Introduction

The global area cultivated with biotech plant crops and the number of globally approved genetically modified (GM) crops and derived products are continuously growing [1]. In the European Union (EU) for example, controlling for the presence of genetically modified organism (GMOs) throughout the food and feed supply chains is necessary to be in compliance with a strict regulatory system [2–4]. A key aspect of this control system is the use by competent authorities of internationally validated, official polymerase chain reaction (PCR) methods for event-specific identification and relative quantification of each individual GM transformation event [3, 5–8].

In real-time PCR, a transgene event-specific assay and a target taxon or endogenous reference sequence assay are combined. The target taxon reference assay—further called reference assay (RA)—is used to quantify the total deoxyribonucleic acid (DNA) copy number and, further, to calculate the relative copy number proportion of the transgenic sequence. The functions of this RA in PCR-based GMO quantification determine the criteria which it has to fulfil: an RA should be taxon-specific, amplify a sequence which is present in a single copy number in the plant genome and amplify uniformly along the varieties within the target taxon.

Maize (Zea mays L.) has a high economical importance [9]. Maize cultured in the world can be divided into five types according to the grain characteristics: flint corn, dent corn, popcorn, sweet corn and waxy corn. Intensive breeding activities have led to many intermediate types between flint and dent corn. Originated in South America and cultured worldwide, the main gene pools involved in modern maize breeding activities are found in North America and Europe. Maize represents one of the most important biotech crops [1]. While only a few GM maize events have been approved in the EU for cultivation, more than 20 transgenic events have been approved so far for import or processing in food and/or feed products in the EU. Currently, 47 MON 810 maize varieties are approved for cultivation in Spain, France and Germany and registered in the Common catalogue of varieties [10].

For maize, various RAs have been developed and validated. The official Community Reference Laboratory (CRL) methods for quantification of different GM maize events employ different RAs, amplifying sequence parts of genes encoding for high mobility group protein gene (Hmg) and alcohol dehydrogenase 1 gene (Adh1) [8]. Other maize-specific methods detecting invertase (Ivr) and zein were developed and validated in-house or in the frame of collaborative projects [11, 12]. Another RA amplifying a region of the starch synthetase type B (SSIIb) gene is implemented in Japanese GMO standards [13–15]. Alternative maize RAs have been developed for reason to overcome possible modifications of the target genes such as copy number differences or mutations [11].

Maize-derived products which have to be tested for GMO presence may contain any of the maize varieties available on the market, which means that the maize RAs must perform uniformly along a broad range of plant varieties, transgenic or conventional. These varieties are produced by different breeding programmes utilising genetic resources from various origins. The maize genome is highly diverse. Single nucleotide polymorphisms (SNPs), duplications, deletions or insertions (indels) and transposon-like elements insertions are the sources of genetic variation determining the variability of different genomic loci between maize inbred lines [16–18]. SNPs are rather common in maize [16, 19, 20]. The average maize sequence diversity is much higher than the observed sequence variation for e.g. human and Drosophila genomes. Two randomly chosen maize loci vary on average in approximately one of 104 bases in contrast with the human genome where the random loci variation is on average one of 1,200–1,900 bases [20].

Variations in the nucleotide sequence in RAs’ loci are likely to affect the GMO quantification. Single base pair substitutions or small indels could lead to primer or probe/template mismatching and further to a decrease in the measured total DNA copy number [21–23]. An SNP leading to a decrease in the measured total DNA copy number was observed in the forward primer sequence of an internationally validated Adh1 assay [21], demonstrating that although the method has been validated, the uniformity of the RA might not always be sufficiently assessed prior to implementation of the assay in routine GMO testing procedures.

This study aims at evaluating the reliability of eight existing maize RAs, four of which are part of official methods for GMO quantification. We analysed the nucleotide sequence variation in the target genomic regions in a broad range of transgenic and conventional varieties and lines: MON 810 varieties cultivated in Spain, conventional varieties and lines from various geographical origin and breeding history. Alternatively, the reliability of the assays was evaluated based on their PCR amplification performance in relation to the observed nucleotide variability.

Methods and materials

Maize samples and sample preparation

Seeds from 14 transgenic MON 810 varieties and five isogenic maize lines were kindly provided by Consejo Superior de Investigaciones Cientificas (CSIC; Spain). Seeds from 73 commercial varieties presenting several commercial and local breeding programmes were kindly provided by Agricultural Institute of Slovenia (Slovenia). Seeds from inbred maize line B73 were kindly provided by Bayer BioScience (Belgium). Seeds from inbred line HiII were kindly provided by European Commission-Joint Research Centre. Another set of 84 varieties was collected originating from Europe, South and North America, South Africa and China and covering the main types of use of maize (grain and forage corn, sweet corn, ornamental corn and popcorn). For additional data on the varieties included in the study, see the Table in the Supplementary material. Certified reference materials (CRMs) for maize GM events of MON 810, Bt176, Bt11, GA21, 1507, MON 863 and NK603 were purchased from the Institute for Reference Materials and Measurements (Geel, Belgium). These CRMs were certificated for GMO content (mass proportion, grammes per kilogramme). Transgene T25 maize material was supplied by Bayer CropScience (Gent, Belgium).

DNA extraction was performed using the DNeasy Plant Mini Kit (QIAGEN, Westburg) starting from 100 mg leaf material obtained from individual plants (Mon810 and conventional maize varieties). DNA was extracted from 250 mg of grain-derived powder as described in Zhang et al. [24] (84 varieties from the worldwide collection).

Maize target sequences

Eight maize RAs targeting five different maize-specific genes were considered in this study (Table 1, A). For Adh1, three publicly available target sequences are considered: one used as RA in the CRL methods for quantification of NK603, MON863 and GA21 (2005 assay, GA212005) maize, and two others used in the CRL methods for T25, respectively, Bt11, GA21 (2007 assay, GA212007), MIR604 and 3272 maize. The latter two assays are more or less one and the same assay, expect that the forward primer for Adh1 in the T25 assay has a single base more in length than the forward primer in the RA for Bt11. Further, in this paper, these two assays will be considered as one and the same. For the purpose of this paper, two different RAs for the same endogenous gene will be distinguished by adding the amplicon length between brackets: the two Adh1 assays will be named Adh1 (70) and Adh1 (136), respectively, while the two zein assays will be named zein (104) and zein (110), respectively (Table 1, A).

Table 1 Published maize reference assays (RAs; A) and used oligonucleotides in polymerase chain reaction (PCR) analyses for the different purposes described in this study: conformation sensitive capillary electrophoresis (CSCE) analysis (B), sequencing analysis (B) and real-time PCR analysis (A)
Full size table

PCR analyses

Unless otherwise stated, all apparatuses, software and specific reagents are from Applied Biosystems. PCRs for conformation sensitive capillary electrophoresis (CSCE) as well as for sequence analysis were performed in 25 µl volumes containing approximately 100 ng DNA, 300 nM from both forward and reverse primers, 200 µM deoxynucleotide triphosphates and 1 U AmpliTaq DNA polymerase. Thermal cycling reactions were performed under the following conditions: initial denaturation at 94 °C for 3 min, followed by 35 cycles of denaturation at 94 °C for 30 s, primer annealing at 57 or 59 °C (depending on the primer pair) for 40 s, and elongation at 72 °C for 40 s to 1 min depending on the amplicon size. The final extension reaction was performed at 72 °C for 5 min, after which, the reaction was held at 4 °C. The PCR products were used either for CSCE analysis or for sequencing. The products used for CSCE were amplified using 3′FAM- or 3′HEX-labelled primers, while for sequencing, the products were amplified with non-labelled primers (see Table 1, B). Before sequencing, the PCR products were purified by means of MSB® Spin PCRapace (Invitek, Belgium). Where appropriate, primers were designed by means of Primer3 v. 0.3.0 [25]. All primer sequences are shown in Table 1. To control the success of the amplification reactions, PCR products were analysed using agarose gel electrophoresis.

CSCE analysis

Pools of DNA were made by combining DNAs from four to eight lines in equal concentrations. The pooled DNAs were used as a template for PCR amplification with the primers targeting the regions of interest (Table 1, B, for Ivr1 and Hmg, see Table 1, A). DNA of inbred line B73 was added to each pool as a reference. After the PCR amplification, heteroduplexes were formed by denaturation followed by re-hybridization under the following conditions: 96 °C for 10 min; 72 cycles (96 °C, 60 °C (−0.5 °C/cycle) for 20 s), 60 °C for 30 min, then cooled down to room temperature. The homo- and heteroduplexes were electrophoretically separated at 18 °C for 2,500 s, at voltage 10 kV using an ABI3130 Genetic Analyzer. The instrument has a capacity of 96 samples automatically injected at 10 kV for 10 s to a 60-cm long, 50-µm diameter capillary filled with the sieving matrix Conformational Analysis Polymer (CAP®). This polymer is composed of 5% native polymer diluted with TTE glycerol tolerant buffer, 4 M urea and 4% sucrose. The relative position of the fluorescent peaks was determined by comparison with an internal 500 ROX Standard. The level of fluorescence was monitored by diluting the samples 120 up to 1,000 times. The fluorescent peak patterns were analysed by using GeneMapper 3.7. All samples were run in duplicates.

Sequencing reactions

Sequencing reactions were performed using the BigDyeTM Deoxy Terminator Sequencing Kit v.1.1 according to the manufacturer’s instructions on an ABI3130-Avant Genetic Analyzer system. For the lane tracking and first pass base calling, Sequencing Analysis software v.5.2 was used. For the SNP characterisation, the sequences were assembled and aligned against the respective reference sequence by means of SeqScape 2.5.

Real-time PCR analyses

DNA concentrations were estimated using the PicoGreen® ds DNA quantification kit (Molecular Probes, The Netherlands), and the FLx800 microplate fluorescence reader (Bio-Tek® Instruments, Inc., Winooski, VT, USA). DNA quantities were related to a calibration curve made of four points (0.2, 2, 20 and 200 ng Lambda DNA). Real-time PCRs amplifying Adh1 (70), Adh1 (136), Ivr, Hmg, zein (110) and SSIIb target sequences were performed as described in original publications (see Table 1, A). All DNA samples from the worldwide variety collection (84 samples) were diluted to the same two concentrations (1 and 0.125 ng/μl). These two sets of dilutions were used as templates in the different real-time PCR systems. Measured Ct values for five endogenous RAs were obtained for both sets of dilutions (total 168 samples) and used for further analysis. Calibration curves for maize DNA copy number quantification were made of four points: 60,000>15,000 > 3,750 > 937.5 haploid T25 maize genomes (100%). For quantifying the GM DNA, the calibration curves were made of five points: 1,600 > 800>200 > 50>2.5 haploid genomes. The calculations of haploid genome copy number were based on extrapolation from the 1C mass of maize: 1 ng of maize DNA equals to approximately 385 maize haploid genomes [26]. For evaluation of the SNP identified in the zein gene sequence, four samples were selected giving different results of sequencing analysis and analysed in five dilutions for all RAs. All samples were analysed in at least two parallels of each DNA dilution.

Statistical analysis of real-time PCR data

For each reference system, a standard curve was constructed by plotting Ct values against log10 of DNA copy numbers of known standards and fitted by linear least square regression. The amplification efficiency in the individual sample was examined by calculating the slopes using the Ct values of different dilutions (1 and 0.125 ng/μl) and the log copy number values. Samples that show bad amplification efficiency in all RAs were completely excluded from further evaluation. Ct values of both sets of dilutions were converted to copy numbers using standard curves. The GeNorm software was used to estimate the stability of the RAs. Copy numbers were used as the input data according to the request of GeNorm. The application automatically calculates a “gene-stability measure” M for all investigated genes in a given set of samples. The parameter M is defined as the average pair-wise variation in measured copy numbers of a particular gene with all other genes of interest [27]. The RA is considered more or less reliable based on the M value; RAs with the lowest M value are more stable than those with higher M value. A genewise normalisation of copy numbers was performed in datasets of both dilutions. The highest copy number value was used as the normalisation factor (highest copy was set to 1). Normalised data were then visualised in two different ways using “R”, a free software environment for statistical computing and graphics. Scatterplots were constructed for each pair of RAs to estimate stability of target copy number estimates using an independent approach; copy numbers obtained in analyses of one gene were plotted against copy numbers obtained in analyses of another gene. Boxplots were constructed to allocate the variability in performance of RAs to individual samples. An outlier in copy number estimation is defined as a true positive, when the overall sample variability in copy number obtained by different RAs was larger than average and when the efficiency of amplification of the RA was less than 15% deviating from the efficiency of amplification of reference material.

Results

Sequence variation analysis through CSCE screening

Screening of the Ivr, Hmg and SSIIb target taxon genes by means of CSCE revealed presence of a single major peak in all of the examined pools. The peaks occurred at positions corresponding to 120, 190 and 380 bp, respectively, for the three targets. The position of the peaks was consistent in all the pools and in all dilutions of the PCR fragments. The presence of a single or of multiple peaks was identified based on the analysis of samples with optimal dilution e.g. displaying fluorescent peaks comparable in height with this of the size marker fragments (Fig. 1). The presence of a single clear and sharp peak corresponds to the homoduplex, which means absence of nucleotide variation in the sequences of the analysed Ivr, Hmg and SSIIb amplicons present in the pools. Amongst the individual genotypes, no nucleotide variation was observed for the conventional varieties and lines as well as for MON 810 varieties.

Fig. 1
figure 1

Results of conformation sensitive capillary electrophoresis screening for the presence of single nucleotide polymorphisms within a pool of relevant polymerase chain reaction fragments, shown as peaks (units of fluorescence, y-axis) in function of the fragment length (x-axis). The fluorescent peaks of the respective fragments are black (FAM-labelled fragments) or grey (HEX-labelled fragments). The major peaks corresponding to the heteroduplexes are indicated with a black arrow, the minor peaks indicating the presence of heteroduplexes (where present) are indicated with a transparent arrow. The transparent peaks correspond to the size marker. The length of the size marker is per step of 50 base pairs (bp), starting with 100 bp (first length marker peak, X-axis on the left) and ending with 450 bp (last length marker peak, X-axis on the right)

Full size image

Multiple peaks were observed in the Adh1 and zein patterns of some of the pools, while in other pools, a single major peak corresponding to the homoduplex was observed. The position of the peaks was consistent in all analysed pools. The major peaks corresponding to the homoduplex occurred at approximately 365 and 320 bp for the Adh1 and zein target sequences, respectively (Fig. 1, upper patterns of Adh1 and zein panels). A second minor peak occurring in the proximity of the major one was identified in the Adh1 pattern of some of the pools (Fig. 1, lower pattern of Adh1 panel). In the zein pattern, four minor peaks with lower mobility than the major one were observed (Fig. 1, lower pattern of zein panel). The presence of more than one fluorescent peak in the CSCE analysis indicates the presence of nucleotide variation between the individuals present in the pool. In case of heteroduplex formation, four peaks should normally be observed. However, sometimes the heteroduplexes migrate together with the homoduplexes and cannot be sufficiently distinguished. Presence of more than one peak is usually sufficient to conclude that nucleotide variation exists [28]. The individuals from the pools displaying the presence of a heteroduplex were further analysed by sequencing to confirm and characterise the observed nucleotide variation. In addition, one individual sample from each pool displaying a single peak was analysed to determine the type of the homoduplex sequence. This has been done for both Adh1 and zein fragments.

Characterization of observed SNP alleles

SNP alleles in Adh1 (70) in conventional and MON 810 varieties

Nucleotide sequences of the individuals from the pools displaying presence of multiple peaks were analysed. Sequence alignments were done against the reference sequence corresponding to the F allele of Adh1 (X04049) [29]. The alignments revealed that the nucleotide sequence for 39 of the analysed genotypes corresponds to the F allele of Adh1. In four of the lines, a heterozygous SNP C/T at position 2039 was observed. In one of the analysed individuals, a homozygous SNP (C > T) was observed (Table 2).

Table 2 Summary of the observed and characterised single nucleotide polymorphisms (SNPs) related to different reference assays in the transgenic and conventional maize varieties and lines
Full size table

Sequencing analysis of the individuals from the pools displaying a single peak revealed that the homoduplex corresponds to the Adh1 F allele. The observed homozygous SNP (C > T) in one out of 44 sequenced genotypes matches the S allele of Adh1 [29].

Sequencing alignment of the obtained sequences and primers used in real-time PCR showed that the single base pair substitution is located in the forward primer attachment site at position +9 (Fig. 2). The observed SNP corresponds to the one described by Broothaerts et al. [21].

Fig. 2
figure 2

Schematic representation of single nucleotide polymorphisms along the Adh1 sequence and location of the single nucleotide polymorphism (SNP) along the sequence of the Adh1 (70) assay. The structure of the F allele of Adh1 is presented according to Sachs et al. [29]. PANZEA refers to nine inbred lines where 38 candidate SNPs were observed in the Adh1 sequence (available from http://www.panzea.org). The base pair substitution observed in this study is presented with an x along the gene sequence. The black cross indicates the SNP occurring in the Adh1 (70) assay and observed in the lines tested in this study. The location of the SNP along the real-time polymerase chain reaction Adh1 (70) assay is shown in details in the figure below. The forward primer sequence is reverse complement to the S allele of Adh1 (X04050) and the base pair in the forward attachment site at 5′ + 9 is A (upper figure). A single base pair substitution A > G or A > G/A occurs in this position of the primer (lower figure). The scale is proportional

Full size image

The nucleotide variation of the considered Adh1 amplicon sequence was studied in more detail in the MON 810 varieties. The sequence analysis revealed that five of the transgenic lines and four of the corresponding isogenic counterparts contain the sequence corresponding to the F allele. In five of the transgenic lines and one of the isogenic lines, the heterozygous SNP allele was observed. None of the varieties contained the homozygous S allele (Table 2). Sequencing analysis was performed on the regions flanking the Adh1 target sequence. The analysis further confirmed that these lines carry the F allele of Adh1. No other nucleotide variation was observed amongst the lines along the Adh1 sequence. The analysed region also covered the region of the transcription initiation, which was used for the design of two other real-time PCR Adh1 assays [11]. The region was uniform among the varieties, and no polymorphisms were found between the lines and the reference sequence (Fig. 2).

SNP alleles in zein (110) sequence in conventional and transgenic varieties

The nucleotide sequence of the zein fragment of the individual samples from the pools displaying multiple peaks was further characterised by sequencing. Sequencing alignments were performed against the sequence corresponding to the AF371266 allele. The alignments revealed that 84 of the conventional varieties correspond to this allele. In 18 of the analysed genotypes, a homozygous base pair substitution (C > G) was observed. In 16 of the genotypes, the heterozygous SNP allele was detected. The three genotypes were observed in the MON 810 and the isogenic lines as well (Table 2). A Basic Local Alignment Search Tool located the SNP at position 125 of AF371266. This corresponds to the regions between the forward primer and the probe attachment site of the zein (110) assay. The sequence of the zein (104) assay remained uniform amongst the examined varieties (Fig. 3).

Fig. 3
figure 3

Schematic representation of the observed single nucleotide polymorphisms in the zein target sequence along the real-time polymerase chain reaction (PCR) assays. The grey arrows and lines represent the primers and probe attachment sites of the zein (110) real-time PCR assay. The white arrows and lines represent the primers and probe attachment sites of the zein (104) assay. The observed single nucleotide polymorphism is presented with a black box on the DNA template

Full size image

Real-time PCR performance testing of reference assays on worldwide collection of varieties

To assess the overall performance of different RAs (Table 1, A), the collection of 84 lines was used. First, comparison of variability in estimated copy numbers and variability in efficiency of amplification was evaluated (Table 3). In the raw data, the difference between extreme Ct values was less than 1 Ct in case of Hmg and Adh1 (136) while higher than 2 Cts in Ivr and Adh1 (70) RAs. Further, the RA stability in real-time PCR amplification was evaluated using GeNorm statistics. Adh1 (70) gives the highest M value, showing low stability of this RA. The variability of amplification in Adh1 (136), zein (110), Ivr and Hmg assays was low in the analysed varieties (Fig. 4).

Table 3 Real-time polymerase chain reaction amplification of the six maize reference systems on 84 maize cultivars representing different geographical origins
Full size table
Fig. 4
figure 4

Stability of reference assays amplification as evaluated in GeNorm. Average stability measure M (y-axis) was calculated for all tested reference assays, plotted on the x-axis. Results from the analyses of the 1 ng/μl DNA dilution set

Full size image

After the genewise normalisation of copy number data, scatter plots were constructed for each pair of assays to show the correlation between copy numbers obtained with two different RAs. Best correlation of results was obtained when using zein (110), Hmg and SSIIb RAs. Extremely dispersed scatter plots and correlation coefficients of less than 0.5 were calculated for Adh1 (70) RA copy number estimations versus all other RAs included in the analysis (Fig. 5).

Fig. 5
figure 5

Comparison of performance of reference assays in worldwide collection of varieties. Each scatter plot was constructed by plotting the copy numbers determined with one reference assay (x-axis) against the copy numbers determined with another reference assay (y-axis) after the genewise normalisation of copy number data. The red line in each scatter plot denotes a perfect correlation (r = 1) of copy numbers between the two assays. The individual correlation coefficients between the assays are also shown in the squares diagonally opposite to individual plots

Full size image

Boxplots of normalised copy numbers in samples were constructed to allocate the variability in performance of the RAs to individual samples. Outliers found using this approach can be candidates for genotypes with sequence modification in the primer/probe target region or changes in target copy number in the genome. Following our control parameters, outliers were identified in Adh1 (70) and Ivr determined copy numbers only. Adh1 (70) was a negative outlier, meaning that numbers of copies are always underestimated, and Ivr was a positive outlier with overestimated copy numbers. In two samples (varieties 43 and 62), outliers for Ivr determined copy numbers were identified indicating potential duplication of the Ivr gene in those genotypes. In five samples (varieties 2, 36, 49, 61, 62 and 72), Adh1 (70) was identified as an outlier in target copy number estimation. No outliers were detected using our criteria in copy number estimation performed by other RAs.

In order to know whether the variability in performance of the RAs depends on the geographical origins (gene pools) of varieties, 73 maize cultivars used for grain production were grouped into five geographical origins corresponding to the main regions of maize culture in the world: South Africa, Asia, South America, North America and Europe. Table 4 shows that the difference of the mean normalised copy numbers between geographical groups is statistically non-significant (p < 0.05) in all analysed RAs if considering only samples with acceptable efficiency of amplification. These results prove our expectations that there is less variation related to the geographical origins (gene pools) than to RA groups. Similar analysis was performed with varieties grouped according to their use: grain and forage corn, sweet corn, ornamental and popcorn (Table 4). Also, in this setup, the differences in copy number estimations between types of use are insignificant for all RA considered.

Table 4 Mean copy numbers and the extreme copy number difference observed on the five maize reference gene systems according to geographical origin and grain utility
Full size table

Discussion

We aimed at studying the reliability of a number of existing maize reference PCR assays which are widely used in GMO testing labs. A reliable RA must be specific for the taxon of interest, uniform along the varieties within the taxon and with low and stable copy number. The specificity of maize RAs was already assessed in silico and experimentally prior to method validation [8]. The uniformity was our main concern for several reasons. First, the maize genome is highly diverse, and SNPs are very common. SNPs occur between the maize inbred lines, which are frequently used in breeding programmes. Second, RAs used for the calculation of the relative GMO content in a sample have to amplify uniformly along the large number of commercially available varieties, which might be present in test samples of either seeds, grains, food or feed. In the EU only, more than 1,000 maize varieties are registered [10], and additional maize varieties might be imported from third countries [30]. The nucleotide stability of the RAs genome targets along the abundance of commercial varieties might not be sufficiently assessed prior to method validation. Variation in the assays’ performance along the varieties might be expected when the assay has to be applied at the analytical lab to analyse unknown samples. This means that some RAs might perform better than other. The commercially cultivated varieties are produced in different years, by different breeders. Given the natural base pair substitution rate of coding maize genome sequences (10−8 bp) [31], it is unlikely that for the production cycle of one variety (approximately 10 years), a significant number of nucleotide substitutions would accumulate in the maize genome potentially leading to non-uniformity of the nucleotide sequence especially in genes used for design of a RA. However, pre-existing nucleotide variation due to the presence of different SNP alleles among the breeding lines could give rise to variation in the nucleotide sequence among the varieties.

The uniformity of the maize RAs was studied in two manners: (1) straightforwardly, by analysing the nucleotide sequence of targeted regions, and (2) indirectly, by statistical analysis of the real-time PCR derived parameters. The first approach gives a direct view on the variation of the nucleotide sequences of the target regions and allows to locate an SNP, if any, along the primer and probe attachment site. The second approach evaluates the uniformity of the PCR performance along the varieties and additionally indicates if the RA amplification is positively or negatively affected. It is used to indicate either nucleotide variability or unstable copy number of the targeted gene.

Reference assays for official GMO control purposes are developed and tested on a limited number of varieties which, in some cases, can result in RAs with lower uniformity [21]. Here, we studied two groups of maize plant materials: conventional commercial varieties and lines bred in different parts of the world on one hand and commercial transgenic lines (MON 810) cultivated currently in Spain on the other hand. The collections of varieties included a wide number of commercial varieties and lines available at the market, representing the variability within the maize botanical taxon.

The nucleotide sequence analysis shows that four of the tested RAs (Adh1 (136), Ivr, Hmg and SSIIb) have genome targets uniform in their nucleotide sequence in the tested varieties (see Table 2). However, from these four assays, the Hmg assay seems to be the most stable, displaying consistent PCR performance and nucleotide sequence uniformity. This complies also with the data obtained by Broothaerts et al. [21] where Hmg showed low variation in the quantified copy number (coefficient of variation) along four maize conventional varieties and seven different transgenic events. The Hmg real-time PCR assay which forms a part of 14 official CRL methods for GMO quantification can be qualified as highly reliable.

The Ivr RA did not indicate non-uniformity of the nucleotide sequence; however, this assay displayed non-consistent PCR results, indicating that the target sequence might occur in more than one copy. Similar suggestions were done by Moor [32]. Ivr displayed overestimated and highly variable copy numbers along a number of varieties. However, the author could not conclude if the sequence exists in multiple copies, and any other literature references were not available.

The SSIIb RA displayed a uniform sequence and relatively good real-time PCR performance. The scattered plots of SSIIb compared to the rest of the tested RAs were not dispersed. However, SSIIb performed slightly worse in GeNorm analysis. Broothaerts et al. [21] demonstrated low variability in the amplification of the SSIIb target sequence comparable to the one of Hmg. Taken altogether, we could consider this method as reliable.

Nucleotide variation was found in the tested target sequences of the Adh1 (70) and zein (110) assays. The nucleotide variation observed in the zein (110) sequence was not located at primer or probe attachment sites, suggesting that no influence on the real-time PCR amplification might be expected. This was consistent with the real-time PCR results displaying a stable performance comparable to this of Hmg. Zein (110) assay partially overlaps with zein (104) assay previously designed by Hernandez et al. [11]. These authors already suggested that zein might occur in more than one copy in some lines. We did not observe any indication that this sequence might have higher copy number. The scattered plots were not dispersed, and the observed variation was very low. Zein RAs are not a part of CRL validated methods. However, both assays are reliable in terms of sequence uniformity.

The single base pair substitution observed in the primer attachment site of Adh1 (70) corresponds to the one reported by Broothaerts et al. [21]. The authors observed that this base pair substitution systematically occurs in homo- or heterozygous form in two non-transgenic varieties and seven transgenic varieties used to prepare CRMs. We analysed a larger number of varieties which potentially could be quantified in unknown test samples. Our results show that indeed the most frequently occurring allele in the tested collections is the one carrying the mismatch between the forward primer and the DNA template. The allele matching the primer sequence was found in only a single individual in a total of 58 characterised varieties and lines (Table 2). The analysis of real-time PCR results revealed that this Adh1 (70) assay performs insufficiently. Adh1 (70) was identified as a negative outlier in terms of measured copy numbers in five varieties. This demonstrates that the Adh1 (70) measured copy number could vary in a large range. Interestingly, no attribution to variability in performance of real-time PCR could be assigned either to geographical origin of varieties or to the type of use of variety. These are the first experimental data and evidences on this matter, which certainly allow dispersing some existing doubts. In summary, the Adh1 (70) assay appears to be highly unreliable due to the frequent occurrence of the SNP in the primer attachment site resulting in unpredictable variation of the measured copy numbers. In contrast to this assay, Adh1 (136) RA demonstrated nucleotide stability in all tested MON 810 varieties and stable PCR performance in all tested conventional maize lines.

Currently, four RAs (Hmg and different Adh1 assays) are part of official CRL validated methods. The methods for quantification of NK603, MON863 and GA212005, as well as a number of additional official methods for gene stacks, rely on the use of the Adh1 (70) assay [8] which was found in this study to be unreliable. Consequently, the reliability of these GMO quantification methods might be questioned.

With this study, we want to stress on the necessity to assess the uniformity of any newly developed RA prior to method validation. The occurrence of SNP variants in the genomic regions used to design a taxon or species RA and the frequency of SNP alleles, if occurring along suitable inbred lines, should be assessed in silico during the method development. In cases where this is not possible e.g. no sequence data available, the occurrence of SNP alleles has to be experimentally assessed in varieties/lines which might be used in GMO analysis. Two groups of such varieties could be distinguished: (1) varieties used to produce CRMs, also representing the varieties which would occur in the unknown samples, and (2) transgenic varieties containing the transgenic event which is targeted and one or more conventional varieties. The development of RAs for GMO quantification would significantly benefit from a harmonised requirement for testing the stable and uniform amplification along a wide range from 20 [8] to several hundreds of varieties [33]. The set of varieties selected to validate the uniformity of RA should represent the existing variation of the maize taxon at a certain moment. These criteria could be extended to other plant taxa relevant for GMO analysis as well.