Functional analysis reveals the possible role of the C-terminal sequences and PI motif in the function of lily (Lilium longiflorum) PISTILLATA (PI) orthologues.

Two lily (Lilium longiflorum) PISTILLATA (PI) genes, Lily MADS Box Gene 8 and 9 (LMADS8/9), were characterized. LMADS9 lacked 29 C-terminal amino acids including the PI motif that was present in LMADS8. Both LMADS8/9 mRNAs were prevalent in the first and second whorl tepals during all stages of development and were expressed in the stamen only in young flower buds. LMADS8/9 could both form homodimers, but the ability of LMADS8 homodimers to bind to CArG1 was relatively stronger than that of LMADS9 homodimers. 35S:LMADS8 completely, and 35S:LMADS9 only partially, rescued the second whorl petal formation and partially converted the first whorl sepal into a petal-like structure in Arabidopsis pi-1 mutants. Ectopic expression of LMADS8-C (with deletion of the 29 amino acids of the C-terminal sequence) or LMADS8-PI (with only the PI motif deleted) only partially rescued petal formation in pi mutants, which was similar to what was observed in 35S:LMADS9/pi plants. In contrast, 35:LMADS9+L8C (with the addition of the 29 amino acids of the LMADS8 C-terminal sequence) or 35S:LMADS9+L8PI (with the addition of the LMADS8 PI motif) demonstrated an increased ability to rescue petal formation in pi mutants, which was similar to what was observed in 35S:LMADS8/pi plants. Furthermore, ectopic expression of LMADS8-M (with the MADS domain truncated) generated more severe dominant negative phenotypes than those seen in 35S:LMADS9-M flowers. These results revealed that the 29 amino acids including the PI motif in the C-terminal region of the lily PI orthologue are valuable for its function in regulating perianth organ formation.

Introduction

The ABCDE model predicts the formation of any flower organ through the interaction of five classes of homeotic genes in plants (Coen and Meyerowitz, 1991; Theissen, 2001; Litt and Kramer, 2010). MADS box genes have a central role in flower development because most ABCDE genes encode MADS box proteins (Coen and Meyerowitz, 1991; Purugganan ; Rounsley ; Theissen ; Theissen and Saedler, 2001; Litt and Kramer, 2010).

B group genes, such as APETALA3 (AP3) and PISTILLATA (PI), have a major role in specifying petal and stamen development (Jack , 1994; Goto and Meyerowitz, 1994; Krizek and Meyerowitz, 1996; Kramer ; Winter ; Hernandez-Hernandez ; Kanno ; Whipple ; Irish, 2009). In Arabidopsis, mutations in AP3 or PI result in identical phenotypes, in which the second whorl petal is converted into a sepal structure and the third flower whorl stamen is converted into a carpel structure (Bowman ; Jack ; Goto and Meyerowitz, 1994). Similar conversions for these petals and stamens are observed in the mutants of the AP3 and PI orthologues from a number of core eudicot species (Sommer ; Tröbner ; Angenent ; van der Krol ; Yu ; Liu ; Vandenbussche ; de Martino ), basal eudicot species (Drea ; Kramer ), and monocot species (Ambrose ; Nagasawa ; Prasad and Vijayraghavan, 2003; Yadav ; Yao ), indicating that the function of the B class genes AP3 and PI is highly conserved during evolution.

Based on analyses of sequence diversity, B group genes have been proposed to have arisen from an ancestral gene through multiple gene duplication events (Doyle, 1994; Theissen , 2000; Purugganan, 1997; Kramer ; Lamb and Irish, 2003; Kim ; Stellari ; Hernandez-Hernandez ). The first duplication is thought to have generated the PI and paleoAP3 lineages. The PI lineage is composed of PI orthologues that contain a highly conserved PI motif identified in most plant species (Kramer ). The paleoAP3 lineage is composed of AP3 orthologues identified in lower eudicots, magnolid dicots, and monocots (Kramer ). The second duplication from the paleoAP3 lineage is thought to have generated the euAP3 and TM6 gene lineages, which have been subject to substantial sequence changes in dicots during evolution (Kramer ; Kramer and Irish, 1999).

Proteins in the euAP3 lineage and in most paleoAP3 lineages are not able to form homodimers. They have to interact with PI to form heterodimers to regulate petal and stamen development (Schwarz-Sommer ; Tröbner ; Riechmann ; Moon ; Vandenbussche ; Yao ). In contrast, the ability of B class proteins to form homodimers has been reported in the gymnosperms as well as in the paleoAP3 and PI lineages of some non-grass monocots (Winter ; Tzeng and Yang, 2001; Hsu and Yang, 2002; Winter ; Kanno ; Tsai , 2005; Chang ). It has been shown that both paleoAP3 and the PI-derived motifs for the paleoAP3 orthologue LMADS1 in lily are essential for the formation of homodimers because the deletion of these two motifs completely abolishes the homodimerization of LMADS1 (Tzeng ). Whether the PI motif identified in PI orthologues is also involved in their homodimerization and important for their function still remains unclear and thus requires additional experimental analysis.

Among monocots, research on the B class genes is mostly limited to maize (Ambrose ), rice (Nagasawa ; Prasad and Vijayraghavan, 2003; Yadav ; Yao ) and the non-grass monocots tulip (Kanno , 2007) and orchid (Hsu and Yang, 2002; Tsai , 2005; Xu ; Guo ; Kim ; Mondragón-Palomino and Theissen, 2008, 2009; Mondragón-Palomino ; Chang ), but the functions of these genes still remain under investigation. Due to the economic importance of the monocot lily (Lilium longiflorum) in the cut flower market, the A, C, D, and E classes as well as one paleoAP3 orthologue (LMADS1) in the B class of MADS box genes have been analysed for function in this flower (Tzeng and Yang, 2001; Tzeng , 2003, 2004; Benedito ; Chen ; Hsu ). In contrast to most plant species, two PI-like genes have been reported in the lily (L. regale and L. formolongi) genome (Winter ; Akita ). These two genes may have redundant functions. However, no further functional analyses of these two genes have been reported. In the present study, the isolation and functional analysis of two putative PI-like MADS box genes from lily (L. longiflorum) are reported. An exploration of the possible function of these two lily PI orthologues in the regulation of perianth formation, as well as the role of the conserved PI motif in homodimerization and function, is conducted, and the results are discussed.

Materials and methods

Plant materials and growth conditions

Plants of lily (L. longiflorum Thunb. cv. Snow Queen) used in this study were grown in the field in Tein Wei County, Chang Haw, Taiwan. The Arabidopsis pi-1 mutant line (CS77) in the Landsberg background was obtained from the Arabidopsis Biological Resource Center, Ohio State University, Columbus, OH, USA. Seeds for Arabidopsis were germinated and grown as described previously (Chang ).

Cloning of LMADS8 and LMADS9 cDNAs from L. longiflorum

PCR products containing partial sequences that showed high sequence identity to PI MADS box genes were identified as described by Tzeng and Yang (2001). Internal gene-specific primers were designed for 3′- and 5′-rapid amplification of cDNA ends (3′- and 5′-RACE) by using the SMART™ RACE cDNA Amplification kit (BD Biosciences Clontech, Palo Alto, CA, USA). The gene-specific primer for 3′-RACE was 5′-AGCGCGGAGATTGATAGGATCAAGAAGG-3′; and that for 5′-RACE was 5′-CCTCAGAAAATCATTCTGCTTCTCCCGGAC-3′. The RACE identified sequences for two genes, one for LMADS8 (GenBank accession no. HQ698550) and the other one that contained a sequence distinct from that of LMADS8 was named LMADS9 (GenBank accession no. HQ698551).

The cDNA for LMADS8 were obtained by PCR amplification using the forward primer 5′-GGGGATCCCCATGGGTCGTGGCAAGATC-3′ and reverse primer 5′-CCGGATCCCTACTTGTCCTCATGTAAATTAGGCTGG-3′. The cDNA for LMADS9 was obtained by PCR amplification using the forward primer 5′-GGGGATCCCCATGGGTCGTGGCAAGATC-3′ and reverse primer 5′-CCGGATCCCTATATGCAAGTTCCATGTCCCTCATA-3′. The specific forward and reverse primers for LMADS8 and 9 contained the generated BamHI (5′-GGATCC-3′, underlined) recognition site to facilitate the cloning of the cDNA.

The cDNA constructs containing truncated MADS boxes (LMADS-M) were obtained by PCR amplification and site-specific mutagenesis. The following 5′ primers with an ATG start codon (in bold) right after the 5′ MADS box domain were used for PCR amplification: 5′-TCTAGAATGTATCAGGTGAACTGCGGC-3′ for LMADS8-M, and 5′-GGGGATCCATGTACCAGTTGAACTCCGGCAAGAAG-3′ for LMADS9-M. The individual 3′ primers used for each construct were 5′-CCGGATCCCTACTTGTCCTCATGTAAATTAGGCTGG-3′ for LMADS8-M, and 5′-CCGGATCCCTATATGCAAGTTCCATGTCCCTCATA-3′ for LMADS9-M. The specific 5′ primers contained the generated XbaI (5′-TCTAGA-3′, underlined) or BamHI (5′-GGATCC-3′, underlined) recognition site to facilitate the cloning of the cDNA.

Construction of LMADS8 with deletion and LMAD9 with addition sequences from L. longiflorum

The cDNA constructs containing the truncated C-terminal or PI motif of LMADS8 (LMADS8-C or LMADS8-PI) were obtained by PCR amplification and site-specific mutagenesis using the forward primer 5′-GGTACCATGGGTCGTGGCAAGATCGAGATC-3′ and two reverse primers with TAG stop codons (in bold) 5′-GGTACCCTATGCAAATTCCAAGTTCCTCAT-3′ and 5′-GGTACTACTGGGAACCGAAATCCCCATCT-3′ for LMADS8-C and LMADS8-PI, respectively. The specific forward and reverse primers for LMADS8-C and LMADS8-PI contained the generated KpnI (5′-GGTACC-3′, underlined) recognition site to facilitate the cloning of the truncated cDNA.

The cDNA constructs of LMADS9 containing an additional C-terminal or PI motif from LMADS8 (LMADS9+L8C or LMADS9+L8PI) were obtained by primer extension techniques for PCR amplification. For LMADS9+L8C construction, the forward primer 5′-GGATCCCCATGGGTCGTGGCAAGATCGA-3′ and five reverse primers 5′-CCATCTTTATGATGATATATGCAAGTTCCATGTCCCTCATATTCTCC-3′, 5′-CTGGGAACCGAAATCCCCATCTTTATGATGATATATGCAAGTTCCATGT-3′, 5′-ACACGAAAGGCCATTGGCATCTGGGAACCGAAATCCCCATCTTTATG-3′, 5′-AATTAGGCTGGATTGGCTGCACACGAAAGGCCATTGGCATCTGG-3′, and 5′-GGATCCTACTTGTCCTCATGTAAATTAGGCTGGATTGGCTGCACACGA-3′ were used sequentially. For LMADS9+L8PI construction, the forward primer 5′-GGATCCCCATGGGTCGTGGCAAGATCGA-3′ and three reverse primers 5′-ACACGAAAGGCCATTGGCATTATGCAAGTTCCATGTCCCTCATATTCTCC-3′, 5′-AATTAGGCTGGATTGGCTGCACACGAAAGGCCATTGGCATTATGCA-3′, and 5′-GGATCCTACTTGTCCTCATGTAAATTAGGCTGGATTGGCTGCACACGA-3′ were used sequentially. The specific forward and reverse primers for LMADS9+L8C and LMADS9+L8PI contained the generated BamHI (5′-GGATCC-3′, underlined) recognition site to facilitate the cloning of the cDNA.

Quantitative real-time PCR

For quantitative real-time PCR, the reaction was performed on an MJ Opticon system (MJ Research, Waltham, MA, USA) using SYBER Green Real-time PCR Master Mix (TOYOBO Co., Ltd). The amplification condition was 95 °C for 10 min, followed by 40 cycles of amplification (95 °C for 15 s, 58 °C for 15 s, 72 °C for 30 s, and then plate reading) and melting (50–95 °C with plate readings every 1 °C). LMADS8 primers 5'-TCCTGTGCCTGCTATTTTATTTAGAC-3′ and 5′-TTATACTATCAAATTACTACATGAACATCAAGAG-3′ were used in lily, and 5′-TCGGTTCCCAGATGCCAATG-3′ and 5′-GTTGCCCGCATAATTACGAATATC-3′ were used in LMADS8 transgenic Arabidopsis. LMADS9 primers 5′-CATGCTTATCGTTACCACTCTACTTG-3′ and 5′-ACTATACTAATAATTACTACATGAACATCTATGC-3′ were used in lily, and 5′-CGAGGAGGCACTAGAGAATGG-3′ and 5′-CGTAACATAAGGGACTGACCAC-3′ were used in LMADS9 transgenic Arabidopsis. The L. longiflorum ACTIN gene was used as a normalization control with the primers Liact-F (5′-ATCCCAGCAGCGTCGCACATCC-3′) and Liact-R (5′-GCCAGATCTTCTCCATGTCATCC-3′) (Wang ). The Arabidopsis housekeeping gene UBQ10 was used as normalization control with the following primers: RT-UBQ10-1 (5′-CTCAGGCTCCGTGGTGGTATG-3′) and RT-UBQ10-2 (5′-GTGATAGTTTTCCCAGTCAACGTC-3′). The data were analysed using the Gene Expression Macro software (version 1.1, Bio-Rad).

Plant transformation and transgenic plant analysis

The full-length cDNAs of LMADS8 and 9 or LMADS8/9-M were cloned into the binary vector pBI121 (Clontech, Palo Alto, CA, USA) under the control of the Cauliflower mosaic virus (CaMV) 35S promoter. These constructs were transformed into Arabidopsis plants using the floral dip method as described elsewhere (Clough and Bent, 1998). Transformants were selected in medium containing 50 μg ml−1 kanamycin and were further verified by PCR and RT-PCR analyses.

Yeast two-hybrid analysis

The cDNA truncations without the MADS box regions for LMADS1, 8, and 9 were amplified by PCR. The primers for LMADS1 were 5′-TTCTGCGAATTCTCCACAGAC-3′ and 5′-CCTACAGAATTCGGGTTTCA-3′; for LMADS8 they were 5′-CATATGTATCAGGTGAACTGCGGCAAGA-3′ and 5′-GAATTCCTACTTGTCCTCATGTAAATTAGGCTG-3′; and for LMADS9 they were 5′-CATATGTACCAGTTGAACTCCGGCAAGA-3′ and 5′-GAATTCCTATATGCAAGTTCCATGTCCCTC-3′. Specific 5′ and 3′ primers contained the EcoRI (5′-GAATTC-3′, underlined) or NdeI (5′-CATATG-3′, underlined) recognition sites to facilitate cloning of the cDNAs. PCR fragments were ligated into the plasmid pGBKT7 (binding domain vector) or pGADT7 (activation domain vector) provided by the MATCHMAKER Two-Hybrid System 3 (Clontech). Recombinant plasmids were transformed into yeast using the lithium acetate method (Gietz ). The transformants were selected on selection medium according to the manufacturer’s instructions. β-Galactosidase activity in the transformants was analysed as described by Tzeng and Yang (2001) and calculated according to Miller (1992).

Electrophoretic mobility shift assay (EMSA)

The cDNA for LMADS1, 8, and 9 used to generate proteins were amplified by PCR. The primers for LMADS1 were 5′-CATATGATGGGGCGGGGAAAGATC-3′ and 5′-GGATCCTCAAGCCAGTCGGAGATCA-3′; for LMADS8 they were 5′-GGAATTCCATATGATGGGTCGTGGCAAGATC-3′ and 5′-GAATTCCTACTTGTCCTCATGTAAATTAGGC-3′; and for LMADS9 they were 5′-GGAATTCCATATGATGGGTCGTGGCAAGATC-3′ and 5′-CCGGATCCCTATATGCAAGTTCCATGTCCCTCATA-3′. Specific 5′ and 3′ primers contained the NdeI (5′-CATATG-3′, underlined), BamHI (5′-GGATCC-3′, underlined), or EcoRI (5′-GAATTC-3′, underlined) recognition sites to facilitate cloning of the cDNAs. The PCR products were cloned into expression vector pET28a (Novagen, Madison, WI, USA) and transformed into expression host BL21(DE3). After induction with 0.4 mM isopropyl-β-D-thiogalactopyranoside (IPTG), the inclusion bodies of LMADS1, LMADS8, and LMADS9 were isolated, solubilized in 6 M urea, and then dialysed with NEB buffer [25 mM HEPES-KOH pH 7.6, 40 mM KCl, 0.1 mM EDTA, 10% glycerol, 5 mM β-mercaptoethanol, 0.5 mM phenylmethylsulphonyl fluoride (PMSF), and 0.5 mg ml−1 lysozyme].

DNA probes were prepared from the promoter of Arabidopsis AP3 by PCR amplification using the reverse primer 5′-GACCATATGCGACCTAAAAAGGAAAAGAGTTGC-3′ and forward primer DIG-5′-GCAACTCTTTTCCTTTTTAGGTCGCATATGGTC-3′ for the 5′ end digoxigenin (DIG)-labelled probe, or forward primer 5′-GCAACTCTTTTCCTTTTTAGGTCGCATATGGTC-3′ for the competitor.

For EMSA, DNA binding reactions were performed in a total volume of 12 μl of binding buffer [10 mM TRIS-HCl, pH 7.5, 1 mM EDTA, 90 μM dithiothreitol (DTT), 2.5 mM NaCl, 4% glycerol, 40 μg ml−1 bovine serum albumin (BSA)] containing 40 μg ml−1 poly(dI–dC), 80 μM PMSF, 50 μg of probe, and 2 μg of proteins. Binding reaction mixtures were incubated for 20 min at room temperature. To separate the protein–DNA complexes, the reaction mixtures were loaded onto a running non-denaturing 5% polyacrylamide gel, which had been pre-run for 1 h at 4 °C and 100 V in 0.5× TRIS-borate-EDTA buffer. Electrophoresis was further carried out at 4 °C and 100 V for 4 h. After electrophoresis, the gels were blotted onto a nylon membrane (Amersham Biosciences, Bucks, UK), hybridized by anti-DIG horseradish peroxidase antibody (Roche, Mannheim, Germany), developed by 0.1× CSPD (Roche, Mannheim, Germany), and exposed on X-ray film. For competition experiments, a 10-fold molar excess of the cold competitor DNAs was included in the binding reactions.

Scanning electron microscopy (SEM)

SEM was performed according to the methods of Tzeng and Yang (2001) and Hsu and Yang (2002). Samples were fixed in 2% glutaraldehyde in 25 mM sodium phosphate buffer (pH 6.8) at 4 °C overnight. After dehydration in a graded ethanol series, specimens were critical point dried in liquid CO2. The dried materials were mounted and coated with gold–palladium in a JBS sputter-coater (model 5150). Specimens were examined in a Topcon scanning electron microscope (model ABT-150S) with an accelerating voltage of 15 kV.

Phylogenetic analyses

The amino acid sequences of the B class MADS box proteins were retrieved via the NCBI server (http://www.ncbi.nlm.nih.gov/). The sequence alignment and phylogenetic tree construction were generated by MEGA 5 software (Kumar ; Tamura ). Multiple sequence alignment of these protein sequences was performed using the ClustalW program (Thompson ) with default parameters. The phylogenetic tree was constructed based on the maximum likelihood method with the following parameters: bootstrap (500 replicates), Jones–Taylor–Thornton (JTT) substitution model (Jones ), uniform rates, complete deletion of gaps/missing data, and Nearest Neighbor Interchange (NNI). The bootstrap consensus tree was inferred from 500 replicates (Felsenstein, 1985). The numbers at each node indicate the percentage of bootstrap values from 500 replications, and values <50% were not shown. The MADS box proteins used in this study included AP3 (NM_115294), ApDEF (AB177941), ApGLO (AB079259), Boi1AP3 (U67453), Boi2AP3 (U67455), CMB2 (L40405), CsAP3a (AY948339), CsAP3b (AY948340), CsPIA1 (DQ231247), CsPIB (DQ231249), CsPIC1 (DQ231250), DEF2 (X67508), DEF (X62810), EGM2 (AF029976), FBP1 (M91190), FBP3 (X71417), GDEF1 (AJ009724), GDEF2 (AJ009725), GGLO1 (AJ009726), HPI1 (AF134114), HPI2 (AF134115), LFGLOA (AB359186), LFGLOB (AB359187), LMADS1 (AF503913), LMADS8 (HQ698550), LMADS9 (HQ698551), LRDEF (AB071378), LRGLOA (AB071379), LRGLOB (AB071380), NMH7 (L41727), NTGLO (X67959), OMADS8 (HM140842), OsMADS2 (NM_001051547), OsMADS4 (NM_001062125), OsMADS16 (NM_001065095), PeMADS6 (AY678299), PhaPI9 (AY748818), PhaPI10 (AY771991), PI (NM_122031), PMADS1 (X69946), PMADS2 (X69947), SLM3 (X80490), TGDEFA (AB094965), TGDEFB (AB094966), TGGLO (AB094967), TM6 (DQ539417), ZMM16 (AJ292959), ZMM18 (AJ292960), and ZMM29 (AJ292961).

Results

Isolation of LMADS8 and LMADS9 cDNAs from lily (L. longiflorum)

To isolate the MADS box genes from L. longiflorum, a strategy that combined RT-PCR and 5′- and 3′-RACE was used. A DNA fragment was amplified by RT-PCR using total RNA from young floral buds as a template. Sequence comparison led to the identification of partial sequences for several MADS box genes. The full-length cDNA sequences for two PI-like B class genes, LMADS8 and 9, were isolated.

The LMADS8 cDNA encoded a 210 amino acid protein that showed high sequence identity to the product of the B class gene PI in monocots (Fig. 1A). LMADS8 showed the highest sequence identity to LFGLOA of L. formolongi (Akita ) and LRGLOA of L. regale (Winter ) (100% and 99%, respectively) (Fig. 1A). In the MADS box domain, 100% of the amino acids were identical between LMADS8 and LFGLOA/LRGLOA (Fig. 1A). LMADS8 also showed 51% identity and 71% similarity to Arabidopsis PI (Fig. 1A). A sequence (KHEHL) that was identical to a highly conserved sequence (KHExL) among the PI orthologues was found within the K box domain (Fig. 1A) (Kramer ). A consensus PI motif (MPFxFRVQPxQPNLQE), which is unique to B class PI proteins (Kramer ; Moon ), was identified in the LMADS8 C-terminal region (Fig. 1A).

Fig. 1.

Sequence information for the LMADS8 and 9 proteins. (A) Sequence comparison of LMADS8 and 9 as well as the related B class MADS domain proteins. The MADS box proteins include LMADS8 and 9 (Lilium longiflorum), LRGLOA and B (Lilium regale), LFGLOA and B (Lilium formolongi), OMADS8 (Oncidium Gower Ramsey), and PI (Arabidopsis thaliana). The first and second underlined regions represent the MADS domain and K domain, respectively. The box in the K domain represents a sequence (KHExL) that is a highly conserved for PI orthologues. The box in the C-terminal region represents the PI motif (MPFxFRVQPxQPNLQE) that is a highly conserved motif for PI orthologues. Amino acid residues identical to LMADS8 are indicated as dots. To improve the alignment, dashes were introduced into the sequence. The names of the LMADS8 and 9 proteins are underlined. This sequence alignment was generated by the ClustalW Multiple Sequence Alignment Program at the DNA Data Bank of Japan (http://clustalw.ddbj.nig.ac.jp/top-e.html). (B) Sequence comparison of LMADS8 and 9 proteins in the C-terminal region. The PI motif is in grey and boxed. Amino acid residues identical to LMADS8 are indicated as dots. To improve the alignment, dashes were introduced into the sequence. (This figure is available in colour at JXB online.)

The cDNA sequence for LMADS9 encoded a 181 amino acid protein that showed a high sequence identity (100% and 98%) to LFGLOB of L. formolongi (Akita ) and LRGLOB of L. regale (Winter ), respectively (Fig. 1A). LMADS9 also showed high sequence identity to LMADS8 (88% total, 95% in the MADS box, and 90% in the K box) (Fig. 1A). Similar to LMADS8, a conserved sequence (KHEHL) was also identified within the K box of LMADS9 (Fig. 1A). Interestingly, LMADS9 seemed to be a truncated form of LMADS8 based on the sequence comparison. Twenty-nine amino acids, including the consensus PI motif, in the C-terminal region of LMADS8 were not present in LMADS9 (Fig. 1A). To confirm this hypothesis further, the cDNA sequences (including the 3'-untranslated region sequence) of LMADS8 and LMADS9 were compared. A small deletion of 24 bp of DNA sequence after the stop codon was found in LMADS9 (Fig. 1B). After this deletion, a sequence of 48 bp that was highly matched to the PI motif of LMADS8 (only five out of 48 nucleotides are different between LMADS8 and LMADS9) was identified in the LMADS9 cDNA (Fig. 1B). Thus, this result strongly supported the conclusion that LMADS9 is a truncated form of LMADS8.

The high sequence identity among LMADS8 and 9 and the PI orthologues from various plant species suggests that LMADS8 and 9 are the putative PI orthologues in lily. The amino acid sequences shown in Fig. 1A and sequences of several other MADS box proteins were used to construct a phylogenetic tree of the B functional genes (Fig. 2). On the basis of this analysis, it was concluded that LMADS8 and 9 are in the PI lineage of the monocot Liliales.

Fig. 2.

Phylogenetic analysis of B class MADS domain proteins. Based on the amino acid sequence of the full-length protein, LMADS8 and 9 were closely related to LRGLOA and B (L. regale), LFGLOA and B (L. formolongi), and TGGLO (Tulipa gesneriana) in the PI group of MADS box genes in the monocot Liliales. The names of the LMADS8 and 9 proteins are shown in bold and underlined. The names of the plant species for each MADS box gene are listed after the protein names. The maximum likelihood (ML) phylogenetic tree was generated based on the maximum likelihood method using MEGA 5 software as described in the Materials and methods. The numbers at each node indicate the percentage of bootstrap values from 500 replications, and the bootstrap values <50% were hidden.

Gene expression of LMADS8 and LMADS9

To explore the relationships between sequence homology and expression patterns for LMADS8 and LMADS9, LMADS8/9 expression was detected using quantitative real-time PCR analysis. The LMADS8 mRNA was absent in the vegetative leaves, root, and stem, but it was detected in the flowers (Fig. 3A). When the expression of LMADS8 in flower organs of the flower buds during different developmental stages (<5, 5, 15, and 25 mm in length) was analysed further, LMADS8 mRNA was consistently highly expressed in the first and second whorl tepals from young to mature flower buds (Fig. 3A–D). LMADS8 mRNA was also detected in stamens during early flower development (<5 mm and 5 mm flower buds) (Fig. 3A, B), but its levels were significantly lower in mature flower buds (15 mm and 25 mm flower buds) (Fig. 3C, D). In contrast, LMADS8 mRNA was not detected in the carpel from young to mature flower buds (Fig. 3A–D). The expression pattern and the strength of the expression for LMADS9 were extremely similar to those observed for LMADS8 (Fig. 3E–H). The LMADS9 mRNA was also absent in the vegetative leaves, stem, and root (Fig. 3E), and the LMADS9 mRNA was highly expressed in the first and second whorl tepals during different developmental stages (Fig. 3E–H). In the stamens, LMADS9 mRNA was also detected only in young flower buds (<5 mm and 5 mm flower buds) (Fig. 3E, F) and was barely detectable in mature flower buds (15 mm and 25 mm flower buds) (Fig. 3G, H). Similar to that of LMADS8, LMADS9 mRNA was also not detected in the carpel from young to mature flower buds (Fig. 3E–H). The expression patterns for LMADS8 and LMADS9 were different from those observed for PI orthologues in Arabidopsis and rice; these orthologues are only expressed in stamens and petals (Goto and Meyerowitz, 1994; Rounsley ; Prasad and Vijayraghavan, 2003; Yao ). Their expression patterns were also different from those of PI orthologues in monocot orchids, which are expressed in all four flower organs (Tsai ; Xu ; Chang ).

Fig. 3.

Detection of expression of LMADS8 and LMADS9 in L. longiflorum. Total RNA isolated from the root, stem, leaf, and flower organs of the first and second whorl tepals, stamen, and carpel of L. longiflorum flower buds at different developmental stages (<5, 5, 15, and 25 mm in length) were used as templates to detect the expression of LMADS8 (A–D) and LMADS9 (E–H) by quantitative real-time PCR. The results indicate that LMADS8 and LMADS9 have identical expression patterns. LMADS8 and LMADS9 mRNAs were absent in the root, stem, and leaf, but were highly expressed in the first and second whorl tepals during all stages of flower development. LMADS8 and LMADS9 mRNAs were detected in the stamen only in young flower buds and were almost undetectable in the carpel. In quantitative real-time PCR, the columns represent the relative expression of these genes. The transcript levels were determined using two to three replicates and were normalized using ACTIN for lily. The error bars represent standard deviations. Each experiment was repeated three times with similar results.

The ectopic expression of LMADS8 and LMADS9 partially converts sepals into petal-like structures in transgenic Arabidopsis plants

To investigate the function of LMADS8 and 9 further, cDNAs for these two genes, driven by the CaMV 35S promoter, were transformed into Arabidopsis for functional analysis.

Five independent 35S:LMADS8 transgenic Arabidopsis plants were obtained. One plant was phenotypically indistinguishable from untransformed wild-type plants, whereas another four plants showed identical altered phenotypes (Fig. 4D). Flowering time, determined by the number of rosette leaves, was not affected in these transgenic plants compared with the wild-type plants. However, flowers produced in the inflorescence of these plants (Fig. 4D) were different from those observed in wild-type plants (Fig. 4A). Unlike in the wild-type flowers (Fig. 4B, C), the partial conversion of the first whorl green sepals into elongated green/white petal-like structures (Fig. 4E, F) was observed in these 35S:LMADS8 flowers. Unlike the wild-type first whorl sepals, which remained tightly associated even after pollination (Fig. 4B, C), these first whorl green/white petal-like structures in 35S:LMADS8 flowers were completely separated immediately after flower opening (Fig. 4E, F). When the epidermal cells of the abaxial side (Fig. 4K) in these first whorl petal-like structures were examined, most of the cells in the converted white portion of the petal-like structures were observed to be morphologically similar to wild-type second whorl petal epidermal cells (Fig. 4M) and distinct from the wild-type first whorl sepal epidermis (Fig. 4J).

Fig. 4.

Phenotypic analysis of transgenic Arabidopsis plants ectopically expressing LMADS8 or LMADS9. (A) A wild-type inflorescence contains flower buds (fb) and mature flowers with normal first whorl sepals (s) and second whorl petals (p). (B) A mature wild-type Arabidopsis flower consists of four whorls of organs, including four sepals (s), four elongated petals (p), six stamens, and two fused carpels. The stamens and carpels were kept inside by the unopened sepals. (C) A sepal was removed from a mature wild-type Arabidopsis flower to reveal the third whorl stamen (st) and fourth whorl carpel (c). s, sepal; p, petal. (D) A 35S:LMADS8 (35S:L8) inflorescence contains flower buds (fb) with green/white sepals in the first whorl and fully opened mature flowers with green/white elongated petal-like sepals (ps) in the first whorl and normal petals (p) in the second whorl of the flower. (E) Close-up view of a 35S:LMADS8 transgenic Arabidopsis flower. Green/white elongated petal-like sepals (ps) and normal petals (p) were produced in the first and second whorls of the flowers, respectively. (F) Close-up view of a 35S:LMADS8 transgenic Arabidopsis flower. Green/white elongated petal-like sepals (ps) and normal petals (p) were produced in the first and second whorls of the flowers, respectively. A petal-like sepal was removed to reveal the inner organs. (G) A 35S:LMADS9 (35S:L9) inflorescence contains flower buds (fb) and fully opened mature flowers with green/white petal-like sepals (ps) in the first whorl and normal petals (p) in the second whorl of the flower. (H) Close-up view of a 35S:LMADS9 transgenic Arabidopsis flower. Green/white petal-like sepals (ps) and normal petals (p) were produced in the first and second whorls of the flowers, respectively. (I) Close-up view of the top of a 35S:LMADS9 transgenic Arabidopsis flower. Green/white petal-like sepals (ps) and normal petals (p) were produced in the first and second whorls of the flowers, respectively. (J) Scanning electron microscopy (SEM) of the surface cells with irregular shapes in the abaxial side of the epidermis of wild-type first whorl sepals. (K) SEM of surface cells of the epidermis in the abaxial side of the first whorl petal-like sepal (ps) of a 35S:LMADS8 (35S:L8-ps) flower, which are similar to the mature wild-type second whorl petal epidermis in M. (L) SEM of the surface cells of the epidermis in the abaxial side of the first whorl petal-like sepal (ps) of a 35S:LMADS9 (35S:L9-ps) flower, which are similar to the mature wild-type petal epidermis in M. (M) SEM of the surface cells of the epidermis in the abaxial side of a mature wild-type petal. Bar=1 mm in A–I, and 10 μm in J–M.

Eight independent 35S:LMADS9 transgenic Arabidopsis plants were obtained. Three plants were phenotypically indistinguishable from untransformed wild-type plants, whereas another five plants showed altered phenotypes (Fig. 4G) similar to those observed in the 35S:LMADS8 plants. Partial conversion of the first whorl green sepals into green/white petal-like structures (Fig. 4G–I) was also observed in these 35S:LMADS9 flowers. However, the petal-like structure in these 35S:LMADS9 flowers (Fig. 4H) was relatively shorter than that in the 35S:LMADS8 flowers (Fig. 4E). Most of the epidermal cells in the converted white portion of the abaxial side (Fig. 4L) in these first whorl petal-like structures were also morphologically similar to wild-type petal epidermal cells (Fig. 4M).

To explore whether this abnormal phenotype was correlated with LMADS8/9 expression in the transgenic plants, quantitative real-time PCR analysis was performed. As shown in Fig. 5, higher LMADS8 (Fig. 5A) and LMADS9 (Fig. 5B) expression was observed in the transgenic plants with the severe phenotype than in the transgenic plants with wild-type-like phenotypes. This result indicates that the phenotypes generated in the 35S:LMADS8/9 transgenic Arabidopsis plants were due to the ectopic expression of the lily LMADS8/9 gene.

Fig. 5.

The detection of gene expression in transgenic Arabidopsis plants. (A) Total RNAs isolated from one severe (L8-4) and one wild-type-like (L8-5) 45-day-old 35S:LMADS8 transgenic Arabidopsis and from one untransformed wild-type plant (WT) were used as templates for quantitative real-time PCR. The result indicates that LMADS8 is expressed at a higher level in the L8-4 than in the L8-5 transgenic plants. LMADS8 expression was undetectable in untransformed wild-type plants. (B) Total RNAs isolated from one severe (L9-4) and one wild-type-like (L9-2) 45-day-old 35S:LMADS9 transgenic Arabidopsis and from one untransformed wild-type plant (WT) were used as templates for quantitative real-time PCR. The result indicates that LMADS9 is expressed at a higher level in the L9-4 than in the L9-2 transgenic plants. LMADS9 expression was undetectable in untransformed wild-type plants. (C) Total RNAs isolated from one strongly complementary (L8/pi-5) and one pi-like (L8/pi-17) 45-day-old 35S:LMADS8/pi transgenic Arabidopsis and from one untransformed pi mutant plant (pi-1) were used as templates for quantitative real-time PCR. The result indicates that LMADS8 is expressed at a higher level in the L8/pi-5 than in the L8/pi-17 transgenic plants. LMADS8 expression was undetectable in untransformed pi mutant plants. (D) Total RNAs isolated from one strongly complementary (L9/pi-10) and one pi-like (L9/pi-18) 45-day-old 35S:LMADS9/pi transgenic Arabidopsis and from one untransformed pi mutant plant (pi-1) were used as templates for quantitative real-time PCR. The result indicates that the LMADS9 is expressed at a higher level in the L9/pi-10 than in the L9/pi-18 transgenic plants. LMADS9 expression was undetectable in untransformed pi mutant plants. (E) Total RNAs isolated from one severe (L8-M-51) and one wild-type-like (L8-M-59) 45-day-old 35S:LMADS8-M transgenic Arabidopsis and from one untransformed wild-type plant (WT) were used as templates for quantitative real-time PCR. The result indicates that LMADS8-M is expressed at a higher level in the L8-M-51 than in the L8-M-59 transgenic plants. LMADS8-M expression was undetectable in untransformed wild-type plants. (F) Total RNAs isolated from one severe (L9-M-51) and one wild-type-like (L9-M-58) 45-day-old 35S:LMADS8-M transgenic Arabidopsis and from one untransformed wild-type plant (WT) were used as templates for quantitative real-time PCR. The result indicates that LMADS9-M is expressed at a higher level in the L9-M-51 than in the L9-M-58 transgenic plants. LMADS9-M expression was undetectable in untransformed wild-type plants. In quantitative real-time PCR, the columns represent the relative expression of these genes. Transcript levels of these genes were determined using two to three replicates and were normalized using UBIQUITIN10. Error bars represent the standard deviation. Each experiment was repeated three times with similar results.

The ectopic expression of LMADS8 and LMADS9 caused different effects in rescuing the pi phenotype in pi-1 mutants

To investigate the function of LMADS8 and LMADS9 further, the cDNAs of these two genes, driven by the CaMV 35S promoter, were transformed into pi-1 plants for complementation analysis. The pi mutation always causes the conversion of the second whorl petal into a sepal structure and the conversion of the third flower whorl stamen into a carpel structure (Fig. 6A, B) (Bowman ; Jack ; Goto and Meyerowitz, 1994).

Fig. 6.

Phenotypic analysis of Arabidopsis pi-1 mutants ectopically expressing LMADS8 or LMADS9. (A) Flowers produced in the inflorescence of a pi-1 mutant plant failed to open during all stages of flower development. (B) Close-up of the unopened pi-1 flower in A. Green sepal-like structures (sl) were produced in the second whorl of this pi-1 mutant flower. S, sepal. (C) Flowers produced in the inflorescence of a 35S:LMADS8/pi (35S:L8/pi) plant containing fully elongated second whorl petals (p) and a first whorl petal-like sepal (ps) with white colour during all stages of flower development. (D) Close-up of a 35S:LMADS8/pi (35S:L8/pi) flower in C that contains fully elongated petals (p) and a petal-like sepal (ps) with white colour in the second and first whorls, respectively. (E) Flowers produced in the inflorescence of a 35S:LMADS9/pi (35S:L9/pi) plant containing second whorl short white petals (p) and a first whorl petal-like sepal (ps) during all stages of flower development. (F) Close-up of a 35S:LMADS9/pi (35S:L9/pi) flower in E that contains short petals (p) and a petal-like sepal (ps) with white colour in the second and first whorls, respectively. (G) Close-up of a fully elongated petal from a 35S:LMADS8/pi (L8/pi) flower (left) and a short petal from a 35S:LMADS9/pi (L9/pi) flower (right). (H) Close-up of a fully elongated first whorl petal-like sepal from a 35S:LMADS8/pi (L8/pi) flower (left) and a short first whorl petal-like sepal from a 35S:LMADS9/pi (L9/pi) flower (right). (I–J) SEM of the surface cells of the epidermis of the second whorl organ of a 35S:LMADS8/pi (35S:L8/pi) (I) and 35S:LMADS9/pi (35S:L9/pi) (J) flower, which are similar to the mature wild-type petal epidermis in L. (K) SEM of the surface cells with irregular shapes in the epidermis of the second whorl organ of a pi-1 (pi) mutant flower, which are similar to the mature wild-type first whorl sepal epidermis in P. (L) SEM of the surface cells of the epidermis of a mature wild-type petal. (M and N) SEM of the surface cells of the epidermis of the first whorl organ of a 35S:LMADS8/pi (35S:L8/pi) (M) and 35S:LMADS9/pi (35S:L9/pi) (N) flower that are similar to the mature wild-type petal epidermis in L. (O and P) SEM of the surface cells with irregular shapes in the epidermis of the first whorl sepal of a pi-1 (pi) mutant flower (O), which are similar to the mature wild-type (WT) first whorl sepal epidermis (P). Bar=0.5 mm in A–H, and 10 μm in I–P.

Thirteen independent 35S:LMADS8/pi transgenic Arabidopsis plants were obtained. Seven of these plants showed identical altered phenotypes during reproductive development, whereas the six other plants were phenotypically indistinguishable from untransformed wild-type plants. In these 35S:LMADS8 transgenic pi-1 plants, the defects of petal formation were fully restored (Fig. 6C, D). In addition, the partial conversion of the first whorl green sepal into a well-expanded green/white petal-like structure (Fig. 6C, D) was observed in these 35S:LMADS8/pi flowers. In contrast, the defect associated with formation of the third whorl stamen was not restored in these 35S:LMADS8/pi flowers (Fig. 6C, D).

Seventeen independent 35S:LMADS9/pi transgenic Arabidopsis plants were obtained. Four plants also showed altered phenotypes, in which petal formation was rescued during reproductive development, whereas the other 13 plants were phenotypically indistinguishable from untransformed wild-type plants. However, the restoration of the second whorl petal defects in 35S:LMADS9/pi (Fig. 6E, F) was weaker than that in 35S:LMADS8/pi flowers. The second whorl white petals in 35S:LMADS9/pi flowers (Fig. 6E, F) were clearly shorter than those in 35S:LMADS8/pi flowers (Fig. 6G). In addition, the partial conversion of the first whorl green sepal into a green/white petal-like structure (Fig. 6E, F) was observed in these 35S:LMADS9/pi flowers. Similarly, the green/white petal-like structures in these 35S:LMADS9/pi flowers were much shorter than those in 35S:LMADS8/pi flowers (Fig. 6H). Similar to what was observed in 35S:LMADS8/pi flowers, the defect associated with the third whorl stamen formation was not restored in these 35S:LMADS9/pi flowers (Fig. 6E, F). This finding clearly indicates that LMADS8 and LMADS9 were able to rescue the petal, but not the stamen, defect in transgenic pi-1 mutant plants.

When the epidermal cells in the second whorl petal-like structures in the 35S:LMADS8/pi (Fig. 6I) and 35S:LMADS9/pi plants (Fig. 6J) were examined, cells morphologically similar to wild-type petal epidermal cells (Fig. 6L) and distinct from the pi-1 second whorl sepal-like structure epidermis (Fig. 6K) were observed. When the epidermal cells in the first whorl petal-like structures in the 35S:LMADS8/pi (Fig. 6M) and 35S:LMADS9/pi plants (Fig. 6N) were examined, cells morphologically similar to wild-type second whorl petal epidermal cells (Fig. 6L) and distinct from the pi-1 (Fig. 6O) and wild-type (Fig. 6P) first whorl sepal epidermis were observed.

To explore whether the phenotype was correlated with LMADS8/9 expression in the transgenic pi-1 plants, quantitative real-time PCR analysis was performed. As shown in Fig. 5, higher LMADS8 (Fig. 5C) and LMADS9 (Fig. 5D) expression levels were observed in the transgenic pi-1 plants with the mutant phenotype rescued than in the transgenic pi-1 plants without complementation. This result indicates that the complementation generated in the 35S:LMADS8/9 transgenic Arabidopsis pi mutants was due to the ectopic expression of the lily LMADS8/9 genes.

The ectopic expression of truncated LMADS8 with the C-terminal sequence or PI motif deleted reduced the ability to rescue the pi phenotype in pi-1 mutants

Because 35S:LMADS8 exerted a stronger rescue effect in the pi-1 mutants than 35S:LMADS9, it is reasonable to propose that this effect may be due to the additional 29 amino acids in the C-terminal region of LMADS8 (Fig. 7A). To investigate the function of this C-terminal region further, two truncated versions of LMADS8, in which either the PI motif (LMADS8-PI) or the 29 amino acids of the C-terminal sequence (including the PI motif) (LMADS8-C) were deleted, were constructed and transformed into pi-1 mutants for functional complementation analysis (Fig. 7B). Seven and three independent 35S:LMADS8-C/pi and 35S:LMADS8-PI/pi transgenic Arabidopsis plants, respectively, that showed identical altered phenotypes for each construct during reproductive development were obtained. Compared with the full-length LMADS8, which caused the complete rescue of the pi-1 mutant in petals (Fig. 7B-1, -2), LMADS8-PI rescued the petal formation for pi-1 mutants in a relatively weaker manner (Fig. 7B-3, -4). The second whorl white petal in the 35S:LMADS8-PI/pi plants was ∼2 mm in length and clearly shorter than that in the 35S:LMADS8/pi or a wild-type flower, which was ∼2.5 mm in length (Fig. 7B-7, -8). Interestingly, LMADS8-C showed a partial rescue of the petal formation for pi-1 mutants (Fig. 7B-5, -6), which was very similar to that observed in 35S:LMADS9/pi transgenic flowers (Figs 6E, F, 7C-5, -6), and the rescued phenotype was weaker than that in the LMADS8-PI and full-length LMADS8 plants. The length of the second whorl white petal in the 35S:LMADS8-C/pi plants (1.5 mm) was longer than the second whorl organ in the pi-1 mutants (1 mm) (Fig. 7B-7, -8) similar to that in the 35S:LMADS9/pi plants (1.5 mm) (Fig. 7C-7), but shorter than that in the 35S:LMADS8-PI/pi (2 mm) and 35S:LMADS8/pi plants (2.5 mm) (Fig. 7B-7, -8). These findings clearly indicated that the different extents of the rescue effect in the pi-1 mutant phenotype by LMADS8 and LMADS9 were mainly due to the additional sequences in the C-terminal region (including the PI motif) of LMADS8 as had been proposed.

Fig. 7.

Phenotypic analysis of Arabidopsis pi-1 mutants ectopically expressing LMADS8 with a C-terminal region deletion or LMADS9 with a C-terminal region addition. (A) The protein structure of LMADS8/9 in the C-terminal region. The LMADS8 protein contains a PI motif (green and boxed region) in the region of the 29 amino acid C-terminus, which was absent in the LMADS9 protein. (B) Flowers produced in the inflorescence of a 35S:LMADS8/pi (35S:L8/pi) plant (B-1, B-2), a 35S:LMADS8-PI/pi (35S:L8-PI/pi) plant (B-3, B-4) plant, and a 35S:LMADS8-C/pi (35S:L8-C/pi) plant (B-5, B-6), containing different degrees of fully elongated second whorl petals (p) and a first whorl petal-like sepal (ps) with white colour during all stages of flower development. (B-7) Length comparison (from left to right) of elongated second whorl organs from a 35S:LMADS8/pi (L8/pi) flower (2.5 mm), a 35S:LMADS8-PI/pi (L8-PI/pi) flower (2 mm), a 35S:LMADS8-C/pi (L8-C/pi) flower (1.5 mm), second (pi-2nd) (1 mm) and first (pi-1st) (1.5 mm) whorl organs of a pi-1 flower, and the elongated second whorl petal of a wild-type (WT) flower (2.5 mm). (B-8) The variation and statistical analysis of the length for the second whorl organs from 35S:LMADS8/pi (L8/pi), 35S:LMADS8-PI/pi (L8-PI/pi), 35S:LMADS8-C/pi (L8-C/pi), pi-1 (pi) mutant and wild-type (WT) plants. The flower organ length was measured for five plants, with two measurements for each plant. Standard deviations are shown at the top of the columns. (C) Flowers produced in the inflorescence of a 35S:LMADS9+L8C/pi (35S:L9+L8C/pi) plant (C-1, C-2), a 35S:LMADS9+L8PI/pi (35S:L9+L8PI/pi) plant (C-3, C-4) plant, and a 35S:LMADS9/pi (35S:L9/pi) plant (C-5, C-6), containing different degrees of fully elongated second whorl petals (p) and first whorl petal-like sepals (ps) with white colour during all stages of flower development. (C-7) Length comparison (from left to right) of elongated second whorl organs from a 35S:LMADS9+L8C/pi (L9+C/pi) flower (2.3 mm), a 35S:LMADS9+L8PI/pi (L9+PI/pi) flower (2 mm), a 35S:LMADS9/pi (L9/pi) flower (1.5 mm), second (pi-2nd) (1 mm) and first (pi-1st) (1.5 mm) whorl organs of a pi-1 flower, and the elongated second whorl petal of a wild-type (WT) flower (2.5 mm). (C-8) The variation and statistical analysis of the length for the second whorl organs from 35S:LMADS9+L8C/pi (L9+C/pi), 35S:LMADS9+L8PI/pi (L9+PI/pi), 35S:LMADS9/pi (L9/pi), pi-1 (pi) mutant, and wild-type (WT) plants. The flower organ length was measured for five plants, with two measurements for each plant. Standard deviations are shown at the top of the columns.

Ectopic expression of LMADS9 fused to the C-terminal sequence or PI motif of LMADS8 enhanced the ability to rescue the pi phenotype in pi-1 mutants

Because both LMADS9 and LMADS8-C, which lacked the additional 29 amino acids of the C-terminal sequence, rescued the pi-1 mutant in a similar matter, it was of interest to determine whether the rescue in the pi-1 mutants would be modulated by the addition of the 29 amino acids of the LMADS8 C-terminal sequence (L8C) or the PI motif alone (L8PI) to LMADS9. For this purpose, two constructs (LMADS9+L8C and LMADS9+L8PI) were produced and transformed into pi-1 mutants for functional complementation analysis (Fig. 7C). Four and seven independent 35S:LMADS9+L8C/pi and 35S:LMADS9+L8PI/pi transgenic Arabidopsis plants, respectively, that showed identical altered phenotypes for each construct during reproductive development were obtained. The results indicated that LMADS9+L8C almost fully rescued the petal formation in pi-1 mutants (Fig. 7C-1, -2), similarly to the LMADS8/pi transgenic plants (Fig. 7B-1, -2). The LMADS9+L8PI plants exhibited a partial rescue of the petal formation in pi-1 mutants (Fig. 7C-3, -4), which was similar to that observed in the LMADS8-PI/pi transgenic plants (Fig. 7B-3, -4). The rescue of the phenotype in the LMADS9+L8PI plants (Fig. 7C-3, -4) was stronger than that in plants expressing full-length LMADS9 (Fig. 7C-5, -6) and weaker than that in plants expressing LMADS9+L8C (Fig. 7C-1, -2). The length of the second whorl white petal in 35S:LMADS9+L8PI/pi plants (2 mm) was clearly shorter than that in 35S:LMADS9+L8C/pi (2.3 mm) or wild-type plants (2.5 mm), but longer than that in 35S:LMADS9/pi plants (1.5 mm) or the second whorl organ in the pi-1 mutants (1 mm) (Fig. 7C-7, -8). These observations further supported the hypothesis that the additional 29 amino acids of the C-terminal sequence, including the PI motif, in LMADS8 play a valuable role in the function of LMADS8 and 9.

The ectopic expression of truncated LMADS8 and LMADS9 cDNAs lacking the MADS box region alters petal and stamen formation in Arabidopsis

MADS box genes have been thought to form dimers and bind to DNA through a DNA-binding domain (the MADS box domain) at the N-terminus (Schwarz-Sommer ; Tröbner ; Riechmann ; Mizukami ). Mutant forms of Arabidopsis AP3 or lily LMADS1 (AP3 homologue) lacking the N-terminal MADS box domain region have been reported to generate ap3 mutations once transformed into wild-type Arabidopsis (Krizek ; Tzeng and Yang, 2001). It has been reported that the ap3-like flowers produced in 35S:LMADS1-M plants are caused by dominant negative mutations through the formation of non-functional heterodimers between truncated LMADS1 and PI proteins in transgenic Arabidopsis (Tzeng and Yang, 2001). To investigate the functions of LMADS8 and 9 further, truncated cDNAs lacking the MADS box domain (35S:LMADS8/9-M) were transformed into Arabidopsis to generate dominant negative mutants.

Six out of the 18 independent 35S:LMADS8-M transgenic Arabidopsis plants showed similar abnormal phenotypes. These plants (Fig. 8A) were phenotypically similar to wild-type plants (Fig. 8B) during both vegetative and inflorescence development. However, they produced flowers with altered phenotypes. Unlike mature wild-type flowers, in which the petals and stamens elongate rapidly during late flower development and result in normal flower opening (Fig. 4B), these 35S:LMADS8-M flowers failed to open during all stages of flower development and remained unsenescent for a long period in the inflorescence (Fig. 8A, C, D). This phenotype was clearly different from that of the wild-type inflorescence, which featured senescent flowers and elongated siliques (Fig. 8B). The mature 35S:LMADS8-M flowers contained all four floral organs when observed upon manual opening (Fig. 8E). However, short and degenerated petal-like structures were observed in the second whorl of the 35S:LMADS8-M flowers (Fig. 8E) that were different from the white elongated petals in the wild-type flowers (Fig. 4B). In addition, short and degenerated stamen-like structures were observed in the third whorl of the 35S:LMADS8-M flowers (Fig. 8E), which were phenotypically different from the elongated stamens in the wild-type flowers (Fig. 4B, C). No pollen was produced from the anther-like structures in the 35S:LMADS8-M flowers (Fig. 8E). Thus, these 35S:LMADS8-M flowers were sterile, producing no seeds.

Fig. 8.

Phenotypic analysis of transgenic Arabidopsis plants ectopically expressing LMADS8-M or LMADS9-M. (A) A 35-day-old 35S:LMADS8-M (35S:L8-M) transgenic Arabidopsis plant produced unopened flowers. (B) A 35-day-old wild-type (WT) plant produced normal flowers and elongated silique (arrowed).(C) Close-up of the unopened flowers in the inflorescence of a 35S:LMADS8-M (35S:L8-M) plant in A. (D) Close-up of an unopened flower in C. (E) When opened manually, short petal-like structures (p), short stamen-like structures (st), and an unelongated carpel (c) were observed in the inner three whorls of a mature 35S:LMADS8-M (35S:L8-M) flower from D. Two sepals (s) were removed to reveal the inner organs. (F) A 35-day-old 35S:LMADS9-M (35S:L9-M) transgenic Arabidopsis plant produced unopened flowers. (G) Close-up of the flowers in the inflorescence of a 35S:LMADS9-M (35S:L9-M) plant in F. Most flowers were unopened during all stages of flower development, whereas slightly opened flowers with relatively long petals were occasionally observed (arrowed). (H) Close-up of an unopened flower in G. (I) When opened manually, short petal-like structures (p), short stamen-like structures (st), and a normal carpel (c) were observed in the inner three whorls of a mature 35S:LMADS9-M flower from H. Two sepals (s) were removed to reveal the inner organs. (J) Close-up of a slightly opened flower in G. (K) Close-up of the flower opened manually from J. Short stamen-like structures (st) along with normally elongated petals (p) and carpel (c) were observed in the inner three whorls. Two sepals (s) were removed to reveal the inner organs. Bar=1 cm in A, B, and F; 0.5 cm in C, G; and 0.1 cm in D, E, and H–K.

As in the 35S:LMADS8-M plants, similar unopened flowers (Fig. 8F–H) with defective short petals and stamens were observed in the flowers of the six 35S:LMADS9-M transgenic plants (Fig. 8I). However, the development of the protruding carpel (Fig. 8H, I) was apparently normal in these 35S:LMADS9-M flowers compared with that in the 35S:LMADS8-M flowers (Fig. 8D, E). Occasionally, flowers with longer petals were also produced in these 35S:LMADS9-M transgenic plants (Fig. 8G, J). However, these flowers were still sterile due to the production of short and defective stamens (Fig. 8K).

As shown in Fig. 5, higher LMADS8-M (Fig. 5E) and LMADS9-M (Fig. 5F) expression levels were observed in the transgenic plants with the severe phenotype than in the transgenic plant that was indistinguishable from the wild-type plants. This discrepancy indicates that the phenotypes generated in the 35S:LMADS8/9-M transgenic Arabidopsis plants were due to the ectopic expression of the lily LMADS8/9-M gene.

Homodimer and heterodimer formation of the LMADS1, 8, and 9 proteins

LMADS1, the PaleoAP3 homologue from the monocot lily, has been reported to be able to form homodimers (Tzeng and Yang, 2001; Tzeng ). Because LMADS8 and 9 showed high sequence homology to monocot PI orthologues, it was interesting to study the possible interaction between LMADS1 and LMADS8/9. Thus, a yeast two-hybrid analysis was performed. As shown in Fig. 9, LMADS8 and 9 were able to form heterodimers with LMADS1. Similar to LMADS1 (Tzeng and Yang, 2001), LMADS8 and 9 were able to form homodimers (Fig. 9B). When the interaction between LMADS8 and LMADS9 was analysed, β-galactosidase activity was also detected (Fig. 9B), indicating that LMADS8 and 9 are able to form not only homodimers, but also heterodimers with each other.

Fig. 9.

Protein interactions among B class proteins of L. longiflorum in a yeast two-hybrid assay. (A) Interaction between LMADS1 and LMADS8/9 proteins in a yeast two-hybrid assay. The yeast strain AH109 co-transformed with truncated LMADS1 in an activation domain plasmid (AD, Prey) and LMADS1, 8, and 9 in a binding domain plasmid (BD, Bait) was streaked on a minimal synthetic dropout (SD) medium without adenine, histidine, leucine, and tryptophan. Yeast cells transformed with LMADS1 in the activation domain plasmid (AD) and the binding domain plasmid (BD) alone were used as controls for background activity. (B) β-Galactosidase activity in yeast cells transformed with combinations of truncated LMADS1, LMADS8, or LMADS9 in either the binding domain plasmid (BD, Bait) or the activation domain plasmid (AD, Prey) was calculated according to Miller (1992). Yeast cells transformed with LMADS1, 8, and 9 in the binding domain plasmid (BD) or the activation domain plasmid (AD) alone were used as controls for background activity. Yeast transformed with vectors (AD+BD) only was used as a negative control.

To confirm further the results obtained from the yeast two-hybrid analysis, an EMSA, a technique established for the investigation of dimerization and DNA binding of MIKC-type MADS domain proteins (Schwarz-Sommer ; Riechmann ; Winter ; Tzeng ), was employed. As shown in Fig. 10, similar to LMADS1, LMADS8 and 9 proteins bound efficiently to the CArG1 probe, and protein–DNA complexes were observed. The ability of LMADS8 homodimers to bind CArG1 was clearly similar to that of LMADS1 and relatively stronger than that of LMADS9 (Fig. 10).

Fig. 10.

Homodimerization of LMADS1, 8, and 9 in electrophoretic mobility shift assays (EMSAs). Equal amounts of LMADS1, LMADS8, and LMADS9 were incubated with the DIG-labelled DNA probe CArG1, and the protein–DNA complexes were subjected to polyacrylamide gel electrophoresis. For the competition assay, a 10-fold molar excess of the cold competitor DNA was included in the binding reactions. The results indicate that the LMADS1, LMADS8, and LMADS9 proteins were able to bind to CArG1 and form protein–DNA complexes. These protein–DNA complexes disappeared and only free probes were seen once the cold competitor DNA was added.

Discussion

To investigate the role of B class MADS box genes in the regulation of flower development in the lily (L. longiflorum), two genes, LMADS8 and LMADS9, were identified and characterized in this study. Based on sequence alignment, the conserved motifs in the C-terminal regions of the encoded proteins, and phylogenetic tree analysis, LMADS8 and 9 were classified as PI-like genes and were found to be closely related to monocot PI orthologues (Figs 1, 2). The deletion of 29 amino acids, including the consensus PI motif in the C-terminal region, in LMADS9 suggested the possibility that LMADS9 might be a truncated form of LMADS8 in lily.

The presence of two PI orthologues in lily is not surprising because duplicated B class genes are often seen in basal eudicots (Di Stilio ) and non-grass monocots, such as Liliales (Liliaceae) and Asparagales (Kramer ; Kanno ). However, the appearance of the duplicated paleoAP3 or PI genes occurred differently in plant species (Kramer ). For example, plants such as tulips and orchids tend to have duplicated paleoAP3 genes rather than PI genes. Two (TGDEFA and TGDEFB), four (PeMADS2, PeMADS3, PeMADS4, and PeMADS5), and three (OMADS3, OMADS5, and OMADS9) paleoAP3 orthologues have been identified in tulips (Tulipa gesneriana, Liliaceae) and orchids (Phalaenopsis equestris and Oncidium Gower Ramsey) (Kanno ; Tsai ; Chang ). Only one PI-like gene is found in these plants. In contrast, relatively few species contain duplicated PI-like genes as found in L. longiflorum. For example, two (LRGLOA and LRGLOB), two (LFGLOA and LFGLOB), and three (MaGLOA1, MaGLOA2, and MaGLOB) PI-like genes have been reported in L. regale, L. formolongi, and Muscari armeniacum, respectively (Winter ; Kanno ; Akita ). Interestingly, as in L. longiflorum, only one paleoAP3 gene is found in these three species, indicating that the duplication events for B class genes occurred more frequently for paleoAP3-like genes than for PI genes in monocots during evolution.

The expression pattern for these two lily PI orthologues, LMADS8 and 9, was interesting. First, their expression patterns were almost identical (Fig. 3). In addition to their close phylogenetic relationship (Fig. 2), this result strongly supported the idea that these orthologues were raised through a gene duplication event and retained similar transcriptional regulation and function in lily. Because ∼88% identity was observed between the LMADS8 and 9 proteins (Fig. 1A), it appeared that there were only few sequence diversifications after duplication. Although high sequence identity was observed, the deletion of 29 amino acids, including the consensus PI motif, from the C-terminal region of LMADS9 was observed when compared with LMADS8. This result revealed that LMADS9 might be a truncated form of LMADS8 in lily, in which the C-terminal region of LMADS9 was lost after gene duplication. This assumption was further supported by the identification of a sequence of 48 bp that was highly matched to the PI motif of LMADS8 (only five out of 48 nucleotides and four out of 16 translated amino acids are different between LMADS8 and LMADS9) in the LMADS9 cDNA after the stop codon (Fig. 1B). This finding strongly indicated that after gene duplication, LMADS9 experienced a 24 bp DNA deletion before the PI motif as well as a point mutation that produced a premature stop codon. This caused the 29 amino acid truncation of LMADS9 compared with LMADS8. Interestingly, similar phenomena were observed for PI orthologues in L. regale and L. formolongi in which two PI-like genes, LRGLOA and LRGLOB in the former and LFGLOA and LFGLOB in the latter, were identified, and the sequence analysis indicated that LRGLOB/LFGLOB was a truncated form of LRGLOA/LFGLOA (Winter ; Akita ) (Fig. 1A). These results indicate the vast similarity of the genome contents and organization for these three species of lilies. Thus, the duplication, point mutation, and deletion events of the two PI-like genes should have occurred before the divergence of these three species of lilies during evolution. Furthermore, truncated forms have also been described for AP3 orthologues (ThtAP3-2a and ThdAP3-2a) in Thalictrum dioicum (Ranunculaceae) (Kramer ; Di Stilio ), indicating that a similar duplication and mutation occurred in both PI-like and AP3-like genes during evolution.

The second interesting characteristic of LMADS8 and 9 is their spatial and temporal expression pattern. The expression patterns for LMADS8 and 9 were different from the expression patterns of PI orthologues in dicots, which are normally expressed in petals and stamens (Goto and Meyerowitz, 1994; Rounsley ; Prasad and Vijayraghavan, 2003; Yao ). In contrast, the mRNAs for LMADS8 and 9 were strongly detected in the first and second whorls tepals during all stages of flower development and in the stamens only during early flower development (Fig. 3). This expression pattern was also different from that of some PI orthologues in non-grass monocots, which are expressed in all four flower organs, such as TGGLO in tulip (T. gesneriana), TriaGLO in Tricyrtis affinis, CsPIc in Crocus sativus, PeMADS6 in P. equestris, DcOPI in Dendrobium crumenatum, and OMADS8 in Oncidium orchid, as well as MaGLOA1, MaGLOA2, and MaGLOB in M. armeniacum (Tsai ; Tsaftaris ; Xu ; Kanno ; Chang ). LMADS8 and 9 also showed a slightly different expression pattern from that of LFGLOB/LFGLOA, which is expressed strongly in whorls 1–3, but very weakly in whorl 4 (Akita ). The expression of LMADS8 and 9 in the first whorl organ is explainable because lily plants produce almost identical first and second whorl tepals. The genes controlling second whorl organ formation in these plants are very probably also expressed in the first whorl organ (Tzeng and Yang, 2001; Kanno ). These results reveal that LMADS8 and 9 may have an important role in regulating all processes involved in tepal formation, while only participating in early stamen formation.

The function of LMADS8 and 9 in flower formation was further revealed by functional complementation analysis. Because an efficient genetic transformation system is still not available for lily (L. longiflorum), the functional analysis for LMADS8/9 was performed in transgenic Arabidopsis in this study. When LMADS8 and 9 were ectopically expressed in Arabidopsis pi mutants, the restoration of the mutant phenotype through the production of normal petal organs in the second whorl along with the green/white petal-like sepals in the first whorl in pi-1 mutants was observed in transgenic plants (Fig. 6). This result provided further evidence to support the conclusion that LMADS8 and 9 simply function as B genes in the PI lineage. This concept was further supported by the production of dominant negative mutations with degenerated short petal-like structures in the second whorl and short stamen-like structures in the third whorl of flowers ectopically expressing LMADS8 or LMADS9 truncated cDNAs lacking the MADS box domain (35S:LMADS8/9-M) (Fig. 8). These results indicate that LMADS8 and LMADS9 are able to participate in the Arabidopsis MADS functional protein complex in the sepals/petals/stamens to perform their functions.

One interesting finding of this study is the different phenotypes generated by the ectopic expression of LMADS8 and 9 or LMADS8-M and 9-M in transgenic Arabidopsis. 35S:LMADS8 completely rescued the second whorl petal formation and partially converted the first whorl sepal into an expanded green/white petal-like structure for Arabidopsis pi mutants, whereas 35S:LMADS9 only partially rescued the petal formation for pi mutants (Fig. 6). In addition, 35S:LMADS8-M generated more severe dominant negative phenotypes than those in 35S:LMADS9-M flowers (Fig. 8), suggesting the LMADS8-M could exert a greater effect in the Arabidopsis MADS functional protein complex than LMADS9-M. Because 29 amino acids, including the consensus PI motif, are missing from the C-terminal region of LMADS9 when compared with LMADS8, LMADS9 is a truncated form of LMADS8 in lily as described above. The different degrees of the phenotypes observed for 35S:LMADS8/pi, 35S:LMADS9/pi, and 35S:LMADS8-M/LMADS9-M suggested a valuable role for these 29 amino acids, including the PI motif, in the C-terminal region in the function of the lily PI orthologues. The absence of this C-terminal region apparently weakened, but did not abolish, the ability of LMADS9 to participate in the Arabidopsis MADS functional protein complex when compared with LMADS8, which contained the entire C-terminal region.

This assumption is further supported by the deletion experiment for LMADS8 and the addition experiment for LMADS9. The results (Fig. 7) indicated that LMADS8-C (with the 29 amino acids of the C-terminal sequence deleted) only partially rescued the petal formation for pi-1 mutants, which was similar to the rescue observed in 35S:LMADS9/pi transgenic plants. The LMADS8-PI (with only the PI motif deleted) also partially rescued the petal formation for pi-1 mutants, and the rescued phenotype was stronger than that in the LMADS8-C plants and weaker than in plants expressing full-length LMADS8. Similarly, LMADS9+L8C (with the addition of the 29 amino acids of the LMADS8 C-terminal sequence) almost fully rescued the petal formation in pi-1 mutants similar to the rescue observed in LMADS8/pi transgenic plants. The LMADS9+L8PI (with the addition of only the PI motif of LMADS8) showed a partial rescue of the petal formation in pi mutants similar to that observed in LMADS8-PI/pi transgenic plants. These results clearly indicated that the more sequence that is deleted from this C-terminal region, the less the degree of the rescue in the pi-1 mutant by the LMADS8 proteins is. In contrast, the more sequences added from this C-terminal region, the greater ability of the LMADS9 proteins to rescue the pi-1 mutation. When considering these data together, the additional 29 amino acids of the C-terminal sequence, including the PI motif, in LMADS8 most probably play a valuable role in the function of LMADS8 and 9 as proposed.

Through yeast two-hybrid analysis (Fig. 9), LMADS8 and LMADS9 were observed to form heterodimers with each other and with the lily paleoAP3 orthologue LMADS1. LMADS8 and LMADS9 were also found to be capable of forming homodimers. Interestingly, the ability of LMADS8 homodimers to bind to CArG1 was relatively stronger than that of LMADS9 homodimers, as indicated by EMSA (Fig. 10), which suggested that the absence of the PI motif in the C-terminal region of lily PI orthologues will affect, to some extent, the ability of these homodimers to bind to the CArG1 sequence and perform their function. Interestingly, this result was similar to that for the lily paleoAP3 orthologue LMADS1. It has been reported that the ability of LMADS1 to form homodimers decreases when the C-terminal paleoAP3 motif is deleted (Tzeng ). Thus, both paleoAP3 and the PI motifs of the paleoAP3 orthologue LMADS1 and the PI orthologue LMADS8 are needed, but not essential, for their homodimers to bind to the CArG1 sequence and function. The finding of similar functional roles for the PI motif of PI orthologues (LMADS8/9) and the paleoAP3 motifs of the paleoAP3 orthologue (LMADS1) provides useful information for the study of homo- and heterodimerization and the function of B class genes during evolution. This postulation was, however, different from the reports that the C-terminal motif of the Arabidopsis PI is dispensable in its floral organ identity function (Piwarzyk ) and that the function of PI orthologues in pea (Pisum sativum) and legume (Medicago truncatula) is not affected in orthologues lacking the C-terminal PI motif (Berbel ; Benlloch ). The present findings suggested that the conserved motifs in the C-terminus of the AP3/PI subfamily of MADS box proteins may play functional roles that diverged between monocots and dicots during evolution.

In summary, two PI orthologues of the B class MADS box genes, LMADS8 and LMADS9, that specify flower development in the lily L. longiflorum were characterized. The mRNAs for LMADS8 and LMADS9 were flower specific and expressed in all stages of first and second whorl tepal development and in the early stage of stamen formation. Different effects on floral organ complementation or conversion observed in 35S:LMADS8/LMADS8-PI/LMADS8-C-pi and 35S:LMADS9/LMADS9+L8PI/LMADS9+L8C-pi, 35S:LMADS8-M, and 35S:LMADS9-M transgenic Arabidopsis plants and the differential ability of LMADS8 and LMADS9 homodimers to bind to the CArG1 sequence suggest a valuable role for the 29 amino acids, including the PI motif, in the C-terminal region in the function of LMADS8 and 9. In addition to LMADS1, the characterization of the LMADS8 and 9 genes provides useful information for understanding the relationships between the lily B class MADS box genes as well as their roles in regulating perianth formation. Because most of the ABCDE (Tzeng and Yang, 2001; Tzeng , 2003, 2004; Chen ; Hsu ) genes of L. longiflorum have been isolated and characterized in the authors’ laboratory, further investigation of the role of ACDE genes in interactions with the B genes characterized in this study should lead to a deeper understanding of flower organ formation in non-grass monocots.