Carlactone-independent seedling morphogenesis in Arabidopsis

Authors

  • Adrian Scaffidi,

    1. School of Chemistry and Biochemistry, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Mark T. Waters,

    1. Plant Energy Biology, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Emilio L. Ghisalberti,

    1. School of Chemistry and Biochemistry, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
  • Kingsley W. Dixon,

    1. Kings Park and Botanic Garden, West Perth, WA, Australia
    2. School of Plant Biology, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
  • Gavin R. Flematti,

    1. School of Chemistry and Biochemistry, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
  • Steven M. Smith

    Corresponding author
    1. Plant Energy Biology, The University of Western Australia, Crawley, WA, Australia
    • School of Chemistry and Biochemistry, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author

For correspondence (e-mail steven.smith@uwa.edu.au).

Summary

Strigolactone hormones are derived from carotenoids via carlactone, and act through the α/β–hydrolase D14 and the F–box protein D3/MAX2 to repress plant shoot branching. While MAX2 is also necessary for normal seedling development, D14 and the known strigolactone biosynthesis genes are not, raising the question of whether endogenous, canonical strigolactones derived from carlactone have a role in seedling morphogenesis. Here, we report the chemical synthesis of the strigolactone precursor carlactone, and show that it represses Arabidopsis shoot branching and influences leaf morphogenesis via a mechanism that is dependent on the cytochrome P450 MAX1. In contrast, both physiologically active Z–carlactone and the non-physiological E isomer exhibit similar weak activity in seedlings, and predominantly signal through D14 rather than its paralogue KAI2, in a MAX2-dependent but MAX1-independent manner. KAI2 is essential for seedling morphogenesis, and hence this early-stage development employs carlactone-independent morphogens for which karrikins from wildfire smoke are specific surrogates. While the commonly employed synthetic strigolactone GR24 acts non-specifically through both D14 and KAI2, carlactone is a specific effector of strigolactone signalling that acts through MAX1 and D14.

Introduction

Strigolactone (SL) phytohormones play key roles in the germination of root-parasitic weeds (Cook et al., 1966, 1972) and the formation of plant–mycorrhizal associations (Akiyama et al., 2005; Besserer et al., 2006). They were subsequently identified as the ‘shoot multiplication signal’ that is transported from the root to repress bud outgrowth (Gomez-Roldan et al., 2008; Umehara et al., 2008), confirming SL as an endogenous regulator of plant growth and development. Additional functions for SLs include the control of secondary growth in the cambium (Agusti et al., 2011), root architecture (Ruyter-Spira et al., 2011; Mayzlish-Gati et al., 2012), adventitious root formation (Rasmussen et al., 2012) and leaf development (Beveridge et al., 1996; Stirnberg et al., 2002; Snowden et al., 2005). Many of these effects are intimately related to auxin-dependent processes, and SL acts at least in part by regulating auxin transport via the efflux carrier PIN1 (Shinohara et al., 2013). SLs therefore have a profound influence on plant productivity, and potentially provide new opportunities for improving crop production.

Genes involved in SL biosynthesis and perception have been identified following the isolation of mutants in rice (Oryza sativa), garden pea (Pisum sativum), Arabidopsis thaliana and Petunia hybrida that exhibit dwarfism and increased numbers of lateral shoots (Brewer et al., 2013; Ruyter-Spira et al., 2013). Key steps in SL biosynthesis from carotenoids were demonstrated by heterologous expression of three plant enzymes (D27, CCD7 and CCD8) in Escherichia coli (Alder et al., 2012). The 40–carbon all-trans-β–carotene underwent isomerization and successive dioxygenase reactions to produce a 19–carbon product named carlactone (CL; Figure S1). CL has the butenolide ring and enol-ether moiety that are typical of SLs, linked in the Z–geometric configuration to a carbon skeleton that may potentially be modified to produce A, B and C rings of functional SLs (Figure 1a). Z–CL generated from E. coli inhibits tillering in SL-deficient rice and stimulates seed germination of the parasite Striga hermonthica (Alder et al., 2012). It is proposed that Z–CL is converted to active SLs in plants by further enzymes, including a cytochrome P450 encoded by the MAX1 gene in Arabidopsis (Booker et al., 2005; Waters et al., 2012a). The activity of such SLs in shoots depends on an α/β–hydrolase (D14/DAD2) together with an F–box protein (MAX2/D3/RMS4) (Hamiaux et al., 2012). D14 hydrolyses GR24, a synthetic SL, through nucleophilic attack by the active-site serine residue on the carbonyl of the butenolide moiety (Hamiaux et al., 2012; Scaffidi et al., 2012; Zhao et al., 2013).

Figure 1.

Structures of strigolactones and chemical synthesis of carlactone.

(a) 1, Strigol, a naturally occurring SL; 2, GR24, a synthetic SL analogue; 3, KAR1, a karrikin derived from burnt plant material; 4, 4–(4–methyl-5–oxo-2,5–dihydrofuran-2–yloxy) benzonitrile, a ‘debranone’ SL analogue. The ABCD rings of the canonical SL structure are indicated.

(b) Synthetic scheme for carlactone, yielding two geometric isomers, Z–CL and E–CL.

Arabidopsis seedling development is also influenced by butenolides, such as the synthetic SL analogues GR24 and debranones, and by karrikins present in bushfire smoke (Flematti et al., 2004; Nelson et al., 2012; Waters et al., 2012d) (Figure 1a). Both karrikins and GR24 inhibit hypocotyl elongation, but, while karrikins promote cotyledon expansion, GR24 does not (Nelson et al., 2010; Waters and Smith, 2013). Likewise, both GR24 and debranones repress shoot branching, but karrikins cannot (Fukui et al., 2011; Nelson et al., 2011; Boyer et al., 2012). Therefore, structurally similar butenolide compounds may have strikingly different bioactivities, which are mediated by partly different response mechanisms. In seedlings, these butenolides act through MAX2, plus one or both of AtD14 and its paralogue KAI2. Notably, while GR24 acts via either AtD14 or KAI2, karrikin is KAI2-specific (Waters et al., 2012c). The function of DLK2, a third paralogue of AtD14 and KAI2 in Arabidopsis, has not yet been established, but conceivably it may exhibit functional redundancy with AtD14 or KAI2 in mediating responses to specific butenolides. Alternatively, as DLK2 transcripts are highly responsive to GR24 and karrikins, and thus serve as a convenient marker for butenolide signalling (Waters et al., 2012c; Waters and Smith, 2013), DLK2 may play a negative feedback role by acting as an enzyme to metabolize butenolide signalling compounds.

Arabidopsis max2 and kai2 mutants, but not Atd14 mutants, exhibit increased seed dormancy and abnormal seedling photomorphogenesis (Nelson et al., 2011; Waters et al., 2012c). These observations suggest that seedling development may be under the control of endogenous butenolides with structures similar to SLs and karrikin, acting primarily through KAI2 rather than AtD14. A role for endogenous SLs in seedling photomorphogenesis has been proposed based on the response of seedlings to GR24 and the fact that MAX2 is required for normal seedling development (Tsuchiya et al., 2010). However, Arabidopsis seedlings lacking enzymes of the known Z–CL pathway (d27, max3 and max4) are morphologically normal (Nelson et al., 2011; Shen et al., 2012; Flematti et al., 2013), so the proposed seedling butenolide is either produced independently of Z–CL, or Z–CL is produced by a different set of enzymes in seedlings. To distinguish between these alternatives, we devised a method of chemical synthesis for CL so that its activity in seedling morphogenesis may be tested.

Results and Discussion

Chemical synthesis of carlactone

Successful preparation of CL was achieved via coupling of the bromobutenolide (6, Figure 1b) and potassium enolate, generated from the aldehyde (5, Figure 1b) and potassium tert-butoxide, with the reaction producing both E and Z geometric isomers (Figure 1b) in a ratio of approximately 6:1. Both isomers were routinely prepared and separated by semi-preparative HPLC, although the overall yield of the reaction is low and all attempts to optimize conditions by varying the solvent, base and temperature failed to increase the proportion of the desired Z–isomer. Nevertheless, we prepared Z–CL in sufficient quantities for biological testing and full characterization. The structure of chemically synthesized Z–CL was determined by both 1H- and 13C-NMR spectroscopy, and was consistent with the data provided for Z–CL isolated from E. coli cells (Alder et al., 2012), resulting in formal confirmation of the Z–CL structure by chemical synthesis.

Activity of carlactone in mature plants

Strigolactone-deficient and SL-insensitive mutants of Arabidopsis exhibit an aberrant leaf shape that is typified by rounder, broader laminas and shorter petioles relative to wild-type (Stirnberg et al., 2002; Booker et al., 2004; Waters et al., 2012c; Challis et al., 2013). To examine the effects of Z–CL on shoot development, we first investigated whether provision of exogenous Z–CL restored the leaf morphology of SL-deficient max1–1 and max3–9 mutants grown under hydroponics. We chose the SL-insensitive Atd14–1 mutant as a negative control because it has a leaf and rosette morphology that closely resembles that of SL-deficient mutants, whereas max2 has more severe defects that may be SL-independent (Waters et al., 2012c). At the whole-rosette level, the effects of GR24 and Z–CL were clear: both compounds restored the rounded leaf and compact rosette phenotypes of max3 mutants towards those of wild-type (Figure S2). However, Z–CL was completely ineffective on max1 plants, and Atd14 mutants were resistant to the effects of both compounds (Figure S2). To examine these effects in more detail, we generated average leaf shape models of leaves from nodes 5–8 using Leaf Analyser software (Weight et al., 2008). Whereas max3 leaves developed a narrower lamina and longer petiole upon treatment with either GR24 or Z–CL, max1 leaves only responded similarly to GR24 (Figure 2a). Again, Atd14 leaves were insensitive to both compounds. To quantify these differences, we determined the leaf length/width ratio, which is significantly lower than wild-type in max1, max3 and Atd14 mutants (Figure 2b). Application of Z–CL restored this aspect of leaf shape in max3 mutants, but not in max1 or Atd14. This response was at least partly mediated by an increase in petiole length, which was fully restored in nodes 5 and 6 of max3 mutants treated with Z–CL and GR24, but not in max1 mutants (Figure 2c). These results demonstrate that MAX1 is necessary for conversion of Z–CL into an active compound that regulates AtD14-dependent leaf development in mature plants.

Figure 2.

Restoration of leaf morphology by Z–CL requires MAX1 and AtD14.

(a) Average leaf shape models produced by LeafAnalyser and derived from nodes 5–8 of plants treated with 1 μm GR24 or 1 μm Z–CL. Six to eight plants were analysed per genotype/treatment combination. Leaf models were scaled relative to the length of mock-treated Col–0.

(b) Ratio of leaf length to leaf width averaged across nodes 5–8 within each plant.

(c) Ratio of petiole length to total leaf length for node 5 (upper panel) and node 6 (lower panel). Values are means ± SE (= 6–8 plants). Shared lower-case letters indicate no significant difference (anova;< 0.01).

Next, we tested the ability of Z–CL to suppress shoot branching, which provided a means to establish whether Z–CL activity is MAX2-dependent. The mutants max1–1, max3–9 and SL-insensitive max2–1 all exhibit increased outgrowth of axillary buds and thus greater numbers of secondary inflorescences. At 1 μm, both GR24 and Z–CL suppressed axillary bud growth in max3 but not in max2, demonstrating that Z–CL operates in a MAX2-dependent manner (Figure 3). However, consistent with its effects on leaf shape, Z–CL was not active in max1. Together, these data imply that MAX1 is necessary for conversion of Z–CL into SLs to suppress shoot branching, as previously hypothesized (Alder et al., 2012; Waters et al., 2012a; Ruyter-Spira et al., 2013). The incomplete restoration of branching in max3 and max1 to wild-type levels by GR24 and Z–CL at 1 μm probably reflects the concentration used, the chemical instability of the active compounds, and their inefficient uptake or transport, thus limiting their concentration within target tissues.

Figure 3.

Z–carlactone represses growth of secondary inflorescences in Arabidopsis via MAX1.

(a) Counts of secondary inflorescences (derived from rosette axillary buds) in 42-day-old Arabidopsis plants grown under hydroponics supplemented with 0.1% acetone (mock), 1 μm Z–CL or 1 μm GR24. Values are means ± SE (= 6 plants). Asterisks indicate a significant difference versus mock-treated controls (**< 0.01, anova).

(b) Close-up of rosettes of max3–9 mock-treated or treated with 1 μm Z–CL. White arrowheads indicate secondary rosette branches; orange arrowheads indicate the primary inflorescence stem. The number of branches, some of which are obscured in the photographs, is indicated in a white circle for each plant. Plants were photographed 35 days after germination. This experiment was repeated twice with similar results; the results from a single experiment are shown here.

Carlactone does not stimulate Arabidopsis seed germination

To investigate the roles of CL at other stages of the plant life cycle, we examined whether CL is active in seed germination and seedling development. Surprisingly, unlike karrikins and GR24, Z–CL did not stimulate germination of primary dormant Arabidopsis seeds. After 1 week at 28°C, 1 and 10 μm GR24 had stimulated the germination of 40 and 80% of wild-type seed, respectively, compared with 8% for untreated controls, whereas equivalent concentrations of Z–CL had no significant effect on seed germination (Figure 4). This finding suggests that Z–CL does not possess the same activity as GR24 in Arabidopsis, in contrast to Striga hermonthica, which is specifically adapted to respond to SLs and whose seed germination is stimulated by both Z–CL and GR24 (Alder et al., 2012). In addition, these results imply that Arabidopsis seed may not possess the ability to convert Z–CL into an active SL, or that natural SL (as opposed to synthetic GR24) does not play a part in promoting Arabidopsis seed germination.

Figure 4.

Carlactone does not stimulate germination of primary dormant Arabidopsis seed. Germination profile of Ler seed sown on 1% agar supplemented with Z–carlactone, GR24 or 0.1% acetone (mock). Values are means ± SE of three independent samples of 75–100 seeds. Each sample was taken from a separate pool of seed, with each pool derived from three parents grown simultaneously under identical conditions (16 h at 22°C/8 h at 16°C day/night cycle and constant 60% relative humidity, with light provided by fluorescent tubes emitting 100–120 μmol photons m−2 sec−1). Primary dormant seeds were harvested promptly after silique ripening, and were stored at −80°C until use. Seed were sown on 6 cm Petri dishes and incubated at 28°C under constant light for 1 week, and germination was monitored daily. Treatments with significantly different germination levels after 168 h are indicated by different lower-case letters (anova;< 0.05).

Carlactone is weakly active in Arabidopsis seedlings

Next, the ability of Z–CL to repress seedling hypocotyl elongation was tested in the same genotypes used to confirm its activity in the repression of shoot branching. As expected, 1 μm GR24 inhibited hypocotyl elongation in wild-type, max3–9 and max1–1, but not max2–1 (Figure 5a). In contrast, 1 μm Z–CL was inactive in all genotypes, and a tenfold higher concentration was required for a significant but relatively weak response. Notably, this activity did not depend on MAX1.

Figure 5.

Carlactone exhibits weak MAX1-independent activity in seedlings.

(a) Hypocotyl growth responses in 4-day-old seedlings grown under continuous red light on half-strength MS medium supplemented with indicated concentrations of Z–CL or GR24. Values are means ± SE (= 3 independent experimental trials of 20 seedlings per genotype/treatment combination in each trial). Asterisks indicate statistically significant differences versus 0.5× MS controls within each genotype (*< 0.05, **< 0.01; anova).

(b) STH7 and DLK2 transcript levels in seedlings grown as described in (a). Transcripts were quantified by quantitative RT–PCR, normalized against CACS (At5g46630) reference transcripts, and scaled to the value for untreated Col–0. Values are means ± SE (= 3 independent pools of >50 seedlings). Asterisks indicate statistically significant differences versus 0.5× MS controls (*< 0.05, **< 0.01; anova).

To further investigate the potential weak activity of Z–CL in seedlings, we investigated the expression of two genes, STH7 (At4g39070) and DLK2 (At3g24420), which are typically induced several-fold by treatment of wild-type seedlings with GR24 and are under-expressed in max2 and kai2 mutants (Waters et al., 2012c) (Figure S3). Treatment of seedlings with 1 μm GR24 increased expression fivefold (STH7) and ninefold (DLK2), whereas 1 μm Z–CL increased expression less than twofold, and even 10 μm Z–CL was only half as active as 1 μm GR24 (Figure 5b). Very similar induction by 1 and 10 μm Z–CL was observed in max1–1 and max3–9 mutants, implying that metabolism of Z–CL was not required for this weak activity.

We therefore considered whether Z–CL may be acting directly as a weak SL, and first tested its ability to repress hypocotyl elongation in max2, Atd14 and kai2 mutants. As expected, 1 μm GR24 inhibited hypocotyl elongation in wild-type (by approximately 60%), and less so in Atd14 and kai2 (by 40 and 50%, respectively) but was inactive in max2 (Figure 6a). In contrast, Z–CL showed relatively weak activity in wild-type and kai2 even at 10 μm (approximately 20% inhibition) and was inactive in max2 and Atd14. We then examined expression of STH7 and DLK2 in response to 10 μm Z–CL in these same mutants plus an Atd14 kai2 double mutant. Both genes showed similar weak responses in wild-type and kai2, but had a greatly reduced response in Atd14 and were unresponsive to Z–CL in max2 and the Atd14 kai2 double mutant (Figure 6b). These results suggest that Z–CL functions directly as a weak SL, acting via MAX2 and predominantly AtD14 rather than KAI2. The lack of response in Atd14 kai2 mutants rules out any redundant role for DLK2 in mediating the CL response in seedlings. Small transcriptional responses to Z–CL in Atd14–1 presumably result from relatively inefficient KAI2-dependent signalling, but it is notable that Z–CL was fully inactive in terms of hypocotyl elongation in Atd14–1, suggesting that any KAI2-dependent signalling is probably physiologically insignificant. In addition, the AtD14-dependent activity of Z–CL may explain, in part, why Z–CL does not stimulate Arabidopsis seed germination, as KAI2 transcripts are at least 100-fold more abundant than AtD14 transcripts in seed (Waters et al., 2012c). This differential expression primes the seed to respond more readily to butenolides, which signal through KAI2, such as karrikins and GR24, rather than those that signal via AtD14, such as endogenous SL.

Figure 6.

Carlactone activity in Arabidopsis seedlings is D14-dependent.

(a) Hypocotyl growth responses in 4-day-old seedlings grown under continuous red light on half-strength MS medium supplemented with indicated concentrations of Z–CL, E–CL or GR24. Values are means ± SE (= 3 independent pools of 20 seedlings/pool). Asterisks indicate statistically significant differences versus 0.5× MS controls (*< 0.05, **< 0.01; anova). For Col–0 treated with E–CL, the difference from that in 0.5× MS was marginally above the significance threshold (= 0.06).

(b, c) STH7 and DLK2 transcript levels in 4-day-old seedlings grown as described above. Atd14–1 and max2–1 are in the Col–0 background; kai2–2 is in the Ler background; Atd14 kai2 is an F3 hybrid. Transcripts were quantified by quantitative RT–PCR, normalized against CACS (At5g46630) reference transcripts, and scaled to the value for untreated Col–0. Values are means ± SE (= 3 independent pools of >50 seedlings). Asterisks indicate statistically significant differences versus 0.5× MS controls (*< 0.05, **< 0.01; anova).

If Z–CL acts directly as a weak SL without metabolic conversion to active SLs, we reasoned that non-physiological E–CL, which should not be converted to an active SL, may also have similar weak activity. We tested E–CL for its ability to repress hypocotyl elongation and to induce expression of STH7 and DLK2 genes. Use of 10 μm E–CL and 10 μm Z–CL resulted in similar weak repression of hypocotyl elongation in wild-type and kai2, and was inactive on max2 and Atd14 (Figure 6a). Furthermore, the STH7 and DLK2 genes were induced similarly by 10 μm Z–CL and 10 μm E–CL, and, again, this induction was largely AtD14-dependent (Figure 6c). Accordingly, we infer that the weak responses of seedlings to both isomers of CL result from inherent activity of the compounds themselves, rather than their downstream metabolic products.

The much weaker effects of CL compared with GR24 demonstrate that CL and/or CL-derived SLs are poor regulators of seedling morphogenesis. Coupled with the normal seedling phenotypes of Arabidopsis SL biosynthetic mutants and the SL-insensitive mutant Atd14 (Nelson et al., 2011; Shen et al., 2012; Waters et al., 2012c; Flematti et al., 2013) (Figure S3), we conclude that seedling development in Arabidopsis is not controlled by canonical SLs derived from CL. Such seedling morphogenesis is independent of D27, MAX3, MAX4 and MAX1, and operates predominantly through KAI2 rather than AtD14. Nevertheless, it is possible that there is a MAX1-independent pathway for the production of SLs, or SL-like compounds, in seedlings. The presence of an alternative to the canonical SL biosynthesis pathway has been proposed based on the production of compounds by Arabidopsis max1 and max4 mutants that stimulate germination of Phelipanche ramosa seeds (Kohlen et al., 2011) and the production of SLs by Physcomitrella patens ccd8 mutants (Proust et al., 2011). In addition, SLs are produced by lower plants (e.g. liverworts and Charales algae) that appear to lack CCD8 genes (Delaux et al., 2012), and no MAX1 gene has been identified in P. patens (Ruyter-Spira et al., 2013). Over-expression of MAX2 partially complements the Arabidopsis max4 branching phenotype, which either reflects non-SL-dependent signalling by MAX2, or an alternative biosynthetic pathway for SLs (Stirnberg et al., 2007). If alternative MAX1-independent pathways for SL biosynthesis exist, they should be active in seedlings but not in mature plants, and, to account for the activity of E–CL, should exhibit relaxed substrate specificity. Thus, expression of different enzymes and response components may generate more complex signalling pathways than currently appreciated.

Crystallography has shown that the active site pocket of KAI2 is appreciably smaller than that of D14, with molecular modelling indicating that the pocket of D14 accommodates GR24 more readily than that of KAI2 (Bythell-Douglas et al., 2013; Zhao et al., 2013). Isothermal calorimetry shows that D14 binds GR24 with high affinity while KAI2 binds the karrikin KAR1 (Kagiyama et al., 2013). Furthermore, KAR1 has been shown by crystallographic data to bind in the active site pocket of KAI2 (Guo et al., 2013). Interestingly, in planta, the KAI2 system responds to the synthetic SL analogues GR24 and debranone, as well as to karrikins from wildfire smoke, and therefore we propose that an endogenous KAI2-specific butenolide derived from a CL-independent pathway has properties similar to those of SLs and karrikins. Karrikin is currently the only example of a butenolide that acts specifically through KAI2, and hence is inactive in shoot branching but acts as a specific seed germination stimulant and seedling morphogen.

Conclusion

Our results raise important questions about the pathways by which butenolides are produced to control various aspects of plant development. Differential signalling via KAI2 and D14 may involve the added complexity of alternative biosynthetic pathways for different signalling butenolides. Whereas GR24 appears to lack specificity, Z–CL (via MAX1-dependent conversion to SLs) provides specificity for D14 signalling, while karrikins provide specificity for the KAI2 pathway and so serve as a surrogate for CL-independent butenolide signalling.

Experimental Procedures

General experimental details

1H- and 13C-NMR spectra were obtained on a Bruker ARX500 or AV600 spectrometer (Bruker, Billerica, MA, USA). Dideuteromethylenechloride (CD2Cl2) was used as the solvent, with residual CH2Cl2H 5.32) or CD2Cl2C 54.0) being used as internal standards. High-resolution mass spectra (HRMS) were recorded using a Waters LCT Premier XE time-of-flight mass spectrometer (Waters, Manchester, UK) with an electrospray ionization source (ESI). Chemical synthesis experiments were performed under an inert atmosphere with solvents dried prior to use. Flash chromatography was performed using Merck silica gel 60 (Merck, Darmstadt, Germany) with the specified solvents. Thin-layer chromatography (TLC) was performed using Merck silica gel 60 F254 aluminium-backed plates. Semi-preparative high-pressure liquid chromatography (HPLC) was performed on a Hewlett-Packard 1050 system (Agilent Technologies, Santa Clara, CA, USA) using an Apollo C18 reversed-phase column (250 mm long, 10 mm internal diameter, 5 μm particle size; Grace-Davison, Grace Davidson, IL, USA). Percentage yields for chemical reactions as described are quoted only for those compounds that were purified by HPLC and the purity assessed by TLC and NMR. Synthesis of GR24 was performed as described previously (Mangnus et al., 1992).

Preparation of E- and Z–carlactone

Potassium tert-butoxide (1.0 g, 8.9 mmol) was added to aldehyde 5 (Figure 1b, 1.6 g, 8.0 mmol) (Wu et al., 2011) and N–methyl-2–pyrrolidone (5 ml) in dry toluene (50 ml), and the mixture was stirred at room temperature (0.2 h). The bromide 6 (Figure 1b, 1.6 g, 8.9 mmol) (Macías et al., 2009) was then added with continued stirring (0.2 h). The reaction was partitioned between water and ethyl acetate. The ethyl acetate was washed with saturated NaCl solution, dried and concentrated to leave a brown oil that was subjected to flash chromatography (EtOAc/hexanes, 1:9) to produce a yellow oil. This oil was subjected to semi-preparative HPLC (C18, MeOH–H2O 17:3) to produce Z–carlactone as the first of the two isomers to elute, with E–carlactone eluting directly afterwards.

Z–carlactone (Z–CL)

Pale yellow oil (15 mg, 1%). TLC (EtOAc:hexane, 1:4 v/v): Rf = 0.64. 1H-NMR (600 MHz, CD2Cl2): identical to that previously reported (Alder et al., 2012). 13C-NMR (150.9 MHz, CD2Cl2): δ 171.1 (1C, C14), 142.3 (1C, C12), 137.9, 137.8 (2C, C6 and C10), 134.6 (1C, C13), 129.0 (1C, C5), 127.1, 127.0 (2C, C7 and C8), 117.2 (1C, C9), 100.3 (1C, C11), 39.5 (1C, C2), 34.0 (1C, C1), 32.9 (1C, C4), 28.6 (2C, C16 and C17), 21.4 (1C, C18), 19.3 (1C, C3), 14.0 (1C, C19), 10.4 (1H, C15). HRMS (m/z): [M + H]+ calculated for C19H26O3, 303.1960; found, 303.1970.

E–carlactone (E–CL)

Pale yellow oil (105 mg, 4%). TLC (EtOAc:hexane, 1:4 v/v): Rf = 0.64. 1H–NMR (500 MHz, CD2Cl2): δ 6.93 (1H, p, = 1.6 Hz, H12), 6.44 (1H, m, H10), 6.00–6.04 (2H, m, H7 and H11), 5.91 (1H, d, = 16.0 Hz, H8), 2.00 (1H, m, H4), 1.96 (3H, d, = 1.5 Hz, H15), 1.78 (3H, d, = 1.3 Hz, H19), 1.67 (3H, m, H18), 1.62 (2H, m, H3), 1.46 (2H, m, H2), 1.00 (6H, d, = 1.8 Hz, H16 and H17). 13C–NMR (125.8 MHz, CD2Cl2): δ 171.7 (1C, C14), 142.9 (1C, C12), 142.0 (1C, C10), 138.4 (1C, C6), 135.3 (1C, C13), 132.2, 125.5 (2C, C7 and C8), 129.0 (1C, C5), 119.8 (1C, C9), 101.0 (1C, C11), 40.1 (1C, C2), 34.7 (1C, C1), 33.5 (1C, C4), 29.2 (2C, C16 and C17), 21.9 (1C, C18), 19.9 (1C, C3), 11.0, 10.0 (2C, C15 and C19). HRMS (m/z): [M + H]+ calculated for C19H26O3, 303.1960; found, 303.1960.

Plant Material

Atd14–1, kai2–2 and the double mutant have been described previously (Waters et al., 2012c). Seeds of max1–1 (Stirnberg et al., 2002), max2–1 (Stirnberg et al., 2002) and max3–9 (Booker et al., 2004) were obtained from the European Arabidopsis Stock Centre (http://arabidopsis.info/). For seedlings grown on 0.5× MS medium, carlactone and GR24 were added from 1000× stock solutions in acetone (1 or 10 mm); an equivalent volume of acetone was added to untreated controls.

Arabidopsis shoot branching assay

Growth of Arabidopsis plants was established under hydroponics as described previously (Waters et al., 2012c). A step-by-step protocol is available online (Waters et al., 2012b). Plants were grown in 0.5× Hoagland's nutrient solution under light provided by white fluorescent tubes emitting 150 μmol photons m−2 sec−1, with a 16 h light/8 h dark photoperiod, and a 22°C light/16°C dark temperature cycle with constant 60% humidity. Treatment commenced 14 days after germination by supplementing the medium with a 1:10 000 dilution of 10 mm Z–carlactone, 10 mm GR24 or 100% acetone (yielding final concentrations of 1 μm, 1 μm and 0.01%, respectively). Treatment was continued for a further 3 weeks, fully replacing the supplemented nutrient solution every 3 days. After this time, plants were provided with non-supplemented 0.5× Hoagland's solution for one further week, and the secondary rosette branches were counted.

Analysis of leaf morphology

Plants were grown under hydroponics and treated with 1 μm Z–CL and 1 μm GR24 as described for the shoot branching assay. Leaves were harvested 26 days after germination, when the first flowers were visible (i.e. 12 days after the start of treatment), laid flat between two sheets of colourless acetate, and scanned using a Microtek ScanMaker 9800XL (http://www.microtekusa.com) at 300 dpi and in 16–bit grayscale. Images were cropped and contrast-enhanced using Adobe Photoshop CS5.5 (http://www.adobe.com), and then imported into LeafAnalyser (Weight et al., 2008). For each genotype/treatment combination, consisting of 6–8 plants, LeafAnalyser was used to calculate a single average leaf model using leaf images from nodes 5–8, which were fully expanded and exhibited the clearest differences in morphology across genotypes. Each leaf model was traced in Adobe Illustrator CS5 to generate the silhouettes presented in Figure 3. To quantify leaf morphology, manual measurements were made using ImageJ software (http://rsbweb.nih.gov/ij/). Leaf length was defined as the distance between the leaf tip and the base of the petiole. Leaf width was measured as the greatest distance across the leaf lamina perpendicular to the proximal/distal axis of the leaf. The petiole/lamina boundary was defined as the inflection point at which the most distal edge of the lamina diverged by a 45° angle (or greater) from the proximal/distal axis.

Hypocotyl elongation assays

Hypocotyl elongation assays were performed as described previously (Waters et al., 2012c).

RNA isolation and transcript analysis

Seedling growth, RNA isolation, cDNA synthesis and quantitative RT–PCR were performed as described previously (Waters and Smith, 2013), with the exception that DNase treatment of total RNA was performed using TURBO DNase (Life Technologies, http://www.lifetechnologies.com) prior to cDNA synthesis. Oligonucleotide sequences for quantitative RT–PCR are listed in Table S1.

Statistical analysis

To compare hypocotyl responses and secondary shoot numbers between treatments within a single genotype, one-way, two-sided anova (Bonferroni t test) was performed. P values were derived from post hoc tests using Dunnett's adjustment for multiple pairwise comparisons to a single untreated control. For comparison of transcript abundance, an identical analysis was performed, except that expression ratios were converted to natural logarithms prior to anova to control for increasing variance with increasing expression level. For analysis of leaf morphological parameters, the effects of genotype, treatment and genotype*treatment were determined by two-way anova (Bonferroni t test) with Tukey's correction for multiple pairwise comparisons. For germination data, percentages were arcsine-transformed prior to anova. Statistical computation was performed using SAS Enterprise Guide 4.3 (www.sas.com).

Acknowledgements

This work was supported by the Australian Research Council (grant numbers LP0882775, DP1096717 and FT110100304). We are grateful to Rowena Long, Nicole Dakin and Santana Royan for technical assistance. We thank David Nelson (Department of Genetics, University of Georgia, USA) and Christine Beveridge (School of Biological Sciences, University of Queensland, Australia) for critical comments on the manuscript.

Ancillary