Gene 221 (1998) 35–43 Cloning vectors for the expression of green fluorescent protein fusion proteins in transgenic plants A.G. von Arnim a,*, X.-W. Deng b, M.G. Stacey a a Department of Botany, HBB 437, The University of Tennessee, Knoxville, TN 37996-1100, USA b Department of Molecular, Cellular and Developmental Biology, Yale University, 165 Prospect Street, PO Box 208104, New Haven, CT 06520-8104, USA Received 10 May 1998; received in revised form 8 August 1998; accepted 10 August 1998; Received by W. Martin Abstract A series of versatile cloning vectors has been constructed that facilitate the expression of protein fusions to the Aequorea victoria green fluorescent protein (GFP) in plant cells. Amino-terminal- and carboxy-terminal protein fusions have been created and visualized by epifluorescence microscopy, both in transgenic Arabidopsis thaliana and after transient expression in onion epidermal cells. Using tandem dimers and other protein fusions to GFP, we found that the previously described localization of wild-type GFP to the cell nucleus is most likely due to diffusion of GFP across the nuclear envelope rather than to a cryptic nuclear localization signal. A fluorescence-based, quantitative assay for nuclear localization signals is described. In addition, we have employed the previously characterized mutants GFP–S65T and GFP–Y66H in order to allow for the expression of red-shifted and blue fluorescent proteins, respectively, which are suitable for double-labeling studies. Expression of GFP-fusions was controlled by a cauliflower mosaic virus 35S promoter. Using the Arabidopsis COP1 protein as a model, we confirmed a close similarity in the subcellular localization of native COP1 and the GFP-tagged COP1 protein. We demonstrated that COP1 was localized to discrete subnuclear particles and further confirmed that fusion to GFP did not compromise the activity of the wild-type COP1 protein. © 1998 Elsevier Science B.V. All rights reserved. Keywords: GFP; Nuclear targeting; Subnuclear architecture; COP1 1. Introduction Fusions between the green fluorescent protein (GFP) from Aequorea victoria and specific passenger proteins have found a wide acceptance in protein targeting research (e.g. Wang and Hazelrigg, 1994; Olson et al., 1995). In plants, subcellular targeting has been investigated primarily by two techniques. On the one hand, recombinant b-glucuronidase (GUS )-fusion proteins have been expressed in transient assays (e.g. Restrepo et al., 1990; Varagona et al., 1992) or in transgenic plants (e.g. von Arnim and Deng, 1994). Typically, the GUS fusion proteins are visualized in situ by histochemi* Corresponding author. Tel: +1 423 974 6206; Fax: +1 423 974 0978; e-mail: email@example.com Abbreviations: CaMV, cauliflower mosaic virus; CCD, charge coupled device; FRET, fluorescence resonance energy tranfer; GFP, green fluorescent protein; gfp, gene encoding GFP; GUS, b-glucuronidase; NLS, nuclear localization signal; TEV, tobacco etch virus; UV, ultraviolet light; wt, wild type. 0378-1119/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved. PII: S0 3 7 8 -1 1 1 9 ( 9 8 ) 0 0 43 3 - 8 cal staining and light microscopy, and the subcellular resolution of this assay is sufficient for distinguishing nuclear from non-nuclear localization. On the other hand, indirect immunofluorescence detection provides a high spatial resolution, but relies on specific antibodies, and requires fixation combined with either tissue sectioning or digestion of the cell wall to facilitate access of the antibody to the antigen. Compared with the aforementioned techniques, GFP-based detection has the advantage that living cells can be examined, allowing real-time imaging, and avoiding fixation artifacts. Moreover, many plant tissues are easily transformable and exhibit little green autofluorescence under ultraviolet ( UV ) excitation. Therefore, wild-type GFP as well as a mutant derivative of GFP with enhanced fluorescence, both of which emit green light upon UV excitation, are detectable with high sensitivity (Haseloff et al., 1997). Additional spectral variants of GFP, most notably the red-shifted S65T mutant and the blue fluorescent Y66H mutant, have been shown to fluoresce in plant cells with characteristics similar to those described in non-plant 36 A.G. von Arnim et al. / Gene 221 (1998) 35–43 cells (Heim et al., 1994, 1995; Davis and Vierstra, 1998). Specifically in plant cells, targeting of GFP-fusions to chloroplasts ( Köhler et al., 1997a), mitochondria ( Köhler et al., 1997b), cytoskeletal structures (Heinlein et al., 1995) and the nucleus (Grebenok et al., 1997) has provided new insights into diverse cellular processes. However, no versatile cloning vectors for the expression of GFP-fusion proteins have become available. Moreover, the adoption of GFP as a tag for nuclear targeting assays has been impeded by the spontaneous uptake of unfused GFP by the nucleus (Grebenok et al., 1997). It has also remained open as to whether transgenic plants expressing GFP-fusion proteins are viable, and whether fusion of GFP to a passenger protein necessarily interferes with the functionality of the passenger protein in transgenic plants. Here, we have created a series of GFP-fusion cloning vectors and established GFP as a reporter for nuclear protein localization assays. A fusion gene between gfp and CONSTITUTIVELY PHOTOMORPHOGENIC 1 (COP1) complements a lethal cop1 loss-of-function allele in Arabidopsis, suggesting that fusion to GFP does not necessarily compromise the function of the fusion partner. Our data further indicate that GFP fusions can provide a reliable means of tagging diverse proteins in live plant cells. carried out with high-fidelity DNA polymerases. The DNA sequence of pAVA319 is available under GenBank Accession No. AF078810. 2.2. Transient expression assays Transient expression assays in onion epidermal cells using particle bombardment (Bio-Rad, Richmond, CA) were carried out as described previously (von Arnim and Deng, 1994). Typically, 2 mg of wild-type GFP or GFP-S65T plasmids or 5 mg of BFP plasmid were used per assay. Tissue was examined on a Zeiss Axiovert 135 inverted microscope with a 100-W mercury arc lamp and the following filter cubes: I, 365-nm excitation; 395-nm beam splitter; 420-nm long-pass emission (Zeiss cube 02); II, Chroma FITC-cube CZ917 with an additional block of infra-red emission (Chroma; 485±10-nm excitation; 505-nm beam splitter; 535±20-nm band-pass emission); III, 365-nm excitation; 395-nm beam splitter; 460±25-nm emission (Chroma); IV, 365-nm excitation; 395-nm beam splitter; 535±20-nm band-pass emission. Images were acquired through a 10× objective with a 12-bit MicroMax cooled CCD camera (Princeton Instruments, Trenton, NJ ) operated by IPLab software (Signal Analytics, Vienna, VA) on a Macintosh 7100 PowerPC. Images were processed for printing via Adobe Photoshop software and printed on a Kodak XLS8600 dye sublimation printer. 2. Materials and methods 2.1. Plasmid construction Standard molecular cloning procedures (Sambrook et al., 1989) were employed, and details of the clone construction are available upon request. All GFP-vectors are based on the expression cassette of pRTL2 (Restrepo et al., 1990). This vector contains a dual 35S promoter from cauliflower mosaic virus (CaMV ), the translational leader sequence from tobacco etch virus ( TEV ), and the 35S polyadenylation signal from CaMV. Except for pAVA68, which contains a wild-type gfp cDNA (Prasher et al., 1992), all gfp cDNAs are based on mGFP4 (Haseloff et al., 1997) and therefore express correctly spliced transcripts in Arabidopsis. Unless stated otherwise in Table 1, all vectors contain the TEV translational enhancer and support the cloning of aminoterminal fusions to GFP via a unique NcoI-site, and the cloning of carboxy-terminal fusions via BglII. An XbaIsite is located beyond the gfp stop codon to facilitate directional cloning. Vectors containing the NcoI cloning site also carry a S1G substitution and a deletion of E6 of GFP, mutations that do not interfere with GFPfluorescence. Fusion protein genes were constructed based on published full-length cDNA sequences using oligonucleotides to introduce suitable restriction enzyme recognition sites. PCR amplification of cDNAs was 2.3. Quantification of nuclear versus cytoplasmic fluorescence GFP-images were acquired with a cooled CCD camera through the FITC-filter cube (Cube II ), adjusting the exposure time to prevent saturation of the CCD. Using IPLab software, the fluorescence level was quantified separately for the nucleus and for the whole cell by circumscribing the respective area as a Region of Interest ( ROI ). Background correction was applied by estimating both nuclear and cytoplasmic fluorescence levels in a neighboring untransformed cell. The fraction of nuclear versus whole-cell fluorescence was expressed as a percentage. Typically, between 10 and 20 cells were imaged per sample. In pilot experiments, we determined that the arbitrarily chosen plane of focus through the cell did not significantly affect our results. Specifically, the variance of the data collected from one GFPexpressing cell at different focal planes is less than the variance of data collected from different cells at a common focal plane. 2.4. GFP expression in Arabidopsis The GFP–COP1 expression cassette was subcloned to the binary T-DNA vector pBIN19 (Bevan, 1984). Arabidopsis thaliana (ecotypes Nossen and Columbia) 37 A.G. von Arnim et al. / Gene 221 (1998) 35–43 Table 1 GFP-fusion cloning vectors Name Fluorescence Seriesa Remarksb pAVA68 pAVA90 pAVA120 pAVA122 pAVA319 pAVA322 pAVA118 pAVA121 pAVA321 pAVA192 pAVA389 Wild-type Wild-type Wild-type Wild-type Wild-type Wild-type S65T S65T S65T Y66H (BFP) Y66H (BFP) I I I I II II I I II I I Splicing defect in Arabidopsis C-fusions only C-fusions only No translational enhancer C-fusions only C-fusions only C-fusions only aRefer to Fig. 1 for restriction enzyme sites in Series I and Series II vectors. bVectors marked ‘C-fusions only’ do not contain the NcoI site displayed in Fig. 1. was transformed by Agrobacterium-mediated T-DNA transfer according to the vacuum infiltration procedure (Bechtold et al., 1993). Kanamycin-resistant transgenic seedlings were grown on GM medium with 1% sucrose ( Valvekens et al., 1988). Seedlings were mounted in water on glass slides and observed through 40× or 63× oil-immersion objectives. All eight independent transgenic lines showed the same subcellular localization pattern for GFP–COP1. In some experiments, seedlings were planted in soil after microscopic examination and grown to seed set without exhibiting any adverse effects in response to the prior microscopic illumination treatment. 2.5. Immunofluorescence and GUS assays The immunofluorescence assay for COP1 using an affinity-purified polyclonal antiserum against Arabidopsis COP1 (McNellis et al., 1994a) or a mouse monoclonal antibody against Arabidopsis COP1 (McNellis et al., 1994b) in conjunction with FITClabelled secondary antibodies was carried out essentially as described (Matsui et al., 1995). In-situ GUS staining was carried out as described previously (von Arnim and Deng, 1994). 3. Results and discussion 3.1. Construction of GFP fusion vectors We have constructed 11 plasmid cloning vectors for the expression of fluorescent protein fusions in plants (Fig. 1 and Table 1). The basic vector (pAVA120) is composed of a dual 35S promoter from CaMV, the translational enhancer sequence of TEV, the GFP coding sequence, and the 35S transcriptional terminator from CaMV. In addition, all vectors carry the ampicillin resistance gene and the high copy number ColE1 origin of replication. Cloning sites are detailed in Fig. 1C. The vectors differ in the restriction enzyme digestion sites available for cloning of protein fusions. While one subset of vectors allows the construction of both aminoand carboxy-terminal fusions via unique NcoI and BglII sites, respectively (pAVA120, 121, 122, 319, 321, 389), the other subset supports carboxy-terminal fusions only (pAVA68, 90, 118, 322; Table 1). The vectors fall into two subsets with respect to cloning sites flanking the expression cassette (Fig. 1). Series I vectors carry identical sites on the left and right of the expression cassette, useful if subsequent subcloning is to remain unbiased with regard to the orientation of the expression cassette. In contrast, Series II vectors carry different restriction sites on either side of the cassette, facilitating directional subcloning. Furthermore, pAVA122 has a deletion of the translational enhancer sequence, which resulted in an approximately threefold reduction in fluorescence intensity in the onion epidermal cell transient expression assay when compared to the most closely related plasmid, pAVA120 (data not shown). This feature may be helpful if overexpression of the protein is toxic to the cell or otherwise undesirable. Vector pAVA68 contains an unmodified GFP cDNA, which lacks the splice site and codon usage modifications required for expression in Arabidopsis (Haseloff et al., 1997). However, GFP expression from pAVA68 is indistinguishable from that of the most closely related vector, pAVA90, in onion epidermal cells and may potentially be superior over pAVA90 in other plant species. 3.2. Fluorescence properties The GFP-fusion vectors differ primarily in the spectral properties of the encoded fluorescent protein. Six of the 38 A.G. von Arnim et al. / Gene 221 (1998) 35–43 Fig. 1. Structure of cloning vectors for expression of GFP fusion proteins. (A) Series I vectors are based on the plasmid pRTL2 (Carrington et al., 1990). (B) Series II vectors are based on pBluescript SK+ (Stratagene). (C ) Nucleotide sequences at the cloning sites around the start and stopcodons of GFP. Amino acids are numbered according to the sequence in Prasher et al. (1992) and are given in single-letter code. vectors contain GFP with wild-type fluorescence, three contain the GFP–S65T mutant, and two contain the GFP–Y66H mutant. Expression of the respective fluorescent proteins in a transient onion epidermal cell assay confirmed the expected fluorescence properties (Heim et al., 1994, 1995). Whereas wild-type GFP fluoresced bright green under both ultraviolet ( UV ) and blue light (not shown), the GFP–S65T mutant fluoresced bright green under blue light but only faintly under UV (Fig. 2). GFP–Y66H (BFP) fluoresced blue under UV, but did not fluoresce under blue light ( Fig. 2). GFP–S65T and BFP show a limited spectral overlap (Heim et al., 1994, 1995). Therefore, these two reporter proteins may be suitable for the concomitant tagging of two different proteins in the same cell. The feasibility of double labelling in the widely popular onion epidermal cell transient expression assay ( Varagona et al., 1992) was tested by co-expressing GFP–S65T and BFP together, as well as individually. As expected, GFP–S65T and BFP showed a limited spectral overlap using the filter combinations described here ( Fig. 2A, left and central column; Fig. 2B). Therefore, double labelling should be feasible, at least if BFP and GFP–S65T do not show one of the following three types of optical interaction. First, it is conceivable that fluorescence resonance energy transfer (FRET ) from BFP to GFP–S65T might result in increased green fluorescence emission by GFP–S65T in the presence of BFP, if BFP and GFP–S65T were physically associated (Mitra et al., 1996; Miyawaki et al., 1997). Second, GFP–S65T, which absorbs blue light, might quench blue fluorescence emission of BFP by reabsorbing blue light. Third, the presence of BFP, which significantly absorbs blue light, might inhibit excitation of GFP–S65T by ‘shading’ GFP–S65T. We tested these possibilities by coexpressing BFP and GFP–S65T in onion epidermal cells at a molar ratio of 2.5:1. This ratio was chosen as a trade-off between the ratio that would maximize interaction between GFP–S65T and BFP (1:1) and the ratio that would result in equal fluorescence intensity (6:1) based on the higher quantum yield and absorption coefficient of GFP–S65T when compared to BFP. We then examined green fluorescence emission under UV excitation as a sensitive measure of any optical interaction. As shown in Fig. 2 (bottom row), the combined fluorescence after co-expression of GFP–S65T and BFP was indistinguishable from the value expected for an additive interaction of the two proteins (Fig. 2B, arrow labelled ‘minus’), indicating that no significant FRET, quenching, or shading occurs between the two proteins. This is not surprising, because FRET relies on the immediate proximity of the two chromophores, whereas quenching and shading should become measurable only under very high concentrations of GFP. Therefore, it may be feasible to employ the fusion vectors in conjunction with the expression assay described here to perform doublelabelling experiments and to examine direct protein–protein interactions in vivo by monitoring FRET. A.G. von Arnim et al. / Gene 221 (1998) 35–43 39 Fig. 2. (A) GFP–S65T and BFP fluorescence are additive upon co-expression in onion epidermal cells. Representative cells expressing GFP–S65T alone (top row), BFP alone (middle row) or both GFP–S65T and BFP together (bottom row), were imaged through filter combinations designed to visualize BFP ( left column, cube III, UVB), GFP–S65T (central column, cube II, BG) and their interaction (right column, cube IV, UVG). (B) Quantification of fluorescence from (A). Note that GFP–S65T fluoresces more brightly than BFP. In the bottom graph, fluorescence values expected in the absence of any interaction (−) or in the presence of FRET or fluorescence quenching (+) are indicated by arrowheads. 3.3. A GFP-based nuclear targeting assay Because GFP is detectable in live cells without the complication of a histochemical assay, this reporter has potential advantages over the GUS reporter enzyme. However, GFP enters the nucleus of diverse plant cells, even in the absence of an added nuclear localization signal ( Fig. 2; Grebenok et al., 1997). To address this potential handicap of GFP, we have employed two strategies. First, we have established a simple, yet quantitative, assay for the measurement of nuclear and total cellular GFP fluorescence levels in live onion epidermal cells (see Section 2). As shown in Fig. 3A and Table 2, approximately 11% of cellular GFP was localized to the nucleus, whereas the remainder was cytoplasmic. The fraction of nuclear GFP was not affected by the expression level of the protein (not shown). We then tested whether GFP might localize to the nucleus because of a cryptic nuclear localization signal. Alternatively, GFP might enter the nucleus by diffusion due to its small size (28 kDa). If GFP contained an NLS, then a GFPdimer, consisting of two GFP domains arranged in tandem within one polypeptide, should result in identical or increased nuclear localization. In contrast, if GFP did not contain an NLS, dimerization should reduce the level of nuclear GFP, because the increased size would inhibit diffusion through the nuclear pore complexes. A head-to-tail dimer of GFP (55 kDa) showed a reduced nuclear localization (Fig. 3B; Table 2), and a trimer of GFP (83 kDa) remained primarily cytoplasmic ( Fig. 3D; Table 2). A fusion between GUS and GFP (Fig. 3C; Table 2), previously proposed to represent bona fide cytoplasmic localization (Grebenok et al., 1997) resembled the GFP-dimer and the GFP-trimer ( Table 2). As expected, when stained for GUS activity, the GFP–GUS fusion appeared cytoplasmic (not shown). These data indicate that GFP does not contain an NLS of its own. Rather, unfused GFP most likely underpasses the size exclusion limit for bidirectional diffusion through nuclear pore complexes, estimated at 40–60 kDa (Görlich and Mattaj, 1996). Our interpretation is consistent with a previous report that lysed protoplasts expressing GFP rapidly lose all nuclear fluorescence (Grebenok et al., 1997). We then compared our nuclear targeting assay with a tobacco protoplast assay described previously (Grebenok et al., 1997) by expressing a fusion between GFP and the nuclear NIa protein from tobacco etch 40 A.G. von Arnim et al. / Gene 221 (1998) 35–43 Fig. 3. Nuclear localization of GFP fusions. (A) GFP; (B) GFP dimer; (C ) GFP–GUS; (D) GFP-trimer; (E ) GFP–NIa. virus ( Restrepo et al., 1990). Whereas exclusively nuclear accumulation of GFP–NIa had been reported using confocal microscopy (Grebenok et al., 1997), we observed that GFP–NIa was strongly enriched in the nucleus, but that a detectable fraction remained cytoplasmic. ( Fig. 4; Table 2). This minor discrepancy may be due to the different cell types employed or to the increased sensitivity of our assay, which, other than a typical confocal imaging protocol, operated at a camera gain-setting of 1, more closely representing quantitative imaging. Because the GFP-based assay is carried out in live cells, affords improved spatial resolution, and is easily quantifiable, it is generally superior to the GUSbased assay. However, for fusion proteins that are smaller than the size exclusion limit for nuclear diffusion, results of the GFP-based assay must be interpreted cautiously. Table 2 Percentage of nuclear fluorescence determined for GFP-fusions in the transient onion epidermal cell assay Protein Nuclear fluorescence (%) Mean±SD GFP GFP-dimer GFP–GUS GFP-trimer GFP–NIa 11.0%±1.2% 7.9%±2.2% 7.3%±2.7% 5.5%±1.5% 58.3%±7.0% a b b b c SD, standard deviation. The letters a, b, and c separate means that are significantly different at the 95% confidence level. 3.4. Transient expression of GFP-fusion proteins To establish whether the newly constructed GFPfusion vectors were suitable for studying the subcellular localization of diverse plant proteins, we created fusions between GFP and each of a series of proteins involved in the light control of seedling development, including Arabidopsis COP9, FUS6, and DET1 (Castle and Meinke, 1994; Pepper et al., 1994; von Arnim and Deng, 1996; Wei et al., 1994). Data on the subcellular localization of these proteins had previously been obtained by independent methods. COP9 and FUS6 are part of a nuclear protein complex of ~550 kDa, but both lack canonical NLSs (Castle and Meinke, 1994; Wei et al., 1994; Chamovitz et al., 1996; Staub et al., 1996). DET1 had previously exhibited nuclear uptake as a GUSfusion protein in Arabidopsis protoplasts, but the NLSs in DET1 remain to be defined (Pepper et al., 1994). COP9 and DET1 were cloned as carboxy-terminal fusions to GFP, and FUS6 was cloned as an aminoterminal fusion (Fig. 4). While GFP–COP9 and FUS6–GFP showed nuclear accumulation to approximately the extent of unfused GFP ( Fig. 4A and B), the nuclear accumulation of GFP–DET1 was poor ( Fig. 4C ). The GFP–COP9 fusion has a predicted molecular mass of 50 kDa. Because a GFP-dimer (55 kDa) showed a reduction in nuclear localization compared to unfused GFP, the nuclear localization of GFP–COP9 cannot be explained easily by simple diffusion, but may be more consistent with a form of active or facilitated transport. It is possible that COP9 contains A.G. von Arnim et al. / Gene 221 (1998) 35–43 41 Fig. 4. Subcellular localization of Arabidopsis COP9, FUS6, DET1, and COP1 proteins as GFP fusions in onion epidermal cells. (A) GFP–COP9; (B) FUS6–GFP; (C ) GFP–DET1; (D) GFP–COP1. (D) is enlarged in order to demonstrate the subnuclear speckles formed by GFP–COP1. The bar corresponds to 10 mm. Arrows point to the nucleus. its own non-canonical NLS, that COP9 enters the nucleus by co-transport with another protein, or that slow diffusion and nuclear retention cooperate in the nuclear accumulation of GFP–COP9. The FUS6–GFP fusion has a predicted molecular mass of 82 kDa, and its diffusion into the nucleus is predicted to be prevented by its size. The incomplete nuclear localization of FUS6–GFP may be mediated by piggy-back transport with a cellular partner protein or by a weak noncanonical NLS. Given that FUS6 and COP9 are part of a nuclear-localized ~550-kDa protein complex in Arabidopsis and cauliflower (Chamovitz et al., 1996; Staub et al., 1996), the efficient nuclear translocation of GFP–COP9 and FUS6–GFP may be dependent on additional subunits of the COP9 complex that are ratelimiting in the heterologous onion cell assay. In our assay, the GFP–DET1 protein demonstrated a limited nuclear uptake (Fig. 4C ). GFP–DET1 has a predicted molecular mass of 90 kDa. DET1 may enter the nucleus by co-transport with a partner protein, and co-transport may be more efficient in the native species Arabidopsis (Pepper et al., 1994) than in the heterologous onion system. The COP1 protein cooperates with DET1, FUS6, and COP9 in repressing photomorphogenic development in Arabidopsis. COP1 had previously been shown to be weakly nucleophilic as a GUS-fusion protein in both onion epidermal cells and in transgenic Arabidopsis (von Arnim and Deng, 1994,von Arnim et al., 1997). Localization of a GFP–COP1 fusion protein in onion epidermal cells confirmed these results. Moreover, GFP–COP1 accumulated in discrete subnuclear particles ( Fig. 4D). The particles were of a defined size and shape and were therefore clearly distinguishable from the irregularly shaped, much larger, and less numerous cytoplasmic inclusion bodies that formed at a high expression level of GFP–COP1 in the cytoplasm (not shown). The distinct subnuclear localization of the GFP–COP1 protein suggested that GFP fusions displayed improved resolution at the subcellular level compared to GUS fusions. 3.5. GFP–COP1 fusion protein in transgenic Arabidopsis After establishing the suitability of the GFP-fusion vector system in transient expression assays, we examined whether GFP-fusion proteins may also be visualized in transgenic Arabidopsis. We selected the GFP–COP1 fusion protein as an example to test whether the subnuclear particles seen in the transient expression assay were reproducible in transgenic Arabidopsis. As expected, GFP–COP1 was detected in transgenic seedlings ( Fig. 5A). Moreover, whereas the cytoplasmic fraction of the GFP–COP1 fusion protein accumulated in a single cytoplasmic inclusion body per cell (not shown), exactly as observed for the GUS–COP1 fusion protein (von Arnim and Deng, 1994), the nuclear GFP–COP1 protein was restricted to subnuclear speckles ( Fig. 5A and B), as previously observed in onion epidermal cells ( Fig. 4D). Typically, one Arabidopsis root epidermal cell nucleus contained approximately 25 evenly distributed speckles. Independently, we have confirmed the confinement of COP1 to subnuclear particles in an immuno- 42 A.G. von Arnim et al. / Gene 221 (1998) 35–43 Fig. 5. Subnuclear localization of COP1 in Arabidopsis. (A) Root epidermal cell nucleus of a 35S:gfp–cop1 transgenic seedling. Note that not all of approximately 25 nuclear speckles are in focus. (B) 4,6-diamidino-2-phenylindole (DAPI )-staining of the nucleus shown in (A). (C ) Immunofluorescence staining of an Arabidopsis mesophyll cell. Note speckled distribution of COP1 within the nucleus. (D) GUS-staining of a 35S:gus–cop1 transgenic Arabidopsis hypocotyl cell. Note nuclear speckles. fluorescence assay on non-transgenic, wild-type, Arabidopsis protoplasts using an anti-COP1 specific polyclonal antiserum (Fig. 5C ) as well as an anti-COP1 monoclonal antibody (not shown). Interestingly, in transgenic Arabidopsis hypocotyl cells, nuclear speckles were also formed by the fusion protein between COP1 and GUS (Fig. 5D). Taken together, these data suggest that the localization of COP1 to subnuclear speckles is functionally significant. To test this hypothesis further and to address whether COP1 retains full activity after fusion to GFP, we crossed the gfp–cop1 transgene into a cop1-null mutant background, the lethal allele cop1–5 (Deng et al., 1992). We were able to derive a family that was homozygous for the cop1–5 mutation and that segregated for the gfp–cop1 transgene. In this family, all phenotypically mutant, cop1-like, seedlings were GFP-negative, whereas all wild-type seedlings were GFP-positive ( Table 3). Moreover, a family segregating for both the cop1–5 mutation and the gfp–cop1 transgene exhibited a reduction in the number of cop1-like seedlings ( Table 3). These data indicate that the gfp–cop1 fusion rescued the cop1 mutation by genetic complementation. Therefore, the fusion of GFP to COP1 did not compromise the activity of the COP1 protein. 3.6. Conclusions (1) The newly constructed GFP-fusion vectors are suitable for visualizing fluorescently labeled proteins at high levels of subcellular resolution in transient expression assays and in transgenic plants. (2) Amino-terminal and carboxy-terminal fusions to GFP may be constructed easily. (3) Spectrally distinguishable GFP derivatives are available for double labelling in situ. (4) A GFP–COP1 fusion protein exhibited biological activities similar or identical to the native protein, as shown by genetic complementation of a lethal cop1 allele in transgenic Arabidopsis. (5) Individual vectors are available from the authors upon request and will be distributed through the Arabidopsis Biological Resource Center (ABRC ) at Ohio State University (http://aims.cps.msu.edu/aims/). Table 3 Seedling phenotypes within Arabidopsis populations segregating for the lethal cop1–5 allele (1) or the gfp–cop1 transgene (3) or both (2) Number 1 2 3 Genotypea Phenotype Ratio of wild-type:mutant cop1 gfp–cop1 Wild-type gfp+ cop1 gfp+ +/− +/− −/− −/− +/− +/− 300 203 165 nd nd 25/25 93 19 59 nd nd 0/19 nd, not done. aGenotype of the parent plant that was selfed to derive the segregating population. −, cop1 mutant allele; +, COP1 wild-type allele. 3.2:1 11.0:1 2.8:1 A.G. von Arnim et al. / Gene 221 (1998) 35–43 Acknowledgement We thank Dr J.C. Carrington for the gift of plasmids pRTL2-GUS and pRTL2-GUS-NIa, Dr J. Chory for a DET1 cDNA, Dr M. Chalfie for a GFP cDNA, Dr J. Haseloff for the mGFP4 cDNA, Drs R. Heim and R. Tsien for the S65T and Y66H mutant GFP cDNAs and L.-H. Ang for construction of plasmid pAVA192. This work was supported by a National Institutes of Health grant (GM47850) to X-W.D. and by a US Department of Energy grant (DE-FG02-96ER20223) to A.G.v.A. References Bechtold, N., Ellis, J., Pelletier, G., 1993. In planta Agrobacterium mediated gene transfer by infiltration of adult Arabidopsis thaliana plants. C.R. Acad. Sci. Paris 316, 1194–1199. Bevan, N., 1984. Binary Agrobacterium vectors for plant transformation. Nucleic Acids Res. 12, 8711–8721. Carrington, J.C., Freed, D.D., Oh, C.S., 1990. Expression of potyviral polyproteins in transgenic plants reveals three proteolytic activities required for complete processing. EMBO J. 9, 1347–1353. Castle, L., Meinke, D., 1994. A FUSCA gene of Arabidopsis encodes a novel protein essential for plant development. Plant Cell 6, 25–41. Chamovitz, D.A., Wei, N., Osterlund, M.T., von Arnim, A.G., Staub, J.M., Matsui, M., Deng, X.-W., 1996. The COP9 complex, a novel multisubunit nuclear regulator involved in light control of a plant developmental switch. Cell 86, 115–122. Davis, S.J., Vierstra, R.D., 1998. Soluble, highly fluorescent variants of green fluorescent protein (GFP) for use in higher plants. Plant Mol. Biol. 36, 521–528. Deng, X.-W., Matsui, M., Wei, N., Wagner, D., Chu, A.M., Feldmann, K.A., Quail, P.H., 1992. COP1, an Arabidopsis regulatory gene, encodes a novel protein with both a Zn-binding motif and a Gb homologous domain. Cell 71, 791–801. Görlich, D., Mattaj, I.W., 1996. Nucleocytoplasmic transport. Science 271, 1513–1518. Grebenok, R.J., Pierson, E., Lambert, G.M., Gong, F.-H., Afonso, C.L., Haldeman-Cahill, R., Carrington, J.C., Galbraith, D.W., 1997. Green-fluorescent protein fusions for efficient characterization of nuclear targeting. Plant J. 11, 573–586. Haseloff, J., Siemering, K.R., Prasher, D.C., Hodge, S., 1997. Removal of a cryptic intron and subcellular localization of green fluorescent protein are required to make transgenic Arabidopsis plants fluoresce brightly. Proc. Natl. Acad. Sci. USA 94, 2122–2127. Heim, R., Prasher, D.C., Tsien, R.Y., 1994. Wavelength mutations and posttranslational autoxidation of green fluorescent protein. Proc. Natl. Acad. Sci. USA 91, 12501–12504. Heim, R., Cubitt, A.B., Tsien, R.Y., 1995. Improved green fluorescence. Nature 373, 663–664. Heinlein, M., Epel, B., Padgett, H.S., Beachy, R.N., 1995. Interaction of tobamovirus movement proteins with the plant cytoskeleton. Science 270, 1983–1985. Köhler, R.H., Cao, J., Zipfel, W.R., Webb, W.W., Hanson, M.R., 1997a. Exchange of protein molecules through connections between higher plant plastids. Science 276, 2039–2042. 43 Köhler, R.H., Zipfel, W.R., Webb, W.W., Hanson, M.R., 1997b. The green fluorescent protein as a marker to visualize plant mitochondria in vivo. Plant J. 11, 613–621. McNellis, T., von Arnim, A.G., Araki, T., Komeda, Y., Miséra, S., Deng, X.-W., 1994a. Genetic and molecular analysis of an allelic series of cop1 mutants suggests functional roles for the multiple protein domains. Plant Cell 6, 487–500. McNellis, T., von Arnim, A.G., Deng, X.-W., 1994b. Overexpression of Arabidopsis COP1 results in partial suppression of light-mediated development: evidence for a light-inactivable repressor of photomorphogenesis. Plant Cell 6, 1391–1400. Matsui, M., Stoop, C.D., von Arnim, A.G., Wei, N., Deng, X.-W., 1995. Arabidopsis COP1 protein specifically interacts with a novel cytoskeleton associated protein, CIP1. Proc. Natl. Acad. Sci. USA 92, 4239–4243. Mitra, R.D., Silva, C.M., Youvan, D.C., 1996. Fluorescence resonance energy transfer between blue-emitting and red-shifted excitation derivatives of the green fluorescent protein. Gene 173, 13–17. Miyawaki, A., Liopis, J., Heim, R., McCaffery, J.M., Adams, J.A., Ikura, M., Tsien, R.Y., 1997. Fluorescent indiators for Ca2+ based on green fluorescent proteins and calmodulin. Nature 388, 882–887. Olson, K.R., McIntosh, J.R., Olmsted, J.B., 1995. Analysis of MAP4 function in living cells using green fluorescent protein (GFP) chimeras. J. Cell Biol. 130, 639–650. Pepper, A., Delaney, T., Washburn, T., Poole, D., Chory, J., 1994. DET1, a negative regulator of light-mediated development and gene expression in Arabidopsis, encodes a novel nuclear-localized protein. Cell 78, 109–116. Prasher, D.C., Eckenrode, V.K., Ward, W.W., Prendergast, F.G., Cormier, M.J., 1992. Primary structure of the Aequorea victoria greenfluorescent protein. Gene 111, 229–233. Restrepo, M.A., Freed, D.D., Carrington, J.C., 1990. Nuclear transport of plant potyviral proteins. Plant Cell 2, 987–998. Sambrook, J., Fritsch, E.F., Maniatis, T., 1989. Molecular Cloning: A Laboratory Manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Staub, J.M., Wei, N., Deng, X.-W., 1996. Evidence for FUS6 as a component of the nuclear-localized COP9 complex in Arabidopsis. Plant Cell 8, 2047–2056. Valvekens, D., Van Montagu, M., van Lijsebettens, M., 1988. Agrobacterium tumefaciens-mediated transformation of Arabidopsis thaliana root explants by using kanamycin selection. Proc. Natl. Acad. Sci. USA 85, 5536–5540. Varagona, M.J., Schmidt, R.J., Raikhel, N.V., 1992. Nuclear localization signal(s) required for nuclear targeting of the maize regulatory protein Opaque-2. Plant Cell 4, 1213–1227. von Arnim, A.G., Deng, X.-W., 1994. Light inactivation of Arabidopsis photomorphogenic repressor COP1 involves a cell type specific modulation of its nucleocytoplasmic partitioning. Cell 79, 1035–1045. von Arnim, A.G., Deng, X.-W., 1996. Light control of seedling development. Annu. Rev. Plant Physiol. Plant Mol. Biol. 47, 215–243. von Arnim, A.G., Osterlund, M.T., Kwok, S.F., Deng, X.-W., 1997. Genetic and developmental control of nuclear accumulation of COP1, a repressor of photomorphogenesis in Arabidopsis. Plant Physiol. 114, 779–788. Wang, S., Hazelrigg, T., 1994. Implications for bcd mRNA localization from spatial distribution of exu protein in Drosophila embryogenesis. Nature 369, 400–403. Wei, N., Chamovitz, D.C., Deng, X.-W., 1994. Arabidopsis COP9 is a component of a novel signaling complex mediating light control of development. Cell 78, 117–124.