Bread wheat (Triticum aestivum L.) is one of the world’s most important crop plants with a total production of over 600 million tons in 2007 [1]. This accounts for over 20% of the world’s food supply. The increase in the global population is expected to increase the demand for wheat to over 800 million tons in 2030 and to more than 900 million tons by 2050 [2]. Meeting this demand is predicted to depend largely on increasing yield through genetic improvement. However, in recent decades, internationally coordinated public wheat breeding efforts have focused on increasing resistance to disease and abiotic stress, while raising the genetic yield potential per se has received scant attention. Moreover, the challenges of increasing the genetic yield potential are considerable [3].

Grain size and grain number are significant components of yield in cereals, and both are major contributors to seed sink strength [4]. Yet, until recently, our understanding of yield at the molecular level was limited and little progress has been made towards the identification of the genetic components that define yield in a quantitative manner [5]. Some of the most significant recent studies towards the understanding of the molecular basis of grain size and number include identification of a number of genes and quantitative trait loci (QTLs) contributing to seed number and size in Arabidopsis[6, 7], rice [8, 9] and maize [10], and the recognition that these genes and/or QTLs are cytokinin synthesis and metabolism genes.

Cytokinins (CKs) are hormones that regulate many developmental and physiological processes in plants [11], including reproductive development and grain yield [7, 8, 12, 13]. In general, cytokinins may affect seed yield by enhancing sink strength and, consequently, increasing seed size, by up-regulating cell cycle related genes, through the action of sugar signalling and by enhancing phloem unloading and sugar import into the endosperm through enzymatic activity of cell wall invertase [12]. Although the beneficial effect of exogenously applied cytokinins on grain yield has been documented for several decades, including for wheat, maize, rice and barley [14–17], the application of cytokinins to field crops is limited due to the substantial cost of cytokinins. Given that changes in the levels of endogenous cytokinins will alter the regulation of many physiological processes resulting in the disruption of normal growth patterns [18], and rapid and dramatic changes in cytokinin concentration have been observed during panicle and early grain development in a number of crop species including wheat [19–21] and rice [8, 22], yield improvement through genetic manipulation of cytokinin synthesis and metabolism genes to artificially manipulate the endogenous cytokinin levels could be of considerable importance.

The last decade has seen the key steps in cytokinin biosynthesis, metabolism, and signal transduction elucidated at the molecular level [23]. The bioactive cytokinin concentration is regulated by several multigene families, including isopentenyl transferases (IPT) for cytokinin biosynthesis, cytokinin oxidases (CKX) for degradation, zeatin O-glucosyltransferases (ZOG) for reversible inactivation, and β-glucosidases (GLU) for reactivation. In Arabidopsis, nine IPT genes, seven CKX genes, three ZOG genes and GLU genes have been identified and functionally verified [24–28]. Their orthologues have also been annotated and/or functionally verified in a number of important crop species. In rice, for instance, eight IPTs and 11 CKXs have been isolated [29–31]. Similarly, 11 IPTs, 13 CKXs, and several ZOGs and GLUs have been isolated from maize [32–34].

Because of the multigene family nature of these cytokinin metabolism genes, each member can be functionally differentiated based on their spatial and temporal expression patterns [31]. Indeed, the temporal and spatial expression, and functional analysis, of specific family members has been confirmed in a number of species. For example, in Zea mays, the temporal and spatial expression, and functional analysis, of specific gene family members, confirmed that maize seed biosynthesises its own endogenous cytokinin, and that ZmIPT2 and ZmCKX1 are the gene family members controlling cytokinin homeostasis in the developing maize grain [32, 35]. In rice, Ashikari et al.[8] showed that a QTL associated with increased grain productivity was in fact a mutated cytokinin oxidase gene, OsCKX2. Most recently, RNAi was used to silence HvCKX1 in barley with a resultant increase in both grain number and grain weight [13].

Due to the pleiotropic nature of cytokinins and the variation in spatial expression of gene family members, manipulation of a particular gene may cause side effects, affecting non target organs or developmental stages. Therefore, detailed expression profiles of all available members of the major gene families involved in cytokinin synthesis and metabolism (i.e. the IPTs CKXs ZOGs and GLUs), should be investigated simultaneously in order to determine the key genes with critical roles in the target organ or developmental stage. While there are examples in the literature showing that there is variation in the spatial and temporal expression patterns of individual cytokinin gene family members [32, 36–41], simultaneous monitoring of the four major cytokinin metabolism gene families has only recently been reported in maize [33], but without detailed information during leaf, floral and seed ontology.

Apart from our preliminary report [41], detailed information regarding the expression profiles and functional territories of these multigene families is lacking for bread wheat, possibly due to the added complexity of its hexaploid nature, such that each gene is represented in triplicate homoeologues on the three genomes [42]. Further, each triplicate homoeologue within the same locus can be differentially regulated by genetic and epigenetic mechanisms in a tissue and developmentally specific manner [43–45]. In general, uniformly expressed homoeologues only account for 20-40% of the loci in bread wheat, with 20-30% of loci presenting at least one silenced homoeologue and, for the rest of the loci, the three homoeologues are differentially expressed across organs and developmental stages [43–46].

Given the above, we hypothesised that cytokinin homeostasis within an organ is co-ordinately regulated by particular homoeologues of different gene family members to allow the precise control of organ development, and that optimum seed yield could be obtained by manipulating the endogenous cytokinin concentration at specific stages of reproductive growth and seed development through genetic manipulation utilising, for example, over-expression, gene silencing, TILLING and MAS techniques.

However, it would be very complicated and essentially impractical to determine the expression profile of each homoeologue of each gene family member simultaneously, particularly when multiple tissue types and time points are involved. Due to this complexity only three TaCKX genes [47–49] and one set of GLU homoeologous genes [50] have been isolated and investigated so far, with little information regarding their expression profiles. No study with regard to IPT and ZOG genes has yet been reported for wheat.

In this study, we report the isolation and simultaneous expression analysis of TaIPT, TaCKX, TaZOG and TaGLU gene family members during flag leaf, spike and seed development by using a novel strategy in which locus-specific PCR primers shared by the three homoeologues were used, and reveal the co-ordinated expression of several members of different multigene families.

Phylogenetic analysis showed that TaCKX1 shares high sequence similarity with orthologues from other monocot species, including HvCKX1 in barley, ZmCKX1 in maize and OsCKX1 in rice (Figure 2), suggesting that comparative phylogenetic analysis based on sequence similarity is a powerful method to identify homologous genes and to predict their physiological functions in target species. Identification of orthologues across species is essential to predict gene function in newly sequenced genomes such as wheat for which a draft genome has recently been released (http://www.cerealsdb.uk.net/search_reads.htm). Indeed, when Gu et al.[34] combined a phylogenetic analysis with comparative expression of Arabidopsis, poplar, maize and rice CKXs in various tissues, they observed a general association of sequence similarity with expression patterns. Phylogenetic analysis has been widely used to classify gene families and predict their functional orthologues, and has been successfully used to generate functional categories within large gene families such as the WRKY and bHLH transcription factor families [60, 61]. However, phylogenetic analysis does not allow the differential expression of homoeologous genes to be determined.

In this study, we isolated most members of the major multi-gene families, TaIPTs, TaCKXs, TaZOGs, and TaGLUs involved in cytokinin synthesis and metabolism. Potential functional homoeologues from the three genomes, A, B and D, were pre-screened based on their expression in selected tissue types and in developmental stages closely related to grain yield determination. By using a novel strategy with specific PCR primers shared among the three homoeologues of each family member, we have been able to reveal detailed expression profiles of family members of the four multigene families during leaf, spike and seed development using qRT-PCR. Such information is required before determining which are the key genes and homoeologues contributing to cytokinin homeostasis within specific organs and developmental stages. These key genes then become the target for further investigation and genetic manipulation toward yield improvement in a breeding programme.

In plants, the rate limiting enzyme in cytokinin biosynthesis is considered to be IPT, which functions by attaching the isopentenyl side chain to the N⁶ moiety of ADP or ATP. At least three of the six IPT genes tested in this study, TaIPT2 TaIPT5 and TaIPT8 showed specific expressed patterns during spike, carpel, seed and leaf development. All three genes were expressed at the highest levels at early seed development immediately after anthesis (Figure 7A), a developmental window coincident with the cell division period, during which the cell number of the endosperm is determined, and during which endogenous cytokinin levels peak [e.g. [19–21]. Expression of IPT family members was also detected in young spikes and developing carpels, suggesting that these genes should be further investigated as players in determining the number as well as the size of the reproductive organs. Interestingly, these TaIPT genes expressed at relatively low levels during the expansion and the fully expanded phase of the leaf. Together with similar results recently reported in maize [32, 33], this suggests that it would be possible to manipulate IPT activity during reproductive development without significantly affecting normal leaf functions. An initial target for enhanced expression could be TaIPT8, which does not express highly in leaves, thereby avoiding creating competition between sink and source.

The CKX gene family members have been shown to express in different plant tissues and to play essential roles in controlling cytokinin levels during plant growth and development [62]. Information about the expression and physiological function of TaCKXs was only recently available and only for TaCKX1, 2 and 3[47–49]. While investigating the expression of ten sets of TaCKX genes, we showed that at least three genes are specifically or preferentially expressed during seed development (Figure 7B). The strong and specific expression of the TaCKX1 gene during early seed development provides a useful candidate target. By perturbing cytokinin homeostasis during this key yield determining stage, i.e. by down regulating the expression of any one of CKX1, 2 or 6 during seed development and disturbing the normal mechanism of cytokinin homeostasis, the resulting elevated cytokinin level may lead to enhanced seed number and size [8]. Indeed, Zalewski et al.[13] were able to increase grain yield of barley by down regulating the TaCKX1 orthologue, HvCKX1, using RNAi.

O-glucosylation is a major step in the metabolism of cytokinins. Since O-glucosylation is reversible and O-glucosides are resistant to cleavage of the N⁶-side chain by CKX, this conversion forms part of the homeostatic control of active cytokinin levels through temporary storage of cytokinins [11]. Interestingly, TacisZOG1 showed elevated expression pre-anthesis (Figure 3), at a time when endogenous O-glucosides start to accumulate in the developing ovule [19]. Extremely high level of expression was detected for two cis-ZOG genes, TacisZOG1 and TacisZOG2-1. The dramatic increase of TacisZOG expression during leaf senescence in the present study agrees with early observations of high levels of O-glucosides in senescing leaves [63] and cytokinin inactivation by O-glucosylation promoted by senescence-related processes in petunia [64]. In contrast, Vyroubalová et al.[33] only detected low expression of ZmcisZOG in old leaves of maize. This difference might be due to differences in cytokinin metabolism during senescence in these species [63].

The β d-glucosidases are members of the GH1 family, which is implicated in the hydrolysis of a number of plant secondary metabolites, including the regulation of the biological activity of cytokinins [65]. Until now, Zm-p60/ZmGLU1 in maize, TaGLU1 in wheat and Bgl4 in Brassica napus are the only enzymes for which the ability to release cytokinins from O-glucoside conjugation has been demonstrated [50, 65–67]. In wheat, only one homoeologous set of GLU genes, TaGLU1a, b and c, has been investigated [50], but with no information shown on in planta expression during reproductive and leaf development. Our qRT-PCR data showed high expression of TaGLU1-1, the equivalent of TaGLU1a in Sue et al. [50], in 0.5 cm spike, 10 cm spike and 2 cm leaf. Using a ZmGLU1P-GUS construct, Gu et al.[68] observed a significant up-regulation of the maize β-glucosidase gene, ZmGLU1, in immature transgenic tobacco seeds. Vyroubalová et al.[33] also detected high expression of ZmGLU in immature ears and young leaves of maize. Therefore, we speculated that this gene may play an important role in reproductive organ initiation and seed development.

Interestingly, and in agreement with the pattern shown in our preliminary study [41], at anthesis, when expression of IPT CKX and ZOG genes was uniformly low in the flag leaf, the TaGLU1 gene showed a marked peak of expression, with over a 300-fold up-regulation. From anthesis, the flag leaf provides significant resources to the developing seed. We postulate that the high β-glucosidase gene activity at this stage could lead to the release of cytokinin from conjugation in the leaf and provide a source of cytokinin for the developing carpels/seeds. Clearly, experiments to support this supposition still need to be performed.

The modest correlation of response regulators with TaIPT gene expression during the early stage of seed development when endogenous cytokinin levels are substantially elevated [19–21] indicates that response regulators may not be good surrogates of endogenous cytokinin content. However, the elevated expression of members of all four gene families at 2 daa is indicative of dynamic turnover of cytokinins at this key stage of seed development.

An interesting trend shown by our data, and that of our preliminary study [41], is that several of the highly expressed genes in the seeds have low expression in the leaves and vice versa, further suggesting that expression of individual family members of the cytokinin synthesis and metabolism genes are tissue specific. As a monocarpic plant, wheat development is characterized by a rapid translocation of metabolites from leaves to developing grains after anthesis. Consequently, delayed leaf senescence at a late stage of seed development may not be beneficial to the final grain yield due to the decreased nutrient translocation from leaves to seeds as suggested by Sýkorová et al.[69]. The independent and variable expression profiles of different members within or among cytokinin regulatory gene families provides useful flexibility for independently manipulating the endogenous bioactive cytokinin levels in reproductive and vegetative tissues to achieve maximal seed yield.

The expression levels shown in this study represent at least one of the three homoeologues for each locus in all cases apart from the TaZOG2-1 TaZOG2-2 TaGLU1-1 and TaGLU1-2 genes, in which the three homoeologous genes showed different expression profiles. The third homoeologue in these cases displayed little activity in the initial screen for highly expressed genes. The expression among homoeologous genes of other cytokinin regulatory genes may also vary, or individual homoeologous genes could even be silenced in some or all of the tissue samples analysed. Homoeologous gene silencing and/or differential expression is a phenomenon widely documented in bread wheat [43–46] and other polyploid species including tetraploid cotton [70], and tetraploid Tragopogon mirus[71].

Given that homoeologous expression can be differentially regulated by genetic and epigenetic mechanisms in a tissue and developmentally-specific manner [43–45], detection and pyramiding of interesting homoeologues for improvement of target traits has recently been used in wheat breeding programs [72]. For instance, in an attempt to improve the bread-making quality of hexaploid wheat, homoeologous recombination has been used to increase the copy number of the stronger flour quality contributing homoeologue, Glu-D1, of the HMW-GS gene. A partially isohomoeoallelic line, in which the chromosome 1A homoeologue was replaced by the chromosome 1D homoeologue, showed significantly improved bread making quality. The improved quality can be explained by the duplication of the Glu-D1 homoeologue [73, 74]. Using a different approach, TILLING in combination with homoeologue-specific primers, Slade et al.[75] obtained a triple homozygote loss-of-function of the Wx gene and obtained a line with waxy phenotype and improved grain quality,

Therefore, from a breeding perspective, the above strategy, of accumulating targeted homoeologues, could also be applied to the cytokinin synthesis and metabolic gene families. To improve seed yield, higher endogenous cytokinin levels could be achieved through tissue-specific increased expression of multiple homoeologues of the IPT genes TaIPT2, TaIPT5 and TaIPT8, or through decreased homoeologue expression of CKX genes TaCKX1 and TaCKX2. This could be achieved using a number of genetic tools, such as over-expression, down-regulation, TILLING and MAS strategies. However, while the differential expression profile of each homoeologue elucidated in this study could be directly used for genetic manipulation towards the improvement of the grain yield, precise genome allocation of these positive homoeologues is important for further understanding the genetic control, and for facilitating the efficient accumulation, of these homoeologues within the same locus and across different loci.

We have demonstrated that the expression patterns of individual members of the TaIPTs, TaCKXs, TaZOGs, and TaGLUs multigene families were tissue and developmentally-specific during spike, ovule, seed and flag leaf development. This supports the suggestions that the four multigene families play important roles in maintaining cytokinin homeostasis during reproductive development and will contribute to grain yield and quality. We suggest that artificial disturbance of endogenous cytokinin levels at specific stages of organ development through targeted manipulation of particular family members is a practical possibility, and have identified several key candidate genes for genetic improvement of grain yield in this species.