Introduction

Narrow-leafed lupin (Lupinus angustifolius L.) is one of four lupin grain crop species that together produce around 1.1 million tonnes of grain annually (FAO 2011). It was first cultivated as a green manure and forage crop in the late nineteenth century in Northern Europe and only became a significant grain crop as domestication traits including reduced seed alkaloids were incorporated through systematic breeding efforts in Europe and Australia (Berger et al. 2012; Brummund and Święcicki 2011; Wolko et al. 2011). Currently, narrow-leafed lupin grain is produced mainly in Australia and Eastern Europe for animal feed though there is increasing interest in its use in the human diet due to its high protein and fibre content, and low glycaemic index (Foley et al. 2011).

Narrow-leafed lupin is a diploid, self-pollinating legume species (2n = 40) with an estimated physical genome size of 924 Mb (1C, cv. Sonet) (Kasprzak et al. 2006). It belongs to the genus Lupinus L. which comprises around 267 species including both annual and perennial species and ranging from tiny herbaceous plants to medium-sized trees (Drummond et al. 2012). Lupinus has among the highest speciation rates known for any genus with particular rapid speciation observed in the Andes (Hughes and Eastwood 2006). Lupinus is part of the basal tribe Genisteae within the phylogenetic clade Genistoid in the Papilionoideae subfamily of legumes (Drummond et al. 2012; Lavin et al. 2005). Lupins are highly diverged from all the agriculturally important legumes and model species, most of which belong to two other Papilionoideae clades: Galegoid (cool season legumes) and Phaseoloid (warm season legumes) clades from which they separated about 56 million years ago (Lavin et al. 2005; Zhu et al. 2005).

Polyploidy has played a crucial role in angiosperm genome evolution including the genus Lupinus. Multiple rounds of polyploidy provide a rich source of gene redundancy that permits rapid diversification of function and/or expression of genes and contributes to the genome plasticity of angiosperms (Leitch and Leitch 2008). A whole genome duplication (WGD) event was associated with the origin of the angiosperms (De Bodt et al. 2005). An analysis of a large sample of plant transcriptomes revealed evidence of a whole genome triplication (WGT) event associated with the early diversification of the eudicots (Jiao et al. 2012). Recent studies confirm the occurrence of a further whole genome duplication (WGD) event early in Papilionoid legume evolution, which may also have included the basal Genistoid clade containing the Lupinus genus (Cannon et al. 2010; Young and Bharti 2012). An additional WGD event occurred in the lineage of Glycine max around 13 million years ago (Schmutz et al. 2010). The occurrence of additional Genistoid-specific polyploidy event(s) in the genus Lupinus is supported by variation in chromosome numbers, nuclear DNA content, duplication of isozyme markers and DNA markers, and duplicated genes in transcriptome and genome survey sequences (Naganowska et al. 2003; Nelson et al. 2006; Parra-Gonzalez et al. 2012; Wolko and Weeden 1989; Yang et al. 2013). However, Nelson et al. (2006) found little evidence of conserved genetic linkage between 58 pairs of duplicated markers which indicated that this additional polyploidy event(s) in the Lupinus lineage was likely to be ancient.

Comparative mapping and synteny analysis are powerful tools for evaluating evolutionary relationships among related taxa and for understanding the structural changes that differentiate the genomes of present-day species (Bertioli et al. 2009; Ellwood et al. 2008; Young and Bharti 2012). Genome conservation of related species has been demonstrated in many plant families with Poaceae, Brassicaceae and Solanaceae being well-known examples (International_Brachypodium_Initiative 2010; Lagercrantz and Lydiate 1996; Tomato_Genome_Consortium 2012). In recent years, a growing number of studies also demonstrated substantial conservation among legume genomes giving hope that genomic information gathered from model genomes can be successfully applied to crop legume improvement (Cannon et al. 2009; Choi et al. 2004; Zhu et al. 2005). However, the ability to transfer knowledge between species reduces as the extent of structural changes increases (Bertioli et al. 2009; Cannon et al. 2009). Since Lupinus diverged from the model legume species early in Papilionoid evolution, it is expected that the many structural changes will differentiate the Lupinus and model legume genomes. Although recent studies confirm the widespread synteny among legumes, the extent of conservation is still not well understood in the Genistoid clade including the economically and ecologically significant genus Lupinus (Lambers et al. 2013). Previous studies revealed short regions of conserved synteny between the L. angustifolius and two model species: Medicago truncatula and Lotus japonicus (Nelson et al. 2006, 2010). Similarly, the structure of the Lupinus albus genome appears to be highly rearranged relative to the M. truncatula genome (Phan et al. 2007). A low-density survey sequence of the L. angustifolius genome was recently reported with a small proportion of scaffolds assigned to linkage groups, but a synteny analysis to other legume genomes was not attempted (Yang et al. 2013). Previous synteny studies suffered from low resolution due to the low number of bridging points, which in case of Nelson et al. (2006) and Phan et al. (2007) was exacerbated by the use of a rudimentary M. truncatula reference genome sequence.

In the present comparative analysis, we employed an improved genetic map of the L. angustifolius genome with an almost threefold increase in the number of sequence-based genetic markers. In addition, an improved assembly of the M. truncatula genome (Young et al. 2011) was enlisted in this updated synteny analysis. The study not only improved the resolution of synteny between these two genomes, but also revealed compelling evidence of one or more ancient polyploidy events shaping the L. angustifolius genome.

Materials and methods

Mapping population

The mapping population used in this study consisted of 112 recombinant inbred lines (RILs), derived from a cross between a domesticated line (83A:476) and a wild type (P27255), and was developed at the Department of Agriculture and Food Western Australia (Perth, Australia). The parental lines were selected on the basis of contrasting phenotypes for domestication loci: Ku (reduced vernalisation requirement for flowering), iucundus (low alkaloid content), lentus and tardus (both for reduced pod shattering), mollis (water-permeable seeds) and leucospermus (visual marker for domestication conferring white flowers and reduced pigmentation in many tissues). This mapping population and phenotyping of the domestication traits and anthracnose resistance locus Lanr1 have been described in detail in previous mapping studies encompassing different subsets of RILs (Boersma et al. 2005, 2009; Li et al. 2011; Nelson et al. 2006, 2010; Yang et al. 2004).

DArT markers

Diversity array technology (DArT) marker assay was performed by Diversity Arrays Technology Pty Ltd (Canberra, Australia) according to Jaccoud et al. (2001) and the protocols described by Kilian et al. (2012). Briefly, a genomic representation of a mixture of DNA of both parental lines was generated after PstI–TaqI digestion (as a complexity reduction method) and distributed on microarray slide. The resulting array was used to genotype the fluorescently labelled RIL individuals, prepared by using the same complexity reduction method. A total of 299 polymorphic loci were scored as present (1) or absent (2) with the aid of dedicated software DArTsoft. Using the parental control samples, the scoring phase was determined for each locus and data converted to ‘A’ (maternal parent allele) and ‘B’ (paternal parent allele). The DNA sequences of 149 DArT non-redundant clones were determined by conventional Sanger sequencing and can be accessed at the GSS database of GenBank (accession numbers KG701214 to KG701371).

New PCR-based markers

New PCR-based STS markers comprised length polymorphism, SNaPshot (Life Technologies Inc., Foster City, USA), cleaved amplified polymorphic sequence (CAPS) or derived-CAPS (Neff et al. 2002). Table S1 details the primer sequences, assay details and parental amplicon sizes for new PCR-based markers. Parental amplicon sequences can be accessed at the GSS database of GenBank (accession numbers KG701372 to KG701468).

Six new gene-based STS primer pairs (mtmt_EST_03396, mtmt_DEG_03488, tRALS, mtmt_GEN_00650, SHK75 and mtmt_EST_03280) were developed within the Sixth European Union Framework Programme’s Grain Legumes Integrated Project (GLIP) and implemented in L. angustifolius using the method described by Nelson et al. (2010). Seven new intron-targeted STS primer pairs designed to amplify orthologous single or low copy genes LG1, LG11, LG13, LG16, LG75, LG96 and LG107) were provided by Prof. Richard Oliver and Dr. Simon Ellwood (Curtin University, Perth, Australia). Thirty gene-based legume anchor primer pairs (prefixed with “Leg”) were developed by Fredslund et al. (2006) and provided by Prof. Jens Stougaard (University of Aarhus, Denmark) (Table S1). The marker assay Gm56 targeting a candidate gene for pod shattering was provided by Dr. Varma Penmetsa (UC Davis, USA).

Eight flowering time gene homologue markers (dFTa, dFTb, dFTc, dSOC1, TFL1, VIN3, VIP3 and VRN1) were developed by designing primers in conserved regions identified by aligning legume ESTs with Arabidopsis thaliana flowering time genes with the aid of Vector NTI (Invitrogen, Carlsbad, California). The primer pair AC123593-13 was designed in the same way except using the M. truncatula BAC clone sequence AC123593 as the template. The resulting PCR amplicons from parents 83A:476 and P27255 were sequenced to confirm the successful amplification of the targeted genes and to identify SNP polymorphisms, which were then assayed in the RIL population using length, CAPS or dCAPS assays (Table S1). An additional flowering time gene homologue marker developed for pea was also used (VRN2; Hecht et al. 2005).

Genetic mapping

The previous version of the L. angustifolius reference map comprised 1,090 marker and trait loci (Nelson et al. 2010). Seven trait loci and the highest quality markers from that map were selected for this current study: 208 PCR-based sequence-tagged site (STS), 157 restriction fragment length polymorphism (RFLP) and 492 microsatellite-anchored fragment length polymorphism (MFLP) markers (864 loci in total; Table S1). When combined with the 353 new DArT and PCR-based STS markers, there were 1,217 loci in total, which were then subjected to linkage mapping.

Linkage mapping was performed with the aid of MultiPoint 2.1 (MultiQTL Ltd, Haifa, Israel), which uses the ‘evolutionary optimisation strategy’ (Mester et al. 2003) to perform multi-locus ordering of linkage groups. We used the approach described in detail by Nelson et al. (2010) and Raman et al. (2012) with some modifications. Briefly, redundant markers were set aside before commencement of clustering analysis. Iterative clustering analysis was conducted at a recombination frequency: rf = 0.12, 0.15 and 0.18 and then increased at 0.02 increments until a maximum of rf = 0.28. At each stage, multi-point analysis was conducted and resulting groups merged as rf was incrementally increased. Jack-knife analysis was performed on the rf = 0.28 linkage groups to identify markers that had a destabilising effect on locus order, which were then temporarily set aside. The remaining markers were used to construct the framework map with genetic intervals size transformed to account for multiple meioses involved in the development of the RIL population and expressed in Kosambi centiMorgans (cM). Redundant markers were then assigned to their representative framework markers and destabilising markers were assigned (or ‘attached’) to the most likely intervals between framework markers.

Synteny analyses

DNA sequences corresponding from the STS markers (PCR-based, RFLP and DArT) used to generate the genetic map of L. angustifolius were used to identify orthologous loci in the M. truncatula reference genome version Mt3.5.1 (Young et al. 2011) using BLASTn analysis with a minimum e-value of 1 e −5 and minimum bit score of 50. The two most significant matches were retained for synteny analysis.

Patterns of conserved locus order between the L. angustifolius and M. truncatula genomes were visualised using a custom ggplot2 script (Wickham 2009), Strudel (Bayer et al. 2011) and Circos v0.63 (Krzywinski et al. 2009). For Circos visualisation, the links were pre-processed with the “bundlelinks” utility script, grouping together markers mapped to the same L. angustifolius linkage group and occupying positions within M. truncatula genome spaced less than 100 kbp apart.

Results

An improved genetic map for Lupinus angustifolius

A new linkage map of narrow-leafed lupin genome incorporating new PCR-based and DArT markers was constructed with the aid of MultiPoint software. The map comprised 1,200 markers and 7 trait loci distributed over 20 linkage groups (NLL-01 to NLL-20) and three small clusters (Fig. 1; Fig. S1; Table S1). A further ten markers remained unlinked (Table S1). The new map was 2,345.7 cM in length with linkage groups ranging from 78 to 187.5 cM and the average spacing between unique framework markers was 2.95 cM (Table 1). The markers are well distributed with just 13 intervals exceeding 15 cM and 1 interval exceeding 20 cM (Fig. 1; Fig. S1). All linkage groups contained at least one marker from the major marker types (PCR-based STS, RFLP, DArT and MFLP). The map included 566 markers with associated DNA sequences that could potentially be used to align the genetic map of L. angustifolius with the reference genome sequence of the model legume M. truncatula.

Fig. 1
figure 1figure 1

Linkage map of the Lupinus angustifolius genome comprising 1,200 markers and seven trait loci distributed over 20 linkage groups (NLL-01 to NLL-20) and three small clusters (Cluster-1 to Cluster-3). Genetic distances are in Kosambi centiMorgans. Framework and redundant markers are presented here. A more detailed genetic map including attached markers is provided in Figure S1

Full size image
Table 1 Summary of the updated reference genetic map of Lupinus angustifolius comprising 20 linkage groups (NLL-01 to NLL-20) and three small clusters
Full size table

Synteny between Lupinus angustifolius and Medicago truncatula genomes

Synteny between the genomes of L. angustifolius and M. truncatula was assessed by comparing the positions of STS markers in the L. angustifolius genetic map with the positions of putatively orthologous sequences in the reference genome of M. truncatula. Of the 566 STS markers mapped in L. angustifolius, 410 (72.4 %) found one or more significant matches in the M. truncatula genome by BLASTn analysis. Of these, 16 markers were excluded from further analyses as they matched repetitive sequences within the M. truncatula genome leaving 394 markers for synteny analysis. Table S2 presents the BLASTn results of 394 markers against the two most significant matches in the M. truncatula genome. The primary match (i.e. the most significant) was considered the ‘best match’ (i.e. the most likely orthologous locus in the M. truncatula genome) for 367 (93.1 %) markers. For 27 (6.9 %) markers the second most significant match appeared to be the best match in the M. truncatula genome on the basis of conserved synteny and colinearity relative to neighbouring markers (Table S2).

Circos visualisation of genome-wide synteny revealed the high level of fragmentation of genome structure between L. angustifolius and M. truncatula with each M. truncatula chromosome sharing syntenic regions with two or more L. angustifolius linkage groups (Fig. 2). Dot-plot analysis was then used to examine these relationships in more detail (Fig. 3). At the segmental level, syntenic regions could be detected in all 20 linkage groups of L. angustifolius and all eight chromosomes of M. truncatula (Mt1–Mt8). The longest conserved block in the two genomes comprised 13 markers (NLL-09 and Mt5; Fig. 3). When evaluating 160 pairwise chromosome comparisons defined as 20 L. angustifolius × 8 M. truncatula chromosomes, there were 53 with at least 3 marker correspondences, indicating a high degree of syntenic fragmentation. Mt4 was the chromosome with the most correspondences (90) with L. angustifolius linkage groups, whereas Mt6 had the fewest (12). Detailed examples of selected regions of the L. angustifolius genome that showed extensive marker colinearity with Mt1, Mt4 and Mt7 are presented in Fig. 4. Four trait loci (tardus, lentus, Lanr1 and Ku) fell within, or adjacent to, conserved syntenic blocks (Fig. 4; Fig. S1). Interestingly, the flowering time gene homologue marker dFTc mapped to the same genetic location as the flowering time locus, Ku, with no recombination detected between the two loci (Fig. 4).

Fig. 2
figure 2

Circos plot of synteny between the Medicago truncatula reference genome and the Lupinus angustifolius genetic map. Lines linking syntenic regions are coloured according to M. truncatula chromosome. The chromosomes and linkage groups are not drawn to scale

Full size image
Fig. 3
figure 3

Dot plot of Lupinus angustifolius linkage groups vs. Medicago truncatula chromosomes. Markers within L. angustifolius linkage groups (x-axis) are presented in sequential order without scale. Positions of loci within M. truncatula chromosomes are drawn in base pair scale (y-axis)

Full size image
Fig. 4
figure 4

Examples of synteny between three linkage groups of Lupinus angustifolius (NLL-08, NLL-10 and NLL-12) and three chromosomes of Medicago truncatula (Mt1, Mt4 and Mt7). The pod shatter resistance gene, Lentus, is located on NLL-08, while the early flowering gene, Ku, is located on NLL-10. Names of loci showing syntenic regions are shown next to the L. angustifolius linkage groups (scaled in Kosambi centiMorgans, cM) and the start position of the homologous sequence on M. truncatula is presented next to each chromosome (scaled in megabases, Mb)

Full size image

Evidence of triplication in the lupin genome

Strikingly, there were several instances of large M. truncatula chromosome segments matching more than one L. angustifolius genomic region (Fig. 3), which were investigated further using Strudel visualisation. Figure 5 shows four of the clearest examples of M. truncatula chromosome segments that each matched two or three regions of the L. angustifolius genome. Figure 6 shows in detail one of the triplicated regions on L. angustifolius NLL-07, NLL-08 and NLL-13 which corresponded to M. truncatula chromosome Mt4.

Fig. 5
figure 5

Examples of duplication and triplication in the Lupinus angustifolius genome detected by comparison with the Medicago truncatula genome. M. truncatula chromosomes (Mt) are shown on the left of each panel and L. angustifolius linkage groups (NLL) on the right, with homologous loci in both genomes connected by lines. Chromosomes and linkage groups have normalised total lengths and drawn to scale within each chromosome (Mb scale) and linkage group (cM scale). Precise positions are presented in Table S2. a The top of Mt2 shares synteny with NLL-04 and NLL-16. b The bottom of Mt3 shares synteny with NLL-03, NLL-14 and NLL-15. c The middle of Mt4 shares synteny with NLL-07, NLL-08 and NLL-13. d The top of Mt5 shares synteny with NLL-06, NLL-17 and NLL-18

Full size image
Fig. 6
figure 6

An example of triplication in the Lupinus angustifolius genome (on linkage groups NLL-07, NLL-08 and NLL-13) relative to one Medicago truncatula chromosome (Mt4). Names of loci showing syntenic regions are shown next to the L. angustifolius linkage groups (scaled in Kosambi centiMorgans, cM) and the start position of the homologous sequence on M. truncatula is presented next to each chromosome (scaled in megabases, Mb). NLL-07 and NLL-08 are presented in reverse loci order and truncated (indicated by a slanted, thick bar)

Full size image

Discussion

The key finding of this study was the first ever observation of duplicated and triplicated regions in the L. angustifolius genome that were present as single copies in the M. truncatula genome (Figs. 3, 5, 6). Such duplications and/or triplications were observed on parts of most L. angustifolius linkage groups, which points to WGD or WGT arising from polyploidy event(s) rather than multiple independent duplications of single chromosomes. Not all of the L. angustifolius genome showed clear duplication and/or triplication; therefore, it is likely that the polyploidy event(s) was ancient and subsequently chromosomes have undergone numerous rearrangements. The conclusion of ancient polyploidy event(s) is supported by previously presented evidence based on diversified genome size and chromosome numbers within the genus, isozyme and DNA marker duplication, and duplicated genes in transcriptome and genome survey sequences (Naganowska et al. 2003; Nelson et al. 2006; Parra-Gonzalez et al. 2012; Wolko and Weeden 1989; Yang et al. 2013).

The timing of the inferred polyploidy event(s) remains an open question. Based on the synteny-based analysis of this study and the above-mentioned published studies, it appears that at least one polyploidy event took place after the divergence of Lupinus from M. truncatula and other Papilionoid legumes around 56 million years ago (Lavin et al. 2005). This question is being addressed in ongoing collaborative projects that are estimating the divergence times of homoeologous gene sequences obtained from transcriptomes of members of the Genistoid clade (C. Hughes, G. Aitcheson, D. Filatov and M. Nelson, unpublished data) and across the legume family (S. Cannon, unpublished data). These ongoing analyses should provide more robust inferences about the timing of polyploidy event(s) in Genistoid genome evolution. The identification of homoeologous gene pairs would be aided by the availability of a high-quality reference genome. The genome survey sequence of L. angustifolius recently reported (Yang et al. 2013) could not resolve gene duplications, highlighting the need for the high-quality reference genome sequence currently being developed for L. angustifolius (Gao et al. 2011). The current analysis will guide the sequence assembly of the L. angustifolius reference genome as it navigates the complexities of a polyploid genome. Owing to the accumulating genomic and transcriptomic data, L. angustifolius may soon become the model genome for the other Genistoid legume species.

The updated reference genetic map of L. angustifolius reported here comprised 1,207 loci, including 352 new, high-quality DArT and PCR-based STS markers. These new markers improved genome coverage with the number of unique framework loci increasing to 795 (Table 1) compared to 637 in the previous reference map (Nelson et al. 2010). The number of intervals >15 cM reduced from 18 (Nelson et al. 2010) to 13 (Fig. 1) in the current map. Intriguingly, three small clusters comprising markers generated by diverse PCR, RFLP and DArT technologies remained despite the increased marker density (Fig. 1). These clusters may represent ends of chromosomes that have low marker coverage, high recombination frequency, structural rearrangements, or a combination of the above. These chromosomal regions may present a challenge to incorporate in the more comprehensive L. angustifolius genome sequencing project that is currently underway (Gao et al. 2011) and may require additional cytogenetic analyses such as the BAC-FISH analysis developed by Lesniewska et al. (2011).

The most significant technical advance in the new map was the increased number of sequence-based markers (394) used as bridging points for comparing the L. angustifolius and M. truncatula genomes (Table S2). This was a substantial increase over the 147 bridging points used in the previous synteny analysis between L. angustifolius and M. truncatula (Nelson et al. 2006). This, along with the availability of an improved genome assembly of M. truncatula, permitted a higher-resolution analysis of synteny between the basal Papilionoid L. angustifolius with the model legume M. truncatula. While the overall impression of high differentiation between the two genomes remained unchanged (Fig. 2), there were much clearer examples of marker colinearity between the two genomes (Figs. 4, 6) compared to the previous study (Nelson et al. 2006). This improved delineation of conserved gene order in the basal Papilionoid genome of L. angustifolius will guide the reconstruction of the ancestral genomes of cool season and warm season currently underway (D. Cook, pers. comm.).

Four trait loci (tardus, lentus, Lanr1 and Ku) in the L. angustifolius genetic map fell within, or closely adjacent to, regions of conserved synteny with M. truncatula (Fig. 4, Fig. S1). For example, the pod shattering gene Lentus on linkage group NLL-08 was located in the synteny block shared with Mt1 (Fig. 4) as well as in a conserved block of Lotus japonicus chromosome 5 (Nelson et al. 2010). It was elsewhere reported that Mt1 and Lj5 show synteny along their entire lengths (Cannon et al. 2006), which brings further strength to the possible synteny exploitation for identification of a candidate gene for Lentus. The region of L. angustifolius NLL-10 containing the flowering time locus Ku shared synteny with the region of M. truncatula Mt7 containing three FT homologues, the floral integrator gene encoding the florigen signal protein in plants (Turck et al. 2008). Intriguingly, the FT-derived marker dFTc showed no recombination with Ku (Fig. 1; Table S1). Further research is underway to determine if FT is indeed the gene underlying the Ku locus or is instead another closely linked flowering time gene such as CONSTANS (Pierre et al. 2011). If FT is demonstrated to be the gene underlying the Ku locus, this would be a strong validation of the synteny approach for transferring genomic knowledge from a model genome to a less well-resourced crop genome.