Apple SEPALLATA1/2-like genes control fruit flesh development and ripening


For correspondence (e-mail;


Flowering plants utilize different floral structures to develop flesh tissue in fruits. Here we show that suppression of the homeologous SEPALLATA1/2-like genes MADS8 and MADS9 in the fleshy fruit apple (Malus x domestica) leads to sepaloid petals and greatly reduced fruit flesh. Immunolabelling of cell-wall epitopes and differential staining showed that the developing hypanthium (from which the apple flesh develops) of MADS8/9-suppressed apple flowers lacks a tissue layer, and the remaining flesh tissue of fully developed apples has considerably smaller cells. From these observations, it is proposed that MADS8 and MADS9 control the development of discrete zones within the hypanthium tissue, and therefore fruit flesh, and also act as foundations for development of different floral organs. At fruit maturity, the MADS8/9-suppressed apples do not ripen in terms of both developmentally controlled ripening characters, such as starch degradation, and ethylene-modulated ripening traits. Transient assays suggest that, like the RIN gene in tomato, the MADS9 gene acts as a transcriptional activator of the ethylene biosynthesis enzyme, 1-aminocyclopropane-1-carboxylate (ACC) synthase 1. The existence of a single class of genes that regulate both flesh formation and ripening provides an evolutionary tool for controlling two critical aspects of fleshy fruit development.


Flowering plants have evolved various fruit types to act as attractants to seed-dispersing organisms. Fleshy fruits are typically derived from floral tissues, which are arranged in whorls of organs comprising sepals, petals, stamens and carpels (Coen and Meyerowitz, 1991). The structure of the fruit is determined by the anatomy of the flower and the positioning of the ovary (Smyth, 2005). While most fruits, such as tomato (Solanum lycopersicum), develop from ovary tissue found in the carpel (true fruit), a small number of fruits, including fruits from the Rosaceae family, such as apple (Malus x domestica) and strawberry (Fragaria x ananassa Duch.), develop from extra-carpellary tissues (accessory fruit; Figure 1). In strawberry, the flesh tissue develops from the receptacle (i.e. is stem-derived), and the true fruit are the achenes. In apple, the flesh is mostly derived from the hypanthium (the tissue positioned beneath the sepals, petals and stamens that surrounds the carpels).

Figure 1.

The floral origin of fleshy fruit.

Flesh from the true fruit tomato is derived solely from ovary tissue (purple) contained within the carpel. Flesh from the accessory fruit apple is derived from the hypanthium (grey). Flesh from the accessory fruit strawberry is derived from the receptacle (yellow).

The development of floral organs relies on correct expression of the ABCE floral homeotic genes (Causier et al., 2010), the majority of which encode MADS box transcription factors. It is proposed that the SEPALLATA (SEP) genes (E function genes in the ABCE flowering model) act as molecular platforms that allow ABC function genes to determine the identity of floral organs in each of the four floral whorls (Jack, 2001). Accordingly, it has been hypothesized that SEP genes played an essential role in the origin of the flower (Zahn et al., 2005). There are four SEP genes in Arabidopsis thaliana. The closely related SEP1 and SEP2 genes and the SEP3 gene are flower-specific, while SEP4 is expressed in all aerial organs and at various developmental stages (Ma et al., 1991; Flanagan and Ma, 1994; Huang et al., 1995; Savidge et al., 1995; Mandel and Yanofsky, 1998). SEP genes show partial redundancy, as single SEP mutants show subtle phenotypes, whereas the triple sep1/2/3 mutant has whorls of sepals and the quadruple mutant develops leaf-like organs in all four whorls (Pelaz et al., 2000; Ditta et al., 2004). Over-expression of SEP genes highlights their role as floral meristem identity genes, with CaMV35S:SEP4 plants exhibiting replacement of the inflorescence meristem with terminal flowers (Ditta et al., 2004) and CaMV35S:SEP3 plants showing early flowering with solitary flowers (Pelaz et al., 2001).

SEP-like genes have been implicated in development of the fleshy fruit of tomato, which is ovary-derived. Transgenic tomato plants in which either SEP1/2-like TM29 or SEP3-like TM5 are suppressed exhibit parthenocarpic fruit development (Pnueli et al., 1994; Ampomah-Dwamena et al., 2002), which suggests a role for SEP-like genes in the repression of tomato fruit development. Also, the SEP4-like SlMADS-RIN (RIN) gene in tomato has exhibited a functional role in fruit ripening and is considered to be a master regulator of this process (Vrebalov et al., 2002). The ripening programme of climacteric fruit, such as tomato, is largely dependent on the presence and abundance of the ripening hormone ethylene. Master regulators such as RIN are proposed to act upstream and independently of ethylene-mediated ripening regulation. All ripening traits are inhibited in rin mutant fruit, including auto-catalytic ethylene production, carotenoid accumulation, softening and volatile production, and the fruits do not ripen in response to ethylene (Vrebalov et al., 2002). Furthermore, expression of genes for auto-catalytic ethylene biosynthesis [namely ACC SYNTHASE 2 (SlACS2) and SlACS4] is suppressed in the rin mutant (Barry et al., 2000), and it has been shown that this suppression is due, at least in part, to RIN binding to and activating the promoters of these genes (Ito et al., 2008; Fujisawa et al., 2011; Martel et al., 2011).

Characterization of SEP genes in other plants has shown a high degree of functional conservation. Seymour et al. (2011) showed that, in the non-climacteric fruit strawberry (Fragaria x ananassa Duch.), suppression of SEP1/2-like FaMADS9 resulted in increased green coloration of petals, similar to the sepalloid petals of sep3 single mutants in Arabidopsis (Pelaz et al., 2001), and ripening was delayed with respect to anthocyanin accumulation, de-greening of achenes and softening, similar to the SEP4-like rin mutant of tomato. Expression of the closest homologue of the RIN gene, SEP4-like FaMADS4, was unaffected in these fruit. It is also of note that, in the most severely affected lines, fruit development did not progress past the immature stage. In banana (Musa acuminata), three SEP3-like genes (MaMADS1, 2 and 4) are highly expressed in pulp (where ethylene production and ethylene-regulated ripening begin), and expression correlates with the climacteric ethylene peak and ripening in general (Elitzur et al., 2010). In peach (Prunus persica), independent over-expression of the SEP3-like gene PrpMADS5 and the SEP1/2-like gene PrpMADS7 in Arabidopsis resulted in early flowering phenotypes (Xu et al., 2008), similar to over-expression of the native genes in Arabidopsis.

In apple, the sequence and expression patterns of a number of SEP-like MADS box genes have been described previously. SEP1/2-like MADS1/MADS8 and MADS9 show high expression from 2 days to 1 week after pollination, MADS3/MADS7 and MADS6 are similar to the petunia (Petunia hybrida) SEP genes FLORAL BINDING PROTEIN9 (PhFBP9) and PhFBP23 (which form a separate clade to SEP1/2, SEP3 and SEP4) and show high expression from 4 to 8 weeks after pollination, and lastly the SEP4-like MADS4 gene is strongly expressed in young fruits (Sung and An, 1997; Yao et al., 1999; Sung et al., 2000). These results suggest that apple MADS box genes have a role in early fruit development. Expression of apple SEP-like genes has not been tested during ripening, and their role in the control of this process has not been determined. Like tomato, ethylene regulates many ripening-related processes in apple; however, some traits progress in the absence of ethylene, albeit slowly (Johnston et al., 2009). The model proposed by Johnston et al. (2009) predicts that early and mid-stage ripening events are controlled by a combination of both developmental and ethylene-mediated regulation. It is feasible that the developmental component of ripening regulation in apple may be mediated by a SEP gene, as conservation of the ripening regulation function within the SEP clade is observed in tomato and strawberry (Vrebalov et al., 2002; Malcomber and Kellogg, 2005; Seymour et al., 2011). Using apple lines with antisense-suppressed SEP-like genes, this study focuses on the role of apple SEP-like genes in both fleshy fruit development and fruit ripening.


Phylogenetic relationships of the apple SEPALLATA-like MADS box genes

To investigate the role of SEP-like genes in apple, the recently published draft apple genome (Velasco et al., 2010) was mined for sequences homologous to the four Arabidopsis SEP genes and the petunia FBP9 gene. It was found that apple contains eight SEP-like genes distributed across six chromosomes (Table S1), representing all sub-clades, with all but two of the predicted genes in the SEP clade already having been described (either published or having had their sequence deposited in GenBank), and with some alleles having duplicate names in the literature. Phylogenetic analysis of the predicted protein sequences with SEP-related proteins from other species identified two closely related homeologous apple genes in all SEP sub-clades (Figure 2a and Table S2). In the SEP1/2 clade, MADS8 (also referred to as MADS1; Sung and An, 1997) and MADS9 clustered with the ripening-related FaMADS9 gene (Seymour et al., 2011). MADS4 (with its putative splice-variant MADS24) and MADS104 clustered within the SEP4 clade, which also contained the tomato gene RIN (Vrebalov et al., 2002). The SEP3 clade contained MADS18 and its putative homeologue MADS118, while MADS6 and MADS7 (also referred to as MADS3; Sung et al., 2000) clustered within the PhFBP9 clade (Figure 2a).

Figure 2.

SEPALLATA genes in apple.

(a) Phylogenetic relationship of apple SEP-like genes with selected MADS box genes from other species, showing the four SEP clades. Full-length protein sequences from apple (Malus x domestica, Md), Arabidopsis (Arabidopsis thaliana, At), petunia (Petunia hybrida, Ph), tomato (Solanum lycopersicum, Sl), peach (Prunus persica, Prp), strawberry (Fragaria x ananassa, Fa), grape (Vitis vinifera, Vv) and banana (Musa acuminata, Ma) were used to construct the tree. Bootstrap values are shown on branches. Scale indicates 0.2 substitutions per site.

(b) Expression of apple MADS box genes during fruit development was measured by quantitative RT-PCR relative to ACTIN. Values are means ± SD (= 4).

Expression of the eight SEP-like apple genes was measured during apple fruit development (Figure 2b). The SEP1/2-like genes MADS8 and 9 showed similar patterns of expression during fruit development, starting high then decreasing in the early stages of development (0–35 days after full bloom), consistent with results obtained by Sung and An (1997) and Yao et al. (1999), followed by an increase on fruit maturation. Expression profiles of the PhFBP9/23-like genes MADS6 and 7 showed low expression early in fruit development, followed by a gradual increase during the latter stages. The SEP3-like gene MADS18 showed high expression at 0 days after full bloom, which decreased during fruit development, while its homeologue MADS118 showed a peak during the initial stages of cell expansion (25 days after full bloom) followed by a decrease in expression. The SEP4-like genes MADS4 and MADS104 showed a consistent level of expression throughout fruit development, with a small increase at ripening.

The effect of MADS8 suppression on flower development

Expression analysis of the apple SEP1/2-like genes and their phylogenetic clustering with the ripening-related FaMADS9 gene suggested that MADS8/9 may play a role in fruit development. To investigate this further, five independent lines of transgenic ‘Royal Gala’ were generated to show suppressed expression of MADS8. Each line contained the CaMV 35S promoter upstream of the full-length MADS8 cDNA in antisense orientation (MADS8as). The five sequence-verified MADS8as lines (lines 2, 5, 9, 18 and 27) all showed normal vegetative development, but exhibited varying degrees of altered phenotype at flowering and during fruit development. The two most severe lines, MADS8as-5 and MADS8as-9, displayed small green petals, similar to sepals (Figure 3a–c), while MADS8as-2 and MADS8as-18 showed a weaker phenotype, with green petals of normal size (Table 1). MADS8as-27 flowers were the same as controls (Table 1). As apple fruit develops from the hypanthium and fruit development was also strongly affected in MADS8as lines, the ultrastructure of the hypanthium was examined using immunolabelling and differential staining microscopy. For immunolabelling, a range of antibodies were tested that were specific to various cell-wall epitopes; among these, one antibody (LM5, specific to (1→4)-β-d-galactan residues in pectin; Jones et al., 1997) showed differential labelling between wild-type ‘Royal Gala’ and MADS8as-9. In wild-type, regions of high fluorescent signal were observed in the epidermis, ovary tissues and inner hypanthium, while a region of low fluorescent signal was observed in the outer hypanthium (Figure 3f). In MADS8as-9, the fluorescent signal was more uniform across the whole tissue, with few or no changes in signal intensity (Figure 3g). Using differential staining with toluidine blue, it was observed that the outer hypanthium of wild-type flowers stained pink (Figure 3d), but this pink-staining was absent in MADS8as-9 flowers (Figure 3e). There were also apparent anatomical changes in the hypanthium, with increased vascular bundles in MADS8as-9 and a change in locule shape from tear drop to a more open triangular shape (Figure 3e).

Table 1. Summary of the physiology of MADS8as lines
MADS8as lineFlower phenotypeFruit morphologyInternal ethylene concentration (μl L−1)Ripening effectsResponse to ethylene
5Sepalloid petals: shape, size and colourNo fruit

Sepalloid petals: shape, size and colour

Altered cell-wall composition of hypanthium tissue layer

Strongly reduced cortex layer

No expansion of cortical parenchyma cells

Approximately 0.1Starch clearance, skin colour change and volatile production all inhibited

Starch and volatile production responded to ethylene

No change in skin colour

2Green petals, extra locules

Moderately reduced cortex layer

Some expansion of cortical parenchyma cells


Skin colour change inhibited

Normal starch clearance

No significant change in skin colour
18Green petals

Moderately reduced cortex layer

Near wild-type expansion of cortical parenchyma cells

Approximately 100

Skin colour change inhibited

Near wild-type for all other ripening parameters

Full response of all ripening traits except skin colour change
27White petals, flowers normal

Near wild-type cortex layer

Near wild-type expansion of cortical parenchyma cells

Approximately 100

Some change in skin colour

Near wild-type for all other ripening parameters

No additional ripening progression when ethylene added
Figure 3.

Phenotype and gene expression of MADS8as-9-suppressed flowers.

(a–c) Flowers of wild-type (a) MADS8as-9 (b) and MADS8as-5 (c) at full-bloom.

(d, e) Cross-sections of wild-type (d) and MADS8as-9 (e) hypanthia stained with toluidine blue. Scale bar = 1 mm.

(f, g) Longitudinal sections of wild-type (f) and MADS8as-9 (g) hypanthia hybridized with LM5 antibodies. Scale bar = 100 μm. hy, hypanthium, ov, ovary, ep, epidermis.

(h–k) Floral development series for wild-type (h, j) and MADS8as-9 (i, k) at balloon stage (h, i) and open flower stage (j, k) showing sepaloid petals in MADS8as-9. f, floral tissue; h, hypanthium tissue.

(l) Expression analysis (l) of apple SEP-like genes in floral and hypanthium tissue at balloon stage, open flower stage (open), 2 days after pollination (DAP) and 8 days after pollination (DAP) in wild-type (WT) and MADS8as-9 flowers. Expression was measured by quantitative RT-PCR relative to ACTIN. Values are means ± SD (= 4).

Characterization of apple SEP-like genes in early fruit development

It has been shown that MADS8 and MADS9 have a high level of expression following fertilization, consistent with a role in early fruit development (Yao et al., 1999). To examine this further, the level of expression of apple SEP-like genes was measured in hypanthium and floral tissues of MADS8as-9 and wild-type flowers. The expression pattern of MADS8/9 in wild-type flowers before and after pollination was consistent with the findings of Yao et al. (1999). In addition, MADS8 exhibited predominant expression in the hypanthium 2 days after pollination, consistent with a role in early fruit development (Figure 3l). Of the other SEP-like genes, MADS6 and MADS104 showed strong and predominant expression in the hypanthium, like MADS8, while the remaining SEP-like genes were expressed in both floral and hypanthium tissues. In the MADS8as-9 line, expression of both MADS8 and MADS9 was strongly suppressed at all time points and in all tissues compared to wild-type, and partial suppression of MADS7 was also observed (Figure 3l).

Effect of antisense suppression on closely related genes

Because the antisense construct contained the complete MADS8 cDNA sequence, it is possible that closely related genes may be affected, especially as gene suppression by RNAi requires only a 21–23 bp region of similarity. To address this, we first compared the cDNA sequence of MADS8 with the DNA sequences of the predicted apple gene models (Velasco et al., 2010). After the gene models for the closely related genes MADS8 and MADS9, the next most similar gene model encodes the MADS7 gene (with 85 and 81% identity over two regions of 228 and 116 bp, respectively), followed by MADS118, MADS104 and the AtAGL6-like gene MADS11 (Table S3a). Expression analysis of these genes, and the remaining apple SEP-like genes, showed that MADS7 was the only gene with consistent down-regulation; however, this down-regulation (2–5-fold) was less than that observed for MADS8 and MADS9 (4–15-fold; Figure 3l and Figure S2a). Some of the other MADS genes with lower homology showed down-regulation but only at individual time points, such as MADS6, which is down-regulated only after pollination, inconsistent with an RNAi suppression phenotype. The MADS8 cDNA sequence was then compared to the whole apple genome sequence. In total, ten genomic loci were identified that showed significant similarity to the MADS8 cDNA sequence (defined as >21 bp in length with 100% identity). Nine loci were predicted to encode MADS box transcription factors (MADS8, MADS9, MADS11, MADS4, MADS118 and MADS18, and the gene models MDP0000156783, MDP0000254052 and MDP0000209705), while the 10th locus had no associated gene models or EST support (Table S3b). Expression analysis of the additional genes identified showed that they were unaffected by the antisense construct (Figure S2a). These data strongly suggest that MADS8 and MADS9 and possibly MADS7 are the only targets of the antisense construct.

Effect of MADS8 suppression on fruit development

All flowers were pollinated to assess whether the apples containing the antisense construct were affected in fruit development and ripening. At fruit maturity, four of the five lines had smaller apple fruit compared to wild-type. When the fruit were cut open, the size of the core was similar to wild-type but there were varying degrees of reduction of cortex tissue, with two lines (MADS8as-5 and MADS8as-9) showing virtually no cortex tissue (Figure 4a,b and Figure S1). The line with the most severe phenotype, MADS8as-5, demonstrated poor fruit set, with only one or two fruits developing each year from more than 50 pollinated flowers. Sufficient numbers of fruits were obtained from the other MADS8as lines, with MADS8as-9 showing the most strongly reduced fruit size compared with the control. Petal abscission from fruit, which normally occurs following pollination, was reduced in MADS8as-5 and MADS8as-9 (Figure 4a,b). Sections from mature wild-type and MADS8as-9 apples showed that the remaining flesh tissue in MADS8as-9 apples had smaller, more densely packed cells compared to those in wild-type cortex tissue (Figure 4c,d). This result, combined with the immunolabelling results (Figure 3f,g), suggests that the reduced cortex in the MADS8as-9 line results from a combination of loss of tissue type and lack of cell expansion in the cortex zone.

Figure 4.

Phenotype and gene expression of MADS8as-9-suppressed fruit. (a, b) Fruit of wild-type (a) and three extreme MADS8as lines (b; left to right: MADS8as-2, MADS8as-9and MADS8as-5) in longitudinal and cross-section. Scale bar = 1 cm.

(c,d) Sections of cortex tissue of wild-type (c) and remnant flesh region of MADS8as-9 (d) stained with Safranin-Fast Green. Scale bar = 500 μm. The inset shows the same tissue at higher magnification: scale bar = 100 μm.

(e) Tissue-specific expression of apple SEP-like genes in three zones (skin, cortex and core) of ripening apples stored for 4 days without added ethylene, measured by quantitative RT-PCR relative to ACTIN. Values are means ± SD (= 4). Asterisks indicate that cortex tissue is unavailable for this line.

Expression of SEP-like genes was measured in three apple tissue zones (skin, cortex and core) of mature fruit from wild-type and each of the transgenic lines. Due to the strong reduction in fruit size, a cortex tissue layer could not be collected for lines MADS8as-9, MADS8as-2 and MADS8as-18. In wild-type apples, MADS8 and MADS9 showed higher expression in the cortex and core compared to the skin (Figure 4e). In MADS8as-9 apples, there was consistent and strong down-regulation of MADS8, MADS9 and MADS7 in all tissues examined, and partial down-regulation of MADS6, but MADS4, MADS104, MADS18 and MADS118 showed decreases in expression only in some tissues at certain time points. Expression of the antisense construct in mature fruit was high in MADS8as-9, MADS8as-2 and MADS8as-18, with MADS8as-9 exhibiting the highest level of expression (Figure S3a), correlating with the degree of severity of the phenotype. Like MADS8as-9, lines MADS8as-2 and MADS8as-18 showed down-regulation of MADS8, MADS9, MADS6 and MADS7 in all tissues in fruits; however, MADS8as-27 did not show consistent down-regulation of any MADS genes across all tissues (Figure 4e and Figure S3b).

Fruit ripening in MADS8 antisense lines

In all five MADS8 antisense lines, there was a range of fruit-ripening effects. The most severe lines, MADS8as-5 and MADS8as-9, showed no ripening progression. MADS8as-2 and MADS8as-18 were delayed and MADS8as-27 showed a reduced response to ethylene. A detailed study of various aspects of ripening was undertaken using fruit representing each ripening category and compared to wild-type fruit (untransformed ‘Royal Gala’). The internal ethylene concentration in MADS8as-9 apples was low but detectable, and, consistent with this, all aspects of ripening were strongly inhibited (Figure 5 and Figure S4). Fruit from the intermediate line MADS8as-2 exhibited a delay in starch clearance in accordance with the low internal ethylene concentration, and the background skin colour change was inhibited (Figure 4). Similarly, MADS8as-18 fruit showed inhibited colour change but other aspects of ripening matched controls, as did the internal ethylene concentration in this line (Figure S4). In contrast, MADS8as-27 fruit had a normal internal ethylene concentration and all ripening traits matched controls with the exception of the background skin colour change, which showed partial inhibition (Figure 5).

Figure 5.

Physiological characterization of MADS8as-9 and MADS8as-27 fruit compared to wild-type and ethylene biosynthesis-suppressed (ACO1as) fruit.

Assessments were performed on mature fruit stored for 2 weeks in the presence (dark bars) or absence (light bars) of ethylene (100 μl L−1). The internal ethylene concentration was measured in the core cavity of the fruits. Results for starch clearance, soluble solids concentration, background skin colour change, aroma volatiles and firmness are shown. Eight fruit were sampled for MADS8as-9, wild-type and ACO1as, and five fruit were sampled for MADS8as-27.

To test whether fruit from the MADS8as transgenic plants were competent to ripen, mature apples (130 days after full bloom) were harvested and stored at 20°C for 2 weeks in the presence of 100 μl L−1 ethylene. As a ripening control, fruit from ethylene biosynthesis-suppressed lines (ACO1as; Schaffer et al., 2007) grown under the same conditions were also assessed. Treatment with ethylene was able to induce starch clearance and/or volatile production in lines MADS8as-9 (Figure 5), MADS8as-2 and MADS8as-18 (Figure S4). However, in all MADS8as lines, the change in background skin colour showed limited or no induction by ethylene (Figure 5 and Figure S4). Phenotypes of all MADS8as lines are summarized in Table 1.

As fruit ripening was severely disrupted in the MADS8as lines, we investigated the expression patterns of known ripening-related genes in apple. In the absence of added ethylene, the ethylene biosynthesis genes ACC SYNTHASE 1 (ACS1) and ACC OXIDASE 1 (ACO1) were strongly suppressed in MADS8as-9 fruit: expression of ACS1 was undetectable and that of ACO1 was reduced more than 1000-fold (Figure 6). Expression of the cell-wall hydrolase gene POLYGALACTURONASE1 (PG1) was also reduced more than 1000-fold, while expression of the volatile-related gene αFS (encoding α-farnesene synthase) was reduced 350-fold (Figure 6). Following ethylene treatment, expression of all four ripening marker genes was induced in all lines, suggesting that ethylene-regulated transcriptional pathways are functional (Figure 6 and Figure S5).

Figure 6.

Tissue-specific expression of ripening marker genes in MADS8as lines and wild-type.

Expression was measured in three zones (skin, cortex and core) of ripening apples stored for 4 days with and without added ethylene using quantitative RT-PCR relative to ACTIN. Values are means ± SD (= 4). Asterisks indicate that cortex tissue is unavailable for this line.

Transactivation of the ACS1 promoter by MADS8 and MADS9

Studies in tomato have demonstrated direct binding of RIN to the SlACS2 promoter. Based on expression profiles and responses to ethylene, the apple ethylene biosynthetic genes ACS1 and ACO1 were selected to test for transactivation by MADS8 and MADS9. For each promoter, a 2–6 kb sequence upstream of the ATG codon was cloned into a firefly luciferase/Renilla luciferase transient assay vector (Hellens et al., 2005). The resulting constructs were then co-infiltrated into Nicotiana benthamiana leaves together with constructs encoding the MADS box transcription factors MADS8, MADS9 and MADS6, driven by the CaMV 35S promoter, both singly and in combination. The promoter of another ripening-associated gene, the cell-wall hydrolase gene PG1 (Tacken et al., 2010), was also tested (Figure 7). Transactivation of between two- and threefold compared to the empty vector control was observed for all three promoters when combined with the MADS9 transcription factor. Additionally, a 1.5–2-fold transactivation of the ACO1 and PG1 promoters was observed with the MADS8 transcription factor. MADS6 did not transactivate either of the ethylene biosynthesis promoters, but a 1.5–2-fold transactivation of the PG1 promoter was observed (Figure 7).

Figure 7.

Promoter transactivation is represented as a ratio of firefly luciferase (LUC) to Renilla luciferase (REN) signals, and normalized to the negative control (empty pHEX2 vector).

MADS box transcription factors were infiltrated singly and in pairs to identify potential dimeric activity. Values are means ± SEM of four independent assays, each comprising four technical replicates.


Although fleshy fruit development has appeared a number of times in evolution, there is very little molecular data to explain how plants have acquired this character (Rohrer et al., 1991; Knapp, 2002). By studying plant species other than traditional model plants that bear ovary-derived fruits, our research has identified an additional role for the SEP class of genes. Expression of an antisense construct for the apple MADS8 gene (a SEP1/2-like gene) caused a loss of flesh tissue and a delay in ripening. In these lines, the phenotype was caused by down-regulation of three closely related SEP-like genes (MADS8, MADS9 and MADS7). In addition, a less homologous MADS box gene (MADS6), which, by sequence similarity, should be unaffected by the transgene, was also down-regulated. This down-regulation was only observed after pollination, suggesting that this was a consequence of the phenotype (i.e. a lack of specific tissue development in the hypanthium) rather than direct silencing per se. By abolishing flesh production through down-regulation of genes in the SEP clade, we provide further insight into how fruit flesh development has evolved.

Mostly in Arabidopsis, SEP genes have been shown to play a role in providing a platform or ‘floral context’ for the regulation of floral organ development through formation of tetrameric complexes with ABC function genes (Causier et al., 2010). In tomato and strawberry, SEP genes have been shown to be key controllers of fruit ripening (Vrebalov et al., 2002; Seymour et al., 2011). In strawberry, strong suppression of FaMADS9, the gene homologous to apple MADS8, leads to repression of receptacle development, similar to the loss of flesh observed in this study (Seymour et al., 2011). Apples and strawberries are accessory fruit derived from extra-capellary tissues (Figure 1). Strawberries bear dry, ovary-derived fruit (achenes) on a fleshy receptacle, whereas apple flesh is derived from a swollen hypanthium. Taken together with the results obtained by Seymour et al. (2011), our results suggest that the SEP1/2 genes play a common role in regulating accessory flesh production. This is further supported by the observations that ovary-derived tissues in MADS8as fruit are not as affected by the antisense construct, and that a reduction in flesh tissue is not observed when this class of gene is suppressed in tomato (Ampomah-Dwamena et al., 2002). To date, the best-characterized gene affecting fleshy fruit development in ovary-derived tissue is the MADS box AGAMOUS-like gene TOMATO AGAMOUS-LIKE 1 (TAGL1). When this gene was suppressed, there was a reduction in pericarp tissue, but the tissue that remained maintained its fleshy character (Vrebalov et al., 2009). It has also been shown that TAGL1 plays a role in determining flesh identity, as over-expression of TAGL1 or the peach homologue (PpPLENA) in tomato causes the floral sepals to become fleshy (Tadiello et al., 2009). However, it should be noted that, despite constitutive expression of the genes, only the sepal tissues became fleshy, which suggests that other key factors are required to determine flesh identity, and these are present in the tomato sepals. The closest apple genes to TAGL1 (MADS14 and MADS15) lie outside the SEP clade (Figure 2), and their expression is unaffected in the MADS8as-9 line (Figure S2). The only fleshless mutant reported so far is fleshless berry (flb) in grape, which has lost the capacity to make the cell layers that are normally required for the formation of the flesh (Fernandez et al., 2006). The identity of the gene affected in this mutant is unknown (Fernandez et al., 2007).

Thus, plants utilize different floral structures to develop flesh tissue in fruits. In apples, the nature of the hypanthium tissue has been the subject of vigorous debate, with two theories proposed based on vascular anatomy and other morphological evidence (Rohrer et al., 1991). The receptacular theory proposes stem-derived origins, such that a fleshy receptacle surrounds the true fruit, while the appendicular theory proposes leaf-derived origins, such that the adnate bases of sepals, petals and stamens (floral tube) encase the true fruit (Pratt, 1988; Rohrer et al., 1991). While this may be considered an outdated discussion based on our current understanding of molecular biology, the data from this study suggest that flesh formation in apples is controlled in a similar way to the receptacle-derived accessory fruit strawberry, which would support the receptacular theory. In addition, if the hypanthium were floral tube-derived, one would expect a more severe floral phenotype in the MADS8as lines. Furthermore, when the floral homeotic B function gene PISTILLATA (PI) is suppressed (Yao et al., 2001), no flesh-related changes are observed (Figure 8a). It is observed that the hypanthium tissue has distinct zones of tissues (Figure 3d,e) from which each of the apple fruit tissues is derived, and that these are independent of floral organ identity genes such as PI. These distinct zones, which are also indicated by differential expression of the SEP-like genes in fruit tissues (Figure 4), may be the foundations of different floral organs. For example, disruption of the ‘petal foundation’ in the MADS8as hypanthium produces sepalloid petals. This foundation hypothesis may also explain the extreme SEP3 over-expression mutants in Arabidopsis, in which stamens and sepals are converted to carpelloid structures (Castillejo et al., 2005), suggesting that the SEP3 gene promotes the ‘carpel foundation’. It may also explain the flb phenotype, as transcriptomic analysis during berry development showed that the SEP3-like gene, VvMADS4, had reduced expression (Fernandez et al., 2007).

Figure 8.

Models of SEP-like gene function with respect to fruit development and ripening.

(a) During flower and fruit development, the ABC floral homeotic genes control the development of the top zone of the flower. When the B function gene PISTILLATA is mutated, the corresponding apple zone is affected, with a whorl of carpel tissue found at the top of the apple. The SEP1/2 genes control the lower flower zone (the hypanthium), with corresponding loss of fruit flesh in the developing apple. Purple represents carpel-derived tissue, orange represents stamen-related tissue, and yellow represents receptacle-derived tissue (hypanthium).

(b) A model of the control of ripening in apples based on phenotypic studies. Full ripening progression in wild-type apples involves breakdown of starch, change in skin colour, production of ethylene and aroma volatiles, and fruit softening. Senescence follows ripening, and the fruit undergoes rotting. In ethylene biosynthesis-suppressed apples (ACO1as) with functional MADS8/9, starch breakdown and skin colour changes occur, but ethylene-dependent softening and volatile production are suppressed. When MADS8/9 is suppressed, no ripening-related changes occur (either ethylene-dependent or -independent), and the fruit dries.

In apple, ripening events are strongly controlled by ethylene (Dandekar et al., 2004; Schaffer et al., 2007). In this study, regulation of ripening by the SEP1/2-like genes appears to act upstream of ethylene, as demonstrated by both the suppressed expression of the ethylene biosynthesis genes in MADS8as-9 fruit and transactivation of these genes by MADS9 in transient assays. This is similar to the effect of the RIN gene in tomato, on which there has been considerable research. The RIN gene is considered to be a master regulator of ripening, and its suppression renders tomato fruit unable to ripen even in the presence of added ethylene. In tomato, RIN is thought to control auto-catalytic ethylene biosynthesis through transcriptional regulation of the tomato auto-catalytic ethylene biosynthesis genes ACS2 and ACS4 (Ito et al., 2008). The apple SEP genes most similar to the well-characterized tomato RIN gene, MADS4 and MADS104, were expressed throughout fruit development, with only a slight induction during ripening (Figure 2), and expression of these two genes was unaffected in the MADS8as lines during ripening (Figure 4). While RIN and the apple SEP1/2-like genes share similarities with respect to ripening regulation, flesh development is unaffected in the rin mutant. As well as full or partial inhibition of some ripening characters, TAGL1-suppressed tomatoes also showed a reduction in pericarp thickness, thus, although it lies outside the SEP clade, the TAGL1 mutant appears to have the closest phenotype to the MADS8as lines. This evidence suggests a plasticity of function both within the SEP clade as well as within the MADS box family, making it important to look widely within the MADS box family when trying to identify functionally related genes in different species.

The low ethylene concentration in MADS8as-9 was equivalent to the ethylene concentration in ACO1as apples in which ethylene biosynthesis is inhibited (Figure 5) (Schaffer et al., 2007). The transient assay results shown here (Figure 7) suggest that, like RIN, MADS9 (and, to a lesser degree, MADS8) transactivates the promoters of ACS1 and PG (Fujisawa et al., 2011; Martel et al., 2011), but, unlike RIN, they also transactivate the ACO1 promoter (Figure 7), thus reinforcing the plasticity of function observed in the SEP clade across fruit species. This activation leads to further expression of both ACO1 and ACS1 by the auto-catalytic loop regulating expression of these genes (Wiersma et al., 2007). It should be noted that, unlike the rin mutant, MADS8as-9 apples undergo ripening when supplied with ethylene. Heterozygous RIN/rin lines also respond to ethylene, albeit at a slower rate (Vrebalov et al., 2002). Although we cannot discount the possibility that the very low transcription levels of MADS8 and 9 observed in the MADS8as-9 lines may allow this response, this result may alternatively suggest that the SEP-like genes are not acting as master regulators controlling competence to ripen but instead control developmental regulation of discrete aspects of apple ripening, including the initiation of ethylene biosynthesis. The sensitivity/dependency ripening model proposed by Johnston et al. (2009) suggests that developmental factors control early ripening events, which are highly sensitive to ethylene (i.e. respond to low ethylene levels), whereas late ripening events are less sensitive to ethylene (i.e. respond to high ethylene levels) and therefore show a greater dependence on ethylene. In MADS8as-9 apples, both early and late ripening events are not completely compensated for by exogenous ethylene either at the physiological level or the gene expression level (Figures 5 and 6), suggesting a need for SEP-like genes to facilitate complete ripening progression. In particular, the lack of developmentally controlled starch clearance and background skin colour changes in MADS8as-9 lines (which normally progress independently of ethylene; Johnston et al., 2009) suggests that apple SEP1/2-like genes control these processes (Figure 8b).

In MADS8as fruit, there is a severe reduction in cortex tissue, which is the bulk of the tissue consumed by seed-dispersing animals. It is noteworthy that loss of the flesh zone is accompanied by loss of other attractive qualities of ripe fruit, i.e. the fruit do not change colour to stand out from the green foliage, no aroma volatiles are produced, and there is no conversion of starch to sugars (Figure 5). In these respects, MADS8as-9 fruit appear to function more like dry indehiscent fruit rather than fleshy fruit. Indeed, when these fruit are left at room temperature for many months (>10), the fruit do not rot like ‘Royal Gala’ apples do, but instead dry out (Figure 8b). In addition to the loss of attractants, the conversion to a dry indehiscent fruit has also changed the structure of the locules from a tight teardrop structure that grips the seed (while being eaten) to a looser structure from which the seeds are easily shaken out (Figures 4 and 8b). This research is of considerable evolutionary interest, as mutation of a single class of genes appears to control the conversion from fleshy fruit to dry fruit, both by determining flesh development and controlling ripening. It is feasible that changes to SEP gene expression patterns or their genetic structure may have contributed to the variation in floral morphology observed in angiosperms (Theissen et al., 2000). As SEP genes appear to be involved in all stages of reproductive development, possibly acting as interaction bridges (Leseberg et al., 2008), it is feasible that their regulation may also contribute to the considerable variation in fruit morphology observed in angiosperms.

Experimental Procedures

Plant material and growth conditions

A binary vector was constructed using the pART7/pART27 system (Gleave, 1992) containing the apple MADS8 gene (Yao et al., 1999) in the antisense orientation (MADS8as) downstream of the CaMV 35S promoter. Agrobacterium tumefaciens strain LBA4404 was transformed with pART27-MADS8as by electroporation, and used to generate five independent transgenic apple lines in Malus x domestica cv. ‘Royal Gala’ (Yao et al., 1995). Transgenic plants were grown in a containment greenhouse alongside untransformed ‘Royal Gala’ controls and transgenic ACC OXIDASE 1 antisense lines (ACO1as) (Schaffer et al., 2007) as a negative control for ethylene-regulated ripening. For each line, the antisense construct was sequence-verified (Figure S6).

Fruit assessment

Fruit from the five MADS8as lines and the ACO1as and wild-type lines were collected at maturity (based on starch clearance and the skin colour of wild-type fruit) and stored in the absence or presence of 100 μl L−1 ethylene in 340 litre ripening bins at 22°C with continuous circulated air for periods of 4 days or 2 weeks. Two analytical replicates of at least four apples were included for each time point. Mature fruit were assessed for weight, background skin colour, firmness, soluble solid concentration, starch clearance and internal ethylene concentration as described by Johnston et al. (2009). Following 2 weeks’ storage in the presence of ethylene, whole apples were weighed and placed into a 1 litre sealed flask for 1 h at 24°C. Dried air was introduced to sweep the headspace (25 ml min−1) for 2 h onto a volatile absorbent trap (80 mg Chromosorb™ 105 60/80 mesh, Volatiles were assessed as described by Nieuwenhuizen et al. (2009). Peaks were converted into ng g−1 values using a mean detector response factor based on a standard mixture containing ethyl butanoate, butyl acetate, 2-methyl butyl acetate, butanol, methyl hexanoate, ethyl hexanoate, hexyl acetate and hexanol in pentane. Component identification was based on calculation of retention indices, mass spectra of authentic standards, and comparison with library spectra (NIST 05,, Wiley 7,, Adams EO,, and in-house).


For toluidine blue-stained samples, sections were cut using a TPI Vibratome Series 1000 vibrating microtome (Technical Products International, Inc.,, and stained with toluidine blue. Immunolabelled samples were prepared as described by Sutherland et al. (2009). Briefly, sections were blocked with bovine serum albumin in PBS-T for 15 min, incubated overnight with LM5 antibodies diluted 1:200 in blocking buffer (0.1% bovine serum albumin (BSA-c, Aurion, in PBS-T (phosphate-buffered saline plus 0.1% Tween 20)), washed in PBS-T, and incubated for 1 h in goat anti-rat antibody conjugated to Alexa Fluor 488 dye (Molecular Probes, diluted 1:600 in PBS. Sections were washed in 2–3 ml PBS-T and mounted in Citifluor AF1 antifade solution (Citifluor, For Safranin-Fast Green (Sigma-Aldrich, samples, cuboid sections (from skin through to core) were cut from the equatorial region of whole apples. Samples were fixed in FAA (4% formalin/50% alcohol/5% acetic acid in water to 100%). Samples were washed twice in 50% ethanol, then dehydrated through a graded ethanol series at 2 h intervals, then 50:50 ethanol/xylene (2 h), two changes of xylene (3 h each), 1:1 wax/xylene (3 h) and two changes of paraffin wax (12 h each), and embedded in Paraplast wax (Oxford Labware, for standard sections, and LR White resin (Hallett et al., 1992) for immunolabelling. Sections of 10 μm thickness were cut using a Leitz 1512 microtome (Leica,, placed on positively charged glass slides, and dried overnight in a slide dryer. Sections were de-waxed in two changes of xylene (5 min each), followed by two changes of absolute ethanol, then air-dried. Safranin-Fast Green staining was performed as described by Sass (1958). Sections were viewed using an Olympus Vanox AHT3 microscope (Olympus Optical,

Phylogenetic analysis

All phylogenetic analyses were performed using Geneious Pro 5.5.6 (Biomatters, (Drummond et al., 2010). Protein alignment (Blosum62 substitution matrix, gap open penalty of 12, gap extension penalty of 2) was based upon predicted full-length protein sequences (GenBank accession numbers shown in Table S2), and phylogenetic tree construction was completed using the maximum likelihood plug-in PhyML (Guindon and Gascuel, 2003) with the JTT amino acid substitution model (Jones et al., 1992) and bootstrap analysis using 1000 datasets.

Quantitative RT-PCR analysis

For fruit development expression (Figure 2), cDNAs as described by Janssen et al. (2008) were used. Four stages of floral development (Figure 3) were assessed: balloon (5 days before full bloom), open flower, 2 and 8 days after pollination. At each stage, flowers were photographed and harvested for gene expression analysis. Flowers were dissected into two samples: the hypanthium (base of the flower cut in half and ovary tissue removed) and floral tissues (sepals, petals, stamens, carpels and ovary tissue removed from the hypanthium). Gene expression for mature fruit (Figures 4 and 6) was performed on fruit 4 days after storage in presence or absence of ethylene. The cortex tissue is most affected in the MADS8as lines; therefore, expression analysis was performed on three tissue zones: skin (peel), cortex and core. The cortex layer was absent or insufficient in fruit from three MADS8as lines (2, 9 and 18); therefore, analysis consists of only two tissue zones in these lines. RNA was extracted and cDNA synthesized as described by Schaffer et al. (2007). Quantitative RT-PCR was performed as described by Tacken et al. (2010). Expression of each gene was measured relative to expression of the apple housekeeping gene ACTIN (Espley et al., 2007). The quantitative RT-PCR primers used are shown in Table S4. To differentiate the homeologous genes, regions of difference between homeologues were targeted for primer design and the quantitative PCR product was sequence-verified.

Promoter transient assays

Genomic DNA from cv. ‘Royal Gala’ was isolated using a DNeasy Plant Mini Kit (Qiagen, according to the manufacturer's instructions. Primers to amplify each promoter were designed using the recently published draft apple genome (Velasco et al., 2010). For ACS1, two alternative 6.5 kb products were amplified using primer 5′- TTAAGGTACCGACCTACCTAGGCTTC-3′ and either 5′ [Phos] CATGGTGGTTAATTTTCTACTGTATGG-3′ or 5′-GTGGTTAATTTTCTACTGTATGG-3′ with iProof DNA polymerase (Bio-Rad, The two products were mixed 1:1 after the PCR column clean-up, and denatured at 95°C for 5 min, then allowed to re-anneal at room temperature for the NcoI overhang to form. The PCR product was then cleaved with KpnI and cloned into pGreenII-0800-LUC (Hellens et al., 2005) digested with KpnI/NcoI. For ACO1, primers 5′-CTAGATGCCATTTCTTAATTTTCAAATGTC-3′ and 5′-GTCGCCATGGCTCTTTGGATTG-3′, with a modified ATG start codon (underlined) to incorporate an NcoI site, were used to amplify a 2 kb region upstream of ACO1 following the cloning method described by Tacken et al. (2010). The PG1 promoter construct was described by Tacken et al. (2010). Transient assays were performed as described by Hellens et al. (2005) using a modified inoculum concentration of 0.75 (absorbance at 600 nm).


The authors would like to acknowledge G. Wadasinghe and J. Ryan for maintenance and care of the apple plants used in this study, Tim Holmes for photography, Tony Corbett (Plant & Food Research, Hawke's Bay, New Zealand) for illustrations, David Ruddell (United States Department of Agriculture, Agricultural Research Service, Wenatchee, WA) for helpful suggestions on hypanthium development, Charles Ampomah-Dwamena and Jason Johnston for critically reading the manuscript, and the New Zealand Ministry of Science and Innovation (previously New Zealand Foundation for Research, Science & Technology (FRST), contract C06X0705; Pipfruit, a juicy future) for funding this research.