• Open Access

Molecular, phylogenetic and comparative genomic analysis of the cytokinin oxidase/dehydrogenase gene family in the Poaceae


(Tel +44 0 1223 342473; fax +44 0 1223 277602;
email james.cockram@niab.com)


The genomes of cereals such as wheat (Triticum aestivum) and barley (Hordeum vulgare) are large and therefore problematic for the map-based cloning of agronomicaly important traits. However, comparative approaches within the Poaceae permit transfer of molecular knowledge between species, despite their divergence from a common ancestor sixty million years ago. The finding that null variants of the rice gene cytokinin oxidase/dehydrogenase 2 (OsCKX2) result in large yield increases provides an opportunity to explore whether similar gains could be achieved in other Poaceae members. Here, phylogenetic, molecular and comparative analyses of CKX families in the sequenced grass species rice, brachypodium, sorghum, maize and foxtail millet, as well as members identified from the transcriptomes/genomes of wheat and barley, are presented. Phylogenetic analyses define four Poaceae CKX clades. Comparative analyses showed that CKX phylogenetic groupings can largely be explained by a combination of local gene duplication, and the whole-genome duplication event that predates their speciation. Full-length OsCKX2 homologues in barley (HvCKX2.1, HvCKX2.2) and wheat (TaCKX2.3, TaCKX2.4, TaCKX2.5) are characterized, with comparative analysis at the DNA, protein and genetic/physical map levels suggesting that true CKX2 orthologs have been identified. Furthermore, our analysis shows CKX2 genes in barley and wheat have undergone a Triticeae-specific gene-duplication event. Finally, by identifying ten of the eleven CKX genes predicted to be present in barley by comparative analyses, we show that next-generation sequencing approaches can efficiently determine the gene space of large-genome crops. Together, this work provides the foundation for future functional investigation of CKX family members within the Poaceae.


Cytokinins (CKs) are a group of plant hormones first identified as promoting cell proliferation and shoot formation. Subsequently, a long list of developmental and physiological functions under the control of CKs has been established, including cell differentiation, leaf expansion, apical dominance, reproductive development, seed germination, delay of senescence, de-etiolation, chloroplast differentiation, plant–pathogen interactions, as well as flower and fruit development (Sakakibara, 2006). A group of cytokinin-degrading enzymes, termed cytokinin oxidase/dehydrogenases (CKXs), are thought to control levels of CK, inactivating the phyto-hormone in a single enzymatic step by cleavage of the N6-substituted isopreneside chain to produce adenine and the corresponding aldehyde (Brownlee et al., 1975). CKXs are characterized by the presence of two conserved regions within their proteins: the flavin adenine dinucleotide (FAD)-binding and cytokinin-binding domains, both of which are essential for fully functional enzymatic activity (Malito et al., 2004). Investigations in various plant species suggest that CKXs respond differently to various stresses and that their expression may be tissue specific (Vaseva-Gemisheva et al., 2004). Recently, positional cloning of the rice yield quantitative trait locus (QTL) Gn1a has shown it to be controlled by OsCKX2, encoding a CKX (Ashikari et al., 2005). Natural OsCKX2 genetic variants were shown to confer increased grain number per panicle, leading to a 21% increase in grain number between the parental line and a near isogenic line harbouring the contrasting OsCKX2 allele. The positive effects of OsCKX2 on grains per panicle are due either to an exonic deletion resulting in truncation of the OsCKX2 protein or to reduced OsCKX2 expression.

Rice, a major provider of global human nutrition, is a member of the grass family (Poaceae). The Poaceae is divided into around twelve subfamilies (with species from the Ehrhartoideae, Panicoideae and Pooideae subfamilies investigated here) and includes other major global crop species such as bread wheat (Triticum aestivum), barley (H. vulgare) and maize (Zea mays). The Poaceae also includes many minor or regionally important crops, such as sorghum (Sorghum bicolour) and foxtail millet (Setaria italica), as well as the model Pooideae species, brachypodium (Brachypodium distachyon). Since their divergence from a common ancestor ∼60 million years ago (mya) (Salse et al., 2008), members of the Poaceae have undergone major changes in the size and structural arrangement of their genomes. For example, while the genomes of brachypodium (355 Mbp), rice (430 Mbp), foxtail millet (515 Mbp) and sorghum (750 Mbp) have remained relatively small, those of barley (5500 Mbp) and bread wheat (16 000 Mbp) are an order of magnitude larger. One of the factors driving increased genome size is polyploidization: while barley (2n = 2= 14), sorghum (2n = 2x = 20), brachypodium (2n = 2x = 10), foxtail millet (2n = 2x = 18) and rice (2n = 2x = 24) are diploid, bread wheat is hexaploid (2n = 6x = 42), with the three subgenomes originating from diploid A, B and D progenitor genomes. Although maize (2500 Mbp) behaves genetically as a diploid with ten pairs of chromosomes (2n = 2x = 20), it actually represents an allotetraploid, resulting from the hybridization of two slightly diverged progenitors ∼5 mya (Wei et al., 2007). Another major factor affecting genome size is the rate of accumulation of repetitive sequence, because of differences in transposable element (TE) activity and redaction/elimination (Bennetzen et al., 2005). The high TE frequency in wheat, barley and maize has been a major factor in the inflation of their genome sizes. Despite these substantial structural changes that have taken place during the evolutionary radiation of the Poaceae, comparative analyses have shown high degrees of colinearity between grass genomes (Gale and Devos, 1998). The ability to ‘cross-reference’ Poaceae genomes provides a framework within which advances in genetic understanding in one species can potentially be transferred to related grass species. This is of particular relevance to barley and wheat, whose large genomes present a major obstacle for the creation of whole-genome sequence assemblies. The ability to align multiple cereal genomes has led to the comparative mapping (and in many cases, map-based cloning) of series of orthologous loci controlling related traits within the Poaceae. Examples include the colinear VRN-1 and VRN-2 flowering time loci in diploid wheat (Triticum durum), tetraploid wheat (Triticum durum), and hexaploid wheat (T. aestivum), barley (H. vulgare) and rye (Secale cereale) (reviewed by Cockram et al., 2007), the photoperiod response PPD-1 loci in barley (Turner et al., 2005) and bread wheat (Beales et al., 2007) and genes controlling anthocyanin in maize and barley (Goff et al., 1992; Cockram et al., 2010a). More recently, the availability of high-density genetic maps and genome sequence assemblies has led to the identification of whole-genome duplication (WGD) events in rice (Salse et al., 2008), brachypodium (International Brachypodium Initiative, 2010), sorghum (Paterson et al., 2009), maize (Wei et al., 2007), bread wheat (Salse et al., 2008) and barley (Thiel et al., 2009), estimated to have occurred ∼20 million years prior to their divergence from a common ancestor (Paterson et al., 2004). After the ancestral WGD, structural rearrangements are thought to have led to diploidization and stabilization of the genome (Paterson et al., 2004). Taken together, inter- and intra-species colinearity provides the evolutionary context with which to investigate the genetic control of related traits within the Poaceae, allowing efficient exploitation of natural and artificial variants within orthologous genes from related grasses.

Several studies have investigated Poaceae CKX family members to varying extents in one or more grass species: Ashikari et al., 2005 (rice CKX family); Bilyeu et al., 2001 (ZmCKX1); Galuszka et al., 2004 [barley and wheat CKX expressed sequence tags (ESTs)]; Houba-Hérin et al., 1999 (ZmCKX1); Schmülling et al., 2003 (rice CKX family, ZmCKX1, HvCKX9); Vyroubalováet al., 2009 (Maize CKX genes, 2008 assembly); Zhang et al., 2010 (rice CKX family, ZmCKX1-3, TaCKX2.1, TaCKX2.2). However, although the cloning of OsCKX2 has highlighted the potential importance of this gene family in plants, a systematic analysis of the CKX family in grasses has been lacking to date. As a basis from which informed investigation of cereal CKX members can be undertaken, we describe the identification, molecular characterization, comparative mapping and phylogenetic analysis of CKX gene families in the Poaceae, with particular emphasis on wheat and barley OsCKX2 orthologs.


Identification of CKX genes in sequenced cereal genomes

Previous studies identified 11 CKX members in rice (Ashikari et al., 2005). To verify copy number in the current rice genome assembly, CKX coding regions (CDS) were used for BLASTn searches of Michigan State University (MSU) Rice Genome Annotation Release v6.1 (http://rice.plantbiology.msu.edu/). As expected, 11 gene models were identified (Table 1). Of these, OsCKX6 and OsCKX7 represent tandemly duplicated genes on chromosome Os2, while OsCKX1 and OsCKX2 are located within ∼570 kb of each other on Os1. The primary predicted splice forms were subsequently used for BLASTn analysis of related grass species.

Table 1.   Rice CKX gene family members, and their homologues identified in the sequenced genomes of brachypodium, sorghum, foxtail millet and maize. Full-length barley CKX genes identified in this study are also shown
GeneChr (Mbp)Stra-ndGene modelGenomic (bp)cDNA (bp)Exo-nsProtein (aa)e-value (% identity)*
  1. *BLASTn comparison between the gene and the closest rice homologue.

  2. Truncated genes.

  3. Re-analysis of gene models SiPROV034880m and SiPROV019828m.

  4. §Oregon Wolfe Barley Recessive allele, FGENESH gene model prediction, utilizing alternative intron 1 splice site.

  5. N/A, not applicable; CKX, cytokinin oxidase/dehydrogenases.

Oryza sativa
 OsCKX11 (4.70)Os01g09260.1247915993532N/A
 OsCKX21 (5.30)Os01g10110.1557616984565N/A
 OsCKX310 (18.20)+Os10g34230.1480915845527N/A
 OsCKX41 (41.30)+Os01g71310.1284215905529N/A
 OsCKX51 (32.79)Os01g56810.1426516055534N/A
 OsCKX62 (6.68)+Os02g12770.1228315063501N/A
 OsCKX72 (6.69)+Os02g12780.1182915753524N/A
 OsCKX84 (26.01)+Os04g44230.1450515995532N/A
 OsCKX95 (17.97)+Os05g31040.1214915665521N/A
 OsCKX106 (22.20)Os06g37500.1293115874528N/A
 OsCKX118 (22.62)Os08g35860.1294918634620N/A
Brachypodium distachyon
 BdCKX12 (4.06)Bradi2g05580.12835156635210.0 (73)
 BdCKX22 (4.53)Bradi2g06030.13435169535640.0 (66)
 BdCKX33 (31.06)Bradi3g29130.13528156655210.0 (90)
 BdCKX42 (57.80)Bradi2g60456.11933157255230.0 (88)
 BdCKX52 (51.12)Bradi2g51530.13838165055490.0 (84)
 BdCKX63 (6.40)+Bradi3g08220.11359135914524e−179 (71)
 BdCKX73 (6.41)+Bradi3g08230.11599159915320.0 (75)
 BdCKX8a5 (19.49)+Bradi5g16080.11781114943827e−122 (66)
 BdCKX8b5 (19.49)+Bradi5g16090.1175843821452e−66 (74)
 BdCKX92 (25.97)Bradi2g27170.12055156655210.0 (87)
 BdCKX10a1 (33.31)+Bradi1g37470.11903161735380.0 (73)
 BdCKX10b2 (3.78)+Bradi2g05210.139939911326e−42 (73)
 BdCKX113 (40.25)Bradi3g37690.12580154245130.0 (80)
Sorghum bicolor
 SbCKX13 (3.35)+Sb03g003280.12465159035290.0 (81)
 SbCKX23 (2.72)+Sb03g002810.14634173425770.0 (77)
 SbCKX31 (19.99)Sb01g019000.14462158765280.0 (86)
 SbCKX43 (72.68)+Sb03g045410.12160157255230.0 (85)
 SbCKX53 (64.20)Sb03g036160.13538164755480.0 (85)
 SbCKX64 (8.09)Sb04g007730.11863161435370.0 (77)
 SbCKX7a4 (8.12)+Sb04g007740.11846160555340.0 (78)
 SbCKX7b4 (8.13)+Sb04g007745.11501501494e−126 (78)
 SbCKX86 (52.15)+Sb06g022930.14869159355300.0 (81)
 SbCKX99 (46.59)+Sb09g018640.11954156955220.0 (82)
 SbCKX1010 (50.61)Sb10g022590.11817165025490.0 (77)
 SbCKX117 (57.12)Sb07g022530.12965156345200.0 (83)
Setaria italica
 SiCKX13 (2.46)+SiPROV006865m2069156935220.0 (81)
 SiCKX23 (1.73)+FGENESH model‡4240166535540.0 (79)
 SiCKX31 (15.84)SiPROV006823m3649157255230.0 (87)
 SiCKX43 (44.77)+SiPROV006863m2378156955220.0 (86)
 SiCKX53 (37.14)SiPROV006381m4286161755380.0 (87)
 SiCKX64 (3.24)+SiPROV033306m1836161435370.0 (78)
 SiCKX74 (3.25)+SiPROV027697mg1772146444880.0 (82)
 SiCKX86 (25.33)+SiPROV0027343m1707124544150.0 (82)
 SiCKX99 (14.87)+SiPROV006866m2087156955220.0 (83)
 SiCKX1010 (34.36)SiPROV028586m1867162035390.0 (79)
 SiCKX117 (29.83)SiPROV006958m3143155735180.0 (83)
Zea mays
 ZmCKX13 (7.59)GRMZM2G146644>4377163275434e−158 (81)
 ZmCKX28 (22.25)GRMZM2G3256123620175535840.0 (75)
 ZmCKX31 (230.01)GRMZM2G4044433223162935420.0 (79)
 ZmCKX4a3 (152.20)GRMZM2G0509971929156055190.0 (84)
 ZmCKX4b8 (159.19)GRMZM2G1672201982157855250.0 (77)
 ZmCKX4c4 (226.39)+GRMZM2G1704464054105683512e−154 (86)
 ZmCKX5a8 (172.36)+GRMZM2G0244763519160555340.0 (85)
 ZmCKX5b3 (191.84)GRMZM5G817173207694253130.0 (88)
 ZmCKX6a4 (230.02)GRMZM2G1144271825162035390.0 (76)
 ZmCKX6b4 (230.11)GRMZM2G1346341825162035390.0 (76)
 ZmCKX74 (230.19)GRMZM2G1620481803160535340.0 (77)
 ZmCKX8a10 (131.7)+GRMZM2G122340797680432670.0 (81)
 ZmCKX8b2 (27.55)GRMZM2G0087924540158755280.0 (82)
 ZmCKX109 (80.49)GRMZM2G3037072747206635680.0 (77)
 ZmCKX111 (207.71)+GRMZM2G3484522775157845250.0 (84)
Hordeum vulgare
 HvCKX2.1§3HN/AHvCKX2.14087170435530.0 (77)
 HvCKX2.23HN/AHvCKX2.23145169535640.0 (77)
 HvCKX7N/AN/AHvCKX71602160215330.0 (77)
 HvCKX9N/AN/AHvCKX91955158155260.0 (84)

Searches of the brachypodium genome (evolutionarily the most closely related to rice of the sequenced grasses analysed here) identified 11 CKX genes, with predicted CDS of 1149–1695 bp (Table 1). In addition, two putative CKX pseudogenes were identified, as assumed by severe truncations in their predicted proteins: the first, BdCKX8b (Bradi5g16090.1, CDS of 438 bp), is a tandem duplication of BdCKX8a (Bradi5g16080.1, CDS of 1149 bp). The second shows sequence similarity to BdCKX10 and is located on chromosome Bd2 (Bradi2g05210.1, CDS of 399 bp). BLASTn analysis of the sorghum genome also identified 11 CKX genes, plus one CKX pseudogene (Sb04g007745.1, CDS of 150 bp) (Table 1). The latter is one of a cluster of three tandemly duplicated genes, colinear with clusters of two CKX genes in rice and brachypodium (CKX6, CKX7). Searches of the foxtail millet genome identified 12 CKX family members (Table 1). Further analysis of two adjacent gene models with high sequence similarity to OsCKX2 (SiPROV034880m and SiPROV019828m) showed they are homologous to the 5′ and 3′ ends of the rice gene. Extraction of the genomic region spanning these two genes, and re-analysis using the gene prediction software FGENESH, resulted in a single-gene model prediction (Figure S1), suggesting an error in the foxtail millet gene annotation. The updated gene model was used in all subsequent analysis, resulting in a total of 11 SiCKX family members. Searches of the current maize genome assembly (v. 5b.60) returned 12 CKX family members, as well as three pseudogenes (CKX4, CKX5 and CKX8) (Table 1). Two maize homologues were identified for CKX4 and CKX6. Single copies were found for CKX1, CKX2, CKX3, CKX7, CKX8 and CKX11.

Although no genome assembly is currently available for barley, BLASTn searches of genomic sequence data from our shotgun next-generation sequencing (NGS) of barley cv. Morex (Nils Stein, unpublished data) identified 14 contigs ranging in size from 1319 to 8641 bp, representing putative homologues of ten (OsCKX1-5, 7-11) of the eleven rice CKX family members (Table S1A). For the majority of CKX genes, single contigs with partial gene coverage were identified. However, OsCKX1 identified two barley contigs, homologous to the 5′ (contig_1005647) and 3′ (contig_1010998) ends of OsCKX1 (3e-159 and 7e−159, respectively). Similarly, four barley contigs with high sequence similarity (≤7e−129) to OsCKX2 were identified. A partial imperfect overlap between contig_47875 and contig_104715 suggests these represent the 5′ and 3′ ends of the same gene, while the remaining discrete contigs (contig_1005571 and contig_2164028) were hypothesized to represent 5′ and 3′ genomic fragments of a second CKX2-like gene in barley. Of the 14 contigs identified, two were predicted to contain a full-length CKX genes: contigs 1021400 and 1008654 (displaying high sequence similarity to OsCKX7 and OsCKX9, respectively) (Table 1). Finally, although 5× sequence coverage of the hexaploid wheat genome (cv. ‘Chinese Spring’) has very recently been made publicly available (http://www.cerealsdb.uk.net/), our analyses find that sequence depth is currently insufficient to provide detailed bioinformatic analysis of the CKX2 genes (data not shown). However, we utilize alternative approaches [EST databases and bacterial artificial chromosome (BAC) library screening] to identify OsCKX2 orthologues in wheat.

Identification of CKX ESTs in the transcriptomes of wheat and barley

Extensive publicly available transcriptome sequence databases are available for wheat and barley, containing ESTs from many different tissues, developmental stages and growth regimes. The grouping of ESTs into transcript assemblies (TAs) provides an invaluable resource for the identification of genes and gene families in large-genome species. Here, the eleven rice OsCKX CDS were used to identify homologues in wheat and barley databases using BLASTn analyses. Homology was verified by back-BLASTn comparison to the rice genome. Putative wheat homologues of OsCKX1, OsCKX2, OsCKX3, OsCKX4, OsCKX5 and OsCKX11 were identified, while searches of barley TAs identified homologues for OsCKX1, OsCKX4, OsCKX5, OsCKX9 and OsCKX11 (Table S2). The ESTs were derived from cDNA libraries constructed from a range of tissues, including callus, roots, leaves and kernels, with 2–8 ESTs identified within each homologous CKX cluster. Although CKX2 homologues were not identified in barley, the four wheat CKX2 ESTs all originated from developing kernel tissue (GenBank accessions BQ235927, BQ238832, BQ903062 and CA705202). In addition, two full-length TaCKX2-like cDNA sequences were identified: TaCKX2.1 (FJ648070) and TaCKX2.2 (GU084177), both of which have recently been cloned by 5′ and 3′ rapid amplification of cDNA ends (RACE) using primers designed from EST CA705202 and coarsely mapped to a deletion bin on chromosome 3D (Zhang et al., 2010).

Molecular characterization of CKX2 homologues in wheat and barley

Among the four wheat ESTs homologous to OsCKX2, two ESTs (BQ238832 and BQ235928) originated from the same cDNA (TaE05040F02), representing forward and reverse sequences spanning the OsCKX2 start and stop codons. To determine the relationship of this wheat cDNA with the two TaCKX2-like genes described by Zhang et al. (2010), the cDNA was sequenced by primer walking. We found the cDNA to be 1865 bp (GenBank accession JF293079), with CDS of 1638 bp, displaying high sequence similarity to OsCKX2 (e = 0.0, identities = 78%). Alignment of the CDS with TaCKX2.1 and TaCKX2.2 (Zhang et al., 2010) showed sequence identity of 84% and 85% identity, respectively, but clearly indicated its origin from a different CKX2 gene (Figure S2). Accordingly, the CKX2 gene represented by cDNA TaE05040F02 has been designated TaCKX2.3. To obtain the complete genomic TaCKX2.3 sequence, as well as the surrounding genomic context for comparative analyses, primers were designed for PCR screening of a subset of a wheat BAC library, identifying four positive BAC clones. HindIII digests of BAC DNAs found two clones to overlap with each other (1991B18 and 2018M8), while the remaining two clones (470M29 abd 400N24) represented nonoverlapping genomic regions (data not shown). Accordingly, we fully sequenced BAC clones 1991B18, 470M29 and 400N24, finding them contain inserts of 157, 128 and 164 kb, with a mean GC-content of 44%, 47% and 69%, respectively. For clone 400N24 (GenBank accession JF292901), four predicted gene models [400.1 (TaCKX2.3), 400.2, 400.3, 400.4] were identified, all of which were found within a gene island of ∼30 kb (Table 2). Alignment of cDNA and genomic sequences found TaCKX2.3 CDS to be arranged over three exons of 646, 422 and 588 bp, with no sequence polymorphisms identified between the coding regions of TaCKX2.3 CDS and gene 400.1. The positions of introns 1 and 2 are perfectly conserved between OsCKX2 and TaCKX2.3, whereas the third intron found in OsCKX2 is absent from TaCKX2.3 (Figure 1). BAC clones 1991B18 and 470M29 each contained a single predicted gene model, annotated TaCKX2.4 (3 exons, CDS = 1659 bp) and TaCKX2.5 (3 exons, CDS = 1638 bp), respectively.

Table 2.   Predicted genes on sequenced wheat TaCKX2 bacterial artificial chromosome clones 400N24 (JF292901), 470M29 (JN381556) and 1991B18 (JN381555)
Wheat geneCDS (bp) [exons]Corresponding TIGR TA/EST singletonOrthologous rice gene modelRice annotationTa versus Os predicted CDS: e-value (% identities)
  1. TA, transcript assembly; Ta, wheat; Os, rice; CDS, coding regions.

400.11656 [3]JF293079LOC_Os01g10110.1OsCKX20.0 (78)
400.2861 [6]TA111747_4565LOC_Os01g10100.1Expressed protein5e−157 (88)
400.31935 [1]NoneLOC_Os01g10090.1Pentatricopeptide0.0 (82)
400.41335 [7]TA81187_4565LOC_Os01g10070.1Expressed protein2e−54 (72)
1991.11659 [3]NoneLOC_Os01g10110.1OsCKX20.0 (78)
470.11638 [3]NoneLOC_Os01g10110.1OsCKX20.0 (78)
Figure 1.

 Diagram of predicted cytokinin oxidase/dehydrogenases 2 intron/exon structure. Exonic sequence conservation surrounding intron/exon boundaries is shown. Species: barley (Hv), wheat (Ta), brachypodium (Bd), rice (Os), sorghum (Sb), maize (Zm) and foxtail millet (Si). The position of the 32-bp deletion within HvCKX2.1 in cv. Oregon Wolfe Barley Recessive is indicated by the triangle, resulting in a premature stop-codon (indicated by a star) relative to cvs. Morex and Oregon Wolfe Barley Dominant.

As no OsCKX2 homologues were identified in barley EST databases, the partial CKX2-like NGS barley genomic contigs were used to identify full-length HvCKX2 gene sequences from cvs. Oregon Wolfe Barley Recessive (OWB-R) and Oregon Wolfe Barley Dominant (OWB-D) by PCR amplification and primer walking. This resolved the four NGS contigs into two genes, designated HvCKX2.1 and HvCKX2.2, with predicted CDS of 1704 and 1695 bp, respectively (Table S1B). Both barley CKX2 genes were predicted to contain three exons located at conserved positions relative to OsCKX2, and like all other Poaceae CKX2 genes investigated here, lack the third intron found in OsCKX2 (Figure 1).

HvCKX2.1 (cv. Morex), HvCKX2.2 (cv. OWB-D), TaCKX2.3, TaCKX2.4 and TaCKX2.5 (cv. Chinese Spring) encode predicted proteins of 567, 556, 551 552 and 545 amino acid residues, respectively (Figure S3). Protein domain analyses predicted all five proteins to possess FAD-binding (≤6.2e-20) and cytokinin-binding (≤1e-108) domains, which are common to all known CKX proteins. Alignment of wheat/barley CKX2 proteins to OsCKX2 showed that while the positions of the predicted stop codons are conserved across the species, the N-terminus shows variation in length (Figure S3. Of the seven known wheat and barley CKX2 proteins, TaCKX2.3 is the most similar to OsCKX2 (identities = 69%), followed by TaCKX2.1/TaCKX2.4/TaCKX2.5 (68%) and HvCKX2.1/HvCKX2.2 /TaCKX2.2 (67%).

Genetic mapping of CKX2 homologues in barley and wheat

Rice chromosome Os1 is colinear with the Triticeae group 3 chromosomes, with the short arms colinear with each other between both clades (Devos, 2005; Salse et al., 2008; Thiel et al., 2009). To support the putative orthology of Triticeae CKX2-like genes with OsCKX2, we genetically mapped the identified OsCKX2 orthologues in wheat (TaCKX2.3, TaCKX2.4, TaCKX2.5), and barley (HvCKX2.1 and HvCKX2.2). TaCKX2.4, TaCKX2.5 and TaCKX2.3 were mapped to chromosomes 3A, 3B and 3D, respectively, using gene-specific primers to amplify products from the nullisomic/tetrasomic (NT) aneuploid series developed from cv. Chinese Spring (Sears, 1966) (Figure 2a). Sequencing TaCKX2.3 PCR products amplified from the wheat varieties ‘Avalon’ and ‘Cadenza’ identified a synonymous single nucleotide polymorphism (SNP) within exon 1 (G + 1084/T). This allowed intra-chromosomal mapping of TaCKX2.3 to chromosome 3D, 2 cm distal to SSR markers GWM456/GWM52, and 28 cm proximal to amplified fragment length polymorphism (AFLP) marker stm02TCAC (Figure 2b). A nonsynonymous SNP in TaCKX2.4 (C + 893/T, exon 2, Ala → Val) was identified in the ‘Opata 85’ × ‘synthetic’ parental lines, placing it 4 cm distal to SSR marker gwm32. Finally, a synonymous SNP within TaCKX2.5 (G + 234/A) was identified between the parental lines of the ‘Helidur’ × ‘TD161′ population, mapping this gene to chromosome 3B between markers gwm533 and gwm285. Alignment of all three mapped TaCKX2 genes to the wheat consensus genetic map (Somers et al., 2004) shows they are located on the short arms of the group 3 chromosomes, close to the nondistal low-recombining regions within which the centromeres are predicted to be located.

Figure 2.

 Genetic mapping of cytokinin oxidase/dehydrogenases genes in wheat and barley. (a) Inter-chromosomal mapping of wheat genes TaCKX2.3, TaCKX2.4 and TaCKX2.5 to the group 3 chromosomes (using gene-specific primers to amplify products from the wheat nullisomic/tetrasomic chromosome addition lines). Deleted chromosomes indicated. W = water control. (b) Intra-chromosomal mapping of TaCKX2.3, TaCKX2.4 and TaCKX2.5 (using the ‘Avalon’ × ‘Cadenza’, ‘Opata 85’ × ‘synthetic’ and ‘Helidur’ × ‘TD161’ populations, respectively), aligned to the consensus genetic map of Somers et al., 2004;. (c) Genetic mapping of barley CKX genes in the ‘Arta’ × ‘H. spontaneum’ population. (d) Genetic mapping of barley CKX genes in the ‘Oregon Wolfe Barley Dominant’ × ‘Oregon Wolfe Barley Recessive’ populations, allowing integration into the consensus map of Close et al., 2009. Colinearity between rice physical maps and homologous barley EST-based markers positioned in the barley consensus genetic map is indicated. HvCKX2.1 and HvCKX2.2 cosegregate with each other on chromosome 3H. No polymorphism was identified for HvCKX1.

In Barley, sequencing HvCKX2.1 PCR amplicons from the parental lines of the OWB mapping population (Costa et al., 2001) identified six polymorphisms between OWB-D (JF293078) and OWB-R (JF293074). Using a 32-bp insertion/deletion (InDel) within exon 1, HvCKX2.1 was mapped to the short arm of chromosome 3H, cosegregating with SNP markers 1_1102, 3_0583 and 3_0721 at 56.4 cm. Full-length sequencing of HvCKX2.2 in OWB-D (JF293076) and OWB-R (JF293077) identified ten polymorphisms (3 exonic, 1 nonsynonymous) across ∼3.2 kb. Genetic mapping of a selected SNP (T + 1583/C, intron 2) showed HvCKX2.2 to cosegregate with HvCKX2.1 on chromosome 3H. Integration of HvCKX2.1 and HvCKX2.2 into the barley consensus map (Close et al., 2009) showed these genes to be located at 50.2, ∼2 cm proximal to the low-recombining region of chromosome 3H predicted to span the centromere (Figure 2c). Of the 455 gene-based genetic markers distributed across 173.2 cm on the 3H consensus map, 75% identify homologous genes on the colinear rice chromosome Os1 (based on highest-scoring matches from BLASTn analyses). A scatter plot of markers on the barley 3H genetic map (cm) versus the rice Os1 physical map (Mbp) shows that HvCKX2 orthologs fit in well with the overall conservation of gene order between the two species (Figure 2c). Indeed, two of the genetic markers found to cosegregate with HvCKX2.1 in the OWB population are orthologous to genes on rice chromosome Os1, 3_0721 (LOC_Os01g10450.1) and 1_1102 (LOC_Os01g11810.1), and are located just 240 and 1120 kb from OsCKX2 (LOC_Os01g10110.1), respectively. In addition to the genetic mapping of the six CXK2 orthologues in wheat and barley, we determined the genetic map positions of eight of the remaining nine barley CKX family members identified in our analysis of the draft genome assembly (Figure 2, mapped polymorphisms listed in Table S3). Using the ‘Arta’ × ‘H. spontaneum’ (A × S) population (Baum et al., 2003), HvCKX3 (42 cm) and HvCKX9 (56 cm) mapped to the short and long arms of chromosome 1H, respectively, while HvCX8 mapped to 95.2 cm, between marker 11_11388 and 11_10786 on the long arm of chromosome 2H. HvCKX4 (131 cm in the A×S population) and HvCKX5 (85.5 cm in the OWB population) both mapped to 3H, while HvCKX7 mapped to the nondistal region of low genetic recombination predicted to span the 6H centromere (cosegregating with 11_10659 at 64.4 cm in the OWB population). Finally, HvCKX10 and HvCKX11 both mapped to the long arm of chromosome 7H in the OWB population, cosegregating with markers 11_20896 (169.3 cm) and 11_10896 (128.3 cm), respectively. No polymorphism was identified for HvCKX1 in the amplicons and mapping populations investigated.

Comparative genomic analysis of cereal CKX gene family

Using the established inter-species grass comparative framework for rice, brachypodium, barley, sorghum, foxtail millet and maize (Devos, 2005; International Brachypodium Initiative, 2010; Paterson et al., 2009; Salse et al., 2008; Schnable et al., 2009; Thiel et al., 2009) (Figure 3), as well as the confirmation of micro-colinearity around each CKX gene (Figures 2d and 4, Table S4), cross-species comparison showed that the identified full-length grass CKX members map to colinear genomic locations. However, chromosome-specific rearrangements within the maize genome have resulted in deletion/disruption of almost all of the duplicated CKX genes arising because of allotetraploidization, with the exceptions of maize orthologues of CKX4 and CKX8 (Figure 3, Table 1).

Figure 3.

 Diagrammatic representation of the position of cytokinin oxidase/dehydrogenases (CKX) genes within the framework of Poaceae inter- and intra-species colinearity. Duplicated segments of the rice genome originating from the ancestral whole-genome duplication (WGD) event are shown in the centre, surrounded by inter-specific genome colinearity between other grass species (adapted from Devos, 2005; International Brachypodium Initiative, 2010; Schnable et al., 2009). The positions of CKX genes are shown (truncated genes indicated by semi-circles), with solid lines linking colinear genes, and dashed lines linking colinear genes involved in the ancestral WGD event. OsCKX8 and OsCKX3 lie within a segmentally duplicated region not involved in the WGD, as determined by our methodology (grey dashed line). Telomeres on the short and long arms are indicated by red and black triangles, respectively. Pseudo-genes ZmCKX4b and BdCKX10b are shown in blue and yellow, respectively.

Figure 4.

 3-D scatter plot illustrating micro-colinearity between physical regions spanning CKX6/7 and CKX10 in sequenced grass species, highlighting regions of segmental duplication thought to have arisen because of the whole-genome duplication (WGD) event in the ancestral grass genome. For each scatter plot, the physical positions of rice (Os) genes in the CKX6/7 region are plotted against the x-axis (the ‘reference genome’). For each scatter plot, orthologous genes at colinear physical positions at the CKX6/7 regions in sequenced cereals (the ‘target genomes’) are plotted against the y-axis, resulting in plots along the horizontal x-y-axis: from left to right: brachypodium (Bd), sorghum (Sb), foxtail millet (Si) and maize (Zm, for which two graphs are shown, because of its allotetraploid genomic structure). The positions of paralogous genes in ‘target genomes’ around the CKX10 region (thought to have arisen because of the WGD event in the ancestral cereal genome) are plotted against the z-axis. Thus, where a homologous gene is present at all three genomic locations, the plotted point is boosted from the horizontal axis, according to its position on the z-axis. The positions of such genes, shared between all three chromosomal regions, are shown in blue. CKX genes are circled in red. Where tandem duplications occur (as is the case for CKX6/CKX7), only the gene with highest homology between species is illustrated. Colinearity in maize is generally less well conserved, likely due to extensive genomic modifications post tetraploidization. CKX, cytokinin oxidase/dehydrogenases.

Duplications within the Ehrhartoideae (rice; Paterson et al., 2004; Salse et al., 2008; Yu et al., 2005; Thiel et al., 2009), Panicioideae (sorghum, maize; Paterson et al., 2009; Wei et al., 2007) and Pooideae (barley, brachypodium, wheat; International Brachypodium Initiative, 2010; Salse et al., 2008; Thiel et al., 2009) originated from a WGD in the putative ancestral cereal genome (Bolot et al., 2009; Paterson et al., 2004; Salse et al., 2008). The most recent rice genome assembly was used to update the WGD blocks within the rice genome (Figure 3). Combining inter- and intra-species colinearity duplications indicated the extent of shared WGD in the remaining grass species relative to the genomic locations of CKX genes (Figure 3). In addition, WGD-derived blocks of colinearity were verified at the micro-colinearity level by analysis of genes surrounding each rice gene (Table S4, Figure 4). Two CKX gene pairs (CKX4/CKX9 and CKX7/CKX10) appear to have evolved as a result of the WGD. In addition, OsCKX3 and OsCKX8 are segmentally duplicated within a small region of 26 gene pairs on rice chromosomes Os10 and Os4 (http://rice.plantbiology.msu.edu/). Although OsCKX1, OsCKX2, OsCKX5 and OsCKX11 are located within regions of intra-specific duplication, no paralogous genes were found in the corresponding chromosomal locations in rice (or any of the other sequenced grass genomes), indicating these may have been lost in the ancestral genome prior to the divergence of the modern-day grass species.

With the aim of clarifying the orthology of TaCKX2.3 with other CKX2-like genes, a 154-kb region of rice chromosome Os1 containing ten genes flanking OsCKX2 on either side was used to investigate the extent of micro-colinearity with the sequenced physical regions of brachypodium, sorghum, foxtail millet, maize and wheat (BAC clone 400N24). Conservation of gene/gene order varied across the species investigated: rice/brachypodium and brachypodium/sorghum shared the highest number of orthologous gene pairs (12), while the lowest conservation was observed between maize chromosomes Zm8 and Zm3, which shared just one gene (Figure 5, Table S5). Gene orientation is conserved in all but one (GRMZM2G014284, maize chromosome Zm3) of the 63 genes conserved between ≥2 species. The four genes on the wheat TaCKX2.3 BAC are conserved in all species, with the exception of maize Zm3, which possessed just one orthologous gene. As expected, the physical sizes of colinear regions varied depending on species: those in rice, brachypodium and sorghum were the smallest, and broadly equal in size (∼120 kb), followed by foxtail millet (170 kb) and maize (450 kb, Zm8). Although the increase in physical size in colinear maize chromosomal regions is predominantly because of the increased abundance of TEs, these regions also contain a greater proportion on noncolinear genes (Figure 5). All of the diploid sequenced cereals investigated possessed a single CKX2 gene at colinear positions.

Figure 5.

 Comparative analysis of physical maps around cytokinin oxidase/dehydrogenases 2 (CKX2) loci in cereal genomes. Sequenced wheat bacterial artificial chromosome clones for TaCKX2.3 (chromosome 3D, JF292901), TaCKX2.4 (chromosome 3A, JN381556) and TaCKX2.5 (chromosome 3B, JN381555) are included. Dashed lines link gene models with significant sequence similarity to genomic regions which lack predicted genes (grey circles). CKX2 genes are highlighted in red. Chromosome orientations (direction to the long arm telomere) are indicated.

Phylogenetic analysis of cereal CKX proteins

To help clarify the evolutionary relationships of the CKX gene family in grasses, a phylogenetic tree was constructed based on protein sequences from the conserved FAD and cytokinin-binding protein domains (Figure 6). After exclusion of truncated CKX family members (Bradi2g05210, Bradi5g16090, Sb04g007745, GRMZM2G817173, GRMZM2G170446, GRMZM2G122340), a total of 66 full-length cereal CKX proteins were analysed, originating from rice (11 members), brachypodium (11), sorghum (11), foxtail millet (11), maize (12), wheat (six) and barley (four). CKX proteins from sequenced grasses grouped into four major phylogenetic clusters, with clades I and II each divided into two subclades. The seven full-length wheat/barley CKX2 proteins are found within clade Ia, with TaCKX2.3 clustering with HvCKX2.1 and HvCKX2.2, while TaCKX2.1 and TaCKX2.2 group separately. The occurrence of two or more CKX2 paralogs is specific to the Triticeae. The overall tree topography of CKX2 proteins agrees with established evolutionary relationships between the grass species investigated, with those from the Poideae (wheat, barley, brachypodium) clustering together, followed by members from the Ehrhartoideae (rice) and Panicoideae (foxtail millet, sorghum and maize). CKX1 proteins from all five sequenced grasses cluster within the second major phylogenetic grouping within clade Ia, supporting their close evolutionary relationship with the CKX2 proteins (as implied by their physical proximity to each other in the five sequenced grass species investigated). Clade Ib includes orthologs of OsCKX6 and OsCKX7, as well as OsCKX10 orthologs. CKX4 and CKX9 proteins group within clade IIa. Clades III and IV contain two (CKX3, CKX8) and one (CKX11) CKX family member, respectively. Phylogenetic analysis of barley and wheat CKX members for which only partial protein sequence data were available showed all members to group within the expected clades (Figure S4).

Figure 6.

 Phylogenetic tree of all 66 full-length cytokinin oxidase/dehydrogenases 2 (CKX2) family members from rice, brachypodium, sorghum, maize and foxtail millet, as well as full-length members from wheat and barley. CKX2 sequences identified in this study are highlighted in bold. TaCKX1 = Feng et al. (2008). Bootstrap proportions (1000 replicates) ≥0.70 are indicated.


Gene duplication, because of genomic processes such as WGD and polyploidization, provides the raw materials upon which mutation, selection and genetic drift result in the evolution of gene families. Following duplication, functional redundancy means that one or more paralog may become silent/nonfunctional because of the accumulation of deleterious mutations or be lost because of other localized genomic rearrangements. This is especially true of CKX family members in the allotetraploid species maize, where in all but two cases, genetic mutation has resulted in the elimination or severe truncation of paralogous CKX genes since the allotetraploidization event ∼5 mya. This agrees with previous reports documenting frequent loss of one duplicated gene (e.g. Bruggmann et al., 2006). Evidence for the loss of paralogous grass CKX genes assumed to have originated during the ancestral grass WGD event is also apparent, with comparative mapping indicating gene loss prior to speciation. Alternatively, paralogous genes may be retained, after which they often diverge in terms of their temporal/spatial expression (subfunctionalization) or function (neo-functionalization) (Wei et al., 2007). Although OsCKX2 is the only member within the grasses for which natural intra-specific variation has been proven to affect phenotype, simultaneous analysis of genomic, phylogenetic, molecular and transgenic data from the Poaceae (and arabidopsis) can provide insights into the evolution and possible function of grass CKX2 orthologs and other Poaceae CKX family members.

Clade Ia

Phylogenetic and sequence analyses show CKX1 and CKX2 to be closely related, grouping within clade Ia. Although they are physically linked in all sequenced grass species investigated, their depth of phylogenetic divergence is the greatest amongst all subclades. Our findings indicate that this local gene duplication arose within the ancestral grass genome, prior to the WGD event. Their older origins, relative to family members duplicated during the ancestral WGD, provide greater evolutionary time for sub- or neo-functionalization to potentially occur. In the grasses, CKX1 genes are expressed in a wide variety of tissues, with TaCKX1 expressed in leaves, sheaths and roots (Feng et al., 2008), HvCKX1 in the roots, leaves and grain (Galuszka et al., 2004; Zalewski et al., 2010) and ZmCKX1 in kernels and the vascular system of seedling shoots, roots and coleoptiles (Brugière et al., 2003; Galuszka et al., 2005; Smehilova et al., 2009). Genetic mapping resolved the rice grain number locus, Gn1, into two closely linked QTL of roughly equal effect: Gn1a (OsCKX2) and Gn1b (Ashikari et al., 2005)While Gn1a is controlled by OsCKX2, Gn1b remains to be identified, mapping distal to marker R3192. OsCKX1, which phylogenetic analysis shows to be closely related to OsCKX2, is located ∼410 kb distal to R3192 and therefore represents a possible candidate for Gn1b. This hypothesis is supported by the recent finding that RNAi silencing of HvCKX1 in barley results in higher grain yield, as well as root weight and height (Zalewski et al., 2010).

Clade 1a: CKX2 orthologs

The following findings indicate true OsCKX2 orthologs have been identified in the six grass species investigated: (i) high levels of sequence conservation at the DNA/protein level, (ii) conservation of intron/exon structure, (iii) close phylogenetic clustering and (iv) macro-/micro-colinear physical/genetic map locations. As OsCKX2 belongs to a gene family of eleven members, it has been suggested that spatial and temporal differences in expression may have contributed to functional differentiation (Ashikari et al., 2005; Zalewski et al., 2010). Indeed, tissue- and organ-specific expression has been demonstrated for arabidopsis CKX genes using promoter:GUS fusions (Werner et al., 2003), with AtCKX enzymes differing in their biochemical properties and subcellular localization (Schmülling et al., 2003). In rice, OsCKX2 expression is low. It is predominantly expressed in the culms, inflorescence meristems and spikelets and was not represented in rice EST libraries prior to map-based cloning (Ashikari et al., 2005). Despite large EST libraries in the species investigated here (ranging from 20 323 accessions in brachypodium to 1 032 224 in maize), evidence of low CKX2 transcript abundance is mirrored in wheat (4 CKX2 ESTs), barley (0), brachypodium (1, DV475865), sorghum (0) and maize (1, CO452786), with all of these transcripts originating from ear/grain tissue. Furthermore, the recently characterized genes TaCKX2.1 and TaCKX2.2 have been found to be expressed predominantly in the wheat culm and spike, (Zhang et al., 2010).

The identification of a Triticeae-specific CKX2 duplication that is likely to have occurred prior to the divergence of wheat and barley ∼11 mya (Huang et al., 2002) suggests the potential phenotypic effects of Triticeae CKX2 variants may be masked, because of functional redundancy. Together with the occurrence of natural rice OsCKX2 gene deletions that result in increased grain number, these findings indicate that intra-specific CKX2 copy number variation may be of importance to agronomic traits in cereals. Indeed, copy number variation of paralogous CONSTANS-like genes that arose during the ancestral WGD event has previously been shown to play a critical role in the agronomic attributes of related grass crops (Cockram et al., 2010b). The establishment of high Triticeae CKX2 copy number will inform future functional validation using reverse genetic approaches, favouring RNA interference (RNAi) approaches over TILLING (McCallum et al., 2000), as combining closely linked TILLed genes in a single genetic background would be problematic. The finding that HvCKX2.2 possesses an exonic deletion that is predicted to result in the truncation of the cytokinin-binding domain in cv. OWB-R indicates EcoTILLING (Comai et al., 2003) may prove fruitful, at least in diploid grasses. Despite the identification of the truncated HvCKX2.2 allele, no grain number QTL overlapping its genetic map position has been reported in the OWB population (Chutimanitsakun et al., 2011), suggesting that HvCKX2.2 is functionally redundant with HvCKX2.1, or that it does not control grain number in barley. Interestingly, analysis of twelve wheat varieties, Zhang et al. (2010) reported an association between increased TaCKX2.1/TaCKX2.2 expression and increased grain number per spike. However, this is opposite to the effects on floral meristem development observed in rice and arabidopsis, where overexpression of CKX2 genes results in reduced floret/flower primordia establishment, while null variants and transgenic plants expressing OsCKX2 antisense constructs display increased grains per panicle (Ashikari et al., 2005). Antagonsitic effects of arabidopsis CKX genes between different tissues have also been observed, in which silencing of CKX genes stimulated shoot growth and repressed root development (Werner et al., 2003). We note that as well as local function in the tissues within which they are synthesized, CKs act as long distance signalling messengers (Kudo et al., 2010) and are able to establish or modulate sink strength (Argueso et al., 2009; Kuiper, 1993; Guivarc’h et al., 2002; Werner et al., 2008). As plant development combines the effects of cell proliferation/expansion with whole plant sink/source balance plant (Vercruyssen et al., 2011), it has been suggested that the observed species-specific differences in the physiological effects of CKX silencing could be due to differences in sink strength (Zalewski et al., 2010). To clarify the function of CKX2 in wheat, an RNAi silencing approach has been initiated to investigate its role in planta.

Clade 1b

Clade Ib contains two to three tandemly duplicated CKX genes (OsCKX6 and OsCKX7 in rice), as well as CKX10 members in all species investigated, with the exception of barley, where no OsCKX6 orthologue was identified. The presence of variable numbers of genes orthologous to the OsCKX6/OsCKX7 cluster in multiple grass species indicates this genomic region is prone to localized genomic rearrangements such as unequal crossing-over (Achaz et al., 2000), which act to increase or decrease copy number of tandemly duplicated genes. In rice, brachypodium, sorghum and maize, CKX10 orthologs are located within regions of intra-specific colinearity with the CKX6/CKX7 cluster, indicating they are related by evolutionary descent after arising as part of the ancestral WGD event. While the lack of ESTs for any of the clade Ib genes in rice, wheat and barley indicates they display low or tissue/treatment-dependent expression, RT-PCR expression analysis in rice found very low OsCKX6 and OsCKX10 transcription in the shoot apical meristem (SAM) (Ashikari et al., 2005).

Clades II–IV

Poaceae CKX4 and CKX9 members all cluster within clade IIa and are likely to have originated because of the ancestral grass WGD event. Little is known about these members, although HvCKX9 has been found to be expressed in the grain and 7-day-old leaves (Galuszka et al., 2004). The deep phylogenetic roots of clades III and IV suggest that they have distinct evolutionary histories relative to clades I and II. Within clade III, grass CKX3 and CKX8 members group together, despite the genomic regions in rice not having been identified as participating in the recent WGD event (based on nonoverlapping regions of duplication and similar Ks distribution). However, these regions are nevertheless involved in a small region of segmental duplication (http://rice.plantbiology.msu.edu/), indicating that they share a common evolutionary origin. This suggests a local duplication, or remnants of an older WGD event, that occurred before the evolutionary divergence of the Poaceae. However, analysis of micro-synteny around CKX3 and CKX10 genes in other sequenced grass species found no further evidence of such duplications, indicating further investigation is needed to establish the evolutionary timing of this event. Finally, clade IV contains just one member (CKX11), which has previously been phylogenetically grouped with AtCKX7 (Zhang et al., 2010), which encodes a nonsecretory cytosolic enzyme in arabidopsis (Kowalska et al., 2010). Although little is known about clade IV members, genomic/cDNA Triticeae CKX11 homologues have been identified here, indicating that CKX11 is expressed in diverse tissues in wheat/barley (root, leaf, grain), while in rice, it is relatively highly expressed in the SAM (Ashikari et al., 2005).

Future prospects

Studies of CKX3 and CKX5 double mutants in arabidopsis show that CKs delay cellular differentiation in the apical meristem and terminal floral cellular differentiation, as well as a promotion of placental growth, leading to an increased seed set per silique (Bartrina et al., 2011). The resulting increase in cytokinin content was associated with an increase in seed yield of over 50%, illustrating the role of sink strength as a yield factor. Recent findings in arabidopsis are also beginning to shed light on the interaction of CKs with other developmental pathways that regulate growth and biomass. Crosstalk between CKs and the brassinosteroid, gibberellin and cell proliferation pathways have been found to affect biomass accumulation in leaves and roots (Vercruyssen et al., 2011). Similarly, the role of CKs in the control of flowering time has been shown to act via transcriptional regulation of the FLOWERING LOCUS T (FT) paralog, TWIN SISTER OF FT (TSF), as well as FD and the floral pathway integrator, SUPPRESSOR OF EVEREXPRESSION OF CONSTANS1 (SOC1) (D’Aloia et al., 2011), allowing integration of CKs into the otherwise relatively well-defined floral molecular pathway. These studies highlight the potential impact that detailed genetic and physiological investigation of the effects of CKX genetic variation may have in crop species. The work presented here provides the molecular and phylogenetic basis from which systematic evaluation of grass CKX gene family function can be undertaken, allowing further understanding of the roles CKs play in the modulation of sink strength, biomass and yield in cereal crops, and the basis from which yield gains comparable to those grained in rice could potentially be achieved in other Poaceae crops. Finally, by identifying what is most likely the complete barley CKX gene family, we demonstrate that NGS shotgun sequencing of the nuclear genome provides an efficient method of determining the gene space of large-genome crop species.

Experimental procedures

Bioinformatic identification and analysis of CKX genes in cereals

CKX copy number in the rice (Oryza sativa ssp. japonica cv. Nipponbare) genome (MSU Osa1 assembly v6.1, http://rice.plantbiology.msu.edu/) was verified by BLASTn (Altschul et al., 1997) searches (match/mismatch scores = 2,−3; gap costs: existence = 5, extension = 2) using OsCKX1 to OsCKX11 CDS (Ashikari et al., 2005). An expectation-value threshold (the number of times that a match, or a better match, occurs by chance within the database) of 1e−40 was used, and the results compared to those reported by Ashikari et al. (2005). While e-values of zero are reported here, these actually represent positive e-values too small for the precision of the software used. Identification of CKX genes within the genome assemblies of sequenced cereals was conducted using OSCKX1 to OsCKX11 CDS for BLASTn searches against the gene models of B. distachyon accession BD21 (assembly v1.0, using sequence data produced by the US Department of Energy Joint Genome Institute, http://modelcop.org/), S. bicolour cv. Moench (v1.0, http://phytozome.net/), Z. mays (cv. B73, v5b.60, http://www.maizesequence.org/) and S. italica cv. Yugu1 (assembly v1.0, using sequence data produced by the US Department of Energy Joint Genome Institute, http://phytozome.net/). Where necessary, gene models were reassessed using FGENESH (http://www.softberry.com/). Barley (H. vulgare) CKX genomic sequence fragments were identified by BLASTn searches of rice CKX CDS versus a 28× NGS genomic survey of cv. Morex (Stein et al., unpublished). In addition, identification of CKX genes within the transcriptomes of barley and wheat (T. aestivum) was conducted using OsCKX1 to OsCKX11 as queries for BLASTn searches of TIGR TA databases (release 2.0, http://www.tigr.org/). To clarify homology groupings, hits for all species were used for back-BLASTn analysis of the rice genome (i.e. used as queries for BLASTn searches of the rice genome determine whether they showed highest homology to the original CKX gene query). Predicted protein sequences were obtained using the VectorNTI Advance package v10.1.1 (Invitrogen, Paisley, UK), and protein domains identified using pfam (http://pfam.sanger.ac.uk/). Inclusion criteria for CKX proteins identified in species with sequenced genomes were the presence of the FAD and cytokinin-binding domains. Protein alignments were performed using CLUSTALW (Thompson et al., 1994), with manual adjustment/editing using GENEDOC v2.6 (http://www.nrbsc.org/gfx/genedoc/) and Jalview v2.6.1 (Waterhouse et al., 2009). CKX genes are prefixed with the corresponding genus and species initials. CKX numbering follows orthology with rice CKX genes (Ashikari et al., 2005). Accordingly, the nomenclature of barley/wheat CKX genes differs to that listed in Galuszka et al. (2004).

Molecular characterization of CKX2 homologues in wheat and barley

Plasmid DNA was extracted using the Qiaprep Kit (Qiagen, Crawley, UK) and the insert sequenced by primer walking (details listed in Table S6), initiated with generic T7 and SP6 primers. Sequencing was performed using BigDye kit v3.1 (Applied Biosystems) and the resulting extension products separated and visualized using a 3730 DNA Analyzer (Applied Biosystems). Sequence editing and assembly was performed using Vector NTI. Inter-chromosomal mapping of TaCKX2 genes was performed using the wheat NT chromosome deletion lines (Sears, 1966). Genomic DNA was extracted using the DNeasy 96 Plant Kit (Qiagen) and used as a template for PCR amplification of TaCKX2 genes using Pfu-turbo (Agilent, Edinburgh, UK) and primers listed in Table S6. PCR products were separated by agarose gel electrophoresis. Amplicon specificity was verified by direct sequencing of PCR products, as described by Chiapparino et al. (2006). Mapped SNPs were identified by direct sequencing of PCR products amplified from parental DNAs of the mapping populations listed below, using primers TaCKX2.3-F4/TaCKX2.3-R4, TaCKX2.4-F1/TaCKX2.4-R1 and TaCKX2.5-F1/TaCKX2.5-R1 (Table S6). Intra-chromosomal mapping of TaCKX2 genes was conducted by fluorescent single-base extension using the ABI Prism SNaPshot Multiplex Kit (Applied Biosystems) as previously described (Cockram et al., 2010b). TaCKX2.3 was mapped in the Avalon × Cadenza doubled-haploid mapping population (http://www.wgin.org.uk/), genotyped using amplification primers TaCKX2.3-F4/-R4 and single-base extension primer TaCKX2.3-SNAPR. TaCKX2.4 was mapped in the ‘Opata 85’ × ‘synthetic W7984′ population (http://wheat.pw.usda.gov/ggpages/maps.html/) using amplification primers TaCKX2.40F1/TaCKX2.4-R1 and extinction primer TaCKX2.4-SNAPF. TaCKX2.5 was mapped in the ‘Helidur’ × TD161′ (RAGT, unpublished data) using amplification primers TaCKX2.5-F1/TaCKX2.5-R1 and extension primer TaCKX2,5-SNAPF. Details of SNPs used for mapping are shown in Table S7. All genetic mapping was performed using JoinMap v3.0 (Van Ooijen and Voorrips, 2004). TaCKX2 primers designed from fully sequenced EST clone TaEo5040F02 were used to PCR screen a subset of the 4× cv. Chinese Spring T. aestivum BAC library (http://cnrgv.toulouse.inra.fr/), using methods described by Alfares et al. (2009). BAC clones were sized and fingerprinted as described by Foote et al. (2004) and selected clones fully sequenced (completed as a service provided by Washington State University). BAC sequence annotation was completed following the protocols of Sabot et al. (2005).

HvCKX2.1 and HvCKX2.2 genomic sequences were obtained using barley NGS contigs to design primers for PCR amplification/primer walking (Table S6). All additional barley genomic contigs are part of a CLC assembly (http://www.clcbio.com/) using 2×730 683 996 sequences (2 × 101 bp; total 147.6 Gb; ∼28× genomic coverage) from three barley (cv. Morex) Illumina-Paired-End libraries (500 bp distance). Genetic mapping was conducted using the Oregon-Wolfe barley (OWB) mapping population (Costa et al., 2001). A 32-bp InDel identified between the parental lines OWB-D and OWB-R was used to genotype the mapping population, by separation of PCR products (amplified using primer pair HvCKX2.1-F4/-R4) on ethidium bromide stained 1% agarose gels. SNP T + 1583/C was used to genetically map HvCKX2.2 by fluorescent single-base extension using the SNaPshot Multiplex Kit in combination with amplification primers HvCKX2.2-F3/-R3 and single base-pair extension primer HvCKX2.2-SNAPF. Genetic mapping was performed using JoinMap v3.0, and map positions integrated into the barley consensus map (Close et al., 2009). Additional barley CKX genes were genetically mapped using the OWB or ‘Arta’ × ‘H. spontaneum’ (Baum et al., 2003) populations, using the primers and polymorphisms listed in Table S3.

Phylogenetic and comparative analysis

Protein alignments were constructed using CLUSTALW, and phylogenetic analysis undertaken using the PHYLIP package, v3.5 (Felsenstein, 1989). Unrooted phylogenies were estimated by the distance matrix method, with tree topographies supported by bootstrapping (1000 replicates). Analysis was based on alignments of FAD- and cytokinin-binding domains separately (for partial predicted proteins based on ESTs or partial NGS genomic sequence), or on both domains simultaneously (for full-length predicted proteins). Macro-colinearity between cereal genomes is as previously described (Devos, 2005; International Brachypodium Initiative, 2010). Verification of micro-colinearity around each CKX member was performed using ±10 genes either side of each OsCKX gene as queries for BLASTn searches of the genomes of each sequenced grass species, using the BLAST parameters described above. Location of rice WGD genome duplications was determined as described by Thiel et al. (2009), updated using the most recent rice assembly and plotted using Circos (Krzywinski et al., 2009). Colinearity between the genetic map of barley chromosome 3H and the physical map of rice chromosome Os1 was performed as previously described (Cockram et al., 2010a).

DNA sequences have been deposited in GenBank: accessions JF292901, JF293074 to JF293079, JF495479 to JF495489, JN381555 and JN381556.


This work was supported by a Biotechnology and Biological Sciences Research Council (BBSRC) Industrial CASE studentship (BBS/S/M/2006/13029) awarded to RAGT, KWS and NIAB. Additional financial support was provided by the NIAB Trust. Barley NGS data were generated by a project funded by German Ministry of Education and Research, BMBF (GABI-BARLEX 0314000). We thank the Cereal Research Centre at the Agriculture and Agri-food Canada for provision of the wheat EST clone, INRA-CNRGV for supply of BAC clones, J Bennetzen for permission to analyse the CKX gene family in the unpublished foxtail millet genome, and Mike Ambrose and Phillipe Vain (John Innes Centre) for images of brachypodium and foxtail millet. The population of doubled-haploid individuals derived from F1 progeny of a cross between cvs. Avalon and Cadenza was developed by Clare Ellerbrook, Liz Sayers and the late Tony Worland (John Innes Centre, UK), as part of a Defra funded project led by ADAS.