Crossing paths: cytokinin signalling and crosstalk

Over 100 years ago, several scientists discovered the existence of substances that were able to induce cell division in cultured or wounded plant tissue (Wiesner, 1892; Haberlandt, 1913; van Overbeek et al., 1941). In 1955, Miller and colleagues (Miller et al., 1955) isolated an active substance from herring sperm and named it kinetin; this was followed by the identification of trans-zeatin (Miller, 1961) and many other cytokinins (reviewed in Mok and Mok, 2001). Cytokinins, which were named for their ability to promote cytokinesis, are a class of phytohormone derived from adenine. Naturally occurring cytokinins can be divided into two groups based on their side chain (Fig. 1A): those with isoprene-derived side chains, which are predominant in plants; and those with aromatic side chains (Sakakibara, 2006).

Early work on cytokinins showed that they promote shoot growth, inhibit root growth, stimulate cell division and induce greening in calli (Miller et al., 1956). Since then, they have been discovered to be important in many developmental processes in plants, including organ formation, leaf senescence, shoot meristem formation and maintenance, apical dominance and seed germination (Mok and Mok, 2001). Cytokinin and auxin have been known to play antagonistic roles in many of these processes since cytokinin was first identified (Skoog and Miller, 1957), and this has remained an important theme in plant biology (Coenen and Lomax, 1997). Here, we provide an overview of cytokinin synthesis, transport and signalling pathways, focusing on recent major advances in our understanding of the crosstalk between cytokinin and auxin, and the interactions between cytokinin and other hormones; several recent reviews have addressed other aspects of cytokinin biology (Argueso et al., 2009; Argueso et al., 2010; Kudo et al., 2010; Frébort et al., 2011; Brenner et al., 2012; Hwang et al., 2012).

Cytokinin biosynthesis is a multi-step process requiring the activity of several different genes. In Arabidopsis thaliana, the model species for plant research, cytokinin biosynthesis is controlled by the ATP/ADP isopentenyltransferase (IPT) genes, which encode rate-limiting enzymes in cytokinin biosynthesis (Fig. 1B). Seven IPT genes (AtIPT1 and AtIPT3 through AtIPT8) are found in Arabidopsis, and they encode enzymes that synthesize the cytokinin precursor, isopentenyladenine (Kakimoto, 2001; Takei et al., 2001). Two cytochrome P450 monooxygenases (CYP735A1 and CYP735A2) then catalyse the hydroxylation of isopentenyladenine-type cytokinins (Takei et al., 2004b). In addition, the LONELY GUY (LOG) gene family encodes enzymes that convert cytokinin from an inactive to an active form. Cytokinin degradation is mediated by the cytokinin oxidases, which were first described over 30 years ago (Pačes et al., 1971); these enzymes are encoded by the CYTOKININ OXIDASE (CKX) gene family (Houba-Hérin et al., 1999; Schmülling et al., 2003).

Cytokinin degradation and synthesis genes are both widely expressed, being active in tissues of both the shoot and the root (Nordström et al., 2004). Analyses of expression patterns using promoter-GUS fusions have demonstrated that cytokinin biosynthesis occurs in various tissues and cells in aerial organs as well as roots (Miyawaki et al., 2004). The IPT and CKX genes are widely expressed throughout the Arabidopsis root and have specific expression patterns (Werner et al., 2003; Miyawaki et al., 2004). Cytokinin biosynthesis and degradation pathways have recently been reviewed comprehensively by Frébort et al. (Frébort et al., 2011).

Although the polar transport of auxin is relatively well understood and characterized, very little is known about how cytokinin is taken up by cells and transported in plants. Diffusion is likely to play an important role, yet evidence from Chenopodium rubrum cultures (Fußeder et al., 1989) and Arabidopsis cell cultures (Cedzich et al., 2008) suggests that cytokinin is also actively taken up by cells, although dedicated cytokinin transporters have yet to be identified. Furthermore, although grafting experiments have long suggested the importance of long-distance cytokinin transport, this phenomenon remains poorly understood, despite significant recent advances. Cytokinin uptake and transport have been reviewed by Kudo et al. (Kudo et al., 2010) but are summarised below.

The PURINE PERMEASE (PUP) gene family encodes broad-affinity transporters that are able to transport several cytokinins. PUP1 was initially identified from suppression cloning in purine transport-deficient yeast and was shown to also recognize cytokinins (Gillissen et al., 2000). Bürkle et al. (Bürkle et al., 2003) showed direct, active uptake of adenine and various cytokinins by PUP1 and PUP2; the expression pattern of PUP1 and PUP2 suggest that they may play a role in the loading and unloading of cytokinins for long-distance transport. The EQUILIBRATIVE NUCLEOSIDE TRANSPORT (ENT) gene family has also been implicated in cytokinin transport in Arabidopsis. Sun et al. (Sun et al., 2005) found a reduction in cytokinin uptake efficiency in ent3 and ent8 mutants. OsENT2, one of the four ENT genes in Oryza sativa, has also been shown to be a functional nucleoside transporter that may play a role in long-distance transport of nucleosides in growing plants (Hirose et al., 2005). In addition, ENT6 has also recently been implicated as a transporter of nucleoside-type cytokinins in Arabidopsis (Hirose et al., 2008). Although these transporters may play a role in cytokinin loading for long-distance transport, their broad affinity and the lack of strong phenotypes in the mutants suggests that they do not contribute to major processes that regulate plant development.

Reciprocal grafting experiments between ipt1;3;5;7 mutants and wild-type plants showed preferential transport of different cytokinin species; trans-zeatin (tz)-type cytokinins were transported from the root to the shoot, while isopentenyl adenine (iP)-type cytokinins moved from the shoot to the root (Matsumoto-Kitano et al., 2008). This is consistent with earlier grafting experiments in tobacco (Faiss et al., 1997) in which cytokinin overproduction in the root stock failed to release lateral branches from suppression or accelerate senescence, suggesting that the long-distance transport of cytokinin is somehow regulated. Further evidence came from the detection of cytokinin in sap and leaf exudates (Takei et al., 2002; Takei et al., 2004a). Long-distance transport of cytokinin through the phloem was recently directly demonstrated by Bishopp et al. (Bishopp et al., 2011b), who observed movement of radiolabelled cytokinins applied to the cotyledons to the root tip. Bishopp et al. were able to disrupt the rootwards flow of cytokinin by inducing callose synthesis in the phloem, demonstrating the importance of symplastic connections in the phloem to acropetal cytokinin transport.

The major components of the cytokinin signalling pathway were identified in a series of seminal papers around the turn of the century. Cytokinin signalling is mediated by a two-component signalling pathway (Fig. 2) similar to the two-component signalling systems (TCSs) found in bacteria. In brief, cytokinin induces autophosphorylation of a histidine kinase (HK) protein, which results in the transfer of a phosphoryl group from a phospho-accepting histidine residue in the kinase domain to an aspartate residue. The phosphoryl is then transferred to a conserved histidine on a histidine phosphotransferase (HP) protein. From there, it is finally transferred to an aspartate in the receiver domain of a response regulator (RR). For a thorough overview of cytokinin signalling, the reader is referred to the recent review by Hwang et al. (Hwang et al., 2012).

The first indication of a link between cytokinin and the bacterial TCS was the discovery that overexpression of CYTOKININ INDEPENDENT (CKI), a gene homologous to the HKs in the bacterial TCS, induced a cytokinin response in plants (Kakimoto, 1996). A few years later, the research of several groups independently converged on the identification of the first cytokinin receptor. Mähönen et al. (Mähönen et al., 2000) identified the gene responsible for the wooden leg (wol) mutation (Scheres et al., 1995) as a histidine kinase involved in two-component signalling in plants. The WOL gene was discovered to be allelic with CYTOKININ RESPONSE 1 (CRE1), which Inoue et al. (Inoue et al., 2001) simultaneously discovered to be a cytokinin receptor, as loss-of-function mutants showed a reduced cytokinin response. Yamada et al. (Yamada et al., 2001) demonstrated the cytokinin-binding activity of WOL/CRE1 in vitro and showed that the wol mutation abolished this activity. Meanwhile, labs working with WOL/CRE1 under the name ARABIDOPSIS HISTIDINE KINASE 4 (AHK4) showed that it could act as a cytokinin sensor in bacteria (Suzuki et al., 2001) and that its histidine kinase activity was cytokinin dependent (Ueguchi et al., 2001). AHK2 and AHK3 were identified as cytokinin receptors shortly afterwards (Hwang and Sheen, 2001).

Arabidopsis mutants lacking all three receptors (AHK2, AHK3 and AHK4) show no cytokinin response in a variety of assays and produce small infertile plants (Higuchi et al., 2004; Nishimura et al., 2004; Riefler et al., 2006). The receptors show differing sensitivity to different cytokinin species, suggesting there may be some specificity in the signalling pathway (Spíchal et al., 2004; Romanov et al., 2006). Recent research has found that a large majority of cytokinin receptors are present on the ER membrane, suggesting that this compartment may play an important role in cytokinin signal transduction (Caesar et al., 2011; Wulfetange et al., 2011). For an overview of our current understanding of cytokinin receptors, the reader is referred to a recent review by Heyl et al. (Heyl et al., 2012).

The Arabidopsis genome encodes five HP proteins (AHP1-AHP5) (Suzuki et al., 1998) that have been shown to have phosphorelay activity in vitro (Imamura et al., 2001; Suzuki et al., 2002; Tanaka et al., 2004). A sixth HP protein, AHP6, lacks the conserved histidine residue and so acts as a pseudo-phosphotransferase; by competing with the true HPs, AHP6 acts as an inhibitor of cytokinin signalling (Mähönen et al., 2006). The AHP proteins are known to shuffle between the cytosol and the nucleus. Although earlier studies in protoplasts (Hwang and Sheen, 2001) reported that this translocation was cytokinin dependent, more recent work by Punwani et al. (Punwani et al., 2010) has shown that the translocation occurs continuously in a cytokinin-independent manner. The quintuple AHP mutant created by Hutchison et al. (Hutchison et al., 2006) has a reduced cytokinin phenotype, though not as severe as the triple receptor mutant. However, Deng et al. (Deng et al., 2010) noted that the ahp2 allele in this mutant was not a null allele; replacing it with a null allele produced a quintuple AHP mutant that was seedling lethal.

The 23 functional response regulators (RRs) in Arabidopsis are divided into three groups, two of which (type A and type B) are involved in cytokinin signalling. Phosphorylation of the type A ARRs acts to stabilize them (To et al., 2007), whereas phosphorylation of the type B ARRs enables them to bind to DNA and initiate transcription of downstream targets, including the type A ARRs. D’Agostino et al. (D’Agostino et al., 2000) showed that the type A ARRs are transcriptionally upregulated by cytokinin. Type A ARRs only have a receiver domain and are generally thought to act as inhibitors of cytokinin signalling. Although genetic analysis has confirmed the negative activity of several of the type A ARRs (To et al., 2004), ARR4 has also been shown to interact positively with phytochrome B (Sweere et al., 2001). However, type B ARRs have additional domains and are therefore able to activate the transcription of cytokinin responsive genes (Sakai et al., 1998; Sakai et al., 2000; Sakai et al., 2001). Work with high-order mutants (To et al., 2004; Mason et al., 2005; Argyros et al., 2008) and overexpression lines (Kiba et al., 2004; Tajima et al., 2004; Ren et al., 2009) has confirmed this general picture.

A novel class of response regulators was identified by Rashotte et al. (Rashotte et al., 2006). The CYTOKININ RESPONSE FACTOR (CRF) family consists of six closely related APETALA2 (AP2) transcription factors, which were identified by a phylogenetic analysis of the ethylene response factor family. Although only CRF2, CRF5 and CRF6 were found to be upregulated by cytokinin, all six CRFs rapidly translocated to the nucleus upon cytokinin application. This upregulation required functional type B ARRs, whereas the migration of the CRFs to the nucleus was found to depend only on HK and HP proteins, thus representing a branching point in the two-component signalling pathway. Microarray analysis showed that the CRFs overlap with the type B ARRs in activating many cytokinin target genes; single and multiple loss-of-function mutations in the CRFs led to defects in cotyledon and leaf development, as well as embryo lethality in the case of crf5 crf6. Recently, published work has shown that the CRFs can form both homo- and heterodimers with one another, and has identified the protein domain required for CRF-CRF interaction and interactions with other cytokinin signalling components (Cutcliffe et al., 2011).

Numerous transcriptomics experiments using microarrays and, more recently, RNA-sequencing have been used to explore the downstream targets of cytokinin signalling (Rashotte et al., 2003; Brenner et al., 2005; Brenner and Schmülling, 2012). Argueso et al. (Argueso et al., 2010) have reviewed some of the transcriptional networks identified by these approaches, but one prominent class of genes that has received special attention includes those that are involved in crosstalk with other hormonal pathways; these genes, which include signalling genes responsive to auxin (SHY2/IAA3, AXR3/IAA17) and gibberellins (GNL/CGA1/GATA22, GNC/GATA21), have recently been reviewed by Brenner et al. (Brenner et al., 2012).

Just as the turn of the century was a fruitful period for our understanding of cytokinin signalling, the new decade has seen an explosion in our understanding of the crosstalk between cytokinin and other hormone pathways, particularly auxin. For over half a century, interactions between auxin and cytokinin have been known to be play a crucial role in many aspects of plant development, often acting antagonistically (Coenen and Lomax, 1997), but we are now beginning to assemble an understanding of this antagonism on a molecular and mechanistic level (Fig. 3; Tables 1, 2). Crosstalk between cytokinin and other hormones is also being unravelled at the same time, although the interactions with auxin have received most of the attention.

Over half a century ago, Skoog and Miller (Skoog and Miller, 1957) demonstrated that in vitro organogenesis could be controlled by varying the relative quantities of auxin and cytokinin in the growth medium. The antagonism between these two phytohormones has since been a continuing theme in plant biology, although the molecular and mechanistic bases for this antagonism have only recently begun to be unravelled. We now know that there is extensive crosstalk between the two hormones at all levels - synthesis, perception and transport - and we are beginning to understand how these networks interact (Fig. 3A) to control a wide variety of plant responses.

The expression of the IPT genes has been shown to respond to levels of auxin. An RT-PCR analysis of the IPT genes by Miyawaki et al. (Miyawaki et al., 2004) found that IPT5 and IPT7 were upregulated within 4 hours of auxin application; Dello Ioio et al. (Dello Ioio et al., 2008) later demonstrated that the upregulation of IPT5 by auxin is mediated by SHY2/IAA3. A few years later, Tanaka et al. (Tanaka et al., 2006) confirmed the effect of auxin on IPT expression while investigating apical dominance in pea, Pisum sativum. Guided by the amino acid sequence of the Arabidopsis IPT proteins, they identified two IPT homologs in pea (PsIPT1 and PsIPT2) that showed increased expression in the second nodal stem following decapitation, suggesting that auxin from the shoot apex normally represses these genes. This was confirmed by incubating excised stem segments with and without auxin; IPT expression only persisted in segments incubated without auxin. Finally, application of the auxin transport inhibitor 2,3,5-tri-iodobenzoic acid (TIBA) around the internode also led to increased IPT expression levels, demonstrating that these genes are normally repressed by auxin transported from the shoot apex.

However, other studies have found that treatment with auxin leads to a reduction in cytokinin biosynthesis. Using quantitative RT-PCR, Takei et al. (Takei et al., 2004b) observed a significant decrease in the expression of the cytokinin biosynthesis gene CYP735A when Arabidopsis seedlings were treated with auxin for 1 hour. Nordström et al. (Nordström et al., 2004) used deuterium labelling to directly measure the synthesis rates of the auxin indole-3-acetic-acid (IAA) and cytokinins; they found that treatment of Arabidopsis seedlings with auxin led to a significant reduction in cytokinin biosynthesis, while increasing cytokinin synthesis using an inducible version of the bacterial ipt gene did not affect IAA pool size or synthesis rate until 36 hours later. Based on this, they concluded that auxin can regulate the pool of available cytokinins but that the effect of cytokinin on auxin is indirect and therefore probably not an effective mechanism to control development.

In addition to regulating cytokinin biosynthesis, auxin has also been shown to affect cytokinin degradation by regulating the CKX gene family. Using semi-quantitative RT-PCR, Werner et al. (Werner et al., 2006) measured the response of the seven Arabidopsis CKX genes to auxin and found that CKX2, CKX4 and CKX7 were weakly downregulated in response to auxin. By contrast, CKX1 and CKX6 were reported to be upregulated by auxin, although they were also strongly downregulated by the application of the polar transport inhibitor 1-naphthylphthalamic acid (NPA), suggesting that their expression depends on polar auxin transport.

Auxin has also been discovered to regulate cytokinin signalling directly. In an elegant experiment, Müller and Sheen (Müller and Sheen, 2008) showed that auxin directly activates two type A ARRs, ARR7 and ARR15, thereby inhibiting cytokinin signalling (Fig. 3B). They noted that the domain of ARR7 and ARR15 expression in Arabidopsis embryos corresponded more with the auxin signalling domain than the cytokinin signalling domain, as reported by a synthetic reporter of two-component signalling, TCS::GFP. Furthermore, treating embryos with the synthetic auxin 2,4-D led to an expansion of the ARR domain. Following this, the researchers identified auxin-responsive elements in the promoters of these two genes and were able to abolish the auxin response by generating point mutations in these response elements. Zhang et al. (Zhang et al., 2011) showed that, contrary to earlier results, the arr7 arr15 double mutant is not embryo lethal. Earlier reports of gametophytic or embryo lethality resulted from reduced pollen production in the arr7 or arr15 heterozygous mutants; the homozygous mutants of either gene produce viable pollen.

Zhao et al. (Zhao et al., 2010) found contrasting results in the shoot apical meristem (SAM), where they reported that auxin represses ARR7 and ARR15; several hours after treatment of the inflorescence apex with the auxin naphthaleneacetic acid (NAA), the transcripts of both genes were reduced by about half. Expression of ARR7 and ARR15 drastically increased following treatments with the protein synthesis inhibitor cycloheximide; this is characteristic of auxin-regulated genes, as they are normally repressed by a bound auxin response factor (ARF). Expression also increased in the arf5/mp mutant, suggesting that ARF5/MP (MONOPTEROS) might mediate the auxin-dependent regulation of these ARRs; chromatin immunoprecipitation (ChIP) and electrophoretic mobility shift assay confirmed that MP binds directly to the auxin-responsive elements (REs) in the ARR15 promoter. Another response regulator, CRF2, was simultaneously identified as a target of MP/ARF5 in a microarray and was confirmed by ChIP (Schlereth et al., 2010). By mutating the auxin REs in the ARR15 promoter, Zhao et al. (Zhao et al., 2010) showed that these elements repress ARR15 in the SAM but activate it in other domains, explaining the discrepancy between their observations and those of Muller and Sheen. These results demonstrate the complexity of the crosstalk between hormones: various components can be up- or downregulated in a context-dependent manner, providing an elegant mechanism for interaction between the two hormone pathways. Owing to the differing response of these type A ARRs, auxin and cytokinin are able to act antagonistically in the embryonic root meristem but synergistically in the SAM.

Conversely, cytokinin regulates auxin biosynthesis. By measuring the incorporation of deuterium into IAA by gas chromatography - multiple reaction monitoring-mass spectrometry (GC-MRM-MS), Jones et al. (Jones et al., 2010) observed that treatment with various cytokinins led to increased auxin synthesis in young leaves, the shoot apex and the root system; this effect was amplified in the cytokinin hypersensitive ARR mutant arr3 arr4 arr5 arr6. By contrast, treatment of the cytokinin-insensitive quadruple AHP mutant ahp1 ahp2 ahp3 ahp4 resulted in a decrease in IAA biosynthesis, suggesting that the relationship between cytokinin signalling and auxin synthesis may not be straightforward; roots of this mutant also had a higher baseline rate of IAA biosynthesis than wild type, further hinting at the complexity of this interaction. Cytokinin treatments did not increase IAA biosynthesis in the axr3-1 mutant, which is hypersensitive to auxin because the increased stability of the AXR3/IAA17 protein; this mutant also has increased steady state levels of IAA. Transcript profiling following cytokinin induction identified many auxin biosynthesis genes as upregulated; genes responsible for IAA storage and degradation via conjugation and deconjugation also responded positively to cytokinin (Table 2). Taken together, these data suggest that degradation of IAA17 is important in maintaining auxin levels and that cytokinin regulates auxin metabolism by increasing the production or stability of this protein.

Several research groups independently discovered that cytokinin also regulates auxin signalling and, in particular, auxin transport. Laplaze et al. (Laplaze et al., 2007) showed that the repression of lateral root primordium initiation by cytokinin happens via changes to the PIN class of auxin efflux transporters (Fig. 3B). Based on observations of transcriptional reporter lines, they reported that the expression of PIN1 and PIN6 in lateral root primordia is reduced and diffuse when plants are treated with cytokinin; cytokinin treatments also led to reduced expression of PIN1, PIN2, PIN3, PIN4 and PIN7 in the shoot. These findings led them to conclude that cytokinin disrupts PIN patterning during lateral root initiation.

Subsequent studies demonstrated that the size of the Arabidopsis primary root meristem is also carefully controlled by a balance between auxin, which controls the rate of cell division, and cytokinin, which controls the rate of cell differentiation (Fig. 3B). In a seminal paper, Dello Ioio et al. (Dello Ioio et al., 2008) investigated the mechanism behind this antagonism. Overexpression of ARR1, a type B ARR, led to a significant reduction in root meristem size. Earlier work (Taniguchi et al., 2007) had identified the Aux/IAA gene IAA3/SHY2 as a target of ARR1. Dello Ioio et al. therefore examined shy2 gain- and loss-of-function mutant plants; while the meristem was reduced in the gain-of-function mutant, the loss-of-function mutant had a larger root meristem along with faster growth, which led to longer roots that closely resembled arr1 mutants. Expression of the SHY2::GUS reporter line was hardly detectable in the arr1 mutant, in contrast with wild type where it was active in the vascular transition zone. Furthermore, induction of ARR1 had no effect on the root meristem in loss-of-function shy2 mutants, consistent with a role for SHY2 downstream of ARR1. Finally, ChIP experiments confirmed a direct interaction between ARR1 and SHY2. Noting that the expression of PIN1, PIN3 and PIN7 in the vascular tissues was downregulated by cytokinin, Dello Ioio et al. speculated that this might be mediated via AHK3, ARR1 and SHY2. Indeed, expression of these PIN genes was cytokinin insensitive in the arr1, ahk3 and shy2 loss-of-function mutants; expression was also reduced in the shy2 gain-of-function mutant. In addition to mediating cytokinin regulation of PIN genes, SHY2 was also found to negatively regulate the cytokinin biosynthesis gene IPT5. Dello Ioio et al. therefore proposed a model in which the size of the root meristem is determined by an antagonism between auxin and cytokinin in the vascular transition zone; this antagonism is mediated by opposing regulation of SHY2, which in turn negatively regulates cytokinin biosynthesis while repressing auxin transport and signalling (Fig. 3B). This dual negative-feedback loop centred on SHY2 establishes a robust mechanism for controlling meristem size. This model was later extended by Moubayidin et al. (Moubayidin et al., 2010) to include upregulation of SHY2 by ARR12 during meristem growth.

Further evidence that cytokinin modulates auxin transport by regulating PIN transporters came from a study by Růžička et al. (Růžička et al., 2009). As cytokinin and 2,4-D, which is poorly transported thanks to its low affinity with PIN transporters, have a similar effect on the root meristem, they speculated that the effect of cytokinin may be mediated via a reduction in auxin transport. Using tobacco cell cultures, they found that cytokinin treatment led to a decrease in auxin efflux. Following this, they showed that cytokinin decreases the expression of PIN1 and PIN3 in the Arabidopsis root meristem while increasing the expression of PIN7. The effect of cytokinin on PIN1 was reduced in the cre1 and cre1 ahk2 cytokinin receptor mutants, confirming that this regulation is dependent on functional cytokinin signalling.

Following this work, a growing body of research describing the regulation of PIN transporters by cytokinin has accumulated. For example, Kalousek et al. (Kalousek et al., 2010), observed rapid upregulation of PsPIN1 and PsAUX1 expression in the axillary buds of pea plants following the application of cytokinin; cytokinin treatment also caused PsPIN1 to become localised to the basal membrane of competent cells. Similar changes were observed following decapitation, in keeping with earlier findings (Tanaka et al., 2006), showing that cytokinin-mediated regulation of polar auxin transport may play a role in regulating apical dominance.

More recently, Bishopp et al. (Bishopp et al., 2011a) described a mutually inhibitory feedback loop between auxin and cytokinin that is involved in patterning the Arabidopsis stele. The vascular cylinder of the Arabidopsis root is precisely and consistently patterned; a central xylem axis is flanked by intervening procambial cells with the phloem poles located perpendicular to the xylem axis. Bishopp et al. observed an auxin signalling maximum throughout the xylem axis and a complementary cytokinin signalling domain in the procambium. Further investigation revealed that treatment with the polar auxin transport inhibitor NPA excluded the auxin response from the stele and altered the expression pattern of the pseudo-phosphotransferase AHP6, which was shown to be regulated by auxin. Because NPA inhibits auxin efflux, Bishopp et al. next investigated the role of PIN transporters in this patterning process. A weak vascular phenotype was found in pin1 but not in the other single pin mutants; however, the pin3 pin7 double mutant showed unstable formation of the auxin signalling maximum. Furthermore, the expression patterns of PIN1, PIN3 and PIN7 were shown to depend on cytokinin signalling. Reduction of cytokinin signalling in the cytokinin receptor mutants or by induction of CKI expression in the vascular domain led to a reduction in the PIN7 expression domain and altered PIN3 expression; lateral localization of PIN1 in vascular cells may also depend on cytokinin signalling, as vascular cells of wol mutants only have basally localized PIN1. Based on these observations, Bishopp et al. proposed a model wherein high cytokinin signalling in the procambial cells generates a bisymmetric pattern of PIN localisation that directs the radial transport of auxin into the central axis, where high auxin signalling promotes the expression of AHP6, which reciprocally restricts the domain of high cytokinin signalling. Thus, a feedback loop (Fig. 3B) involving hormone transport dynamics and mutually inhibitory hormone interactions provides a mechanism to sharpen the boundary between the hormone domains and generate positional information for patterning.

Researchers have recently begun to elucidate some of the ways that cytokinin may regulate the PIN transporters. Zhang et al. (Zhang et al., 2011) compared the protein and mRNA levels of PIN1, PIN3, PIN4 and PIN7 in wild-type plants and in the multiple type A arr3, arr4, arr5, arr6, arr7, arr8, arr15 mutant both with and without cytokinin. The mRNA levels did not differ significantly between the wild-type and mutant plants; PIN1 and PIN3 mRNA levels did not change upon cytokinin treatment, whereas PIN4 mRNA levels increased. By contrast, expression of all three GFP-fusion proteins decreased with cytokinin treatment. These discrepancies led Zhang et al. to conclude that cytokinin-mediated regulation of these PIN transporters occurs primarily at a post-transcriptional level. In line with this, Marhavý et al. (Marhavý et al., 2011) found that cytokinin downregulated PIN1 at the plasma membrane even when it was driven by the constitutive 35S promoter, indicating that this regulation occurs at the protein level. Downregulation of PIN1 by cytokinin occurred very quickly, with a 40% reduction in less than 2 hours. They observed transport of PIN1-GFP to the vacuole and were able to prevent this trafficking by pharmacological depolymerisation of actin or inhibition of vacuolar targeting. The BFA-visualized endocytosis trafficking defective mutants ben1 and ben2, which are both defective in PIN1 endocytosis, did not show a decrease in PIN1-GFP following cytokinin treatment. Furthermore, cytokinin degradation of PIN1 was robust in the ahk2 and ahk3 single and double mutants but disrupted in the cre1/ahk4 mutant; cytokinin also failed to degrade PIN1 in arr2 and arr12 mutants. Together, these data indicate that cytokinin signalling mediated via the AHK4 receptor, ARR2 and ARR12 leads to endocytic trafficking of PIN1 from the plasma membrane to lytic vacuoles for degradation.

Gibberellins are phytohormones involved in a wide range of processes in plant development, from stem elongation and seed germination to regulating meristem size. Exogenous gibberellin application, which increases the size of the root meristem, was found to strongly downregulate ARR1 expression in 5-day-old plants but not in 3-day-old plants. Root meristems were uniformly insensitive to gibberellin 3 days after germination; 2 days later, the meristems of arr1-3 plants remained insensitive while wild-type meristems showed a response. The DELLA protein Repressor of GA 1-3 is expressed in the root transition zone and shows an increase in expression between 3 and 5 days after germination; the loss-of-function rga-24 mutant was found to have decreased levels of ARR1 and consequently of SHY2 5 days after germination. Taken together, these data indicate that a reduction of gibberellin levels 5 days after germination releases ARR1 from repression, which upregulates SHY2 (Fig. 4A) and leads to an increase in cell differentiation, which balances with cell division to control the size of the root meristem (Moubayidin et al., 2010).

It has been known for some time that cytokinin also regulates the biosynthesis of ethylene, which serves as a hormone in plants. As the application of cytokinin to dark-grown plants resulted in the ‘triple response’ that is characteristic of ethylene, Cary et al. (Cary et al., 1995) tested the role of ethylene in the cytokinin response of seedlings. They treated the ethylene-resistant mutants ein1 and ein2 with cytokinin and observed that these mutants were resistant to cytokinin-induced reduction of root and hypocotyl growth. Furthermore, the effects of cytokinin treatment in wild-type plants were reduced by the application of inhibitors of ethylene signalling or biosynthesis. These results demonstrate that the effect of cytokinin on root and hypocotyl growth is mediated by ethylene. Consistent with this, Penmetsa et al. (Penmetsa et al., 2008) observed that ethylene-resistant mutants in Medicago were resistant to cytokinin-mediated effects on root growth.

Following this, Vogel et al. (Vogel et al., 1998) suggested that cytokinin post-transcriptionally increases the activity of the ethylene biosynthesis gene 1-AMINOCYCLOPROPANE-1-CARBOXYLIC ACID SYNTHASE 5 (ACS5). Subsequently, Woeste et al. (Woeste et al., 1999) demonstrated that treatment of etiolated seedlings with either cytokinin or auxin led to increased ethylene levels, while treatment with both hormones had a synergistic effect. Because they found that auxin, but not cytokinin, led to increased levels of ACS4 mRNA, Woeste et al. suggested that cytokinin acts post-transcriptionally to increase ACS4 function.

Continuing work by Joseph Kieber’s group has further elucidated the mechanics of this interaction (Fig. 4B). The Arabidopsis mutants eto1, eto2 and eto3 were identified as ethylene over-producing mutants owing to a constitutive triple-response phenotype (Guzmán and Ecker, 1990; Kieber et al., 1993). Positional cloning identified eto2 and eto3 as mutations in genes from the ACS family (Vogel et al., 1998; Chae et al., 2003), which encode enzymes that catalyse the first committed, and generally rate-limiting, step in ethylene biosynthesis. The eto2 and eto3 mutations are found within the C-terminal domains of ACS5 and ACS9, respectively, and seem to influence post-transcriptional regulation of the proteins. Chae et al. (Chae et al., 2003) showed that eto2 does not affect the activity of ACS5; using a myc-tagged, dexamethasone-inducible ACS5, they were able to show that the eto2 mutation increases the stability of the ACS5 protein. In wild-type plants, cytokinin was found to increase the synthesis and half-life of the myc-tagged construct; exogenous cytokinin also increased the steady-state ACS5 levels in the myc-tagged eto2 mutant, leading Chae et al. to suggest that the effect of cytokinin and the eto2 mutation are at least partially independent.

Shi et al. (Shi et al., 2012) reported that ethylene, in turn, modulates cytokinin signalling. They observed increased transcription of ARR5, ARR7 and ARR15 in the ein3 eil1 double mutant compared with wild type; cold induction of these ARRs was also consistently higher in the mutant than in wild-type plants. These results indicate that the ethylene-responsive transcription factor EIN3 represses these ARRs, a finding that was confirmed by direct binding assays. Furthermore, plants overexpressing these ARRs showed enhanced tolerance to freezing treatments; exogenous cytokinin treatment also enhanced the survival of freezing treatments. Treatment with the ethylene precursor 1-aminocyclopropane-1-carboxylic acid (ACC) led to a dramatic reduction in the expression of these ARRs and blocked their induction by cold temperatures, thereby reducing the freezing tolerance of the plants. Shi et al. therefore proposed that ARR5, ARR7 and ARR15 represent a key node that integrates ethylene and cytokinin signalling in regulating the freezing response of plants.

Abscisic acid (ABA) is a plant hormone involved in seed dormancy, growth inhibition and stress response. Though not much is known about the interaction between cytokinin and ABA, several experiments have indicated that crosstalk occurs (Fig. 4C). Microarray and RT-PCR experiments demonstrated that CKX1, CKX3, CKX4 and CKX6 are downregulated by ABA (Werner et al., 2006). In addition, cytokinin receptor mutants show increased drought and salinity tolerance, a process known to involve ABA; a genome-wide expression analysis identified many upregulated ABA-responsive genes in the ahk2 ahk3 double mutant (Tran et al., 2007). Furthermore, Jeon et al. (Jeon et al., 2010) showed that AHK2 and AHK3 are involved in mediating the cold stress response by inhibiting ABA signalling via type A ARRs, although this may occur independently of endogenous cytokinin.

Originally identified from branching mutants of Arabidopsis thaliana, Pisum sativa and other plants, strigolactones are a class of phytohormones that inhibit the outgrowth of axillary buds. Cytokinins are also known to play a role in axillary branching in the shoot. Increased levels of strigolactones in pea and Arabidopsis lead to reduced cytokinin levels in the xylem sap, although levels in the shoot remain unchanged (Foo et al., 2007). Dun et al. (Dun et al., 2009) found that the buds of strigolactone-insensitive plants also have reduced sensitivity to cytokinin supplied from the vasculature. Cytokinin may also prevent auxin upregulation of MAX4, a strigolactone biosynthesis gene (Bainbridge et al., 2005). In a recent study, Dun et al. (Dun et al., 2012) found that the rms1 and rms4 pea mutants, which are deficient in and insensitive to strigolactones, respectively, had increased levels of PsIPT1 expression in the nodes and internodes of the shoot. The rms1 mutant was also more sensitive to low levels of cytokinin supplied to the vasculature or directly to the bud than wild type; buds grew out at lower levels of cytokinin in rms1 plants than in wild type. Taken together, these data suggest that cytokinin and strigolactone may act antagonistically in regulating bud outgrowth. Consistent with this, exogenous application of the synthetic strigolactone GR24 combined with cytokinin attenuated the effect of cytokinin on bud growth in rms1 but not in rms4 mutants, indicating that the effect of strigolactone on cytokinin is mediated by RAMOSUS4 (RMS4). The antagonism between strigolactone and cytokinin results from their common target, PsBRC1, which encodes a strigolactone-responsive TCP transcription factor thought to act downstream of strigolactones in non-growing buds. PsBRC1 expression, which is negatively correlated with bud growth (Aguilar-Martínez et al., 2007), was found to be upregulated by GR24 and downregulated by cytokinin even in the presence of the protein-synthesis inhibitor cycloheximide, indicating that this regulation does not require protein synthesis. Furthermore, PsBRC1 expression was found to increase upon treatment with cycloheximide, suggesting that it may be repressed by a rapidly turned over protein. Dun et al. (Dun et al., 2012) suggest that strigolactones and cytokinin may interact quite closely with BRC1 or with the stable proteins that regulate it (Fig. 3B; Fig. 4D; Table 3).

The past 5 years have seen significant advances in our understanding of the extensive crosstalk between cytokinin and various other phytohormones. Antagonisms and interactions that have been known from early physiological and anatomical experiments are now being understood at a molecular and genetic level. The picture that is emerging is one of dizzyingly interconnected networks at all levels, from synthesis and degradation to perception and transport. As the number of interconnections and the complexity of the feedbacks grow, it becomes increasingly challenging to understand intuitively the dynamics of these networks and form insightful hypotheses about their behaviour. The recent growth of computational and systems biology approaches is a welcome development that can help to alleviate these difficulties. Modelling hormonal interaction networks and simulating them in silico will be invaluable in helping us to understand their dynamics, allowing us to formulate better hypotheses and even to predict possible novel interactions and partners (Ibañes et al., 2009; Chickarmane et al., 2012; Naseem et al., 2012).

As we learn more about how tightly intertwined the hormone synthesis, metabolism, signalling and transport networks are, the intriguing possibility of considering them as one entity emerges. Auxin has a dominant presence in plant morphogenesis; it is an inescapable player in many developmental processes and a central component of the crosstalk networks, as well as being the first class of phytohormone described. Could it be that the main role of cytokinin and the other hormones is, in fact, to modulate auxin transport, signalling and homeostasis? Given the pervasive crosstalk between auxin and cytokinin, it is possible that most or all of the developmental effects of cytokinin somehow involve auxin, raising the issue of whether there exist any truly auxin-independent cytokinin-driven processes. If so, should auxin be considered the primary driver in plant development? Or should it instead be considered a read-out, integrating information about the total hormonal status of the plant? As we continue to discover how intertwined and self-referential the hormone networks are, we may come to consider both sides of the coin at once and find that it acts as simultaneously as a driver and as a readout.

Constraints of space prevented us from thoroughly discussing all aspects of cytokinin metabolism, signalling and interactions; we apologize to any colleagues whose work was not referenced.