• Open Access

Only half the transcriptomic differences between resistant genetically modified and conventional rice are associated with the transgene

Authors


Correspondence (Tel +34 972 41 98 52; fax +34 972 41 83 99; email maria.pla@udg.edu)

Summary

Besides the intended effects that give a genetically modified (GM) plant the desired trait, unintended differences between GM and non-GM comparable plants may also occur. Profiling technologies allow their identification, and a number of examples demonstrating that unintended effects are limited and diverse have recently been reported. Both from the food safety aspect and for research purposes, it is important to discern unintended changes produced by the transgene and its expression from those that may be attributed to other factors. Here, we show differential expression of around 0.40% transcriptome between conventional rice var. Senia and Senia-afp constitutively expressing the AFP antifungal protein. Analysis of one-fifth of the regulated sequences showed that around 35% of the unintended effects could be attributed to the process used to produce GM plants, based on in vitro tissue culture techniques. A further ∼15% were event specific, and their regulation was attributed to host gene disruption and genome rearrangements at the insertion site, and effects on proximal sequences. Thus, only around half the transcriptional unintended effects could be associated to the transgene itself. A significant number of changes in Senia-afp and Senia are part of the plant response to stress conditions, and around half the sequences for which up-regulation was attributed to the transgene were induced in conventional (but not transgenic) plants after wounding. Unintended effects might, as such, putatively result in widening the self-resistance characteristics because of the transgene in GM plants.

Introduction

The use of genetically modified organisms (GMO) in basic research and crop development is very extensive nowadays. Genetically modified (GM) crops are subject to different legislation worldwide to cover aspects of consumer safety and protection; authorized GM events have been shown to be equivalent to non-GM comparable varieties by targeted analysis of compounds that are relevant to the species and trait that is introduced (OECD, 1993). Additionally, more unbiased profiling technologies, such as metabolomics, proteomics and transcriptomics, have been used to extend the scope of comparative analyses. Various studies report the very limited difference between GM and comparable non-GM plants, in contrast to the major impact of mutagenesis techniques used to generate diversity in conventional breeding approaches (Batista et al., 2008), and more generally to the extensive variation between conventional varieties [see reviews in (Bradford et al., 2005; Chassy et al., 2008; Kok et al., 2008)]. A number of recent publications report the differences, always limited and of diverse identity, between specific GMOs and comparable non-GM plants [(El Ouakfaoui and Miki, 2005; Gregersen et al., 2005; Dubouzet et al., 2007; Cheng et al., 2008; Coll et al., 2008, 2010; Baudo et al., 2009; Abdeen et al., 2010; Barros et al., 2010) (Arabidopsis, maize, rice, soybean and wheat transcriptomes); (Barros et al., 2010; A. Coll et al., 2010; Corpillo et al., 2004; Khalf et al., 2010; Lehesranta et al., 2005; Ruebelt et al., 2006) (Arabidopsis, maize, potato and tomato proteomes); (Catchpole et al., 2005; Kristensen et al., 2005; Baker et al., 2006; Beale et al., 2009; Zhou et al., 2009; Kogel et al., 2010) (Arabidopsis, barley, potato, rice and wheat metabolome)].

Unintended differences in transgenic and non-GM plants can be predictable or unpredictable as a function of whether they are expected and explicable in terms of the present knowledge of plant metabolism and physiology or whether they fall outside our present level of understanding (Cellini et al., 2004). They may occur as a consequence of pleiotropic effects of the integrated DNA on the host plant genome, related to both the transgene coding region and regulatory elements (Filipecki and Malepszy, 2006; Miki et al., 2009). As an example, they can arise as a result of transgene products interacting with the regulation of other genes or the activity of other proteins. These changes depend on the presence and/or expression of the transgene.

They may also occur as a consequence of host gene disruption or DNA sequence rearrangements at the insertion site (Forsbach et al., 2003) or because of interactions between transgene elements and proximal host sequences (e.g. the coding sequences surrounding the insertion site may fall under the influence of transgene promoters leading to sense or antisense transcripts). The integration of a transgene in the host genome is usually a random process, but with the preference for gene-rich regions (Koncz et al., 1992), disruption or modification of the activity of genes might occur. GM plants with undesired transgene position effects are discarded and other events used. These are event-specific unintended effects.

The host plant genome may also be affected by the process to obtain GM plants. Most transformation techniques involve tissue culture, plant cell dedifferentiation, plant regeneration and acclimatization. In vitro culture technology gives rise to somaclonal variation (Larkin and Scowcroft, 1981), which mainly relies on genetic changes (Ngezahayo et al., 2007; Noro et al., 2007), but also on epigenetic and karyotypic changes. One example of changes caused by in vitro culture is potato control and tissue culture-derived tubers, which have been shown to have compositional differences (Shepherd et al., 2006). Stress factors are the main factors described as potentially responsible for somaclonal variation. The host plant genome may also be influenced by the infection with Agrobacterium, particle bombardment or other transformation processes. AFLP and RAPD analysis of transgenic rice plants produced by different transformation techniques have verified the presence of genomic changes (Labra et al., 2001).

Especially in the context of public concern regarding health and environmental safety of GMOs, it is essential to differentiate the real extent of unintended changes produced by the transgene and its expression from (i) event-specific changes that can be minimized or avoided through proper selection and (ii) those which are a consequence of technologies (such as tissue culture) widely used with commercial applications such as freeing plants from viruses or plant micropropagation. In contrast to transgenic plants, conventionally bred crops have not given rise to public safety concerns (Cellini et al., 2004). For research purposes, it is often important to consider the causes of unintended effects of the transgene insertion and expression even if they fall within the natural variation of the species, to avoid misinterpretation of the results, for example in the evaluation of the expression patterns of transformed genes under study.

Here, we used rice as a model cereal to assess unintended differences between GM and conventional isogenic plants and to evaluate their possible causes. Rice is commercialized as inbred lines (not hybrids) often suitable for genetic transformation and regeneration. Both transgenic and isogenic lines can be obtained, and they could have commercial interest. We analysed transcriptomic differences between the conventional line Senia and the GM, fungal-resistant rice line Senia-afp (which constitutively expresses the antifungal protein AFP) obtained by Agrobacterium transformation. We further quantified the impact of different factors on transcriptomic regulation, in particular the insertion site and associated rearrangements, the processes to produce transgenic plants (e.g. in vitro culture) and the transgene.

A detailed understanding of GMO unintended effects and their causes is needed to develop strategies to minimize unintended changes to transgenic plants. In the context of the public concern regarding health and environmental safety of GMOs, this study aims to give a more accurate perspective of the unintended differences between GM and non-GM plants.

Results

Transcriptomics comparison of conventional and AFP-producing GM rice by microarray hybridization

We initially assessed transcriptional differences between the conventional commercial rice line Senia and the GM line, Senia-afp. Senia-afp was produced by Agrobacterium-based transformation of Senia and is self-resistant to fungal infections by expression of the antifungal AFP protein, under the control of the constitutive promoter ubi-P (Coca et al., 2004). In particular, we used the homozygous stabilized line of event R25.14 that has a single and full-length copy of the transgene cassette per haploid genome.

To specifically focus on transcriptional changes related to the transgenic character and avoid the effect of unrelated factors, our approach was based on in vitro-cultured plantlets under highly controlled experimental conditions. Furthermore, three biological replicates, each from leaves of ten different plantlets, were independently analysed in three microarrays per variety. Microarray data are available at the European Bioinformatics Institute (EMBL-EBI) ArrayExpress repository database under accession E-MEXP-2730. The data obtained in the three replicates were collectively analysed using the Robust Multichip Average (RMA) software for gene expression summary values. The estimated log2-fold changes and log odds values (T-test) for differential expression of the data are shown in Figure 1.

Figure 1.

 Volcano plot representation of changes in gene expression in Senia-afp and Senia rice lines. Each point represents one gene in the rice Affymetrix microarray. The log odds for differential expression of all genes (estimated from the Robust Multichip Average analysis of the data) are plotted against the estimated log2-fold changes. Vertical bars indicate the twofold increase or decrease in a given sequence. Sequences further analysed by RT-qPCR are in bold.

The data were subsequently filtered by considering only sequences with higher than twofold levels, Student test P-values below 0.05 and fluorescence intensity values (in the rice line with the highest expression) above 200 units. Microarray results were subsequently validated by reverse transcription–real-time polymerase chain reaction (RT-qPCR). A total of 44 sequences were selected that corresponded to those with the highest fold-values (above fivefold) plus a random selection of sequences regulated down to twofold. Specific qPCR assays were designed and optimized to target each selected sequence. The sequence with probe set ID Os.22606.1.S1_at was not suitable for designing a qPCR assay in the Beacon Designer software default settings and was discarded. qPCRs produced unique amplicons, as assessed by melting curve analysis, and had linearity and efficiency values above 0.99 and 0.90, respectively (mean R2 = 0.998 ± 0.002; mean E = 0.966 ± 0.026). Messenger RNA levels of the 43 selected sequences were assessed in triplicate samples of Senia-afp and Senia leaves by RT-qPCR. The three reference genes, β-actin, elongation factor (EF-Iα) and 18S ribosomal RNA, were used to normalize the data. They had stability values (M) below 0.5 (geNORM v3.4 statistical algorithm, Ghent University, Ghent, Belgium) so were suitable for standardizing gene expression under these experimental conditions. For each sequence, the RT-qPCR results in Senia-afp and Senia were statistically analysed by the Student test coupled to the Benjamini and Hochberg False Discovery Rate multiple testing correction, with P < 0.05 (Table 1). Up to 82% of the RT-qPCR results were in agreement with microarray data (mean P = 0.017 ± 0.016), which is within the expected range (Dallas et al., 2005).

Table 1.   Validation of microarray data Thumbnail image of

The rice GeneChip® microarray used contains probes to query 51 279 transcripts, from which approximately 48 564 represent japonica cultivars. Among the sequences analysed, 196 were differentially expressed in Senia-afp and Senia with at least a twofold increase or decrease in the level of a given transcript (Supplementary Table S2). A total of 12 sequences were represented by two or three probe sets, such that 183 genes were regulated in GM and conventional rice. This corresponded to 0.40% of the analysed sequences. Most regulated sequences were overexpressed in Senia-afp (162 sequences), whereas only 34 were down-regulated in the GM line (i.e. c. 83% and c. 17% regulated sequences, respectively). Eighteen sequences (17 up-regulated and 1 down-regulated) were differentially expressed five- to 13-fold, and an additional sequence was highly up-regulated in Senia-afp (above 1000-fold), with background levels in Senia samples. This sequence corresponds to hygromycin B phosphotransferase, used in Senia-afp as the selection gene under the regulation of the cauliflower mosaic virus 35S promoter. Because it was not an unexpected effect, the hgr sequence was discarded from further expression analyses.

We identified a representative public ID and description of each differentially expressed sequence and determined the biological process, molecular function and cellular component in which these sequences were potentially involved, using the Gene Ontology annotation files (Table 1 and Supplementary Table S2). Among the regulated sequences, 175 were assigned a biological process GO term. Gene annotation enrichment analysis showed significant overrepresentation of sequences related to the response to chemical stimulus and catabolic and glucose metabolic processes (Table 2). Different tools for functionally annotating gene sets and identifying significantly enriched GO categories can give different results (van den Berg et al., 2009), so they should be considered evaluative and not conclusive.

Table 2.   Singular enrichment analysis of sequences differentially expressed in Senia-afp and Senia plants using the AgriGO tool
GO IDGO term (biological process)Annotated/total no. in query listAnnotated/total no. in backgroundAdjusted P-value*
  1. *Based on the Fisher’s statistical method and the Yekutieli FDR multiple test correction method (Benjamini and Yekutieli, 2001).

GO:0070887Cellular response to chemical stimulus6/1757/30 2418.1 e−11
GO:0009056Catabolic process13/175553/30 2410.0053
GO:0044248Cellular catabolic process10/175432/30 2410.029
GO:0006006Glucose metabolic process6/175180/30 2410.042
GO:0042221Response to chemical stimulus9/175394/30 2410.042

Evaluation of the impact of the transformation process on transcriptomic differences

Transgenesis unintended effects may be a consequence of the process applied to transform and regenerate transgenic plants, and it is well known that tissue culture-based technologies give rise to genomic variation. For Agrobacterium-mediated transformation, the process involves cell dedifferentiation to callus, Agrobacterium infection, cell growth in different media (including selection) and regeneration and acclimatization of GM plants. To assess the weight of the transformation process on unintended effects of transgenesis, we analysed null segregants, i.e. non-GM plants that had been subjected to the same transformation process as Senia-afp. After Agrobacterium tumefaciens infection of rice embryogenic callus and selection of hygromicine-resistant cells, regenerated transgenic plants are hemizygous for the transgene. Self-pollination of Senia-afp R25-14 gave rise to homozygous (i.e. the previously analysed Senia-afp event R25-14 stabilized line), hemizygous and non-transgenic plants (25%, 50% and 25%, respectively, according to Mendelian distribution of the transgene). The latter non-GM plants, Senia-afp(−), went through the whole transformation process but did not contain the transgene.

A total of 130 Senia-afp R25-14 hemizygous grains were germinated in vitro, and leaves were individually frozen in liquid nitrogen. A small portion of each sample was analysed by qPCR assays targeting afp, and 31 non-transgenic plants were identified. They were grouped into three biological replicates of ten plants each for further RT-qPCR analysis of the 34 sequences previously shown to be regulated in Senia-afp R25-14 and Senia by microarray hybridization and RT-qPCR. The number of sequences was above the minimal representative sample size required (assuming a margin error of 15% and a confidence level of 95%), and they represented around 20% of those regulated in Senia and Senia-afp, according to the microarray results. The β-actin, EF-Iα and 18S rRNA reference genes (M-values <0.5 in these samples) were used for normalization. Pair-wise comparison of their expression levels in Senia-afp(−) and Senia is shown in Table 3. As expected, most sequences were similarly expressed in Senia and the non-GM Senia-afp(−) line. However, 12 of 34 sequences showed differential expression in Senia and Senia-afp(−), indicating that their regulation in Senia and Senia-afp was not directly linked to the presence or expression of the transgene but to the transformation process.

Table 3.   RT-qPCR expression patterns of a selection of sequences differentially expressed in Senia-afp and conventional Senia plants. T-test-based comparison of their expression levels in Senia-afp(−) vs. Senia; and in three different events of Senia-afp (R25-14, R25-15 and R25-12) vs. Senia. Pair-wise comparisons giving T-test values with statistical significance (P < 0.05) are highlighted in grey. From these data, the differential expression of each sequence in genetically modified and conventional rice was associated to the transformation process (t, dots), event-specific insertion site and associated genomic rearrangements (e, white) or the transgene (T, dense dots) Thumbnail image of

Sequences regulated in Senia and Senia-afp(−) were in most cases similarly expressed in Senia-afp(−) and Senia-afp, and sequences similarly expressed in Senia and Senia-afp(−) were generally regulated in Senia-afp(−) and Senia-afp. The differential regulation of OsA10, OsA11, OsA26 and OsA35 in Senia compared to Senia-afp and Senia-afp(−) was at different levels. OsA32, OsA33 and OsA37 had intermediate expression levels in Senia-afp(−), which resulted in statistically similar values for Senia and Senia-afp(−) and also for Senia-afp(−) and Senia-afp. Note that their variation was below 2.85-fold in the microarray.

Evaluation of transcriptomic differences between Senia-afp and Senia in different GM events

Unintended gene expression differences in Senia-afp and Senia may also originate from host gene disruption, DNA sequence rearrangements at the insertion site or interactions between transgene elements and proximal host sequences. As different GM events arise from independently transformed cells and have different insertion sites and associated rearrangements, we used three different GM events from Agrobacterium-mediated transformation of the same afp construct used above to evaluate to what extent GMO unexpected effects could be associated to these event-specific factors.

Senia-afp lines R25-14, R25-15 and R25-12 were obtained in parallel in the laboratory. They all had a single copy of the transgene (as confirmed by Southern hybridization, J. Messeguer, pers. commun.) and stable lines, homozygous for the transgene, were subsequently obtained. Seeds of each line were in vitro cultured, and leaves were collected at the V2 stage (triplicate samples of ten plants each) for further gene expression analysis. Messenger RNA levels of the afp transgene were assessed by RT-qPCR in all nine samples, using the same β-actin, EF-Iα and 18S rRNA reference genes, which had stability values (M) below 0.5 (geNORM v3.4 statistical algorithm) in these samples. The anova statistical algorithm showed that the three Senia-afp events expressed similar levels of afp mRNA (P-value = 0.096). This excluded any further difference among events being because of different transgene expression levels.

The expression levels of the same 34 sequences with differential expression in Senia and Senia-afp event R25-14 were further assessed in Senia-afp events R25-15 and R25-12 samples by RT-qPCR. Twenty-three of these sequences, i.e. around 70% of the analysed sequences, were regulated in all three events (Table 3), indicating their regulation was independent of the insertion site. However, the expression of 11 sequences was not consistently regulated in the three analysed events. Specifically in Senia and Senia-afp event R25-14, five were differentially expressed so their sole dependence on the presence and expression of the transgene could be discarded. Among sequences differentially expressed in an event-specific manner, OsA30 and OsA32 were located on chromosome 4 (positions 29 280 748–29 281 890 and 33 988 248–33 991 626, respectively), whereas OsA12, OsA14 and OsA41 were on chromosomes 5, 2 and 6, respectively. As shown in Table 1, OsA32 was annotated (oryzain β) but OsA12, 14, 30 and 41 corresponded to expressed proteins with no associated putative function or regulatory pathway.

Sequences induced by wounding and the transgene

Overall, the regulation of 17 of 34 of the analysed sequences could not be associated to the insertion site or the transformation process, so it was attributed to the transgene itself. They were all induced in the GM line. We further assessed the expression patterns of these 17 sequences in non-GM and Senia-afp plants in response to abiotic stress conditions using wounding. Three biological replicates of in vitro-cultured V2 plants, with or without wounding, were analysed by RT-qPCR with the same reference genes as before. The M-values were <0.5 in these samples, which proved their suitability for normalization purposes. As shown in Figure 2, nine sequences were induced in non-GM Senia plants after wounding, while no significant changes were observed for the eight other sequences. In contrast, these sequences were not induced in Senia-afp upon wounding, and two were down-regulated in stress conditions. Control RT-qPCR assays targeting the afp transgene confirmed its similar expression levels in wounded and control Senia-afp plants.

Figure 2.

 Transcriptional response to wounding of 17 sequences with differential expression in Senia-afp and conventional Senia plants associated to the transgene. Pair-wise comparisons giving T-test values with statistical significance (P < 0.05) are highlighted in black (sequences induced in the first versus the second plant/conditions) or grey (sequences down-regulated in the first versus the second plant/conditions).

Discussion

The recent research on assessment of unintended effects of GMO has provided details of transcriptomic, proteomic and metabolomic differences between conventional and a series of transgenic plants, including different species and traits (Hoekenga, 2008; Kok et al., 2008). They consistently show a low degree of unintended variation between GMO and non-GM comparable plants, with no preferential functional categories regulated and at expression levels usually falling into the generally acceptable range, naturally occurring in different cultivars or ecotypes of the species. While the differences between GM and non-GM plants are minor, a detailed understanding of their causes would facilitate the development of strategies to minimize them and would put the social debate on substantial equivalence of GM and conventionally bred plants in a more accurate context.

Transcriptomic comparison of the conventional rice line Senia and the transgenic stabilized line Senia-afp (constitutively expressing the antifungal protein AFP) showed differential expression of around 0.40% in the analysed transcripts. The effects of varying environmental conditions were minimized by analysing leaves of V2 plantlets grown under in vitro homogeneous conditions. The magnitude of regulated transcripts is similar to those reported for other species and transgenes. Minor transcriptional regulation has been reported in the model Arabidopsis species, expressing various marker (El Ouakfaoui and Miki, 2005), herbicide resistance (Abdeen and Miki, 2009) or drought tolerance-related transcription factor (Abdeen et al., 2010) genes. Commercialized GMOs such as insect-resistant MON810 maize have been shown to have very limited and variety-dependent transcriptome regulation (Coll et al., 2008). This event, and also glyphosate-tolerant NK603 maize, had fewer transcriptomic changes than those produced by environmental and conventional breeding factors (Barros et al., 2010; Coll et al., 2010). Similar results have been reported for rice plants producing CsFv antibodies (Batista et al., 2008) and anthranilate synthase α subunit (Dubouzet et al., 2007); glyphosphate-tolerant soybean (Cheng et al., 2008) and wheat plants either producing phytase (Gregersen et al., 2005) or a glutelin subunit (Baudo et al., 2009). So they were considered adequate for use as models to assess the possible causes of transcriptomic unintended effects.

Further analysis of one-fifth of the sequences regulated in Senia and Senia-afp indicated that the transgene is not the sole cause of unintended differences between GM and non-GM isogenic rice plants.

By comparison of transgenic Senia-afp and non-GM Senia-afp(−) plants that had undergone the same Agrobacterium-based transformation process, sequences were identified whose differential expression was definitely not associated to the presence and/or expression of the transgene. Our results show that it is possible to separate the effects of the transgene from those of the transformation process when the appropriate comparator is used, and that the transformation process caused around 35% of the unintended effects observed in Senia-afp and Senia, with 21% and 52% being the lower and upper limits of the 95% confidence interval (CI) calculated according to Newcombe (1998).

The introduction of a transgene into a plant genome usually involves tissue culture technologies to induce dedifferentiation of plant tissues and allow selection and regeneration of an intact plant from a single GM cell. These procedures are associated with stresses such as wounding, osmotic stress, insufficient nutrient, growth regulators and antibiotics (Carman, 1995) and have been shown to induce genetic changes (Filipecki and Malepszy, 2006; Latham et al., 2006). Plant in vitro culture is used as a method to generate genetic variation for breeding purposes (Larkin and Scowcroft, 1981; Veilleux and Johnson, 1998), and GM plants obtained through callus had higher variability than those obtained using the floral dip technique, which does not include tissue culture (Labra et al., 2001). Agrobacterium and other pathogen infections may also cause mutations (Budziszewski et al., 2001; Lucht et al., 2002). This is in agreement with our results associating as much as c. 35% of transcriptional unintended differences between GM and conventional rice to the transformation process. Some sequences differentially expressed in Senia-afp and conventional Senia-afp(−) plants were also regulated in Senia and all three Senia-afp events examined. We could not discount the transcriptional regulation being linked to epigenetic changes (Kaeppler et al., 2000), preferential changes in certain sequences associated to tissue culture technologies and/or genetic differences in the individual Senia plant, which gave rise to Senia-afp and the Senia line as a whole.

We estimate that around 15% [95% CI: (6%, 30%)] of the transcriptomic differences between Senia-afp R25-14 and the conventional comparator Senia line were event specific. They mostly affected unannotated sequences, with alignment to the rice genome predicting their position on different chromosomes. This means that these position effects were not restricted to sequences in the single transgene insertion site. Agrobacterium T-DNA insertions usually result in small deletions of plant DNA at the insertion site and insertion of superfluous DNA, but they can also be associated with large-scale rearrangements or deletions (Latham et al., 2006). Interactions between the transgene and proximal host sequences can also result in unintended effects, which may also have more distant downstream effects (Miki et al., 2009).

Differential expression of only half the sequences [95% CI: (34%, 66%)] regulated in Senia-afp and Senia could, therefore, be directly associated with the transgene or its expression; this represented around 0.20% of the transcriptome in this GMO. Senia transgenic plants harbouring the same construct as Senia-afp, with the only exception being a weak promoter driving the expression of afp, had around 1000-fold lower levels of afp than Senia-afp and, compared to Senia, had similar mRNA levels for 9 of the 17 sequences for which regulation in Senia and Senia-afp was attributed to the transgene (data not shown). This suggests this regulation was caused by the afp expression level.

As in other GM plants [see e.g. (El Ouakfaoui and Miki, 2005; Cheng et al., 2008; Coll et al., 2008)], differentially expressed sequences in Senia-afp and Senia rice lines are involved in multiple biological processes. Regulated sequences were mainly involved in response to chemical stimulus, catabolic processes and glucose metabolic processes. The transformation techniques, the insertion site and the transgene do not seem to preferentially regulate any specific type of sequences among those analysed by RT-qPCR, although the small number of sequences in each group should be noted.

Among the nine sequences grouped under response to chemical stimulus, five are involved in the stress response, i.e. peroxidase reactions and response to oxidative stress (BU673129, putative peroxidase; D21280.1, glyceraldehyde-3-phosphate dehydrogenase; D10425.1, catalase), water deprivation (AK107749.1, oryzain beta) and defence response to bacteria and fungi (AY050642.1, seedling pathogenesis-related protein PR4). The last four genes in this GO term (AK058556.1 and AK062728.1, auxin-responsive SAUR gene family members; 9636.m00152, similar to auxin-induced SAUR-like protein; and 9634.m02249, auxin-responsive Aux/IAA gene family member) are related to the response to auxin. Auxin homoeostasis can regulate the activation of defence responses in various plant species, including Arabidopsis (Navarro et al., 2006; Wang et al., 2007) and rice (Domingo et al., 2009).

Catabolic and glucose metabolic processes are overrepresented GO terms as well. These include the five genes related to the stress response and an enzyme of the phenylpropanoid pathway (BI807677, phenylalanine ammonia-lyase activity, PAL, involved in defence reactions). They also include four genes involved in glycolysis (BU667041, fructose-1,6-bisphosphatase, putative; AK062270.1, aldolase C-1; BI810367, putative fructose-bisphosphate aldolase; and D21280.1), two involved in cellulose biosynthesis (AK120236.1, CESA3, cellulose synthase; and AK067850.1, RSW1-like cellulose synthase catalytic subunit), and an enzyme of the tricarboxylic acid cycle TCA (AK106451.1, phosphoenolpyruvate carboxylase). Plant stress responses are associated with a wide array of mechanisms involving increased demands for energy and its redistribution. Different publications profiling the plant response to pathogens show regulation of primary metabolism genes that likely play a role in providing energy for the resistance response [for review, see (Bolton, 2009) and references therein]. Plant respiration is stimulated during the stress response. This involves glycolysis (a cytosolic pathway converting glucose to pyruvate and resulting in a small net gain of ATP) and the mitochondrial TCA cycle (generating reducing equivalents that are used by the electron transport chain to fuel ATP synthesis). Reduction in photosynthesis and changes in carbohydrate metabolism are also well documented.

Through expression of afp, Senia-afp plants are protected against infection by Magnaphorte oryzae (Coca et al., 2004). Here, we show the unintended modulation of the expression of a set of genes which have been related to the response to stress, with the Senia-afp gene regulation partly mimicking an abiotic stress response on wounding. In response to wounding, conventional Senia plants overexpressed around half the sequences that were up-regulated in Senia-afp vs. Senia in a transgene-dependent manner. Remarkably, these sequences were not further up-regulated in Senia-afp upon wounding. From these results, we consider that unintended overexpression of genes involved in the stress response could potentially enhance the resistance of certain GM plants. In a defensome study of GM Senia rice overexpressing the antimicrobial peptide Cecropine A, Campo and collaborators (Campo et al., 2008) showed altered levels of genes involved in protection against oxidative stress, which correlated with the tolerance of Cecropine A plants to diverse bacterial and fungal pathogens and oxidative stress. The expression of genes encoding antimicrobial peptides such as AFP and Cecropine A in transgenic plants could involve unintended modifications in the transcription levels of host stress-related genes and potentially extend resistance to a broader range of stress conditions.

In conclusion, differences between transgenic Senia-afp and conventional Senia lines were around 0.40% of the transcriptome. Around 35% of these differences were attributed to the procedure followed to obtain GM plants. In vitro culture techniques, cell dedifferentiation and plant regeneration are used to obtain GM plants but are also routinely used for micropropagation, to obtain virus-free material and numerous other commercial applications. In addition, around 15% of the transcriptomic changes in Senia-afp and Senia plants were event specific, associated with the insertion site and genome rearrangements occurring during transformation. The generation of many independent transgenic lines would, therefore, create a range of position effects and allow for the eventual identification of lines in which unfavourable position effects are absent. Finally, only around half the sequences regulated in GM Senia-afp and conventional lines were associated to the transgene. A significant number of the sequences unintentionally regulated in Senia-afp were involved in the response to stress and could, as such, enhance the resistance characteristics of the transgene in GM plants. This should be taken into account in food safety assessment studies.

Experimental procedures

Plant material

GM rice (Oryza sativa L. cv. Senia) lines resistant to fungal infections through constitutive expression of the Aspergillus giganteus antifungal protein AFP were used (Senia-afp) (Coca et al., 2004). Transgenic lines were produced by Agrobacterium-mediated transformation using three independent transformation events, Senia-afp R25.12, Senia-afp R25.14 and Senia-afp R25.15, all harbouring a single copy of the transgene. The expression cassette included the maize ubiquitin constitutive promoter (ubi-P) and the A. tumefaciens nopaline synthetase terminator (nos-T) driving afp expression, and the hygromycin phosphotransferase (hptII) selection gene. Normally, stabilized plants homozygous for the transgene were used, with the commercial japonica line Senia used as the conventional isogenic line. In addition, non-transgenic plants obtained by self-pollination of hemizygous Senia-afp R25.14 plants [Senia-afp(−)] were included in the study.

Seeds were surface sterilized, germinated in vitro and sown in glass tins, three in each, containing 100 mL sterile MS medium (Murashige and Skoog, 1962) supplemented with 3% sucrose and 0.7% agar. They were incubated in a culture chamber at 25 ± 1 °C with a photoperiod of 16 h light/8 h dark under fluorescent Sylvania Cool White lamps. All plants were grown together in the same chamber with the glass tins placed randomly. Rice plantlets were sampled at the vegetative two-leaf stage (V2), immediately frozen in liquid nitrogen and stored at −80 °C. Each sample consisted of two leaves of ten plantlets, without lesions, taking three biological replicates. When required, two 0.5-cm-diameter circular wounds were applied per leaf, 4 h before sampling.

Genomic DNA extraction

Genomic DNA from 50 mg of T1 transgenic rice leaves was extracted using the commercial NucleoSpin® Plant II kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s instructions. It was quantified by UV absorption at 260 nm in a NanoDrop ND1000 spectrophotometer (Nanodrop technologies, Wilmington, DE). The OD 260/280 nm absorption ratios [mean and standard deviation (SD) = 1.93 ± 0.17] were used to confirm the purity of the DNA samples.

Total RNA extraction

Total RNA was extracted using a protocol based on the Trizol reagent (Invitrogen Life Technologies, Carlsbad, CA) and purified with the Qiagen RNeasy MiniElute Cleanup kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. It was quantified by UV absorption at 260 nm in a NanoDrop ND1000 spectrophotometer (Nanodrop technologies). Agarose gel electrophoresis and OD 260/280 nm absorption ratios (mean and SD = 2.09 ± 0.02) were used to confirm the integrity and purity of the RNA samples.

Microarray hybridization and analyses

The GeneChip® Rice Genome Array (Affymetrix, Santa Clara, CA) was used to search for transcriptome differences between Senia and Senia-afp R25.14. The Rice GeneChip contains sequences from the two most common rice cultivars, indica and japonica. It contains 54 168 probe sets to analyse 48 564 transcripts from the O. sativa japonica cultivar group. Sequence information for this array includes public contents from UniGene Build #52 (7 May 2004), GenBank® mRNAs (13 July 2004) and 59 712 gene predictions from TIGR’s osa1 version 2.0 release (http://www.affymetrix.com/).

Three GeneChips were employed to analyse three independent Senia and Senia-afp (R25.14) biological replicates. Hybridization and statistical analysis were performed at the Genomics Unit, Parque Científico in Madrid as previously described (Coll et al., 2008). Briefly, the integrity of total RNA samples was assessed by capillary electrophoresis using a Bioanalyser 2100 (Agilent Technologies, Palo Alto, CA). The GeneChip IVT Labeling kit (Affymetrix) was used to in vitro synthesize biotin-labelled cRNA from complementary DNA (cDNA) obtained with the One-cycle cDNA Synthesis kit (Affymetrix). Fifteen micrograms of biotinylated cRNA were fragmented into sequences of around 100 nt and these hybridized to the GeneChip Rice Genome Array (Affymetrix) in the GeneChip Hybridization Oven 640 (Affymetrix) for 16 h at 45 °C. The chips were subsequently washed and fluorescently labelled with phycoerythrin using the antibody amplification step in the GeneChip® Fluidics Station 450, and the fluorescence quantified using the GeneChip® 3000 scanner device. The data was extracted using the RMA software (Irizarry et al., 2003), which includes background adjustment, quantile normalization and summarization. Gene ontology analysis was performed using the AgriGO computational tool, considering probes with expression changes greater than twofold, P-values below 0.05 and minimal fluorescence signal intensity of 200 units.

Reverse transcription and real-time PCR amplification

The expression of 43 selected sequences, the 18S ribosomal RNA, β-actin and EF-Iα housekeeping genes and the afp transgene were assayed by RT-qPCR. Reverse transcription was performed on 2000 ng total RNA, previously treated with Turbo DNase (Ambion, Austin, TX) using 50 U of MultiScribe Reverse Transcriptase (Applied Biosystems, Foster City, CA) and random hexamer primers (Applied Biosystems) according to the manufacturer’s protocol. For each sample, cDNA was prepared at least in duplicate and the 43 sequences were analysed with all cDNA preparations. The absence of remaining DNA targets was confirmed by real-time PCR analyses of DNase-treated RNA samples.

qPCR assays targeting the 43 selected sequences were developed based on SYBR-Green technology. PCR primers were designed using the Beacon Designer 7.0 software (Premier Biosoft International, Palo Alto, CA), targeting the sequences used for generation of the GeneChip® Rice Genome Array. qPCR assays were performed in a 20 μL volume containing 1× SYBR Green PCR Master Mix (Applied Biosystems), the optimized concentration of primers (see Supplementary Table S1) and 1 μL cDNA. Reaction conditions were as follows: (i) initial denaturation (10 min at 95 °C); (ii) amplification and quantification (50 repeats of 15 s at 95 °C and 1 min at 60 °C) and (iii) melting curve program (60–95 °C with a heating rate of 0.5 °C/s). Melting curve analyses produced single peaks, with no primer–dimer peaks or artefacts, indicating the reactions were specific. All oligonucleotides (Supplementary Table S1) were purchased from MWG Biotech AG (Ebersberg, Germany).

Reactions were run on a 7500 Fast Real-Time PCR System (Applied Biosystems) in duplicate or triplicate. Linearity (R2) and efficiency (E = 10[−1/slope]) (Rasmussen, 2001) of each reaction were within accepted values. The suitability of the housekeeping genes as internal standards was confirmed in our samples through the geNORM v3.4 statistical algorithm, with M-values below 0.5 in all cases.

Senia-afp(−) plants were identified by analysing genomic DNA from T1 transgenic lines (R25.14) by means of qPCR targeting the afp transgene and the β-actin reference gene.

Bioinformatics expression analysis

RT-qPCR data were normalized and statistically analysed using the Genex software v.5.1.1.2 (MultiDAnalyses, Göteborg, Sweden). The Benjamini and Hochberg False Discovery Rate multiple testing correction was applied (Benjamini and Hochberg, 1995).

The Affymetrix software was used to identify a representative public ID and target description of each differentially expressed sequence. Biological processes of these sequences were determined on the basis of the homology of genes with known functions using the Gene Ontology annotation files.

The singular enrichment analysis of sequences differentially expressed in Senia-afp and Senia plants was carried out using the AgriGO computational tool (Zhou and Su, 2007) using Fisher’s statistical method and the Yekutieli false discovery rate (FDR) multiple test correction method (Benjamini and Yekutieli, 2001).

The singular enrichment analysis of sequences regulated by the transformation techniques, the insertion site and the transgene was performed with the AgriGO tool, using the whole set of sequences analysed by RT-qPCR as the customized reference, the hypergeometric statistical test with Yekuleti adjustment test and P-value 0.05.

Acknowledgements

We thank R. Collado, N. Company (UdG) and M. Palaudelmàs (CRAG) for technical assistance; S. Burgess for revision of the English; E. Melé (CRAG) for valuable suggestions; E. Garcia-Berthou (UdG) for assistance with the statistics, and P. Puigdomènech (CRAG) for critically reading the manuscript. This work was financially supported by the Spanish MEC project, ref. AGL2007-65903/AGR. M.M. received a studentship from the Fundación Ramón Areces.

Ancillary