Introduction

Apple (Malus × domestica Borkh.) is one of the most popular perennial tree fruits. A wealth of knowledge exists regarding varietal differences, physiological changes, and horticultural management in relation to fruit maturation, ripening, and quality. Due to its long history and widespread cultivation, and more importantly its out-crossing nature, apple exhibits a high level of heterozygosity and great variation in ripening behavior and quality attributes. The ripening season of apple can differ up to 3 months among the elite apple cultivars under the same weather condition. There are considerable variations of apple fruit texture attributes such as fruit firmness and crispness, which significantly influence consumer preference and fruit industry profitability. Accordingly, apple is one of few fruit crops for which consumers recognize and prefer fruit from specific cultivars.

Fruit ripening, characterized by texture modification, aroma production, and color change, is under tight genetic control (Giovannoni 2004; Seymour et al. 2002), and cultivar-specific patterns of maturation and ripening intimately influence fruit quality. The processes of cell division and expansion have largely attenuated in apple fruit at later stages of maturation (Janssen et al. 2008). However, continuing changes inside the cell and cell wall transform the starchy, hard, and astringent flesh into a state more palatable for human consumption and amenable for natural seed dispersion. In regard to fruit texture changes, microscopic and cellular features such as cell size and density, cell wall properties, cell turgor, and cell–cell adhesion may contribute to the observed phenotypes (Johnston et al. 2002). Currently, a molecular model describing apple fruit maturation, ripening, and texture change is lacking, and its molecular characterization is less represented by other model plant systems due to the unique physiology of pome fruit maturation and ripening.

Plant hormones, especially ethylene, are known to regulate apple fruit maturation, ripening, and quality (Bleecker and Kende 2000; Costa et al. 2005; Harada et al. 2000). Apple cultivars with different allelotypes of two ethylene biosynthesis genes, ACS1 and ACO1, have been shown to correlate with apple fruit firmness in a wide selection of germplasm (Oraguzie et al. 2004; Zhu and Barritt 2008). Recent reports suggest that auxin may also play a critical role in regulating fruit ripening and quality of apple and peach (Li and Yuan 2008; Kondo et al. 2009; Trainotti et al. 2007; Ziliotto et al. 2008). Cell wall metabolism has been closely associated with fruit texture changes, and numerous gene families could potentially contribute to cell wall biosynthesis and modification (Bennett and Labavitch 2008; Harker et al. 1997; Johnston et al. 2002). Many other pathways and physiological processes, such as those related to intra-cellular turgor (water and metabolite transport) and epidermal structural deposition (wax and cutin), may also contribute to fruit texture attributes. Yet little is known regarding the individual genes and pathways differentiating cultivar-specific apple fruit ripening patterns and quality attributes. Apple cultivars ‘Honeycrisp’ (HC) and ‘Cripps Pink’ (CP) exhibit distinct texture attributes and ripening behavior. While fruit of late-ripening CP demonstrate outstanding firm flesh, HC is a mid-early ripening cultivar with extraordinarily crisp but less firm flesh. Both are commonly utilized as apple breeding parents for their exceptional texture attributes and other quality traits.

Several recent studies using various apple microarrays were reported including those focusing on early fruit development (Lee et al. 2007); the roles of ethylene on volatile biosynthesis pathways (Schaffer et al. 2007); the comprehensive transcriptome analysis from flowering to fruit ripening on the apple cultivar ‘Gala’ (Janssen et al. 2008); the effect of different apple root stocks in relation to gene expression in scion tissues (Jensen et al. 2010); and using apple EST array and heterologous tomato array to study the ripening dynamics of ‘Mondial Gala’ (Costa et al. 2010). Various formats of microarray have also developed for transcriptome profiling on other rosaceous crops (Falara et al. 2011; Vizoso et al. 2009; Ziliotto et al. 2008). In this study, a high density long-oligo apple array was designed for and manufactured on NimbleGen platform based on Malus unigene assembly version 4 (http://www.bioinfo.wsu.edu/cgi-bin/gdr/gdr_unigeneV4_project_description.cgi?genus=Malus). Parallel transcriptome profiling was performed on two cultivars at equivalent maturation stages. Unigenes associated with biosynthesis, homeostasis and transport, reception and metabolism of multiple plant hormones as well as unigenes functioning in cell wall modification were identified; their relationships with cultivar-specific phenotypic variations of apple maturation patterns and texture attributes were discussed.

Materials and methods

Physiological characterization of apple fruit ripening

Physiological characterization of weekly fruit samples was used to define the maturity level and maturation process of two apple cultivars, ‘Honeycrisp’ (HC) and ‘Cripps Pink’ (CP, also known as Pink Lady™). Fruit were harvested from two separate commercial orchards close to Wenatchee WA in 2007. Fruit with uniform size and appearance were randomly harvested from a number of trees for weekly maturity testing. Based on the ripening data in previous years, fruit sample collection started approximately 6 weeks before and continued to 2 weeks after the projected harvest date for obtaining fruit with the desired range of maturity. Fruit firmness and crispness were measured on pared fruit surfaces using the Mohr Digi-Test, MDT-1 (Mohr and Associates, Richland, WA, USA), which is a computerized penetrometer with 11-mm probe diameter. Crispness (Cn) values from the MDT-1 are the Fourier transformation of the force profile as the probe goes through fruit cortex tissues, and these values therefore are theoretically similar to the energy released during a bite; a recent study showed there was a good correlation between the Cn value and human sensory evaluation regarding fruit crispness (Evans et al. 2010). Fruit internal ethylene concentration (IEC) was measured using a Hewlett Packard 5880A series gas chromatograph equipped with a Porapak Q column (0.3 cm × 30 cm i.d., 80–100 mesh) and flame ionization detector (Blanpied and Silsby 1992; Argenta et al. 2002). Fruit maturity which is primarily determined by starch pattern indices (SPI) were calculated by averaging the values (1–6 in grade) (Brookfield et al. 1997) of 15 apples. The weekly fruit samples with starch pattern indices close to an average of 3.5 were arbitrarily designated as week 0 or physiological maturity in this study. Fruit with various maturity levels were defined in relation to week 0 samples, i.e., fruit collected at 4, 3, 2, or 1 week(s) before week 0 were retrospectively assigned as week −4, −3, −2, or −1 samples. Values of individual fruit diameter were recorded automatically by Digi-Test; the significance of difference between weekly samples was analyzed by t test (Microsoft Excel). Cortex tissues collected at weeks −4, −2, and 0 were used for transcriptome profiling, and each weekly sample was represented by four biological replicates. Each replicate consisted of a pool of cortex tissues from five apples. Once collected, fruit cortex tissues were immediately frozen in liquid nitrogen and stored at −80 °C until RNA isolation.

RNA isolation

Total RNA was isolated as described by Gasic et al. (2004) with modification. Apple cortex tissue (2–3 g) ground (IKA A 11 basic; IKA®-WORK Inc., Wilmington, NC, USA) in liquid nitrogen was transferred to 50-mL polypropylene tubes with 10 mL 2× CTAB extraction buffer, vortexed, then incubated at 60 °C for 15 min with occasional inversions. Then an equal volume of chloroform/isoamyl alcohol (24:1, v/v) was added and the mixture vortexed for 2 min. The mixture was then centrifuged at 10,000 × g for 10 min at 4 °C and the supernatant transferred to a clean 50-mL polypropylene tube. After repeating this process, the supernatant was transferred to a 15-mL polypropylene tube where a one third volume of 7.5 M LiCl was added; the tube was inverted several times to mix and then incubated at 4 °C for 16 h. Following incubation, tubes were centrifuged at 14,000 × g for 30 min at 4 °C, the supernatant was discarded, and the RNA pellet washed with 750 μL 70 % ethanol. The RNA pellet was re-suspended in 500 μL DEPC water and transferred to a 1.5-mL micro-centrifuge tube. Total RNA was precipitated by adding 1/10 vol of 3 M sodium acetate (pH 5.5) and 2 vol of 100 % ethyl alcohol followed by incubation for 1–3 h at −80 °C. After centrifugation at 14,000 × g for 30 min at 4 °C, the supernatant was discarded; the pellet was washed with 70 % ethyl alcohol and then air-dried for 5 min. The pellet was suspended in 150 μL DEPC water then stored at −80 °C. Isolated total RNA was quantified using an ND-1000 NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and RNA quality was verified by agarose gel electrophoresis.

Microarray design

An isothermal long-oligo apple microarray was designed based on the Malus unigene V4 sequences which are available at the Genome Database for Rosaceae (GDR) (Jung et al. 2008). A total of 260,581 Malus EST sequences were downloaded from NCBI dbEST (Benson et al. 2007), filtered for contamination and low quality sequence, and assembled into 23,284 contigs and 53,200 singletons using CAP3 (Huang and Madan 1999) with an overlap percentage parameter of 90 (−p 90). The unigenes, comprising of the combined contigs and singlets, were computationally annotated for putative function by pairwise comparison against the Arabidopsis thaliana protein database (http://www.arabidopsis.org), and the Uniprot Swiss-Prot and TrEMBL databases (Wu et al. 2006; Mulder et al. 2007) using the BLASTX algorithm. Only matches with an E value of less than 1.0 e−6 were recorded. Based on the similarity search results, 55,960 (73 %) of the Malus unigene sequences had significant matches with proteins from these databases. Duplicate transcripts present in both the contigs and singletons were eliminated by removing singlets with a common protein match with the contigs. Following detection of Open Reading Frames (ORFs) using the EMBOSS getORF program (http://bioweb2.pasteur.fr/docs/EMBOSS/getorf.html), 23,106 contigs and 17,060 singletons, totaling 40,166 unigenes, were available for oligo design. The isothermal oligonucleotides of 50–70-mer were designed from the coding region of the 40,166 unigenes using custom perl scripts. Oligos designed from contigs had three to 10 oligos per contig with an average of seven; those designed from singlets had seven to 10 oligos with an average of eight oligos per singlet. Including control sequences, a probe set containing 190,135 oligos representing 23,997 unigenes was designed for NimbleGen array format. The unigene, oligo sequences, and protein homology results are available at http://www.bioinfo.wsu.edu/cgi-bin/gdr/gdr_unigeneV4_project_description.cgi?genus=Malus .

Microarray fabrication, hybridization, data acquisition, and analysis

Array fabrication, cDNA conversion and labeling, hybridization, image scanning, and data normalization were performed at NimbleGen (http://www.nimblegen.com/). Using a single color labeling system, a total of 24 microarray slides were utilized for transcriptome profiling, i.e., two cultivars × three maturation stages × four biological replicates. The normalized signal intensity based on the internal control probe set were imported from the NimbleGen expression files and were analyzed using Arraystar (http://www.dnastar.com/). After performing an ANOVA and multiple testing correction (Benjamini and Hochberg 1995), the probes (or oligos) that displayed a 2-fold or greater change in transcript abundance, across any two adjacent maturation stages (i.e., week −4, week −2, and week 0) within each cultivar, were extracted and converted to differentially expressed (p < 0.01) unigenes. Tables S1 and S2, displaying the expression patterns of individual identified unigenes, were prepared using CORELDRAW (http://www.corel.com/servlet/Satellite/us/en/Content/1150905725000).

Validation of microarray data by qRT-PCR

Individual unigenes were selected randomly from the list of identified unigenes for data validation by qRT-PCR. EST sequences were obtained from the GDR (Jung et al. 2008). Forward and reverse primers were designed using web-based software Primer3plus (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) and IDT oligo analyzer (http://www.idtdna.com/analyzer/Applications/OligoAnalyzer/). Where possible, an optimum annealing temperature of 59 to 60 °C, GC content 40–60 %, amplicon length 150–180 bp, and primer length 20 bp were applied. As needed, unigenes from multi-member family were analyzed using ClustalW (http://www.ebi.ac.uk/tools/clustalw/) and BLAST (http://www.ncbi.nlm.nih.gov/BLAST/) to choose divergent regions for designing gene-specific primer sets. Primers used in this study were listed in Table S5. Total RNAs isolated from the same tissues used for microarray analysis were treated with DNase I to eliminate co-purified residual genomic DNA and further purified using Qiagen RNeasy columns (Qiagen, Valencia, CA, USA). For a 20-μL reverse transcription (RT) reaction, 2 μg total RNA and 1 μL 100 μM poly dT primer were incubated at 70 °C for 10 min, then put on ice for 2 min. Then 15 U AMV reverse transcriptase and 4 μL of 5× reverse transcription buffer (Promega, Madison, WI, USA), 2 μL of 10 mM dNTP, and 20 U recombinant RNasin® ribonuclease inhibitor (Applied Biosystems, Foster City, CA, USA) were added to a final volume of 20 μL with H2O. The reactions were incubated at 42 °C for 1 h and then at 70 °C for 15 min. For each cortex tissue sample, two independent total RNA isolations were obtained, and thereafter two separate cDNA preparations. PCR reactions were carried out in triplicate for each cDNA preparation and repeated at least once. The volume of quantitative PCR was 15 μL, which contained 1× PCR buffer, 2.5 mM MgCl2, 2.4 μL cDNA (20× dilution from original RT reaction), 200 nM forward and reverse primers, 200 μM dNTP, 0.3 U iTaq DNA polymerase (Bio-Rad Lab, Hercules, CA, USA), and 0.45 μL 2,000× dilution of SYBR I green (Invitrogen, Carlsbad, CA, USA). Reactions started with a denaturation stage for 3 min at 94 °C, then amplified for 40 cycles (94 °C for 15 s and 59 °C for 30 s) using a Bio-Rad iQ5™ real-time PCR detection system (Bio-Rad). Melting curve and amplification efficiency were analyzed for each primer set, and “no template control” and “reverse transcriptase minus control” were routinely included. The sequence of a unigene for actin (Contig13803) was used to design primers as a reference gene for normalization. Data was normalized against an apple actin encoding unigene and expression analysis was performed using the ddCt method (Bio-Rad).

Results

Cultivar-specific fruit maturity progression

Based on weekly maturity data, ‘Honeycrisp’ (HC) fruit harvested on Sep. 2 and ‘Cripps Pink’ (CP) fruit harvested on Nov. 4 were defined as week 0 samples, i.e., HC0 and CP0, when their starch pattern indices (SPI) were close to average values of 3.5 (Table 1). Fruit firmness decreased as maturation progressed; however, the rate of firmness loss varied substantially between these two cultivars (Table 1). At week −4, a difference of fruit firmness between these two cultivars was 11.2 N (Newton), yet a more substantial 33.3-N difference was observed at week 0. Apple fruit crispness measured by the Digi-Test expressed as Cn value also showed remarkable differences between these two cultivars. Most of the Cn values (four out of five) from the fruit of HC were close to or higher than 250, while most of the values (four out of five) for those of CP were close to or lower than 200. In both cultivars, the average values of SPI steadily increased as maturation progressed, and the comparable extent starting from an average value of about 1 to a value close to 3.5 among the five weekly samples were observed. No substantial and consistent ethylene production was observed, indicating that climacteric ripening had not initiated in these fruit. No significant changes of fruit diameter were observed among weekly fruit samples except for week −4 and −3 of HC, indicating fruit have attained their mature size at late maturation.

Table 1 Physiological characterization of fruit maturity and texture attributes
Full size table

Transcriptome profiling of apple fruit cortex tissues

The identification of differentially expressed unigenes was performed within each cultivar and based on ANOVA analyses using a cutoff value of 2-fold of normalized signal intensity and a non-adaptive false discovery rate (FDR) of 0.01. A total of 1,793 differentially expressed unigenes from HC and 1,209 from CP were identified, which represent 7.47 % and 5.04 % of all unigenes deposited on the array, respectively (Table 2). The volcano plots showed an asymmetric shape between two adjacent stages (−4 to −2 and −2 to 0) for HC which reflected more unigenes with “transitional expression patterns”, i.e., those with down- then up-regulated patterns or vice versa; this trend was not observed in CP (Fig. 1 and Fig. S1a). In both cultivars, unigenes showing down-regulated expression patterns substantially outnumbered the ones with up-regulated expression patterns (Table 2 and Fig. S1a). The raw dataset was deposited in GEO with Series accession GSE24523 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE24523).

Table 2 Summary of identified differentially expressed unigenes from two cultivars
Full size table
Fig. 1
figure 1

Volcano plots showing the distribution patterns of identified unigenes from two cultivars. a Identified unigenes from week −4 to week −2 in the cortex tissue of HC. b Identified unigenes from week −2 to week 0 in the cortex tissue of HC. c Identified unigenes from week −4 to week −2 in the cortex tissue of CP. d Identified unigenes from week −2 to week 0 in the cortex tissue of CP

Full size image

Functional categories of differentially expressed unigenes

The differentially expressed unigenes were classified via their functional annotations and grouped into finer functional categories in order to derive biological meaning. Unigenes with genotype-specific expression patterns may contribute to the observed phenotypes of maturation patterns and texture attributes.

Hormonal metabolism and response

Over half of the unigenes in this group (41 out of total 74) were those related to the biosynthesis, metabolism, transport, and response of ethylene and auxin, indicating the central roles of these two hormones during apple maturation. Unigenes implicated in metabolism and response of other hormones including gibberellins (GA), brassinosteroid (BR), jasmonic acid (JA), abscisic acid (ABA), and cytokinins (CK) were also identified (Table 3). About 50 % of the unigenes were shared between these two cultivars, and the majority of them exhibited similar expression patterns (Table 3, Table S1, Fig. 2a, and Fig. S2).

Table 3 Differentially expressed unigenes with the annotated function in plant hormone metabolism and responses
Full size table
Fig. 2
figure 2

Venn diagram demonstrating overlapping differentially expressed unigenes from each cultivar for selected functional groups, and those between cultivars. a Unigenes with annotated functions of hormone metabolism and responses. b Unigenes with annotated functions of cell wall metabolism and modification. c Transcription factor encoding unigenes. d Unigenes with other cellular functions

Full size image

Ethylene biosynthesis and responses

Consistent with the critical roles of ethylene during pome fruit maturation and ripening, the majority of unigenes related to ethylene functions showed up-regulated expression patterns as fruit maturation progressed (Table 3, Table S1, and Fig. S2a). In fact, this was one of a few sub-groups which consisted of more up-regulated unigenes than down-regulated ones. Two unigenes, contig21202 and 21169, annotated as “1-aminocyclopropane-1-carboxylate synthase 7”, and sequence alignment indicated that they encode apple MdACS3 for pre-climacteric ethylene biosynthesis (Rosenfield et al. 1996; Varanasi et al. 2011), showed consistently up-regulated expression patterns in both cultivars, although a higher fold increase was observed in CP. Two identified unigenes (contig13926 and 21937) encoding 1-aminocyclopropane-1-carboxylate oxidase (ACO) were similarly up-regulated, though a much higher fold increase from week −4 to week −2 was observed in HC. A unigene (contig18342) encoding MdETR2 (Wiersma et al. 2007), an ethylene receptor which negatively regulates ethylene responses, exhibited up-regulation but only in CP. Four unigenes encoding the EIN3-binding F-box protein (Contig17900, 18538, 21897, and 21173), a negative regulator of ethylene action (Gagne et al. 2004), were also identified exclusively from CP with up-regulated expression patterns. A unigene (Contig12329) encoding an “Ethylene-overproduction protein 1”, a negative regulator of ethylene evolution (Yoshida et al. 2005), also showed a down-regulated pattern and only in CP.

Auxin biosynthesis, transport, and responses

In contrast to up-regulated expression patterns for most ethylene function-related unigenes, the majority of the identified unigenes associated with auxin function exhibited down-regulated expression patterns, particularly those from CP (Table 3; Table S1 and Fig. S2b). Four unigenes annotated as “auxin efflux carrier component” (contig7586, 18662, Malus_CV794150, and Malus_CV880396), which encode auxin transporter PIN1 and PIN8 homologues (Petrášek et al. 2006), were identified only from CP with down-regulated expression patterns. Additionally, one unigene from each cultivar (Malus_CN945995 from CP and contig17221 from HC) with the similar annotated function showed the comparable up- and then down-regulated expression patterns. Three unigenes with annotated functions of auxin homeostasis regulation were identified from both cultivars. Contig3460 encoding a protein homologous to rice “indole-3-acetic acid-amido synthetase GH3.5”, which catalyzes the synthesis of IAA–amino acid conjugates (Jain et al. 2006), showed down-regulated expression patterns in both cultivars, though with a double-digit fold decrease in HC at later stages (from −2 to 0 week), while another unigene with similar function (Malus_CN909148) was up-regulated in CP. Contig196 encoding a protein homologue to “indoleacetamide hydrolase” (Mazzola and White 1994), which releases bioactive IAA from amide conjugation, was down-regulated only in CP. In addition, more than a dozen unigenes encoding “auxin induced proteins” or “auxin responsive proteins” were identified from both cultivars, but more unigenes showed down-regulated patterns in CP.

Metabolism and responses of other plant hormones

Thirteen unigenes with the annotated functions in gibberellin biosyntheses and responses were identified (Table 3, Table S1, and Fig. S2c). Two unigenes from each cultivar (contig21434 and 21375 from CP, contig21386 and 22883 from HC) which encode “DELLA proteins”, the repressor of GA responses (Chandler et al. 2002), were identified and all showed down-regulated expression patterns except the up-regulated contig22883 in HC at the later stages. Four unigenes (contig16116, 9239, 16825, and 9114) encoding proteins homologous to “gibberellin 3 (or 2)-beta-dioxygenase”, which convert the inactive precursors to the bioactive GA (Thomas et al. 1999), showed mostly up-regulated expression patterns in both cultivars. Contig16116 showed slight down-regulation first and then a double-digit fold change of up-regulated expression in HC, compared to a moderately increased expression in CP. A unigene (contig5705) encoding protein homologous to “rice gibberellin receptor GID1”, a soluble gibberellin receptor (Ueguchi-Tanaka et al. 2005), showed up- and then down-regulated expression patterns in both cultivars, though a greater fold change was observed in HC. Contig8101 encoding another “gibberellin receptor” was identified only from CP with moderate but progressive up-regulation. A total of nine unigenes annotated as “brassinosteroid insensitive 1-receptor kinase 1” were identified from both cultivars. All seven of those identified from CP showed down-regulated expression patterns; similarly, all five from HC were down-regulated except two unigenes which showed slight up-regulation during the early maturation stage. Among four “Jasmonate O-methyltransferase” (contig8892, 17950, 23142, and 17848) encoding unigenes identified from both cultivars, all three from HC (except one value) were up-regulated, but in CP one was down-regulated and another up-regulated. Two unigenes encoding proteins homologous to “cytokinin dehydrogenase” (contig15546 and Malus_CN910875) and three unigenes for “cytokinin-N- (or O-) glucosyltransferase” (contig21874, 21694, and Malus_CN994862), which regulate homeostasis by deactivating CK (Hou et al. 2004; Schmuelling et al. 2003), were identified from both cultivars; more than 50 % of the values indicated up-regulated patterns of three unigenes in HC, but both unigenes from CP showed down-regulation. A unigene for “abscisic acid 8′-hydroxylase (contig1068)” was down-regulated in CP, but displayed an up- and then down-regulated pattern in HC. An “abscisic acid response protein” (contig5535) encoding unigene was down- and then up-regulated only in CP.

Carbohydrate metabolism and cell wall modification

Identified unigenes with annotated functions in polysaccharide metabolism and cell wall modification represent another major transcriptomic change during apple maturation. Overall, comparable numbers of unigenes belonging to this category (57 from CP and 59 from HC) were identified from each cultivar and over 40 % of these unigenes were shared by both cultivars (Fig. S3). Nevertheless, a higher percentage of unigenes (42.9 %) from HC showed up-regulated patterns compared with those from CP (29.8 %) (Table 4, Table S2, and Fig. S3). Across different sub-groups of cell-wall-related genes (Cantarel et al. 2009), fewer unigenes with annotated functions in polysaccharide biosynthesis (glycosyl transferases, GTs) were identified than those implicated in polysaccharide breakdown and cell wall disassembly or remodeling such as glycosyl hydrolases: GHs and polysaccharides lyases and esterases (Table 4 and Fig. S3). Within the GT sub-group, the majority of the unigenes from HC were up-regulated during the later maturation stage from week −2 to week 0, compared to only two unigenes (out of 10) that showed up-regulated patterns in CP. Among GHs and other cell wall modifying proteins, comparable numbers of unigenes were identified between these two cultivars, and in general these unigenes exhibited similar expression patterns. For example, a unigene annotated as polygalacturanase (PG) (contig21179) showed similar up-regulated expression patterns in both cultivars, even with similar fold increases (Table 4 and Table S2). For a few sub-groups including “glucan endo-1, 3-beta-glucosidase”, “carbohydrate esterase”, “polysaccharide lyase”, “expansin”, and “cell wall structural proteins”, equal or more unigenes were with up-regulated expression pattern in cortex tissues of HC, though the extra down-regulated unigenes were commonly identified from CP. For a few gene families, noticeable cultivar-specific regulation patterns were observed (Table 4 and Table S2). Six unigenes encoding “xyloglucan endotransglycosylases/hydrolase (XTH)” (contig17875, 22333, 21660, 21193, 8509, and 21334) were identified from these two cultivars. All four of those from HC showed progressively up-regulated expression patterns during maturation; among five XTHs identified from CP, two unigenes exhibited up-regulated expression patterns, but other three showed up- and then down-regulated patterns. Three unigenes (contig15993, 11693, and Malus_CV793668) annotated as wall-associated-kinase (Verica et al. 2003) were exclusively identified from HC, and two of them showed up- then down-regulated patterns. A unigene (contig13231) encoding a protein homologous to COBRA7 protein (Roudier et al. 2005) and with putative function in cellulose microfibril deposition exhibited down- and up-regulation patterns and only from HC.

Table 4 Differentially expressed unigenes with the annotated function in cell wall metabolism and modification
Full size table

Transcription factors (TFs)

A substantial number of TF-encoding unigenes were identified in cortex tissues during apple fruit maturation, i.e., 139 from HC and 81 from CP. Consistent with the critical roles of ethylene and auxin on apple fruit ripening (as shown in Table 3), more than 20 % (39/182) of TFs encoding unigenes belonged to those specifically responding to these two plant hormones (Table 5 and Table S3). From HC, the relatively equal numbers for both auxin- and ethylene-specific TFs showed either up- or down-regulated patterns. However, from CP, an obvious disparity was observed between these two sub-groups of TF-encoding unigenes: the overwhelming majority of auxin-specific TF-encoding unigenes showed down-regulated expression patterns, compared to more up-regulated values for ethylene-specific TFs (Table 5 and Fig. S4a and S4b). This observation suggests that a deficiency in auxin metabolism and/or transport occurs in CP (see previous section). For several TF families, unigenes were commonly identified from both cultivars, such as “bHLH”, “homeobox-leucine zipper protein”, “MYB”, and “squamosa promoter-binding-like proteins” (Table 5 and Table S3). Nevertheless, unigenes belonging to a few other TF families showed strong cultivar-specific expression patterns, and in most cases they were identified almost exclusively from HC, such as “zinc finger protein”, “WRKY”, and “NAC” (Table 5 and Fig. S4c).

Table 5 Distribution of up-regulated or down-regulated values of unigenes for major TF families
Full size table

Unigenes with other annotated molecular functions

Unigenes with various annotated molecular functions were identified (Table S4). Among them, the largest group consisted of more than 100 unigenes which were annotated as “protein kinase”, “receptor-like protein kinase”, or “phosphatase”. Other groups included those with putative functions in modulating oxidative balances such as “cytochrome p450”, “lipoxygenase”, and “oxidase”; stress-response and other cellular processes such as “heat shock proteins”, “dehydration-responsive and element-binding protein”, “aquaporin” as well as those annotated as “major allergen Mal d1” and “calcium responsive proteins”. For the vast majority of these unigenes, their biological functions and their potential association with apple fruit maturation are largely elusive.

Microarray data validation

Twelve identified unigenes based on transcriptome profiling were randomly selected and their expression profiles were characterized by quantitative reverse transcription PCR (qRT-PCR). As summarized in Table 6 and Fig. S5, among 48 values for 12 unigenes from two cultivars, 87.5 % of the values showed consistency between the microarray data and qRT-PCR result in terms of either up- or down-regulation expression patterns. Due to different chemistries employed by these two approaches, the actual values of fold changes between the microarray analysis and relative expression levels by qRT-PCR may not be comparable in some cases.

Table 6 Data validation by qRT-PCR
Full size table

Discussion

To gain the insight of global transcriptional networks and identify genotype-specific transcriptome changes over apple fruit maturation patterns and fruit texture attributes, parallel transcriptome profiling was performed on two apple cultivars with contrast phenotypes (Table 1). The high-density microarray developed for this study, to our knowledge, contains the largest unigene collection among reported transcriptomic studies on apple or other Rosaceae crops (Costa et al. 2010; Falara et al. 2011; Schaffer et al. 2007; Vizoso et al. 2009; Ziliotto et al. 2008), which offered better coverage and resolution for analyzing transcriptome activities. A reliable dataset was generated using single color labeling system in array hybridization and proper biological repeats, which was mostly verified by qRT-PCR on randomly selected unigenes (Table 6 and Fig. S5). The majority of gene families or functional groups showed comparable regulation patterns between these two cultivars; those unigenes exhibiting cultivar-specific expression patterns could represent the candidates contributing to the observed phenotypes of fruit maturation patterns, ripening season, cell wall features, and texture attributes. However, it is likely that some of the identified unigenes may result from the variations of external conditions at or before sampling time such as temperature or horticultural practices. The presence of more unigenes with “transitional expression patterns” from HC is not clear, which could derive from either external conditions or the cultivar itself as HC was originally bred for cold hardiness but hot and dry summer is common in central Washington State.

Maturation progression and tissue comparability between cultivars

For a meaningful comparison between two cultivars, it is critical that the samples included are with equivalent maturity or at similar developmental stages. Several physiological indicators can practically be used to define apple fruit maturity, such as days after full bloom (DAFB), fruit firmness, background color, internal ethylene concentration (IEC), and fruit starch levels. Fruit texture and maturation pattern are two targeted phenotypes in this study. IEC remains low and fluctuates during apple fruit maturation prior to climacteric ripening. Therefore, fruit starch pattern indices (SPI) (Brookfield et al. 1997), which provide a steady change through the maturation process, were primarily utilized to align the maturation stages between these two cultivars. Several studies indicated that climacteric ethylene production from system II ethylene biosynthesis by the function of MdACS1 had not activated in apple fruit before physiological maturity (Rosenfield et al. 1996; Varanasi et al. 2011; Wiersma et al. 2007; Zhu et al. 2008). Consistent with these results, no up-regulation of MdACS1 was detected in either cultivar; instead, two unigenes encoding MdACS3 and functioning in pre-climacteric ethylene biosynthesis showed similar up-regulation in both cultivars (Table 3 and Table S1). Based on our maturity data, fruit cortex tissues with comparable maturity were selected for transcriptome profiling. For example, similar extent of SPI values from 1 to about 3.5 was observed for both cultivars.

Critical roles of cultivar-specific hormonal regulation

Consistent with the fundamental roles of plant hormones and their crosstalk in plant physiology (Spartz and Gray 2008; Vandenbussche and Van Der Straeten 2007), the identified unigenes associated with plant hormone metabolism and responses were one of the major focal points of transcriptomic changes in apple cortex tissues (Tables 3, 5, S1, and S3; Figs. S2 and S4). Except ethylene biosynthesis, little is known regarding the genotype-specific roles of plant hormones metabolism and response during fruit ripening. Several unigenes encoding negative regulators of ethylene response, such as ethylene receptor and EIN3-binding F-box (EBF) protein (Rosenfield et al. 1996; Gagne et al. 2004), were identified with up-regulated patterns and exclusively from CP, a late-ripening cultivar with extremely firm flesh (Tables 1, 3, and S1). Most unigenes related to auxin metabolism and function showed down-regulated patterns in contrast to up-regulated trends of most ethylene-related genes. The data suggested that the deficiency in auxin transport and availability of its biologically active form may be associated with the prolonged maturation and late-ripening phenotypes of CP (Table 3 and Table S1). Given the synergistic nature between auxin and ethylene interactions (Rahman et al. 2002; Stepanova et al. 2005; Swarup et al. 2002), the variations of auxin metabolism and action may determine the timing and strength of ethylene metabolism. Jasmonate and its methyl ester have been shown to stimulate apple ACS and ACO enzyme activities and gene expression in apple fruit (Fan et al. 1998; Kondo et al. 2009), as similarly demonstrated in vegetative tissues of Arabidopsis (Kazan and Manners 2008; Shinshi 2008). The interplays between GA and other plant hormones in apple fruit tissues are virtually unknown and has been rarely reported, but the role of GA in apple fruit ripening cannot be dismissed based on this dataset. Although ethylene biosynthesis and signaling in apple fruit maturation, ripening, and quality have been extensively studied, ethylene itself is most likely just one node in the plant hormone regulation network (Kondo et al. 2009; Kuppusamy et al. 2009; Shinshi 2008) rather than an independent or isolated pathway. With the progress in elucidating the plant hormone regulation network and their crosstalk in model plant systems (Ding et al. 2008; Murphy et al. 2002; Noh et al. 2001; Swarup et al. 2007; Teale et al. 2006; Yoo et al. 2009), significant advances of understanding the roles of their counterparts in apple fruit are expected.

TFs in transcriptional regulation network of gene expression

Transcriptional regulation is a major control point of gene function, primarily by the actions of transcription factors (TFs) to control the unique sets of genes in response to endogenous and exogenous stimuli (Riechmann et al. 2000). Consistent with the predominant roles of ethylene and auxin in apple fruit ripening regulation, a substantial portion of identified TF-encoding unigenes (39/182) were those specifically responding to these two hormones (Table 5, Table S1, and Fig. S4). Interestingly, more ethylene-specific TF-encoding unigenes identified from CP showed up-regulated expression patterns, suggesting a delayed but essential role of ethylene even for an extremely late-ripening cultivar. In contrast, the majority of auxin-specific TFs showed down-regulated patterns from CP. Whether or not the deficiency of auxin metabolism and transport (Table 3) in late-ripening CP resulted in a less synergistic interaction with ethylene will remain an interesting question. Several TF families, such as NAC, MYB, “zinc finger”, and WRKY, showed larger numbers of up-regulated unigenes in HC (Table 5, Fig. S4, and Table S3). It is likely that some other TFs such as NAC and MYB may be involved in cell wall modification and/or hormone metabolism, but further investigation is needed. It should not be surprising that some of the identified TFs were the result of different environmental factors or horticultural practice in separate commercial orchards. Nevertheless, due to the nature of tree fruit crops, it will be very difficult, if not impossible, to exclude environmental and/or horticultural conditions from sampling strategy. One example is CBF/DREB1 (contig21396), which encodes a stress-regulated transcription factor and was moderately and consistently up-regulated in CP, but showed a double-digit fold change of up-regulation from the week −4 to week −2 stage, and then a double-digit fold decrease from week −2 to week 0 in HC.

Cell wall metabolism and cultivar-specific fruit texture attributes

Plant cell wall metabolism, particularly cell wall degradation and modification, has long been associated with fruit ripening and softening (Brummell and Harpster 2001; Harker et al. 1997; Johnston et al. 2002). Individual genes encoding cell wall enzymes have been studied for their potential roles in fruit ripening and postharvest fruit softening (Brummell 2006; Dotto et al. 2006; Mann et al. 2008; Marín-Rodríguez et al. 2002; Wakasa et al. 2006). Cell wall metabolisms include multiple steps such as substrate generation, synthases and glycosyl transferases activities, secretory pathways, wall assembly, wall dynamics, and wall disassembly (Cantarel et al. 2009). It was estimated that plants devote 10 % of their genome to cell wall biogenesis alone, and among those genes currently annotated as “unknown function”, up to 1,000 of them may encode cell-wall-related proteins (Yong et al. 2005; Penning et al. 2009). From the current dataset, identified unigenes related to cell wall metabolisms represented a major transcriptomic change (Table 4 and Figs. S1c and S3). Coincident with the decreasing fruit firmness as maturation progressed, larger number of unigenes with putative functions in cell wall disassembly were identified, compared to fewer unigenes with putative roles in cell wall biosynthesis (GT) (Table 1, Table 4, and Fig. S3). As GTs are defined as those catalyzing the formation of glycosidic bonds, their functions may encompass the biosynthesis of disaccharides, oligosaccharides and complex carbohydrates, as well as glycosylation of many other molecules beyond cell wall metabolisms including protein, lipid, and plant hormones (Coutinho et al. 2003). A few gene families including XTH, WAK, and a “COBRA-like cell wall protein” showed strong cultivar-specific expression patterns. The apple genome contains at least 11 XTHs, and their activities potentially regulate several aspects of growth and development (Atkinson et al. 2009). An earlier analysis on EST frequency between the apple cultivars ‘Gala’ and ‘GoldRush’ also suggested the roles of XTHs in regulating fruit texture (Park et al. 2006). The roles of expansin, polysaccharide lyases, beta-galactosidases, and polysaccharide esterase encoding genes contributing to genotype-specific textural attributes deserve more investigation. The degradation of hemicelluloses by XTHs may also associate with cultivar-specific fruit firmness and crispness.

Unigenes annotation and their alignment with whole genome sequences

Multiple unigenes belonging to a few gene families (or functional groups), such as “auxin-induced protein 5NG4” and “Jasmonate O-methyltransferease” (Table S1), were observed to share the same annotation and protein ID but with different expression patterns. Blast search of these unigene against recently available apple whole genome sequences (Velasco et al. 2010; http://www.rosaceae.org/projects/apple_genome#publication) suggested that they belong to different individual genes in the same gene family (data not shown). It is likely that the short and incomplete coding sequence and the huge gene family size they belong to are the cause of the ambiguity in annotation. It was also observed that different unigenes were aligned at various coding regions of the same gene model. It is also possible that splicing variants of the same genes or different alleles of the same loci cause multiple ESTs sharing the same protein IDs. Or in some cases, they may simply result from errors in sequencing, assembling, detection, or data analysis.

The full bloom dates for these two cultivars are known to be only a few days apart. However, maturation and ripening of HC advances with an accelerated and condensed process, while CP progresses in a much slow-down and expanded fashion resulting in a separation of ripening time for more than 2 months. From current transcriptomics data focusing on the late maturation stages, it can be hypothesized that the inherent genetic variation at both hormone regulations and cell wall metabolisms primarily regulate apple fruit maturation patterns and fruit texture such as fruit firmness and crispness. For example, the deficiency in auxin transport and less available in biologically active auxin (from homeostasis regulation) may lead to the delayed activation of climacteric ethylene biosynthesis. Conceivably, the expanded maturation period as well as less active cell wall metabolism-related genes (such as XTHs) eventually lead to a late-ripening CP with firm fruit, in contrast to early-ripening HC with less firm but crispy fruit texture. A simplified working hypothesis can be outlined as follows: (1) The genetic variations in plant hormone metabolism and responses, such as ethylene responses, auxin homeostasis, and transport, could fundamentally impact the apple fruit ripening process. (2) Cultivar-specific plant hormone metabolism and crosstalk will activate or suppress a unique set of TFs and/or signal transduction components. (3) Consequently, distinct sets of genes in various metabolic pathways, such as those encoding cell wall modifying enzymes of XTHs, are regulated in a cultivar-specific fashion. The overall outcomes of such differential gene expression ultimately determine the unique features of fruit ripening behaviors and quality attributes. Selected unigenes from current analysis are being aligned to apple genome sequences, and their expression profiles are being investigated in an HC × CP cross population consisting of 170 fruiting trees and other more diverse apple germplasm. The expression patterns of these genes will be analyzed against apple ripening and texture phenotypes to further define the specific function of these candidate genes.