INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The study of bacterial protein export has been greatly facilitated by the use of reporter genes whose products serve as enzymatic markers for cellular location. It is well established that reporter proteins such as alkaline phosphatase (encoded by phoA) and -lactamase (encoded by bla) function optimally only when localized to the periplasmic space. These reporters have been especially valuable in identifying regions of a protein that promote membrane translocation (6, 24) and allow predictions of membrane topology (37, 38).

-Galactosidase functions in the opposite manner of alkaline phosphatase and -lactamase in that it is enzymatically active only when retained in the cytoplasm (15, 18, 21, 40). When fused to appropriate export signals, -galactosidase is localized to cellular locations that render the enzyme inactive. This phenotype has been exploited for the isolation of mutants with restored -galactosidase activity. Many of these mutants ultimately led to the identification of a number of sec genes that encode important components of the Escherichia coli protein export machinery (17, 30, 40).

Another phenotype associated with E. coli strains that produce exported -galactosidase fusions is overproduction lethality, resulting from high-level synthesis of the hybrid proteins (3, 13). It has been observed that -galactosidase is incompatible with components of the export machinery and, as a consequence, the export pathway becomes "jammed" at the translocation step. This jamming event appears to be the result of the conformation assumed by -galactosidase and not of the presence of specific amino acid sequences that are incompatible with the protein export machinery (32, 47). Upon sufficient overproduction of -galactosidase hybrid proteins, the export pathway becomes so severely jammed that other exported proteins accumulate in the cytoplasm in their precursor form (3, 13). The overproduction lethality phenotype has also been exploited to characterize the protein export pathway. Specifically, mutations that reverse the lethal effects of hybrid protein overproduction have been used to reveal important features of the signal sequence of exported proteins (2, 4, 11, 12).

Despite the successes of using -galactosidase fusion proteins for genetic analysis of protein export, biases and limitations to the genetic selections and screens that use this reporter likely exist, potentially limiting the spectrum of export mutants that can be isolated. For example, -galactosidase fusions have been used in an attempt to isolate mutants defective in the insertion of integral membrane proteins. Although this selection yielded an informative class of mutants defective in disulfide bond formation (1), they were not revealing as to the cellular components important for membrane translocation.

It is predicted that use of an alternative reporter system may lead to the identification of additional components of the translocation machinery and facilitate better characterization of the components currently known to be important for protein export. In particular, a reporter protein that is efficiently translocated across the cytoplasmic membrane and yet is active only in the cytoplasmic compartment would be useful for initiating new screens for export-defective mutants.

To this end, we have investigated the use of green fluorescent protein (GFP) from Aequorea victoria as a reporter for protein localization in bacteria. GFP has several features that make it an attractive candidate for protein localization studies in bacteria. For example, the protein is active in E. coli, and it has proven to be a useful reporter for a number of investigations in this microorganism, including monitoring gene expression (45), assessing viability (7), and detecting bacteria in the environment (33). GFP is active as a chimeric protein and has provided details of bacterial cell division and chromosome partitioning (8, 19, 28, 34, 35, 42, 48, 52, 53, 55, 56). An additional characteristic of GFP that is potentially advantageous for protein localization studies is that GFP has a molecular mass of 27 kDa, which is significantly smaller than -galactosidase, whose molecular mass is ca. 130,000 kDa. GFP also functions as a monomer, in contrast to the tetrameric configuration required for -galactosidase activity. The three-dimensional structure of GFP is also known and reveals that GFP attains a relatively uncomplicated "-can" structure (41, 54) not unlike the structure of bacterial porin proteins (29). This fact suggests that GFP would likely be exported from the cytoplasm if fused to appropriate export signals. Also, GFP emits green light following excitation of an internal fluorophore composed of a Ser-Tyr-Gly sequence positioned near the protein's amino terminus. Excitation of GFP-expressing cells can be performed by exposure to long-wave UV light, making detection of GFP activity simple and obviating the need for specific substrates.

Although GFP is widely used to probe the events that occur within living cells (39), including protein localization in eukaryotic cells (20, 22, 43), the use of GFP for bacterial protein export studies has not previously been reported. Results presented in this study confirm that GFP can function as a reporter for protein localization in E. coli and hence provides a new tool for the analysis of bacterial protein export.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and plasmids. MC4100 [araD139 (argF-lac)169 flhD5301 fruA25 relA1 rpsL150 rbsR22] (46) was used as the host for fluorescence assays and cell fractionations along with the sec derivatives MM54 [MC4100, secA51(Ts) leu-59::Tn10] (40), IQ85 [MC100, secY24(Ts) zhe-33::Tn10] (44), and CK2163 (MC4100, secBL75Q) (16). DH5 (Gibco-BRL, Gaithersburg, Md.) was used as the host strain for cloning. XL1-Red (endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac mutD5 mutS mutT) (Stratagene, Inc., La Jolla, Calif.) was used as a host for mutagenesis of plasmid DNA.

Bacterial growth. Cells were cultured in Luria broth (LB) (46) or E minimal medium (50) at 30°C. Ampicillin was added at a concentration of 100 µg/ml when selecting or maintaining transformants. Transformations were performed as outlined by Inoue et al. (25). Bacterial growth was monitored spectrophometrically at an optical density of 600 nm (OD₆₀₀). The synthesis of fusion proteins was induced by addition of 5 µM isopropyl--D-thiogalactoside (IPTG). Cells were also cultured by distributing a lawn of bacteria in LB top agar (0.7% agar) supplemented with ampicillin onto LB plus ampicillin plates and positioning a sterile 25-mm filter disk that had been immersed in dithiothreitol (1 M) at the center of the plate.

Fluorescence assays. Cells were harvested from the surface of LB plates following incubation overnight at 30°C with sterile saline. The cells were washed once in saline, and the OD₆₀₀ of the cultures was determined. Fluorescence was measured at an excitation wavelength of 365 nm using a Dyna-Quant 200 fluorometer (Amersham Pharmacia Biotech, Piscataway, N.J.) modified with an interference filter (Edmund Scientific, Barrington, N.J.) to detect emission at 509 nm. The fluorescence values of three independently grown cultures were averaged and normalized to the density of the cultures. Fluorescence was expressed as arbitrary units. For comparison between bacterial strains, the fluorescence of MC4100 transformed with pMGC2 was set at a value of 1,000. For detection of fluorescence in colonies, plates were incubated for 3 to 4 days at 30°C. Plates were photographed under UV illumination at 385 nm.

Isolation and identification of export-defective mutants. pMGP2 was transformed into XL1-Red, and several ampicillin-resistant colonies were inoculated into 50 ml of LB, cultured overnight and then subcultured for a second overnight incubation. Plasmid DNA was prepared from the culture and transformed into DH5. The transformants were screened on plates containing 1,000 to 2,000 colonies using a long-wave UV lamp (395 nm) to identify those that were fluorescent. A reconstruction experiment was performed by mixing IQ85 and MC4100 transformed with pMGP2 at ratios of 1:1,000 and 1:10,000 and plating the mixed populations onto LB Amp plates at various concentrations. After incubation for 4 to 5 days, plates were observed under UV illumination to identify fluorescent colonies. Both fluorescent and nonfluorescent colonies were restreaked at 30 and 42°C to test for temperature sensitivity.

Cell fractionation. Cells were fractionated after growth to an OD of 1.0. Then, 10 ml of cells was pelleted and resuspended in 0.5 ml of periplasting buffer (20% sucrose, 1 mM EDTA, 30,000 U of Ready-Lyse lysozyme [Epicentre Technologies, Madison, Wis.] per ml). Samples were incubated on ice for 5 min. Spheroplasts were then pelleted by centrifugation at 12,000 × g for 2 min. The supernatant was reserved as the periplasmic fraction. The pelleted spheroplasts were lysed in 1 ml of water containing 400 U of Omnicleave endonuclease (Epicentre Technologies) per ml. The samples were incubated for 5 min at room temperature followed by brief sonication at 30 to 40% power. Unlysed cells were removed by centrifugation at 12,000 × g. The supernatant was removed and centrifuged at 138,000 × g for 1 h. The supernatant was reserved as the cytoplasmic fraction, and the pellet containing the membrane fraction was resuspended in 0.5% Sarkosyl, 10 mM Tris-HCl, and 5 mM EDTA. Cells were further fractionated into inner and outer membranes by a modification of the technique described previously (26). Cells were cultured as described above and disrupted by French pressing. The lysate was centrifuged for 20 min at 8,000 × g to remove any unbroken cells. The clarified lysate was layered atop a 15 to 70% discontinuous sucrose gradient and centrifuged at 100,000 × g for 4 h. A band representing the membrane fraction was isolated, diluted 1:2 with 30 mM Tris-HCl (pH 8.0), and then layered on a 53 to 70% discontinuous sucrose gradient and centrifuged at 247,000 × g for 18 h. The fraction representing the inner-membrane proteins was recovered off the top of the 53% sucrose layer, and the outer membrane fraction was collected off the top of the 70% sucrose layer. Each fraction was layered on a second 53 to 70% discontinuous sucrose gradient and centrifuged as described above. The membrane fractions were again diluted, as described above, and centrifuged for 2 h at 100,000 × g to pellet the membrane proteins. The pellets were washed once with 1 M KCl, followed by an additional centrifugation step. The pellets and KCl washes were reserved for analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Cell labeling and immune precipitation. Transformants were cultured in E minimal medium supplemented with amino acids excluding methionine and cysteine to an OD₆₀₀ of 0.3. The cultures were pulse-labeled with 140 µCi of [³⁵S]methionine and [³⁵S]cysteine (Amersham Pharmacia Biotech, Piscataway, N.J.) per ml for 1 min, followed by a 1-min chase using an equal volume of prewarmed E medium containing 0.8% unlabeled methionine and cysteine. Then, 1-ml samples were immediately combined with 50 µl of ice-cold 100% trichloroacetic acid and incubated on ice for 10 min. Precipitated proteins were pelleted for 2 min in a microcentrifuge followed by a wash with cold acetone. The pellet was resuspended in a solution of 20 mM Tris-HCl (pH 7.5), 2% SDS, and 20 mM EDTA and then boiled for 2 min. Insoluble material was removed by centrifugation, and a 30-µl portion was mixed with 650 µl of immunoprecipitation buffer (50 mM Tris-HCl [pH 8.0], 150 mM EDTA, 2% Triton X-100) and GFP rabbit antiserum. After 1 h of incubation at 4°C with gentle mixing, immobilized protein A (Pierce Chemical Co., Rockford, Ill.) was added, and the samples were incubated for an additional hour at 4°C. The immune complexes were isolated by centrifugation and washed twice with immunoprecipitation buffer and twice again with 10 mM Tris-HCl (pH 8.0). The final pellet was resuspended in 50 µl of 10 mM Tris-HCl (pH 8.0) and 50 µl of SDS sample buffer (62.5 mM Tris-HCl [pH 6.8], 20% glycerol, 10% -mercaptoethanol, 6% SDS, 0.001% bromophenol blue), boiled for 2 min, and centrifuged for 5 min. Then, 10 µl of supernatant was resolved by SDS-7.5% PAGE. Immune-precipitated proteins were visualized using a Bio-Rad GS-363 phosphorimaging system (Bio-Rad, Inc., Hercules, Calif.).

Hybrid protein purification and characterization. MBP-GFP hybrid protein was purified from MC4100 transformed with pMGP2 and pMGC2. One-liter cultures of each strain were grown in LB supplemented with glucose at 2 g/liter at 30°C for 24 h. Cells were harvested by centrifugation, resuspended in 50 ml of column buffer (20 mM Tris-HCl [pH 7.4], 200 mM NaCl, 1 mM EDTA, 10 mM -mercaptoethanol), and lysed by French pressing. The resulting cell lysate was centrifuged for 20 min at 8,000 × g. The supernatant was then centrifuged for 2 h at 100,000 × g to further clarify the lysate. Then, 3 ml of amylose resin (New England Biolabs), prepared as described by the manufacturer, was incubated with the cleared lysate for 20 min at room temperature. The protein-bound resin was transferred to a 10-ml disposable column and washed with 50 ml of column buffer. The hybrid protein was eluted with 10 ml of a 10 mM maltose solution and concentrated using a Centricon YM-50 concentrator (Amicon, Inc., Beverly, Mass.). After quantifying the protein spectrophotometrically, samples were diluted to 1 mg/ml in column buffer. The purity of the protein samples was assessed by SDS-PAGE and staining with Coomassie dye.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction and characterization of MBP-GFP hybrid proteins. The GFP variant GFPuv (9) was used to test the utility of GFP as a reporter for protein localization in E. coli. GFPuv was chosen because of its increased intensity of fluorescence relative to wild-type GFP and the fact that its codon usage has been optimized for expression in E. coli. MBP from E. coli was used as the target protein for fusion with GFP because it is a well-characterized exported protein normally localized to the periplasmic space. Gene fusions between gfp and malE were made as described in Materials and Methods, yielding the constructs shown in Fig. 1A. One construct, carried by pMGP2, encoded a hybrid protein where GFP was tagged onto the carboxy terminus of MBP, designated MBP[SS418]-GFP. Because it was unclear if GFP would function if localized outside of the cytoplasm, a construct was also assembled where gfp was fused to a version of malE that lacked a signal sequence. This gene fusion, carried by pMGC2, encoded the hybrid protein designated MBP[SS418]-GFP that was retained in the bacterial cytoplasm.

View larger version (17K): [in this window] [in a new window]   FIG. 1.   Constructs encoding MBP-GFP hybrid proteins. (A) Constructs carried by pMGP2 (encoding MBP[SS418]) and pMGC2 (encoding MBP[SS418]). (B) Constructs carried by pMGP22 (encoding MBP[SS128]) and pMGC22 (encoding MBP[SS128]). Features: lacI (diagonal stripes), LacI repressor; malE' (narrow vertical stripes), MBP; 'gfp, GFPuv (wide vertical stripes); POlac, lac promoter and operator; SS (horizontal stripes), signal sequence encoding region; and the linker region between malE and gfp (solid black line). The location of the region of gfp encoding the chromophore is indicated by an asterisk. Relevant restriction sites: E, EcoRV; B, BamHI; Bg, BglII; H, HindIII. , region of malE deleted in the constructs. The designation of the hybrid proteins encoded by each construct is shown on the right. The number within the square brackets indicates the number of amino acids of the mature portion of MBP fused to GFP.

View larger version (127K): [in this window] [in a new window]   FIG. 2.   Fluorescence of transformants expressing MBP-GFP hybrid proteins. MC4100 transformed with the following: 1, pMGP2; 2, pMGC2; 8, MM52 [secA(Ts)] transformed with pMGP2; 3, pMGC2; 7, CK2163 (secB) transformed with pMGP2; 4, pMGC2; 6, IQ85 [secY(Ts)] transformed with pMGP2; and 5, pMGC2.

Localization of MBP-GFP hybrid proteins. To determine the cellular location of the MBP-GFP hybrid proteins, cells were fractionated into cytoplasmic, periplasmic, and membrane samples. The fractions were resolved by SDS-PAGE and immunoblotted with antibodies directed against either GFP or MBP. As shown in Fig. 3, MBP[SS418]-GFP was localized primarily in the cytoplasm, as was the signal sequenceless MBP species encoded by pMalc2 (Materials and Methods). In contrast, MBP synthesized with its signal sequence intact, as well as MBP[SS418]-GFP, was localized to the periplasmic space, indicating that GFP hybrid proteins can be efficiently exported out of the cytoplasmic compartment.

View larger version (74K): [in this window] [in a new window]   FIG. 3.   Immunoblot analysis of MBP-GFP hybrid proteins. Cells transformed with plasmids expressing either MBP-GFP hybrid proteins (A and B) or MBP (C) were fractionated as described in Materials and Methods and analyzed by Western blot analysis. Immunoblots were decorated with anti-GFP antibody (A) or anti-MBP antibody (B and C). For panels A and B, lanes represent transformants of the following: lane 1, pJF2 (expressing wild-type MBP); lanes 2, 4, 6, and 8, pMGC2 (expressing MBP[418]-GFP); lanes 3, 5, 7, and 9, pMGP2 (expressing MBP[SS418]-GFP). (C) MBP expressed from: lane 1, pJF2; lanes 2, 4, 6, and 8, pMalc2; lanes 3, 5, 7, and 9, pMalp2. Lane designations: WC, whole cell; P, periplasmic fraction; C, cytoplasmic fraction; M, membrane fraction. The location of the full-length MBP-GFP protein is noted with arrows next to the nearest corresponding molecular weight marker; the asterisk indicates breakdown products detected with anti-MBP antibody.

Activity of MBP-GFP protein in sec mutants. To further characterize the MBP-GFP hybrid proteins with respect to protein export, the gene fusions were introduced to E. coli strains known to be defective in protein export. If MBP-GFP is exported by the normal sec-dependent pathway of protein export, then it was predicted that fluorescence could be detected if the hybrid proteins were expressed in a sec mutant. The MBP-GFP-encoding plasmids were transformed into three well-characterized sec mutants, previously shown to be important for export of a large number of proteins, including MBP. Figure 2 shows that fluorescence was detected whenever pMGP2 was transformed into a secA or secY mutant, even at 30°C, the permissive temperature for growth of these mutants (40, 44). Fluorescence was also detected when pMGP2 was transformed into a secB mutant. secB is a nonessential gene involved in the export of MBP, as well as other E. coli proteins (30, 31). As anticipated, transformants of pMGC2, encoding MBP[SS418]-GFP, all remained fluorescent.

Construction of additional malE-gfp fusions. To further characterize exported GFP hybrid proteins, additional malE-gfp gene fusions were constructed by deleting the region between the BglII and the two BamHI sites on both pMGP2 and pMGC2 (Fig. 1B). Similar to the results seen with the full-length MBP-GFP hybrids, transformants expressing these shorter hybrid proteins showed fluorescence only when expressed without the MBP signal sequence (MBP[SS128]-GFP) or when the exported protein (MBP[SS128]-GFP) was expressed in a secY mutant (Table 1). However, fractionation of the cells revealed that MBP[SS128]-GFP was not localized to the periplasmic space; rather, it was consistently detected in the membrane fraction. Additional fractionation studies revealed that the protein, with a predicted molecular mass of 40 kDa, fractionated with the outer membrane. In addition, an apparent breakdown product was found associated with the inner membrane fraction (Fig. 4). One possibility considered was that this breakdown product represented GFP that had been clipped free of the MBP sequences. However, this product had a molecular mass of 33 kDa, which is significantly larger than GFP.

View larger version (108K): [in this window] [in a new window]   FIG. 4.   Cell fractionation of shortened MBP-GFP hybrid protein. Cells expressing MBP[SS128]-GFP were fractionated into inner and outer membranes as described in Materials and Methods. The immunoblot shown was decorated with anti-GFP antibody. MBP[SS128]-GFP migrated at a predicted molecular mass of 40 kDa and fractionated with the outer membrane. An apparent smaller-molecular-weight derivative of this protein was also found in the inner membrane fraction. Molecular weight (MW) standards are as shown. IM, inner membrane fraction; IM-W, KCl wash of inner membrane fraction; OM, outer membrane fraction; OM-W, KCl wash of outer membrane fraction. Although not shown, OmpA and ATPase were used to monitor fractionation of the outer membrane and inner membranes, respectively.

Expression of full-length MBP-GFP hybrids does not result in overproduction lethality. Although GFP behaves similarly to -galactosidase in that neither protein is active when exported out of the cytoplasm, it remained to be determined if GFP hybrid proteins also resulted in an overproduction lethality phenotype characteristic of exported -galactosidase fusions. To test the effects of induced synthesis of MBP-GFP hybrid proteins, we took advantage of the fact that expression of the malE-gfp gene fusions is under control of the lac regulatory elements (Fig. 1). We observed that upon induction of MBP[SS128]-GFP synthesis with IPTG, transformants continued to grow at a rate similar to cultures grown without induction (data not shown). Microscopic examination of the cells also revealed no morphological changes and no evidence of cell lysis. Induction of hybrid protein synthesis was confirmed by Western blot analysis (data not shown).

Isolation of export-defective mutants. Since cells expressing GFP localized to the periplasmic space do not fluoresce, we predicted that fluorescence could be used as a phenotype to isolate export-defective mutants. Given our understanding of the protein export process, it was predicted that mutations that map to malE-gfp that block export of the hybrid gene product, including those that alter the signal sequence-encoding region (27), should be isolated.

View larger version (9K): [in this window] [in a new window]   FIG. 5.   Signal sequence mutations of MBP. The wild-type malE signal sequence encoding region is shown below the amino acid sequence of the amino-terminal region of the precursor MBP protein along with the sequence of three signal sequence mutations.

View larger version (33K): [in this window] [in a new window]   FIG. 6.   Signal sequence processing of MBP-GFP hybrid proteins. Shown are pulse-labeled MBP-GFP proteins immune precipitated with anti-GFP antibody. Transformants labeled: lane 1, pMGC2; lane 2, pMGP2; lane 3, pMGP2 (mutant M1); lane 4, pMGP2 (mutant M2.1); lane 5, pMGP2 (mutant M2.3). P, precursor (signal sequence unprocessed) form; M, mature form.

Why doesn't GFP function in the periplasmic space? Although it is not immediately obvious why fluorescence of MBP-GFP is not detected when the hybrid protein is localized to the periplasmic space, a number of possible explanations can be considered. For example, the exported protein could be irreversibly inactivated by modification such as proteolytic cleavage. Although full-length MBP-GFP was detected in immune blot analysis, degradation products related to the hybrid protein were also seen (indicated by the asterisks in Fig. 3). In addition, the amount of full-length MBP-GFP protein detected in the periplasmic space was also substantially reduced in comparison to the cytoplasmic version, perhaps explaining the lack of fluorescence.

View larger version (48K): [in this window] [in a new window]   FIG. 7.   Characterization of purified MBP-GFP hybrid protein. Hybrid proteins were purified from the periplasmic space or the cytoplasm and subjected to acid-base treatment, as described in Materials and Methods. (A) MBP[SS418]-GFP isolated from the cytoplasm. Tubes: 1, untreated; 2, acid treated; 3, acid-base treated. (B) MBP[SS418]-GFP isolated from the periplasmic space. Tubes: 1, untreated; 2, acid treated; 3, acid-base treated. (C) Relative fluorescence of tubes shown in panel A. (D) Relative fluorescence of tubes shown in panel B.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We have explored the use of GFP as a reporter for protein localization in E. coli. Because this protein is relatively small and attains a conformation that resembles a porin monomer (41), it was predicted to be compatible with the protein export process of bacteria. To test this prediction and to assess the activity of GFP when localized outside of the cytoplasm, fusions between MBP and GFP (the UV-shifted variant, GFPuv) were constructed. Characterization of these fusions revealed that GFP remained active as long as the protein remained localized to the cytoplasm; export of the protein, however, rendered GFP inactive. Cellular fractionation studies revealed that the MBP[SS418]-GFP hybrid was efficiently exported outside of the cytoplasm. However, although fusions that contained the signal sequence plus 128 amino acids of MBP were exported out of the cytoplasm, they were not faithfully localized to the periplasmic space. Export of the hybrid proteins was also shown to be dependent on the sec pathway since GFP fluorescence was detected in strains carrying different sec alleles. Efficient signal sequence processing of MBP[SS418]-GFP was also confirmed by pulse-chase experiments.

The observation that GFP is active only in the cytoplasm parallels the observation that -galactosidase fusions are likewise inactive when exported out of the cytoplasm. The distinction between Lac⁺ and Lac phenotypes has been successfully used to isolate a number of export-defective mutants. In contrast to -galactosidase, however, overproduction of MBP[SS418]-GFP did not cause a lethal jamming of the protein export machinery. This distinction likely is attributed to the smaller size and relatively simple conformation of GFP compared with -galactosidase. We did observe, however, that expression of the shorter MBP-GFP hybrid protein (MBP[SS128]-GFP) was toxic when overproduced. It is not likely that this toxicity is the result of jamming of the protein export machinery (3, 12, 32, 47), since the hybrid protein was found to be efficiently localized outside of the cytoplasm (Fig. 4). Furthermore, export jamming is, in general, reduced when shorter -galactosidase hybrid proteins are expressed (5), while the opposite was observed for GFP. The toxic effects of exported MBP[SS218]-GFP are not unprecedented, however, since expression of certain LamB fusion proteins is lethal to E. coli by a mechanism that does not involve export jamming (47). These results indicate that not every GFP fusion construct may be suitable for analysis of protein export.

The distinction in GFP activity in different subcellular compartments manifests itself in a striking phenotype between fluorescent and nonfluorescent colonies. The ease with which fluorescence can be detected and quantified makes GFP a potentially useful reporter with which to genetically track bacterial protein localization. To test the utility of GFP as a screenable marker for the isolation of export-defective mutants, pMGP2 was randomly mutagenized, and transformants were screened for fluorescent colonies. Fluorescent colonies were shown to carry plasmids where the MBP signal sequence had been mutationally altered, hence blocking entry of MBP[SS128]-GFP into the export pathway. The nature of the signal sequence mutations conformed to those previously isolated for malE (2, 4). The mutations included those that introduced a positive charge to a region of the signal sequence that requires hydrophobicity (Met18 to Lys) and introduction of a proline that could disrupt a requirement for a specific secondary structure (Ser13 to Pro and Leu10 to Pro). We also noted that the Leu10 to Pro mutation is identical to the previously isolated 10-1 malE signal sequence mutation (2, 4, 10).

The sensitivity of GFP as a reporter for export defective mutants was further tested by performing a reconstruction experiment where mixed cultures of wild-type E. coli and a secY mutant, both expressing MBP[SS418]-GFP, were cultured together on an agar plate. After 3 to 4 days of growth at 30°C, fluorescent colonies representing the secY mutants were clearly visible. Although not tested, the sensitivity of detection of mutants could perhaps be heightened by use of a fluorescence-activated cell sorter to separate fluorescent from nonfluorescent cells, as has been reported for other applications (49). Collectively, these results indicate that GFP can be used as a genetic tool for the isolation of different classes of export-defective mutants. In addition to the use of GFP to yield new insights into the protein export process, it also may serve as a substitute for -galactosidase as a reporter to determine the topological orientation of integral membrane proteins (15, 18, 36).

The observation that GFP is active only when localized to the cytoplasm also raises the related question of why the protein is inactive in the periplasmic space. By purifying MBP-GFP protein from both the periplasmic and cytoplasmic compartments, we were able to show that the lack of fluorescence in the periplasmic space could not be attributed solely to diminished levels of the protein. When equivalent amounts of MBP-GFP purified from both the periplasmic space and the cytoplasm were compared, only the GFP isolated from the latter compartment fluoresced. Furthermore, GFP fluorescence was not irreversibly inactivated in the periplasmic space, since fluorescence could be restored to the protein by treatment under conditions known to alter protein conformation, including acid-base treatment and prolonged incubation in the presence of a reducing agent. This latter observation suggests that disulfide bond formation plays a role in inactivation of GFP. Indeed, two cysteine residues are located in the GFP moiety (one in GFP, the other encoded by the region linking malE with gfp) of MBP[SS418]-GFP that could contribute to either intra- or intermolecular disulfide bonds. Altering the reducing environment of the periplasmic space, however, did not restore fluorescence in vivo. Collectively, these results indicate that the periplasmic environment is not suitable for GFP to fold into a conformation required for fluorescent activity. These results further suggest that GFP may also be a useful tool to study protein folding in the periplasmic space.