Introduction

Peach fruits subjected to prolonged cold storage (CS) to delay decay and over-ripening are susceptible to developing a form of chilling injury (CI) called mealiness/woolliness (WLT; Fig. 1a): a flesh textural disorder characterized by lack of juiciness (Von Mollendorf 1987). Although some microscopic alterations are observed during CS (Luza et al. 1992), no macroscopic injuries are observed at that stage. Visual symptoms of WLT appear upon transfer of the fruits to room temperature for shelf life ripening (SLR) (Crisosto et al. 1999). Chilling-injured fruit do not go through the melting phase of ripening (Brummell et al. 2004b), a second phase of softening characterized by rapid loss of fruit firmness (Lester et al. 1994) in parallel with the climacteric peak, ethylene burst and juiciness development (Zhou et al. 2001). Woolly fruit still softens gradually, but unlike normal fruit has an unusual texture, with cells aggregating in clumps and reduced fracture upon application of pressure such as while eating (Brummell et al. 2004b).

Fig. 1
figure 1

Global gene expression analysis of ChillPeach transcripts in response to CS and subsequent SLR. a Woolly and healthy fruits. b The woolliness index (WI) of sensitive (S) and low sensitive (LS) fruit pools given 2 days SLR after storage for up to 3 weeks at 5 °C (reproduced from Pons et al. 2014). c, d PCA of the global expression profile showing the variation in each treatment, averaged from three replicates. In c, the first principal component (PC1) is shown on the x-axis while the second principal component (PC2) is shown on the y-axis; in d, the first principal component (PC1) is shown on the x-axis while the third principal component (PC3) is shown on the y-axis. M: mature fruits; R: mature fruits after 2–3 days ripening at 20 °C; CS1: M + 1 week cold storage at 5 °C; CS2: M + 2 weeks cold storage at 5 °C; CS3: M + 3 weeks cold storage at 5 °C; CSR1: M + CS1 + R; CSR2: M + CS2 + R; CSR3: M + CS3 + R; S sensitive and LS low sensitive

Full size image

Ultrastructural observation showed extensive changes to the middle lamella of WLT fruit during CS and subsequent SLR (Brummell et al. 2004b; Luza et al. 1992). The contact region between cells in woolly fruit was also smaller and the cells were more spherical, with loose attachment to their neighbors, while the intercellular spaces were enlarged and characterized by the presence of amorphous pectic substances, insoluble polysaccharides, cellulose and hemicellulose (Brummell et al. 2004b; Luza et al. 1992). The amount of pectins in the bigger intercellular spaces (Brummell et al. 2004b) and inside parenchyma cells close to vascular bundles increased dramatically in 5 °C-stored fruits in parallel with the protracted macroscopic CI indices they would develop upon subsequent SLR (Luza et al. 1992).

Most previous molecular work done on cold-induced fruit woolliness has focused on endo-polygalacturonase (endo-PG) and pectin methylesterase (PME) activities (Ben-Arie and Sonego 1980; Brummell et al. 2004a; Buescher and Furmanski 1978). These activities are required for normal ripening in melting flesh cultivars during the melting phase (Brummell et al. 2004a; Fishman et al. 1993; Orr and Brady 1993; Pressey and Avants 1978; Trainotti et al. 2003). Reduced ability to depolymerize insoluble homogalacturonan in the middle lamella and convert it into soluble pectins during CS and subsequent SLR has been attributed to low endo-PG activity and persistent pectin de-esterification by PME (Brummell et al. 2004b; Lurie et al. 1994, 2003). However, other activities and carbohydrates may contribute to cell wall alterations during CS and SLR (Brummell et al. 2004b; Gonzalez-Aguero et al. 2008; Obenland et al. 2003; Vizoso et al. 2009). These include altered pectic arabinose metabolism associated with WLT (Brummell et al. 2004b; Lurie et al. 1994), up-regulation of genes required to control mucilagous pectin structure, production and extrusion, and up-regulation of genes related to non-cellulosic cell wall polysaccharide biosynthesis and lignification that correlate with WLT sensitivity during CS (Pons et al. 2014, 2015). Several transcriptomic and proteomic studies have dissected the molecular mechanisms underlying WLT development during SLR (Falara et al. 2011; Gonzalez-Aguero et al. 2008; Nilo et al. 2010; Pavez et al. 2013; Vizoso et al. 2009). These studies identified different cellular functions as important for development of WLT symptoms, but did not deeply examine the problem and sometimes reported results were contradictory, possibly due to use of different peach varieties and shelf ripening protocols as discussed elsewhere (Pons et al. 2014, 2015). Moreover, these reports were based in a single genotype with or without CS, providing only a limited snapshot of WLT development. Thus, previous studies either failed to associate gene expression with WLT sensitivity or were based on two different peach cultivars with other phenotypic differences than CI responses.

To provide a more comprehensive view SLR after CS, we used pools of fruit from siblings in the Pop-DG population with contrasting sensitivity to develop WLT and performed a comprehensive transcriptome analysis to provide new insights into WLT disorder. The use of pools of samples from Pop-DG selected for contrasting sensitivity to CI compensates for other phenotypic differences randomly segregating in the population (Peace et al. 2005) and has proved useful to map WLT-associated QTLs (Dhanapal et al. 2012; Ogundiwin et al. 2009; Peace et al. 2005) and identify candidate genes for WLT tolerance/sensitivity identified at pre-symptomatic CS stages. Most important, they provide reliable information that can be confirmed in individual siblings of the Pop-DG population and other peach genotypes with different degrees of tolerance to CI (Dagar et al. 2013; Ogundiwin et al. 2008; Pons et al. 2014, 2015). To unveil the processes underlying actual WLT development, a time course transcriptomic analysis was conducted that covered different stages from unripe fruit at harvest through CS and subsequent SLR. Our study identified novel, stage-specific expression patterns whose functional analysis revealed new links between gene function and the woolly/non-woolly phenotype. Finally, medium throughput Fludigm RT PCR validated and extended our results for some individual siblings of the Pop-DG population characterized by different degrees of WLT. The predictive value of many of the woolliness-associated markers identified in the pools was extended to individual siblings with a range of chilling sensitivities.

Materials and methods

Plant material and CI/MI measurements

Fruit mesocarp samples from the following siblings of the Pop-DG mapping population (Peace et al. 2005) were used: 49/59, 84/85, 86/87 and 132/133 with high sensitivity to WLT (S) and 71/72, 88/89, 134/135 and 142/143 with low sensitivity (LS). Genotypes were selected for similar horticultural characteristics outside of sensitivity to woolliness development, although not in the incidence of other CI disorders such as flesh bleeding and flesh browning (Pons et al. 2014). To avoid potential distortions caused by maturity stage, fruits were harvested at the same commercial mature stage (M) as described (Kader and Mitchell 1989), with flesh firmness of 12–14 lb, soluble solids content (SSC) of 11–14 % and tritrable acidity (TA) of 0.5–0.7 %. The ripening stages and further characteristics of these fruits were as reported (Pons et al. 2014). Twelve fruits (M) were allowed to ripen at 20 °C directly after harvest to the edible firmness of 2–3 lb (14 N) and used as controls (R samples). For different degrees of WLT, M fruits were forced-air cooled at 0–2 °C within 6 h of harvest and stored at 5 °C with 90 % relative humidity for 1, 2 or 3 weeks (CS samples) and subsequently allowed to ripen during 2–3 days at 20 °C (CSR samples). The WLT phenotype of each fruit after the different CS periods plus SLR was expressed as the woolliness index (WI, (Crisosto et al. 1999)) and the proportion of measured fruits with WLT as described (Pons et al. 2014). Samples representing at least six fruits from each genotype and treatment were bulked, immediately frozen in liquid nitrogen and stored at −80 °C until RNA isolation.

Microarray hybridization, pre-processing and expression analyses

RNA purification, sample preparation and hybridization to the ChillPeach microarray were performed as described (Ogundiwin et al. 2008). For the microarray experiments, equal amounts of RNA purified from fruit representing each genotype were mixed in the corresponding sensitive (S) and low-sensitive (LS) pools. The RNA pools were hybridized using the ChillPeach microarray (Ogundiwin et al. 2008). All samples were compared using a dye-swap design against a common superpool reference composed of equal amounts of RNA obtained from all mesocarp samples. Three replicates from each sample pool were hybridized, one of them dye-swapped. Data processing and normalization was performed as described (Pons et al. 2014). Raw expression data of 3394 ChillPeach probes that met the threshold for hybridization quality is shown in Table S1.

Differentially expressed genes (Table S2) were identified from the raw dataset using the Significance Analysis of Microarray software (SAM package, (Tusher et al. 2001)) as described (Pons et al. 2014). PCA and 2D-hierarchical clustering were performed on significant data using Acuity (Axon instruments) and functional enrichment was performed and visualized using Matrix2png (Pavlidis and Noble 2003) as described (Pons et al. 2014).

Mid-throughput quantitative RT-PCR analysis using a dynamic array

Expression of 96 genes (Table S3) was analyzed by qRT-PCR using 96.96 dynamic arrays (Biomark Dynamic Array, Fluidigm Corporation) and 62 cDNA preparations corresponding to 31 samples: 13 genotypes of the Pop-DG population stored for 1 week at 6 °C or stored 1 week at 6 °C and shelf-life ripened 2–3 days at 20 °C, plus five pools (CS1-S, CS1-LS, CSR1-S, CSR1-LS and the reference superpool). Two biological replicates were included in each array for each individual genotype or pool, each representing at least three different fruits. Two replicated 96.96 Fluidigm dynamic arrays were used.

For Fluidigm analysis, 93 genes were selected based on our microarray results (Table S4). Three reference genes were selected to normalize qRT-PCR results based on the low variability in expression observed with the ChillPeach microarray under all conditions analyzed in this paper. These included a gene with unknown function (PPN036E09), an ABC1 family protein (PPN076G09) and an esterase/lipase/thioesterase gene (PPN078E12) (Ogundiwin et al. 2008). The flat expression pattern was confirmed by comparative ∆∆Ct as described (Livak and Schmittgen 2001). Fluidigm analysis were performed at the Hospital General lab in Valencia, Spain, and oligo pairs designed against each selected gene and relative gene expression values (Table S4) were determined using PerqExpress as described (Pons et al. 2014).

We extended the validity from results obtained in pools to individual lines with defined WLT index values (Figure S1). For each gene in a predefined expression set, Pearson correlation coefficients and associated p values were calculated between expression values for a gene in individual Pop-DG siblings and the WLT index they have after SLR (Table S5).

Results

Overview of transcriptomic changes: from harvest to cold storage and shelf-life ripening

An earlier report used two pools of peach fruits from the Pop-DG population (Peace et al. 2005) with known contrasting WLT tolerance to show extensive changes in the fruit transcriptome associated with WLT at the pre-symptomatic CS stage (Pons et al. 2014). Sensitive (S) and low sensitive (LS) pools in that experiment differed mostly in the time of onset of visual WLT symptoms (Pons et al. 2014), which was delayed by ~1 week in pools of LS fruits compared to the S pool (Fig. 1b, Pons et al. 2014). This suggested that LS fruits exhibited better chilling tolerance and ripening recovery ability during subsequent SLR than S fruits.

To analyze the molecular changes leading or not to development of visual injury symptoms during SLR after CS, we examined global gene expression by Bulk Segregant Gene Expression Analysis (BSGEA) (Pons et al. 2014) using the ChillPeach microarray (Ogundiwin et al. 2008) on the transcriptomes of the LS and S fruit pools when subjected to SLR after extended CS. This new dataset was analyzed and compared with the reanalyzed transcriptomes from LS and S pools during CS (Pons et al. 2014). To determine the effect of cold on ripening, room temperature control samples were included: Mature-harvested fruits (M) and fruits ripened for 2–3 days without previous CS (R). There were 3394 differentially expressed genes (DEG) in at least some comparisons using two criteria: a false discovery rate (FDR) <5 % and p value <0.05 (Table S1).

Principal component analysis (PCA; Fig. 1c, d) revealed that cold-induced WLT is associated with progressive changes in gene expression. Similarly to those described previously (Pons et al. 2014), the first principal component (PC1, 61 % of the variance) clearly separated fruit samples directly out of CS from M and R fruits (Fig. 1c). However, when CS fruits (in the right side of the PCA) experienced SLR (CSR fruits), the majority of cold-induced transcriptome alterations were not maintained (Fig. 1c, d); rather, CSR fruit samples grouped with non-cold stored fruits (as they had similar values in PC1). This indicates that temperature has a greater effect on the peach transcriptome than genotype or ripening, which are associated primarily with PC2 (16 % of the variance) and PC3 (8 % of the variance). This PCA also indicated that even though most transcriptomic changes caused by cold were reversed during SLR in both S and LS fruit and CSR-S and CSR-LS fruits followed parallel ripening programs, CSR fruits followed transcriptionally-diverging paths from mature (M) and ripe (R) fruits. These differences increased with time in CS (Fig. 1c, d). PC2 and especially PC3 revealed a considerable transcriptomic resilience (Franssen et al. 2011) in CSR-LS fruits, which partially returned to control R expression levels after all CS lengths (Fig. 1d). In contrast, fruit from CSR-S pools diverged even further from control R treatments during SLR. This suggests that during SLR, CSR-LS fruits may have a better ability to resume the molecular ripening program that was partially halted by CS. CSR-S fruits, in contrast, neither terminated the transcriptional response initiated during CS nor reinitiated the standard program during SLR. In addition, samples from cold-stored fruit arranged in PC1 and PC3 according to the eventual injury levels the fruits would suffer during SLR after CS (Fig. 1b, d). The same pattern of sample separation was observed previously when PC1 vs PC2 was plotted (Pons et al. 2014). This suggests that cold-induced transcriptional changes related to cold tolerance/sensitivity may affect how fruits recover from CS and this affects subsequent ripening. In PC3 (Fig. 1d), CSR3-S was closer to M fruits; thus, extended CS had a more dramatic effect on S fruits, which, in part, remained transcriptomically closer to M, with many cold-induced alterations not progressing or preventing ripening.

PCA also indicated that ripening of S fruits was slowed by CS, but to a lesser extent than in LS fruits. In PC3 (Fig. 1d), S fruits with 1 or 2 weeks CS (CS1-S and CS2-S samples) achieved a ripening pattern similar to R-S fruits (as they had similar values in PC3), while CS-LS fruits showed a similar ripening pattern to M-LS fruits, especially at 1 week. Therefore, as observed previously (Pons et al. 2014), the low temperatures used here are quite efficient in slowing at least part the ripening program of LS fruits, but less so in S fruit.

Alterations in post-harvest ripening during SLR after CS

To analyze to what extent CS primes fruits and modifies ripening during SLR, and to identify when these alterations begin, we analyzed the transcriptome of CSR fruits and compared them to the transcriptome of the corresponding R fruits (Fig. 2a; Table S2). This comparison showed that altered ripening during SLR after CS resulted mainly from an earlier global up-regulation of gene expression, while down-regulation occurred later. However, as CS continued, the number of genes downregulated increased, becoming much larger than the number of upregulated genes. This was more evident in S fruits than in LS fruits: expression of 396 genes was affected in CSR1-S fruit in contrast to only 20 genes in CSR1-LS fruits. In agreement with WLT evolution (Fig. 1b), most alterations occurring in CSR-S fruits after 1 week CS either did not occur in LS fruits at all, or occurred only after 2 or 3 weeks CS (Figure S2), together with visual symptoms of WLT.

Fig. 2
figure 2

Differences in the transcriptomic response of S and LS pools to SLR after CS. a Number of genes with altered expression in SLR fruits stored up to 3 weeks at 5 °C (CSR) compared to untreated ripe fruits (R). b Functional categories overrepresented in genes up- or downregulated in CSR fruits compared to R fruits. c Venn diagram showing the number of genes differentially expressed between tolerant and sensitive fruit after three lengths of CS followed by SLR. d Over-represented functional categories (p > 0.05) corresponding to genes differentially expressed between LS and S pools at each length of CS + SLR. Differentially expressed genes were obtained at FDR <0.05 and q value <0.05. Enriched functional categories were considered when Fisher test p values <0.05 and the number of genes was greater than three. R: mature fruits after 2–3 days ripening at 20 °C; CSR1: 1 week CS plus SLF; CSR2: 2 weeks CS plus SLR; CSR3: 3 weeks CS plus SLR; S sensitive pool and LS low-sensitive pool

Full size image

Functional enrichment indicated that RNA translation and protein assembly genes were upregulated in all CSR-S fruits over the corresponding R fruits, while this functional category was enriched in CSR-LS fruits previously exposed to CS for 2 or 3 weeks (Fig. 2b). In both S and LS fruits, the number of expressed genes in the RNA translation and protein assembly category increased in parallel with WLT development (Figs. 1b, 2b). In addition, antioxidant system-related genes had more genes with enhanced expression in CSR1-S samples. Cell wall-related genes were highly expressed in both CSR-LS and CSR-S pools after 2 weeks CS, although the number of genes with altered expression was much greater in CSR2-S fruits than in CSR2-LS fruits (Fig. 2b). In addition, energy production genes were upregulated in CSR3-S fruits compared to R-S fruits, but not in CSR3-LS fruits.

Carbohydrate metabolism genes were downregulated after all CS lengths compared to R fruits, with the number of downregulated genes increasing with the time in CS and severity of WLT symptoms (Fig. 2b). Downregulation of carbohydrate metabolism occurred later and to a lesser extent in CSR-LS fruits. Amino acid metabolism and signaling were also downregulated in CSR-S fruits after 2 or 3 weeks CS (Fig. 2b). Further, genes for post-translational protein modifications and cofactor and vitamin metabolism were downregulated in CSR3-S fruits. In CSR-LS fruits, in addition to carbohydrate metabolism, only genes of unknown function enriched the set of genes downregulated in CSR-LS fruits compared to untreated ripe fruits (Fig. 2b).

Stage-specific differences in the transcriptomic response to SLR after CS between the pools of contrasting genotypes

We next performed one-to-one direct comparisons of the transcriptome of S and LS fruits ripened after each length of CS (Fig. 2c, d; Table S2). Since the primary difference between these pools of fruits was the time of onset of WLT, we reasoned that only WLT-dependent transcriptomic differences should be found between the two pools of fruits. The transcriptomes of S and LS fruits after SLR tended to converge from 1 to 2 weeks CS, consistent both with their converging WI values and with the increased number of genes with altered expression in CSR2-LS over R-LS (Fig. 2c). The set of genes expressed at higher levels in CSR-S fruits after 1 week CS, the stage of maximum differences in WLT (CSR1-S WI = 0.8 vs CSR1-LS WI = 0; Fig. 1b), was enriched in genes related to transport, cell wall and secondary metabolism (Fig. 2d). Genes differentially expressed in CSR2-S samples were also enriched in cell wall-related genes. Interestingly, between 2 and 3 weeks CS, the number of genes differentially expressed between S and LS increased (Fig. 2c), although both S and LS fruits were severely woolly and differences in WI were reduced (Fig. 1b). After 3 weeks CS and SLR, the set of highly expressed genes in CSR3-S fruits were enriched in RNA translation, protein assembly and energy production, while amino acid metabolism, organic acid metabolism, energy production and signal transduction functions (Fig. 2d), although downregulated below R samples (Fig. 2b), were more prevalent in LS than S fruits with the same postharvest record (Fig. 2d).

The impact of CS on subsequent SLR

WLT in CS fruit develops during SLR for 2–3 days at 20 °C, suggesting that expression of specific genes involved in injury could occur during post-harvest ripening. The identification of genes involved in ripening and their expression patterns provide important clues about how CS affects SLR and its causal relationship to fruit quality during SLR. For this purpose, the dataset of DEG was split into two subsets. The first contained 859 ripening-regulated genes previously identified (Pons et al. 2014) as up- or downregulated in R fruits in unchilled ripe fruits after 2–3 days at 20 °C compared to M fruits (Table S2). Peach fruits were partially ripe after 3 days SLR (Borsani et al. 2009), so this dataset was dubbed as early post-harvest ripening genes (eRG). The second dataset, dubbed non-early post-harvest ripening genes (NeRG), consisted of the remaining 2532 genes with no significant expression differences between R and M fruits. The eRG and NeRG datasets across the different sampleswere subjected to bidimensional hierarchical cluster analysis (2D-HCA).

In agreement with the classification criteria, 2D-HCA organized by eRG genes (Figure S3A) separated samples by physiological state (SLR or no SLR), but also according to length of CS before SLR. Thus, M and CS fruits clustered apart from ripened fruits; CSR1-S fruits with R and CSR1-LS, but in a distant sub-cluster; and CSR2 and CSR3 separated from R according to tolerance/sensitivity (WI values). 1 week CS did not dramatically affect expression of eRG in S or LS SLR fruits, which clustered together despite their very different WI.

The 2D-HCA obtained with the NeRG transcript profiles (Figure S3B) indicated that R fruits are closer to M (unripe) fruits than to CS fruits, with or without SLR, and that major differences were detected when their transcriptome was compared to that of chilled fruits (CS and CSR series). Most important, NeRGs segregated WLT fruits from non-WLT fruits (Figure S3B), confirming the PCA analysis results. This indicates that NeRGs may play an important role in development of visual WLT symptoms.

As one purpose of CS is to stop or slow ripening until fruit is returned to room temperature, the eRG were classified depending on whether CS stopped their normal expression during ripening in control fruit. In LS fruits, the ripening response of 700 genes was stopped or delayed during CS (81.49 %) (eRG1, eRG2, eRG5, eRG6; Fig. 3a; Table 1), while in S fruits, only 517 genes were affected during CS (60.19 %) (eRG1, eRG5, eRG6; Fig. 3a; Table 1). The response of eRG during both CS and subsequent SLR (Table 1) indicated that of 700 genes in LS fruits whose characteristic ripening trend was stopped or delayed by one, 2 or 3 weeks CS, 99.29, 91.71 and 78.86 %, respectively, resumed the normal ripening pattern during later SLR. In S fruits, resumed their ripening trend only 79.50, 61.12 or 35.01 % of the 517 genes with ripening-related changes stopped or delayed by one, 2 or 3 weeks CS., respectively.

Fig. 3
figure 3

Expression analysis of early (eRG) and non-early ripening (NeRG) ChillPeach transcripts in response to CS and SLR. Clusters resulting from unsupervised two-dimensional hierarchical clustering of eRG and NeRG are shown. a, b Clusters of eRG and functional categories over-represented in each cluster. c, d Clusters of NeRG and functional categories over-represented in each cluster. Y-axes (a, c) represent the normalized expression ratio (Log2 M) of three biological replicates in relation to a reference pool. The functional categories overrepresented in each cluster (b, d) are shown as a heatmap obtained with matrix2png. Enriched functional categories with Fisher test p values <0.05 are colored in grades of yellow. The number of genes in each cluster is indicated to the right of the heatmap

Full size image
Table 1 Early and non-early ripening genes recovering the R expression levels during SLR after 1, 2 or 3 weeks in CS
Full size table
Table 2 Differences between the S and LS pools in ripening behavior of the 859 eRG according to the length of cold storage
Full size table
Table 3 Differences between the S and LS pools in ripening behavior of the 2532 NeRG according to the length of cold storage
Full size table

The eRG subset that continued ripening during CS at 5 °C, as if at room temperature, accounted for 159 or 342 genes in LS or S fruits, respectively (Table 2). In LS fruits, >85 % of these genes resumed normal ripening-related expression during SLR after all CS lengths, while only 62.87, 43.86 or 24.27 % showed normal ripening expression in S fruits after 1, 2 or 3 week CS (Table 1). Thus, during CS, some components of transcriptomic ripening may become unsynchronized, with some parts proceeding during CS while others were arrested. The degree of unsynchronization is S- or LS-dependent and will impact later SLR.

Classifying NeRG by whether or not they reach the expression levels in non-CS fruit (i.e., similar expression to M or R fruits, which are very similar during the first week) indicated that over 90 % of NeRG in CSR-LS fruits achieved similar expression to M/R fruits after all CS lengths (Table 1). In S fruit, 93.57 (2372), 81.07 (2055) or 56.09 % of NeRG were expressed at the level of M/R fruits after 1, 2 or 3 weeks CS.

To assign functionality to this differential response, we combined the cluster analysis results (Fig. 3a, c) with the genes identified as having altered expression (Fig. 2a). Of particular interest were genes in clusters eRG2 and NeRG8 (Fig. 3a, c). Most genes with enhanced expression in CSR fruits were in clusters eRG2 (~50 % of eRG with enhanced expression) and NeRG8 (~35 % of NeRG with enhanced expression) (Tables 2, 3). The genes in these clusters were enriched in chromatin status and regulation, energy production, RNA post-transcriptional regulation, RNA translation and protein assembly (Fig. 3b, d). A particular characteristic of these clusters is that LS fruits maintain the expression levels observed in M fruit during CS, while the same genes were strongly downregulated in S fruits (Fig. 3a, c). Expression of genes in clusters eRG2 and NeRG8 was markedly stimulated during SLR after being repressed during CS. During SLR, these genes were much more strongly expressed than in M and R fruits and the number of genes with altered expression increased during SLR concomitant with increasing MI and time. Taken together, genes in this class exhibited the strongest response to CS and SLR and are perhaps associated with macroscopic WLT development.

Clustering analysis also indicated that in those eRG and NeRG genes with transcriptional responses reversed during SLR, differences in expression between S and LS fruits were mainly quantitative, except in clusters NeRG1 and NeRG5 (Fig. 3a, c). Most genes with quantitative differences in expression during SLR were in clusters eRG1, eRG5, eRG6, NeRG2, NeRG3, NeRG4, NeRG6 and NeRG7 (Fig. 3a, c; Tables 2, 3). Genes in clusters eRG5, eRG6, NeRG3 and NeRG4 were upregulated during CS, while eRG1, NeRG6 and NeRG7 were downregulated (Tables 2, 3). A functional enrichment analysis (Fig. 3b, d) indicated that genes in CS- upregulated clusters were enriched in antioxidant system, RNA transcription regulation, trafficking machinery, transport, signal transduction, structure maintenance proteins and unknown function. Genes in CS-downregulated clusters were enriched in amino acid metabolism, carbohydrate metabolism, cell wall, lipid metabolism, other nucleic acid metabolic processes, secondary metabolism and signaling.

Four clusters account for the main differences observed between pre-symptomatic S and LS fruits. For eRG3, eRG4 and NeRG2, the response initiated during CS is not reversed during SLR and expression differences during CS and SLR are mainly quantitative (Fig. 3a, c). The genes in clusters eRG3 and eRG4, down- and upregulated during CS in both S and LS fruits, are enriched in other nucleic acid metabolic process, secondary metabolism and signaling (Fig. 3b, d). These continued the ripening-related program during CS and SLR (Table 2). However, the effect of CS on ripening-related expression of these genes is greater in S than in LS fruits, especially after prolonged CS. Genes in cluster eRG4 (Fig. 3a), always expressed at high levels in S fruits, were progressively downregulated in CSR-S fruits with respect to R-S fruits as the time in CS increased. NeRG2 genes were repressed during CS in a manner correlating with CI sensitivity, only recovered part of the expression seen in LS fruits during SLR, and repression during CS was greater with increased duration of CS (Fig. 3c). Clusters NeRG1 and NeRG5 showed quantitative differences, but special behavior. Although the transcriptional response initiated during CS is generally reversed during SLR, the genes in cluster NeRG1 (not enriched in any functional category; Fig. 3d) showed opposite patterns during CS: they were highly upregulated during CS in S fruits, but downregulated in LS fruits. In genes of cluster NeRG5, CS caused downregulation of expression, particularly in S fruits, but after SLR, these genes were more highly expressed in S than in LS fruits (Fig. 3c). As WLT symptoms and time in CS increased, the number of genes with enhanced expression also increased.

Validation and extension of microarray expression profiling to individual lines from the Pop-DG population

To validate the microarray results, we performed medium-throughput qRT-PCR over a set of 93 genes that were randomly selected (with the exception of a few genes of interest) for being differentially expressed after 1 week CS and/or SLR (Table S3) in the pooled fruit samples. The qRT-PCR results obtained from the pools indicate that 73.12 % (68/93) and 75.26 % (70/93) of the genes in CS and CSR samples had the same expression pattern in the microarray as in qRT-PCR, respectively (Table S4). Although the expression level was not confirmed for each gene in each sample, these results confirm the general validity and robustness of our microarray data.

Expression of genes validated in the pools was also measured by qRT-PCR in fruits of 13 individual siblings from the Pop-DG population, including those used in the pools and others showing a range of WLT, that were stored and ripened under identical conditions as used for microarray analysis (Figure S1). To assess whether gene expression profiles were related to WLT sensitivity phenotype, correlations between the relative expression of candidate genes in each sibling and their WLT sensitivity were determined at pre-symptomatic and symptomatic stages (Fig. 4; Table S5). In general, genes for which expression was validated as highly expressed in S pools (in yellow) showed positive correlations with WLT at the pre-symptomatic stage (Fig. 4a), the symptomatic stage (Fig. 4b) or both (Fig. 4c), while genes low-expressed in S pools with respect to LS pools (blue) showed negative correlations. A gene-WLT association was considered significant when p < 0.05 (corresponding to a correlation >0.2; Table S5). The WLT association was confirmed for 63 % of genes in CS samples, 51 % of genes in CSR samples and 38 % for genes selected for differential expression at both CS and SLR (Fig. 4; Table S5). In all comparisons, there was a group of genes that were both highly expressed in LS pools and positively correlated with WLT in individual lines. This indicates that high expression of these genes is not necessarily functionally connected with tolerance degree, but may rather reveal intrinsic differences in the cold response in each line. For genes showing an opposite pattern during CS and SLR (Fig. 4c), positive correlations with WI probably reflect that the effect of cold over these genes is more important for WLT development than ripening, or vice versa.

Fig. 4
figure 4

Correlations between relative expression of candidate genes analyzed by medium throughput Fludigm RT-PCR in each sibling and the WI exhibited after 1 week CS plus SLR. Genes are plotted in X axis; Pearson correlations in Y axis. Pearson correlations were calculated over the sets of genes validated in the pools and differentially expressed during ripening. a 69 genes validated in CS pools, (b) 71 genes validated in SLR pools and (c) 52 genes validated in both CS + SLR pools. Genes highly expressed in S pools are shown in yellow, genes highly expressed in LS pools are shown in blue. Genes with opposite responses during CS and SLR are showed in orange and green

Full size image

Discussion

Transferring fruits to CS induced a large transcriptome and proteome rearrangement in peach fruits consistent with eventual WLT development (Dagar et al. 2013; Falara et al. 2011; Gonzalez-Aguero et al. 2008; Nilo et al. 2010; Ogundiwin et al. 2008; Pavez et al. 2013; Pons et al. 2014, 2015; Vizoso et al. 2009). Although ripening was delayed and the transcriptome was reconfigured by CS, development of visual symptoms of WLT only occurs afterward, during SLR when fruits resume ripening (Crisosto et al. 1999).

To investigate what happens during SLR after CS and what genes/functions are probably involved in WLT development, we expanded our previous work on CS (Pons et al. 2014, 2015) to SLR. An analysis of fruit transcriptome changes in response to CS and SLR is presented for two pools of siblings from the Pop-DG population with contrasting sensitivity to cold-induced woolliness. The transcriptomes of S and LS pools at harvest and after three lengths of CS with or without (the same data used in Pons et al. 2014) subsequent SLR were analyzed together using the ChillPeach microarray (Figs. 1, 3). The expression profiles were validated for 73–75 % of the genes tested (Table S4). The expression patterns of single genes analyzed by qRT-PCR were consistent with microarray data (Table S4), although not confirmed for each gene in each sample as in similar studies (Ogundiwin et al. 2008). Further, the expression pattern of most genes validated in the pools was confirmed in individual fruit/genotypes within the Pop-DG population (Fig. 4). Although there are technical limitations to using microarray (discussed in Pons et al. 2014, 2015) and pooled samples (Zou et al. 2016), the combination of pools from the Pop-DG population with the ChillPeach microarray has proved robust, reproducible (Table S4; Dagar et al. 2013; Ogundiwin et al. 2008; Pons et al. 2014, 2015) and applicable to individuals within the same Pop-DG population (Fig. 4; Pons et al. 2014) and cultivars with different sensitivity toward CI (Dagar et al. 2013; Pons et al. 2015). Thus, data presented here can be interpreted with confidence.

Transcriptome resilience during SLR promotes the resumption of ripening and delays woolliness development

One of the purposes of CS is to stop or slow ripening so that ideally, when fruits are returned to room temperature, they can resume ripening and eventually produce a sound, ripe fruit. However, the intervening cold stress affects subsequent ripening. The response during SLR after CS ranged from: (i) returning to the expression levels they would have without CS, although less efficiently in S fruits to (ii) continuing with the CS-induced changes, although the expression levels in SLR correlated with those in CS and with sensitivity. These responses are comparable with those observed in other tissues. De-acclimation of leaves from cold stress to warm temperatures (Kalberer et al. 2006; Rapacz 2002) mainly reflects a reversal of what occurred during cold acclimation (Oono et al. 2006) and the ability to recover from the cold is inherent in cold sensitivity (Kratsch and Wise 2000). Similarly, when transferred from CS to SLR, CSR-S fruits could not properly terminate their transcriptional response to cold and resume normal ripening (Figs. 1, 2, 3). CSR-LS fruits showed considerable transcriptomic resilience (Franssen et al. 2011), with many genes returning to R expression levels after CS, particularly for short times (Figs. 1c, d, 3a, c; Figs. S2, S3). As the duration of CS increased, the transcriptomes of CSR fruits diverged further from control R treatments and the likelihood of recovery decreased due to sustained negative effects.

We recently reported that quantitative differences in the subset of common cold-responsive genes correlated with sensitivity/tolerance to CI at pre-symptomatic stages (Pons et al. 2015). If this holds true for SLR, genes showing quantitative differences between S and LS pools during CS and subsequent SLR and altered expression between S fruits and R fruits would be good candidates for woolliness tolerance/sensitivity and also for fruit recovery and reactivation of ripening during SLR (Pavez et al. 2013). The best clusters to examine for genes associated with cold tolerance and recovery-specific effects are those including genes highly expressed in the LS pools during CS and SLR. Clusters eRG1, eRG3, NeRG2, NeRG6 and NeRG7 (Fig. 3; Table S2), all downregulated during CS, either returned to normal expression during SLR, especially after shorter CS lengths, and/or were expressed at higher levels in LS fruits than in S fruits. Many genes in these clusters were related to cell wall, amino acid, carbohydrate, lipid and secondary metabolism. This indicates that the transcriptional ability required to achieve normal fruit metabolism and cellular function during ripening was partly restored in LS fruits after CS. Several genes in these clusters (Table S6) were recently reported as associated with cold tolerance in peach fruits (Pons et al. 2014), such as the orthologs of CBF1 (NeRG7), CAMTA2 (NeRG6 and NeRG2), SZF1 (NeRG6), HMGB1 (NeRG6), SKIP (NeRG6), ATAF1 (NeRG6), GI (eRG1) and ICE1 (NeRG2). Although these genes are associated with stress tolerance in Arabidopsis (Buchanan et al. 2005; Cao et al. 2005; Doherty et al. 2009; Feng et al. 2015; Gilmour et al. 1998; Lildballe et al. 2008; Sun et al. 2007) and to growth repression under unfavorable environmental conditions in vegetative tissues (Achard et al. 2008; Fornara et al. 2015; Hsieh et al. 2002; Kodaira et al. 2011; Lildballe et al. 2008), they are also associated with the ability to recover from cold stress. Overexpression of ICE1, CBF1 or CAMTA in plants increased the recovery rate after cold and drought stress in several species (Hsieh et al. 2002; Huang et al. 2013, 2015; Jaglo-Ottosen et al. 1998; Navarro et al. 2011; Pandey et al. 2013). In addition, CAMTA genes regulate drought recovery (Pandey et al. 2013) and CBF genes regulate the springtime de-acclimation phase in leaves and dormant buds (Polashock et al. 2010; Welling and Palva 2008).

Genes associated with cold tolerance and recovery-specific effects could be active in reinitiating ripening. CBF, ATAF1 (or SlNAC4) and bZIP11 orthologs (clusters NeRG6 and NeRG7, with low expression in S; Table S6) also participate in normal ripening of fleshy fruits by regulating processes such as cell wall remodeling, fruit softening, ethylene production, carbohydrate metabolism and secondary metabolism (Lovisetto et al. 2013; Manganaris et al. 2011; Tacken et al. 2010; Zhao et al. 2009; Zhu et al. 2014). Further support that the genes in these clusters are probably involved in reinitiating ripening after CS comes from the identification among them of orthologs of key fruit ripening genes, some described in peach. Examples include orthologs of FRUITFUL (FUL, in eRG3, downregulated in a WLT-sensitive manner) and SQUAMOSA PROMOTER BINDING PROTEIN (SPB)-LIKE 1 (SPL1, in clusters eRG1 and NeRG6, both with low expression in S fruit). Tomato FUL homologs act as key regulators of many ripening processes (Bemer et al. 2012; Fujisawa et al. 2014; Shima et al. 2014) and PpFUL accumulated in peach mesocarp during early fruit development and (Dardick et al. 2010). In peach, a gene encoding a putative TF containing the SBP domain is located in the region in which the major QTL controlling fruit maturation time was mapped (Romeu et al. 2014). Moreover, mutations in the tomato key ripening regulator SlycSBP-CNR (color non-ripening) (Manning et al. 2006) results in fruits with a mealy/woolly texture and complete absence of carotene biosynthesis (Fraser et al. 2001).

Conversely, clusters corresponding to genes upregulated during CS that remain highly expressed in both pooled and individual line S fruits (eRG4, eRG5, eRG6 and NeRG4; Fig. 3; Table S7) are probably related to sensitivity to developing WLT and negatively related to recovery. These clusters include genes whose functions are linked to stress sensitivity and low recovery ability, such as genes encoding members of the SnRK system in clusters NeRG4 and eRG5, previously linked to WLT sensitivity at a pre-symptomatic stage (Pons et al. 2014, 2015), and ortholgs of Arabidopsis SHORT VEGETATIVE PHASE/AGAMOUS-LIKE 24 genes in cluster NeRG4 (Table S7), Transgenic plants overexpressing SnRK1 and with decreased trehalose 6-phospate (in clusters eRG1 and NeRG6; Table S9) are impaired in immediate growth recovery when transferred from cold to warm conditions (Nunes et al. 2013). Furthermore, low expression of PpDAM5 (dormancy-associated MADS-box genes) and PpDAM6, the peach homologues of SVP/AGL24 genes, correlated with the ability of buds to resume growth under favorable conditions (Yamane et al. 2011).

Hormonal de-synchronization of ripening after CS may contribute to woolliness

The SLR response of peach fruit after CS is more complex than a simple recovery leading to ripening. During and after CS, many components of ripening were unsynchronized: some progressed, while others were arrested (Tables 1, 2). Many genes related to hormone biosynthesis, subcellular distribution, hormone sensing and signaling were in gene clusters that were severely altered in S fruits during CS and subsequent SLR compared to R fruits (Tables S6, S7). The most-affected hormone pathways were those related to ethylene and auxin. Proper regulation of these hormones is essential for normal peach ripening (El-Sharkawy et al. 2008; Tatsuki et al. 2013; Tonutti et al. 1997; Trainotti et al. 2007; Ziliotto et al. 2008). In melting Prunus fruits, high auxin levels initiate ripening by inducing system II ethylene production. This triggers the transition from the pre-climateric S3 phase to the climacteric S4I stage, or commercial maturity. Auxins then work independently and synergistically with ethylene to accelerate ripening (S4II or melting phase) during the second phase of softening, characterized by a rapid loss of fruit firmness (Lester et al. 1994) and juiciness development (Zhou et al. 2001).

During SLR following CS, WLT fruits do not achieve melting phase (Brummell et al. 2004a). Reduced ethylene production in WLT fruits during CS and SLR is associated with the bypass of the melting phase that leads to WLT development (Dong et al. 2001; Zhou et al. 2001). Altered expression of key elements related to ethylene biosynthesis, signaling and regulation is found in WLT fruits during both CS and SLR, some of which correlate with WLT sensitivity (Dong et al. 2001; Nilo et al. 2010; Pons et al. 2014, 2015; Vizoso et al. 2009; Zhou et al. 2001).

Woolly peach fruits during CS and SLR have fewer ACO transcripts and less activity and more ACS than juicy fruits (Dong et al. 2001; Zhou et al. 2001). Similarly, mealy CNR tomato fruit mutants had low expression of ACO and FUL genes during ripening, but high expression of ACS (Eriksson et al. 2004). Thus, the effect on ethylene is a common feature of the mealiness/woollines phenotype. The expression of orthologs of EFE/ACO4 (cluster eRG1) and FUL (cluster eRG3; Fig. 3a; Table S6) genes was greater in LS fruits during SLR; this was confirmed by RT-PCR in the pooled samples and in different individual lines from the Pop-DG population (Tables S3, S5). However, both higher and lower expression was found for two orthologs of ACS (NeRG4 and eRG4, respectively; Fig. 3; Table S7) and confirmed by RT-PCR in individual lines (Tables S3, S5). Different ACS isoforms regulate basal ethylene levels (system I) and autocatalytic ethylene production (system II) in fruits such as plum, apricot and tomato (Barry et al. 2000; El-Sharkawy et al. 2008; Muñoz-Robredo et al. 2012) and transition from ethylene system I to system II depends on down-regulation of system I ACS and activation of system II ACS (Barry et al. 2000). Thus, ACS isoforms could be involved in different ethylene production systems. ETO1, associated with chilling sensitivity at the pre-symptomatic stage (Pons et al. 2014) was highly expressed in S fruits during SLR (cluster NeRG4; Table S7). The Prunus salicina ETO1-like protein (Ps-EOL1) contributes to determine the suppressed-climacteric pattern in late plum cultivars (El-Sharkawy et al. 2008; Kim et al. 2015), since down-regulation of Ps-EOL1 is required for transition from system I to system II. In addition, EIN2, a positive regulator of both ethylene signaling, via negative feedback regulation of ethylene system I (Thain et al. 2004) and autocatalytic (system II) ethylene production (Alonso et al. 1999), was also highly expressed in S fruits during SLR (cluster eRG4; Fig. 3; Table S7). Expression of ppEIN2 transcripts increased in CS fruits (Begheldo et al. 2008; Pons et al. 2015) and especially in WLT fruits (Begheldo et al. 2008; Pons et al. 2015). ppEIN2 expression in CS fruits dropped when autocatalytic ethylene production was initiated (Begheldo et al. 2008), while greater ppEIN2 induction correlated with basal ethylene production (Begheldo et al. 2008). Collectively, these findings suggest that transition from ethylene production system I (basal autoinhibitory) to system II (autocatalitical) is affected in fruits developing WLT. This hypothesis is consistent with low ethylene production and lack of ethylene burst in transgenic tomatoes suppressed for FUL or ATAF1/SlNAC4 genes (Shima et al. 2014; Zhu et al. 2014), whose orthologs here were low-expressed in S fruits during CS and SLR (clusters eRG3, NeRG6 respectively; Fig. 3; Tables S6, S7).

Altered expression of genes related to auxins is linked to the WLT phenotype (Pavez et al. 2013) and differential expression of auxin distribution and signaling genes is involved in tolerance/sensitivity at pre-symptomatic stages (Pons et al. 2014, 2015). Several auxin-related genes had altered ripening patterns during SLR, especially in S fruits, reinforcing its role in the WLT phenotype. Many genes related to auxin conjugation, negative regulators of auxin responses, proteasomal signaling and proper auxin distribution (clusters NeRG2, NeRG6 and NeRG7; Table S7) failed to reach the expression in R fruits after being, in general, downregulated by CS. In contrast, genes related to auxin de-conjugation, some nuclear signaling elements and plasma membrane and endoplasmic reticulum auxin efflux carriers (eRG4 and NeRG4; Fig. 3; Table S7) were more expressed in CSR-S fruits after being upregulated by CS.

Although auxins were not determined in this paper, the increased expression of some auxin-related genes is considered diagnostic for increased auxin concentrations in the tissue from which the RNA was obtained (Tadiello et al. 2016). Enhanced expression of genes encoding for Aux/IAA proteins such as IAA16, IAA27/PAP, and IAA17/AXR3 (eRG3; Fig. 3; Table S6) and conjugating enzymes like IAA-amido synthase GH3.1 and IAMT (NeRG6 and NeRG7, respectively; Fig. 3; Table S6) suggested more free auxin in fruit from the most tolerant pool during SLR after CS. Conversely, the higher expression in CSR-S fruits of de-conjugating and biosynthesis enzymes such as IAR3, ECH2 (NeRG4) and IBR1, MES17 and ILL3 (eRG6; Fig. 3; Table S7), the endoplasmic PILS5 and PILS6 (NeRG4; Table S7) and plasma membrane efflux transporters PIN3 and PGP19 (eRG4; Table S7) indicates that cytoplasmic auxin concentrations are low in comparison to R or CSR-LS fruits, while higher concentrations may be found in cell wall and endoplasmic reticulum. The expression profile of PILS 6 was further confirmed by qRT-PCR of pools and individual lines (Tables S3, S5). Enhanced expression of IAA27/PAP in LS fruits (Table S6) and its negative correlation with MI during CS and SLR was further confirmed in individual lines from the Pop-DG population (Table S3; Pons et al. 2014, 2015).

In melting Prunus fruits, high free auxin levels are essential to induce system II ethylene production (El-Sharkawy et al. 2008; Tonutti et al. 1997; Trainotti et al. 2007; Ziliotto et al. 2008). This suggests that auxins in LS fruits during CS and subsequent SLR probably contribute to tolerance and proper ripening. Further experiments are required to confirm this hypothesis.

The altered ripening during CS could be more complex than just an imbalance in ethylene and auxin. Other hormones such as ABA, gibberellins, brassinosteroids and jasmonate may play additional roles in WLT sensitivity and development, as suggested by differential expression of some genes related to their biosynthesis and signaling (Tables S6, S7).

The molecular phenotype of woolliness: alterations in cell wall during development of visual symptoms

Consistent with previous findings (Brummell et al. 2004b; Gonzalez-Aguero et al. 2008; Vizoso et al. 2009), enhanced cell wall (CW) metabolism was associated with WLT development (Fig. 2b, d), but there was also depletion of other CW gene functions (cluster eRG1; Fig. 3b). Some cell wall alterations detected during visual symptom development were initiated during CS and maintained during SLR. CS fruits had pectins that were different from M or R fruit; during SLR, these pectins depolymerized in a manner similar to R fruit when fruits did not develop woolliness, but not when they did (Brummell et al. 2004b). Genes involved in pectin and mucilage modification and hemicellulose biosynthesis, previously associated with sensitivity at the pre-symptomatic stage (Ogundiwin et al. 2008; Pons et al. 2014, 2015), such as BXL1 (eRG4, should be upregulated during ripening), PGIP1 and CSLE1 (eRG6, should be downregulated during ripening), failed to reach R levels after being upregulated by CS, especially in CSR-S fruits (Table S8). Expression of BXL1 and CSL1 in CSR samples is further confirmed by qRT-PCR in pools (Table S3) and individual lines (Table S5) and correlates positively with WLT (Fig. 4). In contrast, CWR genes associated with sensitivity during CS but related to control of mucilage extrusion and production and several genes related to lignification (Pons et al. 2015) returned to normal or even lower expression in S fruits during SLR (Table S8).

Classifying CW genes by putative role (Figure S4) shows that the main differences between S and LS fruits during SLR are due to pectin metabolism genes: S fruit were enriched for highly expressed genes related to pectin degradation, while LS fruits were enriched in genes related to pectin biosynthesis. In addition, S fruits had altered expression in more genes related to extracellular matrix, while LS fruits had more in genes related to biosynthesis of cell wall polysaccharides and precursors.

The association of these enzyme activities with WLT provides new information about mechanistic alterations occurring during WLT development. Loss of endoPG activity is a major mechanism underlying WLT phenotype (Ben-Arie and Sonego 1980; Brummell et al. 2004b). In addition, mealiness/WLT in peach and in the cnr tomato mutant correlated with low pectin methylesterification (Orfila et al. 2002, 2001). Among pectin degradation-related genes highly expressed in S fruits during SLR (Table S8), we find genes such as TBL39 (eRG6), TBL13 (NeRG4), ATC/VIF1 (eRG6), PME1 and PME12 (NeRG4), several pectinases and pectate lyases (eRG4, eRG5, eRG6 and NeRG4) and the ortholog of PGIP1 (eRG6). PIGP1 is associated with sensitivity at the pre-symptomatic stage (Ogundiwin et al. 2008; Pons et al. 2014, 2015). To effectively regulate PG activities, PIPG proteins must bind to a homogalacturonan region characterized by high O-acetylation and low methylation (Sakamoto et al. 2003). TBL proteins are involved in O-acetylation of polysaccharides such as pectins and xyloglucans (Gille and Pauly 2012).

Highly methylated pectins help maintain cell wall-to-cell wall bonds (Willats et al. 2001) and O-acetylation correlates negatively with cell adhesion (Liners et al. 1994). Loss of endoPG activity combined with continued pectin de-esterification by (PME) leads to accumulation of gel-forming pectic compounds associated with the WLT phenotype (Ben-Arie and Sonego 1980). Although PME activity contributes to the WLT phenotype, their transcriptional regulation is not a primary contributor to the methylesterification degree of pectin during SLR (Vizoso et al. 2009). Our results suggest that low methylated pectin during WLT results from repression of pectin biosynthesis, initially synthesized with a high degree of methylesterification (Vincken et al. 2003). Genes highly expressed in CSR-LS fruits include three of the GAUT family (GAUT9, GAUT15 and GATL1, localized in clusters eRG1, NeRG6 and NeRG7) and QUA2 (NeRG6; Table S8). Expression of an ortholog of QUA1 in the GAUT family (Leboeuf et al. 2005) correlated negatively with WLT in the qRT-PCR analyses of individual lines (Fig. 4). Cumulative evidence indicates that GAUT and QUA2, encoding pectin methyltransferases (PMT), are co-regulated during processes leading to fully methylesterified pectin (Mouille et al. 2007). GAUT genes may also be involved in the altered pectin composition of woolly fruits (Brummell et al. 2004b). Ripe tomato fruits from GAUT4-silenced lines had significantly less pectin, with reduced Rha, Gal and Ara (de Godoy et al. 2013). GAUT Arabidopsis mutants had cell walls with reduced galacturonic acid and very reduced cell adhesion (Bouton et al. 2002; Leboeuf et al. 2005).

The cell WLT phenotype thus is very complex and entails many actors (Table S8), including both biosynthesis and degradation of glycan, cellulose, hemicelluloses, cell wall precursors and extracellular matrix modification. Genes such as CHI-B (Del Campillo and Lewis 1992; Eriksson et al. 2004), BXL1 (Arsovski et al. 2009) and RHCA1 (Corbacho et al. 2013) are required for cell wall disruption and separation. Together, results suggest these genes modify the cell wall for the action of other CWR enzymes, which in turn affect cell adhesion and biosynthesis of non-pectic polysaccharides and generate a signal to coordinate cell wall WLT phenotype.

The expanded molecular phenotype of woolliness: progressive down-regulation of genes related to carbohydrate, amino acids, cofactor metabolism and organic acid are related to WLT development

Soluble sugars, organic acids and amino acids contribute to the overall organoleptic quality of fresh peach. During normal postharvest ripening, fermentative metabolism is induced: sucrolytic degradation of sucrose by sucrose synthase is downregulated while sorbitol, sucrose hydrolyzed by invertases and amino acids, may be used as carbon fuel (Borsani et al. 2009; Lombardo et al. 2011). Several glycolysis steps are downregulated and glucose and fructose metabolism occur via glycolysis shunt alternatives, then through the pyruvate orthophosphate dikinase (PPDK) and phosphoenol pyruvate carboxylase [PEPC]–MDH–ME glycolytic bypasses, which maintains pyruvate production for energy and the fermentative pathway, which is required for aromatic volatile development (Sánchez et al. 2012; Sanchez et al. 2013).

Comparison of metabolites and genes/activities measured during normal postharvest ripening of fruits kept at 20 °C (Borsani et al. 2009; Lombardo et al. 2011), fruits cold-stored at 0 °C with subsequent ripening (Lauxmann et al. 2014) and transcript changes observed here in R fruits indicates that, outside of genotypic differences, there is good agreement in metabolites and genes/activities. Serine, trehalose, galactinol and its biosynthetic enzyme GolS did not change during normal ripening (Borsani et al. 2009; Lauxmann et al. 2014; Lombardo et al. 2011). Genes involved in serine, threhalose and galactinol metabolism such as PGDH, TPS, and GolS2 did not change in R fruit during SLR (Table S9). We confirmed the reported upregulation (Borsani et al. 2009; Lauxmann et al. 2014; Lombardo et al. 2011) of transcripts for PK, asparaginase B, and PDC and constant expression of transcripts for UGP2 and NAD-ME. Thus, the transcriptomic analysis indicated that our samples behaved as expected during postharvest ripening.

In general, CS represses and/or stops ripening-induced changes of most genes involved in carbon metabolism, amino acid metabolism and organic acid metabolism. While most resume ripening properly in LS fruits (clusters eRG1 and NeRG6), there is a gradual decrease in recovery with CS time during SLR in S fruits in both the number and expression of genes in these functional categories (Fig. 2; Table S9).

According to our gene expression data, during WLT development in SLR there is a gradual depletion of genes related to amylose and sucrose metabolism affecting both hydrolysis and the formation of sucrose derivatives trehalose and galactinol (Table S9). This fits the observation that in detached fruits, sucrose concentrations decline gradually during CS, especially at CI-inducing temperatures (Wang et al. 2013). High glycolytic flux can be supported only if sufficient carbohydrate is available. Correspondingly, there was progressive depletion of gene expression related to the glycolytic and penthose phosphate pathways in a WLT phenotype-dependent manner (Table S9), followed by progressive reduction in gene expression encoding functions related to several metabolic pathways branching out from central carbohydrate metabolism (Fig. 2), such as the serine-, aspartate- and glutamate-derived amino acids, branched amino acids and cofactor and vitamin metabolism (Table S9). In addition, fermentative pathways are affected in a WLT-sensitive manner consistent with previous studies (Zhang et al. 2011). S fruits had reduced PDC, ADH, ALDH, ACN1/AAE7 (clusters eRG1, eRG3, NeRG2, and NeRG6) during SLR (Table S9; Fig. 3).

Altered carbohydrate and pyruvate metabolism during SLR after CS has been associated previously with WLT phenotype (Pavez et al. 2013). Sucrose synthase, NADP-ME and phosphoenolpyruvate carboxykinase (PEPCK) were downregulated in WLT tissue compared to untreated fruit (Pavez et al. 2013). Further, enhanced expression of genes in fermentative pathways and fermentation products were found in peach fruit at 0 °C, considered a CI-delaying temperature (Lurie and Crisosto 2005), than at 5 °C (Zhang et al. 2011). However, the response of S fruits during CS and SLR at 5 °C here is slightly different than that observed for fruits stored at 0 °C (Lauxmann et al. 2014). These differences can be attributed to differences in the genes, proteins and metabolites between CS temperatures (Wang et al. 2013; Zhang et al. 2010; Zhang and Tian 2009), but also to the time of CS, as those fruits were stored only for 2 or 5 days (Lauxmann et al. 2014). There are analogies between the peach response observed at 0 °C (Lauxmann et al. 2014) and the most tolerant peaches during CS and SLR. The results presented here confirm these results and expand the list of genes in these functional categories (Table S9) and the timing of these alterations, providing a more comprehensive view of metabolism during WLT development.

Long term CS: the injury and the termination of the cold response

A striking result of our microarray analysis was the transcriptomic response observed in S fruits undergoing the longest CS. PCA analysis indicated that extended CS affected S fruits more dramatically: the transcriptome from these fruits looks as if they remained at the M stage in some respects, with many irreversible cold-induced alterations in gene expression that were not reversed during SLR (Fig. 1c, d). After 3 weeks CS, both S and LS fruits were severely woolly, with much reduced differences in WI (Fig. 1b), but there was still a surprising increase in the number of genes showing differences in expression between S and LS fruits (Fig. 2c). Cluster analysis suggested that this is due to a special response observed at 3 weeks CS in genes in clusters NeRG1, NeRG7, eRG2 and NeRG8 (Fig. 3).

Clusters eRG2 and NeRG8 were enriched overall in RNA translation and protein assembly genes, but also in energy production, chromatin status regulation and RNA post-translation regulation genes (Figs. 1, 3; Table S2). These genes were of special interest in defining altered ripening during SLR, given their direct correlation to WLT. Genes in these clusters do not change in LS fruits during CS but correlate with the WLT phenotype during SLR in both S and LS fruit, suggesting that they are a consequence of the WLT phenotype rather than a cause.

Rapid recovery of protein synthesis capacity upon rewarming during SLR after CS is expected as cellular activity resumes. During recovery from stress, translation is efficiently restored in tolerant plants, but less efficiently in sensitive ones. However, after the global repression of protein synthesis transcripts induced by cold in a sensitive-dependent manner (Fig. 3; Pons et al. 2014, 2015; Vizoso et al. 2009), expression of these genes during SLR returns to normal (for LS fruits) or even greater expression in S fruits, especially after 3 weeks CS (Figs. 2, 3; Vizoso et al. 2009). Enhanced expression of genes in these two clusters probably reflects that S fruits were over-woolly and undergoing cell death and senescence responses. During ripening after short or medium CS, fruits that did not develop WLT (tolerant or stored at 0 °C) had higher respiration rates, but increased respiration was associated with longer times and over-woolly fruits stored at 2–4 °C (Campos-Vargas et al. 2006; Fernández-Trujillo et al. 1998). Further, autumn senescence may be preceded by a peak in expression of genes encoding ribosomal components (Andersson et al. 2004) and senescence and ripening are strong regulated by epigenetic and post-transcriptional regulation (Ay et al. 2014; Humbeck 2013; Zhong et al. 2013). Apart from implying that such fruits are entering active senescence, why does increased transcription of the protein synthesis machinery correlate with visual WLT symptoms? One plausible explanation is that accumulation of ribosomal protein transcripts in woolly cells may result from ribosome compositional change and translational prioritization for specific transcripts involved in visual symptoms of WLT phenotype: the cell wall. Translational prioritization of transcripts for cell-wall-related genes associated with root cells having high cell wall activity has been described (Rajasundaram et al. 2014). Abscission (a specific type of cell separation process like WLT) requires rapid and abundant protein synthesis (Abeles 1968) and ribosomal genes are abundant in cells where early events of xylogenesis occurs (Foucart et al. 2006). Further evidence supporting involvement of the translation machinery in WLT injury comes from overexpression experiments of translation initiation factor 5A (EIF-5A) (Hopkins et al. 2008; Liu et al. 2008), which is consistent with this gene showing enhanced and altered expression in S fruits (cluser eRG2; Table S2) and in the mealy Cnr tomato mutant (Eriksson et al. 2004). EIF-5A overexpression results in increased primary and secondary xylem formation (Liu et al. 2008), precocious cell death and development of injury symptoms (Hopkins et al. 2008). It is unclear whether this gene expression enhancement is due to the specific temperature change, the rate of recovery, an indirect response to acceleration of ripening or some combination of these. Nevertheless, these genes presumably contribute to formation of the visible WLT phenotype.

The genes in cluster NeRG1 were highly upregulated during the first 2 weeks CS, but downregulated during the third. This correlated with CSR3-S expression (Fig. 3) and inversely correlated with expression at 3 weeks of genes in clusters NeRG7, eRG2 and NeRG8. Among genes in cluster eRG1, there were several regulators of the transition from vegetative to reproductive stage in flowers such as VOZ1 (Celesnik et al. 2013), NAC052 (Ning et al. 2015) and BRM (Farrona et al. 2011) (Tables S2, S7); the latter plays an important role in mediating the temporary arrest in growth that is induced in Arabidopsis upon perception of stress (Mlynárová et al. 2007). Furthermore, VOZ1, which is repressed during cold acclimation in Arabidopsis, is a negative regulator of CBF genes (Nakai et al. 2013). We associated genes in cluster NeRG7 (Fig. 3) with cold tolerance and effective cold acclimation (as CBF1 is among NeRG7 genes; Table S7). However, after 3 weeks CS, these genes were upregulated. This suggests that these fruits eventually acquired some cold tolerance. However, because they eventually developed injury and cold response is reverted during CS in only some gene clusters, it is more plausible that these fruits entered a de-acclimation stage during CS. Chilling hours accumulation during vernalization may trigger de-acclimation, even under low temperatures (Rapacz 2002). Acceleration of growth interferes with cold acclimation not only through energy competition and water balance, but also through reprogramming of development, which may disturb signal transduction pathways and gene expression involved in cold acclimation (Rapacz 2002). In this sense, it is likely that during long storage, some parts of the cold response are overcome even while cold conditions remain, allowing some aspects of ripening that were delayed by cold to progress in an unsynchronized fashion that produces a fruit that is only partially ripe. This could have strong consequences when fruit is transferred to room temperature to induce ripening.

Medium-throughput qRT-PCR analysis reveals genes that can be used as WLT expression markers

Previous gene profiling experiments provided a picture of the overall transcriptomic differences between woolly and juicy fruits (Falara et al. 2011; Gonzalez-Aguero et al. 2008; Pavez et al. 2013; Vizoso et al. 2009). However, those results did not correlate quantitatively with fruits over a range of WLT phenotypes since the experiments were based in a single genotype (Gonzalez-Aguero et al. 2008; Pavez et al. 2013; Vizoso et al. 2009) or on two genotypes with different cold response that also differed in other phenotypic traits (Falara et al. 2011). Here, in combination with BGSA, we used a new approach to identify WLT associated genes, characterized by the combination of genotypic/phenotypic variation and corresponding changes in the transcriptomes. The inclusion of a set of siblings form the Pop-DG population with different sensitivity to develop WLT permitted us to validate and extend the expression patterns obtained in the pooled samples with contrasting sensitivity to individual lines. Most important, the method has the advantage, given the continuous range of WLT sensitivity, to identify potential expression markers for WLT development at both symptomatic and pre-symtomatic stages. For the later, by assigning to CS samples the protracted WLT of the corresponding CSR samples, we assumed that transcription patterns in the cold have a direct effect on WLT and, therefore, that the protracted WLT incidence can be used to test their association with the injury. We found that in general, genes highly expressed in S pools during CS or CS + SLR showed positive WLT-gene correlations in individual lines, while low-expressed genes showed negative correlations (Fig. 4). Therefore, these genes, especially those with higher correlation values, can serve as potential markers for predicting sensitive and low sensitive genotypes, a tool very valuable for breeding programs. Further experiments can validate expression in other peach phenotypes. This has already been done in part for a subset of these genes, whose value as pre-symptomatic stage WLT markers was recently validated in fruit from other cultivars with different degree of tolerance to WLT (Pons et al. 2015).

Conclusions

This study represents a comprehensive attempt to characterize the molecular bases of WLT development during SLR. Using pools of siblings from the same Pop-DG population, but with contrasting sensitivity to develop the disease, we identified an inventory of genes with different expression during SLR in sensitive and tolerant fruits that may be responsible for the WLT phenotype. We identified what molecular alterations are present and when they occur at the transcriptional level, reconciling the apparently contradictory results of previous WLT studies. Most important, we demonstrated that this holds true for individual lines of the same population. The gene functions and processes differentially regulated in S and LS fruit and their relation with WLT development are presented (Figure S5).