Validation of RT-qPCR reference genes for in planta expression studies in Hemileia vastatrix, the causal agent of coffee leaf rust

f u n g a l b i o l o g y 1 1 5 ( 2 0 1 1 ) 8 9 1 e9 0 1

journal homepage: www.elsevier.com/locate/funbio

Validation of RT-qPCR reference genes for in planta expression studies in Hemileia vastatrix, the causal agent of coffee leaf rust Ana VIEIRAa,b, Pedro TALHINHASa,*, Andreia LOUREIROa, S
ebastien DUPLESSISc,
vio S. PAULOb, Helena Gil AZINHEIRAa Diana FERNANDEZd, Maria do C
eu SILVAa, Octa ~ o das Ferrugens do Cafeeiro/Instituto de Investigac¸a ~o Cient
ıfica Tropical, 2784-505 Oeiras, Portugal Centro de Investigac¸a Computational Biology and Population Genomics Group, Centro de Biologia Ambiental, Faculdade de Ci^encias da Universidade de Lisboa, 1749-016 Lisboa, Portugal c Institut National de la Recherche Agronomique, UMR 1136 INRA/Nancy Universit
e Interactions Arbres/Micro-organismes, Champenoux, France d Institut de Recherche pour le D
eveloppement, UMR 186 IRD/CIRAD-UM2 R
esistance des Plantes aux Bioagresseurs, Montpellier, France a

b

article info

abstract

Article history:

Hemileia vastatrix is a biotrophic fungus, causing coffee leaf rust in all coffee growing coun-

Received 16 December 2010

tries, leading to serious social and economic problems. Gene expression studies may have

Received in revised form

a key role unravelling the transcriptomics of this pathogen during interaction with the

30 June 2011

plant host. Reverse transcription quantitative real-time polymerase chain reaction (RT-

Accepted 4 July 2011

qPCR) is currently the golden standard for gene expression analysis, although an accurate

Available online 18 July 2011

normalisation is essential for adequate conclusions. Reference genes are often used for this

Corresponding Editor: Simon Avery

purpose, but the stability of their expression levels requires validation under experimental conditions. Moreover, pathogenic fungi undergo important biomass variations along their

Keywords:

infection process in planta, which raises the need for an adequate method to further nor-

Basidiomycete plant pathogen

malise the proportion of fungal cDNA in the total plant and fungus cDNA pool. In this

Coffea arabica

work, the expression profiles of seven reference genes [glyceraldehyde-3-phosphate dehy-

Coffee leaf rust

drogenase (GADPH), elongation factor (EF-1), Beta tubulin (b-tubulin), cytochrome c oxidase

Housekeeping gene

subunit III (Cyt III), cytochrome b (Cyt b), Hv00099, and 40S ribosomal protein (40S_Rib)] were

Normalisation factor

analysed across 28 samples, obtained in vitro (germinated uredospores and appressoria) and in planta (post-penetration fungal growth phases). Gene stability was assessed using the statistical algorithms incorporated in geNorm and NormFinder tools. Cyt b, 40S_Rib, and Hv00099 were the most stable genes for the in vitro dataset, while 40S_Rib, GADPH, and Cyt III were the most stable in planta. For the combined datasets (in vitro and in planta), 40S_Rib, GADPH, and Hv00099 were selected as the most stable. Subsequent expression analysis for a gene encoding an alpha subunit of a heterotrimeric G-protein showed that the reference genes selected for the combined dataset do not differ significantly from those selected specifically for the in vitro and in planta datasets. Our study provides tools for correct validation of reference genes in obligate biotrophic plant pathogens, as well as the basis for RT-qPCR studies in H. vastatrix. ª 2011 British Mycological Society. Published by Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ351 214544680; fax: þ351 214544689. E-mail address: [email protected] 1878-6146/$ e see front matter ª 2011 British Mycological Society. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.funbio.2011.07.002

892

Introduction Gene expression analysis is crucial for enhancing our understanding of the signalling and metabolic pathways which underlie cellular and developmental processes. Although several methods have been used to quantify gene expression, including Northern blotting, RNase protection assay, in situ hybridisation, and cDNA microarray technology, Reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR) is considered the gold standard for quantifying gene expression, thanks to its sensitivity, specificity, dynamic range, and high throughput capacity (Freeman et al. 1999; Bustin et al. 2005; Udvardi et al. 2008). This technique can detect very low quantities of a target transcript even if only a few copies are present in the sample (Wong & Medrano 2005). However an accurate normalisation is required to obtain a reliable quantification of the transcript (Huggett et al. 2005; Derveaux et al. 2010). Different procedures to normalise RT-qPCR data have been proposed over the years, evolving from techniques such as sample size, total RNA or genomic DNA quantification, and artificial molecules controls, to techniques that take into account sample variations, such as differences in the quantity and quality of RNA, and efficiencies of reverse transcription or PCR (Huggett et al. 2005; Cruz et al. 2009; Teste et al. 2009; Walker et al. 2009). In this perspective, internal control genes (references genes) are the most commonly used to normalise RT-qPCR data, though the success of this procedure relies on the appropriate choice of control genes (Pfaffl et al. 2004; Dheda et al. 2005). Typically, reference genes should exhibit two major properties: (i) they should be essential for the maintenance of cellular function and viability, and therefore should be constitutively expressed in all tissues; (ii) their transcription should not be affected by the conditions under study (Derveaux et al. 2010). However, some evidences show that almost all genes seem to be regulated under some conditions and there are always some variations in transcript levels, so that none of the commonly exploited genes can be viewed as a universal reference gene (Thellin et al. 1999; € rzenbaum & Kille 2001; Dheda et al. 2005). In many cases, Stu the use of, a single reference gene is inadequate, and can pro€ rzenbaum & Kille 2001; duce wrong biological conclusions (Stu Hu et al. 2009; Teste et al. 2009). Additionally, gene expression studies as well as reference gene validations are mainly limited to human and other well-established model organisms, while non-model species often suffer from lack of background information available (Axtner & Sommer 2009). For all these reasons, a careful evaluation of one or even more suitable reference genes is essential prior to any gene expression profiling study. Currently, several methods such as geNorm (Vandesompele et al. 2002) and NormFinder (Andersen et al. 2004) based on different statistical algorithms have been developed to select the most stable reference genes to appropriately normalise RT-qPCR results (Vandesompele et al. 2002; Andersen et al. 2004; Pfaffl et al. 2004). Genes encoding actin, glyceraldehyde-3-phosphate dehydrogenase, elongation factor, ß-tubulin, and 28S and 18S ribosomal genes are frequently used in molecular studies as reference genes for RT-qPCR without appropriate validation,

A. Vieira et al.

which may jeopardise the reliability of data (Kim et al. 2003; Nicot et al. 2005; Fang & Bidochka 2006; Bohle et al. 2007). In fungi and oomycetes, the determination of the most stable genes for expression studies was only carried out for a limited number of species, with different reference genes validated in each species. These include actin, secretion associated GTPbinding protein (sarA), and cytochrome c oxidase (Cox5) from Aspergillus niger (Bohle et al. 2007), elongation factor (MlpELF1a) and a-tubulin (Mlp-aTUB) from Melampsora larici-populina (Hacquard et al. 2011), elongation factor (EF-1), a glyceraldehyde-3-phosphate dehydrogenase (GADPH ), and tryptophan biosynthesis enzyme (tryp) from Metarhizium anisopliae (Fang & Bidochka 2006), ubiquitin-conjugating enzyme (Ubc), Beta tubulin (Tub-b), and 40S ribosomal protein (WS21) from Phytophthora parasitica (Yan & Liou 2006) and mannosyltransferase activity (ALG9), RNA Pol II transcription factor activity (TAF10), RNA Pol III transcription factor activity (TFC1), and ubiquitin-protein ligase activity (UBC6) from Saccharomyces cerevisiae (Teste et al. 2009). In this way, it is crucial to validate reference genes prior to expression studies, namely in non-model organisms such as Hemileia vastatrix. Hemileia vastatrix is responsible for coffee leaf rust, a disease that can lead to yield losses of 30 % in Coffea arabica if no control measures are applied. This pathogen establishes a biotrophic interaction with its host and is completely dependent of plant living cells to grow and reproduce (Silva et al. 2006; Nunes et al. 2009; Ramiro et al. 2009; Azinheira et al. 2010; Talhinhas et al. 2010). In spite of its biotrophic lifestyle, H. vastatrix uredospores can germinate and differentiate appressoria in vitro, but further differentiation of infection structures (i.e. infection hyphae, haustoria, and sporulation structures) requires the presence of the host plant (Azinheira et al. 2001). In coffee leaves, appressoria formed over stomata differentiate an infection hypha which invades the substomatic cavity, from which the fungus grows colonising the leaf tissues inter- and intra-cellularly, feeding from living coffee cells by specialised structures named haustoria (Silva et al. 1999). Leaf tissues become heavily colonised by the fungus and the infection cycle is completed with the formation of sporogenic hyphae and the release of uredospores, which occurs from 21 d after the infection (Rodrigues et al. 1975; Silva et al. 1999). Therefore, the proportion of H. vastatrix biomass in planta increases along the infection process, which can be a problem for validation of reference genes. In this way, an additional normalisation step is required to successfully validate H. vastatrix reference genes for in planta studies. Although this is a normalisation in conceptual terms, we termed it ‘correction’ to avoid confusion with the common use of ‘normalisation’ in qPCR. It should be noted that, even though quantification of fungal pathogen DNA is frequently carried out by qPCR (Atallah et al. 2007; Mideros et al. 2009), including rust fungi (Boyle et al. 2005; Jackson et al. 2006; Acevedo et al. 2010), this strategy for the validation of RT-qPCR fungal reference genes ‘in host’ is still a novelty. To date, such validation of reference genes following a correction step has been only defined in the poplar rust fungus M. larici-populina (Rinaldi et al. 2007; Hacquard et al. 2011). The main focus of this work was the identification of the best H. vastatrix reference genes suitable for the normalisation

Validation of RT-qPCR reference genes for in planta expression studies

of RT-qPCR data in subsequent expression studies. To achieve this goal, a methodology for correction of H. vastatrix nucleic acid content in planta was established, and eight genes from different cellular functional categories were selected for RT-qPCR primer design and validation as references for H. vastatrix gene expression studies both in vitro and in planta.

Material and methods Fungal and plant material In vitro growth Fresh uredospores (30 mg) of the Hemileia vastatrix isolate CIFC 1065 (race II) were spread in sterile distilled water in Petri dishes to germinate (GU e germinated uredospores sample) or on artificial oil-collodion membranes (Heath & Heath 1976) placed in Petri dishes to differentiate appressoria (A) (Azinheira et al. 2001), and incubated under saturating humidity for 24 h under darkness at 24  C. Fungal material (aprox. 15 mg) was collected by filtration (for the GU sample) or the oil-collodion membranes were used directly for grinding (for the A sample).

In planta growth Both Caturra and Matari cultivars (physiological groups E and b respectively) of Coffea arabica were inoculated with fresh uredospores of Hemileia vastatrix isolate 1065, establishing compatible interactions. The inoculation was performed by spreading the uredospores (1 mg/pair of leaves) on the lower surface of young coffee leaves with a soft camel brush. After an incubation period of 24 h in a dark moist chamber, the plants were moved to greenhouse conditions. The leaves (two pairs of leaves from two different plants for each cultivar) were collected at six time-points after inoculation (1, 2, 3, 7, 14, and 21 d) in order to obtain the in planta samples of H. vastatrix.

893

Controls Control samples consisted in ungerminated Hemileia vastatrix uredospores (U e the calibrator sample), uninoculated young coffee leaves, and freeze-dried Lecanicillium sp. (formerly Verticillium sp.; a hyperparasite isolated from H. vastatrix uredospores) mycelia obtained upon culture in liquid medium (Potato Dextrose Broth, Difco, USA) for 7 d at 25  C without agitation. All experiments were repeated in order to obtain two independent biological replicates. At each collection time, fungal and plant material were collected to evaluate the germination of spores, appressorium formation, and fungal growth in host tissues by light microscopy (Silva et al. 1999), and to perform RNA extraction. The material required for extraction was immediately frozen by immersion in liquid nitrogen and stored at 80  C.

Reference genes Candidate reference genes were selected from the literature representing functional categories related either with metabolism, structure, and secretion. Such functional diversity reduces the risk that the selected genes share the same regulatory mechanisms. Also, a gene predicted to encode a secreted protein was selected from three 454-pyrosequencing Hemileia vastatrix datasets in which the corresponding transcripts showed identical abundance (Talhinhas et al. 2010). The genes putatively encode an elongation factor (EF-1), a beta-tubulin (b-tubulin), a GADPH, an ubiquitin (Ubi), a cytochrome c oxidase subunit III (Cyt III ), a cytochrome b (Cyt b), a 40S ribosomal protein (40S_Rib), and a predicted H. vastatrix secreted protein (Hv00099), as detailed in Table 1. Sequences of the H. vastatrix genes were obtained from several H. vastatrix 454-pyrosequencing datasets (Talhinhas et al. 2010; Fernandez et al. in press), from the literature (Cyt b; Grasso

Table 1 e RT-qPCR primer sequences and amplicon characteristics for Hemileia vastatrix genes assessed in this study. Gene description (and database referencea) EF-1 e deliver of aa-tRNA to the elongation ribosome (FR720601) GADPH e oxidoreductase in glycolysis and gluconeogenesis (contig03273) Cyt III e catalyses the transfer of electrons from reduced cytochrome c to molecular oxygen (contig19515) Cyt b e component of a respiratory chain complex III (DQ022192) b-tubulin e cytoskeletal structural protein involved in cell motility structure and integrity (FR720600) 40S_Rib e protein involved in m RNA synthesis (contig01333) Hv00099 (Specific secreted protein) e unknown function (contig21737) Gpa (Heterotrimeric) e protein involved in the regulation of the signalling pathway (contig00973)

Primer sequence

Amplicon PCR size (bp) efficiency

Boundary

Tm ( C)

F e GTTGTGGAAGTTCGAGACTC R e CAGCAATCAACAAGATGCC F e ACTTGGACAGCTACGAC R e CCATACCAGTGAGCTTCC F e AGTAGATATGAGTCCCTGACC R e CACCTTCAGCACTTACATCC

128

1.96

Exon/exon

87

280

1.92

Exon/intron

87

173

1.93

Exon/intron

80

F e TAGGGGTGACTGGGAATGC R e GGAGCGTACATGACAATAATAGC F e CTAGACATGGTCGTTACCTC R e GCAATGTCACAATGAGCAG

120

1.91

Exon/exon

77

156

1.96

Exon/intron

79

F e CACACGGAAAGATTGGTACG R e CCTTGAGCGAATCAACGG F e CCCAGTCCGAATAATAGCC R e CCAACGCTTATTATCATCCG F e GCTCTGAACGTTGTCCAC R e AAGTCGTATTGGTCAGCC

114

1.97

Exon/intron

83

195

2.03

Exon/exon

81

146

1.98

Exon/exon

86

a From Fernandez et al. (in press), except Cyt b (Grasso et al. 2006) and EF-1 and b-Tubulin (this work; Genbank references).

894

et al. 2006), or using degenerated PCR primers designed from orthologous genes in other rusts, and with the CloneJET PCR cloning kit (MBI Fermentas, Lithuania). The latter included EF-1 and b-tubulin, which were amplified from H. vastatrix DNA based on the following degenerated primers designed after homologous genes in Uromyces fabae and Melampsora lini (EF-1 forward, CBGARCGHGARCGTGGTATCAC; EF-1 reverse, TTVGADATACCRGCYTCRAATTCACC; b-tubulin forward, ACHGGDGCTGGAATGGG; b-tubulin reverse, TTNCC RATRAAHGTSACVGACATCTT).

RT-qPCR primer design and efficiency test Primers were designed using PerlPrimer v1.1.17 (Marshall 2004) with melting temperatures of 60  C, lengths of 18e24 bp, and GC content of 50e60 %. Gradient PCRs (58e68  C) were performed for each gene in order to determine the optimal annealing temperature (data not shown). Exon/ intron boundaries were determined by aligning each cDNA sequence with its corresponding genomic sequence (Hemileia vastatrix DNA sequences of EF-1, Cyt b, and b-tubulin, and orthologous sequences from other rust fungi for the remaining genes). When possible, primers were designed at the junction of two different exons to help distinguishing amplification from potential DNA contamination (Table 1). This strategy was not adapted in the case of genes for which it would generate poor primer quality (Table 1). Nevertheless, the presence of DNA in the RNA samples was also systematically tested. Amplicon size was shorter than 200 bp for all genes tested, except for GADPH, where the need to avoid the amplification of the plant gene led to the generation of a 280-bp amplicon. In the infected coffee samples (in planta), the fungal and plant cDNAs are mixed, so primer sequences must specifically target the fungus and should not amplify plant cDNAs. In addition, H. vastatrix may be hyperparasited by some ascomycete fungi from the genus Lecanicillium, which DNA should neither be amplified. To avoid heterologous amplification of target sequences, an in silico strategy was defined: first, H. vastatrix candidate gene sequences were used as queries to search for orthologous sequences in the NCBI (http://www.ncbi.nlm. nih.gov), the HarvEST Coffee (http://harvest.ucr.edu/) and the Verticillium group database (http://www.broadinstitute. org/annotation/genome/verticillium_dahliae/MultiHome.html) databases using BLASTN and BLASTX algorithms; then matching sequences (best blast hits) were aligned with the corresponding primers to assess their similarity and only primers showing more than a five base pair difference were accepted. Primers were synthesised by STABvida (Caparica, Portugal) with HPLC purification. The primer efficiency was experimentally tested with the LinRegPCR programme developed by Ramakers et al. (2003) which uses a linear regression analysis of fluorescence data from the exponential phase of PCR amplification to determine amplification efficiency (E). LinRegPCR software utilises an iterative algorithm (considering the number of data points, regression coefficient and slope of the regression line) for the selection of the exponential phase in each PCR amplification (Cikos et al. 2007; Rutledge & Stewart 2008). Analysis of RT-qPCR fluorescence data with LinRegPCR

A. Vieira et al.

determined E values for each reaction. Additionally, a combined analysis using an averaged E value (arithmetical mean of E values of all samples) was applied and that value was used in subsequent analyses.

Total RNA isolation and cDNA synthesis Total RNA was extracted from frozen samples using the RNeasy Plant Mini kit (Qiagen, Hilden, Germany), with addition of an in-solution DNase I digestion, according to Appendix E of the RNeasy Plant Mini kit manual. Only RNA samples with 260 nm/280 nm wavelength ratio between 1.9 and 2.1 and 260 nm/230 nm wavelength ratio greater than 2.0 before and after DNase I digestion were used for cDNA synthesis. The quality of RNA samples was also assessed by electrophoresis on 1.2 % agarose gels. RNA quantity was determined with a Lambda EZ201 spectrophotometer (PerkinElmer, Waltham, USA). A control PCR with the Cyt b primers (Table 1) was run on RNA samples to check for the absence of genomic DNA. First-strand cDNAs were synthesised from 1 mg total RNA in 20 ml final volume, using the Omniscript Reverse Transcription kit (Qiagen) and oligo(dT)18 primer (MBI Fermentas, Vilnius, Lithuania) following the manufacturer’s instructions. Each sample was diluted 25-fold and stored at 20  C. Serial dilutions (2, 10, and 100) were performed for the U, 1 d and 21 d samples and used in RT-qPCR for all genes (as described below). Cq values were plotted against the logarithm of dilution. Linearity was verified in the U sample for all dilutions, in the 21 d sample for the 2 and 10 dilution (and also for the 100 dilution for most genes) and in the 1 d sample for the 2 dilution (and occasionally for the 10 dilution). This dilution assay indicated that, even for the 1 d sample (which is the scarcest of all in planta samples for Hemileia vastatrix cDNA), the dilution used in the study was still within the linear dynamic range. On the other hand, it suggested the absence of inhibition at this dilution.

RT-qPCR The RT-qPCRs were performed using an iQ5 real-time thermal cycler (Bio-Rad, Hercules, USA), based on EvaGreen Supermix (Bio-Rad), in 96-well Clear Multiplate PCR Plates (Bio-Rad) sealed with iCycler iQ Optical tape adhesives (Bio-Rad). Each 15 ml reaction comprised 6 ml template, 7.5 ml EvaGreen Supermix, 0.3 ml of each primer (10 mM), and 0.8 ml of sterile distilled water. The reactions were subjected to an initial denaturation step at 95  C during 10 min followed by 40 cycles at 95  C for 10 s and 60  C for 30 s. A melting curve analysis was performed at the end of the PCR run over the range 60e95  C, increasing the temperature in a stepwise fashion by 0.5  C every 10 s. Baseline correction was performed with the iQ5 Optical System Software (Bio-Rad) and the corresponding fluorescence values were exported and then used to determine Cq and efficiency values with LinRegPCR programme. Three negative controls (1 e no template; 2 e cDNA of plant; 3 e cDNA of Lecanicillium sp.) were included in the RT-qPCR run to validate the in silico strategy and ensure that only cDNA of Hemileia vastatrix were amplified. Each RT-qPCR reaction was performed in duplicate and the specificity of the amplicons was

Validation of RT-qPCR reference genes for in planta expression studies

checked by melting curve analysis and by 2.5 % agarose gel electrophoresis.

Assessment of expression stability Correction of Hemileia vastatrix genomic DNA In order to correct the expression levels of H. vastatrix reference genes in the in planta samples according to the fungal biomass, genomic DNA was extracted from the same leaves used for the RNA assays [100 mg of plant material, using the DNeasy Plant mini kit (Qiagen), including a RNase treatment following the manufacturer’s recommendations]. Genomic DNA concentrations were estimated with a Nanodrop 1000 (Thermo Scientific, Waltham, USA), and a qPCR was performed on 60 ng genomic DNA following conditions described above, using Hv00099 primers. Cq obtained (CqDNA Hv00099) were used to correct reference genes Cq (CqcDNA reference genes) using the following formula: Cqcorrected ¼ CqcDNA reference genes  CqDNA Hv00099

(1)

Determination of reference gene expression stability The Cq value of each reference gene obtained in all in vitro and in planta samples (including ungerminated uredospores) following the experimental design mentioned in ‘Fungal and plant material’ were transformed into relative quantities (Q) as compared to ungerminated uredospores (U) using the DCq formula, where DCq ¼ Cqsample  CqU, and E is the efficiency of the primer pair (Pfaffl 2001): Q ¼ EDCq

(2)

Expression stability of candidate reference genes across biological samples was evaluated using Q values in geNorm (Vandesompele et al. 2002) version 3.4 (Microsoft Excel VBA applet; http://medgen.ugent.be/wjvdesomp/genorm), and NormFinder (Andersen et al. 2004) (Microsoft Excel-based tool; http://www-mdl.dk/publicationsnormfinder) following the author’s recommendations.

NormFinder. NormFinder assesses the expression stability of a gene by evaluating its expression variation within tissues or treatments (‘groups’ in NormFinder terminology) compared to variation among these groups (Andersen et al. 2004). The programme algorithm implies the estimation of intra- and inter-group variation and combines both results in a stability value for each investigated gene (Andersen et al. 2004). To select for the best reference genes, four datasets were run in this programme, each dataset containing two groups (the two biological replicates). The first dataset (in vitro assays) is composed of three samples (U, GU, and A). The second is composed of Hemileia vastatrixeCoffea arabica (Caturra) assays and comprises seven samples (U þ in planta samples). The third is composed of H. vastatrixeC. arabica (Matari), as described for the second dataset. Finally, the fourth dataset is composed of all nine samples (U þ in vitro þ in planta samples).

geNorm. geNorm is based on the simple assumption that expression of two ideal reference genes will always have the same ratio among samples regardless of the experimental

895

conditions before the RT-qPCR (Vandesompele et al. 2002). The average expression stability value M for each gene is calculated using the expression data. M is the average pairwise variation of a gene compared with each of the other potential reference genes in one sample. The average M of all genes together is then calculated by stepwise exclusion of the least stable gene until the two most stable genes in the remaining set cannot be ranked any further. geNorm also allows estimating the optimal number of reference genes that should be used for normalisation. It calculates the normalisation factor (NF) based on the geometric mean of the expression of more than one reference gene. geNorm calculates the NFn of the two most stable reference genes based on the geometric mean of the expression data and then the NFnþ1 with the next most stable gene. To determine how many genes should be used for accurate normalisation, the pairwise variation Vn/ nþ1 was determined out of two sequential NFs (NFn and NFnþ1) (Vandesompele et al. 2002). As previously described for NormFinder, four groups were used to determine the most stable reference genes. The first dataset (in vitro assays) is composed of six samples (two biological replicates of U, GU, and A samples). The second is composed of Hemileia vastatrixeCoffea arabica (Caturra) assays and comprises 14 samples (two biological replicates of U þ in planta samples). The third is composed of H. vastatrixeC. arabica (Matari), as described for the second dataset. Finally, the fourth dataset is composed of all 18 samples (two biological replicates of U þ in vitro þ in planta samples).

Validation of normalisation strategies The gene expression profile of a heterotrimeric G-protein a subunit (Gpa) homologue was used to compare and validate the different NFs. The fold change values were calculated using the relative expression formula after Pfaffl (2001), with non-germinated uredospores as control. The standard deviation (SD) of each point was calculated using the following formula (according to the iQ5 Optical System Software instruction manual): SD ¼ SDCq  Q  LnðEÞ

(3)

where SDCq is the standard deviation of the Cq values obtained; Q is defined in equation (2) and E is the efficiency. To assess significance of differences of Gpa expression values normalised by different NFs, a Spearman Correlation was performed using the Statistica 5.0 software.

Results and discussion How to successfully validate Hemileia vastatrix reference genes An accurate normalisation requires the use of several reference genes whose expression variations must be negligible under the investigated conditions. In this study we analysed the gene expression profiles of seven candidate genes (Table 1), previously described as good reference genes for other fungal species (Fang & Bidochka 2006; Yan & Liou 2006; Bohle et al. 2007; Teste et al. 2009; Hacquard et al. 2011). RT-qPCR primers

896

generated amplicons spanning 100e280 bp (Table 1) with GC content between 45 % and 60 % and no secondary structures. RT-qPCR efficiency ranged from 1.92 to 2.03 (Table 1), with little differences between in vitro and in planta samples (maximum difference in efficiency was for Hv00099, ranging from 1.94 for the in vitro samples to 2.03 for the in planta samples) or even between all samples (maximum SD was 0.06 for 40S_Rib, with individual efficiencies ranging from 1.90 to 2.08). No amplification was obtained for any of the negative controls (no template, plant or Lecanicillium sp. cDNA). Melting curve analysis further corroborated these observations (data not shown), namely in situations where primer dimers were formed because of lack of template. Besides these seven genes, Ubi was also tested but was excluded from our analysis because it was not amplified in all tested samples. Taken together these results validated the in silico strategy for primers design and ensured that the Cq value only reflected the transcript abundance of H. vastatrix reference genes. For some genes, the Cq value ranged from 35 to 15 cycles along the infection process in the collected in planta samples (Fig 1). This striking range of Cq reflects the variation of the fungal RNA amounts in the different in planta samples, following the H. vastatrix DNA contents measured in these samples (Fig 1). These values reflect the growth of the fungus as it colonises the plant and the increase in fungal biomass per gram of leaf tissue. As expected, low Cq values were obtained for in vitro samples, ranging from 17.3e18.2 for Cyt III to 24.4e25.8 for b-tubulin, without major differences among samples for each gene (data not shown). Therefore, it became clear that the validation of reference genes for in planta H. vastatrix samples would require correction of the Cq values of the candidate reference genes according to the variation of the fungal

A. Vieira et al.

biomass in the plant, as previously described for the poplar rust pathogen (Rinaldi et al. 2007; Hacquard et al. 2011). Despite this correction, occasionally Cq values for each gene did not follow a strictly parallel line to that of genomic DNA, suggesting variations in the expression levels of such genes (e.g., EF-1 in both replicates for the Coffea arabica (Caturra)eH. vastatrix interaction in the 21 d sample; Fig 1), as reported in other systems (Vandesompele et al. 2002; Andersen et al. 2004; Pfaffl et al. 2004). It becomes obvious and essential to use statistical tools like geNorm and NormFinder to validate the best set of reference genes. However, these tools were developed to assess the stability of reference genes when the amount of RNA is similar under study conditions (Vandesompele et al. 2002; Andersen et al. 2004). So, in order to use these programmes in H. vastatrix in planta samples, the Cq value obtained at each time-point was corrected after the DNA content, as described in formula (1). It should be noted that this additional normalisation (correction) step is only required for the selection of stable reference genes, not in the subsequent gene expression analysis.

Selection of the best candidate reference genes for RT-qPCR analyses The stability of seven candidate genes was evaluated using geNorm and NormFinder algorithms, since each programme has the potential to deliver different results, which led us to follow a combined analysis. Several datasets were analysed: 1 e in vitro stages; 2 e in planta stages using Coffea arabica (Caturra) as a host; 3 e in planta stages using C. arabica (Matari) as a host; 4 e in vitro and in planta samples. As shown in Fig 2 the results of both programmes were consistent, with only

Fig 1 e Transcript levels of reference genes tested, presented as Cq mean value in Hemileia vastatrix samples when inoculated in Coffea arabica leaves (A e Caturra, B e Matari). The DNA curve corresponds to the amplification of the Hv00099 gene from genomic DNA isolated from rust infected coffee leaves.

Validation of RT-qPCR reference genes for in planta expression studies

897

Fig 2 e Gene expression stability values of the seven potential reference genes. The stability values on the right axis were calculated with NormFinder ( ) and the average expression stability values M (-C-) on the left axis were calculated with geNorm after stepwise exclusion of the least stable gene. Genes are plotted from the least (red names) to the most stable expressed genes (green names). (A) in vitro assays; (B) in planta Hemileia vastatrixeCoffea arabica (Caturra) assays; (C) in planta H. vastatrixeC. arabica (Matari) assays; (D) in vitro and in planta assays.

slight differences in the ranking order, except for the in vitro stages samples. For the later, geNorm identified b-tubulin and Cyt b as the most stable combination of reference genes, while NormFinder identified Cyt b and 40S_Rib as the two best reference genes (Fig 2A). The differences in rankings of gene stability using the two algorithms could be due to the different methods used to assess gene stability (Hu et al. 2009; Walker et al. 2009). geNorm selection is based on the pairwise variation between genes so the two most stably expressed genes are therefore those that share an expression profile (Vandesompele et al. 2002). In contrast, NormFinder uses a model-based algorithm that takes into account overall stability as well as the stability of any groups that may be present in the sample set (Andersen et al. 2004). However, these differences in selection of reference genes may not be significant, since all the candidate genes have a M value below the default limit of 1.5, which is the top suitability threshold suggested by geNorm (Vandesompele et al. 2002). In the in planta assays, despite slight differences in the ranking order, both programmes selected the same three genes (GADPH, Cyt III, and 40S_Rib) as the best reference genes (Fig 2B and C). In the fourth dataset, when analysing in vitro and in planta samples together, the ranking order of reference genes was the same with both tools (Fig 2D). The best pair matches with the best three reference genes chosen for the in planta dataset (GADPH and 40S_Rib), but the third most stable is different (Cyt III is replaced by the Hv00099). As in the in vitro dataset, both genes (Cyt III and Hv00099) have an M value lower than 1.5, suggesting that they can be used to normalise studied conditions, even though Cyt III was more stable than Hv00099 in the in planta dataset. Thus, we concluded that in most cases the GADPH and 40S_Rib were the most stable reference genes while

b-tubulin and EF-1 were the least stable, in all tested datasets. Since geNorm recommended the selection of the three most stable genes, we also retrieved the third most stable gene from NormFinder.

Number of reference genes As Vandesompele et al. (2002) stated regarding the costebenefit relationship of the normalisation procedure, there is a trade-off between practical considerations, such as time/ costs, and accuracy. Obviously an accurate NF should not include rather unstable genes like EF-1 and b-tubulin (Fig 2). On the other hand, if all genes are relatively stable, analysing more genes than necessary can be a waste of resources. In this way, it is essential to find the optimal number of reference genes. Vandesompele et al. (2002) suggested that a minimal of three genes were required to a correct normalisation of RTqPCR data and that subsequent additions of a new reference gene will be necessary until the normalisations with NFn and NFnþ1 give similar values (Vandesompele et al. 2002). geNorm has the ability to perform this analysis automatically by calculating the pairwise variation values (Vn/nþ1) between each combination of sequential NFs. A cut-off value of 0.15 is recommended, below which inclusion of an additional control gene does not result in a significant improvement of the normalisation. Yet this is not an absolute value and it can change according to the data (Vandesompele et al. 2002). As shown in Fig 3, the pairwise variation values, for the four datasets, are well above of the recommended cut-off value. However this value was established based on a Spearman correlation between NFn and NFnþ1 in which low variation values, correspond to high correlation coefficients (Vandesompele

898

A. Vieira et al.

Fig 3 e Optimal number of reference genes required for effective normalisation. The pairwise variation (Vn/VnD1) was analysed between the NFs NFn and NFnD1 by geNorm programme to determine the optimal number of reference genes required for RT-qPCR data normalisation.

Fig 4 e Effect of normalisation strategies in the Gpa expression profile relative to ungerminated uredospores. (A) Gpa expression profile for the first assay of Hemileia vastatrixeCoffea arabica (Caturra). (B) Gpa expression profile for the second assay of H. vastatrixeC. arabica (Caturra). (C) Gpa expression profile for the first assay of H. vastatrixeC. arabica (Matari). (D) Gpa expression profile for the second assay of H. vastatrixeC. arabica (Matari). (E) Gpa expression profile for the first in vitro assay. (F) Gpa expression profile for the second in vitro assay. Normalisation was performed using: (i) Best NF for the in planta samples (40S_Rib, GADPH, Cyt III ); (ii) Best NF for the in vitro samples (Hv00099, Cyt b, 40S_Rib); (iii) Best NF for all developmental phases (40S_Rib, GADPH, Hv00099); (iv) worst NF (geometric average of EF-1 and b-tubulin); (v) a single gene often used as reference gene (GADPH ).

Validation of RT-qPCR reference genes for in planta expression studies

et al. 2002). Based on the same premise we performed a Spearman correlation tests and obtained the following results: in vitro (V3/4 ¼ 0.175391; r ¼ 0.987578; p ¼ 0.000); in planta using Coffea arabica (Caturra) as a host (V3/4 ¼ 0.316262, r ¼ 0.972028, p ¼ 0.000); in planta using C. arabica (Matari) as a host (V3/4 ¼ 0.20654; r ¼ 0.972028; p ¼ 0.000); in vitro and in planta together (V3/4 ¼ 0.308546; r ¼ 0.975652; p ¼ 0.000). These results show that the addition of a fourth reference gene to perform normalisation does not provide relevant information, so the use of three genes is enough to accurately normalise expression of genes of interest. However, as stated before the three most stable genes differ between the in planta (40S_Rib, GADPH, and Cyt III ), the in vitro (Cyt b, Hv00099, and 40S_Rib), and the in planta and in vitro (40S_Rib, GADPH, and Hv00099) samples.

The choice of the reference genes affects the normalisation of a gene of interest To understand how the choice of the reference genes could affect the normalisation of a gene of interest, the expression of the Hemileia vastatrix Gpa gene encoding a heterotrimeric Gpa was analysed during the plant infection process. Gpa plays important roles in fungal biology such as regulation of filamentous growth, mating, appressorium formation, and € lker 1998; Lev et al. 1999; Deising et al. 2000; pathogenicity (Bo Li et al. 2007). Fig 4 presents the relative Gpa expression levels when normalised with different methods: (i) NF validated for the in planta samples; (ii) NF validated for the in vitro samples; (iii) NF validated for the in vitro and in planta assays together (global NF); (iv) Worst NF (EF-1, b-tubulin); (v) a single reference gene as it is sometimes used in expression studies (GADPH ). Ratios of expression values were almost identical using the three best NFs (Fig 4, blue, purple, and pink bars). However, some inconsistent values were detected among the 28 samples depending on the NF considered. For example, gene expression was always higher using the best NF for the in planta samples than using the best NF for the in vitro samples for GU, A or 1 d. However, despite these differences in the absolute values, the gene expression profiles remained similar. Thus, we concluded that the use of a stage-specific NF (i.e., specific factor for in vitro or in planta samples) or a global NF (same factor for all samples) were both good strategies to follow, although specific NFs provide a more accurate normalisation of expression profiles. By contrast, when we compared any of the best NFs with the worst one, there were important differences in the gene expression profiles (Fig 4, black bar versus blue, violet or pink bars). In fact, at some time-points after inoculation, this NF likely led to biological misinterpretation. For instance, the Gpa expression was higher at early stages under appropriate normalisation but not expressed at all with the worst NF. Finally, the Gpa expression profile normalised with GADPH was similar to the profiles obtained by using the best NF (Fig 4, green bar versus the blue, pink or violet bars), with significantly higher absolute values at 1 d, corroborating the previous studies showing that a single reference gene is not sufficient for proper normalisation of gene expression profiling by RT-qPCR.

899

Conclusions This study provides the first report of suitable reference genes for expression studies of Hemileia vastatrix genes by RT-qPCR. Normalisation is a critical factor in reporting gene expression data, providing a necessary control for errors associated with sample preparation. Reference genes provide a mean of controlling these errors. The present study tested eight candidate reference genes across 28 H. vastatrix samples representing major developmental stages of the pathogen. We determined distinct NFs that could alternatively be used to assess expression in rust samples in planta and in vitro. We further described a method for correction of in planta H. vastatrix samples to account for variations in fungal biomass along the infection process, which may be adapted to gene expression studies on any organism that requires a living host to differentiate specific structures and that undergoes important in-host biomass changes. On the other hand, we showed that some reference genes commonly used in molecular analysis of fungi such as EF-1 and b-tubulin were not appropriate in our pathosystem and could have led to wrong expression profiles and biological misconclusions. Therefore, selection of appropriate reference genes is a prerequisite before studying any gene of interest. To summarise, this work provides a set of pertinent reference genes that should be used for validation of expression data from RT-qPCR analysis in H. vastatrix.

Acknowledgements We would like to acknowledge the valuable comments and suggestions from Luisa Carvalho (Instituto Superior de Agronomia/Technical University of Lisbon, Lisboa, Portugal), Diogo ~ o das Ferrugens do Cafeeiro/InstiSilva (Centro de Investigac¸a ~ o Cient
ıfica Tropical, Oeiras, Portugal) and tuto de Investigac¸a ^ nTiago Jesus (Centro de Biologia Ambiental, Faculdade de Cie cias da Universidade de Lisboa, Lisboa, Portugal). This re~ o para a Cie ^ncia search was financially supported by Fundac¸a e a Tecnologia (FCT), Portugal (PTDC/AGR-AAM/71866/2006 re des Affaires and SFRH/BPD/47008/2008) and by Ministe
res et Europe
ennes, France, and FCT (Partenariat Etrang e Hubert Curien PHC-Pessoa 14700TF).

references

Acevedo M, Jackson EW, Sturbaum A, Ohm HW, Bonman JM, 2010. An improved method to quantify Puccinia coronata f. sp. avenae DNA in the host Avena sativa. Canadian Journal of Plant Pathology 32: 215e224. Andersen CL, Jensen JL, Ørntoft TF, 2004. Normalization of realtime quantitative reverse transcription-PCR data: a modelbased variance estimation approach to identify genes suited for normalization, applied to bladder and colon cancer data sets. Cancer Research 64: 5245e5250. Atallah ZK, Bae J, Jansky SH, Rouse DI, Stevenson WR, 2007. Multiplex real-time quantitative PCR to detect and quantify Verticillium dahliae colonization in potato lines that differ in response to Verticillium wilt. Phytopathology 97: 865e872.

900

Axtner J, Sommer S, 2009. Validation of internal reference genes for quantitative real-time PCR in a non-model organism, the yellow-necked mouse, Apodemus flavicollis. BMC Research Notes 2: 264. ~ es L, Mendgen K, Azinheira HG, Silva MC, Guerra-Guimara Rodrigues Jr C, Pinto Ricardo C, 2001. Development of infection structures of Hemileia vastatrix on artificial membranes. In: 11th Conference of the Mediterranean Phytopathological Union,
17e20 September 2001. Evora, Portugal, pp. 353e355. Azinheira HG, Silva MC, Talhinhas P, Medeira C, Maia I, Petitot A-S, Fernandez D, 2010. Non-host resistance responses of Arabidopsis thaliana to the coffee leaf rust fungus (Hemileia vastatrix). Botany 88: 621e629. € cke Y, Dalpiaz A, Cordes C, Horn H, Bohle K, Jungebloud A, Go Hempel DC, 2007. Selection of reference genes for normalisation of specific gene quantification data of Aspergillus niger. Journal of Biotechnology 132: 353e358. € lker M, 1998. Sex and crime: heterotrimeric G proteins in fungal Bo mating and pathogenesis. Fungal Genetics and Biology 25: 143e156. Boyle B, Hamelin RC, Seguin A, 2005. In vivo monitoring of obligate biotrophic pathogen growth by kinetic PCR. Applied and Environmental Microbiology 71: 1546e1552. Bustin SA, Benes V, Nolan T, Pfaffl MW, 2005. Quantitative realtime RT-PCR e a perspective. Journal of Molecular Endocrinology 34: 597e601.
A, Koppel J, 2007. Relative quantification of Cikos S, Bukovska mRNA: comparison of methods currently used for real-time PCR data analysis. BMC Molecular Biology 8: 113. Cruz F, Kalaoun S, Nobile P, Colombo C, Almeida J, Barros LM,
MF, Vaslin M, Alves-Ferreira M, 2009. Romano E, Grossi-de-Sa Evaluation of coffee reference genes for relative expression studies by quantitative real-time RT-PCR. Molecular Breeding 23: 607e616. Deising HB, Werner S, Wernitz M, 2000. The role of fungal appressoria in plant infection. Microbes and Infection 2: 1631e1641. Derveaux S, Vandesompele J, Hellemans J, 2010. How to do successful gene expression analysis using real-time PCR. Methods 50: 227e230. Dheda K, Huggett JF, Chang JS, Kim LU, Bustin SA, Johnson MA, Rook GA, Zumla A, 2005. The implications of using an inappropriate reference gene for real-time reverse transcription PCR data normalization. Analytical Biochemistry 344: 141e143. Fang W, Bidochka MJ, 2006. Expression of genes involved in germination, conidiogenesis and pathogenesis in Metarhizium anisopliae using quantitative real-time RT-PCR. Mycological Research 110: 1165e1171. Fernandez D, Tisserant E, Talhinhas P, Azinheira HG, Vieira A, Loureiro A, Petitot A-S, Poulain J, da Silva C, Silva MC, Duplessis S. 454-pyrosequencing of Coffea arabica leaves infected by the rust fungus Hemileia vastatrix reveals in planta expressed pathogen-secreted proteins and plant functions expressed in a late compatible plant-rust interaction. Molecular Plant Pathology, in press. doi:10.1111/J.1364-3703.2011.00723.X. Freeman WM, Walker SJ, Vrana KE, 1999. Quantitative RT-PCR: pitfalls and potential. BioTechniques 26: 112e122. Grasso V, Sierotzki H, Garibaldi A, Gisi U, 2006. Relatedness among agronomically important rusts based on mitochondrial cytochrome b gene and ribosomal ITS sequences. Journal of Phytopathology 154: 110e118. Hacquard S, Veneault-Fourrey C, Delaruelle C, Frey P, Martin F, Duplessis S, 2011. Validation of Melampsora larici-populina reference genes for in planta RT-quantitative PCR expression profiling during time-course infection of poplar leaves. Physiological and Molecular Plant Pathology 75: 106e112. Heath IB, Heath MC, 1976. Ultrastructure of mitosis in the cowpea rust fungus Uromyces phaseoli var. Vignae. Journal of Cell Biology 70: 592e607.

A. Vieira et al.

Hu R, Fan C, Li H, Zhang Q, Fu Y, 2009. Evaluation of putative reference genes for gene expression normalization in soybean by quantitative real-time RT-PCR. BMC Molecular Biology 10: 93. Huggett J, Dheda K, Bustin S, Zumla A, 2005. Real-time RT-PCR normalisation; strategies and considerations. Genes and Immunity 6: 279e284. Jackson EW, Avant JB, Overturf KE, Bonman JM, 2006. A quantitative assay of Puccinia coronata f. sp. avenae DNA in Avena sativa. Plant Disease 90: 629e636. Kim B, Nam H, Kim S, Kim S, Chang Y, 2003. Normalization of reverse transcription quantitative-PCR with housekeeping genes in rice. Biotechnology Letters 25: 1869e1872. Lev S, Sharon A, Hadar R, Horwitz BA, 1999. A mitogen-activated protein kinase of the corn leaf pathogen Cochliobolus heterostrophus is involved in conidiation, appressorium formation, and pathogenicity: diverse roles for mitogen-activated protein kinase homologs in foliar pathogens. Proceedings of the National Academy of Sciences of the USA 96: 13542e13547. Li L, Wright SJ, Krystofova S, Park G, Borkovich KA, 2007. Heterotrimeric G protein signaling in filamentous fungi. Annual Review of Microbiology 61: 423e452. Marshall OJ, 2004. PerlPrimer: cross-platform, graphical primer design for standard, bisulphite and real-time PCR. Bioinformatics 20: 2471e2472. Mideros SX, Windham GL, Williams WP, Nelson RJ, 2009. Aspergillus flavus biomass in maize estimated by quantitative real-time polymerase chain reaction is strongly correlated with aflatoxin concentration. Plant Disease 93: 1163e1170. Nicot N, Hausman J, Hoffmann L, Evers D, 2005. Housekeeping gene selection for real-time RT-PCR normalization in potato during biotic and abiotic stress. Journal of Experimental Botany 56: 2907e2914. Nunes CC, Maffia LA, Mizubuti ES, Brommonschenkel SH, Silva JC, 2009. Genetic diversity of populations of Hemileia vastatrix from organic and conventional coffee plantations in Brazil. Australasian Plant Pathology 38: 445e452. Pfaffl MW, 2001. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Research 29: e45. Pfaffl MW, Tichopad A, Prgomet C, Neuvians TP, 2004. Determination of stable housekeeping genes, differentially regulated target genes and sample integrity: BestKeeper e excel-based tool using pair-wise correlations. Biotechnology Letters 26: 509e515. Ramakers C, Ruijtera JM, Lekanne Depreza RH, Moorman AFM, 2003. Assumption-free analysis of quantitative real-time polymerase chain reaction (PCR) data. Neuroscience Letters 339: 62e66. Ramiro DA, Escoute J, Petitot A-S, Nicole M, Maluf MP, Fernandez D, 2009. Biphasic haustorial differentiation of coffee rust (Hemileia vastatrix race II) associated with defence responses in resistant and susceptible coffee cultivars. Plant Pathology 58: 944e955. Rinaldi C, Kohler A, Frey P, Duchaussoy F, Ningre N, Couloux A, Wincker P, Le Thiec D, Fluch S, Martin F, Duplessis S, 2007. Transcript profiling of poplar leaves upon infection with compatible and incompatible strains of the foliar rust Melampsora larici-populina. Plant Physiology 44: 347e366. Rodrigues Jr CJ, Bettencourt AJ, Rijo L, 1975. Races of the pathogen and resistance to coffee rust. Annual Review of Phytopathology 13: 49e70. Rutledge RG, Stewart D, 2008. Critical evaluation of methods used to determine amplification efficiency refutes the exponential character of real-time PCR. BMC Molecular Biology 9: 96. Silva MC, Nicole M, Rijo L, Geiger JP, Rodrigues Jr CJ, 1999. Cytochemistry of plant-rust fungus interface during the compatible interaction Coffea arabica (cv. Caturra)eHemileia vastatrix (race III). International Journal of Plant Sciences 160: 79e91.
rzea V, Guerra-Guimara ~ es L, Azinheira HG, Silva MC, Va Fernandez D, Petitot A-S, Bertrand B, Lashermes P, Nicole M, 2006. Coffee resistance to the main diseases: leaf rust and

Validation of RT-qPCR reference genes for in planta expression studies

coffee berry disease. Brazilian Journal of Plant Physiology 18: 119e147. € rzenbaum SR, Kille P, 2001. Control genes in quantitative Stu molecular biological techniques: the variability of invariance. Comparative Biochemistry and Physiology Part B 130: 281e289. Talhinhas P, Azinheira HG, Loureiro A, Batista D, Vieira B, PinaMartins F, Tisserant E, Petitot AS, Paulo OS, Duplessis S, Silva MC, Fernandez D, 3rde8th October 2010. Overview of the functional virulent genome of the coffee leaf rust pathogen Hemileia vastatrix. In: 23rd International Conference on Coffee Science (ASIC), pp. 414e422 Bali, Indonesia. Teste M, Duquenne M, Franc¸ois JM, Parrou J, 2009. Validation of reference genes for quantitative expression analysis by realtime RT-PCR in Saccharomyces cerevisiae. BMC Molecular Biology 10: 99. Thellin O, Zorzi W, Lakaye B, De Borman B, Coumans B, Hennen G, Grisar T, Igout A, Heinen E, 1999. Housekeeping

901

genes as internal standards: use and limits. Journal of Biotechnology 75: 291e295. Udvardi MK, Czechowski T, Scheible W, 2008. Eleven golden rules of quantitative RT-PCR. Plant Cell 20: 1736e1737. Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F, 2002. Accurate normalization of realtime quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biology 3: 0034. Walker CG, Meier S, Mitchell MD, Roche JR, Littlejohn M, 2009. Evaluation of real-time PCR endogenous control genes for analysis of gene expression in bovine endometrium. BMC Molecular Biology 10: 100. Wong ML, Medrano JF, 2005. Real-time PCR for mRNA quantitation. BioTechniques 39: 75e85. Yan HZ, Liou RF, 2006. Selection of internal control genes for realtime quantitative RT-PCR assays in the oomycete plant pathogen Phytophthora parasitica. Fungal Genetics and Biology 43: 430e438.