The three-dimensional structure of TNF has been determined at 0.29 nm using the technique of X-ray crystallography. Published data on site-directed mutagenesis and antibody binding may now be assessed in the light of the structure, thus the links between structure and function for TNF may be addressed.

TNF is a compact trimer composed of three identical subunits of 157 amino acids. The main-chain topology for a single subunit is essentially a β-sandwich structure formed by two anti-parallel β-pleated sheets. This mainchain fold corresponds to the ‘jelly roll’ motif observed in viral coat proteins such as VP1, VP2 and VP3 of rhinovirus, or the hemagglutinin molecule of influenza. TNF is the first non-viral protein to contain this motif. The subunits associate tightly about a threefold axis interacting through a simple edge-to-face packing of the β-sandwich to form the solid, conical shaped trimer. A large number of the residues conserved between the amino acid sequences of TNF and lymphotoxin lie within the β-sandwich or at the threefold axis of the trimer. This implies the presence of the same β-sandwich motif in the lymphotoxin monomer and preservation of the edge-to-face mode of trimeric association.

The detailed three dimensional structure for TNF explains a wide range of observations, including data on antibody binding and site directed mutagenesis. The currently available evidence points to a region of biological importance situated at the interface between two subunits on the lower half of the trimer.

The biological action of the cytokine tumour necrosis factor (TNF) is dependent on its interactions with other protein molecules such as cell surface receptors, inhibitors etc. These interactions are governed by the precise arrangement adopted within three dimensions by the amino acids of the linear polypeptide sequence (primary structure) on formation of the correctly folded tertiary structure. Thus in order to understand how the TNF molecule performs its biological function at the level of amino acid interactions, one must not only know the amino acid sequence but also the three dimensional structure.

The three dimensional structure of human TNF has been determined to a resolution of 0.29 nm by X-ray crystallography (Jones et al. 1989) and has been refined to a current R factor of 20.4% (on all data from 0.6 to 0.29 nm with no model for the solvent structure). We present here a brief summary of the general topology and structural characteristics of the TNF molecule followed by a discussion, in the light of the structure, of data published by other authors on site directed mutagenesis and antibody binding.

The structure

The overall shape of a single 157 amino acid subunit of the TNF trimer is wedge-like with height 5.5 nm and maximum breadth 3.5 nm near the base. The main-chain topology is illustrated in Fig. 1A; it is essentially a β-sandwich structure formed by two antiparallel β -pleated sheets. The mainchain fold conforms to that of the classic jelly roll motif (Richardson, 1981, Fig. IB) common in viral capsid proteins. The nomenclature adopted in Fig. 1 for the labels of the secondary structural units follows the established convention for the viral structures (Rossmann et al. 1983).

Fig. 1.

A. Diagrammatic sketch of the subunit fold. β-strands are shown as thick arrows in the amino to carboxy direction and connecting loops are depicted as thin lines. The disulphide bridge is denoted by a lightening flash and a region of high flexibility is cross-hatched. The trimer threefold axis would be vertical for this orientation. B. The jelly-roll motif. The insertion between β-strands B and C is shown in dashed lines, normally the connection between B and C would run straight across at the top of the molecule.

Fig. 1.

A. Diagrammatic sketch of the subunit fold. β-strands are shown as thick arrows in the amino to carboxy direction and connecting loops are depicted as thin lines. The disulphide bridge is denoted by a lightening flash and a region of high flexibility is cross-hatched. The trimer threefold axis would be vertical for this orientation. B. The jelly-roll motif. The insertion between β-strands B and C is shown in dashed lines, normally the connection between B and C would run straight across at the top of the molecule.

The crystallographic data yield a measure of the relative flexibility of the various parts of the structure. The β-strands form a fairly inflexible scaffold; in particular the back β -sheet is situated at the core of the trimer and consequently is particularly rigid, and the C terminus is embedded in the base of this secondary structural unit. As would be expected it is the loops which adorn the outer, solvent-accessible surface of the molecule which exhibit high levels of flexibility/mobility. The N terminus is highly flexible and as far as residue 10 is rather independent of the rest of the molecule. Overall there is a general decrease in rigidity as the core becomes more loosely packed in the upper half of the molecule.

Three TNF monomeric subunits associate non-covalently to form a compact, conical trimer of length 5.5 nm and maximum breadth 5.0 nm. The β-strands of the three individual β -sandwiches lie approximately parallel to the threefold axis. The interaction between subunits related by the threefold axis is through a simple edge-to-face packing of the β-sandwich; the edge of the β-sandwich consisting of strands F and G from one subunit lies across the back β-sheet (GDIB) of a threefold-related subunit (see Fig. 2). This mode of packing produces an extremely tight association between the subunits. Thus the core of the trimer is completely inaccessible to solvent. Loss of solvent-accessible surface area may be equated to hydrophobic free energy gain: each 0.01 nm2 of accessible surface area removed from contact with water gives a free energy gain of approximately 25 cal mol− 1 (1cal=4.184 Joules) (Chothia, 1974). The solvent accessible surface area buried for a single TNF subunit on formation of the trimer is over 20 nm2, the energetic equivalent of about 30 hydrogen bonds. The stability of the TNF trimer is placed in perspective if one considers that for a typical antibody antigen complex the epitope comprises a solvent-accessible area of some 7 to 8 nm2 (Tulip et al. 1989).

Fig. 2.

Edge-to-face packing of β -sandwiches in the TNF trimer. The view, down the threefold axis, shows a narrow slab of the trimer with β-strands represented by ribbons running into and out of the page.

Fig. 2.

Edge-to-face packing of β -sandwiches in the TNF trimer. The view, down the threefold axis, shows a narrow slab of the trimer with β-strands represented by ribbons running into and out of the page.

The overall distribution of residue types in the three dimensional structure of TNF echoes the general rule for proteins; namely the hydrophobic residues cluster in the core of the molecule whilst charged residues decorate the surface. Thus the core of the TNF β-sandwich has the expected filling of tightly intercalating, large, apolar residues. The detailed distribution of residue types becomes particularly noteworthy in the region near the trimeric threefold axis. At the top of the trimer the interaction between subunits involves charged sidechains, towards the center it is mediated by polar sidechains and at the base by EL patch of large apolars. The threefold interaction at the heart of the trimer is produced by the edge-to-face packing of the aromatic rings from a nest of tyrosine residues (59, 119 and 151) centred on Tyrll9. There is 28% sequence identity between sequences of TNF and lymphotoxin (Pennica et al. 1984). The majority of the residues conserved between the amino acid sequences of TNF and lymphotoxin may be classified, on the basis of the TNF structure, as internal to the β-sandwich or at the trimeric interface; the conservation of the residues which form the surface buried in trimer formation is illustrated in Fig. 3. This conservation in lymphotoxin of residues which play a key structural role in TNF strongly implies the presence of the same β-sandwich motif in the lymphotoxin monomer and preservation of the edge-to-face mode of trimeric association (Jones et al. 1990a).

Fig. 3.

Residues buried at the trimeric interface. The residues for a TNF monomer are represented by spheres, with area proportional to their loss of solvent-accessible area on the association of three such monomers to form a trimer. The view is onto the back βsheet. Residues which are absolutely conserved in all TNF and lymphotoxin sequences are stippled. Residues which maintain a particular class identity (e.g. large apolars) are striped. Variable residues remain blank.

Fig. 3.

Residues buried at the trimeric interface. The residues for a TNF monomer are represented by spheres, with area proportional to their loss of solvent-accessible area on the association of three such monomers to form a trimer. The view is onto the back βsheet. Residues which are absolutely conserved in all TNF and lymphotoxin sequences are stippled. Residues which maintain a particular class identity (e.g. large apolars) are striped. Variable residues remain blank.

Structure/function relationship

For the three-dimensional structure of TNF a set of prominent, surface residues, mainly from the loops connecting the β-strands of the β-sandwich, form the antibody-accessible surface of the trimer (Jones et al. 19906). It has been observed that antibodies raised against TNF from one species (e.g. human) do not crossreact with TNF from another species (e.g. mouse), despite a sequence identity in excess of 80% and the ability of TNF to bind to the TNF receptors of other species (Fiers et al. 1987). However, it is immediately obvious from the structure that the most sequence-variable regions of the molecule correspond to the antibody-accessible surface loops. Thus the epitope for an antibody against TNF will always contain some residues which will vary between species thus abolishing antibody binding. This implies that the characteristics of the interaction between TNF and its receptors must somehow differ from those required for binding of an antibody to TNF.

Neutralising antibodies have so far been mapped in two regions of TNF, amino acids 1–15 (Socher et al. 1987) and an epitope involving Argl31 (Fiers et al. 1990). However, antibodies are rather blunt instruments when employed to probe the properties of the TNF molecule; the relative sizes of the TNF trimer (3×157 residues) and a single antibody Fab fragment (about 400 residues) are comparable. A rather more specific probe of the structure/function relationship is provided by site-directed mutagenesis. Fiers et al. (1990) report more than a thousand fold loss of biological activity on mutating serine 86 to phenylalanine. A double mutation of the two external arginines 31 and 32 to asparagine and threonine by Tsujimoto et al. (1987) also resulted in a dramatic loss of biological activity. Yamamoto et al. (1989) have shown that site-directed mutagenesis of histidine 15 may abolish biological activity; this residue is involved in stabilising the threefold interaction in a region close to residues 31-32 and, in terms of the trimer, adjacent to serine 86. In contrast, mutation of tryptophan residues 28 and 114 to phenylalanines does not markedly affect biological activity (Van Ostade et al. 1988) nor does changing histides 73 and 78 to glutamines (Yamamoto et al. 1989). Finally Narachi et al. (1987) have disrupted the disulphide bridge between cysteines 69 and 101 without deleterious effects. This set of functionally tolerated mutations generally lies in the upper portion of the molecule.

In Fig. 4 the various pieces of evidence are combined in a schematic diagram of the trimer structure. From the results of site-directed mutagenesis a region of functional importance appears to be located at the interface between two subunits on the lower half of the trimer. Both neutralising antibodies could serve to block receptor binding throughout this general region by simple steric hindrance. In addition, Creasey et al. (1987) have reported that truncation of the N terminus by four or seven amino acids actually increases biological activity; this would be consistent with the extremely flexible/mobile N terminus causing a slight steric hindrance effect at a receptor site in the region of residues 15, 31—32 and 86. Thus current circumstantial evidence points to a receptor-binding site which involves the interface between two subunits in this region near the base of the TNF trimer.

Fig. 4.

Putative area of receptor binding. This schematic diagram highlights the relative positions of Ser 86 (circle) and Arg 131 (star) on one subunit, also His 15 (pentagon) and residues Arg 31 and 32 (square) on the neighbouring subunit in the trimer.

Fig. 4.

Putative area of receptor binding. This schematic diagram highlights the relative positions of Ser 86 (circle) and Arg 131 (star) on one subunit, also His 15 (pentagon) and residues Arg 31 and 32 (square) on the neighbouring subunit in the trimer.

Chothia
,
C.
(
1974
).
Hydrophobic bonding and accessible surface area in proteins
.
Nature
248
,
338
339
.
Creasey
,
A. A.
,
Doyle
,
L. V.
,
Reynolds
,
M. T.
,
Jung
,
T.
,
Lin
,
L. S.
and
Vitt
,
C. R.
(
1987
).
Biological effects of recombinant human tumor necrosis factor and its novel muteins on tumor and normal cell lines
.
Cancer Res.
47
,
145
149
.
Fiers
,
W.
,
Beyaert
,
R.
,
Brouckaert
,
P.
,
Everaerdt
,
B.
,
Haegeman
,
G.
,
Libert
,
C.
,
Suffys
,
P.
,
Takahashi
,
N.
,
Tavernier
,
J.
,
Van Bladel
,
S.
,
Vanhaesebroeck
,
B.
,
Van Ostade
,
X.
and
Van Roy
,
F.
(
1990
).
In vitro and in vivo action of tumor necrosis factor
. In
Tumor Necrosis Factor: Structure, Mechanism of Action, Role in Disease and Therapy
(ed.
Bonavida
,
B.
and
Granger
,
G.
), pp.
31
37
.
Karger
,
Basle
.
Fiers
,
W.
,
Brouckaert
,
P.
,
Goldberg
,
A. L.
,
Kettelhut
,
I.
,
Suffys
,
P.
,
Tavernier
,
J.
,
Vanhaesebroeck
,
B.
and
Van Roy
,
F.
(
1987
).
Structure-function relationship of tumour necrosis factor and its mechanism of action
. In
Tumour necrosis factor and related cytotoxins. (Ciba Foundation Symposium 131)
pp.
109
123
.
Wiley
,
Chichester
.
Jones
,
E. Y.
,
Stuart
,
D. I.
and
Walker
,
N. P. C.
(
1989
).
Structure of Tumour Necrosis Factor
.
Nature
338
,
225
228
.
Jones
,
E. Y.
,
Stuart
,
D. I.
and
Walker
,
N. P. C.
(
1990a
).
The three-dimensional structure of TNF at 2.9Â resolution
. In
Tumor Necrosis Factor: Structure, Mechanism of Action, Role in Disease and Therapy
(ed.
Bonavida
,
B.
and
Granger
,
G.
), pp.
31
37
.
Karger
,
Basle
.
Jones
,
E. Y.
,
Stuart
,
D. I.
and
Walker
,
N. P. C.
(
1990b
).
The three-dimensional structure of tumour necrosis factor
. In
Molecular and Cellular Biology of IL-1, TNF and Lipocortins in Inflammation and Differentiation.
(Proceedings of the Sclavo International Conference, ed.
L.
Parente
), pp.
321
327
.
Alan R. Liss
:
New York
.
Narachi
,
M. A.
,
Davis
,
J. M.
,
Hsu
,
Y.-R.
and
Arakawa
,
T.
(
1987
).
Role of single disulfide in recombinant human tumor necrosis factor-a
.
J. biol. Chem.
262
,
13107
13110
.
Pennica
,
D.
,
Nedwin
,
G. E.
,
Hayflick
,
J. S.
,
Seeburg
,
P. S.
,
Derynck
,
R.
,
Palladino
,
M. A.
,
Kohr
,
W. J.
,
Aggarwal
,
B. B.
and
Goeddel
,
D. V.
(
1984
).
Human tumour necrosis factor: precursor structure, expression and homology to lymphotoxin
.
Nature
312
,
724
729
.
Richardson
,
J. S.
(
1981
).
Protein anatomy
:
Adv. Protein Chem.
34
,
167
339
.
Rossmann
,
M. G.
,
Abad-Zapatero
,
C.
,
Murthy
,
M. R. N.
,
Liljas
,
L.
,
Jones
,
T. A.
and
Strandberg
,
B.
(
1983
).
Structural comparisons of some small spherical plant viruses
.
J. molec. Biol.
165
,
711
736
.
Socher
,
S. H.
,
Riemen
,
M. W.
,
Martinez
,
D.
,
Friedman
,
A.
,
Tai
,
J.
,
Quintero
,
J. C.
,
Garsky
,
V.
and
Oliff
,
A.
(
1987
).
Antibodies against amino acids 1-15 of tumor necrosis factor block its binding to cell-surface receptor
.
Proc. natn. Acad. Sci. U.S.A.
84
,
8829
8833
.
Tsujimoto
,
M.
,
Tanaka
,
S.
,
Sakuragawa
,
Y.
,
Tsuruoka
,
N.
,
Funakoshi
,
K.
,
Butsugan
,
T.
,
Nakazato
,
H.
,
Nishihara
,
T.
,
Noguchi
,
T.
and
Vilcek
,
J.
(
1987
).
Comparative studies of the biological activities of human tumor necrosis factor and its derivatives
.
J. Biochem., Tokyo.
101
,
919
925
.
Tulip
,
W. R.
,
Varghese
,
J. N.
,
Webster
,
R. G.
,
Air
,
G. M.
,
Laver
,
W. G.
and
Colman
,
P. M.
(
1989
).
Cold Spring Harbor Symposium on Immune Recognition. In press
.
Van Ostade
,
X.
,
Tavernier
,
J.
and
Fiers
,
W.
(
1988
).
Two conserved tryptophan residues of tumor necrosis factor and lymphotoxin are not involved in the biological activity
.
FEBS Lett.
238
,
347
352
.
Yamamoto
,
R.
,
Wang
,
A.
,
Vitt
,
C. R.
and
Lin
,
L. S.
(
1989
).
Histidine-15: an important role in the cytotoxic activity of human tumor necrosis factor
.
Protein Engineering
2
,
553
558
.