Int. J. Peptidehotein Res. 12, 1978, 38-41 Published by Munksgaard, Copenhagen, Denmark N o part may be reproduced by any process without written permission from the author(s)

T I G H T PACKING O F PROTEIN CORES AND INTERFACES

Relation to Conservative Amino Acid Sequences and Stability Of Protein-Protein Interaction

JAKE BELL0

Dept. of Biophysics, Roswell Park Memorial Institute, Buffalo, New York, U.S.A.

Received 1 2 December 1977, accepted for publication 6 February 1978

The tightly packed protein interiors and interfaces are taken to be essentially solids. The tight packing, and the r-6 dependence o f the energy of van der Waals’ interactions may account for the highly conservative core regions of homologous proteins. The hydrophobic free energies used to calculate con formational stability and the association constants for protein-protein interactions are not adequate, since the free energies are obtained from liquid-liquid transfer of model compounds. A n additional term is required, the enthalpy o f fusion. This provides an additional 7 kcal mol-’ for the stabilization of the trypsin-trypsin inhibitor complex.

Key words: hydrophobic; stability; protein; packing; evolution; association; interfaces.

It has been observed that in some regions of proteins there is a very high degree of conser- vation of amino acid sequence among homo- logous proteins of different species. These regions generally have high proportions of apolar side chains. These conserved regions are in addition to the crucial residues in active or binding sites. A few examples illustrate the phe- nomenon. Lenstra er al. (1977) compared the sequences in 24 homologous ribonucleases (23 of pancreatic origin and one of bovine seminal fluid origin); they observed that the “residues which form hydrophobic contacts or shield them, are constant in all ribonucleases or only conservatively replaced”. Similarly in the globin homologs (Perutz et al., 1965; Ptitsyn, 1975) and the cytochromes c (Dickerson et al., 1971) the hydrophobic core residues are in- variant or are replaced by other hydrophobes which fit equally well (Welling et d., 1975). It

38

appears that the integrity of the hydrophobic core will not endure a wide variety of amino acid substitutions. Much attention is being directed to the fit between protein interfaces and the free energy of stabilization of protein- protein complexes. The stability of cores and of interfaces is here proposed to arise, in part, from the extra van der Waals’ interactions in solids over that in liquids.

DISCUSSION

The evidence from numerous X-ray structure analyses, and from other studies, shows pro- teins to be tightly packed (Klapper, 1973; Richards, 1974; Chothia, 1975). In a previous report (Bello, 1977) we noted that the tightly packed, solid-like nature of the protein interior provides the clue for eliminating a discrepancy between the fact that native proteins are folded, compact structures and the fact that the

PROTEIN CORES AND INTERFACES

sum of the thermodynamic factors based on model systems (hydrogen-bonding, hydrophobic effects, entropy of unfolding, etc.) predicts that proteins should be unfolded (Tanford, 1970). A missing term in the calculations was the en- thalpy of fusion, AHt, of the model com- pounds. Since the model systems are liquid and proteins are solid, it is not correct to base the hydrophobic AG on transfer of model com- pounds from water to organic liquid, i.e., liquid-to-liquid, but on transfer from water to organic solid.

There are numerous indicators of fluctu- ations in protein conformation. (See for example, Cooper, 1976; Woodward, 1977, and references therein.) Cooper has shown that fluctuations are consistent with the overall tightly packed solid-like character of proteins. McCammon et ul. (1977) performed acomputer simulation of the dynamics of a protein and found a “range of conformations in the neigh- borhood of the X-ray structure, but not iden- tical with it”. They stated that aspects of internal motion suggest “that the folded pro- tein is fluid-like at ordinary temperatures”, and that the fluctuations “although confined to the neighborhood of the average structure” have diffusional character. They noted that the flu- idity is similar to the motions in the disordered solid state of n-alkanes, below the melting point, and that the “fluctuations of neighbor- ing dihedral angles are generally correlated so as to minimize disturbances.” Despite the manifes- tations of fluctuating structure, proteins are as tightly packed as solids, more tightly packed than liquids. The heat capacity of proteins, 0.3 cal.g-’ deg-’ (Hutchens et ul., 1969) is close to that of organic solids similar to protein groups, and distinctly smaller than that of melted polymers (Yoshida et ul., 1970; Wras- idlo, 1972; Griskey & Foster, 1968).

The enthalpy of fusion is a manifestation of the greater van der Waals’ interaction in solids than in liquids. The non-covalent interactions between apolar groups vary in energy inversely with the 6th power of the distance (London, 1937). For example, a 10% increase in distance reduces the energy about SO%, about 200 cal per mole of CH2 or CH3. Therefore, a small degree of misfit will produce a relatively large reduction in noncovalent interaction energy.

Small loosening of the solid-like structure results in a partial approach toward a liquid state, the “oil-drop”.

Amino acid replacements do not result in uniform small distance changes around the whole side chain. The goodness of fit will vary from point to point. Replacement of phenyl- alanine (3.4A thick and 6.8A diameter ben- zene ring), for example, by an aliphatic side chain (4 A diameter of CH2 or CHB) results in a tighter fit in one direction, and a looser fit in other directions; the latter is considerably more than the 10% change we mentioned above. Thus, such a substitution will cause the loss of much of AHf of the benzene ring (about 2 kcal/ mol) and perhaps result in some repulsive de- stabilization, in the absence of compensating changes in conformation.

Experimental evidence summarized by Tan- ford (1970) shows that for many proteins the AG between native and denatured states is about 10 kcal/mol. But smaller values are known. Recent work of Yutani er ul. (1977) on a mutant a-subunit of tryptophan synthetase of Escherichia coli gives AG of about 2-3 kcal/ mol. Thus, for many proteins the misfit of a small number of side chains, even of one, can result in loss of stability of the native confor- mation. Partial denaturation may require a AG of less than 3 kcal/mol. Partial unfolding may result in loss of function or in easy suscepti- bility to enzymic cleavage. These consequences might follow from small changes in interatomic distances.

Destabilization resulting from the small loosening of the structure produced by mu- tations would not be counterbalanced by the stabilizing effect of increased entropy, since AS resulting from small changes in distance (when the increase in available space is in- sufficient to permit rotation from one energy minimum to another) is small compared with the full AS of free rotation. When the structure becomes loose enough to permit free rotation,’ the resulting AS will be stabilizing a non-native state, likely a partially denatured state. Also, with regard to binding of ligands, the r-6 de- pendence of the energy is important (Pauling &- Pressman, 1945); but the TAS difference be- tween a rather good fit and a very good fit is likely to be small in most cases.

39

J . BELL0

When mutation results in a larger side chain, stabilization may occur if the mutant side chain fits more closely. If the side chain is too large to be inserted because of the repulsive energy, there will be destabilization resulting from the unfavorable AG of interaction of the outward- turned side chain with water. Where the out- ward-turned side chain finds itself in a lipid environment the latter factor would not arise, but the van der Waals’ energy of the normal side chain in the protein interior will be lost. The requirement for tight packing may limit the potentially viable sequences. Of the vast number of sequences conceivable for a given length of polypeptide, many will not pack tightly enough to overcome the entropic driving force of unfolding.

In cases where the backbone is well exposed to solvent, such as residues-15-24 of bovine pancreatic ribonuclease (Kartha et al., 1967; Carlisle et al., 1974), the AHf contribution to the native state is small or non-existent.This is the most highly variable part of ribonuclease (Lenstra et al., 1977). In such sequences the survival of mutations will not be significantly affected by the van der Waals’ energies involved in close-packed protein interiors. Other factors will determine the survival value of mutations.

An oil-drop interior could better accommo- date mutations because the energy penalty for a misfit is smaller (larger interatomic distances) and the fluidity could better compensate for changes. But mutations in an oil-drop might result, because of the fluidity, in loss of activity or specificity, with resultant deterioration of the precision of control processes. This might be especially detrimental t o poikilothermic organisms, since a more fluid protein would be more likely to change conformation with tem- perature, rather than substantially maintain its native structure until the cooperative melting temperature is attained. Very hydrophobic pro- teins might have sufficient hydrophobic stabil- ization even with a higher degree of internal fluidity than more typical proteins. A solid-like interior is more likely to result in a go/no-go situation. Since natural selection has chosen solid-like proteins instead of more fluid poly- mers, the go/no-go situation appears to have a selective advantage.

The AHf factor applies to protein surface groups which become part of the interface on protein-protein association. Chothia & Janin (1975) have calculated that interface groups are as closely packed as protein interiors. Wood- ward (1977) has noted that the interface is a good model for the interior. Chothia (1 975) has calculated the hydrophobic AG for folding of proteins as 25 cal/AZ , and the same was applied to interfaces (Chothia & Janin, 1975). This value was obtained from the free energy of transfer of hydrophobic groups from water to organic liquids. For reasons given above, a con- tribution from AHf should be added to the hydrophobic energy estimates of Chothia and of Chothia and Janin. The magnitude of this contribution will depend on the extent t o which the surface groups in the dissociated proteins are in contact with solvent. If they are folded down onto the underlying atoms, to minimize contact with water, we may consider these groups to be partially tightly packed on their under sides (less tightly than interior groups), and that on association of the proteins the new interactions result in an increment of tightly packed volume. Therefore, part of the AHf term is already present in the dissociated subunits, so that only part of AHf is applicable. The appropriate fraction is not obvious. The work of McCammon et al. (1977) on computer simulation of protein dynamics suggests that surface groups are more mobile than interior groups. A further correction must be made for the polar interface atoms. Most polar interface groups form hydrogen bonds in both the dis- sociated state (intramolecularly or with water) and in the associated state (Chothia & Janin. 1975). Since AHf for hydrogen-bonding groups is largely determined by the hydrogen bonding energy (which is much larger than the van der Waals’ energy of polar groups), to a first ap- proximation we can set at zero the AHf con- tribution of polar groups to the interface ener- getics. We now estimate the AHf contribution to two cases analyzed by Chothia & Janin (1975). For trypsin-trypsin inhibitor 1390 A’ of buried interface gives about 25 kcalmol-’ for AHf. This is based on 19A2 per CH2 group (Chothia, 1974) and 25calg-’ x 14gmol-’ of CH2. Since 56% of the interface is non-polar,

40

PROTEIN CORES AND INTERFACES

the estimate is reduced to 14 kcalmol-' . Be- cause the entire AHf of the hydrophobic group is not applicable, we arbitrarily take one-half, or 7kcalmol-' for AHf. The estimates of Chothia & Janin (1975) indicated that 45 kcal mol-' of association energy is required to com- pensate for the translational/rotational free energy (27 kcal mol-') and to provide the 18 kcal mol-' for the observed dissociation con- stant. The hydrophobic contribution was estimated by Chothia and Janin at 35kcal mol-', leaving a deficit of 10kcalmol-I. Our estimate of AHf of 7 kcal mol-' improves the agreement. But for the insulin dimer, for which the calculation of Chothia and Janin gives a deficit of only 2 kcal mol-' , our cor- rection (again 7 kcal mol-') results in too high an estimate. Considering the numerous approxi- mations involved (e.g., the hydrophobic AG, AHf, and the estimate of entropy changes) close agreement is likely to be fortuitous. Nevertheless, attempts to estimate packing energies must take into account the fusion enthalpy.

REFERENCES

Bello, J. (1977) J. Theor. Biol. 67,335-336 Carli.de, C. H., Palmer, R. A., Mazumdar, S. K., Cor-

inski, B. A. & Yeates. D. G. R. (1974) J. Mol.

Chothia. C. (1974) Nature (Lond.) 248, 338-339 Chothia, C. (1975) Nature (Lond.) 254, 304-308 Chothia, C. & Janin, J. (1975) Nature (Lond.) 256,

Cooper, A. (1976) Proc. Natl. Acad. Sci. U.S.A. 73, 2740-2741

Dickerson, R. E., Takano, T., Eisenberg, D., Kallai, 0. B., Samson, L., Cooper, A. & Margoliash, E. (1971) J. Biol. Chem. 246,1511-1535

Bid. 85,l-18

705-708

Griskey, R. G. & Foster, C. N. (1968) Ann. Tech. Conf. SOC. Plastics Engineers, Tech. Papers, 26th SOC. Plastics Eng. Inc., Greenwich, Conn., pp.

Hutchens, J. O., Cole, A. G. & Stout, J. W. (1969) J.

Janin, J. & Chothia, C. (1976) J. Mol. Biol. 100,

Kartha, G., Bello, J. & Harker, D. (1967) Nature

Klapper, M. H. (1973) Progr. Bioorganic Chem. 2,

Lenstra, J. A., Hopteenge, J. & Beintema, J. (1977)

London, F. (1937) Duns. Faraday SOC. 33.8-21 McCammon, J. A., Gelin, B. R. & Karplus, M. (1977)

Pauling, L. & Pressman, D. (1945) J. Am. Chem. SOC.

Perutz, M. F., Kendrew, J. C. & Watson, H. C. (1965)

Ptitsyn,O. B. (1975) J. Mol. Biol, 88,286-300 Richards, F. M. (1974) J. Mol. Biol. 82.1-14 Tanford, C. (1970) Advan. Protein Chem. 24, 1-94 Welling, G. W., Mulder, H. & Beintema, J. I. (1975)

Woodward,C. (1977) J. Mol. Biol. 111,509-515 Wrasidlo, W. (1972) J. Polymer Sci. 10, 1719-1729 Yoshida, S., Sakiyama, M. & Seki, S. (1970) Polymer

Yutani, K., Ogasahara, K., Sugino, Y. & Matsushiro,

546-550

Biol. Chem. 244,26-32

197-211

(Lond.) 213,862-865

55-132

J. Mol. Biol. 109,185-193

Nature (Lond.) 267,585-590

1003 -1012

J. Mol. Bi01. 13,669-678

Biochem. J. 147,505-511

J. 1,573-581

A. (1977) Nature (Lond.) 267,274-275

Address: Jake Bello, Ph.D. Department of Biophysics Roswell Park Memorial Institute 666 Elm Street Buffalo, New York 14263 U.S.A.

41