Báo cáo khoa học: Structural mobility of the monomeric C-terminal domain of the HIV-1 capsid protein pptx

Structural mobility of the monomeric C-terminal domain of the HIV-1 capsid protein ´ ´ Luis A Alcaraz1,*, Marta del Alamo2, Mauricio G Mateu2 and Jose L Neira1,3,* ´ndez, Elche (Alicante), Spain ´ Instituto de Biologıa Molecular y Celular, Universidad Miguel Herna ´ ´ Centro de Biologıa Molecular ‘Severo Ochoa’ (CSIC-UAM), Universidad Autonoma de Madrid, Spain Biocomputation and Complex Systems Physics Institute, Zaragoza, Spain Keywords flexibility; human immunodeficiency virus; NMR; structure Correspondence ´ J L Neira, Instituto de Biologıa Molecular y ´ Celular, Edificio Torregaitan, Universidad ´ Miguel Hernandez, Avenida del Ferrocarril s ⁄ n, 03202 Elche (Alicante), Spain Fax: +34 966 658 758 Tel: +34 966 658 459 E-mail: jlneira@umh.es *These authors contributed equally to this work (Received February 2008, revised 22 April 2008, accepted 24 April 2008) doi:10.1111/j.1742-4658.2008.06478.x The capsid protein of HIV-1 (p24) (CA) forms the mature capsid of the human immunodeficiency virus Capsid assembly involves hexamerization of the N-terminal domain and dimerization of the C-terminal domain of CA (CAC), and both domains constitute potential targets for anti-HIV therapy CAC homodimerization occurs mainly through its second helix, and it is abolished when its sole tryptophan is mutated to alanine This mutant, CACW40A, resembles a transient monomeric intermediate formed during dimerization Its tertiary structure is similar to that of the subunits in the dimeric, non-mutated CAC, but the segment corresponding to the second helix samples different conformations The present study comprises a comprehensive examination of the CACW40A internal dynamics The results obtained, with movements sampling a wide time regime (from picoto milliseconds), demonstrate the high flexibility of the whole monomeric protein The conformational exchange phenomena on the micro-to-millisecond time scale suggest a role for internal motions in the monomer– monomer interactions and, thus, flexibility of the polypeptide chain is likely to contribute to the ability of the protein to adopt different conformational states, depending on the biological environment Dynamic processes in proteins contribute toward defining their structure and function, including protein folding, association and ligand binding [1] The main challenge in all structural and dynamic studies is to find a relationship between the structural and mobility results, as well as protein function Recent advances in isotopic labelling techniques [2] and NMR spectroscopy [3] have raised interest in protein dynamics as provided by heteronuclear relaxation measurements [4–6] Relaxation of the particular backbone amide 15 N provides details of rotational tumbling, and the movement of the internal N–H bonds [3] allows conclusions to be drawn on the redistribution of conformational entropy upon folding and ⁄ or binding [1] The structural retroviral polyprotein (Gag) of HIV-1 forms the immature capsid, and is subsequently cleaved by the viral protease into several mature proteins: the matrix, the capsid protein of HIV-1 (p24) (CA), the nucleocapsid and p6, as well as the spacer peptides p2 and p1 [7–9] After proteolytic cleavage of Gag, CA reassembles to form the mature capsid [10] In vitro, CA spontaneously assembles into cylindrical structures and cones resembling the viral capsid [11–15] Dimerization through its C-terminal domain (CAC) is a driving force in virus assembly [14–17] Recent studies of the mature capsid lattice have shown that CAC connects through homodimerization the CA hexamers, which Abbreviations CA, capsid protein of HIV-1 (p24); CAC, C-terminal domain of CA, comprising residues 146–231 of the intact protein; CACW40A, mutant of CAC with Ala instead of Trp at position 184 of CA; CSA, chemical shift anisotropy; Gag, the structural retroviral polyprotein of retroviruses; NOE, nuclear Overhauser effect FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS 3299 Dynamics of monomeric CAC L A Alcaraz et al form the mature capsid, and also interacts with the CA N-terminal domain [18] The CA of HIV-1 is formed by two independently folded domains separated by a flexible linker [19–22] The N-terminal domain (residues 1–146 of the intact protein) is composed of five coiled-coil a-helices, with two additional short a-helices following an extended proline-rich loop [19–21] The CAC domain (residues 147–231) is a dimer both in solution and in the crystal form [22,23] Each CAC monomer is composed of a short 310-helix followed by a strand and four a-helices: a-helix (residues 160–172), a-helix (residues 178– 191), a-helix (residues 195–202) and a-helix (residues 209–114), which are connected by short loops or turn-like structures The dimerization interface is formed by the mutual docking of a-helix from each monomer, with the side chains of each tryptophan (Trp184) deeply buried in the dimer interface [22,23] Our previous folding equilibrium analyses indicate that the monomeric CAC mutant Trp184Ala, CACW40A, resembles a transient monomeric intermediate formed during dimerization [24,25] In the present study, for sake of clarity, the mutant is referred to as CACW40A to denote the position of the mutation in the C-terminal domain; in addition, the amino acids of CACW40A are numbered from its first residue (i.e the added N-terminal methionine is Met1, and the second residue is Ser2, which corresponds to Ser146 in the numbering of the intact CA) The CACW40A protein is monomeric, and its structure is similar to that of the subunits in the dimeric, non-mutated CAC, but, in the monomeric form, the segment corresponding to the second helix samples different conformations [26] (Fig 1) At the end of this region, several hydrophobic residues are buried and, as a consequence, the last two helices are rotated compared to their position in dimeric non-mutated CAC Thus, from a structural point of view, only the dimerization interface has substantially changed To determine whether the apparent dynamic character of this region is shown by other polypeptide patches, we have studied the dynamics of monomeric CACW40A Flexibility is often associated with interfaces, and it is well known that complex formation (either in an oligomer or in a more simple substrate– enzyme reaction) can lead to conformational and dynamic changes at some, if not all, of the residues involved [27] In our previous description of the structure of CACW40A, we observed a high flexibility in the region involved in the dimerization interface (as concluded from the absence of signals in the HSQC) [26] In addition, millisecond-to-second dynamics was addressed by following the hydrogen-exchange behav3300 Fig Structure of CACW40A UCSF CHIMERA software was used to render the model from the 2JO0 Protein Data Bank deposited structure: the first a-helix is in blue; the second one in green; and the last a-helix is shown in yellow The single turn of a 310-helix at the N-terminus of the protein is shown in red iour In the present study, we have advanced a step further and describe the pico-to-millisecond dynamics The present study aims to ascertain whether there are regions within the CACW40A that exhibit particular high flexibility (i.e whether the region comprising the dimerization interface in the non-mutated CAC is not the sole highly mobile region) This would indicate a lower energy barrier to structural rearrangements throughout the whole structure The results obtained indicate not only that the dimerization interface displays a high flexibility, but also that the rest of the protein is affected by movements on the pico-to-millisecond time regime This mobility, as shown by the dimeric non-mutated CAC, is important in the virus cycle, as confirmed by structural studies of CAC in the presence of various molecules and agents [28–31] Results Relaxation measurements of CACW40A Mean R1 (= ⁄ T1, the longitudinal relaxation rate) was 2.95 s)1 (range 1.49–3.69) (Fig 2A) (see supplementary Table S1) Residues in the first a-helix presented a mean of 2.90 s)1 (range 2.56–3.69); the second a-helix presented a mean of 3.06 s)1 (range 2.79–3.15); the third a-helix presented a mean of 2.78 s)1 (range 2.26–3.05); and, finally, amino acids in the loop region presented a mean of 3.18 s)1 (range 2.91–3.47) There FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS L A Alcaraz et al Fig Relaxation rates of CACW40A The relaxation rates are shown for (A) R1, (B) R2 and (C) 15N-1H NOE for CACW40A at 11.7 T Sample conditions were 293 K, pH 7.0 in 0.1 M phosphate buffer The cylinders at the top of each panel indicate the three a-helices was no clear correlation between the elements of secondary structure and the values of R1 Similar findings have been found in proteins of similar size at the same magnetic field, such as eglin c [32,33], CI2 [34], and the GAL4 domain [35,36] Mean R2 (= ⁄ T2, the transversal relaxation rates) was 11.9 s)1 (range 6.3–14.7) (Fig 2B) (see supplementary Table S1) Residues in the first a-helix presented a mean of 12.3 s)1 (range 9.1–14.0); the second a-helix presented a mean of 13.2 s)1 (range 11.9–14.2); the third a-helix presented a mean of 11.3 s)1 (range 8.19– 13.5); and, finally, amino acids in the loop region presented a mean of 12.3 s)1 (range 9.2–14.7) As with R1, there was no clear correlation between the elements Dynamics of monomeric CAC of secondary structure and the values of R2 However, it is interesting to note that the values of R2 in CACW40A were clearly higher than those of other proteins of similar size measured at the same magnetic field (eglin c, CI2 or GAL4 with average values of 5.6, and s)1, respectively [32–36]; GAL4 is the most disordered protein, and thus shows the highest values of R2) The mean of the nuclear Overhauser effect (NOE) in CACW40A was 0.60 (range 0.28–0.87) (Fig 2C; see also supplementary Table S1) This mean is lower than the value of 0.79 expected from theoretical considerations at a field strength of 11.7 T [37] These results (together with those of the R2 described above) suggest a high flexibility of the whole backbone of CACW40A; interestingly, a study of dynamics of the C-terminal region of dimeric CAC also shows low NOE values [38], and extensive signal broadening has been observed in the assignment of dimeric non-mutated CAC [31] The residues with low NOE values (< 0.65) in CACW40A were Ile9 (at the C-cap of the 310-helix); Tyr20 (at the beginning of the first helix); Lys26 and Ala30 (at the C-cap of the first helix); Val37 (in the middle of the long disordered loop); Thr44, Val47 and Gln48 (at the long disordered loop); Lys55, Thr56, Ile57 and Leu58 (at the second helix); Ala60, Gly62 (in the type II b-turn); Leu67 and Met71 (in the second helix); and Gly78 and Gly81 (at the C terminus of the protein) For the different regions, the first a-helix presented a mean of 0.73 (range 0.52–0.94); the second a-helix presented a mean of 0.68 (range 0.60–0.85); and the third a-helix presented a mean of 0.70 (range 0.61–0.90) These data suggest that the second and third helices were slightly more mobile than the first one, which agree qualitatively with the last two helices showing a higher rmsd than the rest of the elements of the secondary structure [26] The NOE values of CACW40A were, however, lower than those found in other helical regions of well-ordered proteins of similar size, such as CI2 and eglin c (within the range 0.7–0.8) [32–34], but they were slightly higher than the values observed in fully unfolded proteins (within the range 0.0–0.3) [36,39–41] Next, we decided to use the model-free formalism [42,43] to obtain further insight into the apparent internal mobility of the protein However, the overall tumbling time of CACW40A, sm, must be estimated first Estimation of the overall tumbling time We used two different experimental approaches to estimate the sm to avoid any potential error in the determination of the model-free parameters FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS 3301 Dynamics of monomeric CAC L A Alcaraz et al We first estimated the sm with tensor2, by using the subset of rigid residues (see Experimental procedures), yielding a value of 6.4 ± 0.1 ns The sm was also determined by using the approach developed by Wagner et al [35,36] Briefly, this method assumes that, if the re-orientation of an internuclear 15N-1H vector is a composite function of noncorrelated motions, then the corresponding spectral density functions can be described as a linear combination of spectral density terms characterizing each motion (usually two Lorentzian curves) This assumption leads to a third degree equation in s, one of whose solutions is the sm: 2a x2 s3 ỵ5b x2 s2 ỵ2a 1ịs ỵ 5b ẳ N N Model-free formalism where the coefficients of the cubic equation, a and b, are obtained from the coefficients of the linear regression of the experimental J(xN) (i.e the spectral density function at the Larmor frequency of the 15N) versus J(0) (i.e the spectral density function at MHz) (Fig 3) In CACW40A, the positive solutions to the cubic equation lead to 1.28 ± 0.03 ns and 7.6 ± 0.6 ns The first root is assigned to an internal motion of the protein, and the second is the overall tumbling of the molecule, which is close to the value obtained previously As can be observed, only a small number of the experimental points in CACW40A are close to the crossing point, demarcating the sm boundary of the theoretical Lorentzian curve for the spectral density function Experimental points close to the Fig Relationship between J(xN) and J(0) The theoretical variation between both parameters assuming a simple Lorentzian curve for the spectral density function is also shown Experimental data (filled squares) were fit to a linear function (y = a + bx) with: a = 0.43 ± 0.04 nsỈrad)1 and b = 0.05 ± 0.01 nsỈrad)1, which are used in the third degree equation in s (for details, see text) Both functions intersect at points corresponding to the overall correlation time (sm) and an internal-motion time (se) 3302 boundary imposed by the theoretical curve correspond to residues with fast internal dynamic contributions, whereas those undergoing slower dynamics are located at J(0) values above the limit of the correlation time, as occurs in CACW40A (Fig 3) We also used different theoretical approaches to estimate the sm [44,45], and the results are similar to those described above (data not shown) The value used in the model-free formalism (see below) was 6.4 ± 0.1 ns It is important to indicate that relaxation measurements of the dimeric, non-mutated CAC have been carried out, and the sm obtained is much higher than that reported here [46] In CACW40A, the residues with high S2 (the order parameter) values (S2 > 0.8) were: Arg18, Asp19, Val21, Arg23, Phe24, Tyr25 and Ly26 (all of them belonging to the first helix); Asn51 and Cys54 (at the N-cap of the second helix); Ala64 and Ala65 (in the b-turn between the second and third helices); and Thr72 and Ala73 (at the C-cap of the third helix) (Fig 4A) The first a-helix is the secondary structure element that has the highest number of residues with high S2 values Thus, the high S2 values cluster at the regions of welldefined secondary structure with a lower rmsd [26] On the other hand, CACW40A has a large number of residues with low values of S2, suggesting that those residues are affected by fast movements (relative to sm) The mean ± SD of S2 in CACW40A is 0.56 ± 0.29 (see supplementary Table S2) This number is significantly lower than the average value of 0.86 found in other proteins [47], probably due to the long loop in CACW40A, which is not very well hydrogenbonded to the rest of the structure [48] None of the residues in CACW40A, except Ala65, could be fitted to the simplest model of tensor2 (see supplementary Table S2) Residues Glu15, Lys26, Gly62, Ala73 and Gly81 could be fitted to the second model Amino acids Phe17, Asp19, Arg23 and Gly79 could be analysed with the third one, where an exchange contribution, Rex, is included Residues Gln11, Thr42 and Thr72 were fitted to the fifth model; and the remaining residues could be analysed according to the fourth model, where Rex contributions and fast movements are included A large number of residues (i.e those fitted to models three and four) did experience conformational exchange on a micro-tomillisecond time scale (Fig 4B) In conclusion, most of the residues in CACW40A, and not only those in the loop region, have a fast internal mobility Furthermore, the fast internal corre- FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS L A Alcaraz et al Dynamics of monomeric CAC Fig The model-free approach parameters (A) The order parameter, S2, is shown on the structure of the protein: 0.8 < S2 < (red); 0.6 < S2 < 0.8 (orange); 0.4 < S2 < 0.6 (green) and < S2 < 0.4 (blue) (B) Residues that show an Rex term are shown on the structure of the protein: 10 < Rex < 16 s)1 (red); < Rex < 10 s)1 (orange) and < Rex < s)1 (blue) lation time, se, for the majority of amino acids was similar to the sm (see supplementary Table S2) It could be assumed that those fast se values are due to a wrong election of the diffusion tensor (e.g the diffusion tensor of CACW40A is fully anisotropic) because it is well-known that simplified isotropic models in which anisotropy is neglected can wrongly lead to exchange terms [49] However, similar values of S2, se and Rex to those reported in the supplementary (Table S2) were observed when a fully anisotropic model was used (data not shown) All these findings suggest that the assumptions of the model-free approach are no longer valid in CACW40A (i.e it is not possible to separate the overall tumbling of the molecule and the local fast movements of each 15N-1H bond) Thus, although the model-free approach is very intuitive, we decided to use the reduced spectral intensity formalism to test whether our results (i.e large mobility through all the elements of structure) were not an artifact of the model-free approach Reduced spectral density approach This approach provides insights into the motion of the N–H bond vector at three selected frequencies, x0 (= 0), xN and 0.87xH (Fig 5) As in other proteins [32,33,35,36], the J(0) (i.e the spectral density function at the frequency 0) had the largest samplings of the three explored frequencies The J(0) showed a mean of 3.25 nsỈrad)1 (range 1.7– 4.2 nsỈrad)1) (Fig 5A; see also supplementary Table S3) The J(0) is a sensitive probe of the nanoto-milliseconds motion (i.e very sensitive to the distri- bution of correlation times): low J(0) values indicate enhanced internal mobility on times scales faster than the sm The regions with the lowest values of J(0) in CACW40A were clustered to: (a) the termini of the helices and (b) the polypeptide patches in between (Fig 5A) However, it should be noted that J(0) contains not only information on the nanosecond motions faster than the overall tumbling of the molecules, but also on the exchange contributions [because it relies on R2; see Eqn (2) in Experimental procedures], which increase J(0) In general, values of J(0) above the mean value (3.25 nsỈrad)1) are good candidates for showing enhanced mobility in the millisecond time scale A comparison of Tables S2 and S3 in the supplementary material shows that all residues with J(0) values higher than 3.2 ns did show a Rex contribution in the modelfree approach These residues were Gly12, Lys14, Phe17, Asp19, Tyr20, Val21, Arg23, Tyr24, Thr27, Glu31, Val37, Met41, Thr44, Gln48, Asn49, Ala50, Asp53 to Leu58, Leu67, Met70, Met71 and Gln75 Because J(xN) (i.e the spectral density function at the Larmor frequency of the 15N) and J(0.87xH) (i.e the spectral density function at the 0.87 times the Larmor frequency of the 1H) are independent of R2 [see Eqns (3,4) in Experimental procedures] and less sensitive than J(0) to the distribution of correlation times, they can provide insights into protein dynamics The mean value of J(xN) was 0.58 nsỈrad)1 (range 0.28–0.76 nsỈrad)1) (see supplementary Table S3) The lowest values of J(xN) belong to residues involved in the polypeptide patches between the helices, and the highest ones correspond to the rigid regions The values of J(0.87xH) were very low and only accounted FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS 3303 Dynamics of monomeric CAC L A Alcaraz et al qualitative agreement with the model-free formalism); the three helices appeared rigid but they showed mobility in the pico-to-nanosecond time scale Furthermore, from the high J(0) values, there was evidence of enhanced mobility in the millisecond time regime in residues involved in the protein core and forming the last two helices, which showed Rex and ⁄ or long se values (i.e within the same order of magnitude than sm) in the model-free formalism (see supplementary Tables S2 and S3) Thus, both approaches qualitatively agree in demonstrating a high internal flexibility of the molecule Discussion We first discuss the results obtained within the framework provided by the structural elements of monomeric CACW40A Subsequently, we examine the biological and thermodynamical implications of such a high flexibility Backbone dynamics and the relationship to structure in CACW40A Fig The reduced spectral density approach Values of spectral density functions: (A) J(0), (B) J(xN) and (C) J(0.87xH) versus the protein sequence The cylinders at the top of each panel indicate the a-helices for a 1% of J(0) (Fig 5B) The mean value was 0.0138 nsỈrad)1 (range 0.00138–0.0245) (see supplementary Table S3) The tendency in J(0.87xH) was the opposite to that observed in J(0): the highest values in J(0.87xH) correspond to the termini of the helices and the regions in between, indicating efficient picosecond averaging In conclusion, using the reduced spectral density approach, analysis of the relaxation parameters shows that the regions between helices are highly mobile, but also the rest of the structure has a high flexibility (in 3304 One of the possible uses of 15N backbone dynamics is to predict regions of a protein with sufficient potential flexibility to allow functional events to occur (binding, conformational changes or catalysis) However, experiments with several dozens of proteins [27] demonstrate that there is no easy and general correspondence between the order parameter (S2), the spectral density function [J(x)] and the secondary structural elements of a protein Furthermore, there are no simple rules for the interpretation of the exchange rates (Rex) or the different correlation times (sm, ss or sf) In CACW40A, although the helical elements have the highest order parameters, there is no relationship between S2 and the location of structural elements (Fig 4) Furthermore, the Rex terms are distributed throughout the 3D structure of the protein, and most of them are large (Fig 4); the exception is Tyr25, with an Rex value of 0.5, which indicates that the dynamics of its 15N backbone nuclei is not robustly identified by the used calculation protocol [50] Thus, it appears that the whole protein is experiencing the same type of movements, ranging from pico- to milliseconds Furthermore, there is no correlation between the motions measured by Rex and the motions probed by hydrogen-exchange [26], where only the residues involved in the helices are protected For example, the first helix, which has the highest S2 values and is relatively well-ordered in the pico-to-nanosecond time scale, exhibits extensive ‘opening ⁄ closing’ equilibria on FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS L A Alcaraz et al a much slower time regime than the other helices These equilibria also occur in the other two helices, as shown by the exchange pattern [26], although they are less well-ordered, as judged by the lower S2 The types of movements and the residues involved are described below The pico-to-nanosecond dynamics Residue Ala65 (at the N-terminus of the third helix) is the sole residue that has restricted internal dynamics (model-free formalism) Fast internal dynamics (i.e residues with at least another tumbling time) occurs at the N (Gln11 and Glu15) and at the C-termini of the first a-helix (Lys26); in the long disordered loop (Thr42); and at the N- (Gly62, Ala64), and C-termini of the third helix (Thr72, Ala73) However, it is not possible to establish any correlation between any structural parameter of those residues and the fast dynamics observed The micro-to-millisecond dynamics Most of the residues in CACW40A required an Rex term (model-free formalism) or had long J(0) values (reduced-spectral approach); furthermore, most of the residues in the loop (which forms the second helix in the dimeric non-mutated CAC protein [22,23]) were broad beyond detection in the HSQC experiments [26] Although the arguments could be considered as speculative, the highest Rex values observed in some amino acids of CACW40A (see supplementary Table S2) might be ascribed to the proximity of the particular residue to either aromatic or Cys residues, as described in other proteins [37,50,51] Residues Val37, Met41 and Thr44 belong to the long disordered loop [26], buried within the structure, but only the amide proton of Thr44 is hydrogen-bonded We not know how to ascribe the exchange contribution of Val37 and Met41 to any particular dynamic process In other proteins, similar micro-to-milliseconds exchange contributions have been observed in well-buried protons, and they have been explained as due to buried water molecules [37] Finally, it is important to note that not only were residues belonging to the second helix absent in the NMR spectra of CACW40A, but also they did not appear in the spectrum of the dimeric wild-type protein [29,31], nor did they appear under physiological conditions in the NMR spectrum of another recently reported monomeric mutant [52] These findings suggest that the reported flexibility in the domain is not a particular characteristic of the mutant, but is an intrinsic feature of the whole dimeric CAC domain Dynamics of monomeric CAC Model-free analysis versus spectral density mapping Our results indicate that the relaxation data of CACW40A could not be satisfactorily explained by the model-free method In this formalism, the correlation function (the function describing the movement) of each bond vector is decomposed as the product of the correlation function for overall (global) and internal (local) motions (i.e the internal motions of the bond vectors are independent of the overall rotational movement of the molecule) Furthermore, the internal motions of each bond vector are independent of each other, but the rotational diffusion of the molecule affects each of those bond vectors identically [42,43] On the other hand, spectral density mapping makes no assumptions about the nature of the rotational diffusion (i.e the information on which oscillations for a particular bond vector are associated with global molecular rotation or segmental molecular motions is lost) Thus, based on the spectral density formalism results, we are unable to discern whether the movement of each NH bond is due to local internal or overall tumbling, but we can conclude that the CACW40A has an intrinsically high structural mobility (Figs and 5) To support this conclusion, the ses obtained from the model-free approach for most of the residues are similar (i.e they are not faster) than the overall molecular tumbling of the protein; this means that we cannot strictly separate the overall tumbling of the molecule from the internal motions of each bond vector and, thus, the model-free formalism cannot be rigorously applied This is not the sole example where the use of the model-free formalism has been unsuccessful: this approach cannot be applied on natively unfolded proteins, proteins at high temperatures [27,39,53–55], or, even recently, in otherwise well-behaving proteins [56] Biological and thermodynamic implications Our study on the dynamics of CACW40A indicates that the protein is structurally very flexible, while preserving most of the native scaffold [26] It could be assumed that this flexibility is due exclusively to the mutation; however, although the mutation increases the flexibility (because the quaternary structure is lost), the high flexibility is present in the structure of CAC, as suggested by several studies First, similar dynamic results have been observed for the C-terminal region of dimeric, nonmutated CAC [38], and in residues belonging to its dimerization interface [29,31] Second, it has been observed that: (a) CAC is able to form swapped domains involving the major homology region and the second a-helix [28,57]; (b) CAC is able to bind a peptide FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS 3305 Dynamics of monomeric CAC L A Alcaraz et al forming a five-helical bundle [29]; (c) the second and third helix in CAC appear to be distorted upon binding to lipids [30]; and (d) the fourth helix in CAC is involved in binding to lysyl-tRNA synthetase [31] Thus, these studies show that the CAC domain is able to alter its structure and promote other interactions in the presence of an external agent (lipids, peptides, other regions of the Gag protein, or even other proteins) In the first three examples, the second helix (as in CACW40A) was the main element of secondary structure affected; in the last example, the fourth helix was the element altered The detection of slow dynamics not only at the dimerization interface (residues Glu31 to Ala40), but also in the rest of the protein implies the presence of a small population of pre-existing conformers within the nativestate ensemble This population interacts with other CACW40A monomers forming the dimeric CAC, probably through the side chains of the hydrophobic residues of the long disordered loop, buried to avoid nonspecific hydrophobic interactions [26] There are several examples of proteins in which binding residues are involved in slow-exchange processes [27,58], most likely to facilitate rapid partner-binding, and the recognition of several ligands Internal motions allow amino acids to explore large regions of the conformational space at a very low energetic cost, increasing the chances of successful binding However, are those slow-exchange processes responsible, from a thermodynamic point of view, for the binding of the monomeric species of CAC? We have previously discussed the variation in the free energy of binding as a function of the changes in buried surface area upon dimer formation [59] On the other hand, there are no clear correlations between the enthalpy of binding and the changes in buried surface area [60]; thus, the only thermodynamic magnitude that has not been estimated in CAC is the binding entropy change, DSb The binding entropy, DSb, can be divided into terms defining the solvent (hydrophobic) (DSsol), the conformational flexibility (DScon) and the rotationtranslation portion (DSrt) entropies: DSb = DSsol + DScon + DSrt The DSrt accounts for )50 calỈmol)1ỈK)1 [61,62] The solvent portion of the entropy can be calculated as a function of changes in polar and apolar surface areas of the binding interface, according to: DSsol = DCp ln(T ⁄ 385), where DCp is the heat capacity change of the binding reaction We have previously determined the DCp ()211 ± 10 calỈmol)1ỈK)1 per monomer) and DSb ()230 ± 10 calỈmol)1ỈK)1 per monomer) [59], and then, the contribution from the conformational flexibility to the entire entropy of binding will be: DScon = )234 calỈmol)1ỈK)1 per monomer Because, on average, the entropy cost per amino acid for a folding transition is approximately 3306 5.6 calỈmol)1ỈK)1 [63], the estimated DScon in CAC upon binding of the two monomers is due to the cost of fixing 42 residues This value is much higher than the number of residues present in the long loop, which is disordered in CACW40A (14 residues), and the difference must be associated with: (a) the movements of the last two helices, as observed in the monomeric structure of CAC, and (b) the inherent flexibility for the majority of the residues Thus, the conformational entropy appears to be distributed through the whole structure of the monomeric species, sampling a wider range of dynamic movements, and not only located at the residues in the interface In summary, we suggest that the inherent flexibility of the CAC domain is consistent with the presence of a low thermodynamic barrier to diverse, templateassisted conformational changes, that allow interaction with several macromolecules Experimental procedures Materials Deuterium oxide was obtained from Apollo Scientific (Bredbury Stockport, UK), and the sodium trimethylsilyl [2,2,3,3-2H4] propionate was obtained from Sigma (Madrid, Spain) Dialysis tubing was obtained from Spectrapore (Breda, the Netherlands), with a molecular mass cut-off of 3500 Da Standard suppliers were used for all other chemicals Water was deionized and purified on a Millipore (Barcelona, Spain) system Protein expression and purification The 15N-labelled CACW40A protein was expressed in Escherichia coli BL21(DE3) in LB and purified as previously described [26]; the DNA segment used for the mutant protein encoded for residues 146–231 of CA from HIV-1 (strain BH10) and was cloned as described [24] The protein concentration was calculated from A240 by using the extinction coefficients of amino acids [64] Samples were concentrated at the desired final NMR concentration by using Centriprep Amicon devices (Millipore), with a molecular mass cut-off of 3500 Da Protein structure calculations The determination of the solvent-accessible surface area was obtained using the VADAR web server [65] NMR samples All NMR experiments were acquired on an Avance Bruker DRX-500 spectrometer (Bruker, Karlsruhe, Germany) FEBS Journal 275 (2008) 3299–3311 ª 2008 The Authors Journal compilation ª 2008 FEBS L A Alcaraz et al Dynamics of monomeric CAC equipped with a triple resonance probe and pulse field gradients Sample temperature was calibrated using a 100% methanol standard [66] NMR relaxation measurements NMR relaxation data were collected at 293 K 15N-T1, 15 N-T2 and 1H-15N NOE experiments were acquired using enhanced sensitivity, gradient pulse sequences developed by Farrow et al [67] All spectra were recorded as 128 · K complex matrices with 64 scans per F1 experiment Spectral widths of 1650 and 8000 Hz were used in F1 and F2 respectively A total of 10 data sets were acquired to obtain 15N-T1 rates using relaxation delays of 50, 100 (· 2), 200, 300, 400, 500, 600, 700 (· 2), 850 and 1000 ms, where the experiments at 100 and 700 ms were repeated twice The 15N-T2 measurements were made using delays of 15, 25 (· 2), 50, 100, 150, 175, 225 (· 2), 300 and 425 ms For the T1 and T2 pulse sequences, the delay between transients was s The 1H-15N NOEs were measured by recording interleaved spectra in the presence and in the absence of proton saturation The spectrum recorded in the presence of proton saturation was acquired with a saturation time of s The spectrum recorded without proton saturation incorporated a relaxation delay of s Each experiment was repeated twice Experiments were carried out at two protein concentrations (1 mm and 400 lm) to rule out any possible concentration-dependent effect on the measured relaxation rates, as has been observed in dimeric non-mutated CAC [46] The measured rates were identical at both concentrations within the experimental error (see supplementary Table S1) Data processing and analysis of the NMR relaxation measurements J0ị ẳ 6R2 3R1 2:72rNH ị=3d2 ỵ4c2 ị; 2ị JxN ị ẳ 4R1 5rNH ị=3d2 ỵ4c2 ị; 3ị J0:87xN ị ẳ 4rNH ị=5d2 ị; 4ị and rNH ẳ R1 ðNOE À 1ÞðcN =cH Þ; ð5Þ where c = (xN ⁄ Ö3)(r|| – r^) and d = l0hcNcH ⁄ (8p < r > 3), l0 is the permeability constant of the free space, cN and cH are the gyromagnetic ratios of 15N ()2.71 · 107 radỈs)1ỈT)1) and 1H (2.68 · 108 radỈs)1ỈT)1), h is the Planck constant, xN is the Larmor frequency of the 15 N, xH is the Larmor frequency of the 1H, is the ˚ length of the amide bond vector (1.02 A), and r|| and r^ are the parallel and perpendicular components of the CSA tensor (r||)r^ = )160 p.p.m for a backbone amide [70]) The uncertainties in a particular J(x) are the quadratureweighted sum derived from Eqns (2–5), assuming that errors in the relaxation rate constants are independent Rotational diffusion tensor The spectra were zero-filled in the F1 dimension four times and processed by using a shifted sine window function The same window function was used through all the T1 and T2 experiments Cross-peaks intensities were measured as volumes, with the xwinnmr software package (Bruker) The T1 and T2 values were determined by tting the measured peak-heights to a two-parameter function: Itị ẳ I0 expðÀt=T1;2 Þ; The T1 and T2 relaxation times (or, R1 = ⁄ T1 and R2 = ⁄ T2) and the NOE enhancement of an amide 15N nucleus are dominated by the dipolar interaction of the 15N nucleus with its attached proton and by the chemical shift anisotropy (CSA) The energy of the CSA and the dipolar interaction has a constant value over all the ensemble of spins [68] The spectral density function, J(x), expresses how this energy is distributed over all the spectrum of possible frequencies, x, explored by the spins The measured rates for each NH are related to the J(x) at the nuclear spin frequencies [68], and they can be approximated as (the so-called ‘reduced spectral density mapping approach’) [32,33,69]: An initial estimation of sm and the rotational diffusion tensors were obtained with tensor2 [71], from the subset of residues which accomplished the following criteria [72]: (a) all residues should have a NOE ‡ 0.65 and (b) the residues should satisfy: R2;i ÀhR2 i R1;i ÀhR1 i À