Created: 28th February 1999, last updated: 7th April 1999, © 1999 ABRF
Joelle N. Pelletier, Ingrid Remy, and Stephen W. Michnick
Département de Biochimie, Université de Montréal, Montréal, Canada.
Reprinted from the electronic version of JBT available at http://www.abrf.org/JBT/JBT.html accession #0012.
Address correspondence and reprint requests to Stephen W. Michnick, Département de Biochimie, Université de Montréal, C.P. 6128, Succursalle, Centre-ville, Montréal, Quebec, Canada H3C 3J7 (email: michnick@bch.umontreal.ca).
We describe a strategy for designing protein-fragment complementation assays to detect protein-protein interactions in vivo. The design and implementation of this strategy is illustrated with two assay systems: one based on the enzyme dihydrofolate reductase (DHFR) and implemented in Escherichia coli and a second, the ubiquitin split-protein sensor, which is demonstrated in a transiently transfected mammalian cell line, COS-7. In the DHFR-based assay, N- and C-terminal fragments of murine DHFR fused to GCN4 leucine zipper sequences were coexpressed in E. coli, in which the endogenous DHFR activity was inhibited with trimethoprim. Coexpression of the complementary fusion products restored colony formation. Survival occurred only when both DHFR fragments were present and contained leucine zipper-forming sequences, demonstrating that reconstitution of enzyme activity requires assistance of leucine zipper formation. Activity was then characterized in vitro. In the ubiquitin split-protein sensor, oligomerization domain-assisted reassembly of ubiquitin from fragments results in susceptibility to ubiquitin-specific proteases that recognize the intact ubiquitin structure and cleave between the C terminus of ubiquitin and a fused reporter protein (with DHFR used as the reporter). We demonstrated equilibrium and kinetics of this reassembly by transient expression of ubiquitin fragments fused to GCN4 leucine zipper sequences in COS-7 cells. Fragment-interface point mutations implemented in both systems resulted in a sequential decrease in signal, demonstrating specific fragment reassembly. Replacement of the leucine zippers with other proteins or protein domains or with expression libraries allow study of the equilibrium and kinetic aspects of protein-protein interactions and allow screening of cDNA libraries for binding of a target protein with unknown proteins. (J Biomol Tech 1999;10:32-40)
Key words: dihydrofolate reductase, ubiquitin, leucine zipper, enzyme reconstitution, protein design.
Many processes in biology, including transcription, translation, and metabolic or signal transduction pathways, are mediated by noncovalently associated multienzyme complexes.1 Much of modern biologic research is concerned with identifying proteins involved in these cellular processes, identifying their functions, and determining how, when, and where they interact with other proteins involved in specific pathways. With rapid advances in genome sequencing projects, there is a need to develop strategies to define protein linkage maps, detailed inventories of protein interactions that make up functional assemblies of proteins.2-4
Despite the importance of understanding protein assembly in biologic processes, there are few convenient methods for studying protein-protein interactions in vivo.5,6 A powerful and commonly used strategy, the yeast two-hybrid system, is used to identify novel protein-protein interactions and to examine the amino acid determinants of specific protein interactions.5,7-9 The approach enables screening of cDNA libraries or mutants of individual genes. However, this and other existing methods for in vivo detection require additional cellular machinery that exist only in specific cellular compartments and that usually cannot be used to distinguish between induced and constitutive interactions.
An alternative to these approaches is oligomerization-assisted complementation of fragments of monomeric proteins and detection of fragment reassembly in vivo. The first example of this approach was the ubiquitin-based split-protein sensor (USPS).10 We developed an assay based on similar principles using the enzyme murine dihydrofolate reductase (mDHFR).11 Fragmentation of proteins of known structure allows design of fragments that can reassemble successfully. The flexibility allowed in the design of this approach makes it applicable to situations for which other detection systems may not be suitable.
We first describe the oligomerization-assisted complementation of fragments of mDHFR as an example of functional reassembly of fragments into active enzyme. We designed two fragments of mDHFR, each of which is covalently attached to an oligomerizing domain (ie, leucine zipper). We demonstrate by an Escherichia coli survival assay that leucine zipper interaction is necessary for detecting reassembly of catalytically active mDHFR from its fragments. This method enables detection of subtle changes in enzyme structure, such as point mutations at the interface between the two assembled fragments.
The in vitro activity of the reassembled enzyme was examined. Because leucine zipper formation is crucial in detecting protein-fragment reassembly in the protein-fragment complementation assay (PCA) described, we propose that the strategy could be used to detect other protein-protein interactions. The leucine zippers would be replaced by interacting proteins of interest for detection of associations in vivo. The selection and design criteria applied to DHFR could be used to develop a strategy using any enzyme that would be appropriate for clonal selection or direct detection of protein-protein interactions.
We examine the results of implementation of USPS for mammalian cells to illustrate that PCAs can be easily implemented in a variety of cell types, regardless of their origin. The strategy is based on cleavage of reporter proteins fused to the C-terminal end of ubiquitin by cytosolic proteases (ie, ubiquitinases) that recognize and specifically bind to ubiquitin. The enzymes cleave the reporter protein from the C-terminal end of ubiquitin only if the structure of ubiquitin is intact. Fusion of a reporter protein-ubiquitin C-terminal fragment can also be cleaved by ubiquitinases, although only if coexpressed with an N-terminal fragment of ubiquitin that is complementary to the C-terminal fragment. We and others10 found that reconstitution of observable ubiquitinase activity occurs only if the N- and C-terminal fragments are bound through GCN4 leucine zipper-forming sequences. As well as offering a new approach for the study of kinetic and equilibrium aspects of protein-protein interactions in vivo, the assay could be adapted for detection of protein-protein interactions in a variety of cell types. This technique enables detection of induced (eg, initiated by cell growth or inhibitory factors) and constitutive protein-protein interactions in mammalian cells.
The general features of a PCA are illustrated in Figure 1. In designing a PCA, we first identify a protein that is relatively small and monomeric and for which structural and functional information exists, simple assays exist for its in vivo and in vitro measurement, and overexpression in eukaryotic, prokaryotic, or both cell types has been demonstrated. The gene coding for the target protein is then rationally cleaved into two or more fragments, according to the protein structure, to create subdomains. The resulting new N and C termini are located on the same face of the protein to avoid the need for long peptide linkers and to allow determination of the orientation dependence of the protein-protein interaction under study. The proteins of interest are then fused to each of the created fragments. The resulting fusion proteins are coexpressed in an appropriate cellular context, and an in vivo assay is performed to detect oligomerization-assisted complementation of the protein fragments.
FIGURE 1. General features of a protein-fragment complementation assay. An enzyme can be transformed and transfected into a host cell (top, left) and its activity detected by an in vivo assay (right). Oligomerization domains A and B are fused to N- and C-terminal fragments of the gene for the enzyme (bottom, left). Cotransformation and transfection of oligomerization domain-fragment fusions result in reconstitution of enzyme activity by oligomerization domain-assisted reassembly of the enzyme (right). Reassembly of enzyme does not occur unless oligomerization domains interact.
All reagents used were of the highest available purity. Mutagenic and sequencing oligonucleotides were purchased from Life Technologies (Rockville, MD, USA). Restriction endonucleases and DNA modifying enzymes were from Pharmacia Biotech (Baie d'Urfé, Quebec, Canada) and New England Biolabs (Beverly, MA, USA). For bacterial protein overexpression, E. coli strain BL21 carrying plasmid pRep4 (lac Iq, Qiagen, Mississauga, Ontario, Canada) was transformed with the appropriate DNA constructs.
DNA Constructs: Dihydrofolate Reductase Fusions All final constructs were based on the Qiagen pQE series of vectors, which contain an inducible promoter-operator element (tac), a consensus ribosomal binding site, initiator codon, and nucleotides coding for an N-terminal hexahistidine peptide. The full-length mDHFR is expressed from the Qiagen expression control vector pQE-16. The mDHFR fragments [1,2] and [3] carrying their own in-frame stop codon were produced by polymerase chain reaction methods using pMT3 (derived from pMT2)12 as a template and subcloned into the pQE-32 polylinker at appropriate restriction sites. All final constructs were verified by DNA sequencing. Residues 235 to 281 of the GCN4 leucine zipper (ie, SalI/BamHI 254-bp fragment) were obtained from a yeast expression plasmid pRS316 harboring that sequence10 and subcloned 3' to the hexahistidine tag and 5' to the DHFR fragments, yielding constructs Zip-[1,2] and Zip-[3]. The control expression construct, Control-[1,2], codes only for the histidine-tagged mDHFR F[1,2] without the zipper.
Ubiquitin Constructs Ub-dha, Nub-Zip, and Zip-Cub-dha fusions and destabilizing mutants of the original yeast expression constructs10 were introduced in the HindIII (for Ub-dha and Zip-Cub-dha) or HindIII and XbaI (for Nub-Zip) restriction sites of the mammalian expression vector pcDNAI/Amp (Invitrogen), which contains a cytomegalovirus promoter and enhancer and SV40 and polyomavirus eukaryotic origin of replication. Ub-dha consisted of the full-length ubiquitin protein fused to mDHFR with a hemagglutinin (HA) epitope tag (DHFR-HA [dha]). The Nub-Zip fusions and Zip-Cub-dha fusions consisted of the N- (codons 1 to 37) and C-terminal (codons 35 to 76) portion of ubiquitin, respectively, fused to the residues 235 to 281 of the GCN4 leucine zipper. To increase the stability of the fusions proteins in mammalian cells, the linker joining the zipper and Cub was replaced by a flexible linker coding for (GlyGlyGlyGlySer)4 at the XbaI and BamHI restriction sites of Zip-Cub-dha.
Creation of Stability Mutants and Methotrexate-Resistant Mutants Site-directed mutagenesis was performed according to the method of Kunkel.13 Mutagenesis reactions were carried out on appropriate DNA fragments subcloned into pBluescript SK+ (Stratagene) using oligonucleotides that encode a silent mutation producing or destroying a diagnostic restriction site. Fragments of putative mutants identified by restriction were subcloned back into the appropriate constructs. The mutations were confirmed by DNA sequencing.
Escherichia coli Survival Assay Competent E. coli BL21/pRep4 were transformed with the appropriate constructs and washed twice with minimal medium before plating on minimal medium plates containing 50 µg/mL kanamycin, 100 µg/mL ampicillin, and 0.5 µg/mL trimethoprim. One half of each transformation mixture was plated in the absence of 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG) and the second half in the presence of 1 mM IPTG. All plates were maintained at 37°C for 66 hours.
Escherichia coli Growth Curves Growth curves in liquid medium were obtained using minimal medium supplemented with ampicillin, kanamycin as well as IPTG (1 mM), and trimethoprim (1 µg/mL) where indicated. All experiments were performed in triplicate. Aliquots were withdrawn periodically for measurement of optical density (OD). Doubling time was calculated for early logarithmic growth (OD 600 between 0.02 and 0.2).
Dihydrofolate Reductase Fusion Overexpression and Native Purification Bacteria were propagated overnight in minimal medium under selective pressure and in a small volume, and they were used to inoculate a large volume (500 mL to 1 L) at 100X dilution. Cells were harvested after 24 hours and stored at -80°C. The cell pellet from 250 mL of E. coli cotransformed with appropriate constructs was lysed by sonication in 10 mL of buffer A (100 mM potassium phosphate at pH 8.0, 1 mM phenylmethylsulfonyl fluoride, 10 mM beta-mercaptoethanol). The lysate was clarified by microcentrifugation at 4°C for 10 minutes at top speed. Then, 0.5 mL of equilibrated Ni-NTA agarose was added to the supernatant, and the slurry was gently mixed at 4°C for 1 hour. The mixture was packed onto a column (7 X 12 mm) and washed with 10 mL of buffer A, 10 mL of buffer A plus 5 mM imidazole, 10 mL of buffer A plus 25 mM imidazole, and 10 mL of buffer A plus 50 mM imidazole. The proteins were eluted in 1 mL of buffer A (pH 7.5) plus 200 mM imidazole, and the imidazole was dialyzed out at 4°C against three changes of 300 volumes of buffer A (pH 7.5).
In Vitro Dihydrofolate Reductase Assays DHFR activity was monitored by fluorometry to follow the appearance of tetrahydrofolate (THF) (excitation = 310 nm; emission = 360 nm). Fixed substrate concentrations were 30 µM dihydrofolate (DHF) and 25 µM NADPH; the reaction buffer was freshly prepared buffer A (pH 7.5). Concentrations of substrates and inhibitor (methotrexate [MTX]) were determined spectrophotometrically. Reconstituted enzyme was added to buffer and NADPH; MTX was then added and the reactions initiated by addition of DHF. Initial rates were determined at 37°C under conditions for which less than 15% conversion to product had occurred. Blanks contained dialysis buffer instead of enzyme. Ki (MTX) was calculated by nonlinear regression with the program AXUM by the proportional occupancy binding function: [E·S]/[E]t = ([S]/Kd)/(1 + [S]/Kd + [MTX]/Ki), where E represents reassembled enzyme, and S and Kd (relating to DHF) are treated as unknowns.
Transfection and Western Blotting For each of the constructs, 1.5 µg was used to transfect COS-7 cells grown in six-well tissue culture plates. Transfections were performed using lipofectamine reagent (Life Technologies/Gibco BRL) according to the manufacturer's instructions. The cells were lysed 48 hours after transfection in lysis buffer containing 50 mM Tris (pH 8.0), 150 mM NaCl, 10% sodium dodecyl sulfate (SDS), 1% NP40, 1 mM EDTA, 20 µg/mL aprotinin, 5 mM AEBSF, 50 µg/mL soybean trypsin inhibitor, 0.5 µg/mL leupeptin, 0.7 µg/mL pepstatin, 25 µg/mL antipain dihydrochloride, and 50 mM N-ethylmethylmaleimide (ie, ubiquitinase inhibitor). The cell lysates were centrifuged for 10 minutes, and a portion of the supernatant was separated on 15% acrylamide-SDS gel, transferred to polyvinylidene difluoride (PVDF) membrane (DuPont-NEN), and immunoblotted with anti-HA mouse monoclonal antibody (Boehringer Mannheim) at a concentration of 1 µg/mL. Bound antibodies were detected using horseradish peroxidase-conjugated anti-mouse IgG (Amersham Life Science) and the chemiluminescence detection system (DuPont-NEN).
Pulse-Chase Experiments and Immunoprecipitation Forty-eight hours after transfection, the cells were labeled with Methionine/Cysteine Protein Labeling Mix [35S] (DuPont-NEN) at a concentration of 100 µCi/mL for 30 minutes at 37°C, followed by a chase of 0, 10, and 30 minutes. The cells were washed with cold phosphate-buffered saline, and lysis was performed as described earlier. The supernatants were immunoprecipitated with 3 µg of anti-HA mouse monoclonal antibody (Boehringer Mannheim) and protein A-sepharose (Pharmacia Biotech). Immunoprecipitates were washed three times with lysis buffer and once with lysis buffer without detergents. The immune complexes were then subjected to electrophoresis on a 15% acrylamide-SDS gel and analyzed by autoradiography.
All of the previously listed criteria are met by mDHFR. It is a small (21-kd), monomeric protein of known structure.14 The folding, catalysis, and kinetics of a number of DHFRs have been studied extensively.15 DHFR activity can be monitored in vivo by cell survival in cells grown in the absence of DHFR end products and in vitro. In the in vivo assay, we took advantage of the fact that E. coli DHFR is selectively inhibited by the antifolate drug trimethoprim. Because mammalian DHFR has a 12,000-fold lower affinity for trimethoprim than bacterial DHFR.16 Growth of bacteria expressing mDHFR in the presence of trimethoprim levels lethal to bacteria is an efficient means of selecting for reassembly of mDHFR fragments into an active enzyme.
Design Considerations Comparison of the crystal structures of mDHFR and human DHFR (hDHFR) suggests that their active sites and substrate binding pockets are identical and homologous to those of E. coli DHFR.14,17 DHFR comprises three structural fragments forming two domains: the adenine binding domain (F[2]) and a discontinuous domain (F[1] and F[3]).18,19 The folate binding pocket and the NADPH binding groove are formed mainly by residues belonging to F[1] and F[2]. Residues in F[3] contribute little to substrate binding and catalysis,17 but they do contribute to DHFR stability and to kinetic parameters.20,21
We have designed two fragments of mDHFR consisting of the sequences coding for F[1] and F[2], which are referred to as F[1,2]), and for F[3] to cause minimal disruption of the active site and NADPH cofactor binding sites. Cleavage of mDHFR at the loop that is the junction between F[2] and F[3] and fusion of the native termini produced a circularly permuted protein with physical and kinetic properties very similar to those of the native enzyme, suggesting that radical modifications in this loop are not disruptive to activity.22 The native N terminus of mDHFR and the novel N terminus created by cleavage occur on the same surface of the enzyme,14,17 facilitating N-terminal covalent attachment of each fragment to oligomerization domains such as the leucine zippers (Zip) used in this study. The resulting fusions are labeled Zip-[1,2] and Zip-[3].
Escherichia coli Survival Assays Cotransformation of bacteria with constructs coding for Zip-[1,2] and Zip-[3] (see Fig. 1) was undertaken with the goal of detecting enzymatic activity from reconstituted mDHFR. Figure 2A illustrates the results of cotransformation in the presence of trimethoprim, showing that colony growth under selective pressure is possible only in cells expressing both fragments of mDHFR. There is no growth in the presence of either Zip-[1,2] or Zip-[3] alone. Induction of protein expression with IPTG is essential for colony growth (see Fig. 2A). The presence of the leucine zipper on both fragments of mDHFR is essential, as illustrated by cotransformation of bacteria with both vectors coding for mDHFR fragments, only one of which carries a leucine zipper (see Fig. 2A).
FIGURE 2. (A) Escherichia coli survival assay on minimal medium plates. Control, left side of the plate: E. coli harboring pQE-30 (no inset); right side: E. coli harboring pQE-16, coding for wild-type murine dihydrofolate reductase (mDHFR). (Panel I) Left side of each plate: transformation with construct Zip-[1,2]; right side of each plate: transformation with construct Zip-[3]. (Panel II) Cotransformation with constructs Zip-[1,2] and Zip-[3]. (Panel III) Cotransformation with constructs Control-[1,2] and Zip-[3]. All plates contain 0.5 µg/mL trimethoprim. In panels I through III, plates on the right side contain 1 mM isopropyl-beta-d-thiogalactopyranoside (IPTG). (B) E. coli survival assay using destabilizing DHFR mutants. Panel I Cotransformation of E. coli with constructs Zip-[1,2] and Zip-[3:Ile114Val]. Panel II Cotransformation with Zip-[1,2] and Zip-[3:Ile114Ala]. Inset is a fivefold enlargement of the right-side plate. Panel III Cotransformation with Zip-[1,2] and Zip-[3:Ile114Gly]. All plates contain 0.5 µg/mL trimethoprim. Plates on the right side contain 1 mM IPTG.
Further experiments were performed to control for absence of activity by Zip-[1,2] in the presence of F-[3] (the reciprocal of Control-[1,2] plus Zip-[3]) and for the effect of variables, including changes in expression level, solubility, and susceptibility to proteolysis. The results are all satisfactory and will be published elsewhere (J. Pelletier, K. Arndt, A. Pluckthun, and S. Michnick, in preparation). Growth of control E. coli transformed with the full-length mDHFR is possible in the absence of IPTG because of low levels of expression in uninduced cells.
Stability Mutants We generated point mutants of F[3] to change the efficiency with which the mDHFR fragments assemble to form active enzyme as tests for specific reconstitution of protein fragments. Protein stability can be reduced by changing the side-chain volume in the hydrophobic core of a protein.10,23-26 Residue Ile114 of mDHFR occurs in a core beta strand at the interface between F[1,2] and F[3], isolated from the active site. Ile114 is in van der Waals contact with Ile51 and Leu93 in F[1,2] (15). We made point mutations of the wild-type mDHFR Ile114 to Val, Ala, or Gly. Figure 2B illustrates the results of double transformation of E. coli with construct Zip-[1,2] and the mutated constructs Zip-[3:Ile114Val], Zip-[3:Ile114Ala], or Zip-[3:Ile114Gly] in the presence of trimethoprim. The colonies obtained from cotransformation with Zip-[3:Ile114Ala] grew more slowly than those cotransformed with Zip-[3] or Zip-[3:Ile114Val] (see Fig. 2B, inset). No colony growth was detected in cells cotransformed with Zip-[3:Ile114Gly]. Overexpression of the mutants Zip-[3:Ile114X] was in the same range as Zip-[3], as determined by Coomassie-stained sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (data not shown).
To compare the relative efficiency of reassembly of mDHFR fragments, we measured the doubling time in liquid medium of E. coli transformed with the appropriate constructs. Doubling time of E. coli in minimal medium was essentially constant for all transformants (Table 1). Selective pressure by trimethoprim in the absence of IPTG and induction of mDHFR fragment expression with IPTG reflect the results obtained on solid media. The doubling time measured for cells expressing Zip-[1,2] plus Zip-[3], Zip-[1,2] plus Zip-[3:Ile114Val], and Zip-[1,2] plus Zip-[3:Ile114Ala] were 1.6-fold, 1.9-fold, and 4.1-fold higher, respectively, than the doubling time of E. coli expressing wild-type mDHFR (pQE-16) in the absence of trimethoprim and IPTG.
TABLE 1.
Relative Doubling Times of Escherichia coli With Mutations in Liquid Media
|
|
|||
| No Additions | + TRIM | + TRIM, + IPTG | |
|
|
|||
| Wild-type mDHFR | 1.1 ± 0.2a | 1.0 ± 0.2 | NG |
| Zip-[1,2] + Zip[3] | 1.0 ± 0.2 | NG | 1.7 ± 0.2 |
| Destabilizing mutations | |||
| Zip-[1,2] + Zip[3:Ile114Val] | 1.0 ± 0.2 | NG | 2.1 ± 0.2 |
| Zip-[1,2] + Zip[3:Ile114Ala] | 1.1 ± 0.2 | NG | 4.3 ± 0.3 |
| Zip-[1,2] + Zip[3:Ile114Gly] | 1.1 ± 0.2 | NG | NG |
| Methotrexate-resistant mutationsb | |||
| Zip-[1,2:Leu22Phe] + Zip[3] | 1.2 | ND | 1.6 |
| Zip-[1,2:Leu22Arg] + Zip[3] | 0.85 | ND | 1.6 |
| Zip-[1,2:Phe31Ser] + Zip[3] | 1.0 | ND | 1.3 |
|
|
|||
aThe OD600 at each time point for triplicate samples was plotted to calculate doubling time (according to log A2/A1 = k(T2 - T1), where A = OD600 and k = growth constant). The average of two separate experiments is given. Where indicated, trimethoprim was present at 1 µg/mL and IPTG at 1 mM.
bData from a single experiment with triplicate samples.
IPTG, isopropyl-beta-d-thiogalactopyranoside; ND, not done; NG, no growth; TRIM, trimethoprim.
The presence of IPTG unexpectedly prevented growth of E. coli transformed with wild-type mDHFR. Growth was partially restored by addition of the folate metabolism end products thymine, adenine, pantothenate, glycine, and methionine (data not shown). This effect suggests that induced overexpression of mDHFR was lethal to E. coli when grown in minimal medium as a result of depletion of the folate pool by binding to the enzyme. However, the amount of active reconstituted enzyme in cells is too low to allow toxicity.
The sequential increase in cell doubling times resulting from the mutations directed at the assembly interface (Ile114 to Val, Ala, or Gly) demonstrates that the observed cell survival under selective conditions is a result of the specific leucine zipper-assisted association of mDHFR F[1,2] with F[3]. Because Zip-[1,2] carries most residues involved in substrate binding and activity, it is possible that reconstituted DHFR activity could result from nonspecific interactions with Zip-[3]. However, the point mutants illustrate that cell survival is a result of correct mDHFR fragment reassembly rather than nonspecific interactions of Zip-[3] with Zip-[1,2].
Methotrexate-Resistant Mutants As a further control for understanding the molecular basis for the reassembly of mDHFR fragments into active enzyme, we mutated F[1,2] to incorporate, one at a time, each of five mutations that were previously shown to significantly increase Ki (MTX): Gly15Trp, Leu22Phe, Leu22Arg, Phe31Ser, and Phe34Ser (numbering according to the wild-type mDHFR sequence).27-32 These mutations occur at various positions relative to the active site and relative to F[3], and they have various effects on Km (DHF), Km (NADPH), and Vmax of the full-length mammalian enzymes in which they were studied.
Mutants Zip-[1,2:Leu22Phe], Zip-[1,2:Leu22Arg], and Zip-[1,2:Phe31Ser] allowed bacterial survival with high growth rates when cotransformed with Zip-[3] (see Table 1). This demonstrates that mutations in F[1,2], which essentially encompasses the substrate and inhibitor binding pocket and the active site residues, can be well tolerated and allow efficient fragment complementation. Mutants Zip-[1,2:Gly15Trp] and Zip-[1,2:Phe34Ser] did not allow bacterial survival. The position of Gly15 readily explains this observation, because it occurs at the interface between F[1,2] and F[3], and its mutation appears to destabilize the fragment assembly, reinforcing the importance of the interface interactions in fragment complementation. The reason for lack of growth with mutant Phe34Ser is not so obvious; the 24-fold increase in Km (DHF) observed in the wild-type hDHFR31 may be responsible for this.
Kinetic Parameters of the Reconstituted Enzyme In vitro activity was assayed after native purification of the reconstituted enzyme: Zip-[1,2:Phe31Ser] plus Zip-[3]. Figure 3 illustrates that the rate of turnover, measured by fluorescence emission of the product THF, is inhibited by increasing concentrations of MTX. The Ki (MTX) for the reconstituted enzyme was determined to be 0.7 nM (average of two independent determinations) at 37°C, pH 7.5, using 30 µM DHF and 25 µM NADPH. This value is similar to the Ki (MTX) of 4.4 nM for mDHFR (Phe31Ser), determined at 30°C, pH 7.9, using 50 µM DHF and 100 µM NADPH.30 Inhibition by MTX retains its competitive character relative to dihydrofolate (data not shown). These results suggest that the kinetic behavior of the reconstituted enzyme closely resembles that of the full-length mDHFR (Phe31Ser).
FIGURE 3. Inhibition of the reconstituted murine dihydrofolate reductase (mDHFR) (Phe31Ser) activity by methotrexate (MTX). Semi-logarithmic plot illustrates the inhibitory effect of increasing MTX concentrations on tetrahydrofolate production (measured as fluorescence emission) by the reconstituted enzyme. The blank (no enzyme) was subtracted at each data point.
We demonstrated the application of the USPS10 for the detection of protein-protein interactions in the mammalian COS-7 cell line and illustrated its use in the study of the equilibrium and kinetic aspects of protein-protein interactions in vivo. The strategy is based on cleavage by cytosolic ubiquitinases of a reporter protein (ie, HA-tagged mDHFR) from the C terminus of ubiquitin, to which it is fused. The proteases release the reporter protein only if the structure of ubiquitin is intact. Fusion of a reporter protein-ubiquitin C-terminal fragment can also be cleaved by ubiquitinases, but only if coexpressed with an N-terminal fragment of ubiquitin that is complementary to the C-terminal fragment. As in the DHFR PCA, the two ubiquitin fragments do not recombine efficiently unless oligomerization domains are introduced at the ends of the complementary fragments; the oligomerization domains act as a template for the reassembly of ubiquitin.
We demonstrated the oligomerization-assisted assembly of ubiquitin fragments with GCN4 leucine zipper-forming sequences in COS-7 cells that were transiently transfected with leucine zipper-ubiquitin fragment gene fusions. Constructs consisted of combinations of GCN4 leucine zipper (Zip), ubiquitin N-terminal (Nub) or C-terminal (Cub) fragments, and mDHFR-HA (dha) fusions. The Zip-Cub-dha and Nub-Zip fusions were coexpressed in COS-7 cells, in which oligomerization was detected by measuring the release of mDHFR-HA from the reassembled ubiquitin.
To study the equilibrium reconstitution of ubiquitin structure, cell lysates were analyzed by SDS-PAGE, followed by protein transfer and immunoblotting with anti-HA antibodies (Fig. 4A). The positive control Ub-dha was totally cleaved by ubiquitinases. When expressed by itself, Zip-Cub-dha yielded minor, nonspecific cleavage that was considerably reduced by the replacement of the original linker joining the zipper and Cub10 by a flexible linker coding for (GlyGlyGlyGlySer)4. However, coexpression of Zip-Cub-dha and Nub-Zip resulted in the complete cleavage of Zip-Cub-dha (ie, release of dha).
FIGURE 4. Reconstitution of ubiquitin from its coexpressed fragments. Cub-dha consisted of the full-length ubiquitin protein fused to murine dihydrofolate reductase (mDHFR) with a hemagglutinin (HA) epitope tag (dha). The Nub-Zip fusions and Zip-Cub-dha fusions consisted of the N- (codons 1 to 37) and C-terminal (codons 35 to 76) portion of ubiquitin, respectively, fused to the residues 235 to 281 of the GCN4 leucine zipper. (A) Western blot of the equilibrium reconstitution of ubiquitin. Lane 1 is full-length Ub-dha, lane 2 is Zip-Cub-dha alone, lanes 3 through 6 are coexpressions of Zip-Cub-dha with, respectively, Nub-Zip, Nub(Ile13Val)-Zip, Nub(Ile13Ala)-Zip, and Nub(Ile13Gly)-Zip. Lane 7 is a mock transfection, and M represents molecular weight markers. (B) Autoradiograph of [35S]-pulse labeled immunoprecipitation: "-" is Zip-Cub-dha alone, and I, V, A, and G are coexpressions of Zip-Cub-dha with, respectively, Nub-Zip, Nub(Ile13Val)-Zip, Nub(Ile13Ala)-Zip, and Nub(Ile13Gly)-Zip.
We replaced the leucine zipper-forming sequences with other interacting protein domains, essentially eliminating the problem of the nonspecific cleavage, probably by increasing the stability of the fusion proteins to nonspecific proteolytic cleavage or competition for internal initiation at the DHFR start codon (which is intact in our fusions). These results will be reported elsewhere (I. Remy, N. Johnsson, and S.W. Michnick, in preparation). Point mutations of increasing severity (Ile13 to Val, Ala, and Gly) were introduced at the interface between the two ubiquitin fragments (ie, in the hydrophobic core of ubiquitin) with the goal of destabilizing reassembly. The cleavage was incomplete for the mutants Nub(Ile13Ala)-Zip and Nub(Ile13Gly)-Zip (see Fig. 4A), demonstrating that the observed reconstitution of ubiquitinase activity was caused by specific refolding of ubiquitin from its complementary fragments. The result for the mutant Ile13Val was comparable to the wild-type Nub.
We also demonstrated that, as in yeast, kinetics of leucine zipper formation can be detected in mammalian cells using USPS. COS-7 cells expressing Zip-Cub-dha by itself or coexpressed with Nub-Zip, Nub(Ile13Val)-Zip, Nub(Ile13Ala)-Zip, or Nub(Ile13Gly)-Zip were incubated for 30 minutes with [35S]methionine, followed by a chase for 0, 10, and 30 minutes, extraction of proteins, immunoprecipitation with anti-HA antibody, and SDS-PAGE (see Fig. 4B). Coexpression of Zip-Cub-dha and Nub-Zip resulted in the nearly complete release of dha by the end of a 30-minute pulse (time 0). The cleavage was slower when Zip-Cub-dha was coexpressed with Nub(Ile13Val)-Zip, Nub(Ile13Ala)-Zip, or Nub(Ile13Gly)-Zip. For the Ala and Gly mutants, the cleavage was still incomplete after a 30-minute chase. The rate of release of dha decreased with increasing severity of mutations in coexpressed Nub fragments, demonstrating that the rate of ubiquitin reconstitution depends on the stability of the refolded ubiquitin. This approach has been suggested as a way to determine relative rates of assembly of different proteins.10
We developed and implemented a bacterial PCA based on mDHFR and illustrated the application of a ubiquitin-fragment complementation assay in mammalian cells, in which interacting leucine zippers direct the reconstitution of protein fragments in vivo with concurrent detection of reconstitution of the fragments. In the first case, activity was detected by an E. coli survival assay. In the second case, the mammalian PCA was used to reveal the equilibrium and kinetic aspects of leucine zipper formation.
Development and demonstration of the two PCAs illustrate how this approach can be used to design at a molecular and atomic level of detail a general strategy to detect protein-protein interactions and how the same principles could be applied to development of other assay strategies. The DHFR PCA is a complete system in which no additional endogenous factors are necessary and the results of complementation may be observed directly, with no further manipulation. The USPS system can be used to monitor protein-protein interactions as a function of time at their natural site of interaction in a mammalian cell line, but it requires enzymes constitutively expressed in the host cell. However, this is not a serious limitation. The ubiquitinases probably are expressed in the cytosol and nucleus, and two significant cellular compartments are therefore accessible.33,34 Subcompartmentalization of the ubiquitinases means that USPS can be used in the context of compartmentalization kinetics studies, as has been illustrated.35
Having coupled enzyme activity in a PCA could be advantageous in some cases. In a given PCA, the complementing fragments may not reassemble to take on the native structure of the protein, but rather devise some stable intermediate form, such as a "molten globule." In the case of an enzymatic PCA, the native activity of the reconstituted enzyme may not be attained, precluding use of its activity as a detection tool. However, in a PCA in which the reassembled protein is recognized by an independently folding enzyme, this accessory enzyme may act on the reassembled protein even if it is not in its fully native state.
The DHFR PCA E. coli cell survival assay described is perhaps the simplest direct measurement method. Several aspects of the DHFR PCA distinguish it from all other available methods for studying protein-protein assembly in vivo. With knowledge about protein stability, folding, enzyme kinetics, and regulation, it should be possible to design complementary fragments of DHFR that would allow controlling the stringency of the assay, obtaining estimates of the kinetics and equilibrium constants for association of two proteins, and detecting induced and uninduced protein-protein interactions. We used point mutations of the wild-type Ile114 to Val, Ala, and Gly to demonstrate that the observed effects on E. coli survival and cell growth rates specifically result from reconstitution of DHFR activity. For determining estimates of equilibrium and kinetic parameters for a specific protein-protein interaction, a series of experiments could be performed in which two proteins that are known to interact are fused to the N- or C-terminal domains of the DHFR fragments in the wild-type or destabilizing mutant forms. It would then be possible to perform the DHFR PCA and compare rates of cell growth or MTX binding with those observed for model protein-protein interactions to obtain an estimate of the strength of the interaction. To illustrate this approach, we are making point mutations in the GCN4 leucine zipper for which direct equilibrium and kinetic parameters are known and correlating these known values with parameters derived from the PCA (J.N. Pelletier, K. Arndt, A. Pluckthun, and S.W. Michnick, in preparation).
Another advantage of designing stability mutants of DHFR is to control the stringency of the assay, which is particularly important in cDNA library screening for protein-protein interactions. By assembling the library as fusions to different stability mutants, it may be possible to distinguish weak, nonspecific interactions from high-affinity, specific interactions.
The DHFR PCA should not be limited by the context in which it could be used, as was illustrated for USPS. mDHFR has been expressed in prokaryotic and eukaryotic hosts, and it could be targeted to specific compartments in the cell by the addition of signaling peptide sequences. Induced and constitutive protein-protein interactions could be segregated by the DHFR PCA when an interaction is triggered by a biochemical event. For example, in the case of signal transduction, it is possible to transfect a cell line that responds to a specific hormone by interaction with a cell-surface receptor, with the DHFR PCA components fused to proteins thought to interact after the hormone-receptor interaction. The cells could then be treated with the hormone and tested for reconstitution of DHFR activity.
Besides the simple example of peptide oligomerization studied here, we are investigating applications of the USPS, DHFR, and other PCAs for screening of cDNA libraries for protein-protein interactions. The capabilities of controlling the stringency of the assays, of detecting interactions in appropriate contexts, and of differentiating induced from constitutive interactions make the PCA approach a promising strategy for determining possible functions of gene products based on their interactions with proteins of known function. The fact that such studies could be done in appropriate cells (ie, those in which a particular biochemical pathway is studied) means that PCAs could be used in conjunction with appropriate functional controls to ensure that interactions of known proteins with products of a cDNA library may be more specific and biologically relevant.
This work was supported by the MRC of Canada (grant DGN 059 to S.W. Michnick) and The Burroughs-Wellcome Fund (S.W. Michnick). J.N. Pelletier is a recipient of a Fellowship from Les Fonds pour la recherche en santè du Quèbec. S.W. Michnick is an Awardee of a Burroughs-Wellcome Fund New Investigator Award in the Basic Pharmacological Sciences. We are grateful to Nils Johnsson and Monique Davies for helpful discussions and for carefully reading the manuscript and to Nils Johnsson for providing us with the original plasmids for the USPS and Monique Davies for providing us with the pMT3 plasmids.
1. Reed LJ. Multienzyme complexes. Acc Chem Res 1974;7:40-46.
2. Lander ES. The new genomics--global views of biology. Science 1996;274, 536-539.
3. Evangelista C, Lockshon D, Fields S. The yeast two-hybrid system--prospects for protein linkage maps. Trends Cell Biol 1996;6:196-199.
4. Fromont-Racine M, Rain JC, Legrain P. Toward a functional analysis of the yeast genome through exhaustive two-hybrid screens [see comments]. Nat Genet 1997;16:277-282.
5. Guarente L. Strategies for the identification of interacting proteins. Proc Natl Acad Sci USA 1993;90:1639-1641.
6. Adams SR, Harootunian AT, Buechler YJ, Taylor SS, Tsien RY. Fluorescence ratio imaging of cyclic AMP in single cells. Nature 1991;349:694-697.
7. Chien CT, Bartel PL, Sternglanz R, Fields S. The two-hybrid system: a method to identify and clone genes for proteins that interact with a protein of interest. Proc Natl Acad Sci USA 1991;88:9578-9582.
8. Fields S, Song O. A novel genetic system to detect protein-protein interactions. Nature 1989;340:245-246.
9. Gyuris J, Golemis E, Chertkov H, Brent R. Cdi1, a human G1 and S phase protein phosphatase that associates with Cdk2. Cell 1993;75:791-803.
10. Johnsson N, Varshavsky A. Split ubiquitin as a sensor of protein interactions in vivo. Proc Natl Acad Sci USA 1994;91:10340-10344.
11. Pelletier JN, Michnick SW. A protein complementation assay for detection of protein-protein interactions in vivo. Protein Eng 1997;10:89.
12. Kaufman RJ, Davies MV, Pathak VK, Hershey JW. The phosphorylation state of eucaryotic initiation factor 2 alters translational efficiency of specific mRNAs. Mol Cell Biol 1989;9:946-958.
13. Kunkel TA, Roberts JD, Zakour RA. Rapid and efficient site-specific mutagenesis without phenotypic selection. Methods Enzymol 1987;154:367-382.
14. Stammers DK, Champness JN, Beddell CR, Dann JG, et al. The structure of mouse L1210 dihydrofolate reductase-drug complexes and the construction of a model of human enzyme. FEBS Lett 1987;218:178-184.
15. Blakley RL. Folates and pterins: chemistry and biochemistry of folates. In Blakley R, Cenkovic S (eds): Folates and Pterins: Chemistry and Biochemistry of Folates, vol 1. New York: John Wiley & Sons, 1984:191-253.
16. Appleman JR, Prendergast N, Delcamp TJ, Freisheim JH, Blakley RL. Kinetics of the formation and isomerization of methotrexate complexes of recombinant human dihydrofolate reductase. J Biol Chem 1988;263:10304-10313.
17. Oefner C, D'Arcy A, Winkler FK. Crystal structure of human dihydrofolate reductase complexed with folate. Eur J Biochem 1988;174:377-385.
18. Gegg CV, Bowers KE, Matthews CR. A general approach for the design and isolation of protein fragments: the molecular dissection of dihydrofolate reductase. In Marshak DR (ed): Techniques in Protein Chemistry, vol VII. New York: Academic Press, 1996:439-448.
19. Bystroff C, Kraut J. Crystal structure of unliganded Escherichia coli dihydrofolate reductase. Ligand-induced conformational changes and cooperativity in binding. Biochemistry 1991;30:2227-2239.
20. Perry KM, Onuffer JJ, Gittelman MS, Barmat L, Matthews CR. Long-range electrostatic interactions can influence the folding, stability, and cooperativity of dihydrofolate reductase. Biochemistry 1989;28:7961-7968.
21. Bullerjahn AM, Freisheim JH. Site-directed deletion mutants of a carboxyl-terminal region of human dihydrofolate reductase. J Biol Chem 1882;267:864-870.
22. Buchwalder A, Szadkowski H, Kirschner K. A fully active variant of dihydrofolate reductase with a circularly permuted sequence. Biochemistry 1992;31:1621-1630.
23. Chen X, Rambo R, Matthews CR. Amino acid replacements can selectively affect the interaction energy of autonomous folding units in the alpha subunit of tryptophan synthase. Biochemistry 1992;31:2219-2223.
24. Prevost M, Wodak SJ, Tidor B, Karplus M. Contribution of the hydrophobic effect to protein stability: analysis based on simulations of the Ile-96-Ala mutation in barnase. Proc Natl Acad Sci USA 1991;88:10880-10884.
25. Kellis J Jr, Nyberg K, Sali D, Fersht AR. Contribution of hydrophobic interactions to protein stability. Nature 1988;333:784-786.
26. Kellis J Jr, Nyberg K, Fersht AR. Energetics of complementary side-chain packing in a protein hydrophobic core. Biochemistry 1989;28:4914-4922.
27. Dicker AP, Waltham MC, Volkenandt M, et al. Methotrexate resistance in an in vivo mouse tumor due to a non-active-site dihydrofolate reductase mutation. Proc Natl Acad Sci USA 1993;90:11797-11801.
28. Hussain A, Lewis D, Yu M, Melera PW. Construction of a dominant selectable marker using a novel dihydrofolate reductase. Gene 1992;112:179-188.
29. Simonsen CC, Levinson AD. Isolation and expression of an altered mouse dihydrofolate reductase cDNA. Proc Natl Acad Sci USA 1983;80:2495-2499.
30. Thillet J, Absil J, Stone SR, Pictet R. Site-directed mutagenesis of mouse dihydrofolate reductase. Mutants with increased resistance to methotrexate and trimethoprim. J Biol Chem 1988;263:12500-12508.
31. Schweitzer BI, Srimatkandada S, Gristman H, Sheridan R, Venkataraghavan R, Bertino JR.Probing the role of two hydrophobic active site residues in the human dihydrofolate reductase by site-directed mutagenesis. J Biol Chem 1989;264:20786-20795.
32. Banerjee D, Schweitzer BI, Volkenandt M, et al. Transfection with a cDNA encoding a Ser31 or Ser34 mutant human dihydrofolate reductase into Chinese hamster ovary and mouse marrow progenitor cells confers methotrexate resistance. Gene 1994;139:269-274.
33. Jonnalagadda S, Butt TR, Monia BP, et al. Multiple (alpha-NH-ubiquitin)protein endoproteases in cells. J Biol Chem 1989;264:10637-10642.
34. Wilkinson KD, Lee KM, Deshpande S, Duerksen-Hughes P, Boss JM, Pohl J. The neuron-specific protein PGP 9.5 is a ubiquitin carboxyl-terminal hydrolase. Science 1989;246:670-673.
35. Johnsson N, Varshavsky A. Ubiquitin-assisted dissection of protein transport across membranes. EMBO J 1994;13:2686-2698.