Green fluorescent protein (GFP) has emerged as a powerful new tool in a variety of organisms. An engineered sGFP(S65T) sequence containing optimized codons of highly expressed eukaryotic proteins has provided up to 100-fold brighter fluorescence signals than the original jellyfish GFP sequence in plant and mammalian cells. It would be useful to establish a non-invasive, quantitative detection system which is optimized for S65T-type GFP, one of the brightest chromophore mutants among the various GFPs. We demonstrate here that highly fluorescent transgenic Arabidopsis can be generated, and the fluorescence intensity of whole plants can be measured under non-disruptive, sterile conditions using a quantitative fluorescent imaging system with blue laser excitation. Homozygous plants can be distinguished from heterozygous plants and fully fertile progenies can be obtained from the analyzed plants. In the case of cultured tobacco cells, GFP-positive cells can be quantitatively distinguished from non-transformed cells under non-selective conditions. This system will be useful in applications such as mutant screening, analysis of whole-body phenomena, including gene silencing and quantitative assessments of colonies from microorganisms to cultured eukaryotic cells. To facilitate the elucidation of protein targeting and organelle biogenesis in planta, we also generated transgenic Arabidopsis that stably express the plastid- or mitochondria-targeted sGFP(S65T). Etioplasts in dark-grown cotyledons and mitochondria in dry seed embryos could be visualized for the first time in transgenic Arabidopsis plants under normal growing conditions.
Although plastids and mitochondria are organelles that have their own chromosomes, many nuclear-encoded proteins are synthesized in the cytosol as precursor proteins with transit peptides that target them to and across the plastid or mitochondrial double-membrane envelope. These organelles also perform essential functions related to the dramatic changes in morphology, size and subcellular location that occur during the plant life cycle (Douce 1985; Possingham 1980). To analyze the biogenesis of plastids and mitochondria in higher plants, microscopic observations of living cells at various stages are necessary. Specific fluorescent dyes have made it possible to visualize mitochondria in vivo; however, these dyes have some problems, including problems with penetration, toxicity and photobleaching. Except for chloroplasts (the plastid state that is specifically differentiated for photosynthesis and has red autofluorescence), most plastids are very difficult to observe under living conditions because of the lack of specific probes and the small size of the organelles.
GFP was first cloned and sequenced from the jellyfish Aequorea victoria (Prasher et al. 1992). Wild-type GFP emits green fluorescence when excited with blue or UV light without any additional substrates or co-factors. The S65T mutation (replacement of the serine in position 65 with a threonine) in the chromophore has been shown to result in enhanced brightness, faster chromophore formation, slower photobleaching, and a single excitation peak ideal for fluorescein isothiocyanate (FITC) filter sets (Heim et al. 1995).
An engineered sGFP(S65T) sequence with codons optimal for high expression of eukaryotic proteins has provided up to 100-fold brighter fluorescent signals than the original jellyfish GFP sequence in plant and mammalian cells (Chiu et al. 1996; Haas et al. 1996). For analysis of organelle biogenesis in higher plants, Arabidopsis thaliana is one of the best model systems because of its many characteristics suited for molecular and genetic approaches (Meinke et al. 1998). The fusions of sGFP(S65T) with plastid and mitochondrial targeting sequences have directed the localization of sGFP(S65T) into plastids (Chiu et al. 1996; Isono et al. 1997) and mitochondria, respectively, in transient expression assays. Although GFP has emerged as a powerful new tool (Leffel et al. 1997; Misteli & Spector 1997), it has been suggested that high levels of GFP expression could be toxic to plant growth and development (Rouwendal et al. 1997), especially in Arabidopsis (Haseloff et al. 1997).
In this study, we first evaluated the toxicity of sGFP(S65T) and then tried to establish a whole-plant, non-lesional, quantitative detection system ideal for S65T-type GFP. Finally, we generated transgenic Arabidopsis lines that stably expressed the plastid- or mitochondria-targeted sGFP(S65T).
Generation of transgenic Arabidopsis expressing sGFP(S65T)
We previously demonstrated that our engineered sGFP(S65T) is useful as a vital marker in plant transient expression systems (Chiu et al. 1996; Isono et al. 1997). In some cases, it has been suggested that high levels of GFP expression could be toxic to growth and development in plants (Haseloff et al. 1997; Rouwendal et al. 1997). To evaluate the toxicity of sGFP(S65T), we introduced the constitutive 35S promoter-sGFP(S65T) fusion gene (Fig. 1) into the Arabidopsis genome. We could achieve high levels of GFP expression and obtain morphologically normal and fertile plants at the usual frequency. This result indicates that sGFP(S65T) is not toxic in Arabidopsis.
Non-disruptive fluorescence measurement in whole plants with blue light excitation
Since the fluorescence level of S65T-type GFP is among the highest for GFP variants (Heim et al. 1995) and its optimal excitation wavelength (blue light) is thought to be harmless to organisms compared to UV irradiation, sGFP(S65T) is an excellent candidate for use in a non-lesional, quantitative detection system for transgenic organisms. Figure 2 shows typical fluorescent images of sGFP(S65T)-transformed Arabidopsis scanned by a quantitative fluorescent imaging system. The fluorescent images of bialaphos resistant progenies (adult plants of T2 generation) from three independent transgenic lines were visualized under 488 nm Ar blue laser excitation (Fig. 2a). Only the dark red autofluorescence from the chloroplasts was detected from the wild-type leaf (WT). The four T2 progeny from nA5, one of the highly fluorescent lines, displayed a yellowish-green color, because the green fluorescence from GFP combined with the red fluorescence from the chloroplasts. Although the GFP expression levels in leaves of nA8 and nA9 progenies were low, those in roots of nA9 progeny were significantly higher. After the analysis, the T2 plants were transferred to soil and set T3 seeds normally. Some of the T3 seeds were spread on the bialaphos-free medium under sterile conditions. The fluorescent images of T3 seedlings from two parental lines, nA5–2 and nA5–4, are shown in Fig. 2(b). While the T3 seedlings from the nA5–2 appeared to be all GFP positive, some of the T3 seedlings from the nA5–4 seemed to be wild-type (autofluorescent only).
To measure the fluorescence intensity of whole plants under non-disruptive, sterile conditions, quantitative analysis of these images was performed using the ImageQuant program. Relative fluorescence intensities of 11 independent transgenic lines (nA1 to nA11) and wild-type (WT) plants are shown in Fig. 3(a). Although the green fluorescence intensity of each line (light bars) differed and was absent in wild-type Arabidopsis, the red autofluorescence intensity (dark bars) was almost the same in all lines, including wild-type.
In order to compare the GFP expression levels of leaves and roots in the same living plant, the green fluorescence intensity of these organs was measured separately using the same images as in Fig. 3(a) and plotted in Fig. 3(b). The value of fluorescence in leaves was then divided by that in roots (Fig. 3c). Seven out of 11 lines of transgenic plants gave values around 1.0, and the values in two lines were below 0.1 (nA9 and nA11). This result indicates that the expression levels in leaves and roots of most transgenic plants were the same; however, in some transgenic lines, such as nA9 and nA11, the expression level in roots was significantly higher than that in leaves. For example, although the GFP expression levels in leaves of progeny from nA8 and nA9 seemed to be the same, the expression levels in roots of nA9 progeny were higher than in those of nA8 progeny (see Fig. 2a).
Because the fluorescence intensity of the whole plant was influenced by the angles of the leaves, we normalized the green fluorescence intensity relative to the red autofluorescence intensity. In order to test the reliability of this value, we randomly picked four bialaphos-resistant T2 progeny of the heterozygous T1 plant nA5 (Fig. 2a) and calculated the relative fluorescence intensities (Fig. 3c). The values obtained from nA5–2 and nA5–3 were twice as much as those from nA5–1 and nA5–4. This result suggested that nA5–1 and nA5–4 were heterozygotes while nA5–2 and nA5–3 were homozygotes. To confirm this prediction, the fluorescent image of T3 progeny from the two parental lines nA5–2 and nA5–4 shown in Fig. 2(b) was quantitatively analyzed and fluorescence values for individual T3 progeny are shown in Fig. 3(e) (for nA5–4) and are summarized in Fig. 3(f). The normalized fluorescence values of the cotyledons of the nA5–2 T3 progeny clustered between 3 and 5, as expected for homozygotes, and those of the nA5–4 T3 progeny were comprised of three distinct groups in the ratio of 1:2:1, as expected for heterozygotes (Fig. 3f).
Application of the quantitative fluorescent imaging system
To test the possibility of direct and quantitative selection of transformed cells, cultured tobacco cells were transformed with the 35S-sGFP(S65T) fusion gene and grown on solid medium with or without bialaphos for selection (Fig. 4a). While only the bialaphos-resistant GFP-positive cells grew on the selection medium (+ bialaphos), GFP-positive cells were clearly distinguished from non-transformed cells even when grown on the non-selective plate (– bialaphos). The green fluorescence intensity of GFP-positive and GFP-negative tobacco calli are shown in Fig. 4(b). The fluorescence intensity of these cells could be measured directly from the surface of the plates. The transformation status of each callus was checked by PCR (data not shown).
Although the transgenic line mtA1 had been isolated as a bialaphos-resistant T1 plant, most of the T2 progeny of this line were sensitive to bialaphos (data not shown). We checked the green fluorescence of these and found that the expression of GFP was uneven. To compare the expression patterns of each progeny at a whole plant level, adult plants grown on the bialaphos-free medium were examined using the FluorImager system. As shown in Fig. 5(a), mtA1 progeny displayed a variegated phenotype that was reproducible but varied in pattern from plant to plant. Other progenies, such as that from mtA2, showed uniform expression (lower three plants in Fig. 5a). Since the minimum size for resolution in the FluorImager system is 100 μm, this system is sufficient for detection down to the organ level (see Fig. 7 and the Discussion). For examining the expression pattern of GFP at the cellular level, we used a fluorescent binocular. The uneven expression of GFP was observed in both leaves (Fig. 5b) and roots (data not shown).
Figure 5(c–f) demonstrate the simplicity of screening approximately 250 seedlings on the surface of a plate with a diameter of nine centimeters under non-disruptive, sterile conditions. The green fluorescence images (c and f), red autofluorescence image (d), and (c) and (d) merged image (e) of seven different sGFP(S65T) transformed lines (nA1 to nA7) and wild-type plants (WT) are shown. Although the green fluorescence intensity of each line differs and is absent in wild-type Arabidopsis (Fig. 5c,f), the red autofluorescence intensity is almost the same in all seedlings, including wild-type (Fig. 5d). Because, in some transgenic seedlings, the level of green fluorescence is significantly higher than that of the red fluorescence from the chlorophyll, the green masks the red (Fig. 5e). These results are consistent with those from the adult plants shown in Figs 2(a) and 3(a). Figure 5(f) shows the green fluorescence image of the reverse of the view shown in Fig. 5(c). Because roots lack autofluorescence from chloroplasts, very weak green fluorescence signals can be detected. In the case of nA6, GFP signals from roots can be more easily detected (Fig. 5f) than those from cotyledons (Fig. 5c).
Visualization of organelles in vivo
Since sGFP(S65T) does not show any toxic effects in Arabidopsis, it should be useful for protein localization analysis and visualization of organelles in living plants. To test this idea, we generated transgenic Arabidopsis lines that stably expressed the plastid- or mitochondria-targeted sGFP(S65T) (Fig. 1). In both cases, the GFP fluorescence signals could be detected in roots, leaves and reproductive organs. Figure 6 represents the typical fluorescence images of these data together with transmission images detected using a confocal laser scanning microscope. GFP signals are shown in the far left column (Green) and red fluorescence signals from chlorophyll (a) or the mitochondrial-specific dye (c) are shown in the column indicated as Red. The yellow color shown in the merged column (Merged) indicates the co-localization of green and red fluorescence signals.
In Arabidopsis transformed with the plastid-targeted construct, we examined plastids in both dark-grown (etioplast-containing) and light-grown (chloroplast-containing) cells. Etioplasts are very difficult to detect in living cells. Green fluorescence could be observed in both chloroplasts and etioplasts, whereas red autofluorescence was detected only in chloroplasts as expected (rows a and b of Fig. 6). In addition to morphological observations, the merged image in row (a) confirmed the correct targeting of pt-sGFP(S65T) into chloroplasts. From these data together with the transmission images, both chloroplasts and etioplasts are localized along the inside of cell boundaries. The etioplasts and the chloroplasts differed in number, shape and size. The average numbers of etioplasts and chloroplasts per mesophyll cell were about 20 and 100, respectively. While the average size of chloroplasts was larger than 10 microns, that of etioplasts was smaller than 5 microns.
In transgenic Arabidopsis transformed with mitochondrial-targeted sGFP(S65T), we examined embryos in dry seed. Because of the dehydration necessary for desiccation tolerance, mitochondria in the cells of dry seed are very difficult to detect. For example, preparation of specimens of dry seeds for transmission electron microscopy imposes limitations on the technique. Mitochondria-specific dyes require seed hydration and are very sensitive to photobleaching. In contrast, we detected around 1 micrometer particles in the living cells of dry embryos transformed with mitochondria-targeted sGFP(S65T) (data not shown). To confirm that targeting was correct, we used a dye specific for mitochondria. Row (c ) in Fig. 6 shows the images of this experiment and the result clearly indicates that the mt-sGFP(S65T) is localized to mitochondria in the hypocotyl of a seed embryo. Although green fluorescent particles did not move in the dehydrated cells, mitochondria began to move after 2 days of incubation on solid medium (data not shown). In living cells, while mitochondria were sometimes observed as streaming particles, plastid movements were hardly detected under our observation conditions.
It has been reported that high levels of GFP expression could be toxic to plant growth and development (Haseloff et al. 1997; Rouwendal et al. 1997). In the case of Arabidopsis, one solution to this problem was found to be targeting the GFP into the endoplasmic reticulum (ER) (Haseloff et al. 1997). Transgenic Arabidopsis plants could be recognized under illumination with a hand-held UV lamp. Chromophore mutations such as S65T demonstrated superior characteristics, for example, sixfold brighter fluorescence, fourfold faster chromophore formation, and greater resistance to photobleaching than wild-type GFP (Heim et al. 1995). In addition, while wild-type GFP has its maximum excitation peak at 396 nm and needs UV irradiation to detect, S65T-type GFP responds to blue light excitation, which can be generated using common FITC filter sets and is thought to be harmless to living organisms. This point is extremely important for procedures such as mutant screening and sequential observation of cellular activities that require long exposures to strong light. In this study, we both achieved high levels of GFP expression and obtained morphologically normal and fertile plants at the usual frequency. These results indicate that sGFP(S65T) is not toxic in Arabidopsis. This is consistent with the results in transgenic Arabidopsis transformed with similar synthetic GFP genes (Pang et al. 1996) and with the results in transgenic tobacco (Köhler et al. 1997b; Pang et al. 1996). In our case, some of the transgenic plants were easily distinguished from non-transformed wild-type plants, even with the naked eye, under blue light illumination with an appropriate filter.
Flow cytometric analysis or fluorescence-activated cell sorting is useful to quantitate the expression level of GFP in living cells (Sheen et al. 1995); however, it is only applicable to dissociated cells. Digital imaging spectrophotometry has been used to measure microbial colonies directly from the surface of Petri dishes (Youvan 1994). It would be very useful to set up a comparable non-destructive, quantitative detection system which is suited for S65T-type GFP for multicellular organisms. To quantitate the sGFP(S65T) expression of whole plants, we used a quantitative fluorescent imaging system. Figure 7 summarizes the properties of each type of equipment used. The Microscope category includes a fluorescence microscope with an Hg lamp and a confocal laser scanning microscope with a Kr/Ar laser as a light source. If the objects are smaller than cells, only the microscope suffices for detection (see Fig. 6). One example of an image detected by the fluorescent binocular is shown in Fig. 5(b). The binocular is easy to use and can effectively detect fluorescence ranging from cellular levels to several seedlings. Since the fluorescent binocular uses an Hg lamp as a light source, some disadvantages exist for quantification by image processing with a fluorescent binocular. An Hg lamp is not strong enough to detect the whole plate in a single view and it is hard to keep light intensity at a steady level. The fluorescent imaging system used in this study overcomes these problems because the Ar laser is chosen as a light source. The maximum area of 21.5 × 36.5 cm can be imaged automatically. Individual fluorescence values can be measured even at a density of about 250 seedlings per 9 cm diameter plate (see Fig. 4e). Together these properties allow quantitation of about 3000 seedlings in one scan. Although fluorescence intensity was influenced by the angle of the objects, we were able to distinguish homozygous plants from heterozygous plants by standardizing the green signal intensity with the red autofluorescence intensity and fully fertile progeny could be obtained from analyzed plants (Fig. 3d–f). If growth conditions are exactly the same as in Figs 2 and 3, using the red autofluorescence of chlorophyll for normalization is very convenient because there is no need to treat them with dyes. In cases where growth conditions or mutation alters chlorophyll levels, arranging the samples on the surface of the plate at the same angle eliminates the need to normalize GFP intensity (data not shown). GFP-positive cells were also quantitatively distinguished from non-transformed cultured tobacco cells (Fig. 4). From these results, we conclude that this method should be especially useful for analysis at the population level such as mutant screening and quantitative selection of transformants, including microorganisms and eukaryotic cultured cells even without any drug-resistance markers. The result shown in Fig. 5(a) demonstrates that this system is also useful for analyzing variegated phenotypes such as gene silencing and chimeric plants or tissues.
In most cases the fluorescence intensities of leaves and roots in a given transgenic plant were almost the same but, in some cases, the expression level of GFP in the roots was higher than that in the leaves (Fig. 3b,c). There exist at least two explanations for this phenomenon: one might be leaf-specific silencing and the other might be positional effects of the insertion locus. Whether one of these is true or not, this system should be useful for quantitative analysis concerned with organ specific expression.
Mitochondria and plastids contain their own chromosomes, and the physiological and morphological states of these organelles are dramatically changed during the plant life cycle. Although chloroplasts, for example, have red autofluorescence and some fluorescent probes can stain living mitochondria, most states of the plastid are very difficult to distinguish and specific probes are very sensitive to photobleaching in living cells. It has previously been demonstrated that organelle-targeted GFP is very useful for visualizing individual organelles (Boevink et al. 1998; Di Sansebastiano et al. 1998; Gerdes & Kaether 1996; Grebenok et al. 1997; Gu & Verma 1997; Haseloff et al. 1997; Köhler 1998; Saito et al. 1999). Plastids and mitochondria have been visualized using transgenic tobacco and petunia (Köhler et al. 1997a; Köhler et al. 1997b). Arabidopsis is one of the most ideal model plants for molecular and genetic analyses (Meinke et al. 1998). We generated transgenic Arabidopsis lines that stably expressed plastid- or mitochondria-targeted sGFP(S65T). Some plastids, such as leucoplasts and proplastids, were visualized using ER-targeted mGFP4 (Haseloff et al. 1997). In our case, the GFP fluorescence signals could be detected in all types of plastids, each of which has its own important role. For the mitochondria-targeted sGFP(S65T), the GFP fluorescence signals could be detected and correctly targeted in roots, leaves, reproductive organs, and even in dry embryos. Using these plants, plastids and mitochondria could be visualized in living cells. Thus, we have generated valuable tools that will facilitate the elucidation of protein targeting and organelle biogenesis in planta.
Growth conditions for Arabidopsis thaliana ecotype Columbia have been described previously (Niwa et al. 1997). Dark- and light-grown seedlings were surface sterilized and incubated for 7 days and were covered or not covered with aluminium foil. Embryos in dry seeds were obtained by cutting the seed coat with a razor blade and peeling it off. For dual imaging of GFP and the mitochondrial-specific dye, embryos were incubated with 500 nm MitoTracker Red CMXRos (Molecular Probes) for 3–4 h at room temperature. Growth conditions for the tobacco BY-2 cell line have been described previously by Nagata et al. (1981).
Introduction of sGFP(S65T) expression vectors into Arabidopsis and tobacco
The mitochondria-targeted mt-sGFP(S65T) fusion gene was constructed so that the HaeIII fragment carrying the N-terminal 78 amino acids of the gamma-subunit of mitochondrial F1ATPase of Arabidopsis was inserted into the filled-in SalI and NcoI sites of the 35Somega-sGFP(S65T) plasmid (Chiu et al. 1996). 35Somega-sGFP(S65T), plastid-targeted pt-sGFP(S65T) (Chiu et al. 1996), and mitochondria-targeted mt-sGFP(S65T) fusion genes were inserted into the T-DNA region of the binary vector pSMAB701. The chimeric genes were introduced into a T0 generation of Arabidopsis by infiltration (Bechtold et al. 1993). Bialaphos-resistant T1 transgenic Arabidopsis plants were selected on solid medium supplemented with 40 mg l–1 bialaphos (De Block et al. 1987). Introduction of the chimeric genes was confirmed by Southern blot analysis (data not shown). Cultured tobacco cells transformed with 35Somega-sGFP(S65T) were grown on medium with or without 4 mg l–1 bialaphos. Transformation was checked by PCR using 35SminiL (5′-GCAAGACCCTTCCTCTATATAAGC-3′) and sGFP5′Rv (5′-CCGTCCAGCTCGACCAGGATG-3′) primers (data not shown).
Analysis of sGFP(S65T) expression using a FluorImager imaging system
Whole seedlings and cultured tobacco cells grown under sterile conditions in either a 15 or 20 mm deep plate were scanned from the bottom of the plate using a quantitative fluorescent imaging system (FluorImager SI, Molecular Dynamics) with excitation at 488 nm (Ar laser) and detection with a 515–545 nm band-pass filter for GFP and a 610 nm filter for red autofluorescence from chloroplasts. Adult plants grown in a 20 mm deep plate were transferred to a 15 mm deep plate and then scanned in the same way. The lower limit of resolution of the FluorImager is 100 microns, the maximum area for one scan is 21.5 × 36.5 cm, and the height limit is approximately 20 mm. Each fluorescence intensity was measured with the ImageQuant program. All images were prepared with Adobe Photoshop 3.0 J.
Monitoring of sGFP(S65T) expression by microscopy
Green fluorescence, red mitochondrial-specific MitoTracker dye, red autofluorescence from chlorophyll, and transmission images were monitored using a confocal laser scanning microscope (MRC1024, Bio-Rad) with Kr/Ar laser excitation. The fluorescence signals were detected with the standard triple labelling filters for FITC/Texas Red/Cy5 (a 506–538 nm band-pass filter for GFP, a 578–618 nm band-pass filter for MitoTracker, and a 664–696 nm band-pass filter for chlorophyll fluorescence). The image of variegated expression of sGFP(S65T) was taken with a Leica 35 mm camera using a fluorescent binocular microscope (MZAPO, Leica) with a GFP Plus filter set. All images were prepared with Adobe Photoshop 3.0 J.
We thank Dr Hiroaki Ichikawa for providing his unpublished binary vector pSMAB701; Dr Hiroyuki Anzai for the bialaphos; Dr Julie Nardone for critical reading of the manuscript; Drs Ryoko Kuruto and Kazuo Tsugane for their assistance; and Drs Jen Sheen and Tsukaho Hattori for helpful discussions. This work was supported in part by a Grant-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture of Japan, by a special grant from the President of our University to Y.N.