cold shock response protein

In the solution structure of the ribosome-associated cold shock response protein Yfia of Escherichia coli in the free state two structural segments can be distinguished: a well structured, rigid N-terminal part displaying a betaalphabetabetabetaalpha topology and a flexible C-terminal tail comprising last 20 amino-acid residues. The backbone dynamics of Yfia protein was studied by (15)N nuclear magnetic relaxation at three magnetic fields and analyzed using model-free approach. The overall diffusional tumbling of the N-terminal part is strongly anisotropic with a number of short stretches showing increased mobility either on a subnanosecond time scale, or a micro- to millisecond time scale, or both. In contrast, the unstructured polypeptide chain of the C-terminal part, which cannot be regarded as a rigid structure, shows the predominance of fast local motions over slower ones, both becoming faster closer to the C-terminus.


INTRODUCTION
Living organisms react to the rapidly changing environmental conditions with the production of specialized stress factors while limiting the ribosomal synthesis of most metabolic proteins.Recently, a ribosome-related mechanism of cold shock adaptation has been elucidated in Escherichia coli (Agafonov et al., 1999;2001).Ribosomes, which serve as fundamental checkpoints for sensing changes in nutrient level and in temperature, are composed of two subunits (30S and 50S) that undergo a cycle of association and dissociation during protein synthesis.A sudden drop in temperature causes the bacterium to initiate the production of protein Yfia (protein Y, PY) which instantly associates with the 30S ribosomal subunit while blocking the tRNA-and mRNA-binding channel.Upon cold shock Yfia competes with translation initiation factors IF1 and IF3 that are required for the dissociation of ribosomal subunits (Vila-Sanjurjo et al., 2004).Formation of the Yfia-30S complex provides a storage mechanism for intact 70S ribosomes and allows their rapid reactivation on improving environmental conditions.
Variants of protein Y are widely dispersed in bacteria but homologous proteins can also be found in chloroplasts of plants such as Spinacia oleracea (PSrp-1) (Johnson et al., 1990;Agafonov et al., 1999).Members of the protein Y family are generally thought to be involved in the process of adaptation to changing enviromental conditions.Recently, two solution structures of protein Y from E. coli (Kalinin et al., 2002;Ye et al., 2002;Rak et al., 2002) and one from the bacteria Haemophilus influenzae (Parsons et al., 2001) as well as the crystal structure of a complex of Yfia bound to the E. coli 30S ribosomal subunit (11 Å resolution) (Vila-Sanjurjo et al., 2004) have been published revealing the structural basis for the control of translation initiation.The protein structure resembles a βαβββα folding topology and represents a compact two layered sandwich of two nearly parallel α-helices packed against the same side of a four-stranded β-sheet (Fig. 1).Since the C-terminal 20 amino acids are flexible in solution, the ribosomebound structure was determined by docking a truncated model of protein Y lacking this region into the average electron density.In the crystal complex Yfia resides in close proximity to 16S rRNA residues which are known to be universally conserved markers for the A and P sites.Bases G926, C1400, C1402 and A1493 are protected by the positively charged side chains of amino acids R22, K25, K28, K79, R82 and K86 of Yfia.These residues block tRNA from the A-site and inhibit binding of deacetylated tRNA to the P-site.
Recently the relationship between the internal dynamics of macromolecules and their biological functions has been the subject of much research (Palmer, 2004;Kay, 2005).Information about the dynamics of a protein backbone from 15 N NMR relaxation studies is typically based on the measurements of 15 N longitudinal (R 1 ) and transverse (R 2 ) relaxation rates and the steady-state 1 H-15 N nuclear Overhauser enhancement ({ 1 H}-15 N NOE) giving access to the mobility of N-H vectors (Korzhnev et al., 2001).Owing to the complexity of motions in proteins a specific type of motion was seldom assumed and the experimental data have been usually analyzed by the model-free approach (MFA) (Lipari & Szabo, 1982).Analysis of relaxation data reveals that, with the exception of loops and termini, protein backbone motions in native conformation are severely restricted on the nanosecond timescale.On the other hand, the backbone motions of partially folded proteins usually show a significant contribution of motions with the timescale comparable to or slower than the overall correlation time (Alexandrescu & Shortle, 1994;Brutscher et al., 1997;Ochsenbein et al., 2004).

MATERIALS AND METHODS
Sample preparation.E. coli DNA encoding Yfia (sequence available in the SwissProt DataBank; entry code: YFIA_ECOLI; accession number P11285) was cloned into the vector pET11c.After transfection in BL21(DE3) the protein was expressed and purified as described earlier (Kalinin et al., 2002).A typical NMR sample contained 50 mM KH 2 PO 4 buffer, 50 mM LiCl and 10% (v/v) D 2 O, with a pH of 6.8 and the protein concentration of 0.8 mM.
NMR experiments.NMR data were recorded at 300 K on a Varian INOVA 400, a Varian UNITY+ 500 and a Varian INOVA 600 spectrometers at 9.4, 11.7, and 14.1 T, respectively.Longitudinal (R 1 ) and transverse (R 2 ) relaxation rates were measured at three magnetic field strengths using a sensitivity enhanced [ 15 N-1 H] HSQC pulse sequence (Kay et al., 1992a) with the option of either R 1 or R 2 measurements of 15 N nuclei (Farrow et al., 1994).In the case of R 1 twelve evolution times within the range 10-1450 ms were used.The R 2 relaxation rate measurements were performed with the Carr-Purcell-Meiboom-Gill (CPMG) pulse train.Refocusing time of 650 μs was used and eight evolution times covered the range 10-250 ms.The acquisition parameters for R 1 and R 2 measurements were identical with the exception of the delay between π ( 1 H) pulses used for the cross-correlation effect suppression (Kay et al., 1992b).Delays of 5 ms and 10 ms were used in R 1 and R 2 measurements, respectively.The relaxation delay of 2.5 s was employed in both experiments.

Backbone dynamics of Yfia protein
{ 1 H}-15 N NOEs were measured at 9.4 and 11.7 T with a pulse sequence included in the ProteinPack Varian Inc. (Palo Alto, USA) software.The relaxation delay of 7 s and 1 H saturation of 3 s were employed.Standard NOE measurement (spectra with and without 1 H saturation) was performed at 9.4 T, whereas the dynamic variant of the experiment was chosen for the NOE determination at 11.7 T (Zhukov & Ejchart, 1999). 1 H saturation time was set to 0. 1, 0.15,  0.3, 0.7, 1.3, and 2.1 s.In all relaxation measurements 15 N decoupling during the acquisition was applied using the GARP sequence with γB 2 /2π = 3200 Hz (Shaka et al., 1985).All spectra were processed by the NMRPipe (Delaglio et al., 1995) and analyzed by the XEASY (Bartels et al., 1995) programs.Relaxation parameters were determined by the two-parameter nonlinear least-squares fit of cross-peak heights to a single exponential decay.Errors in the determination of the relaxation rates were obtained from the covariance matrix.Error in the NOE values was estimated from assuming that the uncertainty in the peak heights in the spectra recorded with and without 1 H saturation was equal to rms noise in each of these two spectra (Tjandra et al., 1995).
Analysis of relaxation data.Two relaxation mechanisms have to be taken into account for the nuclear spin relaxation of amide nitrogens in proteins: the chemical shift anisotropy of a nitrogen nucleus and the dipole-dipole interaction between a nitrogen and the hydrogen directly bound to it.Equations describing relaxation parameters in terms of spectral density functions are given as (Korzhnev et al., 2001): where C = ω N ∆σ and other symbols have their usual meaning.In further calculations the values of N-H distance, r NH = 0.104 nm (Case, 1999), and 15 N chemical shift anisotropy, ∆σ = -170 ppm (Tjandra et al., 1996a;1996b), were used.The additional term R ex takes into account the conformational exchange contribution to R 2 resulting from processes in the micro-to millisecond time scale often named the chemical exchange effect (Stone et al., 1992;Korzhnev et al., 2001).Such processes, slower than the molecular tumbling, but fast enough to average chemical shifts, can influence transverse relaxation rates determined using the CPMG method.The R ex contribution to the transverse relaxation rate is pro-portional to the square of the chemical shift difference between exchanging states, ∆δ, and ω N -the Larmor frequency at the fast exchange limit.It should be pointed out that conformational exchange mechanism is able to participate in the transverse relaxation only if ∆δ ≠ 0. This term can be written as R ex = Φω N 2 (Peng & Wagner, 1995).The proportionality factor Φ represents the effectiveness of conformational exchange processes.
Model-free approach spectral density function takes the form (Lipari & Szabo, 1982): The isotropic overall motion is described by the correlation time τ R and internal motion(s) by a generalized order parameter S which is a measure of the degree of spatial restriction of the motion and an effective correlation time τ int corresponding to the rate of this motion.
In the case of anisotropic overall motion the spectral density function becomes more complex.Combining the model-free approach with axially anisotropic overall tumbling (Woessner, 1962) the spectral density function takes the form (Barbato et al., 1992): where α is the angle between the N-H vector and the unique axis of the rotational diffusion tensor.Coefficients A k (α) have the form: The overall correlation times τ k are defined as: and the effective correlation times τ i,k are as follows: int .D  and D ⊥ are parallel and perpendicular components of the rotational diffusion tensor.Additional two parameters, the polar angles θ and ϕ, relate the direction of the unique axis of the diffusion tensor to the molecule fixed coordinate system.The geometric coefficients A k (α) were calculated using the atomic coordinates of the NMR-derived structure of Yfia (PDB accession no.1N3G).
Least-squares procedure used to optimize the model parameters consisted of a minimization through a grid-search of the target function χ given by: where the sum was over eight relaxation parameters (R 1 s and R 2 s at three magnetic field strengths and NOEs at two magnetic field strengths) for each of N residues, and P ij,calc were the appropriate relaxation parameters calculated from the assumed model.The σ ij values were the corresponding standard deviations of experimentally derived P ij,exp .The minimization procedure delivered four global parameters D  , D ⊥ , θ, ϕ and N sets of local parameters S 2 j , τ j,int , and Φ j .Model parameter uncertainties derived in the minimization of target function were obtained as standard deviations from 200 Monte Carlo simulations (Press et al., 1986).

RESULTS AND DISCUSSION
Resonance assignments of 15 N nuclei in Yfia protein have been published recently (Kalinin et al., 2002).Eight 15 N relaxation parameters P exp were measured for 98 backbone amide groups out of 113 residues.The lacking data comprise the Nterminal M1, four prolines (P14, P37, P45, P57), stretch of residues Q31-I35 showing extremely weak correlations, and five other residues with strongly superposed crosspeaks (T2, M3, I5, A94, A104).A graphic representation of the relaxation parameters is presented in Fig. 2 and the actual values are given in the Supplementary Material.All relaxation parameters show the expected dependence on the magnetic field strength: the R 1 s decrease whereas R 2 s and NOEs increase with B 0 .The negative NOE and markedly smaller R 2 values observed for the C-terminal part of the protein are typical for unstructured protein segments (Alexandrescu & Shortle, 1994).Analysis of the 15 N relaxation data was carried out separately for the well-structured N-terminal part (residues 1-91) and the flexible, unstructured C-terminal part (residues 92-113).

N-terminal part of Yfia protein
The MFA with anisotropic overall motion was applied to analyse the dynamics of the N-terminal part of Yfia protein.Relaxation data for 78 residues were fitted simultaneously without any arbitrary omission of residues displaying deviations of R 2 /R 1 ratios from the average and/or smaller NOEs (Clore et al., 1990).The model included global parameters D  , D ⊥ , θ, and ϕ, as well as local, residue-specific parameters, S 2 , τ int , and Φ.
A graphic representation of the model-free approach parameters obtained for the anisotropic model is given in Fig. 3.The S 2 values are related to the local mobility of the N-H vectors on the subnanosecond timescale.For a completely restricted local motion S 2 = 1, whereas S 2 = 0 corresponds to fully unrestricted motion.Secondary structure segments show uniform, large S 2 values which decrease at their ends.This effect is especially well visible for the C-terminal residue of helix α 2 , H90, which can hardly be included in a rigid structural element on the basis of its S 2 value.
Residues in turns between β 2 -β 3 and β 3 -β 4 display increased mobility on the ns-ps time scale (smaller S 2 values) in contrast to those between the β 4 -α 2 secondary structure elements.The loop between β 1 -α 1 shows dispersion of S 2 values, whereas a lack of experimental data precludes any firm conclusion concerning fast motion mobility in the α 1 -β 2 loop.Nevertheless, small S 2 values for W30 (S 2 = 0.74) and N36 (S 2 = 0.58) may point to an increased amplitude of fast local motions in this area.Errors of the NOE values at 9.4 T were estimated as described by Alexandrescu and Shortle (1994).Data were obtained at 9.4 T (•), 11.7 T (), and 14.1 T ().Structure elements are represented by horizontal lines.

Backbone dynamics of Yfia protein
The Φ parameter represents slow, often largescale conformational motions, on the ms-µs time scale comprising several adjacent residues (Korzhnev et al., 2001).Several such areas showing elevated Φ values can be found in the N-terminal part of Yfia protein; residues 4-10 comprising β 1 and a part of the following loop, residues 25-30 and 36 constituting the C-end of α 1 and the following turn, residues 53-55 constituting the C-end of β 3 .Residues 38-42 corresponding approximately to β 2 and residues 60-72 comprising β 4 and the β 4 -α 2 turn also display small but uniform conformational exchange effects.On the basis of the large Φ values for residues W30 and N36 one may speculate that the extremely weak correlations for residues Q31-I35 precluding determination of their relaxation parameters are due to the conformational exchange not fast enough to fully average chemical shifts and resulting in strong broadening of the resonances (Cavanagh et al., 1996).
In fact, a significant number of residues forming the four-stranded β-sheet show a significant amount of slow conformational motions allowing us to conclude that its bigger part undergoes relatively slow, large-scale conformational motion.One can speculate that this conformational flexibility is characteristic only for free Yfia protein and it is probably suppressed when the protein is associated with the ribosome.
In the N-terminal part of Yfia protein increased mobility on the ns-ps timescale often goes together with the increased mobility on the ms-µs timescale.Many residues, especially in turns, show both small S 2 values and large Φ values indicating extremely high mobility of those parts of the protein in a free state.
A short comment should be given on the values of correlation times for internal motions, τ int .Many of them seem to be unexpectedly large and becoming of the same order of magnitude as the effective overall correlation time.Nevertheless, one should take into account that τ int are the most prone of all local MFA parameters (S 2 , τ int , Φ) to experimental inaccuracies.This feature is reflected by the large error bars in the central part of Fig. 3. Equally important is that τ int values are sensitive to the assumed values of r NH and ∆σ used in calculations.For instance, use of r NH = 0.102 nm instead of 0.104 nm could decrease τ int values on average 2.7 times.In general, the pattern of τ int follows that of S 2 .Less local freedom corresponds to longer τ int s and this regularity seems to be self-explanatory.

C-terminal part of Yfia protein
Since the unstructured, flexible C-terminal part cannot be regarded as a rigid rotor, a description of its motion in terms of a single overall correlation time may not be appropriate.In such a case the motion of the N-H vector in the l-th residue can be described by the model-free approach spectral density functions wherein the global parameter τ R is substituted with a local one, τ R,l .(Alexandrescu & Shortle, 1994;Brutscher et al., 1997).Alternatively, the use of a distribution of correlation times allows one to account for intramolecular motions comparable to or slower than overall molecular tumbling (Ochsenbein et al., 2002).A number of models utilizing different distribution functions have been used for the interpretation of nuclear relaxation data in synthetic polymers (Ward & Klein, 1996) and polypeptide chains (Buevich et al., 2001).Since the model utilizing local correlation times describes well the dynamics of an unstructured tail attached to a fully folded fragment -the case of Yfia proteinwe used it in the analysis of the 15 N relaxation data of the C-terminal part of the protein.The results are shown in Fig. 4. Φ values are equal to zero within the accuracy limit and are not included.S 2 values are much smaller than those in the N-terminal part (except residues 90 and 91) pointing to the high preference of fast local motions characterized by τ int over the slower motions described by τ R,l .Since the N-domain anchors the unstructured C-domain, the systematic decrease of both correlation times when approaching the C-terminus reflects the increasing amount of conformational freedom.It strongly supports the conclusion that the C-terminus is unstructured and randomly changes its conformation.The MFA parameters calculated for residues A104-E108, which showed a number of cross peaks characteristic for a helix in NOESY spectra (Ye et al., 2002), do not reflect any increase of conformational stiffness for these residues.

CONCLUSION
The backbone dynamics of Yfia protein is strongly diversified.The overall tumbling of the rigid N-terminal part comprising 91 amino-acid residues is typical for native proteins, whereas the intense local motions within the C-terminal part (22 amino-acid residues) are characteristic for unstructured or denaturated proteins.A simultaneous ap-pearance of so different dynamic behaviours in the same protein molecule is very unusual.
Although the role of Yfia dynamics on the translation inhibition by directly blocking the P-site of the ribosome seems to be minor, it might play a crucial role in the dissociation of protein Y from the ribosome under elevated temperatures (37°C).The slow conformational motion within the N-terminus might be reduced under cold shock allowing strong binding of protein Y to the ribosomal subunit.Under elevated temperatures the internal motion may weaken the binding to the ribosom allowing the competitive binding of initiation factor IF1 to an overlapping binding site.However, the role of the C-terminus is unknown.Its high internal dynamic and flexibility suggest an unknown binding partner.

Figure 2 .
Figure 2. Relaxation data as a function of Yfia protein structure.A -R 1 s, B -R 2 s, and C -NOEs.Error bars correspond to standard deviations of data points in individual fits.Errors of the NOE values at 9.4 T were estimated as described by Alexandrescu and Shortle (1994).Data were obtained at 9.4 T (•), 11.7 T (), and 14.1 T ().Structure elements are represented by horizontal lines.

Figure 3 .
Figure 3. Order parameters S 2 , correlation times of internal motions τ int , and Φ factors characterizing conformational exchange processes along the sequence of N-terminal part of Yfia protein (amino-acid residues 1-91).Error bars represent appropriate standard deviations obtained from Monte Carlo simulations.Structure elements are represented by horizontal lines.

Figure 4 .
Figure 4. Residue-specific correlation times, τ R,i (•) and τ int (), (A) and order parameters (B) obtained for C-terminal part of Yfia protein (amino-acid residues 91-113).Error bars represent appropriate standard deviations obtained from Monte Carlo simulations.Regression lines in part A correspond to correlation coefficients: r = 0.59 for τ R,i and r = 0.79 for τ int .