A Non-cleavable UmuD Variant That Acts as a UmuD′ Mimic*

UmuD2 cleaves and removes its N-terminal 24 amino acids to form UmuD′2, which activates UmuC for its role in UV-induced mutagenesis in Escherichia coli. Cells with a non-cleavable UmuD exhibit essentially no UV-induced mutagenesis and are hypersensitive to killing by UV light. UmuD binds to the β processivity clamp (“β”) of the replicative DNA polymerase, pol III. A possible β-binding motif has been predicted in the same region of UmuD shown to be important for its interaction with β. We performed alanine-scanning mutagenesis of this motif (14TFPLF18) in UmuD and found that it has a moderate influence on UV-induced mutagenesis but is required for the cold-sensitive phenotype caused by elevated levels of wild-type UmuD and UmuC. Surprisingly, the wild-type and the β-binding motif variant bind to β with similar Kd values as determined by changes in tryptophan fluorescence. However, these data also imply that the single tryptophan in β is in strikingly different environments in the presence of the wild-type versus the variant UmuD proteins, suggesting a distinct change in some aspect of the interaction with little change in its strength. Despite the fact that this novel UmuD variant is non-cleavable, we find that cells harboring it display phenotypes more consistent with the cleaved form UmuD′, such as resistance to killing by UV light and failure to exhibit the cold-sensitive phenotype. Cross-linking and chemical modification experiments indicate that the N-terminal arms of the UmuD variant are less likely to be bound to the globular domain than those of the wild-type, which may be the mechanism by which this UmuD variant acts as a UmuD′ mimic.

The umuDC gene products are induced as part of the SOS response and are responsible for much of the UV-induced mutagenesis in Escherichia coli (1). These gene products are subject to an elaborate set of controls that regulate their activity (1). The LexA repressor provides transcriptional control, and there are several proteolytic controls on both the umuD and umuC gene products (1). The homodimeric protein UmuD 2 is the predominant species during the first about 20 -30 min after SOS induction (2). UmuD 2 , together with UmuC, plays a role in a DNA damage checkpoint, decreasing the rate of DNA synthesis and allowing time for accurate repair processes to act (2). This correlates with the cold-sensitive phenotype observed under conditions of over-expression of the umuDC gene products (2,3). As the SOS response proceeds, UmuD 2 binds the RecA/ssDNA 3 nucleoprotein filament. This stimulates the latent ability of UmuD 2 to convert to UmuDЈ 2 by cleaving off its N-terminal 24 amino acids, resulting in UmuDЈ 2 becoming the predominant species. The RecA/ssDNA nucleoprotein filament serves to bring together the active site dyad residues Ser 60 and Lys 97 , facilitating deprotonation of Ser 60 by Lys 97 (4). The activated Ser nucleophile then cleaves the peptide bond between Cys 24 and Gly 25 of UmuD 2 (1).
The wealth of structural data and models available for UmuD 2 and UmuDЈ 2 provide insight into how the two forms of the umuD gene products engage in multiple highly specific interactions ( Fig. 1) (4 -8), including with the ␣, ␤, and ⑀ subunits of the replicative polymerase, pol III (9). Of the two forms, UmuD 2 interacts more strongly with the ␤ processivity clamp (also referred to as ␤ or the ␤ clamp) than does UmuDЈ 2 (9,10). In full-length UmuD 2 , the 39-amino acid N-terminal arms are stably bound to the globular C-terminal domain (4,7) and form a distinct interaction surface. In the cleaved form of the protein, UmuDЈ 2 , the remaining approximately 15 amino acids at the N terminus appear unbound from the body of the protein and solvent-exposed (5,6), revealing the buried portion of the C-terminal globular domain (4,7). A series of truncations at the N-terminal arm of UmuD 2 indicates that the first eight amino acids of UmuD 2 are dispensable for the UmuD 2 -␤ interaction, whereas deleting residues 2-18 results in a substantial decrease in, but not a complete loss of, cross-linking efficiency with the ␤ clamp (10). Thus, the umuD gene products interact with the ␤ clamp via both the N-terminal arms of UmuD 2 and the globular domain of UmuD 2 and UmuDЈ 2 (10). This differential interaction appears to control, at least in part, whether the umuDC gene products act as part of a DNA damage checkpoint or as a translesion polymerase (9,10). These interactions with the ␤ clamp are of particular interest because sliding clamps play a key role in coordinating the multiple DNA polymerases present in cells (11)(12)(13)(14)(15). The eukaryotic DNA sliding clamp proliferating cell nuclear antigen interacts with multiple proteins. These interactions are in part regulated by covalent modification of proliferating cell nuclear antigen with monoubiquitin or the small ubiquitin-like modifier, (SUMO) which is distinct from the role of polyubiquitination in proteolytic degradation (16 -19). Duzen et al. (20) have suggested that UmuD 2 and UmuDЈ 2 play conceptually similar roles in modulating the various clamp interactions.
A version of the canonical ␤ clamp binding motif found in eubacterial polymerases as well as other proteins involved in DNA metabolism was postulated to be present in UmuD at residues 14 -18 ( Fig. 2) (21). A yeast two-hybrid experiment with the motif of UmuD showed, however, that these five amino acids are not sufficient for the interaction with the ␤ clamp (21). Given the fact that this result was obtained utilizing only the five-amino acid motif in UmuD, and cross-linking experiments showed that the region of UmuD between residues 9 and 19 is important * This work was supported in part by National Institutes of Health NCI Grant CA21615 , National Institutes of Health NIEHS Center Grant P30ES02109 from the MIT Center for Environmental Health Sciences, an American Cancer Society Professorship (to G. C. W.), and a postdoctoral fellowship from the Damon Runyon Cancer Research Foundation (to P. J. B.). Some of this work was performed under the auspices of the United States Department of Energy by the University of California, Lawrence Livermore National Laboratory Contract W-7405-Eng-48. 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 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1 and S2 and for interactions with the ␤ clamp (10), we undertook a site-directed mutagenesis analysis of this motif. These studies led to the unexpected discovery of a new class of UmuD variant proteins that fail to undergo cleavage but whose properties resemble those of the cleaved version, UmuDЈ.

MATERIALS AND METHODS
Homology Model of UmuD-The models of the UmuD homodimer were created by the combined use of the program LGA (22) for protein structure comparison and superposition and the AS2TS program (23) for homology model building. An initial model of the UmuD monomer (single chain) was constructed based on the crystal structure of UmuDЈ (Protein Data Bank code 1ay9, chain A) (24). The missing N-terminal arm was modeled by the LGA loop building/grafting procedure (22), using mainly the arm conformation in 1jhh_A, from the x-ray crystallographic structure of LexA (25) as well as the other LexA structures (1jhc, 1jhe, and 1jhf) as template structures to guide the local and overall conformation. In the final alignment (Fig. S1), UmuD Asp 20 had to be inserted into the LexA template (between residues 80 and 81 of LexA), and this was done by the LGA loop building procedure (22). Finally, residues 1-14 are in an extended conformation (i.e. we make no prediction as to the placement of these residues; they are only modeled in a formal sense).
In creating the full UmuD homodimeric complex, we used LGA to superimpose our monomer model onto each of the template chains in the NMR structure of the UmuDЈ homodimer (4) (1i4v, chains A and B, model 1), but some minor clashes occurred that were alleviated by following the LexA homodimer instead (25). This procedure creates a cis (non-domain swapped) conformation of the UmuD homodimer. Because there is a very small "shoulder" region at the top of the arms, the trans UmuD homodimer model could be constructed from the cis UmuD homodimer model by swapping the arms as follows: the first 39 residues in chain A of our trans model were taken from the chain B of the cis model, and vice versa. This process of "arm swapping" was completed after applying the LGA loop building procedure to residues 39 -41 in the shoulder regions. Finally, the LexA structures appear as both "elbows up" (N-terminal arm unbound) and "elbows down" (N-terminal arm bound to C-terminal domain), allowing us to model both conformations (Fig. S1). Thus, we created four models, two cis and two trans, each with an elbows up and an elbows down conformation (it is possible that heterogeneous conformations also occur with one elbow up and one elbow down as in the 1jhh LexA structure). For all the cis and trans models of the UmuD homodimer, the conformations of side chains from residues either not present in 1ay9_A or that presented a steric clash after building the dimeric structures were modeled using the side chain placement program SCWRL (26).
Proteins, Strains, and Plasmids-A plasmid expressing UmuD-3A was constructed in pSG5 using mutagenic primers and the QuikChange kit (Stratagene) (27). Wild-type UmuD and UmuD-3A were purified according to the published procedure (27). The plasmid expressing His-HMK-␤ was a gift from Prof. M. O'Donnell (Rockefeller University), and ␤ was purified according to the published procedure (28). The strains and plasmids used in this study are listed in Table 1. Plasmid pSJS9 was a gift from Prof. Charles McHenry (University of Colorado). Site-directed mutagenesis was performed using the QuikChange kit (Stratagene). Primer sequences are published in the supplementary materials.
Mutagenesis and Survival Assays-SOS mutagenesis assays were performed according to the published method (27). Briefly, cultures of GW8017 harboring various umuDC-expressing plasmids growing exponentially in LB were washed with 0.85% saline, exposed to 25 J/m 2 UV light from a germicidal lamp (General Electric), and then plated on M9 minimal plates with trace arginine (1 g/ml). Colony-forming units were scored after 48 h of growth at 42°C. Survival was determined by plating on M9 minimal plates with 40 g/ml arginine. Non-UV irradiated cultures were treated identically to assess the spontaneous mutation frequency. The data represent the average of at least three determinations.
UV survival curves were obtained after treating cells suspended in 0.85% saline in a Petri dish with the indicated doses of 254-nm light. Each sample was serially diluted and the dilutions were plated on M9 minimal media plates supplemented with 1% casamino acids, 0.005% tryptophan, and 1.5% agar. Plates were incubated overnight at 42°C.
Quantitative Transformation Assays-Transformation assays were performed essentially as described (27). Plasmids (0.1 g) were added to 25-50 l of competent AB1157 cells and incubated on ice for 10 min. After a 5-min heat shock at 37°C, and a further 10-min incubation on ice, transformation mixtures were allowed to recover in 750 l of LB at 37°C for 1.5 h with gentle shaking. Equal volumes were plated on LB plates containing the appropriate antibiotics for incubation under different temperatures as indicated in the figure legends.
Immunoblots-To determine UmuD expression levels, cells were harvested from exponentially growing cultures in LB, lysed by boiling for 15 min, and loaded on 4 -20% SDS-polyacrylamide gradient gels (Cambrex). Electrophoresed proteins were transferred to polyvinylidene difluoride membrane (Millipore) in 10 mM CAPS, pH 8, 10% methanol. After blocking, membranes were probed with anti-UmuD/DЈ, and antibody interactions were detected with SuperSignal substrate (Pierce). Antibodies to UmuD/DЈ were raised against purified UmuD in rabbits (Immunodynamics, Inc., La Jolla, CA). For UV-induced expression and cleavage experiments, an aliquot of about 2.5 ϫ 10 10 cells from an exponentially growing culture at A 600 ϭ 0.2-0.3 was harvested, washed in 0.85% saline, and UV-irradiated at 25 J/m 2 . Irradiated cells were then transferred to LB and grown at 37°C for the times indicated in the figure legends.
UmuD in Vitro Cleavage Assay-RecA/ssDNA nucleoprotein filament-facilitated UmuD cleavage was assayed (27,29,30) in LG buffer for 30 min. Reactions were quenched by addition of SDS-PAGE buffer to 1ϫ, and products were analyzed on 4 -12% gradient polyacrylamide gels. Alkaline cleavage of UmuD was carried out (27,31) in 100 mM glycine, pH 10, 10 mM CaCl 2 , 50 mM NaCl, 10 mM dithiothreitol, and 0.25 g/ml bovine serum albumin for 48 h at 37°C. Reaction products were analyzed by 14% polyacrylamide gel electrophoresis. The extent of UmuD cleavage was quantified using the Line Profile feature of the National Instruments Vision Assistant.
Cross-linking and Chemical Modification-Cross-linking was performed essentially as described (32) with bis-maleimidohexane (BMH, Pierce). Reactions were incubated at room temperature for the times indicated. For chemical modification with 5,5Ј-dithiobis(2-nitrobenzoate) (DTNB) (33), DTNB was dissolved at 2 mM final concentration in 50 mM HEPES, pH 7.5. Reactions were performed with 10 -20 M Fluorescence Determination of Binding Constants-Binding constants between UmuD and ␤ were determined essentially as described, with a PT1 QM-20000 -4SE spectrofluorimeter (Lawrenceville, NJ) (27). The ␤ clamp has a single Trp (residue 122), whereas UmuD has none. The ␤ 2 concentration was constant at 2.5 M. Emission from UmuD or UmuD-3A without ␤ 2 was subtracted from emission of the complex, and the center of the spectral mass was calculated for each [UmuD]. Excitation was at 278 nm, and emission was monitored from 300 to 400 nm. Excitation and emission path polarizers were oriented perpendicularly. The data represent the average of at least three independent experiments Ϯ 1 S.D.

Mutations in the "␤-Binding Motif " of UmuD Do Not Result in Com-
plete Loss of Induced Mutagenesis-We used alanine-scanning mutagenesis to make single alanine mutations in the putative ␤-binding motif in UmuD (Figs. 1 and 2) and investigated the consequences of these variants on known phenotypes of UmuD. These plasmid-borne variants were assayed for their ability to complement a ⌬umuDC null strain for UV-induced mutagenesis. In addition to single alanine variants of UmuD, we constructed one variant with alanines at the first and last two positions in the motif (UmuD-3A, Fig. 2), the positions most conserved among all ␤-binding motifs (21). In the case of UmuC, the analogous mutation in its ␤-binding motif results in a complete loss of UV-induced mutagenesis (11,12). No single mutation or set of multiple mutations in this motif in UmuD failed completely to complement a ⌬umuDC strain. However, plasmids expressing either the F18A UmuD variant, which is located at the top of the arm over the C-terminal globular domain, or the T14A/L17A/F18A ("UmuD-3A") variant resulted in substantial decreases in induced mutagenesis, down to about 15-20% of the wildtype (Fig. 2). Curiously, this decrease in mutagenesis of cells harboring these variants was not accompanied by a corresponding decrease in survival after UV irradiation (Fig. 5); yet, typically, increased mutagenesis due to translesion synthesis by UmuDЈ 2 C is associated with increased survival after treatment with UV.
Cleavage of the N-terminal 24 amino acids from the arm of UmuD to yield UmuDЈ is required to activate UmuC for its role in translesion synthesis (1). Because the UmuD arm harboring these mutations would be removed upon cleavage, we reasoned that the defect in induced mutagenesis of strains expressing the F18A and T14A/L17A/F18A (UmuD-3A) variant proteins might be because of defects in cleavage. Given their positions in the N-terminal arm (Fig. 1), it might be expected that these residues would play a role in properly positioning the arm in the active site for cleavage. We tested whether these mutations in the N-terminal arm of UmuD interfered with cleavage after UV exposure. The F15A and L17A mutants showed a slight decrease in cleavage and an ϳ1.5-fold decrease in induced mutagenesis compared with the wild-type (Figs. 2 and 3 and data not shown). The two UmuD variants (F18A and UmuD-3A) that showed essentially no cleavage up to 3 h after UV exposure (Fig. 3), or even after 14 h (data not shown), resulted in a substantial but not complete reduction in induced mutagenesis (about 15-20% of wild-type). This is in contrast to noncleavable UmuD active site variants that have been assayed previously, which showed essentially complete loss of induced mutagenesis (the limit of detection of this assay is about 1000-fold, or 0.1% of wild-type) (34,35). Thus, there are two groups of non-cleavable UmuD variants, one that renders cells partially mutable and the other that renders cells essentially non-mutable.
The wild-type and UmuD-3A variant proteins were purified to assess their efficiency in in vitro cleavage facilitated by the RecA/ssDNA nucleoprotein filament. Under these conditions, UmuD 2 is cleaved efficiently to form UmuDЈ 2 , whereas UmuD-3A 2 exhibits little detectable cleavage (Fig. 3). Here again, the lack of cleavage is similar to that exhibited by the active site mutant of UmuD 2 , UmuD-S60A 2 (34). We note that there is a lower band present in some of the samples incubated without the RecA/ssDNA nucleoprotein filament. This lower band is often observed in preparations of UmuD, and even in some preparations of UmuD-S60A. However, in the case of UmuD-3A, the intensity of the lower band does not increase after incubation with the RecA/ssDNA nucleoprotein filament.
UmuD forms exchangeable dimers (34), so wild-type UmuD 2 was combined with UmuD-3A 2 , and cleavage was observed (Fig. 3). Because UmuD-3A cannot cleave its own arm, the observed cleavage is likely because of the active site catalytic dyad of UmuD-3A acting on the arm of the wild-type partner, although the reverse is also possible. In this experiment it is also possible that the cleavage observed is due entirely to  a small population of wild-type UmuD 2 homodimers. To eliminate this possibility, UmuD-3A 2 was incubated with the active site variant UmuD-S60A 2 , and some cleavage was still detected (Fig. 3). This slight cleavage must be due to the active site residues of UmuD-3A cleaving the arm of UmuD-S60A, which suggests that the active site of UmuD-3A is proficient for cleavage and that the cleavage defect is isolated to its arm. The mutations in UmuD-3A at the top of the arm may disrupt folding of the arm over the globular domain or may interfere with specific protein-protein contacts required to facilitate cleavage (Fig. 1).
To ensure that the cleavage defect was not because of defective interactions of UmuD-3A 2 with the RecA/ssDNA nucleoprotein filament, we also carried out cleavage under alkaline conditions in the absence of the RecA/ssDNA nucleoprotein filament. The RecA/ssDNA nucleoprotein filament serves to facilitate deprotonation of Ser 60 by a neutral Lys 97 (4,25). In the absence of the RecA/ssDNA nucleoprotein filament, the activation of Ser 60 as a nucleophile can be accomplished under alkaline conditions. Under these conditions, UmuD 2 cleavage is inefficient but can be detected (31). We found that cleavage of UmuD-3A 2 was substantially decreased compared with that of the wild-type (Fig. 3). This suggests that the cleavage defect of UmuD-3A 2 is due to a defect intrinsic to the UmuD-3A 2 variant rather than deficient interactions with the RecA/ssDNA filament.
UmuD-3A Fails to Exhibit the Cold-sensitive Phenotype-Strains with elevated levels of the umuDC gene products exhibit a cold-sensitive phenotype that correlates with a DNA damage checkpoint (2, 3). Cells harboring plasmids overexpressing the cleavable umuD variants T14A and F15A (ϩwild-type umuC) were also cold-sensitive, whereas those overexpressing the L17A variant displayed an intermediate phenotype. The T14A and F15A variants behave similarly to wild-type in terms of their ability to exert the cold-sensitive phenotype, to be cleaved to UmuDЈ, and to act in UV-induced mutagenesis. The cold-sensitive phenotype is substantially enhanced in cells overexpressing the noncleavable variant UmuD-S60A (Fig. 4) (36). Thus, we were surprised to find that strains harboring plasmids expressing the noncleavable umuD arm variants (UmuD-F18A and UmuD-3A) failed to display this coldsensitive phenotype (Figs. 1 and 4).
We hypothesized that the loss of the cold-sensitive phenotype was due to the specific arm mutations of UmuD, regardless of their cleavage defect. To test this, we combined in single constructs either the UmuD-F18A or UmuD-3A arm mutations with the S60A mutation that renders UmuD catalytically inactive. Even though they are not cleavable, the arm mutations F18A and UmuD-3A alleviated the extreme cold sensitivity exhibited by strains with elevated levels of UmuD-S60A (Fig. 4). Although strains harboring plasmids overexpressing the UmuD F18A/ S60A double mutant display a cold-sensitive phenotype that is interme-  diate between that of cells with each corresponding single mutant, the cold-sensitive phenotype of cells overexpressing UmuD-S60A is suppressed by 2 orders of magnitude by the presence of only a single mutation in the N-terminal arm, F18A. These arm variants must disrupt a specific molecular interaction necessary to cause the cold-sensitive phenotype that is independent of whether they can be cleaved.
Simultaneously elevated levels of the umuD, umuC, and dnaN (which codes for the ␤ clamp) gene products cause a lethal phenotype, which has been interpreted as an exaggeration of the cold-sensitive phenotype (37). A strain harboring a plasmid expressing UmuD-3A and UmuC, when combined with high levels of the ␤ clamp, fails to exhibit the synthetic lethal phenotype (Table 2). This suggests that a critical aspect of this complex formation with ␤ is disrupted in the UmuD-3A variant.
Sensitivity to UV Exposure-Given the cleavage defect of the UmuD-3A and F18A N-terminal arm variants, we decided to look more closely at the curious lack of a correlation between UV mutability and survival after exposure to UV that we had noted earlier for cells expressing the F18A and UmuD-3A variants. It is known that E. coli strains harboring a noncleavable UmuD variant are hypersensitive to killing by UV light and are non-mutable (2, 34) (Fig. 5). To determine whether this is also true of the non-cleavable UmuD F18A and UmuD-3A variants, strains harboring plasmid-borne umuD variants were assayed for their resistance to UV light. Strains with plasmids expressing either of the non-cleavable variants F18A or UmuD-3A showed a similar level of resistance to UV light as those expressing wild-type UmuD or a synthetic construct of UmuDЈ (Fig. 5). We suspected that the resistance to killing by UV of the F18A and UmuD-3A variants was because of a specific feature of the arm mutants, unrelated to their cleavage defect. Cells with plasmids expressing the non-cleavable arm variants constructed in the context of the UmuD-S60A active site variant were assayed for their resistance to UV light. Strains with plasmids expressing either the F18A or UmuD-3A arm variants combined with UmuD-S60A exhibited a striking resistance to UV light that was similar to strains with wild-type UmuD (Fig. 5). This suggests that alterations in the N-terminal arm of UmuD are able to suppress the extreme UV-sensitive phenotype of non-cleavable UmuD-S60A, even though they are also non-cleavable. In light of the dramatic changes in the phenotypes of cells expressing the UmuD-3A variant compared with those with previously characterized non-cleavable variants of UmuD (34), we investigated the conforma-   tional consequences of the UmuD-3A variant compared with the wild-type. Chemical Cross-linking and Modification of UmuD Homodimers-We hypothesized that these non-cleavable UmuD variants are able to confer resistance to UV light, as well as to suppress the cold-sensitive phenotype, by mimicking the conformation of UmuDЈ. To examine this possibility, we analyzed the conformation of the N-terminal arm of the UmuD-3A variant compared with that of the wild-type UmuD. UmuD, which possesses a C 2 axis of symmetry, has a single Cys residue, Cys 24 , at the cleavage site in the N-terminal arm. To determine whether UmuD-3A is a UmuDЈ mimic with respect to the position of its arms, cross-linking was performed with the thiol-specific homobifunctional 16-Å cross-linker BMH. Our model allows us to put a lower limit of 20 Å on the distance between these two Cys thiols. This lower limit represents an implausible path for the cross-linker, because it is the direct distance between the two Cys thiols (Fig. 1). Thus, cross-linking should only be detected when the arms are "up," i.e. not bound to the C-terminal globular domain of UmuD. UmuD-3A was more readily crosslinked by BMH than either wild-type UmuD or UmuD-S60A (Fig. 6), suggesting that the arms of UmuD-3A are less likely to be bound to the globular domain of UmuD. Therefore they are more often close enough to be cross-linked.
One of the models of UmuD that we have proposed ("trans, elbows down," see Fig. 1 and supplemental Fig. S1) predicts that the thiol group of the single Cys 24 residue is partially buried under the peptide backbone of the N-terminal arm. However, if the UmuD arms are in an "up," or more flexible, conformation, then the Cys should be more accessible to a thiol-specific reagent. We performed a titration of the Cys residue at the Cys 24 -Gly 25 cleavage site with DTNB. The thiol moiety of UmuD-3A was more reactive to DTNB and therefore slightly more accessible than that of the wild-type UmuD (Fig. 6). We determined that there is 1.0 reactive Cys residue per wild-type UmuD 2 and 1.2 reactive Cys residues per UmuD-3A 2 , supporting the idea that in the UmuD-3A variant, the N-terminal arms undergo a shift in equilibrium to a less bound, arms-up state.
Determination of K d of UmuD and the ␤ Clamp-To quantify the binding of UmuD and the UmuD-3A variant to the ␤ clamp, we determined the K d for this interaction. Surprisingly, we found that although the K d is similar for ␤ binding to either wild-type UmuD (5.5 Ϯ 0.8 M) or UmuD-3A (6.1 Ϯ 0.5 M), the mode of binding is different for each protein. Namely, the fluorescence emission from the tryptophan in ␤ shifts to a longer wavelength upon binding to UmuD, whereas the shift is to a shorter wavelength in the presence of UmuD-3A (Fig. 7). Tryptophan fluorescence emission peaks at a longer wavelength in a polar environment and at a shorter wavelength in a hydrophobic one, indicating that the partially exposed tryptophan in ␤ (Fig. 7B) becomes more solvent exposed upon binding to wild-type UmuD and buried upon binding to UmuD-3A (38). Accordingly, unlike canonical ␤-binding motifs (11,12,39,40), this motif in UmuD is not responsible for the strength of the interaction with ␤, but rather for a qualitatively different mode of binding.

DISCUSSION
Although this work was initiated to determine the role of the putative ␤-binding motif ( 14 TFPLF 18 ) in UmuD function, we found that alterations in the motif do not prevent binding to the ␤ clamp, unlike corresponding mutations in UmuC, DinB, and the pol III ␣ subunit (11,12,21,39,40). Instead, we show here that the UmuD-3A variant alters the N-terminal arm conformation in a way that dramatically changes UmuD activity, and seems to exhibit properties of UmuDЈ, particularly with respect to resistance to killing by exposure to UV light and lack of the cold-sensitive phenotype when overexpressed together with UmuC.
The UmuD-3A variant shows no defect in survival but decreases UV mutagenesis. This suggests that this variant may allow selective bypass of T:T cyclobutane pyrimidine dimers but not (6 -4)-photoproducts, because lethality is associated with T:T cyclobutane pyrimidine dimers and mutagenesis with (6 -4)-photoproducts (1).
How could an uncleaved UmuD mimic the cleaved form, UmuDЈ? Current evidence suggests that the N-terminal arms of UmuD are usually bound to the C-terminal globular domain of the protein, i.e. it is usually in the elbows down conformation (7). Even when the arms are covalently bound to the globular domain, UmuD can be cross-linked to the ␤ clamp with almost no decrease in efficiency (10). When UmuD is cleaved to UmuDЈ, the remainder of the N-terminal arm (residues 25-39) is able to move relatively freely (5,24). The UmuD-3A variant  (20) versus the site of interaction of the canonical ␤-binding motif (green, residues 247 and 360) and the second site of interaction observed in the co-crystal structure with DinB (green, labeled second site; residues 93 and 98) (44,45). The single tryptophan, Trp 122 , is also indicated in yellow. Arrows indicate the dimer interface. The domains are labeled I, II, and III. This image was prepared with VMD (50). seems to have enough flexibility in its arms that it is at least a partial mimic of UmuDЈ. Although residues 14 -18 of UmuD are predicted from our model to be only partially buried (Fig. 1), the UmuD-3A variant may disrupt optimal packing of the arm against the globular domain. The mutations in UmuD-3A are at the point of the N-terminal arm that begins a downward turn over the C-terminal globular domain, and disruption of this turn is consistent with the more extended UmuDЈ arm structure. In addition, the cleavage of UmuD exposes a different surface of UmuDЈ for protein-protein interactions. Thus, the cleavage reaction serves the dual function of removing a portion of the N-terminal arms and presenting a dramatically different surface of the protein for interactions.
We were able to create models of the UmuD homodimer in both the elbows up and the elbows down conformation because both conformations are observed in the LexA structures (25). The LexA structure is in a cis conformation (non-domain swapped) with respect to the positioning of the arms (25). We have noticed, however, that in the UmuDЈ 2 structures the truncated arms point in the trans direction, suggesting that perhaps the trans conformation is actually preferred for UmuD (5,6). It has been shown that UmuD can undergo cleavage in trans ( Fig. 3) (5,41). The model of UmuD most consistent with the available biochemical evidence is one in which the arms are in the trans conformation (7).
In constructing the UmuD-3A variant, we have made a version of UmuD that binds to the ␤ clamp with a similar affinity as the wild-type, but with a subtle change in the specific interaction as evidenced by the strikingly different tryptophan fluorescence emission spectra. This change would not have been detected by many of the techniques commonly used to detect protein-protein interactions, such as co-immunoprecipitation or two-hybrid analysis. Recent evidence suggests that the domains of the sliding clamps are rigid bodies joined by flexible linker regions (42). The single tryptophan of ␤ is on a long flexible loop between rigid Domains I and II (Fig. 7) (43), so UmuD binding at a distal site to the tryptophan could cause a slight conformational rearrangement of the domains that alters the environment of the tryptophan in the loop. The distinct fluorescence spectra are indicative of at least a slight change in the conformation of the ␤ clamp when bound to either wild-type UmuD or UmuD-3A. UmuD-3A exhibits similar biological functions to those of UmuDЈ, yet UmuD-3A retains both the arm length and the strength of the interaction with ␤ of wild-type UmuD. This suggests that the change in conformation of the ␤ clamp, along with the relative conformational freedom of the UmuD-3A N-terminal arms, may be partially responsible for the UmuDЈ-like phenotype of UmuD-3A. UmuD and DinB/pol IV bind to ␤ at overlapping sites ( Fig. 7) (20,44). One of these sites is the hydrophobic channel between ␤ Domains II and III where all known ␤-binding motifs interact (44 -46). Although UmuD possesses a similar motif to a canonical ␤-binding motif that modulates its interaction with the ␤ clamp, and UmuD has been shown to bind to the same site on ␤ as other ␤-binding motifs, it does so in a way that is distinct from other proteins that bind ␤ via their ␤-binding motifs (46).
Processivity clamps play a critical role in controlling traffic at the replication fork and in cell cycle checkpoints. Polymerase binding to the ␤ clamp regulates access to the primer terminus by replicative or translesion synthesis DNA polymerases (13,14). Moreover, it has been shown that UmuD interacts more strongly with ␤ than UmuDЈ does (9). The interaction of UmuD with the ␤ clamp seems to be important for facilitating a DNA damage checkpoint in E. coli (2,3), and the cleavage of UmuD to UmuDЈ may attenuate this checkpoint function (10). The UmuD-3A variant seems to bypass this switch, yet still binds the ␤ clamp. In eukaryotes, covalent modification of proliferating cell nuclear antigen with monoubiquitin or small ubiquitin-like modifier (SUMO) determines whether the cell utilizes DNA repair or potentially mutagenic translesion synthesis (17). Thus, access to sliding clamps is universally important for control and regulation of proteins acting at the replication fork.