Reference Gene Selection for qRT-PCR Analysis in the Sweetpotato Whitefly, Bemisia tabaci (Hemiptera: Aleyrodidae)

Background Accurate evaluation of gene expression requires normalization relative to the expression of reliable reference genes. Expression levels of “classical” reference genes can differ, however, across experimental conditions. Although quantitative real-time PCR (qRT-PCR) has been used extensively to decipher gene function in the sweetpotato whitefly Bemisia tabaci, a world-wide pest in many agricultural systems, the stability of its reference genes has rarely been validated. Results In this study, 15 candidate reference genes from B. tabaci were evaluated using two Excel-based algorithms geNorm and Normfinder under a diverse set of biotic and abiotic conditions. At least two reference genes were selected to normalize gene expressions in B. tabaci under experimental conditions. Specifically, for biotic conditions including host plant, acquisition of a plant virus, developmental stage, tissue (body region of the adult), and whitefly biotype, ribosomal protein L29 was the most stable reference gene. In contrast, the expression of elongation factor 1 alpha, peptidylprolyl isomerase A, NADH dehydrogenase, succinate dehydrogenase complex subunit A and heat shock protein 40 were consistently stable across various abiotic conditions including photoperiod, temperature, and insecticide susceptibility. Conclusion Our finding is the first step toward establishing a standardized quantitative real-time PCR procedure following the MIQE (Minimum Information for publication of Quantitative real time PCR Experiments) guideline in an agriculturally important insect pest, and provides a solid foundation for future RNA interference based functional study in B. tabaci.


Introduction
In recent years, quantitative real-time PCR (qRT-PCR) has been widely utilized for gene expression analysis because of its sensitivity, accuracy, reproducibility, and most importantly, quantitativeness [1][2][3][4]. There is no argument that qRT-PCR is a powerful tool for gene expression analysis, data analysis, and subsequent interpretation. However, interpretation can be challenging due to variation caused by pipetting error and different extraction techniques, transcription and amplification efficiencies among different samples [2,[5][6][7]. Therefore, controlling for internal differences and reducing errors between samples requires the use of reliable reference genes for normalization in gene expression analysis [8].
Traditionally, housekeeping genes including 18S ribosomal RNA, glyceraldehyde-3-phosphate dehydrogenase, elongation factor, ubiquitin-conjugating enzyme, alpha microtubules protein, and beta microtubule protein have been used extensively as endogenous controls for the normalization of qRT-PCR data, but the expression levels of these reference genes can differ under environmental conditions [9]. Based on previous studies, it is evident that the existence of a single universal reference gene suited for all experimental conditions is highly unlikely [5,[10][11][12]. Therefore, selection of reliable reference genes that are consistently expressed under specific experimental conditions is critical for the interpretation of qRT-PCR results. Currently, two Excelbased software tools including geNorm (http://medgen.ugent.be/ ,jvdesomp/genorm/) [10] and Normfinder (http://www.mdl.dk/ publications normfinder.htm) [13], are widely used for evaluating the performance of reference genes. The geNorm program was used to calculate the mean pair-wise variation between an individual gene and all other tested candidate reference genes and the results were shown as expression stability (M). Normfinder is an algorithm for estimation of reference genes among a set of candidates. It ranks the candidate genes based on their expression stability.
The sweetpotato whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) is one of the most destructive insect pests worldwide [14][15]. This whitefly damages many crops by direct feeding and by vectoring 114 plant viruses [16]. Bemisia tabaci has long been thought to comprise morphologically indistinguishable biotypes that often differ in host range, fecundity, insecticide resistance, transmission competency for begomoviruses, and the symbionts they harbor [14,16,17]. Recent studies suggest that most of these biotypes represent genetically distinct cryptic species [18][19], among which the B biotype of the Middle East-Minor Asia 1 and the Q biotype of the Mediterranean group are the most invasive and destructive [20]. Although B. tabaci was first recorded in China in the late 1940s, crop damage caused by this insect did not become serious until the introduction of the B biotype in the 1990s [21]. The Q biotype of B. tabaci was first detected in Yunnan Province, China in 2003 [22]. Since then, the Q biotype has gradually displaced the established B populations and has become the dominant B. tabaci in most of China [23].
To examine the temporal and spatial changes of gene expression in B. tabaci, b-actin and a-tubulin are the most frequently used endogenous reference genes in qRT-PCR analyses [24][25][26][27]. These genes were selected without the companion validation study to evaluate their suitability under specific experimental conditions. Previous studies have demonstrated that the expression of b-actin can be significantly influenced by tissue type [12]. Bustin and his colleagues proposed a MIQE guideline (Minimum Information for publication of Quantitative real time PCR Experiments) [28] to standardize qRT-PCR analysis; reference gene selection is an integral part of their recommendations. In this study, 15 housekeeping genes from a parallel B. tabaci transcriptome study [29] were selected as candidate reference genes. The overall goal of this research is to develop a standardized qRT-PCR analysis in B. tabaci following the MIQE guideline. Specifically, we evaluate the stability and performance of the above mentioned candidate reference genes under different experimental conditions including five biotic factors (host, acquisition of a plant virus, developmental stage, tissue, and whitefly biotype) and three abiotic factors (photoperiod, temperature, and insecticide exposure). The choice and number of reference genes needed under various conditions are investigated and recommended.

Ethics Statement
Bemisia tabaci B biotype strains used in this study were initially collected in the field at Beijing in 2000, and have been maintained in a greenhouse at the Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences. The species in the genus Aleyrodidaeare common agricultural pests and are not included in the ''List of Protected Animals in China''. No specific permits were required for the described field studies.

Biotic Conditions
Host plant. Bemisia tabaci B biotype was maintained on three different host plants including cabbage, tomato, and cucumber [30]. A total of 180 3-day-old adults were collected, snap frozen in liquid nitrogen, and stored at 280uC before qRT-PCR analysis.
Acquisition of a plant virus. Tomato plants infected with Tomato yellow leaf curl virus (TYLCV) were obtained by Agrobacterium tumefaciens-mediated inoculation using a cloned TYLCV genome (GenBank accession ID: AM282874) [31]. Plants were inoculated with the virus at the 3-true-leaf stage. Viral infection of tomato plants was confirmed by the development of characteristic leaf curl symptoms and was further validated by molecular analysis [32]. Viruliferous B. tabaci were obtained by caging non-viruliferous B. tabaci adults with TYLCV-infected tomato plant for a 72 h acquisition access period [33]. Non-viruliferous B. tabaci were obtained by caging nonviruliferous B. tabaci adults with healthy tomato plants for 72 h. A total of 180 3-day-old adult whiteflies from both virus-infected and virus-free tomato plants, respectively, were snap frozen and stored as described earlier.
Developmental stage. Three developmental stages (egg, pupa, and adult) were collected from B. tabaci B biotype maintained on healthy cabbage plants. A total of 900 eggs, 900 pupae, and 300 adults were snap frozen and stored as described earlier.
Tissue. A dissection needle and a stereo microscope (Leica, DFC425) were used to obtain three body regions (head, thorax, and abdomen) from 3-day-old B. tabaci adults (TH-S). These sections were dissected from adults reared on cabbage plants; snap frozen, and stored as described earlier.
Whitefly biotype. Bemisia tabaci B and Q biotype strains were collected from Beijing, China in 2000 and 2008, respectively, and have been maintained in a greenhouse at the Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences [30].

Abiotic Conditions
Photoperiod. A total of 200 3-day-old B. tabaci adults were placed into nine screen cages, respectively, and provisioned with cabbage plants at the 5 to 7-true-leaf stage. These cages were kept in growth chambers (2760.5uC, 60610% RH) with photoperiods (L/D) of 24:0, 0:24, and 14:10, respectively. After 96 h, B. tabaci adults were snap frozen and stored as described earlier.
Temperature. A total of 720 3-day-old B. tabaci adults (80 whiteflies69 replications) were collected from cabbage plant and placed individually into 30 ml specimen tubes. The tubes were then placed in climatic chambers at 4.0, 25.0, and 37.5uC, respectively. After 1 h, the live adults were snap frozen and stored as described earlier.
Insecticide susceptibility. Thiamethoxam susceptible (TH-S) and resistant (TH-R) B. tabaci strains were established from the same populations described previously [34]. Before sample collection, a leaf-dip bioassay [34] was conducted to confirm that the resistance factor [LC50 (TH-R)/LC50 (TH-S)] was over 70fold. A total of 180 adults from both TH-S and TH-R were collected, snap frozen, and stored as described earlier.

Total RNA Extraction and cDNA Synthesis
Total RNA was extracted with a Trizol reagent (Invitrogen, Carlsbad, CA, USA). RNA was quantified by measuring the absorbance at 260 nm with a Nano Vue UV/Vis spectrophotometer (GE Healthcare). The purity of RNA was assessed at an absorbance ratio of OD260/280 and OD260/230, and the integrity was checked with 1% agarose gel electrophoresis. Then, 1 mg of RNA was used to synthesize the first-strand cDNA using the PrimeScriptHRT reagent Kit (Takara Bio, Tokyo, Japan) with gDNA Eraser (Perfect Real Time, TaKara, Shiga, Japan) according to the manufacturer's protocol. The synthesized cDNA was stored at 220uC.

Quantitative Real-time PCR analysis
Quantitative Real-time PCR (qRT-PCR) was performed on an ABI 7500 real-time system. The cDNA of each sample representing one biological replicate was diluted to a working concentration of 17 ng/ml for the qRT-PCR analysis. The melt temperature was 60uC and product contained between 80 and 200 base pairs ( Table 1). The 25 ml reaction system contained 1 ml of diluted cDNA, 11.25 ml of SYBRH Green Real-time PCR Master Mix (TIANGEN, Corp, Beijing, China), and 0.5 ml of each primer. The cycling parameters were as follows: 95uC for 3 min followed by 40 cycles of 95uC for 30 s, 60uC for 30 s, and 72uC for 35 s. A 3-fold serial dilution of cDNA was used to construct a standard curve to determine the PCR efficiency that would be used to convert the quantification cycles (Ct-values) into the relative quantities (relative gene expressions).

Data Analysis
Expression of reference genes was evaluated with two webbased analysis tools: geNorm and NormFinder. geNorm was used to calculate the M stability value as the mean pairwise variation between an individual gene and all other tested candidate genes. The lower the M value, the more stable the reference genes. The value of V n /V n+1 indicates the pairwise variation between two sequential normalization factors and determines the optimal number of reference genes required for accurate normalization. A value below 0.15 indicates that an additional reference gene will not significantly improve normalization. Normfinder evaluates the overall variation of the candidate reference genes under the experimental conditions and estimates the variation between and within groups. For each candidate gene, Normfinder provides a stability value that is a direct and rapid measurement of expression variation. This stability value enables the user to estimate the systematic error introduced when selecting a suitable reference gene.

Expression profiles of candidate reference genes
For each reference gene, a dissociation curve with a single-peak ensured that the primer sets amplified a unique PCR product ranging from 81 to 187 bp. The PCR efficiency was consistently high for candidate reference genes except c-tubulin (75.3%) and   Based on data obtained from three abiotic factors, the ideal reference genes are EF1-a, PPIA, NADH, SDHA, and HSP40.
Normfinder. Normfinder indicated that RPL29, GAPDH, and NADH are the most stable reference genes for host plants, tissues, biotypes, respectively (Table 2). Specifically, for developmental stages and viruliferous/non-viruliferous B. tabaci, SDHA is the most stable reference gene. A similar trend is observed under selected abiotic factors, in which HSP20, HSP40, and EF1-a are ranked as the most stable reference genes for photoperiod, temperature, and insecticide susceptibility, respectively (Table 3).

Discussion
Because it is highly sensitive, specific, accurate, and reproducible, qRT-PCR is, in many ways, superior to conventional methods (northern hybridization and semi-quantitative PCR), and has become an essential tool for gene expression analysis [13,28,[35][36][37][38][39]. qRT-PCR analysis, however, is influenced greatly by the selection of reference genes [10,[40][41]. The endogenous reference genes should be stable across different experimental treatments; otherwise, a variable reference gene can compromise the qRT-PCR analysis by introducing artificial changes or masking true changes in target gene expression. Some commonly used reference genes can vary substantially in response to specific experimental conditions [11,[42][43]. In this study, we used two Excel-based algorithms geNorm and Normfinder to evaluate the stability of 15 candidate reference genes in B. tabaci in response to five biotic factors (host, virus, stage, tissue, and biotype) and three abiotic factors (photoperiod, temperature, and insecticide susceptibility).
A major conclusion of this study is that many of the candidate genes in B. tabaci should not be used as the default reference genes because their expression is highly variable under certain conditions. Our results indicate that the stability of reference gene expression must be validated for each experimental condition under investigation. The ranking of these reference genes differs  somewhat for geNorm and Normfinder, because these programs have different algorithms and different sensitivities toward co-regulated reference genes. Despite the discrepancies, both programs identified a similar set of reference genes suited for the respective experimental conditions. The ideal reference genes in response to biotic factors were RPL29, HSP40, and HSP90 according to geNorm and RPL29 based on Normfinder, respectively. Combing these results, RPL29 is a consensus reference gene that is reliable across a range of biotic conditions (Table 4), and this is consistent with the performance of the other ribosomal protein L32, a widely used single normaliser in gene expression studies [11,[43][44][45][46][47]. Despite subtle ranking differences between geNorm and Normfinder, the ideal reference genes in response to abiotic factors were determined to be EF1-a, PPIA, NADH, SDHA, and HSP40 (Table 4). EF-1a has rarely been used as a normaliser in the past but has recently been selected as a suitable reference gene in salmon [48], humans [47,49], Orthoptera [46,50], and Hymenoptera [43]. PPIA was considered sufficiently stable for normalization in this study, which is consistent with a previous report in human cervical tissues [47].
Another conclusion of our study is that some genes that have been consistently used for the normalization study showed high levels of variation in response to certain treatments. Previously, 18S has been considered an ideal reference gene because the expression level of rRNA appears to vary considerably less than mRNA [51]. In this study, the raw Ct values of 18S ranged from 9.92 to 15.94 depending on insect body region and host plants, suggesting that the expression of 18S can be highly variable and consequently, it could not be used as a reference gene under certain experimental conditions. This result is consistent with some earlier studies on 18S RNA [11,42]. Another commonly used reference gene, actin, encodes a major component of the protein scaffold that supports the cell and determines its shape. The expression of actin is moderately abundant in most cell types, and actin has been used extensively as a reference gene in B. tabaci and in many other insects including the desert locust [46], European honey bee [45], and two species of Collembola [52]. In our study actin was not stable among different tissues (body regions) and hosts; disqualifying actin as a suitable reference gene under these conditions.
In recent years, more researchers have adopted a multiple reference gene approach to analyze gene expression [38,53]. Our results demonstrated that the expression of several reference genes from B. tabaci were consistently stable across selected experimental conditions. However, the best-suited reference genes can be different in response to diverse biotic and abiotic factors (Table 4). Our finding is the very first step toward establishing a standardized qRT-PCR procedure following the MIQE (Minimal Information required for Publication of Quantitative Real-Time PCR) guideline in an agriculturally important insect pest. More importantly, this study provides a solid foundation for future RNAi-based functional study in B. tabaci.