Histone Acetylation Accompanied with Promoter Sequences Displaying Differential Expression Profiles of B-Class MADS-Box Genes for Phalaenopsis Floral Morphogenesis

Five B-class MADS-box genes, including four APETALA3 (AP3)-like PeMADS2∼5 and one PISTILLATA (PI)-like PeMADS6, specify the spectacular flower morphology in orchids. The PI-like PeMADS6 ubiquitously expresses in all floral organs. The four AP3-like genes, resulted from two duplication events, express ubiquitously at floral primordia and early floral organ stages, but show distinct expression profiles at late floral organ primordia and floral bud stages. Here, we isolated the upstream sequences of PeMADS2∼6 and studied the regulatory mechanism for their distinct gene expression. Phylogenetic footprinting analysis of the 1.3-kb upstream sequences of AP3-like PeMADS2∼5 showed that their promoter regions have sufficiently diverged and contributed to their subfunctionalization. The amplified promoter sequences of PeMADS2∼6 could drive beta-glucuronidase (GUS) gene expression in all floral organs, similar to their expression at the floral primordia stage. The promoter sequence of PeMADS4, exclusively expressed in lip and column, showed a 1.6∼3-fold higher expression in lip/column than in sepal/petal. Furthermore, we noted a 4.9-fold increase in histone acetylation (H3K9K14ac) in the translation start region of PeMADS4 in lip as compared in petal. All these results suggest that the regulation via the upstream sequences and increased H3K9K14ac level may act synergistically to display distinct expression profiles of the AP3-like genes at late floral organ primordia stage for Phalaenopsis floral morphogenesis.


Introduction
In Arabidopsis thaliana and Antirrhinum majus, the development of different floral organs is controlled by several classes of floral-organ identity genes [1]. All these genes, except AP2 from A. thaliana, belong to the MADS-box family, with a highly conserved 180-bp sequence of the MADS domain that can bind to the conserved CArG-box [CC(A/T) 6 GG] sequence [2,3]. These MADS-box genes were cloned from a wide range of plant species to explain the floral organ development [1,[4][5][6]. The diversification of MADS-box genes during evolution has been proposed to be a major driving force for floral diversity in land plant architecture [1,7].
Various regulatory strategies have been reported for the expression of MADSbox genes in Arabidopsis, including transcriptional regulation on the upstream sequences or intron regions by transcription factors, with feedback and feedforward loops, and epigenetic regulation by small RNAs [8]. Discrete cis-acting elements on the B-class APETALA3 (AP3) and PISTILLATA (PI) promoters are responsible for their expression in petal and stamen [9][10][11]. In addition, the first and second introns of FLOWERING LOCUS C (FLC) [12] and AGAMOUS (AG) [13], respectively, have a role in regulating spatial or temporal gene expression patterns. For epigenetic control of gene expression, both dimethyl histone H3 lysine-9 (H3K9me2) and dimethyl histone H3 lysine-27 (H3K27me2) are the gene repression markers. In contrast, both trimethyl histone H3 lysine-4 (H3K4me3) and acetyl histone H3 (H3Ac) are the active histone markers.
Orchidaceae is one of the largest families of flowering plants. The high species diversity in orchids is largely due to their adaptation to specialized insect pollination [14]. The orchid flower is spectacular with a gynostemium or column (a fusion of the male and female reproductive organs) and a highly modified petal, the labellum or lip, which offers a landing platform for pollinators [14][15][16].
In Phalaenopsis orchids, four AP3-like and one PI-like B-class MADS-box genes, PeMADS2,6, have been isolated and characterized for their roles in flower morphogenesis [17][18][19]. Two duplication events resulted in the four AP3-like PeMADS2,5. The first, occurring early in the evolutionary history of Orchidaceae, resulted in AP3A and AP3B clades, and the second resulted in four subclades, AP3A1 (PeMADS3), AP3A2 (PeMADS4), AP3B1 (PeMADS2), and AP3B2 (PeMADS5) [20][21][22]. Fluorescence in situ hybridization revealed that the four AP3-like PeMADS2,5 genes are located on different chromosomes of P. equestris, so the four orchid AP3 paralogs may have been resulted from genome duplication [20]. The effects of gene duplication and their differences on gene regulation are important in the diversity and evolution of flowering plants [23][24][25]. At the floral primordia and early floral organ primordia stages, the transcripts of PeMADS2,5 were detected ubiquitously, and then they are constrained to distinctively expressed organs at the late floral organ primordia stage and floral bud stage: PeMADS2 mainly expresses in sepal and petal, PeMADS3 predominantly expresses in petal and lip, PeMADS4 exclusively expresses in lip and column, and PeMADS5 is mainly expressed in petal [17,20]. In contrast, the PI-like PeMADS6 is ubiquitously expressed in sepal, petal, lip, and column [18]. The 'Orchid code' assumes that the differential expression of B-class genes determined the development of sepal, petal, lip, and column [26,27]. Moreover, the 'homeotic orchid tepal' (HOT) model are proposed for the dualistic features of duplicated Bclass MADS-box genes involved in orchid perianth development and growth [20].
Although the five B-class MADS-box genes play important roles in the perianth development in orchids, the regulatory strategies for their distinct expression profiles in various floral organs have not been characterized. In this study, we identified the upstream promoter sequences of PeMADS2,6 in Phalaenopsis orchids and used phylogenetic footprinting to identify conserved motifs among these promoter sequences. We analyzed the promoter activity of the upstream sequences of PeMADS2,6 for driving GUS and luciferase gene expression in various floral organs. In addition, we examined the regulatory effects of the intron region, DNA methylation, and histone modification for their association with the high expression level of PeMADS4 in lip.

Plant materials
All upstream sequences of PeMADS genes were isolated from P. equestris with red sepal, petal and orange lip [17]. P. aphrodite subsp. formosana with white sepal, petal and yellow lip was purchased from Taiwan Sugar Corp. (Tainan, Taiwan) and used in particle bombardment experiments because the white sepal and petal made it easier for GUS staining. All plant materials were grown in the greenhouse at National Cheng Kung University (Tainan, Taiwan) under natural light and controlled temperature from 23˚C to 27˚C.
Isolation of the upstream promoter sequences of PeMADS2,6 Genomic DNA was extracted from young flower buds by the cetyltriammonium bromide (CTAB) method [28]. The upstream promoter sequences of PeMADS2,6 were isolated by use of the Universal GenomeWalker Kit (Clontech, Palo Alto, CA, USA). The desired DNA fragment was obtained by two successive PCR-based rounds of screening the GenomeWalker libraries and checked by agarose gel electrophoresis. The major bands were recovered from gels with use of the Gel DNA Fragment Extraction Kit (Geneaid, New Taipei City, Taiwan), and cloned into the pGEM-T Easy Vector (Promega, Madison, WI, USA). We randomly selected 10 to 12 colonies for sequencing. The promoter sequences were compared to all known DNA sequences with use of the default settings of BLASTN from NCBI (www.ncbi.nlm.nih.gov).
When genome walking could not extend the upstream regulatory sequences for PeMADS3,5 genes, we used BAC clones constructed from P. equestris [29] for promoter identification. Southern blot hybridization was used to identify BAC clones containing various PeMADS genes with the gene-specific probes used in previous study [17,18]. The BAC DNAs of each clone were isolated following the standard method [30], then digested with HindIII and separated by electrophoresis in 0.7% agarose gel. The resolved bands corresponding to promoter sequences with hybridized signals were recovered with use of the Gel DNA Fragment Extraction Kit (Geneaid) and cloned into the pGEM-T Easy vector (Promega). We randomly selected 10 to 12 colonies for sequencing. The promoter sequences characterized are deposited at the NCBI site under the accession numbers: PeMADS2 promoter (KJ127932), PeMADS3 promoter (KJ127933), PeMADS4 promoter (KJ127931), PeMADS5 promoter (KJ127934), and PeMADS6 promoter (KJ127935).

Promoter sequence analysis
We used the PLACE software (A database of Plant cis-acting Regulatory DNA Elements, http://www.dna.affrc.go.jp/PLACE/index.html) to predict the putative CArG box having 10-nt matches with the C(A/T) 8 G sequence [31]. In addition, the CArG box sequence was predicted for a standard of 9 of 10 matches with the core consensus binding site CC(A/T) 6 GG [9] by using a homemade software (designed by Dr. Chih-Hsiung Fu, Department of Engineering Science, National Cheng Kung University).
Since only 1.3-kb upstream regulatory sequence of PeMADS3 was cloned, we then chose the 1.3-kb upstream sequences of PeMADS2,5 for phylogenetic footprinting analysis by FOOTPRINTER [32]. This tool takes into account the evolutionary relationships and the phylogenetic tree analysis. The prediction of a conserved 10-bp or 11-bp motif with a 0-bp mutation allowance was performed and the motif losses were allowed to identify the conserved motif within two promoter sequences. For all other parameters, default values were used.

Construction of chimeric reporter gene fusions
All the promoter sequences of PeMADS2,6 were amplified directly from the genomic DNA of P. equestris by using PCR with the length of 3,249, 1,293, 3,303, 2,062, and 1,514 bp, respectively, and constructed in-frame translational fusions with the GUS reporter gene in pBI221 vector. Serial deletion fragments of PeMADS2,6 promoters were also cloned by PCR-amplification with a series of forward primers and a reverse primer (S1 Table). The resulting PCR products were cloned into pGEM-T Easy vector (Promega) and then digested with SphI and BamHI. pBI221 containing the GUS reporter gene was digested with the same enzymes to replace the cauliflower mosaic virus (CaMV) 35S promoter with the serial deletion fragments of promoter sequences. All constructs were confirmed by sequencing to eliminate possible PCR-introduced mutations. Both the pBI221 vector containing the CaMV 35S promoter-GUS fusion (pBI221) and pBI221 vector containing a promoterless GUS cassette (pBI-PL) were recruited and considered as positive and negative controls, respectively.

Transient transformation by particle bombardment
The promoter deletion-GUS fusion plasmids were isolated by use of the High-Speed Plasmid Mini kit (Geneaid) and coated on gold particles 1.6 mm in diameter by coprecipitation as described [33]. Before particle bombardment, each floral organ was separated from the floral buds and placed on a central core 2 cm in diameter on solid agar medium. Promoter constructs were bombarded into various floral organs by use of Model Biolistic PDS-1000/He (BioRad, Hercules, CA, USA) at 1,100 psi helium gas pressure, 28.5-inch Hg vacuum and 9-cm target distance. After bombardment, floral organs were incubated at 23˚C to 27˚C for 2 days in an incubator with a 10-h/14-h light-dark photoperiod until analyzed by GUS histochemical staining and quantitative dual luciferase assays.

GUS histochemical staining assay
Histochemical staining of GUS activity was performed as described [34]. Tissues for GUS staining were vacuum-infiltrated in GUS assay buffer (1 mg/ml 5bromo-4-chloro-3-indoyl glucuronide [X-Gluc]; 50 mM sodium phosphate, pH 7.0; 10 mM EDTA, pH 8.0; 0.5 mM potassium ferricyanide, 0.5 mM potassium ferrocyanide and 0.1% Triton X-100) and incubated at 37˚C overnight. Stained tissues were cleared of chlorophyll in 70% ethanol and then photographed under a microscope (TMS-F, Nikon). For each promoter-GUS fusion, the GUS staining pattern was analyzed in four independent bombarded buds, and repeated three times independently.

Quantitative dual luciferase assay
Serial-deleted promoter fragments were obtained by digestion of PeMADS4 and PeMADS6 promoters in pBI221 with SphI and BamHI, and then ligated into pJD301, containing a firefly (Photinus pyralis) luciferase gene, to replace the CaMV 35S promoter for the serial pJD-Pe4p and pJD-Pe6p constructs. pJD301_R, with Renilla luciferase gene driven by the CaMV 35S promoter, was an internal control to normalize transfection efficiency.
On day 2 after bombardment, each sample was ground, and then 1 X Passive Luciferase Buffer (Promega, Madison, WI, USA) was added. Luciferase activity was measured by use of the dual-luciferase reporter assay system (Promega) with a Lumat LB 9507 Luminometer (Berthold Technologies, Bad Wildbad, Germany), a 10-sec pre-measurement delay and a 10-sec measurement period for each assay. The relative luciferase activity was calculated as the ratio of firefly to Renilla luciferase activity. For each analysis, two independent buds were bombarded and analyzed, and the bombardments were repeated three times independently. Statistical analysis was performed by T-test, and the differences were considered significant at p,0.01.

Southern blot hybridization
Genomic DNA isolated from sepals, petals, lips, and columns of P. equestris was digested with restriction enzymes HpaII and MspI or DraI and HpaII, resolved on 0.8% agarose gel, and transferred to nylon filters (Amersham Pharmacia Biotech, Piscataway, NJ, USA) by use of a vacuum transfer system (Amersham Pharmacia Biotech). The recognition sites of HpaII and MspI are both CCGG, but digestion of HpaII is blocked by cytosine methylation and MspI is blocked by only cytosine methylation within the first cytosine (CpCGG). Two probes of PeMADS4 for methylation Southern blot assay were used: probe 1 contained the promoter and 59 UTR sequences, and probe 2 contained the 5 th intron regions. The primers were listed in S1 Table. Southern blot hybridization was performed and followed the standard protocol [30] with the 32 P-labeled probes prepared by a PCR strategy.

Bisulfite sequencing
Bisulfite sequencing analysis was carried out with the EpiTect Bisulfite kit (QIAGEN, Hilden, Germany). 2-mg genomic DNAs from the petal and lip of P. equestris were treated with conversion reagents and then cleaned up as described in the manufacturer's instruction. The DNA was served as a template for PCR amplification with the incubation at 94˚C for 5 min, thermocycling for 35 cycles (94˚C for 30 s, 55˚C for 30 s, and 72˚C for 1 min), and finally at 72˚C for 7 min with the primers listed in S1 Table. The PCR products were cloned into pGEM-T Easy vector (Promega) and transformed into Escherichia coli. We randomly selected 10 colonies for sequencing and analysis.

Chromatin immunoprecipitation (ChIP) and real-time PCR analyses
ChIP assay was as described [35]. Chromatin extracts were prepared from the petal and lip of P. equestris treated with formaldehyde. Chromatin was sheared to an average length of 500-1500 bp by sonication and immunoprecipitated with the antibodies anti-H3K4me3 (catalogue no. 04-745, Millipore, Billerica, MA, USA), anti-H3K9me2 (catalogue no. 04-768), or anti-H3K9K14ac (catalogue no. 06-599). The immunocomplexes were harvested with Protein G agarose beads (Millipore) and heated at 65˚C for 5 hours to release DNA cross-linked to the immunoprecipitated proteins. The DNA cross-linked to the immunoprecipitated proteins was analyzed by real-time PCR with the primers listed in S1 Table. The immunoprecipitations were performed twice.
The DNA template was mixed with 2X SYBR Green PCR master mix (Applied Biosystems, Foster City, CA, USA) in an ABI Prism 7000 sequence detection system (Applied Biosystems), and each sample was analyzed in triplicate. Reactions involved incubation at 95˚C for 10 min, and thermocycling for 40 cycles (95˚C for 15 s and 60˚C for 1 min). After amplification, melting curve analysis was used to verify amplicon specificity and primer dimer formation. The amount of DNA after ChIP was quantified and normalized to an internal control ACTIN2 for H3K4me3 and H3K9K14ac or Ta3 for H3K9me2 [36]. Data are mean ¡ SD calculated from three technological and two biological replicates.

Results
Cloning of the upstream sequences of PeMADS2,6 The five PeMADS promoter regions were cloned from genomic DNA by use of GenomeWalker. The PeMADS2,6 promoter fragments obtained were 3,249, 1,293, 422, 2,121, and 1,514 bp, respectively (Fig. 1). However, the 2,121-bp promoter sequence of PeMADS5 could not be amplified from the genomic DNA of P. equestris by using PCR, and both PeMADS3 and PeMADS4 with the promoter fragments ,1.5 kb could not be extended. We then used BAC clones of P. equestris for promoter identification of PeMADS3,5 by Southern blot hybridization with probes from the coding sequences of PeMADS3,5 (Table 1). BAC DNA was isolated and confirmed by restriction enzyme digestion. The resolved bands corresponding to the hybridized signals were recovered and sequenced. After assembly, the PeMADS3 promoter remained at 1,293 bp. For the PeMADS4 promoter, a 4.8-kb DNA fragment from BAC clones was recovered and sequenced, which extended its upstream sequence to 3,303 bp ( Fig. 1, horizontal line box). For PeMADS5, the promoter fragment was replaced with a 2,062-bp fragment (Fig. 1).
Prediction of CArG box in the promoter sequences of PeMADS2,6 We used an in-house developed software to predict the presence of the putative CArG box with a standard of a 9-in-10-nt match with the core consensus CC(A/ T) 6 GG motif [9]. Four CArG boxes were detected at nucleotides 21,548, 2318, 2248, and 2214 in the PeMADS2 promoter region; three at nucleotides 21,912, 21,231, and 2434 in the PeMADS4 promoter region; three at nucleotides 21,263, 2738, and 2175 in the PeMADS5 promoter region, and two at nucleotides 2881 and 2678 in the PeMADS6 promoter region (Fig. 1, Table 2). In contrast, no CArG-box-like sequence was detected in the PeMADS3 promoter sequence.
We used the PLACE database with a standard of 10-nt match with the C(A/ T) 8 G sequence to further examine the CArG box [31]. Five CArG boxes were detected at nucleotides 22,586, 22,236, 21,891, 21,254, and 2750 in the PeMADS2 promoter region; two at nucleotides 21,097 and 2443 in the PeMADS3 promoter region; six at nucleotides 23,258,22,980,22,656,22,523,2559, and 265 in the PeMADS4 promoter region; and two at nucleotides 21,041 and 2862 of the PeMADS5 promoter region; one at nucleotide 2774 in the PeMADS6 promoter region (Fig. 1, Table 2).
The conserved regulatory motifs predicted within the promoter regions of PeMADS2,5 The conserved regulatory motifs are thought to be functional important for gene expression profiles [37][38][39][40]. However, multiple alignment of these promoter sequences using global alignment procedures was failed because the inversions often cause rearrangements of the regulatory elements [41]. Therefore, we examined the 1.3-kb upstream sequences of PeMADS2,5 by phylogenetic footprinting, a method for discovering regulatory elements in a set of regulatory  regions [32] and has been used to promoter analysis for MADS-box genes in Arabidopsis and Orchis italica and for bZIP genes in rice and sorghum [40,42,43]. With prediction of a conserved 11-bp motif with a 0-bp mutation allowance, conservation of four 11-bp motifs was identified between the promoter regions of PeMADS3 and PeMADS4 in different order ( Fig. 2A), while no motifs were conserved between promoter regions of PeMADS2 and PeMADS5 ( Fig. 2A). With prediction of conserved 10-bp motifs, increased conserved motifs were identified between the promoter regions of PeMADS3 and PeMADS4, and three motifs were lined up in those of PeMADS2 and PeMADS5 (Fig. 2B). Interestingly, we noticed that differential conserved 10-bp motif sets were detected between the promoter regions of PeMADS2/PeMADS5 and PeMADS3/PeMADS4 (Fig. 2B), which  suggests that the two lineage of PeMADS2/5 and PeMADS3/4 have diverged for their subfuncationalization after gene duplication. Moreover, the CArG boxes were broadly distributed in the promoter regions of the four PeMADS genes, and no clear correlations of these CArG boxes were detected with their distinct expression profiles (Fig. 2C). Furthermore, with prediction of a conserved 12-bp motif with a 1-bp mutation allowance, motifs were identified between the promoter regions of PeMADS2 and PeMADS5, PeMADS3 and PeMADS4, and PeMADS6 and AtPI, but no motifs were present between the promoter sequences of AtAP3 and PeMADS2,5 (Fig. 2D).

Functional analysis of PeMADS2,6 promoter sequences
We examined the promoter activities of PeMADS2,6 fragments with length of 3,249 bp, 1,293 bp, 3,303 bp, 2,062 bp, and 1,514 bp, respectively, for their ability to drive GUS expression by bombarding them into 1.5-cm floral buds of P. aphrodite subsp. formosana. Notably, all five PCR-amplified promoter fragments could drive GUS expression in the floral organs examined (Fig. 3A-T), similar to their expression patterns at the early floral primordia stage [20]. Moreover, serial deletion clones of the upstream sequences of PeMADS2,6 were constructed for GUS expression assay.
The 208-and 375-bp promoter sequences of PeMADS6 and PeMADS4, respectively, were sufficient to drive GUS expression The PI-like PeMADS6 was expressed ubiquitously in all floral organs. To assess the minimal promoter region of PeMADS6, four deletion clones of the PeMADS6 promoter sequence were resulted, including 1,108-bp, 808-bp, 508-bp, and 208-bp fragments containing 3, 2, 0, and 0 CArG boxes, respectively. Similar to the full length PeMADS6 promoter construct, pBI-Pe6p-1514 ( Fig. 3Q-T), all four deletion promoter sequences of PeMADS6 could drive GUS expression in all the floral organs examined (Fig. 4B-Q), although the expression was slightly decreased in pBI-Pe6p-508 and pBI-Pe6p-208 constructs (Fig. 4J-Q). Moreover, quantitative dual luciferase assay was performed to further examine the differential promoter activities of these serial deletion constructs. The pJD-Pe6p-208 construct was sufficient to drive luciferase expression in all floral organs (Fig. 4R). Extension of the upstream sequence from the pJD-Pe6p-508 to pJD-Pe6p-1514 constructs conferred similar GUS expression in all floral organs (Fig. 4R). Therefore, the 208-bp promoter sequence was a minimal promoter for PeMADS6 expression in all four floral organs. The four AP3-like PeMADS2,5 genes express differentially in the floral organs with distinct patterns [17]. Two to five serial deletion clones for the upstream regulatory sequences of PeMADS2,5 were constructed for GUS expression assay. Similarly, most serial deletion constructs could drive GUS expression in all four floral organs (Fig. 5B-Q, S1-S3 Figures), resembled to the gene expression patterns at the early floral organ primordia stage. Among them, the minimal promoters of PeMADS2,5 were found to be 291 bp, 407 bp, 375 bp, and 122 bp of their upstream regulatory sequences, respectively ( Fig. 5N-Q, S1-S3 Figures).
The promoter sequence of PeMADS4 was chosen for further examination by using quantitative dual luciferase assay to delineate its exclusive expression in lip and column at the late floral primordia stage [20]. All five serial deletion promoter constructs of PeMADS4 with various lengths conferred a 1.6,3-fold increase of luciferase activities in lip and column than in sepal and petal (Fig. 5R), in accordance with the high expression of PeMADS4 in lip and column. The shortest fragment, pJD-Pe4p-375 conferred a 3-fold higher promoter activity in lip and column than in sepal and petal (Fig. 5R), which suggests that the upstream sequences of PeMADS4 was necessary, but not sufficient for its exclusive high expression in lip and column. It is possible that other factors are also required for regulating the lip-and column-specific expression of PeMADS4 at the late floral organ primordia stage of Phalaenopsis orchids.
The 5 th intron of PeMADS4 had no effect on its organ-specific expression pattern The longest introns of AG and FLC in Arabidopsis play a regulatory role for their gene expression [12,13]. It was intriguing to know whether the introns of PeMADS4 may regulate its distinct expression at the late floral organ primordia stage of Phalaenopsis orchids. To test this, we first sequenced the genomic sequence of BAC clones containing PeMADS4, NCKU-PE-btBAC-1105 H24. Then, the genomic sequence was compared to its cDNA sequence, and seven exons and six introns were identified for PeMADS4 with a long 5 th intron of 9,483 bp (Fig. 6A). The 5 th intron contains two conserved CC(A/T) 6 GG motifs (Fig. 6A, rhombus) and 11 C(A/T) 8 G motifs (Fig. 6A, triangles) (Fig. 6A).
To assess the effect of the 5 th intron on PeMADS4 expression, we generated three subfragments of 8-kb, 3.5-kb, and 2-kb fragments by PCR amplification, cloned into the upstream region of the promoter sequence in the pBI-Pe4p-375 construct, and named as pBI-Pe4p-375+8 kb, pBI-Pe4p-375+3.5 kb, and pBI-Pe4p-375+2 kb 5 th intron constructs, respectively (Fig. 6A). The addition of the 8kb fragment intron resulted a sharp decrease of GUS expression in all four floral organs ( Fig. 6F-I) as compared with the pBI-Pe4p-375 native construct (Fig. 6B-E), which suggests that the 8-kb 5 th intron may have a negative effect on PeMADS4 expression. Alternatively, the addition of the 8-kb fragment was too long to affect the transformation efficiency. Otherwise, the GUS expression in all floral organs was not significantly different with the addition of either the 2-or 3.5-kb fragments of the 5 th intron (Fig. 6J-Q), so the 3.5-or 2-kb 5 th -intron sequence showed little or no effects for the exclusive PeMADS4 expression in lip and column.

DNA methylation was not responsible for regulation of PeMADS4 expression
To examine whether the specific expression of PeMADS4 in lip and column was caused by DNA methylation in the regulatory sequences in planta, Southern blot hybridization was performed to analyze the DNA methylation status in the promoter, translation start site, and intron regions of PeMADS4. DNA samples isolated from petal and lip of P. equestris were digested with the methylation-  sensitive endonucleases HpaII or MspI (H or M) (Fig. 7B) and DraI/HpaII (Fig. 7C). Probe 1 containing a 582-bp fragment including a 375-bp promoter sequence and a 207-bp 59-UTR region of PeMADS4 (Fig. 7A, B) was used for DNA samples from petal or lip and similar methylation status were obtained with digestion of HpaII (Fig. 7B, 4.2-and 8-kb fragments) and MspI (Fig. 7B, 4.2-and 5.6-kb fragments). Moreover, hybridization with probe 2, containing a 2,136-bp fragment of the 5 th intron of PeMADS4, gave the same results for all DraI/HpaIIdigested DNA samples from sepal, petal, lip, and column (Fig. 7A, C, 1-and 1. kb fragments), so the HpaII site within probe 2 region was methylated in all four floral organs. Thus, the DNA methylation status was the same in the promoter region, translation start site, and the 5 th intron regions of the PeMADS4 gene for both petal and lip of Phalaenopsis flowers.

2-
To have a single-base resolution of methylation status, bisulfite sequencing technology was performed. Highly methylated cytosine residues were detected within the promoter region and translation start site of PeMADS4 in DNA samples from both petal and lip (Fig. 7D). Therefore, DNA methylation may not play a role if any in tissue specificity of PeMADS4 expression in lip and column.

Concomitant differential histone acetylation for PeMADS4 expression
It is possible that the tissue-specific expression profiles of PeMADS genes not only reside in the DNA-level promoter sequences, but also in the protein-level histone modification. To address this, various histone modifications were analyzed by ChIP assay with antibodies against the gene repression marker H3K9me2 and gene activation markers H3K4me3 and H3K9K14ac. The precipitated DNA samples from both petal and lip were analyzed by real-time PCR with the primer sequences located at the translation start site (ATG) and the 5 th intron regions of PeMADS4 (Fig. 8A). Notably, we detected a 4.9-fold higher H3K9K14ac at the translation start site in lip than in petal (Fig. 8B). In contrast, no differential levels of H3K9me2 and H3K4me3 were detected in the translation start site in both petal and lip (Fig. 8B). Furthermore, no substantial differential levels of H3K9me2, H3K4me3, and H3K9K14ac within the 5 th intron region of PeMADS4 were detected in petal and lip (Fig. 8C). Thus, the increased level of H3K9K14ac on the translation start site of PeMADS4 gene may allow for more access of the transcription factor and the further increased gene expression in lip, thus leading to its optimized expression in Phalaenopsis.

Discussion
In model species Arabidopsis thaliana and Antirrhinum majus, the molecular genetic studies on flower morphogenesis indicate that homeotic B-class MADSbox genes determine petals and stamen identities. A shift model has been proposed through several comparative expression analyses suggesting that a shift expression of B-class genes to the outer perianth is associated with the petal-like organs on the first flower whorl in several monocot species, such as Tulipa gesneriana and Lilium longiflorum [44,45]. To understand the development and evolution of orchid flowers, extensive studies of the molecular phylogeny and expression patterns of candidate B-class MADS-box genes have been performed in several genus, species, and hybrids of orchids. Four ancient orchid-specific clades of AP3-like and one major lineage of PI-like B-class genes were identified and characterized [17,26,46]. Differential expression of AP3-like genes has been associated with the identity of distinct perianth organs, that are the basis of several models used to explain the morphogenesis and evolution of the orchid flowers [20,27].

Histone acetylation and promoter sequences act synergistically to regulate PeMADS4 expression in lip
The minimum promoter sequence necessary for a wild-type AP3 expression pattern in Arabidopsis is localized within the 727-bp fragment upstream of the transcriptional start site [9]. In this study, we showed that the minimal promoter sequences for PeMADS2,6 were 291 bp, 407 bp, 375 bp, 122 bp, and 208 bp of their upstream regulatory sequences, respectively. In addition, the regulation of the promoter sequence and the increased H3K9K14ac level may act synergistically to result in the exclusive high expression of PeMADS4 in lip and column at the late floral organ primordia stage and floral bud stage of Phalaenopsis orchids. A similar regulatory mechanism may be adopted for regulation of the distinct expression profiles of the other three AP3-like paralogs, PeMADS2, PeMADS3, and PeMADS5, in Phalaenopsis flowers to complete the diversified subfunctionalization for orchid floral morphogenesis.
Transient expression assay for promoter analysis by use of particle bombardment Stable transformation of Phalaenopsis orchids is time-consuming and requires considerable human resources, because their long life cycles of two to three years for the transition from the seed germinative to reproductive stages. Transient expression assays, which were carried out by using particle bombardment [47,48] and protoplast transfection [49], have been used to reduce the time for analyzing gene functions in Arabidopsis [50,51], rice [49,52], maize [53], potato [54], soybean [55], tomato [56], wheat [55], and white spruce [57]. Protoplasts retain many signal transduction pathways from the cells which they are derived [58], but the process of protoplast cultivation may change the mRNA expression profiles [59]. In contrast, particle bombardment permits the transient expression within intact tissues of entire plants. However, an important argument for transient expression is that it is easy to overexpress gene constructs because of high copy numbers of plasmid DNA, strong promoters, long expression times, and the independence of the transgene expression on the genomic site of integration [60]. While in stable transgenic plants, a high transgene copy is frequently accompanied with gene silencing and the position effect affects the expression of promoter constructs by flanking sequences in the genome. In Orchidaceae, particle bombardment with histochemical GUS staining have been used to analyze the promoter activities of disease resistance response protein (OnDRRP), Expansin (OnExpansin), and three trypsin inhibitor (OnTI1,3) in leaves and flowers of Oncidium Gower Ramsey, and cytokinin oxidase (DSCKX1) in protocorm-likebodies of Dendrobium Sonia, respectively [61,62]. Here, we analyzed the promoter activities of PeMADS2,6 for driving GUS and luciferase reporter genes by particle bombardment, and the ubiquitous expression in all floral organs may be caused by the high copies of bombarded plasmid DNA and/or the naked DNA lack of chromatin modification. However, the serial deletion sequences of PeMADS4 promoter showed a higher luciferase activity in lip and column than in sepal and petal in contrast to the more-or-less similar lucifease activity detected in all four floral organs driven by PeMADS6 promoter. Moreover, we further examined the DNA methylation and histone modification within the translation start site of PeMADS4 to verify the regulatory strategy for its differential expression pattern. All these results suggested that the transient expression assay by particle bombardment accompanying with DNA methylation and histone modification analyses provides a basic information about the regulatory strategies of these PeMADS2,6 genes exhibiting distinct expression profiles.

375-bp promoter sequence of PeMADS4 was required for its lip and column expression
The 375-bp promoter sequence of PeMADS4 conferred higher luciferase activity in lip and column, which meant that the PeMADS4 promoter was regulated by a lip-and column-specific transcription factor. Several transcription factors have been shown to dominantly express in lip, such as MADS, ARF, C3H, HB-other, YABBY, ZF-HD, bZIP, CO, TALE, HD-ZIP, MYB, and AP2-like families [63]. The cis-acting regulatory elements on the upstream region of PeMADS4 were analyzed by PLACE software, and the CARGCW8GAT for MADS and MYBCOREATCYCB1 or MYBST1 for MYB-binding motifs were predicted within this sequence (S2 Table). It is possible that unidentified motifs were resided in the 375-bp fragment of PeMADS4 promoter for the interaction with above mentioned transcription factors. The exact transcription factors responding for the activation on the 375-bp fragment of PeMADS4 promoter were required for further studies.
The promoter sequences of OrcPI and PeMADS6 showed little similarity Comparing the differential expression patterns of AP3-like genes, PI-like genes showed uniformly and highly conserved expression patterns in all four floral organs in several species of various subfamily of Orchidoideae, such as Cypripedioideae, Epidendroideae, Orchidoideae, and Vanilloideae [20]. OrcPI is a PI-like MADS gene from Orchis italica, a species of Orchidoideae, and expresses in young inflorescences and all floral organs [42]. Various serial deletion of promoter sequence of OrcPI with 1324-bp, 854-bp, 577-bp, and 356-bp upstream sequences can drive the GUS expression in petal tissue in the white Rosa hybrid [42]. However, the upstream regulatory sequences of PeMADS6 and OrcPI could not be aligned together, even for their minimal promoter with 208-and 356-bp fragments, respectively. However, it is intriguing that two 11-bp motifs were detected between nucleotide 2249 and 2173 bp of PeMADS6 promoter and between 2937 and 2641 bp of OrcPI promoter by using the BLAST2 algorithm (S4 Figure). Whether this element plays any roles in the ubiquitous expression of the PI-like genes in orchid flowers awaits further studies.

Histone modification regulated the plant development and stress response
The epigenetic regulation is dynamic and varies between cell types and in response to development stages or environmental stimuli [64]. H3K9me2 is mainly detected in heterochromatin regions and associated with transposable elements (TEs) [65]. In contrast, H3K4me3 is enriched in euchromatic regions and associated with transcribed regions of non-TE genes. Expectedly, typical activating histone modification, such as H3K4me3 and H3K9ac is detected in the same genomic regions [66]. For example, during the vernalization response (exposure to a prolonged period of low temperature), the gene repression marks, H3K9me2 and H3K27me2, are enriched at FLC locus and thereby controlling flowering time, in contrast to the activated state of FLC chromatin with active histone marks, H3K4me3 and H3ac, before prolonged cold exposure [67]. Moreover, several epigenetic regulators are involved in the regulation of floral homeotic genes. Mutation in a H3K4 methyltransferase ATX1 results down-regulation of AP1, AP2, PI, and AG, but not of AP3 and SEP3 [68]. The PRC2-like complexes containing CLF, FIE, EMF2 and MSI1 act on repression of AG by regulating H3K27me3 [69,70], and the emf2 mutant results ectopic overexpression of AG, AP3, AP1, PI, SEP2, and SEP3 [71]. Furthermore, the gene repression marker, H3K27me3, have been shown that its release results in tissue-specific gene activation [72]. Here, we showed that an increased level of H3K9K14ac on the translation start site of PeMADS4 gene may enhance the exclusive gene expression in lip to decipher the lip morphogenesis in Phalaenopsis orchids, although its mechanism is still needed to be investigated.
Transgenic plants by use of the PeMADS2,5 promoter In Arabidopsis, the promoter sequences of AP3 and PI have been used for the floral organ-specific expression in the transgenic approach. The 288-bp promoter sequence of AP3 with a petal-specific domain [9] and the 300-bp fragment of PI with expression in petal and stamen [11] were used to drive an RNAi vector targeting the GUS reporter gene and introduced into a line constitutively expressing GUS, which resulted in reduced GUS expression in petal [73]. However, the promoter sequences of PeMADS2,5 drove the expression of GUS and luciferase reporter genes in the whole flower but not exclusively in distinct floral organs. Thus, the amplified upstream regulatory sequences of PeMADS2,5 could be used as flower-specific, but not practically for floral organ-specific promoters in the application for transgenic plants.
Supporting Information S1 Figure.