Petunia hybrida CAROTENOID CLEAVAGE DIOXYGENASE7 is involved in the production of negative and positive branching signals in petunia.

One of the key factors that defines plant form is the regulation of when and where branches develop. The diversity of form observed in nature results, in part, from variation in the regulation of branching between species. Two CAROTENOID CLEAVAGE DIOXYGENASE (CCD) genes, CCD7 and CCD8, are required for the production of a branch-suppressing plant hormone. Here, we report that the decreased apical dominance3 (dad3) mutant of petunia (Petunia hybrida) results from the mutation of the PhCCD7 gene and has a less severe branching phenotype than mutation of PhCCD8 (dad1). An analysis of the expression of this gene in wild-type, mutant, and grafted petunia suggests that in petunia, CCD7 and CCD8 are coordinately regulated. In contrast to observations in Arabidopsis (Arabidopsis thaliana), ccd7ccd8 double mutants in petunia show an additive phenotype. An analysis using dad3 or dad1 mutant scions grafted to wild-type rootstocks showed that when these plants produce adventitious mutant roots, branching is increased above that seen in plants where the mutant roots are removed. The results presented here indicate that mutation of either CCD7 or CCD8 in petunia results in both the loss of an inhibitor of branching and an increase in a promoter of branching.

The dynamic process that leads to a plant's architecture is regulated by developmental factors and by environmental conditions. Whether or not axillary meristems grow to form branches is one key component of plant architecture. Plants with altered architecture have been important in agronomy since the earliest selections were made by humans. More recent examples are vital to the productivity of our current farming systems. The domestication of maize (Zea mays) and the dwarfing of wheat (Triticum aestivum) and rice (Oryza sativa; as part of the Green Revolution) involved alterations to plant height and branch num-ber that dramatically improved productivity (for review, see Sakamoto and Matsuoka, 2004).
Arabidopsis (Arabidopsis thaliana), rice, pea (Pisum sativum), and petunia (Petunia hybrida) are important model plants in which axillary branching has been studied. The growth habits of these plants show differences when grown under standard floral inductive conditions. This is due, in part, to the differing developmental programs controlling the outgrowth of axillary branches. Petunia (inbred genetic stock V26) produces basal axillary branches between nodes two and eight that begin their growth during the vegetative growth phase (Snowden and Napoli, 2003). Axillary branches may also form in the nodes immediately below the first flower after the floral transition (Napoli et al., 1999). Arabidopsis generally produces axillary branches after flowering, releasing axillary meristems in the rosette and also from cauline leaves (Hempel and Feldman, 1994). Wild-type, tall pea cultivars such as Parvus are very unlikely to produce basal axillary branches at any stage of growth but do branch at the nodes immediately below the first flower (Stafstrom, 1995). Cultivated rice produces basal axillary branches, called tillers, during vegetative growth. The tillers formed early in plant development will produce panicles (flowering branches), and the remainder will senesce (Hanada, 1993). How these differences in development arise is yet to be understood.
Although the overall architecture of plants varies considerably, the genes so far identified that control branching are frequently conserved between species.
In particular, two CAROTENOID CLEAVAGE DIOXY-GENASE (CCD) genes, CCD7 and CCD8, appear to be well conserved among the plant species studied. Mutations in these two genes result in increased branching phenotypes in every species studied to date (Sorefan et al., 2003;Booker et al., 2004;Snowden et al., 2005;Zou et al., 2005;Johnson et al., 2006;Arite et al., 2007). One interesting line of enquiry is to consider whether differences in the regulation or activity of these two genes are involved in the diversity of architecture seen in plants.
Grafting experiments have provided insight into the control of axillary branching, in particular the discovery that signals move from roots to shoots. In petunia, Arabidopsis, and pea, some of the increased branching mutants (ccd7 and ccd8 mutants in particular) can be reverted to a wild-type phenotype by grafting mutant scions onto wild-type rootstocks (for review, see Drummond et al., 2009). Additionally, ccd8 mutant plant lines have been reverted to the wild type by the insertion of a small piece (approximately 2 mm) of wild-type hypocotyl into the hypocotyls of mutant petunia or by insertion of a small piece of epicotyl into the epicotyl of mutant pea (Napoli, 1996;Foo et al., 2001). In Arabidopsis, the ccd7 mutant has been similarly reverted using hypocotyl interstock grafts (Booker et al., 2004). Together, these results suggest the presence of a mobile branch inhibitor produced in wild-type tissue. However, an observation by Napoli (1996) suggested that decreased apical dominance1 (dad1) mutant roots may also have a branch-inducing effect in certain circumstances. A similar result was observed for pea in Parvus by Foo et al. (2001). The discussion presented by Napoli (1996) did not exclude either a branch-inducing or a branch-suppressing signal, although current models generally only consider the presence of a branch inhibitor, and recent efforts have focused on the identification of inhibitors of branching.
Strigolactones have recently been identified as signaling molecules that inhibit axillary branch outgrowth in plants ( Gomez-Roldan et al., 2008;Umehara et al., 2008). Strigolactones were previously identified as signal molecules secreted from roots. When arbuscular mycorrhizae detect strigolactones, they undergo a preinfection hyperbranching response that is thought to aid fungal colonization of the roots, frequently leading to improved nutrient uptake by the plant (Akiyama et al., 2005). The seeds of the parasitic plants Orobanche species and Striga species are also induced to germinate upon detection of strigolactones in the soil, resulting in significant yield losses for some crops (Cook et al., 1966;Siame et al., 1993;Yokota et al., 1998). The production of strigolactones in rice and pea has been shown to require the action of both CCD7 and CCD8 (Gomez-Roldan et al., 2008;Umehara et al., 2008). The discovery that strigolactones can alter branching confirmed a new layer of regulatory complexity in the control of branching that has long been hidden beneath the global plant growth regulators of auxin and cytokinin.
In this study, we have focused on the role of the CCD7 gene in the control of branching in petunia. We have isolated a petunia CCD7 ortholog (PhCCD7) and show that the increased branching phenotype of the dad3 mutant is caused by a lesion in this gene. The phenotype of the dad3 mutant is less severe than that of the petunia ccd8 mutant (dad1), and the double ccd7ccd8 mutant is shown to be additive. These observations are contrasted with what has been observed for other plant species. We show that the regulation of PhCCD7 is similar to that of the PhCCD8 gene, with expression predominantly in root and stem tissue (although at a reduced level) and up-regulation of expression in plants with increased numbers of branches. We also provide evidence for the presence of a branchpromoting signal in mutant roots of petunia. These results suggest that there is an added layer of complexity to the control of branching that is not fully described by current models and indicate that the CCD7 gene may have a role in the diversity of plant architecture.

dad3 Mutants Have Reduced Root and Shoot Growth
An increased branching mutant, dad3, of petunia was identified by Napoli (1996) and characterized by Snowden and Napoli (2003) and Simons et al. (2007). Snowden et al. (2005) reported that the dad1 mutation reduced root mass and also reduced flower size and weight. To determine if a similar effect was observable in the dad3 mutant, we grew wild-type (inbred line V26) and dad3 mutant petunia plants in a hydroponic culture system and quantified the fresh weight of shoots, roots, and flowers. The dad3 mutant plants produced both less root and shoot mass than did the wild type (Table I), as observed previously for dad1 (Snowden et al., 2005). The reduction of mass in the dad3 mutant was not in proportion, with a greater percentage reduction observed in the mass of the roots, leading to an increased shoot-root mass ratio (Table I). Flower weight and size in the same dad3 plants was also reduced compared with those in the wild type (Table I; data not shown).

Isolation of PhCCD7
The phenotypes of the dad1 and dad3 mutants of petunia have many aspects in common, including but  Ongaro and Leyser, 2008). As we had previously identified PhCCD8 as DAD1 (Snowden et al., 2005), it seemed a strong possibility that the mutation causing the dad3 phenotype occurred in a CCD7 ortholog. We identified CCD7 candidate genes in public sequence databases, aligned these using their protein sequences, and designed degenerate primers to conserved regions. These degenerate primers were used in a reverse transcription (RT)-PCR approach to isolate part of the petunia ortholog (PhCCD7). Inverse PCR on genomic DNA was then used to isolate flanking regions, giving a total of 6 kb of DNA sequence at the PhCCD7 locus. A full-length cDNA clone of the gene was amplified from reverse-transcribed mRNA, cloned, and sequenced, allowing confirmation of the intron-exon boundaries (Fig. 1A). The petunia PhCCD7 gene encodes a predicted protein of 621 amino acids. The protein is more closely related to AtCCD7 than any other Arabidopsis CCD protein (Supplemental Fig. S1). The CCD7 genes in petunia, pea, and rice are composed of seven exons, whereas AtCCD7/MAX3 has six (Booker et al., 2004;Johnson et al., 2006;Zou et al., 2006). The positions of four introns are conserved between the CCD7 genes of Arabidopsis, rice, pea, and petunia; the additional intron in petunia and pea splits exon 5 of the Arabidopsis model, and the extra rice intron splits exon 1 of the Arabidopsis model (Fig. 1B). Sequencing of the PhCCD7 gene locus in the dad3 mutant was carried out to determine if a mutation in this gene may be responsible for the mutant phenotype. In the dad3 mutant, we identified a mutation that changes Trp-156 to a stop codon (i.e. TGG to TGA). This mutation would result in a greatly truncated protein (155 instead of 621 amino acids, as occurs in the wild type; Fig. 1A). It is likely that this mutation causes a complete loss of function of PhCCD7 due to the loss of highly conserved sequence motifs, which in a related protein, apocarotenoid-15,15#-oxygenase, form the substrate-binding site (Kloer et al., 2005).
We investigated whether the mutation in the PhCCD7 gene cosegregates with the dad3 phenotype. We generated a segregating population of dad3 plants using a cross between dad3 and a wild-type inbred line (M1) and then self-crossed this F1 plant to give an F2 population. M1 was used instead of our original inbred line (V26) from which dad3 was derived to allow for potential mapping of the DAD3 gene to visual markers if necessary. A single nucleotide polymorphism assay was developed and used to detect the mutant and wild-type alleles of PhCCD7 in our F2 population of plants segregating for dad3. All 29 plants with a dad3 phenotype were homozygous for the mutation in the PhCCD7 gene. This result indicates cosegregation of the PhCCD7 gene with the dad3 mutant phenotype.
PhCCD7 Can Complement the dad3 Mutation, and RNA Interference of PhCCD7 Phenocopies dad3 The relationship between the dad3 mutant and the PhCCD7 gene was confirmed using wild-type (V26) and dad3 petunia carrying PhCCD7-derived transgenes. An overexpression construct was made using a full-length copy of the PhCCD7 cDNA under the control of the cauliflower mosaic virus (CaMV) 35S promoter. A knockout construct was developed using a 368-bp section of the PhCCD7 coding region in an RNA interference (RNAi) hairpin conformation (Fig.  1A). It was noted that regeneration of rooted plants from transformations using the overexpression or RNAi hairpin constructs into both wild-type and dad3 petunia was more difficult than concurrent transformations using the control constructs. The control construct (pHEX4, a 35S:GUS construct) did not alter the phenotype of the seven dad3 and 11 wild-type lines that were generated. The 35S:PhCCD7 construct reverted 16 out of 22 independent kanamycin-resistant lines in the dad3 background to a wild-type phenotype. The same construct in wild-type plants did not substantially alter the phenotype of the nine lines that were generated. The PhCCD7-RNAi construct converted 12 out of 17 wild-type lines to a dad3-like phenotype. This RNAi-induced phenotype was indistinguishable from the phenotype of dad3 mutant controls. The same RNAi construct did not change the phenotype of the 13 dad3 lines that were generated. Representative transgenic and control plants are shown in Figure 1C. These results show that ectopic expression of the PhCCD7 gene can complement the dad3 mutant phenotype and that a knockout of PhCCD7 can phenocopy the dad3 mutant and confirm the identity of the DAD3 gene as PhCCD7.

The Expression of the PhCCD7 Gene Is Primarily in Stem and Root and Is under Feedback Control
The expression of the PhCCD7 gene was investigated in 11-week-old wild-type petunia using quantitative real-time RT-PCR (qRT-PCR) and compared with the expression pattern of the PhCCD8 gene in the same plants ( Fig. 2A). PhCCD8 is expressed most highly in roots, with moderate expression in low stems. PhCCD7 shows weaker expression than PhCCD8 in roots, but its expression is more evenly distributed between the roots and stem and occurs at higher regions of the stem where PhCCD8 expression was not detected. The expression levels of PhCCD7 were also determined in mutant and grafted plants (Fig. 2, B and C); the sources of the RNAs used in these experiments are described by Snowden et al. (2005) and Simons et al. (2007), respectively. PhCCD7 expression was up-regulated in the stems of dad1 and dad2 mutants, although not in dad3 stems. The dad2 mutant has a similar branching phenotype to that of dad1 (Napoli and Ruehle, 1996), but it is not reverted to the wild type by grafting to wild-type rootstocks (Simons et al., 2007). The defective PhCCD7 transcript in the dad3 mutant may explain why an increase in expression is not observed in these plants. In grafted plants, PhCCD7 expression was up-regulated in only the dad1/dad1 (scion/rootstock) and dad2/dad1 plants. In contrast to the PhCCD8 expression, where upregulation was seen only in stem tissue (Simons et al., 2007), in the same grafted plants PhCCD7 expression was increased in both stem and root. The increased PhCCD7 expression was correlated with grafted plants that had an increased branching phenotype (Fig. 2C). When dad1 scions were reverted to a wild-type branching phenotype, no up-regulation of expression of PhCCD7 was observed. These results are similar to those previously observed for PhCCD8 (Snowden et al., 2005;Simons et al., 2007) in that the expression of both genes responds to the branching phenotype of the plants and is not correlated with the genotype of the tissue sampled. This suggests that these genes are coordinately regulated in a feedback system correlated with the branching phenotype of the plant.
The dad1-1dad3 and dad1-2dad3 Double Mutants Are Additive Earlier (Simons et al., 2007), we reported that the dad1dad3 double mutant is additive, producing a plant with further decreased height but without additional branches. The dad1 mutant (specifically, the dad1-1 allele) produces a very high density of branches, such that it is possible that some environmental constraint unrelated to the mutation is limiting further branch outgrowth in this and the double mutant. We have repeated the experiment described by Simons et al. (2007) and extended the analysis to include a double mutant carrying the weak dad1-2 mutant allele and the dad3 mutation with its parental lines (Fig. 3). In this experiment, the total number of branches produced by dad1-1, dad3, and dad1-1dad3 was significantly (P = 0.05) different from each other, and the double mutant was more branched than either single mutant. Double mutants between dad3 and the weak dad1-2 allele also produced significantly (P = 0.05) more branches than either single mutant. As observed in previous experiments, plant height was also reduced in the double mutants (Fig. 3B). The number of branches produced by the double mutants in this experiment confirms the additive nature of the interaction between the CCD7 and CCD8 genes in petunia. Interestingly, we did not observe a strong correlation between height and branch number in this experiment, suggesting that although both traits are usually altered in the dad branching mutants, they may not be coordinately regulated.
Petunia Produces a Branch-Promoting Compound in dad1 and dad3 Roots We investigated the possibility that in dad mutant petunia, a growth-promoting compound is derived from some perturbation of secondary metabolism caused by the loss of the CCD genes in addition to the loss of the growth-suppressing hormone. Previous work has shown that simple wild-type scion over mutant root grafts do not usually produce scions with a greater than wild-type number of branches (Napoli, 1996;Simons et al., 2007). This is the expected result if CCD7 and CCD8 act in the stem to degrade any positive signal. Hence, to test for a branch-promoting signal, we used grafted plants with two root systems, one wild type in genotype and the other mutant. A series of micrografted plants were cultured to promote the formation of adventitious roots from the mutant tissue of the scions (Fig. 4A). Adventitious roots formed on grafted scions within 2 d of grafting but did not inhibit the formation of the graft union between the rootstock and scion. In the grafted plants with two root systems, the mutant roots grew more slowly than the wild-type roots, and the mutant roots tended to be smaller at the time the plants were transferred to soil. In an experiment where we measured the root mass (dry weight) of the plants, the average root mass for wild-type homografted plants was 0.42 6 0.07 g, for dad1 homografted plants it was 0.15 6 0.004 g, and for dad1 over the wild type it was 0.40 6 0.09 g. On double-rooted plants (dad1 scion), the average wild-type root mass was 0.48 6 0.l3 g, and the average dad1 root mass was 0.09 6 0.03 g.
Plants with a dad3 scion and with both dad3 and wild-type roots had a number of branches greater than dad3 scions with only wild-type roots ( Fig. 5A; Supplemental Figs. S2 and S3). Additionally, grafted plants using dad1 or dad1dad3 produced a comparable result to dad3-grafted plants; in each case, mutant over wild-type grafted plants produced more branches if mutant roots were present on the plant than if only wild-type roots were present (Fig. 5, B and C). The amount of branching observed in these double-root plants was intermediate between plants with only wild-type roots or only mutant roots. In this experiment, the wild-type over mutant grafts produced the same number of branches as wild-type over wild-type grafts (P = 0.01). This grafting experiment has been repeated three times using dad1 plants; each experiment showed a similar trend (data not shown).
In order to test whether the graft union between the dad1 scion and the wild-type roots in the double-root plants was functional, water uptake into grafted petunia plants was traced using rhodamine B. Water transport to the scion from either root system was global, with the rhodamine B staining all branches, although with varying intensity (Fig. 4, B-D), indicating that both root systems had a functional vascular connection to the shoot. The number of secondary branches on a primary branch did not correlate with the degree of staining of that branch (data not shown). Furthermore, the asymmetry of the mutant root sys- tem (which, in most cases, grew from only one side of the stem) was not reflected in the staining of branches or the branching pattern in the scion, which both had radial symmetry (Fig. 4,C and D;Supplemental Fig. S3). The data presented above allow us to conclude that the rootstock is contributing water and nutrients to the scion and that there is an active and functional connection between the grafted tissues.

DISCUSSION
In this paper, we have shown that the dad3 mutant of petunia is caused by a mutation in the PhCCD7 gene. The dad3 mutant was complemented by overexpression of the PhCCD7 gene, and reduction of PhCCD7 expression in wild-type V26 plants resulted in a phenotype indistinguishable from the dad3 mutant plants. The mutant allele is likely to be null, introducing a stop codon after amino acid 155, whereas the full-length protein is 621 amino acids. This conclusion is supported by the observation that the phenotype of the RNAi plants was not more severe than that of dad3 mutant plants. PhCCD7 transcript was detected in stems and roots of plants, and increased expression was observed to be correlated with increased branching in mutant and grafted plants. We have also shown that the petunia ccd7ccd8 double mutant displays an additive branching phenotype. Finally, we have shown that mutant scions grafted to a wild-type rootstock having both mutant and wild-type roots produce more branches than those with only wild-type roots. The CCD7 and CCD8 genes have now been cloned from four plant systems that are being studied in relation to the control of axillary branching at the molecular level (Arabidopsis, pea, rice, and petunia; for review, see Drummond et al., 2009).
The expression of both CCD7 and CCD8 has been shown to be necessary for the production of a rootderived branching inhibitor in pea and rice (Gomez-Roldan et al., 2008;Umehara et al., 2008). The PhCCD7 gene is also likely to be involved in a similar process in petunia, and qRT-PCR analysis shows that it is expressed in roots and stems and that this expression  pattern overlaps, but is not identical to, that shown for PhCCD8 by Simons et al. (2007). We show that PhCCD7 has a lower level of expression than PhCCD8 and that PhCCD7 was detectable in more acropetal regions of stem tissue than PhCCD8. Reciprocal grafting results using the dad1 and dad3 mutants have shown that both genes must be functional in the same tissue to produce a plant with wild-type levels of branching (Simons et al., 2007). It is likely that only tissues where both PhCCD7 and PhCCD8 are expressed are important in the control of branching.
As seen for PhCCD8 (Simons et al., 2007), we show that PhCCD7 expression is altered in response to the branching phenotype of the scion, suggesting that regulation of expression of the CCD7 and CCD8 genes is under feedback control. Grafting studies have shown that wild-type root tissue is sufficient to revert ccd7 and ccd8 scions to the wild type. However, the observations that feedback up-regulation of CCD7 (this work) and CCD8 (Simons et al., 2007) expression occurs in stems and that small interstock grafts of stem tissue are sufficient to revert the branching phenotype of ccd8 mutant plants (Napoli, 1996) both suggest that the stem expression (in addition to root expression) of these genes is important in the regulation of axillary branching.
To date, most of the models proposed to explain the action of the CCD7 and CCD8 proteins place them as consecutive steps in a biochemical pathway leading to the production of a single branch outgrowthsuppressing hormone (Booker et al., 2004;Foo et al., 2007;Hayward et al., 2009). The CCD7 and CCD8 proteins are required to produce strigolactones (Gomez-Roldan et al., 2008;Umehara et al., 2008) and have been shown to cleave a range of carotenoids in bacteria (Booker et al., 2004;Schwartz et al., 2004;Auldridge et al., 2006). However, it is possible that CCD7 and CCD8 are involved in the production of multiple biologically active apocarotenoids, for example D'Orenone, an apocarotenoid that alters PIN2 activity (Schlicht et al., 2008).
The branching phenotypes of the ccd7 and ccd8 mutants of Arabidopsis and rice are consistent with the simple model for the action of CCD7 and CCD8 (Booker et al., 2005;Auldridge et al., 2006;Arite et al., 2007). The results for petunia presented in this paper and in other published work (Snowden and Napoli, 2003;Simons et al., 2007) show that the dad3 mutant is less severe than the dad1 mutant. In pea, the only direct comparison of ccd7 and ccd8 mutants (rms5 and rms1, respectively) in the same genetic background used the rms5-3 allele (Morris et al., 2001). The rms5-3 mutation could plausibly result in a protein with partial activity (Johnson et al., 2006). Nevertheless, Morris et al. (2001) reported that the rms5-3 mutant had a less severe phenotype than rms1, and this observation is consistent with the results presented here for petunia and not consistent with a simple linear biochemical pathway.
One possible explanation for the reduced severity of the PhCCD7 mutant phenotype is that the PhCCD7 gene has some genetic redundancy. We have investigated the copy number of CCD7 in petunia and found no evidence for the presence of a second locus, nor Figure 5. Relationship between the graft type and the total number of branches produced. Petunia plants were 12 weeks old and had flowered. Values are means 6 SE; within each graph, bars with the same lowercase letter (a-c) are not significantly different at P = 0.01. Where two genotypes for root are listed, the first is the rootstock and the second is the adventitious root. A, The five possible configurations of dad3 to wild-type (wt) grafts (n $ 7). ND, Not determined. B, The five possible configurations of dad1 to wild-type grafts (n $ 7). C, The five possible configurations of dad1dad3 to wild-type grafts (n $ 6). The plants represented by the data in A and C were grown as part of a single experiment; those in B were grown during a separate experiment.
have we identified any reports or database entries suggesting the presence of a second locus in any other species. Another possibility is functional redundancy in CCD7 activity, perhaps by one of the other CCD proteins. We suggest that the reduced severity of the dad3 (PhCCD7) mutant may result from not simply loss of an inhibitor (strigolactone) but also changes in other compounds (e.g. accumulation of a promoter of branching), and that loss of PhCCD8 could result in a greater increase of promoter signal than loss of PhCCD7. Such a model would propose that either PhCCD7 or PhCCD8 can degrade the precursor(s) that lead to production of an additional signal but that PhCCD7 and PhCCD8 are both required to produce the compound that leads to production of the negative signal (a strigolactone or its derivative).
In Arabidopsis, the ccd7ccd8 double mutant is phenotypically similar to both of its parental lines (Booker et al., 2005;Auldridge et al., 2006). However, in pea, double mutants between rms1 (CCD8) and rms5-3 (CCD7) have a more severe branching phenotype than the rms1 mutant alone (Morris et al., 2001). We have shown that petunia ccd7ccd8 double mutants also produce an additive phenotype, a result not consistent with a simple linear biochemical pathway. The classical interpretation of this result is that the genes lie in separate pathways controlling the same phenotypic characteristic. However, both PhCCD7 and PhCCD8 have been shown by double mutant analysis to be in the same pathway as DAD2, an as yet unidentified gene controlling branching in petunia (Simons et al., 2007). Furthermore, in rice and pea, biochemical analysis of either ccd7 or ccd8 revealed a loss of strigolactones, also supporting the hypothesis that these enzymes act in the same pathway (Gomez-Roldan et al., 2008;Umehara et al., 2008). Since loss of either CCD7 or CCD8 results in loss of the branching inhibitor, then the additive phenotype observed is likely to be the result of a change in another signal molecule.
In this paper, we have presented quantitative data showing that if dad1, dad3, or dad1dad3 mutant scions grafted to wild-type rootstocks are allowed to develop mutant roots in addition to their wild-type rootstocks, then branch numbers are significantly increased over plants with only wild-type roots. This result suggests the presence of a branch-promoting signal produced in the mutant root tissue. Recently, a physiological study of axillary meristem outgrowth in clover concluded that meristem outgrowth was activated in a dose-dependent manner by a root-derived signal (Thomas and Hay, 2007).
One possible explanation for our observations is that strigolactone, produced in the wild-type roots, is blocked from entering the stem or diluted by the presence of the mutant roots. We have shown that the vascular connection from the wild-type roots to the stem is functional. In these grafting experiments, the mutant roots never accounted for more than half of the total root mass, and there was no decrease in the average wild-type root mass. Additionally, physiolog-ical data suggest that only a small amount of wild-type material (approximately 2 mm) is required to revert grafted mutant scions (Napoli, 1996;Foo et al., 2001;Booker et al., 2004), suggesting that the branching inhibitor is potent. This has been confirmed by the biochemical demonstration that strigolactones can act to suppress branching at very low concentrations (Gomez-Roldan et al., 2008;Umehara et al., 2008). In the grafting experiments presented here, there is at least 2 mm of wild-type hypocotyl tissue present in addition to wild-type roots. If there were only a branch-suppressing hormone to consider, the amount of wild-type hypocotyl and root tissue present in the grafted plants should produce sufficient strigolactone to inhibit branching to the extent observed for wildtype controls.
The expression of CCD7 and CCD8 has been shown to be up-regulated in the stem when the scion of the plant has increased branching (Johnson et al., 2006;Simons et al., 2007;Hayward et al., 2009;this work) and in response to auxin in pea, Arabidopsis, and rice (Johnson et al., 2006;Zou et al., 2006;Arite et al., 2007;Hayward et al., 2009). Hayward et al. (2009) also showed that auxin regulates the expression of CCD7 and CCD8 via IAA12. These results suggest that auxin acts as a feedback signal to increase production of strigolactone (by increasing expression of CCD7 and CCD8) when excess branching occurs. However, for the dual rooted grafted plants, which have increased branching, it would be expected that such a feedback signal would act to increase strigolactone production from the wild-type roots of plants with both wild-type and mutant roots. Thus, while the auxin feedback appears to play an important role in the regulation of strigolactone biosynthesis, it does not explain the increased branching observed in plants with both wild-type and mutant roots.
One possible candidate for a positive signal from mutant roots is cytokinin. However, xylem cytokinin has been shown to be reduced in strigolactone pathway mutants for Arabidopsis, pea, and petunia (Morris et al., 2001;Foo et al., 2007;Drummond et al., 2009). The observations that cytokinin is synthesized in the aerial parts of the plant (Nordströ m et al., 2004) and that increasing cytokinin production in the roots did not increase branching (Faiss et al., 1997) suggest that root-derived cytokinin is not the mobile positive branching signal seen in our experiments.
How the diversity of plant architecture is generated at the molecular level is still largely unknown. It appears that the CCD7/CCD8 pathway is conserved in the species examined thus far. However, several different strigolactones have been isolated from different plants , and this may reflect differences in substrates or activities of CCD7 and CCD8. The strigolactone content of plants is complex, and whether the activities of the various compounds are equal with respect to branching has not been characterized. In addition, differences in CCD7 expression between species (Sorefan et al., 2003;Bainbridge et al., 2005;Arite et al., 2007) may also play a role in the regulation of architecture.

CONCLUSION
The data presented here suggest that the metabolic pathway involved in strigolactone production is not simple and that mutation of CCD7 or CCD8 or both leads to increased production of a promoter of branching as well as loss of the inhibitor. A model that includes more than one biologically active molecule from this pathway will allow a more complete understanding of the developmental and environmental controls of plant form.

Genetic Stocks, Grafting, and Plant Growth Conditions
The single dad mutants were isolated by Napoli and Ruehle (1996), and double mutants were generated as reported by Simons et al. (2007). All soilgrown plants were cultured as described by Snowden et al. (2005). The hydroponics system required the germination of surface-sterilized petunia (Petunia hybrida) seeds on 0.53 Murashige and Skoog salts and vitamins solidified with 8 g L 21 agar (Murashige and Skoog, 1962). Four-week-old seedlings were transferred to a hydroponics system (Hydroponics Wholesalers). Each plant was grown in 250 mL of 4 -to 8-mm Hydroton clay pebbles. The nutrient solution was supplied from a 50-L reservoir constantly recirculated (ViaAqua 480) and aerated (Two AquaOne 1-3 1-3 10-cm airstones). The growth medium was a nutrient-complete rich medium. The solution volume was maintained at 50 L using tap water and replaced every 2 weeks. The glasshouse conditions for the hydroponics system were as for soil-grown plants.
Transgenic petunia plants were produced using Agrobacterium tumefaciensmediated transformation of tissue culture-grown plants according to the method of Jorgensen et al. (1996).
Grafting of petunia plants was performed as described previously (Napoli, 1996). Approximately 2 weeks after grafting, the plants were transferred to soil (Black Magic Seed Raising mix [Yates] combined 3:1 with perlite). A plastic divider was used to keep the root systems separate for plants with both mutant and wild-type roots. The plants were grown until approximately 3 months of age. Grafted plants were irrigated with water containing fertilizer as described previously (Snowden et al., 2005). Numbers of leaf nodes and branches were defined and recorded as described by Snowden and Napoli (2003). In all cases, the genotype of each root mass was verified by PCR.

Gene Isolation and Vector Construction
The degenerate primers, designed from an alignment of CCD7 proteins, used to isolate the first fragment of PhCCD7 were 5#-ATGAARAAYGTNGC-NAAYAC-3# and 5#-AANGCCCARTCRTGDATCAT-3#. A PCR regime of 94°C for 2 min, followed by 30 cycles of 94°C for 15 s, 60°C for 30 s, and 72°C for 1 min, followed by 72°C for 5 min was used. The rest of the PhCCD7 gene was then isolated using inverse PCR as described by Snowden and Napoli (1998). A full-length cDNA copy of the gene was isolated using PCR with primers 5#-TCACTAGTACAACTCCTCTAG-3# and 5#-TCCTCCAGCATT-GACCAAGA-3#, cloned into pGEM T-Easy, and used in all future work. To generate the PhCCD7-silencing vector, a 368-bp region from exon 5 was PCR amplified from the pGEM T-Easy clone using primers 5#-GGGGAC-AAGTTTGTACAAAAAAGCAGGCTCAATTTTCTTGAAATAATTGCGGC-3# and 5#-GGGGACCACTTTGTACAAGAAAGCTGGGTCACAAGAAGATAT-CCTTCATCTTC-3#. The Gateway BP reaction with PCR product and pDONR 201 was carried out to generate entry clone pENTRY_PhCCD7. To derive a silencing vector by a single step, Gateway LR reactions were carried out with the pENTRY_PhCCD7 vector and destination vector pTKO2 (Snowden et al., 2005). To generate a vector for the overexpression of the PhCCD7 gene, the cDNA was excised as an EcoRI fragment from the pGEM T-Easy clone and cloned into the EcoRI site of pSAK778, placing the cDNA between the CaMV 35S promoter and the ocs 3# transcriptional terminator.
To generate a vector for the overexpression of GUS-with-Intron (the coding sequence of the uidA gene for GUS containing the tobacco [Nicotiana tabacum] yellow dwarf virus intron), a PCR amplification was carried out on pART27/ 10 (Yao et al., 1996) with primers SP1 (5#-attB1-ATGGTACGTCCTGTAGA-AACC-3#) and SP2 (5#-attB2-TCATTGTTTGCCTCCCTGCTG-3#), where attB1 and attB2 are the DNA sequences recommended by Invitrogen. The resulting 1,897-bp PCR amplification product was recombined with pDONR201 to generate pENTRY_GUS-with-Intron. A Gateway attL 3 attR reaction with pENTRY_GUS-with-Intron and the pHEX2 (Hellens et al., 2005) destination vector was carried out to generate pHEX4, which is 35S:GUS. All Gateway reactions were carried out as recommended by the manufacturer (Invitrogen).

Genotyping of the dad3 Segregating Population
The dad3 segregating population was screened by four-primer multiplex PCR for the presence of the wild-type and/or dad3 alleles of the PhCCD7 gene. The presence of the PhCCD7 gene was detected using the A (5#-CTT-CAAAACTTCTACCACCAGC-3#) and B (5#-GTGACAGGCTTACAGGCA-AC-3#) primers. The dad3 mutant allele was detected using the A and C (5#-CCCGGTGAGTGAACCGT-3#) primers, and the wild-type allele was detected using the B and D (5#-CAGTGACCGGACAGTGG-3#) primers. The C and D primers bind across the position of the nucleotide polymorphism that distinguishes the wild type from dad3, the C primer binds to only the dad3 allele, and the D primer binds to only the wild-type allele. The PCR conditions required to detect each allele reliably are as follows: 13 PlatinumTaq PCR buffer (as supplied; Invitrogen), 500 nM of each primer, 1.5 mM MgCl 2 , 1.5 units of PlatinumTaq, 50 nM deoxyribonucleotide triphosphate in 25-mL reaction volumes, and a PCR cycle of 94°C for 2 min, 30 cycles of 94°C for 15 s, 67.5°C for 30 s, and 72°C for 1 min, with a final extension at 72°C for 5 min.

qRT-PCR
qRT-PCR was carried out largely as described by Snowden et al. (2005) and Simons et al. (2007). qRT-PCR with Sybr Green detection was carried out for PhCCD7, PhCCD8, and the internal control genes Actin, EF-1a, and Histone4. The primers used to amplify all of these genes are listed in an earlier paper (Snowden et al., 2005), except PhCCD7, which was amplified using the following primers: forward, 5#-CTGAAAGGTGGGAAGATGGT-3#; reverse, 5#-CCTTCCCACAAGCATAACAA-3#. Relative expression was calculated using the comparative cycle threshold method (Pfaffl, 2001) with normalization of data to the geometric average of the internal control genes (Vandesompele et al., 2002). Expression levels were then normalized relative to expression in wild-type root (Fig. 2, A and B) or wild-type stem (Fig. 2C), where expression was set to 1. Samples for qRT-PCR were collected from at least six individual plants, and tissues were pooled prior to RNA isolation. Data shown are from one experiment, and a second independent experiment gave essentially the same results. The d1/d2 and d2/d1 grafted plants were not included in the duplicate grafting experiment.
The DNA sequence for the PhCCD7 gene has been deposited in the GenBank data library under accession number FJ790878.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Phylogenetic relationships within the CCD gene family.
Supplemental Figure S2. Primary and secondary branching of grafted plants.
Supplemental Figure S3. Phenotype of grafted plants.