Crystallization of trimeric recombinant human tumor necrosis factor ...

Communication

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 263, No. 26, Issue of September 15, pp. 12816-12819,1988 0 1988 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

Crystallization of Trimeric Recombinant Human Tumor Necrosis Factor (Cachectin)"

The sequence of mature TNF is 28% identical to that of human lymphotoxin, a relatedlymphokine that interactswith the same class of receptors (8). TNF has been isolated, purified, and characterized (3). The matureprotein (157 residues, M , = 17,000) is derived by (Received for publication, March 8, 1988) cleavage of the 76 amino-terminal residues from the prohorMichael J. Eck$#,Bruce Beutlerll, George Kuoll, mone, contains two cysteines believed to form an intrachain James P. Merryweatherll, and disulfide bond, and is not glycosylated. Under native condiStephen R. Sprang$** tions TNF has been reported to exist in oligomeric states with From the $Departmentof Biochemistry, llThe Howard molecular weights ranging from 34,000 to 140,000. Recent Hughes Medical Institute, The Universityof Texas reports, however, indicate that the hormone is active as a Southwestern Medical Center, Dallas, Texas 75235-9050 trimer (4). and 11 Chiron Research Laboratories, In this report we describe the conditions for growth of Emeryuille, California94608 crystals suitable for x-ray crystallographic structure determiCrystals of tumor necrosis factor (TNF) have been nation and describe experiments that confirm the trimeric obtained intwo forms. Rhombohedralcrystals grow in quaternary structure of TNF in solution. 1.8 to 2.0 M ammonium sulfite, pH 7.8 at 21 "C, and MATERIALS ANDMETHODS tetragonal crystals grow in 2.6 M magnesium sulfate, TNF was produced from a synthetic construct, which was expressed pH 5.5 at 25 "C. Analysis of TNFby isoelectric focusing under native and denaturing conditions indicates as an intracellular protein in yeast. The protein was purified as that TNF molecules exist as trimers in solution. The described previously (5) and maintained in storage buffer (0.02 M rhombohedral cachectin crystals belong to space group Tris, pH 8.0, 0.15 M NaC1) at a concentration of 20 mg/ml. Analysis SDS-PAGE and isoelectric focusing were performed on a PharR3 and have unit cell constants a = b = c = 47.65 A by macia PhastSystemTMunit using standard separation and developand a = = y = 88.1'. Density determinations and the ment techniques. PhastGelTM gradient 10-15%gelswere used for space group indicate that the unit cell contains one SDS-PAGE, and IEF 5-8 gels were used for the IEF separations. 51,000-dalton trimer. Thesecrystals are stable in the Two-dimensional, one dimension denatured isoelectric focusing x-ray beamand diffract to at least 1.85 A butare (ODD-IEF) was run with the first dimension under native conditions on a standard IEF 5-8 gel. The sample lane (approximately 5 mm apparentlytwinned by merohedry.Thetetragonal crystals are space group P4&2 or its enantiomorph wide) was then cut out, inverted,and applied across the cathodal end P41212and have unitcell constants a = b = 95.08, c = of the second dimension gel, a PhastGel IEF 3-9, which had been 117.49. The asymmetric unit contains one trimer; the soaked for 20 min in 5% Pharmalyte 3-10,6 M urea, and 0.1% Triton The gel was then run under standard conditions for an IEF crystals are stable in the x-ray beam and diffract to X-100. 3-9 gel and developed as above. beyond 3 A. Chromatofocusing was performed on a Pharmacia LKB Biotech-

Tumor necrosis factor (cachectin) is a metabolically active protein secreted by macrophages in response to certain invasive stimuli, particularly in response to challenge with endotoxin. It is believed to be an important mediator of many infection-related phenomena including induction of fever, shock, and cachexia (see Ref. 1 for review). TNF' is present at elevated levels in the serum of Mycobacterium bouis strain BCG-infected mice treated with endotoxin, and passive immunization of mice against TNF canprevent endotoxininduced death (2). Further, the hormone has been demonstrated to have specific cytolytic effects against certain tumorigenic cells. TNF has been demonstrated to bind with high affinity to an as yet uncharacterized cellular receptor.

nology Inc. FPLCTMsystem using the Mono-PTMHR 5/20 column pre-equilibrated with 25 mM BisTris, pH 6.7. TNF was applied in the equilibration buffer at a concentration of 0.75 mg/ml and eluted at a rate of 0.5 ml/min with a solution containing 10 ml of Polybuffer 74TM/100ml of H20, pH5.0. Buffer exchange and protein concentration were performed in a Centricon-loTM concentrationdevice. Rhombohedral crystals weregrown a t 22 "C in hanging drops equilibrated by vapor diffusion. Typically, 5 pl of the proteinin storage buffer was combined on a siliconized coverslip with an equal volume of a reservoir solution containing 200 mM BES and 1.8-2.0 M ammonium sulfite at pH7.8. The coverslip was then inverted over 1 ml of the reservoir solution contained in the well of a Linbro cell cultureplate and sealed with vacuum grease. Macroseeding was performed by introducing a thoroughly washed crystal (0.1 mm on edge) into a 15-pl hanging drop containing 5 p1 of 3 M BES buffered ammonium sulfite and 10 pl of protein concentrated to 33 mg/ml. Tetragonal crystalswere grownas above but with a reservoir solution containing 200 mM MES and2.4-2.7 M magnesium sulfate a t pH 5.5. Crystal density determinations were performed ina Ficoll 400 gradient essentially as described by Westbrook (6). Partial specific volume wascalculated (7) from the amino acid sequence (8). Crystals were mounted in thin-walled glass capillary tubes, and precession photographs were taken using CuK, radiation from a Rigaku Rotaflex rotating anode generator at 50 kV and 108 mA. Data to3 8, from the rhombohedral crystals and 4.5 A from the tetragonal crystals were collected by 0.75 "C w peak scans a t 0.5 deg/min and 15-5 background counts at room temperature with a Rigaku Rotaflex generator operating at 5.4 kilowatts on a Rigaku AFC-5R diffractometer.

* 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. 5 Supported by National Institutes of Health Medical Scientist Training Program Grant T32 GM08014. ** To whom reprint requests should be addressed. 'The abbreviations used are: TNF, tumor necrosis factor; IEF, isoelectric focusing; BisTris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)-propane-1,3-diol; SDS, sodium dodecyl sulfate, PAGE, RESULTSANDDISCUSSION polyacrylamide gel electrophoresis; ODD-IEF, one dimension denatured isoelectric focusing; BES, N,N-bis(2-hydroxyethyl)-2-amino- Characterization-Purified TNF migrates as a single sharp band with M , = 17,000 (Fig. 1, lane A ) on SDS-polyacrylamide ethanesulfonic acid; MES, 2-(N-morpholino)ethanesulfonicacid. 12816

Crystallization of TNF A 0

C

D E G F p H l s,

I II 111 IV

TNFa

NATIVE-IEF

-.

12817

I

-a 5.82

6.55

SDS-PAGE

FIG. 1. Electrophoretic analysis of TNF. A, SDS-PAGE of the TNF fraction used for crystallization reveals a single sharp band with M , = 17,000. B, washed redissolved crystals also contain primarily the 17-kDa species. C, Pharmacia LKB Biotechnology Inc. low molecular weight calibration markers. D, isoelectric focusing of TNF under native conditions reveals a series of four band groups, each containing from one to four sub-bands with PIS between 6.7 and 5.8. The band groups are uniformly separated, consistent with groups of species differing by integral electrical charge. E , band group IV TNF (isolated by chromatofocusing) becomes extremely heterogeneous after prolonged storage at 4 "C. F, G, IEF calibration markers: plactoglobulin A PI = 5.2; bovine carbonic anhydrase B PI = 5.85; human carbonic anhydrase BPI = 6.55. H, washed redissolved crystals contain primarily band group I protein, which was only a minor component of the TNF fraction used to set up thecrystallization. I , the presence of 6 M urea (favoring dissociation of oligomers) in isoelectric focusing gels greatly simplifies the banding pattern of unfractionated TNF; the complex pattern of lane D is reduced to three major bands: a, b, and c. Further analysis (Fig. 2) has shown that the formation of trimers from species a, b, and c accounts for the complex banding pattern seen in lane D.

4-

NATIVE

B C cbbabb bbb

FIG.2. Two-dimensional ODD-IEF indicates TNF is a trimer. A , in two-dimensional ODD-IEF, performed as described under "Materials and Methods," oligomers separated by PI in the native first dimension dissociate into their component monomers in the denaturing second dimension. The two-dimensional gel therefore gels. However, isoelectric focusing under native conditions reveals a series of four banding groups each containing from shows the monomer content of various oligomeric species. B shows the orientation in which the native IEF gel was applied to thesecond one tofour sub-bands withPI values between 6.7 and 5.8 (Fig. dimension (denaturing IEF) gel. Species a, b, and c have the same 1, lune D).The bandgroups are uniformly separated, consist- PIS asthe respective bands in Fig. 1, lane I. The four bands of group ent with groups of species differing by integral or constant IV dissociate in the second dimension into bands a and b. Band electrical charge. In the hope that a more homogeneous pro- groups I1 and I11 both dissociate into a,b, and c, but group I1 contains tein sample would improve crystal size and quality, we at- proportionately more species c than group 111. Band group I contains only species c. As discussed in the text, this pattern suggests that the tempted to isolate single bands by chromatofocusing. The molecule exists as a trimer under native conditions. Proposed monelution profile of the Mono-P column contained four peaks omer composition of the major bands in the native gel are shown in with no discernible substructure which were shown by IEF to C.

correspond to groups I-IV. An in vitro T N F cytotoxicity assay (9) demonstrated thatall fractions were equally active. Fractions containing pure band group IV TNF were pooled and concentrated to 20 mg/ml. The source of charge heterogeneity has not been determined. SDS-PAGE analysis of band groups I-IV separated by chromatofocusing shows no evidence of degradation of the 17-kDa monomer. Amino-terminalpeptide sequences obtained by automated Edman degradation (13) for both unfractionated TNF and bandgroup IV TNF indicate the lass of the NH2-terminal valine residue (8) from roughly 50% of the molecules? Loss of the amino-terminal valine residue is unlikely to generate significant change in the PI of the subunit. Elman analysis (10) of the unfractionated protein indicated that no reduced thiols were present, thus ruling out heterogeneity in theoxidation state of cysteine residues. It is possible that deamidation of one or more of the 16 asparagine or glutamine residues is responsible for the heterogeneity of recombinant TNF. Isoelectric focusing of TNF in the presence of 6 M urea, which would favor dissociation of oligomers, reveals three major bands (Fig. 1, lune I) along with other minor components. Deamidation is also consistent with the observation that band group IV protein, the least negatively charged group of species isolated by chromatofocusing, be-

' C . Slaughter, M. J. Eck, and S. R. Sprang, unpublished data.

comes markedly heterogeneous (Fig. 1, lune E ) after storage in sample buffer at 4 "C for 2 months (as opposed to the unfractionated sample, which was maintained at -70 "C). The seven band groups visible in this lane are similar in subbanding pattern and bandseparation to thefour band groups in the fresh TNF. Isoelectric focusing under denaturing conditions reveals additional monomeric species which are either absent or present only as minor contaminants in fresh T N F samples (data notshown). The complex banding pattern produced by fresh purified TNF (or samples stored at -70 "C)in IEF gels run under native conditions (Fig. 1, lune D) is proposed to arise from trimers containing different combinationsof the threemajor (and several minor) monomeric species. Substructure within individual band groups is expected because two or more combinations could yield trimeric species with similar PI values. The trimeric structure of TNF is deduced from two-dimensional ODD-IEF gels (Fig. 2) described above by identifying the monomeric species that comprise the individual band groups in the nativeIEF gel. The observed pattern of banding is consistent with a trimeric, but nota dimeric, molecule. The four constituents of band group IV migrate in the native dimension (Fig. 2) with a mean PI of 6.6. These dissociate in the denaturingdimension into two monomeric species, a and

12818

Crystallization of TNF

+:

A

FIG. 3. Crystallization and preliminary diffraction analysis. A, screened zero layerprecession photograph of rhombohedral crystals, hkO zone, rhombohedral setting, p = 15', F = 75 mm. Note pseudo-mirror symmetry about [hhO]axis due to twinning B, rhombohedral crystals of TNF grow to 0.1 mm on edge in a few days; C, screened zero layer precession photograph of tetragonal crystals, h01 zone, p = 12",F = 75 mm.

b, with closely spaced PI values which fall within the range of group IV oligomers. The combination of two monomer species into four unique oligomeric species is expected for trimers; the four bands on the native dimension correspond to the four possible combinations of the two monomers (aaa, aab, abb, bbb). Dimerization of the a and b monomers, on the other hand, would produce only three bands inthe native gel. Likely monomer compositions are indicated for the remaining major bands in Fig.2. Our results are in agreement with recently reported results from neutron-scattering experiments which indicate that TNFis a trimer insolution (11)and with cross-linking studies that indicate that TNF is active as a trimer (4). Crystallization-We have obtained crystals of cachectin in two space groups. Rhombohedral crystals (Fig. 3B) are produced when ammonium sulfite is used as a precipitant while tetragonal crystals have been grown from magnesium sulfate. Rhombohedral crystals were obtained with the unfractionated protein in ammonium sulfite at concentrations ranging from 1.8 to 2.0 M at a pH of approximately 7.8 a t 21 "C. The crystals are highly birefringent and rhombohedral in shape. Crystals first appear after 36 h and grow to 0.1 mm on edge in a few days. Rhombohedra as large as 0.22 mm on edge have been grown by seeding with the 0.1-mm crystals. Crystallization attempts with ammonium sulfate have been entirely 10 mM ammonium unsuccessful, even intrialscontaining sulfite. Also, attempts with ammonium sulfate and either pmercaptoethanol or dithiothreitol as a reducing agent have been unproductive. SDS-PAGE (Fig. 1,lune B ) and IEF(Fig. 1, lane H ) of washed, redissolved rhombohedral crystals indicate that thelattice contains primarily band group I protein, which is present in barely detectable quantities inthe original crystallization solution. This may explain our lack of success with the purified band group IV TNF; crystallization attempts with this fraction resulted in showers of microcrystals. Analysis of precession photographs and diffractometer data indicate a rhombohedral unit cell with a = b = c = 47.65 A, and a = 3/ = y = 88.1", space group R3. By using measured density (1.22 g/cc) and calculated partial specific volume (0.74 cc/g), the number of molecules per unit cell was calculated to be 2.55, not inconsistentwith a trimer in the unit cell, considering possible experimental error in density measurements. For a trimer, the computed Matthew's coefficient, V,, is 2.12 A/dalton, well within the range of reported values (12). The reciprocal lattice shows approximate RTM symmetry (consistent with space group R32); however, the extent of 2-fold

symmetry about the a* and b* axes varies from crystal to crystal, suggesting that the pseudo-symmetry is due to twinning by merohedry (14). R32 symmetry implies six molecules per unit cell, resulting in a? unreasonable value for the packing density ( V , of 1.08 A3/dalton). Twinning fractions (15) estimated from data setsfor selected crystals range from 35 to 45%. Despite the observed twinning, the crystals are well formed and highly birefringent. Measurements from 15min still photographs and preliminary data collection trials indicate that thecrystals diffract to atleast 1.85 A, are stable to a monochromatized x-ray beam, and show a loss of only 10% in intensity of three low order reflections after 70 h of irradiation at room temperature. All attempts to grow untwinned crystals in this space group, including alteration of pH, precipitant, and crystal growth rate, have been unsuccessful. Tetragonal crystalsof TNF were obtained with the unfractionated protein in2.6 M magnesium sulfate at pH5.5. Crystal growth isquite temperature-dependent; the largest single crystals are grown when the crystallization plate is kept at approximately 20 "C for the first 48 h and then shifted to 25 "C. Crystals typically appear after 72 h and grow to 0.2 X 0.2 X 0.8 mm in several days. Analysis of precession photographsand diffractometer dataindicate that the crystals belong to space group P 4 ~ 2 ~ or2 its enantiqmorphP41212 aad have unit cell dimensions a = b = 95.08 A and c = 117.5 A. The crystals diffract to atleast 3.0 A and arestable in the xray beam. A search for heavy atom derivatiyes is in progress. The computed packing density ( V , = 2.6 A/dalton) is consistent with the presence of three monomers in theasymmetric unit or24 monomers per unit tell. A self-rotation function (16) computed with a native 4.5 A resolution data set reveals a local 3-fold axis of symmetry parallel to thea* and b* axes, indicating the presence of trimers of TNF in the asymmetric unit. The two crystal forms of TNF reported here and those reported by other groups (11,17) all contain multiples of three TNF molecules in the unit cell, suggesting that TNF is a trimer in thecrystalline state aswell as in solution. Acknowledgments-We wish to thank Clive Slaughter for the execution and interpretation of amino-terminal sequencing experiments, Marvin Hackert for his kindness in providing his facilities for our initial diffraction experiments, and Elizabeth Goldsmith for helpful discussions. REFERENCES 1. Beutler, B., and Cerami, A. (1986) Nature 320,584-588 2. Beutler, B., Milsark, I. W., and Cerami, A. C. (1985) Science 229,869-871

Crystallization of TNF 3. Agganval, B. B., Kohr, W. J., Hass, P. E., Moffat, B., Spencer, S. A., Henzel, W. J., Bringman, T. S., Nedwin, G. E., Goeddel, D. V., and Harkins, R. N. (1985) J. Biol. Chem. 2 6 0 , 2345-2354 4. Smith, R. A., and Baglioni, C. (1987) J. Biol. Chem. 2 6 2 , 69516954 5. Tracey, K. J., Beutler, B., Lowry, S. F., Merryweather, J., Wolpe, S., Milsark, I., Hariri, R. J., Fahey, T. J., 111, Zentella, A., Albert, J. D., Shires, T. J., and Cerami, A. (1986) Science 2 3 4 , 470-474 6. Westbrook, E. M. (1985) Methods Enzymol. 114,187-196 7. Cohn, E. J., and Edsall, J. T. (1953) The Proteins: Chemistry, Biological Actiuity, and Methods, pp. 370-377, Academic Press, New York 8. Pennica, D., Nedwin, G.E., Hayflick, J. S., Seeburg, P. H., Derynck, R., Palladino, M. A,, Kohr, W. J., Agganval, B. B., and Goeddel, D. V. (1984) Nature 312,724-729

12819

9. Mossman, T. J. (1983) Zmmuml. Methods 65,55-63 10. Riddles, P. W., Blakeley, R.L., and Zerner, B. (1979) Anal. Biochem. 9 4 , 75-81 11. Lewit-Bentley, A., Fourme, R., Kahn, R., Prange, T., Vachette, P., Travernier, J., Hauquier, G., Fiers, W. (1988) J. Mol. Biol. 199,389-392 12. Matthews, B. W. (1968) J. Mol. Biol. 3 3 , 491 13. Hewick, R. M., Hunkapiller, M. W., Hood, L. E., and Dreyer, W. J. (1981) J. Biol. Chem. 256,7990-7997 14. Buerger, M. J. (1960) Crystal-Structure Analysis, John Wiley & Sons, New York 15. Fisher, R. G., and Sweet, R. M. (1980) Acta Cryst. A36,755-760 16. Rossman, M. G. (ed) (1972) The Molecular Replacement Method, Gordon & Breach, New York 17. Hakoshima, T., and Tomita, K-I. (1988) J. Mol. Biol. 201,455457