Int. J. Peptide Protein Res. 39, 1992, 1 1 1-1 16

NMR study on solution structure of the site-specific mutant Leu48+Ala transforming growth factor alphas TAMMY PAGE KLINE and LUCIAN0 MUELLER'

SmithKline Beecham Pharmaceuticals, King of Prussia, Pennsylvania, USA

Received 5 February, accepted for publication 21 September 1991 The NMR spectra of the Leu48-tAla mutant of human transforming growth factor CI were compared to that of the wild-type. All chemical shift changes are 50.02 ppm with the exception of resonances associated with residues 47, 48 and 50 (all 50.07 ppm). Minimal changes were observed for NOEs associated with residues Vall to His45. The weakening of some NOEs associated with the region Ala46-Ala50 may suggest a slightly increased flexibility for this region. Refinement of the previously calculated wild-type structures using distance constraints derived from the L48A mutant had little overall effect. Leu48-Ala50 is ill-defined for both wildtype and mutant proteins. These results suggest that Leu48 has no structural role and thus must be an important factor in the protein-receptor interface. Key Miords: molecular mechanics calculations; NMR structure; site-directed mutagenesis; structure-activity relationships;

transforming growth factor a

Human transforming growth factor alpha (TGF-a) is a ibility on the NMR time scale); 2) an irregular, possibly 50-residue polypeptide implicated as an autocrine fac- flexible segment AsplO-Phel5; 3) a P-sheet consisting tor involved in the growth of many tumor cells (1-6). of Gly19-Cys34; 4) a type I p-turn at Va125-Asp28; 5) a TGF-a is a member of the family of epidermal growth type I1 p-turn at His35-Tyr38; 6) a left-handed twist of factor (EGF)-like peptides (7-9) and most of its bio- the loop Gly40-Glu44: 7) a short antiparallel b-sheet logical activity is expressed via binding to the E G F between Tyr38-Val39 and His45-Ala46; and 8) a highly receptor (10-12). In order to determine the role of flexible C-terminal segment Leu48-Ala50. Other strucTGF-cl in the process of tumorogenesis, it is important tural studies have been done and are consistent with to understand how the polypeptide structure relates to our results (14-19). The second step is to identify which residues are its binding to the receptor and its consequent biological activity. The process of deciphering this structure- essential for receptor binding or biological activity. One function relationship involves several steps. First, the common method of addressing this issue is to mutate structure of the biologically active species in solution specific sites and monitor the effects on binding/activity. must be determined. This step has recently been com- Several investigations have been carried out (20-24) pleted by this group using computer-based refinement and, thus far, binding and activity are always associprocedures with NOE-derived distance constraints (13). ated with each other. A number of residues have been The structure of desValVal TGF-a at pH 6.3 was found established as being important to the structure or bioto possess the following characteristics: 1) a short triple- logical activity of TGF-a. Once it has been established which residues arc esstranded sheet involving the N-terminal segment [VallAsp71 (which, however, appears to maintain some flex- sential, the third step is to determine which role those residues predominantly play, structural or functional. A residue performing a primarily structural role has little contact with the receptor, but its mutation may alter the § This work funded in part by the National Institutes of Health structure of the protein to the extent that binding/ Grant no. GM-39526. activity is affected. A purely functional residue, on the Present address: Squibb Institute for Medical Research, Room other hand, is usually present at the surface of the proD4159, P.O. Box 4000, Princeton, NJ 08543. +

111

T.P. Kline and L. Mueller tein and has little to no involvement with overall struc- Structural calculations ture. In this case, mutation directly affects the interface Distance constraints were generated from resolved between the protein and the receptor. Some residues NOE crosspeaks. Distances involving amide and aromay combine the two roles. For purposes of designing matic protons are difficult to quantitatively measure an agonist or antagonist, it is necessary to determine from volume integrals due to amide exchange and arthe role of each essential residue in the activity of the omatic ring rotation. Therefore, the lower bounds wer: set to 1.8 A and ypper bounds were estima!ed at 2.7 A protein, whether structural, functional, or both. One can predict a residue's role by relating the mu- for strong, 3.5 A for medium, and 4.5 A for weak tation effect to the residue's position in the wild-type crosspeaks. For crosspeaks involving aliphatic protons, - 15% of the volume structure. For example, an essential residue buried in bounds were estimated as + 20 the protein may be predicted to be structural. However, integrals obtained by manual peak picking in FTNMR. the role of some residues may not be obvious from the The intern+ marker for volume integrals was d6,Eof wild-type structure. In order to resolve any ambiguity, Tyr = 2.45 A. For AMBER calculations, a distance of the structure of the mutated protein should be deter- 1 A was added to the upper bound of constraints inmined. Mutation of any purely functional residues volving methyl protons due to the treatmen: of methyl should have no effect on the overall protein structure. groups as pseudoatoms. A distance of 1 A was also Leu48 has been predicted to be an important func- added to the upper bounds of histidine ring proton tional residue of TGF-x (16, 17, 21, 24, 25). Several constraints in order to compensate for ring rotation. studies have shown that any change of this residue, Distance constraints used for WT and L48A refineeven minor, has a dramatic effect on receptor binding ments are listed in a supplemental table available from and activity. It is also completely conserved among the the authors on request. EGF-like growth factors. Examination of the calcuThe programs AMBER and XPLOR were previlated desValVal wild-type structure (13) (WT) reveals ously employed to calculate TGF-u wild-type structhat Leu48 is located in an ill-defined region at the C- tures (13). The objective of using two different molecterminal end of the protein and seems to have no sig- ular mechanics procedures, rather than just one, was to nificant structural role. In this study, we have examined be able to obtain a better sampling of conformational the site-directed desValVal mutant Leu48-+Ala(L48A) space. Final refinement involved minimization of ten (24) by NMR and found little difference from the WT structures by AMBER 3.0 (30) and simulated annealstructure, thereby confirming a functional role for Leu48. ing (3 1) of 12 structures by XPLOR. Two new distance Also, two new constraints were added to the distance constraints were recently added to tbe WT constraints: bounds for the WT structures, resulting in better defi- Leu48 H ~ - A l a 4 6Mep (1.80-5.50 A) and Leu48 HNnition for residues 46-48. Asp47 H, (1.80-3.00 A). The previously calculated WT structures were further refined using the methods described below. The 10 WT structures calculated using AMBER were METHODS refined by molecular dynamics with the following proNMR spectroscopy cedure: the structures were minimized in 500 steps with The L48A mutant of TGF-cx was dissolved to a con- all charges set to zero, heated from 0 to 700 K over centration of 3 mM in 50mM phosphate buffer 2 ps, equilibrated for 5 ps at 700 K, cooled from 700 to (pH = 6.3), 0.1 mM EDTA, and 0.1 mM NaN3. DQF- 0 K over 2 ps, and minimized for a final 500 cycles. A COSY and NOESY spectra were collected on a Bruker flat-bottomed well potential with harmonic sides was AM600 spectrometer at 20". The PECOSY experiment used for NOE-specified bounds such that no penalty was done on a JEOL GSX5OO at the same tempera- was assessed for distances within the bounds. A force constant of 50 kcal/niol A was used to enforce the ture. NOE-derived constraints. For some residues, torsion Assignments angles between C, and Cp protons were fixed by imThe 'H NMR spectrum of the L48A mutant of TGF- posing constraints of 2.9-3.1 (180" 30") fo; the Cg x was assigned in the same manner as that described proton in the trans position and 2.3-2.6A (60", for WT (13). Through-bond connectivities were estab- -60" t 30") for the gauche proton. lished by DQF-COSY (26) and linked sequentially by The 12 WT structures calculated using XPLOR were NOE crosspeaks (27). The strong similarity of the L48A refined with the dynamic simulated annealing procemutant 'H NMR spectrum to that of WT facilitated dure. In addition to the distance constraints, some C,H these assignments. Also, the NOE identification pro- and some CoH torsion angles were fixed for an intergram IDNOE and spin diffusion at long mixing times val of 30" with a quadratic force field and a force were used to resolve ambiguities in assigning NOE constant of 10 kcal/(mol rad2). The force constantofor crosspeaks. These methods have been described pre- the distance constraints was set to 50 kcal/(mol A2). viously (13). A PECOSY (28, 29) spectrum collected in Due to the high degree of similarity of the L48A D20 gave 3 J ~ z .coupling ~p constants. NOESY spectra to that of WT, the newly refined WT

x,

A

*

112

*

Transforming growth factor structures, after modification of residue 48, were used as starting structures for refinement with the L48A distance constraints. The same procedures utilizing AMBER and XPLOR were used as described above. RMS deviations for the WT structures were calculated between one structure, randomly chosen, and each of the remaining structures for the backbone atoms (N, C,, C, 0).Either the average of these calculated values (average RMS deviation) or the range of RMS deviations is reported. The same was done for the L48A structures. RESULTS AND DISCUSSION

Chenzical shifts Few significant chemical shift changes were observed in the 'H NMR spectrum of TGF-a upon mutation of Leu48 to Ala. Aside from resonances arising from the local area around residue 48, all chemical shift changes are 5 0.02 ppm (Fig. 1). The marginal effect of the mutation upon the protons in the vicinity of residue 48 suggests no major alteration in average structure for this region. With the exception of HN of Ala50 ( + 0.05 ppm, presumably due to sensitivity of the terminal HN chemical shift to solution conditions), the chemical shift changes for residues Ala46, Leu49, and Ala50 are 1 0 . 0 2 ppm. The CH, and CHBresonances of Asp47 undergo more significant shifts (-0.04 and -0.03, + 0.06 ppm, respectively), but considering its proximity to the mutated residue, this is probably a result of a change in environment rather than any backbone conformational change. 3 J ~ c r , coupling ~p constants were also measured from a PECOSY experi-

Arnide

Protons

tl

ment. The resulting similarity of values measured for residues possessing a CpHz center with corresponding values for WT (13) indicates that the mutation has no effect on H-C,-CpH torsional angles.

Further rejnement of WT structures For WT at pH 6.4, the Asp47 and Leu48 HN resonances were coincident and, therefore, two crosspeaks involving these residues were unassigned at the time when calculations were performed for the previously reported WT structures. The different chemical shift of the residue 48 H N resonance after mutation resulted in the unambiguous assignment of the crosspeaks AIa48 H~-Ala46 MeD and Ala48 H ~ - A s p 4 7 H,. Data obtained for WT at pH 3.4 and 5.6 demonstrate that these crosspeaks are also present for the unaltered protein as well and are of similar intensity. Distance constraints derived from these two newly assigned crosspeaks involving HN of residue 48 were added to the WT constraints and the WT structures were further refined. No significant conformational change was observed, however, the local region around residue 48 became better defined. The average RMS deviation for residues 46-48 decreased from 1.25 A to 0.95 A for the AMBER structures and from 1.02 A to 0.73 A for the XPLOR structures. L48A structural NOES and calculations The observed NOE crosspeaks for the L48A mutant are very similar in identity and intensity to those observed for WT (Fig. 2). A few exceptions, however, were attributable to local effects of the L48A mutation. In general, the major effect was a weakening of a few

Alpha

Protons

7.c

7.:

L 4

8

9.c

H

"

8.5

t a

9.c

Wild Yype

FIGURE 1 Chemical shifts of wild-type (WT) change in chemical shift occurs.

vs

Wild Type

Leu48-Ala (L48A) TGF-a for amide and alpha protons. Diagonal line indicates position where no

113

T.P. Kline and L. Mueller

.

8

P

i

O

a P

'

.o

a

0

a

D -

*

'!r .

9 .*5

9:o

8:s

8 ,' 0

1-

a Y E

n

7.5

PPm

0 V

'I

I

,

9.5

I

9.0

8.5

2.3

'.5

ment using distance constraints derived from L48A NOESY spectra. No significant change was detectable in the overall structure of the protein (Fig. 3). Also, RMS deviations for residues 16-46 of the LA8A structures fall in the samemvaluerange as those of the WT structyes (1.0 to 3.0 A for AMBER structures and 0.8 to 2.0A for XPLOR structures). The region 48-50 remained ill-defined. Specifically for residues 47 and 48, Ramachandran diagrams ($ vs $)for the WT and L48A structFres are similar. All NOE violations are less Jhan 0.23 A for AMBER structures and less than 0.40 A for XPLOR structures. The purpose of this study was not a comparison of the AMBER and XPLOR methods, but we think it is worth noting that, given the same constraints, AMBER seems to generate a slightly greater variety of conformations. This is particularly noticeable in the region Leu(Ala)48-AlaSO (Fig. 3). Also, RMS deviations for the whole molecule are considerably lower for the XPLOR structures (see above). We are uncertain of why this is so, but it is of interest to note that whereas AMBER uses bond angle force constants in the range of 70-80 kcal/(mol rad'), XPLOR uses the considerably higher Kanglcof SO0 kcal/(mol rad2). Also, NOE violations are greater for the XPLOR structures than for the AMBER structures. This suggests that there is some parameter(s), possibly Kangle,that is dominant over KNOEin the XPLOR refinement pathway and may drive the structures towards conformity.

PCV

Stnrcture-firnction relutiortship The relationship between structure and function of T G F - r (and other members of the EGF-like family) is not currently understood. Particularly, it is not known whether binding to the receptor and signal transduction occur through one or more domains of the molecule or through which residues. Recent site-directed mutagenNOEs. These include Asp47 H.-Val39 Me,,, Ala- esis studies (20-24) have reported several mutants with (Leu)48 H~-Leu49 HN, and Leu49 H N - A ~ ~ S O HN. low or no activity. Residues predicted to perform strucAlso, NOEs observed between Asp47 Hp protons and tural roles include the 6 cysteines, 2 glycines (Gly19, Val39 Me,.I for WT are no longer present for L48A. Gly40), Va133, and probably Tyr38. These results suggest a possible increase in flexibility It has been proposed (24) that two domains of the for 47-50 and weakening of the association between TGF-r molecule are involved in the protein-receptor residues 47 and 39. interface. The first domain consists of residues Phel5, As noted above, the chemical shifts and the NOESY Phel7, and Arg42. Although located in nonscquential spectrum of L48A are very similar to that of WT, sug- regions, in the calculated WT structure these three resgesting that there is no significant structural change. idues are grouped together. They are conserved throughHowever, in addition to those NOEs discussed above, out the EGF-like family and their mutations exhibit there are small differences in constraints throughout the decreased activity. The second domain involves the molecule, presumably as a result of inherent experimen- C-terminal end of the protein, Leu48 in particular. Mutal error (i.e., inconsistency of sample and spectrome- tation to Ala, Ile, or Met results in dramatic loss of ter conditions; errors in calculating volume integrals). activity (21, 24). The similarity of the chemical shifts Molecular mechanics calculations were used to deter- and NOE spectra of the WT and L48A proteins is mine if the change in constraints ( W T j L 4 8 A ) taken as strongly suggestive that the mutation of Leu48 has lita whole would have any aggregate effect on the calcu- tle effect on TGF-ct structure. Molecular mechanics calculations provided no evidence of any conformalated WT structures. The 22 revised WT structures were modified at res- tional changes, although as the region 48-50 is not idue 48 and then used as initial structures for refine- well-defined, a change in the local microenvironment

FIGURE 2 Fingerprint region of NOESY spectra (T,,~ = 150 nis) for .A) rvild-type (WT) and B)Leu48-.Ala (L48A) mutant proteins of TGF-r. Crosspeaks associated with residues 47 and 48 in A are enclosed in a box. The same region in B differs due to changes in chemical shift.

114

Transforming growth factor

3

B

FIGURE 3 Stereoview of Cys34-Ala50 backbone of TGF-a structures calculated by A) AMBER molecular dynamics and B) XPLOR simulated annealing. Light tracing - Four WT structures. Dark tracing - Same four structures after refinement with L48A NOE constraints. Position of residue 48 is indicated by arrow.

cannot be ruled out. The great flexibility of Leu48-Ala50 in solution suggests that shape of the Leu48 sidechain is the most important factor in the protein-receptor relationship, rather than local backbone conformation. Residues Asp47 and Tyr38 may also play some functional role in this domain. Two other residues in the vicinity of Leu48 may have mixed structural/functional roles. The calculated structures suggest that Asp47 is likely involved with the short &sheet in the C-terminal domain and thus may be important for proper orientation of the Leu48 sidechain at the interface. Due to its proximity to Leu48, it also may directly interact with the receptor. Site-directed mutagenesis studies have produced mixed results (21, 23, 24). The position of Tyr38 within the C-loop suggests that it may be essential for the structure of this domain. However, this residue is also close to Asp47 (4 NOES) and Leu48 could easily be positioned nearby

upon receptor binding. The possible functional roles of Tyr38 and Asp47 remain to be resolved.

ACKNOWLEDGMENTS We thank Drs. R. Reid, J. Feild, and M. Anzano for purifying and assaying the sample. We appreciated the acccss to the AM600 spectrometer in Dr. Wand's lab at Fox Chase Cancer Research Institutc, Philadelphia. Many thanks to Drs. K. Kopple and C. Peishoff for helpful discussions.

REFERENCES I . Delarco J.E. & Todaro, G.J. (1978) Pror. Nutl. Acned. Sci. USA 75, 4001-4005 2. Sporn, M.B. & Todaro, G.J. (1980). N . Engl. J . Med. 303, 878-

880 3. Moses, H.L. & Robinson, R.A. (1982) Feed. Proc. 41,3008-301 1

115

T.P. Kline and L. Mueller 4. Rosenthal, A,. Lindquist, P.B.. Bringman. T.S. Goeddel, D.V. &

Derynek, R. (1986) CeN 46, 301-309 5. Coffey, R.J., Der>nck. R.. Wilcox. J.N.. Bringman. T.S.. Goustin, A.S., Moses, H.L. & Pittelkow. M.R. (1987) .l'crrirre lot id mi/ 328, 817-820 6. Murthy, U., Anzano. M.A. & Grcig. R.G. (1989) I v r . J . Crrrrcer 44, 110-115 7. Burgess, A.W. (1989) Brir. Med. Bull. 45. 401-424 8 . Shoyab, M.. P h v m a n , G.D.. McDonald. V.L.. Bradle?, J.G. & Todaro, G.J. (1989) St-ietre 243, 1074-1076 9. Campbell, I.D.. Baron, M., Cooke, R.M.. Dudgeon. T.J.. Fallon. A., Harvey, T.S. & Tappin, M.J. (1990) Biochern. Plinrm. 40. 35-40 10. Todaro, G.J., Fryling. C. & DeLarco. J . E . (1980) Proc-. .2brl. A m d . Sci. LISA 77, 5258-5262 1 I . Carpenter, G., Stoseheck, C.M., Preston. Y..k & DeLarco. J.E. (1983) froc. h M A c ~ SC;. . L5.4 80. 5627-5630 12. Massague, J. (1983) J . Biol. Chem. 258. 13611-13620 13. Kline, T.P., Brown. F.K., Bro\vn. S.C.. Jeffs. P.\V.. Kopple. K.D. & Mueller. L. (1990) BfochemiJrr:l'29. 7805-7813 14. Kohda, D., Shiniada, I., Mikake. T.. Funs. T. & Inagaki. F. (1989) Biochertiiw), 28, 953-958 15. Montelione, G.T., Winkler. M.E., Burton, L.E.. Rinderknecht. E., Sporn. M.B. & Wagner. G. (1989) Proc. . V d . .4cr2d. Sci.l'S.4 86, 1519-1523 16. Tappin, M.J.. Cooke, R.M.. Fitton, J.E. & Campbell. I.D. (1989) Europerrri J . Biochern. 179, 629-637 17. Campbell, I.D., Cooke, R.M.. Baron. M.. Harye). T.S. & Tappin, M.J. (1989) frog. Gronrh fcicror.7 Res. 1, 13-22 18. BroLvn, S.C.. Mueller. L. & Jeffs. P.W. (1989) Biochemittrj. 28. 593-599 19. Katz, B.A.. Seto. M.. Harkins. R.. Jenson. J.C. & Sykes. B.D. (1989) in 7kechtiiques irt Protein Cherriistrj, (Hugli. T.E..ed.). pp. 223-232, Academic Press, Ne\v York 20. Porchio, A.F., Twardzik. D.R.. Bruce. A.G.. W'izental. L.. Madisen, L., Ranchalis, J.E.. Hu, S.L. & Todaro. G. (1987) G e m 60, 175-182

116

21. Lazar. E.. Watanabe, S., Dalton, S. & Sporn, M.B. (1988) Mol. Cell. Bid. 8, 1247-1252 22. Lazar. E.. Vicenzi, E., Van Ohherghen-Schilling, E., Wolff, B., Dalton. S . . Watanabe. S. & Sporn, M.B. (1989) Mol. Cell. B i d . 9. 860-864 23. Defeo-Jones. D., Tai. J.Y., Wezgrzyn, R.J., Vuoeolo, G.A., Baker. A.E..Payne. L.S., Garsky, V.M., OSff, A. & Riemen, M.W. (1 988) Mol. CeN. B i d . 8. 2999-3007 21. Feild, J.. Reid, R., Kline, T.P., Greig, R. & Anzano, M.A. (1992) Biorhe/vicril J., in press 25. Burgess. A.W., Lloyd, C.J., Smith, S., Stanley, E., Walker, F., Fahri. L.. Sinipson, R.J. & Nice, E.C. (1988) Biochemistry 27, 4977-4985 26. Rance. M., Sorensen, O.W., Bodenhausen, G., Wagner, G., Ernst. R.R. & Wuthrich, J . (1983) Biochem. Biophys. Res. C O / ~ I / ~117. ~ I U479-485 ~. 17. LVurhrich. K . (1 986) N M R .f Proteir1.r arid Nzicleic Acids, WileyInterscience, New York, and references therein 28. Mueller. L. (1987) J . Mqn. Resoti. 72, 191-196 29. Marion. D. & Bax., A . (1988) J . M u g ~ i .Reson. 80, 528-533 30. Weiner, S.J., Kollman. P.A.,Case, D.A., Singh, C.U., Ghio, C., Alagona. G., Profeta. S. & Weiner, P. (1984)J. A m . Chem. Soc. 106.765-784.; Weiner, S.J.. Kollman, P A , Nguyen, D.T. Case, D.A. ( 1 986) J . Cmipur. Chem. 7, 230-252 3 1 . Nilges. M., Clore. G.M. & Gronenborn, A.M. (1986) FEBS Len. 229. 3 17-324

Address: Dr. Tminij. P. Klirie SmithKline Beechani Pharmaceuticals Mail code L-940 P.O. Box 1539 King of Prussia, PA 19406-0939 USA