helix metal ion binding sites of calcium-binding proteins fold pairwise .... been assigned tentatively to Ile-58 due to its metal exchange ...... Pitman, London. 27.
THEJOURNAL OF BIOLOGICAL CHEMISTRY 0 1987 by The American Society of Biological Chemists, Inc.
Vol. 262, No. 13, Issue of May 5, pp. 6248-6256 1987 Printed in d.S.A.
Oncomodulin ‘H NMR AND OPTICAL STOPPED-FLOWSPECTROSCOPICSTUDIES CONFORMATION AND METAL-BINDING PROPERTIES *
OF ITS SOLUTION
(Received for publication, October 20,1986)
Thomas C. Williams, David C. Corson, and BrianD. Sykes From the Medical Research Council of Canada Group in Protein Structure and Function and the Department of Bimhemisty, University of Alberta, Edmonton, Alberta, Canudn T6G 2H7
John P.MacManus From the Division of Biological Sciences, National Research Council, Ottawa, Ontario, Canada KIA OR6
As deduced from its ‘HNMR spectrum, oncomodulin’s solution conformation is very similar to the tertiary structure of other single domain 2-site calciumbinding proteins of the troponin C class. Despite its extensive amino acid sequence homology with parvalbumins, however, oncomodulin differssignificantly Ln(II1) exchange from these proteins in its Ca(I1) characteristics. Although the relative affinity of Lu(II1) for the EF site of Ca2-oncomodulinwas normal, BLU:EF/BCn:EF being 175 f 15,displacement of Ca(1I)from the CD site was not favored, BLu:CD/@Ca:CD being 1.2 f 0.1. Lineshape analyses of several ‘H NMR resonances generated by the Lu(II1) titration of Ca2-oncomodulin indicated that Ca(I1)+ Ln(II1) exchange at the CD site was 15-20 s-’, approximately 100 times faster than exchange at the CD site of parvalbumins. Analyses of the distribution of metal-bound oncomodulin species showed that Ca(I1) + Lu(II1) exchange was cooperative, thecoefficient of cooperativity being estimated as 5 f 1. Thekinetics of therelease of Yb(II1) from oncomodulin as measured by optical stopped-flow techniques corroboratedthe observed cooperativity in metal binding; the off-rate constant of Yb(II1) from the EF site of Yb2-oncomodulin was 0.0036 s-’, approximately 19 times slower than therelease of Yb(II1)from the EF siteof CalYbl-oncomodulin. We attribute part of the reduced preference of small Ln(II1)s for the CD site of oncomodulin to a combination of this site’s inherent incompressibility (Williams,T. C., Corson, D. C. & Sykes, B. D. (1984) J. Am. Chern. SOC.106,56985702) and the Glu .-, Asp substitution at sequence position 59, the residue which chelates metal at the -X coordination position. Like theCD site inoncomodulin, site I11 in troponin C hasnot only a lower affinity for calcium relative to the CD site of parvalbumins but also aspartic acid at its -X position; a water molecule bridges thegap between bound metal and the carboxyl group of the relatively short side chain of Asp-114 (Herzberg, 0.& James, M. N. G . (1985)Biochemistry 24, 5298-5302). Hence, we suggest that Asp-59 in oncomodulin binds metal only indirectly through an
-
* This work was supported by the Medical Research Council of Canada Group in Protein Structure andFunction, the Alberta Heritage Foundation for Medical Research (equipment grant for the NT-3OOWB spectrometer, and fellowship and research allowance to T. C. W.), and the National Research Council of Canada. 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 18U.S.C. Section 1734 solelyto indicate this fact.
intervening water molecule, a proposal which is consistent with the CD site’s reduced affinity forions the size of Ca(I1) or smaller. Oncomodulin, a calcium-binding protein extractedboth from oncogenic and fetal tissue (1-5), is classified as a 8lineage parvalbumin based on extensive amino acid sequence homology (6, 7). However, this member of the superfamily of calcium-binding proteins differs markedly from parvalbumins in two respects. First, parvalbumins’ three contiguous helixloop-helix regions, which are folded even in the apo form into a compact highly structured macromolecular chelator, form two high-affinity Mg(II)/Ca(II) binding sites that experience only small conformational changes during metal chelation (8, 9); oncomodulin, however, appears to have only one highaffinity Mg(II)/Ca(II) site and undergoes major conformational changes accompanying chelation of metal at itsCa(I1)specific site (10). Second, parvalbumins do not interact significantly with any other protein; oncomodulin, by contrast, mimics calmodulin in its ability to activate several enzymes (11, 12) and to stimulate DNA synthesis in Ca(I1)-deprived cells (13). The pairing of one high-affinity calciumlmagnesium site with one calcium-specific site within a single domain is extremely unusual. Ordinarily, the 2-4 contiguous helix-loophelix metal ion binding sites of calcium-binding proteins fold pairwise into one or two structurally discrete domains, each site within a given domain having an affinity for Ca(I1) nearly matching the other site in the same domain. Oncomodulin’s apparent ability to incorporate aspects of calcium-buffer proteins and calcium-modulator proteins into a single-domain two-site unithasnot gone without skepticism. Klee and Heppel (14) have shown that oncomodulin fails to stimulate CAMPphosphodiesterase activity; however, Mutus et al. (15) attribute this failure to the inherently lesser sensitivity of bovine brain phosphodiesterase used in their assays. MacManus et al. (10) have suggested that the Ca(I1)-specific region of oncomodulin is its CD’ site. The CD site of The abbreviations used are: CD, the helix C-loop-helix D region of oncomodulin (approximately residues 39-71); EF, the helix E-loophelix F region of oncomodulin (approximately residues 7S108); y., the y methyl group of valine which, if deuterated, confers absolute S configuration to OC; J, scalar spin-spin coupling constant; HDO, hydrogen deuterium oxide; DzO,deuterium oxide; j3”xy, the overall microscopic stability constant of metal x at sitey in a protein when n sites are filled by metal x and 2 - n sites are filled by calcium(I1); re1 pnZ, relative to the overall microscopic stabilityconstant of Ca(I1)a t the same site;Ln(III), thegeneric symbol for lanthanide(II1) ions; klEF, the off-rateconstant of Ln(II1) from the EF site of oncomodulin when the CD site is occupied by Ca(I1); kZEp, the offrate constant of Ln(II1) from the EF site of oncomodulin when the
6248
‘H NMR and Optical Stopped-flow Studies of Oncomodulin
6249
parvalbumins, although quite capable of binding Mg(1I) or the smaller Ln(III)s, is a relatively rigid structure making it a poorer chelator of these ions compared to their affinities for the EF site (9, 16-18). However, Henzl et ul. (19) have shown that Ca(I1) + Ln(II1) exchange in oncomodulin is similar to displacement of Ca(I1) from parvalbumin when the midsize lanthanides ( i e . Tb(II1) and Eu(II1)) are used. Therefore, in order to probe the differences in the metal ion binding properties of oncomodulin’s two sites, we have chosen the closest lanthanide analogues of Mg(II), i.e. Lu(II1) and Yb(III), to monitor Ca(I1) + Ln(II1) exchange. EXPERIMENTAL PROCEDURES
Protein Preparatwn-Oncomodulin was purified from rat hepatomas (Morris 5123tc) as detailed elsewhere (1); its purity was confirmed by comparison of its amino acid composition with its composition as calculated from its sequence (6), its UV absorption spectrum, and its banding pattern following sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Rat parvalbumin was isolated and purified as described elsewhere (20,21). ’H NMR Studies-”H NMR spectraof oncomodulin were recorded on a Nicolet NT-300WB spectrometer operating in thepulsed Fourier transform mode with quadrature detection. Typical acquisition parameters were as follows: spectral width f2000 Hz; pulse width, 8 ps; acquisition time, 2.05 s; free induction decay filter, Bessel (f3000 Hz). Homodecoupling was used to suppress the residual HDO resonance during acquisition. Postacquisition processing included a Lorentzian to Gaussian lineshape transform via a double-exponential multiplication of the free induction decay. The calcium(I1) form of oncomodulin (0.5mM, 150 mM KC1 in D20, pH 6.7,40 ‘C) wastitrated with microliter additions of an EDTA-standardized LuC13 solution as described elsewhere (19). For those ‘H NMR resonances showing changes in slow exchange, lineshapes were evaluated by a curve analysis program which yielded normalized areas of the contributing resonances. For those singlet resonances showing changes in intermediate exchange, lineshapes were evaluated by a two-site exchange program which yielded the fraction of each contributing resonance and the mean lifetime. Calculation of the Lu-oncomodulin binding constants andcoefficient of cooperativity was based on a curve-fitting program derived from distribution of species plots (see “Appendix”). Optical Stopped-flow Studies-The kinetics of the release of ytterbium(II1) from oncomcdulin were measured using a Gibson-Durrum 13000 stopped-flow apparatus interfaced to a Nicolet Instruments Explorer IIIA digital oscilloscope as detailed elsewhere (17, 18). This method entailed measuring the increase in absorbance at 570 nm upon chelation of Yb(II1) by xylenol orange. A Yb(II1)-containing solution of oncomodulin (-10 pM protein, 150 mM KCl, 15 mM Pipes, pH 6.6, 23’C) was mixed with a solution of excess xylenol orange. The absorbance record of the reaction was analyzed using multiple exponential functions to fit the kinetics of formation of the Yb(II1). xylenol orange complex, the rate-limiting step being the release of Yb(II1) from oncomodulin (17,18). The concentration of oncomodulin was determined by UV spectroscopy, using Eikm= 2.7 (22) and amino acid analysis; the concentration of Yb(II1) in the stock solution was determined by standardizationagainst EDTA using xylenol orange as indicator (23). RESULTS AND DISCUSSION
The ‘H NMR Spectrum of Cup-Onconodulin-The 300MHz ‘H NMR spectrum of rat oncomodulin is shown in Fig. 1; it is very similar to spectra obtainedfor other parvalbumins under similar conditions (9,24-28). Most importantly, several spectral regions give clear indications of the presence of secondary and tertiary structuralelements. First, many of the methyl resonances of aliphatic residues (such as leucine, valine, and isoleucine) are shifted significantly upfield from their random-coil positions, 10 or 11 being especially wellresolved in the +0.7 to -0.2 ppm range (Fig. 1, region a). Second, two of three methionine cCH, singlet resonances are CD site is occupied by Ln(II1); PCD, the off-rate constant of Ln(II1) from the CD site of oncomodulin when the EF site is occupied by Ln(II1).
rn
b
a
FIG. 1. The 300”Hz ‘H NMR spectrum of rat Caz-oncomodulin (Morris hepatoma 5123tc) in DzO (0.5 mM, 160 mM KCl, pH 6.68, 45OC). Regions a, b, and d show resonances of residues shifted upfield due to secondary ring-current effects: a, valine, leucine, and isoleucine methyls; b, methionine methyls; and d, phenylalanine ring protons. Regions c and f show resonances of residues in the metal-binding loops and &sheetcontactsshifted downfielddue to the anisotropic effects of nearby carbonyl or carboxyl groups: e, aCH protons; and f, main chain amide NH protons.Region e shows the C2H resonance of His-107 and all other long-lived NH resonances not exchanged by solvation in D20 (approximately 24 h at 45°C at pH 6.7). The resonances a t 1.31 and 0.15 ppm, marked by x, are impurities; chemical shifts are referenced to the temperatureand pH-calibrated impurity resonance at 0.15 ppm relative to sodium 4,4-dimethyl-4-silapentane-l-sulfonate. Relative to the vertical scale of the bottom truce, the vertical scale of the middle truce is X 3 and the vertical scale of the top truce is X 9.
also shifted significantly upfield; the resonance tentatively assigned to Met-105 appears at 1.904 ppm, and the resonance tentatively assigned to Met-86 appears at 1.849 ppm (Fig. 1, region b ; see “Tentative Assignment of ‘H NMR Resonances to Specific Residues” below). Third, a number of aromatic resonances from phenylalanine residues are shifted upfield, including the well-resolved resonances in the 6.8-5.8 ppm range from metu and pura protons of at least 3 rings (Fig. 1, region d ) . Such upfield shifts are related closely to protein conformation. The interaction of aromatic rings with other side chains in a protein is governed primarily by electrostatic interactions of ring edges, which carry a 6+ charge, or ring faces, which carry a 6- charge, with the dipoles or induced dipoles of other side chains, including the methyl groups of valine, leucine, isoleucine, or methionine (29-31). Ring edges and methyl groups, therefore, prefer being positioned near ring faces. Currents generated in aromatic rings by the applied magnetic field produce secondary magnetic moments which shift the resonances of protons on nearby side chains. These
6250
‘H NMR and Opticat Stopped-flow Studies
ring-current effects lead to characteristic upfield shifts observed for ring edges and methyl groups positioned near ring faces (32-34,40). Therefore, oncomodulin (like parvalbumin) has a well-defined tertiarystructure arising from packing aromatic and aliphatic hydrophobes into a central core of interhelical contacts, In addition, several proton resonances are shifted downfield, five aCHs into the 5.7-5.0 ppm range and atleast two amide NHs into the10.2-9.8 ppm range (Fig. 1, regions c and f). Deshielding of main chain proton resonances is attributed to their proximity to magnetically anisotropic carbonyl groups (35), asfound in regions of secondary structure such as@-sheets, reverse turns,oraspartic acid turns (36-39, 41), or electric field effects (42). It is, therefore, highly likely that the short@-sheet structureformed between the loops of paired metal-binding sites in all other calciumbinding proteins is also present in oncomodulin. The appearance of approximately 8 slowly exchanging NH resonances in the 9.0-7.8 ppm region (Fig. 1, region e) also indicates that oncomodulin has other relatively solvent-inaccessible regions of higher order structure (probably a-helices) quite comparable to parvalbumins. Tentative Assignment of ’H NMR Resonances to Specific Residues-Of more than 30 well-resolved resonances in regions a-f in the spectrum of oncomodulin (Fig. l), 12 were assigned tentatively to specific residues in oncomodulin’s primary sequence: 6 to methyl groups, 3 to phenylalanine ring protons, 2 to aCHs, and1to theC2Hof histidine. The doublet at -0.17 ppm was assigned to theys methyl group of Val-106 by comparison to theanalogous resonance in spectraof parvalbumins (9, 18, 24-28). According to the x-ray structure of carp parvalbumin (43,44), Val-106 and Phe-102 form part of the hydrophobic surface of helix F which interacts with Phe30 on the hydrophobic side of helix B; hence, calculations of the ring-current contributionsto thechemical shift were used initially to assign the most upfield-shifted resonance in carp parvalbumin to the ys-CH3 of Val-106 (25). Subsequently, this resonance’s sensitivity to Mg(I1) exchange, Ca(I1) exchange, Ln(II1) exchange, C-terminal digestion with carboxypeptidases A and B, and pH titration, and its appearance as a triplet in the spectrum of the pike I1 isoform (which has isoleucine in position 106)*have been used to corroborate the assignment to Val-106 in other parvalbumins (9, 18, 19, 24). The triplet resonance at 0.51 ppm is uniquely that of an isoleucine’s 6CH3group; not quite so obvious, however, is its specific assignment to one of six isoleucine residues in oncomodulin. In ’H NMR spectra of most parvalbumins, two methyl triplet resonances are shifted upfield into the 0.7-0.0 ppm region (9, 18, 45); the more upfield of these two resonances in the spectrum of rat parvalbumin (at 0.25 ppm) has been assigned tentatively to Ile-58 due to its metal exchange sensitivity (9). The side chains at positions 58 and 97, related by an approximate 2-fold molecular symmetry axis, penetrate the hydrophobic core of parvalbumin from the @-sheetbetween the two metal-binding sites (43,44). Consequently, the resonances of the methyl groups on these side chains are expected to appear upfield due to ring-current shifts. However, position 58 in oncomodulin is occupied by leucine; the triplet at 0.51 ppm is, therefore, assigned tentatively to Ile97. The four singlet methyl resonances in the 2.1-1.8 ppm range were assigned tentatively to specific methionine residues and the N-acetyl group on the basis of their sensitivity to Ca(I1)-+ Ln(II1) exchange (see below). The broadest singlet (1.85 ppm), sensitive to exchange at both sites, was assigned to thecCH3 of Met-86, this residue being located between the
* T. C. Williams andB. D. Sykes, unpublished results.
of
UncomoduZin
CD and EF metal binding loops as surmised from the crystal structure of carp parvalbumin (43, 44); the singlet at 1.90 ppm, sensitive to exchange primarily at the EF site, was assigned to €Met-105,this residue being located near the end of helix F; the singlets at 2.09 ppm and 2.10 ppm, not sensitive to EFsite exchange and only weakly sensitive to exchange at the CD site, were assigned to €Met-38 andthe N-acetyl group, respectively. These assignments are compatible with the presence of a broad Met tCH, singlet resonance at 1.88 ppm assigned tentatively to €Met-86 in the spectrum of rat parvalbumin (9), a narrow Met &Ha singlet at 1.91 ppm assigned to €Met-105in the spectrum of pike I1 parvalbumin,’ and the weak sensitivity of carp parvalbumin’s N-acetyl methyl group to metal exchange at theCD site (18). The threeupfield-shifted triplet resonances in the aromatic region were assigned to two phenylalanine residues. By comparison with the analogous resonances in the spectra of parvalbumins (assigned by correlations to x-ray-based calculations of ring-current shifts (25, 26, 43, 44)), the tripletat 5.81 ppm in thespectrum of Caz-oncomodulinwas assigned to the para proton of Phe-24, the triplet at 6.30 ppm was assigned to the para proton of Phe-29, andthe triplet at 6.47 ppm was assigned to themeta protons of Phe-29. The resonance at 5.09 ppm, assigned to one of the aCHsin the metal-binding loops, must arise from isoleucine, valine, glycine, or threonine due to its unique appearance as a doublet. However, there areno threonines or valines in themetalbinding loops. Furthermore, glycine, although present, would show (uCHresonances having a geminal scalar coupling constant of approximately 15-20 Hz; the resonance at 5.09 ppm has an 11-Hz coupling. Therefore, this resonance was assigned to Ile-97, the only residue in the loops capable of generating such a doublet. This assignment is consistentwith the observation by two-dimensional J-correlated andnuclear Overhauser effect-correlated NMR spectroscopic techniques of two uniquely downfield-shifted aIle protons between 4.8 and 4.7 ppm in thespectrum of rat La2-parvalbumin.3The other four of rat parvalbumin’s six isoleucines are located in a-helices and have aCH resonances in the 3.6-3.4 ppm range, consistent with their expected upfield position relative to the randomcoil chemical shift of 4.22 ppm for the aCH of isoleucine (36, 46). The marked downfield shifts of these resonances are consistent with the x-ray analyses of the crystal structure of carp parvalbumin (43, 44); theaCH of the loop residue between the -Y and -6 coordination positions projects inward and lies within 2.5 A of the 6-carboxyl group of glutamic acid at the-Z coordination site. The poorly resolvedtriplet aCH resonance at 5.69 ppm was assigned to Cys-18 bycomparison to spectra of other &lineage parvalbumins, nearly all of which have a single cysteine residue a t position 18 and theanalogous resonance a t approximately 5.6-5.9 ppm. a-Lineage parvalbumins have neither cysteine nor this analogous resonance. Assignment to cysteine is not inconsistentwith observations that cysteine has among the most downfield of shifted aCH resonances, appearing at 5.8 ppm in ‘H NMR spectra of lysozymes (47) and between 6.1 and 5.1 ppm in spectra of snake venom neurotoxins (48). NH resonances which persist in D20 are usually ascribed to secondary structural elements such as a-helicesor P-sheets in which hydrogen bonding retards their exchange with solvent. Most of the NH resonances present in region e of the spectrum of Ca2-oncomodulin(Fig. 1)were assumed to arise from the six putative a-helical regions due to their relative upfield-shifted positions (36). By the end of Lu(II1) titration, most of these resonances had disappeared, indicating their T. C. Williams and B. D. Sykes, manuscript in preparation.
6251
' H NMR and Optical Stopped-flow Studies of Oncomodulin relative ease of exchange under these experimental conditions (approximately 48 h at 45 "C, pH 6.7). At the midpoint of the putative &sheet formed by the metal-binding loops lie Leu58 and Ile-97.By the very nature of antiparallel&sheet structure, the main chain carbonyls of residues 57 and 96 are turned away from the @-sheet contactsbetween the loops and are directed inward toward the bound metal ions, serving to chelate metal at the -Y coordination position (39, 43). Because NH exchange in amides is believed to occur via an imidic acid intermediate (i.e. 0-protonation rather than N protonation), the carbonyl oxygen of a peptide bond must be solvent accessible in order to facilitate H-+D exchange at its amide nitrogen (49, 50). The apparently solvent-excluded orientation of their carbonyl oxygens and the additional intraresidue hydrogen bonding afforded by the assumed conformation of Tyr-57 and Lys-96 (43, 51) may partially account for the observed retardation of NH exchange attributed to residues 58 and 97.If the hydrogen bonds inthis region retarded sufficiently the exchange of the involved NHs, an aCH/NH scalar coupling would be detectable. Although the analogous Ile-97 doublet resonance at 5.12 ppm in the 'H NMR spectrum of rat Cap-parvalbumin appeared initially as a triplet in a sample not deliberately pre-exchanged in DpO (9), no coupling was observed for the aIle-97 resonance at 5.09 ppm in the spectrum of Ca2-oncomodulin. We suggest, therefore, that of the two most downfield-shifted NH resonances, only one derives from the @-sheet contactsbetween the loops. The singlet resonance at 7.96 ppm (Fig. 1, region e ) was assigned to the CpH of oncomodulin's only histidine, His-107, because of its sensitivity to the slight drop in pH (ApH < -0.5) which accompanied the Lu(II1) titration of unbuffered Ca2-oncomodulin.The chemical shift of the protonated form of His-107 in Lup-oncomodulin (approximately 8.4 pprn), the estimated chemical shift of its basic form (7.5-7.6 ppm (24)), and the chemical shift of the CalLul-form at pH 6.7 (8.15 ppm) were used to estimate the pK, of His-107 as 7.1-7.2. In addition, slight line broadening of this resonance at pH 6.7 indicated that the exchange rate between protonatedand unprotonated forms was similar to thatof His-106 in pike I11 parvalbumin. Oncomodulin's His-107 and pike I11 parvalbumin's His-106 are located at thesame sequence position near the end of helix F. Their nearly identical pK, values (approximately 7.2 for His-107 and 7.1 for His-106 (24)) indicated that their microenvironments may also be similar. Because the pK, of the imidazole ring of free histidine is 0.7-0.8 pH unit lower than thepK, of histidine at theend of helix F, the protonated form of His-107 in oncomodulin (like His-106 of pike I11 parvafbumin) is probably involved as a proton donor in a hydrogen bond (24). Determination of the Relative Affinities of Lutetium(III) for Cap-oncomodulin-The displacement of calcium ions hy lanthanide (Ln) ions in proteins that chelate two metals generates four discrete metal-bound species: one form in which both metal-binding sitesare filled by Ca(II), one form in which both sites are filled by Ln(III), and two nonequivalent forms in which the Ca(I1) of only one site is displaced by Ln(II1). Equilibrium among these four forms is analogous to the equilibrium which describes the protonation of a diprotic acid (52-54). Lutetium(II1) titrationof Ca,-oncomodulin was, therefore, modeled bythe equilibrium illustrated in Scheme I where A is Cap-oncomodulin,B and C are the two nonequivis alent CalLul forms, and D is Lup-oncomodulin, re1 PILu:EF the relative affinity of Lu(II1) for the EF site of Cap-oncomodulin, re1 is the relative affinity of Lu(II1) for the CD site of Ca,-oncomodulin, re1 p2Lu:EF is the relative affinity
D
A
SCHEME I
of Lu(II1) for the EFsite of CalLul-oncomodulin in which the ~ ~ relative affinity CD site is Lu(II1) bound, and re1 @ 2 ~ uis, the of Lu(II1) for the CD site of Ca,Lu,-oncomodulin in which the EF site is Lu(II1) bound? If the sites exchange Lu(II1) independently then re1 BILU:EF = rel
and re1 B'L.:cD
/ 3 2 ~ u , ~ ~
=
re1 B'L.:cD;
however, if Lu(II1) exchange is cooperative then re1 $Lu:EF = c(re1 B 1 ~ , , ~ and ~ ) re1 P 2 ~ , , =c ~c(rd P'L~:CD)
where c is the coefficient of cooperativity in metal binding. Titration of the Ca(I1) form of oncomodulin with Lu(II1) caused changes throughout its 'H NMR spectrum; however, only those resonances which were well resolved during the entire course of the titration were analyzed in terms of exchange between the various Ca(I1) and Lu(II1) forms of the protein. The response of the threemost-upfield methyl proton doublets is illustrated in Fig. 2C. The doublet resonance of the ys methyl of Val-106 (-0.17 ppm, trace a ) showed slow exchange with the doublet at -0.13 ppm (trace d ) during the first phase of Lu(II1) titration (i.e.from a Lu(II1):oncomodulin ratio of 0 + 1); during the second phase of the titration (i.e. from a Lu(II1):oncomodulin ratio of 1 + 3), exchange of the doublet a t -0.13 ppm with the doublet of the Lu2 form at -0.14 ppm was in the intermediate regime as evidenced by the broadening of this resonance as its chemical shift moves progressively upfield (traces e-i).Similarly, the doublet at 0.0 ppm experienced slow exchange with a doublet resonance at +0.01 ppm (trace d ) during the first phase but fastexchange with the doublet of the Lu, form during the second phase, shifting back upfield to 0.0 ppm (trace k). Likewise, the doublet at 0.05 ppm underwent slow exchange with the doublet resonance of the predominant Ca,Lul form of oncomodulin at 0.09 ppm (trace d ) before broadening during intermediate-to-slow exchange with its counterpart in the Lup form, its chemical shift being estimated a t 0.2 ppm from the increase in the complexity of the overlapping multiplet resonances in this region in thespectrum of Lu2-oncomodulin.The areas of these resonances, determined from lineshape analyses (Fig. 3, top), were plotted as a function of the Lu(II1):oncomodulin ratio (Fig. 3, bottom). In a similar fashion, the singlet resonances of €Met-105 (1.90 ppm) and €Met-86 (1.85 ppm) were evaluated and plotted as function a of added Lu(II1). Although the resonances of the N-acetylgroup and tMet-38were insensitive to thefirst phase of Lu(II1) titration, the intensitiesof tMet-105 and tMet-86 were each affected by slow exchange with upfield-shifted resonances during this phase. During the second phase, however, the intermediate resonance of €Met105 (1.89ppm) appeared in fast exchange with a resonance of Because metal binding to each chelation site ina calcium-binding protein involves at least 5 successive ligand-metal interactions, 6s rather than Ks are used to symbolize the overall (versus stepwise) microscopic binding constants.
’HNMR and Optical Stopped-flow Studies of Oncomodulin
6252
k
n
C A 6 FIG.2. Lu(1II) titration of Caa-oncomodulin as monitored by ‘HNMR spectroscopy at 40 “C. Panel A, a comparison of the downfield-shifted aCH region of ( a ) Cat-oncomodulin and ( b ) Lup-oncomodulin.Panel B , the two phases of the titration as illustrated for *Met-86:bottom, first phase (Lu:oncomodulin = 0.00 ( a ) ,0.56 ( b ) , 0.83 ( c ) , 1.12 ( d ) ) ;top, second phase (Lu:oncomodulin = 1.40 (e), 1.68 (f),1.95 (g), 2.23 ( h ) ,2.51 (i), 2.79 ( j ) , 3.07 ( k ) ) .Panel C, the spectral changes induced in the most upfield-shifted methyl region (Lu:oncomodulin ratios are the same as in panel B ) .
the Luz form slightly downfield, and thetMet-86 intermediate resonance (1.82 ppm) experienced intermediate exchange with its Luz form resonance at 1.84 ppm. The differences in the rates of exchange between metal-bound forms during the first and second phases of the titration are illustrated for the tMet86 resonance in Fig. 2 B. The downfield-shifted aCH resonances also showed significant changes during both early and late phases of Lu(II1) titration (Fig. 2A). Most of these changes, however, resulted in rather severely overlapping multiplets rendering them difficult to analyze quantitatively. Nonetheless, one of the two overlapping aCH resonances of Caz-oncomodulincentered at approximately 5.38 ppm (Fig. 2 A , trace a ) experienced slow exchange with an intermediate aCH resonance at approximately 5.45 ppm; this resonance of the CalLul form then experienced intermediate-to-slow exchange with the aCHresonance of the Luz form at 5.25 ppm. Lineshape-derived areas of the 5.25-ppm resonance were plotted as a function of the Lu(II1):oncomodulin ratio (Fig. 3, bottom). Analyses of the species distribution plot yielded the following metal-binding parameters for Ca(II):Lu(III) exchange in these oncomodulin ~ 1.2 f 0.1, and c complexes: re1 / 3 l L u : ~=~ 175 f 15, re1 / 3 ’ L u : ~ = = 5 k 1. For a more detailed description of the use of distribution plots to extract binding parameters from NMR data see the “Appendix.”The relatively poor fit of the CalLuloncomodulin dataset (represented by filled circles) at Lu(II1):oncomodulinratios greater than 1.2 was attributed to the inaccuracies of resonance area measurements for the doublet at 0.09 ppm as it broadened and shifted downfield due to effects of intermediate-to-slow exchange with Luzoncomodulin. In addition, the area of this broadened reso-
nance was distorted by its proximity to the sharp resonance of the impurity at 0.15 ppm. Kinetics of the Dissociation of Lu(III)- and Yb(III)-onconodulin Complexes-Table I summarizes the rate constants determined for the release of Yb(II1) from oncomodulin and, for comparison, rat parvalbumin. At a Yb(II1):oncomodulin ratio of 0.7, only one reaction was detected. Release of Yb(II1) from this oncomodulin complex wasapproximately 100 times faster than the release of Yb(II1)from the EF site of rat parvalbumin but slower than the release from rat parvalumin’sCD site by a factor of approximately 2. At a Yb(II1):oncomodulin ratio of 2.3, two reactions were detected, one having a rate nearly equal to that seen for the reaction at the 0.7 ratio and the other being about 10 times slower. The relative intensity of the slower component was 1.27times that of the faster. Inthe preceding section, analysis of Ca(II):Lu(III) exchange by ’H NMR methods indicated at a Lu(II1):oncomodulin ratio of2.3 the sample should be a 57:421 mixture of the Luz, CalLul, and Caz forms of oncomodulin. From this, the relative concentration of the Luz form of oncomodulin was predicted to be 1.33 times that of the CalLul-oncomodulin, inclose agreement with the relative intensities of the slow and fast components as determined from the kinetic analysis. Therefore, we attribute the faster component to release of Yb(II1) from the EF site of CalYbloncomodulin ( k l E F ) and the slower component to release of Yb(II1) fromthe EF site of Ybz-oncomodulin(12’EF). Analyses of the two-site exchange of the Met-86 singlet resonances in traces f-h of Fig. 2B yielded a value of approximately 15-20 s-’ for the rate constant of conversion between the CalLul and Luz forms of oncomodulin. Because the rate-determining
' H NMR and Optical Stopped-flow Studies Oncomodulin of
6253
modulin-Because the absolute affinities of a metal ion for a ligand can be expressed as
P
= kdk-8,
the affinities of Lu(II1) for the EFsite of oncomodulin can be calculated from the koffvalues as determined in this work if k,, can be determined. The rate at which lanthanides form complexes in aquo systems depends primarily on the metal ion itself and is insensitive to the nature of the complexed ligand (55, 56). Therefore, k,, for the complex formation reaction between Lu(II1) (or Yb(II1)) andprotein can be estimated from the dehydration rate of the Lu(II1) (or Yb(II1)) aquo cation (i.e. -1.6 X lo7 M-' s" (55)). For example, the absolute affinity of Lu(II1) for the EFsite of carp parvalbumin was calculated from the measured kOff,1.2 x s-' and k, as estimated above, yielding 1.3 x 10" M-'; @Lu:EF calculated from the measured absolute affinity of Ca(I1) for the EF site of carp parvalbumin (2.5 X 10' M" (57)) and the measured relative Lu(III)/Ca(II) affinity (0.012) was 2.1 X 1O'O M-', in very closeagreement with the affinity estimated from kinetics (17). Nonetheless, conformational changes in the protein may influence the k,, rate constants so that the following calculated absolute affinities should be considered estimates only. Hence, B'L":EF
= 1.6 X 10' M-' S-'/o.070
j 3 Z ~ u : a=
1.6
X
8-'
lo7 M-' s"/0.0036
= 2.3 X 10s M"
s-' = 4.4
X
IOg M"
where P'L,,:EF is the affinity of Lu(II1) for the EF site of Ca,Lu,oncomodulin and P'L,,,EFis the affinity of Lu(II1) for the EF site of Lu,-oncomodulin. The coefficient of cooperativity calculated from these kinetically derived values of is
LU ( 1111 : ONCOMODULIN
FIG. 3. Curve analysis and fitting of Lu(II1):oncomodulin titration data. Top, the most upfield-shifted methyl region of the 'H NMR spectrum of oncomodulin: ( a )experimental spectrum (same as trace b in Fig. 2C); ( b ) simulated spectrum; and (c) simulated individual resonances (CalLul form is shown in gray). Bottom, the species distribution plot: the NMR resonance intensities are plotted as points (A, average of the threemost upfield-shifted methyl doublets of Caz-oncomodulin;A,average of the eMet-105 and cMet-86 singlet resonances of Cap-oncomodulin; 0, the methyl doublet resonance (0.09 ppm) of CalLul-oncomodulin; 0, the average of the two methyl doublet resonances (approximately 0.01 and -0.13 ppm) of CalLuloncomodulin and Lu2-oncomodulin;and 0,the aCH resonance (5.25 ppm) of Lu2-oncomodulin); the calculated distribution profiles are displayed as curves (re1 ~'L":cD = 1.2, re1 B'LU:EF = 175, and e = 5; see "Appendix"). The deviation of the 0 data set from its calculated profile is explained fully in the text.
B2Lu:EF/b'Lu:EF
= 19.
Averaging the kinetic and NMR-derived values for the coefficient of cooperativity yielded c = 12 f 7. The absolute affinity of Ca(I1) for the EF site of Cap-oncomodulin,calculated from PILu:EF and re1 PILu:EF, was 1.3 X IO6 M-'. Because the exchange rate between the Ca,Lu, and Lu2 forms of oncomodulin (15-20 s-') was limited primarily by the off rate of Lu(II1) from the CD site, P 2 L U : c ~ ,estimated from kZcDand knas defined above, was9.1 X lo5 M-'. PILu:CD, estimated from P2Lu,cD and the NMR-derived coefficient of cooperativity, was 1.8 X lo5 M-'. Finally, P2ca:c~,estimated from P'L,,:cDand re1 P'L,:cD, was 1.5 X 10' M-'. These calculated affinity constants are summarized in Table 11. CONCLUSION
step in the exchange between these two forms is the release of Lu(II1) from the CD site of Lu2-oncomodulin, k2cD = 1520 s-1. Calculation of the Absolute Metal Ion Affinities for Onco-
The 'H NMR spectrum of Cap-oncomodulinindicates that its conformation is similar to both parvalbumins and calmodulins (9,58-63), homologous regions of secondary and tertiary structure being present in all of these proteins. The hydro-
TABLEI Kinetics of Yb(III) and Lu(III) dissociation from rat oncomodulin and rat parvalbumin Oncomodulin
Parvalbumin
Yb(II1):protein
k'ep
0.074
k2m
R2.1-
k,
PCD
s-'
0.7' 0.066 2.3d
s-'
0.0036 s-'
17.5 1.27
&
2.5 s-'
e
km
REF cnb
0.129 s-'
0.00072 s-'
4.8
0.33 s-'
0.001 s-'
1.42
'R2,' is the ratio of the amplitudes of the two kinetic phases, characterized by rate constantsklEFand k2Ep, which were used to evaluate the optical stopped-flow data. REF,CDis the ratio of the amplitudes of the phases characterized by rate constants kEFand k c D . For parvalbumin, the kinetic parameters were averaged for Yb(II1):protein ratios of 0.5 and 1.0. For parvalbumin, the Yb(II1):protein ratio was 2.5. e kZCD was calculated from lineshape analyses of selected 'H NMR resonances at various Lu(II1):oncomodulin ratios during the Lu(II1) titration of Ca2-oncomodulin.
‘H NMR and Optical Stopped-flow Studies
6254
of Oncomodulin
TABLEI1 Stabilib constants (@) of the calcium(IZ) and lutetium(IIZ) complexes of rat oncomodulin Site B’c. re1 @ILu P’ru P’ru M’
M’
CD
1.5 X
105”
1.2
1.8 x
105b
MI
9.1 X 105”
EF 1.3 x 175 2.3 X 10“ 4.4 x 109’ Calculated from j3’Lu:CDand re1 PILU:CD. Calculated from ~ z L u , c oand c. Calculated from the NMR-derived kZCDand k,, as estimated from the rate of dehydration of the Lu(II1) aquo cation (seetext (44,45)). * Calculated from @lLu:Epand re1 j3’Lu:EF. e Calculated from ~ ‘ E Fand k,,,as described in the text. Calculated from k 2 and ~ k,. as described in the text.
phobic corewhich characterizes the intersite contacts of parvalbumin’s six helices is apparently present in oncomodulin as well; we may, therefore, conclude that the helices themselves are, for the most part, intact. In addition, the @-sheet contacts found between the paired metal-binding loops of all other troponin C class calcium-binding proteins also appear intact FIG. 4. Drawing of the CD loop of oncomodulin as derived in oncomodulin. However, oncomodulin’s behavior during Lu(II1) exchange from a composite of the x-ray structures of carp parvalbumin (Refs. 2 4 and 25; main chain and liganding side chains) and is abnormal by comparison with parvalbumins.Although both turkey skeletal troponin C (Ref. 38;Asp-1 14 side chain at the oncomodulin andtheparvalbumins lose two bound cal- -Xposition and the water molecule that bonds this side chain cium(I1) ions in response to Lu(II1) chelation, oncomodulin’s to the metal ion). The main chain is depicted as a rope, bound metal preferenceforLu(II1) at onesiteis relatively weak. The as the large sphere near the center, parvalbumin’s liganding oxygens stronger of its two binds Lu(II1) with an affinity intermediate as circles with diagonal lines, and oncomodulinsAsp-59 and associated oxygen as blackened circles. Two side chains are deliberately between those of parvalbumin’s CD and EF sites. By analogy water to illustrate the proposed relative superimposed at position 59 in order with parvalbumins, oncomodulin’s stronger Lu(II1) binding orientation of oncomodulin’s unique Asp-59. site is the EF helix-loop-helix structure. This is consistent with the conclusion that oncomodulin’s Ca(I1)-specific site is Glu-6hGly-60 substitutions relative to parvalbumins) may the CD site (lo), smaller metal ions such Mg(I1) as or Lu(1II) contribute toward theCD site’s apparent calcium specificity having lower affinities for the relatively rigid cavity of the CD through alterations in inter-site stabilization or ion’s theouter site (19). spherecharge-chargeinteractionswithnonchelating side Chelation of Lu(II1) at the CD site is quite interestingfor chains, the relative importance of these factors has not been two reasons. First, the affinityof this lanthanide ion for this determined. However, the calculated affinity of calcium for site of Ca2-oncomodulin is remarkably low, 1.8 X lo5 “I, the CD site inoncomodulin is comparable to the affinitiesof being approximately equal to that of Ca(I1) as determined calcium-specific sites in calmodulins and troponin Cs (lo4from NMR-monitored Ca(II):Lu(III) exchange experiments lo6 ”I), corroborating earlier suggestions of oncomodulin’s of unusual pairingof one calcium-specific site with one (see Table 11). In parvalbumins,thestabilityconstant calcium/ Lu(II1) at the CD site is 109-10’0 M-’. Second, the binding of magnesium site. of Ln(II1)s at the CD site appears to increase the affinity metal bound at the EF site, that is unlike metal ion exchange Acknowledgments-We thank Dr. 0.Herzberg and Prof.M. N. G . in parvalbumins (9), positive cooperativity appears to exist. James for access to the MMS-X molecular graphicssystem and their We suggest that a major factor contributing toward the CD permission to useselected atomic coordinates of turkeyskeletal site’s decreased affinity for ions of small ionic radius (and, troponin C in making Fig, 4 of this paper, D. Bacon for assistance in molecular graphics using his RASTERID graphics proconsequently, toward increasedselectivity for largerions such generating gram, M. Natriss for analyzing amino acid samples, Prof. B. Kraas Ca(I1)) issubstitution of a liganding aspartic acid in tochvil for numerous discussionsof species distribution calculations, oncomodulin for what, in all other parvalbumins, isglutamic Prof. C. M. Kay for critically reading the manuscript of this paper, acid at the -X metal ion chelation position (7). Shortening and P. D’Obrennan for transforming several detailed computer graphone of the liganding arms combined with the CD site’s inher- ics into clearer figures. ent incompressibility disfavors chelation of small ions. The APPENDIX -X chelation position of site 111, the metal ion-binding site most homologous to the CD site of parvalbumins as judged In thisAppendix, we describe the use of species distribution by amino acid sequence alignment, is occupied by aspartic plots t o determine from NMR data themicroscopic metal ion acid in all sequenced troponin Cs (7). Herzberg and James binding constants and cooperativity coefficient of two-site (39) have Getermined the structureof turkey skeletal troponin calcium-binding proteins. The distribution of protein among C to 2.2-A resolution by x-ray crystallographic techniques. the four metal-bound proteinspecies generated duringa titraCa(I1) bound to siteI11 is chelated toAsp-114 only indirectly, tion experiment canbe represented as fractional populations, one of Asp-114’s y-carboxyl oxygens being hydrogen bonded a A = A / ( A + B + C + D ) = AIP-1 (1) to the oxygen of a water molecule which is itselfdirectly = B/Ptow (2) bonded to thecalcium ion. Based on this structure of troponin C’s site 111, Fig. 4 illustrates the proposed indirect chelation ac = C/Ptoial (3) by Asp-59 of the CD site calcium(I1) in oncomodulin. Ala~ = D/Pto,I (4) though other factors (e.g. the CD loop’s Ile-5bLeu-58 and
' H NMR and Optical Stopped-flow Studies of Oncomodulin
0
I
2
4
3
6255
5
M total 'Ptotol
FIG. 5. Species distribution plot for a four-component system as shown in Scheme I. The component profiles are labeled A through D. re1 /31hCD = 1, c = 1, and Pmw = 0.5 mM. re1 B'L.:EF was varied from 10 to 250 as shown, the intermediate values being 20,40, 80, and 160. The critical points as discussed in the text are marked to show their variation as a function of re1 B'I,~:EF.
where P,,, is the total analytical concentration of protein, and A-D are the concentrations of each of the four forms as per Scheme I. Each of these four protein species can also be represented as a simple function of one of two microscopic binding constants. Forexample, in the case of Lu(II1) titration of oncomodulin A = B/t%MF = C/BZMF
(5)
B = ABlMp = D/C&MF
(6)
C = A&MF = D/cB~MF
(7)
D = BcBZMF = CC&MF
(8)
2.0
I
0.
where = re1 p2 = re1 plLu:CD, MF is the ratio[Lu(III)] /[Ca(II)], [Ca(II)]is the concentration of free Ca(I1) displaced ac + 2 a d . P,t,l, and c is the by Lu(II1) equal to (ole coefficient of cooperativity. Making the appropriate substitutions of Equations 5-8 into Equations 1-4,one can express the distribution of species in terms of the three microscopic binding parameters (Dl, p2, and c ) and MF,
+
FIG. 6. Top, surface representation of the array of O values L~, calculated as described in the text. The limits over which re1 BILu:EF Be = BIMF/Q (10) and re1 B ~ L ~ : c Dwere varied are indicated on the flat plane beneath the curued surface; the flat plane which intersects the curved surface OLC = BzMF/Q (11) corresponds to an OLE, value of 0.72, the lower limit of this critical point as measured for the Ca2-oncomodulin/Lu(III) system. Bottom: f f = ~ @I@&F'/Q (12) panel A , regions of the re1 @'L,,.EF, re1 B'L~CDplane that yielded the experimentally determined values for the critical points when the where Q is the partition function. For a given set of binding coefficient of cooperativity was 1:0.72 (YE- < 0.74 (diagonal lines), parameters, one can generate a plot of the distribution of 0.13 < O L A
