J. CHEM. SOC. PERKIN TRANS. 11 1985 Nuclear Magnetic Resonance Study of Ln3+ Complexes with Aspartate and Glutamate Residues. Thermodynamic and Structural Analysis Nadhge Jamin, Daniel Baron, and Nicole Lumbroso-Bader * Universit4 Pierre et Marie Curie (Paris 6)and LASIR, CNRS, 2 rue flenri Dunant, 94320 Thiais, France Binding of aspartate and glutamate residues with the lanthanide cation Yb3+ has been studied in D,O, involving two kinds of peptide backbones: N-acetylamino acid methyl esters and cyclic dipeptides. A method is described for the quantitative analysis of the lanthanide-induced shifts (LIS) from which the best model of solution, the binding constants, the induced shifts, and the geometry of the complexes have been determined. Aspartate and glutamate derivatives are found to form both 1 :l and 2:l complexes (peptide: cation).Backbone flexibility and, to a lesser extent, side-chain length seem to govern the magnitude of the binding constants. Only one oxygen of the carboxylate group is involved in the complex. The Yb3+ 0-distances are found to be different according to the Asp (2.70-2.75A)and Glu (2.50-2.55 A) residues. The predominant conformer has an extended side-chain so no chelation occurs with the peptide backbone. The complexation of cations by amino acid residues is an important biochemical process, as this interaction can be involved in the transport of cations or in enzymatic activations or inhibitions. The aim of this study was to define the influence of structural factors on the binding strength of the residues and also to determine the geometry of the complexes in an aqueousmedium.We first focused our attention on the intrinsic complexation of a single residue, aspartate or glutamate, in a dipeptidebackbone, without interaction of any other residue or peptide end. Since various types of peptide flexibilities can be found, two different kinds of peptides were selected: (i) rigid diketopiperazines [cyclic dipeptide (l),Gly-Asp or Gly-Glu]; and (ii) N-acetylamino acid methyl esters (2) (where more rotational freedom occurs). 'H and 3Cnuclear magnetic resonance lanthanide-induced shift (LIS)measurements seemed a suitable technique for this investigation.1-3 As regards biological activity, the possibilities and limits of substituting Ln3+ for Ca2+ has already been discussed extensively. Furthermore, the induced shifts are large enough to allow an accurate determination of the binding constants.The Yb3+ ion was chosen as a probe because the shifts it induces often have no substantial contact interaction and consequently are of use in the structural elucidation of the complexes.'*2*4*s Pr3+, for which the effective axial symmetry of the magnetic sus- ceptibility tensor seems best dem~nstrated,~.' was not convenient because of difficult experimental conditions: its LIS are at low field in water and many 'H signals overlap with the HDO or other signals, so that detailed thermodynamic analysis becomes very difficult. The first part of this paper deals with the quantitative analysis of LIS in order to derive the stoicheiometry of the complexes and their binding constants, and to discuss the results with respect to the backbone and side-chain flexibilities.The structural investigation, carried out from the calculated LIS in the complexes, is reported in the second part. ExperimentalPerchlorate salts of the lanthanides were either of commercial origin (Alfa-Ventron) or prepared by addition of a solution of Ln2(S04)343H20 in D20 to a solution of Ba(ClO,),. The concentrations of the salt solutions were determined through the usual EDTA titration procedures.' Me-CO-NH-CH-COLOMe I R (2) Peptides were purchased from Interchim and Bachem, except cyclo(G1y-Asp) and cyclo(Gly-Glu), which were synthesized according to the literat~re.~*'~ Two series of peptide solutions were prepared, one 0.1~ (for the I3C and 'H spectra) and the other 0.01~(for the 'H spectra only).It is well known that analysis of data from different concentration ranges gives an improvement in the thermody- namic results. The concentration of the salt was varied up to 0.1~. The precipitation of the lanthanide hydroxides occurs at cu. pH7," but the products of hydrolysis can be formed before.12 Thus a relatively low pH was selected, but one that was allowed to have [C0,-]/[C02H] = 4 (pH = 4.6 for aspartate derivatives and 5.1 for the glutamates). These pH readings were actual meter readings, i.e., uncorrected for deuterium isotopic effects (pH = pD + 0.4).13 N.m.r.spectra were recorded at 27 f 2°C on Bruker WH 90 and Varian CFT 20 spectrometers. Most of the peak assignments were carried out according to literature data.14 There were still a few problems, which were solved in the following way: 'H signals of CH,, and CHzv in Glu residues -proton-selective decoupling experiments at 400 MHz; C,and C, of Glu -pH variations of the chemical shifts; C--O resonances in N-Ac-amino acid OMe -proton-selective decoupling experiments on acetyl Me and 0-Me; C--O in cyclo(G1y-Asp) -coupled spectrum (at 62.9 MHz).Owing to the low solubility of cyclo(Gly-Glu), cu. 5 x lW2~, the C--O resonances in this compound were assigned by analogy with cyclo(G1y-Asp) data. Proton and carbon-13 chemical shifts are reported in Table 1.2 J. CHEM. SOC. PERKIN TRANS. II 1985 Table l.-Chemical shifts in D20(p.p.m., ref. DSSNa) NAcAspO Me pH = 4.6 61, 613C 179.82ASP or Glu 2.72 40.78 4.66 52.98 176.57 Ester Me 3.75 55.68 Acetyl G== 176.70 Me 2.03 24.41 Table 2. Experimental data Complex n' 'H 13C IA81max(P. p.m. 1 'H 13C Yb3+complexes ~y~lo(Gly-A~p) NAc AspOMe cyclo(Gly-Glu) NAcGluOMe 30 32 39 36 10 9 10 14 6.15 5.25 7.53 7.57 17.80 15.51 21.13 15.69 Eu3+ complex cyclo(Gly-Asp) 30 11 2.14 42.0 Number of experimental solutions. Sodium 2,2-dimethyl-2-silapentane-5-sulphonate(DSS) was used as an internal standard. Thermodynamic Analysis Method.-In our systems, shift ratios AS,/AS, are fairly constant until (Yb3+ J = 0.1~~both in dilute and concentrated solutions.Therefore, the extrapolation of these ratios to p (ionic strength) = 0, as in some work^,'^-'' was not necessary here. Furthermore, in other works on histidine18 or carboxylic acids,12 no significant effect on the chemical shifts due to ionic strength variations (of the same order of magnitude as in our systems) was found. Two models of solution can be used, the simplest one involving a single complex PLn2+ (model I) and the other one two complexes PLn2 + and P2Ln+ (model 11) as previously shown for some amino acids in D20.19-2i From these models, some improvements can be made, taking into account, for example, a macroscopic medium effect arising from loose interactions between C104-and the complexed ligands.So a secondary term, proportional to the salt concentration, could be added either in model I or in model 11. Concerning model 11, an average LIS value in the two complexes was taken as a first approximation,20 A&, = A&-,(pLn) = AGC,(P2Ln),but a more sophisticated model could be considered with two different values of A6c,.19*21 Calculation of the thermodynamic and spectroscopic parameters, using the four most refined models, required sufficient data: in fact we obtained 150-230 LIS for the whole set of 10-13 sites (Table 2). NAcGluOMe Cyclo(Gly4lu) pH = 5.1 pH = 5.1 61, 513C 6tH 613, 182.81 182.80 2.33 35.14 2.34 34.04 41.83 2.05 29.45 ;:$ 32.06 4.30 54.69 4.41 55.37 4.16 56.77 172.88 177.00 172.66 171.20 171.19 4.06 46.73 4.05 46.45 3.75 55.60 177.00 2.03 24.27 For more coherent results, all the data from the different magnetic sites were processed simultaneously.In that way, the K, and K, values are better than when averaging the values obtained separately from each nucleus. This procedure is required particularly when different ranges of peptide concentration are used. Furthermore, to get coherent A6c,/A6c, ratios, the A&, have to be calculated from the same K,, K2 set. Three considerations now have to be discussed. (i) Each nuclear site ('H or I3C) can give thermodynamic information if the uncertainty about the frequency measurement, in a given sample, is small enough when compared with the maximum LIS for this probe.In our systems, the precision is better than 1%for all the 'H and two 3Csites. (ii) The most sensitive data are also the most affected by uncertainties about pH, temperature, concentration, etc. Consequently, when the mixing of data from different nuclei is required, the best way is to normalize, i.e., to divide these data by the maximum LIS of each probe. (iii) In order to calculate the best values of the binding constants in the peptide concentration range 0.01--0.1~,we chose to give about the same weight to the data coming from the two lines of solutions, the dilute and the concentrated ones. Hence we first processed all of the data, carefully studying the standard deviation for the 13C nuclei for which the spectroscopic precision was the worst.Except for the case of N-Ac-Glu-OMe, we finally kept all the data because the standard deviations for the nuclei in question never exceeded twice the mean standard deviation for the whole system. As the maximum value of LIS for 'H sites came from 0.01~ solutions, a further correction had to be made for the I3C data, for which only 0.1~solutions were available [equation (l)]. Results and Discussion.-(a) Models of solution. Normalized shift variations were analysed using a least-squares method to determine the best fit for models I, 11, and the improved ones. Models involving a secondary effect on the chemical shifts (medium effects or when A8C,,P2Ln# ASC,,pLn) give results that are physically meaningless.Probably both secondary effects should be introduced as neither correcting term is dominant. This would imply three spectroscopic parameters for each magnetic site, requiring more I3C data than could be obtained. As a consequence, the only two models discussed further involve either one or two complexes with, in the latter case, the approximation of an average A&, in the two complexes. Results are reported in Tables 3 and 4. J. CHEM. soc. PERKIN TRANS. 11 1985 The (PLn + P2Ln) model gives the best normalized standard the complexes also depend on the model. In fact, concerning the deviation. A statistical F-test22 was used to corroborate this geometrical analysis, the choice of model would not matter if result. The sum of the squares of deviation (S.S.) associated with the LIS values of each site were proportional from one model to model I was taken as the total S.S.and that of model I1 as the the other one, but this was not the case in the systems under residual S.S. Moreover the number of linear parameters that study: ratios from the analysis using model I were not could approximate to a binding constant in the relationship in agreement with the experimental A6c-A6c, values. In A6 =f(P,, Ln,) has been chosen as 3 rather than 1. In these contrast, the results obtained from model I1 fit these conditions the validity of the second model, compared with the experimental ratios well (Table 5). This shows that = first one, was demonstrated (with a confidence level better than is a good approximation of these systems.99% for each of the five studied systems). The confidence level was also calculated independently for (b) Binding Constants.-Binding constant values are the each site to make sure that it was not systematically too low for average values in the range 0-0.1~ for Yb(ClO,),. It was any one of them. This was not the case, even for nuclei far from assumed that the ionic strength effects would not significantly the ytterbium cation (see Table 4). modify the K,ratios for similar compounds. Taking the two complexes into account not only improved Two structural factors may affect the K, binding constants, the fit of the data but also increased the validity of the the backbone flexibility and, to a lesser extent, the side-chain parameters.Besides the binding constants, the chemical shifts of length: (i) for N-acetylamino acid methyl esters, K, is twice as large as for cyclic dipeptides (see Table 3). The larger degree of freedom in the first compounds may facilitate their binding. (ii), Table 3. Binding constants for Yb3+ complexes (hi-') K, is increased by 2&30% for glutamate over the aspartate derivatives. One more CH, group may increase the Ligand Model K, K2 6' complexation either because of carboxylate basicity or of CyClO(Gly-ASp) 1:l 29 0.0704 conformational freedom. 1:l + 2:l 34 17 0.0273 As K, is the last supplied parameter, its value can take into NAc AspOMe 1:l 66 0.0568 account other secondary effects (cf:Models ofsolution);it is then 1:l + 2:l 70 6.1 0.0465 difficult to discuss K2/K, ratios.Cyclo(G1 y-Glu) 1:1 86 17 0.0517 In order to test any intrinsic backbone binding, a cyclic1:l + 2:l 48 0.03 54 NAcGluOMe 1:l 88 0.0338 inactive dipeptide, cyclo(G1y-Ala), was also investigated.A weak interaction with Dy3+, the lanthanide that induces the 1:l + 2:l 87 3.4 0.0295 largest LIS, was detected, the constant K, being less than lw'.'Q For normalized chemical shifts. Table 4. Induced shifts Aaci (p.p.m.) and normalized standard deviation (mi) Complex 1:l Complexes Complex 1 :1 Complexes 1:l + 2:l 1:l + 2:l -A8ci Qi -A6Ci Qi -A6Ci Qi -A&i Qi CyClO( Gl y- Asp) C yclo(G1 y-Glu) ASP or Glu CO, -CY 36.47 0.063 23.64 0.019 34.53 23.13 8.01 0.036 0.052 0.053 27.96 20.15 8.56 0.025 0.022 0.035 23.71 0.085 15.43 0.036 6.84 0.041 5.67 0.020 8.01 0.072 6.60 0.024 4.57 0.059 4.87 0.037 8.14 0.072 6.70 0.022 4.93 0.057 5.21 0.037 7.28 0.083 4.69 0.031 3.64 0.040 3.01 0.021 3.97 0.071 3.32 0.022 2.06 0.054 2.20 0.034 5.49 0.084 3.53 0.032 2.08 0.040 1.69 0.032' 2.26 0.061 1.47 0.020 0.70 0.043 0.59 0.059 1.60 0.046 1.03 0.033 0.79 0.041 0.65 0.044 0.40 0.061 0.36 0.033 0.45 0.050 0.48 0.037 NAc AspOMe NAcGluOMe ASP 22.83 0.058 19.95 0.030 29.02 0.027 25.88 0.020 or 23.28 0.027 19.84 0.031 Glu 8.12 0.036 8.02 0.026 14.45 0.070 12.64 0.042 5.85 0.032 5.22 0.025 5.89 0.041 5.64 0.048 4.82 0.030 4.73 0.029 4.45 0.064 3.89 0.035 3.55 0.031 3.16 0.024 2.99 0.042 2.89 0.046 2.02 0.036 2.01 0.033 Methyl ester 2.52 -0.47 0.073 0.123 2.21 -0.42 0.041 0.095 1.76 c 0.033 1.57 0.022 -0.43 0.047 -0.41 0.041a c Acetyl c=o Me C 0.88 -0.30 0.045 0.099 0.76 -0.27 0.040a 0.067 0.51 c 0.050 0.48 0.051 H -0.39 0.049 -0.37 0.041 c a Confidence level (model I1 compared with model I) less than 95%.Confidence level (model I1 compared with model I) less than 90%. Site removed from analysis (experimental induced chemical shifts d 0.08p.p.m.). 4 J. CHEM. SOC. PERKIN TRANS. 11 1985 Table 5. Ratios ASc/A6cal-in the two models of solution and experimental A6t/A6m2-in 10% peptide solutions Cy~lo(Gly-Asp)-Yb~+ model PLn /&,, -model PLn + P,Ln Aw~c,l-(experimental l@lM) 2 0.650 0.220 0.653 0.279 0.642 0.276 H:: Ca 0.223 0.200 0.283 0.198 0.278 0.197 0.109 0.062 H;':GoAsp 0.150 0.140 0.149 0.062 0.141 0.148 0.063 HZ c, 0.044 0.011 0.044 0.015 0.044 0.015 Thus for Asp and Glu derivatives, the intrinsic complexation of the peptide backbone could reasonably be neglected.We also wondered whether chelation could occur,involvinganother ligand group apart from the carboxylate. A first indication of this was given by comparison of the cyclo(G1y- Asp) binding constants with Eu3+ (K,= 132~')and Yb3+ (34~'). The larger value observed with Eu3+ (a light lanthanide) generally demonstrates that only one ligand group interacts with the cati0n.23024 This feature will be of interest for the subsequent structural analysis.Structural AnalysisMethod.-Structural Analysis. Our aim was to compare the structures of the four 1: 1 peptide complexes, looking for some differences that could explain the Kl values and indicate whether chelation mrs or not. The paramagnetic induced shifts (ie., LIS corrected for diamagnetic terms) can be used for a structural elucidation if the contact contribution is small, which is generally the case for Yb3+. Moreover, in order to simplify the analysis, axial symmetry of the magnetic susceptibility tensor is required. Rapidly interconverting states of the complexes 6v2 and/or small values of the non-axial effects26 should allow the use of an approximation of an axial symmetry. For some light lanthanides, axial symmetry has been often demonstrated, particularly in the case of Pr3+,while recently Delepierre et d.' observed strong non-axial effects with Tm3 + and Er3+, two heavy lanthanides as is Yb3+.However, in this latter work, the differences in the LIS ratios between Pr3+ and Yb3+,involving aquo ions, were not so important.Moreover, other studies have shown the validity of the axial ap-proximation along the lanthanide series 2o or that the behaviour of Yb3+ is very close to one of the light Under these conditions the axial symmetry assumption seemed a reasonable approximation in order to compare the structures of the four peptide complexes. Before we describe the procedure for analysing the data, we will first summarize the determination of the diamagnetic term and the choice of the geometric data, i.e., bond angles, interatomic distances, and conformations of the substrates.(a) Diamagnetic term (Ad,). We measured, in a few solutions, the chemical shifts induced by Lu3+ (AS,) in the substrate. As the binding constants for Yb3+ and Lu3+ have nearly the same value^,^^^**^^ the peptide proportions ql and q2in the 1: 1 and AS Adi = q' +,q2 NAcGluOMeYb3+ A6C/A6Cm2 -A6,1A6Cm2-model model (experiment a1 PLn PLn + P,Ln 1Wd CY 0.806 0.767 0.769 0.280 0.310 0.303HY CII 0.202 0.202 0.201 0.166 0.183 0.172HI3C. 0.122 0.122 0.121 Ha 0.070 0.078 0.079 C-OO," 0.061 0.061 0.060 MA, 0.018 0.019 0.018 the 2: 1 complexes were evaluated for each solution using the Yb3+ binding constant values, and Adi given by equation (2).The average values for different samples are reported in Table 6. The diamagnetic contribution is small for 'H (less than 0.10 p.p.m.) but significant for 13Cespecially for those nuclei close to the site of interaction (up to 8 p.p.m. for the carboxylate carbons). (b) Geometric data. Bond angles and interatomic distances were selected from the literature on crystallographic results or conformational parameter values 30-39 (see Table 7). The diketopiperazine ring was assumed to be rigid and planar.35 However, for the N-acetylamino acid methyl ester backbone, different conformations have to be considered.For amide and ester groups, trans and cis planar configurations, respectively, were ch~sen.~~~~~ We assumed a free rotation of the methyl groups, six staggered conformations around the N-C, bond, and three eclipsed ones around the Ca-C' axis. For the side-chains, three staggered conformations around the Csp3-Csp3bonds were assumed, but for the Cg(orar~-COZ-axis either staggered or eclipsed conformations could be adopted, according to a bidentate or a monodentate binding of the carboxylate group, respectively. In fact, as the induced chemical shifts of the p-protons in the cyclo(G1y-Asp) complex are different, especially in the case of Eu3 + (see Table 6), we first assumed-a dissymetrical complexation, i.e.,a monodentate one.Therefore three eclipsed conformations around the CB(ory)-C02-axis were investigated; however, the staggered conformations were also tested in a few cases. The final number of conformers was 9 for cyclo(G1y-Asp), 27 for cyclo(Gly-Glu), and 162 for NAcAspOMe and NAcGluOMe. For the last compound, conformers from C,-C' rotation have not been considered, as the induced chemical shifts of the methyl esters are too small to be included in the analysis. (c) Data analysis. Figure 1 shows the internal Cartesian co- ordinates chosen with the carboxylate in the x,~plane and a positive y co-ordinate for the adjacent carbon. The direction of the principal magnetic axis was calculated for each lanthanide position (p, a,A) taking into account the distances between the ion and the two oxygen atoms, with the weighting factor for each oxygen inversely proportional to the square of its distance to Ln3+.Previous work had shown the validity of such an appro~imation.'~*~'"~ To locate the lanthanide cation, space was explored between two spheres of radius p = 2.30 and 2.90 A (a reasonable domain)4i1g*20*4swith steps of 0.05 A for p and 20" for o and A. For each point (p, a, A), the geometric factors of the McConnell and Robertson relationship 46 were calculated for all the magnetic sites i and conformers j [equation (3)]. A mean-squares method was applied to the paramagnetic terms Api J. CHEM. SOC. PERKIN TRANS. II 1985 Tak 6. Diamagnetic (Adi) and pseudo-contact (Apj terms (p.p.m.). Api = A&, -Adi a.Ytterbium complexes Cyclo(Gly-Asp) NAcAspOMe Cyclo(Gly4lu) NAcGluOMe Ad APi Mi APi Ad APi Ad APi 5.15 -28.79 3.95 -23.90 8.12 -36.04 7.39 -33.27 C 0.57 -20.25 -0.47 -19.37 H 0.06 -8.64 0.04 -8.05 C 0.60 -16.03 0.20 -12.84 -1.07 -4.59 -0.79 -4.43 Asp or Glu H 0.06 -6.66 0.04 -4.92 0.06 -4.790.07 -5.710.04 -6.74 0.04 -5.264:C -0.39 -4.30 -0.48 -3.42 -0.15 -2.85 -0.52 -2.64 H 0.10 -3.42 0.08 -2.97 0.03 -2.24 0.03 -2.04 -0.25 -3.28 -0.18 -2.03 -0.03 -1.66 -0.27 -1.31 -0.10 -1.37 0.03 -0.60 C -0.04 -0.99 -0.02 -0.63 H 0.04 -0.40 0.02 -0.50 Methyl ester C 0.28 0.16 H 0.01 0.40 Acetyl 0.17 -0.60 0.10 -0.58 C 0.12 0.15 H 0.01 0.36 b. Cyclo(Gly-Asp)-Eu3 complex. Api values+ f-co7-55.07 GlY G== -1.47 ASP CH;, C -51.58 H -1.34 CH, C -0.57 G== -2.33 H -0.34H61 -1.76 H6, -2.30 analysis based on a single conformer.In all cases, the validity of the analysis involving two conformers, with respect to the one involving a single conformer, was checked by a statistical F-test.22 Apt = + (1 -aj)GFi'] (4) /o=c.J 'c Results and Discussion Figure 1. The axis system for the Ln3+ complexes. Comments should first be made on the comparison of Yb3+-and Eu3+-induced shifts. The regular increase in Yb3 + shifts, according to the proximity of the co-ordination site, is not observed for Eu3+ shifts; in this case, a positive shift for the carboxylate and a negative one for the adjacent carbon show that these shifts include an important contact contribution and, (Table 6) in order to determine the variance for each conformer. consequently, these two shifts should be discarded from the Only the lowest variance value was retained, V (p, a, h) analysis of the Eu3 + complexes.corresponding to the 'best' conformer.Concerning the four ytterbium complexes, the analysis was Finally, for each sphere the lanthanide co-ordinates o and h carried out by first taking into account a single conformer and were refined, to a resolution of lo,for all the sub-minima for all the magnetic sites. Results are reported in Table 8. It is worth which Vsub.min<1.5 Vmin (Vmin being the smallest over-all noting that the Ln3+ 0-distances are 2.7G2.75 A for variance of the 13 spheres).Thus the best geometry of the the aspartate derivatives and 2.50-2.55 A for the glutamates. complex (conformer and Ln3 + co-ordinates) was determined The values are consistent with a monodentate complexation with a resolution of 1" for o and h and 0.05 A for p. of the carboxylate group and an extended structure of the best Another analysis was carried out, taking into account two conformers [(I) and (II)] (see also Figure 2). So no chelation conformersj and I (assuming the co-ordinates of the cation to occurs with the peptide backbone. However, this analysis leads be the same for both). The variances were then calculated for all to rather large values of R (0.084.14), R being the Willcott the combinations of two conformers j and 1 with a mean- agreement factor4' used to check the validity in such asquares method applied to equation (4), oj being the weight of structural analysis [equation (S)].Moreover, an importantconformerj. For N-acetylamino acid methyl esters, on account of the great number of combinations of two conformers (about 13 OOO), the minimal variance was determined by only exploring the space around the lanthanide positions indicated by the 6 J. CHEM. SOC. PERKIN TRANS. 11 1985 Table 7. Bond angles (") and bond lengths (A) Plane (Asp) Diketopiperazine ring Mean value Ca-C' 1.509;" 1.519' 1.52 C,-N 1.452;" 1.460' 1.46 C-N 1.334;" 1.32gb 1.33 A NC'C, 118.9;" 118.6' 119 A CaNC' 126.0;" 127.9' 127 A C'CaN 115.1;" 113.4' Ca-C8 1.532-1.545;' 1.551 ' 114 1.544-1.548;d 1.548" A 1.542 1.54 systematic deviation is observed for the carbon adjacent to the C'CaC6 107.6-108.0;'' 112.4" carboxylate group (from 1.3 to 4.6 p.p.m.).113.1' 110 At first sight and without dismissing the pseudo-contact Peptide group model, these discrepancies could arise from the presence of a second conformer or from some staggered conformations N-Ca 1.4730 1.47 around the carboxylate group. In fact, the results of the two- N-C' 1.3230 1.32 C'=o 1.2430 1.24 conformer analysis, involving the four ytterbium complexes, A were unsatisfying: large deviations for the carbon adjacent to H3CC'N 11730 117 the carboxylate were still present and the improvement in the R A factors was too small with respect to the loss of one degree of CaNC' 12330 123 freedom.In addition, calculations for the staggered conform- + Ester group ations, in the cyclo(Gly-AspkYb3 system, led to slightly Ca-C' 1.5330*35 1.53 larger R values than for the eclipsed ones. Moreover, the C'-O 1.36;30 1.3435 1.35 deviation for C, was not reduced. Thus, we had to take into 0-Me 1.45;30 1.4735 1.45 consideration contact shifts, at least for nuclei close to the A CaC'0 ii5;30 11137 113 lanthanide ion. The removal from the analysis of the induced shifts of the carboxylate carbon and of the adjacent one (as inA C'OMe 108;36 10637 107 the case of europium complex) then gave acceptable R values Asp or Glu (4.3-5.5%, see Table 9). It should be observed that this procedure does not change the main features of the structure of c8-c~ 1.53 30 1.53 the ytterbium complexes, i.e., Ln3+ 0-distances, mono- nCaC gCr 115 30 115 dentate complexation, and extended side-chains [see Table 9 A and Figures 3 and (2a and c)]. 107;30lWJ8 108 Two slight differences should be noted, the o angles (most of HC8C7C-O 1.2537*39 1.25 them being now ca.180') and the N-acetylamino acid methyl c-co 1.5239 1.52 ester conformers, which differ from those previously determined A cco 118-119;39 117j7 118 only by a Ca-N rotation (a rotation defined by nuclei far from the site of interaction and thus not accurately positioned). These structural results could have been strengthened by relaxation measurements but this was not technically possible Table 8.Structural results for the Yb3+ complexes. One-conformer analysis involving all the n.m.r. 'H and 13C sites Cyclo(Gly-Asp) NAcAspOMe C yclo(G1 y-Glu) NAcGluOMe PlA 2.70 2.75 2.50 2.55 01" 122 117 142 138 A/" 25 -26 4 0 Rl% 10.8 8.1 13.6 14.0 cp -90"" cp + 90" yl -60"' " cp and w defined following the IUPAC-IUB conventions (Biochemistry, 1970.9, 3471). Table 9.Structural results for the Yb3 + complexes: one-conformer analysis after removal of the carboxylate and adjacent group 3C Cy~lo(Gly-Asp) NAcAspOMe Cyclo(G1 y-Glu) NAcGluOMe PIA 2.70 2.70 2.55 2.50 w/O 185 153 182 177 V" -35 -46 -29 -12 Rl% 5.3 4.3 5.5 5.0 ~p -150" cp -90" w -60" J. CHEM. soc. PEWN TRANS. 11 1985 Fm 2 Conformationsof the ligand in the Yb3+complexes.(a) Cyclo(G1y-Asp); (b)NAcAspOMe;(c)Cyclo(Gly4lu); and (d)NAcGluOMe.(b)and (d),Oneanformer analysis involving all the n.m.r.'H and "C sites Figure 3. Conformations of the ligand in the Yb3+ complexes (one-conformer analysis after removal of the carboxylate and adjacent group I3C). (a)NAcAspOMe; (b)NAcGluOMe. for nuclei in 0.1~peptide solution; moreover, too few 'H nuclei were available to discriminate the structures of the complexes just from the Tl measurements. Finally, broadening effects were in agreement with the order of the Yb3+ distance as deduced from the analysis of the LIS. On considering the thermodynamic results, it is worth noting that the Yb3+***O-distances, shorter in glutamatederivatives than in the aspartates, are consistent with the K, binding constant values, which are larger by 2&30"/, in the case of the glutamate complexes.These results are probably a consequence of the carboxylate basicity, unless some other interaction with the peptide backbone is taking place. In fact, whatever residue or peptide backbone is involved, the predominant conformer has an extended sidechain and, consequently, no chelation occurs with the peptide backbone. The fact that Kl for the N-acetylamino acid methyl esters is roughly twice that for the cyclodipeptides could arise from the larger degree of conformational freedom of the noncyclic backbones, which are not involved in the complexation. Finally, with the europium complex, it seems that the contact contribution is so large that the analysis based on pseudo- contact shifts fails, even when removing more than two induced shifts.Thus, to obtain structural information about this complex without specifying a parameter (lanthanide co-ordinates or substrate conformation, for example) does not seem possible. Conclusions Aspartate and glutamate side-chains form 1: 1 and 2: 1 complexes with lanthanide ions in an aqueous medium. A model assuming an average LIS value for the two complexes yields a good fit of the data, the normalized standard deviation being less than 5% This study suggests that two factors affect the binding constants: the peptide backbone flexibility and, to a lesser extent, the side-chain length. With Yb3+ as a probe, the main features of the structure of the ytterbium complexes can be obtained, using all the proton and the 13Cdata, but removal of the two carbons closest to the lanthanide allows a refinement of the results.In the four compounds, the complexation of the carboxylate is monodentate, the Yb3+ 0-distances being shorter in the Glu derivatives (2.50-2.55 A) than in the Asp (2.70-2.75 A).With the predominant conformation of the side-chain being extended, no chelation occurs with the peptide backbone. References 1 F. Inagaki and T. Miyazawa, Prog. Nucl. Magn. Res. Spectrosc., 1980, 14, part 2. 2 J. Reuben, Prog. Nucl. Magn. Res. Spectrosc., 1973, 9, part 1. 3 E. Nieboer, Struct. Bonding (Berlin), 1975, 22, 1. 4 A. D. Sherry and E. Pascual, J.Am. Chem. Soc., 1977,99,5871. 5 R. E. Davis and M. R. Wilcott 111, ‘Nuclear Magnetic Resonance, Shift Reagents,’ ed. R. E.Sievers, Academic Press, New York, 1973, p. 143. 6 B. A. Levine and R.J. P. Williams, Proc. R. Soc. London, Ser. A, 1975, 345, 5. 7 M. Delepierre, C. M. Dobson, and S. L. Menear, J. Chem. Soc., Dalton Trans., 1981, 678. 8 ‘Comprehensive Analytical Chemistry,’ eds. C. L. Wilson and D. W. Wilson, Elsevier, Amsterdam, 1960, vol. I B, p. 356. 9 G. R. Brubaker and N. Y. Sakkab, Bioinorg. Chem., 1974,3,243. 10 M. R. Harnden, J. Chem Soc. C, 1967, 2341. 11 A. Glasel, Prog. Znorg. Chem., 1973, 18, 383. 12 A. D. Sherry, C. Yoshida, E. R. Birnbaum, and D. W. Darnall, J. Am. Chem. SOC.,1973,95, 3011. 13 P. K. Glasoe and F.A. Long, J. Phys. Chem., 1960,64, 188. K. Wuthrich, ‘NMR in Biological Research Peptides and Proteins,’ North-Holland, Amsterdam, 1976, pp. 42, 170. C. D. Barry, A. C. T. North, J. A. Glasel, R. J. P. Williams, and A. V. Xavier, Nature (London), 1971, 232, 236. C. D. Barry, C. M. Dobson, R. J. P. Williams, and A. V. Xavier, J. Chem. SOC.,Dalton Trans., 1974, 1765. C. F. G. C. Geraldes and M. H. Mendonca-Dias, J. Magn. Reson., 1981, 45, 394. J. CHEM. SOC. PERKIN TRANS. Ir 1985 18 D. H. Sachs, A. N. Schechter, and J. S. Cohen, J. Biol, Chem., 1971, 246,6576. 19 B. T. Pennington and J. R. Cavanaugh, J.Magn. Reson., 1978,29,483. 20 G. A. Elgavish and J. Reuben, J. Magn. Reson., 1981,42, 242. 21 J. Mossoyan, M. Asso, and D. Benlian, J.Magn.Reson., 1982,46,289.22 C. J. Brookes, I. G. Betteley, and S.M. Loxston, ‘Mathematics and Statistics for Chemists,’ John Wiley, New York, 1966, p. 365. 23 M. Cefola, A. S.Tompa, A. V. Celiano, and P. S. Gentile, Znorg. Chem.: 1962,1,290. 24 G. R. Choppin and W. F. Strazik, Znorg. Chem., 1965,4, 1250. 25 W. Dew. Horrocks Jr., J. P. Sipe 111, and D. Sudnick, ‘Nuclear Magnetic Resonance, Shift Reagents,’ ed. R. E. Sievers, Academic Press, New York, 1973, p. 53. 26 J. W.M. De Boer, P. J. D. Sakkers, C. W. Hilbers, and E. De Boer, J. Magn. Reson., 1977, 25, 455. 27 A. D. Sherry, C. A. Stark, J. R. Ascenso, and C. F. G. C. Geraldes, J. Chem. Soc., Dalton Trans., 1981, 2078. 28 G. R. Choppin and A. J. Graffeo, Znorg. Chem., 1965,4, 1254.29 R. S. Kolat and J. E. Powell, Znorg. Chem., 1962, 1, 293. 30 T. S. Ooi, R. A. Scott, G. Vandekooi, and H. A. Scheraga, J. Chem. Phys., 1967,46,4410. 31 R. Degeilh and R. E. Marsh, Acta Crystallogr., 1959, 12, 1007. 32 E. Sletten, J. Am. Chem. Soc., 1970, 92, 172. 33 M. Cotrait, M. Ptak, B. Busetta, and A. Heitz, J. Am. Chem. Soc., 1976,98, 1073. 34 C. F. Lin and L. E. Webb, J. Am. Chem. Soc., 1973,95, 6803. 35 B. Pullman and A. Pullman, Adu. Protein Chem., 1974, 28, 347. 36 A. C. Hopkinson, K. Yates, and I. G. Csizmadia, J. Chem. Phys., 1970,52, 1784. 37 P. W. Borthwick, Acta Crystallogr., Sect. B, 1980, 36, 628. 38 J. L. Derissen, H. J. Endeman, and A. F. Peerdeman, Acta Crystallogr., Sect. B, 1968, 24, 1349. 39 H. Einspahr and C. E. Bugg, Acta Crystallogr., Sect. B, 1981,37,1044. 40 R. M. Moravie, Thesis, University of Paris-6, France, 1975. 41 G. Hawkes, D. Leibfritz, D. W. Roberts, and J. D. Roberts, J. Am. Chem. Soc., 1973,%, 1659. 42 D. J. Chadwick and D. H. Williams, J. Chem. Soc., Perkin Trans. 2, 1974, 1202. 43 B. A. Levine, J. M. Thornton, and R. J. P. Williams, J. Chem. Soc., Chem. Commun., 1974,669. 44 C. M. Dobson, L. 0.Ford, S. E. Summers, and R. J. P. Williams, J. Chem. Soc., Faraday Trans. 2, 1975,71, 1145. 45 S. P. Sinha, Struct. Bonding (Berlin), 1976, 25, 69. 46 H. M. McConnell and R. E. Robertson, J.Chem. fhys., 1958,29,136 1. 47 M. R.Willcott 111, R. E. Lenkinski, and R.E. Davis, J. Am. Chem. SOC.,1972, 94, 1742. Receivzd 8th August 1983; Paper 311387