NMR study of thymulin, a lymphocyte differentiating thymic nonapeptide. Conformational states of free peptide in solution.

The nonapeptide less than Glu-Ala-Lys-Ser-Gln-Gly-Gly-Ser-Asn (formerly called serum thymic factor) is a factor produced by the thymic epithelium, which needs a zinc ion to express its immunoregulatory properties. We report here on 1H and 13C NMR investigation of the conformational properties of the free peptide in aqueous medium and in dimethyl sulfoxide-d6 solution by a combination of homo- and heteronuclear one- and two-dimensional experiments. The various resonances have been assigned in a straightforward manner on the basis of 1H,1H COSY spectroscopy for the recognition of the proton spin systems; two-dimensional NOESY spectra with the correlation peaks across amide bonds and for the amino acid sequence assignment; amide bonds and for the amino acid sequence assignment; 13C,1H COSY experiments using selective polarization transfer from 1H- to 13C-nucleus via the 13C,1H long-range couplings for the attribution of the carboxyl and carbonyl groups; and 13C,1H COSY experiments with selective polarization transfer via the 13C,1H direct couplings for the assignment of all the aliphatic carbons. Other experiments such as pH-dependent chemical shifts, combined use of multiple and selective proton-decoupled 1H and 13C NMR spectra, the temperature and the concentration dependence of the proton shifts of the amide resonances, the solvent dependences of peptide carbonyl carbon resonances, and comparison of the spectra with three different analogues were performed. In aqueous solution, the data are compatible with the assumption of a highly mobile dynamic equilibrium among different conformations, whereas in dimethyl sulfoxide-d6, a more rigid structure is found involving three internal hydrogen bonds. These observations provide an insight into the conformational tendencies of this peptidic hormone in two different media.

The nonapeptide <Glu-Ala-Lys-Ser-Gln-Gly-Gly-Ser-Asn (formerly called serum thymic factor) is a factor produced by the thymic epithelium, which needs a zinc ion to express its immunoregulatory properties. We report here on 'H and 13C NMR investigation of the conformational properties of the free peptide in aqueous medium and in dimethyl sulfoxide-d6 solution by a combination of homo-and heteronuclear one-and two-dimensional experiments. The various resonances have been assigned in a straighforward manner on the basis of (i) 'H,'H COSY spectroscopy for the recognition of the proton spin systems; (ii) two-dimensional NOESY spectra with the correlation peaks across amide bonds and for the amino acid sequence assignment; (iii) 13C,lH COSY experiments using selective polarization transfer from 'H-to 13C-nucleus via the 13C,lH long-range couplings for the attribution of the carboxyl and carbonyl groups; and (iv) 13C,lH COSY experiments with selective polarization transfer via the 13C,lH direct couplings for the assignment of all the aliphatic carbons.
Other experiments such as (i) pH-dependent chemical shifts, (ii) combined use of multiple and selective proton-decoupled 'H and 13C NMR spectra, (iii) the temperature and the concentration dependence of the proton shifts of the amide resonances, (iv) the solvent dependences of peptide carbonyl carbon resonances, and (v) comparison of the spectra with three different analogues were performed. In aqueous solution, the data are compatible with the assumption of a highly mobile dynamic equilibrium among different conformations, whereas in dimethyl sulfoxide-de, a more rigid structure is found involving three internal hydrogen bonds. These observations provide an insight into the conformational tendencies of this peptidic hormone in two different media.
Thymulin, formerly called serum thymic factor, is a metallopeptidic hormone, selectively produced by thymic epithelial cells, able to induce T-cell markers, and functions on immature cells (1). It is a nonapeptide which was first isolated from * This research was supported by the Centre National de la Recherche Scientifique and the Institut National de la Sant6 et de la Recherche MBdicale. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
To whom correspondence should be addressed. porcine serum by Bach and co-workers (2, 3) and ultimately shown to be present in calf thymus extract. Its amino acid sequence was determined (<Glu-Ala-Lys-Ser-Gln-Gly-Gly-Ser-Am), and the synthetic peptide was shown to be fully biologically active (3) in the presence of a zinc ion (4). Nevertheless, a comparative study of the conformational tendencies of free and complexed peptide is of interest to determine the conformation-activity relationships between the two important species.
In this work, we have attempted to obtain information on the conformational states of the nonapeptide in aqueous medium and in dimethyl sulfoxide solution by means of oneand two-dimensional 13C,lH NMR and CD spectroscopy. The spectra of several analogues with Ala4, Alas, and Nva5 substitution were also recorded for comparison with thymulin. Complete assignments of the spectra were made by 'H,lH and 13C,lH COSY and 'H,lH NOESY experiments. A conformational analysis of the nonapeptide has been achieved by NMR through measurements of the concentration, temperature, and solvent dependence of chemical shifts.
The combined use of NMR and CD spectroscopy indicates strongly that the nonapeptide in aqueous solution is flexible and is in rapid equilibrium between multiple conformations. In dimethyl sulfoxide-ds, the peptide adopts a partially folded structure stabilized by three intramolecular hydrogen bonds.

EXPERIMENTAL PROCEDURES
Materials-Synthetic serum thymic factor was provided by the Institut Choay, Paris, France (5), whereas the three analogues, Ala4, Ala', and Nva6, were a generous gift from Dr. D. Blanot (University of Paris-Sud, Orsay, France). They ,were synthesized as described previously (6, 7), purified on pBondapak Cis, and eluted with 50 mM ammonium formate buffer at pH 3.5. All the materials were dissolved in HzO at neutral pH, lyophilized, and then dissolved in DzO or dimethyl sulfoxide-de immediately before the measurements were made.
One-dimensional NMR Instrumentation-DC1 and NaOD solutions were used to adjust the pH which was measured with a Knick digital pH-meter, calibrated with standard buffers (Merck Titrisol), and an Ingold microelectrode. The pH values are not corrected for deuterium isotope effects.
The concentration, temperature, and solvent dependence of the chemical shifts was studied by one-dimensional NMR experiments on a Briiker WM-250 spectrometer equipped with Aspect 2000 Computer (v0('H) 250 MHz, v0(13C) 62.8 MHz). The usual 'H spectrometer conditions were 2,800-Hz sweep width, 16,384 data points, and 32 scans. The HOD signal was considerably reduced by pre-7784 saturation between scans. 13C NMR spectra were recorded with quadrature detection and broad-band proton decoupling. 30,000-40,000 transients are necessary for a 0.05 M solution of peptide. Other conditions were: spectra width, 15,000 Hz; size, 16,384 data points; repetition time, 0.5 s. Both 'H and 13C chemical shifts were measured in parts/million with 4,4-dimethyl-4-silapentane-l-sulfonate or tetramethylsilane as reference. Several proton homodecoupling experiments were performed automatically by a microprogram edited with one frequency list, similar to those of NOE' difference experiments.
Two-dimensional NMR Instrumentation-All two-dimensional experiments were performed on a Bliiker AM-400 apparatus (a('H) 400.13 MHz, vO(l3C) 100.58 MHz) by using the standard Briiker microprograms and quadrature detection in both dimensions. A 'H,'H COSY 45 was used for minimization of diagonal peaks.
For 'H,'H COSY 45" experiments, the sequence 90"-t1-45" acquisition was performed; relaxation delay was 2 s; and a 90" pulse for 8.6 ps was used. For a dimethyl sulfoxide concentration of 0.005 M, the acquisition time was 256 ms; spectral width in F1 and FZ was 4000 Hz; 512 experiments with 64 scans of 2048 points were performed; data points in tl were zero-filled to give a (1024 X 1024) data matrix; and sine-bell apodization was performed in both dimensions. For a DzO concentration of 0.01 M, the acquisition time was 0.320 ms; spectral width was 1600 Hz in both dimensions; and 256 experiments with 32 scans of 1024 points were performed.
For 'H,lH NOESY experiments, a 0.005 M solution of the nonapeptide was prepared in dimethyl sulfoxide-ds (CEA loo%, isotopic purity). The sample was degassed by several freeze-thaw cycles on a high vacuum line to remove dissolved oxygen. The sequence 90"-tl-90"-tm-9O" acquisition was performed with three mixing times (0.2, 0.4, and 0.6 s). A 15% random variation of mixing time was applied to cancel scalar correlation effects. A relaxation delay of 5 s was inserted between scans to ensure quantitative correlation peak intensities. Other conditions were: 90" pulse for 8. Results are expressed in molar ellipticity [8] = 3300 At degrees cmz/ dmol. H" proton and [13C]carbonyl group two-dimensional NMR correlation.

'H NMR Assignments in
The correlation peaks of the Lys3 spin systems in the COSY map draw some remarks: (i) a small intensity of cross-correlated peaks between a-and one @-proton indicative of a small value of the corresponding coupling constant and (ii) lack of @/y and y/6 cross-correlation peaks probably due to small coupling constants.
T h e spin decoupling and the pH-dependent chemical shift experiments were consistent with the assignment indicated in Fig. 1. For example, the titration of the Asng spin system was observed at pH g 3.1 (doublet of doublet centered at 4.522 ppm corresponding to the proton Ha; multiplet structure at 2.803 and 2.674 ppm for the two protons HB). The four methylene protons of the Lys3 side chain are also pH-dependent at pH E 10.8. The remaining resonances of the spectrum are essentially pH-insensitive. I3C NMR Assignments in Aqueous Medium-Spin system assignment and sequence determination of peptide residues are a challenge to NMR spectra analysis. Several combined two-dimensional homo-and heteronuclear shift correlations are helpful in solving this problem. To this end, we have used the two-dimensional shift correlation technique developed by Bax and Morris (8) which was also reported elsewhere (9) as a convenient method of assigning CO signals. Two separate heteronuclear shift-correlated two-dimensional experiments using the selective polarization transfer from the proton to the carbon 13C-nucleus via the scalar couplings were performed on the nonapeptide aqueous solution for assignment of the carboxyl and carbonyl region. In the peptide sequence, each carbonyl 13Ci0 is coupled with three kinds of protons: two NH, two a-protons, and one or two @-protons (Fig. 2). The long-range scalar couplings (-4 to +7.5 Hz) can be used to assign the carbonyl groups and to obtain the connectivity of the peptide residues. The two-dimensional heteronuclear shift correlation between 13C0 carbons and protons was performed by the selective polarization transfer technique via the k 5 -H~ average long-range coupling. In aqueous solution, all amide groups are exchanged; each carbonyl group is only Cross-peaks yield the complete 13C0 assignment.
coupled to the a-and ,&protons of its own residue and to the a-proton of the neighboring peptide residue. All cross-correlated peaks are observed and allowed the attribution of the carbonyl groups (Fig. 3).
The <Glul proton spin system is characterized by the crosscorrelated peak at 184.7 ppm between the <Glul a-proton and the C*O carbonyl group. The cross-peak affords a distinction between the two C60 carbonyl groups of <Glul and Gln5. Moreover, the <Glu' and Gln5-CO groups at 177.2 and 176.2 ppm are correlated to the Ala2 and Gly6-H" protons, respectively.
The recognition of the Ser4, Sera, Gly6, and Gly7 residue spin systems is straightforward, thanks to the presence of cross-peaks relative to the Lys3-Ser4, Ser4-Gln', Gln5-Gly6, Gly7-Sers and SerS-AsnQ dyads. All /3-protons are also assigned, and the conclusion.is consistent with the 'H,lH COSY map.
The pH dependence of 13C resonances (Fig. 4) was studied to obtain eventual long range interactions resulting from titrations of the different charged groups. The COOH-terminal Asn9-COO-carboxylate group and the Lys3-NH$ ammonium group define two titration domains at pH = 3.0 and 11.0. Three CO groups titrate at acidic medium. The AsnQ-COO-peak at 176.5 ppm undergoes a downfield shift of 2.8 ppm; similarly the AsnQ-CYO at 177.3 ppm titrates downfield (0.9 ppm), whereas the Sera-CO is shifted upfield by 0.5 ppm by the deprotonation of the Asng-COO-group. The only resonance affected at basic pH corresponds tQ Lys3-C0 at For assignment of the aliphatic 13C region, the value of 135 Hz was used as the average 'Jc.H direct coupling in the heteronuclear shift-correlated experiment. Cross-correlation peaks in the 13C,lH COSY map (Fig. 5 ) allow the complete assignment of side chain aliphatic carbons.
The Ser-13C8signal which undergoes a small downfield shift of 0.1 ppm in acidic medium was identified as Se1.8-13CP on the basis of the proximity effect of the Asng-COO-terminal group. The pH dependence for aliphatic carbons of the nonapeptide is similar to that of the carbonyl groups: the AsnQ-13C signals titrate downfield at pH 3.1, and all the Lys3-13C signals experience an upfield shift at pH 11. Other remaining resonances are essentially pH-insensitive (Fig. 4).
'H NMR Spectra in Dimethyl Sulfoxide-d6-Because of its high polarity, water might destabilize intramolecular hydrogen bonds and cause an unfolded structure with amides exposed to the solvent. In nonaqueous environments, particularly dimethyl sulfoxide, an aprotic solvent, the intramolecular hydrogen bonds and the folded conformation are in general less perturbed than in water. Consequently the tendencies to intramolecular hydrogen bond formation and electrostatic interactions are greater in dimethyl sulfoxide than in water. For this reason, the conformation and complexation study of the nonapeptide was performed in dimethyl sulfoxide by examining, in particular, the amide NH resonances and their variation with temperature and concentration. Furthermore, the conformations of hormone peptides in dimethyl sulfoxide are probably more relevant to their physiological activities and to the conformations at the receptor site than to the conformations in water because the environment in the organic solvent is more similar to the lipid phase (11). Thus, it seems that dimethyl sulfoxide solutions may mimic in some respects the medium at the surface of cell membranes, where receptors for the thymulin peptide may be located. The use of freshly prepared solutions shows that the tendency for the line widths of resonance to broaden with increasing concentration is negligible. The present line width data provide no evidence for a change in aggregation of the peptide over the concentration range 0.7-12 mM. The concentration dependence of the chemical shifts of the different NH protons in dimethyl sulfoxide has been examined. The small variations of the chemical shifts observed (upfield of 0.07 ppm for Lys3-NH and 0.05 ppm for Gln5-NH) at concentrations ranging from 0.7 to 12 mM showed that there were only very few and negligible intermolecular associations. As the concentration is raised, however, concentration dependence of chemical shifts is observed for the Amg-NH and Gly6-NH resonances, which shift downfield by 0.10 and 0.12 ppm, respectively. This observation indicates the possibility of the intermolecular associations. Consequently, all our spectra were obtained in dilute solution (10 mM) in order to minimize peptide aggregation.
Exactly as for the aqueous solution, the connectivities within the spin systems of each amino acid residue in the thymulin were obtained by a two-dimensional COSY spectrum (Fig. 6) giving a direct assignment for the proton resonances of <Glu', Ala2, Lys3, Gln5, and Amg. The absence of cross-peaks between the Lys3-@/y-and y/&proton couples is indicative of small average scalar coupling values.
The Ser4/Sera and Gly6/Gly7 spin systems were identified on the basis of NOESY experiments which involved the observation of the NOE effects between the amide and aproton of neighboring amino acid residues (12). Some NOE correlation peaks between the a-proton and the amide NH proton of the following neighboring residue are shown in Fig.  7. This is the case of Gln5-NH/Ser4-Ha, Ams-NH/Sera-Ha, and Sera-NH/Gly7-H". These cross-correlation peaks allow the sequence attribution and are used to distinguish between Ser4/Ser8 and Gly6/Gly7 residues. Additional selective homodecoupling experiments complete this attribution. The difference spectrum between coupled and uncoupled spectra (Fig.  8) was used to calculate all coupling constant values in the ABX spin systems of Gly6 and Gly7 (Fig. 8, a and b) and Ser4 and Sefl (Fig. 8, c and d). 'H NMR data and the rotamer distribution of the Ca-CB bond in the Ser4, Sera, and Asng residues are given in Table I. One of the purposes of peptide conformational analyses is the elucidation of the intramolecular hydrogen bonds, which is of great help in gaining knowledge about the different conformers. The most widely used procedure is to measure the influence of the temperature on the amide proton chemical shifts. Since the concentration dependence over the range of 0.7-12 mM indicates that intermolecular interactions were negligible in such solutions, the temperature coefficients of chemical shifts of amide proton resonances may be used for distinguishing between "exposed" and "hydrogen-bonded" amide   Tentative assignment on the basis of the general attribution by Percentages of the C"-CB rotamers I, 11, and I11 estimated from selective @-deuteration (13-15). protons. Exposed protons exhibit larger temperature coefficients than do hydrogen-bonded protons (17,18). The temperature coefficients (d&/dT) for all w i d e protons of the nonapeptide are given in Table 11. The one-and two-dimensional nuclear Overhauser effect experiments were also performed with several mixing times. Excepting the short-range NOE effects used for the Ser4/Sefl and Gly6/Gly7 assignment (Fig. 7), no long-range NOE effect was observed in any of the cases considered. This result seems to indicate the absence of cross-relaxation due to the dipoledipole interaction between two protons. In this case, th? average interproton distances are probably more than 3.0 A (19). 13C NMR Spectra in Dimethyl Sulfoxide-d6-To avoid an eventual aggregation in dimethyl sulfoxide, the concentration of the peptide solution must be quite low. In these conditions, two-dimensional heteronuclear shift correlations would be exceedingly time-consuming.

TABLE I1 3JNH.caH (Hz) coupling constants and temperature dependence coefficients dG/dT ((ppml°C) X I @ ) of the nonapeptide in dimethyl sulfoxide-d6
Nevertheless, 13C NMR was used to delineate the internal hydrogen bonds in peptides (20,21). It has been shown that solvent titrations using the solvent pair dimethyl sulfoxideds/D,O can discriminate between solvent-exposed and solvent-shielded peptide carbonyls. Indeed, D20 is a good proton donor, whereas dimethyl sulfoxide-& is a proton acceptor, and the resonances of solvent-exposed carbonyls shift downfield on addition of D20 to dimethyl sulfoxide-d6 solutions to a greater extent than do those of solvent-shielded carbonyls. The attribution of the 13C resonances in dimethyl sulfoxide was then achieved by the progressive addition of dimethyl sulfoxide-d6 to DzO. The results are listed in Table 111. All resonances shift downfield in the range of 4 ppm to more than 7 ppm. Gly7-CO is seen to shift least and is taken as the internal reference. Lys3-CO, Gln5-CO, and Sers-CO exhibit a small shift relative to Gly7-CO, whereas all the remaining resonances show substantial shift.

DISCUSSION
Nonapeptide in Aqueous Solution-The nonapeptide has only two titratable protons in the pH range studied (1.70-12.20); they are attributed to the terminal a-carboxyl group and to the c-amino group of the lysine residue. Dissociation shifts of these ionizable groups produce characteristic changes in the chemical shifts of resonances emanating from proton and carbon atoms in the proximity of the titratable group. These parameters, as a function of pH, due to the direct influence of changes in electron density about the individual atoms or indirectly from an altered conformation of the peptide, are able to yield rich information about intramolecular interactions and through-bond and through-space influences (22). One of these indications is obtained from the dissociation constants (pK,). Thus, the two pK, values obtained are close to that found in small peptides or amino acids (23), whereas the chemical shifts of many of the assigned resonances are very similar to the corresponding chemical shifts observed in these peptides. These findings indicate that the Asn' carboxyl group and the Lys3 e-amino group are not involved in salt-bridge formation or hydrogen bonds with positively or negatively charged groups.
However, the small downfield shift observed for the <Glu'-C"H and Ala2-C"H resonances near pK, 2 10.6 could be due to spatial closeness of the c-amino group and may reflect through-space effects. These observations agree well with the interesting behavior encountered in basic medium. At pH > 11, some of the resonances corresponding to the -Ala2-LysS- whereas a marked intensity decrease of the positive maximum at 217 nm is observed (Fig. 9). All the observations made (a very small proximity titration effect upon pH variation, absence of NOE effects throughspace, and strong negative peak at 205 nm in the CD spectrum) converge to the conclusion that the nonapeptide probably does not adopt any highly preferred folded conformation in DzO. This nonapeptide is probably flexible and assumes multiple conformations in rapid equilibrium with no substantial contribution of structural features such as @ turns, internal hydrogen bonds, oi hydrophobic interactions.
The unexpected titration deviation at basic pH could suggest that the lysine chain may be giving privileged conformation in this extreme pH domain.
Peptide in Dimethyl Sulfoxide Solution-As shown in Table   11, when the temperature was raised from 20 to 60 "C, the signal assigned to the trans-carboxamide proton of Amg-NHz (24) and the signal of the Gly6-NH proton were little affected. Similarly, the Ams-NH resonance shows small variation, whereas the Lys3-NH move upfield by 6.5 X ppm/"C. The lack of effect of the temperature variation on the three truns-Asng-NHz, Gly6-NH, and Amg-NH establishes that they are solvent-protected and strongly suggests that they are at least partly involved in intramolecular interactions.
From the solvent titration of peptide CO chemical shifts, it appears that the Lys3-CO, Gly7-CO, and Gln5-CO moieties are solvent-shielded and become candidates for intramolecular hydrogen bonding to be appropriately paired with the delineated peptide NH moieties of the above 'H studies.
The temperature dependence and solvent perturbation studies of the peptide NH protons and CO carbons suggest that a possible folded structure (Fig. 10) for the monomeric nonapeptide could contain a 10-membered H bond encompassing the Ser4-Gln5 sequence (@-turn), an 11-membered H bond between the truns-Asng-NHz side chain carboxamide and the Gly7-CO, plus a 13-membered H bond between the Amg-NH and the Gln5-CO. The small value of the Sers-CO perturbation can be explained by a rather weak shielding effect exerted by the Sers-OYH bond on it. However, the absence of cross-relaxation between proton groups in the oneand two-dimensional NOE experiments indicates that this folded structure is probably not predominant or is in rapid equilibrium with other random or open structures. The medium coupling constants observed ( J N H -C W ranging from 5.2 to 7.5 Hz, Table 11) are effectively consistent with fluctuating conformations in rapid equilibrium (25).
The vicinal coupling constants 3 J a p~ and 3 J a @~ in Ser4, Ser', and Asng indicate a high percentage of the rotamer I11 (xl = 60") for the three C"-Cp bonds (Table I). In the case of the Ser4 residue, this observation is compatible with the @l turn of the folded structure (26).

CONCLUSION
The NMR and CD data presented here strongly indicate that the nonapeptide in aqueous solution is flexible and assumes multiple conformations in rapid equilibrium. Temperature dependence and solvent perturbation studies in dimethyl sulfoxide-d6 solution suggest the existence of more defined conformational properties with some amount of a folded structure stabilized by three intramolecular hydrogen bonds. The 3-9 COOH-terminal part of the nonapeptide is solvent-shielded, whereas the <Glu1-Ala2 NH2-terminal part is solvent-exposed. Although the free nonapeptide hormone is not active in uiuo, the present studies could help in understanding how the complexation by zinc can modify the ternary structure of the thymulin-zinc complex, making it highly active. Such a comparison was considered very important for the structure-activity relationship since the thymic hormone is virtually devoid of biological activity. Work is now in progress to determine more accurately the exact secondary structure of the folding conformation and the conformational influence of Zn2+ complexation.