Ethylene-responsive transcription factors interact with promoters of ADH and PDC involved in persimmon (Diospyros kaki) fruit de-astringency

Plant materials and treatments

‘Mopan’ (astringent-type) persimmon (Diospyros kaki) fruit were obtained from a commercial orchard at Fangshan (Beijing, China). The fruit were transported to Zhejiang University (Hangzhou, Zhejiang, China) on the second day after harvest. Uniform fruit free of visible defects were chosen for the de-astringency experiments, including CO₂ and ethylene treatments.

For the ethylene treatment, the fruit were harvested in 2009, with mean colour index, firmness, and soluble solids content of 14.70, 52.31 N, and 17.88%, respectively. The fruit were divided into two lots of 180 fruit each. The control fruit were held at 20 °C free of any treatment for the same time periods as used for the ethylene-treated fruit. Continuous ethylene (100 µl l^–1) in air (2 d) was applied to the second batch to remove the astringency of the fruit. For the CO₂ treatment, the fruit were harvested in 2010, with mean colour index, firmness, and soluble solids content of 18.11, 43.08 N, and 18.76%, respectively. The fruit were divided into three batches (200 fruit in each) and the first two batches of fruit were treated with 95% CO₂ for 1 d or 2 d, separately, according to Matsuo and Ito (1977). The third batch of fruit was stored in air at 20 °C for 4 d as the control. After treatment, the fruit were held in air at 20 °C.

All of the treatments and controls were performed with three biological replicates. At each sampling point, bulk fruit flesh samples (without skin and core) were taken from three replicates each of four fruit. The samples were frozen in liquid nitrogen and stored at –80 °C for further use.

For tissue-specific gene expression, stems, leaves, and petals were collected at full bloom, frozen in liquid nitrogen, and stored until use.

Fruit firmness and TSS

Fruit firmness and TSS (total soluble solids) were recorded at each sampling time. Fruit firmness measurements were made with a TA-XT2i texture analyser (Stable Micro Systems, UK), and the penetration measurements were carried out according to our previous report (Yin et al., 2012). Firmness was measured at two positions 90° apart at the fruit equator, after the removal of 1mm peel. Fruit TSS were measured from samples from both ends of each fruit, using a N20 digital hand-held refractometer (Atago, Japan), according to the methods described in Yin et al. (2008). Fruit firmness and TSS were measured on each of 10 fruit.

Soluble condensed tannins content

Soluble condensed tannins (SCTs) were chosen as the main de-astringency index and were also used for comparing the effectiveness of different treatments. SCTs were measured using frozen samples, according to the method described previously (Yin et al., 2012), where fruit flesh was extracted with methanol and the SCTs were measured with the Folin–Ciocalteu reagent (Sigma). The results were expressed as tannin acids equivalents g^–1 fresh weight.

Acetaldehyde and ethanol

Acetaldehyde and ethanol production were measured with a gas chromatograph (Lunan Chemical Engineering Instrument Co. Ltd., model GC-14B, Shandong, China), fitted with a FID column (HP-INNOWAX). A 2g frozen fruit sample was ground in liquid nitrogen and the ground powder was mixed with 5ml saturated NaCl solution. Three ml of the mixture were transferred to 10ml air-tight vials with crimp-top caps. Before measurement, the vials were incubated at 60 °C for 1h in a water bath. Then, 0.2ml head-space gas was removed from each vial for acetaldehyde and ethanol measurement. The injector, detector, and oven temperatures were 140, 140, and 100 °C, respectively. Sec-butyl alcohol was added to each vial as an internal control. The results were calculated using standard curves for acetaldehyde and ethanol, respectively.

ADH and PDC activities

ADH and PDC enzymes were extracted and assayed according to the methods described by Botondi et al. (2012). One gram of persimmon fruit flesh was homogenized with 3ml 0.02M KH₂PO₄ buffer (pH 7.5) (including 2mM thiamine pyrophosphate, 2mM MgCl₂, 1mM β-mercaptoethanol), on ice. The homogenate was centrifuged at 17 000 rcf for 20min at 4 °C and the supernatant was filtered with PD-10 Desalting Columns (GE Healthcare Bio-Sciences AB) and the eluate was used to measure PDC and ADH activity. Three biological replicates were performed at each sampling point.

ADH activity was measured using the increment in NADH by reading the absorbance at 340nm, in 1ml of reaction buffer [0.68ml 100mM KOH-glycine (pH 9.0), 0.07ml 11.4mM NAD⁺, 0.15ml 2M ethanol, 0.10ml enzyme extract].

PDC activity was measured using the decrease in NADH. The 1ml reaction mixture consisted of 0.4ml 100mM MES (pH 7.5), 0.1ml 10mM thiamine pyrophosphate, 0.1ml 10mM MgCl₂, 0.1ml 2.6mM NADH, 0.1ml ADH (containing 7 kU ml^–1), 0.1ml of enzyme extract, and 0.1ml 200mM pyruvate.

One activity unit (U) was defined as the change in one unit of absorbance at 340nm min^–1 and the results were expressed as specific activity (mU g^–1 flesh). All the reagents were the products from Sigma Aldrich Co. (USA).

RNA extraction and cDNA synthesis

Total RNA was extracted from various frozen tissues (2.0g for flesh tissues, 0.8g for leaves, 1.2g for stems, 1.6g for petals) with the method developed by Chang et al. (1993). The trace contaminating genomic DNA in total RNA was removed swith TURBO Dnase (Ambion). cDNA synthesis was initiated from 1.0 µg DNA-free RNA, using iScript^TM cDNA Synthesis Kit (Bio-Rad). For each sampling point, three biological replicates were used for RNA extraction.

Gene isolation and sequence analysis

ADH and PDC genes were isolated both using degenerate primers constructed from information for other species and also the persimmon EST database [published by Nakagawa et al. (2008) and available from NCBI (National Center for Biotechnology Information)]. The EST databases was blasted using the sequences isolated from using degenerate primers. The redundant sequences were excluded and non-redundant EST sequences were chosen for further experiments. The novel hypoxia (high CO₂)-responsive ERF gene was isolated based on RNA-Seq. RNA-Seq was performed and the sequence was assembled and annotated by the Beijing Genome Institute (BGI) (Shenzhen, China). RNA isolation and the preparatory procedures were as previously reported by Feng et al. (2012). 26.7 million sequence reads were performed for each library and the sequences were compared by BLASTX to the public protein database.

Rapid amplification of cDNA ends (RACE) was chosen to amplify the 3’ end and/or 5’ end UTR (untranslated region) of ADH, PDC, and ERF, using a SMART^TM RACE cDNA Amplification Kit (Clontech). The sequences of primers used for cloning and RACE experiments are described in Supplementary Table S1 at JXB online.

The gene sequences were translated with online software (http://web.expasy.org/translate/) and were confirmed with the BLAST methods in Genbank. The deduced amino acid sequences of homologous genes in other plant species were downloaded from NCBI. Bootstrap Neighbor–Joining trees were generated with ClustalX (v 1.81), using the default parameters.

Oligonucleotide primers and real-time PCR

Oligonucleotide primers for real-time PCR analysis were designed with primer3 (v. 0.4.0,http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). The specificity of primers was determined by examining the melting curves and re-sequencing the PCR products. The sequences of oligonucleotide primers are described in Supplementary Table S1 at JXB online.

Real-time PCR was carried out using a CFX96 instrument (Bio-Rad). The PCR protocols were according to our previous reports, with Ssofast EvaGreen Supermix kit (Bio-Rad) (Yin et al., 2012). The relative abundance of each gene was calibrated with samples from mature fruit (for tissue specific studies) or day 0 fruit (for post-harvest experiments). Actin, a housekeeping gene, was used as the internal control (Akagi et al., 2009b).

Transient over-expression in persimmon leaves

In order to determine the roles of DkADH and DkPDC genes in the regulation of SCTs, an unstable transformation system (transient expression) was adapted. Full-length DkADH1 and DkPDC2 coding sequences were constructed into pGreenII 0029 62-SK (Hellens et al., 2005). Both of the constructs were electroporated into Agrobacterium tumefaciens GV3101. The transient expression system (Agrobacterium culture, infiltration buffer) was according to the protocols used for tobacco in our laboratory (Espley et al., 2007; Yin et al., 2010).

Transient expression in persimmon leaves was performed with the ‘Mopan’ cultivar, the same cultivar used for astringency removal experiments. Suspensions of Agrobacterium, harbouring either DkADH1 or DkPDC2, were prepared with infiltration buffer and infiltrated into persimmon leaves, on the tree. Negative controls were included in the same leaves. Three days after infiltration, the infiltrated leaves were sampled and used for SCT analysis. 0.1g of tissue from each of the infiltrated leaves was taken for SCTs analysis, using ten single leaf replicates in total.

Genome walking and transient assay

Genome walking and a transient assay were performed according to our previous report (Yin et al., 2010). Promoters of DkADH1 and DkPDC2 were isolated with a GenomeWalker kit (Clontech), using genomic DNA from ‘Mopan’ persimmon leaves. The sequences of DkADH1 and DkPDC2 are described in Supplementary Fig. S5 at JXB online. Full-length DkERF9 was inserted into pGreen II 0029 62-SK vector (SK), while the promoters of DkADH1 (1021bp) and DkPDC2 (919bp) were inserted into pGreen II 0800-LUC vector. As the DkADH1 promoter has an endogenous NcoI site, a 12bp pGem-T easy vector sequence was incorporated between the DkADH1 promoter and luciferase reporter. All of the constructs were electroporated into Agrobacterium tumefaciens GV3101. The transient assay was performed with Nicotiana benthamiana leaves. Agrobacterium cultures were prepared with infiltration buffer (10mM MES, 10mM MgCl₂, 150mM acetosyringone, pH 5.6) to an OD₆₀₀ from 0.7 to 1.0. Agrobacterium culture mixtures of TFs (1ml) and promoters (100 µl) were infiltrated into tobacco leaves by needle-less syringes. Tobacco plants were grown in a greenhouse, under natural light with daylight extension to 16h. Three days after infiltration, firefly luciferase and renilla luciferase were assayed using the dual luciferase assay reagents (Promega). For each TF-promoter interaction, three independent experiments were performed (at least six replicates in each experiments).

Statistical analysis

Principal component analysis (PCA) was applied to analyse the correlations between de-astringency (soluble tannin concentrations) and gene expression with AlphaSoft version 11.0 (Alpha MOS, Toulouse, France). Software and data manipulations were according to our previous report (Zhang et al., 2011). Statistical significance of differences was calculated using Student’s t test.

Isolation of novel hypoxia (high CO₂)-responsive ERF genes

Four hypoxia responsive DkERF genes (DkERF1, DkERF4, DkERF5, and DkERF6) were characterized in our previous studies (Yin et al., 2012). Two novel hypoxia-responsive DkERF genes, designed as DkERF9 (JX117848) and DkERF10 (JX145122) were newly isolated after comparing RNA from 2 d CO₂-treated fruit and control fruit using RNA-Seq technology. The RPKM (Reads per kb per million reads) values of DkERF9 and DkERF10 were 0 in 0 d fruit and 1.5692 and 450.2169 in 2 d CO₂-treated fruit, respectively (data not shown), indicating that both DkERF9 and DkERF10 were possibly high-CO₂/hypoxia-inducible. Phylogenetic analysis indicated that hypoxia responsive DkERF genes belong to different subfamilies. DkERF1 belongs to subfamily I, while DkERF4-6 belongs to subfamily VIII (according to Nakano et al., 2006). The two newly isolated hypoxia (high CO₂)-responsive DkERF genes were clustered in subfamily IV (DkERF9) and VII (DkERF10) (see Supplementary Fig. S1 at JXB online). Within the six DkERF genes, only DkERF10 had a conserved MCGGAII motif at the N-end of the deduced protein.

Fruit de-astringency and ERF genes expression

The relationship between hypoxia/high-CO₂-responsive ERF genes and fruit de-astringency was evaluated in ‘Mopan’ persimmon fruit using both CO₂ (95%) and ethylene (100 µl l^–1) treatments. Both treatments promoted de-astringency, indicated by the decreasing concentration of SCTs (Fig. 1). However, the effects of CO₂ and ethylene treatment on fruit ripening were different. Ethylene accelerated fruit texture change, and fruit softened from an initial value of 52.31 N to 14.19 N after ethylene treatment compared with 45.15 N in control fruit after 3 d in storage. By contrast, the firmness of CO₂-treated fruit was similar to that of control fruit (Fig. 1). The TSS content decreased generally during both treatments, however, a larger reduction was found in CO₂-treated fruit (Fig. 1).

Fig. 1.

Effects of CO₂ and ethylene treatments on SCTs, firmness, and TSS of ‘Mopan’ persimmon fruit at 20 °C. Mature fruit were treated with CO₂ (~95%, v/v, open diamonds, 1 d), ethylene (100 µl l^–1, open circles, 2 d), and air (control, open squares), separately. The horizontal arrow represents the end day of the treatment, when the fruit were transferred to 20 °C, free of ethylene. Error bars represent SEs from three (SCTs) or 10 (firmness and TSS) biological replicates.

Acetaldehyde and ethanol, the main products of anaerobic fermentation, were also analysed in CO₂- and ethylene-treated fruit. Acetaldehyde, the main compound that interacts with the SCTs to render them insoluble, was only detectable during treatment, and was undetectable once the treatments were removed (Table 1). The acetaldehyde content was much higher in CO₂-treated fruit (5.94 µg g^–1 at the end of 2 d continuous CO₂ exposure) than ethylene-treated fruit (0.79 µg g^–1 at the end of 2 d continuous ethylene exposure). Ethanol was generally detectable, although the de-astringent treatments also enhanced its accumulation several fold, and it was at its highest concentration in CO₂-treated fruit (Table 2).

Real-time PCR analysis indicated the two newly isolated DkERF genes, DkERF9 and DkERF10, were both induced by the different de-astringent treatments (Fig. 2). Expression of DkERF10 was more responsive to these treatments than DkERF9 and the other ERFs.

Fig. 2.

Transcriptional analysis of hypoxia responsive DkERF9 and DkERF10. Transcripts of DkERF9 and DkERF10 were measured by real-time PCR in response to CO₂ and ethylene treatment. Mature fruit were treated with ethylene (100 µl l^–1, open circles, 2 d) and air (control, open squares), separately. A horizontal arrow represents the end day of the treatment, when the fruit were transferred to air at 20 °C. Day 0 fruit values were set as 1. Error bars indicate SEs from three biological replicates.

Isolation of anaerobic fermentation-related structural genes

Three ADH (DkADH1–3, JX117841–JX117843) and five PDC (DkPDC1–3, JX117844–JX117846; DkPDC4, DC592775; DkPDC5, JX117847) cDNAs were isolated from ‘Mopan’ persimmon fruit (astringent type). DkADH1, DkADH2, and DkPDC2 were full-length coding sequences (CDS), while others were partial CDS.

Phylogenetic analysis of the deduced amino acid sequences showed that DkADH1 clustered close to LeADH1, LeADH5, and StADH1; DkADH2 was close to AtADH2 and LeADH4; while DkADH3 was similar to PaADH1 and CmADH2 (Fig. 3). Unlike DkADH sequences, DkPDC coding sequences could be divided into two groups, with DkPDC1, DkPDC3, and DkPDC4 in one group with homology to LePDC1 and CmPDC1; while DkPDC2 and DkPDC5 were similar to each other and closely related to VvPDC1 (Fig. 3). Tissue-specific expression analysis revealed that DkADH and DkPDC genes were differentially expressed in various persimmon tissues (see Supplementary Fig. S2 at JXB online).

Fig. 3.

Phylogenetic analysis of ADH (a) and PDC (b) genes. The deduced amino acid sequences were obtained from the NBCI and the accession numbers are indicated. At, Arabidopsis thaliana; Cm, Cucumis melo; Cs, Citrus sinensis; Fa, Fragaria ananassa; Le, Lycopersicon esculentum; Md, Malus domestica; Nt, Nicotiana tabacum; Os, Oryza sativa; Pa, Prunus armeniaca; Pc, Pyrus communis; Pp, Prunus persica; St, Solanum tuberosum. The phylogenetic trees were constructed using ClustalX (v 1.81).

Enzymatic and transcriptional response of ADH and PDC during fruit de-astringency

ADH activities were induced by both the 1 d CO₂ and 2 d ethylene treatments. The levels in the other tissues were very high but the fruit levels were lower, but still at effective levels of activity. With the CO₂ treatment, activity was induced from 1 d and reached a peak at 3 d (Fig. 4), and increased more gradually in ethylene-treated fruit (Fig. 5). Expression of DkADH1 mRNA was particularly responsive to these treatments, and paralleled the changes in enzyme activity. DkADH2 and DkADH3 expression was relatively more constant and at much lower levels. There was a stronger differential response to ethylene compared with CO₂ (Figs 4, 5).

Fig. 4.

Changes in ADH enzyme activity and mRNA amounts from specific genes in response to CO₂ treatment. Mature fruit were treated with CO₂ (~95%, v/v, open diamonds, 1 d) and air (control, open squares), separately. The horizontal arrow represents the end day of the treatment, when the fruit were transferred to air at 20 °C. Day 0 fruit values were set as 1. Black columns and white columns represent relative mRNA abundance of the gene transcripts in control and ethylene treated fruit, respectively. Error bars indicate SEs from three biological replicates.

Fig. 5.

Changes in ADH enzyme activity and mRNA amounts from specific genes in response to ethylene treatment. Mature fruit were treated with ethylene (100 µl l^–, open circles, 2 d) and air (control, open squares), separately. The horizontal arrow represents the end day of the treatment, when the fruit were transferred to air at 20 °C. Day 0 fruit values were set as 1. Black columns and white columns represent relative mRNA abundance of the gene transcripts in control and ethylene-treated fruit, respectively. Error bars indicate SEs from three biological replicates.

The PDC activities exhibited similar patterns, which increased in CO₂-treated fruit, the increase being observed from 1 d (Fig. 6), while in ethylene-treated fruit there was a 1 d delay before PDC increased (Fig. 7). Among the PDC gene family, the transcript abundance of DkPDC1 and DkPDC2 was increased by more than 50-fold (Figs 6, 7). DkPDC3 and DkPDC4 mRNA accumulation increased only in the CO₂-treated fruit and the levels were relatively lower than for DkPDC1 and DkPDC2. DkPDC5 was partially repressed by exogenous ethylene treatment (Fig. 7) The above expression patterns from the 1 d CO₂ treatment (Figs 4, 6) were confirmed in the 2 d CO₂-treated fruit (see Supplementary Fig. S3 at JXB online).

Fig. 6.

Changes in PDC enzyme activity and mRNA amounts from specific genes in response to CO₂ treatment. Mature fruit were treated with CO₂ (~95%, v/v, open diamonds, 1 d) and air (control, open squares), separately. A horizontal arrow represents the end day of the treatment, when the fruit were transferred to air at 20 °C. Day 0 fruit values were set as 1. Black columns and white columns represent relative mRNA abundance of the gene transcripts in control and ethylene-treated fruit, respectively. Error bars indicate SEs from three biological replicates.

Fig. 7.

Changes in PDC enzyme activity and mRNA amounts from specific genes in response to ethylene treatment. Mature fruit were treated with ethylene (100 µl l^–1, open circles, 2 d) and air (control, open squares), separately. A horizontal arrow represents the end day of the treatment, when the fruit were transferred to air at 20 °C. Day 0 fruit values were set as 1. Black columns and white columns represent relative mRNA abundance of the gene transcripts in control and ethylene-treated fruit, respectively. Error bars indicate SEs from three biological replicates.

PCA analysis was performed to evaluate the relations between de-astringency and expression of ADH/PDC genes, with the mean data for soluble tannin contents and levels of gene expression. The analysis showed that, for both treatments, PC1 and PC2 accounted for over 99% of the total variation, and PC1 was dominant with 99.757% and 99.754% for ethylene and CO₂ treatments, respectively (Fig. 8). As expected, de-astringency treatments showed negative correlations with soluble tannin contents. In the ethylene treatment, DkADH1, DkADH3, DkPDC1, and DkPDC2 were distributed along the negative axis of PC1, and were far from SCTs (Fig. 8). Similarly, with the CO₂ treatment, DkADH1, DkPDC1, and DkPDC2 were the most distant from SCTs (Fig. 8). As PC1 was the main contributor to the total variation, these results suggest that DkADH1, DkPDC1, and DkPDC2 are most likely to be involved in de-astringency (decrease in SCTs).

Fig. 8.

PCA analysis of the relationship between DkADH and DkPDC gene expression and persimmon fruit de-astringency. Soluble tannin contents indicate the extent of fruit de-astringency. Ethylene treatment (a) and CO₂ treatment (b). The mean data were used for PCA analysis.

Transient over-expression of DkADH1 and DkPDC2 in persimmon leaves

Persimmon is a perennial fruit and it takes years to generate transgenic fruit with stable transformation technologies, thus a transient overexpression system was chosen for quick gene functional analysis. DkADH1 and DkPDC2 constructs, driven by the CaMV 35S promoter, were introduced into Agrobacterium, which was then infiltrated into persimmon leaves. Analysis of SCTs in leaves indicated that the effects of DkADH1 and DkPDC2 in decreasing SCTs were different. The levels of SCTs in the leaves infiltrated with DkADH1 were similar to those in leaves infiltrated with empty vectors (SK, Fig. 9). Transient expression of DkDPC2, on the other hand, led to a small but significant (P <0.05) decrease in levels of SCTs in the leaves (Fig. 9).

Fig. 9.

Transient expression of DkADH1 and DkPDC2 in ‘Mopan’ persimmon leaves. Soluble tannins contents in leaves infiltrated with DkADH1 or DkPDC2, driven by the CaMV 35S promoter. SK represents empty vector. Error bars indicate SEs from ten biological replicates (*P <0.05).

Association and interaction of hypoxia-responsive ERFs and DkADH1 and DkPDC2

Linear correlation analysis was performed on gene expressions of DkERF genes and the target genes. Except for DkERF6, expressions of other DkERF genes were closely related with DkADH1 and DkPDC2 (see Supplementary Fig. S4 at JXB online).

A transient assay indicated that DkERF9 and DkERF10 acted as activators, while the other five ERF genes had less effect on DkADH1 and DkPDC2 promoters (Fig. 10). Activity of the DkPDC2 promoter was induced by interaction with DkERF9, while activity of the DkADH1 promoter was only induced by DkERF10 (Fig. 10).

Fig. 10.

In vivo interaction of hypoxia-responsive ERF gene and target promoters. 100 µl samples were infiltrated into Nicotiana benthamiana leaves. Firefly luciferase and renilla luciferase were assayed 3 d after infiltration. For each TF-promoter interaction, three independent experiments were performed (at least six replicates in each experiment). The ratio of LUC/REN of the empty vector (SK) plus promoter was used as calibrator (set as 1). Error bars indicate SEs from six biological replicates.

Both CO₂ and ethylene treatments accelerate persimmon fruit de-astringency

By utilizing the astringent type persimmon ‘Mopan’, two de-astringency treatments, CO₂ and ethylene, were compared. The results showed that CO₂ was more effective than ethylene in inducing de-astringency, indicated by the decrease in SCTs (Figs 3, 34) and the increase in acetaldehyde formation (Tables 1, 12), which is in agreement with previous reports (Arnal and Del Río, 2003; Pesis and Ben-Arie, 2006; Salvador et al., 2007). This is consistent with acetaldehyde being the main compound involved in the insolubilization of SCTs.

Isolation of hypoxia-responsive ERF genes from persimmon

In model plants, HRE and RAP type ERF genes were recently shown to be the key regulators in plant responses to hypoxia (Gibbs et al., 2011; Hinz et al., 2010; Licausi et al., 2010, 2011a; Sasidharan and Mustroph, 2011; Yang et al., 2011). Comparative analysis of the predicted amino acid sequences encoded by the HRE and RAP genes revealed that all of the five proteins had conserved N-ends, initiated by the MC domain (MCGGAII), which has recently been shown to be involved in controlling the stability of regulators of the plant oxygen-sensing mechanism (Gibbs et al., 2011; Licausi et al., 2011a; Sasidharan and Mustroph, 2011). In persimmon, four ERF genes, DkERF1, DkERF4, DkERF5, and DkERF6 were characterized as inducible by both low oxygen and ethylene, and proposed to be de-astringency-related in our previous study (Yin et al., 2012). By utilizing RNA-Seq methodology, two additional novel hypoxia (high CO₂)-responsive genes, DkERF9 and DkERF10, were isolated. However, only DkERF10, encoded the conserved MC domain. Phylogenetic analysis also indicated that DkERF10 was closely related to HRE and RAP genes, and, like them, belongs to subfamily VII. Thus, the phylogenetic and domain analyses indicated that DkERF10 had a high probability of being functionally involved in low oxygen sensing.

Real-time PCR indicated that all six DkERF genes were responsive to hypoxia/high CO₂. Of these six, DkERF10 was the most responsive to the de-astringent treatment, especially to CO₂ treatment (greater than a 2000-fold increase, Fig. 2). Moreover, some of hypoxia/high CO₂-responsive ERF genes were also modulated by ethylene, with different patterns. DkERF4, 5, 9, and 10, were responsive to both ethylene and CO₂ (present results and Yin et al., 2012b), while DkERF1 and 6 only seemed to respond to CO₂. This observation is similar to that for the hypoxia-response ERF genes in Arabidopsis, where HRE1 is both responsive to ethylene and hypoxia (Hinz et al., 2010; Yang et al., 2011), while Rap2.12 is not regulated by 1-aminocyclopropane-1-carboxylic acid (ACC) (Licausi et al., 2011b).

Characterization of persimmon structural genes related to anaerobic fermentation

As mentioned previously, acetaldehyde is one of the main compounds that render the SCTs insoluble and cause de-astringency. Synthesis of acetaldehyde is generally catalysed by PDC, which converts pyruvate to acetaldehyde. ADH is then involved in the potentially reversible interconversion of acetaldehyde and ethanol (Strommer, 2011). In persimmon fruit, both PDC and ADH have been proposed as candidates involved in acetaldehyde formation (Yamada et al., 2002). In plants, PDC has been associated mostly with acetaldehyde formation (Nguyen et al., 2009), whereas ADH is best known for its role in the anaerobic response (Strommer, 2011), although it has also been linked in fruit to the regulation of the formation of volatile six-carbon alcohols (Speirs et al., 1998; Zhang et al., 2010). Phylogenetic analysis of persimmon DkADH and DkPDC genes revealed that the isolated genes were clustered with different homologues from Arabidopsis, tomato, tobacco, etc (Fig. 1). However, it is not possible to predict gene function simply based on phylogenetic tree analysis, particularly as the genes with high similarity were found to have different expression patterns in tissue-specific analysis (e.g. DkPDC3 and DkPDC4 were similar in sequence, but differentially expressed in various tissues).

The expression patterns of DkADH and DkPDC genes were monitored during de-astringency. The up-regulation by ethylene of DkADH1,3 and DkPDC1,2 is similar to the response of ADH and PDC genes to ethylene in other fruit, e.g. apricot PaADH (EU395433), and melon CmADH1 (DQ288986) and CmADH2 (DQ288987) (Manríquez et al., 2006; González-Agüero et al., 2009). During CO₂ treatment, no matter whether for 1 d or 2 d durations (Fig. 6; see Supplementary Fig. S2 at JXB online), DkADH1, DkPDC1, DkPDC2, and DkPDC3 were transiently induced, similar to tomato LeADH2 (X77233) and tobacco TobADH1 (X81853) and TobPDC1 (X81854) genes (Longhurst et al., 1994; Bucher et al., 1995). The expression patterns of these genes have also been shown in previous reports, to coincide with the increase in related enzyme activities (Tamura et al., 1999). DkADH1, DkPDC1, and DkPDC2 were substantially more responsive to both persimmon de-astringency treatments than the other gene family members. Furthermore, the increase in expression of DkADH1, DkPDC1, and DkPDC2 occurred in parallel with the changes in ADH and PDC enzyme activities (Figs 4, 45, 46, 47), suggesting that they might be the key genes involved in persimmon de-astringency, and this suggestion is supported by the PCA results (Fig. 8). Transient expression of DkADH1 and DkPDC2 in persimmon leaves had different effects on leaf SCTs contents. Leaves infiltrated with DkPDC2 exhibited a small, but significant reduction in levels of SCTs, while DkADH1 had no affect (Fig. 9). Although further functional analysis needs to be performed with stable transformation systems, all the present results strongly indicate that DkADH1, DkPDC1, and DkPDC2 should be considered as the most important candidates for persimmon fruit de-astringency.

DkERF genes involved in the transcriptional regulation of persimmon DkADH1 and DkPDC2 and in relation to de-astringency

Transcription factors (TFs) act as the switches for regulating target genes (Riechmann et al., 2000). In model plants, expression of ADH and PDC genes is modulated by novel hypoxia-responsive ERF genes. Our results indicate that, with the exception of DkERF6 (see Supplementary Fig. S4 at JXB online), changes in transcript abundance of DkERF genes and the target genes were closely correlated. Using the transient assay, gene–gene screening indicated that only DkERF9 and DkERF10 are involved in regulating the promoters of DkPDC2 and DkADH1 target genes (Fig. 10). These results were similar to the finding obtained by RNAi methodology that Arabidopsis HRE1 is an activator of ADH1, PDC1, and PDC2 (Yang et al., 2011). Thus, DkERF9 and DkERF10 might act as the major regulators of persimmon fruit low oxygen signalling, and thus be the key regulator genes in causing de-astringency.

Interestingly, the present results indicate that two ERF genes act as switches for different targets. DkERF10 interacted with the DkADH1 promoter, while DkERF9 only induced activation of the DkDPC2 promoter. These results suggest that different ERF genes may have different targets, which is in contrast to the observations for HRE1 in Arabidopsis. This indicates that these two DkERF genes have different regulatory roles in persimmon de-astringency.

DkERF10 is a homologue of Arabidopsis HRE and RAP genes, and belongs to subfamily VII. Thus, it is not surprising that DkERF10 was characterized as a regulator of persimmon fruit hypoxia response and de-astringency. On the other hand, DkERF9 belongs to subfamily IV, which comprises the DREB subfamily. DREB type ERF genes are the main regulators of plant responses to abiotic stresses, such as drought and heat (Lata and Prasad, 2011). However, the roles of many DREB-type ERF genes have not yet been elucidated. Based on the present results, in addition to members of subfamily VII, subfamily IV ERF genes may also be involved in plant hypoxia responses.

In conclusion, both CO₂ and ethylene treatments were effective in reducing ‘Mopan’ persimmon fruit astringency, by promoting the increase of acetaldehyde, ADH and PDC activities and the decrease of SCTs. Three genes, DkADH1, DkPDC1, and DkPDC2, are the most likely candidates for increasing acetaldehyde concentrations and causing de-astringency, and the role of DkPDC2 in decreasing SCTs was functionally confirmed. Two hypoxia-responsive ERF genes, DkERF9 and DkERF10, are involved in the regulation of the DkPDC2 and DkADH1 promoters, respectively. Based on the present findings, we propose that both of DkERF9 and DkERF10 are involved in the persimmon fruit anoxia response and de-astringency and increase acetaldehyde accumulation by modulating different target genes. As there are at least three hypoxia responsive ERF genes in Arabidopsis (Hinz et al., 2010; Licausi et al., 2010), additional hypoxia-responsive ERF genes may exist in persimmon fruit. Furthermore, the effects high CO₂ and ethylene on the DkERF transcriptional regulation of DkADH and DkPDC genes need to be investigated further.