Specialisation within the DWARF14 protein family confers distinct responses to karrikins and strigolactones in Arabidopsis

Seed germination is a critical event in the plant life cycle. Many species exhibit physiological seed dormancy, which limits germination under favourable but transient environmental conditions that may not support long-term survival (Finkelstein et al., 2008). Wildfires present a brief opportunity for plants to exploit reduced competition for light, water and nutrients, and the dormant seed of a wide taxonomic range of species exhibit enhanced germination following smoke exposure (Roche et al., 1997; Chiwocha et al., 2009). The butenolide 3-methyl-2H-furo[2,3-c]pyran-2-one, or KAR₁, was identified in smoke as a bioactive compound that defines a family of related molecules known as karrikins (Flematti et al., 2004). Karrikins are potent germination stimulants, which are active in some species at concentrations as low as 1 nM (Flematti et al., 2004; Long et al., 2010). They also increase the sensitivity of seedlings to light, potentially enhancing seedling establishment and survival in the post-fire environment (Nelson et al., 2010). Elucidating the molecular genetic basis for responses to karrikins is key to exploiting their potential application in restoration ecology and agriculture (Daws et al., 2007; Stevens et al., 2007).

Germination of dormant Arabidopsis thaliana seed is generally promoted by karrikins, although karrikins cannot overcome the germination requirement for light and de novo synthesis of the phytohormone gibberellin (Nelson et al., 2009). To discover the molecular components of karrikin perception in Arabidopsis, we initiated a genetic screen for karrikin insensitive (kai) mutants. Two such mutants, exhibiting increased seed dormancy that could not be alleviated by karrikins, carried mutations in the MAX2 gene. Additional max2 alleles also conferred increased dormancy and insensitivity to karrikins, confirming that MAX2 is required for karrikin responses (Nelson et al., 2011). MAX2 is most renowned for its role in mediating responses to strigolactones, a class of plant-synthesized compounds that are exuded from roots, triggering the germination of parasitic weeds and promoting hyphal branching in arbuscular mycorrhizal fungi (Yoneyama et al., 2007; Dor et al., 2011a; Dor et al., 2011b). Recently, strigolactones were shown to be endogenous plant hormones that inhibit the outgrowth of axillary buds (Sorefan et al., 2003; Gomez-Roldan et al., 2008; Umehara et al., 2008) and influence root architecture (Kapulnik et al., 2011; Koltai, 2011; Ruyter-Spira et al., 2011). Two carotenoid-cleavage dioxygenases, CCD7/MAX3 and CCD8/MAX4, as well as a cytochrome P450, MAX1, are involved in strigolactone biosynthesis in Arabidopsis. Mutations in orthologous genes in rice, petunia and pea indicate that strigolactone control of shoot branching is conserved in angiosperms (Sorefan et al., 2003; Snowden et al., 2005; Arite et al., 2007). All of the max mutants share an increased shoot branching phenotype, but only max2 is strigolactone insensitive, implying that MAX2 is involved in the strigolactone response. Both karrikins and strigolactones induce similar effects at the germination and seedling stages in a MAX2-dependent manner (Nelson et al., 2011). Thus, despite their different origins, the karrikin and strigolactone signalling pathways both converge upon MAX2.

Consistent with a common signal transduction mechanism, karrikins and strigolactones share partial structural similarity. KAR₁ and KAR₂, the most active karrikins in Arabidopsis, have a butenolide moiety in common with the D-ring of strigolactones (supplementary material Fig. S1). The D-ring is necessary, but not sufficient, to stimulate the germination of parasitic weeds and to promote hyphal branching in arbuscular mycorrhizal fungi (Akiyama et al., 2005; Akiyama et al., 2010). Several simple butenolide variants are ineffective in seed germination stimulants of Arabidopsis (Nelson et al., 2011), indicating that the butenolide ring alone is not sufficient for bioactivity. Despite the molecular similarity, plant responses to karrikins and strigolactones differ in several respects. The synthetic strigolactone GR24 is highly effective in promoting germination of parasitic weeds (Striga and Orobanche spp.), but karrikins are not (Nelson et al., 2009). Both karrikins and GR24 are effective light-dependent inhibitors of hypocotyl elongation in Arabidopsis, but karrikins are much more effective than GR24 in promoting seed germination of Arabidopsis and Brassica tournefortii (Nelson et al., 2009; Nelson et al., 2010; Tsuchiya et al., 2010). Most strikingly, karrikins do not recover normal shoot branching in strigolactone-deficient mutants of Arabidopsis or pea (Nelson et al., 2011). Therefore, plants have the ability to distinguish between karrikins and strigolactones at various stages of development.

Given the similarities between karrikins and strigolactones, we reasoned that they might have related signalling components in addition to MAX2. More specifically, we hypothesized that plants must possess a mechanism to discriminate between the two classes of butenolides. Recently, it was shown that the rice DWARF14 (OsD14) gene, a member of the α/β hydrolase superfamily, is required for strigolactone-dependent control of shoot branching and acts in the same pathway as other strigolactone-related genes (Arite et al., 2009; Gao et al., 2009; Liu et al., 2009). This raised the possibility that related proteins within the same family might be involved in karrikin responses. Here, we demonstrate that two members of the DWARF14 family provide a molecular basis for differentiation of strigolactone and karrikin signalling.

Synthesis of KAR₁, KAR₂ and GR24 was performed as described (Mangnus et al., 1992; Goddard-Borger et al., 2007).

The spt-1 mutant, which carried the kai2-1 allele, was a gift from D. Smyth (Monash University). The kai2-1 allele was isolated from the spt-1 background by outcrossing to Ler and selecting for KAR-insensitive F₂ individuals that lacked the stunted gynoecium phenotype of spt mutants. All other seeds were obtained from the European Arabidopsis Stock Centre (NASC). The Atd14-1 mutant was isolated from the Wisconsin DsLox T-DNA insertion collection (NASC ID: N913109). The kai2-2 allele was derived from the Institute of Molecular Agrobiology (IMA) Ds insertion lines collection (N100282). The dlk2 mutants were isolated from the SALK T-DNA insertion collection: dlk2-1 (N679066), dlk2-2 (N657226) and dlk2-3 (N665057). The max2-8 mutant was described previously (Nelson et al., 2011). Genotyping primers are listed in supplementary material Table S1.

Plants were grown in peat, vermiculite and perlite mixture (6:1:1) under fluorescent lamps emitting 100-120 μmol photons s^–1 m^–2 with a 16-hour light/8-hour dark photoperiod and a 22°C light/16°C dark temperature cycle. Germination tests were performed under constant white light at 20°C (Nelson et al., 2009; Nelson et al., 2010). For hypocotyl elongation and cotyledon expansion assays, surface-sterilised seeds were sown on solidified 0.5× MS media (pH 5.9) and stratified in the dark at 4°C for 72 hours. At 20°C the seeds were exposed to white light for 3 hours, transferred to dark for 21 hours, and then exposed to continuous red light (LED, λ_max=652 nm, 20 μmol photons s^–1 m^–2) for 4 days. Hypocotyls and cotyledons were measured using ImageJ (http://imagej.nih.gov/ij/). Karrikin and GR24 stock solutions (1000× or 2000×) were prepared in acetone; equivalent acetone volumes were added to untreated controls.

Total RNA was isolated from seedlings using the Qiagen RNeasy procedure. RNA isolation from seed, DNase treatment, cDNA synthesis and qRT-PCR were performed as described (Nelson et al., 2010). For primers, see supplementary material Table S2.

Seeds were sown on rock wool plugs held in black 1.5-ml microcentrifuge tubes (bottom end and cap removed). The tubes were placed in 24-well plastic boxes (I5100-43, Astral Scientific, NSW, Australia) containing Hoagland’s nutrient solution (Heeg et al., 2008), and the boxes placed in the dark at 4°C for 3 days. Seeds were germinated under the growth conditions described above. Five days after germination, seedlings were thinned to one seedling per tube. One week later, individual plants were spaced out to four plants per box, with three boxes per genotype/treatment combination. Boxes were randomised with respect to position on the shelf. At this point, the media were supplemented with either 5 μM GR24 or 0.05% (v/v) acetone, and changed every 4-6 days as required. Rosette leaves were counted upon bolting, and secondary rosette branches (>5 mm in length) were counted when the primary inflorescence ceased growth.

KAI2 homologues were identified with BLASTP searches of GenBank protein databases using the Arabidopsis KAI2 amino acid sequence as a query (http://www.ncbi.nlm.nih.gov). To identify DLK2 orthologues in monocots, the Arabidopsis DLK2 sequence was used in a separate BLASTP query. Additional sequences were obtained from the Plant Genome Database (http://www.plantgdb.org) and the incomplete JGI Marchantia genome sequencing effort (supplementary material Table S3). Sequences were sampled from a broad taxonomic spread and screened for duplication and truncation. Full-length sequences were then aligned using MAFFT (http://mafft.cbrc.jp/alignment/software) using the default settings, and the alignment was conservatively edited to remove regions of poor alignment using PFAAT (http://pfaat.sourceforge.net). MrBayes v3.1.2 (Ronquist and Huelsenbeck, 2003) was used to infer Bayesian trees (random start tree, eight chains of temperature 0.2, WAG substitution matrices, four discrete categories of gamma distribution substitution rate). Maximum likelihood phylogenies were produced using PHYML v2.4.4 (Guindon and Gascuel, 2003) (100 replicates, WAG matrix, four gamma categories, alpha parameter re-estimation for each replicate).

For comparisons of branch numbers per rosette leaf between different genotypes, one-way, two-sided ANOVA (Bonferroni t-test) was performed. For hypocotyl lengths, branching assays and transcript levels, the effects of genotype, treatment and genotype×treatment were analysed by two-sided ANOVA. P-values were derived from post-hoc tests using Tukey’s correction for multiple pairwise comparisons. For germination data, seed germination percentages were arcsine transformed prior to analysis. Statistical analysis was performed with SAS Enterprise Guide 4.3 (SAS, Cary, NC; www.sas.com).

In rice, OsD14 is considered necessary for strigolactone signalling because the increased tillering phenotype of Osd14 mutants cannot be reversed with exogenous GR24 (Arite et al., 2009). We searched for mutations in the closest Arabidopsis homologue of OsD14, At3g03990, which we term AtD14. We obtained one mutant allele, Atd14-1, which carries a T-DNA insertion within the first 100 bp of the AtD14 coding region, rendering full-length transcripts undetectable (supplementary material Fig. S2A). Homozygous Atd14-1 mutants have reduced stature and increased numbers of secondary rosette branches (Fig. 1A). Knockdown of AtD14 transcripts by artificial microRNAs leads to a similar increased branching phenotype (supplementary material Fig. S2B,C).

Upon flowering, some axillary buds at the base of rosette leaves activate and grow out to form secondary branches. In strigolactone signalling mutants, inhibition of such outgrowth is lost, and more rosette branches develop. To test whether the Atd14-1 branching phenotype is due to insensitivity to strigolactones or to a deficiency in strigolactone biosynthesis, we grew plants in hydroponic medium supplemented with GR24. The strigolactone biosynthetic mutant max3 responded to GR24 application with a significant reduction in secondary bud outgrowth (P<0.001), but Atd14-1 and max2-1 were insensitive to GR24 (Fig. 1B). These data are consistent with the rice Osd14 mutant phenotype and suggest that, like MAX2, AtD14 is necessary for strigolactone response.

To test whether Atd14-1 mutants are also defective in karrikin responses, we first examined the effects of karrikins on seed germination. In our hands, the Arabidopsis Col-0 ecotype exhibits little seed dormancy and consequently shows relatively little enhancement in germination by karrikins (Nelson et al., 2009; Nelson et al., 2011). The Atd14-1 mutant exhibited a germination profile indistinguishable from wild type. By contrast, max2-1 showed substantially delayed germination that was insensitive to karrikins or GR24 (Fig. 1C). Because of the weak dormancy of Col-0, a more sensitive assay for karrikin responses is the inhibition of hypocotyl elongation in seedlings grown under red light (Nelson et al., 2010; Nelson et al., 2011). Wild-type seedlings grown in the presence of KAR₁, KAR₂ and GR24 exhibited a shortened hypocotyl, with GR24 being the most potent compound (Fig. 1D). However, although Atd14-1 hypocotyls responded normally to karrikins they were substantially less sensitive to GR24: on average, wild-type seedlings grown on 1 μM GR24 were 37% of the height of seedlings grown without GR24, as compared with 62% for Atd14-1 (Fig. 1D).

To examine this response further, we assayed hypocotyl length and cotyledon size over a range of GR24 concentrations. In wild-type seedlings, increasing concentrations of GR24 induced smaller cotyledons and a shorter hypocotyl (Fig. 1E). Although this trend was generally also true for Atd14-1 seedlings, the response was much weaker, with inhibition of cotyledon expansion only detected at 10 μM GR24. Furthermore, Atd14-1 hypocotyls were longer than wild-type hypocotyls at all concentrations tested (Fig. 1D,E). It is noteworthy that whereas KAR₁ and KAR₂ promote cotyledon expansion (Nelson et al., 2010), we find that GR24 has the opposite effect and inhibits cotyledon growth.

Further, we measured the accumulation of STH7 transcripts, which increase in abundance in Landsberg erecta (Ler) seedlings in response to KAR and GR24 (Nelson et al., 2010; Nelson et al., 2011). The STH7 transcript responses to karrikins in Atd14-1 and Col-0 seedlings were comparable (Fig. 1F). However, the induction of STH7 transcripts by GR24 was significantly lower in Atd14-1 mutants than wild type (P<0.001), consistent with the decreased hypocotyl response. Two additional transcripts showed a similar pattern to STH7 (supplementary material Fig. S2D).

Together, these data indicate that AtD14 is dispensable for karrikin responses, but is essential for normal strigolactone responses in seedlings and mature plants.

The rice genome encodes an uncharacterised protein similar to DWARF14, termed D14-LIKE (Arite et al., 2009). We reasoned that DWARF14 homologues might have attained distinct functions during evolution, so we searched for mutations in the Arabidopsis orthologue of D14-LIKE (At4g37470) to ascertain whether this protein is important for karrikin or strigolactone signal transduction. We obtained two mutant alleles via two independent routes. First, we discovered that spt-1, a weak mutant allele of the bHLH transcription factor SPATULA (Alvarez and Smyth, 1999), was insensitive to karrikins whereas three additional null spt alleles responded normally (data not shown). Therefore, we suspected that a second mutation in the spt-1 background was responsible for the karrikin-insensitive phenotype, and we renamed this unknown mutation kai2-1. Second, we used reverse genetics to identify a Ds transposon insertion allele of At4g37470, which we found was also karrikin insensitive. Given the genetic proximity of SPATULA (At4g36930) to At4g37470 (∼202 kb), we speculated that kai2-1 might carry a mutation in At4g37470; upon sequencing we discovered a G-to-A transition, modifying a conserved glycine residue to glutamic acid (Fig. 2A, supplementary material Fig. S3). As kai2-1 and the insertion mutant were allelic (supplementary material Fig. S4), we concluded that At4g37470 was KAI2 (Fig. 2A) and named the insertion allele kai2-2.

Both kai2 alleles are in the Ler background, which exhibits higher levels of seed dormancy than Col-0 (van Der Schaar et al., 1997; Nelson et al., 2009). We found that whereas 1 μM KAR₁, 1 μM KAR₂ and 10 μM GR24 strongly promoted the germination of primary dormant Ler seed, both kai2 alleles were fully insensitive to all three compounds (Fig. 2B). Like max2, the germination of fresh kai2 seed on water-agar was negligible, but this enhanced dormancy could be overcome with cold stratification in the presence of nitrate (Fig. 2B). To examine additional karrikin responses, we performed hypocotyl elongation assays. We first noticed that kai2 seedlings had longer hypocotyls than wild type, a trait shared with max2 (Fig. 2C). Crucially, although kai2 hypocotyls were fully insensitive to karrikins they were still responsive to GR24 (Fig. 2C). This contrasts with max2 seedlings, which were insensitive to both karrikins and GR24. To confirm this differential sensitivity on the transcript level, we measured the accumulation of two KAR- and GR24-responsive transcripts, STH7 and KUF1, in seedlings. Both kai2 alleles showed the same pattern of transcriptional marker response: no induction by karrikins but a strong positive response to GR24 (Fig. 2D). In addition, both kai2 and max2 exhibited decreased levels of STH7 and KUF1 transcripts relative to wild type, indicating that both MAX2 and KAI2 are required for normal expression of these two marker genes.

Together, these data suggest that KAI2 is required for responses to karrikins and is necessary for responses to exogenous strigolactone in seeds but not in seedlings. These results also indicate that KAI2 is necessary for normal seed germination and seedling development, whereas AtD14 is not.

KAI2 and DWARF14 belong to a family of poorly characterised proteins classified as α/β hydrolases (Pfam domain: Abhydrolase_6 PF12697). The only characterised member of this family is RsbQ from Bacillus subtilis. RsbQ is an energy-stress response regulator required for the activity of a protein phosphatase that ultimately regulates the transcription of a stress response operon (Brody et al., 2001). At the amino acid level, AtD14 and KAI2 are respectively 38% and 39% identical to RsbQ and both share the putative hydrolase catalytic triad of Ser96, Asp219 and His247 residues (supplementary material Fig. S3). To investigate the evolution of DWARF14 and KAI2, we carried out a phylogenetic analysis of related proteins. Beyond land plants, the most similar KAI2 homologues are found in eubacteria. Both Bayesian inference and maximum likelihood analyses produced consensus trees with highly similar topologies (Fig. 3), with the land plant sequences forming two sister clades. The DWARF14 clade contains AtD14 and OsD14, while the D14-LIKE clade contains the rice founding member OsDLK1 and its Arabidopsis orthologue KAI2.

The genome of the moss Physcomitrella patens encodes multiple KAI2 homologues, some of which are weakly assigned to the base of the D14-LIKE clade. However, none could be definitively assigned to the DWARF14 clade, which contains a sequence from just one non-seed plant, Selaginella moellendorffii (Fig. 3). By contrast, the D14-LIKE clade is taxonomically diverse, encompassing mosses, liverworts, lycophytes, gymnosperms and angiosperms. Such diversity within the D14-LIKE clade suggests that this protein is the ancestral form within the land plants, and that DWARF14 sequences arose from a gene duplication event within the vascular plant lineage. In addition, a subclade of the DWARF14 clade contains members from the angiosperms and at least one gymnosperm; we named this subclade DLK2, as it contains a second Arabidopsis protein with similarity to AtD14. Thus, there was another gene duplication event within the DWARF14 clade, which presumably occurred after the evolution of seed plants. The long branches within the DLK2 subclade suggest that there has been substantial sequence divergence within this group, while the ‘core D14’ sequences are relatively more conserved (Fig. 3). This distinction might reflect the acquisition of a novel function by DLK2 in angiosperms, while D14 function was constrained. Together, our phylogenetic analysis suggests that the DWARF14 clade contains more recently divergent homologues, whereas the D14-LIKE clade is evolutionarily more ancient.

To assess whether the Arabidopsis DLK2 protein has a role in karrikin or strigolactone signalling, we isolated three mutant T-DNA insertion alleles in the corresponding gene, At3g24420 (supplementary material Fig. S5A,B). All three dlk2 mutants exhibited normal hypocotyl elongation responses to KAR₁, KAR₂ and GR24 (supplementary material Fig. S5C). Furthermore, there was no discernible effect of dlk2 mutations on seedling morphology or the number of rosette branches (supplementary material Fig. S5D,E). We conclude that DLK2 does not play a crucial role in karrikin or strigolactone signalling in Arabidopsis.

The increased shoot branching pattern of Atd14-1 and the increased seed dormancy of kai2 are both common to max2 mutants. To analyse the shared phenotypes more closely, we compared each mutant at different stages of development. At the 7-day seedling stage, Atd14-1 resembles wild type, but max2-1 seedlings have an elongated hypocotyl and hooked, epinastic cotyledons (Fig. 4A). By contrast, kai2 seedlings are indistinguishable from max2-8 (Fig. 4B), a similarity that mirrors the germination phenotypes (Fig. 2B). Later during vegetative development, Atd14-1 closely resembles max2-1, with shortened leaf petioles and broader, more highly lobed leaves (Fig. 4C). By contrast, the leaves of kai2 are frequently elongated with curled margins, a phenotype that is most severe in the kai2-2 allele (Fig. 4D). Although max2 also exhibits some leaf curling, the overall leaf shape and rosette size of max2 are distinct from kai2 (Fig. 4D). Upon flowering, Atd14-1 has an increased branching phenotype similar to max2 (Fig. 4E,G), but kai2 has a branching pattern that is indistinguishable from wild type, indicating that strigolactone signalling is unperturbed (Fig. 4F,G). Thus, KAI2 is not required for branching control, consistent with KAI2 being necessary for karrikin responses and our previous finding that karrikins cannot substitute for strigolactones in the control of axillary bud outgrowth (Nelson et al., 2011). Together, these data show that Atd14-1 resembles max2 during the adult growth phase and that kai2 is similar to max2 only during the seed and juvenile phases. These broad phenotypic similarities suggest that MAX2 functions together with AtD14 and KAI2, but at distinct stages of plant development.

To test for genetic interaction between KAI2 and MAX2, we generated a kai2-1 max2-8 double mutant and analysed levels of STH7 and KUF1 transcripts in seedlings. As expected, the GR24 responses observed in kai2-1 were abolished in the double mutant, but otherwise the transcript profiles were indistinguishable from those of either single mutant (Fig. 4H). Furthermore, kai2-2 max2-8 hypocotyl responses resemble those of max2 (supplementary material Fig. S6), consistent with KAI2 and MAX2 operating in the same signalling pathway.

Compared with wild type, kai2 exhibits reduced germination on water-agar, but Atd14-1 does not (Figs 2, 3), showing that KAI2, but not AtD14, has an important role in seed germination. Previously, we discovered that primary dormant Ler seed is substantially more sensitive to karrikin than to GR24 in a germination assay, requiring 10-fold more GR24 than karrikin to stimulate germination (Nelson et al., 2009). We reasoned that spatiotemporal regulation of KAI2 and AtD14 expression might be responsible for the differential sensitivity of seed to karrikins and GR24. We measured transcript levels in Ler seed that had been water imbibed for 24 hours, and found that KAI2 transcripts were ∼100-fold more abundant than AtD14 transcripts (Fig. 5A), suggesting that KAI2 dominates in seeds. Simultaneously, we discovered that KAI2 and AtD14 transcripts respond minimally to KAR or GR24, but that DLK2 transcripts increase in abundance several fold (Fig. 5A), making DLK2 a serendipitously good marker for KAR and GR24 responses. To confirm the differential sensitivity of Ler seed to KAR and GR24 on the transcript level, we exposed seed for 24 hours to different concentrations of KAR₂ and GR24. Based on STH7 and DLK2 transcript abundance, Ler seeds are 10- to 100-fold less sensitive to GR24 than to KAR₂ (Fig. 5B). Taken together, these results suggest that KAI2 predominately mediates the stimulation of germination by butenolides, and that KAI2-dependent signalling is promoted by KAR in preference to GR24. Notably, these findings are consistent with the insensitivity of kai2 seed to GR24 (Fig. 2B).

Whereas kai2 seeds do not respond to GR24, kai2 seedlings evidently do (Fig. 2C). We hypothesized that AtD14 might mediate the GR24 responses in kai2 seedlings, and found that in seedlings AtD14 transcripts were slightly more abundant than KAI2 transcripts, in contrast to imbibed seed (Fig. 5C). Although AtD14 transcripts responded positively to light, the relative induction was considerably lower than for KAI2 transcripts, which increased 5-fold in light-grown seedlings over dark-grown controls (Fig. 5C). DLK2 transcripts, however, were unresponsive to light. Together, these results show that KAI2 and AtD14 transcripts are similarly expressed in seedlings, and the particular sensitivity of KAI2 to light might explain the known light requirement for karrikin effects on germination and seedling development (Nelson et al., 2009; Nelson et al., 2010).

To further investigate the regulation of KAI2, AtD14 and DLK2 transcripts in seedlings, we assessed the response of each to KAR and GR24. As in seed, neither KAI2 nor AtD14 transcript levels varied in response to KAR or GR24 in wild-type seedlings, nor was there any evidence of KAI2- or MAX2-dependent regulation of KAI2 or AtD14 expression (Fig. 5D). By contrast, and as observed in imbibed seed, DLK2 transcripts responded positively to both KAR and GR24 and were ∼6-fold downregulated in kai2 and max2 (Fig. 5D). Thus, whereas DLK2 is among those genes (such as STH7 and KUF1) that respond to KAR and GR24 and may enact physiological responses, KAI2 and AtD14 are more uniformly expressed, performing early in the signal transduction process. Importantly, both KAR- and GR24-mediated regulation of DLK2, STH7 and KUF1 require MAX2, but only KAR-mediated regulation requires KAI2. These findings provide additional evidence for KAI2 and AtD14 specificity and for both proteins acting upstream of MAX2.

As the residual response of Atd14-1 seedlings to GR24 might reflect the redundant activity of KAI2, we generated kai2-2 Atd14-1 double mutants. Such seedlings were insensitive to both karrikins and GR24 in a hypocotyl elongation assay (Fig. 6A). In addition, the expression patterns of STH7, KUF1 and DLK2 transcripts in kai2-2 Atd14-1 seedlings resembled those of untreated kai2-2 seedlings, even in the presence of GR24 (Fig. 6B). IAA1 transcripts, which are positively auxin responsive and are upregulated in max2 (Hayward et al., 2009; Nelson et al., 2011), were also upregulated in kai2-2, but not in Atd14-1. Significantly, GR24 was able to suppress IAA1 transcript levels in all genotypes except the double mutant (Fig. 6B). These data indicate that both KAI2 and AtD14 can mediate GR24 responses, but that KAI2 is functionally dominant in seedlings, as exemplified by the long hypocotyls and aberrant transcript levels of untreated kai2 seedlings. We suggest that the stronger responses of wild-type seedlings to GR24 over karrikins result from the combined activity of both AtD14 and KAI2: whereas GR24 signals via both KAI2 and AtD14, karrikins only act via KAI2. However, this potential dual signalling role of KAI2 in seedlings might be physiologically unimportant: it is possible that KAI2 does not mediate responses to endogenous strigolactones given that the strigolactone-deficient mutants max1, max3 and max4 do not share any phenotypes with kai2 mutants.

With the discovery that karrikin and strigolactone signalling pathways both require MAX2 (Nelson et al., 2011), it remained unclear how plants discriminate between these two structurally similar but physiologically distinct growth regulators. Here, we show that the independent perception of karrikins and strigolactones results from evolutionary specialisation within a divergent family of α/β hydrolase-like proteins.

Karrikins and strigolactones are two known types of plant signalling compounds that share a butenolide ring. Karrikins have only been identified in smoke and have not been reported as endogenous plant compounds. Our findings provide further support for the existence of an endogenous, non-strigolactone signal that acts through the butenolide signalling system to influence seed germination and early seedling development. First, karrikin response is broadly conserved among angiosperms (Flematti et al., 2004; Stevens et al., 2007; Long et al., 2011), but many such species (including Arabidopsis) are not associated with fire-prone environments and thus are not under an obvious selective pressure to retain a capacity to respond to smoke signals per se. We now know that at least two loci that facilitate this signalling mechanism, KAI2 and MAX2, have been maintained in the Arabidopsis genome. Second, both max2 and kai2 mutants have similar phenotypes in terms of seed germination, seedling morphology and STH7, KUF1, DLK2 and IAA1 transcript levels. These mutant phenotypes are opposite to the effects of karrikin or strigolactone application on wild type, consistent with the failure of a signalling system to respond to a putative endogenous butenolide. Third, the germination and seedling phenotypes of max2 and kai2 are unlikely to be due to a defect in strigolactone signalling because the strigolactone-deficient max1, max3 and max4 mutants are normal in these respects (Nelson et al., 2011). Moreover, kai2 seedlings still retain the capacity for strigolactone response through AtD14, yet show a reduced photomorphogenic phenotype not seen in strigolactone-deficient mutants or in Atd14 seedlings. We propose that plants from fire-prone regions might have recruited an endogenous signalling mechanism, which is involved in seed germination and seedling development and still present in plants more broadly (including Arabidopsis), to detect and respond to karrikins found in smoke.

The normal strigolactone responses of kai2 mutants suggest that, whereas MAX2 has a diverse role throughout plant development, the function of KAI2 is more specialised. Recently, KAI2 was independently identified in a screen for light signalling mutants (Sun and Ni, 2011). The hyposensitive to light (htl) mutants exhibited reduced sensitivity to light of all wavelengths, indicating a general lesion in seedling photomorphogenesis; however, no germination phenotype was reported, nor was any similarity to max2, which also exhibits the same broad photomorphogenesis phenotypes (Shen et al., 2007). Interestingly, KAI2/HTL appears to be a target of the transcriptional regulator HY5, a central component of light signalling that acts downstream of multiple photoreceptors (Oyama et al., 1997; Ulm et al., 2004; Sun and Ni, 2011). The MAX2 and HY5 signalling pathways are likely to be separate with respect to hypocotyl development (Tsuchiya et al., 2010), suggesting that KAI2 might operate in the MAX2 pathway and HY5 in another, but with the capacity to regulate KAI2 transcriptionally. Although the precise relationship between KAI2 and HY5 is unclear, it is consistent with the demonstrated crosstalk between light and karrikin signalling and the observation that hy5-1 seedlings have weak karrikin responses (Nelson et al., 2010).

Recently, it was demonstrated that the moss Physcomitrella patens responds to GR24 and has the genetic machinery for strigolactone biosynthesis (Proust et al., 2011). As the P. patens genome also encodes a MAX2 orthologue (Waters et al., 2011) and several candidate D14-LIKE homologues (Fig. 3), it appears that the strigolactone signalling system is evolutionarily ancient, arising before the divergence of the bryophytes and the vascular plants some 420 million years ago (Kenrick and Crane, 1997). However, given that the genome of the liverwort Marchantia polymorpha encodes two KAI2 orthologues but no DWARF14 orthologues, it is plausible that strigolactone signalling was not the original function of this protein family in the first land plants. In Arabidopsis, it is clear that KAI2 mediates the activity of both karrikins and GR24, whereas AtD14 is specific for GR24 and, presumably, endogenous strigolactones. If the more relaxed chemical specificity of KAI2 is common to its orthologues in other species, then the ancestral protein might have had a more diverse role. Elucidating the most likely ancestral function of the DWARF14 family of α/β hydrolases will require a close examination of the responses of mosses and liverworts to butenolides and the effects of gene-specific knockouts. We propose that, in angiosperms such as rice and Arabidopsis, gene duplication and specialisation within the DWARF14 family has led to distinct signalling functions for KAI2 and D14 orthologues. In basal land plants, the division between karrikins, strigolactones and potentially other butenolide-containing compounds might be less distinct.

The overlapping phenotypes of max2, kai2 and Atd14-1 suggest that KAI2 and AtD14 act together with MAX2. MAX2 is an F-box protein that interacts with other core components of the SCF complex, a class of E3 ligases that ubiquitylates specific proteins and so marks them for degradation (Woo et al., 2001; Stirnberg et al., 2007). In the SCF complex, the F-box protein confers substrate specificity, acting as an adapter between the core subunits and the target protein (Somers and Fujiwara, 2009). Previously, we postulated that a D14 homologue might regulate MAX2 activity in a similar manner to the gibberellin (GA) signalling system (Nelson et al., 2011). The GA receptor GID1, which is also an α/β hydrolase superfamily member, promotes the binding of growth-repressing DELLA proteins to the F-box protein SLY1 in a GA-dependent manner, thus regulating DELLA degradation by the SCF complex (Harberd et al., 2009; Sun, 2010). We now extend this model to include AtD14 and KAI2 as central signalling components for strigolactones and karrikins, respectively (Fig. 7). Although direct protein interactions between MAX2 and either KAI2 or AtD14 have not yet been demonstrated, both MAX2 and KAI2 are localised to the nucleus (Shen et al., 2007; Stirnberg et al., 2007; Sun and Ni, 2011). In this GA signalling-like scenario, both AtD14 and KAI2 can function with MAX2, but the relative response of the SCF^MAX2 complex to karrikins and strigolactones is dependent, in part, on the relative availability of AtD14 and KAI2 proteins and their respective signalling butenolides. Notably, KAI2 and AtD14 are differentially expressed during development, whereas MAX2 is constitutively expressed at a steady level (supplementary material Fig. S7). With regard to karrikin and strigolactone signalling, there is evidence for a ‘division of labour’ between KAI2 and AtD14 in modulating aspects of MAX2 activity. Specifically, our data suggest that KAI2 is responsible for the strigolactone-independent functions of MAX2. We propose that during seed germination and seedling establishment, KAR (or an endogenous butenolide) promotes KAI2-MAX2 interaction, leading to degradation of growth repressor proteins. At later stages, most notably during flowering in Arabidopsis when secondary bud outgrowth is initiated, increased AtD14 activity would allow strigolactone-dependent functions of MAX2 to dominate.

As hypothetical hydrolases, AtD14 and KAI2 might generate an active signal from strigolactones and karrikins. We note that the serine residue of the catalytic triad could accommodate a receptor-ligand interaction by means of a nucleophilic attack upon an enol-ether functionality within the ligand, a mechanism proposed for both strigolactones and karrikins (Chiwocha et al., 2009; Zwanenburg et al., 2009). However, no catalytic activity has been demonstrated for either KAI2 or RsbQ (Brody et al., 2001; Sun and Ni, 2011), and, in our hands, purified KAI2 does not exhibit any catalytic activity towards KAR₁ in vitro. Other lines of evidence are consistent with a hypothesis that KAI2 and AtD14 are ligand-binding proteins. First, the Bacillus subtilis RsbQ protein has been shown to bind the serine protease inhibitor PMSF, and thus the postulated active site of RsbQ is thought to bind ‘a hydrophobic small compound’, the identity of which is unknown (Kaneko et al., 2005). Additionally, although GID1 and DWARF14-like proteins are not related on the amino acid sequence level, both contain the topologically similar α/β hydrolase fold (Kaneko et al., 2005; Shimada et al., 2008). In GID1, valine replaces histidine in the catalytic triad, which compromises its hydrolytic activity while providing a binding site for gibberellins (Ueguchi-Tanaka et al., 2005). Thus, KAI2 and AtD14 might perform a hybrid role in which the substrate is hydrolysed, followed by a conformational change in protein structure that could modulate MAX2 activity. KAI2 and AtD14 could bind their ligand directly or in partnership with MAX2 and/or specific repressor proteins. To date, we have not been able to detect any binding of KAR₁ with KAI2 in vitro, implying that the interaction might be highly transient or that other factors (such as protein partners) are involved. Although the details of the molecular interactions require further investigation, it is now clear that D14 and KAI2 proteins allow plants to discriminate between strigolactones and karrikins to trigger appropriate environmental and developmental responses.

We thank Matthew Hodges (University of Oxford, UK) for assistance with phylogenetic analysis; John Bowman and David Smyth (Monash University, Australia) for providing sequence data, spt mutants and valuable discussion; and Winslow Briggs (Carnegie Institution, Stanford, USA), Rowena Long, Emilio Ghisalberti (UWA, Australia), Christine Beveridge, Elizabeth Dun and Philip Brewer (UQ, Australia) for valuable discussion.