Next-generation sequencing for next-generation breeding, and more


(tel +1 706 542 1271; email

The need to accelerate tree breeding programs for a bio-based economy has never been greater, as the demand for lignocellulosic biomass from diminishing land, soil and water resources continues to grow. Advances in molecular genetics and biotechnology have long promised to accelerate forest tree improvement. However, decades of research on quantitative trait loci (QTL), and more recently on genome-wide association studies (GWAS) have not yet identified robust molecular markers that are of value to tree breeders (Neale & Kremer, 2011). On the biotechnology front, while significant progress has been made in genetic engineering (GE) of agronomic traits, such as lignin properties, field evaluation and commercial deployment of GE trees remain difficult in many parts of the world due to regulatory burdens. By a creative blend of the above approaches, Vanholme et al. (pp. 765–776) in this issue of the New Phytologist demonstrate that it is feasible to exploit natural variation of Populus nigra to identify defective variants of genes predicted by prior GE research to impact, in this case, lignin properties. Individuals carrying naturally defective alleles can then be incorporated directly into breeding programs, thereby bypassing the need for transgenic manipulation. The method, called ‘Breeding with Rare Defective Alleles’ (BRDA), offers a reverse genetics complement to emerging genomic selection for targeted improvement of quantitative traits.

‘The proof-of-concept study by Vanholme et al. demonstrates how knowledge gained from molecular, biochemical and transgenic research can be translated to the field for breeding applications.’

The core of BRDA is essentially candidate gene-based single nucleotide polymorphism (SNP) discovery, using high-throughput next-generation sequencing (NGS; Marroni et al., 2011). The candidate gene approach was popular in association studies before the wide adoption of NGS for genome-wide SNP discovery. However, like GWAS, such studies typically yield nonfunctional or small-effect variants that are of limited value (Ingvarsson & Street, 2011). Instead of genetic association, the BRDA approach as used by Vanholme et al. aims to identify ‘functional polymorphisms’ in candidate genes that are predicted to impact specific traits. Examples of functional polymorphisms include SNPs or small indels that give rise to premature stop codons (Gill et al., 2003) or splice variants (Thumma et al., 2005). Because BRDA targets genes that have previously been functionally characterized, the often daunting task of phenotyping in association studies can be reduced to more targeted analyses on a subset of the study population, more akin to those employed in GE research, to confirm the gene–trait relationship. Controlled crosses between individuals homozygous and heterozygous for the defective allele are then used to validate the observed genotype–phenotype association, and to identify suitable accessions for entry into existing breeding programs.

The natural poplar mutant characterized by Vanholme et al. harbors a premature stop codon in the hydroxycinnamoyl-CoA:shikimate hydroxycinnamoyltransferase 1 gene, which is predicted to encode a truncated protein (PnHCT1Δ73). HCT channels the 4-coumaroyl-CoA thioester, a common precursor for a wide array of natural products, into the lignin branch of phenylpropanoid metabolism. In HCT-silenced transgenic Arabidopsis and alfalfa (Medicago sativa), lignin accrual and plant growth were both severely compromised (Hoffmann et al., 2004; Chen et al., 2006). In these plants, the unusual hydroxyphenyl (H) monolignol unit accumulated to very high levels, accounting for c. 60–86% of the residual lignin, at the expense of guaiacyl (G) and syringyl (S) units typically found in dicots (Chen et al., 2006; Besseau et al., 2007). By contrast, the poplar mutant and its PnHCT1Δ73-homozygous progeny showed normal growth with attenuated lignin phenotypes. Lignin content was not changed but the H lignin fraction increased from negligible to 2–7%, with small and sometimes insignificant decreases in G lignin and little change in S lignin.

Vanholme et al. offer two possible explanations for the mild mutant phenotypes. One scenario is that the mutant HCT1, though expressed at normal levels, functions inefficiently based on xylem crude protein assays. The question then remains whether the reduced HCT activities of mutant xylem extracts reflect compromised PnHCT1Δ73 catalysis, or an inactive PnHCT1Δ73 with functional redundancy conferred by other HCTs, or both. The C-terminal Δ73 truncation spans several conserved residues implicated in the substrate binding pockets of shikimate and hydroxycinnamoyl moieties, based on the crystal structure of the PnHCT1 ortholog from Coffee canephore (Lallemand et al., 2012). However, the residues predicted to be most important for HCT catalysis were unaffected in PnHCT1Δ73, supporting a possibility that the truncated protein retained activity. The second scenario is that the residual HCT activities of mutant xylem extracts were contributed by HCT6, a genome duplicate of HCT1 (Tuominen et al., 2011). By mining a high-resolution transcriptional map of poplar wood formation (Schrader et al., 2004), Vanholme et al. found that expression of HCT6 precedes, and is stronger than, that of HCT1 during early xylem development. Their finding of an age-dependent increase of H lignin deposition in the mutant is consistent with the interpretation that both duplicates participate in lignin biosynthesis, albeit with spatiotemporal specificities. While additional research is needed to address the above scenarios and the mechanistic impact of the PnHCT1Δ73 mutation, the work by Vanholme et al. highlights the value of natural mutants with defective alleles for investigating duplicate gene function.

Another surprising finding of Vanholme et al. is the differential effect of PnHCT1Δ73 mutation on the two downstream monolignol branchways; while G lignin accumulation was slightly reduced, S lignin accrual remained essentially unchanged. At face value, this is counterintuitive given the upstream position of HCT1 in the pathway and inconsistent with reports from other HCT-silenced transgenic plants (Hoffmann et al., 2004; Chen et al., 2006). However, HCT along with 4CL (4-coumarate:CoA ligase) were recently shown to form supramolecular complexes in vivo, anchored by two cytochrome P450 membrane proteins cinnamate 4-hydroxylase (C4H) and coumaroylshikimate 3-hydroxylase (C3H), presumably for effective coupling of consecutive enzymatic reactions in the lignin pathway (Bassard et al., 2012). Interestingly, enzyme complexes of various C4H–C3H isoform configurations exhibited superior catalytic efficiency and broader substrate acceptance than the individual enzymes in vitro (Chen et al., 2011). These studies suggest that the lignin biosynthetic pathway is likely far more complex than the ‘metabolic grid’ currently known (Vanholme et al., 2012). Given the propensity of P450 proteins in the phenylpropanoid pathway to form membrane-associated multi-enzyme complexes (Ralston & Yu, 2006), it is conceivable that there exist other complexes involving F5H (ferulate 5-hydroxylase) and/or additional lignin enzymes/isoforms that channel the pathway intermediates for S lignin biosynthesis. Such complexes could potentially bypass the HCT steps, or exhibit altered enzyme kinetics due to PnHCT1Δ73 mutation. Paradoxically, the notion that the S lignin branch may operate independently of G lignin pathway has been proposed in alfalfa (Chen et al., 2006). Transgenic manipulation of lignin pathway genes has shown that plants possess a remarkably flexible lignin biosynthesis machinery, with new genes, enzymatic steps or lignin monomers continuing to be reported (Vanholme et al., 2012). The work by Vanholme et al. suggests that there may be more to discover from naturally occurring mutants with defective lignin genes.

Continued NGS-based SNP discovery in natural populations for GWAS or genomic selection will likely facilitate identification of defective candidate genes for BRDA. In principle, screening for defective alleles can also extend to promoter or other noncoding sequences for mutations in cis regulatory elements that may affect expression or stability of the target transcripts. The proof-of-concept study by Vanholme et al. demonstrates how knowledge gained from molecular, biochemical and transgenic research can be translated to the field for breeding applications. Because harmful mutations are usually eliminated by natural selection, BRDA is more likely to uncover milder and more fitness-compatible mutants than those like the lignin-reduced but growth-compromised HCT transgenics derived from GE. Naturally occurring mutants are rare, but can be extremely valuable for breeding. A well-known example is the cad mutant (clone 7-56) of loblolly pine (Pinus taeda), widely used in loblolly pine breeding programs in the United States (MacKay et al., 1997; Gill et al., 2003). The work by Vanholme et al. shows that BRDA can accelerate discovery of mutants that are of value not only to tree breeders, but also to basic research.