Introduction
Ricinus communis L. (castor bean) is a drought-resistant species which is well adapted to hot climates. It is an important oilseed crop because of its unique oil composition, up to 94% of which comprises the fatty acid ricinoleic acid (12-hydroxy-cis-9-octadecenoic acid) (Gong et al., Reference Gong, Li, Ma, Indu Rupassara and Bohnert2005). This fatty acid confers unique properties to the oil and the biodiesel produced from it (Conceicao et al., Reference Conceicao, Candeia, Silva, Bezerra, Fernandes and Souza2007; Salimon et al., Reference Salimon, Noor, Nazrizawati, Firdaus and Noraishah2010; Anjani, Reference Anjani2012). R. communis oil is used in the chemical industry worldwide, but its production is limited by insufficient supply of castor bean seeds (Severino et al., Reference Severino, Auld, Baldanzi, Candido, Chen, Crosby, Tan, He, Lakshmamma, Lavanya, Machado, Mielke, Milani, Miller, Morris, Morse, Navas, Soares, Sofiatti, Wang, Zanotto and Zieler2012). India, Brazil and China are the major oil crop producers in the world and, recently, biodiesel production has received wide attention in Brazil, where the Program for Production and Use of Biodiesel (PNPB) has identified castor bean as the ideal species for oil production which, in addition, could help to promote social development in the semi-arid regions of Brazil (Cesar and Batalha, Reference Cesar and Batalha2010). Its agronomic importance stimulated molecular research in R. communis related to seed maturation (Cagliari et al., Reference Cagliari, Margis-Pinheiro, Loss, Mastroberti, de Araujo Mariath and Margis2010; Sánchez-García et al., Reference Sánchez-García, Moreno-Pérez, Muro-Pastor, Salas, Garcés and Martínez-Force2010; Loss-Morais et al., Reference Loss-Morais, Turchetto-Zolet, Etges, Cagliari, Körbes, Maraschin, Margis-Pinheiro and Margis2013), plant development (O'Leary et al., Reference O'Leary, Fedosejevs, Hill, Bettridge, Park, Rao, Leach and Plaxton2011), and responses to biotic and abiotic stresses (Wei et al., Reference Wei, Dai, Wang, Chuan, Gou and Chen2010; Maciel et al., Reference Maciel, Salles, Retamal, Gomes and Machado2011). Characterization and expression profiles of several genes from R. communis have been reported (Cagliari et al., Reference Cagliari, Margis-Pinheiro, Loss, Mastroberti, de Araujo Mariath and Margis2010; Chileh et al., Reference Chileh, Esteban-García, Alonso and García-Maroto2010; Wei et al., Reference Wei, Dai, Wang, Chuan, Gou and Chen2010; Maciel et al., Reference Maciel, Salles, Retamal, Gomes and Machado2011). In this study we focus on seed germination and seedling growth, since a better understanding of the biochemical and molecular aspects of these processes in R. communis is crucial for the breeding of high-yielding varieties adapted to various growing environments (Severino and Auld, Reference Severino and Auld2013).
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) is the method of choice for gene expression profiling studies, due to its high sensitivity, accuracy and speed. Accurate gene expression measurements involve normalization to correct for differences in sample input due to errors in measurements, pipetting and differences in enzymatic efficiencies (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002). Internal control genes or ‘reference genes’ are used to normalize for such differences. Reference genes are presumed to be expressed stably across different tissues and at various developmental stages and growth conditions. However, the stability of used reference genes needs to be properly validated for accurate and reliable normalization of gene expression data, rather than to trust their presumed expression stability. This is an important requirement for a reliable RT-qPCR experiment (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002). In the past decade several tools have become available to determine which candidate reference genes are stably expressed across a set of samples, e.g. GeNorm (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002), BestKeeper (Pfaffl et al., Reference Pfaffl, Tichopad, Prgomet and Neuvians2004) and NormFinder (Andersen et al., Reference Andersen, Jensen and Ørntoft2004). These tools have helped to identify stable reference genes in a range of plant species (Artico et al., Reference Artico, Nardeli, Brilhante, Grossi-de-Sa and Alves-Ferreira2010; Jarosová and Kundu, Reference Jarosová and Kundu2010; Dekkers et al., Reference Dekkers, Willems, Bassel, Van Bolderen-Veldkamp, Ligterink, Hilhorst and Bentsink2012; Pettengill et al., Reference Pettengill, Parmentier-Line and Coleman2012). However, so far, no validated reference genes are available for RT-qPCR studies during germination and early seedling establishment in R. communis.
In this study, the expression of 17 candidate reference genes was analysed across a set of 14 samples – using three biological replicates per sample – that include several tissues at different stages under different environmental conditions, including light/dark, water/nutrient solution and temperature treatments. Furthermore, to validate the candidate reference genes for analysis of gene expression purposes, and to illustrate their usefulness, we analysed the expression levels of three R. communis genes, oil-body lipase 2 (OBL2), chlorophyll a/b binding protein (light-harvesting complex – LHCB3) and α-amylase (AMY3), which are involved in primary energy metabolism. This is the first systematic report on the selection of reference genes for R. communis, providing useful guidelines for future accurate gene expression profiling experiments by RT-qPCR.
Materials and methods
Plant material
Ricinus communis L. seeds (genotype MPA11) used in this work were kindly supplied by Empresa Baiana de Desenvolvimento Agrícola S.A (EBDA), Salvador-Bahia, Brazil. After seed-coat removal, dry seed samples were collected and additional seeds were allowed to imbibe and germinate, using paper rolls as substrate at 25°C in the dark. In the first part of the experimental design, germinated seeds were allowed to grow on paper rolls at 25°C, in the dark, for 9 d. After collection, dry seeds and the roots, endosperm and cotyledons of 3-, 6- and 9-day-old germinated seeds were immediately frozen in liquid nitrogen, freeze-dried, ground and stored at − 80°C prior to analysis. Three biological replicates of 30 seeds each were used. In the second part of the experimental design, germinated seeds were transferred from paper rolls to moist vermiculite and were allowed to grow at 20°C in continuous light for 10 d. Half of the 10-day-old seedlings were then transferred to an incubator at 35°C with continuous light, while the remaining seedlings were kept under the original conditions. After 4 d, the roots and green cotyledons (three biological replicates of 15–18 seedlings each) were collected, immediately frozen in liquid nitrogen, freeze-dried, ground and stored at − 80°C prior to analysis. Details of the different samples are shown in Table 1.
Table 1 Description of tissue samples used for reference gene validation
aThree biological replicates of 30 seeds. bThree biological replicates of 15–18 seeds.
RNA extraction and cDNA synthesis
Total RNA was extracted from 20 mg of dry material using the RNeasy Plant Mini Kit (Qiagen, Hilden, North Rhine-Westphalia, Germany), including the DNase digestion of genomic DNA performed using the RNase-Free DNase Set (Qiagen), according to the manufacturer's instructions. RNA quantification and quality control were performed spectrophotometrically using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, Delaware, USA). Samples showing A260/A280 and A260/A230 ratios of 1.8–2.2 were used for subsequent analysis. RNA quality was further inspected by the integrity of ribosomal RNA bands on a 1% agarose gel. Sharp and intense 18S and 28S ribosomal RNA bands, without visible degradation, confirmed the suitability of the isolation method. The first-strand cDNA was synthesized from 1 μg of total RNA using the iScriptTM cDNA synthesis kit (Bio-Rad, Hercules, California, USA) according to the manufacturer's instructions. Reverse transcription was performed at 37°C for 40 min followed by 85°C for 5 min. The cDNA was diluted 20 times for the use in RT-qPCR reactions. All cDNAs were stored at − 20°C prior to the analysis.
Reference gene selection, primer design and measurement of amplification efficiency
Potential reference genes were selected based on available data in the literature on commonly used reference genes in RT-qPCR studies in Arabidopsis, tomato, Jatropha curcas L. and R. communis. We started our selection for candidate reference genes by a literature search for reference genes commonly used in R. communis, which includes: 18S, protein Pob (POB), NADH-ubiquinone oxidoreductase (NADH_OXI) and elongation factor 1-beta (EF1B) (Eastmond, Reference Eastmond2004; Chen et al., Reference Chen, Turner, He, Nguyen, McKeon and Laudencia-Chingcuanco2007; Cagliari et al., Reference Cagliari, Margis-Pinheiro, Loss, Mastroberti, de Araujo Mariath and Margis2010; H.L. Li et al., Reference Li, Zhang, Guo, Li and Peng2012; Arroyo-Caro et al., Reference Arroyo-Caro, Chileh, Kazachkov, Zou, Alonso and García-Maroto2013; Loss-Morais et al., Reference Loss-Morais, Turchetto-Zolet, Etges, Cagliari, Körbes, Maraschin, Margis-Pinheiro and Margis2013). In addition to these candidates, 40S ribosomal protein (40S), 60S acidic ribosomal protein (60S), β-tubulin (β-TUB), translation elongation factor G (EFG) and structural constituent of ribosome (SCR) genes were also selected because of their use as reference genes for other species (Fig. 1 and see supplementary Table S1, available online) (Le et al., Reference Le, Aldrich, Valliyodan, Watanabe, van Ha, Nishiyama, Guttikonda, Quach, Gutierrez-Gonzalez, Tran and Nguyen2012; X.S. Li et al., Reference Li, Yang, Zhang, Zhang and Wood2012; Rapacz et al., Reference Rapacz, Stepień and Skorupa2012). However, only few of these studies have validated the expression stability of these putative reference genes (Cagliari et al., Reference Cagliari, Margis-Pinheiro, Loss, Mastroberti, de Araujo Mariath and Margis2010). Since the number of candidate reference genes was very low we extended our search and looked for putative reference genes that were described as stably expressed in seeds of the model plant Arabidopsis thaliana, tomato and also for putative reference genes in the closely related species J. curcas. For Arabidopsis, tomato and J. curcas candidate genes, the sequences of possible orthologues were identified through a TBLASTX against the R. communis translated nucleotide database (NCBI, http://www.ncbi.nlm.nih.gov/). Only sequences that showed high similarity (E-value < 1e–70) were considered putative orthologues in R. communis and were selected for primer design. This selection consisted of genes for ubiquitin-conjugating enzyme E2 (UBI_E2), serine/threonine protein phosphatase 2a regulatory subunit A (PP2AA1), Type 2A phosphatase activator TIP41 (TIP41) and GTP-binding protein Sar1 (SAR1) used for Arabidopsis and tomato seeds (Dekkers et al., Reference Dekkers, Willems, Bassel, Van Bolderen-Veldkamp, Ligterink, Hilhorst and Bentsink2012) and actin (ACT), ubiquitin (UBI), α-tubulin (α-TUB) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) used for developing J. curcas seeds (Fig. 1, supplementary Table S1) (Gu et al., Reference Gu, Yi, Tian, Sangha, Hong and Yin2012; Fan et al., Reference Fan, Li, Lu, Li and Yin2013; Rocha et al., Reference Rocha, Soares, Costa, Costa, Soares, Nogueira, Domont and Campos2013). After an initial quality assessment, primer pairs for 17 putative reference genes remained (supplementary Table S1). The gene identifier (GI) number, gene name and description, amplicon length, primer efficiency and primer sequences (reverse and forward) are listed in supplementary Table S1. Primers for the 17 potential reference genes were designed using CLCbio software (CLCbio, Aarhus, Denmark) with melting temperatures (T m) of 58–62°C, primer lengths of 18–22 bp and amplicon lengths of 80–200 bp. Information about all primers used in this study is summarized in supplementary Table S1. The PCR amplification efficiency was evaluated based on a standard curve generated by twofold serial dilutions of a pooled cDNA sample. The specificity of the primers was verified by loading the amplicons on a 2.5% agarose gel and melting curve analysis.
Figure 1 Flowchart showing the steps followed in this study from the selection of putative reference genes to gene expression stability analysis.
RT–qPCR conditions
For RT-qPCR experiments, reactions were prepared in a total volume of 10 μl, containing 2.5 μl of cDNA, 0.5 μl of primer-mix (from a 10 μM working solution), 5 μl of iQ SYBR Green Supermix (Bio-Rad) and 2 μl of water. The RT-qPCRs were run on a CFX (Bio-Rad). The following qPCR programme was used for all PCR reactions: 95°C for 3 min, followed by 40 cycles of 95°C for 15 s and 60°C for 30 s. Melting curves were recorded after cycle 40 by heating from 65 to 95°C, increasing the temperature stepwise every 5 s by 0.5°C.
Evaluation of reference gene expression stability: GeNorm and NormFinder
The expression levels of 17 candidate reference genes were determined by the number of cycles (Cq) required to reach fluorescence above a specific threshold level. GeNorm (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002; Hellemans et al., Reference Hellemans, Mortier, De Paepe, Speleman and Vandesompele2007) and NormFinder (Andersen et al., Reference Andersen, Jensen and Ørntoft2004) were used for expression stability evaluation of the candidate reference genes across all tested samples.
GeNorm ranks the candidate reference genes according to the expression stability index, M. The most stable genes show the lowest M values. The pairwise variation (V n /V n+1) was calculated to determine the optimal number of reference genes (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002). Alternatively, Normfinder (Andersen et al., Reference Andersen, Jensen and Ørntoft2004) was used to identify the optimal reference genes. Normfinder ranks the putative reference genes according to their stability value (Andersen et al., Reference Andersen, Jensen and Ørntoft2004).
Reference gene validation: normalization of three target genes
Three genes involved in primary energy metabolism were used to demonstrate the usefulness of the candidate reference genes and to validate them. Oil-body lipase 2 (OBL2), chlorophyll a/b binding protein (light-harvesting complex–LHCB3) and α-amylase (AMY3) expression levels were normalized using the most stable candidate reference genes across all tested samples. qBase (Biogazelle, Ghent, Belgium) was used to normalize the expression levels using these reference genes and the most stable candidate reference genes across three sample subsets: root, cotyledons and endosperm samples. qBase allows gene expression levels to be normalized by using up to five reference genes (Hellemans et al., Reference Hellemans, Mortier, De Paepe, Speleman and Vandesompele2007). Statistical analyses were conducted using one-way ANOVA and the Tukey post-hoc test with IBM SPSS Statistics® (Windows version 19, SPSS Inc., Chicago, Illinois, USA) and Microsoft Excel 2010®. Values of P< 0.05 were considered to be statistically significant.
Results and discussion
Primer specificity, efficiency and expression profile of the 17 candidate reference genes
Gene expression measurements by RT-qPCR rely on normalization by stably expressed reference genes. A set of validated reference genes is of utmost importance to ensure reliable and accurate data normalization. However, such a set is not available for R. communis. Therefore we selected a range of putative reference genes and tested their expression stability over a diverse set of samples encompassing seed germination and seedling establishment in R. communis. A successful approach to identify stably expressed genes, and thus potential reference genes, in a given sample panel is to mine microarray data (Czechowski et al., Reference Czechowski, Stitt, Altmann, Udvardi and Scheible2005; Popovici et al., Reference Popovici, Goldstein, Antonov, Jaggi, Delorenzi and Wirapati2009; Dekkers et al., Reference Dekkers, Willems, Bassel, Van Bolderen-Veldkamp, Ligterink, Hilhorst and Bentsink2012; De Oliveira et al., Reference De Oliveira, Breton, Bastolla, Camargo, Margis, Frazzon and Pasquali2012). To the best of our knowledge, microarray data for seed germination and seedling establishment in R. communis are not yet available. Therefore, we selected 17 putative reference genes from a literature search. These include four candidate reference genes commonly used in R. communis and 13 candidate reference genes used in other species (Fig. 1).
To evaluate the primer amplification specificity of these putative reference genes, electrophoresis was performed with the product of an RT-qPCR on a pooled cDNA template with the primers for each candidate reference gene. Additionally, analysis of the melting curves confirmed that all primers generated only a single amplicon (see supplementary Fig. S1, available online) and that neither primer–dimers nor unexpected products were found. The absence of genomic DNA was confirmed by comparing RT-qPCR with the cDNA template with that of RNA samples which were not reverse transcribed (minus RT control). Most of the minus RT controls did not show any amplification. In some samples, low signals were detected in the minus RT control, but these were at least 7 Cq (quantification cycles) higher than with the cDNA samples. This is clearly above the limit of 5 Cq as proposed by Nolan et al. (Reference Nolan, Hands and Bustin2006). Amplification efficiency ranged from 88.77% for GAPDH to 111.6% for β-TUB, and the coefficient of determination (R 2) varied from 0.9709 to 0.9997 (see supplementary Table S1). This indicates that the reaction conditions are optimal and that the results obtained are highly repeatable.
The expression levels of the 17 candidate reference genes across all samples are presented as Cq values (Fig. 2 and supplementary Table S1) which represent the relative abundance of a particular transcript. The mean values of most candidate reference genes were between a Cq of 21 (EF1B) and 29 (SCR). This difference represents a 150-fold higher abundance of EF1B over SCR. The mean Cq value of 18S was 8.41. SAR1 showed the lowest variation in expression across all tested samples (Cqmax – Cqmin = 6.22), whereas β-TUB showed the highest variation (Cqmax – Cqmin = 13.34) (see supplementary Table S1).
Figure 2 (colour online) Quantification cycle values (Cq) of all 17 candidate reference and 3 target genes across all samples. The horizontal line across the boxes depicts the median Cq value.
Gene expression stability analysis
First, gene expression stability was analysed across all samples, aiming at selecting reference genes that could be used to normalize between tissues. Then the samples were subdivided into three tissue-specific subsets, i.e. root, cotyledon and endosperm samples (Table 1). These four sample panels were analysed using two methods to assess the gene expression stability: GeNorm and NormFinder. GeNorm calculates the average expression stability (M) based on the average pairwise variation between all tested genes. A lower value of M indicates more stable gene expression and vice versa (Vandesompele et al., Reference Vandesompele, De Preter, Pattyn, Poppe, Van Roy, De Paepe and Speleman2002). The ranking order according to the M value is shown in Fig. 3 and Table 2. For more heterogeneous sets of samples (all tested samples), M values lower than 1 were used (Hellemans et al., Reference Hellemans, Mortier, De Paepe, Speleman and Vandesompele2007). M values lower than 0.5 were used to identify stably expressed reference genes in relatively homogeneous sets of samples (tissue-specific).
Figure 3 Expression stability values (M) of 17 candidate reference genes. (a) All R. communis samples; (b) root samples in different developing stages and temperatures; (c) cotyledon samples in different developing stages and temperatures; (d) endosperm samples in different developing stages. A lower M value indicates more stable expression.
Table 2 Expression stability analysis of 17 candidate reference genes by GeNorm
POB, PP2AA1, ACT and SAR1 were defined as the four most stable reference genes (0.575 < M< 0.625) for the entire set of samples (inter-tissue) (Fig. 3a). GeNorm also determines the optimal number of reference genes (GeNorm V < 0.15) required for calculating an accurate normalization factor, which is based on pairwise variation (V n /V n+1). In order to accurately normalize gene expression data for the entire set of samples, the use of the three most stable genes is suggested (V 3/V 4< 0.15) (Fig. 4a, Table 2). For accurate normalization of root and endosperm subset samples two reference genes should be used (Fig. 4b and 4d; Table 2), while for cotyledons the optimal number is three (Fig. 4c; Table 2). NADH_OXI and α-TUB were the most stable genes across all root samples, UBI_E2, TIP41 and PP2AA1 for cotyledons and TIP41 and SAR1 for endosperm samples (Fig. 3; Table 2). As expected, the most stable genes showed the lowest variation in expression across the entire sample panel (Cqmax – Cqmin ≈ 6–7) while the least stable genes showed the highest variation (Cqmax – Cqmin ≈ 9–13) (Fig. 2, Table 2, supplementary Table S1).
Figure 4 Determination of optimal number of reference genes based on pairwise variation (V) analysis of normalization factors of the candidate reference genes. (a) All R. communis samples; (b) root samples in different developing stages and temperatures; (c) cotyledon samples in different developing stages and temperatures; (d) endosperm samples in different developing stages. The V n /V n+1 values were calculated to determine the optimal number of reference genes for use in RT-qPCR data normalization.
NormFinder is a freely available Visual Basic application for Microsoft Excel, which automatically calculates the stability value for all candidate reference genes tested on a sample set containing any number of samples organized in any given number of groups (Andersen et al., Reference Andersen, Jensen and Ørntoft2004). NormFinder ranks all the candidate reference genes according to their stability value in a robust manner which shows low sensitivity toward co-regulation of the candidate reference genes (Andersen et al., Reference Andersen, Jensen and Ørntoft2004). The ranking of reference genes based on the NormFinder algorithm for the entire set of samples indicated ACT, PP2AA1 and POB as the three most stable genes (Table 3), which is very consistent with the output obtained from GeNorm (Fig. 3a, Table 2). Furthermore, NormFinder indicated ACT, PP2AA1 and POB as the three most stable genes across all root samples, UBI_E2, PP2AA1 and POB for cotyledons and ACT, EFG and 18S for endosperm samples (Table 3).
Table 3 Expression stability analysis of 17 candidate reference genes by NormFinder
Additionally, genes encoding the ribosomal units 18S, 40S and 60S were suggested by NormFinder as suitable reference genes for normalization of RT-qPCR data in R. communis (Table 3). In order to rule out a potential sensitivity of this algorithm to co-regulation of these ribosomal genes, we re-calculated the stability value of all candidate reference genes without the gene expression data of 40S. However, the removal of 40S hardly affected the candidate reference gene output of NormFinder (Table 4).
Table 4 Expression stability analysis of 17 candidate reference genes by NormFinder (excluding 40S)
Expression profiles of three functional genes
To evaluate our sets of reference genes we measured and normalized the expression of three target genes involved in energy metabolism. OBL2 is an oil-body-associated lipase from R. communis endosperm that is induced upon seed germination and is responsible for the hydrolysis of a range of triacylglycerols (Eastmond, Reference Eastmond2004). OBL-like proteins are present in many species and they probably play an important role in regulating lipolysis during germination (Eastmond, Reference Eastmond2004). LHCB3 (light-harvesting chlorophyll a/b) is responsible for the conversion of free energy from absorbed light into ATP, reduced ferredoxin (Fd) and NADPH that can be used to drive metabolism in the chloroplast and cytosol (Foyer et al., Reference Foyer, Neukermans, Queval, Noctor and Harbinson2012). Down-regulation or disruption of any member of the LHC family reduces responsiveness of stomatal movement to abscisic acid (ABA) and therefore results in a decrease in tolerance to drought stress in Arabidopsis thaliana (Xu et al., Reference Xu, Liu, Yan, Liu, Jiang, Shen, Wang and Zhang2012). α-Amylase is involved in the degradation of starch and it has been demonstrated that in Arabidopsis leaves it is induced by biotic and abiotic stress and secreted (Doyle et al., Reference Doyle, Lane, Sides, Mudgett and Monroe2007). The expression levels of the three target genes were measured and normalized by: (1) using ACT, POB and PP2AA1 as reference genes; and (b) using tissue-specific reference genes for which the expression stability value, M, was below the recommended value of 0.5 for each subset. The results show that OBL2, LHCB3 and AMY3 genes were expressed in all samples (Fig. 5) and the mean Cq values varied between 27.23 (OBL2) and 29.96 (LHCB3) (Fig. 2, supplementary Table S1). LHCB3 showed the highest variation in expression across all tested samples (Cqmax – Cqmin= 13.3), whereas AMY3 showed the lowest variation (Cqmax – Cqmin = 6.81). The observed wide range of Cq values for LHCB3 is due to the fact that photosynthetic tissues, such as 14-day-old greening cotyledons, showed relatively higher expression as compared to the non-photosynthetic endosperm tissues (data not shown). Expression levels of OBL2 were more than a 1000-fold higher in 9-day-old endosperm as compared to the expression in dry seeds (Fig. 5g). This is consistent with the function of the gene, which regulates lipolysis during germination (Eastmond, Reference Eastmond2004). To illustrate the effect of reference gene set selection on the normalization and the accuracy of the RT-qPCR data, we compared the normalized expression data of the target genes obtained by using ACT, POB and PP2AA1 as reference genes with the values obtained by using tissue-specific reference genes. For most of the genes, a high correlation was found between the normalized data obtained by using ACT, POB and PP2AA1 as reference genes and the data normalized by the respective tissue-specific reference genes (Fig. 5). The correlation coefficient (R 2) varied from 0.4348 (AMY3, roots) to 0.9986 (OBL2, endosperm) (Fig. 5a–i). We only found a weak correlation for OBL2 (Fig. 5a) and AMY3 (Fig. 5c) in the root sample panel. However, it is important to highlight that the combination of genes suggested for the entire dataset by GeNorm (as well as Normfinder) is the same as that suggested by NormFinder for the root subset and thus similar! This difference is likely caused by the different algorithms used in both methods.
Figure 5 Relative expression of OBL2 (a, d and g), LHCB3 (b, e and h) and AMY3 (c, f and i) in root (a–c), cotyledon (d–f) and endosperm (g–i) samples. Black bars represent the expression levels normalized by using ACT, POB and PP2AA1 reference genes as the optimal combination selected by GeNorm to normalize an inter-tissue (heterogeneous) sample panel. White bars represent the expression levels normalized by using reference genes found as the optimal combination selected by GeNorm for the sample-specific sets: NADH_OXI and α-TUB in root samples (a–c), UBI_E2, TIP41 and PP2AA1 in cotyledon samples (d–f) and TIP41 and SAR1 in endosperm samples (g–i). Averages and standard errors of three biological and two technical replicates are shown. Expression levels were normalized towards the first column (R3, C3 and Dry) for roots, cotyledons and endosperm sample panels, respectively. Results are expressed as mean ± SD. Letters above the bars indicate significant differences between different samples by Tukey's HSD (P< 0.05).
The fact that most stable genes suggested by GeNorm are very consistent with those indicated by NormFinder for the entire sample panel and for the cotyledon subset, highlights the strength of the selection of the reference genes in our study. However, little overlap was observed for endosperm samples (Tables 2 and 3). Therefore, we also normalized the expression of the three studied target genes by using the most stable genes suggested by NormFinder (ACT and EFG) for endosperm samples (Fig. 6). The expression pattern of OBL2 and LHCB3 were extremely consistent between all three sets of reference genes used to normalize the expression levels (Fig. 6a, b), however the relative expression levels of OBL2 were slightly higher for the samples normalized by using TIP41 and SAR1 reference genes (white bars) as compared to the samples normalized by using ACT, POB and PP2AA1 (black bars) and ACT and EFG [grey bars (red online)] as reference genes for the subset of endosperm samples (Fig. 6a). The expression pattern and expression levels of AMY3 for the samples normalized by using the optimal combination selected by GeNorm to normalize the complete inter-tissue sample panel (ACT, POB and PP2AA1) and the optimal combination selected by NormFinder for the subset of endosperm samples (ACT and EFG) were more similar to each other than to the values obtained by using TIP41 and SAR1 as reference genes (Fig. 6c). Taken together, these results suggest that the approach used to select reference genes for R. communis is robust and provides useful guidelines for future accurate gene expression profiling experiments by RT-qPCR.
Figure 6 (colour online) Relative expression of OBL2 (a), LHCB3 (b) and AMY3 (c) in endosperm samples. The black bars represent the expression levels normalized by using ACT, POB and PP2AA1 reference genes as the optimal combination selected by GeNorm to normalize an inter-tissue (heterogeneous) sample panel. The white bars represent the expression levels normalized by using TIP41 and SAR1 reference genes found as the optimal combination selected by GeNorm for the subset of endosperm samples. The grey bars (red online) represent the expression levels normalized by using ACT and EFG reference genes found as the optimal combination selected by NormFinder for the subset of endosperm samples. Averages and standard errors of three biological and two technical replicates are shown. Expression levels were normalized towards dry seeds. Results are expressed as mean ± SD. Letters above the bars indicate significant differences between different samples by Tukey's HSD (P< 0.05).
Conclusions
This study reports a systematic analysis aimed at determining the optimal combination of reference genes for the normalization of gene expression data for R. communis seed germination and seedling establishment, providing useful guidelines for future accurate gene expression profiling experiments by RT-qPCR. Genes described in the literature as stably expressed in J. curcas (ACT), Arabidopsis and tomato (TIP41, SAR1 and PP2AA1) were proven to be stable for R. communis seed and seedling samples (Fig. 1 and Tables 2 and 3). Therefore these genes are likely candidates for successful gene expression studies in seeds and seedlings of a broader range of species. More importantly, our data showed that some of the widely used reference genes were not suitable reference genes for our samples (Table 2), highlighting the importance of a proper validation of candidate reference genes for each study. Normalization of expression data of three target genes involved in energy metabolism highlights the reliability of the selected reference genes in our study. The approach used in this study to identify stably expressed genes, and thus potential reference genes, was successfully applied for R. communis and it provides important guidelines for finding suitable reference genes for RT-qPCR studies in seeds and seedlings of other species (especially in those cases where extensive microarray data are not available).
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0960258514000294
Authors’ contributions
P.R.R. carried out the physiological and molecular experiments, data processing, statistical analysis and draft of the manuscript. L.G.F. and R.D.C. participated in the design of the study. B.J.W.D., W.L. and H.W.M.H. participated in the design of the study, coordination and critical reading of the manuscript. All authors read and approved the manuscript.
Acknowledgements
The authors would like to thank Empresa Baiana de Desenvolvimento Agrícola S.A (EBDA) for kindly supplying the seeds used in this work.
Financial support
Financial support was provided by Conselho Nacional de Desenvolvimento Científico e Tecnológico – CNPq (grant number 200745/2011-5), Petróleo Brasileiro S.A. (grant number 0050.0062805.10.9), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – CAPES – Brazil, and the Netherlands Organization for Scientific Research – NWO.
Conflicts of interest
None.
