RAG1-DNA binding in V(D)J recombination. Specificity and DNA-induced conformational changes revealed by fluorescence and CD spectroscopy.

The RAG1 and RAG2 proteins together constitute the nuclease that initiates the assembly of immunoglobulin and T cell receptor genes in a reaction known as V(D)J recombination. RAG1 plays a central role in recognition of the recombination signal sequence (RSS) by the RAG1/2 complex. To investigate the parameters governing the RAG1-RSS interaction, the murine core RAG1 protein (amino acids 377-1008) fused to a short Strep tag has been purified to homogeneity from bacteria. The Strep-RAG1 (StrRAG1) protein exists as a dimer at a wide range of protein concentrations (25-500 nM) in the absence of DNA and binds with reasonably high affinity and specificity (apparent K(D) = 41 nM) to the RSS. Both electrophoretic mobility shift assays and polarization anisotropy experiments indicate that only a single StrRAG1-DNA species exists in solution. Anisotropy decay measured by frequency domain spectroscopy suggests that the complex contains a dimer of StrRAG1 bound to a single DNA molecule. Using measurements of protein intrinsic fluorescence and circular dichroism, we demonstrate that StrRAG1 undergoes a major conformational change upon binding the RSS. Steady-state fluorescence and acrylamide quenching studies reveal that this conformational change is associated with a repositioning of intrinsic protein fluorophores from a hydrophobic to a solvent-exposed environment. RSS-induced conformational changes of StrRAG1 may influence the interaction of RAG1 with RAG2 and synaptic complex formation.

The genes encoding the variable domains of immunoglobulins or T cell receptors are generated during lymphocyte differentiation by a somatic recombination reaction known as V(D)J recombination (1). The reaction is initiated by DNA double strand breaks created at the junction between two coding segments (termed V, D, or J) and their flanking recombination signal sequences (RSSs). 1 Cleavage is followed by a complex repair process that results in imprecise joining of the two coding segments and typically precise joining of the two RSSs. The RSSs consist of two conserved sequence elements, the heptamer (consensus 5Ј-CACAGTG-3Ј) and the nonamer (consensus 5Ј-ACAAAAACC-3Ј), separated by a poorly conserved spacer sequence of either 12 or 23 base pairs. Efficient recombination occurs only between a 12-RSS and a 23-RSS, a phenomenon known as the 12/23 rule (for review, see Ref. 2). Recognition of 12/23-RSSs and concerted cleavage at the RSScoding sequence border is performed by a complex of the RAG1 and RAG2 proteins (for reviews, see Ref. 3 and 4), the lymphoid-specific products of the recombination-activating genes RAG1 and RAG2 (5,6). Binding and cleavage of DNA by the RAG1⅐RAG2 complex is facilitated by the ubiquitously expressed architectural DNA-binding proteins HMG1 and -2 (7,8).
Deletion mutagenesis has established the minimal "core" domains of murine RAG1 (residues 384 -1008 of the 1040 aa RAG1 protein; Fig. 1A) and RAG2 (residues 1-383 of the 527 aa RAG2 protein) required for recombination activity in transfected nonlymphoid cell lines (9 -12). Most biochemical studies of V(D)J recombination have been performed with the core RAG proteins because of their enhanced solubility compared with the full-length proteins.
Surface plasmon resonance (13) and in vivo one-hybrid (14) experiments demonstrated specific recognition of the nonamer by the RAG1 nonamer binding domain and a less significant contribution of the heptamer to binding. In the one-hybrid experiments, RAG2 only modestly enhanced RAG1 DNA binding (14), implying that RAG1 by itself was capable of significant, sequence-specific DNA binding. Subsequently, numerous studies making use of electrophoretic mobility shift assays (EMSAs) and, in most cases RAG proteins fused to large Nterminal tags (glutathione S-transferase (GST) or maltosebinding protein (MBP)), led to the identification of a stable RAG1⅐RAG2⅐RSS complex termed the SC (for signal complex) (15)(16)(17)(18). These studies also indicated that RAG1 alone binds the RSS with low specificity and that RAG2 substantially enhances the affinity and specificity of the interaction. Such findings have led to the suggestion that the RAG1-RSS interaction is unlikely to be biologically significant in the absence of RAG2 (15). Only weak, nonspecific DNA binding by RAG2 alone has been reported (17), but recent evidence supports a direct interaction between RAG2 and DNA in the RAG1⅐RAG2⅐RSS complex (19). Footprinting experiments (15,20,21) reveal that RAG1 generates protection over the nonamer and the flanking portion of the spacer, whereas with both RAG proteins, protein-DNA interactions extend throughout the spacer and heptamer (21). All footprinting experiments utilized MBP-RAG1 core fusion proteins.
The stoichiometry of the RAG proteins in the SC is controversial. Some experiments indicated the presence of a dimer of RAG1 (18,22,23,24), whereas others argue for the presence of three or four RAG1 subunits (25,26). RAG2 is present in the SC either as a monomer (18) or as a dimer (23,26). The nonamer binding domain of RAG1 is thought to contact the nonamer, whereas a central domain of RAG1 (aa 528 -760) has been shown to display some specificity for the heptamer (27). In addition, a C-terminal domain of RAG1 has been shown to cross-link to the coding flank (28). Recent experiments indicate a division of labor between RAG1 monomers in the SC; one contacts the nonamer, whereas another engages the heptamer at the site of cleavage (25,29).
We currently do not know the equilibrium constant that governs the association of RAG1 monomers into dimers. Furthermore, since most DNA binding studies have been performed by EMSA, we do not know the equilibrium constants governing reactions involving RAG1 alone or RAG1 in conjunction with RAG2 and HMG proteins in forming complexes with the RSS in solution. The equilibrium dissociation binding constant for the interaction of a MBP-RAG1 core (aa 384 -1008) fusion protein with 12-or 23-RSS oligonucleotides was determined by EMSA to be close to 100 nM (22). In this study, however, the MBP-RAG1 fusion protein was capable of forming several protein-DNA complexes with various protein stoichiometries and various DNA binding affinities, and the determined dissociation constant reflected the overall binding without specification of the individual components of reaction.
We have now used fluorescence spectroscopy and a RAG1 protein lacking a bulky tag to study the RAG1-RSS interaction in solution, thereby overcoming some of the limitations inherent in the EMSA methodology and any complications caused by a large fusion partner. We find that the RAG1-RSS interaction is of higher affinity and specificity than previously reported. In addition, we document a substantial conformational change in RAG1 induced by interaction with the RSS. Based on these findings, we propose that the RAG1-RSS interaction may play an important role in vivo, especially in light of the observation that for a portion of the cell cycle, developing lymphocytes express RAG1 in the virtual absence of RAG2 protein (30).

EXPERIMENTAL PROCEDURES
DNA Oligonucleotides-Unlabeled and fluorescently labeled deoxyoligonucleotides were synthesized and high performance liquid chromatography-purified by Integrated DNA Technologies Inc. (Coralville, IA). The 12-RSS substrate was made by annealing (top strand) 5Ј-GTCGA-CCACAGTGCTACAGACTGGAACAAAAACCCTGCAG-3Ј with its complement. The nonspecific DNA substrate was made by annealing (top strand) 5Ј-GTCGACTGGCCATCTACAGACTGGAGCGGCCGCACTG-CAG-3Ј with its complement. The fluorescein fluorophore was attached at the 5Ј ends of the top strands using a 5Ј-fluorescein phosphoramidite that attaches 6-carboxyfluorescein to the oligonucleotide via a C6 chain linker. Annealing of oligonucleotides to generate double-stranded DNA was performed in binding buffer (BB; 10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 5 mM MgCl 2 ) by heating the complementary oligonucleotides mixed in equimolar amounts for 5 min at 95°C followed by slow cooling to room temperature.
Expression and Purification of Strep-RAG1 (StrRAG1)-Murine core RAG1 was expressed as a Strep-tag II fusion in the context of the pASK-IBA5 vector (Sigma Genosys). The catalytic core of murine RAG1 from aa 377-1008 was amplified using primers that created an NdeI site at the 5Ј end of the gene and a XhoI site at the 3Ј end. These PCR products were subcloned into the corresponding sites in pASK-IBA5, placing the Strep tag at the N terminus of the protein.
The fusion protein was expressed in the BL21 strain of Escherichia coli. Expression cultures were started by adding 20 ml of overnight culture per 1 liter final culture volume and grown at 25°C with rapid shaking (Ͼ200 rpm) until the A 600 was ϳ1. Expression was induced by the addition of anhydrotetracycline to a final concentration of ϳ1 M and ZnCl 2 to a final concentration of 0.1 mM. Cultures were grown for an additional 12-16 h, at which point the A 600 was usually ϳ5 absorbance units. Pelleted bacteria were resuspended in ice-cold lysis buffer (20 mM Tris (pH 8.0), 100 mM NaCl, 0.5 mM EDTA, 5 mM 2-mercaptoethanol, 2 M ZnCl 2 , 10% glycerol, and protease inhibitor mixture (phenylmethylsulfonyl fluoride, pepstatin A, aprotinin, leupeptin). Cells were lysed by sonication on ice, and debris and insoluble material were pelleted by ultracentrifugation at 26,000 rpm for 1 h at 4°C in a SW-41 rotor (Beckman). Cleared lysate was added to Fast-Flow Q-Sepharose (Amersham Biosciences) 50 -75-ml bed volume column preequilibrated in the lysis buffer before the addition of lysate. After loading, the column was washed with two bed volumes of lysis buffer containing 200 mM NaCl, and RAG1 was eluted with 3 bed volumes of lysis buffer containing 500 mM NaCl. Q-Sepharose elution fractions were applied to a 7-10-ml bed volume Streptactin-Sepharose column (Sigma Genosys) pre-equilibrated in lysis buffer containing 500 mM NaCl, and after washing with lysis buffer containing 500 mM salt, Strep-RAG1 was eluted in lysis buffer (500 mM salt) containing 2.5 mM desthiobiotin. The eluted Strep-RAG1 protein was typically Ͼ90 -95% pure. Aggregated Strep-RAG1 was separated from dimers and other oligomers using a Superdex 200HR column (Amersham Biosciences). Fractions containing the dimeric protein were dialyzed against BB. MBP-RAG1 core (aa 384 -1008) and glutathione S-transferase-RAG2 (aa 1-383) protein expression and purification were performed as described in (22,31).
Coupled Cleavage Assay-Coupled cleavage reactions (50-l final volume) contained 10 ng of body-labeled substrate containing both a 12-RSS and a 23-RSS, 100 ng of each RAG protein, 30 ng of HMG2, and 5 mM MgCl 2 and were incubated for 2 h at 37°C using buffer conditions described previously (32). Reaction products were resolved on a 4% native Tris borate EDTA polyacrylamide gel.
EMSA-To create the EMSA probe, 2 M 12-RSS double-stranded oligonucleotide DNA was 5Ј end-labeled using T4 polynucleotide kinase (New England Biolabs) and [␥-32 P] ATP (3000Ci/mmol) (PerkinElmer Life Sciences) and purified on a 5% native polyacrylamide gel. StrRAG1 or MBP-RAG1 was incubated with radioactively labeled 12-RSS oligonucleotide in the presence or absence of unlabeled 12-RSS or nonspecific DNA in BB supplemented with 10 ng/ml heparin and 0.5% glycerol at 25°C for 20 min, and samples were resolved on 4% native polyacrylamide gels in 0.5ϫ Tris borate (45 mM Tris (pH 8.9), 45 mM borate) at room temperature. For better separation of the multimeric species formed by MBP-RAG1, a discontinuous 3.5-8% polyacrylamide gel was used. Gels were dried and quantitated using a PhosphorImager (Molecular Dynamics). The fraction bound at saturation, maximal fraction bound, F MAX ϭ 1, is set constant, whereas we expressed the normalized fraction bound as the ratio between the fraction bound at each data set and the actual constant value of the fraction bound at saturation. The data were fit to a Hill binding isotherm with n ϭ 2.
Steady-state Fluorescence Measurements-The StrRAG1 intrinsic emission fluorescence spectra were recorded using a SLM-8000 L format spectrofluorometer equipped with a 750-watt Xenon arc lamp. All fluorescence measurements were performed in 100-l quartz cuvettes (Starna, Atascadero, CA) at 25°C constant temperature maintained by a circulating water bath. The StrRAG1 intrinsic emission fluorescence spectra were recorded with an excitation wavelength of 280 nm using an 8-nm band pass for both the monochromator and the emission long pass. All fluorescence emission spectra were recorded between 300 and 420 nm using 1-nm steps and 2-s integration times. Dilution of StrRAG1 samples was done in ice-cold BB buffer filtered twice through 0.2-m sterile filters (Gelman Laboratory) followed by a 10-min incubation at 25°C. For static experiments, mixing of StrRAG1 with DNA or acrylamide was done inside the quartz cuvette followed by a 10-min incubation at 25°C constant temperature before recording. Each spectrum was corrected (33) for background, photobleaching (5-15%), and dilution relative to a control sample for which the same amount of protein was incubated in parallel with a corresponding volume of buffer. Photobleaching was always reversible, indicating that no photo products are generated in the reaction. The fluorescence emission intensity of all spectra were expressed as a ratio relative to a standard rhodamine reference excited with the same source and whose emission output was a constant Ϸ46,500. If sample absorption at excitation wavelength exceeded 0.001, inner filter corrections were also applied according to Lakowicz (34). For the time course experiment shown in Fig. 7b, DNA was mixed with protein rapidly with a long capillary tip with the cuvette in the holder and an automatic shutter opened immediately for recording. The emission at 327 nm was recorded at 1-s intervals.
Steady-state Fluorescence Polarization Measurements-Polarization anisotropy measurements were performed on a PTI C-61 T-format fluorometer equipped with UV-transmitting Glan-Thompson plane polarizers and a circulating water bath to control the cell temperature. All measurements were performed using the same quartz cuvette used for emission scans and temperature set at 25°C. For tryptophan fluorescence, the samples were excited at 295 nm, and emission was collected at 327 nm, whereas for fluorescein the excitation wavelength was set at 492 nm, and emission was recorded at 520 nm. The recorded intensities were corrected for variations in lamp intensity, detector, and monochromator wavelength-dependent sensitivity. G-factor corrections were included in all calculated anisotropy values. For the time course experiment shown in Fig. 7a, fluorescence intensities were recorded at 1-s intervals. A 1-ml quartz cuvette with a round bottom was used to allow magnetic stirring. To the solution containing 50 nM 12-RSS fluoresceinlabeled DNA, either a concentrated StrRAG1 solution or BB buffer supplemented with 50% glycerol (control reaction) was added in a volume less than 1/20 of the total sample, without mixing. The unmixed components were incubated together for 100 s, and then mixing was initiated by rapid stirring.
Circular Dichroism Measurements-Circular dichroism spectra either of DNA or DNA incubated with StrRAG1 was recorded using a JASCO J715 spectropolarimeter. The buffer used for incubation was 10 mM Tris-Cl (pH 7.5), 10 mM NaCl, 40 mM KCl, 5 mM MgCl 2 . The spectra were recorded at 25°C between 200 and 300 nm. For analysis of the spectra of DNA and StrRAG1 mixtures, the separately recorded DNA spectra were subtracted, and then the spectra were converted to molar ellipticity based upon the molar concentration of StrRAG1 present in each sample.
Time-resolved Fluorescence Measurements-Time-resolved fluorescence intensity and anisotropy decay kinetics were measured in the frequency domain using an SLM-AMINCO 48000 multiharmonic Fourier (MHF) spectrofluorometer (Spectronic Instruments). An argon-ion laser (Coherent, Innova 90-6) operating at 488.0 nm was used as the excitation source. An interference filter (Oriel) was placed in the excitation beam path to minimize extraneous plasma tube super radiance from reaching the detection system. The sample fluorescence was monitored in the typical L-format after passing through a 515-nm longpass filter. The Pockels cell modulator was operated at a 5 MHz base repetition rate. Typically, data were acquired for 60 s between 5 and 150 MHz (30 total frequencies), and at least 10 discrete multifrequency data sets were acquired for a given sample. For the excited-state intensity decay measurements, we used a dilute solution of rhodamine 6G dissolved in water as the reference lifetime standard (3.85 ns). Magic angle polarization conditions were used for all excited-state intensity decay measurements to eliminate bias arising from fluorophore rotational reorientation. The excited-state fluorescence lifetimes and rotational reorientation times were recovered from the frequency-domain data using a commercially available non-linear least squares software package (Globals Unlimited) wherein the true uncertainty in each datum was used as the frequency-dependent weighting factor. The data were fitted for single double and triple exponential decay with limiting anisotropy set nonvariable to 0.4 (35), but only the double exponential fitting gave more than 98% correlation with our recorded data points and confident 2 values between 0.05-0.2.
Fluorescence Data Analysis-For every set of anisotropy data, a minimum of three experiments was performed under identical conditions, and the data points shown are average values. Each averaged data point was considered only if the sample S.D. was less than 5-7% with respect to the calculated mean. The fitting of averaged data points was performed using a typical Hill binding isotherm (36), where r obs is the observed anisotropy of the reporter species, r b is the basal anisotropy of the reporter species in the absence of ligand, r sat is the maximal anisotropy reached at saturation, [L] is the free ligand concentration varied in the reaction, expressed as monomeric species, n is the Hill coefficient expressing the degree of cooperative ligand binding to the reporter species, and K D is the apparent dissociation constant comprising cooperative interaction factors, which reflects the ligand concentration at half-maximal anisotropy. To fit our data values, Equation 2 was written as a function of [L] tot , total concentration of the ligand species, and [R] tot , total concentration of the reporter species in solution.
The real solutions f c ϭ [R(L) n ]/[R] tot of Equation 3 were substituted into the fluorescence anisotropy expression to be fitted to, The best fit for StrRAG1 binding curve to fluorescein-labeled 12-RSS DNA was obtained with the Hill coefficient n ϭ 1.8. The solution calculations and fitting were performed using Maple 7.0 and Origin 6.0 software. For experiments in which tryptophan anisotropy was monitored while adding unlabeled 12-RSS, because StrRAG1 quantum yield changes with DNA binding we calculated the correction factor R ϭ F/F 0, where F 0 is the relative quantum yield (derived from the integral value of fluorescence corrected spectra) of free StrRAG1, and F is the relative quantum yield of DNA bound StrRAG1 at saturation.
Using Equation 5, which relates the fraction bound protein to anisotropy (34,37), combined with Equation 2 for DNA binding, we obtained the equation used for fitting our tryptophan anisotropy data in Fig. 5 where The calculations of the apparent molecular weight of protein-DNA complexes using the rotation reorientation time were performed using the original Perrin equation, where ϭ 0.94 centipoise P, the viscosity of the aqueous medium at 25°C, M r is molecular weight, v ϭ 0.73 ml/g, the specific volume for the anhydrous protein, h is the degree of hydration (h ϭ 0 for an anhydrous sphere, h ϭ 0.2-0.4 ml/g for the typical hydration of most protein molecules), T ϭ 298 K, and R ϭ 8.31 ϫ 10 7 erg K Ϫ1 mol Ϫ1 ; one can predict based on the rotation reorientation time Global ϭ 48.03 Ϯ 10.05 ns the size of the StrRAG1⅐12-RSS complex to be between Ϸ175 kDa (using h ϭ 0) and 120 kDa (using h ϭ 0.4). Acrylamide quenching data represented as Stern-Volmer plots show an upward curvature concave towards the y axis. This is a characteristic feature for fluorophores quenched both by collisional and static quenching mechanisms. The fractional remaining fluorescence (F 0 /F) versus acrylamide concentration [Q] was fit to a second-order function (38,39), where K D and K S are the apparent dynamic and static quenching constants that characterize the ensemble of fluorophores (Trp and Tyr) present in StrRAG1.

RESULTS
Catalytically Active, Dimeric RAG1 Protein Lacking a Bulky Tag-In our study, we used a bacterially expressed and purified murine RAG1 protein (aa 377-1008) fused at its N terminus to a short (8 aa) Strep-II tag (Sigma Genosys, Inc.) that allows affinity purification with Streptactin-coupled Sepharose (Fig. 1a). Three purification steps result in a StrRAG1 protein of greater than 98% purity (Fig. 1b). The elution profile from the final gel filtration column reveals a prominent peak at the position expected for a StrRAG1 dimer (compare Fig. 2, a and c; the predicted molecular mass of the monomer is 72 kDa). When the purified dimeric protein was reanalyzed on the gel filtration column, only the dimer peak was observed, indicating that the protein remains stably associated in dimeric form (Fig.  2b). The intrinsic polarization anisotropy of the purified StrRAG1 at concentrations from 25 to 500 nM is constant, consistent with the existence of protein as a dimer at all concentrations within this range (data not shown). In a standard coupled cleavage assay, the StrRAG1 protein was approximately as active as highly purified, dimeric MBP-core RAG1 (aa 384 -1008) (Fig. 1c).
By EMSA, StrRAG1 Binds Specifically to the RSS, Forming a Single Protein-DNA Complex-We first tested the DNA binding ability of StrRAG1 by EMSA. Increasing amounts of StrRAG1 (or for comparison, MBP-RAG1) were incubated with a constant amount of labeled DNA bearing a consensus 12-RSS sequence. The binding buffer used throughout this study contains a moderate salt concentration (50 mM NaCl and 5 mM Mg 2 Cl) and was chosen to minimize nonspecific RAG1-DNA interactions (13,15). Under these binding conditions, StrRAG1 (Fig. 3a) gives rise to one predominant shifted species, whereas MBP-RAG1 (Fig. 3b) yields at least three complexes of distinct mobility, consistent with a previous study (22). Binding by StrRAG1 was quantified, and the data points were fit to a single Hill binding isotherm equation (Equation 1; "Experimental Procedures"), giving a dissociation constant of 92 Ϯ 12 nM. To assess binding specificity, a competition assay was performed. 200 nM StrRAG1 samples were individually incubated with 75 nM labeled 12-RSS probe followed by the addition of either unlabeled 12-RSS DNA or an oligonucleotide of the same length but lacking both nonamer and heptamer consensus sequences (nonspecific DNA). The 12-RSS is an effective competitor, whereas greater than a 6-fold excess of nonspecific competitor results in only an Ϸ30% decrease in the shifted complex (Fig. 3c). This 30% decrease may be the result of a fraction of StrRAG1 that binds non-specifically to DNA as well as the presence of a high concentration of polyanionic charge contributed by the phosphate groups of DNA, which may affect the overall stability of the complex in the electrophoretic field within the gel matrix. Gel-sieving effects combined with prolonged exposure of protein-DNA complexes to an electrophoretic field (3 h) and electrophoretic buffer may affect the results obtained using the EMSA methodology (40,41). These perturbations may especially affect the interactions between two macromolecular species when opposite charge attraction is a major factor in the stabilization of the complex. Such interactions have been implicated in the RAG1-RSS interaction (13,14).
In Solution, StrRAG1 Binds with Higher Affinity and Specificity to the 12-RSS Than Detected by EMSA-To circumvent the disadvantages associated with EMSAs, we used fluorescence polarization to investigate interactions between StrRAG1 and DNA in solution. Fluorescence anisotropy provides an indirect measure of the rotational diffusion of molecules in solution, which is in turn dependent on their size and shape (larger molecules have lower tumbling rates and have higher anisotropy values) (34). The first experiments made use of double-stranded 12-RSS DNA or nonspecific oligonucleotides labeled at the 5Ј end of one strand with fluorescein (12-RSS-Fl, nonspecific-DNA-Fl), which allowed measurement of the anisotropy of the DNA molecules (excitation wavelength 492 nm, emission recorded at 520 nm). 12-RSS-Fl or nonspecific-DNA-DNA-Fl (50 nM) was incubated with increasing amounts of StrRAG1, and the fluorescein polarization anisotropy was determined (Fig. 4a). The curve for 12-RSS-Fl has a sigmoidal shape with increasing anisotropy values that reach saturation around 100 nM StrRAG1 (monomeric protein concentration) and an increase of anisotropy of 90 -100% over the value in the absence of protein. The best fit of the data points (correlation ϭ 96%; sum of squares ϭ 0.65 ϫ 10 Ϫ4 ) was achieved using a Hill binding isotherm with n ϭ 1.8 (Equations 3 and 4), yielding an apparent K D ϭ 41 Ϯ 8 nM (reflecting monomeric protein concentration). This value for K D corresponds to an affinity of interaction higher than that determined by EMSA experiments. The increase in 12-RSS DNA anisotropy observed in this experiment is due solely to the formation of protein-DNA complexes, which increase the molecular size reported by the fluorophore attached to the DNA. In Fig. 4b  ear increase in anisotropy, corresponding to nonspecific binding of StrRAG1 to DNA. The specificity of the StrRAG1-DNA interaction was also examined by anisotropy competition experiments. First, 50 nM 12-RSS-Fl was incubated with 100 nM StrRAG1, and then either specific 12-RSS or nonspecific unlabeled DNA was added in increasing amounts to the mixture (Fig. 4c). Similar to the results of EMSA competition experiments, only the 12-RSS was capable of dissociating the 12-RSS-Fl⅐StrRAG1 complex and reducing the anisotropy to values characteristic of 12-RSS-Fl DNA alone. Even a 100-fold excess of nonspecific DNA resulted in only a slight reduction (12-15%) in the anisotropy of the complex.
In conclusion, polarization anisotropy experiments using  (lanes 10 -14). Labeled probe and competitor DNA were added simultaneously to the StrRAG1 protein.
allow direct calculation of the fluorophore life-time and the rotational reorientation time (when the emitted fluorescence comes through a vertically oriented polarizer), which is the inverse value of the rotational diffusion coefficient of the molecule (tumbling rate of molecule).
We first measured the lifetime of the fluorescein fluorophore attached to the 12-RSS or nonspecific DNA (Table I) and found that it was not significantly changed by the addition of StrRAG1 or HMG2 (a nonspecific DNA binder) from the standard 3.6 -4.1-ns value reported for fluorescein alone in pH 7.5 aqueous solutions (42). These important control measurements demonstrate that neither DNA bases nor protein amino acid side groups interact with fluorescein or significantly change the fluorophore environment when these proteins interact with the DNA. We then used frequency domain differential phase and polarized modulation ratio measurements to determine the timeresolved anisotropy decay of the fluorescein-labeled 12-RSS or nonspecific DNA in the presence or absence of StrRAG1 or HMG2 (Table I). Fitting of the anisotropy decay curves was performed using a double exponential equation (35,43) (44,45). Importantly, much lower global rotational reorientation times are reported for nonspecific DNA in the presence of StrRAG1, either at 50 or 100 nM protein concentrations, confirming the lower representation of protein-DNA species that form non-specifically in solution in our assay conditions (Table  I). From the calculated K D , one can estimate that more than 85% of the 50 nM 12-RSS DNA is present in protein-DNA complexes in the mixture with 100 nM StrRAG1 (saturation conditions). Hence, the global rotation reorientation time obtained in this reaction predominantly represents that of the protein-DNA complex. Using the rotation reorientation time Global ϭ 48.03 Ϯ 10.05 ns obtained under saturation binding conditions and the original Perrin equation (Equation 7; see "Experimental Procedures"), we can estimate the size of the StrRAG1⅐12-RSS complex to be between Ϸ175 kDa (using h ϭ 0 ml/g, an anhydrous sphere) and 120 kDa (using h ϭ 0.4 ml/g, the maximal degree of hydration) (46,47). Given the apparent molecular masses of the protein (72 kDa) and 12-RSS-Fl DNA (31 kDa), we conclude that StrRAG1 is most likely present as a dimer in these complexes.
Tryptophan Fluorescence Anisotropy of StrRAG1 Increases in the Presence of the 12-RSS-We wanted to monitor by fluorescence anisotropy the effect of adding DNA to StrRAG1 having as a reference the protein present in solution. This set of polarization anisotropy experiments took advantage of the intrinsic fluorophores in the StrRAG1 protein (primarily its 6 tryptophan residues) to serve as the fluorescent reporters (excitation wavelength 295 nm, emission recorded at 327 nm). Hence, these experiments measured changes in the size of the protein complexes whether or not they were bound to DNA. StrRAG1 (250 nM) was incubated with increasing concentrations of unlabeled 12-RSS, and tryptophan anisotropy was measured (Fig. 5). With the 12-RSS, anisotropy increased abruptly and reached a saturation value about 20% greater than that of the protein alone. The data fit a rectangular hyperbola, with an apparent K D ϭ 16.5 Ϯ 7 nM. This apparent K D value reflects the 12-RSS DNA concentration present in solution at half-maximal increase in anisotropy. Because the protein binds as a dimer, this K D value is expected to be approximately one-half of the K D value obtained from the data points in Fig. 4a, which reflects the concentration of monomeric StrRAG1 present in the reaction under similar conditions. The increase in tryptophan anisotropy is predominantly due to an increase in the molecular size of a significant population of StrRAG1 molecules in the presence of 12-RSS DNA. This increase can be due either to formation of protein-DNA complexes or to formation of higher order protein-protein oligomers not associated with DNA. In the presence of nonspecific DNA, there is an increase in anisotropy to levels 5-6% above the value for the free protein, and the increase is linear and nonsaturable.
Binding of StrRAG1 to DNA Causes Major Changes in the Protein Intrinsic Fluorescence Spectrum-We next wanted to ask whether DNA alters the conformation of the StrRAG1 protein. First, we tested if the intrinsic emission fluorescence spectrum (excitation at 280 nm, emission at 350 nm) of StrRAG1 protein is altered by DNA. Upon incubation of 100 nM StrRAG1 with increasing amounts of 12-RSS oligonucleotide, a drastic decrease of emission fluorescence intensity (fluorescence quenching) occurred, saturating at 50 -60% of the fluorescence in the absence of DNA (Fig. 6a). As with the increase in tryptophan polarization anisotropy (Fig. 5), this decrease in fluorescence caused by 12-RSS DNA can be fit to a rectangular hyperbola (Fig. 6c), but the effect does not quantitatively par-allel the DNA binding phenomenon. It is important to note that since the quenching effect occurs at concentrations of DNA lower than those expected to yield a significant amount of protein-DNA complexes, it is possible that interaction with the 12-RSS induces a change in protein configuration (see below). The decrease in fluorescence intensity induced by 12-RSS is associated with a slight shift of the maximum emission intensity ( max ) toward the red portion of the spectrum from 335 to 340 nm. A small decrease in fluorescence emission intensity was obtained in the presence of nonspecific DNA (Fig. 6, b and  c), but the quenching was not saturable, and there are no detectable spectral changes of max in this case. As a control for the quenching potential of the amine groups of DNA bases, we also used an amount of bovine serum albumin that gives a fluorescence intensity similar to that of 100 nM StrRAG1; the addition of increasing amounts of 12-RSS DNA resulted in minimal quenching (Fig. 6c).
Next, we performed kinetic experiments to determine whether quenching and binding occur simultaneously. First, 50 nM 12-RSS labeled with fluorescein was mixed with 50 nM StrRAG1, and the change in DNA anisotropy was measured with time (Fig. 7a). 10 -15 s after mixing, anisotropy reached levels expected from the static equilibrium experiment (Fig. 4a) and then remained unchanged for 13 min. This indicates that protein-DNA complexes are formed very rapidly. In the second experiment, 50 nM StrRAG1 was incubated with 10 nM 12-RSS or with nonspecific DNA, and intrinsic StrRAG1 fluorescence was measured with time. The 12-RSS induced a complex, slow multiphasic quenching of StrRAG1 fluorescence, quite distinct from that induced by nonspecific DNA. The first phase, which lasts ϳ200 s, was characterized by a substantial (17%) decrease in fluorescence. In the second phase, lasting between 200 and 700 s, fluorescence increased transiently, perhaps because of transient changes in fluorophore environment caused by protein conformational changes. Surprisingly, StrRAG1 intrinsic fluorescence quenching continued even after 800 s but at a much reduced rate. Although the kinetic behavior of StrRAG1 intrinsic fluorescence will require further study, it is clear that DNA binding (Fig. 7a) occurs much more rapidly than fluorescence quenching (Fig. 7b). We conclude that fluorescence quenching predominantly reflects conformational changes induced by the 12-RSS DNA.
Acrylamide Quenching Indicates That 12-RSS Binding Causes StrRAG1 Fluorophores to Become More Exposed to Solvent-To investigate the mechanism by which 12-RSS DNA acts as a quencher of StrRAG1 intrinsic fluorescence, we tested the effect of acrylamide on the StrRAG1 fluorescence in the presence or absence of DNA. Acrylamide is a small soluble molecule that quenches both tryptophan and tyrosine fluores-  cence and has access to regions of a protein where a macromolecule like DNA does not. Therefore, adding acrylamide after protein-DNA complex formation tests whether the intrinsic protein fluorophores become more exposed to the solvent interface (in the case of a conformational change) or less exposed to solvent molecules (if they are involved in stable complexes either with DNA bases or with other surrounding amino acids) relative to the protein fluorophores in the absence of DNA. Increasing concentrations of acrylamide were added to 600 nM StrRAG1 in the absence or presence of 150 nM specific 12-RSS or nonspecific DNA, and protein intrinsic emission fluorescence was measured. In Fig. 8, the ordinate represents the ratio between the original sample fluorescence in the absence of acrylamide (F o ) and the sample fluorescence (F) at the indicated concentration of acrylamide on the abscissa (Stern-Volmer plot) (38). In Table II, we present the values of the quenching constants derived from fitting our data to the quadratic Equation 8 (see "Experimental Procedures" and "Discussion"). The upward curvature of the Stern-Volmer plots indicates the combined effect of dynamic and static quenching (see "Experimental Procedures"). The slope of the curve, however, is increased by the presence of DNA (and more by 12-RSS than nonspecific DNA), indicating that DNA causes some of the amino acid fluorophores to become more accessible to the quencher, as reflected in the increased values of their dynamic quenching constants (Table II). This result is also consistent with the red shift of the spectra that occurs upon the addition of DNA. StrRAG1 has a wealth of fluorophores (6 Trp and 20 Tyr residues), and it is, therefore, difficult to address where in StrRAG1 quenching occurs. In summary, these experiments indicate an important role for dynamic (solvent mediated) quenching of intrinsic fluorophores of StrRAG1 upon interaction with DNA, and that this effect is stronger with specific than nonspecific DNA.
Circular Dichroism Provides Further Support for a DNAinduced Conformational Change in StrRAG1-We then asked if the 12-RSS-induced conformational change in StrRAG1 observed by fluorescence could also be detected by circular dichroism (CD). Fig. 9 presents the close and far UV CD spectra of 500 nM StrRAG1 protein alone or in the pres-  (48). Upon 12-RSS addition, the molar ellipticity at both of these characteristic wavelengths decreases dramatically, saturating at a DNA concentration of 50 -60 nM (Fig. 9b). When nonspecific DNA was used, the spectral changes observed were smaller (Fig. 8, c and d;  compare b with d).
Together, the StrRAG1 fluorescence and CD spectral changes induced by 12-RSS DNA lead to the conclusion that RSS sequences induce major conformational changes in StrRAG1. In addition, because these changes are much greater with 12-RSS than nonspecific DNA, sequence-specific interactions apparently underlie much of the observed changes. DISCUSSION RSS Binding by RAG1 in the Absence of RAG2-In an early study using an in vivo one-hybrid assay, core RAG1 was found to interact specifically with a tandem array of eight 12-RSS elements, and RAG2 appeared to enhance binding only modestly (14). Based on this and other findings, it was suggested that RSS recognition might be a two-step process consisting of RAG1 binding followed by recruitment of RAG2 (13,14). The results of subsequent in vitro experiments provided at least two types of arguments against the general validity of this model. First, it was found that the initial rate of DNA cleavage could be enhanced by preincubating MBP-RAG1 with RAG2 but not by preincubating MBP-RAG1 with the DNA (49). Second, studies relying primarily on EMSA methodology and RAG1 proteins with a bulky tag suggested that the RAG1-RSS interaction was not sufficiently specific to be physiologically relevant (see the Introduction). The only study to investigate the equilibrium binding of highly purified, dimeric RAG1 core protein used EMSA methodology and documented an approximate 10fold specificity for the RSS and a reasonably high affinity (K D ϭ 100 nM) (22). However, as confirmed here, the MBP-RAG1 core protein used in those experiments forms three species with the RSS in which the protein is present in various oligomeric states.
We show here that a RAG1 core protein with only a short eight amino acid tag is soluble and forms stable dimers in solution in the absence of DNA. Both EMSA and solution DNA binding assays demonstrate that this protein binds the RSS as a dimer and that it does so specifically and with moderately high affinity (in solution, apparent K D ϭ 41 Ϯ 8 nM). Furthermore, the complex that forms with the 12-RSS is relatively resistant to competition with nonspecific DNA (Fig. 4c) and by a number of different measures is distinct from that formed upon interaction with nonspecific DNA. These findings lead us to reconsider the relevance of the two-step model for RSS recognition noted above.
In developing lymphocytes in the G 1 phase of the cell cycle, RAG1 and RAG2 are co-expressed, with RAG2 apparently in molar excess (50). In the S, G 2 , and M phases of the cell cycle, however, RAG2 is rendered unstable by phosphorylation by a cyclin-dependent kinase (51,52), and RAG2 levels drop dramatically, whereas RAG1 levels are almost unchanged (30). We would like to propose that in S, G 2 , and M phase cells, RAG1 can bind to RSSs in the absence of RAG2. Furthermore, we propose that as these cells complete mitosis and enter G 1 and RAG2 levels begin to rise, RAG2 can be recruited to RSSs by pre-bound RAG1, creating a catalytically competent complex. This does not exclude the possibility that some and perhaps the majority of RSS recognition is performed by RAG1⅐RAG2 complexes. Instead, our results lead to the idea that a pattern of RAG1-RSS interactions laid down in S/G 2 /M may be relevant to the subsequent targeting of the recombination reaction.
A DNA-induced Conformational Change in RAG1-It has been proposed previously that RAG2 can induce a conformational change in RAG1 that facilitates binding of RAG1 to the RSS and perhaps also the appropriate folding of the RAG1 DDE active site (3,18,27). Until the experiments reported here, however, there have been no studies of conformational changes in the RAG proteins either induced by one another or by DNA. We have been able to study the interactions of StrRAG1 with DNA in solution by monitoring protein intrinsic fluorescence, circular dichroism spectra, and fluorescence polarization. These studies would likely have been considerably less informative if RAG1 had been attached to bulky MBP or glutathione S-transferase tags due to the spectral contributions of the fusion partners. Two pieces of evidence support the conclusion that RSS DNA induces a conformational change in RAG1. The addition of RSS DNA to StrRAG1 results in (i) strong quenching of intrinsic StrRAG1 protein fluorescence (Fig. 6) and (ii) a dramatic alteration in the CD spectrum of the protein (Fig. 9). We consider each of these in turn. The decrease in fluorescence intensity and red shift of max observed when StrRAG1 interacts with 12-RSS DNA could have two, not mutually exclusive, explanations.
(i) Dynamic quenching could occur if some of the main fluorophores (Trp and Tyr residues) of the protein undergo a change in environment from the hydrophobic core to the solvent interface of the macromolecule, where they are subject to dynamic quenching by solvent molecules. The red shift of max when StrRAG1 was incubated with 12-RSS DNA is in itself an indication that the dielectric constant ⑀ of the environment of some of the protein fluorophores increases, consistent with the red shift reported for the indole ring in cyclohexane versus water (53). This can only be caused by a conformational change in the protein structure. Such changes would also increase accessibility of the protein fluorophores to the small molecule quencher acrylamide, and hence, one would expect stronger acrylamide quenching for strRAG1 plus DNA than for strRAG1 alone. This is exactly what is observed in Fig. 8.
(ii) Static quenching could occur if some of the internal fluorophores of StrRAG1 became involved in direct complexes with amines of DNA bases. Alternatively, the DNA molecule could be an indirect static quencher if upon binding to StrRAG1, it induces a protein conformational change that causes charged amino acid residues to form direct complexes with Trp or Tyr fluorophores (the charge transfer that occurs in such complexes forbids the decay of the excited state and prevents photon emission). Direct static quenching by DNA does not necessarily involve a protein conformational change. In these cases, stronger acrylamide quenching should be observed for StrRAG1 alone than for the 12-RSS⅐StrRAG1 complex since fluorophores involved in static quenching cannot be quenched further by acrylamide. This scenario, however, is in conflict with the data that we present in Fig. 8. This reasoning was applied in assigning the static and dynamic acrylamide quenching (K S, K D ) values presented in Table II, which were simply derived as solutions of the quadratic Equation 8 (see "Experimental Procedures").
The kinetics of 12-RSS-induced quenching of StrRAG1 fluorescence shows a complex pattern (Fig. 7b) and occurs considerably more slowly than protein-DNA complex formation. Quantitatively, the 12-RSS specifically and efficiently quenches StrRAG1 intrinsic emission fluorescence, with the effect reaching saturation at DNA concentrations where only a fraction of the protein is engaged in forming StrRAG1⅐DNA complexes (Fig. 6c). Together, our observations are best explained by a model in which dimers that dissociate from the DNA either retain the newly acquired conformation or further modify their configura- tion (as the transient increase in fluorescence in Fig. 7b suggests) and, hence, are as susceptible or even more susceptible to solvent quenching than those bound to DNA. Therefore, it is possible that the 12-RSS acts as a "catalyst" to induce major conformational changes in StrRAG1 whether bound to the DNA or not. We note that the oligomerization state of other DNA-binding proteins such as repressor (54) and human immunodeficiency virus integrase (55) has been shown to be influenced by the presence of DNA. In the case of human immunodeficiency virus integrase, time-resolved anisotropy measurements of its tryptophan rotation reorientation times show that integrase exists in various oligomeric states in solution, and Mg 2ϩ and DNA cause dissociation of the tetrameric form to the monomeric form that is found in complex with DNA (55,56).
The addition of 12-RSS DNA and to a lesser extent nonspecific DNA results in a dramatic change in the CD spectrum of StrRAG1 (Fig. 9). A possible explanation for the DNA-induced decrease in both 209-and 222-nm peaks (indicative of ␣ helicity (57)) is the movement of the peptide groups from an orientation almost parallel to the helix axis towards a tilted configuration with carbonyl groups pointing outwards, due to a drastic change in the protein backbone. Such a reorientation of the peptide groups has been predicted to decrease the amplitude of long wavelength CD peaks of proteins with ␣ helix content (58).
The stoichiometry of pre-cleavage RAG1⅐RAG2⅐RSS complexes is controversial, with two reports suggesting that these complexes contain more than two (presumably four) monomers of RAG1 (25,26), and another suggesting that they contain only a single dimer of RAG1 (24). If the former model is correct, then since DNA-bound StrRAG1 dimers show no propensity to associate with other StrRAG1 dimers, it is likely that stable association of two dimers of the RAG1 core with an RSS requires RAG2. The situation might be different for the fulllength RAG1 protein, which contains an additional zinc-dependent dimerization domain (59).
In conclusion, we demonstrate that the core RAG1 protein containing a small epitope tag binds the RSS as a dimer and undergoes a conformational change upon doing so. Our results provide a starting point for future studies of possible RAG2induced conformational changes in RAG1.