Introduction

Anthocyanins are a group of flavonoids that are responsible for various colors in plant organs, and a high level of diversification in Plantae (Wrolstad 2004; Landi et al. 2015; Mushtaq et al. 2016). They play a significant role in the attraction of pollinators, frugivores, and dispersers; the repulsion of herbivores and parasites; the provision of anti-viral and anti-microbial activities, and resistance to the harmful effects of ultraviolet light (Schaefer et al. 2004; Stintzing and Carle 2004; Wrolstad 2004). These secondary metabolites are also essential to the color quality of fresh and processed fruits and vegetables and have been extensively studied by horticulturists and food scientists. Moreover, naturally occurring anthocyanins in plants have been given wide attention in recent years (Hidalgo and Almajano 2017), and are recognized for their high antioxidant capacity and reported positive human health effects (Guerrero et al. 2010; Wallace 2011; Bochi et al. 2015). Anthocyanins have become sought-after natural products and are in large demand for their medicinal and industrial uses. However, because only a few plants naturally produce anthocyanins, their widespread use is limited (Petrella et al. 2016). This limitation has prompted researchers to seek alternative renewable resources for anthocyanin biosynthesis and other plant secondary metabolites. Studies have shown that plant cell culture could be a useful method for producing natural secondary metabolites, and is also a perfect model system to study the biosynthesis of plant secondary metabolites (Dörnenburg and Knorr 1995; Oksman-Caldentey and Inzé 2004; Isah et al. 2018). Therefore, anthocyanin production in plant cell culture is an achievable technology that is being pursued industrially and academically (Zhang and Furusaki 1999; Davis et al. 2012; Simões et al. 2012; Karaaslan et al. 2013).

Spine grape (Vitis davidii [Rom. Caill.] Foëx.), also called the Chinese bramble grape and Davids Rebe, is a wild East Asian grape species (Meng et al. 2012). A majority of spine grape fruits are violet black and contain abundant flavonoids, especially anthocyanins, which is an excellent material for fruit processing and wine making. To take advantage of this valuable wild grape, a previous attempt was made to produce anthocyanins in cell culture. Two cell lines were obtained from young embryo-derived callus tissue, which could be subcultured in the long term and thrived during long-term maintenance (Lai et al. 2014). A few studies have reported that secondary metabolites, such as anthocyanins, were produced in grape cell cultures (Decendit and Mérillon 1996; Sato et al. 1996; Zhang and Furusaki 1999; Simões et al. 2012). As in other plant cell culture systems, several barriers to commercial utilization have not been overcome in grape cell culture, such as slow plant cell growth, low production rates, metabolite channeling, and poor metabolic regulation under certain conditions (Abdullah et al. 2005), which almost certainly reflects the lack of knowledge of the basic regulation of secondary metabolism in cultured plant cells, and anthocyanin production (Decendit and Mérillon 1996; Sato et al. 1996). Anthocyanins are derived from the flavonoid biosynthetic pathway. The anthocyanin biosynthetic pathway may be the most studied secondary metabolite pathway in plants, and the genes associated with anthocyanin biosynthesis, can be obtained from almost every biosynthetic step (Holton and Cornish 1995; Winkel-Shirley 2001; Solfanelli et al. 2006; He et al. 2010). Little is known about the mechanisms that govern the function and interaction of the structural genes and regulatory factors of anthocyanin biosynthesis in the cell, especially when cultured under specific conditions. Therefore, research on molecular regulatory mechanisms of structural genes and regulatory factors associated with anthocyanin biosynthesis at a cellular level is significant for both fundamental and applied issues.

Gene expression analyses are currently used to study the regulatory mechanisms of secondary metabolite pathways in plants. Real-time quantitative polymerase chain reaction (RT-qPCR) has been used in many analyses at the transcription level and is the most appropriate method for detecting changes in gene expression, because it has good repeatability, specificity, high sensitivity, and a wide dynamic range (Pfaffl and Hageleit 2001; Ginzinger 2002; Gutierrez et al. 2008; Freitas et al. 2017). However, substantial problems that affect the experimental data are inherent in RT-qPCR. One strategy to reduce potential errors is the introduction of reference genes to calibrate the RT-qPCR data (Ginzinger 2002; Huggett et al. 2005; Borges et al. 2012). A variety of housekeeping genes, such as actin, 18S rRNA, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), tubulin, and elongation factor 1-α (EF1-α), have been commonly used as reference genes (Libault et al. 2008; Hu et al. 2009). However, several studies have indicated that the expression of housekeeping genes could vary with different experimental treatments, and misrepresent the nature of the expression of a gene of interest (Thellin et al. 1999; Nicot et al. 2005; Jian et al. 2008). Thus, it is critical to systematically select one or more appropriate stably expressed reference genes to accurately quantitate gene expression for a specific treatment. To address this issue, two spine grape cell lines were used to establish a model system to study the production of anthocyanin and some other secondary metabolites. The expression profiles of the genes associated with the anthocyanin biosynthesis pathway were analyzed by RT-qPCR to facilitate understanding of the mechanisms involved in this process at the cellular level. Almost all current RT-qPCR studies in grapes used a tree or part of a tree as the plant material and used reference genes for normalization (Bogs et al. 2006; Reid et al. 2006; Espinoza et al. 2007; Gutha et al. 2010; Gamm et al. 2011; González-Agüero et al. 2013; Monteiro et al. 2013; Borges et al. 2014). There has not been a study to date that has identified a particular set of stable reference genes for grape cell culture. Thus, it was necessary to confirm the most stable reference gene(s) for future gene expression studies of spine grape cells under specific experimental conditions.

In the present study, the expression stability of 15 reference gene candidates was assessed using RT-qPCR with eight different experimental treatments of spine grape cells, and reliable reference genes to normalize the relative gene expression were confirmed.

Materials and methods

Spine grape cell lines

Two cell lines, designated as DLR and DLW, were obtained from spine grape (Vitis davidii [Rom. Caill.] Foëx.) callus tissue. These calluses had been subcultured in vitro for over 2000 d (96 generations) and were induced from one spine grape immature embryo as described by Lai et al. (2014). DLR is red-mauve in color, and DLW is light yellow-green (Fig. 1).

Figure 1.
figure 1

Two cell lines of spine grape (Vitis davidii [Rom. Caill.] Foëx.). (a) DLR cell line with red-mauve color. (b) DLW cell line with light yellow-green color.

Full size image

Experimental treatments

The expression stability of the reference gene candidates was assessed in eight independent experiments. (1) The cell cultures were cultured on Murashige and Skoog (MS) solid medium (Murashige and Skoog 1962) with 1.0 mg L−1 2,4-dichlorophenoxyacetic acid (2,4-D; Bio Basic Inc., Markham, ON, Canada) and harvested at 15, 25, 35, 45, or 55 d. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (2) The cell cultures were cultured on MS solid medium with 1.0 mg L−1 2,4-D under white light (color temperature is 6500 K, 62.5 μm m−2 s−1; Philips, Shanghai, China), warm light (color temperature is 3000 K, 62.5 μm m−2 s−1; Philips) and darkness for 25 d. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (3) The cell cultures were cultured on MS solid medium with 1.0 mg L−1 2,4-D for 23 d and irradiated with ultraviolet B (HuaQiang, Nanjing, China) for 0 h, 3 h, or 6 h at an intensity of 110 μW m−1, and then cultured for another 48 h. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (4) The cell cultures were cultured in MS solid medium with 1.0 mg L−1 2,4-D and removed from culture room (temperature 25°C) on day 23, and then cultured at 4°C, 25°C, or 37°C for 48 h. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (5) The cell cultures were removed from MS solid medium with 1.0 mg L−1 2,4-D on day 23, and cultured in MS liquid medium with 0 mg L−1, 20 mg L−1, or 40 mg L−1 cinnamic acid (Sangon Biotech, Shanghai, China) for 48 h. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (6) The cell cultures were transferred from MS solid medium with 1.0 mg L−1 2,4-D on day 23 and cultured in MS liquid medium supplemented with 0 mg L−1, 400 mg L−1, or 800 mg L−1 chitosan (Bio Basic Inc.) and harvested after 48 h in the presence of chitosan. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (7) The DLR cell line was subcultured on MS solid medium with 1.0 mg L−1 2,4-D and 0.0 mg L−1, 0.05 mg L−1, 0.5 mg L−1, or 1.5 mg L−1 kinetin (KT; Bio Basic Inc.) and harvested on day 25. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. (8) Two different cell lines (DLR and DLW cell lines) were both cultured on MS solid medium with 1.0 mg L−1 2,4-D and harvested on day 25 and day 35. All of the harvested cells were immediately frozen in liquid nitrogen and stored at − 80°C. Total RNA was isolated from all eight independent experiments for the following experiments. Three biological replications were performed for each treatment for all samples. All of MS solid media were supplemented with 30 g L−1 sucrose (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China), 6 g L−1 Agar powder (AOBOX, Beijing, China). All of MS liquid media were supplemented with 30 g L sucrose. All media were adjusted to pH 5.8 using 1.0 mol L−1 NaOH and were autoclaved (G154DW; Zealway Instruments, Inc., Wilmington, DE) at 121°C, 0.11 MPa for 20 min.

Total RNA isolation and cDNA synthesis

Of the harvested cells, 0.1 g was used for the following total RNA isolation. Total RNA was isolated using a Plant RNA Extraction Kit (Applygen Technologies, Inc., Beijing, China) according to the manufacturer’s instructions. Removal of the residual genomic DNA was done with an RQ1 RNase-Free DNase Kit (Promega®, Madison, WI). The RNA purity and concentration were automatically calculated at 260/280 nm using a NanoDrop™-2000 spectrophotometer (Thermo-Fisher Scientific®, Waltham, MA), and the RNA integrity was confirmed by agarose gel electrophoresis on a 1.5% (w/v) agarose gel (BIOWEST®, Nuaillé, France). For each treatment including the control, three biological replicate RNA samples were isolated from each mixed cell culture, and equal amounts were pooled to form the test sample. According to the manufacturer’s instructions, cDNA from each representative sample was synthesized from 1000 ng of total RNA using a PrimeScript™ RT reagent Kit (Perfect Real Time; Takara Bio, Inc., Shiga, Japan) anchored with an oligo(dT)18 primer in a 20-μL volume reaction.

Selection of candidate genes

A total of 22 reference gene candidates were selected according to previous studies in Vitis vinifera, including actin (actin-7, other hits include ACT1, ACT2), clathrin adaptor complex subunit mu (AP-2, AP47), peptidyl-prolyl cis-trans isomerase (Cyclophilin), SAND family protein (SAND), V-type proton ATPase subunit G (VAG), EF1-α, GAPDH, tubulin alpha-3/alpha-5 chain (α-tubulin), tubulin beta-1 chain (β-Tubulin), ubiquitin-60S ribosomal protein L40 (UBQ-L40), phosphoenolpyruvate carboxylase, housekeeping isozyme (PEP), peptidyl-prolyl cis-trans isomerase CYP20-2 (CYP), NADH dehydrogenase subunit 5 (NAD5), Alcohol dehydrogenase 2 (ADH2), 60S ribosomal protein L18-2 (60SRP), ubiquinol-cytochrome-c reductase complex assembly factor 1 (UQCC), V-type proton ATPase 16-kDa proteolipid subunit (VATP16), 18S ribosomal RNA (18S rRNA), T-complex protein 1 subunit beta (VvTCPB), small nuclear ribonucleoprotein (SMD3), ubiquitin carrier protein E2 C (UBE2), and ubiquitin-conjugating enzyme E2-17 kDa (UBQ). This group of genes was investigated as the associated reference genes in each experiment.

Design of RT-qPCR primers and validation of gene amplification

Primers for some of the candidate genes in this experiment were available from the literature, and others were designed according to de novo assembly sequences from the transcriptome of spine grape cell cultures (data not shown). These latter primer pairs were designed using the online program Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/), with an optimal Tm of 60°C, GC% between 40 and 60%, primer length of 18–24 bp, and amplicon length between 104 and 287 bp. The specificity of all primer pairs was initially assessed by RT-qPCR and indicated by a melting-curve with a single peak. A single amplification band of the expected size for each gene was further checked by 2.0% (w/v) agarose gel electrophoresis with ethidium bromide (EB) staining. The electrophoretic bands and melting-curve were the criteria used to select 15 reference gene candidates from the 22 preliminary candidate genes. The expression profiles of these 15 genes, actin, AP-2, cyclophilin, EF1-α, GAPDH, VAG, SAND, α-tubulin, β-tubulin, UBQ-L40, NAD5, ADH2, 60SRP, 18S rRNA, and UBQ, were further analyzed in subsequent experiments. The primer pair sequences, melting temperature, amplicon size, and amplification efficiency are shown in Table 1.

Table 1. List of candidate genes, primers, and amplicon characteristics
Full size table

Real-time quantitative PCR

Real-time quantitative polymerase chain reactions were performed in 96-well microtiter plates with the Light Cycler® 480II System (Roche, Basle, Switzerland) using the SYBR® Premix Ex Taq™ II Kit (Takara Bio, Inc., Shiga, Japan) to verify dsDNA synthesis. The PCR reactions were completed in 20-μL volumes, and each reaction contained 10 μL of SYBR® Premix Ex Taq™ II (Takara Bio Inc.), 1.0 μL each of the upstream and downstream primer (10 μM), 2 μL of cDNA, and 4 μL of sterile nuclease-free water. The RT-qPCR reaction conditions were as follows for PCR tubes in 96-well plates: a denaturation step at 95°C for 30 s; 40 cycles of 5 s at 95°C, and 30 s at 60°C; then, the melting-curve with denaturation at 95°C for 5 s, cooling at 60°C for 1 min, and gradual heating at 0.11°C per s, up to 95°C, and finally rapid cooling, at 2.2°C s−1, down to 50°C. All RT-qPCR reactions were carried out with three replicates, and each primer pair was checked with a non-template control (NTC).

The expression stability of the 15 reference gene candidates was determined by RT-qPCR. The LightCycler® 480 Software Version 1.5 (Roche) automatically determined each gene crossing point (Cp), and the amplification efficiency (E). Standard curves were produced with RT-qPCR using 5-fold serially diluted mixed cDNAs from all experimental samples as the templates. Table 1 shows a summary of the results.

Statistical analyses and validation of the reference genes

Data analysis was performed for the eight treatments. The RT-qPCR Cp data were obtained using LightCycler® 480 Software Version 1.5 (Roche, Basle, Switzerland). The expression stability of the 15 reference gene candidates for each experimental group was analyzed with three statistical methods, including geNorm (Vandesompele et al. 2002), NormFinder (Andersen et al. 2004), and BestKeeper (Pfaffl et al. 2004), and the stability rankings of the 15 reference gene candidates were comprehensively assessed using RefFinder (http://150.216.56.64/referencegene.php) for the different sample datasets. Finally, geNorm was used to calculate the pairwise variation of two sequential normalization factors (V(n/n + 1)) as the standard deviation of the logarithmically transformed NFn/NFn + 1 ratios, and the optimal number of reference genes to accurately normalize the RT-qPCR data was confirmed. A threshold value of 0.15 was recommended by geNorm to standardize the expression stability. It has been proposed that a pairwise variation value for n genes of less than 0.15 indicated that it was unnecessary to add another reference gene.

Normalization of DFR

Dihydroflavonol 4-reductase gene (DFR; GenBank accession number KF915803) was cloned from spine grape cell cultures and was used as a gene of interest to investigate the availability of the selected reference genes in RT-qPCR. In different cell cultures of spine grape with different concentrations of cinnamic acid for different durations (two of eight experimental conditions were randomly selected), the relative expression of DFR was analyzed using one, two, or three of the most stable reference genes, in addition to the least stable reference gene, to normalize the data. These reference genes were comprehensively ranked and validated as indicated by RefFinder. The primer pair was designed forward: 5-ACCAACTGCCAGTGTACGATGA-3 and reverse: 5-GCTTGCTCAGCCAGTGTCTTG-3, to determine the quantitative expression of DFR.

Results and discussion

Real-time quantitative polymerase chain reaction is an important method and efficient tool for the study of gene expression profiles (Nolan et al. 2006). However, the reliability of RT-qPCR data depends extensively on the appropriate reference gene(s), and the expression of these reference gene(s) must not change with different experimental conditions (Czechowski et al. 2005). However, because of the complexity of experimental treatments, one gene cannot maintain constant expression under all conditions (Wong and Medrano 2005). For example, EF1-α, GAPDH, actin, and SAND were recommended as reference genes for normalization in grape berry development studies (Reid et al. 2006), UBC, VAG, and PEP were used in grapevine leaves, and EF1, CYP, and UBC were the most stable internal genes for use in grapevine leaves treated with P. chlamydospora (Borges et al. 2014). In another experiment, NAD5 and Actin were used as the most stable genes to normalize the gene expression data from grapevine leaves (Gutha et al. 2010). Therefore, stable reference gene(s) for RT-qPCR must be identified for each new set of treatments (Wu et al. 2016).

Grapes are an economically important fruit crop worldwide. Many studies of grape molecular biology, including gene expression studies, were done during the last few decades. Some reference genes have been validated to normalize the target gene expression data in grape gene expression studies (Selim et al. 2012; Upadhyay et al. 2015; Tashiro et al. 2016; Katayama-Ikegami et al. 2016). However, an analysis of gene expression at the transcriptional level in spine grape using RT-qPCR has not yet been reported in the literature. The present study is the first effort focused on systematically validating stable reference genes for RT-qPCR analysis in spine grapes, particularly in cell culture.

Reference gene candidates and their amplification efficiency

Twenty-two candidate genes were selected for investigation from the relevant literature (actin, AP-2, cyclophilin, EF1-α, GAPDH, VAG, SAND, α-tubulin, β-tubulin, UBQ-L40, PEP, CYP, NAD5, ADH2, VATP16, 60SRP, UQCC, 18S rRNA, VvTCPB, SMD3, UBE2, and UBQ). These genes had been successfully used to normalize RT-qPCR data in previous studies with Vitis vinifera (Table 1). Primers were used for some candidate genes that were available from sources in the literature (actin, AP-2, GAPDH, β-tubulin, 18S rRNA, VvTCPB, and UBQ), and others were designed from the transcriptome de novo assembly sequences from spine grape cell cultures (data not shown). The amplification specificity of the 22 reference gene candidates was assessed by RT-qPCR using a similar pooled cDNA sample, which consisted of a mixture of equal amounts of each sample. Fifteen genes were selected from the 22 genes for an expression stability analysis as indicated by the specific amplification, the expected length of the single products in agarose gel electrophoresis with EB staining (Fig. 2), and a melting-curve with a single peak. The 15 genes selected were actin, AP-2, cyclophilin, EF1-α, GAPDH, VAG, SAND, α-tubulin, β-tubulin, UBQ-L40, NAD5, ADH2, 60SRP, 18S rRNA, and UBQ. Standard curves for these 15 genes were constructed using five-fold serial dilutions of a pooled cDNA sample. The PCR amplification efficiencies varied from 1.903 to 2.071 (Table 1).

Figure 2.
figure 2

Specificity of real-time quantitative polymerase chain reaction amplification and amplicon length. The amplicons of 22 genes were separated by 2.5% (w/v) agarose gel electrophoresis with ethidium bromide staining. M represents DL2000 DNA marker (TaKaRa). 1~22 represents, actin-7 (actin, other hits include ACT1, ACT2), AP47 (AP-2, clathrin adaptor complex subunit mu), cyclophilin (peptidyl-prolyl cis-trans isomerase), elongation factor 1-α (EF1-α), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), V-type proton ATPase subunit G (VAG), SAND family protein (SAND), tubulin alpha-3/alpha-5 chain (α-tubulin), tubulin beta-1 chain (β-tubulin), ubiquitin-60S ribosomal protein L40 (UBQ-L40), phosphoenolpyruvate carboxylase, housekeeping isozyme (PEP), peptidyl-prolyl cis-trans isomerase CYP20-2 (CYP), NADH dehydrogenase subunit 5 (NAD5), alcohol dehydrogenase 2 (ADH2), V-type proton ATPase 16-kDa proteolipid subunit (VATP16), 60S ribosomal protein L18-2 (60SRP), ubiquinol-cytochrome-c reductase complex assembly factor 1 (UQCC), 18S ribosomal RNA (18S rRNA), T-complex protein 1 subunit beta (VvTCPB), small nuclear ribonucleoprotein (SMD3), ubiquitin carrier protein E2 C (UBE2), and ubiquitin-conjugating enzyme E2-17 kDa (UBQ), respectively.

Full size image

Expression profiling of the selected reference genes

To determine the range of variation of the selected reference genes, the expression profiles of all 15 reference genes were analyzed according to their crossing point (Cp) for all samples (Fig. 3). The 18s rRNA gene had the highest abundant expression with a mean Cp of 9.4, and SAND had the lowest expression with a mean Cp of 29.9. The mean Cp values of the other genes ranged from 23.1 to 29.8. In general, the lowest range of variation is the most favorable. The range of variation of the Cp values of 18s rRNA was less than three amplification cycles, and it was less than four cycles for cyclophilin. The largest variance was for AP-2, with more than 12 cycles. The variation of the Cp values in the others was between four and eight cycles. The analysis revealed obvious changes in the expression of the selected reference genes in each treatment. To ensure the reliability of these reference genes, the data had to be grouped according to treatment.

Figure 3.
figure 3

Real-time quantitative polymerase chain reaction crossing point (Cp) values for 15 selected reference genes. Values are given in the form of Cp values across all experimental samples. The boxes represent mean Cp values, the bars represent standard deviations.

Full size image

Expression stability of the 15 reference genes

The expression of the 15 reference genes varied extensively by treatment, therefore, a statistical analysis was necessary to investigate the stability of these genes, and to ascertain the optimal number of reference genes needed to calibrate the gene expression for each treatment. To obtain more detailed expression profiles for the selected reference genes, all samples were grouped into eight treatments according to culture duration, illumination conditions, UVB exposure duration, temperature, cinnamic acid, chitosan, KT concentrations, and cell lines. The expression stability of the 15 reference genes for the eight treatments was calculated using geNorm, NormFinder, and BestKeeper statistical algorithms. The appropriateness of the genes was comprehensively evaluated with RefFinder, which is an integrated tool combining geNorm, Normfinder, comparative Delta Ct, and BestKeeper. The geometric mean of the individual gene weights was used to evaluate the comprehensive ranking of the reference genes, which considered each method’s rank and stability value.

GeNorm analysis

GeNorm software was used to evaluate the expression stability of the 15 reference genes, and these values (M) are displayed in Table 2. Gene expression stability is inversely correlated with the M value, in which a lower M value indicates more stable expression, while a higher M value indicates less stable expression. For culture duration, the M value of ADH2 was the highest, and that of AP-2 and SAND was the lowest, which showed that ADH2 was the most unstable gene, while AP-2 and SAND were the most stable. For illumination conditions, 60SRP and 18S rRNA had the most stable expression, and UBQ-L40 was the most unstable. For UVB exposure, AP-2 and VAG had the most stable expression, and ADH2 was the most unstable. For temperature differences, VAG and 60SRP had the most stable expression, and cyclophilin was the most unstable. For cinnamic acid exposure, AP-2 and 60SRP had the most stable expression, and cyclophilin was the most unstable. For chitosan treatments, VAG and α-tubulin had the most stable expression, and AP-2 was the most unstable. For KT exposure, ADH2 and α-tubulin had the most stable expression, and cyclophilin was the most unstable. In general, EF1-α and β-tubulin were the most stable genes among the 15 reference genes, but ADH2 was the least stable gene in different cell line samples of spine grape.

Table 2. Ranking of 15 candidate reference genes in eight subsets of spine grape (Vitis davidii [Rom. Caill.] Foëx.) cell cultures as calculated by geNorm
Full size table

NormFinder analysis

The data for the eight treatments were further assessed with NormFinder software and compared with the geNorm results, which indicated that the lower the expression stability values, the more stable the gene expression. Table 3 shows that the top five most stable genes were α-tubulin, EF1-α, VAG, SAND, and AP-2 for culture duration, while the most unstable gene was ADH2. The geNorm results for the top five stable genes were in excellent agreement with NormFinder, but the rank order was different. However, the most unstable gene was the same. For most other treatments, NormFinder and geNorm both identified the same top five stable genes, except for illumination conditions, UVB exposure, and KT treatment (Tables 2 and 3). Interestingly, the changes in the rank order did not affect the agreement regarding the least stable gene by the two software packages for any treatment.

Table 3. Ranking of 15 candidate reference genes in eight subsets of spine grape (Vitis davidii [Rom. Caill.] Foëx.) cell cultures as calculated by Normfinder
Full size table

BestKeeper analysis

The reference gene expression stability was ranked by the BestKeeper algorithm on the basis of the standard deviation (SD) and the variation coefficient of the Cp value. A lower SD value indicated more stable gene expression, and a higher value indicated greater instability. If the SD value was greater than 1.0, it was considered inappropriate as a reference gene. The results of BestKeeper analysis of the eight data groups are displayed in Table 4. For culture duration, SD values of the 14 reference genes were less than 1.0, which indicated little variation in gene expression, and ADH2 with an SD of 1.48 was the most variable. For illumination conditions, UVB treatment, and temperature, all of the reference gene candidates showed SD values less than 1.0. For cinnamic acid exposure, cyclophilin, 18S rRNA, ADH2, β-tubulin, NAD5, and 60SRP genes had SD values less than 1.0. For chitosan, 14 reference gene candidates had SD values less than 1.0, but AP-2 with an SD of 4.24 was the most unstable. For KT treatments, the SD values of cyclophilin, 18S rRNA, and NAD5 were less than 1.0, but the other genes had SD values greater than 1.0. For the cell lines, actin, 18S rRNA, cyclophilin, SAND, GAPDH, and EF1-α had SD values less than 1.0, while SD values of the other genes were greater than 1.0. The reference gene ranks according to BestKeeper were different than those ranked by geNorm and NormFinder (Tables 2, 3, and 4).

Table 4. Ranking of 15 candidate reference genes in eight subsets of spine grape (Vitis davidii [Rom. Caill.] Foëx.) cell cultures as calculated by Bestkeeper
Full size table

RefFinder analysis

The reference gene rankings among the above three algorithms were different; therefore, it was not possible to find an ideal all-purpose statistical method, so another method was necessary to optimize the data. The comprehensive algorithm RefFinder was utilized to assess the geNorm, NormFinder, and BestKeeper results. RefFinder is a practical online integrated tool, which is a collection of four currently available algorithms, including the comparative ΔΔCt method, geNorm, Normfinder, and BestKeeper. RefFinder was used to provide an overall final ranking of the expression stability for the reference genes in all of the treatments according to the weighted geometric mean of the rankings of each individual algorithm. The rankings of the reference genes from four algorithms, and the comprehensive RefFinder ranking were obtained by inputting the Cp values into this integrated tool. The results from all treatments according to RefFinder are shown in Table 5. The stability rankings of the reference gene candidates for each treatment by RefFinder shared a high uniformity with the rankings obtained from geNorm and Normfinder (Tables 2 and 3). The top four or five genes were basically the same with slight differences in the ranking. However, some obvious differences were apparent between the RefFinder and BestKeeper rankings (Table 4). These analyses illustrated that the geNorm and Normfinder rankings satisfied the accuracy requirements of RT-qPCR. These analyses also indicated the potential value of RefFinder for optimizing reference gene choices for RT-qPCR.

Table 5. The comprehensive ranking of 15 candidate reference genes in eight subsets of spine grape (Vitis davidii [Rom. Caill.] Foëx.) cell cultures calculated by RefFinder
Full size table

Confirmation of the optimal number of reference genes

To enhance the applicability of reference gene candidates obtained by the above-combined analyses, the optimal number of reference genes required for accurate normalization was suggested by geNorm and was based on the pairwise variation (V(n/n + 1)) between two sequential normalization factors NFn and NFn + 1. More reference genes are recommended until the V(n/n + 1) value reaches a threshold of less than 0.15 (Vandesompele et al. 2002). The optimal number of reference genes for the eight treatments is shown in Fig. 4. The analysis indicated that the pairwise variation of the V2/3 values for the eight treatments was less than the threshold of 0.15. Therefore, the two most stable reference genes for the eight test groups are shown in Table 6, based on the above analyses, thus eliminating obstacles from genes with large expression variability. The ideal reference gene combinations for culture duration were SAND and VAG, and α-tubulin and β-tubulin for illumination conditions, UBQ and VAG for UVB exposure, VAG and 60SRP for temperature, AP-2 and 60SRP for cinnamic acid, α-tubulin and UBQ for chitosan, actin and ADH2 for KT, and β-tubulin and EF1-α for the cell lines.

Figure 4.
figure 4

Pairwise variation (V(n/n + 1)) analysis of the candidate reference genes. Determination of the optimal number of reference genes for normalization via pairwise variation (V(n/n + 1)) calculated by geNorm software. A threshold value of 0.15 was adopted below, which represents n genes adequate for normalization.

Full size image
Table 6. The most stable reference genes and the best combinations for eight experimental groups
Full size table

These ideal reference gene combinations for eight different experimental treatments at the cellular level were obviously different from the predecessor’s researches in grape plants. For example, some reference genes were utilized for normalization in grape berry development studies (Reid et al. 2006), such as α-tubulin, actin, AP-2, cyclophilin, EF1-α, GAPDH, SAND, β-tubulin, and UBQ-L40. Three genes, VAG, CYP, and PEP, were the most stable reference genes to be used under biotic and abiotic stress in Vitis vinifera (Borges et al. 2014). The gene NAD5 was validated as a stable reference gene for grapevine leaves (Gutha et al. 2010). The expression stability of seven genes was studied in pterostilbene synthesis of grapevine (Gamm et al. 2011), including ADH2, UBE2, VATP16, 60SRP, UQCC, SMD3, and 18S rRNA. The gene VvTCPB was found to be the most stable gene to study grape genotypes and phenological stages (Monteiro et al. 2013) and GAPDH and UBQ were the most reliable genes to study the effect of genotype in susceptible and resistant Vitis vinifera cultivars (Monteiro et al. 2013). Therefore, it was very necessary to validate the most stable reference gene(s) before trials are conducted under different experimental conditions. It was interesting that the 18S rRNA gene showed highly stable expression under illumination conditions or with UVB treatment, but also had a low Cp value (< 10 cycles) with the large quantity transcripts in these samples. Previous studies showed that some problems might be encountered when using this gene as a reference in gene expression analysis of chicory (Maroufi et al. 2010). Therefore, the 18S rRNA gene was not used as a reference gene in the following study.

Demonstration of the usefulness of the recommended reference genes

Dihydroflavonol 4-reductase (DFR, GenBank accession number KF915803), is a member of the short-chain dehydrogenase family, and catalyzes the NADPH-dependent reduction of dihydroflavonols to flavan 3,4-diols (Petit et al. 2007), which is the last common step in the biosynthesis of flavan-3-ols and condensed tannins (Trabelsi et al. 2011). For different cell cultures of spine grape with different culture durations, SAND, VAG, and AP-2 were used as the most stable reference genes, while ADH2 was the most unstable gene. The expression profiles of DFR showed similar variation with little differences when normalized with SAND and SAND + VAG (Fig. 5a, b). The trend of DFR expression first decreased, then increased, and decreased again, which indicated that two stable reference genes were adequate for valid normalization. When normalized with SAND + VAG + AP-2, the expression profiles of DFR increased at 45 d (Fig. 5c), which illustrated that the precision of RT-qPCR could be reduced when an unnecessary reference gene was added. However, the opposite result was obtained with ADH2 as the reference gene, that is, the expression of DFR was increased at first and then decreased (Fig. 5d), which showed that it is necessary to select reliable reference genes for normalization.

Figure 5.
figure 5

Relative expression levels of the target gene dihydroflavonol 4-reductase (DFR) normalized with different reference gene(s). (a, b, c, d) Relative expression of DFR in different spine grape (Vitis davidii [Rom. Caill.] Foëx.) cell lines at different culture times. (e, f, g, h) Relative expression of DFR under different concentrations of cinnamic acid.

Full size image

For different cell cultures of spine grape grown in different concentrations of cinnamic acid, AP-2, 60SRP, and GAPDH were utilized as the three most stable genes for normalization, while Actin was the lowest ranked gene. The expression profiles of DFR indicated similar variation when AP-2, AP-2 + 60SRP, AP-2 + 60SRP + GAPDH, or actin were used as the reference gene(s) (Fig. 5e–h). A higher concentration of cinnamic acid led to greater expression of DFR. The expression of DFR was slightly different when normalized with AP-2, AP-2 + 60SRP, and AP-2 + 60SRP + GAPDH. However, DFR expression was over-estimated more than two-fold in cell cultures with 20 mg L−1 or 40 mg L−1 cinnamic acid when normalized with Actin, which illustrated that normalization with an inappropriate reference gene had an adverse impact on the reliability of the experimental results.

According to the above testing, the DFR expression profiles were almost the same when one or two of the most stable reference genes were utilized. To improve the veracity of RT-qPCR, the number of reference genes must be considered in the standard revision of the test results. However, when the pairwise variation V2/3 values were less than the threshold of 0.15, no additional reference genes should be introduced. The precision of RT-qPCR could even be reduced if unnecessary reference genes are added. In the present study, a combination of two stable reference genes was confirmed for specific experiments. The DFR expression profiles changed if inappropriate reference genes were used, which indicated that different reference gene(s) were necessary to normalize data for different treatments because factors such as the RNA quantity and quality, reverse transcription efficiency, and PCR amplification will all affect the accuracy of the RT-qPCR. The use of inappropriate reference genes may cause bias and even result in misleading conclusions.

Conclusions

This study is the first attempt to identify reference genes for RT-qPCR in cell cultures of spine grape. The expression stability of 15 reference gene candidates was surveyed using a large number of test samples from eight treatments. The results showed that different appropriate reference genes and different numbers of reference genes must be used to normalize the data. The DFR expression analysis verified the reliability of the validated reference genes and confirmed that validated reference genes were essential to improve the accuracy of RT-qPCR. The results of the present study offer the most robust platform for the most precise and broad application of RT-qPCR to investigate gene expression associated with anthocyanin biosynthesis in spine grape cell cultures.