Mutating aspartate in the calcium-binding site of α-lactalbumin: effects on the protein stability and cation binding

Transcription

1 Protein Engineering vol.14 no.10 pp , 2001 Mutating aspartate in the calcium-binding site of α-lactalbumin: effects on the protein stability and cation binding Sergei E.Permyakov 1, Vladimir N.Uversky 1,2, Dmitry B.Veprintsev 1, Alexandra M.Cherskaya 1, Charles L.Brooks 3, Eugene A.Permyakov 1 and Lawrence J.Berliner 4,5,6 1 Institute for Biological Instrumentation of the Russian Academy of Sciences, Pushchino, Moscow region , Russia, 2 Department of Chemistry and Biochemistry, University of California, Santa Cruz, CA 95064, 3 Departments of Veterinary Biosciences and Biochemistry and 4 Departments of Chemistry and Molecular and Cellular Biochemistry, The Ohio State University, Columbus, OH 43210, USA 5 Present address: Department of Chemistry and Biochemistry, University of Denver, Denver, CO 80208, USA 6 To whom correspondence should be addressed. berliner@du.edu The residue Asp87, which is in the calcium-binding loop of bovine α-lactalbumin (α-la) and provides a side-chain carboxylate oxygen for ligand Ca(II) co-ordination, was substituted by either alanine or asparagine. The physical properties and calcium-binding affinities were monitored by intrinsic fluorescence and circular dichroism spectroscopy. D87A α-la displayed a total loss of rigid tertiary structure, a dramatic loss in secondary structure and negligible calcium affinity [Anderson et al. (1997) Biochemistry, 36, ]. On the contrary, D87N α-la displayed native-like secondary structure with a somewhat destabilized tertiary structure. When the well-documented N-terminal methionine was enzymatically removed from D87N α-la [Veprintsev et al. (1999) Proteins: Struct. Funct. Genet., 37, 65 72], the structure appeared to more closely resemble native α-la. Remarkably, the thermal transition mid-temperature of apo-desmetd87n α-la was ~31 C versus native apo- α-la (~25 C), probably due to negative charge compensation in the calcium co-ordination site. On the other hand, the transition mid-temperature of Ca(II)-bound desmetd87n α-la was ~57 C versus native α-la (~66 C), which was related to a decreased Ca(II) affinity (K ~ versus ~ /M at 40 C, respectively). These results reaffirm that alanine substitution in site specific mutagenesis is not always a prudent choice. Substitutions must be conservative with only minimal changes in functional groups and side-chain volume. Keywords: α-lactalbumin/calcium binding/fluorescence/ thermal stability Introduction α-lactalbumin (α-la) is a small (M r ), globular, Ca(II)- binding milk protein (for a review, see Permyakov and Berliner, 2000). In the lactating mammary gland α-la is involved in catalysis of the final step of lactose biosynthesis as a modifier of galactosyltransferase (E.C ) specificity (Hill and Brew, 1975). The protein consists of two structural domains (Pike et al., 1996): a large α-helical domain and a small β-sheet domain, which are positioned on either side of a Ca(II)-binding loop, comprised of residues The loop is closed by a disulfide bridge 73 91, which also connects both domains. The Ca(II)-binding domain of α-la represents a structure of helix loop helix type, different from typical EF-hand (Stuart et al., 1986) due to a shortened binding loop (10 residues instead of 12). The calcium ion in the X-ray structure is co-ordinated by seven oxygen ligands which form a distorted pentagonal bipyramidal structure. Calcium coordination involves the carboxylic oxygens of Asp residues 82, 87, 88, carbonyl oxygens of Lys79 and Asp84, and one or two water molecules. A secondary Ca(II)-binding site, located in the vicinity of the primary site, was revealed by X-ray crystallography in human α-la (Chandra et al., 1998). It is formed by side-chain oxygens of Thr38, Gln39, Asp83 and the carbonyl oxygen of Leu81, arranged in tetrahedral structure. Although the process of Ca(II) binding by α-la is well characterized both structurally and quantitatively (Permyakov et al., 1981; Griko and Remeta, 1999), the role of the process in lactose synthesis is still far from clear (Permyakov and Berliner, 2000). Nevertheless, the binding of Ca(II) to α-la is necessary for correct folding of the protein (Rao and Brew, 1989), which emphasizes a significance of the Ca(II)-binding loop of α-la for its folding. In order to evaluate contributions of each of the Ca(II)-coordinating carboxylic charges into the process of the Ca(II) binding by bovine α-la, we previously prepared a set of α-la mutants with ligating Asp residues substituted by Ala (Anderson et al., 1997). Measurements of Ca(II) affinity of these mutants revealed that D87A and D88A were unable to bind calcium ions, which we interpreted to mean that residues Asp87 and Asp88 are required for effective Ca(II) binding. Both mutants exhibited a total loss of tertiary structure with a partial loss of secondary structure. The structural rearrangement could be caused by transition to a molten globule state or by incorrect folding of the mutant α-las, caused by their decreased calcium affinity as predicted by the work of Rao and Brew (Rao and Brew, 1989). In any case, the observed distortion of α-la structure, induced by alanine substitution, makes accurate measurement of calcium affinity difficult and prevents the delineation of the calcium site determinants. In order to minimize structural distortion in mutant bovine α-las, we substituted Asp87 with Asn. Asparagine lacks the non-compensated negative charge of a carboxylate group, yet has the same side-chain volume and geometry. Studies of the physico-chemical properties of the resulting D87N α-la by means of intrinsic protein fluorescence and circular dichroism (CD) spectroscopy allowed us to measure both the electrostatic and geometric contributions of the side chain for calcium binding. Materials and methods Materials Bovine α-lactalbumin (lot 128F-8140) and Aeromonas proteolitica aminopeptidase (E.C , lot 27H4050), N-(2- Oxford University Press 785

2 S.E.Permyakov et al. hydroxyethyl)piperazine-n -(2-ethanesulfonic acid) (HEPES) and ethylene glycol-bis(β-aminoethyl ether)-n,n,n,n tetraacetic acid (EGTA) were purchased from Sigma Chemical Co. (St Louis, MO). All other chemicals were reagent grade or better. Solutions were prepared from double-distilled, demineralized water. Protein concentrations were evaluated spectrophotometrically using an extinction coefficient, E 1% 280 nm 20.1 (Kronman and Andreotti, 1964). Preparation of recombinant α-lactalbumins Recombinant proteins D87A and D87Nα-LA were expressed in E.coli as previously described (Anderson et al., 1997; Peterson et al., 1999). As a consequence of protein expression the recombinant protein contained an extra methionine residue at the N-terminus, which is known to destabilize α-la structure, thermal stability and calcium affinity (Ishikawa et al., 1998; Chaudhuri et al., 1999; Veprintsev et al., 1999). In order to eliminate this perturbation the N-terminal Met was removed by A.proteolitica aminopeptidase enzyme (Wilkes et al., 1973; Prescott and Wilkes, 1976) treatment as described in our earlier work (Veprintsev et al., 1999). The efficacy and precision of this procedure were confirmed by mass spectrometry and CD. Fluorescence and CD spectroscopy Fluorescence measurements were performed on a Perkin-Elmer LS-50B spectrofluorimeter. All spectra were corrected and fitted to log-normal curves (Burstein and Emelyanenko, 1996) using non-linear regression analysis (Marquardt, 1963). The maximum positions of the spectra were obtained from the fitted data. Protein fluorescence was excited at 280 nm. Temperature scans were performed stepwise, allowing the sample to equilibrate at each temperature for at least 5 min. Temperature was monitored directly inside the cell. The fraction of conversion from the native to the thermally denatured state was calculated from the plots of the temperature dependence of emission intensity at a fixed wavelength as previously described (Permyakov and Burstein, 1984). Calcium-binding affinity was measured using a spectrofluorimetric CaCl 2 /EGTAtitration method described earlier (Permyakov et al., 1981). The accuracy of the determination of the calcium-binding constants was approximately half an order of magnitude. Estimation of parameters of theoretical models was carried out according to Marquardt (Marquardt, 1963). Protein concentrations were between 2 and 6 µm. CD measurements were performed on a AVIV 62DS or Jasco J-500A spectropolarimeter. Typical instrument conditions were: scan rate, 5 nm/min; time constant, 8 s. The pathlength was 0.1 mm for the far-uv studies and 5 mm for the near- UV region. Protein concentrations were in the 25 to 100 µm range. Results and discussion Comparison of D87N with D87A α-la Figure 1A and B depict the near- and far-cd spectra, respectively, of apo- and Ca(II)-saturated native α-la, D87A α-la, D87N α-la, and des-met D87N α-la at 5 C (10 mm HEPES or Tris, ph 8.0). Figure 1A demonstrates the expected recovery of tertiary structure in D87N upon the enzymatic removal of the N-terminal methionine yielding desmetd87n α-la. This was the case for both apo- and Ca(II)-saturated desmet D87N α-la. The far-uv CD spectra (Figure 1B) of all the D87N mutants differ slightly from that of the native protein, suggesting a Fig. 1. Near-UV CD (A) and far-uv CD (B) spectra for native, D87A, D87N and desmetd87n α-las in the presence and absence of calcium (1 mm CaCl 2 or 1 mm EGTA, respectively) at 5 C (10 mm HEPES or Tris, ph 8.0). Protein concentrations: µm for D87A, µm for other proteins. retention of secondary structure. The calcium form(s) are slightly more structured than the apo-form as manifested in a decreased 208/222 nm ratio. In contrast, D87A α-la demonstrates a pronounced decrease in secondary structure, which indicates a destabilization of the protein. Despite similar secondary structures among the other species, D87N α-la is characterized by an apparent loss in rigidity in the aromatic amino acid environment, as reflected by a decrease in the near- UV CD, with and without calcium (Figure 1A). Thus, the tertiary structure of D87N α-la seems to be somewhat disordered compared to the native protein. In contrast to D87N α-la, these near-uv CD signals are totally absent for D87A α-la, indicating lack of tertiary structure. It is noteworthy that D87A α-la does not display any temperature dependent conformational transitions as monitored by intrinsic tryptophan fluorescence, CD and calorimetry (Anderson et al., 1997; Veprintsev et al., 1999). Note also that the CD spectra of D87A α-la are essentially identical in the absence and presence of calcium which is consistent with the absence of calcium affinity (Anderson et al., 1997). At the same time, the CD signals for both D87N α-la and desmetd87n α-la (both in the near- and far-uv regions) change with calcium content, confirming that these mutants possess tertiary structure and calcium-binding ability. Thus, as shown earlier (Anderson et al., 1997), the D87A mutation totally eliminates near-uv CD tertiary structure and distorts secondary structure; also this form exhibits a loss of thermal stability and calcium affinity. On the other hand, the isomorphous substitution of Asn for Asp87 does not significantly perturb protein structure. Thus, 786

3 Mutating Asp in the Ca(II)-binding site of α-lactalbumin Fig. 3. Spectrofluorometric Ca(II) titration of desmetd87n α-la at 40 C (10 mm HEPES, ph 8.0). Fluorescence intensity at 350 nm as a function of Ca(II) to protein molar ratio. The solid curve fitted to the experimental points is calculated from a simple one-site binding scheme as previously described (Permyakov et al., 1981). The excitation wavelength was 280 nm. The protein concentration was 2.5 µm. Fig. 2. Thermal denaturation of desmetd87n versus native α-la as followed by intrinsic tryptophan fluorescence (10 mm HEPES, ph 8.0). (A) Emission maximum wavelength, (B) fraction of transition for the apo-forms (1 mm EGTA) and calcium-bound (1 mm CaCl 2 ) forms. Native α-la: apoform (open circle), Ca(II)-bound form (filled circle); desmetd87n α-la: apo-form (open triangle), Ca(II)-bound form (filled triangle). The excitation wavelength was 280 nm. The protein concentration was 2 6 µm. desmet D87N α-la is the best model for studying the effects of neutralizing a charged carboxylate in the Ca(II)-binding loop of α-la. Thermal denaturation of desmetd87n α-la Figure 2A and B compare fluorescence thermal denaturation curves for the apo- and calcium-bound forms of desmetd87n α-la and native α-la, respectively. Note that the fractional conversion data were obtained from changes in fluorescence emission intensity at a fixed wavelength after Permyakov and Burstein (Permyakov and Burstein, 1984). It is clear that apodesmet D87N α-la is ~7 C more stable than native apo-α-la (Figure 2B). In addition, apo-desmetd87n α-la is ~6 nm red-shifted (Figure 2A) compared with the native protein, which indicates a more mobile and/or more polar tryptophan environment. The same conclusions were drawn from the near-uv CD data at 5 C (Figure 1A), where the ellipticity at 270 nm for apo-desmetd87n was reduced by ~32% compared to apo-native α-la, which suggests a partial loss of tertiary structure in desmetd87n α-la. At the same time, the secondary structures of both forms are identical as judged from far-uv CD spectra (Figure 1B). Calcium-loaded desmetd87n α-la at low temperature is characterized by ~8 nm red-shifted tryptophan fluorescence spectrum versus native Ca(II)-loaded α-la, reflecting similar phenomena observed with their apo-forms (Figure 2A). In contrast, Ca(II)-loaded desmetd87n is ~9 C less stable than calcium-saturated native α-la (Figure 2B). Note that a partial destabilization of Ca(II)-loaded desmetd87n α-la resulted in a decrease in the near-uv CD band, while the spectra of both species were superimposable in the far-uv CD (Figure 1A, B). Note that experiments were performed at 1 mm Ca(II) to ensure calcium saturations of all α-la species except the D87A mutant. This comparison suggests an identity in secondary structure concomitant with differences in tertiary structure. In order to reconcile these opposite changes in thermal stability between native and desmetd87n α-las in their two metal-bound forms, it is necessary to consider the role of charge and consequently calcium-binding affinity within the calcium-binding loop. Calcium affinity of desmetd87n α-la The calcium-binding affinity of desmetd87n α-la was measured by direct titration with calcium chloride as monitored by changes in intrinsic tryptophan fluorescence at 40 C (10 mm HEPES, ph 8.0) which is above the thermal transitions for both apo-forms and below the thermal transitions for the Ca(II)-bound forms (see Figure 2). This temperature was chosen because it offered the most sensitive spectral changes between the apo- and calcium-bound forms allowing the most accurate determinations of calcium affinity. The experimental points were fit to a simple one-site binding scheme: K α-la Ca(II) Ca α-la Ca(II) where K Ca is a calcium association constant. The calcium affinity value calculated for desmetd87n α-la from Figure 3 yielded K Ca /M which is to be compared with /M for native α-la (Veprintsev et al., 1999). The free energy of calcium binding to the native protein is, in part, derived from the charge neutralization by the binding of calcium to the negative carboxylate of D87 (Stuart et al., 1986). We assume that removal of this charge in the (desmet)d87n mutant no longer provides this increment of binding free energy. We evaluated whether the decreased calcium-binding constant for desmetd87n α-la is consistent with a loss in enthalpy upon calcium binding to the Asp87 carboxylate, which is a contribution to the total free energy of calcium binding G Ca. The free energy change upon calcium binding is G Ca RT ln K Ca. Therefore, upon mutation, an estimate of the change in G Ca is: δ G Ca RT ln(k D87N / K nat ) µ 11 kj/mol, 787

4 S.E.Permyakov et al. Table I. Spectral and thermodynamic parameters for desmetd87n and native α-la in the presence and absence of Ca(II) Protein Ca(II)/EGTA λ max (nm) θ 270 (deg cm 2 /dmol) t 1/2 ( C) K Ca (40 C) (/M) desmetd87n α-la EGTA Ca(II) Native α-la EGTA Ca(II) λ max, fluorescence spectrum maximum position at 5 C; θ 270, molar ellipticity at 270 nm, 5 C; t 1/2, thermal denaturation half-transition temperature; K Ca (40 C), calcium association constant at 40 C (calculated from fluorescence titration experiments). The conditions were: 10 mm HEPES, ph 8.0, 1 mm CaCl 2 or1mmegta. where K D87N and K nat are the calcium-binding constants for the mutant (desmetd87n) and native α-las, respectively (see Table I). The experimental accuracy in the equilibrium constant is approximately half an order of magnitude; therefore the error in δ G Ca is approximately 2 kj/mol. This 11 kj/mol reduction in binding free energy, compared with G Ca µ 54 kj/mol for the native protein, corresponds roughly to an equal contribution to the five to six oxygen co-ordinating sites in the calcium-binding loop (Stuart et al., 1986). Assuming that the free energies of either apo-native or apodesmetd87n α-las were the same at 40 C, since both species were thermally denatured by intrinsic fluorescence criteria, the difference in the thermal transitions of their respective calciumbound states is consistent with thermodynamic calculations using the T 0 and C p values taken from differential scanning calorimetry data for α-la at ph 8.0 (Griko and Remeta, 1999). In contrast to the results with D87A α-la, where our CD measurements revealed the total disorganization of tertiary structure, a major loss of secondary structure, the inability to bind calcium and a loss of a co-operative thermal transition (Anderson et al., 1997), desmet D87N α-la retains each of these structural and functional features. The failure of D87A α-la to fold into a functional form, as reflected in negligible calcium affinity, demonstrates the necessity of calcium binding for correct folding and disulfide bond formation as pointed out earlier by Rao and Brew (Rao and Brew, 1989) and Chandra et al. (Chandra et al., 1998). An unexpected property of desmetd87n α-la was a substantial increase in the thermal stability of the apo-form (~7 C), despite a moderate decrease in tertiary structural rigidity (~32% decrease in molar ellipticity at 270 nm). This observation is consistent with the differential scanning calorimetry studies of Griko and Remeta (Griko and Remeta, 1999), who suggested that the stability of apo-α-la was significantly reduced due to negative charge charge interactions in the calcium-binding site. Hence, virtually any substitution of one of these residues by a non-anionic side chain should increase the stability of the apo-form. In summary, the residue change from Asp to Asn was absolutely minimal from both packing volume and ligand coordination considerations. The Asn side chain still allows for calcium co-ordination via the same carbonyl oxygen that exists in the native protein. The only significant changes were the removal of a negative charge and the potential loss of an alternative oxygen atom for ligand co-ordination. Since the metal-free native protein is characterized by five charged carboxylates (Asp 82, -83, -84, -87 and -88) in the calciumbinding loop region, three of which directly co-ordinate the cation and one of which is not that distant), it is clear that electrostatic repulsion is a key factor in folding and stabilization 788 of the apo- and calcium-bound forms, respectively. This was supported by recent X-ray structures of bovine α-la which clearly displayed a retreat of Asp87 from the other liganding side chains, as well as the loss of the two calcium co-ordinating waters in the apo-form (Chrysina et al., 2000). Consequently, electrostatic repulsion would appear to be a major driving force in the movement of Asp87 away from the vacant calcium co-ordinating site the apo-protein and the loss of the two key structural water molecules, initially noted in the X-ray structure of human α-la (Acharya et al., 1991), that bridges key Asp residues in the calcium-binding loop. One water molecule forms potential hydrogen bonds to Asp82 and Asp87. The second water bridges Asp84 and Asp87. Hence, the charge compensation afforded by either the Asp to Asn mutation in the apo-form or by Ca(II) binding in the native protein would appear to predict an increase in thermal stability. In addition, we cannot rule out the contribution of additional hydrogen bonding donor capacity resulting from the introduction of an Asn residue, which could potentially contribute additional stability. Furthermore, overall protein stability is also related to the stability of the unfolded state which this latter mutation could potentially influence. The consideration of side-chain volume is also potentially important when choosing amino acid substitution in protein engineering. The respective average volume of buried residues for the amino acids under consideration here are: Ala, 92 Å 3 ; Asp, 125 Å 3 ; Asn 125 Å 3 (Chothia, 1975). Initially it might seem difficult to assess quantitatively the degree to which a 26% relative void in packing volume between Ala and Asn impacts on protein structure and stability at a single site. However, residue volume cannot be overlooked as a real contribution to the remarkable differences in stability observed between the Ala and Asn mutants, since D87A α-la was unable either to fold properly or bind calcium. Furthermore, desmet D87N α-la resembles most other structural and functional features of the native protein, while eliminating one repulsive negative charge, yet retaining the ability to pack and hydrogen bond with other residues in the vicinity. These results reaffirm that alanine substitution in site specific mutagenesis is not always a prudent choice for gaining a mechanistic understanding of protein function. Substitutions must be conservative with only minimal changes in functional groups and side-chain volume. Acknowledgements We are indebted to Drs Luybov A.Wasserman and Patricia J.Anderson for providing some of the initial data for this study. This work was supported in part by grants from the Russian Foundation for Basic Research (No ) and U.S.P.H.S. (No. GM 56970). References Acharya,K.R, Ren J., Stuart,D.I., Phillips D.C. and Fenna,R.E. (1991) J. Mol. Biol., 221,

5 Mutating Asp in the Ca(II)-binding site of α-lactalbumin Anderson,P.J., Brooks,C.L. and Berliner,L.J. (1997) Biochemistry, 36, Burstein,E.A. and Emelyanenko,V.I. (1996) Photochem. Photobiol., 64, Chandra,N., Brew,K. and Acharya,K.R. (1998) Biochemistry, 37, Chaudhuri,T.K. et al. (1999) J. Mol. Biol., 285, Chothia,C. (1975) Nature, 254, Chrysina,E.D., Brew,K. and Acharya,K.R., (2000) J. Biol. Chem., 275, Griko,Y.V. and Remeta,D.P. (1999) Protein Sci., 8, Hill,R.L. and Brew,K. (1975) Adv. Enzymol. Relat. Areas Mol. Biol., 43, Ishikawa,N., Chiba,T., Chen,L.T., Shimizu,A., Ikeguchi,M. and Sugai,S. (1998) Protein Eng., 11, Kronman,M.J. and Andreotti,R.E. (1964) Biochemistry, 3, Marquardt,D.W. (1963) J. Soc. Ind. Appl. Math., 11, Permyakov,E.A. and Burstein,E.A. (1984) Biophys. Chem., 19, Permyakov,E.A. and Berliner,L.J. (2000) FEBS Lett., 473, Permyakov,E.A., Yarmolenko,V.V., Kalinichenko,L.P., Morozova,L.A. and Burstein,E.A. (1981) Biochem. Biophys. Res. Commun., 100, Peterson,F.C. Anderson,P.J., Berliner,L.J., Brooks,C.L. (1999) Protein Express Purif., 15, Pike,A.C., Brew,K. and Acharya,K.R. (1996) Structure, 4, Prescott,J.M. and Wilkes,S.H. (1976) Methods Enzymol., 45, Rao,K.R. and Brew,K. (1989) Biochem. Biophys. Res. Commun., 163, Stuart,D.I., Acharya,K.R., Walker,N.P., Smith,S.G., Lewis,M. and Phillips,D.C. (1986) Nature, 324, Veprintsev,D.B., Narayan,M., Permyakov,S.E., Uversky,V.N., Brooks,C.L., Cherskaya,A.M., Permyakov,E.A. and Berliner,L.J. (1999) Proteins: Struct. Funct. Genet., 37, Wilkes,S.H., Bayliss,M.E. and Prescott,J.M. (1973) Eur. J. Biochem., 34, Received February 13, 2001; revised June 28, 2001; accepted July 16,