Regulation of Escherichia coli SOS mutagenesis by dimeric intrinsically disordered umuD gene products

Abstract

Products of the umuD gene in Escherichia coli play key roles in coordinating the switch from accurate DNA repair to mutagenic translesion DNA synthesis (TLS) during the SOS response to DNA damage. Homodimeric UmuD₂ is up-regulated 10-fold immediately after damage, after which slow autocleavage removes the N-terminal 24 amino acids of each UmuD. The remaining fragment, UmuD′₂, is required for mutagenic TLS. The small proteins UmuD₂ and UmuD′₂ make a large number of specific protein–protein contacts, including three of the five known E. coli DNA polymerases, parts of the replication machinery, and RecA recombinase. We show that, despite forming stable homodimers, UmuD₂ and UmuD′₂ have circular dichroism (CD) spectra with almost no α-helix or β-sheet signal at physiological concentrations in vitro. High protein concentrations, osmolytic crowding agents, and specific interactions with a partner protein can produce CD spectra that resemble the expected β-sheet signature. A lack of secondary structure in vitro is characteristic of intrinsically disordered proteins (IDPs), many of which act as regulators. A stable homodimer that lacks significant secondary structure is unusual but not unprecedented. Furthermore, previous single-cysteine cross-linking studies of UmuD₂ and UmuD′₂ show that they have a nonrandom structure at physiologically relevant concentrations in vitro. Our results offer insights into structural characteristics of relatively poorly understood IDPs and provide a model for how the umuD gene products can regulate diverse aspects of the bacterial SOS response.

The bacterial SOS response is a tightly regulated reaction to stress-induced DNA damage (1). It is temporally divided into an early, relatively accurate DNA repair phase and a later, more mutagenic damage-tolerance phase (2). This timing is regulated in part by products of the umuD gene. The initial product, UmuD₂, is a homodimer composed of 139 amino acid subunits that appears early after SOS induction (2). Damage-induced RecA:ssDNA nucleoprotein filaments mediate a slow autocleavage of UmuD₂ that is mechanistically similar to the inactivation of the LexA repressor (3). The N-terminal 24 amino acids of each subunit of UmuD₂ are removed, leaving a homodimer of the C-terminal 115 amino acid subunits, UmuD′₂ (3). UmuD′₂ activates UmuC, the catalytic subunit of the Y family DNA polymearse (Pol) V, for mutagenic translesion DNA synthesis (TLS) (4–6).

For such small proteins, UmuD₂ and UmuD′₂ make a remarkable number of specific protein–protein contacts, many, but not all, of them to DNA polymerases. Both proteins have been shown to interact with UmuC (5), DinB (the Y family DNA Pol IV) (7), and three subunits of the replicative DNA Pol III (8). Additionally, both interact with RecA:ssDNA nucleoprotein filaments (3, 9).

However, despite the nearly identical primary structures of UmuD₂ and UmuD′₂, their interactions with the same partner can differ in affinity and functional significance. Only UmuD₂ prevents DinB-induced -1 frame shifts (7), whereas only UmuD′₂ activates UmuC for TLS (4–6). UmuD₂ interacts preferentially with the β-processivity subunit of Pol III, whereas UmuD′₂ favors the α-catalytic subunit (8). RecA:ssDNA interacts with UmuD₂ to promote cleavage to UmuD′₂ (3), whereas UmuD′₂C requires trans RecA:ssDNA for efficient TLS (9). UmuD₂ may be degraded by either Lon (10) or ClpXP proteases (11), whereas UmuD′₂ must first exchange into the UmuD′D heterodimer to be degraded by ClpXP (12). The fact that such small proteins (no more than 30 kDa as dimers) can make so many specific protein–protein interactions is intriguing, and high-resolution structural studies were undertaken in an effort to find an explanation.

The x-ray (13) and NMR (14) structures of the cleaved form, UmuD′₂, offer some insight. Although both methods indicate that UmuD′₂ has an overall β-sheet fold, a detailed comparison between the two structures reveals substantial differences (14). The shape of the protein is less globular in the NMR structure (14), and the protease active site residues are only poised for catalysis in the x-ray structure (13). The structural differences suggest that UmuD′₂ may have considerable plasticity.

Although no high-resolution structural data are available for UmuD₂, we have recently proposed four energy-minimized symmetrical models of UmuD₂ (15). Previous single-cysteine studies of UmuD₂, which probed the structure of UmuD₂ in solution at physiologically relevant concentrations, have generally been consistent with our structural models. Single-cysteine derivatives of many amino acids that are predicted to be close to the dimer interface robustly cross-link to covalent dimers (16, 17). Interestingly, some positions that are predicted to be far away from the dimer interface also cross-link (16, 18). However, these residues can come together if the N-terminal arms are in an intermediate conformation (15). These results suggest that UmuD₂ may interchange among multiple conformations in solution.

We used circular dichroism (CD) spectroscopy to compare the secondary structure of UmuD₂ with that of the known β-sheet protein UmuD′₂, but we were surprised that, at physiological concentrations, both protein spectra resemble a random coil more than the expected β-sheet. These results are typical of intrinsically disordered proteins (IDPs), which have significantly less secondary or tertiary structure in vitro than other proteins (19–28). This class of proteins has been previously called natively denatured, natively unfolded, and intrinsically unstructured, among other names (21). Examples of these proteins include proteases, signaling factors, and other protein and nucleic acid-binding proteins (20, 24). IDPs often have important roles in regulation because of an ability to alter their precise structure, and therefore they function in response to changes in the cellular environment (25). Many of these proteins obtain more typical secondary structure in vivo, although some remain disordered (22, 23). The actual structure of an IDP is poorly understood, although a completely random and extended structure is probably rare (19, 27).

Unlike most previously characterized IDPs, UmuD₂ and UmuD′₂ form stable homodimers at a wide range of concentrations in vitro. Additionally, previous single-cysteine studies show that both proteins have a nonrandom structure at physiological concentrations (16–18, 29). Our results provide a rare opportunity to probe the actual structure of proteins that appear unfolded by CD spectroscopy.

Results

UmuD₂ and UmuD′₂ Have Extremely Different CD Spectra at μM and mM Concentrations.

As part of our effort to compare the undetermined structure of UmuD₂ to the known structure of its derivative UmuD′₂ (13, 14), we measured the CD spectrum of UmuD′₂ at 5 μM, which is the concentration found in SOS-induced cells (7). We were startled to discover that the CD spectrum of UmuD′₂ at the physiologically relevant concentration more closely resembles a random coil than the expected β-sheet (Fig. 1A). In an attempt to reconcile these results with the two previous high-resolution analyses of UmuD′₂, which had revealed it to be a β-sheet-rich protein (13, 14), we took the CD spectrum of UmuD′₂ at the high, nonphysiological protein concentration used to solve the NMR structure (14). Consistent with NMR and crystallography (13, 14), the CD spectrum of UmuD′₂ at 2 mM displays more typical β-sheet character (Fig. 1A). Examination of full-length UmuD₂ reveals the same striking anomaly, a CD spectrum resembling a random coil at 5 μM and one consistent with a β-sheet-rich protein at 2 mM (Fig. 1B).

Fig. 1.

CD and limited proteolysis of umuD gene products. (A and B) CD spectra of 5 μM (dashed line) and 2 mM (solid line) UmuD′₂ (A) or UmuD₂ (B) at 25°C. (C) Limited proteolysis of 5 μM UmuD′₂ (lanes 1–3) and 5 μM UmuD₂ (lanes 4–6) at 37°C for 5 min. Lanes 1 and 4, 5 μM protein with no protease; lanes 2 and 5, proteins preequilibrated at 10 μM and diluted 1:1 with 5 mg/ml chymotrypsin; lanes 3 and 6, proteins freshly diluted to 10 μM 1 min before 1:1 dilution with 5 mg/ml chymotrypsin.

We then examined the effect of dilution on the susceptibility of UmuD′₂ and UmuD₂ to limited proteolysis by chymotrypsin over a 5-min time window. Consistent with the CD results, UmuD′₂ or UmuD₂ that has been preequilibrated at 10 μM results in more complete proteolysis than UmuD′₂ or UmuD₂ that has been freshly diluted from a 2 mM stock (Fig. 1C). The extent of degradation of freshly diluted UmuD′₂ is ≈60% of that of preequilibrated UmuD′₂, whereas freshly diluted UmuD₂ is degraded at ≈85% the level of preequilibrated UmuD₂.

The CD and proteolysis results at physiological concentrations are typical of IDPs, which lack significant α-helix and β-sheet structure in vitro (19–28). We therefore used PONDR protein disorder-prediction programs (30) to test the similarity of UmuD′₂ and UmuD₂ to known disordered sequences. We found that the extreme N terminus of UmuD′₂ and much of the C-terminal regions of both UmuD′₂ and UmuD₂ are predicted to be disordered [supporting information (SI) Fig. 7]. Nevertheless, both proteins are active in vitro at physiologically relevant concentrations (SI Fig. 8).

Crowding Agents and Specific Protein–Protein Interactions Induce Secondary Structure in UmuD₂ and UmuD′₂.

To test whether the β-sheet-rich CD spectra of UmuD₂ and UmuD′₂ at mM concentrations result from specific self–self interactions or from more general crowding effects, we took the CD spectra of umuD gene products in the presence of the osmolytic crowding agent proline (31). Proline at 200 mM increases the secondary structure of both UmuD′₂ (Fig. 2A) and UmuD₂ (Fig. 2B). Less profound but consistent results are obtained with 2.5 M glucose (SI Fig. 9). Interestingly, other crowding agents such as PEG 8000, glycerol, and NaCl up to 1 M did not increase the secondary structure of UmuD₂ or UmuD′₂ (data not shown).

Fig. 2.

Induced secondary structure of umuD gene products. (A and B) CD spectra of UmuD′₂ (A) or UmuD₂ (B) in the absence (dashed line) or presence (solid line) of 200 mM proline. (C) CD spectrum of UmuD₂ alone (dashed line) or in the presence of DinB (solid line). (D) CD spectrum of UmuD₂ alone (dashed line) or in the presence of the β-subunit of Pol III (solid line). The CD signal of DinB or β alone was subtracted from that of the complex to obtain the spectrum of bound UmuD₂.

We have previously shown that UmuD₂ interacts with DinB (K_D = 0.64 μM) (7) and with the β-subunit of DNA Pol III (K_D = 5.5 μM) (15). To test whether these interactions induce secondary structure in UmuD₂ at μM concentrations, we took the CD spectrum of 50 μM UmuD₂ in the presence of 50 μM interacting protein. After subtracting the signal from the interacting protein alone, the resulting spectra of UmuD₂ in the presence of DinB (Fig. 2C) or of the β-subunit (Fig. 2D) reveal a nearly identical increase in β-sheet content. Because both the β-subunit and DinB have a more typical secondary structure than UmuD₂ (data not shown), it is likely that the increase in secondary structure of the complex is mostly due to an increased β-sheet content in UmuD₂. However, we cannot rule out the possibility that binding of UmuD₂ may cause a conformational change in the interacting proteins as well (7, 15).

UmuD₂ and UmuD′₂ Are Dimeric at Physiologically Relevant Concentrations.

Although the data in Figs. 1 and 2 would be consistent with a coupled folding and dimerization model, several lines of evidence are inconsistent with monomeric umuD gene products at physiological concentrations. Gel filtration of UmuD′₂ or UmuD₂ shows that their elution volume is between the expected size of a dimer and that of a trimer (SI Fig. 10A) (5). Extensive evidence suggests dimeric forms of UmuD₂, UmuD′₂, and UmuD′D (5), whereas no evidence for trimers has been found. Both UmuD₂ and UmuD′₂ elute slightly earlier than expected for a globular dimer, representing a Stokes' radius that is ≈12% greater than expected for a 25- to 30-kDa globular protein. This increase is more indicative of a molten globule form of UmuD₂ and UmuD′₂ than a fully unfolded protein (28, 32). Guanidinium-denatured UmuD′ and UmuD behave as monomers, eluting earlier than their native counterparts and just before denatured chymotrypsin (14 kDa) (SI Fig. 10B). In 6 M guanidinium, UmuD and UmuD′ have Stokes' radii ≈75% greater than expected for globular monomers, between the radius expected for a premolten globule and a fully unfolded random coil (28, 32).

Native PAGE of UmuD₂ and UmuD′₂ at 500 nM (20 μl) and 5 μM (2 μl) shows that the proteins are dimeric at both uninduced and SOS-induced physiological concentrations (SI Fig. 10C). The major UmuD₂ band runs nearly identically to a UmuD derivative, UmuD(F94C)₂, that has been covalently cross-linked in the dimeric form (33). An equimolar mixture of UmuD and UmuD′ at these concentrations shows a predominant intermediate band corresponding to the UmuD′D heterodimer, rather than two distinct monomeric bands. The theoretical pI of all of these proteins is 4.5, making charge effects negligible. A Ferguson plot of UmuD₂ and UmuD′₂ compared with native PAGE standards shows that both UmuD₂ and UmuD′₂ migrate most similarly to the 45-kDa size standard (Fig. 3), which is consistent with gel filtration and inconsistent with a monomeric form of UmuD or UmuD′ at physiological concentrations.

Fig. 3.

Ferguson plot of native PAGE size standards (gray circles), UmuD′₂ (filled square), and UmuD₂ (filled triangle) was produced as described previously (51). The best fit of the plot of −K_T versus molecular mass is to y = 7.3408x^−0.6868. R = 0.958. Solving for the molecular mass of UmuD gives an estimate of 46 kDa and for UmuD′ an estimate of 49 kDa. The difference is not statistically significant. Native gel standards are 545 kDa jack bean urease hexamer, 440 kDa equine spleen ferritin, 272 kDa jack bean urease trimer, 232 kDa bovine liver catalase, 140 kDa bovine heart lactate dehydrogenase, 66 kDa BSA, and 45 kDa chicken egg white ovalbumin. Where more than one data point is present, multiple protein isoforms were analyzed.

In an effort to determine the K_D of UmuD₂ and UmuD′₂ homodimers, equilibrium analytical ultracentrifugation was performed at three rotor speeds. The best fit of the data is to a single-species model (Fig. 4 A and C). The predicted molecular mass of UmuD′ at 20 μM is 25.4 kDa, compared with the monomer molecular mass of 12.5 kDa (Fig. 4A). The same model for UmuD at 40 μM (Fig. 4C) results in a fitted molecular mass of 31.0 kDa, in comparison with the predicted monomeric molecular mass of 15.1 kDa. If data are fit to a monomer–dimer equilibrium model, the K_D generated is infinitely low. Residuals, although somewhat nonrandom (Fig. 4 B and D), are small and do not improve with fits to other theoretical models. The lower limit of K_D determination for monomer–dimer equilibrium by using analytical ultracentrifugation is ≈10⁻¹¹ M (34). Thus, despite the CD spectra at low concentrations, both UmuD₂ and UmuD′₂ are dimers with K_Ds of <10 pM, which is in the range of the K_D of the related protein LexA (35).

Fig. 4.

Equilibrium analytical ultracentrifugation of UmuD′₂ and UmuD₂. (A and C) Results are shown for 20 μM UmuD′₂ (A) and 40 μM UmuD₂ (C). Data for three different speeds (16,000 rpm, filled circles; 20,000 rpm, dark gray circles; and 30,000 rpm, light gray circles) of Beckman Coulter rotor AN-50 Ti plotted with the best fit theoretical curve (single species of dimeric molecular mass) overlaid. (B) Residuals from data fitting to A. (D) Residuals from data fitting to C.

A Covalently Linked Variant of UmuD₂ Has a CD Spectrum Resembling a Random Coil.

To confirm that the random coil CD signal of UmuD₂ does not require a monomeric species, we analyzed the spectrum of disulfide cross-linked UmuD(F94C)₂, which binds the β-subunit of DNA Pol III in a similar manner to wild type (33). Surprisingly, although this variant cross-links nearly quantitatively (Fig. 5A), it shows slightly less propensity for secondary structure than an otherwise equivalent mock-treated sample of UmuD(F94C) (Fig. 5B).

Fig. 5.

Cross-linking does not constrain the secondary structure of UmuD₂. (A) Extent of cross-linking of UmuD(F94C). Lane 1, mock-treated UmuD(F94C), no reductant; lane 2, cross-linked UmuD(F94C), no reductant; lane 3, mock-treated UmuD(F94C), with 1 mM DTT; lane 4, cross-linked UmuD(F94C), with 1 mM DTT. Positions of molecular mass markers are to the left of the gel. (B) CD spectra of cross-linked (dashed line) or mock-treated (solid line) UmuD(F94C). (C) Native gel electrophoresis of physiological and high concentrations of umuD gene products. Lanes 1–3, 5 μM; lanes 5–9, 2 mM; lanes 1 and 5, UmuD₂; lanes 2 and 7, UmuD′₂; lanes 3 and 9, UmuD′D.

We have no evidence of stable higher order oligomers of UmuD₂ or UmuD′₂ at 2 mM, wherein the CD spectrum shows considerable secondary structure, and UmuD′₂ at this concentration has been shown to be dimeric (14). Native PAGE of 5 μM (20 μl) and 2 mM (0.5 μl) UmuD₂, UmuD′₂, and UmuD′D shows that all proteins have a consistent retention factor regardless of the starting concentration (Fig. 5C).

Discussion

These studies have led us to conclude that, at physiologically relevant concentrations, UmuD′₂ and UmuD₂ share structural characteristics with IDPs. Little is known about the precise structures of IDPs, although efforts to further characterize them have begun (28). In the case of UmuD₂ and UmuD′₂, a considerable amount of structural information is already available from solution studies at physiologically relevant concentrations (16–18, 29). Consistent with a flexible structure, cross-linking of single-cysteine derivatives of UmuD₂ by slow, gentle methods such as dialysis shows that most derivatives will cross-link to form covalent UmuD₂, with only a few positions that react much more or less than average (17). However, certain amino acid positions are consistently more solvent-exposed than others, and faster methods of cross-linking can better distinguish residues that are near the homodimer interface, suggesting that UmuD₂ is likely to have a flexible but nonrandom structure in solution (16–18). The high-resolution structures of UmuD′₂ (13, 14) both may have relevance to its structure in vivo, although inside the cell umuD gene products are likely to be surrounded by interaction partners that may influence their actual structure (Fig. 6A).

Fig. 6.

A model for sequential protein–protein interactions by IDPs. (A) UmuD₂ and UmuD′₂ make a variety of distinct protein–protein interactions. The relative binding affinities, if known, are represented by thick arrows (for strong interactions) or thin ones (for weak interactions). (B) Model for sequential protein–protein interactions with an IDP. An IDP may first bind to one interaction partner (1), which stabilizes a particular conformation. If a second binding interface becomes exposed in this conformation, another protein may now bind (2). The second binding event could destabilize the first protein–protein interaction, causing the original protein to exit the complex and possibly exposing a different interface. If so, a different partner (3) can bind at this site. (C) UmuD₂ and UmuD′₂ may act as interchangeable protein–protein interaction domains for E. coli Y family DNA polymerases. Y family DNA polymerases have conserved catalytic domains (large boxes), and many eukaryotic ones have extended interaction domains (lines on bottom three representations). Although these interaction domains are missing in the two E. coli Y family DNA polymerases, both of them interact with umuD gene products, which may serve as interchangeable protein–protein interaction domains in a streamlined bacterial genome.

UmuD₂ and UmuD′₂ share characteristics of hub proteins, which are represented in the interactomes of many organisms and make a large number of protein–protein contacts (36–38). Hub proteins have been found to have a larger degree of disorder than the general proteome (36–38), and those proteins that are relatively well ordered often have disordered binding partners (36). The high degree of disorder has been proposed as a mechanism to enable a large number of protein–protein interactions, especially transient interactions that are separated temporally or spatially (39). We suggest that the properties of IDPs might provide a simple mechanism for temporally ordering multiple protein interactions in hub proteins. An initial interaction may constrain the conformations of an IDP in such a manner as to expose a preferential binding interface for a second protein. After the second protein binds, the structure may change again to expose or occlude other binding interfaces (Fig. 6B). It is not known whether multiple interactions with umuD gene products occur simultaneously or in a stepwise fashion, although their role in timing regulation suggests that interactions may be transient.

The crystal structures of several DNA polymerase catalytic domains have been solved, but N-terminal or C-terminal protein–protein interaction domains are often removed to enable crystallization (40), possibly due to a tendency toward disorder in these regions. Although UmuC and DinB do not have these interaction domains, they both interact with disordered umuD gene products (5, 7). We suggest that, instead of being fused to a particular DNA polymerase, UmuD₂ and UmuD′₂ may act as interchangeable interaction domains for the two Y family DNA polymerases in Escherichia coli, thus allowing for a streamlined genome while maintaining the regulatory sensitivity of a disordered interaction module (Fig. 6C). A flexible structure that can adapt to multiple distinct protein–protein interactions helps explain how the small umuD gene products can make many specific interactions, and a posttranslational modification further differentiates these interactions (Fig. 6A).

Although IDPs are often involved in protein–protein contacts, few are known homodimers in solution. A stable quaternary structure in the absence of a rigid secondary structure is counterintuitive to the current protein-folding paradigm (41). However, limited examples are present in the literature, including the E. coli MazE antitoxin (42) and the human papillomavirus protein E7 (43). Both of these proteins have unfolded domains in addition to more rigid dimerization domains. Similarly, many of the residues in umuD gene products that are predicted to be ordered are at the dimer interface (SI Fig. 7).

Structural analysis of a dimeric protein can distinguish between a coupled folding-dimerization event and temporally separated folding and binding steps (44). Calculations for UmuD′₂ based on its NMR structure (14) obtained by using the program MOLMOL (45) show that UmuD′₂ has 75 Å of accessible surface area per residue and 22 Å of interface area per residue. These results suggest that the monomeric forms of UmuD and UmuD′, if they are ever present in solution, would be disordered and may undergo some disorder-to-order transition upon homodimerization (44). However, recent data show that a disorder-to-order transition is not necessary for homodimerization of one IDP, the T cell receptor ζ-subunit (46).

UmuD₂ and UmuD′₂ share homology with the dimerization domains of certain bacterial transcription factors (47). Transcription factors often have large regions of intrinsic disorder, either in their DNA-binding domains or in protein-interaction domains (48). The seemingly unrelated tendencies for transcription factors to be homodimeric and intrinsically disordered suggest that more homodimers with a large degree of structural plasticity may be found soon.

Materials and Methods

Materials.

RecA protein was purchased from New England Biolabs. High-molecular-weight native PAGE standards were obtained from GE Healthcare. Other protein standards and copper phenanthroline were purchased from Sigma–Aldrich. Native PAGE gels were obtained from Bio-Rad.

Protein Purification.

Purification of UmuD′₂, UmuD₂, and UmuD(F94C)₂ (49) and cross-linking (50) were performed as previously described. A plasmidencoding UmuD(F94C) was produced from pSG5 by using the Stratagene QuikChange site-directed mutagenesis kit (49). Protein concentration was determined by using the Bio-Rad protein assay. DinB was a kind gift from Daniel Jarosz (49). ClpXP and the β subunit of Pol III were generously provided by the Baker Laboratory at Massachusetts Institute of Technology (11) and the Beuning Laboratory at Northeastern University (15), respectively.

CD Spectroscopy.

CD was performed on an Aviv Model 202 spectrometer. Spectra were recorded at 25°C; each data point represents the average of 3 s of data collection. Proteins at physiological concentrations were monitored by using a 350-μl 0.1-cm cuvette (Hellma), and proteins at ≥50 μM were recorded by using a 4-μl 0.01-cm cuvette (Wilmad). Spectra of umuD gene products alone were recorded in buffer consisting of 10 mM Na₃PO₄ (pH 6.8), 100 mM NaCl, 0.1 mM EDTA, and 1 mM DTT. For interaction studies, the buffer was 50 mM Hepes (pH 7.4), 100 mM KCl, 5 mM MgCl₂, and 5% glycerol. The buffer spectrum was subtracted from that of the protein.

Limited Proteolysis.

UmuD′₂ or UmuD₂ was diluted to 10 μM in CD buffer and either incubated on ice for 2 h or used within 1 min. Proteolysis reactions were begun by adding 10 μl of 5 mg/ml chymotrypsin to 10 μl of UmuD′₂ or UmuD₂ and incubating at 37°C for 5 min. Reactions were stopped by addition of 4 μl of 6× SDS/PAGE-loading buffer [1× is 25 mM Tris·HCl (pH 6.8), 5% glycerol, 0.1% bromophenol blue, 2% SDS, and 1 mM DTT] and freezing in liquid nitrogen. Proteins were run on 4–20% Tris-glycine gels (Cambrex), stained with 1× SYPRO Orange (Molecular Probes) in 7.5% acetic acid, and quantified by using ImageQuant software.

Protein-Disorder Prediction.

Access to PONDR was provided by Molecular Kinetics. VL-XT is copyright 1999 by the Washington State University Research Foundation, all rights reserved. PONDR is copyright 2004 by Molecular Kinetics, all rights reserved.

Gel-Filtration Chromatography.

Gel filtration was performed by using a 100-ml Superdex 75 column on an Akta FPLC system (GE Healthcare). One milliliter of 5 μM protein solution was injected; UmuD₂ and UmuD′₂ were 5 μM at injection. The buffer described for CD spectroscopy was used as a running buffer. For denatured gel filtration, UmuD′₂, UmuD₂, and each size standard were denatured separately in CD buffer plus 6 M guanidinium hydrochloride for 2 h. Denatured samples were centrifuged for 1 min in a microcentrifuge at 16,000 × g to pellet aggregates prior to injection. CD buffer plus 6 M guanidinium hydrochloride was used for elution.

Native PAGE.

Proteins were diluted into 1× PAGE-loading buffer lacking SDS, incubated for 30 min at 25°C, and run at 20 V at 4°C overnight. Cross-linked UmuD(F94C)₂ was diluted into 1× PAGE-loading buffer lacking both SDS and DTT. Gels were soaked in 0.05% SDS for 30 min and stained with 1× SYPRO Orange (Molecular Probes) in 7.5% acetic acid after running. Ferguson plots were calculated as described by using 5 μM UmuD₂, UmuD′₂, and UmuD′D (51).

Sedimentation Equilibrium.

Experiments were performed on a model XL-I analytical ultracentrifuge with an AN-50 Ti rotor at 20°C (Beckman Coulter). Proteins were dialyzed against three changes of 500-ml CD buffer at 4°C over 12 h. The reference solution was the final dialysis buffer. Protein gradients were monitored by interference. Each rotor speed was centrifuged for 12 h, and WinMatch software was used to confirm equilibrium. Rotor speeds were 16,000, 20,000, and 30,000 rpm; the same protein samples experienced all three rotor speeds. Only the last scan for each speed was used in the data analysis. Protein concentration was determined by direct analysis of each sample after the last scan. Data analysis was performed by using the software WinNonlin.