Identification and Validation of Reference Genes for Transcript Normalization in Strawberry (Fragaria × ananassa) Defense Responses

Strawberry (Fragaria spp) is an emerging model for the development of basic genomics and recombinant DNA studies among rosaceous crops. Functional genomic and molecular studies involve relative quantification of gene expression under experimental conditions of interest. Accuracy and reliability are dependent upon the choice of an optimal reference control transcript. There is no information available on validated endogenous reference genes for use in studies testing strawberry-pathogen interactions. Thirteen potential pre-selected strawberry reference genes were tested against different tissues, strawberry cultivars, biotic stresses, ripening and senescent conditions, and SA/JA treatments. Evaluation of reference candidate’s suitability was analyzed by five different methodologies, and information was merged to identify best reference transcripts. A combination of all five methods was used for selective classification of reference genes. The resulting superior reference genes, FaRIB413, FaACTIN, FaEF1α and FaGAPDH2 are strongly recommended as control genes for relative quantification of gene expression in strawberry. This report constitutes the first systematic study to identify and validate optimal reference genes for accurate normalization of gene expression in strawberry plant defense response studies.


Introduction
Transcriptomic analyses are essential in understanding complex molecular processes occurring in plants. Although global evaluation techniques such as microarrays or RNAseq provide a representative snapshot of a transcriptome, these techniques can only be practically applied to a limited number of tissues, treatments or time points. The data found by global expression techniques need to be then considered carefully, typically using relative quantification of gene expression by quantitative reverse transcription (RTqPCR). This method is used as a primary source of in-depth molecular expression information for a smaller set of gene candidates due to its wide range of quantification, reproducibility, and higher precision and accuracy [1], [2], [3]. However, this approach requires knowledge of stably expressed reference genes for data normalization of target genes under specific experimental conditions. Failure to use an appropriate reference or internal control gene may result in biased gene expression profiles, as well as low reproducibility. Consequently, either only gross changes in gene expression level are declared statistically significant, or the pattern of gene expression is inaccurately characterized [4], [5].
To date, some of the best known and most frequently used reference gene transcripts for RTqPCR in plants and animals include those coding for 18S rRNA, glyceraldehyde-3-phosphate dehydrogenase, elongation factor-1a, actin, and aand b-tubulin [6], [7], [8], [9], [10], [11], [12], [13]. These genes have been recognized as stably expressed housekeeping genes, and they have been historically used as reference genes in many plants when normalizing RNA-gel blots and semi-quantitative PCR. However, numerous reports have indicated that transcript accumulation is not always consistent under some experimental conditions or across tissues. Such variation, may introduce a significant level of error in interpreting the actual expression pattern of a target gene [14], [15]. Identification of most appropriate and highly-stable internal reference genes for normalization in any given experimental plant system is a prerequisite and compulsory step to obtain reliable and reproducible results from RTqPCR. A strong reference is the foundation of accurate RTqPCR analyses following the golden rules which have been detailed recently in Udvardi et al. [16].
Over the last few years efforts have been made to identify suitable reference genes for quantification of gene expression in model plant species such as Arabidopsis [17]. Efforts have been extended to crop plants such as pea [18], banana [19], [20], sulla [21], zucchini [22], and citrus [23]. However, reference genes still have yet been identified and tested in other species of high agricultural interest including strawberry (Fragaria spp), a small fruit crop of great value throughout the world (FAOSTAT Agriculture Data [http://faostat.fao.org/, updated 7 aug 2012]).
Due to its broad horticultural importance and relatively close relationship to other valuable rosaceous crops, strawberry has been proposed as a model for functional genomics and transgenic studies within the Rosaceae [24], [25]. Strawberry's rapid cycling, fast growth and relative transformability make it an attractive system for functional evaluation of genes associated with plant traits not testable in Arabidopsis. An increasing number of molecular studies are being reported in this species. Many of these studies have performed RTqPCR analysis using traditional reference genes to describe a wide variety of molecular events occurring in strawberry. The technique has been used to query gene expression during plant development, fruit ripening, aroma production, and responses to many biotic and abiotic stresses [26], [27], [28], [29], [30]. However, the suite of strawberry reference genes has not been carefully vetted to determine their optimal suitability for comparative expression analyses across a range of conditions, tissues or treatments.
It is necessary to identify candidate genes specifically chosen for transcript normalization for the conditions under study [31], [32]. Also, when using only one reference gene, its stability cannot be properly evaluated. The use of multiple reference genes does not only produce more reliable data, it permits an internal evaluation of the stability of these reference transcripts as well.
In the present study a subset of candidate reference genes for strawberry RTqPCR normalization in plant defense studies were identified and tested. Candidates were evaluated across a range of forty-eight situations distributed over seven experimental conditions including fruit ripening stages, biotic stress after Colletotrichum acutatum infection, and treatments with plant hormones such as SA and MeJA. Also, different cultivars of strawberry (Fragaria 6 ananassa), and growth conditions were tested. Recommendations for the use of these candidate genes are provided to ensure an accurate normalization of transcript level under a given condition in strawberry gene expression studies by RTqPCR.

Selection of Candidate Reference Genes in Strawberry for Gene Expression Analysis
Candidate genes were selected for further analysis based on inhouse data and information obtained from a range of microarrays experiments ( [33], Amil-Ruiz et al., unpublished). Specific strawberry transcripts have been identified that exhibit a high degree of stability among biological replicates and in varying experimental conditions. Due to the fact that low abundance transcripts generally show high variation in their basal expression [34] they were not considered further. The analysis was performed only with candidates whose primers match prescribed conditions described below. In addition, the analysis sought to examine transcripts representing a cross-section of functional diversity to avoid a putative co-regulation effect among genes that may respond in parallel in response to a particular experimental assay. Such precautions are a prerequisite for one of the statistical procedures (the geNORM algorithm) reported to identify stably expressed genes [4].
Under all of these restrictive conditions, thirteen preselected candidate genes were identified (Table 1). These genes encode molecular components associated with a wide variety of biological functions in plant cell physiology such as 18S rRNA (gene FaRIB413), a ribosome component; GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE (genes FaGAPDH1 and Fa-GAPDH2), an essential enzyme for carbohydrate metabolism in cytoplasm; ELONGATION FACTOR-1a (gene FaEF1a), a component of the protein synthesis machinery; ACTIN (gene FaACTIN), a-TUBULIN (gene FaTUBa) and b-TUBULIN (gene FaTUBb), major components of microfilament and microtubule of the cytoskeleton, respectively; the UBIQUTIN CONJUGATING EN-ZYME E2 (gene FaUBQ1), a basic component of the ubiquitinmediated protein tagging system; chromatin remodeling protein CHC1 (gene FaCHC1), an essential part of the chromodomain remodeling complex; an S-adenosyl-L-methionine-dependent methyltransferase (gene FaMT1), an enzyme implicated in secondary metabolism; a strawberry ortholog of the Arabidopsis AtBZIP61 regulatory transcription factor (gene FaBZIP1); a mitochondrial import inner membrane translocase (gene FaTIM1); a protein with a forkhead-associated domain and unknown molecular function (gene FaFHA1). In addition, the FaWRKY1 gene, a previously reported strawberry gene known to respond to all the different biological conditions used in this study [30], was chosen as a target gene to test the validity of these strawberry candidate genes as good reference genes in RTqPCR analyses.
Primers designed of candidate reference genes. The RTqPCR primer pairs for each putative reference gene, as well as for FaWRKY1, were designed based on common criteria, and were tested to generate clear and unique PCR products in RTqPCR reactions (Table 1 and Figure S1).
All primers were designed from the CDS of the selected genes, avoiding regions of conserved sequence similarity to other genes. For genes belonging to gene families or with identified paralogs present in the genome of the diploid woodland strawberry (F. vesca) [35], the least conserved region was used to ensure amplification of a single gene by PCR. In four cases (FaEF1a, FaTUBa, FaTUBb and FaACTIN), it was not possible to differentiate between either multicopy or nearly identical genes although unique amplicons were obtained. In six cases including the control gene (FaGAPDH1, FaTUBb, FaBZIP1, FaTIM1, FaFHA1, FaWRKY1) primers were designed to span an exon-exon junction.
To ensure maximum specificity and efficiency during PCR amplification, primers were designed to have melting temperatures over 70 uC, and were required to generate short amplicons, usually between 100 and 200 bp ( Table 1). The most appropriate annealing temperature for every primer pair was calculated by RTq-gradientPCR, and only primer pairs with optimal efficiency at annealing temperatures of above 65uC were considered for subsequent RTqPCR analyses. The primer pair for gene FaRIB413 was previously designed in our group [29], and tested to meet all of the above criteria. The specificity of the primers was tested by PCR using first-strand cDNAs synthesized from total RNA isolated from the biological samples.
The PCR efficiency of each primer pair was calculated using LinRegPCR, a method that utilizes absolute fluorescence data captured during the exponential phase of amplification of each real-time PCR reaction [36]. Table 1 shows the calculated PCR efficiencies for the primer pairs studied. Each efficiency value represents an average 6 SD calculated from 192 amplification plots (i.e. two technical replicates of two biological replicates of a total of 48 different experimental conditions). For all primer pairs, values ranged from 1.712 to 1.925, with low standard deviation. These values indicated comparable amplification efficiencies among the 96 diverse cDNA samples tested (Table 1), and suggested that the designed primer pairs efficiently amplified their target genes. Therefore, the mean primer pair efficiency value was considered for all subsequent studies, including estimations of the relative expression level of the reference genes under evaluation. Candidate reference genes were evaluated by RTqPCR analyses in response to the experimental conditions summarized in Table 2. Samples from different strawberry varieties were also examined. Two independent biological replicates were performed for each experimental condition. Between 10 and 18 independent samples per experiment were analyzed. In addition, two technical replicates corresponding to two biological replicates were used in this study. The generated results were subjected to the following analytical methods: analysis of ''Stability index'' [10], geNORM [4] implemented in qBASEplus software [37], NormFinder [9], BestKeeper [38], and the comparative D-Ct [39].
Statistical analysis of gene expression by ''stability index'' calculation. Figure 1 shows the expression level of candidate reference genes in the seven experimental conditions named in Table 2. Mean Cq values for each transcript in every experimental condition, together with coefficient of variation (CV), slope, and stability index (SI), according to Brunner, (2004) [10] are given in Table 3.
The analysis of variation, as reflected in the coefficient of variation (CV), showed highly predictability of all candidate reference genes in every of the seven experimental conditions. Considering them together, almost all CV values were below 6%. Exceptions were genes FaGAPDH1 and FaGAPDH2, which deviated substantially during ripening, and genes FaTUBa, FaGAPDH1, FaBZIP1 and FaTIM1, within the ''all together'' conditions ( Table 3).
The mean expression level for each gene was regressed against the overall means for the different samples ( Figure S2). The slope of the predicted regression lines provided an estimate of the degree to which the gene is sensitive to general expression-promoting conditions. Assuming that both consistent transcript levels among samples (low slope) and high predictability (low CV) are desired, the ''stability index'' (SI) (product of slope and CV) is used to describe transcript stability as in Brunner, (2004) [10]. Transcripts with the lowest stability index will usually provide the best reference genes or controls.
The results show that several predicted candidate genes show a favorable SI in each of the main areas studied (Table 3, marked by asterisks). During fruit ripening candidates FaRIB413, FaCHC1 and FaTUBb showed low SI values (0.011, 0.018, and 0.024, respectively). Genes FaGAPDH1, FaTUBa and FaUBQ1 also appear to be excellent reference genes for fungal infection studies in red fruit (SI of 0.007, 0.028, and 0.074, respectively). In vegetative tissues challenged with the fungus, variations in number and diversity of convenient reference genes was also found. Thus, genes FaUBQ1 (SI, 0.065) and FaRIB413 (SI, 0.083), were found to be the best candidates for normalization on crown tissue of cultivar Camarosa but genes FaTUBa (SI, 0.014), FaACTIN (SI, 0.020), FaRIB413 (SI, 0.072), FaBZIP1 (SI, 0.084), FaEF1a (SI, 0.084) were also very good candidates on petiole tissue of this cultivar. However, on petiole tissue from cultivar Andana, the set of predicted good candidate reference genes is not the same. The best candidates were genes FaGAPDH1 (SI, 0.009), FaGAPDH2 (SI, 0.051), FaACTIN (SI, 0.054), FaEF1a (SI, 0.059), FaMT1 (SI, 0.060), and FaFHA1 (SI, 0.085). Only genes FaACTIN and FaEF1a were found to be the reasonable reference genes for normalization in petiole tissue of both strawberry cultivars. In addition, genes FaUBQ1, FaGAPDH2, and FaRIB413 were found to be the optimal reference genes for SA and JA studies either in in-vitro plants (SI, 0.047, 0.055, and 0.093, respectively) in cell suspension treatments (SI, 0.081, 0.022, and 0.073, respectively), as well as across different cultivars. Genes FaGAPDH1 (SI, 0.045) and FaTIM1 (SI, 0.013) were also found to be appropriate candidates for the invitro plants and cellular suspension experiments, respectively. Also, we have considered an ''all together'' analysis where all seven experimental variables have been examined. In this analysis, gene FaACTIN showed the lowest stability index (SI, 0.015), and appears to be the best overall reference gene following this analytical method.
Expression stability and calculation of hypothetical normalization factor by geNorm PLUS . The stability coeficient (M values) and the coefficient of variation (CV values) of each gene are inversely related to their expression stability. These values were calculated using qBase software [37] but taking into account the previously calculated specific PCR efficiency of each gene. The average stability coefficient (M A ), defined as the average value of the M values (average pairwise variation of a gene with all other tested reference genes of all combinations of a gene and highranking reference genes) of the relative quantities of the thirteen genes under evaluation were analyzed with geNormPlus (qBase software, [4], [37]). Figure 2 represents the average stability coefficients (M A ) of the thirteen candidate reference genes tested from every analyzed condition. All thirteen genes showed acceptable expression stabilities (M A #1), as described for heterogeneous samples [37], with the exception of genes FaBZIP1 and FaGAPDH1 when all seven experimental conditions were analyzed together. Table 4 shows transcripts ranked by their M A and CV values. The M A results revealed that optimal candidate reference genes differed among the analyzed experimental conditions. Thus, FaACTIN (0.182) seems to be the most stable gene in fruit ripening analyses, meanwhile FaTIM1 (0.594) in the ''all together'' conditions. A similar result was obtained when CV values were considered.
The optimal and the minimal number of reference genes needed to calculate a hypothetical optimal normalization factor suitable in each analyzed condition was determined, as described by Vandesompele [4]. Figure 3, shows that the optimal number (V) of these needed reference genes differed in each experimental conditions but a combination of them is assumed to be an ideal reference gene. Thus, in fruit ripening analyses, V 5/6 was the lowest pairwise variation value (0.041). Therefore, the hypothetical normalization factor in these experimental conditions would be the geometric mean of the five or six more stable genes (see Figure 2 and Table 4, for the ranking of more stable genes for this and other experimental condition). Other lowest pairwise variation values were, V 11/12 (0.03) for the infected fruit experiment, V 8/9 (0.036) and V 11/12 (0.047) for Camarosa crown and petiole infected tissues, respectively, V 9/10 (0.035) for Andana infected petioles, V 9/10 (0.043) for hormonal treatment of in-vitro plants experiment, V 6/7 (0.053) for elicited cellular suspensions, and finally, V 7/8 (0.086) when all experiments were considered together.
In practice, however, the number of genes required should be low enough to make experimental procedures affordable, yet high enough to merit confidence in the conclusions. This means that if the pairwise variation value for n genes is below the recommended cut-off of 0.15, additional genes will not likely contribute to improved normalization [4]. Thus, the minimal number of reference candidates in each single experiment was determined to be two, for all the experimental conditions tested (marked with an arrowhead in Figure 3), but four in the all-together conditions. In each experimental condition, these genes were FaACTIN and   Evaluation of expression stability by DCt method, Normfinder and BestKeeper approaches. In order to accurately assess the usefulness of the thirteen candidate reference genes, other analytical methods were applied to the same data set. These include the comparative DCt method [39], which ranks the reference genes by their mean standard deviation in the pairwise comparisons. The NormFinder [9] method was also used. NormFinder ranks the set of candidate normalization genes according to their expression stability in a given sample set and a given experimental design. The Bestkeeper algorithm [38] performs pairwise comparison using the geometric mean of the Cp (Cq), values, and this one was also implemented. Table 5 shows the results obtained from all three methods. Both DCt and NormFinder analyses indicated a similar set of ideal reference genes for each experimental condition. Essentially, the best were FaTIM1 for ripening, FaEF1a for infected fruits, FaEF1a and FaGAPDH2 for Camarosa crown and petiole infected tissues, respectively, FaACTIN for Andana infected petioles, FaRIB413 for in-vitro hormone-treated plants, FaRIB413 for cellular suspension treatments, and finally, FaEF1a when all the experiments were analyzed together. Similar results were also obtained when BestKeeper algorithm was used.

Combination of All Five Methods Used for Selective Classification of Reference Genes by RankAggreg
Combined stability measurements were generated by merging all five approaches (''stability index'', geNorm PLUS , DCt method, Normfinder, and BestKeeper) to establish a consensus rank of reference genes by applying RankAggreg [40]. The input to this statistical package was a matrix of rank-ordered genes according to the different stability measurements previously computed by each of the five methods described above. RankAggreg calculated Spearman footrule distances and the software reformatted this distance matrix into an ordered list that matched each initial order as closely as possible. This consensus rank list was obtained by means of the Cross-Entropy Monte Carlo algorithm present in the software.
As shown in Figure 4, results of the merged data revealed that the most appropriate reference genes from all the preselected candidates tested for normalization are FaRIB413 and FaACTIN for analysis of strawberry fruit ripening, FaEF1a and FaACTIN for defense response studies in fruit, FaEF1a and FaGAPDH2, and

Validation of the Selected Superior Reference Genes
In order to validate the selected superior reference genes, the relative expression level of the strawberry gene encoding the transcription factor FaWRKY1 (AtWRKY75 ortholog, [30]) was determined in all the experimental sets of evaluated conditions. The strawberry gene FaWRKY1 acts as positive regulator of defense response during compatible and incompatible interactions in Arabidopsis and, very likely, FaWRKY1 is an important element mediating defense responses to C. acutatum in strawberry. We also know that the FaWRKY1 gene is significantly upregulated in strawberry tissues under C. acutatum attack, and after SA and MeJA treatments ( [30], Amil-Ruiz et al., unpublished data).
To analyze the bias effect on target expression analysis by selection of an inappropriate reference gene, FaWRKY1 was normalized to either a combination of the two best candidates ranked by RankAgreg algorithm as recommended by geNorm (Figures 3 and 4), or the least recommended one. FaWRKY1 primer sequences and other characteristics are listed in Table 1. As predicted, the reported expression profile of FaWRKY1 is strongly affected by the choice of the reference gene. Thus, in the strawberry fruit ripening conditions (RCF) as well as for infected petioles of cultivar Camarosa (FCP) and elicited cellular suspensions (HCC), the expression level values were similar to those previously reported ( [30]) when the reference genes were the two most recommended ones (FaRIB413 and FaACTIN, FaGAPDH2 and FaRIB413, FaEF1a and FaRIB413, respectively), either individually or combined as geometric mean (Figures 5a, 5d and  5g). To the contrary, a strong bias in the FaWKRY1 expression pattern was obtained when the least recommended gene (FaGAPDH1 in all three cases) was used for normalization. From these data the use of FaGAPDH1 as reference gene somehow neutralizes the detectable induction of FaWRKY1 during fruit ripening and senescence, in the response to infection and after elicitation with SA and MeJA compounds.
Interestingly, in other three experimental conditions (FCF, FCC and HCY) the use of the least stable reference gene (FaMT1, FaBZIP1 and FaACTIN respectively) seemed to have opposite influence in the detection of accurate expression values of the FaWRKY1 target gene. In this case the induction of this target gene was artificially high (Figures 5b, 5c and 5f). This is probably due to slightly but opposite variation of the corresponding reference mRNA levels during the analyzed process. These variations have significant impact in the final relative quantification of the expression of the target gene. Only 'Andana' petioles under fungal infection (FAP experiment) showed insignificant differences

Discussion
This work has mainly been focused to the evaluation of a set of potential strawberry reference genes for plant-defense response studies. Therefore, a variety of biological samples representing experimental conditions used to evaluate plant defense responses were used. The effect of natural pathogen infection as well as fruit senescence and decay are represented by experiments in this report through the analysis of fruit ripening and fruit natural infection by C. acutatum. Other tissues from Camarosa and Andana strawberry cultivars under fungal infection conditions were also included in this study, allowing comparisons between vegetative tissues within a cultivar, and between same tissues in different cultivars. Also, strawberry cultivars grown under contrasting contexts (in-vitro plants and cellular suspensions) were compared after treatment with either SA or JA, two phytohormones implicated in the activation of two well-known plant defense signaling pathways. A reference candidate with stability across such a range of conditions would be likely to perform well in narrower comparisons.
Several statistical procedures and software packages have been implemented to test which reference gene is best suited for transcript normalization in a given subset of biological samples. Each algorithm has its own strengths and limitations, so consensus among multiple tests provides great confidence that the results will be accurate and widely applicable.
Two methods, DCt and ''stability index'', perform studies about the variation of DCt in pairwise genes or simple Ct, respectively. The comparative DCt method ranks the reference genes by their mean standard deviation in the pairwise comparisons, while the ''stability index'' approach introduces statistics and linear regression analysis to rank the candidates by the product of the coefficient of variation and slope of regression of gene means against overall means for the different samples. In the latter method (Table 3, Figure S2), although genes with the lowest ''stability index'' values represent the best option for normalization, many of the other strawberry candidate genes may also be considered acceptable as controls based on the SI value obtained in this study. In addition, the level of expression of the reference genes compared to that of the genes being analyzed is an important factor to be considered in certain cases [10]. Thus, the two most stably expressed strawberry genes in all seven experiments together exhibited the greatest range in steady-state transcript accumulation. FaRIB413 was detected at relatively high levels due to its role as a structural component of the ribosome (mean Cq = 8.542), whereas FaACTIN was expressed at a much lower level (mean Cq = 24.011). Therefore, they may be considered as appropriate reference genes to test target genes with high or low transcript levels, respectively. Indeed, the FaRIB413 RNA has been demonstrated to be an appropriate internal control for strawberry expression studies across several tissues and experimental conditions, using RNA-gel blots or RTqPCR analyses [41], [29], [30]. FaRIB413 has also been recommended for studies of strawberry genes expressed at relatively low levels, but it must be diluted up to 4000 times in order to equilibrate this transcript to general expression levels an achieve comparative Cq analyses [29]. Using the ''stability index'' method, it appears clear that FaACTIN may serve as a non-diluted reference instead of FaRIB413.
The geNORM program (Table 4) uses pair-wise comparisons and geometric averaging across a matrix of reference genes and biological samples to determine the best reference. The program calculates the expression stability value (M A ) and allows accurate normalization of RTqPCR data [4], [37]. However, this approach leaves the method vulnerable to errors due to co-regulation, which tends to select those genes with the highest degree of similarity in their expression profiles [9]. On the other hand, it has the advantages that it is minimally affected by expression intensity of the reference genes [42] and it can determine the optimal number of genes (V) required to accurately normalize RTqPCR data based in pairwise variation [4]. Accordingly, two common well established sets of candidates with relatively high and low stability values were detected in all experimental conditions. FaEF1a always appears well positioned in the experimental conditions tested, and FaACTIN is stably expressed in ripening and mostly all infection conditions (except in crown tissue of cultivar Camarosa).
In contrast, the FaGAPDH1 and FaBZIP1 transcripts mostly showed high M A values (a lower stability) in all conditions. FaFHA1 is stably expressed in all conditions except in all infected tissues from cultivar Camarosa, and FaRIB413 is also stable but only in infected crown and petiole tissues from the same cultivar. On the other hand, the FaTIM1 transcript presented high M A values in all conditions except the two fruit experiments, where its accumulation was stable. The FaMT1 transcript presented low  Unlike geNORM, NormFinder is not affected by correlated expression of the candidate genes (Table 5). However, the latter gains in robustness as the sample number is increased, while geNorm doesn't need large sample size since it uses pair-wise comparisons. The Bestkeeper algorithm also performs pairwise comparisons using the geometric mean of the Cp (Cq) values, but different expression levels can generate heterogeneous variance between groups, and this can invalidate the use of Pearson correlation coefficient [43], [5]. The results from these two methodologies coincide with that of DCt method, and taken together these results indicate that gene FaEF1a seemed to be the most stably expressed reference gene meanwhile genes FaGAPDH1 and FaBZIP1 were the least stable ones.

Recommended Reference Genes in a Strawberry-defense Response Context
We have applied RankAggreg [40] to establish a consensus rank of reference genes by combination of all five above methods. This approach strengthens the value of the recommended candidates to normalize target gene expression in any of the conditions here described. Thus, results in Figure 4 show genes recommended in each particular experiment, suggesting they can be used as superior reference genes for transcript quantitation. Taken together, we propose genes FaRIB413, FaACTIN, FaEF1a and FaGAPDH2 as superior reference genes for accurate transcript normalization in strawberry (Fragaria 6 ananassa) under the described experimental conditions.
The genes proposed here have been reported in previous strawberry studies (see Table 1), although no experimental work was performed to validate their usefulness as RTqPCR reference genes in a variety of tissues, treatments or conditions. As previously stated, the FaRIB413 gene has been extensively used for northern and RTqPCR normalization in strawberry [41], [29], [44], [30], Table 4. Reference genes ranked in order by their average expression stability (MA) and coefficient of variation (CV) respectively.  Figure 3. Determination of the number of genes required to calculate a hypothetical normalization factor. Pairwise variation (Vn/n+1) analysis was carried out to determine the number of reference genes required for accurate normalization. An asterisk indicates the lowest V value in each experiment. An arrowhead indicates the minimum number of genes required to pass the suggested cut-off value (0.15) [4]. See Table 2

Plant Materials and Growth Conditions
Plant material, Fragaria 6 ananassa cultivars Chandler, Camarosa and Andana were used. Colletotrichum acutatum, a major strawberry pathogen was used for natural infection and controlled inocculation. All the plant culture and growth conditions, C. acutatum experimental conditions, and treatments with chemicals have been previously described [29], [30], and are summarized in Table 2. Briefly, strawberry cellular suspensions (cv. Chandler) were prepared from in vitro growing calli. Five days old cell suspensions were treated with MeJa (0.1 mM), SA (0.75 mM) or water (as control). Alicuots were taken at 2 hour intervals and cells were frozen in liquid nitrogen. Samples at 4 and 6 hours were used in this work because they match with a strong relative expression of the FaWRKY1 target gene, and many other strawberry genes currently under study in our lab. Axenic in-vitro plants from cv. Camarosa were aseptically sprayed with water, MeJa (2 mM) and SA (5 mM) solutions and collected at 12, 24 and 48 hours posttreatment. Strawberry fruits were collected from a growing field in several ripening stages and pooled by stage. Red stage strawberry fruits naturally-infected by Colletotrichum acutatum and exhibiting different increasing degrees of fungal necrotic lesions were collected and fruits having similar symptoms were pooled. No specific permissions were required for these activities. None human manipulation was applied to strawberry field prior to sample collection. Field studies did not involve endangered or protected species. Eight-week-old strawberry plantlets were placed in 20 cm diameter plastic pots containing sterilized peat and grown for a minimum of six additional weeks prior to mock or pathogen inoculation by spraying a spore suspension of 10 6 CFU ml 21 . Crowns and petioles were collected 1, 3, 5 and 7 days after treatment. All samples were flash frozen in liquid nitrogen and stored at 280uC until needed.

RNA Preparation for RTqPCR
Total RNA from strawberry fruits and vegetative tissues, as well as cell suspension cultures, was isolated according to Manning [53], treated with DnaseI (Invitrogen) to remove the residual contaminating DNA, and further purified with the RNeasy MinElute Cleanup Kit (QIAGEN). Extracted RNA samples showed high degree of purity, without residual contamination by organic compounds, accordingly to Accerbi et al. (2010) [54]. RNA samples were tested to be free of genomic DNA contamination after DNase I treatment by performing a qPCR analysis using the primer pairs corresponding to the FaGAPDH2 and FaRIB413 genes. Amplicons corresponding to these two genes were undetectable in the RNA samples after 40 cycles as confirmed by qPCR or by agarose gel electrophoresis (data not shown). These results indicated that amplicons generated by RTqPCR analysis were produced only from synthesized cDNA. Purified RNA was quantified by the NanoDrop 1000 Spectrophotometer (Thermo scientific) and the integrity checked by agarose gel electrophoresis and the Agilent 2100 Bioanalyzer (Agilent Technologies, Deutschland). All the samples showed RIN values over 8 (data not shown) and therefore were deemed suitable for RTqPCR analysis.
To ensure equal concentrations of RNA in all samples prior to the RT reactions, samples were diluted to 200 ng/ul and reassessed three times in a serial dilution of 1:0, 1:5 and 1:25, to ensure fidelity of the measure. First-strand cDNA synthesis was carried out by the iScript cDNA Synthesis kit (Bio-Rad) using as template 1 mg of purified total RNA per 20 mL of reaction volume. RT reactions were diluted 5-fold with nuclease-free water prior to be used in the qPCR.

Real-time qPCR
Specific primer pairs set for the genes tested were designed using Oligo Primer Analysis software version 6.65, tested by dissociation curve analysis, and verified for the absence of non-specific amplification. More details are provided in results. RTqPCR runs were performed in MyIQ and iCycler real-time PCR systems (Bio-Rad) using 96-well plates and 20 mL final reaction volume per well. Two mL template cDNA was added to the PCR reaction mixture containing 0.4 mM of each primer and 10 mL of 26 SsoAdvanced TM SYBRH Green supermix (Bio-Rad). The protocol was: an inicial step of enzyme activation/DNA denaturation of 95uC for 1 min, followed by 40 cycles of 95uC for 15 sec, 65uC for 15 sec and 72uC for 15 sec, and a final standard dissociation protocol to obtain the melting profiles. Data were adquired by means of the MyIQ v1.004 and iCycler v3.1 softwares (Bio-Rad).  Table 2  Computational Data Analysis Data analysis strategy is described in detail in the results section. Reaction efficiency calculus was done using LinRegPCR version 2012.3 [36], [55]. Resulting mean PCR efficiencies per amplicon were taken. Reference genes validation was performed using previously described software applications, included the MS Excel Figure 5. Transcript level relative quantification of the FaWRKY1 transcription factor. FaWRKY1 gene expression was analyzed in strawberry under the seven independent experimental conditions used in this study. Error bars show standard deviation calculated from two biological replicates. Normalization factors were calculated as the geometric mean of the expression levels of the two most stable reference genes as recommended in Figure 4 for each single experiment. Normalization to each gene individually is also shown. Additionally, the least stable reference gene was used for normalization of each experiment to demonstrate the effect of unstable reference genes in the quantification of the relative amount of target gene mRNA. Every sample was calibrated with their corresponding mock sample (see Table 2 for experimental details). Black lines linked to the X axis have been added to f and g to illustrate range of gene induction. doi:10.1371/journal.pone.0070603.g005 VBA applets NormFinder v0.953 [9] and BestKeeper v1 [38], and the geNorm [4] algorithm provided in qBasePlus v2.4 package [37]. Other statistical procedures were performed with the free software R v2.15.2 (http://www.R-project.org), with the packages RankAggreg 0.4-3, clValid 0.6-4 and gtools 2.7.0; and SPSS software ver 15.0 for Windows.  Table 2 for  descriptions of tissue samples, represented here by abbreviations.  (TIF)