Small-angle X-ray scattering reveals architecture and A₂B₂ stoichiometry of the UvrA–UvrB DNA damage sensor

Abstract

The UvrA–UvrB (AB) protein complex operates in the bacterial nucleotide excision repair pathway as the main sensor of DNA damage. Crystallographic analysis of the AB complex revealed a linear UvrB–UvrA–UvrA–UvrB arrangement of subunits with an internal two-fold axis that became incorporated into the crystal. Here, we have used small-angle X-ray scattering (SAXS) to show close correspondence between the crystal structure and the entity in solution. This result confirms the number and disposition of subunits in the crystallographic model and rules out other possible arrangements suggested by packing in the crystal. The current SAXS analysis failed to detect significant changes to the structure as a function of nucleotide. Proteins 2013. © 2012 Wiley Periodicals, Inc.

INTRODUCTION

The UvrA–UvrB (AB) protein complex serves as the DNA damage sensor in the bacterial nucleotide excision repair (NER) pathway (reviewed in Ref.1). The AB complex plays key roles in three important early stages of NER to identify and process DNA damage. In the first stage, the AB complex scans the genome for damaged DNA, with the task of discriminating native from damaged DNA performed primarily by UvrA; in stage 2, UvrA within the AB complex binds tightly to DNA around the lesion; and in stage 3, UvrA is evicted from the AB complex and UvrB takes its place at the lesion. A series of processing reactions then ensue that lead to the restoration of the original DNA sequence.

Damage to genomes from a wide variety of sources is a common occurrence. Left uncorrected, DNA damage can lead to permanent, disruptive, changes in the genome and can cause diseases in humans. Thus, all organisms deploy a battery of DNA repair pathways with overlapping specificities.2–4 One of these, the NER pathway, is unique in its ability to repair a diverse set of lesions that alter the structure of DNA in profoundly different ways. In eukaryotes, NER is the only pathway that is capable of removing bulky adducts.5–7 Mutations in this pathway can give rise to many human diseases including xeroderma pigmentosum, Cockayne syndrome, and trichothiodystrophy.8 Given the importance of NER to stability of the genome, it is crucial that its logic and molecular mechanisms be well understood. To this end, we are studying the NER pathway in bacteria, which displays mechanistic similarities to the corresponding eukaryotic pathway although executed by unrelated proteins.

A large body of thoughtful studies has revealed a great deal about the two main players in bacterial NER, UvrA and UvrB.1, 9–19 However, description of two aspects, the oligomeric state of the AB sensor and the nature of nucleotide-dependent structural changes, remains incomplete. Our recent crystallographic analysis of the AB sensor concluded that its composition is A₂B₂, with the two UvrB molecules located at the periphery of the complex ∼ 75 Å away from the expected site of the lesion. Based on the arrangement of subunits in the AB sensor, we proposed a mechanism by which two UvrB molecules are delivered to DNA, distant from the site of damage initially, only to be reduced to one copy when the UvrB–UvrC complex forms on the lesion.20, 21 As we noted,20 our results diverge from the long-standing view that the stoichiometry of the AB sensor is A₂B₁,1, 22, 23 a value obtained from densitometric analysis of fractions collected from velocity sedimentation glycerol gradient experiments.22 This stoichiometry estimate has had far-reaching implications for the mechanism of NER because it neatly explains the manner in which an asymmetric A₂B₁ ensemble could target an asymmetric lesion (lesions usually target one DNA strand) for processing.1 However, more recently, atomic force microscopy, fluorescence resonance energy transfer measurements, and biochemical studies challenged this view when they provided evidence for involvement of two UvrB molecules during the early stages of NER.21, 24, 25 Nevertheless, a reliable estimate of the composition of the AB sensor complex proved a challenge to obtain.22, 23

Nucleotide-dependent structural changes in UvrA and the AB complex is the second incompletely understood aspect of these two important players in NER. Crystallographic studies have revealed the architecture of the UvrA dimer and the organization of the two types of nucleotide binding sites (total of four) contained within. The current set of structures of UvrA shows both nucleotide binding sites that are either occupied by ADP or vacant,9, 10, 12, 23 except for a recent structure that displays one site filled with ADP and one site empty.20 Furthermore, although both types of sites of the UvrA dimer found in the crystal structure of the complete AB complex are filled with nucleotide, the conformation of the UvrA dimer within this complex is distinct from that in all other structures. Gaps in our current understanding leave the nature of ATP-dependent structural changes in UvrA to enlightened speculation.

A complete understanding of the composition of the AB sensor and its nucleotide-dependent dynamics is essential for achieving deeper insights into the mechanism of NER. In prior work, we described the X-ray crystal structure of the AB sensor. However, as the complete A₂B₂ sensor crystallized as part of the crystal, our analysis required that we designate one heterotetramer, from among several in the crystal, as the physiological entity.20 Here, we describe a small-angle X-ray scattering (SAXS) study of the AB complex and of isolated UvrA. SAXS (reviewed in Ref.26) is well suited to shed light on the issues raised above because it delivers low-resolution structural information that can be used to evaluate prospective models. Furthermore, as data are measured from molecules in solution, the resulting interpretation is uncomplicated by crystal packing considerations. The SAXS data show that the AB complex exhibits an elongated shape in solution. Furthermore, the SAXS-derived shape of the AB sensor in solution unambiguously points to one of the possible packing arrangements in the crystal as the relevant entity, as it rules out all others.20 This result has the important overall consequence of validating the disposition and number of UvrB molecules observed in the crystal structure. Our results argue against the presence of only a single UvrB molecule in the complex. The current SAXS data, however, do not reveal large changes in the structure of the AB complex or isolated UvrA as a function of nucleotide.

MATERIALS AND METHODS

Protein biochemistry

Geobacillus stearothermophilus (Bst) UvrA and UvrB were expressed and purified as described previously.11 UvrA was dialyzed into 20 mM Tris-HCl pH 7.5, 250 mM NaCl, 5% (v/v) glycerol, 5 mM tris(2-carboxyethyl)phosphine (TCEP), 5 mM MgCl₂, and 2 mM nucleotide (ADP or ATP). Following dialysis, the protein concentration was quantified by Bradford assay (Pierce Coomassie Plus Protein Assay). Samples for SAXS were prepared by diluting a concentrated stock to 1, 2, and 4 mg/mL using dialysis buffer. Aliquots of the dialysis buffer were collected to serve as blanks for the SAXS measurements.

The AB complex was formed by mixing UvrA with two-fold molar excess of UvrB in 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 5% (v/v) glycerol, 5 mM TCEP, 5 MgCl₂, and 2 mM nucleotide (ADP or ATPγS) and purified by Superdex 200 size exclusion chromatography (GE Healthcare), in the same buffer except that ATP was substituted for ATPγS. Peak fractions were pooled, concentrated, and quantified by Bradford assay. The samples were then diluted to 1, 2, and 4 mg/mL using gel filtration buffer. Aliquots of the column buffers were collected and used as blanks for the SAXS measurements.

SAXS data collection and analysis

SAXS data were measured at three different protein concentrations (1, 2, and 4 mg/mL) at the Bio-CAT beamline 18ID-D, Advanced Photon Source, Argonne National Laboratory. Immediately before each measurement, experimental samples (protein and the matched buffer blanks) were centrifuged at 20,000g for 10 min at 4°C to remove aggregates that might be present in the samples. A measurement series was initiated by first recording the scatter due to the empty flow cell, followed by that due to the buffer blank and the experimental protein sample. In all cases, the optical path of the X rays through the sample was 1.5 mm. To enable capture of the desired q (momentum transfer) range (0.006 < q < 0.345 Å⁻¹ where q = 4π sin θ/λ, 2θ is the scattering angle and λ is the X-ray wavelength), we set the sample-to-detector distance to 2361 mm. SAXS data were measured by exposing samples for 1 s; 15 such exposures were recorded. To minimize the extent of radiation damage during the measurement, a sample volume larger than that exposed to the beam was continuously pumped through the flow cell. The samples were maintained at 10°C during the course of the experiment. Following the experiment, we remeasured the concentration of protein in the samples and observed no changes due to the exposure to X rays.

SAXS data were processed with IGOR Pro (WaveMetrics) using scripts developed by the Bio-CAT staff and the ATSAS software package.27 Comparison of the scattering curves obtained from the 15 1-s exposures indicated that there was no significant radiation damage. Therefore, the scattering curves from the 15 measurements were averaged to improve signal-to-noise, and the resulting averaged curve was carried forward in our analysis.

Guinier analysis28 was performed using data where q•R_g < 1.3. A plot of ln I(q) against q2 yielded a straight line, from which the radius of gyration (R_g) and forward scattering intensities (q = 0), I(0), were extracted (Table I). At this stage, we noted a slight increase in R_g at the highest protein concentration (4 mg/mL), likely due to aggregation of the protein in the sample. Thus, further analyses were performed using data collected at lowest protein concentration (1 mg/mL) where this effect was not seen.

Table I. Forward Scattering Intensities, Radii of Gyration, and Maximal Particle Dimensions of UvrA–UvrB Complex and UvrA Determined from Guinier Analysis and Pair Distribution Function at Varying Protein Concentrations

Samples	Concentration (mg/mL)	I(0)a(Guinier)	Rg (Å)a(Guinier)	Qmax·Rga(Guinier)	I(0)b(P(r))	Rg (Å)b(P(r))	Dmax (Å)b(P(r))
a Calculated from Guinier analysis of small-angle data range in which q·Rg < 1.3. b Derived from pair distribution function.29 To minimize potential aggregation, only data collected from the lowest sample concentration were used in the analysis.
AB·ATP	1.00	9.533E-04	62.84	1.234	9.536 E −04	64.37	213.7
	2.00	1.742E-03	63.03	1.237
	4.00	3.957E-03	66.06	1.297
AB·ADP	1.00	7.895E-04	61.13	1.285	7.956 E −04	63.38	215.7
	2.00	1.847E-03	63.36	1.287
	4.00	3.560E-03	65.30	1.282
A·ATP	0.50	3.259E-04	45.15	1.247	7.210 E −04	45.96	163.7
	1.00	7.247E-04	46.10	1.273
	2.00	1.416E-03	47.53	1.269
A·ADP	1.00	6.950E-04	44.61	1.275	6.975 E −04	44.91	154.2
	2.00	1.463E-03	47.35	1.264
	4.00	2.874E-03	49.28	1.271

The pair distribution function was calculated with the program GNOM29 using SAXS data measured from the 1 mg/mL protein samples. This calculation requires an accurate estimate of the maximum particle dimension (D_max), which we determined using the automated program DATGNOM.27 The obtained D_max value was then used in the program GNOM with the scattering data systematically truncated at low and high q values. These calculations were performed while restraining P(r) to zero at r_max.

Ab initio bead model reconstructions were performed in DAMMIN in “FAST” mode without imposed symmetry (P1). DAMMIN represents the protein as an ensemble of beads or dummy atoms. Starting from a random assembly, DAMMIN uses simulated annealing to construct a model that gives rise to a scattering curve that best fits the experimental data. We repeated this procedure to obtain 50 independent models; these were superimposed,30 averaged, and evaluated using DAMAVER.31 Superposition of atomic coordinates from the X-ray crystal structures into the SAXS-derived model was performed using SUPCOMB.30 Reconstructions were also attempted using the “SLOW” mode of DAMMIN and DAMMIF.32 Despite the slightly improved chi values of the individual models, however, the resulting averaged models from these reconstructions were not as well connected as that obtained through use of the FAST mode. Also, the SLOW mode derived reconstructions showed poorer fits to the crystallographic model as judged by the normalized spatial discrepancy (NSD) calculated in SUPCOMB.

Scattering curves from atomic coordinates for comparisons to experimental data were calculated with the program CRYSOL.33R_g and D_max values from the various models for the AB sensor were calculated using CRYSOL and MOLEMAN2.34 Distribution functions for these structural models were calculated in GNOM using scattering curves computed in CRYSOL from input coordinates, and D_max values computed within MOLEMAN2.

To validate our data measurement and processing procedures, SAXS data were also collected on cytochrome C (4 mg/mL), a well-behaved protein standard. Analysis of cytochrome C data revealed that the experimental scattering curves and associated R_g are in good agreement with those calculated from the crystal structure (PDB code: 1HRC).

RESULTS AND DISCUSSION

The AB complex adopts a highly elongated shape in solution

To obtain insights about the shape of the AB DNA damage sensor in solution, we measured SAXS from preparations of the complex that had been purified away from isolated UvrA and UvrB. The data, recorded as 15 1-s exposures at three different protein concentrations: 1, 2, and 4 mg/mL [Fig. 1(A)], were of high quality as indicated by three measures. First, comparison of the 15 individual scattering curves revealed close juxtaposition with no outliers (not shown). This indicates that our samples experienced no radiation damage during the measurement. Second, a Guinier plot (ln I(q) against q²) revealed a straight line in the low-q region [Fig. 1(A), inset] for all three protein concentrations. Third, radius of gyration (R_g), a quantity extracted from the Guinier plot, revealed consistent values (62.84 Å, 63.03 Å, and 66.06 Å for the 1, 2, and 4 mg/mL measurements, respectively, see also Table I). These last two measures indicated that our preparations showed little aggregation and/or interparticle interference, especially at low protein concentrations. The increase in R_g with the 4 mg/mL measurement suggests that the sample might have aggregated or exhibited interparticle attraction at high concentration. Therefore, we have restricted our analysis to the data collected at the lowest protein concentration.

Figure 1. X-ray scattering curves, pair distribution function, and SAXS-derived molecular envelope of the AB complex. (A) Curves representing SAXS from the AB complex measured at three different protein concentrations. Inset: the corresponding Guinier plots (ln I(q) against q²). (B) Pair distribution function, P(r), calculated from the scattering data collected from 1 mg/mL sample concentration. (C) Close correspondence between the bead model calculated from the SAXS data and the UvrA₂–UvrB₂ crystal structure (Model #1, Fig. 2(A), the tetramer is generated by a crystallographic two-fold axis from the PDB 3UWX). Superposition of the UvrA₂–UvrB₂ crystal structure (Model #1)20 onto the SAXS-derived bead model shows good agreement. The SAXS-derived bead model is shown using semitransparent gray spheres with a radius of 11 Å. The AB crystal structure is shown using ribbon representation. The two UvrA protomers are shown in different shades of green, whereas the two UvrB molecules are shown in different shades of magenta. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 2. Evaluation of five potential AB models using SAXS. The left panel shows scattering curves computed from five potential models for the AB complex (red) in comparison to the experimental (gray) curve. The right panel depicts the P(r) curve calculated from the models (red) and experimental data (gray). The AB structural models are displayed in a surface representation with the two UvrA molecules colored in different shades of green, and the two UvrB molecules colored in different shades of magenta. R_{g coord.} and D_{max coord.} are calculated directly from atomic coordinates using the program MOLEMAN2.34R_{g calc. scatt.} is from pair distribution function29 calculated using scattering curves derived from coordinates.33D_{max calc. scatt.} is from the scattering curves calculated from coordinates using the program CRYSOL.33

An important overall conclusion from inspection of the scattering data is that the AB complex is a highly elongated entity in solution. This is evident from the pair distribution function, P(r), calculated from the experimental data [Fig. 1(B) and Ref.29], which in addition, provided an estimate of the longest chord (D_max) in the damage sensor: 213.7 Å.

We next used the SAXS data collected from the AB sensor, in combination with prior structural and biochemical data, to identify the conformation and composition of the sensor in solution. The issue of the precise identity of the physiological entity arose during determination of the crystal structure when we observed that the AB sensor complex crystallized with one AB heterodimer in the crystallographic asymmetric unit (PDB 3UWX).20 As it is well established that UvrA is a dimer in solution,22, 35 we concluded that the physiological entity must be composed of two copies of UvrA and two copies of UvrB. However, inspection of the packing in the crystal revealed four possible candidates for the biologically relevant entity. To identify the species in solution, we analyzed these four models, plus a fifth model that contained two copies of UvrA and one copy of UvrB (referenced by a large number of studies1, 22, 23) in the context of five parameters derived from the SAXS data (the scattering curve, the radius of gyration (R_g), the maximum particle dimension (D_max), the pair distribution function (P(r)), and the solvent envelope). The five models are shown in Figure 2. As noted previously, Model #1 displays a dimer configuration (termed the “closed groove” conformation) not observed in any previous structure of UvrA, which all resemble the “open tray” conformation.20 In this context, none of the other possible models suggested by packing in the crystal (Models #2, #3, and #4) contain the dimer seen in the open tray conformation. Also, consideration of A₂B₁ as a candidate (Model #5) was restricted to a species that contained the dimer observed in the closed groove configuration of UvrA.20

Comparison of the scattering curves calculated from the five candidate models of the AB complex reveals that only Model #1 agrees with the measured SAXS data [Fig. 2(A)]. Models #2, #3, and #4 display a poor fit over the entire resolution range [Fig. 2(B–D)]. Notably, the X-ray scattering calculated from the A₂B₁ model (Model #5, [Fig. 2(E)]) shows a poor fit at very low resolution; this feature is especially informative as this is the range that contains size information and where one would expect close agreement if this model represented AB complex in solution.

We obtained further support for Model #1 by consideration of other SAXS-derived quantities. Analysis of the linear Guinier region (q•R_g < 1.3) of the experimental scattering curve provided an estimate of the R_g of the AB complex in solution (62.84 Å). This value compares well with the value calculated from coordinates of Model #1 (64.2 Å). In contrast, the corresponding values for Models #3, #4, and #5 diverged from that obtained from the experiment (Table I and Fig. 2). We note that the R_g calculated from Model #2 is closer to the experimental value than those of Models #3, #4, and #5, thus potentially making a case for the physiological relevance of this model. However, Model #2 can be eliminated from consideration as it is organized around UvrA dimer contacts to a domain (the “insertion” domain) that we have shown previously to be dispensible for dimerization.11 Model #1 is also favored by comparison of the SAXS-derived D_max value against the value calculated from the coordinates (213.7 vs. 220 Å). Moreover, the large divergence of D_max values calculated from the other models eliminates them from consideration. In addition, analysis of the pair distribution function, P(r) from the various models, argues in favor of Model #1 as the physiological entity. As illustrated in Figure 2, only the pair distribution function calculated from Model #1 agrees with the curve calculated from the experimental data.

To further explore the elongated shape suggested by the one-dimensional SAXS data, we performed an ab initio calculation of the three-dimensional shape of the AB sensor from the scattering curve. This procedure, as implemented in the computer program DAMMIN, uses simulated annealing to modify a starting volume filled with dummy atoms, so as to maximize the agreement between a scattering curve calculated from this volume and that measured from the experimental sample.36 The starting volume is chosen to be a sphere, with a radius equal to one-half of the longest chord from the P(r) curve calculated from the data. We repeated this procedure 50 times to obtain independent models; these were superimposed, averaged, and filtered to the average volume. In this context, two types of reconstructions were performed, the first without imposition of symmetry (P1) and the second with imposition of two-fold symmetry (P2, the crystal structure of the A₂B₂ tetramer is two-fold symmetric). The models that emerged form the P1 and P2 calculations were very similar in appearance. We decided to focus our analysis on the P1-derived reconstruction as the AB sensor may not obey two-fold symmetry in solution. This P1-derived calculation produced an average volume with dimensions of approximately 80 × 80 × 210 Å [Fig. 1(C)]; the NSD between individual models used in averaging were in the range of 0.81–0.89 as calculated by SUPCOMB.30

Comparison of the volume that emerged from the procedure above, visually and quantitatively, points to close agreement with Model #1 and rules out other models [Fig. 1(C)]. The overall dimensions of the SAXS-derived envelope (80 × 80 × 210 Å) follow those seen in the crystal structure (80 × 90 × 210 Å). Superposition of the UvrA₂–UvrB₂ heterotetramer (Model #1) to the bead model revealed a good fit visually; this fit was characterized by a normalized special discrepancy (NSD) of 1.3 (Ref.30). We note that the SAXS-derived volume, which represents the hydrated volume of the complex in solution, is slightly smaller than that expected for UvrA₂–UvrB₂. In addition, upon superposition, some parts of the crystallographic model lie outside the SAXS volume. We speculate that this could be due to a combination of factors, including intrinsic flexibility of the AB sensor, partial dissociation of the complex in solution, errors in measurements and calculation, and the low resolution of the scattering data.

Taken together, the five experimental quantities extracted from the SAXS data establish that the shape of the AB sensor and the disposition of the UvrB molecules seen in the crystal structure (Model #1) reports faithfully on the species in solution.

Nucleotide-dependent changes in the AB damage sensor or UvrA were not observed by SAXS

Both UvrA and Uvr B are ATPases, and ATP plays crucial, but incompletely understood roles, in genome scanning, lesion recognition, and UvrB loading onto the lesion DNA.1, 37–41 To obtain insights into how nucleotide affects the conformation of UvrA and the AB complex, we measured SAXS data for isolated UvrA and the AB complex in the presence of ATP and ADP. Data were collected at three different protein concentrations (1, 2, and 4 mg/mL) in buffer containing either ATP or ADP. As illustrated in Figure 3, the scattering profiles of both UvrA and the AB complex in the presence of the two nucleotides showed small differences. At this time, we have chosen not to interpret these differences as significant because of our recent discovery that UvrA might be natively interconverting in solution between open and closed forms with quite different structures11, 20; interconversion is likely to be more pronounced when ATP is present, as it is in some of our measurements. Our two crystal structures of UvrA (isolated11 and part of the UvrAB complex20) give a hint as to the nature of these large structural changes. However, to understand the effects of nucleotide on the structure of UvrA, conformationally homogeneous preparations of UvrA with defined nucleotide state are needed. We are currently developing a procedure that will provide us with such samples using a combination of biochemical trapping and site-directed mutagenesis.

Figure 3. Nucleotide-dependent conformational changes of UvrA and AB complex were not observed by SAXS. SAXS curve and pair distribution function of AB complex (A and B) and isolated UvrA (C and D) measured from 1 mg/mL sample in the presence of ADP and ATP are very similar. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

While this manuscript was under review, a new crystal structure of Bacillus subtilis UvrB was reported (Webster et al. Nucleic Acids Res. 2012. Epub 2012/07/04. doi: 10.1093/nar/gks633. PubMed PMID: 22753105). The most notable feature of this new structure is the presence of a two-fold symmetric UvrB dimer in the crystallographic asymmetric unit (PDB 3V4R); biochemical methods established the solution relevance of this arrangement. We note that the architecture and stoichiometry of the UvrAB sensor (PDB 3UWX), as described above, are consistent with the new UvrB dimer structure. In this regard, we speculate that the arrangement of UvrB in the complete UvrAB sensor (PDB 3UWX) and that found in the new UvrB dimeric structure may represent different, sequential, stages of the damage recognition pathway.

Acknowledgements

The authors are grateful to Barak Akabayov and Sabine Akabayov for their advice on SAXS data processing; and Sabine Akabayov for sharing scripts used in systematic evaluation of pair distribution function. The authors thank the staffs at Bio-CAT especially Liang Guo for assistance during data collection and for advice on data processing. Bio-CAT is a National Institutes of Health-supported Research Center RR-08630. The content is solely the responsibility of the authors and does not necessarily reflect the official views of the National Center for Research Resources or the National Institutes of Health.