures see Czechowski et al. (2005).

Sequence alignment and phylogenetic analysis.
MAFFT version 5 (Katoh et al., 2005) was used to create an alignment of the Arabidopsis DUF231 protein sequences available at TAIR and of other plant TBR homologs available at NCBI. The alignment was processed with Boxshade 3.21 (http://www.ch.embnet.org) and arranged manually into the final layout (cf. supplemental Fig. S4B, Table S2). Prior to computing a phylogeny, the highly variable and gapped N-terminal region was removed manually, and all columns with 50% or more gaps were removed with the software REAP (Hartmann and Vision, 2008). RAxML v7 was then used to compute an unrooted maximum likelihood phylogeny from the masked alignment using the PROTCATWAG model (Stamatakis, 2006

Accession numbers.
Sequence data from this article can be found in the Arabidopsis Genome Initiative under the following locus identifiers: AT5G06700 (TBR), AT3G12060 (TBL1), AT1G60790 (TBL2) and AT5G01360 (TBL3). The mutant tbr sequence is available from GenBank (ID XXX).

Supplemental Material
The following materials are available in the online version of this article.  Table S1. Specifications and primer sequences of mapping markers and the tbr CAPS marker.

Supplemental
Supplemental Table S2. Partial alignment of the 46 Arabidopsis DUF231 proteins.
Supplemental Table S3. Genes co-expressed with TBR or TBL3.
Protocol S1. Positional cloning of the tbr locus     Unrooted, bootstrapped tree of the TBR-like (TBL) gene family in Arabidopsis. MAFFT (version 5) was used to create an alignment of the protein sequences available at TAIR. Highly variable, gapped N-terminal regions were removed and a maximum likelihood phylogeny was generated with RAxML model PROTCATWAG (Stamatakis, 2006), and bootstrapped (n = 1000 trials) to generate the final tree. Bootstrap values >70 are shown. Segmental / tandem gene pairs are highlighted in blue, and genes mentioned in the text in red.  Individual sugars and uronic acids (UA) of wild-type and tbl3 stems are expressed as a percentage of the total noncellulosic cell wall sugar (mean values ± standard errors; n = 4).
Significant decreases as deduced from Student t-test (p ≤ 0.05) are marked with an asterisk. activity isolated from etiolated seedlings. The mean values of PME activity are expressed as release of methanol in nmol/g fresh weight (FW) derived from wild-type, tbr and tbl3 seddlings, respectively. Mean values ± standard error bars are shown (n=5). Significant decreases (p ≤ 0.01) as deduced from Student t-test are marked with an asterisk.