Identification of Residues in the L 1 Region of the RecA Protein Which Are Important to Recombination or Coprotease Activities *

Using a combinatorial cassette mutagenesis procedure we have introduced a large number of single and multiple amino acid substitutions into an area of the RecAprotein defined by residues 152-159. This sequence overlaps the disordered loop 1 region (Ll) in the RecA crystal structure which has been hypothesized to be involved in DNA binding. Assays for recombinational DNA repair and Led coprotease activities identify Glu’” as the only one of these 8 residues which is critical to RecA function. Several other mutations observed at nearby residues support the identity of Glu’% as the most important of the 14 residues in the area defined by Prola’ to Metla. In addition, Glyl“ and GlulS8 appear to be hot spots for the occurrence of mutation-induced constitutive coprotease activity.

Using a combinatorial cassette mutagenesis procedure we have introduced a large number of single and multiple amino acid substitutions into an area of the RecAprotein defined by residues 152-159. This sequence overlaps the disordered loop 1 region (Ll) in the RecA crystal structure which has been hypothesized to be involved in DNA binding. Assays for recombinational DNA repair and L e d coprotease activities identify Glu'" as the only one of these 8 residues which is critical to RecA function. Several other mutations observed at nearby residues support the identity of Glu'% as the most important of the 14 residues in the area defined by Prola' to Metla. In addition, Glyl" and GlulS8 appear to be hot spots for the occurrence of mutation-induced constitutive coprotease activity.
The R e d protein from Escherichia coli is a multifunctional enzyme that plays two distinct catalytic roles related to cell survival following DNA damage. First, RecA catalyzes a postreplicational strand exchange activity between homologous DNA molecules so that information from the undamaged DNA is used to restore information on the damaged homologous partner (for review see Refs. 14). Second, in response to DNA damage RecA becomes activated for a coprotease function in which it facilitates the autoproteolysis of the cellular LexA repressor ( 5 , 6). This event increases the expression of a number of genes including r e d , collectively referred to as the SOS genes, which are involved in cell survival following DNA damage (7,8). To carry out either of these activities the RecA protein must bind both ATP and DNA to form an activated nucleoprotein filament. Even though a number of r e d mutants show coincident decreases in both recombination and coprotease activities, there is both genetic and biochemical evidence showing that these two activities are separable. A number of r e d mutants have been shown to be coprotease-proficient but compromised for recombination activity (9,10). In addition, several mutations result in RecA proteins that display a constitutive coprotease activity (catalysis of LexA cleavage in the absence of DNA damage) yet have varying effects on the recombination activity (11,12, this study). We have now designed a series of plasmids in which expression of recA is regulated by Pgac, thereby removing r e d from SOS regulation. The basal level of expression is sufficient for genetic studies of all RecA activities.
* This work was supported by National Institutes of Health Grant GM 44772 (to K. L. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with This system also provides a more complete separation of the coprotease function from other activities and has allowed identification of r e d mutants that are defective for L e d cleavage but remain active for other functions.' To investigate the structural requirements for the many activities catalyzed by RecA we have initiated mutagenesis studies within targeted regions of the protein (13)(14)(15). The availability of the x-ray crystal structure of RecA (16,17) provides an opportunity to address detailed questions regarding the functional and/or structural roles of specific amino acid residues within these targeted regions. The structure of R e d was solved in the absence of bound DNA, and Story et al. (16) have suggested that two disordered regions, L12 and L2, are involved in the interaction of the RecA oligomer with DNA. We have introduced a large number of single and multiple amino acid substitutions into a stretch of 8 residues (152-159) that flank and are contained within region L1. Assays for both the DNA repair and LexA coprotease activities show that RecA function is most sensitive to mutation at Glu'". The DNArepair activity shows a moderate sensitivity to mutation at Gly15?, whereas the other six positions tolerate high levels of substitution. Twentynine of the 149 unique mutants in this study display constitutive coprotease activity yet have widely varying effects on DNA repair activity, Our results are discussed in terms of possible roles for this region in the catalysis of LexA autodigestion and DNA repair.
EXPERIMENTAL PROCEDURES Materials-All media (LB broth, LB agar, 2 x YT) were prepared as described (18) and were supplemented with 100 pg/ml ampicillin when appropriate. MacConkey-lactose plates were prepared according to the manufacturer (Difco) and contained 0.5% lactose and 100 pg/ml ampicillin. Stock solutions (4 mg/ml) of o-nitrophenyl P-D-galactopyranoside (ONPG; Sigma) were made in 2-buffer (19). Protein concentrations in cell extract supernatants were determined using the Bio-Rad protein assay kit. Mitomycin C was from Sigma or Boehringer Mannheim. Restriction enzymes, polynucleotide kinase, T4 DNAligase, and Klenow DNA polymerase I large fragment were from New England Biolabs. Modified T7 DNA polymerase (Sequenase version 2) was from U. S. Biochemical Corp. All buffers were as recommended by the manufacturer. Serum containing polyclonal antibodies against R e d (rabbit anti-RecA) was prepared commercially (East Acres Biologicals, Southbridge, MA) using gel-purified RecA protein as antigen (13).
Strains and Plasmids-Strain X90 (20) was used for general purposes, e.g. plasmid construction and propagation and as a r e d + control strain. Strain DE1663' was used for in vivo assays of the DNA repair and coprotease activities of all plasmid-borne r e d mutants. DE1663' was constructed by mating strain DE1663, a A(red-srZR) 306:TnlO A(lac-argF) U169 sulA.211 ma1B::TnS (A cI ind-l recAolp::Zac Z Y ) derivative of AB1157, with strain DE1781 which carries an F' lac pZacPL8 lac2 4505::Tn5proA+B+ episome. Both DE1663 and DE1781 were generous gifts from Don Ennis (NIH). Plasmid pTRecA322 (5,740 base pairs; Fig. l), which contains an M13 origin of replication and carries the r e d gene under control of P, , is a derivative of pTRecA103 (13). Unique restriction sites were introduced at positions that flank region L1 of the r e d sequence using standard procedures for oligonucleotide-directed mutagenesis (13,21). An SalI site was created within codons 143-145, and a BssHII site was created within codons 168 and 169.
Plasmid pZ150 (22) is pBR322 that contains an M13 origin of replication and was used as a r e d -control plasmid for the in vivo assays described below.
Cassette Mutugenesis-Cassette mutagenesis was performed using a modification of described procedures (13,23,24) to introduce a calculated degree of mutation into two sets of four contiguous codons ( L y P to Ile'55 and G~u '~~ to Ile"'). Two 86-base oligonucleotides corresponding to the sequence encoding residues Val'43 to were made using an Applied Biosystems 392 DNABNA synthesizer and were gel purified (21). One of the oligonucleotides was synthesized such that the 12 bases corresponding to residues 152-155 had a 73% probability of being wild type and a 27% probability of being any of the other 3 bases (13,23). All other bases were the wild type r e d sequence. The second oligonucleotide contained the above probability of mutation at the 12 bases corresponding to residues 156-159. Both oligonucleotides contained an SalI site at the 5' end (plus 3 bases beyond the recognition sequence) and a BssHII site at the 3' end (plus 2 bases beyond the recognition sequence). Enzymatic second strand synthesis, digestion with SalI and BssHII, and gel purification of the resulting double-stranded cassettes were performed as described (24). Cassettes were ligated into SalI-BssHIIdigested pTRed322 backbone, and products were transformed into DE1663' (25). Transformants were selected on LB-ampicillin plates and restreaked once.
Measurement of DNA Repair Activity in Viuo-The recombinational DNA repair proficiency of r e d mutants was analyzed using two in vivo assays: 1) cell survival in the presence of mitomycin C, and 2) cell survival following exposure to different doses of W light. Cell survival in the presence of mitomycin C was assayed by spotting 3 pl of overnight cultures diluted 1/40 onto LB-ampicillin plates containing either 0.3 or 0.6 pg/ml mitomycin C and onto control plates containing no mitomycin. Following overnight incubation at 37 "C cell growth was ranked on a scale of 0-4. Each plate contained a positive (DE1663'/pTRecA322) and a negative (DE1663'/pZ150) control strain. The use of these two concentrations of mitomycin C allowed us to identify r e d mutants that displayed a partially functional repair activity. For W survival assays cultures were diluted 1/40 and streaked across LB-ampicillin plates that had been marked off into seven time zones (0-60 s). Cells were exposed to 0.67 J/m2 UV light (model WGL-25; U W , Inc., San Gabriel, CA) for the appropriate length of time, wrapped in foil to prevent DNA repair by photoreactivation, and incubated at 37 "C overnight. Positive and negative controls were included on each plate. Fractional survival at 30 s was calculated using the slope of a line resulting from a plot of relative growth versus time.
Assays were performed in the absence of isopropyl-1-thio-0-D-galactopyranoside. Under these conditions basal expression from the plasmid-borne P,,-red gene is approximately 20-fold higher than basal expression of chromosomal r e d in wild type E. coli (13). This increased basal expression and the inability of DNA-damaging agents to up-regulate r e d expression have little or no effect on the ability of in vivo assays to distinguish different classes of r e d mutants based on measurement of recombinational DNA repair activity (13).
Measurement ofRed-mediated L e d Cleavage-Strain DE1663' carries the lacZ and Y genes under control of the r e d operator/promoter, and therefore measurement of P-galactosidase activity is directly related to the extent of RecA-mediated LexA cleavage. Two assays were used for the measurement of @galactosidase activity. First, cultures carrying plasmid-borne r e d mutants were grown overnight in LBampicillin media at 37 "C, diluted 1/40 in the same media, and 0.5 pl spotted onto two MacConkey-lactose plates, one containing a sublethal dose of mitomycin C which is sufficient for SOS induction (0.05 pg/ml) and one that contained no mitomycin C. Plates were incubated overnight at 37 "C. The second assay provides a more quantitative measure of RecA-mediated L e d cleavage by determination of P-galactosidase activity in extracts of cultures carrying the r e d mutants. Cultures were grown overnight at 37 "C, diluted 1/100 in LB-ampicillin media, and grown for 2 h at 37 "C. To half of each culture (1.5 ml) mitomycin C was added to a final concentration of 0.5 pg/ml, and all samples, both (+) and (-1 mitomycin C, were grown for an additional 40 min. Cultures were then chilled on ice for 20 min, centrifuged for 5 min, washed with 1 ml of 10 m M NaC1, and resuspended in 1 ml of Z-buffer (19). Cells were sonicated on ice for 30 s, centrifuged for 20 min, and the supernatant was stored at 4 "C. P-Galactosidase activity in these supernatants was measured using ONPG as described (19). Units of p-galactosidase are defined as mol of ONPG hydrolyzed/min, and activity is expressed as unitdpg protein.
Cellular Level of Mutant R e d Proteins-Western blot analysis was performed as described previously (13) for each r e d mutant. For this analysis cultures were grown in the absence of isopropyl-1-thio++galactopyranoside. None of the mutants described in this study showed a steady-state level of R e d protein which was significantly different from wild type RecA (data not shown).
DNA Sequencing-Amino acid substitutions were determined by DNA sequence analysis of the plasmid-borne mutant r e d genes. Sequencing was performed using modified T7 DNA polymerase (Sequenase version 2) on either double-stranded template (Magic Mini-prep DNA purification system; Promega) or single-stranded template that was isolated following infection of cultures containing mutant r e d plasmids with an M13-derived helper phage, RV-1 (26). We determined the sequence of the cassette insert as well as several bases flanking the SalI and BssHII restriction sites.

RESULTS
463 transformants were screened for DNA repair and LexA coprotease activity. DNA sequence analysis showed that 254 mutants contain one of the following: 1) silent mutation(s1; 2) an insertion or deletion; 3) nonsense codon(s); or 4) the wild type recA sequence. Data for the remaining 209 mutants (149 unique) are presented below.
Previous studies suggest that cell survival following DNA damage by UV or mitomycin C is more a measure of the recombination proficiency of RecA than its ability to induce the SOS response (27,28). Two results, in particular, lend strong support to this claim: 1) transformation of pTRecA322 into a strain deleted for r e d in which the SOS system is irreversibly repressed (ArecA Zed31 allows survival following exposure to UV or mitomycin C at approximately 80% of the level seen in either a ArecA Zed' strain carrying the same plasmid or a recA' Z e d ' strain3; and 2) a strain deleted for r e d in which the SOS system is expressed constitutively (ArecA Zed def) shows survival at no more that 0-5% the level of a red' l e d ' strain' (29). Therefore, using survival following DNA damage as a measure H. G. Nastri and K. L. Knight, unpublished data. of the recombinational proficiency of RecA we separated all mutants into three general phenotypic categories defined as follows. recA' showed growth in the presence ofboth 0.3 and 0.6 pg/ml mitomycin C and survival of W irradiation (fractional survival at 30 s 2 0.95) similar to positive control cells (DE1663'/pTRecA322 or X90). r e d -showed no survival following exposure to mitomycin C or W (fractional survival at 30 s 5 0.05) greater than negative control cells (DE1663'/pZ150). recA'" showed intermediate levels of survival.
Preliminary screens for L e d coprotease activity were performed using MacConkey-lactose plates (*mitomycin C). More quantitative ONPG assays were used t o separate all mutants into three categories: coprt' (constitutive activity), coprt' (activity induced following mitomycin C-dependent DNA damage), and coprt-. These categories are defined as follows. Coprt' represents noninduced @-galactosidase activity at least 1.5-fold higher than wild type RecA. Coprt' represents noninduced activity greater than negative control cells (DE1663'/pZ150) followed by some degree of mitomycin C-dependent induction. Coprt-represents noninduced and induced levels of activity similar to negative control cells. Coprt' mutants were divided into three categories based on the specific activity of P-galactosidase in the absence of mitomycin C: strong, >20 units/pg; moderate, c20 and 210; weak, 4 0 and 25.9.
L y~'~~" P o s i t i o n 152 tolerates a high level of mutation with no adverse effects on either the DNA repair or coprotease activities. The Lys -Ala recA+/coprt+ mutant indicates that no specific side chain information is required at this position for either activity (Table I). Several other substitutions allow full activity and are limited to positively charged (Arg and His) or uncharged, polar residues (Gln, Asn, Ser, and Thr).
A significant inhibition of the DNA repair activity results from mutation t o a negatively charged (Glu) or hydrophobic Ole) side chain (Table 11, N-82 and N-83). In contrast, although the Glu mutation decreases coprotease activity this function is unaffected by the Ile substitution. The deleterious effects of both substitutions are suppressed by a secondary change to Lys at nearby positions (see below; Tables I and V, mutants N-43 and N-45). No single substitution at position 152 was found which completely inhibited RecA functions.
A l~~~~-P o s i t i o n 153, like 152, accommodates a number of substitutions that have no effect on either RecA activity. Both polar (Ser, Thr, and Gln) and nonpolar (Leu and Val) residues are allowed. Substitution by either Pro or Gly is also allowed, indicating that no particular constraints exist regarding the positioning of the polypeptide backbone at this location.
Like position 152, introduction of a negatively charged Glu side chain results in a decrease in DNArepair activity (Table 11, N-84). However, this mutation increases coprotease activity to a moderately constitutive level (see Table IV). No single substitution was found at this position which completely inhibited RecA activity.
G l~'~~-T h i s is the only one of the 8 targeted residues for which the wild type side chain appears to be critical for the maintenance of full RecA activity. Although we picked up only four unique single mutations, they indicate that rather strict chemical and steric constraints are in effect at this position. A G~u '~~ -Asp mutation results in an appreciable decrease of both activities (Table 11, N-85), whereas substitution with an isosteric Gln residue completely inactivates RecA functions (Table 111, N-121). These two mutants suggest that both the presence and precise positioning of a negative charge at position 154 are critical to RecA function. Interestingly, defects in the DNA repair uersus coprotease activities of the G~u '~~ -Asp mutant are differentially suppressed by second site mutations (see Table V, N-85, N-95, N-55; see below).
Mutation of G~u '~~ to either Lys or Arg completely inactivates RecA (Table 111, N-122 and N-123). Other mutations at this position that probably account for a recA-phenotype (Gly, Ala, Tyr, and Val) have been inferred from multiple substitution mutants (see below).
Zle'55-Position 155 tolerates a number of substitutions of varying chemical character and size with little or no adverse effect on either RecA activity. Substitution with other nonpolar residues of varying size (Ala, Val, and Leu), polar residues Cys and His, and the aromatic Phe and Tyr residues have no detectable effect on DNA repair or coprotease activity (Table I).
However, although an Ile'55 -Met mutation has no effect on the DNA repair function, it has a dramatic effect on coprotease activity, resulting in a strong coprt" mutant (Table IV, N-19). The fact that this is the only one of 8 single mutations which has an exclusive effect on the coprotease function suggests that there are specific chemical and/or steric requirements at this position which differentiate between the DNA repair and coprotease capabilities of RecA.
Substitution of Ile'55 with Asn has a slight inhibitory effect on both activities (Table 11, N-86). Although we found no single substitution that completely inactivates RecA function, mutation to Thr appears to correlate with complete inhibition of both activities. A double mutant carrying Ile155 + Thr and Lys15' + Thr (Table 111, N-128) scores as recA-/coprt-even though the latter mutation by itself has no effect on RecA function (Table   GLu'~~--A large number of mutations are tolerated at residue 156, indicating that virtually no constraints, chemical or steric, exist at this position (Table I). The fact that Gly allows full activity indicates that there is no essential information regarding side chain identity or polypeptide backbone conformation. In addition to Gly and a conservative Asp mutation other single substitutions that have no effect on the DNA repair activity include hydrophobic (Ala, Val, and Ile) and positively charged (Arg and Lys) residues (Table I). Mutation to Gln and Leu occurs in multiple substitution mutants that also score as recA'.
Similar to residue 155 there appears to be very specific requirements that allow an exclusive effect on coprotease activity a t position 156. Mutation to either Lys or Arg has no effect on DNA repair activity, yet only the Lys substitution results in a strong coprt' mutant (Table IV, N-29). Lys is the only one of seven recA' single mutants which shows such an effect.
No single mutation was observed to inhibit RecA activities completely. For each of the four recA-/coprt-multiple mutants carrying a change at position 156 the phenotype can be attributed to one of the other mutations (Table 111, N-142, N-143,

GZY'~~--I~ addition to
G~u~~~ this is the only other one of the 8 targeted residues which shows some sensitivity to mutation. Only two single mutants were observed, Cys and Asp, both of which result in decreased DNA repair activity (Table 11, N-87 and N-88). Interestingly, both of these mutations resulted in constitutive coprotease activity (see below). The Gly157 + Asp mutant corresponds to recA1602, and our results match the previous characterization of this mutant as having partial DNA repair activity and moderate constitutive coprotease activity (11,12).
Unlike position 154 several substitutions at position 157 result in little or no inhibition of DNA repair activity. For example, although Gly157 -Ser was not obtained as a single mutant, the three multiple mutants with this change score either as recA' or show very high partial activity (Table I, N-52  and N-77; Table 11, N-106). Changes to either Trp or Arg are observed in two triple mutants that retain a significant amount I, N-6).
N-147, N-148).    N-117 and N-118). Also, substitution to Ala is seen in four multiple mutants, one of which retains a very high partial activity ( Table 11, N-1131, the others maintaining moderate levels of activity (Table 11, N-97, N-104, and N-105). Although we cannot say with certainty that single changes at position 157 to Ala, Trp, or Arg would not by themselves decrease DNA repair activity, these results suggest that position 157 tolerates at least a moderate level of mutation.

u t a n t T P K A E I E G E I G D S H M m u t a n t T P K A E I E G E I G D S H
In contrast to the DNA repair activity, mutations at position 157 had a significant effect on coprotease activity, frequently resulting in coprt' mutants. Of the 20 mutants in this study which contain a substitution at G~Y '~~, 10 are coprt' (Table IV), a result suggesting that this position is particularly sensitive to mutation-induced constitutive coprotease activity (see below).
The two multiple mutants that contain Gly'57 + Cys provide another example of suppressors that differentially alter defects in the DNA repair versus coprotease activities of the primary mutant (Table V, N-87, N-70, N-120; see below).
Gl~~~~-Position 158 supports a high level of mutation, and there appear to be very few constraints at this position regarding DNA repair activity. Single mutants that score as red ' include positively charged (Lys), uncharged polar (Gln and both recombinational DNA repair and LexA coprotease activities are described under "Experimental Procedures." Assays were repeated at least Sequences of the 39 red"-mutants observed in this study are shown . No red'" mutants were observed with 2 four substitutions . Assays for three times for each mutant . Asterisks indicate coprt' mutants (see Table IV) . Numbers in parentheses refer to the number of mutants with the indicated substitutions which arose independently, and for these mutants the scores for UV survival were averaged and include the standard error of the mean . Standard errors are not shown for all P-galactosidase measurements but approximate those shown for the positive and negative controls .  A m ) . and hydrophobic (Leu) residues (Table I) (Table I) . Even the large. aromatic residues Tyr and Phe no effect on DNA repair showed constitutive coprotease activpermit a moderate to high level of repair activity, although it y. The frequency of coprt' mutants with changes at this posithese changes occur in multiple mutants (Table 11. N-100 and tion and the varying chemical nature of the substitutions and no data are presented because all mutants showed no DNA repair or coprotease activity greater than the negative control. Numbers in parentheses refer to the number of mutants with the indicated substitution(s) which arose independently. resulting in this phenotype suggest that position 158, like 157, is sensitive to mutation-induced constitutive coprotease activity (see below). Our mutant N-30 ( G~u '~~ + Lys) corresponds to recA1219, which has also been characterized previously as recA+ with a strong constitutive coprotease activity (12).
Again, suppressor mutations were found which differentially effected the primary defects in the DNA repair or coprotease activities of position 158 mutants (Table V, see below).
IZe159-As for positions 152, 153, 155, and 156, position 159 supports a fairly high level of mutation with no effect on either RecA function. Single changes to Val, Met, and Phe, as well as substitution to Trp which occurs in a double mutant, result in recA+lcoprt+ mutants (Table I, N-34, N-35, N-36, N-65). A Tyr mutation occurs in a double mutant with very high partial DNA repair activity and wild type-like coprotease activity (Table 11, Several multiple mutants with changes at position 159 show no effect on DNA repair activity but result in coprt' mutants (Table IV, N-66, N-71, N-73, N-77). Another shows a significant decrease in DNA repair activity yet still gives rise to a low level constitutive coprotease activity (Table  IV, N-119). The frequency of coprt' mutants with changes at position 159 is lower than those with mutations at positions 157 and 158, and it may be that the coprt" phenotype of these multiple mutants is explained by the substitutions at positions other than 159 (see below). N-102).

' RecA Region L1
Thr appears to be a deleterious mutation as the double mutant Glu15'j + Val/Ile'59 + Thr has low partial activities (Table  11, N-1031, whereas the G~u '~~ + Val change by itself has no effect on RecA functions (Table I, N-24). Although we did not pick up as many substitutions at position 159 as for other residues in this study, we did not find any that correlate specifically with a r e d -or coprt-phenotype.
Mutations a t Nontargeted Residues-Although our procedure was designed to mutate two sets of four contiguous residues (152-155 and 156-159) we picked up several fortuitous mutations at nontargeted positions.
A Gly1'j0 -* Ala substitution completely inactivates both RecA functions (Table 111, N-1241, suggesting a n important role for the rotational flexibility of the polypeptide backbone at this position. However, we observed a double mutant with a substitution at position 160 which maintains a very low level of DNA repair activity (Table 11, N-107, G~u '~~ + Gly/Gly''jo -* Ser). In this case the backbone conformational flexibility lost as a result of the Gly1'j0 + Ser mutation may be partially compensated for by the substitution of Gly for Glu at position 158. Interestingly, this double mutant also displays a moderate constitutive coprotease activity that is further inducible by mitomycin C ( Table IV).
The only other nontargeted mutation that can be correlated with a recA-lcoprt-phenotype is Gly1'j5 + Ser (Table 111, N-133). This substitution occurs in a double mutant along with Ile'55 + Val, a mutation that by itself has no effect on RecA function (Table I, N-20). This result suggests that the flexibility of the polypeptide chain at position 165 is also important to both RecA activities.
An Asp"' + Asn substitution allows full RecAfunction (Table   I, N-37), whereas an Asp"jl -Val change still allows a moderate level of DNA repair activity and wild type-like coprotease activity (Table 11, N-89). This result indicates that not only is a formal negative charge not essential, but the general steric and chemical constraints are somewhat lax at position 161. A Ser"j2 + Phe recA+lcoprt+ mutant suggests that this position would likely support a high level of mutation (Table I, Substitution of His1'j3 with Leu results in a mutant RecA with a moderate level of DNA repair activity and no decrease in coprotease activity (Table 11, N-90) indicating that the wild type side chain a t this position is not essential for RecA function.
Finally, a Pro'51 -* Ser mutation in the recA'lcoprt' double mutant (Table I, N-40) suggests that the rotational constraints imposed on the polypeptide backbone by the wild type side chain at this position are not critical to RecA function.
MuZtipZe Substitution Mutants-For the most part the DNA repair phenotype of multiple mutants can be understood in terms of the corresponding single mutations. Several mutants, however, proved to be exceptions to this general trend. Based on the repair phenotypes of two r e d ' single mutants (LysI5' -* Gln and Ala'53 + Leu; Table I, N-3 and N-121, one might predict that the corresponding double mutant would be recA'. However, this double mutant (Table 11, N-92) shows a repair activity only marginally above the negative control. Likewise, the re& double mutant N-126 (Table 111) carries two substitutions ( L~S '~~ + Asn and Ala'53 + Val), which by themselves score as recA' (Table I, N-7 and N-11). These results suggest some deleterious interaction between the side chains a t positions 152 and 153 in the two double mutants which does not occur when only one of these positions carries a substitution.
A similar observation was made regarding the DNA repair phenotype of one particular triple mutant (N-116, Table 11) which carries three conservative changes, Glu15'j + Asp, G~u '~~ + Asp, and IleI5' + Leu. Based on the occurrence of these N-38).

TABLE rV
Coprt' mutants phenotype is indicated by fractional survival following exposure to W for 30 s. p-Galactosidase activity is a direct measure of the LexA coprotease Amino acid substitutions, LexA cleavage, and DNA repair activities are shown for the 29 coprt' mutants observed in this study. The repair activity and was determined as described under "Experimental Procedures." 6-Galactosidase and W survival assays were repeated at least three times for each mutant. Standard errors for all P-galactosidase assays are not shown but approximate those shown for the positive and negative controls.  substitutions in other recA' and recA+" mutants (see Table I, N-28, N-56, N-58, N-69, N-72, and Table 11, N-100, N-101, N-102, N-104), one might predict only a minimal effect on RecA function. However, N-116 shows a very low DNArepair activity. The coprotease activity is uneffected in this triple mutant.
A different result that supports our observation of an unexpectedly high tolerance for mutation at most positions in this region occurs in several triple mutants that carry combinations of hydrophobic substitutions. These include the following mutants with the changes indicated at positions 156,158, and 159, respectively: N-78 (Val, Leu, and Val), N-79 (Val, Leu, and Leu), N-81 (Gly, Val, and Val) and N-115 (Val, Phe, and Val). Whereas the wild type RecA sequence in this region contains two negative charges in a stretch of 5 residues ('5511e-Glu-Gly-Glu-Ile'5g), these four mutants create a run of hydrophobic residues which, despite the dramatic chemical and steric differences from wild type, support wild type-like DNA repair and coprotease functions.
The results in the sections above regarding the effects of mutations on DNA repair activity are summarized in Fig. 2.
No single substitution at positions 152,154, 159 or any of the nontargeted residues resulted in a coprt" mutant. Of the nine single mutants obtained at Lys15' (Tables I and 11) all showed inducible coprotease activity. At G~u~~~ all four of the single mutants have coprt phenotypes that correspond to the DNA repair phenotypes, three score as red-lcoprt- (Table 111, N-121, N-122, and N-1231, and the other is partially functional for both activities (Table 11, N-85). At position 159 all three single mutants score as recA'lcoprt' (Table I, N-34, N-35, and N-36).

Suppressor mutants
Amino acid substitutions, L e d cleavage, and DNA repair activities are shown for six sets of primary mutants and corresponding suppressor mutants. Second site suppressors were found for eight single mutants having alterations in DNArepair, Ledcoprotease, or both activities. Assays are described under "Experimental Procedures." Numbers in parentheses refer to the number of mutants with the indicated substitution which arose independently. Most of the coprt" mutants found in this study (21 of 29) contain multiple amino acid substitutions. Six of these 21 carry substitutions that by themselves result in a coprt' phenotype: N-70, N-93, N-96, N-98, N-110, and N-119. The other multiple mutants carry substitutions that by themselves score as coprt' or did not occur as single mutants. Although it is tempting to ascribe the coprt' phenotype to specific substitutions in these multiple mutants it may be that only the particular combination of changes is actually responsible for the phenotype.
To assess the likelihood that mutations that result in constitutive coprotease activity will occur at any one of the residues from 152 to 159 we plotted the &action of coprt' mutants observed as a function of the total number of mutants containing any substitution at each position (Fig. 3). This shows, for example, that 20 mutants contain a substitution at position 157, with 10 being coprt'; 44 contain a substitution at position 158, with 13 being coprt'; and 35 contain a substitution at position 152, with only 4 scoring as coprtc. These data suggest that coprt' mutants are more likely to result from substitution at positions 157 and 158 than other residues in this area. Because Fig. 3 includes multiple as well as single mutants it may actually underestimate the importance of positions 157 and 158 regarding the occurrence of coprt' mutations. Although analysis of single mutants would avoid this problem, we have only a limited number of single substitution mutants at certain positions. Despite this, we note that both single mutants at residue 157 and three of four single mutants at position 158 are coprt'. The difference between the mutant side chains, Cys uersus Asp 3.4f0.5 8.6k1. 7 1.00 1.6f0.2 1.61t0.2 0 at residue 157 and Lys, Gln and Asn at residue 158, suggests that there is little specificity required for a coprt' mutant at these positions, an observation that is in sharp contrast to positions 155 and 156 (see above). Studies of a larger pool of single mutants at all residues in this region will help clarify this issue. Second Site Suppressor Mutations-An advantage of using a combinatorial mutagenesis procedure is in the ability to pick up second site suppressor mutations. In this study we have found eight primary single mutants with alterations in DNA repair, coprotease, or both activities that are suppressed by one or more second site mutations (Table V). Anumber of these second site mutants show differential suppression of the repair uersus coprotease activities.
We found two examples of suppressor mutations that corrected severe defects in both the DNA repair and coprotease activities: 1) a secondary Ile155 + Lys change (N-43) corrects the defects in a Lyd5' + Glu mutant (N-82), and 2) a Ser16' -Phe change (N-55) corrects the defects caused by a G~u '~~ -Asp mutation (N-85). A Ser16' + Phe mutation by itself has no detectable effect on either activity (Table I, N-38). For the G~u '~~ + Asp mutation (N-85) we found another second site suppressor, Ile155 + Met (N-95), which corrects only the defective coprotease activity. The Ile'55 + Met change by itself results in a very strong constitutive coprotease activity with no effect on DNA repair (Table IV, N-19).
We also obtained several second site mutants that suppressed the constitutive coprotease activity of the original Substitutions that allow activity are listed above the wild type sequence; those that result in a r e d -phenotype are listed below. For those listed above, the roman font indicates full activity, and italics indicate partial activity. This summary is derived from all single mutants as well as those multiple mutants for listed.
which it is likely that the phenotype results from the single substitution decreases DNArepair by 50% but does not effect the coprotease function. Therefore, the Glu15* --j Lys change in mutant N-45 can also be viewed as a suppressor that specifically corrects a defect in DNA repair activity. Interestingly, the Lys15' -+ Ile change is the same second site mutation that Liu et al. (30) have recently characterized as a suppressor of the coprt' Gln's4 + Lys mutant, recA1202. 3) A third mutation at G~U "~ (Gln, N-31) results in a low level constitutive activity. We found three

RecA Region Ll 26319
different mutants that suppressed this activity, yet each had no effect on DNArepair (N-63, N-74, N-76). Two other suppressors also caused decreases in the DNA repair activity (N-105, The final examples of suppressor mutants in Table V again show different effects on the DNA repair versus coprotease activities. A Gly157 --f Cys substitution (N-87) results in a moderate coprte mutant and a 50% decrease in DNA repair activity.
A secondary + Lys change (N-70) corrects the DNA repair defect but also increases the level of constitutive coprotease activity. Another mutant that carries two secondary changes suppresses the constitutive coprotease activity and increases, but does not fully restore, the DNA repair activity N-117). (N-120).

DISCUSSION
By extensively mutagenizing an area of the RecA protein which overlaps the disordered L1 region we have identifed specific residues that play important roles in either the recombinational DNA repair or LexA coprotease activities. Considering the recombinational DNA repair activity, the identification of G~u '~~ as the most important residue in this area can be correlated with the position of the side chains for residues 152-156 in the RecA structure. Of these five residues, only G~u '~~ extends inward toward the helical axis of the RecA protein filament (Fig. 4). The remainder of the side chains extend either toward a neighboring monomer or away from the filament structure toward surrounding solvent. However, nearest neighbor analysis of all atoms in the GluI5* side chain (maximum interaction distance = 4 A) reveals no compelling structural reason for the rather strict constraints that we find at this position. It is interesting t o speculate, therefore, that the importance of G~u '~~ lies in its interaction with DNA, a constraint that is missing from the current RecA crystal structure (16).
Although there is good evidence that the primary DNA substrate binds deep within the helical groove of the RecA filament away from L1 (31), the positioning of the second DNA when bound to RecA is not yet known. Story et al. (16) have speculated that disordered region L2 (residues 196-209) makes up part of the primary DNAsite, and L1 (residues 157-164) makes up part of the secondary DNA site. In Fig. 4 the inner surface of the RecA oligomer is shown for three contiguous subunits. This image shows that although residues within region L2 can create a continuous surface that runs along the inside of the RecA filament, those within region L1 appear to create a continuous surface that runs along the upper part of the protein filament. If the side chain of G~u '~~ plays a role in binding the second DNA substrate, positioning of this DNA should be such that it is readily accessible to the primary DNA, and G~u '~~ might, therefore, be expected t o extend toward the helical axis of the protein filament. Because the binding of both DNA substrates to RecA likely occurs via the polyphosphate DNA backbone (321, Glu'" may either attenuate DNA binding through repulsive interactions with the DNA backbone or may potentiate binding via cation bridging. Another consideration regarding a DNA binding function for region L1 is that, other than G~u '~~, it may be the polypeptide backbone atoms that make the contacts with DNA, and the role of the other side chains has to do with positioning these contacts thereby making the identity of the amino acid less important. Most of the mutants with a substitution at position 154 (18 of 21) are completely inactivated for both repair and coprotease functions. However, the fact that the double mutant G~u '~~ 4 A~p A l e '~~ "-f Met ( Table 11, N-95) maintains wild type-like coprotease activity but shows a significant inhibition of DNA repair implies that the identity of the side chain at position 154 is not as important for coprotease as for recombinational DNA repair activity. Study of a larger population of single mutants at G~U '~~ w i l l help to resolve this issue. We performed a nearest neighbor analysis of all atoms in residues 152-156 and found that certain interactions that are seen in the RecA crystal structure are unlikely to be important determinants of function. For example the E-NH, group of Lys15z is within bonding distance of the Glu15'j side chain. However, both of these residues support a variety of substitutions, indicating that this is not a functionally important interaction. The GlulS6 side chain is also within H-bonding distance of Ty? on the neighboring subunit, but both positions support mutations that eliminate this interaction yet have no inhibitory effect on RecA function (15 and this study).
is within van der Waals distance of Phe2I7 on the neighboring subunit. Although the recombination activity of RecA shows a strict requirement for either Phe or Tyr at position 217 (E), supports a variety of substitutions, suggesting that this interaction is not essential to RecA function. The crystallographic B-factors are somewhat high for side chain atoms in residues 152 to 156 (range = 36-68), indicating a fair degree of uncertainty regarding their position in the structure. Regardless of the precise positioning, our mutagenesis data indicate that the identity of residue 154 is clearly more important to the recombinational DNA repair activity of RecA than the identity of the other 7 targeted residues.
Analysis of the LexA coprotease activity showed that 20% of all mutants (29 of 149) display a constitutive phenotype. At certain positions there is a strict specificity regarding the substitutions that give rise to coprt' mutants. For example, the noninduced and induced coprotease activities of the Glu16'j + Lys mutant are 22.4 and 25.7 (Table I) in contrast to 5.0 and 10.8 for the G W 6 + Arg mutant, demonstrating a dramatic specificity regarding the side chain requirement at this position for constitutive coprotease activity. A similar situation occurs at Ile'55, where substitution to Met is the only one of eight red' single mutants which effects coprotease function, resulting in strong constitutive activity. In contrast to the specificity observed at positions 155 and 156, a number of different substitutions at G~Y'~'I and G~u "~ give rise to coprt" mutants.
Yu and Egelman (33) recently performed electron microscopic studies of RecA.LexA.DNA complexes and suggest that region L1 is one of the main areas of contact between the RecA and L e d proteins. This raises the possibility that many of the mutations we have found effect coprotease activity by altering part of the LexA binding site. Met and Lys substitutions at residues 155 and 156, respectively, may create new higher affinity interactions with LexA repressor. However, because of their position at the subunit interface in the RecA filament, a specific conformational shift in the oligomeric structure may also contribute to the activity of these mutants. For GlylS7 and Glu15* an increase in the affinity for LexA by the removal of bad contacts may underlie the more frequent occurrence of coprt" mutants at these positions. This could be achieved by removal of the Glu side chain at position 158 or through a more general conformational mechanism resulting from any number of different substitutions at GIY'~~.
If the L1 region forms part of both the LexA and secondary DNA binding sites one might expect to find an equal number of mutations that disrupt one activity or the other. However, although we found a number of mutants that show an exclusive inhibitory effect on DNA repair, we found no mutants with such an effect on the coprotease activity. These results are consistent with the idea that determinants in the L1 region for DNA binding and catalysis of recombination are significantly more stringent than for LexA binding and repressor autodigestion.
Biochemical studies have been performed on purified mutant RecA proteins that contain either of two different mutations responsible for constitutive coprotease activity; + Lys oligomers as seen in the crystal structure (16). The central boxed area corresponds to the interfilament contact region defined by Story et al. (16 This region contains several residues at which mutations lead to a coprt' phenotype ( G W + Lys, T h P + Ile, Gln184 + Lys, or Gly301 -Asp a -j Cys mutation also result in coprtc mutants. These 3 residues are highlighted in all six subunits in each filament and are clearly distan Ser) or impose a temperature dependence on coprt' activity (IleZ9 "-f Val). Specific mutations at Ile'" and G~u '~~ (fecA441(3$-36), red730 (36), and recA1211 (37,38)) or GlnIs* -Lys (red1202 (37,38)). Although wild type RecA has a strong preference for single-stranded DNA and either dATP or ATP as cofactors for coprotease activity, both of these mutations create proteins that have a more relaxed specificity for NTP and nucleic acid cofactors. The crystal structure of RecA (16) shows that both the Glu3' and GlnIs4 side chains protrude outward from the helical RecA oligomer toward neighboring filaments, and therefore neither would be expected to play a direct role in ATP or DNA binding. However, Story et al. (16) proposed that the bundles of R e d filaments seen in the protein crystal are a mechanistically relevant structure and that mutation of either Glu3' and GldM could dismpt the formation of these inactive bundles, increasing the pool of active RecA filaments, thereby allowing the use of otherwise inappropriate cofactors. of two adjacent RecA filaments as seen in the crystal structure (16). The central boxed area is the site of contact between neighboring filaments and contains several residues at which mutations cause a coprtc phenotype (Glu3*, Thr3', Glr~'~*, Gly301) or impose a temperature dependence on coprtC activity (IleZs8) (16). However, other mutations that give rise to a coprtc phenotype do not lie within this area of interfilament contact (12,391. This is, in fact, the case for the coprt' mutants that we have identified in the L1 region (Fig. 5). Despite their distance from the i n~~~a m e n t contact region, these mutations may still cause a change in either the formation of inactive bundles of RecA ~Iaments or the oligomeric structure of the RecA filament itself. We are currently investigating the oligomeric properties of several purified coprt" RecA proteins.
In addition to the targeted residues 152-159 we picked up several fortuitous mutations at nontargeted residues which allow further insights into the importance of specific amino acids in this region. Although we have characterized far fewer changes at these positions, substitutions at Asp'", S e P , His'63, and Met1%, which cause partial or no decrease in either activity indicate that these wild type side chains are not critical to RecA function. Meah and Bryant (40) suggested that His1* is not absolutely required for RecA function because a purified Hisx* Ala mutant protein carries out DNA strand exchange and DNA-de~ndent NTP hydrolysis. Our results support their suggestion since we find that a functional r e d phenotype is not dependent on a His residue at position 163. Most mutants in this study which carry substitutions at Gly'60 are red-, suggesting that this residue may be critical for RecA activity. However, we did find one double mutant containing a G1yl6O -+ Ser change which still maintains a low level of activity. This is the position of the recA-mutant r e d 1 (Gly'60 -+ Asp). Previously, Bryant (41) showed that mutant RecAproteins carrying either an Asp or Asn substitution at position 160 catalyze some RecA activities in vitro, although both of these mutants are red-. We find that in uiuo RecA function is not entirely dependent on a Gly residue at position 160, and it may be that characterization of more substitutions would show this position to support some level of mutation. In summary, our results show that in the sequence defined by Pro'" to Met1'&, which includes all 9 residues in disordered region L1, G~u '~~ is most critical for R e d function, We have isolated a number of second site suppressor mutants that effect either or both the DNA repair and coprotease activities. The occurrence of a relatively large number of suppressors in a small, defined area suggests a functional relationship between residues in this region of the protein structure, supporting the idea that this area forms part of the L e d and secondary DNA binding sites. Also, the fact that some of these suppressors differentially effect the recombinational repair uersus coprotease activities reinforces the idea that these functions, although overlapping, are clearly separable.