Introduction

Strigolactones (SLs) are a class of phytohormones that also act as rhizosphere signals for arbuscular mycorrhizal fungi and root parasitic plants (Al-Babili and Bouwmeester 2015). To date, approximately 30 SLs have been characterized from various plant species (Yoneyama et al. 2018b). Natural SLs consist of an (R)-configurated butenolide ring (D-ring) linked by an enol ether bridge to a less conserved second moiety (Jia et al. 2018; Wang and Bouwmeester 2018). SLs are classified into two groups, canonical and non-canonical SLs, based on the structure of the second moieties. Canonical SLs possess a tricyclic lactone ring (ABC ring), while non-canonical SLs lack the A, B, or C ring. Canonical SLs are further divided into strigol- and orobanchol-type SLs, which have an α- and a β-oriented C ring, respectively (Fig. 1a).

Fig. 1
figure 1

Lotus SLs production of WT and the candidate gene mutants were analyzed by LC–MS/MS. a Structures of L. japonicus SLs. 5-Deoxystrigol (5DS) is a canonical SL possessing a tricyclic lactone ring (ABC ring). Lotuslactone (LL) is a non-canonical SL lacking C ring. b 5DS and LL in root exudates of WT and line 30130539 (dsd) were analyzed. c 5DS and LL in root exudates of WT and line 30092673 (lld) were analyzed. MRM chromatograms of 5DS (the precursor ion [M + H]+, red: 331.15/216.00, blue: 331.15/97.00, green: 331.15/234.00, m/z in positive mode) and LL (the precursor ion [M-AcO + H]+, red: 373.00/276.00, blue: 373.00/244.00, green: 373.00/216.00, m/z in positive mode) are shown

Full size image

SL biosynthesis is initiated from the conversion of β-carotene to carlactone by the sequential action of DWARF27, CAROTENOID CLEAVAGE DIOXYGENASE 7 and CAROTENOID CLEAVAGE DIOXYGENASE 8 (Alder et al. 2012; Seto et al. 2014). In Arabidopsis, carlactone is converted by the cytochrome P450 (CYP) enzyme MORE AXILLARY GROWTH 1 (MAX1) (CYP711A) into carlactonoic acid (CLA) that is methylated to methyl carlactonoate (Abe et al. 2014). Methyl carlactonoate is oxygenated by the 2-oxoglutarate-dependent dioxygenase (2OGD) enzyme LATERAL BRANCHING OXIDOREDUCTASE (LBO), to an unidentified SL-like compound (Brewer et al. 2016). In rice, one of MAX1 homologs, Os900, converts carlactone to an orobanchol-type SL, 4-deoxyorobanchol, and another MAX1 homolog, Os1400, hydroxylates 4-deoxyorobanchol to orobanchol (Zhang et al. 2014). The conversion of carlactone to CLA is a common reaction catalyzed by MAX1 homologs as found recently (Yoneyama et al. 2018a). In sorghum, a sulfotransferase, LOW GERMINATION STIMULANT 1, is likely to be involved in a strigol-type SL, 5-deoxystrigol (5DS) production (Gobena et al. 2017).

The model legume Lotus japonicus produces canonical 5DS that was isolated as the first hyphal branching factor for arbuscular mycorrhizal fungi (Fig. 1a) (Akiyama et al. 2005). Non-canonical lotuslactone (LL) is the second branching factor identified in L. japonicus (Fig. 1a) (Xie et al. 2019). Despite that 5DS is considerably more active than LL in hyphal branching (Akiyama et al. 2010), L. japonicus roots exude both SLs (Xie et al. 2019). LL lacks the C-ring and has a seven-membered cycloheptadiene A-ring as in medicaol (Tokunaga et al. 2015). Its B-ring is highly oxygenated with hydroxy, acetyloxy, and oxo groups. The large structural difference between 5DS and LL implies that these two SLs are produced by branched biosynthetic pathways. Recently, we showed that LjMAX1 (CYP711A9) catalyzes the oxidation of carlactone to 18-hydroxyCLA via CLA, and that 18-hydroxylated carlactonoates are common biosynthetic precursors for 5DS and LL in L. japonicus (Mori et al. 2019). In addition, the ljmax1 mutant was deficient in both 5DS and LL. These findings indicate that the branching point in the biosynthesis of 5DS and LL is positioned downstream of LjMAX1. However, nothing is known about the pathways downstream of MAX1 and the branching point to produce 5DS and LL in L. japonicus.

In this study, we attempted to discover novel genes encoding enzymes involved in the individual biosynthetic pathways of 5DS and LL in L. japonicus. The pathway from 18-hydroxyCLA to 5DS is expected to include oxidative cyclization step(s) in the BC ring formation as observed for Os900. The same is likely true for the B ring formation in LL biosynthesis. In addition, the construction of LL requires ring expansion in the A-ring and substitution of oxygenated functionalities in the B-ring. We, therefore, focused on genes encoding CYP and 2OGD enzymes as candidates involved in such processes. The results we report here are the identification of a CYP and a 2OGD genes responsible, respectively, for the biosyntheses of Lotus SLs by transcriptome profiling and mutant screening, and their transcriptional responses to phosphorus availability and a phytohormone gibberellin.

Materials and methods

Plant growth and treatment

Lotus japonicus L. ecotype Gifu B-129 provided by the National BioResource Project (L. japonicus and Glycine max; https://www.legumebase.brc.miyazaki-u.ac.jp/index.jsp) was used as the WT. LORE1 mutants were obtained from the LORE1 insertion mutant resource (Fukai et al. 2012; Urbański et al. 2012). Genotyping of LORE1 mutants was performed using primers listed in Supplementary Table S1. Seedlings were grown hydroponically as described previously (Mori et al. 2019). Phosphate and gibberellin treatments and CLA feeding were performed using shaded centrifuge tubes (50 ml) containing 40 ml hydroponic solutions.

Liquid chromatography-tandem mass spectrometry (LC–MS/MS) analysis

Measurements of SL levels in root exudates were performed with UHPLC (Nexera X2, Shimadzu) and an ion trap instrument (amaZon SL, Bruker) using authentic standards as described previously (Yoneyama et al. 2018a; Mori et al. 2019).

RNA-sequencing (RNA-seq)

Sequencing library construction and RNA-seq on an Illumina HiSeq 2000 sequencer (paired-end, 100 bp) were performed by Macrogen Japan (Kyoto, Japan) using total RNA extracted from roots using an RNeasy Plant Mini Kit (Qiagen).

Cloning

Total RNA was extracted from roots of the WT using an RNeasy Plant Mini Kit (Qiagen). The first strand cDNA was synthesized using SMARTScribe Reverse Transcriptase included in a SMARTer rapid amplification of cDNA ends (RACE) 5′/3′ Kit (Clontech). 5′- and 3′-RACE to obtain partial sequences of DSD were performed using the kit with 5′-UTR-GSP2 and 3′-UTR-GSP1 primers (Supplementary Table S2). The full-length coding sequences were amplified from the first strand cDNA by nested-polymerase chain reaction (PCR) using PrimeSTAR Max DNA polymerase (Takara) with the first primers 5′-UTR-DSD/LLD-F and 3′-UTR-DSD/LLD-R and the second primers Nested-5′-UTR-DSD/LLD-F and Nested-3′-UTR-DSD/LLD-R. The full-length cDNAs were subcloned into the cloning vector pTAC1 (DynaExpress) for sequencing or the expression vector pUB-GW-GFP for Agrobacterium-mediated transformation (Maekawa et al. 2008) by an In-Fusion HD Cloning Kit (Takara) using INF-DSD/LLD-F and INF-DSD/LLD-R primers. Sequencing was performed by Eurofins Genomics (Tokyo, Japan) using the Sanger method. The phylogenetic tree was constructed using the MEGA7 program (Kumar et al. 2016). The homologous sequences of DSD and LLD in representative plant species were obtained from Phytozome v12.1 (https://phytozome.jgi.doe.gov/pz/portal.html).

Agrobacterium rhizogenes-mediated transformation

LORE1 mutants were transformed by A. rhizogenes strain AR1193 harboring pUB-GW-DSD-GFP, pUB-GW-LLD-GFP, or pUB-GW-GFP according to the method of Okamoto et al. (2013). Composite plants consisting of transformed hairy roots expressing GFP were used for further analysis.

Quantitative reverse-transcriptase PCR (qRT-PCR) analysis

Total RNA extracted from roots as above was treated by TURBO DNA-free Kit (Ambion), and reverse transcribed using a High-Capacity RNA-to-cDNA Kit (Applied Biosystems). qRT-PCR was performed on an ABI PRISM 7300 sequence detection system using a Power SYBR Green PCR Master Mix (Applied Biosystems) with primers listed in Supplementary Table S3. Ubiquitin4 expression was used as an internal standard.

Statistical analysis

Student’s t-test was applied to evaluate the significance of the results obtained.

Accession number

The accession numbers of the DSD and LLD sequences are LC508723 and LC508724, respectively.

Results and discussion

In general, SL production is promoted by phosphate (Pi) starvation (Umehara et al. 2010; Yamada et al. 2014). By contrast, the supply of Pi negatively regulates SL levels in roots and root exudates with a concomitant downregulation of multiple SL biosynthetic genes (Umehara et al. 2010; Liu et al. 2015; Zhang et al. 2018). To check the Pi-dependent change in SL levels in L. japonicus, we quantified SL levels in root exudates of seedlings continuously starved or supplied with Pi by LC–MS/MS analysis (Supplementary Fig. S1a). The levels of 5DS and LL in Pi-supplied seedlings were significantly decreased to 7.5% and 4.7% of continuously Pi-starved control, respectively (Supplementary Fig. S1b). To screen candidate oxygenase genes involved in the biosyntheses of Lotus SLs, we performed transcriptome profiling by RNA-seq on roots of Pi-starved or supplied seedlings grown under the same conditions as above (Supplementary Table S4). Among 655 genes annotated as CYP (473 genes) and 2OGD (182 genes) in the L. japonicus genome database, 49% of genes (CYP 224 genes, 2OGD 99 genes) were downregulated in Pi supplied roots compared to Pi starved roots. These downregulated oxygenase genes were ranked in descending order based on the degree of expression change. To identify genes responsible for the biosyntheses of 5DS and/or LL, we next performed a screen for SL-defect mutants from the LORE1 insertion mutant resource (Fukai et al. 2012; Urbański et al. 2012). We retrieved LORE1 mutant lines containing exonic insertions in the higher ranked genes through Lotus Base (https://lotus.au.dk). The homozygous LORE1 mutants were grown hydroponically without Pi, and analyzed for SL production in root exudates. Through a screen of the top 29 candidates, we found that two mutant lines 30130539 and 30092673 were defect, respectively, in 5DS and LL production. Both SLs were detected in root exudates of the WT. Only LL was detected in line 30130539 homozygous for a putative CYP gene Lj0g3v0346399 (Fig. 1b). We named this gene DSD after 5DS-defective phenotype. Only 5DS was detected in line 30092673 homozygous for a putative 2OGD gene Lj1g3v4997280 (Fig. 1c). We named this gene LLD after LL-defective phenotype.

The coding sequence (CDS) of DSD and LLD genes were cloned from cDNA of the WT ecotype Gifu B-129. Based on the partial coding sequence available, the missing cDNA sequences were obtained by 5′- and 3′-RACE. Using protein sequences encoding these DNA fragments as query, OrthoDB (https://www.orthodb.org/) searches were performed against Medicago truncatula and Glycine max. By comparing the gene sequences of orthologs in M. truncatula and G. max, we predicted the full-length mRNA sequences of DSD and LLD, and designed 5′-untranslated region (UTR) and 3′-UTR primers. The full-length CDSs of the two genes were obtained by PCR using the primer sets. The CDSs of DSD and LLD were 1455 bp encoding for a 484-amino acid protein, and 1071 bp encoding for a 356-amino acid protein, respectively (Supplementary Table S5). Phylogenetic analysis revealed that DSD belongs to CYP722C subfamily (Nelson et al. 2008) (Supplementary Fig. S2). Other legumes have both CYP722A and CYP722C, while Arabidopsis and gramineous plants have only CYP722A and CYP722B, respectively. LLD falls within the DOXC55 clade of 2OGDs as shown by a phylogenetic tree (Kawai et al. 2014) (Supplementary Fig. S3). The next clade DOXC54 includes LBO.

In the dsd mutant, the LORE1 retroelement was inserted in the first exon, 66 bp after the start codon of DSD in contig CM0131 on chromosome 4 (in build 2.5) (Fig. 2a). In the lld mutant, the LORE1 retroelement was inserted in the second exon, 357 bp after the start codon of LLD (Gene ID chr4.LjT16L01.240.r2.d in build 2.5) in contig LjT16L01 on chromosome 4 (Fig. 2b). To directly confirm the involvement of DSD and LLD in 5DS or LL-defective phenotype of the mutants, we used A. rhizogenes-mediated transformation to generate transgenic roots expressing the WT cDNA sequence. The full-length CDSs were cloned behind the LjUBIQUITIN promoter in the vector pUB-GW-GFP. The constructs were introduced into roots of the mutants by A. rhizogenes harboring the vectors. 5DS was not detected in root exudates of dsd transformed with empty vector, while 5DS production was restored in transformed dsd harboring the DSD CDS (Fig. 3a). Likewise, LL was not detected in root exudates of lld transformed with empty vector, while LL production was restored in those transformed with pUB-GW-LLD-GFP (Fig. 3b). Taken together, we demonstrated that the exonic mutations in DSD and LLD genes caused SL deficiency in the mutants.

Fig. 2
figure 2

Location of the LORE1 insertion in DSD and LLD genes. aDSD locates in contig CM0131 on chromosome 4 and has eight exons. The dsd mutant has the LORE1 retroelement in the first exon. bLLD locates in contig LjT16L01 on chromosome 4 and has four exons. The lld mutant has the LORE1 retroelement in the second exon

Full size image
Fig. 3
figure 3

Restoration of Lotus SLs production in the LORE1 insertion mutants expressing the WT cDNA sequence by A. rhizogenes-mediated hairy root transformation. Root exudates and authentic standards were analyzed by LC–MS/MS. adsd was transformed by pUB-GW-DSD-GFP (dsd-DSD) and pUB-GW-GFP (empty vector, dsd-EV). blld was transformed by pUB-GW-LLD-GFP (lld-LLD) and pUB-GW-GFP (empty vector, lld-EV). MRM chromatograms (left) and full-scan spectra of fragment ions (right). MRM chromatograms of 5DS (the precursor ion [M + H]+, red: 331.15/216.00, blue: 331.15/97.00, green: 331.15/234.00, m/z in positive mode) and LL (the precursor ion [M-AcO + H]+, red: 373.00/276.00, blue: 373.00/244.00, green: 373.00/216.00, m/z in positive mode) are shown

Full size image

To check the Pi-dependent change in transcript levels of DSD and LLD as well as other known SL biosynthetic genes, we analyzed their transcript levels by qRT-PCR in the WT roots grown under the same condition in the RNA-seq analysis as above. Our analysis revealed that the supply of Pi reduced transcript levels of the SL biosynthetic genes, LjD27, LjCCD7, LjCCD8, LjMAX1, DSD and LLD, but did not affect that of the SL signaling gene D14 (Supplementary Fig. S4). It was recently discovered that gibberellin signaling negatively regulates SL biosynthesis in rice and Arabidopsis (Ito et al. 2017; Lantzouni et al. 2017). We investigated whether the same mechanism is involved in the regulation of SL biosynthesis in L. japonicus using the WT seedlings grown with or without active gibberellic acid (GA3) in the hydroponic solutions. GA3 treatment reduced not only transcript levels of LjCCD7, LjCCD8, LjMAX1, DSD and LLD in roots but also 5DS and LL levels in root exudates (Supplementary Fig. S5). These results showed that SL biosynthesis is negatively regulated by phosphate and gibberellin in L. japonicus.

We performed feeding experiments with exogenous CLA using dsd and analyzed labeled products in root exudates by LC–MS/MS. We could not detect the peak of labeled 5DS, but could detect 18-hydroxyCLA, indicating that DSD is an enzyme in the later stage of 5DS biosynthesis downstream of LjMAX1 and its enzymatic function was not C-18 hydroxylation of CLA (Supplementary Fig. S6).

In conclusion, we identified novel oxygenase genes DSD and LLD involved, respectively, in the biosyntheses of canonical 5DS and non-canonical LL in L. japonicus. Further characterization of DSD and LLD enzymes, including identification of substrates and products, will advance our understanding of SL biosynthetic pathway. The availability of the dsd and lld mutants will provide the opportunity to assess the functional differences between canonical and non-canonical SLs in rhizosphere signaling. These studies will provide insights into how and why plants produce the two classes of SLs.

Author Contribution Statement

NM and KA conceived and designed research. NM conducted experiments. NM, TN and KA analyzed data. NM and KA wrote the manuscript. All authors read and approved the manuscript.