Breeding with rare defective alleles (BRDA): a natural Populus nigra HCT mutant with modified lignin as a case study


  • Bartel Vanholme,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Igor Cesarino,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Geert Goeminne,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Hoon Kim,

    1. Department of Biochemistry, and the DOE Great Lakes Bioenergy Research Center, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
  • Fabio Marroni,

    1. Istituto di Genomica Applicata, Udine, Italy
    Search for more papers by this author
  • Rebecca Van Acker,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Ruben Vanholme,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Kris Morreel,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Bart Ivens,

    1. Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    2. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    Search for more papers by this author
  • Sara Pinosio,

    1. Istituto di Genomica Applicata, Udine, Italy
    Search for more papers by this author
  • Michele Morgante,

    1. Istituto di Genomica Applicata, Udine, Italy
    2. Dipartimento di Scienze Agrarie e Ambientali, Università di Udine, Udine, Italy
    Search for more papers by this author
  • John Ralph,

    1. Department of Biochemistry, and the DOE Great Lakes Bioenergy Research Center, University of Wisconsin, Madison, WI, USA
    Search for more papers by this author
  • Catherine Bastien,

    1. INRA – Unité Amélioration, Génétique et Physiologie forestières, Olivet, France
    Search for more papers by this author
  • Wout Boerjan

    Corresponding author
    1. Department of Plant Biotechnology and Bioinformatics, Ghent University, Gent, Belgium
    • Department of Plant Systems Biology, Flanders Institute for Biotechnology (VIB), Gent, Belgium
    Search for more papers by this author

Author for correspondence:

Wout Boerjan

Tel: +32 (0)9 33 13 881



  • Next-generation (NG) sequencing in a natural population of Populus nigra revealed a mutant with a premature stop codon in the gene encoding hydroxycinnamoyl-CoA : shikimate hydroxycinnamoyl transferase1 (HCT1), an essential enzyme in lignin biosynthesis.
  • The lignin composition of P. nigra trees homozygous for the defective allele was compared with that of heterozygous trees and trees without the defective allele. The lignin was characterized by phenolic profiling, lignin oligomer sequencing, thioacidolysis and NMR. In addition, HCT1 was heterologously expressed for activity assays and crosses were made to introduce the mutation in different genetic backgrounds.
  • HCT1 converts p-coumaroyl-CoA into p-coumaroyl shikimate. The mutant allele, PnHCT1-Δ73, encodes a truncated protein, and trees homozygous for this recessive allele have a modified lignin composition characterized by a 17-fold increase in p-hydroxyphenyl units.
  • Using the lignin pathway as proof of concept, we illustrated that the capture of rare defective alleles is a straightforward approach to initiate reverse genetics and accelerate tree breeding. The proposed breeding strategy, called ‘breeding with rare defective alleles’ (BRDA), should be widely applicable, independent of the target gene or the species.


Trees capture carbon dioxide and incorporate most of the carbon into lignocellulose that can later be used to produce bioenergy and biomaterials. At the dawn of the bio-based economy, the genetic improvement of trees is gaining increasing interest (Harfouche et al., 2012). Sexual hybridization by conventional breeding is still the most commonly used method. This involves crosses of parental genotypes selected from genetically diverse germplasm and the subsequent screening of the progenies for beneficial phenotypes. Lower sequencing costs have allowed association genetics and genomic selection to enter the arena of forest tree domestication (Harfouche et al., 2012; Resende et al., 2012). For these methods, a large number of natural accessions is genotyped and single nucleotide polymorphisms (SNPs) or groups of SNPs associated with phenotypes of interest are identified. However, SNPs associated with quantitative traits typically explain only a minor fraction of the phenotypic variation. Although defective (nonfunctional) alleles are expected to explain a larger portion of this variation, such alleles may be kept at a low frequency as a consequence of negative selection (Manolio et al., 2009). These rare alleles will most often be recessive and present in heterozygous condition in randomly mating populations. Hence, they remain undetectable in most conventional breeding programs that use only a limited number of parental genotypes.

In a previous study, with lignin biosynthesis genes as targets, we set out to investigate whether rare defective alleles were present in the poplar germplasm and how frequent they were (Marroni et al., 2011). Lignin is an amorphous polymer that intercalates between cellulose and hemicellulose in the secondary cell wall. Its hydrophobicity makes it an excellent compound to waterproof the cell walls of the water-conducting cells of the xylem, but simultaneously makes it one of the main obstacles to the efficient processing of lignocellulosic biomass into pulp and fermentable sugars (Vanholme et al., 2010). The pathway towards the lignin building blocks is well documented and consists of a metabolic grid that modifies phenylalanine in several steps to ultimately produce the monolignols p-coumaryl, coniferyl, and sinapyl alcohols that are then polymerized by radical coupling mechanisms to result in p-hydroxyphenyl, guaiacyl, and syringyl (H, G, and S) units in the lignin polymer in the cell wall (Boerjan et al., 2003) (Supporting Information, Fig. S1). Here we describe the characterization of a natural mutant harboring a defective gene encoding a truncated hydroxycinnamoyl-CoA : shikimate hydroxycinnamoyl transferase 1 (HCT1; accession JF693234; EC This enzyme is key in the pathway, as it channels the phenylpropanoids towards the biosynthesis of the methoxylated monolignols coniferyl alcohol and sinapyl alcohol, as shown in Arabidopsis (Hoffmann et al., 2003). Its role in poplar is unclear as it has recently been shown that phenylpropanoids such as p-coumarate can be converted to caffeic acid by the C4H/C3H complex, and, as such, bypass HCT (Chen et al., 2011; Fig. S1). In line with the position of HCT in the lignin pathway, the natural poplar mutant has an altered lignin composition. The strategy presented here can be used to study gene function in both model and nonmodel plants and opens up important perspectives to using defective alleles in breeding programs.

Materials and Methods

A detailed description of phenolic profiling and cell wall characterization can be found in Methods S1. These methods are presented briefly here.

Substrates and products

Chemicals were purchased from Sigma-Aldrich unless otherwise stated. All solvents used for mass spectrometry were ULC/MS grade (Biosolve, Valkenswaard, the Netherlands) and water was produced by a DirectQ-UV water purification system (Millipore S.A.S).

Plant material

All Black poplar (Populus nigra L.) trees described in this paper were derived from a population of 768 trees collected from natural provenances in Europe and grown in a common garden at the INRA station in Orléans, France (Marroni et al., 2011). Dormant woody cuttings of c. 20 cm, derived from field-grown trees, were planted in the glasshouse (photoperiod, 16 : 8 h, light : dark; 23–25°C) in pots (diameter 22 cm, height 21 cm; three cuttings for each genotype per pot). After rooting, plants were transferred to individual pots. The following genotypes were grown for analysis: one P. nigra homozygous for the PnHCT1-Δ73 allele (clone 71030-501; also designated Δ73/Δ73), six P. nigra genotypes heterozygous for the PnHCT1-Δ73 allele (genotypes SPM45, BSL39, VDL47, BSL01, RHN23 and PGS06; also designated Δ73/+), and 10 P. nigra genotypes without the PnHCT1-Δ73 allele (BSL28, BSL37, VDL29, VDL34, SPM37, BSL12, VDL06, LOW17, SPM26 and SPM57; for convenience called ‘wildtype’ clones or +/+ throughout the text).

Phenolic profiling

The outer xylem of the debarked 2-month-old stems was scraped with a scalpel, ground in liquid nitrogen and extracted with 1 ml methanol and subjected to solid phase extraction (SPE). After SPE extraction, the lyophilized pellet was dissolved in 50 μl water. An 8 μl aliquot was subjected to LC-MS and LC-MS/MS performed on a Waters Acquity UPLC system (Waters Corp., Milford, MA, USA) connected to a Synapt HDMS quadrupole time-of-flight (Q-TOF) mass spectrometer (Micromass, Manchester, UK). Chromatographic separation was performed on an Acquity BEH C18 column (2.1 mm × 150 mm, 1.7 μm; Waters Corp.) using a water-acetonitrile gradient elution. Full scan data were recorded in negative centroid V-mode, the mass range was set between m/z 100 and 1000 with a scan speed of 0.2 s per scan using Masslynx software (Waters Corp.). Principal component analysis (PCA) and S-plots were generated via Masslynx 4.1.

Lignin analyses

Chemical characterization of the cell wall was performed on debarked dormant woody cuttings harvested from 1-yr-old stems of field-grown trees. Subsamples of 5 mg grinded material were transferred to 2 ml vials and subjected to sequential extractions, each time for 30 min, at near-boiling temperatures in water (98°C), ethanol (76°C), chloroform (59°C), and acetone (54°C). The remaining cell wall residue was dried under vacuum and weighed again. Lignin was quantified according to the acetyl bromide method (Dence, 1992), optimized for small plant quantities. Absorbance was measured at 280 nm using a NanoDrop® ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). The lignin concentrations were calculated using the law of Bouguer–Lambert–Beer (A = ε × l × c, with ε = 20.09 l g−1 cm−1 (Iiyama & Wallis, 1988) and l = 0.1 cm). The lignin monomeric composition (H/G/S) was investigated with thioacidolysis as previously described (Robinson & Mansfield, 2009).

Whole-cell-wall NMR

Nuclear magnetic resonance was performed on the same wood samples that were used for lignin analysis. Gel-state samples were prepared as previously described (Kim et al., 2008; Kim & Ralph, 2010). The ball-milled cell walls were collected directly into the NMR tubes and gels formed using dimethyl sulfoxide-d6/pyridine-d5 (DMSO-d6/pyridine-d5) (4 : 1). NMR experiments for the whole-plant-cell-wall gel-state samples were also performed as previously described (Kim et al., 2008; Kim & Ralph, 2010). NMR spectra were acquired on a Bruker Biospin (Billerica, MA, USA) Avance 500 MHz spectrometer fitted with a cryogenically cooled 5 mm triple resonance (1H, 13C, 15N; ‘TCI’) gradient probe [optimized for 1H with both the (internal) 1H coil and preamp cooled, as well as high 13C performance due to both the 13C coil and the preamp being also cooled]. Volume integration of contours in heteronuclear single-quantum correlation spectroscopy (HSQC) plots used Bruker's TopSpin 3.1 (Mac version) software.

Heterologous expression

The coding sequence (CDS) of PnHCT1 (1302 nt) was synthesized by GenScript (Piscataway, NJ, USA) and the truncated sequence (nt 1–1083) was generated by PCR using the synthesized coding sequence as template. In an alternative approach, a premature stop codon was inserted in the full-length CDS using PCR-based, site-directed mutagenesis. The 4CL4 gene of Arabidopsis thaliana (At3g21230) was ordered from the Arabidopsis Information Resource as a gateway-compatible open reading frame clone (G21871). It later turned out to miss part of its 5′-end and the truncated sequence was adjusted by PCR. Gateway AttB recombination sites were incorporated into all clones by a subsequent PCR. All polymerase reactions were performed with iProof high fidelity polymerase (Bio-Rad). Final PCR products were inserted into the pDONR221 vector using the Gateway cloning strategy (Invitrogen), and the attB × attP recombination mixture obtained was used to transform DH5α Escherichia coli cells. Transformed cells were selected on lysogeny broth (LB) plates supplemented with kanamycin (50 mg l−1) and the presence of the insert was confirmed by colony PCR. All selected-entry clones were sequenced before being used in attL × attR recombination reactions during which the insert was transferred towards the appropriate destination vectors (423GAL-ccdB and 425GAL-ccdB; Addgene plasmids 14149 and 14153, respectively; Alberti et al., 2007) to create yeast expression vectors. The LR reaction mixture was used to transform DH5α cells, which were selected on LB plates supplemented with carbenicillin (100 mg l−1). Positive clones were confirmed by PCR and plasmids were extracted by mini-prep. Transformation of Saccharomyces cerevisia strain BY4742 was performed by the lithium acetate/polyethylene glycol method. Yeast cells transformed with the recombinant vector for 4CL4 (425GAL-4CL4) but the empty vector for HCT (423GAL-ccdB) were used as negative controls. Transformants were selected on dropout (DO) medium containing 26.7 g l−1 minimal synthetic defined (SD) medium with the appropriate DO supplement (Clontech, Saint-Germain-en-Laye, France).

Reverse transcription polymerase chain reaction (RT-PCR)

Recombinant yeast cultures were produced and gene expression was induced as described earlier. After 6 h of induction, cells were harvested by centrifugation (1700 g, 1 min) and total RNA was extracted using the RNeasy® Mini Kit (Qiagen), according to manufacturer's instructions. One microgram of total RNA was treated with RQ1 RNase-free DNase (Promega) and used as a template for the cDNA synthesis using iScript™ cDNA Synthesis Kit (Bio-Rad). RT-PCR was performed to check the expression of both HCT and 4CL4, using the following primers: HCT_F, HCT_R, 4CL4_F, and 4CL4_R. The cycling conditions were 95°C for 3 min, 35 cycles of 95°C for 30 s, 55°C for 30 s and 72°C for 1 min, and a final extension of 72°C for 5 min, and the expected amplicon length was 246 bp for HCT and 517 bp for 4CL4.

Enzyme assay

Five individual recombinant yeast colonies were combined and grown overnight in one 5 ml liquid DO medium containing glucose at 30°C with shaking. Cells were harvested and washed with deionized water before inoculation of 15 ml DO medium containing galactose to induce 4CL4 and HCT transcription. The amount of inoculum was calculated to reach an OD600 of 0.1 after inoculation. Subsequently, the yeast cultures were grown at 30°C with shaking for 16 h. Feeding with specific compounds was by adding 150 μl from a 100 mM stock solution (1 mM final concentration) of the appropriate compound. The cultures were incubated at 30°C with shaking, and samples for metabolite profiling were harvested 24 and 48 h after feeding (sample volume = 1.5 ml) and lyophilized. The pellet was resuspended in 1 ml 90% methanol and incubated in an Eppendorf thermomixer at 70°C for 10 min while shaking (750 rpm). After centrifugation at 20 000 g for 5 min, 900 μl supernatant was transferred to a 2 ml Eppendorf tube and lyophilized. The pellet was treated with 100 μl water and 100 μl cyclohexane. After 10 min of centrifugation (20 000 g), 80 μl of the aqueous phase was retained for analysis. For reversed-phase LC, 10 μl of the aqueous face was subjected to LC-MS and LC-MS2 on a Waters Acquity UPLC system connected to a Thermo LTQ XL mass spectrometer (Thermo Scientific, Waltham, MA, USA). Chromatographic separation was carried out as for phenolic profiling. The eluent was directed to the mass spectrometer, via electrospray ionization (ESI) in negative mode. MS source parameters were as follows: capillary temperature, 300°C; capillary voltage, −24 V; source voltage, 3.5 V; source current, 100 μA; sheath gas flow, 30; aux gas flow, 20; sweep gas flow, 5. The mass range was set between m/z 95 and 500. The reaction products were characterized based on their masses, retention times, and fragmentation spectra.

cDNA synthesis and quantitative PCR (qPCR)

Ten centimeter stem sections, obtained 1 m from the apex of 6-month-old glasshouse-grown poplars were harvested, debarked and snap-frozen in liquid nitrogen. Xylem was scraped from debarked and liquid nitrogen-frozen stem sections and further ground in liquid nitrogen. RNA was extracted by Trizol (Invitrogen) and resuspended in 40 μl RNase free water. RNA concentration and 260/280 ratios were determined with a NanoDrop® ND-1000 spectrophotometer (NanoDrop Technologies, Inc.). One microlitre was checked for degradation on a 2% 0.5X TAE (Tris-Acetate-EDTA) agarose gel. cDNA was synthesized from 1 μg total RNA with the Iscript cDNA synthesis kit of Bio-Rad. Pipetting for qPCR was performed with the Janus Automated Workstation (Perkin Elmer, Zaventem, Belgium) using the LightCycler 480 SYBR Green I Master mix (Roche). Quantitative PCR reactions were performed with the LightCycler 480 (Roche) using the following conditions: preincubation at 95°C for 10 min, 45 cycles of 95°C for 10 s, 60°C for 15 s and 72°C for 15 s. The gene-specific primers were based on the primers described previously (Shi et al., 2010) with minor modifications according to available sequence information for P. nigra HCT genes. 18S-rRNA was chosen as a housekeeping gene in order to compare data from different cDNA samples. The sequence of each primer is given in Table S1.

Introducing the mutation in different genetic backgrounds

To obtain a full-sib P. nigra population that segregates for the recessive phenotype caused by the Δ73 allele, the male P. nigra clone 71030-501 homozygous for Δ73 was crossed with two female trees heterozygous for the Δ73 allele (71104 and SPM12, named families EP11F1 and EP11F2, respectively). Male and female flowering branches were collected in February from the original copies of the three parental clones at their sites of origin in France. The branches were stored at 2°C until being used within 1–2 wk after collection. Male scions from 71070-501 were placed in 2 l bottles filled with water in individual pollen forcing chambers. Pollen was collected, sieved to clean it of debris and dried in small vials over Drierite® (Hammond Drierite Co., OH, USA) in capped jars. Pollen, tested for viability through an in vitro germination test, was rehydrated for 2–3 h before use by dispensing the pollen into a 50 × 15 mm Pyrex® Petri dish and placing it inside a 90 × 50 mm Petri dish, adding c. 1 ml of water to the larger dish, and covering the dish to slowly raise the internal relative humidity and allow the pollen to gradually rehydrate. Female floral branches were established for rooting in large plastic boxes filled with fresh water that was daily changed. The boxes were placed in polyethylene enclosures before floral expansion. Flower development was checked every day. At full flower expansion, while stigmas were exposed, pollen from 71030-501 was dusted on female flowers using brushes. Pollination of 10–12 catkins per cross was repeated twice at 2 d intervals as needed to catch all stigmas in a receptive condition. At the onset of capsule dehiscence/seed-shed, cotton from individual crosses was collected, checked for presence/absence of seeds and stored in the fridge at 7°C. The contents of the capsules from a single cross were treated individually to extract seeds from cotton. For EP11F1 and EP11F2 families, 350 and 115 seeds, respectively, were sown in individual containers filled with 20% sand, 40% peat and 40% ground bark. They were maintained the first 3 wk in a glasshouse under mist before being transferred outside where they were irrigated by aspersion until the end of the growing season.

Forty-three progenies were obtained from EP11F2 and 115 from EP11F1 and fresh leaves were sampled for genotyping by Sanger sequencing of the HCT locus. Fifty-centimetre-tall trees were debarked and the wood lignin was analyzed by thioacidolysis as described earlier.

HCT1 activity assays

For each of the two available families (EP11F2 and EP11F1), 20 trees were selected, of which 10 were heterozygous for the Δ73 allele, and 10 were homozygous for the defective allele. Xylem tissue was scraped from 15 cm debarked frozen stem sections harvested above the first node above the soil of 1-m-tall stems. The scraped xylem was further ground in liquid nitrogen with a mortar and pestle and the powder was extracted on ice for 1 h with 20 mM Tris–HCl buffer, pH 7.5, containing 10 mM DTT, 1% polyvinylpolypyrrolidone (PVPP), 15% glycerol and 1× cOmplete Mini Protease Inhibitor Cocktail (Roche). After centrifugation at 4°C for 10 min at 20 000 g, the total soluble protein concentration was measured in the supernatant by the Bradford method (Bradford, 1976). The reaction mixture contained 20 mM Tris–HCl buffer, pH 7.5, 1 mM DTT, 100 μM p-coumaroyl-CoA (or 50 μM p-coumaroyl-CoA and 50 μM caffeoyl-CoA), 100 μM shikimic acid and 10 μg xylem proteins in a total volume of 50 μl. Samples were incubated at 30°C for 30 min and terminated by boiling for 5 min. Reaction products were directly analyzed by LC-MS as already described.

Next-generation (NG) sequencing and bioinformatics

HCT gene sequences of Populus trichocarpa were obtained from JGI v1.1 (Tuskan et al., 2006) and Phytozome v2.2 ( The corresponding P. nigra sequences were obtained through NG sequencing of four P. nigra individuals (total coverage > 100×). Paired-end sequencing libraries were constructed with an insert size of c. 500 bp. Library preparation and sequencing followed the manufacturer's instructions (Illumina, This allowed complete P. nigra-specific sequences to be obtained for all genes of the HCT gene family, with the exception of HCT3 and HCT4, which were only partially covered. The coding portion of the HCT genes was determined based on information available on JGI and Phytozome and compared with results obtained by the gene prediction software genemark (Lomsadze et al., 2005). Obtained candidate sequences were manually corrected using ClustalW sequence alignments in BioEdit v7.1.3 (Hall, 1999). Tissue-specific expression profiles of the different HCT genes were obtained from eFP browser (Wilkins et al., 2009) and PopGenIE (Sjodin et al., 2009).


Populus nigra 71030-501 is homozygous for PnHCT1-Δ73

Marroni et al. (2011) screened for rare defective alleles of lignin biosynthesis genes in a natural population of 768 P. nigra trees by NG sequencing. One of the genes of interest was HCT, coding for an acyltransferase that is a key branch-point enzyme in the lignin biosynthetic pathway, as shown in Arabidopsis. The gene consists of three exons and the deduced coding sequence of 1302 bp encodes a protein of 433 aa (theoretical pI/MW: 6.15/47.84) containing both an HXXXD- and a DFGWG-motif, which are typical for proteins of the BAHD-family to which it belongs (D'Auria, 2006). Phylogenetic analysis revealed its orthology to HCT1 (POPTR_0003s18210) of P. trichocarpa, which, in differentiating xylem tissue, is the highest expressed member of a small gene family (= 7) (Shi et al., 2010); the gene was accordingly named PnHCT1. Among the 13 different SNPs in this gene found in the population of 768 P. nigra trees, one (C1083A) introduced an early stop codon (C361*). The truncated protein (theoretical pI/MW: 6.5/39.79) lacks 73 aa at the C-terminal end (including the conserved DFGWG-motif) and the allele was accordingly named PnHCT1-Δ73. Despite its low allele frequency (2.98%), one tree homozygous for the PnHCT1-Δ73 allele was identified in the population (clone 71030-501; Δ73/Δ73). To study the effect of the point mutation, six genotypes heterozygous for the PnHCT1-Δ73 allele (Δ73/+; SPM45, BSL39, VDL47, BSL01, RHN23 and PGS06) and 10 wildtypes (+/+; BSL28, BSL37, LOW17, VDL29, VDL34, SPM37, BSL12, VDL06, SPM26 and SPM57) were selected from the natural P. nigra population, and for each genotype four biological replicates were grown in the glasshouse. Although the Arabidopsis hct mutant is severely impaired in growth (Besseau et al., 2007), no obvious effects on height, diameter or overall phenotype of the young trees could be attributed to the Δ73/Δ73 genotype under glasshouse or field conditions (Fig. S2). However, differences in growth could have been easily masked by the overall large phenotypic variation that was observed among the genetically diverse individuals.

P. nigra 71030-501 accumulates H-containing oligolignols

Xylem tissue was harvested from 2-month-old stems of the glasshouse-grown population and the aromatic metabolites were analyzed by UPLC-MS. Over the different genotypes, a total of 4600 peaks were detected. The 68 chromatograms were aligned and analyzed by multivariate statistics. Whereas the metabolic profiles of genotypes heterozygous for the PnHCT1-Δ73 allele did not separate from those of the wildtype genotypes, the Δ73/Δ73 genotype had a clearly distinct profile and clustered separately in PCA plots (Fig. 1a). This indicated that the truncation of HCT1 had a metabolic effect despite the absence of an easily observable phenotype. S-plots were used to highlight the differences between the genotypes (Fig. 1b) and compounds that were significantly different in abundance in Δ73/Δ73 as compared with the other genotypes (Δ73/+ and +/+) were identified (Tables 1, S2). Interestingly, most of the compounds accumulating in Δ73/Δ73 were not detectable in the +/+ and Δ73/+ genotypes. Lignin sequencing revealed these molecules to be H-containing oligolignols. In these oligolignols, H units were, in most cases, positioned as phenolic end-groups. This specific position of H units is presumably the consequence of the high oxidation potential of p-coumaryl alcohol, which has a negative effect on the oxidative-driven lignin biosynthesis (Lapierre, 2009). The accumulation of H-containing oligolignols is consistent with the specific position of HCT in the lignin biosynthetic pathway (Fig. S1). Other compounds upstream of HCT did not accumulate in the mutant, indicating they were efficiently processed towards downstream compounds such as p-coumaryl alcohol. Although most of the compounds that were reduced in abundance in Δ73/Δ73 were oligolignols as well, they were exclusively composed of G and S units.

Table 1. Top 10 list of metabolites accumulating or reduced in abundance in Populus nigra clone 71030-501 (Δ73/Δ73) compared to the other genotypes (Δ73/+ and +/+)
Compound m/z FormulaΔmDaFold-change vs +/+Fold-change vs Δ73/+
  1. Ten +/+ (BSL28, BSL37, VDL29, VLD34, SPM37, BSL12, VDL06, LOW17, SPM26 and SPM57) and six Δ73/+ plants (SPM45, BSL39, VDL47, BSL01, RHN23 and PGS06) were used. Four biological replicates were analyzed for each genotype. Details concerning their characterization as well as a complete list of differentials can be found in Table S2. Compounds with similar m/z but different retention time are isomers. ??, unknown or structure not fully conclusive from MS/MS data; ΔmDa, mass accuracy; X, unknown moiety; −, compound absent in +/+ or Δ73/+.

H(8–O–4)G(8–O–4)G′ ??539.1906C29H31O10−1.1
H(8–O–4)G(8–O–4)S′ ??569.2012C30H33O11−1.1
G(e8–O–4)S(8–5)G′ ??581.2017C31H33O11−0.60.3220.262
Figure 1.

Principal component analysis (PCA) and S-plot of the metabolites. (a) PCA scatterplot of the first two principal components (x- and y-axes, respectively) using the metabolic profiles of one Δ73/Δ73 clone (Populus nigra 71030-105), six Δ73/+ genotypes (SPM45, BSL39, VDL47, BSL01, RHN23 and PGS06) and 10+/+genotypes (BSL28, BSL37, VDL29, VLD34, SPM37, BSL12, VDL06, LOW17, SPM26 and SPM57). Four biological replicates were analyzed for each genotype. (b) S-plot showing the differences in compound abundance between the Δ73/Δ73 clone and the +/+ or Δ73/+ genotypes. Compounds with high discriminatory ability are colored red and were further identified based on their m/z value and their specific MS/MS fragmentation spectra (see Table S2; +/+ or Δ73/+ = +1; Δ73/Δ73 = −1).

P. nigra 71030-501 has a modified lignin composition

Metabolic profiling showed the accumulation of H-containing oligolignols in Δ73/Δ73. To investigate to what extent the H units were incorporated into the higher-molecular-weight lignin polymers, we determined the abundance of H, G and S units that are linked by ether bonds in lignin by thioacidolysis. This analysis of xylem from dormant woody cuttings revealed an overall increase in H units from barely detectable in Δ73/+ and +/+ trees to 6.98% in Δ73/Δ73 (c. 17× increase; Fig. 2a, Table 2). The shift towards H units was largely at the expense of G units, which were reduced from 35.03 to 35.23% for Δ73/+ and +/+, respectively, and to 26.98% in Δ73/Δ73 (Fig. 2a, Table 2) resulting in an increased S/G ratio. Similar results were obtained when wood from actively growing trees was used (glasshouse-grown, 2 months after sprouting from woody cuttings), although in these samples the percentage of H units was lower (3.72%, in the Δ73/Δ73 trees; c. 16× increase), and consequently the drop in G units was no longer statistically different (Table 2). The total yield of thioacidolysis monomers (H + G + S), representing the level of uncondensed units in lignin, was not significantly different in Δ73/Δ73 compared with Δ73/+ and +/+. In line with this observation, there was no change in the lignin content, as analyzed by the acetyl bromide method (Fig. 2b), although the variation in lignin concentrations was large between individual trees as a result of the high genetic variation in Populus (Ismail et al., 2012).

Table 2. Relative thioacidolysis monomer composition in Populus nigra clone 71030-501(Δ73/Δ73) compared with the other genotypes (Δ73/+ and +/+)
  1. Five +/+ genotypes (BSL12, LOW17, SPM26, SPM37 and VDL06) and five Δ73/+ plants (BSL01, BSL39, RHN23, SPM45 and VDL47) were analyzed. Four biological replicates were analyzed for Δ73/Δ73, and two biological replicates were analyzed for each Δ73/+ and +/+ genotype. Values statistically significant from the +/+ are depicted with an asterisk (*, < 0.01). Seven progeny trees were analyzed for both genotypic classes (Δ73/Δ73 and Δ73/+) of each of the crosses. H, p-hydroxyphenyl; G, guaiacyl; S, syringyl; Old, woody cuttings; Young, 2 months after sprouting from woody cuttings; EP11F1, Populus nigra clone 71030-501 × clone 71104; EP11F2, P. nigra clone 71030-501 × clone SPM12.

+/+ 0.39 ± 0.0935.23 ± 3.1763.20 ± 3.18
Δ73/+ 0.56 ± 0.1035.03 ± 2.8664.41 ± 2.93
Δ73/Δ73 6.98 ± 0.95*26.98 ± 0.42*66.04 ± 1.37
+/+ 0.22 ± 0.0634.71 ± 2.2365.07 ± 2.22
Δ73/+ 0.34 ± 0.0633.46 ± 3.5266.02 ± 3.56
Δ73/Δ73 3.72 ± 0.50*34.71 ± 1.4366.17 ± 1.39
Δ73/+ 0.20 ± 0.0832.23 ± 1.8567.57 ± 1.79
Δ73/Δ73 3.60 ± 0.8929.93 ± 2.9166.47 ± 2.20
Δ73/+ 0.21 ± 0.1133.45 ± 2.2566.35 ± 2.32
Δ73/Δ73 2.00 ± 0.2531.44 ± 2.2866.57 ± 2.31
Figure 2.

Lignin composition and amount. (a) Percentage of p-hydroxyphenyl (H units; gray), guaiacyl (G units; black) and syringyl (S units; white) monomer yields after thioacidolysis. Values are averages obtained for one Δ73/Δ73 clone (Populus nigra 73010-501), five Δ73/+ genotypes (BSL01, BSL39, RHN23, SPM45 and VDL47) and five +/+ genotypes (BSL12, LOW17, SPM26, SPM37 and VDL06). For each of the genotypes, four biological replicates were analyzed. (b) Percentage of acetyl bromide lignin concentrations. Black bars, lignin amount relative to the dry stem; gray bars, lignin amount relative to the purified cell wall residue. Values are the averages obtained for one Δ73/Δ73 genotype (73010-501), five Δ73/+ genotypes (BSL01, BSL39, RHN23, SPM45 and VDL47) and six +/+ genotypes (BSL12, LOW17, SPM26, SPM37, VDL06 and VDL29). For each of the genotypes, two biological replicates were analyzed. The error bar represents the standard deviation.

The shift towards H-rich lignin was confirmed by two-dimensional (2D) 13C–1H correlation (HSQC) NMR spectra comparing the Δ73/Δ73 genotype with four +/+ genotypes (Figs 3, S3, S4, Table S3). Whereas both S and G aromatic resonances were easily observed, only traces of H units could be detected from the whole-cell-wall gel sample of the wildtypes (0.5%, Fig. 3c, H-contour below the plotted level). Δ73/Δ73 lignin had a remarkable increase in H units (estimated by contour volume integration at c. 16×) that was, as noted in the thioacidolysis data, largely at the expense of G units (Table S3). In contrast to the thioacidolysis data, an increase in S units was found in Δ73/Δ73 lignin. However, both methods measure H/G/S distributions on entirely different fractions of the lignin. The NMR value represents the entire cell wall lignin and is slightly biased towards end-units, whereas the thioacidolysis value represents only the monomer fraction released by cleaving β-ether bonds in the lignin. The side-chain region (Fig. S4) shows interunit distributions that are consistent with the changes in the aromatic monomer profile. Lignins from Δ73/Δ73 had spectra with obvious compositional changes. One of the β-aryl ether structures, LA-S (β-syringyl ether units), was slightly (relatively) increased, whereas the LA-H/G (β-p-hydroxyphenyl or β-guaiacyl ether units) were relatively decreased as a result of the lower abundance of G units, even though H units were increased. Phenylcoumarans LB, the abundance of which typically parallels that of G units, was decreased in the mutants; S units cannot couple at their 5-positions as a result of to the methoxyl substitution (Ralph & Lu, 2004; Ralph & Landucci, 2010). Finally, as whole-cell-wall gel samples for the NMR study were used, detailed polysaccharide (hemicelluloses and cellulose) information was also obtained, but no significant differences were observed (data not shown).

Figure 3.

Whole-cell-wall NMR. Partial short-range two-dimensional 13C–1H (heteronuclear single-quantum correlation spectroscopy (HSQC)) correlation NMR spectra (aromatic region) of lignin isolated from two biological replicates of Populus nigra 71030-501 (Δ73/Δ73) (a, b), and a control genotype (VDL06; +/+) (c). Semiquantitative volume integrations are given. Spectra of additional control genotypes (BSL37, VDL29, and LOW17) are in Figs S3 and S4, and semiquantitative (comparative) data are given in Table S3. Correlation peaks are colored the same as their assigned structures (although the H3/5 and PB3/5 correlations overlap with G correlations and are only indicated); H, p-hydroxyphenyl; G, guaiacyl; S, syringyl units; PB, p-hydroxybenzoate unit acylating lignin side-chains; correlations from cinnamyl alcohol (X1) and cinnamaldehyde (X2) end-groups are also in this region and are assigned/colored.

Confirming cell wall composition in homozygous genotypes with different genetic background

In order to confirm the cosegregation between genotype and phenotype, we generated full-sib offspring by crossing Δ73/Δ73 with two different Δ73/+ genotypes (71104 and SPM12; cross EP11F1 and EP11F2, respectively). This provided us with additional trees that were homozygous and heterozygous for the Δ73 allele but with a more comparable genetic background. The progeny of both crosses was genotyped and no obvious visible phenotypes were observed that distinguished Δ73/Δ73 from Δ73/+ trees, although large biological variation was still observed, as is common in tree breeding (Fig. S5). After 3 months of growth in the nursery, seven Δ73/Δ73 and seven Δ73/+ trees derived from each of the two crosses with comparable height were selected for analysis of lignin composition. Trees were debarked and analyzed by thioacidolysis. The H/G/S composition of the lignin was comparable to that found for young developing wood of the parental clone 71030-501. Interestingly, although in both progenies the abundance of H units increased in the Δ73/Δ73 compared with the Δ73/+ genotypes, the degree of accumulation was considerably different in both crosses (c. 18× and c. 10× increase for EP11F1 and EP11F2, respectively; Table 2). This observation indicates that the expressivity of the phenotype is highly dependent on the genetic background and is consistent with the large genetic variation in the poplar germplasm.

Transcript abundances of HCT genes are not changed in genotype 71030-501

Based on the observation that knockdown of the single-copy Arabidopsis HCT has dramatic effects on plant growth (Besseau et al., 2007; Li et al., 2010), combined with the fact that the P. trichocarpa ortholog PtHCT1 is the highest expressed member of the gene family in xylem (Shi et al., 2010), a severe phenotypic knockdown effect was expected in trees homozygous for the PnHCT1-Δ73 allele. However, although there was an up to 18-fold increase in H units in these trees, their frequency remained surprisingly small (c. 7% by thioacidolysis, c. 8% by NMR). This could point to functional redundancy, for example, caused by elevated expression of other HCT gene family members. To verify this, we analyzed the transcript abundances of the different HCT genes in scraped xylem of 6-month-old branches of +/+, Δ73/+ and Δ73/Δ73 plants. The qPCR results confirmed that among the amplified HCT genes, PnHCT1 has highest expression levels in xylem tissue. PnHCT6 expression could be detected in this tissue as well, but at considerably lower levels (Fig. S6a), and according to PopGenIE, PnHCT6 is mainly expressed in young developing xylem (Fig. S6b). Both PnHCT1 and PnHCT6 cluster together with biochemically characterized HCT genes in phylogenetic trees. The other genes of the family cluster with the closely related hydroxycinnamoyl-CoA : quinate hydroxycinnamoyl transferase (HQT) genes, which are involved in the biosynthesis of the phenolic antioxidant chlorogenic acid, at least in solanaceous species (Niggeweg et al., 2004; Hamberger et al., 2007). All amplified HQT-related HCT genes were expressed at considerably lower levels in xylem. Similar expression profiles for the different genes were obtained based on cDNA from Δ73/+ and +/+ genotypes, indicating that the transcript abundances of the different members of the gene family were not affected by the point mutation in HCT1 and that the point mutation has no effect on the gene expression or mRNA stability of PnHCT1-Δ73.

Molecular function of PnHCT1

To further support its role in lignin biosynthesis, we verified whether PnHCT1 catalyzed the same reaction as the Arabidopsis HCT. To this end, PnHCT1 was heterologously expressed in yeast and cell lysates were incubated with the appropriate substrates. p-Coumaroyl shikimate was detected after feeding with p-coumaroyl-CoA. However, shikimate conjugates were not detected after feeding with cinnamoyl-, caffeoyl-, or feruloyl-CoA. An alternative approach was tested by expressing PnHCT1 together with the 4-hydroxycinnamoyl-CoA ligase 4 (4CL4) from Arabidopsis in yeast cells. As 4CL4 has low substrate specificity (Hamberger & Hahlbrock, 2004), different CoA-substrates could be tested in the assay. Feeding the transformed yeast cultures with p-coumarate and shikimate resulted in the accumulation of different p-coumaroyl shikimate isomers (Fig. 4b, Table 3). In addition, compounds consisting of shikimate coupled to two molecules of p-coumarate were detected. These products were not detected in the absence of p-coumarate, or when using yeast cells transformed with an empty vector. In this experimental setup, caffeic and ferulic acids were esterified by shikimate, although to a lower extent than p-coumarate. This could partly be a result of 4CL4, which has lower specificity towards both compounds (Hamberger & Hahlbrock, 2004), and consequently limits the supply of appropriate CoA-building blocks to HCT. However, as these products remain below the detection limit when HCT-expressing yeast was fed with CoA-esters, we can conclude that p-coumaroyl-CoA is the preferred substrate. We also tested whether shikimate could be replaced by quinate, as is the case for several described HCT enzymes (Hoffmann et al., 2004; Kim et al., 2012). However, in our experimental setup, acyl group transfer to quinate was below the level of detection. To extrapolate our findings towards more physiologically relevant conditions, xylem proteins were extracted from trees that were hetero- or homozygous for the defective allele. When protein extracts of Δ73/+ trees were fed with p-coumaroyl-CoA and shikimate, p-coumaroyl-shikimate accumulated to high amounts, whereas significantly lower amounts of p-coumaroyl-shikimate were obtained when a similar experiment was performed with protein extracts of Δ73/Δ73 trees (Fig. 4c). Comparable results were obtained in two different populations, indicating that the results were independent of the genetic background.

Table 3. Detection of shikimate esters by ultra-performance liquid chromatography-MS ( UPLC-MS) after feeding yeast cells expressing the Arabidopsis 4CL4 in combination with PnHCT1
SubstrateProducttR (min) m/z Peak areaSE
  1. Feeding was with shikimate or quinate in combination with p-coumaric acid, caffeic acid, ferulic acid or sinapic acid. Values are averages of four replicates. No product was detected when quinate (data not shown) or sinapic acid was used. shiA, shikimate; SE, standard error; ND, not detected.

p-coumaric acid + shiAp-coumaroyl shiA8.30319577 51942 626
9.8331923 1622674
Di-p-coumaroyl shiA17.60465136 54721 202
17.8046510 7921131
Caffeic acid + shiACaffeoyl shiA6.4033530 9852060
Di-caffeoyl shiA13.70497100991
Ferulic acid + shiAFeruloyl shiA9.303494933336
Sinapic acid + shiA585ND
Figure 4.

Molecular function of Populus nigra HCT1. (a) PnHCT1 converts CoA-esters to shikimate esters. The reaction is given for p-coumaroyl-CoA as a representative example, but caffeoyl-CoA and feruloyl-CoA can be used as substrate as well. (b) Detection of 5-O-p-coumaroyl shikimate by ultra-performance liquid chromatography-MS (UPLC-MS). Black, chromatogram at m/z 319 after incubation of yeast cells expressing the Arabidopsis 4CL4 in combination with PnHCT1 in the presence of p-coumaric acid and shikimate as substrates. The peaks at 8.3 and 9.8 min are the trans- and cis-isomers of 5-O-p-coumaroyl shikimate, respectively. Gray, empty vector control, where 4CL4 was expressed in the absence of PnHCT1. Insert: MS2 spectrum of p-coumaroyl shikimate (m/z 319) in negative mode. The retention time and the MS2 were confirmed by authentic reference compounds. (c) Accumulation of p-coumaroyl-shikimic acid upon feeding xylem protein extracts with p-coumaroyl-CoA and shikimic acid. Stem tissue of two independent crosses were analyzed: EP11F1 (71072-501 × 71104) and EP11F2 (71072-501 × SPM12). Gray bars represent the progeny homozygous for the defective allele (Δ73/Δ73), whereas white bars represent progeny heterozygous for the defective allele (Δ73/+). Error bars, ± SD.


In this paper we describe the functional characterization of HCT by means of a rare recessive allele (PnHCT1-Δ73) in P. nigra. HCT is essential for the synthesis of G and S lignin monomers (Hoffmann et al., 2003), and plants homozygous for the recessive allele have an altered lignin composition characterized by a 10- to 18-fold increase in H units depending on the developmental stage and genetic background. The increase was largely at the expense of G units and resulted in a higher S/G ratio, a feature that was also observed in Populus alba × grandidentata transgenics down-regulated in C3H (Ralph et al., 2012), the enzyme catalyzing the conversion after HCT, but significantly differs from what has been observed in alfalfa (Medicago sativa) C3H- and HCT-down-regulated plants, where the S/G ratio remained fairly constant (Reddy et al., 2005; Ralph et al., 2006; Shadle et al., 2007). The reason why the increase in H unit content is at the expense of G units in poplar is currently unclear, but could, for example, be a result of the high efficiency of F5H, which guides the metabolic flux towards S units (Meyer et al., 1998).

As G and S units are formed downstream of HCT, blocking this step was expected to have a dramatic effect on lignin composition, as was observed for the Arabidopsis hct mutant, in which the H unit content was raised to c. 85% (Besseau et al., 2007). Despite the up to 18-fold increase in H units in the Δ73/Δ73 mutant as compared with heterozygous genotypes, the overall H unit content remained low, indicating that the plant has a way of coping with this mutation. It is possible that the truncated enzyme has residual activity to shuttle p-coumaroyl-CoA towards caffeoyl-CoA. Supporting this hypothesis is the fact that the putative catalytic site (HxxxD) is not affected by the point mutation. However, the 73 aa C-terminal end, which is deleted in the product of the Δ73 allele, contains the conserved DFGWG-motif that covers the binding pocket (Fig. S7) and could play a role in guiding the substrates in the active site. Similar truncations in related BAHD proteins or specific point mutations in the DFGWG-motif resulted in enzymes with drastic reductions in activity, improper folding, and subcellular mislocalization (Suzuki et al., 2003; Bayer et al., 2004; Yu et al., 2008). Unfortunately, we were never successful in showing expression of PnHCT1-Δ73 in yeast by RT-PCR, suggesting that the mRNA is unstable and rapidly degraded in yeast. Although similar post-transcriptional processing can occur in plants (Stellari et al., 2010), the PnHCT1-Δ73 transcript abundances remained similar to those of the wildtype allele in P. nigra. We therefore hypothesized that the truncated protein has some residual enzymatic activity. The ability of protein fractions extracted from the Δ73/Δ73 mutant xylem to produce p-coumaroyl-shikimate in vitro is in favor of that hypothesis. Another explanation for the absence of a more drastic phenotype could be the functional redundancy within the gene family, although we showed that the point mutation in PnHCT1 does not result in altered transcript abundances of this gene or the other members of the HCT family. Interestingly, the accumulation of H units was less pronounced in young developing wood, where HCT6 rather than HCT1 seems to be expressed. Thus, it is possible that HCT6 provides sufficient HCT activity for normal growth and development in the presence of truncated HCT1. RNAi strategies that target multiple gene family members are needed to answer this question. Finally, it has recently been shown in P. trichocarpa that the C4H/C3H complex catalyzes the 3-hydroxylation of p-coumaric acid (Chen et al., 2011), and the formed product (caffeic acid) can serve as a substrate for poplar 4CL to form the corresponding CoA-ester (Allina et al., 1998). Interestingly, this new pathway apparently cannot fully compensate for the deficiency of the HCT classical pathway, as can be concluded from the accumulation of H units in the Δ73/Δ73 trees.

For trees, only few examples exist where rare natural defective alleles have been associated with a phenotype. The first natural mutant in a tree species was described by MacKay et al. (1995), who fortuitously discovered a null allele in the cinnamyl alcohol dehydrogenase (CAD) gene by analyzing CAD allozyme patterns in a segregating population of loblolly pines (Ralph et al., 1997). The corresponding mutant allele was later discovered to be caused by a two-base-pair insertion in exon 5 of the CAD gene (Gill et al., 2003). Another example comes from association genetics of wood quality in eucalyptus, where Thumma et al. (2005) found a correlation between microfibril angle and SNPs in the cinnamoyl-CoA reductase (CCR) gene. Cloning of the causal allele revealed that the SNP caused a splice variant of the CCR transcript.

In contrast to these studies, our strategy is the first specifically to search for rare defective alleles in target genes, and provides the first poplar mutant in which the rare allele is demonstrated by reverse genetics to cause a wood phenotype when homozygous. The ability to detect and access rare defective alleles in wild germplasm opens up perspectives to accelerate tree breeding programs. Indeed, NG EcoTILLING enables multiple heterozygous carriers of defective alleles within different genetic backgrounds to be identified. Sexual hybridization will result in homozygosity of the defective locus, without inbreeding depression (Fig. S8). The derived breeding approach, BRDA, can exploit the potential of these rare natural variants with major effects on the phenotype. BRDA is a straightforward complementary approach to association genetics and genomic selection to accelerate breeding. In addition, it is an alternative strategy to validate information obtained by gene-silencing approaches and allows reverse genetics to be used in economically important nonmodel organisms for which techniques such as transformation are not available. This approach can be particularly useful for genes for which a genetically engineered knockout or knockdown has proven that loss of function can lead to trait improvement (Boerjan, 2005; Harfouche et al., 2012), and it has the additional advantage that natural alleles are genetically stable, whereas the expression of a transgene can be silenced. The foreseen reduction in the cost of NG sequencing will make it possible to screen much larger populations than the 768 accessions analyzed in this study. The vast genetic variation in the natural poplar germplasm (Ismail et al., 2012) ensures that increasing the sample size will not saturate the allele discovery. The full exploitation of the strategy will involve the collection of thousands of unrelated individuals and their arrangement into appropriately organized pools to maximize the likelihood of detecting the desired mutations while minimizing the sequencing effort. The germplasm collection may become permanently shared resources, to exploit the strategy for different sets of genes and for different traits. As sequencing costs decrease, resequencing complete genomes of thousands of individuals will become possible, reducing the search for promising parental genotypes to a mere bioinformatic search for the desired mutants.

In summary, we introduce BRDA as a new tool for breeding based on the capture of rare defective alleles from wild germplasm. BRDA can be widely used in species for which large and genetically diverse natural provenances still exist. The fact that rare natural alleles can now be identified by NG sequencing increases the economic value of genetic diversity and will hopefully help in its protection.


This work has been supported by the European Commission within the Seventh Framework Programme for Research (FP7), Project ENERGYPOPLAR (FP7-211917), and NOVELTREE (grant agreement no. 211868), by grants from the Bijzonder Onderzoeksfonds-Zware Apparatuur of the Ghent University for the FT-ICR-MS (grant no. 174PZA05) and from the Hercules Foundation for the Synapt Q-TOF (grant no. AUGE/014), by the U.S. Department of Energy's Great Lakes Bioenergy Research Center (Department of Energy Office of Science grant no. BER DE–FC02–07ER64494), and by the Multidisciplinary Research Partnership Ghent Bio-Economy. R.V. is indebted to the FWO (Fonds voor Wetenschappelijk Onderzoek –Vlaanderen) for a postdoctoral fellowship. R.V.A. is indebted to the IWT (Institute for the promotion of Innovation through Science and Technology in Flanders) for a predoctoral fellowship. We are grateful to Tessa Moses for help with the heterologous expression in yeast.