Reference gene selection in multiple Passi � ora edulis tissues and under cold stress for RT-qPCR analysis


 Background: Passiflora edulis is a tropical fruit with high edible and medicinal values that is widely cultivated in southern China. The development of new P. edulis variety with cold resistance and research on the mechanism of cold resistance are the basis for further promoting the cultivation of this species. In a previous study, our laboratory discovered a cold-resistant variety Pingtang 1, which is used as a material for the study of cold resistance mechanism in P. edulis. However, functional genes of P. edulis have not been well studied, especially via relative quantitative analysis. In addition, research on reference genes in P. edulis has not been reported.Results: We used three tools to test the expression stability of 10 candidate reference genes in multiple tissue samples and cold-treated samples of P. edulis. We found that Ts and EF1 showed the highest expression stability in all samples. Further analysis showed that HIS and EF1 were stably expressed in tissue samples, whereas UBQ and EF1 were stably expressed in cold-treated samples. Interestingly, EF1 was stably expressed in not only multiple tissue samples but also cold-treated samples. To validate the selected reference genes, we used the target gene ICE1 and investigated its expression level in cold-treated leaves. Interestingly, the expression of ICE1 increased with increasing cold treatment times.Conclusions: In this study, we successfully selected reference genes from among 10 candidates for P. edulis RT-qPCR data normalization. We selected Ts and EF1 as reference genes for normalization in all samples. HIS and EF1 were ideal for data normalization in tissue samples, whereas UBQ and EF1 were ideal for cold-treated samples. Our work substantially benefits the study of functional genes in P. edulis.


Abstract
Background: Passi ora edulis is a tropical fruit with high edible and medicinal values that is widely cultivated in southern China. The development of new P. edulis variety with cold resistance and research on the mechanism of cold resistance are the basis for further promoting the cultivation of this species. In a previous study, our laboratory discovered a cold-resistant variety Pingtang 1, which is used as a material for the study of cold resistance mechanism in P. edulis. However, functional genes of P. edulis have not been well studied, especially via relative quantitative analysis. In addition, research on reference genes in P. edulis has not been reported.
Results: We used three tools to test the expression stability of 10 candidate reference genes in multiple tissue samples and cold-treated samples of P. edulis. We found that Ts and EF1 showed the highest expression stability in all samples. Further analysis showed that HIS and EF1 were stably expressed in tissue samples, whereas UBQ and EF1 were stably expressed in cold-treated samples. Interestingly, EF1 was stably expressed in not only multiple tissue samples but also cold-treated samples. To validate the selected reference genes, we used the target gene ICE1 and investigated its expression level in coldtreated leaves. Interestingly, the expression of ICE1 increased with increasing cold treatment times.
Conclusions: In this study, we successfully selected reference genes from among 10 candidates for P. edulis RT-qPCR data normalization. We selected Ts and EF1 as reference genes for normalization in all samples. HIS and EF1 were ideal for data normalization in tissue samples, whereas UBQ and EF1 were ideal for cold-treated samples. Our work substantially bene ts the study of functional genes in P. edulis.

Background
Passi ora edulis (P. edulis), or passion fruit, is a perennial vine that belongs to the Passi oraceae family. P. edulis, which origined in South America, is now widely cultivated in southern China because of its high edible value, medicinal e cacy and ornamental properties; additionally, P. edulis is an important raw material in the juice and pulp industry [1][2][3][4][5][6]. Passion fruit, a tropical fruit, provides a variety of nutrients, such as carbohydrates, vitamins and minerals, which are essential for biological activities [5]. P. edulis was introduced in China at the beginning of the last century [2]. In recent years, researchers have studied the adaptability of P. edulis to cold environments, providing a theoretical basis for expanding the cultivation area of P. edulis in China [1,2]. Although the edible value and medicinal e cacy of P. edulis have increased, only a few studies on the functional genes of P. edulis have been reported, especially on gene expression patterns under cold conditions. Previously, our laboratory had discovered a cold resistant variety Pingtang 1, which has been used as a material for research on cold resistance mechanisms in P. edulis. In the future, further research will be focused on the cold resistance mechanism and functional genes in P. edulis, that are associated with avor, color, nutritional elements, medicinal compounds and environmental adaptability. However, before such studies can be performed, suitable internal control genes for reverse transcription quantitative polymerase chain reaction (RT-qPCR) data normalization in P. edulis must be selected.
RT-qPCR, as a technology that has high throughput, high accuracy, and high sensitivity, is widely applied in functional gene expression pattern studies [7,8]. Nonetheless, using RT-qPCR to investigate gene expression levels requires stably expressed reference genes for data normalization, and the expression stability of reference genes directly affects the accuracy of relative quantitative analyses [7,[9][10][11][12]. In theory, reference genes should have stable expression patterns in different samples and under different experimental conditions, yet no reference gene has satis ed these conditions simultaneously [8,11,[13][14][15][16]. To our knowledge, a candidate reference gene may exhibit distinct expression patterns in different tissues within a species, and it may also exhibit various expression patterns under different conditions even in the same tissue [15,17,18]. For example, Actin 6 shows distinct expression levels under different types of abiotic stress in Pinus massoniana Lamb [18]. In Liriodendron chinense, Ubiquitin 1 exhibits various expression levels in multiple tissues [19]. Therefore, comprehensive evaluation and selection of appropriate internal control genes is a prerequisite for obtaining accurate and reliable relative quantitative results.
To date, many tools have been used to evaluate the expression stability of candidate reference genes. The most commonly used tools are geNorm, NormFinder and BestKeeper [20][21][22]. These three tools have been successfully used for the selection of reference genes in amaranth (Amaranthus tricolor), Setaria viridis, Carex rigescens, Lolium multi orum, Euscaphis konishii Hayata, P. massoniana, Alopecurus aequalis Sobol, Carica papaya L., Olea europaea L., Cucurbita pepo, Brassica napus, L. chinense and Populus [18,19,[23][24][25][26][27][28][29][30][31][32][33]. The use of multiple tools to select internal control genes for RT-qPCR data normalization has become a basic and indispensable process in the study of the expression patterns of functional genes. However, the studies on reference gene selection in P. edulis have not yet been performed, which has hindered the study of functional gene expression patterns in P. edulis.
In this work, we used 3 tools, namely, BestKeeper, geNorm, and NormFinder, to comprehensively evaluate the expression stability of 50S (50S ribosomal protein), Ts (threonine synthase), UBQ (ubiquitin), OTU (OTU domain-containing protein), 18S (18S ribosomal RNA), OmpH (Outer membrane protein), EF1 (elongation factor 1 alpha), eIF5A (eukaryotic translation initiation factor 5A), HIS (histone H3), and Liom (Theobroma cacao L-isoaspartate O-methyltransferase mRNA) and then to select genes with high expression stability for application in RT-qPCR data normalization in P. edulis. To further understand the scope of the application of reference genes in P. edulis, we classi ed 60 samples into two subsets, namely, tissue samples and cold-treated samples. Moreover, we used ICE1 (inducer of C-repeat-binding factor (CBF) expression) as a target gene to validate the internal control genes that we selected. We investigated the expression level of ICE1 in leaves that were treated with cold stress for different lengths of time. This work will provide reliable internal control genes for the study of functional gene expression levels in P. edulis, especially under cold stress.

Results
Ampli cation E ciency Calculation and Primer Speci city Veri cation Using LinRegPCR 2014.x software, we calculated the mean ampli cation e ciency of all primers based on row data obtained by RT-qPCR, and the mean ampli cation e ciencies of the primers ranged from 1.762 (HIS) to 1.963 (18S) ( Table 1) [34]. The primer speci city was validated by melting curve analysis. As shown in Fig. 1, each melting curve had only a single peak. This result indicated that the ampli cation was speci c, and the RT-qPCR data could be used for further analysis.  (Fig. 2). The results indicated that 18S had the highest expression level among the samples and that OmpH had the lowest expression level. Moreover, 18S showed the least variable expression levels among the 60 samples (maximum Ct value -minimum Ct value = 1.94 Ct), while OmpH showed the most variable expression levels (maximum Ct value -minimum Ct value = 12.07 Ct) (Fig. 2). Apparently, 18S can be used as a reference gene, but this needs further analysis. In studies on rice (Oryza sativa) and Dimocarpus longan Lour, 18S also showed small variable expression levels, but it cannot be used as reference gene [35,36].

Results of the geNorm Analysis
The M value was used to evaluate the expression stability of genes by geNorm software. Genes with relatively small M values have high expression stabilities. An M value greater than 1.5 indicates an unstable gene [20]. In all samples, Ts and EF1 had the most stable expression patterns (Fig. 3a). Among tissue samples, the best choices for normalization were Ts and HIS (Fig. 3b). For cold-treated samples, UBQ and HIS showed the most stable expression patterns (Fig. 3c). OmpH and 18S showed the least stable expression patterns under these conditions ( Fig. 3a, b, c).
The pairwise variation V n/n+1 was calculated to determine the minimum number of reference genes for accurate normalization [20]. If V n/n+1 values exceed 0.15, an additional internal control gene must be included for normalization [20]. We found that V 2/3 was less than 0.15 in all samples, tissue samples, and cold stress-treated samples (Fig. 4). This suggested that the minimum number of reference genes for accurate normalization was two.

Results of the NormFinder Analysis
The stability values were calculated by NormFinder to evaluate gene expression stability. Low stability values indicate high expression stability [21]. We found that Ts had the most stable expression level in all samples (Fig. 5a). In tissue samples, HIS had the highest expression stability, followed by EF1 (Fig. 5b). Among cold-treated samples, UBQ showed the most stable expression pattern, while EF1 had the second most stable pattern (Fig. 5c). Notably, Ts, HIS, and UBQ were also recommended as internal control genes by geNorm software in all samples, tissue samples, and cold-treated samples, respectively (Fig. 3a, b, c).
However, there were some differences between the geNorm analysis and NormFinder analysis. For example, in all samples, EF1 ranked rst in geNorm but fourth in NormFinder ( Fig. 3a and Fig. 5a). In tissue samples and cold-treated samples, EF1 ranked second in NormFinder and ranked fourth and fth in geNorm, respectively (Fig. 3b, c and Fig. 5b, c).

Results of the BestKeeper Analysis
In the BestKeeper algorithm, genes with a standard deviation (SD) value more than 1 are regarded as unacceptable [22]. In addition, r 2 (r represents the Pearson correlation coe cient) values close to 1 indicate a gene with an increasingly stable expression pattern [22]. In all samples, 50S, HIS, Liom and OmpH were unsuitable for data normalization (SD value > 1), while Ts exhibited the most stable expression level followed by EF1 (Table 2). Similarly, 50S, Liom and OmpH were unsuitable for data normalization in tissue samples, while EF1 and HIS showed the highest expression stability in this case ( Table 2). For the cold-treated samples, OmpH was unsuitable for normalization (SD value > 1), while UBQ showed the highest expression stability, followed by Ts (Table 2).
We found that Ts was suggested as a reference gene by BestKeeper, NormFinder and geNorm in all samples (Fig. 3a, Fig. 5a, Table 2), and UBQ had the highest expression stability calculated by the three tools in cold-treated samples (Fig. 3c, Fig. 5c, Table 2). Among tissue samples, the expression stability of HIS ranked rst in geNorm and NormFinder, and second in BestKeeper (Fig. 3b, Fig. 5b, Table 2). Notably, OmpH showed unstable expression levels in all samples, tissue samples and cold-treated samples (Fig. 3, Fig. 5, Table 2). However, the rankings of these 10 genes were slightly different among the three algorithms. For example, in tissue samples, the Ts gene ranked rst in geNorm, third in NormFinder and fth in BestKeeper (Fig. 3b, Fig. 5b, Table 2). Among cold-treated samples, the EF1 gene ranked fth in geNorm, second in NormFinder and third in BestKeeper (Fig. 3c, Fig. 5c, Table 2). EF1 also ranked differently in the analysis results of the three tools among all samples (Fig. 3a, Fig. 5a, Table 2). The reason for this phenomenon is that these tools use different algorithms to evaluate the expression stability of genes.

Reference Gene Determination
Through the analyses of the three tools, two reference genes were found to be su cient for accurate normalization in all samples, tissue samples and cold-treated samples (Fig. 4). Our results suggested that Ts and EF1 was the best combination for accurate normalization in all samples (Table 3). Moreover, HIS and EF1 were selected as reference genes in tissue samples (Table 3). Under cold stress, UBQ and EF1 were the best choices for accurate normalization (Table 3). Notably, among all samples, tissue samples and cold-treated samples, OmpH was always the least stably expressed gene, indicating that OmpH was unsuitable for RT-qPCR data normalization under these conditions (Fig. 3, Fig. 5, Table 2). Although the 18S gene possessed least variable expression levels among the 60 samples (Fig. 2), geNorm and NormFinder analysis showed that 18S had an unstable expression pattern ( Fig. 3 and Fig. 5). We suggest that 18S should not be used as a reference gene.

Veri cation of Reference Genes
To verify the reference genes Ts and EF1, we used ICE1 as a target gene and Ts and EF1 as internal control genes. ICE1 is a transcription factor that can regulate the expression of the downstream CBF gene, thereby improving plant resistance to cold stress [37]. In the leaves of P. edulis, when we used Ts and EF1 as internal control genes, we found that the expression level of ICE1 improved with increasing cold treatment times (Fig. 6). This phenomenon was consistent with the ndings of a previous study [37].
This suggests that ICE1 may increase the stability of P. edulis to cold stress.

Discussion
Reference genes are used for RT-qPCR data normalization. The expression stability of reference genes can directly in uence the accuracy of RT-qPCR, and reference genes should sustain stable expression levels in various tissues and under distinct experimental conditions [7,38]. To our knowledge, the expression levels of genes vary with changes in tissue, species and experimental conditions [11]. Therefore, a gene must be strictly screened before being used as a reference gene to ensure the accuracy of RT-qPCR analysis. In addition, functional genes of P. edulis have not been well studied, especially by relative quantitative analysis of functional genes, and research on reference genes in P. edulis has not been reported. [1,2]. To enhance the reliability of functional gene expression research, we systematically selected reliable internal control genes for RT-qPCR data normalization in multiple tissue samples and cold-treated samples of P. edulis.
In this work, we selected 10 candidate reference genes, namely, 50S, Ts, UBQ, OUT, OmpH, EF1, 18S, eIF5A, HIS, and Liom, and we used RT-qPCR to obtain their Ct values in 60 samples. Through analysis based on Ct values, we found that 18S had the highest expression level (5.99 Ct) and the smallest expression variation (1.94 Ct), while OmpH had the lowest expression level (26.75 Ct) and the largest expression variation (12.07 Ct) (Fig. 2). However, we could not use 18S a reference gene due to its low expression variation. According to previous studies, genes with low expression variation may not be suitable for data normalization, and reference genes must be evaluated comprehensively by using different tools [39,40]. For example, research conducted on rice and D. longan showed that 18S cannot be used as an internal control gene [35,36]. In particular, in research conducted on Bixa orellana, although 18S had the highest expression level and the smallest expression variation, further analysis results showed that 18S was the most unstable gene in multiple tissues and at different seed development stages [8]. This result indicated that Ct values of candidate reference genes were not enough to determine the expression stability.
Through further analysis by geNorm, NormFinder, and BestKeeper, Ts and EF1 were found to be ideal for data normalization in all samples ( Table 3). The protein encoded by Ts can catalyze the last step of threonine biosynthesis, and there are many reports about its structure, expression and function [41,42]. However, there are few reports about Ts as an internal control gene. EF1 is stably expressed in not only P. edulis, but also other plants. For example, EF1 was used as an internal control gene in Ilex paraguariensis leaves and potato (Solanum tuberosum L.) under abiotic stress [43,44]. To our knowledge, the expression level of reference genes may change with changes in species, experimental conditions, and tissue types.
For example, although EF1 was suitable for data normalization in P. edulis, potato under abiotic stress, drought-treated leaves of I. paraguariensis, and Betula luminifera, this gene was unsuitable for data normalization in Camellia sinensis under metal stress and in leaves of Camellia sinensis [38,[43][44][45][46]. Overall, Ts and EF1 have the highest expression stability, but if we classify all samples into different subsets, there may be other genes in each subset with higher expression stability than Ts and EF1.
To determine whether there were other genes that had higher expression stability than Ts and EF1 in different sample subsets, we classi ed the 60 samples into two subsets, namely, tissue samples and cold-treated samples. In previous studies on reference gene selection in different species, we found that the expression patterns of candidate reference genes may differ between nonstress and abiotic stress conditions [23,47]. Under nonstress conditions, HIS and EF1 was the best combination for data normalization in tissue samples (Table 3). In L. multi orum, HIS also showed a stable expression pattern under acidic aluminum stress and heavy metal stress, while it could not be selected as an internal control gene in Populus tomentosa due to its poor expression stability [26,48]. Under cold stress conditions, UBQ and EF1 were su cient for data normalization in P. edulis (Table 3). UBQ was also used as a reference gene in cotyledons of Cunninghamia lanceolata [11]. However, this gene was unsuitable for data normalization in Bursaphelenchus mucronatus and lettuce (Lactuca sativa) [49,50]. Compared with Ts, HIS had higher expression stability in tissue samples and UBQ had higher expression stability in coldtreated samples. This nding indicates that HIS and UBQ can be used as internal control genes in only speci c samples and experimental conditions. Moreover, 18S and Liom were unstably expressed in this study, and 18S also showed unstable expression patterns in amaranth, B. orellana and Stellera chamaejasme (Fig. 3, Fig. 5, Table 2) [8,23,51]. However, 18S could be used for data normalization in Bursaphelenchus mucronatus and C. sinensis under metal stress [45,49].
As a nutritious tropical fruit, P. edulis is widely cultivated in southern China [2]. However, cold resistance has been an important factor limiting the expansion of the P. edulis cultivation area. Our laboratory discovered a cold-resistant variety, Pingtang 1, which is used as sample material for the study of cold resistance mechanisms. Our previous research has shown that the ICE1-CBF-COR pathway plays a crucial role in the cold tolerance of P. edulis [2]. ICE1 is a transcription factor that can regulate the expression of the downstream CBF gene, thereby improving plant resistance to cold stress [37]. To validate the reference genes that we selected, we used ICE1 as a target gene and Ts and EF1 as internal control genes, to investigate the expression level of ICE1 in cold-treated leaves of P. edulis. We found that the expression level of ICE1 increased with increasing cold-treatment times (Fig. 6). In previous studies, increased expression of ICE1 enhanced plant resistance to cold stress [37,52]. Therefore, the increase in ICE1 expression may enhance the adaptability of P. edulis to cold stress.
Our work will facilitate the study of functional gene expression levels in P. edulis. We selected stable and reliable internal control genes and applied these genes in RT-qPCR data normalization in different tissue samples and under cold stress conditions in P. edulis.

Conclusions
In this study, we successfully selected reference genes from among 10 candidates for P. edulis RT-qPCR data normalization. In all samples, Ts and EF1 were the best reference genes for data normalization; HIS and EF1 were ideal for tissue samples, and UBQ and EF1 were the best internal control genes for coldtreated samples. In conclusion, this work provides reliable and effective reference genes for gene expression level analysis in P. edulis.

Plant Materials and Stress Treatment
Plant materials were sampled from adult P.  [20]. At the same time, the pairwise variation (V n/n+1 ) value given by geNorm is useful for determining the minimum number of internal control genes [20]. By combining intergroup and intragroup variations in gene expression levels, NormFinder evaluates genes based on stability values [21]. A low stability value indicates high expression stability. In the BestKeeper algorithm, the coe cient of determination (r 2 , r represents the Pearson correlation coe cient) and the SD are used to select reference genes [22].

Veri cation of Reference Genes
To test the reference genes Ts and EF1, we chose ICE1 as a target gene and investigated the expression pattern of the ICE1 gene in the leaves of P. edulis under cold stress. The leaves were collected from P. edulis treated with cold stress (4 ℃) for 0 h (CK), 1 h, 4 h, 8 h, 24 h, and 72 h. The primers for ICE for RT-qPCR were 5´-AAGCCACAGACAACAAGGCA-3´ (forward) and 5´-AGCACTGCGGACATTCTTCA-3( reverse). All further steps are described above.    Pairwise variation (V) of ten candidate reference genes calculated by geNorm. If Vn/n+1 values exceed 0.15, an additional internal control gene must be included for normalization. In total samples, tissue samples, and cold-treated samples, V2/3 values were less than 0.15, indicating that two reference genes were su cient for RT-qPCR data normalization.