Tryptophan super-repressors with alanine 77 changes.

The binding of L-tryptophan to Escherichia coli tryptophan aporepressor enables the holorepressor complex to bind operator DNA tightly. The side chain of residue alanine 77 is located in one of the most flexible regions of Trp repressor, between residues critical for binding DNA. Codon-directed mutagenesis was used to make genes encoding mutant Trp repressors with each of the 19 naturally occurring amino acid changes of Ala77. The 19 mutant proteins are made at the same steady-state levels as wild type. Sensitive challenge phage assays show that 7 of the 19 mutant proteins (Cys, Ser, Val, Leu, Thr, Ile, and Lys) are more active than wild-type protein when tryptophan is limiting in vivo. Among these 7 mutant super-aporepressors, proteins with Cys and Ser changes also are super-holorepressors, because they repress better than wild-type holorepressor when tryptophan is in excess. These results and others suggest that super-aporepressors associate more poorly than wild-type aporepressor with nonspecific DNA. Consistent with this idea, these 7 changes are predicted to disrupt the tertiary structure of aporepressor, but have more limited effects on the structure of holorepressor.

Trp aporepressor binds 2 molecules of the corepressor ligand, L-tryptophan, to form a holorepressor complex with increased affinity for specific DNA sites on the Escherichia coli genome. Trp repressor is a symmetric dimer formed from two identical monomers 107 amino acid residues in length (Gunsalus and Yanofsky, 1980). Comparisons of the crystal structures of aporepressor and holorepressor (Schevitz et al., 1985;Zhang et al., 1987;Lawson et al., 1988;Otwinowski et al., 1988) suggest that activation of the TrpR dimer involves the repositioning of flexible DNA "reading heads" (a-helices D and E) with respect to its rigid hydrophobic core (a-helices A, B, C, and F). The flexible reading heads in aporepre!sor are about 25 A apart, and become separated by 30-34 A in holorepressor, enabling them to contact successive major grooves of operator DNA ( Fig. 1; Schevitz et al., 1985;Luisi and Sigler, 1990;Arrowsmith et al., 1991). The turn between a-helices D and E ( L e~~~-G l y~~-A l a~~-G l y~' ) shows the greatest variation in coordinates for defined residues in the orthorhombic structures of aporepressor and holorepressor (Lawson et al., 1988).
* This work was supported by National Institutes of Health Grants GM34150 (to P. Y.) and GM 12629 (to D. N. A.). 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.
$ Current address: Dept. of Biochemistry and Molecular Biology, University of California, Berkeley, California 94720.
To whom correspondence and reprint requests should be addressed.
Several mutations that change amino acids in the flexible D/E turn have been isolated. Whereas changes of residue Gly7' to Ser and Asp result in a loss of repressor function, the change Ala77 + Val improves function. Expression of the mutant Val77 protein in the absence of exogenous tryptophan inhibits the growth of a host with a mutation in the trpA gene that decreases but does not abolish tryptophan synthase activity. In contrast, expression of wild-type repressor in an otherwise isogenic host does not inhibit growth (Kelley and Yanofsky, 1983). Also, the Ala77 + Val change has been isolated as a second-site revertant of a change (GlyS5 + Arg) that decreases TrpR activity . However, purified Val77 protein has the same operator-binding activity as wild-type TrpR (Ala77), both in the presence and absence of tryptophan. Even more puzzling, the Val77 protein aporepressor requires as much tryptophan to be activated for operator binding as wild-type repressor in vitro (Hurlburt and Yanofsky, 1990).
To understand why the mutant Val77 protein appears to be more active than wild type in vivo, but not in vitro, we have examined its phenotype in vivo, using challenge phage assays (Benson et al., 1986;Bass et ab, 1987;Bass et al., 1988;Arvidson et al., 1991a). In addition, we have asked whether other changes of this residue may help elucidate the role residue Ala77 plays in TrpR function. Here we describe the construction and phenotypic characterization of plasmids producing mutant aporepressors with each of the possible amino acid changes at residue Ala77. Although 12 of these 19 changes decrease TrpR activity, 7 result in mutant proteins with super-aporepressor phenotypes. When tryptophan is limiting, lower levels of these proteins are required for repression. Two changes, to cysteine and serine, also have super-holorepressor phenotypes. Lower levels of these proteins are required for repression when tryptophan is present in excess. Phenotypic differences imply that these 7 changes affect different combinations of equilibria between free aporepressor, free repressor, and their complexes with both specific (operator) and nonspecific DNA.

EXPERIMENTAL PROCEDURES
Bacteria, Phage, and Plasmids-Bacterial strains are derivatives of E. coli K12 and Salmonella typhirnurium LT2. E. coli strain CJ236 (dut-I, ung-I, thi-I, relA-I (pCJ105); Bio-Rad Laboratories) was used as the host for preparing uracil-containing single-stranded plasmid templates for site-directed mutageneses. Plasmid pCJ105 is a derivative of the F episome that permits phage M13 adsorption and confers resistance to chloramphenicol. E. coli XSO/F'lacP (recA-argE-am thi-A(proAB-lac), NalR, RIP; Amann et al., 19831, carrying an F' episome that produces high levels of Lac repressor, was used as the transformation recipient for mutant plasmids. E. coli CG103/ RSF2001 (Alac-pro AtrpR-504 thi-I recA-56 srkTnl0 (XCLG145)) was used as the host for rapid @-galactosidase assays. It carries a specialized transducing X prophage with a fusion of the wild-type E. coli trp promoter/operator, attenuator, and start of trpE to lacZ. The episome RSF2001 is a derivative of F with an insertion of a small a b FIG. 1. The Ala7' + Val change creates a steric clash in the crystal structure of aporepressor. Conformations of the helixturn-helix (DNA-binding) motif in the crystal structures of Trp aporepressor and holorepressor are represented in two dimensions. a, in the aporepressor crystal structure, the replacement of the methyl side chain of alanine 77 with the bulkier isopropyl side chain of valine (dotted lines) introduces a steric clash between the side chains of residue 77 and residue Ile8' in e-helix E. b, in contrast, the holorepressor conformation can accommodate this change. Note that 1) the side chain of Ala77 is in a hydrophobic environment in both structures, 2) the replacements of the alanine side chain with the larger side chains of Thr, Val, Leu, and Ile also might be accommodated by the holorepressor structure, but certainly not by the aporepressor structure, and 3) the larger side chain of might cause steric problems with both structures. Modified from Marmorstein and Sigler (1989), with permission. kanamycin resistance determinant (Heffron et al., 1977). For some measurements, it was necessary to use a derivative of this strain that also carries plasmid pMS421 (Gardella et al., 1989), a plasmid with the pSClOl origin and determinants for spectinomycin and streptomycin resistance that overproduces Lac repressor. S. typhimurium strain MS1868/F'lacp (leuA-414(um) hsdSB r-; Benson et al., 1986) was used as the host for infections with challenge phage P22 Kn9 0-ref2 arc-amH1605 and its derivatives with symmetric changes in the trp operator (Bass et al., 1987). The 0-ref2 operator is a minimal, consensus Trp repressor binding site with the sequence: 5' GAAC-TAGTTAACTAGTTG. Plasmid pPY2000, an ampicillin-resistant derivative of pBR322, carries a phage M13 origin of replication and expresses wild-type TrpR protein from the lacUV5 promoter; plasmid pPY1999 is an otherwise isogenic derivative of pPY2000 missing the 434-hp' BamHI fragment with the trpR gene (Arvidson et al., 1991a).
Olikonucleotide-directed Mutagenesis and DNA Sequence Analysis-Derivatives of plasmid pPY2000 with changes a t codon 77 were made by codon-directed mutagenesis. Synthetic oligonucleotides were used to prime the synthesis of the sense strand of single-stranded plasmid pPY2000 DNA containing a low level of uracil. Singlest,randed plasmid template was prepared as described by Del Sal and Schneider (1987), after infection of a dut-ung-E. coli host CJ236 with wild-type M13 phage. Transformation of the double-stranded product into the dut+ ung+ E. coli recipient, X90/F'laclQ, permits a st.rong enrichment against the template strand (Kunkel, 1985).
Most changes of codon 77 were made with the mutagenic primer, 5' GAACTCGGCNNSGGCATCGCG. The oligonucleotides, 5' GA-ACTCGGCNACGGCATCGCG (representing Tyr, His, Asn, and Asp changes), 5' GAACTCGGCYYCGGCATCGCG (Phe, Leu, Ser, Pro), and 5' GAACTCGGCATGGGCATCGCG (Met) were used to obtain changes not found after mutagenesis with the first primer. Mutations in trpR on plasmids carried by transformants of XSO/F'lacP were identified by sequencing the sense strand of the 180-bp SalI-MluI region of trpR, using the dideoxy method (Sanger et aL, 1977). Of the 58 transformants screened after mutagenesis with the first primer, 39 (67%) were mutant. Similar efficiencies were obtained for the other primers, except for the unique primer, which yielded one mutant plasmid among seven independent transformants.
Site-directed mutagenesis often results in secondary, untargeted mutations. These untargeted mutations may result from the relatively low fidelity of M13 replication, mispriming by contaminating RNA o r DNA, replication errors in the synthesis reaction in uitro, or errors The abbreviations used are: bp, base pair(s); IPTG, isopropyl 8- related to the repair of apyriminidinic sites produced in the dut' host. T o ensure that our results do not reflect the phenotypes of secondary, untargeted mutations, the small 180-bp SalI-MluI fragment from each original mutant plasmid was subcloned into an otherwise wildtype plasmid SalI-MluI backbone from plasmid pPY2OOOS prior to characterization. To construct plasmid pPY2000S, the oligonucleotide 5' CGTGAGTTAAAAAATGAGCTCGGCGCAG was used to make a silent change of codon 74 on plasmid pPY2000 from GAA to GAG, which creates a unique SacI site overlapping codons 73 and 74 of trpR. The small 180-hp SalI-MluI fragment from one mutant plasmid with the desired change was subcloned into an otherwise wild-type pPY2000 backbone. The unique SacI site in pPY2OOOS lies between the unique Sal1 and MluI sites; restriction of ligation mixes containing a mutant SalI-MluI fragment and the pPY2OOOS backhone with SacI provides an enrichment for recombinant plasmids. Restriction enzymes and phage T4 DNA ligase (Stratagene Cloning Systems) were used in the mutagenesis and subcloning procedures under conditions specified by the supplier. Quantitation of Mutant Proteins-Minimal tryptophan drop-out (TDO) medium (Bass et al., 1987) consists of M9 medium (Smith and Levine, 1964) supplemented to 20 Fg/ml L-arginine HCI, L-aspartic acid, L-histidine HCl, L-isoleucine, L-leucine, L-lysine HCI, L-methionine, L-phenylalanine, L-threonine, and L-tyrosine (Sigma), and to 0.4% glucose (Difco). E. coli CG103 witb each mutant derivative of pPY2000 was grown in TDO medium supplemented with 1 pg/ml tryptophan, 50 pg/ml proline, 100 pg/ml thiamin, and 400 pg/ml carbenicillin to a density of 2 x 108/ml. Cells (1 ml) were pelleted by centrifugation, resuspended in 70 pl of sample buffer with 2% sodium dodecyl sulfate, and disrupted by heating for 5 min at 70 "C. Proteins in cell lysates (10 pg/lane) were separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose electrophoretically (Arvidson et ai., 1991a). Bound repressor was detected by sequential incubation with a 1:lOOO dilution of primary rabbit anti-aporepressor antibody (Gunsalus et al., 1986), the kind gift of Rob Gunsalus, and a secondary goat anti-rabbit antibody conjugated with alkaline phosphatase, using the Picoblue screening reagent kit (Stratagene Cloning Systems). Under these conditions, TrpR represents about 0.1% of the total cell protein.
Modified Challenge Phage Assays-Challenge phages are derivatives of temperate Salmonella phage P22 that afford a direct selection for proteins that bind specific DNA sites (Benson et al., 1986;Bass et al., 1987Bass et al., ,1988Arvidson et al., 1991a). After infection of a sensitive sup" host, a challenge phage with the trp operator will establish lysogeny if and only if its trp operator is bound by active Trp repressor produced by the host. If the challenge phage operator is free of repressor, the infecting P22 genome will produce antirepressor, develop lytically, and result in death of the infected cell. If the operator is hound by repressor, the infecting P22 genome will lysogenize and confer a kanamycin-resistant phenotype upon its surviving host. Because the fractional occupancy of the trp operator increases with increasing concentrations of Trp holorepressor, the efficiency of lysogeny of a host infected with a trp challenge phage increases in response to both increasing levels of aporepressor and increasing levels of the corepressor, L-tryptophan (Bass et al., 1987, and Fig. 3, below). The Salmonella host we use for challenge phage assays carries two plasmids, pPY2000, which produces Trp repressor from the lacUV5 promoter, and the F'lacP episome. The level of TrpR expression in this host is low in the absence of IPTG, an inducer of Lac repressor, because a high level of F-encoded Lac repressor inhibits transcription of trpR from the lacUV5 promoter.
Challenge phage assays were modified from previous procedures (Bass et al., 1987(Bass et al., , 1988 to permit many simultaneous assays. Serial dilutions of phage-infected cells were made using a programmable automatic pipetting device to transfer multiple aliquots simultaneously to the wells of a sterile microtiter dish. In addition, multiple aliquots of these serial dilutions were spotted simultaneously onto solid culture media in large (150-mm diameter) Petri dishes.
To measure repression as a function of the amount of plasmidencoded Trp repressor required for the efficient lysogenization of the challenge phage, overnight cultures of MS1868/F'lacp carrying pPY2000 or each mutant plasmid were diluted 100-fold into 5 ml of LR medium with 50 pg/ml ampicillin and grown to a density of 2.5 X 108/ml at 37 "C. IPTG was added to a final concentration of M to induce the expression of Trp repressor, and cells were grown an additional 30 min at 37 "C, then placed on ice for 15 min. To an aliquot (135 p l ) of each culture, an equal volume of challenge phage P22 Kn9 0-ref2 arc-arnH1605 at a titer of 10'n/ml was added, to give a multiplicity of infection of approximately 26 phage/cell. After adsorption of the phage for 15 min at 25 "C, infected cells were diluted seven successive 5-fold dilutions in the wells of microtiter plates, and 4 pl of each 5-fold serial dilution was spotted on green tryptophan drop-out plates with various concentrations of IPTG, ampicillin (100 pg/ml), kanamycin (25 pg/ml), and with or without tryptophan (100 To measure repression as a function of the amount of tryptophan (corepressor) required for the efficient lysogenization, 4 pl of each serial dilution was spotted on green tryptophan drop-out plates with M IPTG, ampicillin (100 pglml), kanamycin (25 pg/ml), and tryptophan at various concentrations.
To determine whether mutant repressors have new binding specificities, cultures of MSl868/F'lac~ carrying each mutant plasmid were grown to a density of 2.5 X 108/ml at 37 "C, IPTG was added to M, and cells were grown for an additional 30 min a t 37 "C. Cultures were infected with each of the eight different challenge phages derived from P22 Kn9 0-ref2 arc-amH1605 that carry symmetric changes in the consensus trp operator which prevent the binding of wild-type repressor (Bass et al., 1987). Infected cells were diluted in the wells of microtiter plates, and 4 pl of each serial dilution was spotted on green tryptophan drop-out plates with lo-' M IPTG, ampicillin (100 pglml), kanamycin (25 pglml), and tryptophan (40 For each challenge phage assay, the efficiency of survival is the titer of surviving cells divided by the titer of input cells (assayed on green plates with ampicillin); averages of the results from a t least three independent experiments are shown. Efficiencies of survival varied less than 5-fold from experiment to experiment.
Rapid P-Galactosidase Assays--Rapid &galactosidase assays were conducted as described by Arvidson et al. (1991b). Each CG103 strain with plasmid pMS421 and a derivative of pPY2000 was grown in TDO medium with 400 pg/ml carbenicillin and with or without 1 pg/ ml tryptophan in the wells of microtiter plates. Exponential cultures were lysed from without with phage T4 in 8-galactosidase assay buffer, and liberation of o-nitrophenol from o-nitrophenyl-@-D-gaIactoside was monitored as A414 over time with a microtiter plate spectrophotometer. Activities were calculated using the equation pg/ml). pdml).
where b is the slope of linear least squares fit to the plot of A414 versus time (rnin), and f (0.2 in our assays) is the fraction of cells added to the total volume of lysate. In all cases, averages of three or more independent determinations varied less than 20%. A high concentration of the relatively slowly hydrolyzed penicillin analog, carbenicillin, was chosen for these assays, because hosts carrying some of the mutant pPY2000 derivatives (in particular, those producing more active aporepressors or holorepressors) segregate a high fraction of daughter cells that have lost the plasmid in parallel assays with ampicillin. This choice of &lactam improves but does not solve the problem of plasmid loss. Assays with carbenicillin in the presence of tryptophan concentrations greater than 4 pg/ml or the absence of plasmid pMS421 also result in a high frequency of loss of some pPY2000 derivatives and yield higher, more variable 8-galactosidase activities.

Site-directed Mutagenesis of Codon 77"
To change codon 77 of the trpR gene, we used a degenerate oligonucleotide to mutagenize plasmid pPY2000, a plasmid which expresses Trp repressor from the lac promoter. The mutagenized template was transformed into E. coli host XSO/F'lucP. This E. coli host produces a high level of Lac repressor from its F' episome that can turn down the potentially lethal expression of mutant trpR genes.
As summarized in Table I, we sequenced plasmid DNA isolated from 58 transformants and found mutant plasmids with 14 different codons representing 12 different amino acids at position 77. To make the remaining 7 possible amino acid changes, we used 3 additional synthetic oligonucleotides to prime the synthesis of the sense strand of pPY2000, as with the more degenerate primer.

Mutant Repressors Are Made at Normal Levek in Vivo-
T o measure the relative steady-state levels of mutant repressors with Ala7' changes in vivo, derivatives of the E. coli host

aP0 A H R K E D Q N Y C T S G M W F P I L V A
. " _ -"" " .. All 19 mutant Trp repressors with Ala" changes are made at similar steady-state levels. Proteins present in mutant and wild-type repressor-producing derivatives of E. coli host CG103 were resolved by polyacrylamide gel electrophoresis, transferred electrophoretically to nitrocellulose, and identified by reaction with rabbit anti-TrpR antibody. The amino acids present at position Ala77 in mutant proteins made by derivatives of plasmid pPY2000 are indicated by single-letter amino acid code; apo, purified aporepressor (12.5 ng); A, protein present in CG103(pPY1999), a deletion derivative of pPY2000 without the trpR gene (negative control); A, wildtype protein present in CG103(pPY2000). This assay is linear to 40 ng of Trp repressor per lane (data not shown). The bands in the lane between lanes T and S are prestained molecular weight standards (Bio--Rad Laboratories Catalog No. 161-0305). strain, CG103, with each of the mutant plasmids were grown to exponential phase in TDO medium with 1 pg/ml tryptophan, and the relative amounts of intracellular TrpR protein were quantitated using a Western immunological assay. As shown in Fig. 2, all of the mutant repressors are made in amounts similar to that of the wild type, indicating that none of these changes results in a protein significantly less stable than the wild type in vivo.
Two Challenge phage assays done with an excess of exogenous tryptophan and varying amounts of Trp repressor allow us to measure the relative DNA-binding activities of mutant holorepressors to an idealized, minimal trp operator in vivo (Bass et al., 1988). This is the case for mutant TrpR proteins with increased activities arising from each of three different Glu + Lys changes.2 Fig. 3 shows that the efficiency of lysogeny of the host MS1868/F'laca (pPY2000) with a trp challenge phage in the presence of excess tryptophan increases with increasing amounts of IPTG, an inducer of Lac repressor. Without IPTG, this host makes such low levels of E. coli Trp repressor that it cannot be lysogenized efficiently. Presumably, the low background level of Trp repressor encoded by the single-copy trpR gene on the host Salmonella chromosome is not sufficient to promote efficient lysogeny, although it accounts for a significant background frequency. With high IPTG, this host produces enough Trp repressor to allow efficient lysogenic development of the challenge phage.
Under this first set of conditions, two mutant repressors, and SeP, are more active than wild-type repressor, and four mutant repressors, V a P , Leu77, T h P , and Ile77, have activities close to that of the wild type (Fig. 3a). The mutant Met77, L Y S~~, and Gly77 repressors have lower activities; however, when they are made in sufficient amounts, they permit efficient lysogenization (Fig. 3b). The mutant His77 repressor has barely detectable activity, and all other mutant repressors are inactive in this assay. These results rank active mutant holorepressors in the order: Cys > Ser > Val, Leu, Ala > Ile, Thr > Lys, Met > Asn > Gly > His.

Seven Mutant Repressors Are Super-aporepressors-Under
a second set of conditions of the challenge phage assay, in which corepressor is limiting (in the absence of exogenous tryptophan), different results are obtained. As shown in Fig.  4, although our control host that produces wild-type protein cannot support efficient lysogenization under these conditions, seven hosts that produce mutant repressors survive challenge phage infections at higher frequencies. These mutants are ranked in the order of activity: Cys > Val, Ile > Leu > Thr > Lys > Ser > Ala. Fig. 5 shows the results of a third set of challenge phage assays. In this experiment, we compared the ability of small amounts of each of the mutant repressors to be activated by tryptophan. Cells induced for repressor expression with 10 p~ IPTG were infected with challenge phage, and survivors were titered on plates with varying concentrations of tryptophan. Again, the same seven mutant repressors are more active than wild-type under these conditions and rank in the order: Cys, Val, Ile, Leu > Thr > Ser > Lys > Ala > Met.
This ranking is nearly identical with that found under the second set of assay conditions (no added tryptophan) and confirms that seven mutant proteins with Ala77 changes are less dependent on corepressor than wild-type TrpR in uivo.
However, we note that the relative activity of one mutant protein, L Y S~~, is different under these two sets of conditions (Figs. 4 and 5 ) .
Residue Ala77 is in the helix-turn-helix (DNA-binding) motif of Trp repressor. One reason why several mutant Trp repressors with Ala77 changes may be more active than wild type is that, like the changes of Ala8' + Ser and Alaso + Thr, these changes add functional groups to a side chain that may participate in new interactions with the operator and result in altered DNA-binding specificities. Therefore, we used the challenge phage assay to screen for the ability of mutant repressors to bind mutant operators that the wild-type repressor cannot bind well, as described by Bass et al. (1988). We found that each of the mutant repressors binds either the same set of operators as wild type or a proper subset of the operators bound by the wild type and has a wild-type or restricted specificity of binding (data not shown). None of the mutant repressors with Ala77 changes has a new or extended specificity of binding.
Interaction with the Natural trp Operator-As an independent assay of the relative activities of these mutant proteins, we measured their abilities to repress a single copy trpllnc Trp repressor protein in the absence of exogenous tryptophan. Because the defined (TDO) medium used for the growth of survivors of challenge phage infection has an excess of both phenylalanine and tyrosine, but no tryptophan, the concentration of free intracellular tryptophan is further limited by the slower synthesis of 3-deoxy-~arabino-heptulosonate 7-phosphate (a common precursor of the aromatic amino acids) under these conditions (Brown and Somerville, 1971). TrpR protein is expected to exist predominantly as aporepressor, not holorepressor, in this assay; consequently, the 7 mutants with activities higher than the wild type under these conditions are called "super-uporepressors." However, we do not wish to imply that mutant aporepressor species can effect repression under these conditions (see text). fusion operon carried by an integrated h prophage in E. coli. Unlike the challenge phage assays, which measure repression of a minimal symmetric TrpR binding site, this assay measures repression of the wild-type E. coli trp operator, a more complex binding site composed of multiple, tandem repressor dimer binding sites (Kumamoto et al., 1987). T o maximize the sensitivity of this assay, each of the plasmids with codon 77 changes was transformed into E. coli host CG103 carrying a second plasmid, pMS421, which produces high levels of Lac repressor to turn down transcription of trpR (Gardella et al., 1989). In the absence of pMS421, constitutive expression of wild-type and many mutant Trp repressors in this host results in the near complete repression of the trpllac fusion operon (see Arvidson et al., 1991a, for example) and makes it even more difficult to observe phenotypic differences between mutant and wild-type Trp repressors. Table I shows that the 19 mutant proteins with Ala77 changes are roughly grouped into three different classes on the basis of their ability to repress the E. coli trpllacZ fusion operon in CG103(pMS421) ( Table I) show that all of the active, mutant repressors have increased activities in the presence of added corepressor; none is tryptophan-independent.
The relative activities of mutant repressors determined by this /3-galactosidase assay agree with those determined by the challenge phage assay, although the relatively insensitive 6galactosidase assay reveals the super-repressor phenotype of only two of the mutants, Cys and Val. The six mutant repressors with activities better than, or comparable to, wild type in the challenge phage assay (Cys, Ser, Val, Leu, Thr, Ile) are the most active in this @-galactosidase assay. Five mutant repressors (Lys, Met, Asn, Gly, His) show intermediate levels of activity, and the remaining mutants show little or no detectable activity in both assays.

DISCUSSION
Residue alanine 77 is located in the center of the turn of the "helix-turn-helix" DNA-binding motif of E. coli tryptophan repressor and is critical for TrpR function. The majority (12/19) of Ala77 changes results in either inactive (Arg, Asp, Gln, Glu, Phe, Pro, Trp, and Tyr) or only partially active (Asn, Gly, His, and Met) proteins. Neither amino acids with acidic side chains (Asp, Glu) nor amino acids with the four largest side chains (Arg, Trp, Tyr, Phe) are acceptable at this position. Because the Pro77 protein has almost undetectable activity, the conformation of the turn between a-helices D and E also must be critical for function.
Changes of Ala77 that impair TrpR activity do not alter its steady-state level of production. All of the 19 different mutant proteins with Ala77 changes are made at similar steady-state levels as the wild-type protein. Presumably, all single amino acid changes affecting residue Ala77 permit folding to a protease-resistant state. In contrast, all of the changes of residue Val", which frames the tryptophan binding pocket, result in repressors that are made at lower steady-state levels in vivo (Arvidson et al., 1991a).
When tryptophan is limiting, 7 mutant aporepressors with Ala77 changes are more active than wild-type aporepressor (Figs. 4 and 5). At high tryptophan concentrations, the 7 super-aporepressors fall into three groups on the basis of their relative activities in the challenge phage assay. Under these conditions, Cys and Ser holorepressors are more active than wild type; Val, Leu, Ile, and Thr are about as active as wild type (Fig. 3a); and Lys holorepressor is less active than wildtype (Fig. 3b). In contrast, none of the single amino acid changes of residue Val5' improves TrpR function (Arvidson et al., 1991a).
How do these changes affect TrpR function? As modeled thermodynamically in Fig. 6a, TrpR protein is distributed among six different states i n uiuo, related by a series of linked equilibria involving aporepressor (A), tryptophan (T), nonspecific DNA (D), and operator DNA (0) (Arvidson 1989;adapted from von Hippel, 1979). When tryptophan is saturating, TrpR protein populates only 3 of the 6 states, free holorepressor (AT), holorepressor bound to nonspecific DNA (ATD), and holorepressor bound to operator DNA (ATO) (Fig. 6b). The activity of a mutant holorepressor in the presence of excess tryptophan reflects the competition between its ability to bind nonspecific and operator DNAs and involves only two equilibrium dissociation constants, KATD and KATO. The ability of holorepressor to bind operator DNA measured in vivo is related to these constants by the equation  (Segel, 1976). For experiments in vivo in which [Dl is held constant, differences in apparent binding reflect differences in this ratio.
The results of challenge phage infections with excess tryptophan rank the activities of four mutant holorepressors, GluX3 + Lys, Glu18 + Lys, Glu4' + Lys, and Ala77 + Val i n uiuo, in the same order as the magnitudes of Kapp determined for these holorepressors i n uitro (Hurlburt and Yanofsky, 1990)'; these assay conditions provide us with a relative measure of Kapp i n uiuo. Therefore, and Ser77 holorepressors must have lower Kapp values than wild type; Thr77, and Val77 holorepressors have Kapp values similar to wild type; and LYS'~ and Met77 holorepressors must have higher Kapp values.
One of these mutant super-repressors, Val77, has been purified and examined i n vitro. Using a nitrocellulose filterbinding assay, Hurlburt and Yanofsky (1990) have found that KAT, KAo, KATO, and KATD for Val77 protein are indistinguishable from those for wild-type TrpR. These results agree well with our physiological data; we see no effect of the Ala77 "-* Val change on Kspp in vivo (Fig. 3a). Because the pathways A --f AT + AT0 and A -+ A 0 + AOT involve the same net change in free energy, if KAT, K A O , and KAT0 are the same for Ala77 and Val77, KAoT must be the same. Therefore, the superuporepressor phenotype of Val77 must be due to depletion of the AD state. If K A T and KATo are the same for Ala77 and Val77, the only equilibria that could be affected by the AlaT7 "* Val change are described by K A D and KADT. The Val77 change must result in an increase in KAD, and, because the pathways A --f AT + ATD and A "+ AD + ADT involve the same free energy change, a commensurate decrease in KADT. Together, these effects will depopulate [AD], which, in turn, will result in an increase in [ATO]. A corollary of this deduction is that a significant fraction of wild-type aporepressor must be associated with nonspecific DNA i n vivo.
Mutations that increase the activity of a repressor by affecting the interaction of a repressor-ligand complex with nonspecific DNA are not without precedent. Super-repressor mutants with Ala77 changes, which show tighter apparent binding to the trp operator with decreased corepressor concentrations, are analogous to a subset of EacP mutants, which show tighter apparent binding to the lac operator with increased inducer concentrations (Jobe et a i , 1974;von Hippel, 1979).
If we assume that other Ala7' changes, like the Ala77 + Val change, do not have pronounced affects on KAT (Marmorstein and Sigler, 1989) (Fig. 3). Consistent with this model, we can define conditions for LysI7 under which these opposing effects on aporepressor and repressor activities balance each other, and behaves much like the wild type (Fig. 5b). This interpretation is supported directly by a comparison of the orthorhombic crystal structures of aporepressor and holorepressor (Fig. 1). In both forms, the side chain of Ala77 points toward the interior of the protein and away from the DNA-binding surface. In the aporepressor structure, the helix-turn-helix motif is collapsed against the hydrophobic core; in the repressor structure, the motif is rotated away from the core. In aporepressor, the side chain of AlaI7 is almost completely buried; in repressor, it is more solvent-exposed. Thus, the AlaI7 + Val change should cause a steric clash in the aporepressor crystal structure, but not in the holorepressor crystal structure ( Fig. 1; Marmorstein and Sigler, 1989;Luisi and Sigler, 1990). Lengthening the Ala77 side chain by mutation should result in proteins that cannot assume this aporepressor conformation.
Consistent with this picture, there is a correlation between the specific volumes of mutant side chains at position 77 and phenotype. Changes to Ser and Cys involve the smallest increases in side chain volume and result in proteins that are super-holorepressors. The side chains of Ile, Leu, Thr, and Val are somewhat larger; mutant proteins with these changes are super-aporepressors, but have holorepressor activities similar to Ala77. These intermediate-size residues are predicted to disrupt the aporepressor tertiary structure, yet are tolerated by the holorepressor structure. The protein, with the largest functional side chain, is a better aporepressor, but a worse holorepressor, than wild type. This large side chain should interfere with folding of both aporepressor and holorepressor (Fig. 1). The Met77 protein, with a side chain almost as long as lysine, also has decreased holorepressor activity (Fig. 3). Mutant repressors with larger side chains (Arg, Phe, Tyr, Trp) have little detectable holorepressor activity.
An alternative explanation of the ValI7 phenotype has been suggested by Marmorstein et al. (1991). Using an assay that measures the ability of TrpR protein to protect an idealized trp operator from dephosphorylation by alkaline phosphatase, they find different results: for V a P repressor, KATO is 2.3-fold higher and KAo is 8-fold lower than wild type. However, their data cannot explain why the protein requires less tryptophan for activation i n vivo, because they find that KAT is the same for V a P as for wild type (Marmorstein 1989), and the pathways A + A 0 + AOT and A + AT + A T 0 must involve the same free energy change. Their data argue that assembly of the repressor/operator complex should be impaired 2.3-fold by either route; i.e. an 8-fold increase in KAO must be accompanied by an 18-fold decrease in KAOT, because (KAT)(KATo) = (KAo)(KAoT). A mutation that increases KAo 8-fold should not have a significant effect on repression by increasing the relatively small A 0 population, unless this complex prevents transcription initiation. More important, these experiments were done under conditions i n vitro that do not reflect the physiological state in uiuo. At higher salt concentrations, the super-repressor phenotype of the mutant Glu4' + Lys repressor observed both in uiuo (Kelley and Yanofsky, 1983) and i n vitro Hurlburt and Yanofsky, 1990) is also revealed by the phosphatase inhibition assay (Marmorstein et al., 1991).
In addition to the 7 super-repressors we have described in this paper, only 4 other single amino acid changes are known to increase TrpR activity. All 4 of these changes arise by single base pair transitions and are predicted to make the DNA-binding surface of the repressor dimer more basic (Kelley and Yanofsky, 1983;. In contrast, genes encoding 4 of the 7 mutant super-aporepressors with Ala77 changes (including C Y S~~, the most active super-repressor) cannot arise from the wild-type trpR gene by mutation of a single nucleotide base pair. Because classical methods of chemical mutagenesis give rise to mutant codons with tandem double or triple base pair changes only very rarely, such methods would preclude the isolation of the majority of mutant repressors with Ala77 changes that increase activity. Thus, our results suggest both that a significant fraction of possible single amino acid changes throughout the protein will increase repressor activity and that our more biochemically thorough approach of isolating mutants with each of the possible single changes at each residue position of Trp repressor will yield dividends that classical genetic approaches cannot.
We have used an assay i n vivo based on a genetic selection to deduce that mutant Trp aporepressors have decreased affinities for nonspecific DNA. Such affinities are likely to be in the millimolar range, a range that precludes direct and accurate measurement i n vitro by many standard biochemical methods. It is clear that novel combinations of biochemical and genetic methods will be necessary to probe the relationship between Trp repressor structure and function.