Selection and validation of appropriate reference genes for RT-qPCR analysis of flowering stages and different genotypes of Iris germanica L

Iris germanica L. is a perennial herbaceous plant that has been widely cultivated worldwide and is popular for its elegant and vibrantly colorful flowers. Selection of appropriate reference genes is the prerequisite for accurate normalization of target gene expression by quantitative real-time PCR. However, to date, the most suitable reference genes for flowering stages have not been elucidated in I. germanica. In this study, eight candidate reference genes were examined for the normalization of RT-qPCR in three I. germanica cultivars, and their stability were evaluated by four different algorithms (GeNorm, NormFinder, BestKeeper, and Ref-finder). The results revealed that IgUBC and IgGAPDH were the most stable reference genes in ‘00246’ and ‘Elizabeth’, and IgTUB and IgUBC showed stable expression in ‘2010200’. IgUBC and IgGAPDH were the most stable in all samples, while IgUBQ showed the least stability. Finally, to validate the reliability of the selected reference genes, the expression patterns of IgFT (Flowering Locus T gene) was analyzed and emphasized the importance of appropriate reference gene selection. This work presented the first systematic study of reference genes selection during flower bud development and provided guidance to research of the molecular mechanisms of flowering stages in I. germanica.

Expression stability of reference genes. To further select the most appropriate reference gene for RT-qPCR-based analysis in the investigation of flower development across three different I. germanica cultivars, four software programs, GeNorm, NormFinder, BestKeeper and RefFinder were used to analyze the expression stability of each reference gene.
GeNorm analysis. GeNorm program was used to evaluate the expression stability of the 8 candidate reference genes by calculating average expression stability (M) values based on the average pairwise variation among all the tested genes. According to GeNorm algorithm, stably expressed genes had M values below 1.5, and a relatively low M value indicates a relatively stable expression 10 . In this study, all of the tested genes showed high expression stability, with M-values of < 1.5, indicating that they all conformed to basic requirements for function as reference genes. IgUBC and IgGAPDH were the most stable reference genes in both '00246' and 'Elizabeth' , while IgTUB and IgGAPDH were identified as the most stable in '2010200' . In terms of the total samples tested, IgUBC and IgGAPDH were recommended as the most stable reference genes. In contrast, IgUBQ with the highest M value was identified as the least stable reference gene in all of the samples (Fig. 2). The optimal number of reference genes was also measured by determining the pairwise variation between sequentially ranked genes (Vn/Vn + 1) based on the GeNorm algorithm (Fig. 3). Generally speaking, a cutoff of 0.15 (Vn value) has been recommended as the threshold to determine the optimal number of reference genes 10 . Our results reveal that the V2/3 values of the '00246' , '2010200' and 'Elizabeth' samples were lower than 0.15 ( Fig. 3), suggesting that two reference genes were sufficient for accurate normalization. However, the value of 0.15 should not be a fixed threshold, and higher cutoff values of Vn/n + 1 have been shown in several reports 19,20 .
Our data showed small variation between V2/3 and V3/4 across all the samples, suggesting that two reference genes were sufficient for normalization (Fig. 3), which was similar to results in bermudagrass 21 and Kentucky bluegrass 22 .
NormFinder analysis. NormFinder is used to determine the stability value of reference genes, based on interand intragroup variance in different sample groups 11 . The stability value is then calculated, with a relatively low stability value meaning that the gene is relatively stable. Reference gene stability values were calculated by   Table 2. IgUBC and IgGAPDH were the two most stable genes among the total group, while IgUBQ was the least stable. The top two most stably expressed genes were IgUBC and IgGAPDH in both '00246' and '2010200' , and IgUBC and IgTUB in 'Elizabeth' . The ranking order generated by this method was slightly different from that of GeNorm.
BestKeeper analysis. BestKeeper evaluates the stability of reference genes based on the standard deviation (SD) and coefficients of variation (CV) of Cq values, with relatively low SD and CV representing relatively high stability 12 . The results of BestKeeper analysis are listed in Table 3. IgGAPDH and IgUBC were recommended as the most stable genes in '00246' , 'Elizabeth' and across all the samples, which was similar to the results from the GeNorm and NormFinder analysis. In '2010200' samples, IgTUB and IgACT6 were detected as the most stable genes via BestKeeper analysis, whereas IgACT6 was ranked fourth by GeNorm and sixth by NormFinder.
RefFinder analysis. RefFinder (http:// www. leonx ie. com/ refer enceg ene. php) was used to obtain the comprehensive rankings of reference genes by integrating three common analysis programs: GeNorm, NormFinder and BestKeeper 23 . The final comprehensive rankings of the three algorithms were integrated by RefFinder and the results are shown in Table 4. Across all the samples, the ranking order was (from the most stable to the least stable) as follows: This order is similar to the results of the GeNorm and NormFinder analysis. IgGAPDH and IgUBC were ranked as the most stable genes in '00246' and 'Elizabeth' , and IgTUB and IgUBC were the most stable genes in '2010200' . On the other hand, IgUBQ was the most unstable gene in all the samples, '00246' and '2010200' , and IgEF1β was the least stable gene in 'Elizabeth' . In the all samples, IgUBC and IgGAPDH were purported to be the most stable reference genes, while IgUBQ showed the least stability.

Validation of the selected reference genes.
To validate the reliability of the reference genes, the relative expression patterns of IgFT were examined using different combinations of reference genes in the three cultivars. The two most stable reference genes (IgGAPDH and IgUBC for '00246' and 'Elizabeth' , IgTUB and IgUBC for '2010200') and the least stable reference genes (IgUBQ for '00246' and '2010200' , IgEF1β for 'Elizabeth') selected from the analyses described above were used either alone or in combination for RT-qPCR analyses. As shown in Fig. 4, although the overall relative expression patterns of IgFT showed similar trends, differences were found when the data were normalized to those of the different reference genes. When the least stable gene IgUBQ was used as the reference gene, the normalized expression levels of IgFT in '00246' and '2010200' significantly decreased compared with those normalized using IgGAPDH or IgUBC alone, the combination of IgGAPDH + IgUBC (for '00246'), IgTUB or IgUBC alone or the combination of IgTUB + IgUBC (for '2010200')  www.nature.com/scientificreports/ (Fig. 4A, B). However, when the least stable gene IgEF1β was used for normalization, the expression level of   www.nature.com/scientificreports/ IgFT dramatically increased compared to that of IgGAPDH, IgUBC, or the combination of IgGAPDH + IgUBC in 'Elizabeth' (Fig. 4C). The combination of IgGAPDH/IgUBC was recommended as the optimum pair of reference genes for '00246' and 'Elizabeth' , and IgTUB/IgUBC was the best suited pair of reference genes for accurate normalization in '2010200' . IgUBC was the most suitable reference gene for three different I. germanica cultivars.

Discussion
Gene expression analyses are extremely important for revealing the molecular mechanisms that regulate important plant traits 24,25 . RT-qPCR has become the most powerful technique for quantification studies at the mRNA transcript level 13 . Selecting the appropriate reference genes is a necessary prerequisite for reliable RT-qPCRbased analysis. Ideally, an accurate reference gene should display stable expression in different tissues, in different organs, at different developmental stages and across different treatments 26 . In this study, we performed a systematic evaluation of 8 reference genes at different flowering stages and different genotypes of I. germanica cultivars. This study is the first attempt to identify the reference genes suitable for RT-qPCR normalization in flowering stages of I. germanica. The expression stability of various reference genes differed among the cultivars. Similar results were reported in tree peony, Panax ginseng and strawberry 16,27,28 . Different genetic backgrounds and biological processes between cultivars may have important effects on the expression stability of reference genes. Indeed, the selection and validation of reliable reference genes for quantitative analysis of gene expression analysis were necessary for the different cultivars. Three programs GeNorm, NormFinder and BestKeeper, which are based on different algorithms and analytical procedures, are widely used to select the most reliable reference genes by researchers 29,30 . In this study, we found discrepancies in the reference gene stability rankings and validation data generated by the three different algorithms above. The rankings created by GeNorm and NormFinder were similar, but they showed quite distinct differences from the ranking obtained by BestKeeper. For instance, IgACT6 was ranked among the top two stable genes by BestKeeper in '2010200' but was ranked in the middle or bottom portion by GeNorm and NormFinder. Moreover, across all the samples, IgACT6 was ranked among the top four stable genes by BestKeeper, whereas it was ranked seventh by GeNorm and NormFinder. These results are similar to those of many previous studies, possibly due to the different principles among the algorithms 13,31 . RefFinder, a comprehensive statistical program that integrates data from GeNorm, NormFinder, and BestKeeper, is used to evaluate the overall stability of reference gene expression 32 Based on the comprehensive analysis by Ref-finder, IgGAPDH and IgUBC for '00246' and 'Elizabeth' and IgTUB and IgUBC for '2010200' were identified as the most stable reference genes for RT-qPCR of target gene expression studies. These results suggest that all 8 reference genes exhibited differential stability among the three cultivars.
In this study, we evaluated 8 genes that have been widely used as candidate reference genes in many species. The results indicate that it is better to select different reference genes according to different biological samples. Based on the results of our study involving different flower developmental stages, IgUBC, IgGAPDH and IgTUB were good candidates for normalization in all of the samples. Similar studies have also been conducted in other species, such as Rhododendrons 33 , Chinese cabbage 34 , Chrysanthemum lavandulifolium 35 and Silybum marianum 36 . Moreover, UBQ was determined to be one of the most stable reference genes under NaCl and Pb stress in Iris. lactea var. chinensis 37 , but this gene was ranked as the least stable reference genes in both '00246' and '2010200' in our study, which was similar to the findings in flower buds of Iris bulleyana 38 and Rhizophora apiculate 39 . EF1α and EF1β were determined to be the best suitable reference genes for all samples of various tissues in soybean 40 . In our study, these two genes ranked very low in all the samples, and IgEF1β exhibited the most unstable expression values in 'Elizabeth' , which was similar to results in Moso bamboo 41 . ACT6 and PGK were determined to be the most stable reference genes for proper normalization in flower buds of Iris bulleyana 38 and Chrysanthemum across ploidy levels 42 and meiosis and somatic tissues of wheat 43 , while these two genes were found to be not well suited in our study, similar to reports in I. lactea var. chinensis 37 . TUB, a member of the Tubulin gene family, has also been widely used as a reliable reference gene in Primula forbesii 44 and peach 45 . Similarly, in our study, IgTUB was determined as the most stable reference gene in the flowers of '2010200' . However, this gene ranked very low under all the tested conditions in Iris bulleyana 38 . GAPDH has been reported as the most stable reference gene under PEG and cold stress in I. lactea var. chinensis 37 , but it showed unstable expression under various environmental conditions in garlic plants 46 and under PEG and NaCl treatments in Glehnia littoralis 47 . In our study, IgGAPDH was the most stable reference gene across all flower developmental stages in '00246' . In addition, IgUBC was ranked first in 'Elizabeth' and was also the most stable reference gene in all the samples, which was similar to the findings in all samples of I. lactea var. chinensis 48 . These results highlight the fact that the choice of reference genes for normalization should be specific. Even though the samples belong to the same type and are from the same species (but belong to different lines), they may have different sets of reference genes. Therefore, it is necessary to select and verify reliable reference genes for quantitative gene expression analysis, whether for different species or for different cultivars.
To illustrate the reliability of the reference genes, the expression levels of the IgFT gene were normalized by using the most stable or least stable reference genes. The results showed that the relative expression level of IgFT exhibited a clear pattern in all three cultivars when the stable reference genes IgGAPDH, IgUBC, and IgTUB or a combination of them were used (Fig. 4). The relative transcript abundance presented conflicting results when the least stable genes, IgUBQ or IgEF1β were used. Therefore, the selection of suitable internal control genes is critically important for normalization of target gene expression data by RT-qPCR.
In summary, the current study provides the first comprehensive analysis of the selection of stable reference genes as internal controls for RT-qPCR-based analysis of target gene expression in different flowering stages and different genotypes of I. germanica cultivars. IgGAPDH combined with IgUBC was recommended as the optimal reference gene in '00246' and 'Elizabeth' , while IgUBC/IgTUB was identified as the best combination for '2010200' . www.nature.com/scientificreports/ This research is the first report on the validation of candidate reference genes across flower developmental stages of three different I. germanica cultivars, which will provide basic data for research on the molecular mechanism underlying flower development in this species, and lays a foundation for similar studies in other related species.  Quantitative real-time PCR (RT-qPCR). qPCR was performed using a Mastercycler ep realplex 2S device (Eppendorf, Germany) in conjunction with SYBR Premix Ex Taq II (TaKaRa, Dalian, China). Reactions were performed in a total volume of 20 μL containing 5 μL of diluted cDNA, 0.6 μL of each of forward and reverse primer (10 μM), 10 μL of 2 × SYBR Premix and 3.8 μL of ddH 2 O. The amplification program comprised an initial denaturation step (95℃ for 2 min), followed by 40 cycles of 95℃ for 5 s, 60℃ for 30 s, and 72 ℃ for 30 s, and a melting curve protocol (60-95 ℃ with a temperature increment of 0.5 ℃ s −1 ). Each RT-qPCR was performed for three biological and three technical replicates, and negative controls were included for each primer pair. Amplification efficiency (E) and correlation coefficient (R 2 ) values were obtained from standard curves generated using a tenfold diluted cDNA series, the starting quantity of cDNA was 500 ng/μL 50 .

Materials and methods
Data analysis. The stability of the eight candidate reference genes was assessed using GeNorm 10 , Nor-mFinder software 11  Validation of selected reference genes. To validate the influence of the choice of different reference genes on the final normalized outcome, the relative expression levels of IgFT which plays an important role in promoting flowering 51 in three cultivars were analyzed using individual stably expressed or unstably expressed genes or a combination of stable reference genes, as determined by GeNorm 48 . The primers used for IgFT are presented in Table 1. The fold change of gene expression was calculated using the 2 −∆∆Ct method 24 .