Validation of reference genes for gene expression analysis in olive (Olea europaea) mesocarp tissue by quantitative real-time RT-PCR

Quantitative reverse transcription PCR (qRT-PCR) is a well-established procedure to study changes in gene expression levels, due to its high sensitivity, reproducibility and large dynamic range[1–6]. In qRT-PCR experiments using a relative quantification approach, the expression level of the target genes are normalised using internal control genes known as reference genes to derive changes in gene expression levels. This normalisation strategy improves the fidelity of the quantification process by controlling any variation in the expression level of the biological samples that might have been introduced due to various factors such as RNA integrity, initial sample amount, reverse transcription efficiency etc. Some of the most common and best known housekeeping genes involved in basic cellular and metabolic processes that have been used as candidate reference genes over the last few decades in plants and animals include GAPDH, 18S or 26S RNA, EF1-alpha, ubiquitin carrier protein, actin, α-tubulin, β-tubulin and TATA-Box binding protein[3, 5, 7, 8]. It is assumed that these reference genes have constant level of expression in different tissues and under different treatments and has no inter-individual variability[9, 10]. Experimentally, it is impossible to find a single, ideal reference gene for normalisation in various samples, under different conditions as the transcription of any gene will not be absolutely resistant to fluctuations in the cell cycle or nutrition status[5, 7]. It has been shown that in many experiments the use of a single reference is not acceptable as it would likely produce erroneous conclusions in expression patterns[11, 12]. Recent reports have also shown that the most commonly used traditional reference genes may be inappropriate for normalization in qPCR experiments due to their expression variability under different experimental conditions[8, 13]. The importance of expression stability in the choice of reference genes has prompted the development of various software packages including GeNorm[4], BestKeeper[14] and NormFinder[15] to identify such genes. Therefore, for accurate analysis of RNA transcription it is crucial to choose reference genes that have been shown to be minimally regulated in a given species, in a given organs/tissues, at a given developmental period and under specific environmental conditions.

In plants a number of reference gene validation reports have been published covering both model and agriculturally important crop species including, Arabidopsis[8], rice[16], wheat[5] grapevine[3], barley[7], soybean[9] and cotton[10]. In Arabidopsis thaliana, analysis of the Affymetrix ATH1 microarray data has revealed that there are hundreds of potential reference genes that outperform traditional reference genes in terms of expression stability, where most of these genes are expressed at much lower levels than the traditional reference genes. This list of new Arabidopsis reference genes were successfully employed to search for reference genes in unrelated species such as Vitis vinifera[17] and Coffea arabica[18].

The olive tree (Olea europaea) represents one of the oldest agricultural tree crops with cultivation of olives beginning in the Mediterranean basin more than 3000 years ago[19]. Over the past few decades an increased awareness of the reported health benefits of olive oil over other fats and oils[20–24] has seen olive oil consumption increase globally[19, 25] and cultivation of olives has spread globally into new areas such as Australia, North America and South America[19]. In the past few years high throughput next generation sequencing (NGS) technologies have been used to study the transcriptome of olive drupes during different developmental stages and tissues[26–28]. These sequencing data are a valuable resource for gene discovery and characterisation in olives. However, in order to normalise the expression data of genes in olives it is important to identify stable reference genes which show consistent expression within olive samples for a specific set of chosen experimental conditions and tissue types in olives.

In olives, an analysis of cDNAs that are associated with alternate bearing in olive was conducted wherein seven commonly used reference genes [Olest34, alpha-tubulin, beta-tubulin, beta-actin, 26S rRNA, 18S rRNA and GAPDH] were chosen to identify the reference gene which is least spatially and temporally variable[28]. GAPDH was decided to be used as an appropriate reference gene for the major tissues (leaves, fruits and pedicels) of olive to normalise the copy numbers of the cDNAs tested[28]. Although the authors did provide data to justify this selection there was not an in-depth analysis of potential reference gene candidates.

More recently, two studies were conducted to identify stable reference genes in olives during fruit development and ripening. Nonis et al.[29] focused on the stability of 13 putative reference genes evaluated on 21 samples collected over different developmental stages of olive fruits and leaves subjected to wounding from the cultivar Frantoio during the 2009 season. Thirteen candidate reference genes belonging to 8 gene families were chosen based on their stability in other plants as well as the availabilty of putative ESTs for these genes in a publicly available database of olive ESTs[26]. Results showed that serine/threonine-protein phosphatase catalytic subunit-1 (PP2A1) was the most transcriptionally stable reference gene, followed by GAPDH2, while GAPDH1 showed the widest variation and was considered the least stable gene. Another study was conducted on the validation analysis of 29 reference genes at 12 different sampling points of olive fruit development from the cultivar, Istrska belica using two evaluation approaches, GeNorm and RefFinder programs[30]. Combining both the programs together, TIP41-like family protein (TIP41) and TATA binding protein (TBP) were identified as the two most stable reference genes during fruit development in this cultivar.

Comparing the results from the two papers[29, 30], it is interesting to note that few reference genes such as EF1-alpha, GAPDH, 14-3-3 protein performed differently in these two experiments during the olive fruit development with differences in their M-values and ranking order clearly indicating that the stability of the reference genes must be validated with each experimental setup. It is to be noted that though both the studies identified stable reference genes during olive fruit development and ripening, these experiments were conducted on different olive cultivars, with the former study conducted on the cultivar Frantoio while the latter was on the cultivar Istrska belica. This indicates that even for different cultivars the stability of the reference genes should be assessed.

Thus, the current study involved the use of two software packages, GeNorm (qBase Plus) and BestKeeper, to validate the expression stability of eight candidate reference genes chosen based on their stability in olive tissues as well as in other plants, during mesocarp development across three different olive cultivars Barnea, Frantoio and Picual independently and together during the 2009 season by qPCR.

This is one of the first reports conducted to assess the expression stability of candidate reference genes at different developmental stages of the olive mesocarp and the first to assess the stability of these genes across different cultivars. For an accurate comparison of mRNA levels in different samples, it is crucial to normalize the expression of target genes against appropriate reference genes. An ideal reference gene should be expressed at constant level in all types of cells, at any time in cell cycle and differentiation and/or with any sample treatment[9, 10, 31]. Traditional reference genes such as EF1-alpha and OUB2 (involved in basic cellular processes) or those encoding actin and tubulin (involved in cell structure maintenance) have been frequently used in qPCR experiments[3, 5, 7, 8]. However research has shown that these genes may be inappropriate for normalization in qPCR experiments due to their expression variability under different experimental conditions, confirming the need to validate expression stability of reference genes in given species and organs/tissues under specific experimental conditions[8, 13]. Normalization with multiple reference genes has become a gold standard in qPCR expression analysis[3, 9, 10] and also a requisite according to MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) guidelines[2]. To this end, a number of software packages have been developed to assess the stability of candidate reference genes in different biological experimental settings, including, GeNorm, BestKeeper and NormFinder. However, the validation of reference genes is not very common in plant research[10, 16, 18].

Recently two studies were conducted to identify stable reference genes in olives during fruit development and ripening[29, 30]. Although stable reference genes were suggested for the olive fruit developmental stages, a few reference genes, such as EF1-alpha, GAPDH, 14-3-3 protein performed differently in these two experiments during the olive fruit development with differences in their M-values and ranking order. These differences clearly show that the stability of the reference genes must be validated with each experimental setup. As two different cultivars were used in these two studies it is possible that differences in cultivars might explain these differences in gene expression between the reference genes indicating that even for different cultivars the stability of the reference genes should be assessed. Therefore, in this study a combination of three different olive cultivars Barnea, Frantoio and Picual were assessed both individually and together to evaluate the expression stability of eight candidate reference genes during olive mesocarp development.

The primers that were designed for the eight reference genes amplified single PCR products of the expected size from the olive cDNA pools as shown by gel electrophoresis (Additional file1) and melt-curve analysis (Additional file3) suggested that single products were amplified and that only TIP2 showed the formation of some primer-dimers. This specificity was confirmed by sequencing all eight PCR products which confirmed their identity to sequences used for designing primers for the reference genes.

Previous research has shown that analysis of expression stability of reference genes using different combinations of GeNorm, BestKeeper and NormFinder can result in minor to substantial discrepancies in the final ranking of the suitable reference genes which is typically explained by the differences in the mathematical models associated with each program[3, 17]. In this study the results obtained by the two algorithms did not show much discrepancy and both the programs were compared for the final choice of suitable reference genes. The ranking of the eight candidate reference genes based on their stability (M) and correlation coefficient values as calculated using GeNorm (qBase Plus) and BestKeeper algorithms respectively across the three cultivars together and as individual cultivars are shown in Tables 2 and3. According to GeNorm analysis in the three cultivars together, the two traditional reference genes GAPDH and EF1-alpha were the most stable with lowest M-values, followed by the reference gene PP2A, 60S RBP L18-3 and PTB were placed in the middle of the ranking, while OUB2, TUBA and TIP2 displayed inappropriate expression stability with higher M-values and thus appear to be regulated in these tissues.

According to BestKeeper analysis of the three cultivars together, PP2A was ranked as the most suitable reference gene with the highest correlation coefficient of 0.805, followed by EF1-alpha and GAPDH. PTB and 60S RBP were placed in the middle of the ranking, while OUB2, TUBA and TIP2 showed SD ≥1 and thus were considered inconsistent. Although TIP2 has the highest correlation coefficient of 0.832, it cannot be used for normalisation because of its unacceptable SD value (SD > 1).

Thus, GAPDH, EF1-alpha and PP2A were determined to be the three most stable reference genes analysed across the three cultivars when analysed together, albeit with slightly different rankings using the different analytical approaches. Many studies with similar findings, wherein different software packages identify the same reference genes as the most stable genes but not in the exact same ranking order have been reported in both animals and plants[31, 32].

These findings are in broad agreement with the Nonis et al.[29] data which showed that PP2A1 was the most transcriptionally stable reference gene during olive fruit development, followed by GAPDH2, while GAPDH1 showed the widest variation and was considered the least stable gene. BLASTx of the GAPDH contig [GenBank:EF506530.1] used in this study showed that it is 97% identical to Glycine max GAPDH Subunit A [GenBank:NP_001238484.1]. BLASTx of the two different GAPDH transcripts used by Nonis et al[29] showed that [OLEEUCl022518|Contig2] and [OLEEUCl004899|Contig2] are 89-90% identical to Glycine max GAPDH Subunit C [GenBank: NP_001237544.1]. The PP2A contig [OLEEUCl010038 contig1] used in this study is 98% identical to Glycine max PP2A-2 [GenBank: XP_003532043.1]. BLASTx of the two different PP2A transcripts used by Nonis et al[29] showed that [OLEEUCl021775|Contig2] is 92% identical to Glycine max PP2A-3 [GenBank:XP_003533904.1] and [OLEEUCl021848|Contig 2] is 97% identical to Glycine max PP2A-1 [GenBank: XP_003534171.1]. Both PP2A transcripts used by Nonis et al[29] performed well in that study[29]. These studies clearly suggests that the PP2A gene family is stably expressed in olive fruit developmental stages. In addition, PP2A was also identified as a suitable reference gene in cotton across different plant organs[10] as well as in Arabidopsis[8]. GAPDH, a traditional reference gene, was considered the most suitable reference gene in coffee leaves under drought-stress and in different cultivars[18], however performed poorly across tissues and organs of tomato at different developmental stages[33]. EF1-alpha was found to be stably expressed in olive fruit developmental stages in two different studies, however with differences in M-values and ranking orders[29, 30]. Under conditions of biotic and abiotic stress, EF1-alpha was found to be very stably expressed in potato[34], while EF1-beta was found to be the most stable in soybean[35]. EF1-alpha was also found to be stable in expression across different tissues of rice[16].

PTB and 60S RBP L18-3 were placed in the middle of the ranking order across the three cultivars as a whole by both software packages, again differing in their order (Table 2). PTB was identified as one of the most stable reference gene in cotton during fruit development[10]. 60S RBP was identified as one of the suitable reference genes to normalize gene expression data in two different grapevine organs (leaves and berries)[3]. Resetic et al[30] also placed 60S RBP in the middle of the ranking order for assessing the average expression stability during olive fruit development.

Both GeNorm and BestKeeper ranked OUB2, TUBA and TIP2 as poor performers as reference genes across the three cultivars as a whole. Nonis et al[29] has ranked OUB2 in the middle of the ranking order with an M value of 0.48 while UBQ10 displayed a higher M value of 0.78 in olive fruit tissues by Resetic et al[30]. UBQ14 was identified as one of the most stable reference gene across different plant organs in cotton[10]. UBQ10 exhibits very stable expression in Arabidopsis[8] however performed poorly as a reference in soybean[9] and in grapevine[17]. TUBA displayed a higher M value of 0.80 in olive fruit tissues by Resetic et al[30]. TUBA was identified as being very stably expressed across development in soybean while was highly unstable across tissues and organs of tomato at different developmental stages[33]. It is interesting to note that though TIP41 was chosen as one of the best reference genes in olive fruit tissues in the cultivar Istrska belica[30], this reference gene performed poorly in the present study. While TIP2 outperformed several traditional reference genes in Arabidopsis across different tissues, organs and developmental stages[8], it performed poorly in grapevine. This analysis clearly suggested that reference genes are regulated differently in different plant species and/or cultivar[10] and also highlights the importance of validating putative reference genes in different species/tissues/cultivar/conditions.

GeNorm and BestKeeper was further used to evaluate the expression stability of the eight reference genes in each cultivar independently. It was interesting to note that the M values varied dramatically across the cultivars where M values in Barnea samples ranged from 0.9-2.4 while in Frantoio and Picual samples they ranged between 0.5-1.85. In Barnea, GAPDH, EF1-alpha and PP2A were determined to be the three most stable reference genes which is consistent with the analysis performed across the three cultivars together. In Frantoio, slight variations between the two algorithms were observed. Combining data from both the algorithms GAPDH, EF1-alpha and 60S RBP L18-3 were the top three stable reference genes identified in Frantoio while 60S RBP L18-3, OUB2 and PP2A were identified as the top three stable reference genes in Picual. All three cultivars ranked TIP2 as the least stable gene. Thus it is observed that though few best/worst performing reference genes are common amongst cultivars, their ranking orders do vary and the top ranking genes should be used for their respective cultivars for accurate normalization. This analysis further validated our earlier statement that reference genes may show slightly different expression stability across cultivars.

GeNorm also provides a measure for the optimal number of stable controls that should be used for normalization based on pairwise variation analysis between subsequent normalisation factors. According to GeNorm V, a combination of six most stable reference genes was calculated as being optimal for gene expression studies across the three cultivars when analysed together with the lowest pairwise variation value of 0.18 (Figure 2B). In Barnea, Frantoio and Picual, individually, the lowest pairwise variations were for V5/6, V3/4 and V5/6, respectively (Figure 3B, D, F). According to Vandesompele[11] the optimal cut-off V number should be around 0.15, however many other studies using this application have resulted in higher pairwise variations[7, 36–38]. The GeNorm threshold is not a strict cut-off value but it’s an ideal value to provide guidance to researchers to determine the optimal number of reference genes and that the observed trend of changing pairwise variation values is equally informative[9, 39]. Depending on the aim of the study, the optimal number of stable controls that should be used for normalization against the target genes should be decided. In case of a comparatively small study, it is impractical to use excessive numbers of reference genes for normalization and thus the minimal use of three most stable reference genes is recommended for calculating the normalization factors[11, 31, 40]. Therefore GAPDH, EF1-alpha and PP2A were recommended as the three most stable reference genes for normalization against the target genes across the three cultivars together and in the cultivar Barnea. In Frantoio, GAPDH, EF1-alpha and 60S RBP L18-3 were recommended as the three most stable reference genes while in Picual, 60S RBP L18-3, OUB2 and PP2A were recommended as the three most stable reference genes.

In this study we have investigated the expression of eight candidate reference genes at different developmental stages of the olive fruit across three different olive cultivars independently and together, in an attempt to identify most suitable reference genes for normalizing gene expression across cultivars. GAPDH, EF1-alpha and PP2A were found to be the most stable reference genes in olive mesocarp tissues across the cultivars and in the cultivar, Barnea. GAPDH, EF1-alpha and 60S RBP L18-3 were found to be most stable reference genes in the cultivar Frantoio while 60S RBP L18-3, OUB2 and PP2A were found to be most stable reference genes in the cultivar Picual. In summary, this is one of the first reports on the evaluation of candidate reference genes across three different O. europaea cultivars and will provide guidance to other researchers to select reference genes for normalization against target genes by qPCR in this species.