Translational repression by bacteriophage MS2 coat protein expressed from a plasmid. A system for genetic analysis of a protein-RNA interaction.

The coat protein of bacteriophage MS2 is a translational repressor. It inhibits the synthesis of the viral replicase by binding a specific RNA structure that contains the replicase translation initiation region. In order to begin a genetic dissection of the repressor activity of coat protein, a two-plasmid system has been constructed that expresses coat protein and a replicase-beta-galactosidase fusion protein from different, compatible plasmids containing different antibiotic-resistant determinants. The coat protein expressed from the first plasmid (pCT1) represses synthesis of a replicase-beta-galactosidase fusion protein encoded on the other plasmid (pRZ5). Mutations in the translational operator or in coat protein result in constitutive synthesis of the enzyme. This permits the straightforward isolation of mutations in the coat sequence that affect repressor function. Because of the potential importance of cysteine residues for RNA binding, mutations were constructed that substitute serines for the cysteine residues normally present at positions 46 and 101. Both of these mutations result in translational repressor defects. Chromatographic and electron microscopic analyses indicate that the plasmid-encoded wild-type coat protein forms capsids in vivo. The ability of the mutants to adopt and/or maintain the appropriate conformation was assayed by comparing them to the wild-type protein for their ability to form capsids. Both mutants exhibited evidence of improper folding and/or instability as indicated by their aberrant elution behavior on a column of Sepharose CL-4B. Methods were developed for the rapid purification of plasmid-encoded coat protein, facilitating future biochemical analyses of mutant coat proteins.

At late stages of infection of Escherichia coli by bacteriophage MS2, the viral coat protein binds a stem-loop structure in the viral RNA which contains the ribosome binding site of the replicase gene, thus repressing replicase synthesis (1). The translational operator consists of a 21-nucleotide stem-loop whose specific three-dimensional structure forms the coat protein-binding site (2,3). This small piece of RNA may also serve as the signal for encapsidation of the RNA genome. Although the basic rules for the structure of the MS2 translational operator have been investigated, we know very little about which aspects of coat protein structure account for its * This work was supported by a grant from the National Science Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
repressor activity. Apart from the putative interaction between a uridine residue of the operator and a cysteine residue of coat protein (4)  Media and p-Galactosidase Assays-Media for the growth of bacteria and methods for the assay of P-galactosidase were as described by Miller (13). Indicator plates contained LB medium and 5-bromo-4-chloro-3-indolyl-@-n-galactoside (X-gal)' at a concentration of 40 pg/ml. Selection of the repressor-positive phenotype was performed on minimal medium (M9) containing 0.2% glycerol and 0.03% phenyl-/3-D-galactoside (P-gal). In a galE-host (such as CSH41) production of P-galactosidase is lethal when the cells are grown in the presence of P-gal (explained in Ref. 13). Note that induction of the lac promoter that drives expression of both the coat and the replicase-1acZ genes is unnecessary, since the chromosomal lac genes (including lac1) are deleted in CSH41. Solution assay of @-galactosidase activity was performed using onitrophenyl-P-n-galactoside as described by Miller (13 Images were recorded at 75 kV in a transmission electron microscope (Hitachi H600).

RESULTS
Construction of Plasmid Recombinants-The structures of the plasmid recombinants used in this study are shown in Fig.  1. The rationale of their design was to create a situation in which the synthesis of a replicase-P-galactosidase fusion protein encoded by one plasmid is under translational control of coat protein encoded by a second plasmid. This permits one to take advantage of the genetic selection and screening schemes which were devised during the study of the lac operon (13). In order to stably maintain both types of recombinant in the same bacterial strain they were constructed from plasmids which belong to different incompatibility groups and confer resistance to different antibiotics.
Thus, pCT1 and its mutant derivatives pCTC46S and pCTClOlS were constructed from pUC19 (9), which contains a ColEl-type origin of replication and confers resistance to ampicillin. These plasmids produce the MS2 coat protein under control of the lac promoter.
The two mutants (C46S and ClOlS) replace each of the two cysteine residues with serines. They were constructed because of the assertion that a cysteine is necessary for RNA binding activity (4,16).
The pRZ5 and pRZ6 constructs produce a replicase+ galactosidase fusion protein also under control of the lac promoter ( Fig. 1). They contain a P15A replication origin (from pACYCl84, Ref. 12) and render cells resistant to chloramphenicol.
Plasmid pRZ5 contains a synthetic translational operator corresponding to the sequence of the wild-type operator of the MS2 replicase gene (l-3,8).
On the other hand, pRZ6 was constructed with a mutant sequence designed to resemble the operator of the related RNA phage, GA (17). The sequence differences are shown in Fig. 1. Both operators consist only of the 21-nucleotide sequence previously determined to contain all of the binding activity for coat protein in vitro (3). The amino acid sequences of the MS2 and GA coat proteins are about 60% identical. They bind operators that are homologous but which differ in the critical -5 position as well as at position -6. This should prevent the MS2 coat protein from efficiently binding the GA operator based on its inability to tightly bind similar MS2 operator variants (2, 3).

Production of /3-Galactosidase Activity in Strains Bearing the Various Recombinant
Plasmids-Escherichia coli strain CSH41 was transformed with pUCter3, pCT1, pCTC46S, or pCTClOlS and pRZ5 or pRZ6. The resulting eight strains are listed in Table I. The ability of each of them to produce /3galactosidase was determined: 1) by solution assay using the o-nitrophenyl$-D-galactoside substrate as described by Miller (13); 2) by the intensity of the blue color of colonies on medium containing X-gal; and 3) by growth on minimal medium containing P-gal. Table I shows the results of these experiments.
By these criteria this two-plasmid system faithfully represents the translational repression observed in MS2infected cells. The presence of pCT1, which produces wildtype coat protein, results in about a 30-50-fold repression of P-galactosidase synthesis from pRZ5. Failure to produce coat protein (e.g. pUCter3), or the production of a defective coat protein (e.g. pCTC46S) results in loss of repression of pgalactosidase synthesis. Mutation of the translational operator (as in pRZ6) also results in a constitutive phenotype. For reasons that are not understood, the mutations in pRZ6 also result in a 2-fold increase in expression of @-galactosidase in the absence of coat protein (pUCter3 samples in Table I).
The operator variant present in pRZ6 permits only partial repression by wild-type coat protein.
Purification of Plasmid-produced Coat Protein-The ability of coat protein to be interconverted between forms of 2.5 X lo6 daltons (capsids) and forms of dramatically lower molecular weight (monomer molecular weight is 1.4 X 104) provides a basis for the purification of plasmid-produced coat proteins. Sepharose CL-4B has previously been used in the purification of intact MS2 virus (18). The exclusion limit of this gelfiltration matrix is appropriate for the separation of the viral particle from species of higher and lower molecular weights. As a first step in the purification of coat protein from CSH4l(pCTl), an extract of the cells was prepared as described under "Materials and Methods" and applied to a column of Sepharose CL-4B. Fractions were assayed for the presence of coat protein by SDS-polyacrylamide gel electrophoresis and Western blotting of fraction aliquots. The elution profile is shown in Fig. 2B. Electrophoretic analysis (Fig.  3) of the peak fractions from the Sepharose column indicates a substantial enrichment for coat protein. A second passage through the Sepharose column of pooled fractions containing coat protein resulted in an additional enrichment. Material from the peak fractions was pooled, concentrated by ammonium sulfate precipitation, dialyzed against Sepharose column buffer, and acetic acid was added to a final concentration of 50%. This acid-denatured material was applied to a 1.5 X 45cm column of Sephadex G-75 in 10 mM acetic acid, 50 mM NaCl. Fraction aliquots were subjected to SDS-polyacrylamide gel electrophoresis and stained with Coomassie Blue. Contaminants were of higher molecular weights and eluted from the column before coat protein.
The purity of coat protein thus isolated was determined by gel electrophoresis. Fig. 3 indicates the purity of coat protein at each stage of the purification procedure. 3, pooled coat protein-containing fractions after a single passage over Sepharose CL-4B; 4, pooled coat protein-containing fractions after a second passage through the Sepharose column; 5, purified coat protein after chromatography on Sephadex G-75; 6, purified MS2 standard. FIG. 2. Elution behavior of purified MS2 virus and coat proteins produced by the plasmids used in this study. In each case the amount of coat protein (in arbitrary units) was determined by Western blot analysis of fraction aliquots followed by scanning densitometry. Each profile of plasmid-produced coat protein represents an extract from a 250-ml bacterial culture containing, in the cases of pCT1 and pCTClOlS, about 0.5 mg of coat protein, and, in the case of pCTC46S, about 0.1 mg of coat. Lysozyme elutes at fraction 37. Analysis of the Plasmid-produced Coat Proteins-It was possible that the repressor-defective phenotype of C46S and the partially defective phenotype of ClOlS were the results of a failure of the proteins to adopt or maintain the native conformation. It has been previously demonstrated that mutations that prevent proper folding, or that result in decreased thermodynamic stability, often also result in rapid intracellular degradation and reduced steady state levels of the specific protein (for example, see Ref. 19). To determine the amounts of coat protein produced by each of the plasmid constructs, equal quantities of cell extract from each strain were applied to a 17.5% polyacrylamide gel containing SDS and fractionated by electrophoresis. After transfer to a nitrocellulose membrane, the coat proteins were visualized using rabbit anti-MS2 serum and ""I-protein A. Fig. 4 shows the results of this experiment. Strains containing pCT1 produced coat protein at levels corresponding to about 2 mg/liter of culture. The C46S mutant directs the synthesis of coat protein at levels that are reduced about B-fold. By analogy to Pakula et al. (19) C46S may be defective for folding and/or stability.
To assess more directly the extent to which the mutant proteins adopt the native conformation, we subjected the mutant coat proteins to chromatography on a column of Sepharose CL-4B. We make the assumption that only properly folded coat proteins are capable of forming virus-like particles. This seems reasonable, since the formation of a capsid requires multiple intersubunit contacts. The position of elution of coat proteins in crude cell extracts was determined by SDS-polyacrylamide gel electrophoresis and Western blotting of column fractions. The coat protein produced by pCT1 coelutes with authentic, purified virus (Fig. 2 23) suggesting that it is capable of assembly into capsids. The product of pCTC46S, however, elutes at a position corresponding with a much smaller molecular weight (Fig. 2C), indicating a failure to form capsids.
The results of chromatography with the product of pCTClOlS also show evidence of an assembly defect (Fig.  20). Although a fair fraction (about 20%) of the coat protein elutes at the position characteristic of viral particles, some of it elutes at a position corresponding to a higher molecular weight. A third peak coelutes with a lysozyme standard and thus seems to correspond either to the coat monomer or to low order oligomers. This behavior might be explained by a failure of some of the ClOlS protein to properly fold, or by a reduced thermodynamic stability. We do not yet know which, if any, of these defects accounts for the partial loss of repressor activity observed with this mutant in uiuo.
The elution behavior of the plasmid-encoded coat proteins suggested that they were capable of forming virus-like particles. In order to confirm this assertion we subjected peak fractions from the Sepharose CL-4B chromatography of the product of pCT1 to electron microscopic analysis. We also imaged purified MS2 viral particles for comparison. Fig. 5 shows that the virus-like particles produced by pCT1 are essentially indistinguishable from intact MS2, suggesting that coelution of plasmid-encoded coat protein with MS2 indicates formation of a typical capsid-like structure. DISCUSSION These results show that the two-plasmid expression system mimics the translational repression observed in MSB-infected cells. Repression of translation of the hybrid replicase-/?galactosidase sequence depends on the presence of coat protein. Repression can be reduced or eliminated either by mutation of the coat protein gene or by mutation of the translational operator. This provides a convenient means for the isolation of a variety of coat mutants altered in their ability to bind wild-type or mutant operators.
Uhlenbeck and his colleagues (4) have presented evidence suggesting that a coat protein sulfhydryl group is involved in the formation of a transient covalent bond with the pyrimidine ring at position -5 in the translational operator. The results of these experiments do not permit us to conclude which (if either) of the 2 cysteine residues participates in this reaction. Replacement of either cysteine residue with serine results in reduced repressor activity. Since substitution of Cys-46 by serine results in complete loss of repression, one might be inclined to propose (as others have, Ref. 16) that Cys-46 is the residue involved in formation of the postulated covalent adduct. The assignment of either of the 2 cysteines as RNA contact sites, however, is compromised by the fact that both C46S and ClOlS result in apparent stability or assembly defects. One cannot tell to what extent this accounts for lost repressor activity, as opposed to direct disruption of a protein-RNA contact. Instability or improper folding could result in a gross deformity in coat protein structure. Since the repressor is apparently a dimer (20), any disruption of the ability to dimerize would also indirectly affect binding activity. Since neither of the cysteines is involved in disulfide bonding (16), it was not obvious that their substitution with serine would be so disruptive.
The potential function of the cysteine residues as sites of contact with RNA is further brought into question by the results of codon-directed mutagenesis experiments (21) that show that although some substitutions are as disruptive as serine (specifically, arginine and tryptophan), some others (phenylalanine, leucine, methionine, tyrosine, alanine, and valine) can be substituted for Cys-46 with little or no loss of repressor function. Cys-101, however, is more sensitive to substitution. Of 18 substitutions tested so far, only cysteine and arginine are able to support normal repression. Thus, while it must be regarded as unlikely that Cys-46 is an RNA contact residue, the higher sensitivity of Cys-101 to substitution suggests that it could represent a site of interaction. The replacement of Cys-101 with arginine could establish new interactions in place of those that are lost by the removal of cysteine. More experiments are required to establish whether Cys-101 is a contact site and whether a transient covalent bond contributes to RNA binding affinity in this case.
These results also indicate that mutations of the translational operator can yield a constitutive phenotype. Plasmid pRZ6 contains an operator sequence modified in two positions to resemble the operator of the group II RNA phage GA (Ref. 17,and Fig. 1). The rules of MS2 translational operator structure predict that these changes should result in a several hundred-fold decrease in K, in vitro (3). However, only a partial loss of repression is observed when the pCTl/pRZG strain is compared to pUCter3/pRZG. Perhaps coat protein exhibits a higher affinity for the mutant operator in vivo than that predicted by the in vitro binding experiments, or maybe the repressor is present in sufficient excess compared to its binding site on the hybrid replicase-lacz mRNA to result in the unexpectedly high level of repression. Since the plasmidproduced coat protein forms capsids in Go, it is also possible that the operator containing RNAs are packaged into viruslike particles. The formation of such complexes is probably nearly irreversible and could result in a higher apparent affinity in vivo than is observed in vitro, where conditions favoring capsid formation are avoided.
Note also that even in the absence of repressor (pUCter3), the operator mutations in pRZ6 give rise to higher levels of p-galactosidase activity than the wild-type operator sequence in pRZ5. The mutations may improve the efficiency of the replicase ribosome binding site, perhaps by changes in the Shine-Dalgarno sequence, or because of changes in RNA tertiary interactions that make the site more accessible. Caution must be exercised when interpreting the results of such experiments, since mutations which increase the inherent translational efficiency of the replicase-P-galactosidase mRNA might mimic the operator-constitutive phenotype without actually affecting the strength of the coat proteinoperator interaction. In this case, however, comparison of the ratios of derepressed to repressed enzyme levels reveals a clear loss of repression in pRZ6 compared to pRZ5 (50-fold versus 7-fold). The level of derepression is easily distinguished from wild-type by colony color on X-gal plates, making it possible to isolate mutations in the coat protein gene that suppress defects in the operator sequence. The ability to quickly isolate milligram quantities of plasmid-produced coat protein will facilitate the biochemical analysis of coat protein variants produced by genetic means.