Karrikins (KAR₁, KAR₂), GR24^5DS, and GR24^ent-5DS were prepared as described (Flematti et al., 2007; Goddard-Borger et al., 2007; Scaffidi et al., 2014) and dissolved as 10 mM stock solutions in acetone. Epibrassinolide (Sigma E1641), gibberellic acid (GA₄ from L. N. Mander, Australian National University), 3-indoleacetic acid (Sigma I2886), (+)-cis, trans-abscisic acid (AG Scientific A-1103) and (±)-jasmonic acid (Sigma J2500) were dissolved in acetone as 5, 10, 10, 10, and 50 mM stock solutions, respectively.

The DLK2 promoter sequence was defined as the 3566 bp of genomic sequence spanning the annotated transcriptional start site of DLK2 (At3g24420) and 103 bp downstream of the annotated 3′ UTR of the preceding gene (At3g24430). We also included the DLK2 5′UTR (31 bp). This sequence was amplified by PCR using Phusion polymerase (New England Biolabs). Oligonucleotides were (5′–AAAAAAGCAGGCTCAAACGCGATAACCTTTTCA–3′) and MW446 (5′–CAAGAAAGCTGGGTGCTTAAGTACAAGAGTTTTG–3′); regions of homology to Arabidopsis genomic DNA are underlined. Gateway-compatible attB recombination sites were added in a further round of PCR and the resulting product cloned into pDONR207. This intermediate plasmid was recombined with the binary vector pHGWL7 (Karimi et al., 2002), inserting the DLK2 promoter sequence upstream of firefly luciferase coding sequence.

The DLK2:LUC construct was introduced into Arabidopsis Ler background by floral dipping. Primary transformants were selected on 20 μg ml^-1 hygromycin B. Six lines that segregated 3:1 for hygromycin resistance were propagated to homozygosity and subsequently screened for LUC activity in response to KAR and racemic GR24. The most robustly responding line was then crossed with kai2-2 (Ler) (Waters et al., 2012) and experiments were performed on the F₃ generation homozygous for both kai2-2 and the DLK2:LUC transgene.

Twenty milligrams of Arabidopsis seeds were imbibed in 1 ml water supplemented with KAR or acetone for 48 h at 20°C under continuous light. RNA extraction and quantitative PCR was conducted as described (Waters et al., 2012). Oligonucleotides for LUC transcripts were 5′–ATTCTTTATGCCGGTGTTGG–3′ and 5′–TGTTGAGCAATTCACGTTCA–3′.

Recombinant firefly luciferase in buffered aqueous solution (Sigma L9420) was diluted to 10^-8g μl^-1 with lysis solution (25 mM Tris-phosphate pH 7.8, 2 mM DTT, 2 mM 1,2-diaminocyclohexane-tetraacetic acid (DACTAA), 10% glycerol, 1% Triton X-100). A 10-fold dilution series in lysis solution was made, ranging from 10^-8 g μl^-1 to 10^-18 g μl^-1. Each luciferase stock solution was further diluted 1 in 20 with lysis solution and 20 μl of each dilution was transferred to a white opaque 96-well assay plate (Sigma-Aldrich, CLS3912) in triplicate. Triplicate lysis solution served as background control.

A POLARstar OPTIMA (BMG LABTECH) was used to measure luminescence. The injector was rinsed with 4.5 ml water, then primed with 1 ml Luciferase Assay Reagent (LAR; 15 mM K₂PO₄/KH₂PO₄ pH 7.8, 25 mM Gly-Gly, 4 mM EGTA, 15 mM MgSO₄, 2 mM ATP, 1 mM DTT, 0.1 mM Coenzyme A, 120 μM luciferin) (Dyer et al., 2000). The instrument was programmed to inject 100 μl LAR (260 μl/second) one well at a time, to shake the assay plate for 5 s after each injection (1 mm shaking width, 600 rpm), and to measure luminescence signal for 5 s using the top optic (gain 4095). The assay plate was read from the lowest to the highest luciferase concentration, to avoid light contamination caused by higher-concentration luciferase samples. After measurements were completed, the mean luminescence signals produced by the background control lysis solution was subtracted from each luminescence reading produced by luciferase samples, to obtain the net reading for each sample.

Dry Arabidopsis seeds (Landsberg erecta; 2.5 mg) were distributed with a home-made seed scoop into 1.2 ml tubes in eight-tube strips (Astral Scientific, I1720-00) held by a rack. The seeds were collected at the bottom by brief centrifugation. Compound stocks (1000-fold) in acetone were diluted in water and added to each tube (100 μl). The seeds were resuspended in treatment solutions by flicking the tubes. The seed tubes were incubated at 20°C under continuous light for between 24 and 72 h.

After incubation, the seed tubes were centrifuged at 4000 rpm for 5 min to collect seeds at the bottom. Treatment solution was removed from each tube using a multi-channel pipette set at 75 μl, with care taken not to remove any seeds. Two, 1 mm-diameter stainless steel balls, were distributed into each tube. Lysis solution (80 μl) was added to each tube. The seed tissues were ground in lysis solution using a mixer mill at 30/s for 1 min twice. The seed extracts were centrifuged at 4000 rpm for 10 min to pellet tissue debris. The supernatant (20 μl), containing extracted luciferase enzyme, was transferred to a white opaque 96-well assay plate (Sigma-Aldrich, CLS3912). Triplicate lysis solution served as background control.

Luminescence was measured as described in the previous section. The plate reading direction was perpendicular to the biological replicates loading direction, to avoid time-dependent bias.

The net readings were obtained as described in the previous section. The mean net luminescence reading was taken for mock and each treatment. Fold change in LUC activity was calculated by the following formula:

Standard errors of mean net luminescence readings were calculated and scaled to the fold change in LUC activity.

Arabidopsis Ler seeds (0.6 ml) were sown directly on soil (peat:vermiculite:perlite 6:1:1) in 20 rectangular pots (Garden City Plastics, PUNSTX, volume 400 ml) distributed across two trays. The seeds were stratified for 3 days in the dark at 4°C, before being transferred to a climate-controlled growth room (8 h light/16 h dark photoperiod, 22°C light/16°C dark temperature cycle, 100–150 μmol m^-2 s^-1 PAR, 60% relative humidity). Rosette tissue from 7-week old plants prior to flowering was harvested, weighed (circa 200 g FW), and frozen in liquid nitrogen. Leaf tissue was ground in liquid nitrogen with a mortar and pestle, and extracted with 80% (v/v) methanol in water at 4°C overnight (10 ml per gram FW). The next day, the methanol/water extract was filtered with Whatman filter paper (18.5 cm, No. 4) to remove tissue debris. Methanol was removed under reduced pressure at 40°C. The remaining aqueous extract was diluted with water to 100 ml, and extracted with ethyl acetate (3 × 100 ml). The aqueous layer was concentrated under reduced pressure at 40°C to 10 ml and stored at -20°C. The combined ethyl acetate extract was evaporated to dryness under reduced pressure to give 0.8 g solid material. The ethyl acetate extract was re-constituted with 10 ml of purified water on a rotating wheel overnight at 4°C. A total of 0.1% of each extract (annotated as “stock”), and dilutions of 1/5 and 1/25 were tested with the DLK2:LUC assay.

Significance groupings were determined using one-way ANOVAs based on Honestly Significant Differences (HSD) Test. The analyses were performed in R Program v3.2.3, using the package “agricolae.”

To measure DLK2 expression while avoiding laborious RNA extraction and quantitative PCR steps, we fused the DLK2 promoter (defined as 3566 bp upstream of the transcriptional start site, plus 31 bp of 5′UTR) to the firefly luciferase (LUC) gene to make the reporter DLK2:LUC (Figure 1A). We chose such a long intergenic region to avoid excluding potential upstream regulatory elements. We reasoned that, in a wild type background, KAR would activate LUC gene expression and induce luciferase activity, but that this response would be absent in kai2 mutants. Accordingly we generated transgenic Arabidopsis expressing the reporter construct in Ler ecotype, and then crossed a suitably responding transgenic line with kai2-2. We first tested induction of LUC gene expression upon KAR treatments using quantitative RT-PCR, in comparison with the endogenous DLK2 gene in Arabidopsis seeds. In imbibed seed, DLK2 transcripts increase in response to KAR via KAI2, while signaling via D14 is low or absent (Waters et al., 2012). We found that both DLK2 and LUC transcripts were induced by KAR₂ treatments (Figure 1B). While levels of LUC transcripts were lower than DLK2 transcripts (relative to CACS reference transcripts), the patterns of induction in response to KAR₂ were very similar.

We then adopted a luciferase assay system in a 96-well plate format to increase throughput. We generated an enzymatic activity standard curve for the assay system using recombinant firefly luciferase standards (Figure 1C). In our hands, the system detection limit is 10^-13 g of luciferase, with a linear response between 10^-13 g to 10^-9 g of luciferase enzyme, which compares favorably with published data (Dyer et al., 2000).

We used a cell-free luciferase detection method to avoid blocking of luminescence signals by seed tissues (Figure 1D). Compounds for treatment were dissolved in water and applied to the DLK2:LUC reporter seeds. After imbibition, cell lysate was prepared from the seeds, and luminescence was measured by an automated plate reader. The luminescence signal produced by a particular treatment was expressed relative to the signal produced by seeds treated with water alone (mock treatment).

To demonstrate the sensitivity of the DLK2:LUC assay, we applied a concentration gradient of the two karrikins KAR₁ and KAR₂ over a time-course of 72 h to the DLK2:LUC [Ler] reporter seeds. Using this method, KAR response can be detected after 24 h (Figure 2A). The system is substantially more sensitive to KAR₂ treatment since 10 nM KAR₂ induced a fourfold change in activity within 24 h, while 1 μM KAR₁ was necessary for a similar response. This preference for KAR₂ is consistent with multiple responses in Arabidopsis (Nelson et al., 2011; Waters et al., 2012, 2015a), as well as for endogenous DLK2 transcripts themselves (Figure 2B). To demonstrate the specificity of the DLK2:LUC assay system in differentiating responses to KAR and SL, we applied KAR₁, KAR₂, and two enantiomers of the synthetic SL analog GR24 to DLK2:LUC [Ler] and DLK2:LUC [kai2] seeds for 72 h. GR24^5DS is a synthetic SL with stereochemistry consistent with natural SLs that act preferentially through D14, whereas its non-naturally configured enantiomer GR24^ent-5DS operates largely via KAI2 (Scaffidi et al., 2014; Waters et al., 2015a). As expected, DLK2:LUC [Ler] seeds responded strongly to KAR₁, KAR₂ and GR24^ent-5DS, but only marginally to GR24^5DS (Figure 2C). Importantly, the responses to KAR₁, KAR₂, and GR24^ent-5DS were eliminated in the DLK2:LUC [kai2] seeds. These data demonstrate that induction of LUC activity is ligand-specific and KAI2-dependent.

To investigate further the specificity of the DLK2:LUC assay system, we applied a variety of plant hormones [epibrassinosteroid (epi-BR); gibberellic acid (GA₄); auxin (IAA); abscisic acid (ABA); jasmonic acid (JA)] and the germination stimulant KNO₃ over a period of 72 h. We also included the two enantiomers of GR24. We found that while 100 nM KAR₂ and 10 μM GR24^ent-5DS induced a 10-fold increase in luciferase activity after 48 h, the other tested compounds were essentially inactive (Figure 2D). Therefore DLK2:LUC activity is not affected by these known plant hormones or KNO₃. There was a limited response to GR24^5DS after 72 h, which may indicate increasing expression of the SL receptor D14 after prolonged seed imbibition.

Having validated the sensitivity and specificity of the DLK2:LUC assay, we used it to test whether compounds extracted from Arabidopsis tissue could activate the KAI2-MAX2-SMAX1 signaling pathway. We reasoned that leaf material would be a good source of KL because this was a simple way to gather relatively large amounts of tissue. In addition, the defective leaf development phenotype of kai2 mutants indicated that KAI2-dependent signaling is active beyond the seed and seedling stages where large amounts of material would be more difficult to obtain.

We extracted metabolites from Arabidopsis rosettes with 80% (v/v) methanol in water. After removing the methanol by evaporation, we added ethyl acetate and partitioned the crude extract into two layers: aqueous (water) and organic (ethyl acetate). The partitioning method was originally optimized to extract KAR₁ from aqueous solutions. We applied a dilution series of each extract to the DLK2:LUC [Ler] and DLK2:LUC [kai2] reporter seeds. We found that unknown metabolites in the aqueous layer increased DLK2 expression fourfold within 48 h in a KAI2-dependent manner, whereas metabolites in the organic layer were inactive in this assay (Figure 3A). As such, we infer that compounds present in the Arabidopsis leaf extract can activate the KAI2-MAX2-SMAX1 signaling pathway. To our surprise, the active compound(s) appeared to be water-soluble and inefficiently extracted into ethyl acetate. This is in marked contrast to KAR₁ and SLs, which are usually extracted into organic solvents such as ether or ethyl acetate (Yasuda et al., 2003; Flematti et al., 2004).

We considered that the organic layer might in fact be active, but also might contain inhibitors that prevent the activation of DLK2:LUC. To detect such inhibitors, we compared the relative activity of KAR₂ dissolved in water versus that of KAR₂ dissolved in the organic layer (Figure 3B). We found that KAR₂ dissolved in the organic layer induced DLK2:LUC to approximately half the level of KAR₂ dissolved in water, suggesting that inhibitors were likely present but were insufficient to completely suppress DLK2:LUC activation, even at low KAR₂ concentration. Because we only observed activity in the aqueous fraction (Figure 3A), it is unlikely that there was appreciable activity in the organic layer that was suppressed by inhibitors. We cannot prove conclusively the absence of any activity in the organic layer because any extraction or recovery process might exclude some compounds. However, on the basis of these results, we conclude that the active compound(s) is primarily water-soluble.