Plastid-expressed 5-enolpyruvylshikimate-3-phosphate synthase genes provide high level glyphosate tolerance in tobacco


For correspondence (fax +00 636 7376060; e-mail


Plastid transformation (transplastomic) technology has several potential advantages for biotechnological applications including the use of unmodified prokaryotic genes for engineering, potential high-level gene expression and gene containment due to maternal inheritance in most crop plants. However, the efficacy of a plastid-encoded trait may change depending on plastid number and tissue type. We report a feasibility study in tobacco plastids to achieve high-level herbicide resistance in both vegetative tissues and reproductive organs. We chose to test glyphosate resistance via over-expression in plastids of tolerant forms of 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS). Immunological, enzymatic and whole-plant assays were used to prove the efficacy of three different prokaryotic (Achromobacter, Agrobacterium and Bacillus) EPSPS genes. Using the Agrobacterium strain CP4 EPSPS as a model we identified translational control sequences that direct a 10 000-fold range of protein accumulation (to >10% total soluble protein in leaves). Plastid-expressed EPSPS could provide very high levels of glyphosate resistance, although levels of resistance in vegetative and reproductive tissues differed depending on EPSPS accumulation levels, and correlated to the plastid abundance in these tissues. Paradoxically, higher levels of plastid-expressed EPSPS protein accumulation were apparently required for efficacy than from a similar nuclear-encoded gene. Nevertheless, the demonstration of high-level glyphosate tolerance in vegetative and reproductive organs using transplastomic technology provides a necessary step for transfer of this technology to other crop species.


Glyphosate is a broad-spectrum, post-emergence herbicide that blocks plant growth by inhibiting the production of aromatic amino acids, leading to arrest of protein production and prevention of secondary compound formation. Glyphosate inhibits 5-enolpyruvylshikimate- 3-phosphate synthase (EPSPS), a nuclear-encoded chloroplast-localized enzyme (della-Cioppa et al., 1986) in the shikimic acid pathway of plants and microorganisms. EPSPS normally catalyzes the reaction of shikimate-3-phosphate (S3P) and phosphoenolpyruvate to yield 5-enolpyruvylshikimate-3-phosphate. Glyphosate does not bind to free EPSPS enzyme (Anderson et al., 1988; Padgette et al., 1988), but rather forms an EPSPS–S3P–glyphosate complex that competes for phosphoenolpyruvate binding (Franz et al., 1997; Gruys et al., 1992). Glyphosate also inhibits import of EPSPS into the chloroplast (della-Cioppa and Kishore, 1988), which may contribute to the herbicide mode of action.

Transgenic plants engineered with glyphosate tolerance have been developed by over-production of wild-type EPSPS (Shah et al., 1986) or EPSPS enzymes with decreased affinity for glyphosate (Comai et al., 1985; Padgette et al., 1996). However, commercial production of glyphosate-resistant crops with acceptable levels of tolerance to the herbicide have been achieved only using the latter approach. The commercial herbicide Roundup is used only when weed control is needed. For example, in 1997 US growers used only Roundup on 83% of resistant soybeans (Kishore and Shewmaker, 1999). Thus commercial production of glyphosate-resistant crops has greatly simplified field management strategies, resulting in increased yield.

It has been shown that the presence of an alanine instead of glycine at amino-acid residue 96 (G96A) of the glyphosate-resistant Escherichia coli EPSPS interferes with glyphosate binding by a mechanism common to all EPSPS enzymes tested. All EPSPS enzymes containing the corresponding G96A mutation have an increased apparent Km (appKm) for phosphoenolpyruvate relative to the wild-type EPSPS (Kishore et al., 1986; Padgette et al., 1988; Stalker et al., 1985). The glyphosate-resistant EPSPS gene currently utilized in nuclear transgenic crops is from Agrobacterium strain CP4 because of this enzyme's high catalytic activity and very low affinity for glyphosate (Barry et al., 1992; Padgette et al., 1996).

The possibility of outcrossing herbicide resistance transgenes to related weeds or neighboring non-transgenic crops is a concern in field management strategies. The potential of plastid transgenic (transplastomic) plants to reduce these outcrossing concerns (Scott and Wilkinson, 1999), due to maternal inheritance of plastids in most crop plants, may provide a potential alternative management strategy in some cases. Recently, plastid transformation technology was used to introduce into the tobacco plastid genome a mutant petunia EPSPS (Padgette et al., 1991) that is glyphosate tolerant but less catalytically active than the wild-type enzyme (Daniell et al., 1998). Transplastomic plants carrying this gene showed vegetative glyphosate tolerance up to 5 mm (≈8 oz acre−1 Roundup). However, glyphosate treatments of significantly higher rates in commercial crops are required for weed control, necessitating resistance of crop plants to at least ≈64 oz acre−1 Roundup.

We report here a comprehensive feasibility study testing the utility of plastid-expressed EPSPS genes to provide very high levels of glyphosate tolerance in plants. For this study, we have over-expressed glyphosate-tolerant forms of prokaryotic (Bacillus, Achromobacter and Agrobacterium) EPSPS genes in plastids, and tested expression by immunological, enzymatic and in vitro and in planta analysis. EPSPS was used as a model to identify plastid translational control sequences that direct a 10 000-fold range in protein accumulation, up to >10% EPSPS in soluble leaf protein. Our results show that very high-level glyphosate tolerance is attainable, although levels of resistance in vegetative and reproductive organs differed. Paradoxically, higher levels of plastid-expressed EPSPS protein accumulation were apparently required for efficacy than from a similar nuclear-encoded gene. Importantly, we show proof of biological containment of the transgenes by strict maternal inheritance of resistance in reciprocal crosses.


Chimeric EPSPS genes

For testing glyphosate tolerance via plastid expression, well characterized bacterial EPSPS genes were employed. The CP4 gene was chosen because it is the most highly glyphosate-resistant EPSPS identified to date, and provides high resistance when expressed as a nuclear-encoded plastid-targeted enzyme in transgenic crops (Franz et al., 1997). Inspection of the native CP4 sequence indicated only ≈45% plastid-preferred codons, according to the data of Bonham-Smith and Bourque (1989). To test the effect of amino-acid codon usage on plastid expression, a synthetic CP4 gene was constructed that carries predominantly plastid-preferred codons (≈77%). Additionally, a glyphosate-tolerant Achromobacter strain LBAA EPSPS (Hallas et al., 1988), carrying a glycine-to-alanine mutation at amino acid 100 (LBAA G100A), and the naturally resistant Bacillus wild-type aroE gene (Henner et al., 1986) were tested because of their high kinetic binding of phosphoenolpyruvate and low affinity for glyphosate.

To achieve high levels of EPSPS protein accumulation in plastids, a systematic series of chimeric genes were tested. All constructs (Figure 1a) carried the strong, constitutive plastid PrrnPEP promoter (Svab and Maliga, 1993) and the 3′-mRNA stability element Trps16 (Staub and Maliga, 1994). To test the influence of different ribosome-binding site (RBS) regions on EPSPS gene translation, the plastid rbcL gene RBS and the bacteriophage gene 10 leader (G10L) RBS were compared (constructs pMON30123 and pMON38773, respectively). The rbcL and G10L RBS regions differ in their length, putative Shine–Dalgarno (S–D) sequences and the spacing of the S–D to the translation start codon (Figure 2). The plastid rbcL RBS has been used extensively in plastid transgenes. However, the G10L sequence has been shown to significantly increase the translatability of foreign genes in E. coli (Olins et al., 1988) and plastids (Staub et al., 2000). In constructs pMON45259 and pMON49218, the 14 N- terminal amino acids of the GFP protein, which is efficiently translated in plastids (Sidorov et al., 1999), was N-terminally fused to CP4. This fusion was predicted to enhance translation or increase fusion protein stability. Plasmid pMON49218 carries the additional Prrn−62 NEP promoter element downstream of PrrnPEP. The Prrn−62 NEP promoter was shown to be transcriptionally active in non-green tissues (Hajdukiewicz et al., 1997), and was used to test for improved expression in reproductive organs.

Figure 1.

Integration of chimeric EPSPS genes into the tobacco plastid genome.

(a) The chimeric EPSPS genes in the pMON vectors are expressed from the Prrn promoter with additional control elements as described in the text, and the Trps16 3′-end needed for mRNA stability. Numbers to the right (in kb) indicate the predicted hybridizing bands when digested with BamHI and probed with P1.

(b) The chimeric EPSPS genes are cloned next to the selectable spectinomycin-resistant gene aadA, in transformation vector pPRV111B (Zoubenko et al., 1994). Note that the transformation vector extends from EcoRI (RI) to BglII (Bg). The transgenes are targeted to the intergenic region between the plastid trnV and rps7/3′-rps12 operon by homologous recombination within the flanking regions. Arrows above the genes indicate the direction of transcription.

(c) Total cellular DNA from both wild-type tobacco (Nt) and the transplastomic lines (Nt-pMON number) were digested with BamHI, electrophoresed and blotted onto nitrocellulose. The 2.4 kb EcoRI/EcoRV DNA fragment used as probe (P1) hybridizes to a 3.3 kb BamHI plastid DNA fragment in the wild-type line (Nt) and to transgenic fragments whose sizes are indicated in Figure 1(a). Note the lack of the 3.3 kb wild-type fragment in transplastomic lines indicating a homoplasmic population. Size (in kb) of selected molecular-weight markers is shown on right. Abbreviation: Ba, BamHI; RI, EcoRI; RV, EcoRV; Bg, BglII; MCS, multiple cloning sites.

Figure 2.

Comparison of the rbcL and gene 10 leader ribosome-binding site (RBS) regions.

The DNA sequences that are complementary to the anti-Shine–Dalgarno (anti-S–D) of the plastid 16S rRNA (rrn16 gene) are shown by vertical connecting lines, with potential G–U pairing shown by a dot. The methionine (ATG) start codon is underlined. The nucleotide (nt) spacing between the S–D and the start codon is noted. The rbcL and the gene 10 RBS sequences are from Svab and Maliga (1993) and Studier et al. (1990), respectively. The nucleotides in lower case (cc) were changed to create an NcoI restriction site.

Plastid transformation and identification of transplastomic plants

The chimeric EPSPS genes were cloned into chloroplast transformation vector pPRV111B (Zoubenko et al., 1994; Figure 1b) linked to the selectable marker gene aadA, which confers resistance to spectinomycin and streptomycin. The chimeric genes are targeted to the duplicated inverted repeat region, between the plastid trnV gene and the rps7/3′-rps12 operon, via homologous recombination within the flanking plastid DNA sequences in the transformation vector. Tungsten microparticles were coated with the various transformation vectors and bombarded into tobacco leaves. Putative transformants were selected on spectinomycin medium, and screened on streptomycin-containing medium to eliminate spontaneous mutants as described (Svab and Maliga, 1993). Transformed lines were monitored by PCR or Southern blot analysis during one to three rounds of plant regeneration on spectinomycin medium until homoplasmic plants were obtained.

Southern blot analysis was used to verify homoplasmy in all transplastomic lines (representatives shown in Figure 1c). Total cellular DNA was isolated from leaf tissue of sterile tissue culture-grown plants, digested with BamHI, electrophoresed and subsequently transferred to nitrocellulose. The BamHI-digested DNA was then probed with the EcoRI/EcoRV ptDNA flanking sequence (Probe P1). The hybridization pattern predicted for insertion of the transgenes is shown in Figure 1(a). The ptDNA probe hybridized exclusively to the wild-type 3.3 kb DNA fragment in the control non-transformed tobacco plant. In the transplastomic lines, no wild-type DNA is present and only transgenic bands diagnostic for insertion of the foreign genes into the plastid genome are observed. This analysis confirms a homoplasmic population of transformed plastids.

Expression of CP4 in leaves

Using anti-CP4 antibodies to probe Western blots, CP4 protein accumulation in leaves of homoplasmic transformants was analyzed. The data from leaves of mature R0 plants and 14-day-old in vitro-grown R1 seedlings were similar; representatives are shown in Figure 3 and Table 1. Quantification of the immunoreactive bands was based on a dilution series of purified CP4 protein.

Figure 3.

Expression of CP4 in 14-day-old seedling leaves of transplastomic lines.

(a) Total soluble protein (amounts indicated above the lanes) on the Western blot was probed with anti-CP4 polyclonal antiserum. A dilution series of purified CP4 standard was used for quantification. Note the mobility difference of the GFP-CP4 fusion proteins.

(b) Total cellular RNA (5 µg per lane) on the Northern blot was probed for CP4 (top blot). The monocistronic CP4 and the dicistronic CP4-aadA messages are indicated. The membrane was stripped and reprobed with an rrn16 gene probe to determine equal loading (bottom blot). Wild-type tobacco (Nt) was used as a control.

Table 1.  Expression of EPSPS genes and whole plant glyphosate tolerance levels
Specific activitya
(nmol min−1 mg−1)
(oz. acre−1)
(oz. acre−1)
No glyphosate1 mm glyphosate
  • ND: not determined due to the unavailability of antibodies.

  • a

    Specific activity is measured as the amount of exogenously added phosphoenolpyruvate converted to Shikimate-3-phosphate (S3P) per unit protein in the plant extract. Addition of glyphosate tests sensitivity of the EPSPS enzyme to glyphosate.

Wild-type (Nt) endogenous EPSPS 3.4000
Nt-pMON10154Nuclear FMV promoterCP4 native0.0419.018.412896
(nuclear control)
Nt-pMON30123PrrnPEP/rbcLCP4 native0.0014.0000
Nt-pMON30130PrrnPEP/rbcLCP4 synthetic0.0026.2000
Nt-pMON38773PrrnPEP/G10LCP4 native0.216.76.73216
Nt-pMON38798PrrnPEP/G10LCP4 synthetic0.317.214.73216
Nt-pMON45259PrrnPEP/G10L14aaGFP-CP4 synthetic>10.0689.6633.512864
Nt-pMON49218PrrnPEP + NEP/G10L14aaGFP-CP4 synthetic>10.0771.9690.012896

The Western blot results show a dramatic difference in CP4 protein accumulation among the various lines. The CP4 genes expressed from the Prrn/rbcL (Nt-30123 and Nt-30130) had barely detectable CP4 that accumulated to only approximately 0.001 and 0.002% of total soluble protein, respectively (Figure 3a; Table 1). A dramatic increase of ≈200-fold CP4 protein accumulation was observed in constructs utilizing the G10L for translation. Lines Nt-38773 and Nt-38798 accumulated CP4 to approximately 0.2 and 0.3% of total soluble protein, respectively (Figure 3a; Table 1). These results also suggest that optimization of plastid codon usage may lead to increased protein accumulation. In the Nt-45259 and Nt-49218 plants, the GFP-CP4 fusion protein accumulates to >10% of total soluble protein, an additional ≈50-fold increase in expression. Preliminary protein pulse-labeling experiments performed using these same tissues indicated a marked increase in the rate of translation between the G10L-containing construct in Nt-38798 and the GFP-CP4 fusion construct in Nt-45259 (data not shown). Therefore an overall 10 000-fold improvement in CP4 protein accumulation was achieved by manipulation of the translation signals driving the chimeric CP4 genes. The control nuclear transgenic line, Nt-10154, which shows a high tolerance level to glyphosate (see below), accumulated relatively low amounts of CP4 protein in leaves (0.04% total soluble protein; Table 1).

Northern blot analysis was performed to determine if changes in CP4 mRNA accumulation could account for the dramatic changes in plastid protein accumulation. Total cellular RNA extracted from the same leaf samples as above was used for the analysis. The CP4 coding region probe identified both monocistronic CP4 and dicistronic CP4-aadA mRNA in the transplastomic lines (Figure 3b). Dicistronic mRNA is a consequence of inefficient transcription termination at the Trps16 3′-end, as shown previously (Staub and Maliga, 1994). The Northern blot was stripped and reprobed for plastid rrn16 mRNA as an RNA loading control. Total CP4 mRNA accumulation differed by less than fourfold among the transplastomic lines, as expected since each of the chimeric genes carries the same promoter. This fourfold mRNA difference clearly cannot account for the 10 000-fold difference in protein accumulation.

Expression of CP4 in reproductive organs

Glyphosate is transported and accumulates in floral tissues and reproductive organs. Conversely, the number of plastids and plastid genomes is expected to decrease in these non-photosynthetic tissues (Bendich, 1987; Pyke and Page, 1998). For this reason it was important to analyze the expression of EPSPS in floral tissues of the transplastomic plants. Total cellular protein and RNA was extracted from immature anthers, ovaries, flower petals and mature leaves from transplastomic lines Nt-45259 and Nt-49218, and used in Western and Northern blot analysis, respectively.

The Western blot analysis indicated that CP4 protein levels were approximately 10-fold lower in flower petals and at least 50-fold lower in both immature anthers and ovaries than in mature leaf tissue (Figure 4a). Both transplastomic lines had similar CP4 accumulation in each tissue type.

Figure 4.

Expression of the CP4 in reproductive tissues.

(a) Total soluble protein (amounts indicated above the lanes) on the Western blot was probed with anti-CP4 polyclonal antiserum. A dilution series of purified CP4 standard was used for quantification. Note additional bands in flower petals that may represent CP4 degradation products.

(b) Total cellular RNA (6 µg per lane) on the Northern blot was probed for CP4 (top blot). The monocistronic CP4 and the dicistronic CP4-aadA messages are indicated. A dilution series of Nt-49218 RNA was used for quantification. Abbreviations: L, mature leaf; IA, immature anther; O, ovary; P, flower petal. Wild-type tobacco (Nt) was used as a control.

Northern blot analysis of the same floral tissues using the CP4 coding region probe revealed both monocistronic CP4 and dicistronic CP4-aadA mRNA (Figure 4b), as observed previously. Both transgenic lines accumulated high levels of CP4 total mRNA in leaf tissues, and much less in the floral tissues. Quantification of total CP4 mRNA using a dilution series of a leaf sample indicated that the floral tissues accumulated from ≈10-fold less (in petals) to ≈50-fold less (in immature anthers) total CP4 mRNA than in leaves. Therefore decreased CP4 protein accumulation in floral tissues paralleled the decreased level of CP4 transcripts.

EPSPS specific activity

Measurement of EPSPS enzyme activity allows a quantitative comparison between EPSPS genes in the different transplastomic lines (Table 1). Developmentally uniform leaves of 4-week-old plants derived from seedlings were used for the analysis. The EPSPS activity was measured as the amount of exogenously added radiolabeled phosphoenolpyruvate that is converted to EPSP per unit protein in the plant extract. Activity in the presence of the competitive inhibitor, glyphosate, is used to measure the tolerance level of the EPSPS enzyme.

Wild-type tobacco plants carry the endogenous nuclear-encoded EPSPS and have a low level of enzyme activity. The activity is completely inhibited by the addition of glyphosate, as expected for the wild-type, glyphosate-sensitive enzyme. In contrast, for transplastomic lines that express CP4 the specific activity of the enzyme increases dramatically as the CP4 accumulates. In each case there is no significant effect of glyphosate on EPSPS enzyme activity, indicating tolerance to glyphosate. Importantly, very high EPSPS activity in the GFP-CP4 fusion protein with no inhibition by glyphosate suggests that the addition of N-terminal amino acids does not significantly affect enzyme activity. Indeed, CP4 carrying an N-terminal chloroplast transit peptide retains full activity in vitro (della-Cioppa et al., 1986). The Achromobacter and Bacillus EPSPS had enzyme activities comparable to the very highly expressed CP4 enzymes, indicating that these genes were probably also well expressed in plastids.

EPSPS activities among plastid- and nuclear-expressed CP4 genes were compared to determine if significant differences exist. For the comparison, nuclear transgenic lines were selected in tissue culture for glyphosate resistance after particle bombardment with plasmid pMON10154 which carries a constitutively expressed CP4 gene driven by the strong figwort mosaic virus promoter. Representative transgenic lines were chosen at random for further study; enzyme activity from line Nt-10154 was typical (data not shown) and had an activity similar to the moderately expressed plastid transgenic lines Nt-38773 and Nt-38798.

Glyphosate tolerance in transplastomic plants

In fully sensitive plants, glyphosate induces severe chlorosis of newly emerged leaves and meristem tissues and severe stunting, and completely blocks normal floral development and seed set. To test the efficacy of the plastid-expressed EPSPS genes, glyphosate tolerance was evaluated for the transplastomic, nuclear transgenic and wild-type tobacco plants. The level of vegetative tolerance for each line was scored 2 weeks after spray, with minimum tolerance levels defined as slight chlorosis of newly emerged leaves, and stunting. Reproductive tolerance levels were scored at the end of anthesis, with tolerance defined as seed set at least 50% of untreated plants. Note that commercial field levels of Roundup tolerance are set at a minimum of 64 oz acre−1 for most dicot species.

The results of Roundup spray treatment are shown in Table 1. Wild-type plants showed severe vegetative damage and died within several weeks of being sprayed at the lowest dose tested (16 oz acre−1). For transplastomic plants expressing the CP4 protein, vegetative and reproductive tolerance levels increased with higher CP4 protein levels. For example, Nt-38773 and Nt-38798 lines, which have intermediate CP4 accumulation levels, survived moderate doses of Roundup (32 oz acre−1 vegetative, 16 oz acre−1 reproductive). In contrast, the Nt-45259 and Nt-49218 that express CP4 at >10% total soluble protein survived extremely high levels of Roundup in vegetative tissues (128 oz acre−1) and lower levels in reproductive organs (64 and 96 oz acre−1, respectively). The Achromobacter EPSPS in Nt-45204 plants provided high-level Roundup tolerance, whereas the Bacillus EPSPS in the Nt-45201 line provided only moderate tolerance (Table 1). This may be due to differences in expression patterns of the two genes; however, no further analysis was performed.

The nuclear transgenic line Nt-10154 that expressed CP4 to relatively low levels in young leaves (≈0.04% of total soluble protein) was also tested for whole-plant resistance to Roundup. Interestingly, this line showed very high resistance in both vegetative and reproductive organs (Table 1; see Discussion).

Maternal inheritance of glyphosate resistance in seedlings

To evaluate maternal inheritance of glyphosate resistance we developed a seedling bleaching assay. Wild-type seedlings were tested for germination and extent of bleaching at several levels of glyphosate. Glyphosate at 200 µm was seen to completely bleach the wild-type seedlings with no effect on germination frequency.

Transplastomic lines were grown to maturity in the greenhouse and allowed to self-seed, or were crossed with wild-type tobacco. Figure 5 shows an example of the assay using line Nt-45204. The seedlings from Nt-45204 self-pollination were uniformly green, indicating tolerance to glyphosate. Similarly, uniform resistance was observed in seedlings from a cross using the transplastomic line as the female recipient and wild-type tobacco as pollen donor. In contrast, when the transplastomic line served as the pollen donor all of the seedlings assayed were sensitive to glyphosate and were bleached.

Figure 5.

Maternal inheritance of the EPSPS transgene in transplastomic line Nt-45204.

Self seed from the wild-type control (WT♀ × WT♂), transplastomic line (Nt-45204♀ × Nt-45204♂) or reciprocal crosses with the transgenic line as pollen recipient (Nt-45204♀ × WT♂) or pollen donor (WT♀ × Nt-45204♂) were germinated on medium containing 200 µm glyphosate. Note that resistant seedlings are uniformly green whereas sensitive seedlings are bleached.


Glyphosate-resistant crops generated through nuclear transgenic technology are currently marketed throughout the USA and several other countries. We report here the development of very high levels of glyphosate tolerance via plastid transformation technology, which has the potential advantage of increased biological containment by apparent elimination of the transmission of transgenes through pollen (Daniell et al., 1998; Scott and Wilkinson, 1999). Although not considered as a commercial target, tobacco was chosen as a model plant due to routine plastid transformation technology (Maliga, 1993). Agrobacterium CP4 EPSPS was used for most of our studies because of its success in nuclear transgenic plants and the availability of specific antibodies. Additionally, the in vitro enzyme activity, in planta tolerance and seedling germination assays allowed efficacy testing of multiple plastid-expressed prokaryotic EPSPS genes. Although a Petunia EPSPS gene was recently expressed in plastids (Daniell et al., 1998), high levels of glyphosate resistance were not achieved.

Significant improvements in existing plastid gene expression elements were necessary to achieve very high levels of glyphosate tolerance. An overwhelming 10 000-fold increase in protein accumulation was achieved by modification of only translational control sequences. First, a dramatic increase of ≈200-fold in protein accumulation resulted from replacement of the rbcL RBS (Svab and Maliga, 1993) with the RBS from the heterologous bacteriophage T7 G10L (Figure 3a; Table 1). The G10L sequence contains one more base match to the anti-S–D sequence on the plastid 16SrRNA, more bases between the S–D and the ATG start codon, and a larger RBS region compared to the rbcL leader (Figure 2). Data from E. coli suggest that longer S–D sequences, which may create a more stable environment for the mRNA−16SrRNA interaction, and the position of the S–D sequence relative to the start codon, may affect the rate of translation (Gold, 1988). As plastids share similarities with prokaryotes in their translational apparatus (Harris et al., 1994), it is not surprising that some RBS preferences similar to E. coli may also be preferred in plastids. Furthermore, the enhanced translation governed by the G10L is not gene-specific, as a similarly expressed human protein also accumulated to very high levels in plastids (Staub et al., 2000). Second, a further increase in protein accumulation of ≈50-fold was observed in the transplastomic lines that express the GFP-CP4 fusion. GFP was chosen for the fusion because it was previously shown to accumulate to high levels in plastids (Sidorov et al., 1999). Preliminary data from pulse-chase experiments indicate that the 14 amino acid N-terminal GFP fusion to CP4 increases the rate of translation. Further experiments are under way to determine the mechanism of translation rate enhancement conferred by the GFP sequences. Interestingly, the high rate of CP4 protein translation and accumulation in the transplastomic plants had no apparent phenotypic consequence, as no differences in growth or development from wild-type plants were observed.

Bacterial genes have been routinely expressed to high levels in the plastid compartment (McBride et al., 1995). Nevertheless, we constructed a synthetic CP4 EPSPS containing predominantly plastid-preferred codons. Expression of this gene was enhanced ≈1.5-fold relative to the native CP4 gene that has only ≈45% plastid-preferred codons. This increased expression was moderate compared to the improvements discussed above suggesting that codon optimization was not critical in this case; rather, translation initiation is the major rate-limiting step for CP4 expression. We also tested EPSPS genes from other prokaryotic sources. Although no attempt was made to optimize codon usage of these genes, the Achromobacter LBAA gene driven by the G10L sequences proved fully efficacious, while the Bacillus aroE gene driven by the rbcL RBS sequence provided moderate protection to glyphosate.

Although the level of vegetative and reproductive tolerance to glyphosate increased with increasing CP4 levels, tolerance in reproductive tissues was consistently lower than in vegetative tissues of the same line (Table 1). This result is probably explained by lower protein levels in the reproductive cells targeted by glyphosate, due to a decrease in plastid copy number and/or transcriptional/translational activity in those tissues (Figure 4). Paradoxically, the nuclear transgenic line used as control that expresses a constitutive plastid-targeted CTP-CP4 to relatively low levels (0.04% total soluble protein in leaves), provided high level of glyphosate tolerance similar to that of transplastomic lines expressing CP4 to >10% of tsp in leaves (Table 1). One reason for this discrepancy between protein level and tolerance may be that the nuclear-encoded gene is expressed at a high enough level to confer resistance in the appropriate cell types, whereas the plastid transgene is not. As the floral and vegetative meristems are known to act as a sink for glyphosate (Franz et al., 1997; Padgette et al., 1996), these meristematic cells may be more susceptible due to lower plastid numbers in these cells. Future research in the transplastomic approach will be designed to identify cell type-specific expression elements that may be useful to circumvent this potential effect.

Experimental procedures

Construction of chimeric EPSPS genes and transformation vectors

For plastid transformation, two Agrobacterium strain CP4 EPSPS genes were used. The native CP4 gene has been described (Barry et al., 1992). The synthetic CP4 gene was constructed using synthetic oligonucleotides to contain predominantly plastid-preferred codons according to the data of Bonham-Smith and Bourque (1989). The CP4 genes were driven by a variety of expression signals:

pMON30123 and pMON30130 carry the native and synthetic CP4 gene, respectively, driven by the plastid-encoded polymerase (PEP) promoter from the rrn16 gene (Prrn) with the ribosome-binding site (RBS) from the plastid rbcL gene (PrrnPEP/rbcL;Svab and Maliga, 1993).

pMON38773 and pMON38798 carry the CP4 native or synthetic gene, respectively, driven by the Prrn PEP promoter (position 1–100, Svab and Maliga, 1993) with the RBS from the G10L (PrrnPEP/G10L). The G10L sequences (Studier et al., 1990) from position +22 to the start codon were included; the C at position +23 was converted to G to destroy the XbaI site. The AT at positions +62 and +63 were converted to CC to create an NcoI restriction site.

pMON45259 and pMON49218 each carries the CP4 synthetic gene fused at its 5′ end to the first 14 amino acids of the GFP protein. The gfp-CP4 synthetic gene in pMON45259 is expressed from the PrrnPEP/G10L sequences. To drive expression of the gfp-CP4 gene in pMON49218, the PrrnPEP sequence was extended to include the nuclear-encoded polymerase (NEP) promoter fused immediately downstream, followed by the G10L. The Prrn−62 NEP from positions −33 to +15 was used (Hajdukiewicz et al., 1997), except that the A at position −26 was converted to T, and the T at position −2 was converted to G to eliminate potential ATG start codons. The C at position +23 of the G10L which was converted to a G in plasmids pMON38773 and pMON38798 was maintained as a C in plasmid pMON49218.

pMON45201 carries the coding region of the wild-type Bacillus aroE gene (Henner et al., 1986) driven by PrrnPEP/rbcL (Svab and Maliga, 1993).

pMON45204 carries the coding region of the EPSPS gene from Achromobacter strain LBAA (22), driven by the PrrnPEP/G10L as in pMON38773.

All of the chimeric EPSPS genes contain the Trps16 3′-end (Staub and Maliga, 1994) required for mRNA stability. The chimeric EPSPS genes were cloned into pPRV111B (Zoubenko et al., 1994) which carries the aadA gene, conferring spectinomycin resistance, as a selectable marker, and targets the chimeric genes to the plastid inverted repeat region.

Nuclear transformation vector pMON10154 carries the CP4 native gene expressed from the constitutive FMV promoter and the petunia HSP70 leader, and has the E9 terminator. Targeting to plastids is by the chloroplast transit peptide of the petunia EPSPS translationally fused to the N-terminus of the CP4 protein (CTP-CP4).

Growth of plants and selection for transformants

Nicotiana tabacum cv. Petit Havana (tobacco) was maintained aseptically on phytagel-solidified medium containing MS salts (Murashige and Skoog, 1962) and Gamborg's B5 vitamins (Gamborg et al., 1968) with 3% sucrose at 24°C with 16 h photoperiod. Transformation, selection and regeneration of plastid-transformed plants was as described (Staub et al., 2000; Svab and Maliga, 1993). A minimum of three plastid-transformed lines from each construct were obtained and further characterized. As all plastid transgenics had similar expression patterns of the transgene, due to the lack of position effects as expected for plastid transgenes, data for a single representative line are shown.

The homozygous nuclear-transformed tobacco line was generated in Nicotiana tabacum cv. Samsun by Agrobacterium- mediated transformation with vector pMON10154. This line was selected directly in tissue culture for glyphosate resistance (data not shown).

For seed germination assays, seeds of self and reciprocal crosses from plastid transformants with wild-type plants were aseptically grown on phytagel-solidified MS medium with 3% sucrose, supplemented with 200 µm glyphosate, at 24°C with 16 h photoperiod.

Glyphosate tolerance spray tests

Homoplasmic plastid transformants were clonally propagated in sterile culture and then planted in soil in the greenhouse. The plants were sprayed with the herbicide Roundup (active ingredient isopropylamine salt of glyphosate, 41.0%), at the 5–6-leaf stage about 2 weeks after transplanting. Vegetative injury and fertility data were obtained by visual observation 2 weeks after the spray and at maturity, respectively. Individual plants were scored for tolerance to Roundup using a scale of 0 (normal plant) to 4 (dead plant) for vegetative injury, and a scale of 0 (fertile, no delay in maturity, lots of seeds) to 4 (sterile, extreme delay in maturity, no seeds) for fertility. The injury data for all treated plants were averaged; average values of less than 2.0 were considered tolerant to the treatment.

Gel blot analysis

Isolation of total cellular DNA and Southern blot analysis, and RNA isolation and Northern blot analysis were performed as described (Staub et al., 2000; Svab and Maliga, 1993). For CP4 mRNA probing, native and synthetic CP4 coding region DNA was mixed at a 1 : 1 ratio (based on radioactivity). Total cellular protein was isolated and Western blot analysis was performed as described (Staub et al., 2000). Antibody to CP4 was used at 1 : 1000 dilution.

Tissues for Northern and Western blot analyses of floral organs were collected from mature flowering plants grown in soil in a greenhouse. Leaf samples were from a fully expanded middle leaf, immature anthers were from floral buds about 4 days before anthesis, and ovary and the corolla section of petals were from post-anthesis flowers.

EPSPS enzyme analysis

Frozen leaf tissue was extracted in ice-cold buffer containing 100 mm Tris–HCl, 10% glycerol, 1 mm EDTA, 1 mm benzamidine, 1 mm dithiothreitol, 1 mm 4-(2-aminoethyl)-benzenesulfonyl fluoride, 0.1 mm leupeptin pH 7.4 at 1 ml buffer per 20 mg tissue. The tissue was homogenized, centrifuged at 0.02 g, then the extract was desalted using a Pharmacia Nap-5 column. The desalted extract was concentrated through a Millipore Biomax membrane (10 000 molecular weight cut-off) to less than half original volume and assayed for activity as follows. A 10 µl aliquot of the extract was added to 30 µl EPSPS assay mix at 25°C. To initiate the reaction, 10 µl [C-14]phosphoenolpyruvate was added (final concentrations in the assay: 50 mm HEPES pH 7.0, 5 mm potassium fluoride, 1 mm shikimate-3-phosphate, 0.5 mm PEP, 0.1 mm ammonium molybdate). After 5–30 min the reaction was quenched by addition of 50 µl acidic ethanol (9 : 1 ethanol : 0.1 m acetic acid pH 4.5). For glyphosate inhibition studies, assays were performed as above except that reactions contained 1 mm glyphosate.

The EPSPS assay reactions were centrifuged at 14 000 r.p.m. for 10 min before injection of 30 µl onto a Synchropack AX-100 column. (mobile phase: 0.235 m potassium phosphate pH 6.5 at 1 ml min−1). Radioactive peaks were detected using a Packard C525 flow scintillation analyzer. The percentage conversion of radiolabeled PEP to EPSP was determined in duplicate reactions and the values averaged. For determination of specific activity (nmol min−1 mg−1) the amount of protein added in the 10 µl extract was determined by the Lowry method. Glyphosate inhibition was determined by comparing specific activity in the absence or presence of glyphosate.


We thank Mary Migas and Rebecca Lemus for technical support, Sheng-Zhi Pang for construction of the pMON30123 and pMON30130 vectors, Bensong Xie for conducting glyphosate spray tests, Conrad Halling for computer analysis, and Pal Maliga, Rutgers University, for the pPRV plasmid. We also thank Paul Feng, Charley Romano, Doug Sammons and Nancy Biest Taylor for discussion, and Ganesh Kishore, Michael Montague, Charles Armstrong and Kenneth Barton for support throughout this work.