Cytokinin is required for escape but not release from auxin mediated apical dominance

Auxin produced by an active primary shoot apex is transported down the main stem and inhibits the growth of the axillary buds below it, contributing to apical dominance. Here we use Arabidopsis thaliana cytokinin (CK) biosynthetic and signalling mutants to probe the role of CK in this process. It is well established that bud outgrowth is promoted by CK, and that CK synthesis is inhibited by auxin, leading to the hypothesis that release from apical dominance relies on an increased supply of CK to buds. Our data confirm that decapitation induces the expression of at least one ISOPENTENYLTRANSFERASE (IPT) CK biosynthetic gene in the stem. We further show that transcript abundance of a clade of the CK-responsive type-A Arabidopsis response regulator (ARR) genes increases in buds following CK supply, and that, contrary to their typical action as inhibitors of CK signalling, these genes are required for CK-mediated bud activation. However, analysis of the relevant arr and ipt multiple mutants demonstrates that defects in bud CK response do not affect auxin-mediated bud inhibition, and increased IPT transcript levels are not needed for bud release following decapitation. Instead, our data suggest that CK acts to overcome auxin-mediated bud inhibition, allowing buds to escape apical dominance under favourable conditions, such as high nitrate availability. Significance Statement It has been proposed that the release of buds from auxin-mediated apical dominance following decapitation requires increased cytokinin biosynthesis and consequent increases in cytokinin supply to buds. Here we show that in Arabidopsis, increases in cytokinin appear to be unnecessary for the release of buds from apical dominance, but rather allow buds to escape the inhibitory effect of apical auxin, thereby promoting bud activation in favourable growth conditions.


INTRODUCTION
Branches develop from axillary meristems (AMs), established in the axils of leaves made by the primary shoot apical meristem (SAM). Axillary meristems typically produce a few leaves before entering a dormant state as an axillary bud. The degree of branching is determined by the subsequent reactivation of these buds. Branching impacts upon light harvest, biomass and seed yield, but must be modulated by the availability of resources and competing growth requirements in the roots. The ability of plants to regulate bud outgrowth in response to environmental parameters is therefore of considerable adaptive significance and must be finely tuned. This is achieved in part by a network of hormones, which together integrate environmental and endogenous inputs. A central player is auxin, an inhibitor of bud outgrowth, which is synthesised in young leaves (Ljung et al., 2001) and transported rootward in the polar auxin transport stream (PATS). Apical decapitation, removing the main auxin source, results in the sustained outgrowth of buds in the axils of leaves below, which can be prevented by applying auxin to the decapitation site (Thimann and Skoog, 1933). Radio-label experiments have demonstrated that auxin acts indirectly, as it does not enter buds in significant amounts (Hall and Hillman, 1975;Morris, 1977;Prasad et al., 1993;Booker et al., 2003;Petr a sek and Friml, 2009).
Two general mechanisms for indirect inhibition by auxin have been proposed. Firstly, the canalisation-based hypothesis states that for a bud to activate it must establish its own PATS into the main stem (Prusinkiewicz et al., 2009). Activation is prevented by high auxin in the main stem, making it a weak sink for auxin from the bud (Sachs, 1981(Sachs, , 2000. In this model, shoot apices compete for access to the main stem PATS. The second hypothesis proposes that auxin affects the level of a second messenger, which moves into the bud and regulates bud activity. Strigolactone (SL) and cytokinin (CK) are both good candidates for this second messenger. Strigolactone biosynthetic genes are upregulated by auxin in the main stem (Sorefan et al., 2003;Foo et al., 2005;Johnson et al., 2006;Zou et al., 2006;Arite et al., 2007;Hayward et al., 2009;Zhang et al., 2010;Waters et al., 2012) and direct application of SL to buds can inhibit their activity (Gomez-Roldan et al., 2008;Brewer et al., 2009;Hamiaux et al., 2012). In contrast, CK biosynthetic genes are downregulated by auxin in the main stem (Tanaka et al., 2006) and direct application of CK to buds causes activation (Wickson and Thimann, 1958;Pillay and Railton, 1983;Cline et al., 1997). In pea, there is a strong correlation between bud outgrowth and the CK:SL ratio in buds, evidenced using exogenous hormone application and SL mutant analysis. This is proposed to be mediated by the opposite effects of SL and CK on the expression of BRANCHED1 (Braun et al., 2012;Dun et al., 2012), a member of the TEOSINTE BRANCHED1, CYCLOIDEA and PRO-LIFERATING CELL FACTORS 1 and -2 (TCP) gene family with a well-supported role in inhibiting bud activity (Doebley et al., 1997;Aguilar-Mart ınez et al., 2007;Mart ın-Trillo et al., 2011).
An alternative mechanism for the action of SL has been demonstrated, which strongly supports the canalisationbased model for bud regulation. A primary response to SL is the rapid depletion of the auxin transporter PIN1 from the plasma membrane, thus compromising canalisation of auxin transport from the bud into the main stem (Bennett et al., 2006;Crawford et al., 2010;Shinohara et al., 2013). This mode of action explains why, in several species, the addition of SL can only inhibit bud activity in the presence of a competing auxin source (Prusinkiewicz et al., 2009;Crawford et al., 2010;Liang et al., 2010;Ward et al., 2013), and why SL can sometimes promote bud activation (Liang et al., 2010;Shinohara et al., 2013;Ward et al., 2013).
Our understanding of the mechanism of action of SL depends heavily on mutants defective in SL synthesis and signalling. In contrast, most data on the role of CK in branching are based on monitoring the effects of the addition of CK on bud activity, and on correlations between CK levels or biosynthetic gene expression, auxin levels and bud activity (Cline, 1994;Turnbull et al., 1997;Bangerth et al., 2000;Nordstr€ om et al., 2004;Tanaka et al., 2006). Although these data support the second messenger hypothesis, it is clear from analysis of the action of SL that they are equally consistent with alternative explanations.
Genetic approaches to understanding the action of CK in shoot branching have been difficult, because many of the genes involved are present in large families (for review see Sakakibara, 2006;Hwang et al., 2012), and higher-order mutations exert pleiotropic effects, particularly on shoot growth and meristem activity (Higuchi et al., 2004;Leibfried et al., 2005;Miyawaki et al., 2006;Tokunaga et al., 2012). Here we describe a targeted approach to investigate the role of CK synthesis and response in shoot branching in Arabidopsis. Our results suggest that CKs play little part in auxin-mediated bud repression and release from apical dominance, but rather they provide a mechanism for buds to escape apical dominance and activate even in the presence of auxin.

Isopentenyltransferases contribute to branching in intact Arabidopsis plants
Isopentenyltransferases (IPTs) catalyse an early step in CK biosynthesis, encoded by nine genes in Arabidopsis (Kakimoto, 2001;Takei et al., 2001a). IPT1 and IPT3-IPT8 are ATP/ADP IPTs and contribute to iP-and tZ-type CK synthesis (Miyawaki et al., 2006). As iP-and tZ-type CKs comprise the major CKs in Arabidopsis, these IPT genes were chosen for further investigation. IPT4, IPT6 and IPT8 are predominantly expressed in floral tissues, and are therefore unlikely to contribute to bud activation (Takei et al., 2001a;Miyawaki et al., 2004). The remaining four genes, IPT1 (At1g68460), IPT3 (At3g63110), IPT5 (At5g19040) and IPT7 (At3g23630), are expressed in roots and vegetative shoot tissues, and could be involved in shoot branching.
In pea, CK is proposed to act as a second messenger for auxin because auxin in the PATS represses expression of PsIPT1 and PsIPT2, whilst decapitation reduces stem auxin levels, resulting in increased IPT expression and levels of iP-and tZ-type CK in the stem (Tanaka et al., 2006). To test whether Arabidopsis IPT genes are similarly responsive to decapitation, the four selected IPT genes were assayed using an isolated two-node system (Ongaro et al., 2008;Prusinkiewicz et al., 2009). Two-node stem segments bearing the intact apex and inactive buds were collected and left for several days to allow the stem to elongate (Prusinkiewicz et al., 2009). The buds on these isolated segments grow out as normal if the apex is removed ( Figure S1). Stem tissues were harvested 0 and 6 h after decapitation and expression levels of IPT1, IPT3, IPT5 and IPT7 determined relative to the reference gene UBC21 using quantitative real-time PCR (qPCR) (Figure 1a). Decapitation resulted in an eight-fold increase in IPT3 compared with intact controls (P ≤ 0.01), while the remaining IPT transcripts were not significantly affected.
Next, the effect of IPT loss-of-function on branching was investigated. Rosette branches were counted at maturity in single and higher-order mutant combinations of the four IPT genes (Miyawaki et al., 2006) (Figure 1b). Single mutants showed significant reductions in branching: ipt1, ipt3, ipt5 and ipt7 all formed less than one branch on average (P < 0.008 for all genotypes compared with the wild type). The ipt3,5,7 mutant ( Figure 1c) formed a similarly low number of branches and ipt1,3,5,7 formed no branches. This is consistent with the well-established role for CK in promoting branching.
We used grafting to test whether the IPT genes are required in the root or shoot to promote branching in intact plants. As ipt1,3,5,7 is severely stunted, reciprocal grafts were performed between the ipt3,5,7 triple mutant and the wild type. At maturity, self-grafted ipt3,5,7 mutants possessed reduced branching compared to wild-type controls (P ≤ 0.01), and the presence of either a wild-type root or shoot reciprocally grafted to ipt3,5,7 was sufficient to confer wild-type levels of branching ( Figure 1e). Cytokinins produced in roots can therefore compensate for reduced CK synthesis in ipt3,5,7 shoots, suggesting that CKs produced in the whole plant, rather than CKs produced in the shoot alone, contribute to branching in intact plants.
We next tested whether these IPT genes are required for auxin-mediated apical dominance by analysing the decapitation response of the ipt mutants. In particular, ipt3 might have reduced branching as IPT3 expression is responsive to decapitation (Figure 1a), and IPT expression in the stem correlates with CK levels and bud outgrowth in pea (Tanaka et al., 2006). Mature plants were decapitated at the base of the bolt and rosette branch numbers were counted after 1 week ( Figure 1d). Surprisingly, most of the mutants, including ipt3, responded as the wild type, producing between 6.4 and 7.3 rosette branches. The exceptions to this were ipt3,5,7 (2.4 AE 0.3 branches, P < 0.008), and ipt1,3,5,7 (0.3 AE 0.1, P < 0.008).
Inspection of ipt3,5,7 and ipt1,3,5,7 showed that many of the axils lacked visible axillary buds, suggesting a failure in bud initiation, which would clearly preclude bud activation. We chose ipt3,5,7 and the three cognate single mutants to quantify this axil phenotype in more detail. All genotypes were found to have some empty axils in more basal rosette nodes (Figure 2a-e) when examined by eye. There was some evidence of an increased number of empty axils in ipt3 (Figure 2a,b), which had 20.3% empty axils compared with 11.5% in the wild type (P < 0.001) (Figure 2f). Both ipt5 ( Figure 2c) and ipt7 ( Figure 2d) had similar numbers of empty axils to the wild type. In ipt3,5,7 the majority of rosette axils bore no visible bud (69.6%, P < 0.001 compared with the wild type) (Figure 2a,e). The reduction in decapitation-induced branching observed in ipt3,5,7 (and ipt1,3,5,7) ( Figure 1c) is therefore probably due to compromised bud development. Consistent with this, scanning electron microscopy (SEM) revealed frequently empty axils in ipt3,5,7 (Figure 2g).
Taken together, the ipt mutant analysis suggests that CK synthesis is required to allow high levels of branching in intact plants but not for activation of buds following (e) Total number of active branches (≥1 cm) in reciprocal shoot/root grafts between the wild type and ipt3,5,7 at the onset of senescence (n = 9-19). For (a) statistical comparisons were made between intact and decapitated samples using t-tests; asterisks denote a significance level of P < 0.01 (**). For (b), (d) and (e) statistical comparisons were made between the wild type and each mutant using Mann-Whitney tests with a Bonferroni correction to adjust for multiple comparisons; asterisks (**) denote a significance level of P < 0.0083 in (b) and (d), or P < 0.0167 in (e). decapitation. To confirm that ipt3,5,7 buds are capable of responding to apical auxin, we investigated the response of ipt3,5,7 buds to apical 1-naphthalene acetic acid (NAA) in an isolated nodal assay (Chatfield et al., 2000). The outgrowth of wild-type buds was delayed by apical auxin application as expected, and no significant differences were found between ipt3,5,7 and the wild type over time in either treatment ( Figure 3). Together, these data suggest that ipt3,5,7 buds can grow with wild-type kinetics and are auxin responsive, and therefore their reduced branching following decapitation ( Figure 1c) is likely to be due to defects in bud initiation. These results suggest that the main role of CK in Arabidopsis branching is to allow the activation of buds in the presence of high auxin, thereby escaping apical dominance. Reduced CK synthesis, as in the ipt mutants, does not increase the sensitivity of buds to apical auxin, and high IPT activity is not required to activate buds following decapitation. These results are consistent with previous observations that CKs have relatively little effect on the outgrowth of isolated buds, but they can overcome the suppressive effect of apical auxin (Chatfield et al., 2000). Figure 3. ipt3,5,7 buds are responsive to apical auxin. Isolated nodal segments bearing one bud were treated apically for 4 days with 1 lM NAA or mock treatment. The mean bud length AE SE is shown (n = 7-31). No statistically significant differences were found between genotypes over time in NAA treatments or mock treatments. Statistical comparisons were carried out using t-tests or Mann-Whitney tests.

Identification of auxin-and CK-responsive genes in buds
To investigate the effect of CK on buds inhibited by apical auxin, we compared the transcriptomes of mock-treated buds with those supplied with apical auxin alone or apical auxin combined with basal CK (6-benzylaminopurine, BA) in isolated nodes using the Affymetrix Arabidopsis gene chip ATH1. Buds were harvested after 18 h, allowing sufficient time for transcriptional changes to occur but before major size differences become apparent. Comparison of the two treatments that activate buds, mock versus simultaneous auxin plus CK, revealed only two genes with statistically significant differences in expression: At1g75450, encoding a cytokinin oxidase (CKX5) (Schm€ ulling et al., 2003;Werner et al., 2003), and At2g33830, a bud dormancy-associated gene (DRM2) (Tatematsu et al., 2005). Both genes were more highly expressed in buds treated with auxin plus CK than mocktreated buds.
To identify genes associated with bud activation, we focused on transcripts that were statistically significantly two-fold or more downregulated by apical auxin compared to mock-treated, but two-fold or more upregulated by simultaneous supply of basal CK compared to auxin alone. The number of genes meeting these criteria was 220 (Data S1). To select genes that might be particularly relevant for CK-mediated bud activation, we ranked them according to their upregulation in auxin plus CK compared with auxin alone. This produces a list in which the top 12 genes are upregulated five-fold or more (Table 1). Of these 12, 11 have at least one ARR1 response element (AAGATT) within 2000 bp upstream of their start codon, consistent with their CK responsiveness (Sakai et al., 2000(Sakai et al., , 2001Taniguchi et al., 2007). Included in the 12 are CKX5 and five members of a six-gene clade of type-A ARRs. The type-A ARRs are primary CK response genes, and are generally considered to be involved in feedback downregulation of CK signalling (Brandstatter and Kieber, 1998;Imamura et al., 1998;D'Agostino et al., 2000;To et al., 2004To et al., , 2007. This type-A ARR clade, comprising ARR3 (At1g59940), ARR4 (At1g10470), ARR5 (At3g48100), ARR6 (At5g62920), ARR7 (At1g19050) and ARR15 (At1g74890), was selected for further analysis of CK-mediated bud activation.

A clade of the type-A ARR family modulates intact branching patterns
We first verified the expression profile of the five ARR genes using qPCR. All genes were upregulated 2.5-to 11fold in buds treated with basal CK and apical auxin compared with apical auxin alone ( Figure S2).
To investigate the role of this type-A ARR clade in bud activation, the hextuple arr3,4,5,6,7,15 mutant branching phenotype was examined in mature plants left intact ( Figure 4a) or decapitated (Figure 4b). Similar to Table 1 The top 12 transcripts downregulated by apical auxin and upregulated by auxin and cytokinin (CK). All transcripts were classed as significantly (two-fold or more) decreased in auxin-treated buds compared with mock-treated buds, and increased fivefold or more in auxin and CK-treated buds compared with auxintreated buds  ipt mutants, intact arr3,4,5,6,7,15 plants formed significantly fewer rosette branches than the wild type (2.5 AE 0.1 versus 3.9 AE 0.1, respectively, P ≤ 0.001) but there was no significant difference in decapitationinduced branching.
We used the isolated one-node assay to assess the CK response of arr3,4,5,6,7,15 buds. Mock-treated mutant buds activated with normal kinetics and were inhibited by apical auxin, as for the wild type ( Figure 5). However, the ability of CK to overcome inhibition by apical auxin was compromised in arr3,4,5,6,7,15. Over time, there was no significant difference in length between mutant buds treated with apical auxin and those treated with apical auxin and basal CK, whereas wild-type buds treated with apical auxin and basal CK were significantly longer (P < 0.01) than their apical-auxin-only equivalents from 48 h onward (Figure 5b).
To determine whether transcriptional responses to CK are impaired in arr3,4,5,6,7,15 buds, we used several genes from Table 1 as markers and analysed their expression in wild-type and arr3,4,5,6,7,15 buds treated with or without basal CK in the presence of apical auxin (Figure 5c). Despite the fact that CK-treated wild-type buds activate while their arr3,4,5,6,7,15 equivalents do not, CK treatment resulted in the upregulation of all the CK-responsive transcripts tested in the mutant buds at least as strongly as in wild-type buds.
Together, these results suggest that this type-A ARR clade is required for CK-mediated bud activation but not for bud transcriptional responses to CK. Furthermore, the results are consistent with the conclusion from the IPT gene analysis that CK is not required for release from auxin-mediated apical dominance, but rather acts to overcome the inhibitory effects of apical auxin, promoting branching in intact plants.

Strigolactone does not rescue decapitation-induced bud outgrowth in CK mutants
It has previously been suggested that bud activity represents a read-out of the ratio of CK to SL in buds, with both being influenced by auxin in the main stem (Dun et al., 2012). This may explain why in many species SL is unable to inhibit buds in the absence of a competing auxin source, as high CK levels resulting from low auxin might render the buds resistant to SL. According to this hypothesis, Figure 5. arr3,4,5,6,7,15 buds have impaired outgrowth to CK but retain normal gene expression responses. (a) Wild-type and (b) arr3,4,5,6,7,15 isolated nodal segments bearing one bud were treated for 8 days with mock, 1 lM NAA (apically), 1 lM 6-benzylaminopurine (BA; basally) or combined 1 lM NAA (apically) and 1 lM BA (basally). The mean AE SE is shown (n = 20). Statistical comparisons shown were made between NAA and NAA + BA treated buds using t-tests and Mann-Whitney tests as the data did not always follow a normal distribution. Asterisks denote a significance level of P < 0.05 (*), P < 0.01 (**) or P < 0.001 (***); N.S, not significant. (c) Expression of cytokinin-responsive genes in isolated buds treated with NAA and BA as in (a)  arr3,4,5,6,7,15 buds may respond normally to decapitation despite their lack of CK response because of low levels of SL. If this is the case, isolated arr3,4,5,6,7,15 buds should fail to activate in the presence of high SL. We therefore tested the response of arr3,4,5,6,7,15 buds to the synthetic SL GR24 ( Figure 6). As previously described by Crawford et al. (2010), wild-type buds activated similarly with or without basal SL, and SL applied in combination with apical auxin inhibited buds to a greater extent than auxin alone (P ≤ 0.01). arr3,4,5,6,7,15 buds responded similarly to the wild type. There was a slight reduction in SL-treated arr3,4,5,6,7,15 bud elongation relative to mock, but this was not statistically significant. Additionally, arr3,4,5,6,7,15 buds exhibited some resistance to SL, as there was no effect of SL on bud activation in the presence of apical auxin. The finding that wild-type and arr3,4,5,6,7,15 buds alike activate under high SL suggests that depletion of auxin in the main stem alone is sufficient to trigger sustained bud activity, with no requirement for low SL and/or high CK, consistent with the canalisation-based hypothesis for bud activation.

Branching response to high nitrate is impaired in higherorder CK mutants
Our data support the hypothesis that the main role of CK in branching is to drive bud activation in the presence of high auxin. One situation where this might be physiologically relevant is during changes in nitrate availability. Low nitrate availability is known to increase in the proportion of biomass allocated to roots versus shoots (Drew, 1975;Scheible et al., 1997) and high N is associated with bud activation (McIntyre and Hunter, 1975;McIntyre and Cessna, 1991;Ding et al., 1995;McIntyre, 2001;Liu et al., 2011;de Jong et al., 2014). As nitrate is known to promote CK synthesis (Takei et al., 2001b(Takei et al., , 2002(Takei et al., , 2004, CK might drive the activation of additional shoot branches despite the high main stem auxin contributed by already active shoots. To test this, we compared total branch numbers of CK mutants with the wild-type under high-and low-nitrate conditions (Figure 7). Although ipt3 and ipt5 single mutants have similar branching levels to the wild type when grown under these conditions, the ipt3,5,7 and arr3,4,5,6,7,15 mutants have significantly fewer active branches than the wild type on high nitrate. In contrast, there was no difference between the wild type and the mutants on low nitrate. These data suggest that CK is not required to support branching on low nitrate but that it may be important to enhance branching under high nitrate.

DISCUSSION
Many of the genes involved in CK biosynthesis and signalling are members of large families, bringing the combined problems of functional redundancy and pleiotropy to mutant analyses. To investigate the role of CK in shoot branching in Arabidopsis we selected mutants of only the most relevant members of the IPT family of CK biosynthetic genes and the ARR family of CK signalling genes based on transcriptional profiles. The fact that we observed similar mutant phenotypes for the selected IPT and ARR genes suggests that this strategy was effective, though some differences were observed. For example, our results support a role for IPT-mediated CK synthesis in bud initiation (Figure 2; Wang et al., 2014) that apparently does not require the ARRs involved in bud activation. (a) Wild-type and (b) arr3,4,5,6,7,15 isolated nodal segments bearing one bud were treated with mock, 0.5 lM NAA (apically), 5 lM GR24 (basally) or combined 0.5 lM NAA and 5 lM GR24 (basally). The mean AE SE on day 6 is shown (n = 20). Statistical comparisons shown were made using t-tests or Mann-Whitney tests as the data did not always follow a normal. Asterisks denote a significance level of P < 0.01 (**); N.S., not significant. Figure 7. The branching plasticity response to high N is reduced in higherorder cytokinin mutants. Plants were grown on soil-free medium under high (9 mM NO 3 ) or low (1.8 mM NO 3 ) nitrate conditions and the total number of active rosette and cauline branches (≥1 cm) counted at the onset of senescence. The mean AE SE is shown (n = 11-12). The statistical comparisons shown were made between the wild type and each mutant on high nitrate using Mann-Whitney tests with a Bonferroni correction to adjust for multiple comparisons; asterisks denote a significance level of P < 0.0125 (**). No significant differences were found between the wild type-and ipt3 or ipt5 on high nitrate or between the wild type and any mutants on low nitrate. Type-A ARR multiple mutants have CK-response phenotypes in roots Zhang et al., 2011). It is therefore possible that some of the effects observed in shoots are influenced by pleiotropy. Furthermore, it cannot be excluded that other family members may play roles in bud activation, for example rarer CK types dependent on IPT2 and IPT9 may be involved. Nonetheless, the combined dataset demonstrates the relatively direct effects of CK on bud activation, and that CK synthesis and signalling contribute to branching in Arabidopsis. Particularly compelling in this regard are the wild-type activation and growth kinetics of the CK-related mutant buds in response to decapitation, demonstrating the general vigour of the buds, and the CK resistance of arr hextuple mutant buds in isolated nodal stem segments, where there are no roots.

The role of CK in auxin-mediated apical dominance
Our results in Arabidopsis are consistent with a substantial body of evidence that CK promotes bud outgrowth in many species. This, in combination with correlative studies, has led to the hypothesis that auxin inhibits bud outgrowth by restricting the supply of CK to buds. This hypothesis is well supported in pea by the transcriptional responses of two IPT genes to decapitation and auxin supply, and changes in CK levels in the stem (Tanaka et al., 2006;Shimizu-Sato et al., 2009). Auxin repression of stem IPT gene expression is conserved across species, for example in chrysanthemum (Chen et al., 2013) and rice (Minakuchi et al., 2010). We found similar results in Arabidopsis, but only for IPT3 (Figure 1a). Interestingly, these results contrast with those in roots, where auxin has little effect on IPT3 but upregulates IPT5 and IPT7 (Miyawaki et al., 2004). The auxin responsiveness of different IPT family members may therefore be determined by tissue-specific factors.
The transcriptional response of IPT3 to decapitation is consistent with the idea that auxin-mediated bud inhibition involves a reduced supply of CK. Similarly, our observation that buds activated by decapitation have virtually identical transcriptomes to buds activated by basal CK supply in the presence of apical auxin could reflect an important role for high CK in bud activation following decapitation. However, our mutant analysis does not support this hypothesis. Multiple ipt mutant buds were unaffected in their response to apical auxin or to presumed auxin depletion in the main stem PATS following decapitation. Furthermore, arr3,4,5,6,7,15 buds, despite being severely compromised in CK-mediated bud activation, are fully capable of responding to apical auxin and decapitation. In Arabidopsis it therefore seems that CK is not required for release from auxin-mediated apical dominance following decapitation, but rather it allows buds to escape apical dominance and activate even when auxin in the PATS is high.
In pea it has been suggested that bud activity is determined by the ratio of CK and SL in buds (Braun et al., 2012;Dun et al., 2012). Auxin-mediated changes in SL levels could explain the normal activation of arr3,4,5,6,7,15 mutant buds following decapitation (Figure 4b) and their inhibition by apical auxin (Figure 5b). However, this seems unlikely because SL is unable to prevent the activation of arr3,4,5,6,7,15 buds on isolated nodes ( Figure 6). Our results from Arabidopsis are therefore not consistent with CK:SL ratios regulating bud activity, highlighting the need to better understand the mechanism by which CK activates buds.

Transcriptional CK responses in buds
There are several hypotheses about how CK promotes bud outgrowth (for a review see M€ uller and Leyser, 2011). Most involve CK-induced transcription of genes such as those involved in cell-cycle regulation or meristem function, and there is good evidence that the CK signalling network regulates the transcription of such genes (Riou-Khamlichi et al., 1999;Suzuki et al., 2001;Dewitte et al., 2007;Braun et al., 2012;Dun et al., 2012).
We identified 220 genes that are significantly (two-fold or more) downregulated by apical auxin and also significantly (two-fold or more) upregulated by simultaneous basal CK supply (Table S1). This includes several type-A ARRs and CK metabolism genes, which have been repeatedly identified as CK-responsive in meta-analyses of CK microarrays (Brenner et al., 2012;Bhargava et al., 2013). The list also includes several genes involved in DNA replication, but not classical CK cell-cycle targets such as cyclin D3 (Riou-Khamlichi et al., 1999) or known CK-regulated meristem regulators such as WUS (Lindsay et al., 2006;Gordon et al., 2009). However, the list does include three members of the FANTASTIC FOUR (FAF) family, of which FAF2 has been implicated in regulating shoot meristem size. Over-expression of FAFs results in reduced meristem size and reduced WUS expression, and FAF2 is expressed in the WUS domain of the shoot apical meristem (Wahl et al., 2010). FAF1 and FAF3 can also influence meristem size when ectopically expressed, but their native expression pattern is primarily in the vasculature. FAF3 has previously been identified as robustly inducible by CK (Bhargava et al., 2013).
An additional hypothesis for the mechanisms of action of CK in bud activation is promotion of auxin export from buds. This could be at the transcriptional level, and, consistent with this hypothesis, several genes involved in auxin biosynthesis and auxin transport are upregulated in the auxin plus CK treatment. This includes the auxin transporter ABCB21 (Table 1) (Kamimoto et al., 2012).

The role of type-A ARRs in bud activation
To assess the role of the upregulated type-A ARRs in the control of bud growth we analysed a hextuple mutant lacking all members of the upregulated clade. The mutant buds were strongly resistant to activation by basally applied CK, consistent with their reduced branching phenotype in intact plants. The inability of these buds to activate in response to CK raises interesting questions about the mechanism of action of CK in promoting bud growth. The hextuple mutant buds are apparently fully responsive to CK at the transcriptional level (Figure 5c), suggesting that these transcriptional changes are insufficient to activate buds in the absence of the type-A ARR clade. Given the reduced branching phenotype of the mutant, but normal growth kinetics in active branches, a likely explanation is that there are essential type-A ARRdependent non-transcriptional roles for CK in bud activation.

Post-transcriptional CK responses
In roots, CK reduces the abundance of some PIN auxin efflux carriers at the plasma membrane via a post-transcriptional mechanism (Marhav y et al., 2011;Zhang et al., 2011). For PIN1, this effect is dependent on AHK4 and two type-B ARRs (Marhav y et al., 2011). Type-A arr mutants are hypersensitive in their PIN responses to CK, suggesting that they function as negative regulators of CK with respect to this non-transcriptional pathway (Zhang et al., 2011). In light of this our results appear paradoxical, because loss of multiple type-A ARRs should lead to hypersensitivity to CK yet arr3,4,5,6,7,15 buds exhibit CK resistance and intact arr3,4,5,6,7,15 plants have reduced branching, similar to the loss of certain IPT CK biosynthetic genes.
This paradox can be resolved if the primary target for CK signalling in bud regulation is plasma membrane accumulation of PIN proteins, and bud activation requires canalisation of auxin transport from the bud to the stem, as proposed in the canalisation-based model for branching control. In roots, there is some evidence of differential sensitivity to CK of PINs on different cell faces (Marhav y et al., 2014). If this is also the case for shoots, then low levels of CK might promote the establishment of auxin transport canalisation out of buds by removing PINs differentially from non-rootward cell faces. If arr3,4,5,6,7,15 buds and/or shoots are hypersensitive to CK this could result in broader PIN removal, including rootward PINs, reducing the ability to canalise auxin transport out of the bud. This idea is highly speculative, and it certainly cannot be assumed that the behaviour of PIN in roots and buds will be the same; however, it does provide a possible explanation for the molecular and physiological phenotypes of the arr3,4,5,6,7,15 mutants, which can be tested at a cellular level.

The response of CK and branching to the environment
Our data strongly suggest that in Arabidopsis CK drives increased bud activation when auxin levels are high. This solves a major problem with apical dominance as a mechanism for modulating shoot architecture. Under this model, if a vigorous apex exports auxin and inhibits the buds below it, apical dominance would be strong under favourable conditions, restricting the plant's capacity to produce branches despite resource abundance. A canalisation-based mechanism for apical dominance is less problematic, because in this case bud activation is essentially a competitive process; conditions that increase the vigour of the primary apex will also increase the vigour of lateral buds. The mode of action for CK suggested by our data allows for the exploitation of favourable conditions through CK-mediated escape from apical dominance. There is evidence that CK synthesis is increased by high nitrate (for review, see Kiba et al., 2011) at least partly via the upregulation of IPT gene expression (Miyawaki et al., 2004;Takei et al., 2004), and that CK can act as a systemic signal for nitrate status (Ruffel et al., 2011), consistent with our grafting results (Figure 1e). Our findings suggest that an N-induced increase in CK contributes to increased branching because both ipt3,5,7 and arr3,4,5,6,7,15 have wild-type-levels of branching when grown under lownitrate conditions, but they are compromised in their ability to activate more buds when nitrate levels are higher (Figure 7). In the case of ipt3,5,7, this could be partly due to bud initiation defects, but this is not the case for arr3,4,5,6,7,15.
Further consideration of this mechanism also provides a possible function for the downregulation of CK in the stem by auxin. If unchecked, escape from auxin-mediated apical dominance could lead to the activation of all buds. This could be prevented by downregulation of CK biosynthesis by additional auxin, supplied by the newly activated branches. Thus, auxin and CK can act in a feedback loop, the equilibrium of which can be adjusted by auxin-independent CK synthesis driven by external inputs such as nutrient availability or light.

Plant lines and growth conditions
Col-0 was used as the wild-type controls and all mutant lines were on a Col-0 background. Seeds were stratified for 3-5 days at 4°C. The ipt1-1, ipt3-2, ipt5-2 and ipt7-1 single, double, triple and quadruple mutant combinations used were those described in Miyawaki et al. (2006). The arr alleles are those previously described in Zhang et al. (2011). All soil-grown plants were sown on F2 compost in P40 trays treated with Intercept 70WG (both Levington, http://www.scottsprofessional.co.uk) then transferred to the glasshouse or controlled environment rooms. Glasshouse conditions comprised a long-day photoperiod (16-h light/8-h dark) and an average temperature range of 15-25°C. Controlled environment conditions comprised either long-day or short-day (8-h light/16-h dark) photoperiods, an average light intensity of 170 or 100 lmol m À2 sec À1 for soil-grown and sterile-grown plants, respectively, and an average temperature range of 17-21°C.

Statistical analyses
Based on the assumption that branch numbers do not follow a normal distribution, the Mann-Whitney test was used for statistical comparisons of non-parametric data and a Bonferroni correction applied for multiple comparisons. Kolmogorov-Smirnov and Shapiro-Wilk tests for normality were used to determine if other datasets followed a normal distribution. For parametric distributions, statistical analyses were performed using two-tailed t-tests and interpreted using Levene's test for homogeneity of variance.

Intact and decapitated branch counts
For ipt mutant branch counts (Figure 1b,c), plants were soil-grown under short days in controlled-environment rooms for 6 weeks then transferred to long-day glasshouse conditions. After flowering, plants were left intact or decapitated at the base of the bolt and rosette branch numbers ≥0.5 cm recorded after 1 week. For arr3,4,5,6,7,15 mutant branch counts (Figure 5a), intact plants were grown under glasshouse conditions and the number of branches ≥0.5 cm were recorded when plants had developed at least two full siliques. The decapitation assays shown in Figure 5(b) were performed as per Greb et al. (2003).

Grafting
Reciprocal grafts between the wild type and ipt3,5,7 (Figure 1e) were performed using the transverse cut and butt alignment method as per Turnbull et al. (2002) with the following modifications. Seeds were germinated on Arabidopsis thaliana salts (ATS) medium (Wilson et al., 1990) containing 0.8% bacto-agar and grown on vertically mounted Petri dishes under sterile short-day controlled-environment conditions as outlined above. Seven-dayold seedlings were used for grafting and 7-9 days after grafting, plants showing root growth and no adventitious roots were transferred to sand and Terra-Green on a high-nitrate regime as outlined below. The total number of active rosette and cauline branches (≥1 cm) were counted at the onset of senescence when the oldest siliques had ripened.
ipt axil characterisation Axils were examined by eye via light microscopy (Figure 2a-f) and buds classified as present if at least one bud leaf was discernible. For SEM (Figure 2g,h), plants were soil-grown for 6 weeks under short-day conditions in controlled environment rooms as described. Emerging bolts, roots and rosette leaf blades were removed and the remaining rosette stems fixed with 4% paraformaldehyde, washed in phosphate buffer then dehydrated in an acetone series up to 100%. Critical point drying was performed at the York University Technology Facility. Specimens were subsequently coated with gold/palladium and imaged using a JEOL JSM-6490LV microscope (http://www.jeol.co.jp/en/).

One-node assays
Plants were grown under sterile long-day conditions as described in Bennett et al. (2006). Cauline stem segments bearing the basal inactive bud were assayed as per Chatfield et al. (2000). For CK treatment, BA (Sigma, http://www.sigma.com) was dissolved in DMSO and applied basally as a 10009 stock to a final concentra-tion of 1 lM. For auxin treatment, NAA (Sigma) was dissolved in 70% ethanol and applied apically as a 10009 stock to a final concentration of 0.5 or 1 lM. GR24 (LeadGen Labs, http://www.leadgenlabs.com/) was dissolved in acetone and applied as a 10009 stock to a final concentration of 5 lM. For mock treatments, DMSO or 70% ethanol was applied basally or apically, respectively, at 0.1% v/v.

Microarray study
One-node segments were treated with apical NAA alone, apical NAA and basal BA or mock controls as described. Buds were harvested onto liquid nitrogen after 18 h. Total RNA was extracted using the RNeasy Micro Kit (Qiagen; http://www.qiagen.com/). For each microarray, 300 ng total RNA input was used. The RNA labelling and preparation for hybridisation was carried out using an Ambion MessageAmp III aRNA amplification kit (http://www.lifetechnologies.com/) according to the manufacturer's instructions. An Affymetrix GeneChip Poly-A RNA control kit was used for controls (http://www.affymetrix.com/). The Affymetrix Arabidopsis gene chip ATH1 was hybridised. Three biological replicates were analysed. Statistical analyses of microarray data were performed using R software version 2.15 (http://www.r-project.org/) and the Affy (Gautier et al., 2004), Affycoretools and Bioconductor Limma packages. Raw data were normalised using the robust multichip analysis (RMA) algorithm. Differential gene expression between groups was then determined by fitting a linear model to the data using lmfit with subsequent comparisons made using the mak-eContrasts function. Transcripts with a q-value (Hochberg and Benjamini, 1990) of less than 0.05 and a fold change of two or more were classed as significantly differentially regulated. Microarray data were deposited at Gene Expression Omnibus (http:// www.ncbi.nlm.nih.gov/geo/), accession number GSE59741.

Quantitative real-time PCR
For IPT gene expression analysis (Figure 1a), two-node assays were set up as described and left intact for 4-6 days. Only segments bearing inactive buds were used. Plants were left intact or decapitated at the apex for 6 h, nodal stems (with axillary buds removed) harvested onto liquid nitrogen or RNAlater (Ambion, http://www.invitrogen.com/) into three pools of 10-15 segments each. Total RNA was extracted using the RNeasy Plant Mini kit (Qiagen) and subjected to DNAse treatment using the Turbo DNAfree kit (Ambion) as per the manufacturer's instructions. The RNA was quantified using a NanoDrop 1000. For cDNA synthesis, 1 lg of total RNA was reverse transcribed with Superscript II (Invitrogen, http://www.invitrogen.com/) according to the manufacturer's instructions. Quantification of transcript levels was carried out using SYBR Green reactions with 5 ng cDNA in a 20 ll volume on a Light Cycler 480 II (Roche, http://www.roche.com/) relative to the reference gene UBIQUITIN-CONJUGATING ENZYME 21 (UBC21, At5g25760). Three technical replicates were run for each biological replicate and averaged. Means shown represent the average of biological replicates. Expression levels were calculated using the Light Cycler 480 II software and the second-derivative maximum method assuming equal primer efficiencies.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article. Figure S1. Decapitation response of buds in an isolated two-node assay. Figure S2. Quantitative real-time PCR confirmation of ARR gene expression in buds treated with basal cytokinin and apical auxin. Table S1. Quantitative real-time PCR primer sequences Data S1. Transcripts downregulated by apical auxin and upregulation by auxin and cytokinin. Methods S1. Supporting experimental procedures.