Salt bridge relay triggers defective LDL receptor binding by a mutant apolipoprotein

Salt bridge relay triggers defective LDL receptor binding by a mutant apolipoprotein Charles Wilson i t, Ted Mau l, Karl H Weisgraber 2, Mark R Wardell 2t, Robert W Mahley 2 and David A Agard' 1 Howard Hughes Medical Institute, Department of Biochemistry and Biophysics, and Graduate Group in Biophysics, University of California, San Francisco, CA 94143-0448, USA and 2 Gladstone Foundation Laboratories for Cardiovascular Disease, Cardiovascular Research Institute, Department of Pathology, University of California, San Francisco, CA 94140, USA Background: Apolipoprotein-E (apo-e), a 34kDa blood plasma protein, plays a key role in directing cholesterol transport via its interaction with the low density lipoprotein (LDL) receptor. The amino-terminal domain of apo-e forms an unusually elongated four-helix bundle arranged such that key basic residues involved in LDL receptor binding form a cluster at the end of one of the helices. A common apo-e variant, apo-e2, corresponding to the single-site substitution Argl58-*Cys, displays minimal LDL receptor binding and is associated with significant changes in plasma cholesterol levels and increased risk of coronary heart disease. Surprisingly, the site of mutation in this variant is physically well removed (>12A) from the cluster of LDL receptor binding residues. Results: We now report the refined crystal structure of the amino-terminal domain of apo-e2, at a nominal resolution of 3.00A. This structure reveals significant conformational changes relative to the wild-type protein that may account for reduced LDL receptor binding. Removal of the Arg158 side chain directly disrupts a pair of salt bridges, causing a compensatory reorganization of salt bridge partners that dramatically alters the charge surface presented by apo-e to its receptor. Conclusions: It is proposed that the observed reorganization of surface salt bridges is responsible for the decreased receptor binding by apo-e2. This reorganization, essentially functioning as a mutationally induced electrostatic switch to turn off receptor binding, represents a novel mechanism for the propagation of conformational changes over significant distances. Structure 15 August 1994, 2:713-718 Key words: apolipoprotein, cholesterol metabolism, electrostatic switch, LDL receptor Introduction The high affinity binding of apolipoprotein-e (apo-e) by cell-surface receptors, including the low density lipoprotein (LDL) receptor, allows lipoproteins associated with apo-e [such as very low density lipoproteins (VLDL), high density lipoproteins (HDL), and chylomicron remnants] to be targeted for endocytosis and intracellular degradation [1]. Interference with such receptor-mediated processing can cause lipoproteins to accumulate in the plasma and can ultimately lead to the formation of atherosclerotic plaques. The apo-e gene is one of the most polymorphic human genes characterized to date and mutations that alter LDL receptor binding are known to have significant effects on cholesterol levels and the risk of coronary artery disease [2]. Apo-E2 is a commonly occurring point mutant of apo- E, initially identified by its altered electrophoretic mobility [3]. Relative to apo-e3 (the wild-type protein), the most common apo-e2 isoform is characterized by the substitution Argl58--4Cys [4]. This mutation (present in approximately 8 % of the population) lowers LDL receptor binding to <2 % of normal levels, although the protein appears to bind to lipoproteins with the same affinity and specificity as the wild-type protein [5]. While plasma cholesterol and LDL concentrations are generally lowered in people expressing the apo-e2 protein (presumably as a consequence of the up-regulation of LDL receptors) [6], a subpopulation of apo-e2 homozygotes are predisposed to type III hyperlipidemia, a lipoprotein disorder associated with premature atherosclerosis [7]. Apo-E appears to contain a 22 kda amino-terminal domain responsible for LDL receptor binding and a 10 kda carboxy-terminal domain involved in lipoprotein binding [8]. Crystallographic studies have shown that the LDL receptor binding domain is arranged as an extremely elongated four-helix bundle (the helical segments extend up to 36 residues or 54A in length) [9]. A cluster of key arginine and lysine residues required for high affinity LDL receptor binding decorate the solvent-accessible face of one of the bundle helices. Several lines of evidence suggest that receptor binding is driven by electrostatic complementarity between this group of positively charged amino acids (spanning residues 136-150) and a set of negatively charged aspartates and glutamates in the short disulfide-rich repeats of the LDL receptor (see [1] for review). Because of its clear role in modulating receptor binding, it had been assumed that Arg158 (the site of the apo-e2 mutation) must be positioned near the other basic amino acids known to be directly involved in binding. It was unclear whether the mutation functioned by interacting directly with some complementary residue on the LDL receptor or *Corresponding author. Present addresses: tdepartment of Molecular Biology, Massachusetts General Hospital, Boston, MA 02114, USA and *Department of Haematology, University of Cambridge, Cambridge, CB2 2QH, UK. Current Biology Ltd ISSN 0969-2126 713

714 Structure 1994, Vol 2 No 8 only indirectly affected receptor binding. Surprisingly, the crystal structure of the LDL receptor binding domain did little to elucidate the mechanism of the E2 mutation. The structure revealed that position 158 is separated from the basic cluster by well over 10A [9]. To understand the structural basis for defective receptor binding by apo-e2, we have crystallized and solved the structure of the amino-terminal domain of the mutant protein. Our results indicate that the apo-e2 amino acid substitution induces a concerted change in salt bridge conformation that is propagated down the length of the helical bundle. This conformational transition induces a major change in the electrostatic potential surrounding the cluster of basic residues, potentially explaining the poor binding of apo-e2 to the LDL receptor. Results and discussion Comparison with the wild-type molecule To understand the structural basis for defective LDL receptor binding by the apo-e2 protein, we have analyzed conformational differences between the wildtype and mutant proteins. Although the resolution of the apo-e2 data is limited, the analysis is based on the use of the well-determined structure of the apo-e3 wild-type fragment (Table 1) which should provide confidence for the observed structural changes. A stereoview of the electron density in a 2Fo-F c map surrounding this region is shown in Fig. 1. As can be seen, the majority of the density is continuous and the model is well positioned. The mutation results in rearrangements of both the backbone and side chain atoms in a large zone around the substituted residue, Arg158 (Fig. 2). For the sake of clarity, we shall consider structural changes in backbone atoms and side chain atoms separately. Direct comparison of the two structures is complicated by two factors. First, the extended loop between helix 2 and helix 3 of the bundle (residues 82-91) is quite poorly defined in the electron density for both the wild-type and E2 structures. As a consequence, its position is relatively unconstrained during simulated annealing refinement. For this reason, conformational differences in this region are not considered to be informative and have been omitted from Fig. 3. Second, shrinkage of the apo-e2 unit cell along the c- axis forces a slight change in crystal contacts which leads to dispersed conformational changes in those side chains that make protein-protein contacts. Fortunately, the zone surrounding the apo-e2 mutation site is solvent-exposed in the crystal and thus local conformational changes are not influenced by contacts with symmetry-related protein molecules. The apo-e2 mutation has no major effect on most of the protein backbone (Figs 2 and 3). The root mean square (rms) deviation between apo-e2 and apo-e3 for the relatively unaffected residues 27-75 and 111-155 is only 0.42A, within the error expected from analysis of the R-factor dependence upon resolution [10]. Table 1. Statistics for data collection and structure refinement. Native apo-e2 Space group P2 1 2 1 2 1 P2 1 2 1 2 1 Cell dimensions (a,b,c, A) 40.7, 54.0, 85.4 41.1, 53.9, 83.9 Resolution (A) 2.25 3.0 Diffraction data Number of unique reflections >2c0 6899 2749 <I>/<Cy> 43.5 32.1 Completeness (10-3.5 A) 100 95.5 Completeness (10-3.0 A) 100 67.9 X-PLOR refinement statistics No. of protein atoms (non-hydrogen) 1172 1167 No. of water molecules 76 0 Rcryst (8-3 A, for all reflections l/l > 2) 0.172 0.195 Rms deviations from ideality Bond lengths (A) 0.017 0.015 Bond angles ( ) 3.2 3.6 Certain localized regions of the backbone, however, including residues 92-96 and 156-163 are significantly distorted (rms deviation 1.06A; Figs 3 and 4). The most severely affected portion of the molecule is the amino-terminal half of helix 3 (residues 92-104). This part of the helix is displaced away from the remainder of the bundle (Fig. 2) as a slight kink near residues 105-108 becomes exaggerated in the mutant protein. Surprisingly, the region thought to be in direct contact with the LDL receptor (136-150) shows essentially no change in backbone configuration, while most of the backbone perturbations are on the other side of the molecule. Salt bridge rearrangement The most striking structural change resulting from the Arg-4Cys mutation is a concerted conformational transition in the side chains directly surrounding position 158. The rms deviation between equivalent atoms within 10A of the mutated residue is 1.6A, significantly higher than that for all atoms (<1.OA). The major effect of the mutation is to disrupt an interconnected network of salt bridges that runs down one face of the four-helix bundle. This network is made up as a linear chain of charged residues, with each amino acid forming bifurcating salt bridges to oppositely charged neighbors on either side of it. The salt bridge path alternately includes residues from the amino-terminal end of helix 3 and the carboxy-terminal end of helix 4, effectively binding them together. In the wild-type protein, the set of salt bridges can be written as: + - + - + Arg92 -* Glu96 - Arg158 "-Aspl 54"-Arg103-Asp1 51 helix: 3 3 4 4 3 4 At the center of this native salt bridge network, the guanidinium group of Argl58 pairs with the side chain carboxylates of both Glu96 and Asp154. Substitution of the positively charged Arg158 by a neutral cysteine in the apo-e2 variant forces its two salt bridge partners

Defective receptor binding by mutant apolipoprotein Wilson et al. 715 Fig. 1. Stereoview of 2Fo-F electron density map in the region around the apo-e2 mutation (Argl 58-Cys). The density is contoured at 1.25a above the mean. Cys158 and Arg150 are shown in red. (Figure produced using the MidasPlus program [17,181.) to seek alternative stable conformations, ultimately disrupting most of the surrounding native salt bridges (Figs 4 and 5). Glu96 swings up and away from residue 158, maintaining a single salt bridge to Arg92. Asp154 moves down towards the previously unpaired Arg150. Of the five salt bridges indicated above for the wildtype protein, only one (Arg92E(-Glu96) remains intact in the apo-e2 structure. As a result, the helical interface is no longer spanned and stabilized by salt bridges (with the exception of the pairings farther away from the mutation site which are retained in the mutant structure: Asp107<--Arg147--Asp10 0). In fact, it is most likely that the observed alterations in local backbone conformation are also a direct result of this dramatic salt bridge rearrangement. Fig. 2. Comparison of the protein backbones for the LDL receptor binding domain of wild-type (apo-e3, white) and mutant (apo-e2, cyan) amino-terminal 22 kda fragments. The apo-e2 mutation (Argl58-Cys) is shown in red and the putative receptor binding residues (136-150) are highlighted in yellow. The structures were superimposed by a leastsquares fit using the backbone atoms of residues 27-48, 64-75 and 138-1 53 and have an rms deviation of only 0.35 A. The loop between helix 2 and helix 3 (82-91) is poorly determined in both structures and, for apo-e2, the connection is shown as a discontinuous line to make this more obvious. The distorted amino-terminal end of helix 3 is found in the upper right corner. (Figure produced using the MidasPlus program 117,181.) While lacking a structure of the protein in complex with the LDL receptor, we can speculate on the mechanisms by which receptor binding is disrupted for the E2 mutant. Site-directed mutagenesis has shown that a number of basic residues in the region 136-150 are required for full receptor binding [11,12]. Replacement of one of these key residues, Arg150, by alanine reduces LDL receptor binding to one-quarter of normal values [13]. In the apo-e3 structure, Arg150 is solventexposed, presumably accessible for interaction with complementary acidic residues on the LDL receptor. In the E2 structure, however, this residue has moved out of the highly positive region of helix 4 and is paired with Asp154. Arg150 thus serves as the replacement salt bridge partner for Asp154 upon mutation of Arg158. The net result of this rearrangement is a significant change in the electrostatic potential surrounding the receptor-binding region of helix 4 (Fig. 6). Thus, while the charged side chain of Argl58 does not contribute directly to the positive region surrounding the

716 Structure 1994, Vol 2 No 8 <i: C 0 2.0-1.5- > 1.0 _ 0.5 UA. I I I I I 20 40 60 80 100 residue number 120 140 160 ISO Fig. 3. Rms deviation between wildtype and apo-e2 structures for backbone atoms (N, Ca, C) plotted against residue number. The region of the disordered loop between helices 2 and 3 (residues 82-91) has been omitted for clarity. Fi. 4. Stereoview of the wild-type (white) and apo-e2 (cyan) structures in the zone surrounding the mutated residue 158. (Figure produced using the MidasPlus program 117,18].) receptor-binding helix, its presence is required to maintain the important Arg150 in an unpaired, receptor-accessible conformation. Previous chemical modification experiments with apo-e2 are consistent with the hypothesis that a saltbridge-mediated conformational change accounts for altered receptor binding [14]. The thiol-specific reagent cysteamine reacts with Cys158 to form a lysine analog which is capable of forming a single salt bridge. This chemical modification raises LDL receptor binding to approximately 10% that of the wild-type protein. The fact that LDL receptor binding is not completely restored may be explained by the observation that the modified cysteine residue is incapable of forming bifurcating salt bridges such as those formed by the wild-type arginine residue. As such, the native salt bridge network can only be partially restored. While the E2 mutation has had little or no effect on main chain conformation in the LDL receptor binding region of apo-e, alterations of side chain conformations have been dramatic. The spatially distant substitution of cysteine for arginine at 158 has caused an extended network of salt bridges to be disrupted. As a consequence, the structural alteration that began at 158 spreads in a 'domino effect' to encompass key side chains in the receptor binding region. The net result is a significant alteration in the electrostatic field and sidechain conformation in the LDL receptor binding segment. These changes would seem sufficient to account for the reduced binding of the apo-e2 mutant.

Defective receptor binding by mutant apolipoprotein Wilson et al. 717 Fig. 5. Comparison of the salt bridge networks in the apo-e2 (left) and wildtype apo-e3 (right) structures. Salt bridges dashed lines) are indicated together with the distance between charged atoms (in A). (Figure produced using the MidasPlus program 117,181.) Fig. 6. Electrostatic potential map calculated for (a) the wild-type protein and (b) the apo-e2 mutant protein. The DELPHI program (Biosym, San Diego, CA) was used to calculate an approximate solution to the linearized Poisson-Boltzmann equation. The protein and solvent dielectrics were set to 2 and 80, respectively. The ionic strength was set to 150mM (mimicking blood plasma). Only formal protein charges were included in the calculation. Positive (cyan) and negative (red) contours in the potential are evaluated at +2 and -2 kt/e-, respectively. The observed conformational change suggests that the receptor may directly interact with only a subset of those residues that appear to be important for binding, while the remainder may be required to maintain the integrity of the native salt-bridge network. Biological implications Apolipoprotein-E (apo-e) is a major component in most of the lipoprotein classes including chylomicrons, very low density, low density and high density lipoproteins (VLDL, LDL and HDL). Functioning as a specific, high-affinity ligand for cell-surface receptors, apo-e plays a fundamental role in mammalian lipid and cholesterol metabolism. There are three major apo-e isoforms (apo-e2, apo-e3, and apo-e4 [3]) which result from single amino acid mutations within the structural gene [4]. The most common form, apo-e3, is considered normal for LDL receptor binding and particle specificity. The commonly occurring apo-e2 isoform (Argl58-4Cys) retains <2 % of normal binding to the LDL receptor but shows the same specificity as the wild-type protein [5]. In people homozygous for the apo-e2 allele

718 Structure 1994, Vol 2 No 8 (~1 % of the population), remnant lipoproteins accumulate in the plasma, a condition known as dysbetalipoproteinemia [15]. As a consequence of environmental or genetic factors, a subpopulation of apo-e2 homozygotes will develop type III hyperlipidemia which is characterized by severely elevated cholesterol and triglyceride levels and premature atherosclerosis [7]. The structural data described here provide a molecular model for the decrease in LDL receptor binding activity associated with the apo-e2 variant. The Argl58-4Cys mutation disrupts a salt bridge network, thereby dramatically altering the surface charge presented by apo-e to its receptor. In the future, it may be possible to design drugs directed against type III hyperlipoproteinemia by engineering compounds that bind specifically to the Cys158 site of apo- E2 and introduce positively charged groups capable of fully restoring the native salt bridges. Materials and methods Protein production and crystallization The 22kDa thrombolytic fragment of apo-e2 was isolated as described previously using blood plasma from a single human donor [16]. Crystals of the E2 mutant were obtained via vapor diffusion by the hanging drop method. Conditions developed for the native apo-e3 protein [9] [using 15% polyethylene glycol 400 (BDH), 20mM sodium acetate, ph 5.3, 0.2% A-n-octylglucopyranoside (Calbiochem), and 0.1% A3-mercaptoethanol] also yielded crystals of the E2 variant. Unfortunately, for several reasons including limited protein availability, it was only possible to produce two quite small crystals. Both wild-type and mutant proteins crystallized in the P2 1 2 1 2 1 space group with nearly identical a and b unit cell dimensions (see Table 1). The c- dimension of the E2 mutant (c = 83.9 A) is somewhat smaller than that of the wild-type protein (c=85.4a). Data collection Data were collected using a Rigaku automated four circle diffractometer (AFC5R), equipped with an MSC cryo-cooling device. Wild-type and mutant apo-e crystals are extremely sensitive to radiation damage. To minimize radiation-induced decay, crystals were quick-frozen in a stream of boiling liquid nitrogen and all data were collected at, or below, -150 0 C. Crystals of the apo-e2 variant were significantly smaller than those obtained for the wild-type protein (smaller than 0.1 mm in each dimension), and usable data could be only collected from a single crystal to a nominal resolution of 3.0A (as opposed to 2.25A for the wildtype apo-e3). As is typical of diffractometry, only a single asymmetric unit of data was collected. Lack of sufficient crystals precluded the extension of the data to higher resolution. Molecular replacement and refinement The 2.25A refined structure of the wild-type 22kDa fragment provided starting phases and coordinates for the structural analysis of the apo-e2 mutant. Following an initial rigid-body minimization to compensate for displacement along the shrunken c-axis, several cycles of simulated annealing molecular dynamics, B-factor and position refinement, and manual rebuilding of the structure were performed. Data collection and refinement statistics are reported in Table 1. The final R-factor for the apo-e2 variant structure is 19.5% (8-3.0A data). With the exception of the large disordered loop between helices 2 and 3 (residues 82-91) and glycine residues, the 4,-s backbone conformational angles were all within or very close to allowed limits. Thus, despite the rather low resolution of the structural analysis, the final structure appears to be reasonably well refined. Rigid-body and fiull-atom refinement were performed using X- PLOR, version 2.0 [17]. The coordinates have been deposited with the Brookhaven Protein Data Bank as entry 1LE2. Acknowledgements: We thank J Newdoll and the UCSF Computer Graphics Laboratory (supported by NIH RR-1081) for help in preparing figures. This work was supported by the Howard Hughes Medical Institute (CW, TM, DAA), the Fannie and John Hertz Foundation (CW) and NIH Program Project Grant HL41633 (KHW,MRW,RWM). References 1. Mahley, R.W. (1988). Apolipoprotein E: cholesterol transport protein with expanding role in cell biology. Science 240, 622-630. 2. Davignon, J., Gregg, R.E. & Sing, C.F. (1988). Apolipoproteinr E polymorphism and atherosclerosis. Arteriosclerosis 8, 1-21. 3. Utermann, G., Langenbeck, U., Beisigen, U. & Weber, W. (1980). Genetics of the apolipoprotein E system in man. Am. J. Hum. Genet. 32, 339-347. 4. Rail, S.C., Jr., Weisgraber, K.H. & Mahley, R.W. (1982). Human apolipoprotein E. The complete amino acid sequence. J. Biol. Chem. 257, 4171-4178. 5. Weisgraber, K.H., Innerarity, T.L. & Mahley, R.W. (1982). Abnormal lipoprotein receptor-binding activity of the human E apoprotein due to cysteine-arginine interchange at a single site. J. Biol. Chem. 257, 2518-2521. 6. Sing, C.F. & Davignon, J. (1985). Role of the apolipoprotein E polymorphism in determining normal plasma lipid and lipoprotein variation. Am. J. Hum. Genet. 37, 268-285. 7. Mahley, R.W. & Rail, S.C. (1989). Type III hyperlipoproteinemia: the role of apolipoprotein E in normal and abnormal lipoprotein metabolism. In The Metabolic Basis of Inherited Disease. (6th Edn, Scriver, C.R., Beauclat, A.L., Sly, W.S. & Volle, D., eds), Vol.1, pp. 1195-1213, McGraw Hill, New York. 8. Wetterau, J.R., Aggerbeck, L.P., Rail, S.C. & Weisgraber, K.H. (1988). Human apolipoprotein E3 in aqueous solution.. Evidence for two structural domains. J. Biol. Chem. 263, 6240-6248. 9. Wilson, C., Wardell, M.R., Weisgraber, K.H., Mahley, R.W. & Agard, D.A. (1991). Three-dimensional structure of the LDL receptor-binding domain of human apolipoprotein E. Science 252, 1817-1822. 10. Luzzati, V. (1952). Traitement statistique des erreurs dans la determination des structures cristallines. Acta Crystallogr. 5, 802-810. 11. Innerarity, T.L., Friedlander, E.J., Rail, S.C., Jr., Weisgraber, K.H. & Mahley, R.W. (1983). The receptor-binding domain of human apolipoprotein E. Binding of apolipoprotein E fragments. J. Biol. Chem. 258,12341-12347. 12. Weisgraber, K.H, et al., & Sparrow, J.T. (1983). The receptor-binding domain of human apolipoprotein E. Monoclonal antibody inhibition of binding.. Biol. Chem. 258, 12348-12354. 13. Lalazar, A., et al., & Vogel, T. (1988). Site-specific mutagenesis of human apolipoprotein E. Receptor binding activity of variants with single amino acid substitutions.. Biol. Chem. 263, 3542-3545. 14. Innerarity, T.L., Weisgraber, K.H., Arnold, K.S., Rail, S.C. & Mahley, R.W. (1984). Normalization of receptor binding of apolipoprotein E2. Evidence for modulation of the binding site conformation.. Biol. Chem. 259, 7261-7267. 15. Utermann, G., Hees, M. & Steinmetz, A. (1977). Polymorphism of apolipoprotein E and the occurrence of dysbetalipoproteinemia in man. Nature 269, 604-607. 16. Aggerbeck, L.P., Wetterau, J.R., Weisgraber, K.H., Mahley, R.W. & Agard, D.A. (1988). Crystallization and preliminary X-ray diffraction studies on the amino-terminal (receptor-binding) domain of human apolipoprotein E3 from serum very low density lipoproteins. J. Mo. Biol. 202, 179-181. 17. BrOnger, A.T., Kuriyan, J. & Karplus, M. (1987). Crystallographic R factor refinement by molecular dynamics. Science 235, 458-460. 18. Ferrin, T.E., Huang, C.C., Jarvis, L.E. & Langridge, R. (1988). The MIDAS display system. J. Mol. Graphics 6,13-27. 19. Huang, C.C., Petterson, E.F., Klein, T.E., & Langridge, R. (1991). Conic: a fast renderer for space-filling molecules with shadows. J. Mol. Graphics 9, 230-236. Received: 13 Apr 1994; revisions requested: 5 May 1994; revisions received: 8 Jun 1994. Accepted: 15 Jun 1994.