Plant Molecular Biology 8: 355-359, (1987) © Martinus Nijhoff Publishers, Dordrecht - Printed in the Netherlands
355
Biotechnology news and views Stanton B. Gelvin
Accurate expression of genes in transgenic plants: do fundamental differences exist between eukaryotes? Because of the availability and increasing ease of use of T-DNA-derived vector systems, as well as the establishment of electroporation as a reliable way of introducing D N A into plant cells, many laboratories are using transgenic plants to investigate basic processes of plant molecular biology. It was established early on that prokaryotic genes (often the source of antibiotic resistance markers) under the control of their own promoters do not function in plant cells. Functional chimaeric genes could be constructed, however, by fusing the promoters and upstream elements of constitutive plant genes (such as those from the T-DNA or cauliflower mosaic virus [CaMV]) or of regulated plant genes (such as those from rbcS or cab) to the prokaryotic coding sequences. Perhaps it was not surprising that prokaryotic transcriptional regulatory signals could not function in plants, considering the m a j o r differences in the mechanisms of gene expression between prokaryotes and eukaryotes. The question remained, however, whether genes from other eukaryotic organisms, such as animals or yeast, could be accurately expressed in plants, or whether there was even a difference between monocots and dicots in the ways in which D N A is expressed. Whereas the answers to these questions have not yet been clearly defined, recent studies have suggested that all eukaryotes may not express genes in the same ways using the same regulatory signals. et al. (Cell Early studies by Barton 32:1033-1043, 1983) indicated that a yeast alcohol dehydrogenase gene could not be expressed in tobacco crown gall tumors. Langridge et al. (EMBO J 3:2467-2471, 1984) showed, however, that two maize genes encoding 19 and 21 kd zein
proteins could be accurately transcribed in yeast. The transcripts initiated at the same nucleotides both in yeast and in maize. It is well known that yeast and animal transcriptional regulatory and R N A processing signals, while similar, may often differ significantly (see e.g. Nevins, Ann Rev Biochem 52:441-466, 1983). It is therefore not unreasonable to expect that genes from higher and lower eukaryotes may at times be expressed either correctly or aberrantly in these heterologous systems. Most animal and plant genes contain 'TATA' boxes and many have 'CAAT' boxes making up their promoter elements. The plant intron junction 'GT/AG' consensus signals are followed as in animal systems, although the polypyrimidine stretches often found upstream of most 3' splice sites in animals are often lacking in plant genes. Because of the presumed similarity of higher animal and plant transcription regulatory and R N A processing signals, one may expect that the expression of animal and plant genes, while perhaps not developmentally or environmentally regulated in the same ways, would at least be accurate in the respective heterologous systems. Indeed, Brown et al. (EMBO J 5:2749-2758, 1986) have recently shown that in vitro SP6 transcripts of two plant genes encoding a wheat amylase and a pea legumin are accurately spliced in a H e L a cell nuclear extract. Lariat RNAs similar to those seen as a result of the splicing of animal and yeast introns were also detected using these plant RNAs. In addition, an analysis of 168 plant intron sequences indicated that putative lariat branch point consensus signals very similar to those found in animal and fungal introns could be located in plant introns. The authors cautioned, however, that a final analysis of plant RNA splicing would have to await the development of plant in vitro splicing systems. Despite the heat-regulated expression of a
356 Drosophila heat shock p r o m o t e r / N P T I I fusion gene in tobacco calli (Spena et al., EMBO J 4:2739-2743, 1985), the accurate transcription and processing of animal genes in plant cells has been difficult to achieve. Koncz et al. (EMBO J 3:1029-1037, 1984) showed that a chicken c~-actin gene was not expressed in tobacco crown gall tumors, and that the initiation of transcription of a chicken ovalbumin gene was not accurate in the same system. Barta et al. (Plant Mol Biol 6:347-357, 1986) more recently investigated the processing of a h u m a n growth hormone (hGH) transcript in tobacco and sunflower crown gall tumors. In order to express the animal gene in plant cells, the authors placed the entire h G H gene (including introns and polyadenylation signal sequences) between the nopaline synthase (nos) promoter and transcription termination sequences. The fusion gene correctly initiated transcription from the nos promoter, but the h G H transcript did not utilize the animal termination signal; rather, the R N A ended using the nos poly A signal. In addition, the h G H introns were not spliced from the primary transcript in plants, although they could be an in vitro H e L a cell nuclear extract. These results suggest that, at least for some animal genes, the termination and splicing signals cannot be accurately recognized, if at all, in vivo in plant systems. As opposed to the similarities between putative animal and plant splicing signals discussed above, it has often been noted that the termination of transcription in plants may differ from that in animals: plant transcripts tend to have multiple termination sites, and the AAUAAA consensus signal found in most animal genes is more variable in plant genes (plants tend to use, in addition, AAUAAU and other variations of the animal consensus sequence). Given the variation in the ability to express and process animal genes correctly in plant cells, one may wonder whether fundamental differences in these processes may exist between monocots and dicots. Numerous studies (see e.g. Murai et al., Science 222:476-482, 1983 and Sengupta-Gopalan et al., Proc Natl Acad Sci 82:3320-3324, 1985 for phaseolin; Chen et aL, Proc Natl Acad Sci 83:8560-8564, 1986 for /3-conglycinin) have indi-
cated that certain specific dicot genes can be correctly expressed and developmentally regulated in other dicots. L a m p p a et al. (Nature 316:750-752, 1985) demonstrated that a monocot (wheat) cab transcript could be correctly initiated in transgenic tobacco plants. The gene correctly showed light/dark and tissue-specific regulation in its new host. Matzke et al. (EMBO J 3:1525-1531, 1984) demonstrated that a 19 kd maize zein gene could be accurately initiated in sunflower crown gall tumors. Goldsbrough et al. (Mol Gen Genet 202:374-381, 1986) showed that a similar 19 kd zein gene, as well as a 15 kd zein gene, could be both correctly initiated and terminated in sunflower tumors. More recently, however, Keith and Chua (EMBO J 5:2419-2425, 1986) presented data suggesting that the efficiency o f splicing of monocot and dicot transcripts, as well as the accuracy of polyadenylation of these transcripts, may differ in transgenic dicots. When a wheat rbcS gene was transferred to tobacco, it was not expressed under the control of its own promoter. When linked to a CaMV 35S promoter, it was expressed. Although the correct splice sites were used, the intron was removed inefficiently. In addition, multiple novel polyadenylation sites, as well as the correct one, were used. The inefficiency of intron removal from monocot genes in transgenic dicots was also demonstrated using a maize Adh-1 gene in tobacco. This is in contrast to the accurate and efficient processing of a pea rbcS transcript in transgenic tobacco. The authors suggested that the differences seen between the processing of m o n o c o t and dicot RNAs in dicots could reflect a difference in the signals used between these groups of plants to effect these processes. Because a relatively small number of animal and monocot genes have been examined for accurate expression and processing of their transcripts in dicots, it is at present difficult to draw general conclusions about the ability of genes from various systems to be expressed in plants. Perhaps the differences seen reflect specific properties of individual genes, or even differences in the experimental systems (in vivo vs. in vitro, tumor vs. non-transformed calli vs. regenerated plants, tobacco vs. Petunia vs. sunflower, etc.). The accurate ex-
357 pression of any given gene in transgenic plants may, at this time, still be considered unpredictable.
Interspeeies and interkingdom genetic exchange The ability of many bacterial species to exchange D N A and of Agrobacterium species to transfer D N A to plant cells, as well as the close association many phytopathogens and symbionts have with their hosts, raises questions as to how much genetic exchange may have taken place between closely related or widely divergent organisms over the course of evolution. The prokaryotic nature of the Agrobacterium Ti- and Ri-plasmids, contrasted with the efficient expression of T-DNA genes in eukaryotic plant cells, has often raised the question of whether the T-DNA originated from plant genes that 'escaped' and were scavenged by the bacterium, or whether the genes are of bacterial origin and have evolved transcriptional regulatory sequences necessary for expression in an eukaryote. Although some evidence for both arguments exists, several recent reports suggest that not only have many T-DNA genes had a prokaryotic origin, but also that the exchange of genetic information between plants and members of the family Rhizobiaceae over the course of evolution may have been more extensive than previously thought. The most compelling arguments for the prokaryotic origin of T-DNA genes derives from two observations: 1) Opines have not been reproducibly detected in higher plants (although the reader should refer to a recent paper by Christou et al. (Plant Physiol 82:218-221, 1986) in which it was shown that opines could be synthesized by calli of some plant species under certain growth conditions) and 2) A. tumefaciens uses the same two-step auxin biosynthetic pathway as does another tumorigenic phytopathogen, Pseudomonas syringae pv. savastanoi. This pathway has not been reported in plants. The relatedness of the bacterial auxin biosynthetic genes, as well as those encoding enzymes for cytokinin biosynthesis, between the two species has been demonstrated in several reports in which the D N A sequences of the respective genes have been compared. Yamada et al. (Proc
Natl Acad Sci 82:6522-6526, 1985) have shown considerable homology between the P. savastanoi iaaM locus (encoding tryptophan monooxygenase) and the tmsl locus of the A. tumefaciens T-DNA. Similar strong homologies exist for the Pseudomohas iaaH gene and the Agrobacterium tms2 gene, which encode the second enzyme in the auxin biosynthetic pathway, indoleacetamide hydrolase. Sequence similarities also exist between the B savastanoi p t z and the A. tumefaciens tmr and tzs genes encoding enzymes involved in cytokinin biosynthesis (Powell and Morris, Nucleic Acids Res 14:2555-2565, 1986). An additional interesting report by Yamada et al. (Proc Natl Acad Sci 83:8263-8267, 1986) may explain how the T-DNA hormone biosynthetic genes were transferred from R savastanoi to A. tumefaciens. These authors have characterized several IS. elements that frequently insert into the Pseudomonas iaaM gene. An additional copy of one of these elements lies 2.5 kbp downstream from iaaH. Sequences very similar to this IS element are found in the T-center region of the octopine-type T-DNA. The authors suggest that a c o m p o u n d transposable element, composed of these IS elements flanking the iaaM and H genes, was introduced at some time in evolution into A. tumefaciens, where it integrated into a plasmid. After considerable rearrangement of the IS elements, the auxin biosynthetic genes may have remained relatively intact to become part of the TDNA. These genes subsequently may have evolved eukaryotic transcriptional regulatory signals that permitted their expression in plants. (There is no evidence for the transfer of D N A between P. savastanoi and plants. The iaaM and H genes have maintained prokaryotic regulatory signals for expression in the bacterium). Close association of bacteria of the family Rhizobiaceae and plants that may have led to genetic exchange has also been suggested by Carlson and Chelm in a recent article (Nature 322:568-570, 1986). Whereas most bacteria contain a certain form of glutamine synthetase (called GSI), bacteria of this family contain two forms of the enzyme. GSI is similar to that found in other prokaryotes, whereas the GSII from Bradyrhizobium japonicum that they studied closely resembled
358 that found in plant species. The authors conclude that, despite the presence of introns in the plant GSII genes and their absence from the B. japonic u m gene, the additional bacterial gene probably derived from some progenitor plant gene that was transferred to the bacterium. These data, together with several reports demonstrating the presence of T-DNA sequences from the A. rhizogenes Riplasmid in certain non-transformed Nicotiana species (White et al., Nature 301:348-350, 1983; Furner et al., Nature 319:422-427, 1986), suggest that the transfer and stabilization of genetic information between closely associated but unrelated species may not be an u n c o m m o n occurrence.
How to normalize the quantitation of gene expression in transgenie plants The elucidation of the transcriptional regulatory sequences of plant genes has become easier during the past few years by the use of transgenic plants. Cloned genes or gene fragments can be isolated from plant genomic D N A and the putative regulatory elements manipulated in vitro. Either the mutagenized gene or a fusion gene composed of the putative transcriptional regulatory elements joined to an easily assayable marker gene (e.g. CAT, NPTII, or nopaline synthase) can then be returned to plant cells using disarmed T-DNA vectors and plants regenerated. The effect of the mutation or manipulation on gene expression can thus be assessed at the whole plant level, and the D N A sequences mediating the responses to environmental (light, heat shock, anaerobiosis, etc.), developmental, or tissue-specific signals may thus be identified. With increasing experimentation, it has now become obvious that introduced genes are not usually expressed at the same (steady state) levels from plant to plant. These differences in levels of expression may range from 2 5 - 5 0 fold for the same gene (Nagy et al., EMBO J 4:3063-3068, 1985), and may not even correlate with gene copy number. (The only published exception that this author is aware of is the less than two-fold variation in the expression in Petunia of a gene encoding the c~' subunit of ~3-conglycinin (Chen et aL, Proc Natl
Acad Sci 83:8560-8564, 1986.) These differences in levels of expression have usually been ascribed to 'position effects' of gene integration, and investigators have in the past attempted to control for these effects by normalizing the activity of the manipulated gene to that of a linked gene. The assumption has been that, because of the proximity of the two genes (generally within a few kilobases), they should exist in the plant at the same copy number and be subject to the same position effects. Several difficulties with this line of reasoning have been pointed out. The 'normalization' gene often used is one for antibiotic resistance. Because the expression of this gene is required for growth of the plant tissue on toxic levels of antibiotics, it may be inappropriate to correlate the activity of an unselected gene with that of a selected one. Even a comparison of the level of expression of two closely linked, unselected genes has yielded surprising results. Karcher et al. (Mol Gen Genet 194:159-165, 1984), while analyzing the relative steady-state levels of five transcripts derived from the TR region of the T-DNA in a cloned crown gall tumor line, noted that the relative abundance of the various TNAs varied depending upon whether the tissue was grown as a liquid suspension or as a callus on solidified agar media. Clearly, position effects could not be evoked to explain these results. The differences appeared to derive from the physiological growth conditions of the tumor cells. More recently, An (Plant Physiol 81:86-91, 1986) investigated the relative expression of two closely linked genes (NPTII and CAT), both of which were under control of the same nopaline synthase promoter. An analysis of 40 transformed tobacco calli revealed a greater than 70-fold variation between the relative activities of the two genes. In this study, the nos-CAT fusion gene was unselected whereas the n o s / N P T I I fusion gene was selected. Despite this possible selection artifact, the enormous variation seen in the relative expression of those two genes was surprising. The author suggested that it may be necessary to examine a large number of individually transformed calli to obtain a statistically significant estimate of the activity of an introduced gene. Whereas this approach of assaying or pooling a large number of individuals may be a reasonable
359 approach for millions of electroporated plant cells or hundreds of tranformed calli, it may be technically unrealistic for regenerated plants. This columnist would like to stimulate a discussion of this subject and would appreciate suggestions, which will be published in future issues of Biotechnology News and Views, as to how to overcome these problems in assaying the transcriptional activity of
genes introduced into transgenic plants. Please address your suggestions to: Dr Stanton Gelvin Department of Biological Sciences Purdue University West Lafayette, I N 47907 USA
355
Biotechnology news and views Stanton B. Gelvin
Accurate expression of genes in transgenic plants: do fundamental differences exist between eukaryotes? Because of the availability and increasing ease of use of T-DNA-derived vector systems, as well as the establishment of electroporation as a reliable way of introducing D N A into plant cells, many laboratories are using transgenic plants to investigate basic processes of plant molecular biology. It was established early on that prokaryotic genes (often the source of antibiotic resistance markers) under the control of their own promoters do not function in plant cells. Functional chimaeric genes could be constructed, however, by fusing the promoters and upstream elements of constitutive plant genes (such as those from the T-DNA or cauliflower mosaic virus [CaMV]) or of regulated plant genes (such as those from rbcS or cab) to the prokaryotic coding sequences. Perhaps it was not surprising that prokaryotic transcriptional regulatory signals could not function in plants, considering the m a j o r differences in the mechanisms of gene expression between prokaryotes and eukaryotes. The question remained, however, whether genes from other eukaryotic organisms, such as animals or yeast, could be accurately expressed in plants, or whether there was even a difference between monocots and dicots in the ways in which D N A is expressed. Whereas the answers to these questions have not yet been clearly defined, recent studies have suggested that all eukaryotes may not express genes in the same ways using the same regulatory signals. et al. (Cell Early studies by Barton 32:1033-1043, 1983) indicated that a yeast alcohol dehydrogenase gene could not be expressed in tobacco crown gall tumors. Langridge et al. (EMBO J 3:2467-2471, 1984) showed, however, that two maize genes encoding 19 and 21 kd zein
proteins could be accurately transcribed in yeast. The transcripts initiated at the same nucleotides both in yeast and in maize. It is well known that yeast and animal transcriptional regulatory and R N A processing signals, while similar, may often differ significantly (see e.g. Nevins, Ann Rev Biochem 52:441-466, 1983). It is therefore not unreasonable to expect that genes from higher and lower eukaryotes may at times be expressed either correctly or aberrantly in these heterologous systems. Most animal and plant genes contain 'TATA' boxes and many have 'CAAT' boxes making up their promoter elements. The plant intron junction 'GT/AG' consensus signals are followed as in animal systems, although the polypyrimidine stretches often found upstream of most 3' splice sites in animals are often lacking in plant genes. Because of the presumed similarity of higher animal and plant transcription regulatory and R N A processing signals, one may expect that the expression of animal and plant genes, while perhaps not developmentally or environmentally regulated in the same ways, would at least be accurate in the respective heterologous systems. Indeed, Brown et al. (EMBO J 5:2749-2758, 1986) have recently shown that in vitro SP6 transcripts of two plant genes encoding a wheat amylase and a pea legumin are accurately spliced in a H e L a cell nuclear extract. Lariat RNAs similar to those seen as a result of the splicing of animal and yeast introns were also detected using these plant RNAs. In addition, an analysis of 168 plant intron sequences indicated that putative lariat branch point consensus signals very similar to those found in animal and fungal introns could be located in plant introns. The authors cautioned, however, that a final analysis of plant RNA splicing would have to await the development of plant in vitro splicing systems. Despite the heat-regulated expression of a
356 Drosophila heat shock p r o m o t e r / N P T I I fusion gene in tobacco calli (Spena et al., EMBO J 4:2739-2743, 1985), the accurate transcription and processing of animal genes in plant cells has been difficult to achieve. Koncz et al. (EMBO J 3:1029-1037, 1984) showed that a chicken c~-actin gene was not expressed in tobacco crown gall tumors, and that the initiation of transcription of a chicken ovalbumin gene was not accurate in the same system. Barta et al. (Plant Mol Biol 6:347-357, 1986) more recently investigated the processing of a h u m a n growth hormone (hGH) transcript in tobacco and sunflower crown gall tumors. In order to express the animal gene in plant cells, the authors placed the entire h G H gene (including introns and polyadenylation signal sequences) between the nopaline synthase (nos) promoter and transcription termination sequences. The fusion gene correctly initiated transcription from the nos promoter, but the h G H transcript did not utilize the animal termination signal; rather, the R N A ended using the nos poly A signal. In addition, the h G H introns were not spliced from the primary transcript in plants, although they could be an in vitro H e L a cell nuclear extract. These results suggest that, at least for some animal genes, the termination and splicing signals cannot be accurately recognized, if at all, in vivo in plant systems. As opposed to the similarities between putative animal and plant splicing signals discussed above, it has often been noted that the termination of transcription in plants may differ from that in animals: plant transcripts tend to have multiple termination sites, and the AAUAAA consensus signal found in most animal genes is more variable in plant genes (plants tend to use, in addition, AAUAAU and other variations of the animal consensus sequence). Given the variation in the ability to express and process animal genes correctly in plant cells, one may wonder whether fundamental differences in these processes may exist between monocots and dicots. Numerous studies (see e.g. Murai et al., Science 222:476-482, 1983 and Sengupta-Gopalan et al., Proc Natl Acad Sci 82:3320-3324, 1985 for phaseolin; Chen et aL, Proc Natl Acad Sci 83:8560-8564, 1986 for /3-conglycinin) have indi-
cated that certain specific dicot genes can be correctly expressed and developmentally regulated in other dicots. L a m p p a et al. (Nature 316:750-752, 1985) demonstrated that a monocot (wheat) cab transcript could be correctly initiated in transgenic tobacco plants. The gene correctly showed light/dark and tissue-specific regulation in its new host. Matzke et al. (EMBO J 3:1525-1531, 1984) demonstrated that a 19 kd maize zein gene could be accurately initiated in sunflower crown gall tumors. Goldsbrough et al. (Mol Gen Genet 202:374-381, 1986) showed that a similar 19 kd zein gene, as well as a 15 kd zein gene, could be both correctly initiated and terminated in sunflower tumors. More recently, however, Keith and Chua (EMBO J 5:2419-2425, 1986) presented data suggesting that the efficiency o f splicing of monocot and dicot transcripts, as well as the accuracy of polyadenylation of these transcripts, may differ in transgenic dicots. When a wheat rbcS gene was transferred to tobacco, it was not expressed under the control of its own promoter. When linked to a CaMV 35S promoter, it was expressed. Although the correct splice sites were used, the intron was removed inefficiently. In addition, multiple novel polyadenylation sites, as well as the correct one, were used. The inefficiency of intron removal from monocot genes in transgenic dicots was also demonstrated using a maize Adh-1 gene in tobacco. This is in contrast to the accurate and efficient processing of a pea rbcS transcript in transgenic tobacco. The authors suggested that the differences seen between the processing of m o n o c o t and dicot RNAs in dicots could reflect a difference in the signals used between these groups of plants to effect these processes. Because a relatively small number of animal and monocot genes have been examined for accurate expression and processing of their transcripts in dicots, it is at present difficult to draw general conclusions about the ability of genes from various systems to be expressed in plants. Perhaps the differences seen reflect specific properties of individual genes, or even differences in the experimental systems (in vivo vs. in vitro, tumor vs. non-transformed calli vs. regenerated plants, tobacco vs. Petunia vs. sunflower, etc.). The accurate ex-
357 pression of any given gene in transgenic plants may, at this time, still be considered unpredictable.
Interspeeies and interkingdom genetic exchange The ability of many bacterial species to exchange D N A and of Agrobacterium species to transfer D N A to plant cells, as well as the close association many phytopathogens and symbionts have with their hosts, raises questions as to how much genetic exchange may have taken place between closely related or widely divergent organisms over the course of evolution. The prokaryotic nature of the Agrobacterium Ti- and Ri-plasmids, contrasted with the efficient expression of T-DNA genes in eukaryotic plant cells, has often raised the question of whether the T-DNA originated from plant genes that 'escaped' and were scavenged by the bacterium, or whether the genes are of bacterial origin and have evolved transcriptional regulatory sequences necessary for expression in an eukaryote. Although some evidence for both arguments exists, several recent reports suggest that not only have many T-DNA genes had a prokaryotic origin, but also that the exchange of genetic information between plants and members of the family Rhizobiaceae over the course of evolution may have been more extensive than previously thought. The most compelling arguments for the prokaryotic origin of T-DNA genes derives from two observations: 1) Opines have not been reproducibly detected in higher plants (although the reader should refer to a recent paper by Christou et al. (Plant Physiol 82:218-221, 1986) in which it was shown that opines could be synthesized by calli of some plant species under certain growth conditions) and 2) A. tumefaciens uses the same two-step auxin biosynthetic pathway as does another tumorigenic phytopathogen, Pseudomonas syringae pv. savastanoi. This pathway has not been reported in plants. The relatedness of the bacterial auxin biosynthetic genes, as well as those encoding enzymes for cytokinin biosynthesis, between the two species has been demonstrated in several reports in which the D N A sequences of the respective genes have been compared. Yamada et al. (Proc
Natl Acad Sci 82:6522-6526, 1985) have shown considerable homology between the P. savastanoi iaaM locus (encoding tryptophan monooxygenase) and the tmsl locus of the A. tumefaciens T-DNA. Similar strong homologies exist for the Pseudomohas iaaH gene and the Agrobacterium tms2 gene, which encode the second enzyme in the auxin biosynthetic pathway, indoleacetamide hydrolase. Sequence similarities also exist between the B savastanoi p t z and the A. tumefaciens tmr and tzs genes encoding enzymes involved in cytokinin biosynthesis (Powell and Morris, Nucleic Acids Res 14:2555-2565, 1986). An additional interesting report by Yamada et al. (Proc Natl Acad Sci 83:8263-8267, 1986) may explain how the T-DNA hormone biosynthetic genes were transferred from R savastanoi to A. tumefaciens. These authors have characterized several IS. elements that frequently insert into the Pseudomonas iaaM gene. An additional copy of one of these elements lies 2.5 kbp downstream from iaaH. Sequences very similar to this IS element are found in the T-center region of the octopine-type T-DNA. The authors suggest that a c o m p o u n d transposable element, composed of these IS elements flanking the iaaM and H genes, was introduced at some time in evolution into A. tumefaciens, where it integrated into a plasmid. After considerable rearrangement of the IS elements, the auxin biosynthetic genes may have remained relatively intact to become part of the TDNA. These genes subsequently may have evolved eukaryotic transcriptional regulatory signals that permitted their expression in plants. (There is no evidence for the transfer of D N A between P. savastanoi and plants. The iaaM and H genes have maintained prokaryotic regulatory signals for expression in the bacterium). Close association of bacteria of the family Rhizobiaceae and plants that may have led to genetic exchange has also been suggested by Carlson and Chelm in a recent article (Nature 322:568-570, 1986). Whereas most bacteria contain a certain form of glutamine synthetase (called GSI), bacteria of this family contain two forms of the enzyme. GSI is similar to that found in other prokaryotes, whereas the GSII from Bradyrhizobium japonicum that they studied closely resembled
358 that found in plant species. The authors conclude that, despite the presence of introns in the plant GSII genes and their absence from the B. japonic u m gene, the additional bacterial gene probably derived from some progenitor plant gene that was transferred to the bacterium. These data, together with several reports demonstrating the presence of T-DNA sequences from the A. rhizogenes Riplasmid in certain non-transformed Nicotiana species (White et al., Nature 301:348-350, 1983; Furner et al., Nature 319:422-427, 1986), suggest that the transfer and stabilization of genetic information between closely associated but unrelated species may not be an u n c o m m o n occurrence.
How to normalize the quantitation of gene expression in transgenie plants The elucidation of the transcriptional regulatory sequences of plant genes has become easier during the past few years by the use of transgenic plants. Cloned genes or gene fragments can be isolated from plant genomic D N A and the putative regulatory elements manipulated in vitro. Either the mutagenized gene or a fusion gene composed of the putative transcriptional regulatory elements joined to an easily assayable marker gene (e.g. CAT, NPTII, or nopaline synthase) can then be returned to plant cells using disarmed T-DNA vectors and plants regenerated. The effect of the mutation or manipulation on gene expression can thus be assessed at the whole plant level, and the D N A sequences mediating the responses to environmental (light, heat shock, anaerobiosis, etc.), developmental, or tissue-specific signals may thus be identified. With increasing experimentation, it has now become obvious that introduced genes are not usually expressed at the same (steady state) levels from plant to plant. These differences in levels of expression may range from 2 5 - 5 0 fold for the same gene (Nagy et al., EMBO J 4:3063-3068, 1985), and may not even correlate with gene copy number. (The only published exception that this author is aware of is the less than two-fold variation in the expression in Petunia of a gene encoding the c~' subunit of ~3-conglycinin (Chen et aL, Proc Natl
Acad Sci 83:8560-8564, 1986.) These differences in levels of expression have usually been ascribed to 'position effects' of gene integration, and investigators have in the past attempted to control for these effects by normalizing the activity of the manipulated gene to that of a linked gene. The assumption has been that, because of the proximity of the two genes (generally within a few kilobases), they should exist in the plant at the same copy number and be subject to the same position effects. Several difficulties with this line of reasoning have been pointed out. The 'normalization' gene often used is one for antibiotic resistance. Because the expression of this gene is required for growth of the plant tissue on toxic levels of antibiotics, it may be inappropriate to correlate the activity of an unselected gene with that of a selected one. Even a comparison of the level of expression of two closely linked, unselected genes has yielded surprising results. Karcher et al. (Mol Gen Genet 194:159-165, 1984), while analyzing the relative steady-state levels of five transcripts derived from the TR region of the T-DNA in a cloned crown gall tumor line, noted that the relative abundance of the various TNAs varied depending upon whether the tissue was grown as a liquid suspension or as a callus on solidified agar media. Clearly, position effects could not be evoked to explain these results. The differences appeared to derive from the physiological growth conditions of the tumor cells. More recently, An (Plant Physiol 81:86-91, 1986) investigated the relative expression of two closely linked genes (NPTII and CAT), both of which were under control of the same nopaline synthase promoter. An analysis of 40 transformed tobacco calli revealed a greater than 70-fold variation between the relative activities of the two genes. In this study, the nos-CAT fusion gene was unselected whereas the n o s / N P T I I fusion gene was selected. Despite this possible selection artifact, the enormous variation seen in the relative expression of those two genes was surprising. The author suggested that it may be necessary to examine a large number of individually transformed calli to obtain a statistically significant estimate of the activity of an introduced gene. Whereas this approach of assaying or pooling a large number of individuals may be a reasonable
359 approach for millions of electroporated plant cells or hundreds of tranformed calli, it may be technically unrealistic for regenerated plants. This columnist would like to stimulate a discussion of this subject and would appreciate suggestions, which will be published in future issues of Biotechnology News and Views, as to how to overcome these problems in assaying the transcriptional activity of
genes introduced into transgenic plants. Please address your suggestions to: Dr Stanton Gelvin Department of Biological Sciences Purdue University West Lafayette, I N 47907 USA
