Identification and Validation of Reference Genes for qRT-PCR Analysis of Petal-Color-Related Genes in Rosa praelucens

The flower’s color is regarded as one of the most outstanding features of the rose. Rosa praelucens Byhouwer, an endemic and critically endangered decaploid wild rose species, is abundant in phenotypic diversity, especially in flower color variation, from white to different degrees of pink. The mechanism underlying this variation, e.g., the level of petal-color-related genes, is worth probing. Seven candidate reference genes for qRT-PCR analysis, including tubulin α chain (TUBA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), histone H2B (Histone2A), eukaryotic translation elongation factor 1-α (EEF1A), 60S ribosomal protein (RPL37), eukaryotic translation initiation factor 1-α (EIF1A), and aquaporins (AQP), were detected from the transcriptome datasets of full blooming flowers of white-petaled and pink-petaled individuals, and their expression stabilities were evaluated through qRT-PCR analysis. According to stability rankings analysis, EEF1A showed the highest stability and could be chosen as the most suitable reference gene. Moreover, the reliability of EEF1A was demonstrated via qRT-PCR analysis of six petal-color-related target genes, the expression patterns of which, through EEF1A normalization, were found to be consistent with the findings of transcriptome analysis. The result provides an optimal reference gene for exploring the expression level of petal-color-related genes in R. praelucens, which will accelerate the dissection of petal-color-variation mechanisms in R. praelucens.


Introduction
Rosa praelucens Byhouwer refers to critically endangered wild rose species endemic to Shangri-La county in northwestern Yunnan, China [1,2].As a well-known alpine ornamental plant [3] and an important rose germplasm resource with cold tolerance [4] and aphid resistance [5], R. praelucens has drawn considerable attention [6][7][8][9][10][11][12] since it was found to be the only extant natural decaploid species (2n = 10x = 70) with the highest number of chromosomes in the genus [13].R. praelucens is abundant in phenotypic variation, among which flower color variation is most conspicuous [6,14].In addition to the common color fading with the opening process of a single flower, the colors of full-bloom flowers from different individuals within or among populations show significant differences.Most individuals have flowers that are in different degrees of pink, while some individuals have white flowers.
Flower color has a vital impact on the attractiveness of rose flowers for insects, and has an important aesthetic value for humans.Inherited from different genetic background species, most flower colors, except blue, are naturally present in roses [15], mostly identified by the constituent profile of the chemicals, e.g., flavonoids and anthocyanins.Limited flower colors can be obtained within a species because of genetic constraints [16].In roses, yellow and orange colors are caused by carotenoid pigment accumulation, while red colors are due to anthocyanin accumulation, mostly glycosylated xantho-cyaniding [17,18].Various homologues of anthocyanin biosynthesis pathway genes are accumulated in the process of pigment accumulation in the rose petals of R. chinensis 'Old Blush' when buds redden rapidly [17], and the unigene of the Cy-3,5-diglucoside-catalyzing enzyme (RhGT1) [19] decreases in the fully flowering stage when the petal color becomes lighter.This is completely different from in petunia.The formation of white-or red-colored petunia flowers is determined by the enzymatic competition between FLS and DFR, showing a close relationship to the disequilibrium of FLS and DFR expression [20].
Loss of duplicate gene expression and function, cisand trans-acting effects, RNAmediated pathways, and regulatory networks, etc., might have resulted in commonly seen phenotypic variation in polyploid wild plants and domesticated crops [21].Polyploidy results in an immediate elevation in gene dosage throughout the genome.With more gene copies, polyploids may have elevated (or altered) gene expression, which would exert potential influence on a lot of phenotypic traits [22].Also, epigenetic-regulated gene silencing might also immediately affect the gene expression for polyploid plants [23].For some dose-dependent genes, e.g., ETH, their expression has some differences between the diploid roses and the tetraploid roses [17].As for the decaploid R. praelucens, there is nothing known for explaining its phenotypic diversity, especially the flower color variation.As far as gene expression is concerned, it is necessary to screen the key related genes and their intraspecific expression first for elucidating the color variation within this decaploid species, which will then also be helpful for us to understand the effect of the most popular ploidy level, tetraploidy, on the phenotypic variation in modern roses.
Quantitative real-time PCR (qRT-PCR) is the most prominent and extensively used technique to quantify the expression of candidate genes owing to its high sensitivity, sufficient reproducibility, and wide quantification range, as well as ease of use [24].However, a lack of experimental standardization, including data normalization, directly influences the reproducibility and integrity of biological replications of qRT-PCR experiments [25].Data normalization is often established by containing stably expressed reference genes to correct an assay for sample-to-sample variation.In addition, the most popular reference genes are the "housekeeping genes", including actin (ACT), tubulin (TUBA and TUBB), eukaryotic translation elongation factor 1-α (EEF1A), ubiquitin (UBQ), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH).However, more and more studies have indicated that many of these genes are inappropriate for qRT-PCR analysis, as their expression could probably be altered by developmental and environmental factors [26][27][28].It is widely accepted that reference genes are specific to experimental situations and even the ploidy levels [29], and suitable reference genes have to be experimentally detected for each species and for each certain experiment [28,30,31].
In roses, PP2A and UBC were shown to be more stable and suitable than those traditional housekeeping genes as references for the study of stress [29].Different genes were chosen as the references under different postharvest conditions for more accurate and widespread application of qRT-PCR analysis in rose flowers, e.g., RhUBI1 for dehydration treatment and receptacles, RhTUB2 for exogenous ethylene, RhUBI6 for petals, etc. [32].As for the expression of petal-color-related genes in roses, there is no special selection of reference genes and the traditional ones adopted for Chrysanthemum [33], such as ACT [17] and GAPDH [34], are often used.
In this work, based on the transcriptome data of R. praelucens samples, with the aim of studying petal-color-related genes and their expression, we selected seven candidate reference genes and investigated their expression stability.The results provide optimal reference genes for exploring the level of petal-color-related genes in R. praelucens, as well as in modern roses, and also make contributions to understand the molecular mechanisms of intraspecific flower color diversity in plants.

Plant Materials and Sampling
Flowers at three different floral development stages of R. praelucens with white and pink petals (Figure 1) were used for RNA-seq.All the sampled plants were naturally distributed in Xiaozhongdian town of Shangrila county in southwestern Yunnan Province, China.The environment and climate of their habitats were almost identical.The flowers at the same development stage removed of sepals and the bottom three-quarters of the hypanthium from one plant were mixed as one sample and immediately frozen in liquid nitrogen.Three plants with the same petal color at almost the same age and growth status were sampled for each replicate, and then a total of 18 samples were gathered and preserved at −80 • C within a refrigerator until further experiments were conducted.

Plant Materials and Sampling
Flowers at three different floral development stages of R. praelucens with white and pink petals (Figure 1) were used for RNA-seq.All the sampled plants were naturally distributed in Xiaozhongdian town of Shangrila county in southwestern Yunnan Province, China.The environment and climate of their habitats were almost identical.The flowers at the same development stage removed of sepals and the bottom three-quarters of the hypanthium from one plant were mixed as one sample and immediately frozen in liquid nitrogen.Three plants with the same petal color at almost the same age and growth status were sampled for each replicate, and then a total of 18 samples were gathered and preserved at −80 °C within a refrigerator until further experiments were conducted.

Candidate Reference Genes, Petal-Color-Related Genes Selection, and Primer Design
Because color difference looks most significant and development is relatively stable at the full-blooming stage before senescence, the transcriptome dataset of the full-blooming stage was used for screening the candidate reference genes for color variation in line with the relative stability of the expression patterns of the genes (fragments per kb per million reads; FPKM) [35].The seven commonly used candidate reference genes including tubulin α chain (TUBA), GAPDH, histoneH2B (Histone2A), EEF1A, 60S ribosomal protein (RPL37), eukaryotic translation initiation factor 1-α (EIF1A), and aquaporins (AQP) were selected to be potential reference genes for exploring the relative expression of petal-color-related genes in R. praelucens (Table 1).Additionally, according to the functional annotation of DEGs (differentially expressed genes) from the transcript analysis, six genes related to petal color variation at the full-blooming stage were assigned for validation of the most reliable reference genes.Expression of all of them was up-regulated in transcriptome libraries of pink-flowered R. praelucens compared with white ones (Table 2).Primers of the candidate reference genes and the petal-color-related genes were designed with the use of the Primer Premier 5.0 software [36], as presented in Table 3.

Candidate Reference Genes, Petal-Color-Related Genes Selection, and Primer Design
Because color difference looks most significant and development is relatively stable at the full-blooming stage before senescence, the transcriptome dataset of the full-blooming stage was used for screening the candidate reference genes for color variation in line with the relative stability of the expression patterns of the genes (fragments per kb per million reads; FPKM) [35].The seven commonly used candidate reference genes including tubulin α chain (TUBA), GAPDH, histoneH2B (Histone2A), EEF1A, 60S ribosomal protein (RPL37), eukaryotic translation initiation factor 1-α (EIF1A), and aquaporins (AQP) were selected to be potential reference genes for exploring the relative expression of petal-color-related genes in R. praelucens (Table 1).Additionally, according to the functional annotation of DEGs (differentially expressed genes) from the transcript analysis, six genes related to petal color variation at the full-blooming stage were assigned for validation of the most reliable reference genes.Expression of all of them was up-regulated in transcriptome libraries of pink-flowered R. praelucens compared with white ones (Table 2).Primers of the candidate reference genes and the petal-color-related genes were designed with the use of the Primer Premier 5.0 software [36], as presented in Table 3.

RNA Extraction
In line with the specifications of the manufacturer, the total RNA of each sample was extracted using the RNAQUEOUS KIT Ambion-1912 (Oebiotech, Shanghai, China).The extracted RNA concentration and quality (OD260/OD280 value) were evaluated using a Nan-oDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA).Then, the integrity was confirmed with agarose gel electrophoresis (1%) stained with ethidium bromide.

Real-Time Quantitative RT-PCR
Quantification was carried out with a two-step reaction process including reverse transcription (RT) and PCR.Each RT reaction consisted of 0.5 µg RNA, 2 µL of 5× TransScript all-in-one SuperMix for qPCR, and 0.5 µL of gDNA Remover, in a total volume of 10 µL.The reactions were conducted within a GeneAmp ® PCR System 9700 (Applied Biosystems, Foster City, CA, USA).The reaction conditions were 15 min at 42 • C and 5 s at 85 • C. The 10 µL RT reaction mix was ten-time diluted with nuclease-free water, and then stored at −20 • C.
Real-time PCR was carried out with a LightCycler ® 480 II Real-time PCR Instrument (Roche, Basel, Switzerland) with 10 µL of PCR reaction mixture including 1 µL of cDNA, 5 µL of 2× PerfectStartTM Green qPCR SuperMix, 0.2 µL of forward primer, and 0.2 µL of reverse primer, as well as 3.6 µL of nuclease-free water.The primer sequences were synthesized by TsingKe Biotech, Beijing, China (Table 2).The reactions were subject to incubation within a 384-well optical plate (Roche, Basel, Switzerland) at 94 • C for 30 s, followed by 45 cycles of 94 • C for 5 s, 60 • C for 30 s.In addition, each sample was run in triplicate for performing the analysis.At the termination of the PCR cycles, a melting curve analysis was carried out, aiming to confirm the specific generation of the expected PCR product.Next, cycle threshold (Ct) values of each gene were automatically computed and then saved [37].

Analysis of Expression Stability
The expression stability of the candidate reference genes was assessed with three common software programs, geNorm version 3.4 [38], NormFinder version 20 [39], and BestKeeper version 1 [40], using the Ct values calculated from qRT-PCR data in accordance with the 2 −∆∆CT method [41].Firstly, geNorm was employed to compute the average expression stability (M value) of each candidate reference gene according to pairwise expression ratios and to identify the optimal number of reference genes by pairwise variation value (Vn/n+1, where n represents the reference gene number) [24,38].Usually, a gene can be chosen as a reference gene when its M value is below 0.15.The lower the M value, the more stable the gene [37].Secondly, NormFinder was applied to compute the stability value (SV) of the reference genes through analyzing the intra-and inter-group variation [39].Those exhibiting a lower level of expression than the average SVs were thought to be the most stable candidate reference genes [37].Bestkeeper, on the basis of the standard deviation value (SD) is usually further adopted for evaluating the expression stability of candidate reference genes [40].The programs were run in line with their respective manuals' instructions.The geometric mean method [42] was adopted for comprehensively ranking the expression stability of the candidate reference genes from the above software.Data analysis, statistics, and graphing were carried out by Microsoft Excel 2016 and Data Processing System [43].We obtained statistical differences through an independent t-test at p < 0.05.

Validation of Reference Gene
Additionally, the reliability of the top reference gene at the comprehensive ranking list was validated by comparing the expression pattern of six petal-color-related target genes (Table 2) via qRT-PCR analysis normalized by the selected reference gene with that in the transcriptome analysis.In addition, the qRT-PCR method remained the same as mentioned above, with three biological replicates.

Isolation of Candidate Reference Genes from Transcriptome Analysis
Based on the RNA-seq transcriptome data of R. praelucens, the FPKM value of the seven candidate reference genes are presented in Table 1.Except for the FPKMs of TUBA Genes 2024, 15, 277 6 of 12 and GAPDH, which were a little bit lower than 60, those of the other genes were higher than 100 (Table S1).

Primer Specificity and Expression Profile Analyses of Candidate Reference Genes
The seven candidate reference genes were amplified through qRT-PCR experiments, aiming to confirm the specificity of their primers.Their melting curves all showed a distinctive single peak without a primer dimer (Figure 2).All amplicons revealed good repeatability, indicating that the primers had strong specificity, reliability, and accuracy.As a result, all of the specific primers were shown to be appropriate for performing further qRT-PCR detection.
that in the transcriptome analysis.In addition, the qRT-PCR method remained the same as mentioned above, with three biological replicates.

Isolation of Candidate Reference Genes from Transcriptome Analysis
Based on the RNA-seq transcriptome data of R. praelucens, the FPKM value of the seven candidate reference genes are presented in Table 1.Except for the FPKMs of TUBA and GAPDH, which were a little bit lower than 60, those of the other genes were higher than 100 (Table S1).

Primer Specificity and Expression Profile Analyses of Candidate Reference Genes
The seven candidate reference genes were amplified through qRT-PCR experiments, aiming to confirm the specificity of their primers.Their melting curves all showed a distinctive single peak without a primer dimer (Figure 2).All amplicons revealed good repeatability, indicating that the primers had strong specificity, reliability, and accuracy.As a result, all of the specific primers were shown to be appropriate for performing further qRT-PCR detection.

Stability Analysis of Candidate Reference Genes
In accordance with the result of the pairwise variation value (Vn/n+1) between sequential normalization factors (Figure 4), because most of the Vn/n+1 values, e.g., V2/3, V3/4, V4/5, V5/6, are far less than 0.15, two reference genes might be required for the expression study of the target petal-color-related genes for R. praelucens.

Stability Analysis of Candidate Reference Genes
In accordance with the result of the pairwise variation value (Vn/n+1) between sequential normalization factors (Figure 4), because most of the Vn/n+1 values, e.g., V2/3, V3/4, V4/5, V5/6, are far less than 0.15, two reference genes might be required for the expression study of the target petal-color-related genes for R. praelucens.

Stability Analysis of Candidate Reference Genes
In accordance with the result of the pairwise variation value (Vn/n+1) between sequential normalization factors (Figure 4), because most of the Vn/n+1 values, e.g., V2/3, V3/4, V4/5, V5/6, are far less than 0.15, two reference genes might be required for the expression study of the target petal-color-related genes for R. praelucens.The average gene expression stability values (M value; Table 4) evaluated by geNorm showed that EEF1A exhibited the lowest M value (0.576) and EIF1A had the second lowest one (0.585), while those of RPL37, GAPDH, Histone2A, and TUBA were 0.608, 0.656, 0.714, and 0.728, respectively, all lower than 1.5.Only the M value of AQP was much higher than 1.5.Thus, EEF1A and EIF1A can be regarded as the most stably expressed candidate reference genes.In addition, the other four might also be suitable.The average gene expression stability values (M value; Table 4) evaluated by geNorm showed that EEF1A exhibited the lowest M value (0.576) and EIF1A had the second lowest one (0.585), while those of RPL37, GAPDH, Histone2A, and TUBA were 0.608, 0.656, 0.714, and 0.728, respectively, all lower than 1.5.Only the M value of AQP was much higher than 1.5.Thus, EEF1A and EIF1A can be regarded as the most stably expressed candidate reference genes.In addition, the other four might also be suitable.The expression stability values (SVs) computed by NormFinder (Table 4) indicated that the average SV of the seven candidate reference genes was 0.375.So, EEF1A, EIF1A, RPL37, GAPDH, and TUBA were more stably expressed, while Histone2A and AQP, with SVs of 0.399 and 1.296, respectively, were unstable.
Among the seven candidate reference genes, only the standard deviation (SD) of the gene expression of TUBA and AQP calculated by BestKeeper was lower than 1.0.The SDs of GAPDH and EEF1A were 1.086 and 1.114, respectively, while those of EIF1A, RPL37, and Histone2A were even higher.
Despite geNorm and NormFinder generating almost the same results for the stable reference genes in R. praelucens, the result computed by BestKeeper was somewhat different, since the computation was based on different criteria.The result of the GM method adopted for integrating the results from the three methods confirmed that EEF1A was the most stable candidate reference gene with the lowest GM value of 1.587, followed by EIF1A and TUBA, both with a GM value of 2.884.They can be used as suitable and reliable reference genes for studying petal-color-related genes in R. praelucens, while AQP and Histone2A were not so stable.

Validation of Selected Reference Gene EEF1A
To demonstrate the utility and stability of EEF1A as the reference gene in the study of petal-color-related genes in R. praelucens, the expression of the six selected target genes engaged in the petal color variation were evaluated.In addition, the expression pattern of these genes in the full-blooming flowers of white and pink individuals via qRT-PCR analysis, were normalized to that of the most suitable reference gene, EEF1A, according to the 2 −∆∆Ct method (Table S3).Based on the results, expression of the six petal-color-related genes was significantly up-regulated in the full-blooming flowers of pink-petal-colored individuals compared with their expression in those of the white ones (p < 0.01) (Figure 5), conforming to the result of transcriptome analysis (Tables 2 and S4).The expression pattern of these possible target genes was consistent with their transcriptomic analysis of RNA-seq.Thus, EEF1A is an appropriate reference gene for the gene expression of the petal-color-related genes in R. praelucens.

Discussion
Flower color is one of the most prominent characteristics of roses [44].With more gene copies, polyploids may have changed gene expression, and thus, affected many phenotypic traits [22].The decaploid R. praelucens has rich intra-species diversity in flower color, from white to different degrees of pink [6,14].The mechanism underlying this, especially the expression of related genes, needs to be deeply investigated.The expression of the dose-dependent genes has some differences between the diploid roses and the tetraploid ones [17,29].For the expression analysis of flavonoid/anthocyanin synthesis pathway-related genes, different reference genes have been used, e.g., AQP in Iris hollandica [45,46], ACT in Muscari botryoides [47] and in R. chinensis ('Old Blush') [17], 18S in hybrid Dendrobium [48], and GAPDH in hybrid roses [34], but none of them were specifically selected and validated.
Herein, we screened seven candidate reference genes in R. praelucens, including EEF1A, EIF1A, TUBA, GAPDH, AQP, Histone2A, and RPL37, and revealed that EEF1A was the most appropriate reference gene for the expression of petal-color-regulated

Discussion
Flower color is one of the most prominent characteristics of roses [44].With more gene copies, polyploids may have changed gene expression, and thus, affected many phenotypic traits [22].The decaploid R. praelucens has rich intra-species diversity in flower color, from white to different degrees of pink [6,14].The mechanism underlying this, especially the expression of related genes, needs to be deeply investigated.The expression of the dose-dependent genes has some differences between the diploid roses and the tetraploid ones [17,29].For the expression analysis of flavonoid/anthocyanin synthesis pathway-related genes, different reference genes have been used, e.g., AQP in Iris hollandica [45,46], ACT in Muscari botryoides [47] and in R. chinensis ('Old Blush') [17], Genes 2024, 15, 277 9 of 12 18S in hybrid Dendrobium [48], and GAPDH in hybrid roses [34], but none of them were specifically selected and validated.
Herein, we screened seven candidate reference genes in R. praelucens, including EEF1A, EIF1A, TUBA, GAPDH, AQP, Histone2A, and RPL37, and revealed that EEF1A was the most appropriate reference gene for the expression of petal-color-regulated genes.The expression pattern of six petal-color-regulated genes, determined via qRT-PCR analysis, using EEF1A as reference gene for normalization, were consistent with the transcriptome sequencing result, confirming the reliability and suitability of applying it as the reference gene in R. praelucens.The selected new reference gene in this study, EEF1A, was different from those that had previously been screened and validated for the petal-color-related gene expression of either Asiatic hybrid lilies, i.e., TUB, [49], or Anemone obtusiloba, i.e., UBQ [50], which reconfirmed that suitable reference genes had to be detected for each species and for each experiment [28,30,31].
Most flower colors in roses, especially the pink and red, are determined by chemicals such as flavonoids and anthocyanins [44,[51][52][53][54].The differences in petal color, brightness, and saturation of modern roses are mainly due to the change in expression quantity of genes such as CHS, DFR, and ANS, etc. [34].The FLS gene in R. rugosa 'Hunchun' is closely related to that in R. multiflora [55].Different homologues of anthocyanin biosynthesis pathway genes are related to the petal color variation in R. chinensis 'Old Blush' [17].As far as the key transcription factors are concerned, MYB, bHLH, and WD40 are the three most possible transcription factors regulating the expression of key enzyme genes in plants [56][57][58].Among them, MYB transcription factors, such as MYB1 [59,60] and R2R3-MYB [61,62] have been proved to be key transcriptors in rose anthocyanin biosynthesis.Among the six candidate petal-color-related genes, DN11691_c0_g1_i1_3, DN14168_c0_g1_i1_6, and DN12207_c0_g1_i2_9 were the most significantly up-regulated genes in the pink-petaled individuals.Therefore, these genes may indicate key candidate genes or transcriptors for deciphering the petal color variation in R. praelucens.As a result, functional characterization of these genes is needed in order to show their roles and clarify petal color molecular mechanisms in R. praelucens.The discovery of reference genes may speed up the molecular dissection of petal-color-variation mechanisms in R. praelucens.

Conclusions
To conclude, R. praelucens Byhouwer is an endemic and critically endangered decaploid wild rose species.It has rich intraspecific flower color variation from white to different degrees of pink.With the purpose of selecting reference gene for further study on the expression of petal-color-related genes, seven candidate reference genes, screened from a transcriptome dataset for qRT-PCR normalization of expression of petal-color-related target genes in R. praelucens, were investigated.EEF1A was shown to be the most stable reference gene.Its reliability was validated via qRT-PCR analysis of the expression pattern of six petalcolor-related DEGs, which were identical to the RNA-seq results.DN11691_c0_g1_i1_3, DN14168_c0_g1_i1_6, and DN12207_c0_g1_i2_9 were the most obviously up-regulated genes in the pink flowers.They might be the vital target genes for the intraspecific petal color variation in R. praelucens.

Supplementary Materials:
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/genes15030277/s1,Table S1: List of the seven candidate reference genes with their FPKM expression values; Table S2: Summary of the qRT-PCR Ct values of seven candidate reference genes; Table S3: qRT-PCR Ct values of the six target petal-color-related genes; Table S4: List of the six petal-color-related genes with their FPKM expression values.

Figure 1 .
Figure 1.Phenotypic variation in R. praelucens and the samples used for RNA-seq and qRT-PCR validation.W: white phenotype; P: pink phenotype.

Figure 1 .
Figure 1.Phenotypic variation in R. praelucens and the samples used for RNA-seq and qRT-PCR validation.W: white phenotype; P: pink phenotype.

Figure 2 .
Figure 2. Melting curves of the seven candidate reference genes in R. praelucens.

Figure 2 .
Figure 2. Melting curves of the seven candidate reference genes in R. praelucens.

Genes 2024 , 13 Figure 3 .
Figure 3. qRT-PCR Ct values for all candidate reference genes in six samples and 3 replicons each.

Figure 3 .
Figure 3. qRT-PCR Ct values for all candidate reference genes in six samples and 3 replicons each.

Figure 3 .
Figure 3. qRT-PCR Ct values for all candidate reference genes in six samples and 3 replicons each.

Figure 4 .
Figure 4. Pairwise variation analysis to identify the optimal number of reference genes in R. praelucens.

Figure 4 .
Figure 4. Pairwise variation analysis to identify the optimal number of reference genes in R. praelucens.

Table 1 .
The details of candidate reference genes screened from transcriptome libraries of R. praelucens.

Table 2 .
FPKM of the petal-color-related genes in the full opening flowers of different R. praelucens.

Table 3 .
Primer sequences for the qRT-PCR expression analysis of candidate reference genes and petal-color-related genes of R. praelucens.

Table 4 .
Stability ranking of candidate reference genes in R. praelucens using geNorm, NormFinder, BestKeeper software and the geometric mean (GM) method.