Transcriptome profiling of differentially expressed genes in floral buds and flowers of male sterile and fertile lines in watermelon

Abstract

Background: Male sterility is an important mechanism for the production of hybrid seeds in watermelon. Although fruit development has been studied extensively in watermelon, there are no reports on gene expression in floral organs. In this study, RNA-sequencing (RNA-seq) was performed in two near-isogenic watermelon lines (genic male sterile [GMS] line, DAH3615-MS and male fertile line, DAH3615) to identify the differentially expressed genes (DEGs) related to male sterility.

Results: DEG analysis showed that 1259 genes were significantly associated with male sterility at a FDR P-value of < 0.01. Most of these genes were only expressed in the male fertile line. In addition, 11 functional clusters were identified using DAVID functional classification analysis. Of detected genes in RNA-seq analysis, 19 were successfully validated by qRT-PCR.

Conclusions: In this study, we carried out a comprehensive floral transcriptome sequence comparison of a male fertile line and its near-isogenic male sterile line in watermelon. This analysis revealed essential genes responsible for stamen development, including pollen development and pollen tube elongation, and allowed their functional classification. These results provided new information on global mechanisms related to male sterility in watermelon.

Fig. 1

Morphological differences in floral organs of a male sterile line (DAH3615-MS) and a male fertile line (DAH3615). The flower (a) and floral bud (c) of the male sterile line, DAH3615-MS, are smaller than those (b, d) of the male fertile line, DAH3615. The male sterile flower (a) shows distinctive defective stamen and absence of pollen grains. The images were taken at 63 d after sowing

Fig. 2

Identification of differentially expressed genes (DEGs) in a male sterile line (DAH3615-MS) and a male fertile line (DAH3615). (a) Venn-diagram of significantly different DEGs detected in the male fertile line and the male sterile line using two types of statistical tests such as Fisher’s exact test and 2-way ANODEV [False discovery rate (FDR) adjusted P-value < 0.01]. In the Fisher’s exact test, two tissues’ samples, bud and floral bud, were respectively employed. (b) Hierarchical clustering analysis tree with k = 2. Blue and green lines represent DAH3615-MS and DAH3615, respectively. (c) Heatmap with hierarchical clustering using logged TMM-normalized gene expressions of the 1259 DEGs (FDR adjusted P-value < 0.01) from the analysis of deviance (ANODEV). Regions 1 and 2 show highly significant differences between DAH3615-MS and DAH3615. (d) Line plots show the gene expression of the top 10 genes from ANODEV. Red and blue lines represent floral bud and flower, respectively. The y-axis represents logged TMM-normalized gene expressions, which were calculated in the edgeR

Fig. 3

Characterization of significantly detected differentially expressed genes (DEGs) using chromosomal location and biasness. Fertility biased: DAH3615-biased expression; Sterility biased: DAH3615-MS-biased expression. (a) Pie charts show the biasness of significantly detected DEGs. The upper two proportional pie charts were created using the whole-annotated genes and the lower two pie charts were created using 1,259 DEGs from the analysis of deviance (ANODEV). (b) Βiasness of significantly detected genes. (c) Manhattan plot represents the chromosomal location of significantly detected genes. The y-axis represents log ₁₀ false discovery rate (FDR) adjusted P-value. The red dotted line is a cut-off line (FDR adjusted P-value < 0.01)

Fig. 4

Technical validation of 19 randomly selected differentially expressed genes (DEGs) detected by RNA-sequencing (RNA-seq) using qRT-PCR. (a) Relative heatmaps of RNA-seq and qRT-PCR. Gene expression from the two platforms was normalized by the quantile normalization method. (b) Box-plots of 19 DEGs detected by qRT-PCR. The y-axis represents the gene expression level, which is the −2^∆ΔCt value of qRT-PCR compared to the control gene