• Open Access

The pleiotropic effects of the bar gene and glufosinate on the Arabidopsis transcriptome


* Correspondence (fax 613 759 1701; e-mail mikib@agr.gc.ca)


The Arabidopsis transcriptome was studied using the Affymetrix Arabidopsis ATH1 GeneChip in wild-type plants and glufosinate-tolerant transgenic plants expressing the bialaphos resistance (bar) gene. Pleiotropic effects were specifically generated in the transcriptomes of transgenic plants by both the bar gene and glufosinate treatments. In the absence of glufosinate, four genes were differentially expressed in the transgenic lines and another 80 genes were differentially expressed in the presence of glufosinate, 29 of which were specific to transgenic plants. In contrast, the number of differentially expressed genes specific to wild-type plants was 194 during the early response at 6 h of glufosinate treatment, and increased to 3711 during the late response at 48 h. Although the wild-type plants undergo extensive transcriptional reprofiling in response to herbicide-induced stress and, finally, plant death, the transgenic plants appear to activate other detoxification processes to offset the toxic effects of the residual herbicide or its derivatives. This study provides the first description of the pleiotropic effects of the bar gene and glufosinate on the plant transcriptome.


Phosphinothricin (l-PPT), also known as glufosinate-ammonium, is the active ingredient in several non-selective herbicide products, such as Basta, Liberty, Finale, Buster and Herbiace. Because l-PPT is structurally similar to l-glutamate, it is able to bind to the active site of glutamine synthetase (GS) and act as a competitive inhibitor of the enzyme (Block et al., 1987; Wohlleben et al., 1988). GS is an essential enzyme in plant nitrogen metabolism, because it assimilates ammonium through the conversion of l-glutamate to l-glutamine. Ammonium is generated in a variety of metabolic processes and serves as the nitrogen donor for the biosynthesis of all nitrogenous organic compounds needed for plant growth and development (Crawford and Arst, 1993; Daniel-Vedele et al., 1998). The mechanism of l-PPT toxicity is believed to involve ammonia accumulation to phytotoxic levels, followed by the impairment of photosynthesis causing plant death (Wendler et al., 1990; Wehrmann et al., 1996; Dan Hess, 2000).

The bialaphos resistance (bar) and phosphinothricin acetyltransferase (pat) genes from Streptomyces hygroscopicus and S. viridochromogenes, respectively, encode the enzyme phosphinothricin acetyltransferase (PAT) which inactivates l-PPT by transferring the acetyl group from acetyl-coenzyme A (acetyl-CoA) to the free amino group of l-PPT, yielding N-acetyl-l-PPT (Thompson et al., 1987; Wohlleben et al., 1988; D’Halluin et al., 1992). The enzymes perform comparably in plants and are highly specific for l-PPT (Wehrmann et al., 1996). Crops resistant to l-PPT have been created with the bar and pat genes using Agrobacterium- and particle bombardment-mediated transformation, and are now widely grown. In addition to Arabidopsis, these include transgenic maize, rice, barley, oat, wheat and cotton (Fromm et al., 1990; Christou et al., 1991; Somers et al., 1992; Weeks et al., 1993; Akama et al., 1995; Cheng et al., 1997; Keller et al., 1997; Tingay et al., 1997). Furthermore, both the bar and pat genes are preferred plant-selectable marker genes in several plant species (Wehrmann et al., 1996). Among the herbicide-resistance genes, they are the most extensively used as selectable marker genes in the scientific literature (Miki and McHugh, 2004).

An understanding of the unintended effects of genes (Cellini et al., 2004), such as pat or bar, is essential for a true comparison of transgenic plants with their non-transgenic progenitors. The functional analyses of unknown genes in transgenic plants in which pat or bar have been co-transformed and co-expressed as selectable marker genes are dependent on this knowledge. Studies with other selectable marker genes and reporter genes have shown that the insertion and expression of neomycin phosphotransferase type II (nptII) for kanamycin resistance and UidA for β-glucuronidase (GUS) reporter activity do not produce unintended effects on the transcriptome of Arabidopsis plants under normal growth conditions or under abiotic stresses, such as heat, cold, salt and drought (El Ouakfaoui and Miki, 2005). These data show that transgenic plants created with these marker genes are transcriptionally identical to non-transgenic plants. Comparable data are needed for each marker gene used in the study of transgenic plants.

The effects of the pat and bar genes on the transcriptome have not yet been investigated in the presence or absence of l-PPT. Therefore, a large number of questions related to the detailed mechanism of action, the range of pleiotropic effects and the progression of events following exposure to l-PPT remain largely unanswered, despite the widespread use of the genes and herbicides containing l-PPT. In this work, we used the Affymetrix ATH1 GeneChip to profile global gene expression patterns associated with the insertion and expression of the bar gene in the model plant Arabidopsis thaliana. We found that the expression and insertion of the bar gene produced changes in global gene expression in both the absence and presence of l-PPT. We report the identification of genes that are differentially expressed between wild-type (WT) and bar-expressing plants in response to 6-h and 48-h treatments with glufosinate-ammonium, providing new knowledge on the mechanisms behind l-PPT toxicity and plant responses to its presence.


The bar gene has limited effects on the transcriptome

Three transgenic Arabidopsis lines (BAR1, BAR2, BAR3) with single insertions of the bar gene (35S-bar-35S) were generated by Agrobacterium-mediated transformation using the pCAMBIA3300 transformation vector (CAMBIA, Canberra, Australia; http://www.cambia.org). They were compared with the progenitor WT line by percentage seedling germination, average leaf number, root length and fresh weight. There was no significant difference between the three transgenic lines and the WT plants (data not shown), indicating that the bar gene did not impart an obvious visible phenotype. The data confirmed several previous studies of transgenic plants containing bar or pat genes (Wehrmann et al., 1996; Oberdoerfer et al., 2005).

Plants were grown in Murashige and Skoog (MS) medium for 7 days, and total RNA was extracted from 200 individual seedlings per transgenic line and non-transformed controls. Microarray analysis using the Affymetrix ATH1 GeneChip was performed in triplicate, and the data were normalized within each group using the Robust Multichip Average (RMA) (Millenaar et al., 2006). Quality control was performed using the AffylmGUI package of BioConductor (Gentleman et al., 2004; Wettenhall et al., 2006). Using a P value cut-off of 0.05 and a criterion of 1.5-fold change, the data revealed a very small number of differentially expressed genes between each of the three transgenic lines (BAR1, BAR2 and BAR3) and the non-transformed control (7, 18 and 32 genes, respectively, representing 0.028%, 0.065% and 0.128% of the Arabidopsis genome). Four genes [At3g30720, At5g45820 (CIPK20), At5g02810 (PRR7) and At2g23840] were found to be significantly repressed only in the three transgenic lines (mean decrease of –1.7- to –3.2-fold).

Inhibition of seedling growth and development by glufosinate

In preparation for microarray experiments designed to examine the effects of l-PPT on the transcriptome, we first determined the most suitable level of glufosinate (see ‘Experimental procedures’) needed to inhibit seedling growth, and the duration of treatment. Seven-day-old seedlings of WT and transgenic lines (BAR1, BAR2, BAR3) were germinated on MS medium and transferred to glufosinate-containing medium (concentration range, 0.1–100 µg/mL). The total fresh weight was measured after 2 weeks. Figure 1a shows that the growth of WT seedlings was inhibited at concentrations of glufosinate above 1 µg/mL, whereas all of the transgenic lines were resistant to concentrations up to 50 µg/mL glufosinate. At 100 µg/mL glufosinate, the transgenic lines showed partial inhibition of growth (Figure 1a) accompanied by root growth inhibition and yellowing (data not shown). These results clearly show that 25 µg/mL glufosinate is the optimal concentration required to distinguish between glufosinate-tolerant and WT phenotypes under our experimental conditions.

Figure 1.

Inhibition of growth by glufosinate (PPT). One-week-old seedlings from the three transgenic lines BAR1, BAR2 and BAR3 and the wild-type (WT) plant were transferred to medium supplemented with glufosinate (concentration range, 0.1–100 µg/mL) and the total fresh weight of 10 seedlings was measured after 10 days (A). The fresh weight (B) and root length (C) of groups of 10 seedlings transferred to medium containing 25 µg/mL glufosinate were measured at intervals up to 96 h. Error bars indicate standard errors.

To determine the optimal duration of glufosinate treatment, 5-day-old WT and transgenic lines (BAR1, BAR2, BAR3) were treated with 25 µg/mL glufosinate, as described previously, followed by measurements of fresh weight and root length at intervals up to 96 h. As shown in Figure 1b,c, both parameters revealed differences between the WT and transgenic lines, starting at 24 h. These differences were significant by 48 h, as the WT seedlings ceased root development and failed to gain fresh weight, whereas the transgenic seedlings continued to develop and grow. Furthermore, the data indicated that 6-h and 48-h treatments should be adequate to distinguish between early and late molecular events related to glufosinate toxicity.

Responses of WT transcriptome to glufosinate

Early responses

For microarray experiments, WT samples consisted of approximately 200 seedlings, treated or untreated with 25 µg/mL glufosinate for 6 and 48 h. Three replicates were performed for each sample, which were grown in the same growth chamber to minimize experimental and sample-to-sample variation. Microarray analysis was performed using the Affymetrix ATH1 GeneChip and the data were normalized using RMA (Millenaar et al., 2006). Using a P value cut-off of 0.05 and a two-fold change, the data revealed an early transient response of the transcriptome to glufosinate at 6 h, and more widespread changes that subsequently occurred by 48 h.

Although the glufosinate-induced phenotypes were not apparent at 6 h in WT plants, previous studies have shown that glufosinate induces more rapid changes at the molecular level. For example, it is well known that glufosinate is absorbed in the first hour by leaves and roots, and that inhibition of GS activity reaches 90% within 30 min, followed by a ≥ 10-fold accumulation in ammonia within a few hours (Ullrich et al., 1990; Lacuesta et al., 1992; Steckel et al., 1997). As shown in Figure 2, our microarray data revealed that the early effect on the transcriptome can be seen by 6 h by the differential expression of 246 genes, or 0.86% of the total genes annotated in the Arabidopsis genome, of WT plants. Of these, 194 differentially expressed genes were specific to WT plants (70% were up-regulated and 30% were down-regulated) and only 52 genes were also differentially expressed in the treated transgenic BAR1 plants (Table S1, see ‘Supporting Information’). Of the 194 genes, the differential expression of 84 genes (43%) was transient at 6 h and not observed at 48 h.

Figure 2.

Differential expression of genes in the early response (6 h) and late response (48 h) to glufosinate (PPT) in wild-type (WT) and transgenic BAR1 plants. The total number of differentially expressed genes is shown in parentheses. Those which are specific to each condition are shown in the individual circles and those which are expressed under both conditions are shown in the overlapping areas. RNA was extracted from 10-day-old WT and BAR1 plants untreated and treated with 25 mg/mL glufosinate at 6 and 48 h.

Based on FunCat assignments available through the Munich Information Centre for Protein Sequences (MIPS) Arabidopsis database (mips.gsf.de/proj/thal/db/), the 194 differentially expressed genes were classified into functional categories, the most extensive being metabolism (16.6%), protein with binding function or cofactor requirement (15.3%), cell rescue (defence and virulence) (10.1%), interaction with environment (8.1%), cellular transport (7.8%) and transcription (6.2%). Almost 5% of the genes could not be classified into known functional categories (Figure 3a). As previous studies have reported the transient inhibition of nitrate reductase following GS inhibition (Trogisch et al., 1989), the repression of genes involved in nitrate metabolism and assimilation, i.e. NIA1 (nitrate reductase; At1g77760) and NIR1 (ferredoxin-nitrite reductase; At2g15620), was expected. The expression of several stress-related genes was also affected, including six heat shock proteins and four oxidative stress response-related genes. The primary response included the rapid induction of transcription factors: eight transcription factors were up-regulated, including members of ethylene-response factor/APETALA 2 (ERF/AP2) (At1g64380, At4g32800, At1g77640, At4g17490 and At2g44840), MYB transcription factor (At2g23290), basic helix–loop–helix (bHLH) (At1g51140) and scarecrow-like protein (At4g17230). Expression profiling did not reveal the activation of selective pathways at this early time point, but rather the induction of several regulatory proteins that could be essential to drive the secondary and subsequent responses that follow in time. To verify the microarray results, 10 genes that were significantly up-regulated by glufosinate treatment were tested for transcription levels by reverse transcriptase-polymerase chain reaction (RT-PCR) analysis. As shown in Figure 3b, the induction of the selected genes was confirmed, providing confidence in the microarray data.

Figure 3.

Functional classification of glufosinate (PPT)-responsive genes in the wild-type Arabidopsis plant. (A) Functional categories of genes differentially expressed at 6 h. (B) Confirmation of microarray results. Reverse transcriptase-polymerase chain reaction (RT-PCR) was conducted using specific primers for the 10 genes indicated and β-tubulin (β-tub) as a control. RNA was extracted from untreated and treated wild-type plants at 6 h.

Late responses

At 48 h of glufosinate treatment, extensive changes in gene expression were observed, including the differential expression of 3762 genes, or 13.5% of the total genes annotated in the Arabidopsis genome. Of these, 3711 genes were specific to WT plants and, again, only 51 were also differentially expressed in the treated BAR1 plants (Figure 2; Table S1, see ‘Supporting Information’). Of the 3711 genes, 1765 (47.5%) were up-regulated and 1946 (52.5%) were down-regulated. Only 111 genes of the 3711 were also differentially expressed at 6 h; however, the expression level increased by 20–120-fold in some cases. The other 3600 genes reflected the late response at 48 h in WT plants.

As shown in Figure 4, the most abundant functional categories of genes were metabolism (17.8%), protein with binding function or cofactor requirement (17.4%), protein fate (7.6%), cellular transport (6.6%) and transcription (4.8%). Almost 12.9% of the late-response genes could not be classified into a known functional category. The differential expression of several genes involved in amino acid metabolism, glutamine metabolism and ammonium assimilation was observed during the late response, including members of the GS1 family and GS2 gene. Furthermore, several other genes involved in detoxification processes were induced. These included five genes previously described as detoxification-related genes (At1g28480, At1g56650, At2g29420, At2g29490 and At2g30540) and others that have been hypothesized to function in detoxification processes, including members of glutathione transferase (GST), cytochrome P450, ABC transporters and uridine diphosphate (UDP)-glucose transferase (see Table 1) (Schaeffner et al., 2002; Baerson et al., 2005; Pilon-Smits, 2005). The expression levels of UDP-glucosyl transferase (At1g05680) and cytochrome P450 (At4g37370) increased up to 74-fold and 36-fold, respectively, at 48 h (Table 1). Six other transcription factors were also up-regulated, including members of WRKY (WRKY46 and WRKY25), no apical meristem (NAM) family protein (At5g63790, At1g77450 and At1g01720), ERF/AP2 (At1g43160), MYB57 (At1g56650) and the heat stress transcription factor (Hsf) family (At4g18880). The up-regulation of genes known to be involved in the biosynthesis of auxin, such as indoleacetic acid (IAA)-amino acid hydrolase 1 (ILL5) (At1g51780), and jasmonic acid biosynthesis, such as 12-oxophytodienoic acid reductase 2 (OPR2) and lipoxygenase 3 (LOX3), was observed. The differential expression of senescence-related genes and photosynthesis-related genes was also observed. These data demonstrate the coordination and integration of several specific pathways following GS inhibition, as a secondary reaction to the initial response to glufosinate seen at 6 h (Figure 5).

Figure 4.

Functional classification of glufosinate-responsive genes in the wild-type Arabidopsis plant at 48 h.

Table 1.  List of selected glufosinate-responsive genes hypothesized to be involved in detoxification processes
Locus IDNameDescriptionFold change
6 h48 h
ABC transporter
 At3g60160ATMRP9ABC transporter family protein 14.5
 At1g15520 ABC transporter family protein 14.3
 At3g59140ATMRP14ABC transporter family protein 14.2
 At1g71330ATNAP5ABC transporter family protein2.19.1
 At3g13100ATMRP7ABC transporter family protein 7.7
 At3g47780ATATH6ABC transporter family protein 5.4
 At4g18050 ABC transporter family protein 4.7
 At2g36380 ABC transporter family protein 3.6
 At3g47730ATATH1ABC transporter family protein 3.3
 At3g53480 ABC transporter family protein2.23.2
 At1g67940ATNAP3ABC transporter family protein 3.1
 At2g26910 ABC transporter family protein –3.3
 At1g51500CER5ABC transporter family protein –3.3
 At2g13610 ABC transporter family protein –3.4
 At1g17840 ABC transporter family protein –4.2
Cytochrome P450
 At4g37370CYP81D8Cytochrome P45015.336.6
 At2g45570CYP76C2Cytochrome P450 76C2, putative 28.6
 At3g26830PAD3Cytochrome P450 71B15, putative (CYP71B15) 27.6
 At2g34500CYP710A1Cytochrome P450 family protein 17.6
 At5g67310CYP81G1Cytochrome P450 family protein; member of CYP81G 11.2
 At3g28740 Cytochrome P450 family protein 10.8
 At3g26210CYP71B23Cytochrome P450 71B23, putative 7.2
 At2g30770CYP71A13Cytochrome P450 71A13, putative 6.9
 At3g14680CYP72A14Cytochrome P450, putative 6.3
 At1g64950CYP89A5Cytochrome P450, putative 5.3
 At3g03470CYP89A9Cytochrome P450, putative 5.1
 At5g36220CYP81D1Cytochrome P450 81D1 3.7
 At4g15110CYP97B3Cytochrome P450 97B3, putative –3.0
 At3g44970 Cytochrome P450 family protein –3.1
 At4g00360CYP86A2Cytochrome P450, putative –3.4
 At4g39510CYP96A12Cytochrome P450 family protein –4.3
 At4g13770CYP83A1Cytochrome P450 family protein –4.6
 At2g34490CYP710A2Cytochrome P450 family protein –4.9
 At3g53130LUT1Cytochrome P450 family protein –5.3
 At5g45340CYP707A3Cytochrome P450 family protein5.8 
 At5g63450CYP94B1Cytochrome P450, putative2.4 
UDP-glucosyl transferase
 At1g05680 UDP-glucuronosyl/UDP-glucosyl transferase15.874.4
 At3g11340 UDP-glucuronosyl/UDP-glucosyl transferase 27.7
 At3g46660 UDP-glucuronosyl/UDP-glucosyl transferase3.023.3
 At2g15490 UDP-glucuronosyl/UDP-glucosyl transferase2.818.4
 At2g36750 UDP-glucuronosyl/UDP-glucosyl transferase3.417.8
 At4g34135 UDP-glucuronosyl/UDP-glucosyl transferase4.113.3
 At1g22400 UDP-glucuronosyl/UDP-glucosyl transferase3.112.5
 At2g15480 UDP-glucuronosyl/UDP-glucosyl transferase6.011.9
 At2g30140 UDP-glucuronosyl/UDP-glucosyl transferase 11.4
 At4g15490 UDP-glucuronosyl/UDP-glucosyl transferase 10.0
 At1g05560UGT1UDP-glucose transferase (UGT75B2)3.49.4
 At4g10960 UDP-glucose 4-epimerase, putative 9.4
 At5g49690 UDP-glucuronosyl/UDP-glucosyl transferase 9.2
 At2g36770 UDP-glucuronosyl/UDP-glucosyl transferase 8.3
 At2g36790 UDP-glucuronosyl/UDP-glucosyl transferase6.06.9
 At3g46670 UDP-glucuronosyl/UDP-glucosyl transferase 6.0
 At2g36970 UDP-glucuronosyl/UDP-glucosyl transferase 5.2
 At3g16520 UDP-glucuronosyl/UDP-glucosyl transferase –2.4
 At2g30150 UDP-glucuronosyl/UDP-glucosyl transferase –2.4
 At2g31790 UDP-glucuronosyl/UDP-glucosyl transferase –2.7
Glutathione S-transferase
 At2g29490ATGSTU1Glutathione S-transferase, putative2.210.4
 At2g29460ATGSTU4Glutathione S-transferase, putative 52.8
 At1g17170ATGSTU24Glutathione S-transferase, putative 30.0
 At2g29480ATGSTU2Glutathione S-transferase, putative 28.8
 At2g29470ATGSTU3Glutathione S-transferase, putative 19.9
 At1g74590ATGSTU10Glutathione S-transferase, putative 12.0
 At1g49860ATGSTF14Glutathione S-transferase, putative 6.9
 At2g29420ATGSTU7Glutathione S-transferase, putative3.36.0
 At1g02930ATGSTF6Glutathione S-transferase, putative 5.5
 At1g69930ATGSTU11Glutathione S-transferase, putative 4.7
 At1g10360ATGSTU18Glutathione S-transferase, putative –4.2
 At1g78370ATGSTU20Glutathione S-transferase, putative –13.6
 At1g28480 Glutaredoxin family protein4.310.4
 At4g33040 Glutaredoxin family protein 9.1
 At3g62960 Glutaredoxin family protein 5.8
 At2g47880 Glutaredoxin family protein 4.7
 At1g03850 Glutaredoxin family protein 3.9
 At5g40370 Glutaredoxin, putative 2.8
 At2g30540 Glutaredoxin family protein3.52.3
 At4g15690 Glutaredoxin family protein –2.4
 At5g18600 Glutaredoxin family protein –2.9
 At1g06830 Glutaredoxin family protein2.1 
MATE family protein
 At2g04040 MATE efflux family protein 29.8
 At2g04050 MATE efflux family protein 10.5
 At3g23550 MATE efflux family protein 8.7
 At1g71140 MATE efflux family protein 7.3
 At2g04070 MATE efflux family protein 3.4
 At5g17700 MATE efflux family protein –3.2
 At1g15180  MATE efflux family protein –4.8
 At4g08770 Peroxidase, putative, identical to class III peroxidase ATP38 6.6
 At4g36430 Peroxidase, putative 4.0
 At4g37530 Peroxidase, putative, identical to cDNA peroxidase ATP37 3.7
 At2g18140 Peroxidase, putative, similar to peroxidase ATP6a 3.5
 At4g11290 Peroxidase, putative, identical to peroxidase ATP19a 3.1
 At5g05340 Peroxidase, putative, similar to peroxidase P7 3.1
 At1g68850 Peroxidase, putative, identical to peroxidase ATP23a 3.0
 At5g40150 Peroxidase, putative, identical to peroxidase ATP26a –2.9
 At1g71695 Peroxidase 12 (PER12) (P12) (PRXR6) –3.7
 At5g58390 Peroxidase, putative –4.1
Transcription factor
 At3g10500ATAF1No apical meristem (NAM) family 3.8
 At1g01720 No apical meristem (NAM) family protein2.03.5
 At5g63790 No apical meristem (NAM) family protein2.63.2
 At1g56650PAP1Myb family transcription factor (MYB75)3.13.4
 At1g62300WRKY6WRKY family transcription factor 4.2
MATE, multidrug and toxic compound extrusion; UDP, uridine diphosphate.
Figure 5.

Early and late events after glufosinate treatment in the wild-type plant.

Downstream effects of GS inhibition by glufosinate

In plants, there are two isoforms of GS: the cytosolic form (GS1), encoded by a multigene family, and a plastidic form (GS2), encoded by a single gene (Hirel and Gadal, 1980). In the late response of Arabidopsis, the GS1 family (GLN1;1, GLN1;2, GLN1;3, GLN1;4, GLN1;5; Ishiyama et al., 2004), excluding GLN1;5, was up-regulated, whereas GS2 was repressed. Furthermore, 14 other genes hypothesized to be linked to this pathway were also differentially expressed (Table 2). The underlying signal transduction and mechanistic pathways are unclear at this time, but presumably result from the stress caused by the inhibition of GS activity, subsequent ammonia accumulation and the lack of essential amino acids.

Table 2.  List of glufosinate-responsive genes hypothesized to be involved in ammonia assimilation and glutamine biosynthesis
Locus IDNameDescriptionFold change
6 h48 h
At5g18170GDH1Glutamate dehydrogenase 1–2.1 
At2g41220GLU2Glutamate synthase 4.3
At3g17820ATGSKB6Glutamine synthetase (GS1) 2.4
At5g35630GS2Glutamine synthetase (GS2) –2.4
At5g16570GLN1; 4Glutamine synthetase 3.4
At1g66200ATGSR2Glutamine synthetase 2.1
At5g37600ATGSR1Glutamine synthetase 12.3
At3g22200POP24-Aminobutyrate aminotransferase 2.5
At2g37500 Arginine biosynthesis protein –2.7
At4g34710ADC2Arginine decarboxylase (ADC) –6.2
At5g11520ASP3; YSL4Aspartate aminotransferase. 5.3
At3g55610 δ1-pyrroline-5-carboxylate synthetase B –2.5
At2g05990ENR1; MOD1Enoyl-(acyl-carrier protein) reductase –3.9
At3g57560 N-Acetylglutamate kinase –2.2
At3g47450 Nitric oxide synthase –2.4
At5g46180delta-OATOrnithine aminotransferase, putative 4.4
At4g01900GLB1P II nitrogen sensing protein –3.5
At3g30775ERD5Proline oxidase 3.8
At5g38710 Proline oxidase, putative 7

As GS plays a major role in ammonium assimilation, which then serves as the nitrogen donor for the biosynthesis of all nitrogenous organic compounds in plants (Crawford and Arst, 1993; Daniel-Vedele et al., 1998), it would be expected that the inhibition of GS would have profound downstream effects on pathways leading to the metabolism of nitrogenous molecules, such as amino acids, proteins, nucleic acids and chlorophyll. This was confirmed by the functional analysis of the categories of genes affected by glufosinate treatment. A large set of genes involved in important metabolic pathways (16.6%–17.8%) (see Figures 3a and 4) were affected, including genes involved in nitrogen metabolism, nucleotide/nucleobase metabolism, phosphate metabolism, carbon compound and carbohydrate metabolism, lipid and fatty acid metabolism, metabolism of vitamins/cofactors/prosthetic groups and secondary metabolism.

Amino acid and chlorophyll synthesis.  Our list of glufosinate-responsive genes was compared with the list of genes in the AraCyc database of metabolic pathways (http://www.arabidopsis.org/biocyc/) (Mueller et al., 2003). Of the genes that are responsive to glufosinate treatment at 6 and 48 h, 455 were identified in the AraCyc list (Table S4, see ‘Supporting Information’). Of these, 52 genes belong to the amino acid metabolism category. The early response included the down-regulation of only one gene, i.e. asparagine synthase (At5g65010), involved in asparagine biosynthesis. The extent of repression reached 20-fold after 48 h. The late response included the up- or down-regulation of 52 genes involved in the biosynthesis and degradation of the other amino acids, excluding leucine and histidine. As the effect was not limited to glutamine, the data provided evidence for the existence of cross-talk among genes in different metabolic pathways. As expected, one consequence was the repression of 15 genes involved in chlorophyll and chlorophyllide biosynthesis during the late response.

Photosynthesis.  The expression of 32 genes involved in photosynthesis was found to be down-regulated during the late response (see Table 3). These included members of the light-harvesting chlorophyll a/b-binding family (Cab/LHC), such as LHCA2 (At3g61470), LHCA2*1 (At1g19150), LHCA3*1 (At1g61520), LHCA5 (At1g45474), LHCB2.2 (At2g05070), LHCB2:4 (At3g27690), LHCB3 (At5g54270), LHCB4.2 (At3g08940), LHCB4.3 (At2g40100), LHCB5 (At4g10340) and LHCB6 (At1g15820). The Cab superfamily in Arabidopsis consists of 20 different proteins (Jansson, 1999), six of which are associated with photosystem I (PSI) (Lhca1–6) and 14 with photosystem II (PSII) (Lhcb1–6 and their isomers). The primary function of this family is the absorption of light through chlorophyll excitation and the transfer of absorbed energy to photochemical reaction centres (Heddad et al., 2006). Consistent with our microarray data, several other studies have linked glufosinate exposure to the inhibition of photosynthesis (Sauer et al., 1987). Glufosinate is known to indirectly inhibit the light reaction in photosynthesis (Sauer et al., 1987; Wild and Wendler, 1991; Lacuesta et al., 1992; Dan Hess, 2000). The mechanism of this inhibition is not yet clear.

Table 3.  List of glufosinate-responsive genes hypothesized to be involved in photosynthesis
Locus IDNameDescriptionFold change
At3g27690LHCB2:4Chlorophyll a/b-binding protein–71.2
At2g39470 Photosystem II reaction centre PsbP family protein–23.0
At1g19150LHCA2*1Chlorophyll a/b-binding protein–17.0
At3g08940LHCB4.2Chlorophyll a/b-binding protein–13.6
At5g54270LHCB3Chlorophyll a/b-binding protein–12.1
At3g55330 Photosystem II reaction centre PsbP family protein–9.5
At3g54890 Encodes a component associated with photosystem I–8.7
At1g52230 Photosystem I reaction centre subunit VI–7.4
At4g28660 Photosystem II reaction centre W (PsbW) family protein–7.2
At3g47470CAB4Chlorophyll a/b-binding protein 4–7.1
At4g05180 Oxygen-evolving enhancer protein 3, chloroplast–6.4
At2g34430LHB1B1Chlorophyll a/b-binding protein–6.2
At3g50820 Encodes a protein which is an extrinsic subunit of photosystem II–5.6
At4g02770 Photosystem I reaction centre subunit II–5.0
At3g16140 Photosystem I reaction centre subunit VI–4.9
At1g29910CAB3Chlorophyll a/b-binding protein 2–4.6
At5g57030LUT2Lycopene ɛ-cyclase–4.5
At3g15850FAD5Fatty acid desaturase family protein–4.4
At4g15510 Photosystem II reaction centre PsbP family protein–4.4
At5g66570 Encodes a protein which is an extrinsic subunit of photosystem II–4.4
At4g30950FAD6ω-6 fatty acid desaturase, chloroplast–4.4
At4g10340LHCB5Chlorophyll a/b-binding protein CP26–3.9
At1g55670 Photosystem I reaction centre subunit V–3.8
At2g05070LHCB2.2Chlorophyll a/b-binding protein–3.8
At2g40100LHCB4.3Chlorophyll a/b-binding protein–3.6
At1g61520LHCA3*1Chlorophyll a/b-binding protein–3.5
At5g47110 lil3 protein, putative–3.3
At1g45474LHCA5Chlorophyll a/b-binding protein–3.2
At3g61470LHCA2Chlorophyll a/b-binding protein–3.0
At1g67740PSBYPhotosystem II core complex proteins psbY–2.9
At5g01530 Chlorophyll a/b-binding protein CP29–2.9
At2g30790 Photosystem II oxygen-evolving complex 23–2.1
At5g07920ATDGK1Diacylglycerol kinase 1 (DGK1)2.5
At4g14690ELIP1Chlorophyll a/b-binding family protein2.8
At5g59820RHL41Zinc finger (C2H2 type) family protein (ZAT12)10.1

The early response did not include the alteration of photosynthesis-related genes; however, the expression of a zinc finger protein DFL2 (‘DWARF IN LIGHT 2’) involved in high light and cold acclimation was found to be up-regulated at 6 and 48 h, and appears to be one of the first early responses. This gene encodes a GH3-related gene involved in red light-specific hypocotyl elongation. The analysis of sense and antisense transgenic plants suggests that DFL2 is located downstream of red light signal transduction and determines the degree of hypocotyl elongation (Takase et al., 2004). Therefore, the early induction of this regulatory protein may provide an early signal in the pathway leading to the inhibition of photosynthesis by glufosinate.

Sugar metabolism.  Other differentially expressed genes in the late response included 28 genes involved in sucrose and cellulose biosynthesis. Our data were consistent with the report of decreased levels of sucrose following GS inhibition (Gordon et al., 1999). Metabolite profiling analysis of Medicago truncatula root nodules also showed a reduction of fructose, fructose-6-phosphate and glucose-6-phosphate after glufosinate treatment (Barsch et al., 2006).

Regulatory pathways responding to glufosinate

Hormone synthesis and signalling.  Our microarray analysis revealed that glufosinate treatment caused an increase in the transcript levels of genes involved in plant hormone biosynthesis and signal transduction pathways. We observed the early up-regulation of three genes involved in the biosynthesis of auxin and jasmonic acid, i.e. ILL5, OPR2 and LOX3. The late response included the elevation of transcript levels of key enzymes involved in jasmonic acid biosynthesis, such as lipoxygenase 1 (LOX1, At1g55020) and 12-oxophytodienoic acid reductase 3 (OPR3, At2g06050) (Schaller et al., 2000; Ziegler et al., 2000). Genes involved in ethylene biosynthesis were also up-regulated, including members of the 1-aminocyclopropane-1-carboxylate (ACC) synthase family (ACS2: At1g01480, ACS8: At4g37770 and At5g59530) (Yang and Hoffman, 1984). Genes involved in abscisic acid biosynthesis and signal transduction pathways, such as aldehyde oxidase 3 (AAO3: At2g27150) and protein phosphatase 2C (ABI2: At5g57050), were found to be up-regulated (Seo et al., 2000). The auxin biosynthesis pathway was also affected by glufosinate, in which some genes were induced, such as the IAA-amido synthase (At2g23170 and At1g59500) (over 50-fold), IAA-amino acid hydrolase 1 (ILR1: At3g02875 and ILL5: At1g51780), nitrile aminohydrolase-nitrilase (NIT2: At3g44300, NIT3: At3g44320 and NIT4: At5g22300) and cytochrome P450 (CYP79B2: At4g39950) (Cheng et al., 2006). Other genes, such as IAA4 and IAA8, were repressed. The complex profile of differentially expressed genes in plant hormone biosynthetic pathways suggests that they coordinate many of the secondary responses to glufosinate.

Trancriptional regulation.  Almost 5% of glufosinate responsive genes belong to the transcription factors and DNA-binding proteins. Examples include the plant-specific WRKY transcription factors, in which 19 genes were up-regulated; NAM/NAC family, in which 24 genes were up-regulated; basic region/leucine zipper motif (bZIP) transcription factors, in which nine genes were up-regulated; ERF/AP2 family, in which eight genes were up-regulated and one was down-regulated; MYB family, in which 12 genes were up-regulated and 12 were down-regulated; and bHLH family, in which four genes were up-regulated and nine were down-regulated. In addition, a number of genes associated with cellular communication and signal transduction mechanisms (6.5%), including receptor-like protein kinases and protein kinases, were also significantly regulated by glufosinate.

Two WRKY transcription factors, i.e. WRKY25 and WRKY46, were up-regulated during the early and late responses, and 17 other members of the WRKY family were up-regulated during the late response. This family of regulatory proteins has been reported to be implicated in different stress responses in many species, and is strongly and rapidly up-regulated in response to wounding, pathogen infection, senescence or abiotic stresses (Eulgem et al., 2000). The function of several WRKY genes has already been elucidated. WRKY70 is known to be implicated in salicylic acid-dependent defence responses (Yu et al., 2001). WRKY22 and WRKY29 have been identified as important downstream components of a mitogen-activated protein kinase (MAPK) pathway that confers resistance to both bacterial and fungal pathogens (Asai et al., 2002). WRKY6 and WRKY53 have been shown to be involved in leaf senescence (Hinderhofer and Zentgraf, 2001; Robatzek and Somssich, 2002). Furthermore, recent microarray analysis reported the up-regulation of several WRKY genes in response to other herbicides and xenobiotic compounds (Baerson et al., 2005; Madhou et al., 2006; Manabe et al., 2007).

Three NAC transcription factors were up-regulated by glufosinate during both early and late responses, and 21 others were induced specifically during the late response. Members of this family have been implicated in various aspects of plant development, including apical shoot, stem and leaf development (Souer et al., 1996). More recently, NAC domain genes have also been implicated in defence responses to biotic and abiotic stresses. The potato StNAC and Arabidopsis ATAF1 and ATAF2 genes have been shown to be induced by pathogen attack and wounding (Collinge and Boller, 2001). Members of the Brassica napus NAC (BnNAC) family of genes have also been shown to be differentially regulated in response to biotic and abiotic stresses, indicating an important role in the mediation of the transcriptional responses to diverse biotic and abiotic stresses (Hegedus et al., 2003).

We also observed the differential expression of ERFs, which belong to the AP2 family of transcription factors with AP2/ERF domains. Two ERFs were transiently up-regulated during the early response and another was up-regulated during the early and late responses. Another seven were up-regulated and one was down-regulated specifically during the late response. ERF proteins play a key role in the integration of the ethylene and jasmonate signalling pathways to activate ethylene/jasmonate-dependent responses to pathogens and various environmental stresses (Chen et al., 2002; Lorenzo et al., 2003).

Interestingly, genes belonging to WRKY, NAC and bZIP families were only up-regulated by glufosinate treatment, whereas members of the MYB and bHLH (MYC) families were both up- or down-regulated. Similarities in gene expression patterns between MYB and MYC families may indicate possible cooperative interactions that regulate glufosinate-responsive genes. Several examples of cooperative interaction of bHLH (MYC) and MYB regulatory proteins have been reported in plants (Roth et al., 1991; Goff et al., 1992; Martin and Paz-Ares, 1997; Grotewold et al., 2000; Payne et al., 2000; Nesi et al., 2001). Similar combinatorial interactions between transcription factors may be involved in controlling and integrating the diverse responses to glufosinate.

Response of the transgenic transcriptome to glufosinate treatment

Our data revealed that the early response of the WT and transgenic transcriptomes involved a small number of genes and did not translate into visible phenotypes. The transcriptomes of the transgenic BAR1 plants after 6 h of treatment revealed only 81 differentially expressed genes in the presence of glufosinate. This was approximately one-third of the number of differentially expressed genes described earlier in WT plants under identical conditions (Figure 2). Of these, 29 genes (36%) were specific to the transgenic plants and 52 genes (64%) were also differentially expressed in 6 h-treated WT plants. Of the 52 common genes, 44 (85%) were transiently expressed at 6 h and eight (15%) had similar or greater expression levels at 48 h. The late response to glufosinate seen in WT plants at 48 h was essentially eliminated in transgenic BAR plants. It was limited to 80 differentially expressed genes, which is 50-fold lower than that in WT plants under identical conditions. Of these, 51 genes (64%) were again common to both transgenic and WT plants, and 29 (36%) were again specific to transgenic plants (Tables S2 and S3, see ‘Supporting Information’). Interestingly, eight genes were shown to be up-regulated in the treated WT and transgenic plants at both 6 and 48 h. These genes were found to be induced in response to wounding; four had been described previously as wound-responsive genes, including wound-responsive gene 3 (WR3) (At5g50200), flavin-adenine dinucleotide (At1g30720), serine protease inhibitor (At5g43580) and β-fructosidase (At3g13790). Moreover, the eight genes were found to be induced by at least one biotic stress condition (http://bbc.botany.utoronto.ca) (Toufighi et al., 2005). These findings may reveal the activation of general stress responses in both tolerant and non-tolerant plants as a first transient response to the presence of glufosinate. Several recent studies have shown that plants are able to develop nonspecific detoxification systems against chemicals, including certain herbicides and antibiotics (Yuan et al., 2007).

It is interesting that 29 differentially expressed genes were specific to transgenic plants at both 6 and 48 h (Table S3, see ‘Supporting Information’; Figure 2). Using a P value cut-off of 0.05 and a criterion of two-fold change to distinguish differential expression, these appeared to be separate and distinct sets of genes. Lowering the stringency criteria only identified the possible overlap of one or two genes. Genes specific to transgenic plants at 6 h include the following: 10 genes in unknown functional categories; five detoxification-related genes (three peroxidase, oxidoreductase and cytochrome P450); two disease resistance genes, including zinc finger transcription factor; and four genes involved in the biosynthesis and metabolism of carbohydrate [myo-inositol oxygenase (MOX2), At2g19800; aldose 1-epimerase, At3g47800; sugar transporter, At1g08920; fructose-bisphosphate aldolase, At4g26530]. The induction of four genes in transgenic lines BAR1, BAR2 and BAR3 was confirmed by RT-PCR (see Figure 6). Among those specific to transgenic plants at 48 h, 28 of the 29 genes were up-regulated, and included seven genes in functional categories, one detoxification gene (peroxidase PER57), three cytochrome P450 genes potentially involved in detoxification processes, two defence-related proteins, three late embryogenesis protein genes, two lipid transfer protein genes and a DREB transcription factor that appears to stimulate cytokinin biosynthesis. These results show that both unique sets of genes were represented by detoxification-related genes, such as P450, peroxidase and transporter genes, which may be specifically and temporally activated in transgenic plants by glufosinate, the acetylated form of glufosinate (N-acetyl-l-glufosinate) or other intermediates in the detoxification pathway. It is not known whether different subsets of genes respond differentially to the different metabolic intermediates.

Figure 6.

Confirmation of microarray results. Reverse transcriptase-polymerase chain reaction (RT-PCR) was conducted using specific primers for the four genes and β-tubulin (β-tub) as a control. RNA was extracted from untreated (–PPT) and treated (+PPT) wild-type plants and the three transgenic lines BAR1, BAR2 and BAR3 at 6 h.

As expected for WT plants that are experiencing herbicide-induced stress, the differential expression of detoxification genes was more extensive than in the transgenic plants. Twenty-two detoxification-related genes were observed during the early response, and 148 genes were observed during the late response (see Table 1). Of the 22, only three genes were transiently up-regulated, whereas the expression of 19 genes was sustained and/or up-regulated into the late response, in some case reaching a 74-fold increase in expression level. As only four of these were up-regulated in both transgenic and WT plants, the majority are probably responding to the widespread stress and damage in WT plants induced by the herbicide in the absence of PAT activity.

The set of detoxification genes expressed in WT plants included the following subfamilies (Table 1): GST, cytochrome P450, ABC transporters, multidrug and toxic compound extrusion (MATE) proteins, alternative oxidase (AOX), steroid sulphotransferase (ST) and ACC-oxidase (ACO). Many of the identified genes have been described previously as detoxification-related genes, such as PAP1 (At1g56650), ATGSTU1 (At2g29490) and ATGSTU7 (At2g29420), whereas others have been hypothesized to function in detoxification processes (Schaeffner et al., 2002; Baerson et al., 2005; Pilon-Smits, 2005) These data indicate that the early and late responses in the WT plant involve the activation of detoxification processes as expected, but these may differ from the processes activated in the transgenic plants. In the transgenic plants, it is possible that the detoxification genes may include those that are specific to the products of glufosinate metabolism.


Bacterial PAT, coded by the pat or bar genes, has been used extensively for the development of new herbicide-tolerant crops and as a selectable marker in the transformation process (Miki and McHugh, 2004). Yet, the unintended effects of the gene on the plant transcriptome have not been examined in detail using rigorous profiling strategies (Cellini et al., 2004). In this study, we have identified pleiotropic effects of the bar gene on the Arabidopsis transcriptome and document the transcriptome responses to the herbicide glufosinate and its derivatives.

The results reveal that the transcriptomes of the three transgenic plants expressing the bar gene differ from their WT counterparts by 7, 18 and 32 genes individually, but only four of these genes are differentially expressed in all three transgenic plants. Other studies that have utilized expression profiling strategies have shown that the insertion and expression of nptII has no pleiotropic effects on the transcriptome of Arabidopsis plants, including the dramatic transcriptional reprogramming that occurs in response to abiotic stresses (El Ouakfaoui and Miki, 2005). Furthermore, the expression of the gus/uidA gene also generates no pleiotropic effects on the transcriptome (El Ouakfaoui and Miki, 2005). These experiments reveal the stability of the Arabidopsis transcriptome to T-DNA insertion and show that marker genes conferring novel traits do not generally perturb the global expression of plants.

As expected, differences between the transcriptomes of the WT and transgenic lines were dramatic under the influence of glufosinate. Differences were found that were specific to transgenic lines and did not reflect the normal responses of the WT plants to glufosinate. As these genes were not activated in WT plants, we must consider the possibility that they are responding to other acetylation products of PAT or metabolic derivatives of l-PPT. A previous study has shown that the expression of the bar gene can be negatively correlated with the viability of Oregon Wolfe barley dominant hybrids (Bregitzer et al., 2007). The authors suggested that the PAT enzyme may acetylate glutamate, a structural analogue of glufosinate. If this is true, the substrate specificity of PAT may be less specific than previously thought, resulting in pleiotropic effects. Alternatively, unknown downstream derivatives of glufosinate may have toxic effects on plants. This is not expected, as studies on several glufosinate-tolerant plants have identified N-acetyl-l-glufosinate following glufosinate treatment (Tshabalala, 1993; Burnett, 1994; Rupprecht and Smith, 1994; Thalacker, 1994; Rupprecht et al., 1995; Stumpf, 1995; Allan, 1996), and several studies have provided evidence for the stability of these intermediates in the treated plant (OECD, 2002). Although the source of toxicity in transgenic plants is unknown, our microarray data clearly revealed the activation of several stress-related genes, including detoxification-related genes, in glufosinate-tolerant plants that were exposed to the herbicide substrate. It is possible that the sensitivity of the profiling strategy used here revealed key responses to glufosinate and its derivatives that offset any phenotypic effects in the plants. This interpretation seems likely, because our previous research has shown that other herbicides, in particular imidazolinone, do not generate similar pleiotropic effects in tolerant Arabidopsis carrying the csr1-1 mutation, whereas WT Arabidopsis shows the activation of detoxification genes as an early response to imidazolinone (Manabe et al., 2007).

The early and late responses to glufosinate included many diverse genes known to be implicated in detoxification pathways in the presence and absence of bacterial PAT. Our data demonstrated broad-ranging effects of glufosinate on several metabolic pathways, including nitrogen assimilation and metabolism, that could result in the accumulation of toxic metabolites, including ammonium. Other microarray studies have also shown the induction of several detoxification genes in response to nitrogen and nitrogen limitation in WT plants (Peng et al., 2007). Plants have evolved detoxification systems that can protect them against a broad range of chemicals, including herbicides (Yuan et al., 2007). Interestingly, recent studies have shown that more than 300 biotypes of weeds have evolved resistance to one or more of the major groups of herbicides (Yuan et al., 2007). Detoxification genes, such as P450 genes, have been linked to naturally occurring herbicide resistance in weeds, and ABC transporters have been used to achieve nonspecific resistance to chemicals as diverse as antibiotics (Yuan et al., 2007). The significance of these systems was demonstrated by the finding that several P450 genes might be involved in multiple herbicide resistance, and that a single P450 gene in transgenic plants can confer resistance to up to 13 different herbicides (Robineau et al., 1998). The induction of detoxification-related genes by different herbicides now appears to be a common response (Hirose et al., 2005; Madhou et al., 2006; Manabe et al., 2007). Our study showed that many of the detoxification-related genes were not induced in the transgenic plants expressing PAT, and were specific to WT plants. Future studies are needed to understand the different detoxification pathways that occur in WT plants and plants in which glufosinate is metabolized through acetylation.

GS activity is essential for plant growth and development (Limami et al., 1999), and the inhibition of GS by glufosinate has been well studied (Wehrmann et al., 1996). The consequence is plant death resulting from the coincidental occurrence of ammonia toxicity and the perturbation of pathways downstream of GS. The integration and cross-talk among the pathways that are affected during this process are presumably very complicated and not yet well understood. Gas chromatography-mass spectrometry-based metabolite profiling analysis of Medicago truncatula root nodules treated with glufosinate showed a reduction in glutamine, glutamate, asparagine and alanine, but elevated levels of leucine, valine, methionine, threonine and isoleucine (Barsch et al., 2006). Our microarray data provide a first step in an understanding of this process by identifying the sequence of transcriptome changes that occur on exposure to glufosinate. The functional classification of genes has revealed the coordinated regulation of several pathways. The initial response includes the regulation of some metabolic pathways other than glutamine biosynthesis, the activation of detoxification processes and the activation of general stress and hormonal regulation pathways. These are directly linked to plant death by the regulation of several metabolic pathways, including the inhibition of the biosynthesis of several amino acids, inhibition of photosynthesis and the alteration of development through leaf senescence-related processes (Figure 5). The effectiveness of glufosinate as a herbicide lies not only its specificity to GS, but also in the broad effects of ammonium toxicity. Transgenic plants may be useful in revealing the pathways that interact with the acetylated derivatives of the herbicide and may shed light on the pleiotropic effects that are now becoming recognized as by-products of the use of the bar gene.

In conclusion, this study has shown that the application of glufosinate to Arabidopsis progressively and precisely results in the dramatic reprogramming of the transcriptome. The insertion of the bacterial bar gene into Arabidopsis is responsible for only a small number of changes to the transcriptome; however, exposure to glufosinate activates a specific set of genes that are unique to transgenic plants and not WT plants, raising speculation that glufosinate or a metabolic derivative of glufosinate activates unique detoxification pathways to offset any effects on plant growth and development.

Experimental procedures

Transgenic Arabidopsis and vectors

The pCAMBIA3300 transformation vector (CAMBIA), which harbours the pat gene driven by a tandem repeat of the cauliflower mosaic virus (CaMV) 35S promoter and the polyA signal of CaMV 35S, was introduced into Agrobacterium tumefaciens strain GV3101. This bacterial strain was used for the transformation of Arabidopsis WT Col-0 via the floral dip method (Clough and Bent, 1998). The copy number of the inserted gene was determined by Southern blot and segregation analysis.

Glufosinate treatment and growth measurements

T3 seeds were surface sterilized with 25% (v/v) commercial Clorox (final concentration of 1.3% sodium hypochlorite) and 0.05% (v/v) Triton X-100 (Fisher Scientific, Hampton, NH, USA) for 20 min, and then rinsed four times with distilled water. Seeds from independent transgenic lines and the progenitor WT Col-0 line were germinated in MS medium for 7 days. The seedlings were then transferred to either MS medium (control) or MS medium containing glufosinate-ammonium (Sigma-aldrich.com 45520) (concentration range, 0.1–100 µg/mL). The plates were orientated in a vertical (for root growth measurements) or horizontal position and data were collected at intervals up to 2 weeks. Root length measurement was conducted using an image analysis system, as described by Buer et al. (2000). Total fresh weight was measured for a group of 10 seedlings. The leaf number was counted under a microscope at the end of the experiment.

Microarray hybridization and analysis

RNA was extracted using an RNeasy Plant Mini Kit (Qiagen, Germantown, MD, USA), according to the manufacturer's protocol. The quality of the RNA was determined using a 2100 Bioanalyser (Agilent Technologies, Santa Clara, CA, USA). cRNA synthesis, hybridization and scanning were performed at the Botany Affymetrix GeneChip Facility, University of Toronto (Toronto, ON, Canada). Microarray analysis was performed in triplicate and the data were normalized within each group using RMA (Millenaar et al., 2006), with the R software packages by Bioconductor (http://www.bioconductor.org/) (Gentleman et al., 2004). The lists of differentially expressed genes were generated by the affylmGUI package (Gentleman et al., 2004).


The authors are grateful to Dr Nicholas Tinker for reviewing the manuscript before submission. The research was supported through a research contract to Agriculture and Agri-Food Canada (AAFC) from the Plant BioSafety Office and Feeds Section of the Canadian Food Inspection Agency.