Reticulate evolution in Panicum (Poaceae): the origin of tetraploid broomcorn millet, P. miliaceum

Summary Sequence data and genomic in situ hybridization identify the New World diploid Panicum capillare as an ancestral genome of the Old World tetraploid cereal broomcorn millet, P. miliaceum.


Introduction
Identification of the wild ancestors of crops is essential for understanding their evolutionary history, including during and after domestication changes (Hammer, 1984;Vaughan et al., 2007). About a third of plant species-both domesticated and wild-are polyploids (Wood et al., 2009). Identification of the ancestral genome donors of polyploid crops, including Panicum miliaceum, can allow resynthesis to broaden the genetic basis and identify novel adaptive genotype combinations.
The grass genus Panicum (Panicoideae, Poaceae) is variously circumscribed (see e.g. Clayton andRenvoize, 1986, Aliscioni et al., 2003). Interpreted broadly, it includes c. 500 pantropical and a few temperate species, with both C 3 and C 4 species. Several basic chromosome numbers have been reported, of which the most common are x=9 and x=10 (Aliscioni et al., 2003). A molecular phylogeny of primarily New World species has facilitated infrageneric classification, including the proposal that the genus Panicum should be restricted to a group of c. 100 species (Panicum sensu stricto), all of which have basic chromosome number x=9 (Aliscioni et al., 2003). The taxonomic position of many Old World species, and the relationships within the core Panicum clade that includes P. miliaceum L., remain unresolved (Triplett et al., 2012).
Panicum miliaceum was cultivated widely across Eurasia in prehistory (Jones, 2004;Hunt et al., 2011;Zohary et al., 2012), but is a minor cereal today in terms of global economic importance, with a global production of 5 Mt, about 1% of that of wheat or rice (calculated from statistics sourced from http://faostat.fao.org/site/567/DesktopDefault. aspx?PageID=567#ancor; last accessed 13 March 2013, and http://www.fao.org/docrep/W1808E/w1808e0l.htm#annex; last accessed 13 March 2013). However, proso millet remains a locally important staple and source of food security in semiarid regions where other cereals fail. The crop was introduced into North America in the 18th century (Dekker et al., 1981) and is now grown primarily for the high-value market niche as birdseed (Graybosch and Baltensperger, 2009).
Panicum miliaceum is known only as a tetraploid species (2n=4x=36) and has been suggested to be an allotetraploid as it has been found to show exclusive bivalent formation at meiosis (Hamoud et al., 1994). No wild tetraploid progenitor of domesticated P. miliaceum has been identified. Weedy forms, which may include the wild ancestor, are found across Eurasia, from Northeastern China to the Aralo-Caspian basin (Zohary et al., 2012), in Central Europe (Scholz, 1983;Scholz and Mikoláš, 1991), and in North America (Bough et al., 1986). Alternatively, these weedy forms may have arisen by back-mutation from the domesticate, as with fatuoid oats (Ellstrand et al., 2010;Gressel, 2010).
Phylogenetic relationships are widely studied using diverse nuclear and chloroplast DNA sequences and, for analysis of polyploids, in situ hybridization. 45S rDNA fragments are very widely applicable at all taxonomic levels (Hollingsworth et al., 2009;China Plant BOL Group, 2011), whereas single-or low-copy genes such as the floral regulatory FLO/LFY gene orthologue (Bomblies and Doebley, 2005), and the endosperm starch synthesis gene granulebound starch synthase I (GBSSI) have shown useful variation within groups of related taxa. Conserved chloroplast genes have also provided primer sequences of widespread taxonomic applicability in defining relationships and maternal ancestors in reticulate complexes (Hollingsworth et al., 2009), although in some lineages where hybridization is involved, plastid and nuclear phylogenies may be inconsistent (Fehrer et al., 2007). In hybrids and polyploids, genomic in situ hybridization has proved valuable to identify the genome donors and give indications about relationships between genomes or taxa such as Nicotiana (Patel et al., 2011), Leymus, and Psathyrostachys (Orgaard and Heslop-Harrison, 1994).
The present study aimed to examine relationships between five Panicum species using nuclear and chloroplast gene sequences and genomic in situ hybridization. Genomic identity was determined in two tetraploid species (P. miliaceum and P. repens) and tested the hypothesis that the diploid species P. capillare is one of the ancestral genome donors of P. miliaceum.

Materials and methods
Plant material and DNA extraction Germplasm samples were provided by the Vavilov Research Institute, St Petersburg, Russia, and by the USDA-ARS North Central Regional Plant Introduction Station, Ames, Iowa, USA. Details of the accessions used are given in Table 1. Genomic DNA was extracted from leaf tissue of single plants ground under liquid nitrogen using a modified CTAB protocol (Rogers and Bendich, 1994), or using a Qiagen DNeasy Mini Plant Kit according to the manufacturer's protocols.
Phylogenetic analysis of the FLO/LFY and GBSSI sequences included outgroups obtained from Genbank (accession numbers as in Fig. 1). Sequences for FLO/LFY and GBSSI in P. virgatum were obtained by BLAST searches of the P. virgatum genome sequence (Panicum virgatum v0.0, DOE-JGI, www.phytozome.net/panicumvirgatum accessed 13 March 2013). Exon sequences (following the annotation of the Genbank sequences) were aligned in MEGA5 (Tamura et al., 2011), using the Muscle algorithm for codons (Edgar, 2004) and optimized by hand.
For the ITS5 nuclear sequence analysis, a region spanning about 190 bp before the conserved 165 bp of the 5.8S rRNA gene and 202 bp after the gene, was obtained from each of 28 species of Panicum sensu lato, including 22 from Panicum sensu stricto, identified in Genbank (accession numbers as Fig. 1C) or from new sequences of PCR products amplified from P. repens and P. sumatrense using the primers 5ʹ-GGAAGGAGAAGTCGTAACAAGG-3ʹ and 5ʹ-TCCTCCG CTTATTGATATGC-3ʹ (China Plant BOL Group, 2011). Genbank accession numbers are HF934104 and HF934106. Sequences were aligned in MEGA5 (Tamura et al., 2011) using the Clustal algorithm, with manual optimization: inspection suggested that Clustal performed better than the Muscle algorithm for this sequence.
DNA sequence evolution models were tested for each of the FLO/LFY, GBSSI, and ITS5 sequence alignments in MEGA5 (Tamura et al., 2011), with complete gap deletion, which is appropriate for alignments with gaps spanning multiple nucleotides (see authors' guide to the software). The most likely sequence evolution model (that with the lowest Bayesian Information Criterion score-the criterion recommended in the authors' guide to the software) was used in each case to estimate the Maximum Likelihood tree. Statistical testing of the robustness of each topology was performed by bootstrap resampling of the sequence alignments, with the default program value of 500 replications. Bootstrap values for each interior branch of the original tree represent the percentage of resampled trees in which the branch is maintained.
The chloroplast DNA matK and rbcL sequences were obtained from Genbank for all available species of the genera Panicum (as recognized by the submitters, so generally sensu lato, the more distant species from P. miliaceum forming an outgroup) and Setaria. Sequences were aligned separately for each locus using Geneious R6 and manual optimization. For matK, a region of c. 1601 bp was analysed, comprising 1542 bp of the chloroplast-encoded group II intron maturase K flanked by 14 and 45 bp regions from the trnK-UUU transfer RNA sequence. A region of 590 bp within the chloroplast rbcL gene was analysed. FASTA alignments were converted to rdf format using Converter v.1.11 (Michael Campana, Boston, USA, pers. comm.). To visualize relationships between the chloroplast gene haplotypes, Median Joining networks (Bandelt et al., 1999) were constructed in Network 4.6 (http://www.fluxus-engineering.com/sharenet.htm).
In situ hybridization Seeds were germinated on filter paper moistened with bottled drinking water (Ashbeck, Tesco) in Petri dishes for c. 4 d until roots were 8-12 mm long. Before germination, husks were removed mechanically from seeds if necessary. Seedling root tips were pretreated with 8-hydroxyquinoline for 1 h at room temperature and the tubes were transferred to iced water for 10 h before fixation in fresh 3:1 ethanol:acetic acid. Chromosome spreads were made essentially following the methods in Schwarzacher and Heslop-Harrison (2000). Chromosome spreads were pretreated with pepsin (typically 10 μg ml -1 , 20 min at 37 °C). The 45S rDNA clone used was pTa71, and the 5S rDNA was pTa794, both originating from wheat. Genomic DNA and cloned probes were labelled with digoxigenin-11-dUTP or biotin-11-dUTP using the Invitrogen Bioprime CGH labelling kit or (for the plasmid pTa794) PCR labelling with M13 primers. In situ hybridization followed standard conditions. Typically, 120 ng of each labelled genomic DNA probe was used for each slide, and 0.5-3 μg of unlabelled DNA from P. miliaceum (or, for the P. repens slides, P. repens) to increase the specificity of hybridization of the probe. Chromosome preparations were denatured at 78 °C for 7 min and the final stringent washes were in 0.1 × SSC at 42 °C without formamide. Probe hybridization sites were detected with avidin conjugated to Alexa564-and anti-digoxigenin conjugated to FITC. After counterstaining with DAPI and mounting in Citifluor, images were collected with a ProgresC12 camera using exposure times of 10-40 s for FITC and Alexa564, and images were processed using Adobe Photoshop CS5 using only cropping, and functions that affected the whole image equally. Gaussian deblur functions were used where indicated.

FLO/LFY sequence analysis
The FLO/LFY primers amplified a fragment around 900 bp, with minor length variations both between and within Panicum species. Panicum repens and P. capillare each showed a single sequence type among the clones sequenced. Panicum miliaceum and P. sumatrense showed variation among clones that divided the clones into two distinct groups of sequences for each taxon, defined by the presence of particular indels and substitutions. Such variants could represent alleles on homologous chromosomes, homeologues from different parental genomes in an allopolyploid, and/or duplicated genes. Analysis of 10 accessions of P. miliaceum showed that the two FLO/LFY sequence types were present in all cases, suggesting the homologous-allele explanation is unlikely. Two sequences for P. virgatum (Pavirv00028202 and Pavirv00001023) were also found that showed high sequence similarity to other sequences in the dataset and are annotated as encoding FLO/LFY proteins. Two sequence types are also found in P. miliaceum for GBSSI. We have previously established (Hunt et al., 2010;Hunt et al., 2013) that these represent two homeologues of GBSSI. As for FLO/LFY, a single GBSSI sequence type was found for P. capillare, and two sequence types for P. virgatum (Pavirv00057044 and Pavirv00054240). In contrast to the FLO/LFY data, two GBSSI sequence types for P. repens were found, and only a single sequence type for P. sumatrense.
Each distinct sequence for each species was represented in the data matrices. The aligned matrices of 10 taxa (FLO/ LFY) and 9 taxa (GBSSI) were 741 and 1572 bp long, respectively. The resulting maximum likelihood trees are shown in Fig. 1. Both the FLO/LFY and GBSSI trees identify two pairs of sister taxa. Firstly, the single P. capillare sequence for each locus forms a clade with one of the two P. miliaceum sequences, with >99% bootstrap support in each case (for GBSSI, the homeologue previously designated GBSSI-L; Hunt et al., 2010.) Secondly, the other P. miliaceum sequence type ('GBSSI-S' at that locus) is sister to one of the P. repens sequences (the only P. repens sequence for FLO/LFY), though bootstrap support for this pairing is weak.
In the FLO/LFY tree (Fig. 1A), one of the two P. miliaceum (A), P. repens, and one of the two P. virgatum (Pavirv00028202) sequences form a clade with weak support, whereas the other P. virgatum (Pavirv00001023) sequence is more distantly related to the other Panicum sequences. The two P. sumatrense sequence types form a clade with relatively strong (81%) bootstrap support. In the GBSSI tree (Fig. 1B), the sole P. sumatrense sequence is sister to the second P. repens sequence (R).
In the ITS5 tree (Fig. 1C), P. miliaceum sequences and P. lanipes form a clade which is weakly recovered as sister to a clade containing P. capillare along with P. hallii and P. hirticaule, whereas P. repens, P. virgatum and P. sumatrense are dispersed on three other branches.

Chloroplast DNA haplotype analysis
The alignment of Panicum and Setaria sequences for the matK locus was 1201 bp in length, of which 119 sites showed variation; the rbcL alignment was 590 bp long, of which 28 sites were variable. The set of species for which sequence data are available differs between the two loci, but in both cases includes the five Panicum species that are the focus of this study and the outgroup taxon S. italica. The medianjoining networks for matK ( Fig. 2A) and rbcL (Fig. 2B) both show two clusters, one including P. miliaceum, P. capillare, P. sumatrense, and P. repens, and various other species which are (Aliscioni et al., 2003) or can be presumed to be Panicum sensu stricto. The number of nucleotide changes separating these four species is small (maximum 6). The P. repens and P. sumatrense haplotypes are identical for rbcL (also shared by P. paludosum and P. subalbidum), and separated by a single mutational step for matK. Two P. miliaceum haplotypes were recovered for rbcL, separated by one mutational step. One of these, shared with the species P. dichotomiflorum and P. flexile, was also a single base change from the P. capillare haplotype and from the P. repens-P. sumatrense-P. paludosum-P. subalbidum haplotype. In the matK network, the P. miliaceum and P. capillare haplotypes lie on their own branch, separated by two mutational steps, with the P. capillare haplotype lying two steps in from a node which is a further single step away from the P. sumatrense haplotype, and more distantly linked to haplotypes of several other species. Despite P. virgatum lying in Panicum sensu stricto, both matK and rbcL sequences are distant from other Panicum sensu stricto species (consistent with Triplett et al., 2012). Indeed, for rbcL, P. virgatum and other Panicum sensu lato species are not monophyletic with the main Panicum sensu stricto clade, relative to the Setaria species: this may reflect the limitations of this marker for taxonomic and phylogenetic resolution.

In situ hybridization
Chromosomes from P. miliaceum (2n=4x=36) were small, varying from 2.2-6.0 μm in length at metaphase (Fig. 3), and mostly submetacentric, with four conspicuous satellited NOR (nucleolar organizing region)-bearing chromosomes (Fig. 3J showing the 45S rDNA locations by in situ hybridization). There were no blocks of heterochromatin when viewed either in phase-contrast or following staining with DAPI. The P. repens accession (until 2012 the only one available from the NPGS/GRIN USDA-ARS germplasm collection) was tetraploid (2n=4x=36). Chromosomes were similar in morphology (Fig. 4) to P. miliaceum, with two NORs (Fig. 4G) labelling with 45S rDNA along most of the short arm of one pair of P. repens chromosomes. Panicum capillare (2n=2x=18, two NORs, not shown) and P. sumatrense (2n=4x=36, six NORs, not shown) chromosomes were generally similar in range of sizes and morphology to those of P. miliaceum.
To test the genome composition of the tetraploid species and the genomic relationships identified from the FLO/LFY, GBSSI, ITS5, and chloroplast DNA analyses, total genomic DNA was used as a probe to chromosomes of the tetraploid species P. miliaceum and P. repens. In metaphases of P. miliaceum (Fig. 3A-I) total genomic DNA of P. repens labelled a broad centromeric region in half (18) of the chromosomes, mostly slightly larger than the average chromosome size. Panicum capillare genomic DNA labelled the other half of the chromosomes with a more uniform signal, often with some exclusion from the centromere. No differentiation of P. miliaceum chromosomes was seen when probed with P. sumatrense (not shown), although technical limitations for a negative in situ hybridization result could not be discounted. Metaphases of P. repens were probed with genomic DNA of P. miliaceum, P. capillare, and P. sumatrense (Fig. 4A-F). Compared with the hybridization of genomic DNA to chromosomes of P. miliaceum, there was less differentiation of hybridization, and weaker signal. Panicum capillare DNA labelled 18 chromosomes slightly more strongly than the remaining chromosomes ( Fig. 4A-D, green). The P. sumatrense DNA probe (red, Fig. 4E, F) labelled the arms of 18 P. repens chromosomes more strongly, as well as centromeric regions of most chromosomes. Genomic P. miliaceum DNA (Fig. 4A-D red; E, F green) probed all chromosomes with some differentiation, perhaps with slightly stronger hybridization to about half the chromosomes. Both the P. sumatrense and P. miliaceum probes sometimes labelled rDNA sites, identified in DAPI images by their weaker staining and constricted morphology. Aliscioni et al. (2003) revised the infrageneric classifications within the genus Panicum using molecular and morphological data. Investigations within the section Panicum, particularly involving Old World species, are limited and are largely morphological. Most treatments have not taken into account the reticulate evolution, although it is clear that there is variation in ploidy within and between species, and that reticulate relationships demand further investigation (Triplett et al., 2012).

Discussion
The DNA sequence data (Figs 1 and 2) were used to suggest genome relationships in the group, which were then tested by in situ hybridization in the tetraploid taxa (Figs 4  and 5). Many of the nodes on the trees (Fig. 1) show low bootstrap support, which indicates that phylogenetic inferences are tentative. Plant molecular phylogenetics depends on markers with appropriate variation to resolve clades within the taxonomic group at the level of interest. This is particularly problematic for lower taxonomic levels (within genera and species; Dong et al., 2012). Hollingsworth et al. (2009) have recommended the 2-locus combination of rbcL and matK (the chloroplast markers used here, Fig. 2) for species identification (barcoding). The data in Fig. 2 show that these sequences perform reasonably well in enabling species discrimination, but do not enable robust phylogenetic inferences. It is labour-intensive to develop markers de novo for a given study that discriminate diagnostically between the set of taxa of interest. Aliscioni et al. (2003), using the chloroplast ndhF gene, were also unable to resolve many of the species or sections within Panicum. Nonetheless, the trees generate useful hypotheses for testing genomic relationships, reinforced by the relatively high congruity between the loci. Chloroplast haplotype minimum spanning networks were informative for suggesting female ancestors.
The nuclear DNA sequences and in situ hybridization results here suggest that the exclusively tetraploid (2n=4x=36) species Panicum miliaceum and the tetraploid accession of P. repens used here are both of allotetraploid origin. It has previously been established that genomic DNA from a tetraploid species or hybrid is able to distinguish the ancestral genomes when probed to its own metaphases: in both the tetraploids Nicotiana × sanderae and N. debneyi, there was differentiation of the two ancestral genomes (Patel et al., 2011) as we have shown here for both P. miliaceum and P. repens.
The 36 chromosomes of 4x P. miliaceum were labelled differentially along their length with genomic DNA from P. repens and P. capillare (Fig. 3), with some further differentiation in centromeric regions. As the signals differentiated half (18) of the chromosomes, we conclude that the probes are identifying whole chromosome sets (genomes) from the ancestral species. The hybridization and nuclear DNA results suggest that P. capillare, or a close relative, contributed one of the ancestral genomes. This extends the inference of genome identity between these taxa made by Triplett et al. (2012). The close relationship of the P. miliaceum and P. capillare matK haplotypes (Fig. 2) further suggests that P. capillare may have been the maternal parent, although the resolution provided by the chloroplast DNA sequence variation is not sufficient to be conclusive.
The inference of the New World native P. capillare as a diploid ancestor of the Old World native P. miliaceum is a surprising finding. The evidence that P. capillare is introduced in its Old World distribution seems to be largely circumstantial, and hangs on its continental (i.e. non-boreal) distribution in North America, south of the Great Lakes, its absence from Siberia and eastern Asia, and occurrence as a weed rather than in any natural habitat in Europe and Asia (Tom Cope, Kew, UK, personal communication). We can speculate that the maternal ancestor of P. miliaceum could have been a closely related Old World species, rather than P. capillare itself. However, current taxonomic knowledge of the genus is insufficient to determine which species are in the same clade as P. capillare. This highlights the need for germplasm collection and phylogenetic work on the Old World Panicum sensu stricto group and study of its biogeographical relationship to the New World group.
The similarity between the P. capillare and P. miliaceum L-homeologue GBSSI sequences is extremely high, not only in the exon sequence (99.3%) but also in the intron sequence (94.5%). In contrast, the intron sequences between the two P. miliaceum GBSSI homeologues cannot be aligned because of extensive indels (Hunt et al., 2013). This further emphasises the close relationship between the P. capillare genome and one of the P. miliaceum genomes. We have previously demonstrated evidence for post-polyploidization functional changes in the P. miliaceum GBSSI homeologues derived from P. capillare. The GBSSI gene in P. capillare produces a functional protein that synthesizes endosperm amylose. In P. miliaceum, the wild-type gene copy encodes a protein with reduced functionality, through unknown genetic or epigenetic mechanisms (Hunt et al., 2013). The other GBSSI homeologue in P. miliaceum is fully functional and determines the amylose-synthesis capacity.
The range of chromosome numbers reported in P. repens has led to previous speculation whether these represent an autopolyploid series or allopolyploid cytotypes (Majure, 2009). The allopolyploid origin of the tetraploid P. repens accession analysed here is supported by the differential labelling of two sets of chromosomes (Fig. 4), which show weak but consistent differentiation based on x=9. This is supported by DNA sequence (GBSSI) results. The data from three nuclear DNA sequences (Fig. 1) and in situ hybridization (Fig. 4) provide good evidence for P. sumatrense, or a close relative, contributing one of the two P. repens genomes and being the female ancestor donating the plastid sequences (Fig. 2). Both P. capillare and P. miliaceum label 18 chromosomes slightly more strongly, suggesting the other genome is more related to P. capillare than the other species tested (Fig. 4). The low differentiation of hybridization of the genomic DNA probes to all 36 chromosomes of P. repens suggests that two genomes are closer to each other than are the two P. miliaceum genomes (compare Fig. 3 and Fig. 4). Furthermore, only one FLO/LFY sequence type was found, whereas two GBSSI sequence types were present (Fig.1). Most tetraploid species in the Hordeeae Martinov (syn. Triticeae Dumort.) are allopolyploids, although a few have an autotetraploid origin (e.g. Dasypyrum breviaristatum; Galasso et al., 1997), and Labelled with P. capillare genomic DNA probe (green) and P. miliaceum probe (red) with weak differentiation of two sets of 18 chromosomes (interpreted as those from one ancestral genome (D). (E, F) Metaphase and adjacent interphase probed with genomic DNA of P. miliaceum (green) and P. sumatrense (red). Panicum sumatrense DNA (red) labels 18 chromosomes more strongly particularly in broad centromeric regions, whereas P. miliaceum (green) complements the P. sumatrense probe, with the other 18 chromosomes labelled slightly more strongly. Differentiated green and red regions are seen in the interphase nucleus (red stain precipitate present around the nucleus. (G) A metaphase and interphase hybridized with 45S rDNA (red) showing two major sites in both. Whereas differentiation of chromosome sets with the genomic DNA probe is not as strong as seen in P. miliaceum (Fig. 3), the result suggests P. repens is an allopolyploid and P. sumatrense is one of the ancestral genomes.
Further collections and analysis (including of modes of reproduction) of the range of 2x, 4x, n=9, and n=10 cytotypes reported from P. repens worldwide are needed to determine the taxonomic integrity of the taxon; the USDA/ARS National Plant Germplasm System GRIN (USDA, http:// www.ars-grin.gov/npgs/; last accessed 2010) lists no available accessions of P. repens, and only one of P. capillare, compared with 850 accessions for the wheat A genome ancestor Triticum monococcum ssp. aegilopoides. It is possible that taxonomic revision of P. repens may be required as with the recent revision of the model grass Brachypodium distachyon. Three cytotypes (2n=10, 2n=20, and 2n=30) were previously thought to represent an autopolyploid series within this single species. Novel cytogenetic and DNA sequence analysis showed that B. distachyon should be restricted to the 2n=10 cytotype. The 2n=20 cytotype is now recognised as a diverged diploid species (named B. stacei), and the 2n=30 cytotype as an allotetraploid (named B. hybridum) derived from B. distachyon and B. stacei (Catalán et al., 2012).
For both FLO/LFY and GBSSI, two distinct P. miliaceum sequences were identified. For the 45S rDNA sequence fragment spanning the 5.8S rRNA gene, our sequence and 180 other P. miliaceum sequences from Genbank (including 175 from Y. Xu, Jilin, China) were placed on a single well-supported branch (Fig. 1C). As there are two NOR loci (4 rDNA sites), one from each genome (Fig. 3G), this provides evidence for homogenization of the ancestral DNA sequences, and replacement of all copies by a variant similar to that in P. capillare. Wendel et al. (1995) have investigated 45S rDNA homogenization in tetraploid cotton, showing the replacement of the locus from one of the Gossypium ancestors of several tetraploid species including G. hirsutum and G. barbardense, leaving only one rDNA variant in the tetraploid. It seems likely that the P. miliaceum rDNA loci, as in cotton (Wendel et al., 1995), have homogenized by interlocus concerted evolution in the allopolyploid, and it is probable this involves the entire repeat.
In many major crop species, knowledge of the wild relatives and ancestors has proved valuable for introduction of new agronomic characters by broadening the genetic base available to breeders (e.g. Tester and Langridge, 2010;Heslop-Harrison and Schwarzacher, 2012). Furthermore, diversification of the number of species in agriculture is an important contribution to future food security, development of niche markets, and increasing dietary variation. Given the worldwide need for sustainable rain-fed agriculture and increasingly limited fresh water for irrigated agriculture, there is interest in improvement and increased exploitation of proso millet (Graybosch and Baltensperger 2009). Furthermore, in suitable climates and with added nitrogen and good weed control, growth of two crops per year gives increased food security, and is associated with less soil loss during fallow periods. The short time from planting to maturity makes proso millet a good candidate to introduce into such rotations, as a catch-crop, or when establishment of another crop fails. As well as abiotic stress resistance, the Panicum species show resistance to biotic stresses including mildew and thrips (data not shown) although, as with most minor crops, there are few systematic studies. Finally, there are also grain quality attributes such as absence of gluten and waxy starches. Thus as a grain crop, P. miliaceum has a range of useful genetic properties and gene alleles which, along with those of its wild relatives, are worthy of further characterization and exploitation in breeding programmes.