Identification of quantitative trait loci controlling fibre length and lignin content in Arabidopsis thaliana stems

Abstract

Fibre properties and the biochemical composition of cell walls are important traits in many applications. For example, the lengths of fibres define the strength and quality of paper, and lignin content is a critical parameter for the use of biomass in biofuel production. Identifying genes controlling these traits is comparatively difficult in woody species, because of long generation times and limited amenability to high-resolution genetic mapping. To address this problem, this study mapped quantitative trait loci (QTLs) defining fibre length and lignin content in the Arabidopsis recombinant inbred line population Col-4×Ler-0. Adapting high-throughput phenotyping techniques for both traits for measurements in Arabidopsis inflorescence stems identified significant QTLs for fibre length on chromosomes 2 and 5, as well as one significant QTL affecting lignin content on chromosome 2. For fibre length, total variation within the population was 208% higher than between parental lines and the identified QTLs explained 50.58% of the observed variation. For lignin content, the values were 261 and 26.51%, respectively. Bioinformatics analysis of the associated intervals identified a number of candidate genes for fibre length and lignin content. This study demonstrates that molecular mapping of QTLs pertaining to wood and fibre properties is possible in Arabidopsis, which substantially broadens the use of Arabidopsis as a model species for the functional characterization of plant genes.

Introduction

Wood properties are of great significance to the multitude of nations that produce pulp and paper products. Variation in pulp properties, notably fibre morphology and physical properties such as fibre length, width, strength, and coarseness, have a profound influence on the properties of paper products (Dinwoodie, 1965; Page and Seth, 1980). In woody plants used for pulp production, the traditional method of improving fibre quality has been through tree breeding (Via et al., 2004). Recent work has shown that woody plant fibre cell dimensions and strength are under genetic control (Yu et al., 2001). Lignin, a major chemical component of wood, also plays an important role in defining the properties of pulp-based products. The lignin content of pulp fibres has to be carefully controlled in order to balance paper strength and optical qualities, and removal of lignin in the pulping process is both costly and environmentally detrimental (Odendahl, 1994; Biermann, 1996). Moreover, lignin content is also a critical parameter in the generation of biofuels from cellulosic materials (Mansfield et al., 2012). Cellulosic biomass, for example from woody species, has become increasingly important as a source of bioenergy that is not in competition with food production (Somerville et al., 2010).

Genetic engineering has demonstrated that direct gene manipulation can improve biomass production or wood properties in woody species (Eriksson et al., 2000; Pilate et al., 2002; Ranocha et al., 2002; Bjurhager et al., 2010; Li et al., 2011; Zawaski et al., 2011), but due to strong resistance in the marketplace to the use of genetically modified organisms, it would be preferable to obtain similar results through traditional breeding strategies.

Long generation times, variable growth conditions, and comparatively scarce genomic resources have limited the use of high-resolution genetic mapping in tree breeding, although each of these limitations is addressed in emerging tree genomic systems, such as Populus trichocarpa (Tuskan et al., 2006; Jansson and Douglas, 2007). A genetic model plant such as Arabidopsis thaliana, on the other hand, is well suited for genetic studies involving large numbers of individuals over multiple generations, because of its small size and short life cycle. Furthermore, a large number of Arabidopsis recombinant inbred line (RIL) populations are readily available from stock centres and some of them have been thoroughly characterized genetically (Reiter et al., 1992; Lister and Dean, 1993). Arabidopsis RIL populations have been used for fast and accurate identification of numerous quantitative trait loci (QTLs) controlling traits such as circadian rhythm (Swarup et al., 1999), the effects of various nutrients and minerals (Bentsink et al., 2003; Harada et al., 2004; Harada and Leigh, 2006; Reymond et al., 2006; Waters and Grusak, 2008; Zeng et al., 2008; Ghandilyan et al., 2009), resistance to predators (Pfalz et al., 2007), stress responses (McKay et al., 2008), biomass accumulation (Lisec et al., 2008), heterosis (Lisec et al., 2009), and salicylic acid pathway responses (Alcazar et al., 2009).

The initial identification of QTLs in Arabidopsis has enabled subsequent detailed genetic studies to unravel the molecular mechanisms controlling the traits associated with the QTLs, and these insights have often contributed to a better understanding of angiosperm plant biology in general (El-Assal et al., 2001; Kroymann et al., 2003; Mouchel et al., 2004; Masle et al., 2005; Werner et al., 2005; Bentsink et al., 2006; El-Din Zhang et al., 2006).

Here, we have explored the feasibility of detecting QTLs in an Arabidopsis RIL population for the traits of fibre length and lignin content in inflorescence stems. We have described the use of high-throughput phenotyping protocols for both traits and the molecular mapping of significant QTLs controlling fibre properties and lignin content in any plant species.

Materials and methods

Plant material

A set of 98 RILs derived from a cross between Columbia (Col-4) and Landsberg erecta (Ler-0) (Lister and Dean, 1993) was obtained from the ABRC (Stock number CS1899). Seeds were surface sterilized, plated on 0.5× MS agar medium (2.1g l^–1 MS salts, 10g l^–1 sucrose, 8g l^–1 agar, pH 5.8) and vernalized for 48h at 4 °C. Seedlings were transferred to Professional Pro-Mix ‘BX’/Mycorise Pro (Premier Horticulture, Rivière du Loup, Canada) supplemented with 0.3% (v/v) Nutricote 14-14-14 Type 100 fertilizer (Chisso-Asahi Fertilizer Co., Tokyo, Japan) after 10 d of continuous light. Plants were grown under a 16h/8h day/night cycle at 21 °C day and 18 °C night temperature. Light was maintained at 200 µmol s^–2 m^–2. Post-flowering inflorescence stems were collected for fibre quality analyser (FQA) processing or lignin content measurement. The set of lines was grown five times, in five separate experiments, and the resulting stems were analysed separately. Three sets of plants were used for fibre length measurements and two sets of plants were used for lignin content measurements. A single plant per line was used in the first FQA experiment and two plants per line were bulked in all subsequent experiments.

Isolation of fibres from plant stems

Samples (10mg) of air-dried stem material taken from the bottom 5cm of one stem or each of two stems, node and internode, were placed in a 20ml test tube for the pulping reaction. The stem sample was compacted using a glass rod before adding 2ml each of distilled water and acetic acid (glacial) and heating in a boiling water bath for 2min. After adding 2ml of 30% hydrogen peroxide, the samples were returned to the boiling water bath for 90min. The resulting solution was then carefully decanted to retain the cooked stem tissue within the tube, and the tissue was rinsed gently with distilled water three times to remove residual reagents. The delignified stem tissue was transferred to a screw-cap conical plastic centrifuge tube with 35ml of distilled water and then agitated vigorously to disintegrate fibre bundles and form a homogenous fibre suspension. The suspension was subsequently filtered through a Britt Dynamic Drainage Jar using 3 l of water, stirred using an overhead stirrer at 200 r.p.m. (TAPPI, 1992) to collect fibres retained on the 200-mesh screen (105 µm opening). The retained fibres were rinsed off the filter mesh, collected in a 50ml centrifugal tube and diluted to a total volume of 50ml with distilled water. The fibre suspensions were inspected visually and fibre bundles, if present, were removed manually using a fine needle. Triplicate measurements were performed on 10mg samples taken from individual stems of the same line.

Fibre quality analyser measurements

Fibre length was determined using a FQA (OpTest; Hawkesbury, ON, Canada) with a cytometric flow cell and image analysis system capable of rapidly and accurately measuring fibre curl, kink, and length distributions (Olson et al., 1995; Roberston et al., 1999). Five millilitres of the fibre suspension was dispensed into a 600ml plastic beaker for FQA analysis. This sample was then diluted automatically to exactly 600ml by the FQA. The fibre input was adjusted to a targeted events s^–1 measurement range of 25–40 fibres s^–1. The length weighted fibre length is reported.

Removal of extractives

The inflorescence stems of Arabidopsis accessions were first ground using a microball mill to pass through a 80-mesh screen. Prior to extraction, the ground samples were dried in a vacuum oven at 40 °C for 48h and conditioned in a vacuumed desiccator over phosphorus pentoxide overnight. The ground stems were then extracted using a rapid extraction by washing method. The procedure was based on Morrison’s quick extraction method (Morrison, 1972) with a reduction in sample particle size and weight, and passage through a fine filtration membrane. Approximately 0.1g of dried 80-mesh sample was put in a test tube, soaked with 15ml of distilled water, heated in a water bath at 65 °C for 30min with occasional shaking, and then filtered hot through a dry and pre-weighed 0.45 µm nylon membrane using a Millipore filter. The residue was first washed with 20×2ml of deionized water and then sequentially with 20×1ml each of ethanol, acetone, and diethyl ether. The residue was then transferred to a pre-weighed aluminium pan in preparation for lignin content measurement.

Lignin content measurement

The procedure used was as described by Chang et al. (2008). Dried 80-mesh samples (5±1mg) of the ground and extracted Arabidopsis stems were weighed to the nearest 0.01mg, and then digested with 1.0ml of 25% acetyl bromide in acetic acid in a 70 °C water bath for 30min with shaking at 10min intervals. After cooling at room temperature, the samples were stored in an ice bath for 5 to 120min, and 5.0ml of acetic acid was added to each sample. After mixing, 30 µl aliquots of the mixed content, in triplicate, were transferred to the wells of a 96-well quartz microplate. Transfer of the samples to the microplate was completed within 5min before sequentially adding 40 µl of 1.5M NaOH, 30 µl of 0.5M hydroxylamine hydrochloride and 150 µl of acetic acid to each well using a ten-channel multiple pipette. The absorbance of the solutions in the wells at 280nm was measured using a Perkin-Elmer Wallac 1420 microplate reader. A blank was included to correct for background absorbance by the reagents.

Data analysis

Data analysis was carried out using the R statistical language (R Development Core Team, 2009). Broad sense heritability was estimated using the analysis of variance component in the lme4 package. Variance components for the lines and for the experiments were calculated with the model:

lmer(TRAIT~ (1|LINE) + (1|EXPERIMENT))

and used to calculate the heritability with the formula:

H² = Var(lines)/[Var(lines)+Var(Residual)]

Best linear unbiased predictor (BLUP) values for each line were also calculated with the lme4 package. Variance components for the lines and the experiments were calculated using the same model as for the heritability, and the random effects for each line were extracted with the ranef function.

A composite interval method was performed with the R/qtl (Broman et al., 2003) analysis package. The imputation method was selected, with 256 draws, an error of 0.001, and 2.5 cM steps. The composite interval mapping was done with the imputation method, with three co-variables and a 10 cM window:

cim(cross, method=‘imp’, window=10, n.marcovar=3)

Calculations were performed on each experimental data set independently and also on the calculated BLUP values for fibre length and lignin content using the same method. The significant thresholds were determined by performing 1000 permutations. The multiple QTL model was devised using the fitqtl function. Results obtained from the composite interval mapping analysis were used to construct the QTLs and their positions were used in a simple additive model. The resulting model was used to assess the effect of each QTL on the trait (percentage of explanation of the observed total variance).

Analysis of variance (ANOVA) was performed to test the association of the genotype of selected markers with the observed trait, using the gee and multcomp packages.

Candidate gene selection

The list of all protein-encoding genes included in the intervals delimited by FQ2 and FQ5 (see Results) was screened for genes with enriched expression in the second-internode data set of the Bio-Array Resources for Plant Biology’s expression browser (http://bar.utoronto.ca/welcome.htm). The selected cut-off was 1.5 on the log₂ scale. Additional genes were also retained for their known connection to fibre and cell-wall development or auxin/gibberellin connection. The selected list of genes was then screened for mismatches at the amino acid level between the resequenced Col-0 genome (Cao et al., 2011) and the Ler-0 genome (Gan et al., 2011), using the Blastx algorithm (Altschul et al., 1997). Genes showing at least one change at the amino acid level were considered as candidates. For the LQ2 interval (see Results), the same method was applied, but additional genes were limited to lignin metabolism/synthesis.

Results

Fibre length measurements

The Col-4×Ler-0 RIL population created by Lister and Dean (1993) was chosen to produce the stem tissues used for the analysis. To collect material for fibre length measurements, the RIL population was grown in three independent biological samples. A single, central inflorescence stem from each line was used in the first experiment, referred to as Fibre1, whilst two stems from two individuals per line were pooled together in the two subsequent experiments, referred as Fibre2 and Fibre3. Great care was taken to minimize contamination from other cell types, typically leading to a very clean fibre preparation (data not shown). The FQA analysis yielded considerably higher fibre length values in the Col-4 parental line than in the Ler-0 parental line, as shown by a t-test (P <0.001), the mean value across the three experiments being 0.977 and 0.757mm, respectively.

The fibre length values determined across the entire set of RILs showed a wider range than the values observed for the parental lines, which indicated the possibility of transgressive segregation (Fig. 1). The shortest measured fibres within the RILs were 0.621mm long, 83.9% of the minimum Ler value, whilst the longest were 1.287mm, 122.5% of the longest Col value (Table 1). Broad sense heritability for the trait was estimated to be 62.3%. The histograms from the individual experiments for Fibre1, Fibre2, and Fibre3 (Fig. 1A, B, and 1C, respectively) showed a similar structure. In each, a main peak was associated with shorter fibres, as indicated by the median values in Table 1, with a trailing end towards the longer values.

Fig. 1.

Distribution of fibre length values (mm) in the three experiments: Fibre1 (A), Fibre2 (B), and Fibre3 (C).

Table 1.

Fibre length measurements (mm) in the RILs and Col and Ler parents.

Experiment	Parents	RIL
		Fibre length	Mean	Mean	Median	Range	SD
Fibre1	Col	1.051	0.926
	Ler	0.800
				0.943	0.909	0.643–1.287	0.143
Fibre2	Col	0.935	0.837
	Ler	0.740
				0.873	0.852	0.621–1.267	0.131
Fibre3	Col	0.945	0.839
	Ler	0.732
				0.840	0.817	0.650–1.181	0.115

Lignin content measurements

For lignin content measurements, the RIL population was grown twice to harvest material for biological replicas, labelled Lignin1 and Lignin2. Measured differences in lignin content between the parent lines were inconclusive; a 2% difference was seen between the parents in Lignin1 but almost none in Lignin2 (Table 2). This was confirmed by a t-test, which did not show a statistically significant difference between the parents (P >0.60). The mean value of both experiments was a lignin content of 19.65% (w/w) (Table 2).

Table 2.

Lignin content measurements in the RILs and Col and Ler parents All values are given as the percentage of lignin based on unextracted weight (w/w).

Experiment	Parents	RIL
		Lignin content	Mean	Mean	Median	Range	SD
Lignin1	Col	17.74	18.73
	Ler	19.73
				17.09	17.15	13.13–19.39	1.39
Lignin2	Col	20.40	20.57
	Ler	20.75
				18.07	18.09	13.80–20.98	1.64

In contrast, the mean lignin content values in individual lines of the RIL population ranged from 13.13 to 20.98%. Broad-sense heritability was estimated at 32.3%, possibly indicating a relatively limited genetic influence. The distribution of lignin content values in the Lignin1 and Lignin2 datasets was very similar, as shown in Figs 2A and B.

Fig. 2.

Distribution of the values for lignin content (%, w/w) in the two experiments: Lignin1 (A) and Lignin2 (B).

QTL analysis for fibre length

A genetic map based on the segregation of 676 single- feature polymorphisms (SFPs) in the Col-4×Ler-0 RIL population (Singer et al., 2006) was used to perform a QTL analysis using the fibre length values of 98 lines of the RIL collection, first by interval mapping (data not shown) and then by composite interval mapping (CIM) on each data set independently, using R/qtl (Jansen, 1993; Zeng, 1993, 1994; Jansen and Stam, 1994).

Two significant fibre length QTLs (P <0.05) were mapped to similar genome positions in all three data sets (Fibre1, Fibre2, and Fibre3): based on the chromosome to which they each mapped, they are referred to hereafter as FQ2 and FQ5, respectively.

FQ2 was located on the bottom arm of chromosome 2, with the maximum LOD score positioned between 49.5 and 55 cM (Fig. 3) in the map defined by Singer et al. (2006). The maximum LOD score was observed in the Fibre2 data set at 9.09, and LOD scores in Fibre1 and Fibre3 were also highly significant, with LOD scores of 8.63 and 5.67, respectively (Fig. 3). The extent of the QTL varied slightly between the data sets, but the overall confidence interval could be calculated based on a 2 LOD score drop from the summit of the peak. This placed FQ2 between 43.3 and 57 cM, an interval covering 21 SFP markers (Fig. 3, Table 3).

Fig. 3.

Distribution over chromosomes 2 and 5 of the LOD scores for fibre length. LOD was calculated independently by CIM for the three fibre experiments using a 10 cM window and three covariates. The LOD score for Fibre1 is in blue, Fibre2 in green, Fibre3 in dark red, and BLUP in yellow. The red line shows the most conservative statistically significant threshold (P <0.05), calculated for Fibre1 as 4.35. Under each graph is presented the QTL interval corresponding to a 2 LOD score drop.

Table 3.

Summary of the QTLs identified by composite interval mapping The positions and marker intervals are based on the map of Singer et al. (2006).

QTL	Experiment	Chromosome	Marker interval	Position (cM)	LOD
Fibre length
FQ2	Fibre1	2	C2_079-C2_089	53.4	8.63
	Fibre2	2	C2_075-C2_091	52.4	9.09
	Fibre3	2	C2_071-C2_088	49.1	5.67
FQ5	Fibre1	5	C5_002-C5_008	3.5	10.07
	Fibre2	5	C5_002-C5_008	3.0	8.85
	Fibre3	5	C5_008-C5_018	8.0	7.27
Lignin content
LQ2	Lignin1	2	C2_067-C2_077	45.9	5.37
	Lignin2	2	C2_056-C2_072	40.8	5.84

The second significant QTL, FQ5, was located in the distal part of the top arm of chromosome 5. The maximum LOD score was found between 3 and 8 cM, with a maximum value of 9.90 in the Fibre1 experiment (Fig. 3). The estimated extent of FQ5 with a 2 LOD score drop was between 1.5 and 11.6 cM, which encompassed 17 SFP markers (Table 3).

In order to assess the accuracy of the mapping, BLUP values were calculated from the three experiments and used to perform the CIM calculation, producing results very similar to those obtained with each individual experiment, as shown in Fig. 3.

Fig. 4 showed that the measured fibre length was strongly correlated with the genotype of two markers (C2_081 and C5_008) in all three experiments, a fact further confirmed by ANOVA (Supplementary Table 1 at JXB online). These markers were selected as proxies for FQ2 and FQ5, as they were both found as part of the interval identified in all three Fibre experiments. Plants having the Ler allele of FQ2 and the Col allele of FQ5 displayed the largest fibres (mean 0.947–1.086mm), whilst the opposite allele combination yielded the shortest fibres (mean 0.764–0.821mm). Fibre lengths in individuals possessing purely Col or Ler alleles at these sites fell in between these extremes.

Fig. 4.

Box plots of fibre length sorted by two markers each linked to QTLs for the Fibre1 (A), Fibre2 (B), and Fibre3 (C) experiments. The box plots were drawn after sorting the population according to the genotype of the markers C2_081 and C5_008, showing the difference in average fibre length (mm) of each combination. The graphs clearly show that individuals with the combination Ler C2_082 and Col C5_008 had the longest fibres, whilst individuals with the combination Col C2_082 and Ler C5_008 had the shortest fibres. The x-axis shows the genotype of the markers C2_082 and C5_008 and the y-axis shows fibre length.

Further analysis with the R/qtl package using a multiple QTL model (Haley and Knott, 1992; Sen and Churchill, 2001) was performed to estimate the contribution of each QTL to the observed variance, based on a model devised for each data set. In the three data sets, the contribution of FQ5 to the total variance (23.96–29.96%) was consistently higher than that of FQ2 (15.91–22.18%). Importantly, the estimated combined contribution of both QTLs amounted to approximately one-half of the total variance (45.79–50.58% depending on the data set).

QTL analysis for lignin content

The genetic map by Singer et al. (2006) was also used to locate QTLs controlling lignin content. Interval mapping and CIM revealed a strong LOD score peak on chromosome 2 in both independent experiments, indicative of a significant QTL, henceforth referred as LQ2 (Fig. 5). The margins, defined by a 2 LOD score drop, located the QTL between 35.3 and 47.9 cM, an interval containing 23 SFP markers. As for the fibre experiment, BLUP values were calculated from both lignin experiments and used to perform the same CIM calculation. The results obtained with the BLUPs were very similar to those seen with individual experiments (Fig. 5). A multiple QTL model was again used to determine the contribution of the significant QTL to the observed variance (Table 4), and this revealed that between 24.80 and 26.51% of the measured variance in lignin content was attributable to LQ2 in the two independent experiments. Sorting the individual lines based on a single marker (C2_067) showed that individuals carrying the Ler allele of this marker had, on average, less lignin (16.50–17.47%) than individuals carrying the Col allele (17.92–19.00%) (Fig. 6), a fact also confirmed by ANOVA (Supplementary Table 1B). The marker C2_067 was selected as a proxy for the QTL LQ2, as it belonged to the interval defined in both Lignin experiments.

Fig. 5.

Distribution over chromosome 2 of the LOD score for lignin content. LOD was calculated independently by CIM for the two lignin experiments using a 10 cM window and three covariates. The LOD score for Lignin1 is in green, Lignin2 in pink, and the BLUP in khaki. The red line shows the most conservative statistically significant threshold (P <0.05), calculated for Lignin1 as 4.39. Under the LOD score curve are indicated the intervals obtained from Lignin1 and Lignin2 in their respective colours.

Table 4.

Summary of the QTL in the multiple QTL mapping model QTLs are identified by their location. The % variance is the percentage of the total variance explained by the QTL. The LOD score given is for the entire model.

QTL	Experiment	% Variance	P value	LOD score
Fibre length
FQ2	Fibre1	22.18	3.33e–08
FQ5	Fibre1	28.40	1.03e–09	13.76
FQ2	Fibre2	21.83	3.39e–09
FQ5	Fibre2	23.96	8.03e–10	15.46
FQ2	Fibre3	15.91	6.87e–06
FQ5	Fibre3	29.96	4.80e–09	11.83
Lignin content
LQ2	Lignin1	26.51	7.45e–07	5.49
LQ2	Lignin2	24.80	4.71e–06	4.70

Fig. 6.

Box plots of the lignin content values sorted according to a QTL-linked marker foe the Lignin1 (A) and Lignin2 (B) experiments. The box plots were drawn after sorting the population according to the genotype of the marker C2_067, showing the difference in average lignin content. The graphs clearly show that individuals with the Ler marker had the lowest lignin content.

Annotated genes within the QTL intervals

As all the SFP markers used in the construction of the Singer et al. (2006) genetic map correspond to positions in the fully sequenced genome of Arabidopsis, all annotated genes within the QTL intervals identified by CIM could readily be identified. The QTL FQ2 was flanked by the markers C2_071 and C2_091, which corresponded to the genes At2g28060 and At2g37050 according to the AGI version 10 of the Arabidopsis genome. This interval is 3.6Mb long and contains 1005 loci encoding proteins. The second significant fibre length QTL FQ5, flanked by C5_002 and C5_018, was located in a 2.7Mb interval between At5g01150 and At5g08580, which contains 789 protein-encoding genes. Finally, LQ2, the QTL controlling lignin content, was located between the markers C2_056 and C2_077, corresponding to At2g23900 and At2g30240, a 1.12Mb interval that contains 672 protein-encoding loci.

Although our genetic approach could link entirely uncharacterized genes to fibre development, it is be worthwhile mining the identified intervals for plausible candidate genes, as the use of increasing genomic resources could further narrow down the selection and accelerate the cloning of the respective QTLs. On the basis of currently available resources, we selected candidate genes in the following way. First, genes with an association with fibre and cell-wall development were compiled. This list was amended for genes involved in auxin or gibberellin metabolism and signalling, two hormones with roles in fibre development (Zhong and Ye, 2001; Dayan et al., 2010; Ragni et al., 2011). At the same time, the Bio-Array Resources for Plant Biology’s expression browser (Toufighi et al., 2005) was used to identify genes whose expression was enriched in the stem second-internode dataset. Finally, all the selected genes were screened for predicted amino acid polymorphisms between the Columbia and Landsberg erecta alleles. This selection resulted in 51 genes for FQ2 and 27 for FQ5. A complete list is provided in Supplementary Table 2 (at JXB online).

Some of the selected genes were highly conspicuous. For example, a gene involved in the metabolism or deposition of cell-wall polymers other than cellulose, FRAGILE FIBER 8/IRREGULAR XYLEM 7 (FRA8/IRX7) encoding a glycosyltransferase involved in the synthesis of xylan, could affect fibre length (Brown et al., 2007). Remarkably, the fra8 mutant also displays thinner secondary cell walls in fibres (Zhong et al., 2005). Among the genes found in the FQ2 interval, there were several members of the CELLULOSE SYNTHASE-LIKE (CSL) family: CSLA7, CSLB1, -B2, -B3, and -B4, and CSLD1. CSLD1 is a gene whose activity is important for pollen-tube cell-wall development and growth (Bernal et al., 2008), while CSLA7 encodes a protein reportedly involved in the production β-mannans (Liepman et al., 2005). The function of β-mannans within the cell wall remains unknown, but CSLA7 activity is crucial for proper embryo development. A csla7 null allele causes arrest during embryogenesis, disrupting normal cell division and patterning of the embryo, precluding further analysis of the protein function at later stages (Goubet et al., 2003). However, the precise functions of the CSLB genes have not been investigated. Also, genes involved directly in cellulose synthesis deposition or related processes, whose knockout mutant phenotypes may affect many tissues, may quantitatively influence fibre properties in their allelic variation, as found in FQ2 and FQ5. The TRICHOME BIREFRINGENCE (TBR) gene is located inside FQ5. TBR is involved in synthesis of the secondary cell wall, as is its homolog TBR-LIKE 3 (TBL3), which is also present in FQ5. Mutant tbr and tbl3 plants display reductions in crystalline cellulose levels but not in total cellulose content (Bischoff et al., 2010). Furthermore, more members of the TBL gene family were found in FQ5 (TBL35) and FQ2 (TBL43, 45). In addition, lignins play major roles in determining the mechanical properties of the cell wall, and mutants affected in lignin biosynthesis or deposition may display fibre abnormalities. Among the candidate genes identified were four laccase-encoding genes: one in FQ2, LAC2, and three on FQ5, LAC10, -11, and -12. Laccases are ubiquitous copper oxidases, present in fungi and plants (Messerschmidt and Huber, 1990) and have been used for their lignin-degrading activity in fibre pulping (Mayer and Staples, 2002). In Arabidopsis, the laccase gene family comprises 17 members, of which only two have been subjected to detailed study: LAC4 and -17. Mutations in these genes result in reduced general lignification, a phenotype particularly pronounced in the interfascicular fibres in the lac17 mutant (Berthet et al., 2011). Another gene involved in lignin metabolism in FQ2 is the cinnamate 4-hydroxylase (C4H) At2g30490. A mutation in this gene, reduced epidermal fluorescence 3 (ref3), results in collapsed stem vasculature and reduced lignin content (Schilmiller et al., 2009).

A link between fibre defects and auxin transport defects is plausible in the light of cell-dimension changes in auxin-transport-inhibited plants (Mattsson et al., 1999) and has recently been supported by the walls are thin 1 (wat1) mutant in Arabidopsis (Ranocha et al., 2010), as well as by a mutation in the LIKE AUXIN RESISTANT1 (LAX1) gene, which is located in FQ5 and encodes an auxin influx carrier of the AUX1 family (for review, see Kramer, 2004). Moreover, the ATP BINDING CASSETTE B1 gene encoding an ABC transporter, also part of an auxin transport system, is located in the FQ2 interval (Wu et al., 2010). Gibberellins are also known to influence cell dimensions and Dayan et al. (2010) demonstrated a role of gibberellins in the development of fibres through modulation of GA 2-oxidase and GA 20-oxiydase activity. These findings prioritize GA 2-ox-dase in FQ2 (GA2OX3) and GA 20-oxydase in FQ5 (GA20OX3) as candidate genes for further studies.

Other genes deserve close attention because of their related mutant phenotypes. The SCAR/WAVE 1 encodes an activator of the ARP2/3 complex responsible for the nucleation of the actin filaments and, together with the other SCAR genes, is necessary for normal stem growth (Zhang et al., 2008). BELLRINGER/PENNYWISE (BLR/PNY) encodes a BELL1-like transcription factor directly involved in the patterning of the inflorescence stem. A brl/pny mutant shows defects in shoot development and displays an altered stem anatomy (Byrne et al., 2003; Smith and Hake, 2003). BLR/PNY can control internode cell elongation through modulation of the activity of the pectin methylesterase PME5 (Peaucelle et al., 2011). Interestingly, At2g36700 and At2g367210 encode a pair of pectin methylesterases and At5g04970 encodes a pectin methylesterase inhibitor, although no particular phenotypic functions are known for them. FOLYLPOLYGLUTAMATE SYNTHETASE 1 (FPGS1) is a plastidial enzyme belonging to a small gene family. Whilst the fpgs1 mutant has no discernable phenotype, the double mutant of fpgs1 and fpgs3 is dwarfed, indicating a role in stem growth (Mehrshahi et al., 2010).

Finally, another group of genes located in the FQ5 interval might also affect fibre properties through their roles in cell proliferation or cell division. One of these is the CELL-DIVISION CYCLE 48C (CDC48C) gene, one of three CDC48 genes in Arabidopsis. The roles of CDC48 genes in cell division and cytokinesis have been characterized in detail (Rancour et al., 2002; Park et al., 2008). Furthermore, the SIAMESE (SIM) protein has a cyclin-binding motif and interacts with several cell-cycle regulators (Kasili et al., 2010; Van Leene et al., 2010). SIM controls the endoreduplication in trichomes and other organs (Walker et al., 2000).

Eighteen genes from the LQ2 interval were selected as candidate genes because of suspected or demonstrated impact on lignin biosynthesis or by their enriched expression in the stem second- internode in association with predicted amino acid polymorphisms between Columbia and Landsberg erecta. Among these, the CCR6 gene, At2g23910, encodes a putative cinnamoyl CoA reductase, a member of a family of oxidoreductases that includes enzymes responsible for the final step in monolignol biosynthesis (for review, see Boerjan et al., 2003). Studies in tobacco and eucalyptus have linked CCRs and lignin biosynthesis (Lacombe et al., 1997; Ralph et al., 1998). Although the precise function of CCR6 is unknown, it is most strongly expressed in immature seeds, and microarray surveys have shown it to be co-expressed with flavonol biosynthesis genes (Yonekura-Sakakibara et al., 2008). Among the 11 CCR genes annotated in Arabidopsis, only CCR1 (At1g15950) has been characterized extensively. A ccr1 mutant seems to suffer from reduced lignin content in the stem (Jones et al., 2001), although more recent work suggests that this phenotype may be the result of delays in lignin deposition rather than an inability to execute the reaction (Patten et al., 2005; Laskar et al., 2006).

Discussion

Whilst QTLs affecting fibre properties have been mapped in several plant species, including maize (Cardinal et al., 2003; Krakowsky et al., 2006, 2005), sorghum (Shiringani and Friedt, 2011), and, most notably, cotton (Mei et al., 2004; Chee et al., 2005a , b ; Draye et al., 2005; Ulloa et al., 2005; Qin et al., 2008; Chen et al., 2009; Paterson et al., 2011; Zhang et al., 2011), application of similar approaches to trees has been hindered by the length of time required for the tree to mature, combined with the difficulty in predicting final wood properties in mature trees based on the properties of juvenile wood.

Although QTL mapping in crop plants and trees directly benefits breeding efforts, molecular access to relevant genes can be achieved much faster in genetic model plants, such as Arabidopsis, where biological properties and genomic resources enable unmatched genetic resolution and molecular access to underlying genes, although it can suffer from its strength, as its simplicity might mask complex epistatic interactions found in larger crop genomes. Furthermore, Arabidopsis has also demonstrated its suitability for providing molecular access to conserved plant biological mechanisms (Liepman et al., 2010; Wienkoop et al., 2010; Zhang et al., 2011), and, as an arboreal growth habit is not a monophylogenetic trait but has arisen independently many times in both angiosperm and gymnosperm lineages (Bowe et al., 2000; Chaw et al., 2000; Soltis et al., 2002), our understanding of the molecular mechanisms operating in the cells of trees is expected to benefit from genetic discoveries in Arabidopsis as much as in other plants.

In this study, we explored the possibility of identifying QTLs relevant for important wood traits in an established RIL population of Arabidopsis as a first step towards a molecular understanding of the underlying cellular mechanisms. Our reproducible mapping of two QTLs controlling fibre length in Arabidopsis accessions have demonstrated the feasibility of the approach, as well as of the high-throughput phenotyping technology that was employed in the study. Rapidly developing mass-sequencing and positional cloning technologies are well suited to move on towards the identification of the molecular alleles underlying these two QTLs in the future. At the same time, Arabidopsis QTL analysis can be expanded towards accessions with more strongly divergent fibre traits.

Our results for genetic control of lignin content seemed to indicate a limited genetic influence, but it was still sufficient to successfully perform QTL mapping. The results of CIM analysis were similar in the Lignin1 and Lignin2 experiments (Figs 4 and 6), which together identified the first QTL affecting lignin content in the Arabidopsis stem. As with fibre properties, studies on genetic control of lignin content have been conducted in other species, such as pine (Sewell et al., 2002), maize (Cardinal et al., 2003; Krakowsky et al., 2006, 2005), barley (Grando et al., 2005) and rice (Xie et al., 2011), research that might be complemented with molecular access to the relevant genes in Arabidopsis. As part of our mapping efforts, we have incorporated a novel, microscale, acetyl bromide-based method for determination of lignin content (Chang et al., 2008) developed for assaying large numbers of Arabidopsis plants. The results reported here encourage its application for screening a wider range of Arabidopsis accessions to identify extreme divergences of lignin content in this species. Genetic analysis of such genotypes may in turn lead to the identification of more numerous and stronger QTLs for this trait.

Previous studies have identified a number of factors influencing fibre properties, including hormonal regulation (Zhong and Ye, 2001; Dayan et al., 2010; Ragni et al., 2011), activation of a number of key transcription factors (Ko et al., 2007; Yamaguchi et al., 2011) and cytoskeletal functions (Burk and Ye, 2002). Although mutations in any of these pathways could affect fibre length, and thus generate the spurious impression that in the background of all these known influences no new determinants of fibre length might be found, this is not an uncommon situation for new genetic searches. Such searches, however, regularly find new influences and their underlying molecular bases, which can then be linked to the network of previously known mechanisms.

Whether the QTLs affect genes with a known relationship to fibre properties or entirely new genes, we can exclude that there is any obvious hormonal, cytoskeletal, or other known influence overtly segregating in the Col-4×Ler-0 RIL population that would just impinge on fibre properties. Furthermore, we explored the fidelity of FQA measurements by comparing them with microscopically measured fibre lengths in a large number of Arabidopsis accessions (R.P. Chandra, H.X. Chang, G. Soong, A. Capron, T. Berleth, and R.P. Beatson, unpublished data). We also did not observe any correlation with plant height or stem elongation, either in this study or in a systematic survey of 150 Arabidopsis accessions (R.P. Chandra, H.X. Chang, G. Soong, A. Capron, T. Berleth, and R.P. Beatson, unpublished data). This is supported by the fact that there is no straightforward correlation between plant height and cell size. For example, mutants defective in fibre development are not necessarily reduced in height (Zhong et al., 1997). Furthermore, although the erecta mutation reduces plant height, cortex cells in the mutant have been described to be longer (Torii et al., 1996). In conclusion, the absence of any recognizable correlation between fibre length and any other trait in the segregating population underscores the need to find relevant genes underlying the QTLs FQ2 and FQ5 to gain a more comprehensive understanding of fibre length regulation.

It is noteworthy that, for both the fibre length and lignin content traits, the variation within the population vastly exceeded the variation between the parental genotypes, a phenomenon that has been observed in other similar studies involving RIL populations (Reymond et al., 2006; Coluccio et al., 2011; Sanyal and Randal Linder, 2011). Multiple QTLs may mask each other’s effects within the genome of a parental line but then become fully recognizable in a segregating population. Thus, it is worthwhile searching for QTLs in established RILs first, which provides the advantages of a high density of DNA markers and thorough characterization of the parental genotypes. In the case of Col and Ler, the two most widely used Arabidopsis accessions, their fully sequenced genomes and the availability of bacterial artificial chromosome (BAC) libraries are invaluable aids for identifying the molecular basis of each QTL. The ability of one of these BACs (or an overlap of several BACs) to confer altered fibre length properties to another accession would be the entry point for the molecular genetic dissection of fibre length control.

Association genomics studies are presently underway in woody species such as Populus and Eucalyptus, whose genomes have been fully sequenced. These analyses seek to correlate single-nucleotide polymorphism allele patterns with variance in a wide range of wood properties, including fibre length and lignin content (reviewed by Nieminen et al., 2012; Mizrachi et al., 2012). As gene sequence and overall synteny are largely conserved between Arabidopsis and Populus (Tuskan et al., 2006), for example, it will be interesting to compare the results of such genome-wide analyses with the gene content of the Arabidopsis QTL intervals detected here.

Supplementary data

Supplementary data are available at JXB online.

Supplementary Table S1. ANOVA testing for correlation between traits and selected markers.

Supplemental Table 2: List of candidate genes for the three mapped QTLs.

Glossary

Abbreviations:

BAC

bacterial artificial chromosome

BLUP

best linear unbiased predictor

CIM

composite interval mapping

FQA

fibre quality analyser

QTL

quantitative trait locus

RIL

recombinant inbred line

SFP

single-feature polymorphisms.