Recognition of T-rich single-stranded DNA by the cold shock protein Bs-CspB in solution.

Cold shock proteins (CSP) belong to the family of single-stranded nucleic acid binding proteins with OB-fold. CSP are believed to function as 'RNA chaperones' and during anti-termination. We determined the solution structure of Bs-CspB bound to the single-stranded DNA (ssDNA) fragment heptathymidine (dT7) by NMR spectroscopy. Bs-CspB reveals an almost invariant conformation when bound to dT7 with only minor reorientations in loop beta1-beta2 and beta3-beta4 and of few aromatic side chains involved in base stacking. Binding studies of protein variants and mutated ssDNA demonstrated that Bs-CspB associates with ssDNA at almost diffusion controlled rates and low sequence specificity consistent with its biological function. A variation of the ssDNA affinity is accomplished solely by changes of the dissociation rate. 15N NMR relaxation and H/D exchange experiments revealed that binding of dT7 increases the stability of Bs-CspB and reduces the sub-nanosecond dynamics of the entire protein and especially of loop beta3-beta4.

INTRODUCTION

A rapid reduction in growth temperature induces the cold shock response of many prokaryotic organisms by down-regulation of the expression of most proteins (1) and transient up-regulation of a small set of proteins (2,3). These induced proteins include the transcription factor NusA, initiation factor IF2, ribosomal proteins L7/L12 and S6, ribosome-binding factor RbfA, prolyl isomerase PPiB (4) and, most prominently, the family of cold shock proteins, CSP (1). CSP stimulate the transcription of cold shock inducible genes (5) and the initiation of translation by destabilizing non-productive secondary structures in mRNA at low temperature (6). Therefore, CSP are denoted as RNA chaperones, because binding of mRNA is sequence unspecific (7–9). Destabilizing of secondary structures of single-stranded nucleic acids takes place in anti-termination, which leads to the induction of various cold shock genes at low temperature (4,10–12). In cell-free transcription and translation, however, CSPs block the bulk protein expression (13). The 5′-untranslated regions (5′UTRs) of Csp mRNAs contain high affinity binding sites for ribosomes, which are responsible for a very effective initiation of CSP translation (14,15). This redistribution of the ribosomes might block the translation of other proteins at low temperature (16).

In Bacillus subtilis, three CSP (Bs-CspB, Bs-CspC and Bs-CspD) have been identified, which can complement each other in vivo, but a knock out of all three CSP genes is lethal (17). The structure of the 67 amino acid encompassing protein Bs-CspB revealed a β-barrel with an OB-fold (18) formed by five antiparallel β-strands (19,20). The structure of the CSP is conserved in mesophilic, thermophilic and hyperthermophilic bacteria, such as Ec-CspA from Escherichia coli (21–23), Bc-Csp from Bacillus caldolyticus (24) and Tm-Csp from Thermotoga maritima (25). One unique feature of their structures is the numerous solvent exposed aromatic side chains (W8, F9, F15, F17, H29, F30, F38 in Bs-CspB) from which F15, F17, F27 and H29 significantly increase the thermodynamic stability of the protein (26,27). They are part of the nucleic acid-binding motifs RNP1 and RNP2, which have been identified by sequence alignment (28,29). Our detailed structural knowledge of OB-folded nucleoprotein complexes in solution is restricted to telomere DNA-binding proteins (30–32).

In the present paper we determined the conformation of Bs-CspB bound to the single-stranded DNA (ssDNA) fragment dT7 by NMR spectroscopy in solution. The structural analysis of the Bs-CspB/dT7 complex dominated by hydrophobic interactions between aromatic protein side chains and ssDNA bases is accompanied by a mutational analysis of the thermodynamics and kinetics of the complex formation by equilibrium and stopped-flow fluorescence spectroscopy. High affinity binding with nanomolar dissociation constants correlates with a low dissociation rate and an invariant, almost diffusion controlled association rate for wild-type Bs-CspB when compared with numerous protein variants. H/D exchange experiments and 15N NMR relaxation of free and ssDNA-bound Bs-CspB revealed the changes in stability and dynamics of the Bs-CspB/dT7 complex when compared to free Bs-CspB.

MATERIALS AND METHODS

Wild-type Bs-CspB from B.subtilis and its variants were expressed and purified as described previously (33) with minor modifications. Deoxyoligonucleotide dT7 and other ssDNA fragments were purchased from MWG Biotech AG (Germany). Bs-CspB and dT7 concentrations were determined by the absorbance at 280 and 260 nm using extinction coefficients of 5800 M−1 cm−1 (27) and 58 800 M−1 cm−1, respectively. The extinction coefficients of various ssDNA sequences were calculated as sum of the extinction coefficient of the individual nucleotides.

NMR measurements

NMR spectra for structure determination were acquired at 15°C on Bruker DRX500, Avance 600, DMX750 and Avance 900 spectrometers in 50 mM Na-cacodylate, 3 mM MgCl2, pH 7.0. NMR spectra for the relaxation study were recorded at 25°C in 100 mM Na-cacodylate, pH 7.0. External referencing of proton chemical shifts was achieved by using 2,2-dimethyl-2-silapentane-5-sulfonate (DSS).

Formation of the Bs-CspB/dT7 complex was accomplished by titration of 0.8 mM uniformly labeled 15N Bs-CspB with unlabeled dT7 at 500 MHz. Most of the backbone amides experience fast exchange with respect to the NMR chemical shift timescale and therefore could be directly assigned by tracking the gradually shifting crosspeaks starting from the unbound protein. The mean weighted change (34) in chemical shifts between apo and holo Bs-CspB of the backbone resonances were calculated according to Δ(δMW(1H,15N)) = (((δ(1H))2 + 1/25(δ(15N))2)/2)0.5 and Δ(δMW(1Hα,13Cα)) = (((δ(1Hα))2 + 1/10(δ(13Cα))2)/2)0.5. Assignment of crosspeaks in the slow chemical exchange regime was carried out with standard 3D 15N-edited NOESY- and TOCSY-HSQC spectra at 500 MHz. Standard triple resonance experiments [HNCACB (35), CBCA(CO)NH (36) and HNCO (37)] for the backbone assignments were recorded on a 13C,15N-labeled sample of Bs-CspB in complex with unlabeled dT7 at 600 MHz in 93% H2O/7% D2O. Aliphatic side chain assignment was performed using H(C)(CO)NH- and C(CO)NH-TOCSY experiments (37) in H2O as well as HCCH-COSY experiments in D2O (38). Additionally, aromatic side chains were assigned from (Hβ)Cβ(CγCδ)Hδ and (Hβ)Cβ(CγCδCɛ)Hɛ spectra (39).

A 2D 15N-edited NOESY at 750 MHz and a 3D 15N-edited NOESY-HSQC (40) at 500 MHz with mixing times of 150 ms were recorded with a sample containing 15N-labeled Bs-CspB and unlabeled dT7. 1H homonuclear 2D NOESY spectra in D2O with mixing times between 80 and 150 ms were acquired at 500 MHz on a fully unlabeled sample. Additionally, 3D 13C-edited HMQC-NOESY spectra were recorded in D2O with altered 13C carrier frequencies to observe aliphatic or aromatic carbons, respectively. Attempts to assign resonances and intramolecular NOEs of the dT7 oligonucleotide with 2D double-half-filtered NOESY (mixing times between 100 and 250 ms) and 2D double-half-filtered TOCSY spectra at 600 and 900 MHz (41,42) or 2D 13C/15N F1-filtered NOESY experiments (43) at 600 MHz were unsuccessful. Intermolecular NOEs between Bs-CspB and dT7 were determined by 2D 13C F1-edited, F2-filtered NOESY spectra (42) but remained also unassigned. NOEs identified in the double-half-filtered (intramolecular NOEs of dT7) and filtered-edited (intermolecular NOEs) spectra assisted in identification of these crosspeaks in the homonuclear and 15N or 13C-edited NOESY experiments from which distance restraints for the structure calculation were derived. Subsequently, these NOEs were omitted from the latter. Additionally, coupling constants were obtained from an HNHA experiment (44) and residual dipolar couplings (RDCs) were determined in a weakly oriented sample of 15N Bs-CspB in 18 mg/ml filamentous phage Pf1 (Asla). HN-N RDCs were measured from IPAP-[1H-15N]-HSQC experiments (45). For the comparison of translational diffusion of Bs-CspB and Bs-CspB/dT7, a PFG-SLED pulse sequence (46) was employed.

Potential hydrogen bond donors were identified from H/D exchange experiments by dissolving lyophilized 15N Bs-CspB/dT7 complex in D2O and subsequent collection of a series of 2D 15N HSQC spectra. Protection factors (P) were derived from P = kint/kex, where kint is the intrinsic exchange rate constant and kex is the observed exchange rate constant of a backbone amide. kint was calculated on the basis of model peptides (47) and kex was obtained by fitting a single exponential function without offset to the intensity decay of crosspeaks in the recorded 2D HSQC spectra.

For the 15N relaxation study of free and dT7 bound Bs-CspB NMR spectra for the determination of R1, R2 and heteronuclear NOE (hNOE) were recorded at a Bruker DRX500 at 25°C as described recently (48). Model-free analyses and iterative optimization of Lipari-Szabo motional parameters were performed with the program MODELFREE (49) (version 3.1) as described by using an isotropic tumbling model (48).

Structure calculation and analysis

NOE crosspeaks were manually classified as strong, medium or weak according to their intensities and converted into distance restraints of <3.0, 4.0 or 5.5 Å, respectively. Scalar 3JHN,Hα coupling constants of either <6.0 Hz or >8.0 Hz were restrained to adopt phi torsion angles between −80° and −40° or between −160° and −80°, respectively (45,50). Slow exchanging hydrogens were identified from amide proton exchange experiments. For each of the assigned 23 hydrogen bonds the distance between the amide proton and the acceptor was restrained to <2.3 Å and the distance between the amide nitrogen and the acceptor to <3.3 Å.

This structural information served as an input for the calculation of 120 structures using restrained molecular dynamics with XPLOR-NIH-1.2.1 (51). A three-stage simulated annealing protocol (52) with floating assignment of prochiral groups (53) was carried out using the following simulation procedure. For conformational space sampling 60 ps with a time step of 3 fs were simulated at a temperature of 2000 K, followed by 80 ps of slow cooling to 1000 K, and 50 ps of cooling to 100 K, both with a time step of 2 fs. A conformational database term for both backbone and side chain dihedral angles (54) with the described modification (55) was included in the target function in order to improve the stereochemical properties of the structures. After simulated annealing the structures were subjected to 1000 steps of energy minimization, the final 500 steps without conformational database potential. The 60 lowest energy structures were subject to refinement with HN-N RDCs as described previously (56) and the 18 lowest energy structures having no NOE distance restraint violations >0.15 Å and no φ-angle restraint violations >0.5° were selected for further characterization. The geometry of the structures, structural parameters and elements of secondary structure were analyzed using the programs DSSP (57) and PROCHECK (58,59) revealing 89.8% of the residues in the core region of the Ramachandran plot, 9.7% in the additionally allowed regions and 0.5% in the generously allowed regions. The molecular structure figures were generated using MolMol (60) and PyMOL [DeLano, W. (2003), ]. This family of 18 structures has been deposited in the Protein Data Bank under the accession code 2F52.pdb (RCSB035462).

Fluorescence quenching

The intrinsic fluorescence of Bs-CspB Trp8 is quenched upon binding to nucleic acids and therefore was used as a sensitive probe to monitor binding in equilibrium and kinetic experiments. Binding affinities were determined by equilibrium titration experiments as described previously (27) whereby protein concentrations of 25, 100 or 200 nM were used depending on dissociation constants (KD) of the respective protein variant to dT7 or of wild-type Bs-CspB to the various ssDNA fragments. To obtain the stoichiometry of the respective nucleoprotein complexes protein concentrations between 2 and 10 μM were used.

Association kinetics were monitored by quenching of the intrinsic tryptophan fluorescence of Bs-CspB by the ssDNA fragments in a sequential stopped-flow mixing device and analyzed as described elsewhere (27). Protein concentrations of 30 or 300 nM were used depending on the KD of the respective Bs-CspB variant and dT7 end concentrations up to 1.2 μM were employed.

RESULTS

NMR titration of Bs-CspB by heptathymidine dT7

Various affinities and stoichiometries have been reported for different nucleoprotein complexes with Bs-CspB (7–9,27). Therefore, a systematic search for a high affinity ssDNA fragment was performed (Table 1) including potential physiological target sequences (Y-box and cold box fragments). The heptathymidine fragment dT7 revealed the lowest KD of 1.8 nM at a 1:1 stoichiometry and was therefore selected for further investigations. An NMR titration of 15N enriched Bs-CspB sample with unlabeled dT7 was performed and a superposition of the first and last HSQC spectrum is depicted in Figure 1. Most amide protons showed fast exchange on the NMR chemical shift time scale allowing a direct assignment of the amide resonances in the Bs-CspB/dT7 complex from the gradual shift during the titration. The following residues stand out from the mean weighted change in chemical shifts Δ(δMW(1H,15N)) and Δ(δMW(1Hα,13Cα)) of the backbone resonances (34) between apo and holo Bs-CspB: β-strand 1 (K7, W8 side chain, N10), β1–β2 loop (S11, E12), RNP1 motif (K13, G14, F15, F17, G16, V20), β-strand 3 (D25), RNP2 motif (V26, F27, V28, H29, F30), β3–β4 loop (S31, I33, G35, F38, K39, T40, L41, E42), β-strand 4 (G54), β4-β5 loop (R56), and β-strand 5 (G57, P58, Q59 backbone and side chain). These residues of Bs-CspB facilitate the binding of dT7 either by direct interaction or by conformational rearrangements remote from the binding interface. Significant Δ(δ(1H)) values for the Hβ protons include F15, D25, F30, Q34 and N62.

Table 1

Dissociation constants and stoichiometries of Bs-CspB/ssDNA complexes from tryptophan fluorescence titration experimentsa

Fragment	Number of nucleotides (%T)	ssDNA sequence	Bs-CspB:ssDNA	KD (nM)
dT4	4 (100)	TTT T	1:1	3205 ± 150
dT5	5 (100)	TTT TT	1:1	940 ± 10
dT6	6 (100)	TTT TTT	1:1	326 ± 20
dC6	6 (0)	CCC CCC	1:1	12600 ± 1000
dT7	7 (100)	TTT TTT T	1:1	1.8 ± 0.4
Y-box5b	5 (40)	ATT GG	1:1	5300 ± 1000
Y-box7b	7 (57)	TAT TGG T	1.4:1	530 ± 50
Y-box12b	12 (25)	CTG ATT GGC CAA	1.9:1	2030 ± 390
Y-box25b	25 (28)	ATC CTA CTG ATT GGC CCA GGT GCT G	3:1	3920 ± 45
dcb1c	7 (57)	TTA TTA G	1:1	20 ± 3
dcb1_ac	13 (69)	ATT ATT TTT GTT C	2:1	44 ± 2
dcb1_bc	13 (8)	GAG CAA GAA TAG G	2:1	1160 ± 90
dcb1_brc	13 (8)	GGA TAA GAA CGA G	2:1	1090 ± 110
dcb2_ac	17 (41)	TTT AAG AAG AAA GTT TT	2:1	460 ± 34
dcb2_bc	16 (43)	GAG TTT TGG TCT TGA A	2:1	43 ± 2
dcb2_brc	16 (43)	AAG TTC TGG TTT TGA G	2:1	30 ± 2
dcb2ds_13c	13 (46)	GTT TTG TAA GAG T	2:1	66 ± 21

aAll experiments were performed at 15°C in 50 mM Na-cacodylate, pH 7.0, and 100 mM KCl.

bssDNA fragments containing the Y-box motif ATTGG (underline).

cssDNA fragments deduced from the 5′-UTR of Bs-CspB mRNA.

Figure 1

The 2D 1H/15N HSQC spectrum of free (black) and dT7-bound (red) Bs-CspB in 50 mM Na-cacodylate, 3 mM MgCl2, pH 7.0 at 15°C. The spectra were recorded at a Bs-CspB-concentration of 0.8 mM and a final dT7 concentration of 1.2 mM.

Structure determination of the protein conformation of Bs-CspB in complex with dT7

The experimental constraints, which were employed for the structure calculations of Bs-CspB in complex with dT7 are summarized in Table 2. Possible backbone hydrogen bond donors have been determined from the protection factors (47) derived by H/D exchange experiments (Figure 2). Several NMR experiments were performed to assign the dT7 resonances in the complexed form and to identify intermolecular NOE effects between Bs-CspB and the oligonucleotide dT7, including 2D 13C(ω1, ω2) double half-filtered NOESY and 2D 13C/15N(ω1)-filtered NOESY experiments of 13C/15N Bs-CspB in complex with unlabeled dT7. Additionally, homonuclear or 15N-edited 1H NOESY spectra of partially deuterated 15N-labeled Bs-CspB in complex with protonated dT7 were analyzed. Owing to the lack of dispersion of the homonucleotide sequence of dT7, an unambiguous assignment of the resonances of either uncomplexed or complexed dT7 was not successful. Therefore, the conformation of dT7 in complex with Bs-CspB could not be determined by NMR spectroscopy.

Table 2

Experimental restraints and structural statistics (for the calculation of the tertiary structure of Bs-CspB in complex with the ssDNA fragment dT7)

Number of experimental restraints
Distance restraints from NOEs
Interresidual NOEs	Sequential	(|i − j| = 1)	283
	Medium long range	(|i − j| < 5)	135
	Long range	(|i − j| ≥ 5)	411
Intraresidual NOEs			381
Dihedral angle restraints φ/χ1			29/2
RDC restraints (DN,HN)			50
Hydrogen bond restriants			23
Molecular dynamics statistics
Average energy (kcal/mol)	Etot 31.0 ± 3.0
	Ebond 0.9 ± 0.1
	Eangle 10.2 ± 1.2
	Eimproper 2.7 ± 0.6
	Evdw 7.6 ± 1.5
	ENOE 3.3 ± 1.2
	Ecdih 0.1 ± 0.02
	ERDC 6.1 ± 1.1
r.m.s.d. from ideal distance (Å)	Bonds 0.00096 ± 0.00007
	NOE 0.012 ± 0.001
r.m.s.d. from ideal angles (degree)	Bond angles	0.190 ± 0.012
	Improper angles	0.287 ± 0.016
r.m.s.d. from dipolar couplings (Hz)	HN-N couplings	0.071 ± 0.014
Atomic r.m.s.d. (Å) of 18 refined Bs-CspB structuresa
All residuesb	0.54 (backbone)	0.94 (heavy atoms)
Secondary structurec	0.21 (backbone)	0.55 (heavy atoms)
Pairwise backbone r.m.s.d. (Å) between different Bs-CspB structures
	NMR(bound)d	X-ray(bound)e	X-ray(free)f	NMR(free)g
NMR(bound)	0	1.44h (0.87)c	1.34 (0.80)	1.68 (1.08)
X-ray (bound)	—	0	0.92 (0.37)	1.90 (1.37)
X-ray(free)	—	—	0	1.66 (1.28)
NMR(free)	—	—	—	0

aThe final force constants used in the structure calculation were 1000 kcal·mol−1·Å−2 for the bond length, 500 kcal·mol−1·rad−2 for the bond angles and improper angles, 50 kcal·mol−1·Å−2 for the NOE distance restraints, 55 kcal·mol−1·rad−2 for the ϕ-angle restraints, and 1.0 kcal·mol−1·Hz−2 for the RDCs.

bCalculated for the final set of 18 structures (residue 1–67).

cCalculated for the elements of regular secondary structure (residues 2–10, 15–19, 26–32, 46–53, 59–65).

dAverage structure from this study.

eCrystal structure of ssDNA-bound Bs-CspB (2ES2.pdb).

fCrystal structure of the free Bs-CspB (1CSP.pdb) (20).

gNMR average structure of the free Bs-CspB(1NMG.pdb) (19).

hCalculated for residues 1–67.

Figure 2

Protection factors of amide protons from H/D exchange of free (hatched bars) and dT7-bound (solid bars) Bs-CspB. Missing bars indicate amides, which got fully deuterated in the dead-time of the experiment (5 min). The solid bars on top represent the five β-strands of Bs-CspB.

An overlay of the protein backbone of the 18 lowest energy structures of the Bs-CspB conformation in complex with dT7 is given in Figure 3 and compared with the NMR structure of apo Bs-CspB (19). The conformation of the five strands of the β-barrel reveals only small differences and serves as a preformed scaffold for ssDNA recognition. Backbone deviations of the two NMR families of structures occur in loop β1–β2 and loop β3–β4. The latter loop remains less well defined with bias in its orientation towards loop β1–β2. We are to some extent limited in the analysis of the side chain conformations of those aromatic residues, which are involved in stacking interactions with dT7 (see below), because the conformation of the nucleic acid could not be calculated and the missing intermolecular NOEs cause overvaluing of the intramolecular NOEs. Additionally, a comparison of the side chain conformations of F15, F27, F29 and F38 is not straightforward because they are not very well defined in the solution structure of free Bs-CspB (19). Beside those, W8 is well defined in both the free and dT7 bound form and shows an almost perpendicular orientation. This is consistent with the strong quench of the tryptophan fluorescence observed upon ssDNA binding (see below).

Figure 3

Backbone superposition of the 18 lowest energy NMR structures of free [blue, 1NMF.pdb (19)] and dT7-bound (yellow) Bs-CspB. The five β-strands and their connecting loops are indicated. The r.m.s.d. values of dT7-bound Bs-CspB of all 67 residues are 0.54 Å (backbone) and 0.94 Å (all heavy atoms). Main deviations between apo and holo form of Bs-CspB are located in loop β1–β2 (L12) and loop β3–β4 (L34).

Mutational analysis of the thermodynamics and kinetics of Bs-CspB/dT7

Based on the above-described structural elucidations, a set of protein variants has been selected to verify the role of the respective residue in dT7 binding by equilibrium and time resolved fluorescence spectroscopy. In the Bs-CspB/dT7 complex, the intrinsic fluorescence of W8 is 80% quenched compared to the apo protein. This sensitive probe allows the determination of dissociation constants (KD) as well as association (kon) and dissociation (koff) rate constants of the wild-type nucleoprotein and its variants. A stoichiometric titration of Bs-CspB monitored by fluorescence quenching and the analysis of NMR chemical shifts of backbone resonances as described elsewhere (27) revealed a 1:1 stoichiometry for Bs-CspB/dT7 within a concentration range of nano- to millimolar (data not shown). Table 3 shows a summary of KD, kon and koff values of Bs-CspB/dT7 and 18 nucleoprotein variants derived from fluorescence experiments.

Table 3

Dissociation constants as well as association and dissociation rate constants of wild-type Bs-CspB (WT) and Bs-CspB variants with dT7 and wild-type Bs-CspB with dT7 derived ssDNA fragmentsa,b

Protein + ssDNA	KD × 10−9 (M)	kon × 108 (M−1 s−1)	koff (s−1)c	koff (s−1)d
WT Bs-CspB + dT7	1.8 ± 0.4	3.37 ± 0.11	1.7 ± 0.6	0.6 ± 0.6
K7A + dT7	6.4 ± 0.6	3.03 ± 0.01	5.3 ± 2.0	1.9 ± 0.2
K13Q + dT7	25.0 ± 3	1.45 ± 0.03	5.5 ± 0.2	3.6 ± 0.5
F15A + dT7	135 ± 4	1.58 ± 0.08	30.0 ± 1.6	21.3 ± 1.7
F15Y + dT7	1.4 ± 0.4	3.44 ± 0.10	2.2 ± 0.3	0.5 ± 0.2
F17A + dT7	345 ± 63	n.d.	n.d.	n.d.
D25A + dT7	21.4 ± 0.5	n.d.	n.d.	n.d.
F27A + dT7	286 ± 7	2.18 ± 0.08	103 ± 2	62.3 ± 3.8
H29Q + dT7	104 ± 6	2.41 ± 0.01	35.1 ± 2.3	25.1 ± 1.6
F30A + dT7	208 ± 11	4.29 ± 0.03	177 ± 6	89.2 ± 5.3
F30W + dT7	1.3 ± 0.3	4.08 ± 0.06	1.0 ± 0.4	0.5 ± 0.1
G35A + dT7	1.1 ± 0.3	3.18 ± 0.02	1.8 ± 1.1	0.3 ± 0.4
G35P + dT7	118 ± 5	2.03 ± 0.04	72.5 ± 1.1	24.0 ± 1.5
F38A + dT7	228 ± 9	3.34 ± 0.01	137 ± 4	76.2 ± 3.2
G54A + dT7	1.3 ± 0.3	2.62 ± 0.07	1.3 ± 0.8	0.3 ± 0.1
G54P + dT7	220 ± 9	n.d.	n.d.	n.d.
R56A + dT7	11.8 ± 1.4	1.39 ± 0.01	10.0 ± 1.4	1.7 ± 0.2
P58A + dT7	1.3 ± 0.2	4.08 ± 0.06	1.0 ± 0.7	0.5 ± 0.1
Q59A + dT7	2.7 ± 0.4	3.43 ± 0.01	6.1 ± 0.7	0.9 ± 0.1
WT + CTTTTTC	33.7 ± 4.1	2.58 ± 0.01	25.3 ± 2.1	8.7 ± 1.0
WT + CTCTTTC	3.9 ± 0.2	2.74 ± 0.01	4.6 ± 1.1	1.1 ± 0.1
WT + CTCTCTC	10.8 ± 0.9	2.66 ± 0.01	6.5 ± 1.2	2.9 ± 0.2
WT + CTCTTCC	66.2 ± 4.8	2.20 ± 0.02	36.1 ± 2.4	14.6 ± 1.1
WT + CTCCTTC	12.5 ± 0.6	2.64 ± 0.01	13.0 ± 0.6	3.3 ± 0.2

aW8 could not be varied, because the quench of its fluorescence emission was used for the determination of KD, kon, and koff.

bKD values were determined according to a 1:1 stoichiometry of protein versus oligonucleotide.

cthe dissociation rate constant koff was determined from the offset of the linear fit of experimentally observed rate constant kobs = kon [ssDNA] + koff at various ssDNA concentrations.

dthe dissociation rate constant koff was calculated by koff = kon KD.

The dissociation constants show that aromatic residues at positions 15, 17, 27, 29, 30 and 38 are essential for a tight binding of dT7. Substitutions with non-aromatic residues at these positions result in an at least 50-fold increase of KD, whereas substitutions by aromatic residues (position 15 and 30) lead to wild-type affinities towards dT7. The latter two substitutions are based on a sequence alignment of Bs-CspB with Bc-Csp and Tm-Csp (27). A change in the electrostatics and polarity at positions K7, K13, D25, R56 and Q59 lead to less pronounced changes in the nucleotide binding affinity. The conformational freedom allowed by glycine at positions 35 and 54 is not required for a tight binding as tested by the alanine variants. A strong reduction of possible backbone Φ angles at these positions by substitution with proline, however, increased KD significantly. G35 and G54 are part of the loops β3–β4 and β4–β5, respectively, confining the binding surface of Bs-CspB, and for which sufficient conformational freedom is required for tight binding. The P58A variant remains silent in this respect.

The dissociation constant at equilibrium reflects the ratio between the kon and koff rate constants of complex formation. To gain deeper insights, both rates have been determined experimentally by rapid mixing of the protein solutions with various amounts of oligonucleotides in a stopped-flow fluorescence spectrometer. Under pseudo first-order conditions (27), kon is the slope of the ssDNA concentration dependence of kobs = kon · [ssDNA] + koff and therefore better determined than the offset koff, which results from an extrapolation to 0 M ssDNA. Therefore, koff is often determined by koff = kon · KD. These values are also given in Table 3. All kon and koff rate constants of the protein variants have in common that kon is almost invariant whereas koff varies up to a factor of 100 according to the respective KD. Consequently, the reduced affinity of mutated Bs-CspB towards dT7 results from an increased dissociation rate. A set of experiments, where binding kinetics were determined for wild-type Bs-CspB and cytosine doped T-rich ssDNA fragments (bottom of Table 3), leads to the same conclusion.

15N NMR relaxation study of the dynamics of Bs-CspB/dT7

Complex formation e.g. in the framework of an induced fit model should be reflected in a change of the protein dynamics. Therefore, the dynamics of Bs-CspB/dT7 were determined by an extended Lipari-Szabo analysis of 15N relaxation rates and hNOE effects. Figure 4 depicts a comparison of hNOE, general order parameter S2, and chemical exchange contributions (Rex) to R2 of Bs-CspB/dT7 with free Bs-CspB. hNOE and S2 reflect large amplitude motions of each backbone N-H vector on a nanosecond-to-picosecond timescale and Rex on a millisecond-to-microsecond timescale. The overall rotational correlation time for Bs-CspB and Bs-CspB/dT7 was very similar (4.4 and 4.0 ns, respectively). Consequently, the observed 1:1 stoichiometry of the Bs-CspB/dT7 complex in solution is realized by protein and ssDNA monomers. The vast majority of backbone amides show increased hNOE and S2 values upon dT7 binding. All residues in the five β-strands show this trend indicating a loss of internal flexibility of the entire protein chain in Bs-CspB/dT7. F30 is the only backbone amide, which has a strongly decreased S2 value in the complex (0.65) compared with Bs-CspB (0.85). In contrast, S2 values of the N-terminus of loop β3–β4 (F38-E43) increase upon binding to values of the adjacent β-strands. This indicates a loss in dynamics and explains the better-defined NMR structure of this region compared to the free form (Figure 3). The order parameters of all other aromatic residues involved in binding show high values in both the free and complexed form of Bs-CspB.

Figure 4

15N relaxation and internal motion parameter of free (48) (open symbols) and dT7-bound (closed symbols) Bs-CspB. (a) heteronuclear NOE hNOE and (c) transversal relaxation rates R2, as well as (b) order parameters S2 and (d) chemical exchange contributions Rex to R2. S2 and Rex were obtained from extended Lipari-Szabo analyses of longitudinal relaxation rates R1 (data not shown), R2 rates and hNOE using the program MODELFREE (49). Solid and gray bars represent the five β-strands of Bs-CspB.

The Rex chemical exchange contributions to R2 show remarkable differences for free Bs-CspB and Bs-CspB/dT7. In the free form of Bs-CspB, all backbone amides reveal Rex contributions from global unfolding and refolding of the protein even under native buffer conditions (refolding rate constant kf = 1070 s−1 and the unfolding rate constant ku = 12 s−1) (33,48,61,62). Upon binding, all Rex rate constants beside W8 drop below 2 Hz indicating that the addition of dT7 stabilizes the native state (see below). As a consequence, the population of unfolded peptide chains drops far below 1% and no Rex contributions from global folding are detectable. The remaining Rex contributions in Bs-CspB/dT7 originate from local unfolding mainly located in the loop regions. These loops remain dynamic on a millisecond-to-microsecond timescale even in the protein/oligonucleotide complex.

H/D exchange of amide protons in Bs-CspB/dT7

Amide proton exchange is a sensitive measure for the global and local stability of a protein (63). Protons with the highest protection factors P are indicative of global unfolding, whereas protons with lower P additionally exchange by local breathing. The Gibbs free energy of unfolding of Bs-CspB is 12 kJ/mol, which is evident by the small protection factors of the amide protons (hatched bars in Figure 2). Most amides had fully exchanged before the first 2D 1H-15N HSQC could be recorded after dissolving protonated 15N protein in D2O. Bs-CspB/dT7 revealed increased protection factors of amide protons in β-strands up to 105. Even amides in loop β3–β4 are well protected. Missing bars in Figure 2 represent amides with protection factors still too small to be detected with the performed experimental setup.

DISCUSSION

Only one NMR-based complex structure of a nucleoprotein from the OB-fold family has been reported (31,32). The presented NMR structure and dynamics of Bs-CspB in complex with dT7 provides now important insights into the molecular grounds of complex formation for the first representative of the CSP family. Based on the NMR data, pinpointed biochemical experiments were performed to understand the recognition of ssDNA by CSP in solution. Although CSP are believed to function in vivo by interaction with RNA, we had to choose ssDNA fragments for the current study because the quality of NMR spectra of Bs-CspB/RNA complexes was not sufficient for high-resolution structural investigations. A strong line broadening of many protein resonances during titration with the RNA fragment UUAUUAG reflect changes in the conformation dynamics of the complex, which require further investigations. We expect that the main findings in terms of structure and dynamics of Bs-CspB in complex with ssDNA will hold for Bs-CspB/RNA complexes because of the following two observations. During NMR titration of Bs-CspB with TTATTAG (Table 1) and UUAUUAG, the chemical shifts of the same residues were affected by the nucleic acids as with dT7 (64). The KD of Bs-CspB in complex with UUAUUAG was increased by a factor of 3 compared to TTATTAG but remained in the sub-micromolar range indicating a high affinity for the RNA fragment.

Preferential binding of thymine rich ssDNA by Bs-CspB

Seventeen oligonucleotides were tested to identify an ssDNA fragment with high binding affinity at 1:1 stoichiometry. Table 1 confirms the earlier proposed assumption that Bs-CspB covers 6–7 nt with a preference for thymine rich stretches (7). Starting from dT4, an additional thymidine decreases the KD by a factor of 3 until the seventh thymidine, which causes a drop by a factor of 200. Above 12 and 25 nt, the stoichiometry increases from 1:1 to 2:1 and 3:1, respectively. Several fragments were derived from the putative binding site of CSP to the 5′-UTR region of the Bs-CspB mRNA (cold box1 and cold box2) (15,65). These fragments show the general trend described above indicating that at least under in vitro conditions neither a preferential binding towards the cold box nor to the sense or anti-sense sequence stands out. The Y-box25 fragment had been earlier investigated with limited conclusions due to the low binding affinity, a 3:1 stoichiometry and poor NMR spectra (27). Shorter Y-box25 derived fragments showed low binding affinity.

Bs-CspB and all variants showed a strong quench of the intrinsic fluorescence, indicating that W8 is close to the binding site or at least changes the conformation of the indole ring upon binding. This is confirmed by the NMR structure and the NMR titration data, where Δ(δMW(1H,15N)) of the W8 side chain shows one of the highest values whereas the backbone is almost not affected (Figure 1). Wild-type Bs-CspB revealed a low preference for thymidine compared to cytidine at positions 1, 3, 4, 5 and 7 of the heptaoligomere and a moderate preference at position 6.

Independent from the analyzed protein variant and ssDNA fragment (Table 3) the association rate constant varied only marginally between 1.39 × 108 M−1 s−1 and 4.29 × 108 M−1s−1. In contrast, koff covered a wide range between 1.0 and 177 s−1 revealing a good correlation between the variation of koff derived from stopped-flow experiments and the KD values determined at equilibrium. Therefore, we propose from this extended dataset that Bs-CspB associates with single-stranded nucleic acids at a constant rate with a low sequence specificity. This might be important for functioning as an ‘RNA chaperone’ or for mapping the nucleic acid for a high affinity binding site. A low affinity of Bs-CspB because of a missing interaction with an essential side chain at the protein-binding site or because of a less favorable base composition at the nucleic acid is only reflected in an increased koff.

Structural elucidations of Bs-CspB/dT7

The structural consequences of the tight binding of Bs-CspB in solution have been investigated by NMR spectroscopy. To localize the binding site for dT7 we employed both the change in protein chemical shifts during NMR titration experiments (blue in Figure 5) and the mutational data (yellow in Figure 5). Differences in the backbone conformation between free and complexed Bs-CspB are mainly located in the loops β1–β2 and β3–β4. These stretches correspond to the group of residues with high Δ(δMW(1H,15N)) values including the proposed RNP1 (13–20) and RNP2 (26–30) motives (28,29). The mutational data revealed the importance of the side chains of F15, F17, F27, H29, F30 and F38 (yellow in Figure 5). All these residues showed also high Δ(δMW) values of the backbone resonances. Additionally, the side chains of W8, F15, D25, F30, Q34, Q59 and N62 stand out from changes in chemical shifts upon binding. For the here reported representative of CSP, the tertiary structure of the five β-strands are almost invariant upon binding. From those OB-fold proteins, where the structure of both the free and nucleic acid-bound state are known, large conformational changes were mainly reported for loop β1–β2 and for the relative orientation of several mainly invariant OB-fold domains recognizing larger stretches of ssDNA or ssRNA (30). We can conclude that CSP also use the β-barrel as scaffold for ssDNA recognition, which is a general feature of OB-fold proteins, and loop β3–β4 to extend the ssDNA recognition site. With only 67 residues, Bs-CspB is one of the smallest OB-fold proteins.

Figure 5

Nucleic acid-binding site of Bs-CspB determined by NMR spectroscopy and site-directed mutagenesis at (a) front view and (b) back view. The following residues experienced substantial chemical shift changes of their NMR resonances upon binding to dT7 and are indicated in blue: 7, 10–17, 20, 25, 26–31, 33, 35, 38–42, 54, 56–59 (backbone) and 8, 25, 34, 59, 62 (side chain). Yellow highlighted side chains illustrate aromatic residues, for which the KD increased above 100 nM after substitution with alanine. For this illustration the NMR structure with the lowest energy was used.

These structural elucidations in solution of the Bs-CspB conformation and ssDNA-binding site in Bs-CspB/dT7 are in very good agreement with the structure of Bs-CspB/dT6 determined by X-ray crystallography (2ES2.pdb; Max et al., manuscript to be published elsewhere). All residues, which are in contact with the ssDNA via the backbone showed high Δ(δMW) values of the amides and/or the Hα/Cα nuclei. All residues with side chains involved in stacking interactions or hydrogen bonds stand out as well from the Δ(δMW) and Δ(δ(1H)) values. The superposition of the 18 lowest energy NMR structures and the crystal structure is given in Figure 6. It revealed an averaged root mean square deviation (r.m.s.d.) for the heavy backbone atoms of residues 1–67 of 1.44 Å (Table 2). Remaining deviations are located in the loops β2–β3, β3–β4 and β4–β5. It is unlikely that binding of the ssDNA causes the deviation of loop β2–β3 because the C-terminal part of loop β2–β3 (Q21–D25) depicts the same deviation between NMR structure and X-ray structure for the free form and between X-ray structures of Bs-CspB determined at different resolutions (1CSP.pdb and 1CSQ.pdb). For a comparison of the side chain conformations between solution and the crystal structure we have to consider on the NMR side the missing dT7 moiety. Still, W8 is well defined and adopts in solution a conformation different from the crystal structure and the free protein. F38 in loop β3–β4 is involved in dT7 binding shown by the increased KD of the F38A variant, the increase of the backbone S2 indicating a reduction in dynamics and the exchange protected amide proton of adjacent G37 in Bs-CspB/dT7. Beside W8 and F38, all aromatic side chains involved in dT7 stacking form in solution a very similar ssDNA recognition pattern compared to the crystal.

Figure 6

Backbone superposition of the 18 lowest energy NMR structures (yellow) of Bs-CspB/dT7 and the crystal structure (green) of Bs-CspB/dT6 (2ES2.pdb). The side chain conformations of the aromatic residues, which facilitate binding of dT7, are depicted and labeled.

Dynamics and stability of Bs-CspB/dT7

Binding of nucleic acids changes the dynamics and stability of the interacting protein. For an induced fit mechanism one would expect a dynamic recognition site (66). A reduction in the sub-nanosecond dynamics upon binding has been reported, for example, for the DNA-binding loop of the Lac-repressor head piece, which forms a rigid hinge helix upon binding to cognate DNA operators (67). Additionally, the protection factors of the amide protons of this protein increased globally after complex formation by five orders of magnitude (68). Therefore, a change in the relaxation and H/D exchange rates upon binding and a comparison of the derived order parameters, chemical exchange rates and protection factors with the values of the free protein are additional indicators for ssDNA binding to sole chemical shift changes.

Correlating the backbone r.m.s.d. values of the family of presented Bs-CspB structures in the Bs-CspB/dT7 complex with general order parameters S2 reveal those regions, which do not converge during structure calculation due to sub-nanosecond dynamics. Residues 34–41 in loop β3–β4 exhibit increased internal dynamics in the free protein (Figure 4) but only G35, E36 and G37 remain dynamic in the Bs-CspB/dT7 complex indicating that binding of dT7 reduces the local dynamics of the C-terminal end of loop β3–β4. All side chains involved in ssDNA binding are presented by free Bs-CspB on a backbone, which is rigid on the sub-nanosecond timescale.

Bs-CspB functions as a CSP implying a strong temperature dependence of the nucleic acid affinity. Indeed, we found a strong increase of KD with increasing temperature of Bs-CspB/dT7: 1.8 ± 0.4 nM at 15°C, 60 ± 4 nM at 25°C and 464 ± 31 nM at 35°C. This suggests a strong entropic contribution counterbalancing the gain in Gibbs free energy upon ssDNA binding. Thermodynamic contributions from the folding–unfolding equilibrium of CspB in this temperature range are expected to be small because binding and stability of CspB are not correlated (27) and the increase of the population of the unfolded state of free CspB between 15 and 35°C is below 5% (69). From the reduction of backbone dynamics in Bs-CspB/dT7, revealed by comparing S2 values of free and bound Bs-CspB, the contribution to the entropy change of the protein chain upon binding can be estimated on a nanosecond-to-picosecond timescale (70,71). The sum of this conformational entropy change associated with the individual N-H bond vectors results in −470 J/(mol·K).

Two further observations by NMR spectroscopy show that dT7 stabilizes Bs-CspB globally during complex formation: a significant reduction of chemical exchange rate constants Rex (Figure 4) and an increase of protection of amide protons against exchange (Figure 2). We attribute the change in Rex to a shift of the equilibrium between folded and unfolded Bs-CspB towards the native, dT7 bound state. At 0.7 mM Bs-CspB, the pseudo first-order association rate constant exceeds 200 000 s−1 and the population of uncomplexed Bs-CspB at a KD of 1.8 nM is too small to give rise to Rex contributions from the binding reaction or the folding–unfolding reaction of free Bs-CspB assuming the following simple scheme, where U denotes the unfolded state of Bs-CspB: A loss in detectable Rex contributions to the transverse relaxation rate because of a stabilization of Bs-CspB by ethylene glycol has been reported previously and discussed (48,62). For the most protected amide protons, only the global unfolding of the entire polypeptide chain will allow exchange with deuterium and the rate of exchange must be much slower than the folding rate to derive thermodynamic stability data from H/D exchange rates (EX2 limit). This has been confirmed for free Bs-CspB by the pH dependence of the exchange rate between pH 6 und 7 (data not shown). A KD in the nM range increases the protection factor by several orders of magnitude, if only the unfolded state U is the exchange competent form, which has been observed for example for the Lac-repressor headpiece during binding with DNA operators (68). For Bs-CspB, the equivalent global effect was detected (Figure 2). However, the protection factors increased only by a factor of 1000, which is about two orders of magnitude lower than expected for the simple three-state model depicted above. Therefore, we assume that despite the low dissociation constant, the Bs-CspB/dT7 complex remains dynamic so that there is still enough local unfolding present to explain the observed H/D exchange rates.