344 Functional Plant Biology D.Xiao et al. (see Fig.S2).The first-strand ofcDNA was synthesised by reverse stress treatments)and were also treated as one group for transcribing I ug of total RNA in a final reaction volume of analysis of candidate reference gene stabilities.Data or mean 20 uL using the M-MLV reverse transcriptase (Takara,Dalian, Ct values obtained from gRT-PCR were transformed to China)according to the manufacturer's protocol and diluted quantities with PCR efficiency derived straight from 1:10 before use in gRT-PCR assays.The concentration and amplification plots using LinReg (ver.7.0;Amsterdam,The quality ofeach RNA and cDNA sample was also measured using Netherlands)software (Ramakers et al.2003).The normalised the nucleic acid analytic apparatus K6000 (Bio Photometer, data were imported and analysed by two stability analysis Eppendorf,Germany). programs for reference genes,geNorm ver.3.4 (Vandesompele et al.2002)and NormFinder (Andersen et al.2004)for ranking Primer design and gRT-PCR the reference genes Specific primer pairs were designed for 10 single genes(GAPDH, geNorm determines the gene stability measure (M)value Tub_d,Cyclophilin (CyP),EFI-a,18S rRNA,poly ubiquitin for each gene,based on the average pair-wise variation for a enzyme(UBO),UBC30,Pentatricopeptide repeat(PPR),PP2A particular gene with all the other tested genes.Thus,genes can be and Malate dehydrogenase(MDH))and three orthologous genes ranked according to their expression stability through the (ACTIN,ACTIN-1 and ACTIN-2)of the ACTIN gene family.The stepwise exclusion of the least stable gene.The genes with an sequences of the genes were retrieved from Arabidopsis thaliana M value were arbitrarily suggested to be lower than 1.5;genes (L.)Heynh.and blasted against EST(expressed sequence tag) with the lowest M values have the most stable expression.A pair- libraries (Brassica)via nucleotide blast.The Arabidopsis wise stability measure aims at determining the benefit of adding consensus sequences were used for comparison with B.rapa extra reference genes for the normalisation process.For this,an EST sequences,if available,to reveal the exon-intron structure. arbitrary cut off value of0.15 for pair-wise variation (Vn/Vn+1) When no B.rapa L.ssp.chinensis (L.)Hanelt EST sequence of normalisation factor (NF)(NFn and NFn+1)is calculated, was available the gene structure was obtained by comparison of reflecting the effect of including additional(n+1)genes. the A.thaliana sequence with Brassica napus L.and Brassica The second different statistical algorithm software, oleracea L.EST sequences assuming that the exon-exon NormFinder,generates a stability value for each gene,which boundaries are conserved between A.thaliana and B.napus/ is a direct measure for its estimated expression variation.It ranks B.rapa/B.oleracea (see Fig.S3).Sequence comparisons were the stability level of each candidate gene and highlights the most done by DNASTAR,Lasergene 9.1 (Lasergene,Madison,WI stably expressed gene with the lowest stability value by using a USA)and primers were designed using Primer Express 2.0 model-based variance estimation approach. software (PE Applied Biosystems,Foster,CA,USA)under default parameters.Primer sequences and exon-exon junctions Results of the reference genes are listed in Table 1.Primer specificity and Expression profiling of the candidate genes DNA contamination were visualised by separating PCR products from cDNA and DNA on agarose gel (Fig.S1A)and when only a Melting curve analysis of the amplification products confirmed single band was observed,the band was purified to be template for that the primers amplified a single PCR product(Fig.S3).The preparing the standard curves(which takes into account primer standard curves for each of the candidate reference genes were efficiency)using gRT-PCR (Ramakers et al.2003).The PCR found to have R2>0.995 (Table 1),indicating a strong reaction efficiencies (E)were calculated using the equation linear relationship between the detected Ct values and the E=10(-1/slope)(Lekanne Deprez et al.2002).This calculation corresponding relative amount of cDNA in all the PCR method results in efficiencies ranging from 95 to 108%.The reactions.Based on the slopes of the standard curves,the 13 standard curve was generated using a dilution series of the Dsl gene assays were found to have PCR efficiencies >90.7% (leaf developmental stage 1)sample over at least five dilution (Table 1).It was apparent that each candidate reference gene points (Fig.S4).gRT-PCR was performed in triplicate on a had variable Ct values in the wide variety of samples tested,as the 32-position carousel (Light Cycler)with the Light Cycler-RNA Ct values ranged widely from 14.56(18S rRNA)to 39.53(PP24) amplification kit SYBR Green I(Roche,Mannheim,Germany) across all samples (Table 2).None of the 13 candidate reference and conducted in 25 uL reaction volumes containing 30ng uL genes had a uniform expression over all samples tested.As a cDNA sample,along with an RNA template control in parallel consequence,for accurate normalisation of gene expression in for each gene.The thermal cycling consisted of 95C for 2 min different experimental conditions,specific sets ofreference genes and 40 cycles of95C for 20s,55C for 20s and 72C for 20s.After are needed. the PCR a melting curve was generated to check the specificity of the amplified fragment.Data analysis was performed with geNorm analysis the Rotor-gene 6 ver.6.1 software(Applied Biosystems).All the The geNorm program conducts sequential elimination ofthe least cycle threshold(Ct)values from one gene were determined at stable gene in any given experimental group,resulting in the the same threshold fluorescence value of 0.2.The single-peak exclusion of all but the two most stable genes in each strategic melting curves obtained using the 13 primer pairs to amplify the group(Table 3).For all 20 samples tested,ACTIN and CyP were candidate reference genes are displayed in Fig.S1B. the most stable genes,followed by UBC30,EFI-a and UBO. In contrast,PP2 A was the least stable gene tested.The two most Data analyses stably expressed genes when comparing samples of different The 20 samples were divided into four strategic groups developmental stages were ACTIN and CyP that were also most (developmental stages,different tissues,biotic and abiotic stable when all 20 samples were analysed together,followed by(see Fig. S2). Thefirst-strand of cDNA was synthesised by reverse transcribing 1 mg of total RNA in a final reaction volume of 20 mL using the M-MLV reverse transcriptase (Takara, Dalian, China) according to the manufacturer’s protocol and diluted 1 : 10 before use in qRT–PCR assays. The concentration and quality of each RNA and cDNA sample was also measured using the nucleic acid analytic apparatus K6000 (Bio Photometer, Eppendorf, Germany). Primer design and qRT–PCR Specific primer pairs were designed for 10 single genes (GAPDH, Tub_a, Cyclophilin (CyP), EF1-a, 18S rRNA, poly ubiquitin enzyme (UBQ), UBC30, Pentatricopeptide repeat (PPR), PP2A and Malate dehydrogenase (MDH)) and three orthologous genes (ACTIN, ACTIN-1 and ACTIN-2) of the ACTIN gene family. The sequences of the genes were retrieved from Arabidopsis thaliana (L.) Heynh. and blasted against EST (expressed sequence tag) libraries (Brassica) via nucleotide blast. The Arabidopsis consensus sequences were used for comparison with B. rapa EST sequences, if available, to reveal the exon-intron structure. When no B. rapa L. ssp. chinensis (L.) Hanelt EST sequence was available the gene structure was obtained by comparison of the A. thaliana sequence with Brassica napus L. and Brassica oleracea L. EST sequences assuming that the exon-exon boundaries are conserved between A. thaliana and B. napus/ B. rapa/B. oleracea (see Fig. S3). Sequence comparisons were done by DNASTAR, Lasergene 9.1 (Lasergene, Madison, WI, USA) and primers were designed using Primer Express 2.0 software (PE Applied Biosystems, Foster, CA, USA) under default parameters. Primer sequences and exon-exon junctions of the reference genes are listed in Table 1. Primer specificity and DNA contamination were visualised by separating PCR products from cDNA and DNA on agarose gel (Fig. S1A) and when only a single band was observed, the band was purified to be template for preparing the standard curves (which takes into account primer efficiency) using qRT–PCR (Ramakers et al. 2003). The PCR reaction efficiencies (E) were calculated using the equation E = 10(–1/slope) (Lekanne Deprez et al. 2002). This calculation method results in efficiencies ranging from 95 to 108%. The standard curve was generated using a dilution series of the Ds1 (leaf developmental stage 1) sample over at least five dilution points (Fig. S4). qRT–PCR was performed in triplicate on a 32-position carousel (Light Cycler) with the Light Cycler-RNA amplification kit SYBR Green I (Roche, Mannheim, Germany) and conducted in 25 mL reaction volumes containing 30ng mL–1 cDNA sample, along with an RNA template control in parallel for each gene. The thermal cycling consisted of 95
C for 2 min and 40 cycles of 95
C for 20s, 55
C for 20s and 72
C for 20s. After the PCR a melting curve was generated to check the specificity of the amplified fragment. Data analysis was performed with the Rotor-gene 6 ver. 6.1 software (Applied Biosystems). All the cycle threshold (Ct) values from one gene were determined at the same threshold fluorescence value of 0.2. The single-peak melting curves obtained using the 13 primer pairs to amplify the candidate reference genes are displayed in Fig. S1B. Data analyses The 20 samples were divided into four strategic groups (developmental stages, different tissues, biotic and abiotic stress treatments) and were also treated as one group for analysis of candidate reference gene stabilities. Data or mean Ct values obtained from qRT–PCR were transformed to quantities with PCR efficiency derived straight from amplification plots using LinReg (ver. 7.0; Amsterdam, The Netherlands) software (Ramakers et al. 2003). The normalised data were imported and analysed by two stability analysis programs for reference genes, geNorm ver. 3.4 (Vandesompele et al. 2002) and NormFinder (Andersen et al. 2004) for ranking the reference genes. geNorm determines the gene stability measure (M) value for each gene, based on the average pair-wise variation for a particular gene with all the other tested genes. Thus, genes can be ranked according to their expression stability through the stepwise exclusion of the least stable gene. The genes with an M value were arbitrarily suggested to be lower than 1.5; genes with the lowest M values have the most stable expression. A pair￾wise stability measure aims at determining the benefit of adding extra reference genes for the normalisation process. For this, an arbitrary cut off value of 0.15 for pair-wise variation (Vn/Vn + 1) of normalisation factor (NF) (NFn and NFn + 1) is calculated, reflecting the effect of including additional (n + 1) genes. The second different statistical algorithm software, NormFinder, generates a stability value for each gene, which is a direct measure for its estimated expression variation. It ranks the stability level of each candidate gene and highlights the most stably expressed gene with the lowest stability value by using a model-based variance estimation approach. Results Expression profiling of the candidate genes Melting curve analysis of the amplification products confirmed that the primers amplified a single PCR product (Fig. S3). The standard curves for each of the candidate reference genes were found to have R2 0.995 (Table 1), indicating a strong linear relationship between the detected Ct values and the corresponding relative amount of cDNA in all the PCR reactions. Based on the slopes of the standard curves, the 13 gene assays were found to have PCR efficiencies 90.7% (Table 1). It was apparent that each candidate reference gene had variable Ct values in the wide variety of samples tested, as the Ct values ranged widely from 14.56 (18S rRNA) to 39.53 (PP2A) across all samples (Table 2). None of the 13 candidate reference genes had a uniform expression over all samples tested. As a consequence, for accurate normalisation of gene expression in different experimental conditions, specific sets of reference genes are needed. geNorm analysis The geNorm program conducts sequential elimination of the least stable gene in any given experimental group, resulting in the exclusion of all but the two most stable genes in each strategic group (Table 3). For all 20 samples tested, ACTIN and CyP were the most stable genes, followed by UBC30, EF1-a and UBQ. In contrast, PP2A was the least stable gene tested. The two most stably expressed genes when comparing samples of different developmental stages were ACTIN and CyP that were also most stable when all 20 samples were analysed together, followed by 344 Functional Plant Biology D. Xiao et al