Transcriptional dynamics of the developing sweet cherry (Prunus avium L.) fruit: sequencing, annotation and expression profiling of exocarp-associated genes

Abstract

The exocarp, or skin, of fleshy fruit is a specialized tissue that protects the fruit, attracts seed dispersing fruit eaters, and has large economical relevance for fruit quality. Development of the exocarp involves regulated activities of many genes. This research analyzed global gene expression in the exocarp of developing sweet cherry (Prunus avium L., 'Regina'), a fruit crop species with little public genomic resources. A catalog of transcript models (contigs) representing expressed genes was constructed from de novo assembled short complementary DNA (cDNA) sequences generated from developing fruit between flowering and maturity at 14 time points. Expression levels in each sample were estimated for 34 695 contigs from numbers of reads mapping to each contig. Contigs were annotated functionally based on BLAST, gene ontology and InterProScan analyses. Coregulated genes were detected using partitional clustering of expression patterns. The results are discussed with emphasis on genes putatively involved in cuticle deposition, cell wall metabolism and sugar transport. The high temporal resolution of the expression patterns presented here reveals finely tuned developmental specialization of individual members of gene families. Moreover, the de novo assembled sweet cherry fruit transcriptome with 7760 full-length protein coding sequences and over 20 000 other, annotated cDNA sequences together with their developmental expression patterns is expected to accelerate molecular research on this important tree fruit crop.

Figure 1

Growth and development of the sweet cherry ‘Regina’ fruit analyzed in this study. (a) Fruit mass and surface area from flowering to maturity. Stage I, cell division and expansion; Stage II (gray shading), seed development and pit hardening; Stage III, cell expansion. Color change from green to red occurred between 59 and 66 DAFB (arrow). (b) Mass of CM per fruit and calculated rate of CM deposition. Data points in a and b show the average of 30 measurements; error bars represent s.e. (not visible if smaller than symbols). Time is given in DAFB. (c) Representative photos of the analyzed fruit and sample codes identifying the RNA-seq samples. The numbers in images indicate the developmental age of the fruit in DAFB. Sample codes contain information on fruit age in DAFB (3–94), tissue type (G, whole ovaries after removal of other floral organs; E, exocarp-enriched tissue; M, mesocarp only) and replicate number (1 or 2) if applicable. Photos not to scale.

Figure 2

Summary of the RNA-seq experiment, pre-processing of raw reads and de novo assembly of the sequence data. Details are given in section on ‘Material and methods’, Supplementary Method S1 and Supplementary Table S2.

Figure 3

Length distribution of assembled sweet cherry ‘Regina’ contigs in Group 1 (‘high abundance’, 34695 contigs, length 200–12 485 bp) and Group 2 (‘low abundance’, 32 712 contigs, 107–1482 bp). Length distribution of predicted transcripts in the P. persica genome (v.1.0) (28 702 sequences, 96–15 738 bp) is shown for reference. The x-values give the center of each bin; bin width is 100 bp, except for the first bin which is from 1 to 98 bp. Note logarithmic scale of the y-axis; bins with 0 sequences not shown.

Figure 4

Distribution of the mapped reads between the contigs in Groups 1, 2 and 3 in each of the 24 RNA samples from sweet cherry ‘Regina’ fruit. For sample codes, see Figure 1.

Figure 5

GO terms of 7628 Group 1F contigs with predicted full-length open reading frames. GO terms in categories biological process, molecular function and cellular component were retrieved from combined graph analyses performed in Blast2GO platform (sequence filter 100, score alpha 0.6, node score filter 100). The GO terms are sorted in descending graph score order; numbers in parentheses indicate annotation levels.

Figure 6

Selected expression patterns within the sweet cherry ‘Regina’ fruit skin transcriptome. (a) All 29 955 contigs in Group 1F (G1F) were first clustered in five clusters applying the NG algorithm on the normalized expression patterns. Clusters NG1, NG3 and NG4 are shown. (b) Each NG cluster was reclustered applying the QT clustering algorithm (cluster diameters adapted to data, minimum cluster size 20 contigs). Numbers in parentheses indicate the number of contigs in each cluster. Selected clusters are shown. Sample codes as in Figure 1. The complete set of cluster plots available as Supplementary Fig. S3.