Site-specific Substitution of Glutamate for Aspartate at Position 59 of Rat Oncomodulin*

Replacement of the aspartate residue at position 59 of rat oncomodulin by glutamate by oligonucleotide-directed mutagenesis binding sites, respectively. This treatment assumes that there is no cooperative interaction between the sites. Evaluation of KEF and Kco was performed using FITFUN. M$+ binding constants were determined by competition with the Ca2+ ion. For these experiments, 1.0 mM M$+ was included in both the upper chamber and in the buffer passed through the lower chamber. The stock M$+ solution was prepared by dissolution of purified magnesium metal with Ultrex@ grade HCI (J. In of competing M$+ ion, the dissocia- tion for M$+, KMe, apparent dissociation Ca2+, Kc.,, through equation, Kc, dissociation Caz+. KM,, for binding at site. The error KM, 10%.

presented at the FASEB meeting in Las Vegas, NV, May 1-5, 1988. 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 18 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence should be addressed. Tel.: 505-646-3418. Ca2+ ion is approximately octahedral with five of the six ligands donated by residues within the loop and the sixth ligand donated by a proximal residue in the COOH-terminal helix. This motif has been labeled the "EF hand" by Kretsinger (3), the spatial relationship of the flanking helices resembling that between the outstretched thumb and forefinger of the right hand. First observed in carp parvalbumin, E F hands have subsequently been observed in the crystal structures of bovine intestinal calcium-binding protein (41, skeletal troponin C (5, 6), and calmodulin (7).
Calcium ion-binding domains can be divided into two classes on the basis of their affinity for M$+ (1,2,8,9). Those which display affinity for both Ca2+ and M$+ a t physiological pH and ionic strength are said to be "Caz+/ M$+" sites, while those having negligible affinity for M$+ under physiological conditions are said to be "Ca2+ specific." Ca2+-specific sites occur in proteins which trigger physiological phenomena. For example, two of the four domains in skeletal troponin C and all four of the sites in calmodulin fall into this category. Ca2+/M$+ sites, on the other hand, are believed to play a more passive role, acting as Ca2+ buffers. Parvalbumin, a muscle-associated protein containing two Ca2+/M$'-binding domains, is believed to function in the relaxation of fast-twitch skeletal muscle. Despite extensive investigation, the ultimate origin of specificity in calcium ionbinding domains is not completely understood.
Oncomodulin, the parvalbumin-like calcium-binding protein first described and characterized by MacManus and Whitfield (lo), represents a potentially useful system in which to examine this question. Named for its frequent association with neoplastic tissue (10)(11)(12) and its ability to stimulate cyclic nucleotide phosphodiesterase in a manner reminiscent of calmodulin (13,14), oncomodulin is normally expressed only in the cytotrophoblast of fetal placental tissue (15)(16)(17). The protein is unusual in that it pairs a single Ca2+-specific site (the CD site)' with a single Ca2+/M$+ site (the EF site) (18)(19)(20). We have embarked upon a program of site-directed mutagenesis aimed at identifying residues in the protein that are critical to the calcium specificity of the CD domain. To this end, we have cloned and expressed the oncomodulin coding sequence in Escherichia coli and have initiated oligo-The abbreviations used are: CD site, metal-binding site in oncomodulin or parvalbumin flanked by the C and D helices; EF site, metal-binding site in parvalbumin flanked by the E and F helical regions; PBS, phosphate-buffered saline; Tween 20, polyoxyethylenesorbitan monolaurate; tac promoter, a hybrid promoter formed by fusion of the tryptophan and lactose promoters; PAGE, polyacrylamide gel electrophoresis, SDS, sodium dodecyl sulfate; D59E, sitespecific variant of recombinant oncomodulin in which aspartate 59 has been replaced by glutamate; Y65F, site-specific variant of recombinant oncomodulin in which tyrosine 65 has been replaced by phenylalanine; MES, 2-(N-morpholino)ethanesulfonic acid; HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid.

18751
This is an Open Access article under the CC BY license.
The potential importance of the -X ligand, the so-called "fifth ligand," in modulating the affinity and specificity of calcium-binding loops has been recognized for some years. Kretsinger (1) noted the invariant glutamate at this position in the CD-and EF-binding domains of the pawalbumins. More recently, Herzberg and James (21) discussed the presence of shorter side chains, aspartate or serine, at the equivalent position in all four domains of calmodulin and in the two regulatory domains of troponin C. The significance of their observation stems from the fact that, while the carboxyl of glutamate can coordinate directly to the bound metal ion, the oxygen atom(s) of the shorter side chains coordinate metal ions indirectly via a bound water molecule. Since the stabilities of metal-chelate complexes are determined in part by the extent of dehydration of the bound ion, the identity of the fifth ligand in a given binding domain could be crucial in determining the affinity and, hence, specificity of that domain.
Rat oncomodulin displays extensive sequence homology with rat parvalbumin. The aligned sequences are identical at 55 of 108 positions (22-24). Nevertheless, an aspartate residue occupies the fifth coordination position in the CD site of oncomodulin, the putative Ca2+-specific site, whereas parvalbumin contains a glutamate at this position. Consistent with the hypothesis of Herzberg and James (21), it has been suggested (20) that this aspartyl side chain is too short to directly coordinate to the metal ion and must, therefore, be coordinated indirectly via a bridging water molecule. Work by Sykes and his colleagues (25, 26) on parvalbumin has clearly shown that the CD-and EF-binding domains of parvalbumin differ in their interactions with metal ions of different sizes. The EF site appears to be relatively flexible and will accommodate lanthanide ions of diverse size. On the other hand, the CD site is more rigid, more cryptand-like, exhibiting a marked preference for the larger lanthanide ions. Williams et al. (20) suggested that the substitution of aspartate for glutamate at position 59 of oncomodulin further increases the preference of the site for larger ions and accounts for much of the increased calcium specificity of the oncomodulin CD site. According to this hypothesis, binding of MgZ+ ion might be disfavored relative to calcium ion because the smaller ion would be less completely desolvated than Ca2+ when bound in this larger relatively rigid binding domain.
On the basis of these arguments, we chose to replace the aspartyl residue at position 59 of oncomodulin with a glutamate residue, the residue present in all parvalbumins examined to date, in order to determine whether a more parvalbumin-like site resulted. The 7Fo-SD~ Eu3+ excitation spectra2 obtained with fully bound parvalbumin and oncomodulin are superficially similar (27); however, there are several important differences. We were particularly interested in learning whether the substitution of glutamate for aspartate at position 59 would afford a protein more closely resembling a parvalbumin.
In this paper, we describe the cloning and expression of oncomodulin, production of site-specific variants by oligonucleotide-directed mutagenesis, and characterization of the proteins using Eu3+ luminescence spectroscopy and Ca2+binding measurements.
7F0 and 5Do are term symbols denoting the ground electronic state and first excited state, respectively, of the Eu3+ ion. 'Fe5Do indicates the electronic transition that occurs upon absorption of a photon of the appropriate energy.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotides, synthesized on a Beckman DNA synthesizer, were purified by electrophoresis through a 20% polyacrylamide gel containing 7.0 M urea in the presence of Tris-borate-EDTA buffer. The bands of lowest mobility were excised and extracted for use as primers in sequencing and oligonucleotide-directed mutagenesis and as end-labeled probes.
Europium sesquioxide (Eu203) was the generous gift of Molycorp, Inc. Solutions of EuC13 were prepared by dissolution of the sesquioxide in concentrated HCl, followed by removal of the excess HCl. They were standardized by titration with EDTA at pH 6.0 to a Xylenol Orange end point as described by Lyle and Rahman (28). All other chemicals were reagent grade or better.
Electrophoresis-Discontinuous SDS-PAGE was performed in 7.5-20% gradient gels using the Laemmli buffer system (29). Gels were silver-stained by the method of Morrissey (30). Urea gradient gel electrophoresis was performed as described by Creighton (31), and the gels were stained with a 1:5 dilution of Bio-Rad protein assay reagent. Agarose gel electrophoresis was performed as described in Maniatis et al. (32).
Immunochemistry-Polyclonal antibodies against rat oncomodulin were raised in a male New Zealand White rabbit by multiple subcutaneous injections (33) of tumor-derived oncomodulin emulsified with Ribi adjuvant (Ribi ImmunoChem Research, Inc., Hamilton, MT).
Electrophoretic transfer of proteins to nitrocellulose membranes was performed as described by Erickson et al. (34). Prior to use, the nitrocellulose replicas were blocked for 1-2 h at room temperature in 5% (w/v) nonfat dry milk in 50% (v/v) calf serum, 50% PBS. Primary and secondary antibodies were diluted into the same solution containing 0.05% (v/v) Tween 20 (35) in addition. Washes were performed with PBS containing 0.05% Tween 20 (PBS/Tween).
Expression of oncomodulin in E. coli was detected immunologically employing the method of Helfman et al. (36), probing first with an adsorbed preparation of rabbit anti-oncomodulin (final dilution 1:lOOO) and then with an affinity-purified preparation of '2SI-labeled goat anti-rabbit IgG (Du Pont-New England Nuclear). Antibodies directed against E. coli antigens were removed from the rabbit antiserum before use by adsorption with an E. coli lysate immobilized on Bio-Rad Affi-Gel 10.
Spectroscopy-UV absorption spectra were acquired on a Perkin-Elmer 320 UV-visible spectrophotometer. Eu3+ 7Fo-8'Do excitation spectra were acquired with the laser spectrofluorimeter described previously (27, 37), employing an emission wavelength of 618 nm. Spectral deconvolution and curve fitting were performed as described previously (27) employing the public procedure FITFUN, within the PROPHET computer resource sponsored by the Department of Research Resources of the National Institutes of Health.
Binding Measurements-Dissociation constants for Ca2+ at pH 7.4 were determined at room temperature (22 * 1 "C) by the flow dialysis technique of Colowick and Womack (38), using a sample cell based on their design. A flow rate of 2.0 ml/min was maintained through the lower chamber by means of a Rainin peristaltic pump. The eluate was sampled 5 min after each addition of Ca2+, and three 0.4-ml aliquots were counted in 5 ml of 3a70B scintillation mixture (Research Products, Inc.). 15 pCi of 45CaC12 (Du Pont-New England Nuclear, 31 mCi/mg of Ca2+) was used per determination, and the protein concentration in the upper chamber was 100 pM. Depletion of 46CaZ+ from the upper chamber during the course of the experiments, which lasted less than 2 h, was less than 5%. No correction was applied. Bound metal ion was removed from our protein preparations by adding 2 eq of Eu3+ and then passing the resulting solution over a 1 X 18-cm column of EDTA-agarose (39) at a flow rate of 0.3-0.5 ml/ min. The absence of a luminescence signal at 580 nm due to the ' F e 'Do transition indicated that removal of Eu3+ by this procedure was complete; and the residual Ca2+ content was C0.1 g atoms/mol of protein, as determined by flame atomic absorption employing a CZH,/ N,O flame. Buffers employed in binding or luminescence studies were likewise passed over an EDTA-agarose column, reducing the level of contaminating Ca2+ to less than 1 pM.
The binding data were tit to a parametric form of a two-site Scatchard equation, where r is the number of moles of Ca2+ bound per mol of protein and KEF and KCD represent the dissociation constants for the EF and CD binding sites, respectively. This treatment assumes that there is no cooperative interaction between the sites. Evaluation of KEF and Kco was performed using FITFUN.
M$+ binding constants were determined by competition with the Ca2+ ion. For these experiments, 1.0 mM M$+ was included in both the upper chamber and in the buffer passed through the lower chamber. The stock M$+ solution was prepared by dissolution of purified magnesium metal with Ultrex@ grade HCI (J. T. Baker Chemical Co.). In the presence of competing M$+ ion, the dissociation constant for M$+, KMe, is related to the apparent dissociation constant for Ca2+, Kc.,, through the equation, where Kc, is the intrinsic dissociation constant for Caz+. In the calculation of KM,, the [Mg2+] was not corrected for binding at the other site. The resulting error in KM, is less than 10%.

RESULTS
Cloning of Rat Oncomodulin cDNA-Recombinant DNA manipulations were performed in accordance with National Institutes of Health guidelines. The monograph by Maniatis et al. (32) served as our primary resource for common procedures.
mRNA was isolated from monolayers of 5123tc (cultured in RPMI 1640 medium supplemented with 5% horse serum) by a minor modification of the procedure described in Maniatis et al. (32). Enrichment for poly(A+) RNA was achieved by two cycles of binding and elution from oligo(dT)-cellulose. Double-stranded cDNA was synthesized from 10 pg of mRNA by the procedure of Gubler and Hoffman (40). After attaching ClaI linkers, the cDNA population was inserted into the ClaI site of Bluescript. Transformation of Library Efficiency@ E.
coli DH5a (Bethesda Research Laboratories) with 10 ng of the resulting preparation yielded approximately 90,000 recombinants, which were replica-plated onto nitrocellulose filters (Millipore) and screened sequentially with end-labeled oligonucleotides complementary to the 3' and 5' ends of the coding sequence (24).
Twenty-one colonies from a cDNA library of approximately 90,000 recombinants hybridized to a probe for the sequence immediately downstream from the COOH-terminal codon. This frequency suggests that the oncomodulin message corresponds to approximately 0.023% of the total hepatoma mRNA, substantially higher than the frequency reported by Gillen et al. (24). Of the 21 positive clones, 20 hybridized to a probe for the region immediately upstream from the start codon and presumably contained the entire oncomodulin coding sequence. Plasmid DNA was isolated from seven of the positive colonies, digested with ClaI, and analyzed by agarose gel electrophoresis. The insert size varied from 548 to 725 base pairs. The plasmid harboring the shortest insert, pRH4, was selected as the starting point for expression in E. coli. Sequence analysis revealed that, relative to the sequence reported by Gillen et al. (24), the pRH4 insert lacked 165 base pairs from the 3' end but contained an additional 53 base pairs at the 5' end of the gene.
Expression of Rat Oncomodulin Coding Sequence-The strategy employed for expressing oncomodulin in E. coli is diagrammed in Fig. 1. The vector chosen for expression was pKK223-3 (Pharmacia LKB Biotechnology Inc.) Prior to inserting the oncomodulin sequence into pKK223-3, the unique EcoRI site in the vector was replaced by a KpnI site using linkers, affording pKK223-3.2. A sample of this plasmid was then prepared for cloning by linearizing with Hind111 and filling in the 5' overhangs with the Klenow fragment of DNA polymerase. The oncomodulin cDNA insert was liberated from pRH4 as a BamHI-Sal1 fragment, likewise filled in with Klenow, electrophoretically purified, and ligated into the blunt-ended vector. Following transformation of E. coli DH5a, recombinants were identified by colony blot hybridization using an end-labeled probe for the 5' end of the cDNA. The desired orientation was verified by restriction analysis.
The resulting construction, labeled pKK223-3.20M, was prepared for unidirectional deletion with exonuclease I11 by sequential treatment with KpnI and SmaI. The 3' overhang generated by KpnI protected the promoter region from deletion, since Ex0111 will only initiate digestion at blunt ends or 5' overhangs. Deletions were performed according to instructions supplied by Stratagene with the Bluescript vector. Following recircularization with DNA ligase, the preparation was used to transform Library Efficiency@ E. coli DH5a.
Approximately 20,000 colonies were screened for oncomodulin expression by colony Western blot (36). Those affording the 11 strongest signals were chosen for further characterization and labeled pMHl through pMH11. Liquid cultures of all 11 clones were harvested in midlog phase, and crude extracts were prepared using an equal number of cells from each culture. Serial dilutions of the extracts were spotted onto nitrocellulose, probed with adsorbed rabbit anti-oncomodulin, and 'z51-goat anti-rabbit IgG and autoradiographed. pMH4 and pMHll displayed the highest levels of expression. Oncomodulin production by pMH6 and pMH7, the next highest overproducers, was lower by a factor of approximately 2.
Although the recombinant oncomodulin expressed by pMH6 and pMH7 co-migrates with tumor-derived oncomodulin during nondenaturing PAGE, the majority of the oncomodulin expressed in pMH4 and pMHl1 had a lower relative mobility (Fig. 2). However, all four proteins co-migrate during SDS-PAGE, and all four plasmids have the same coding sequence reported by Gillen et al. (24). Until we understand the origin of the altered electrophoretic mobility, we are using pMH7 as our expression vector for wild-type recombinant oncomodulin.
pMH6 and pMH7, which produce -3% of the total protein as oncomodulin, were found to have 12 and 18 nucleotides, respectively, between the Shine-Dalgarno sequence and the start codon. pMH4, which produces just over twice as much of the recombinant protein, has 16 nucleotides between the ribosome-binding site and the start codon. I t appears, then, that the optimal separation is approximately 16 nucleotides. Purification of Recombinant Oncomodulin-E. coli DH5a harboring either pMH7 or a site-directed variant thereof was grown to late log phase at 37 "C with continuous shaking in 1-liter cultures of LB broth containing 50 pg/ml ampicillin. In DHSa, maximal expression of recombinant oncomodulin was observed even in the absence of isopropyl-1-thio-P-Dglucopyranoside. Each liter of culture yielded 3.5-4.0 g of cell paste, which was collected by centrifugation, washed once with PBS, and stored a t -80 "C prior to use.
Bacterial cell lysis was achieved by treatment with lysozyme (0.5 mg/g of cell paste) and passage through a French pressure cell. Employing a modification of the procedure by MacManus (41), the preparation was subjected to heat treatment at 80 "C, ion exchange chromatography on DEAE-Sepharose (Pharmacia), and gel filtration through Sephadex G-75 (Pharmacia). The progress of an isolation of recombinant oncomodulin from pMH7, as monitored by SDS-PAGE, is presented in Fig.  3. The material collected from the gel filtration column is highly purified. Only a single minor contaminant is visible (running just ahead of the major band) on an overloaded silver-stained SDS gel.
Protein concentrations were determined spectrophotometrically for solutions of recombinant oncomodulin. An extinction coefficient a t 275 nm of 3650 M" cm" was determined for the wild-type protein on the basis of careful and repeated titration of the protein with Ed+. Extinction coefficients were obtained for the site-specific variants discussed below with the Bio-Rad protein assay, employing solutions of known absorbance and using wild-type recombinant oncomodulin of known concentration as a standard.
Site-specific Mutagenesis of Rat Oncomodulin-The Blue- script cloning vector contains the intergenic region of bacteriophage M13, thus permitting production of single-stranded plasmid in E. coli infected with a helper phage (VCS-Ml3, purchased from Stratagene). We utilized this feature of Bluescript to produce the coding strand of pRH4. To increase the frequency of the mutant phenotype, we employed the strategy of Kunkel (42,43), producing our single-stranded DNA in a host that was deficient in dUTPase (dut-) and uracil-Nglycosylase (ung-). To the resultant uracil-containing singlestranded template, we annealed a 5'-phosphorylated oligonucleotide bearing a single nucleotide mismatch responsible for either the mutation of aspartate 59 to glutamate or tyrosine 65 to phenylalanine. Double-stranded circular DNA, produced by extension with T 4 DNA polymerase and recircularization with T 4 DNA ligase, was used to transform E. coli DH5a.
Recombinants harboring the desired mutation were detected by the method of Norris et al. (44). Twenty colonies selected a t random were grown in 1.5-ml cultures, harvested in late log phase, and lysed by the method of Holmes and Quigley (45). NaOH and NaCl were added to the lysates to denature the plasmid DNA, and samples of the six lysates were spotted onto three nitrocellulose filters. Following neutralization and wash steps, the filters were probed with the approriate end-labeled mutant oligonucleotide for 4 h a t 30 "C. After a brief rinse at room temperature, single 5-min washes were then performed at T, -12 "C, T, -6 "C, or T,,, in 6 X SSC, with each filter being treated at a different temperature.
The melting temperature T, was estimated from the composition of the particular oligonucleotide using the formula given by Maniatis et al. (32). Autoradiographic signals which persisted at the higher stringencies were presumed to originate from colonies bearing the mutant genotype. Plasmid DNA was isolated from these colonies and sequenced to verify the presence of the altered genotype. Mutants bearing either a tyrosine * phenylalanine mutation a t position 65 (pRH12) or an aspartate + glutamate substitution a t position 59 (pRH13) were isolated in this manner.
Expression of the mutant alleles required that we shuttle the altered sequence from Bluescript into our expression plasmid, pMH7. T o accomplish this, we made use of unique PstI and EcoRI sites at codons 16 and 101, respectively, in the oncomodulin coding sequence (24). Thus, we treated pRH12 and pRH13 sequentially with PstI and EcoRI, isolated the appropriate fragments from a 1.4% agarose gel, and cloned them into pMH7 that had likewise been treated with PstI and EcoRI and gel purified. The resultant plasmids were sequenced to verify the presence of one, and only one, nucleotide modification. The recombinant protein containing the tyrosine + phenylalanine substitution is subsequently denoted as Y65F, while the protein containing the aspartate + glutamate substitution is referred to as D59E.
Characterization of the Mutant Oncomodulins-The recombinant oncomodulins we have isolated (wild-type and mutations alike) behave similarly to the tumor-derived protein during purification, withstanding heat treatment at 80-83 "C for 5 min and fractionating similarly on ion exchange and gel filtration. Moreover, the mobilities of all three recombinant proteins under denaturing or nondenaturing conditions are indistinguishable from the tumor-derived protein. These observations imply: 1) that the structure and stability of recombinant oncomodulin must be very similar to those of the tumor-derived protein; and 2) that the two site-specific modifications we have introduced have not significantly diminished the stability of the polypeptide.
The latter conclusion is substantiated by urea gradient gel electrophoresis. This technique employs electrophoretic mobility as a conformational probe (31). In general, the mobility of proteins through a polyacrylamide matrix decreases upon denaturation, since the unfolded form has a greater volume. Thus, if protein is loaded across the top of a polyacrylamide gel containing a horizontal gradient of urea and subjected to electrophoresis, a wave in the mobility is observed (Fig. 4). If the observed mobility is plotted as a function of urea concentration in the gel, the concentration of urea required for denaturation (the midpoint) can be determined. The values obtained for wild-type oncomodulin and the D59E and Y65 variants were 3.5, 3.6, and 3.4 M, respectively, suggesting that the three proteins have similar conformational stabilities. Interestingly, the protein bands split in the transition region. This phenomenon suggests (31)  The polyacrylamide gel, which was cast as described by Creighton (31), incorporated a 0-6.0 M gradient of urea and a 16-12% gradient of acrylamide. The purpose of the reverse acrylamide gradient was to compensate for variation in mobility resulting from increased viscosity a t higher urea concentrations. 0.050 M MES, pH 6.5, containing 1 mM EDTA, served as the gel and reservoir buffer. The EDTA was included to ensure that the proteins would be in their apo forms. The three proteins were loaded a t 2-h intervals. Electrophoresis was performed a t 300 V a t 4 "C. The gel was stained with Bio-Rad protein assay reagent. 0 proline residues at positions 21 and 26. At higher temperatures, the discrete bands in the transition zone are replaced by a broad smear, indicating more rapid interconversion of conformers.
Eu3+ Luminescence Properties of the Recombinant Oncomodulins-The utility of Eu3+ as a spectroscopic probe derives in part from its ability to undergo an electronic transition near 580 nm involving the 7Fo ground state and the 5Do excited state (46). This transition is nondegenerate, so that each distinct Eu3+ species in solution (or each binding site in a multisite protein) should give rise to a*single peak having a typical line width on the order of 5 A. Since the 7Fo+5Do transition is highly forbidden (e = 0.01 M" cm-'1, it can only be studied in dilute solutions of biological materials by employing an intense laser source and measuring the luminescence (at 618 nm) that accompanies return of the ion to the ground state. The display of luminescence intensity versus the wavelength of absorbed light is referred to as the Eu3+ 7Fo+5Do excitation spectrum. This transition has been used to study such calcium-binding proteins as thermolysin (471, troponin C (48), calmodulin (49,50), phospholipase A (51), parvalbumin (27,(52)(53)(54)(55) and oncomodulin (27,55).
T h e 7 F e 5 D o spectrum of fully bound recombinant oncomodulin is presented in Fig. 5B. In most respects, it mimics that observed with the tumor-derived protein (Fig. 5A). At PH 5.0, it is dominated by a slightly asymmetric peak at 5795 A. Although this "low pH" spectrum arises from Eu3+ ions bound at both the CD and EF sites, only a single peak is observed, presumably because the two spectral contributions are nearly coincident. In fuFher analogy to the tumor-derived protein, the peak at 5795 A is replaced a t higher pH values by a broad intense spectrum centered at 5781 A. When the intensity at 5781 8, is plotted uersus pH (Fig. 8), a pK, for this spectral transition of 6.0 jI 0.1 is extracted. This value is lower than the pK, determined for the tumor preparations (6.3 0.1) and, together with the somewhat greater intensity of the high pH spectrum, may signal an altered conformation at the NH2 terminus of the recombinant protein.
Substitution of a glutamyl residue for the aspartyl residue normally present at position 59 of oncomodulin, via sitespecific mutagenesis, alters the Eu3+ 7FO-+6D0 spectrum in several respects. Selected spectra from a Eu3+ titration of apoD59E, performed at pH 5.0, are presented in Fig. 6. At low ratios of Eu3+:protein, a single sharp peak is present, centered at 5795 A. As the titration proceeds, however, a prominent shoulder appears at 5791 A. Thus, the two metal ion-binding sites in D59E are filled sequentially by Eu3+, with occupation of the EF site occurring first. This behavior was noted previously with the tumor-derived protein (27) and occurs with the wild-type recombinant protein as well (data not shown).
For purposes of comparison, the spectrum of the D59E mutant protein containing two equivalents of Eu3+ at pH 5.0 is presented in Fig. 7, together with the corresponding spectra for the wild-type recombinant oncomodulin and rat and pike parvalbumins. The D59E spectrum displays a shift in the position of the CD signal to lower wavelength relative to the spectrum of the wild-type protein. In fact, the location of the CD signal, at 5791 A, is quite similar to that observed for the CD site in pike parvalbumin (37). The intensities and line widths differ, however, accounting for the somewhat different overall appearance of the spectrum.
The 7Fo_t5Do spectrum of the fully bound D59E variant is displayed as a function of pH in Fig. 5C. Qualitatively, the behavior resembles that observed with the wild-type oncomodulin. However, there are two noteworthy differences. The position of the high-pH peak observed with the mutant protein is shifted to higher wavelength. As observed w$h pike and rat parvalbumins, the spectrurg appears at 5784 A for the mutant protein rather than 5781 A, as observed with tumorderived and wild-type recombinant oncomodulins. Furthermore, the pH dependence is altered, with the spectral transition now occurring at substantially higher pH. The pK, value  determined for the D59E mutation is 6.8 (Fig. 8), as compared with 6.0 for the wild-type protein. Since the corresponding pK, value for a parvalbumin-type site is ~8 . 2 , the observed shift is in the expected direction for a mutation producing a more parvalbumin-like CD-binding domain.
The pH dependence of the 7FO+5D0 spectrum observed with the Y65F mutant, in which phenylalanine replaces tyrosine at position 65, is presented in Fig. 5 0 . It is important to note that the appearance of the low pH spectrum, peak position of the high pH spectrum, and the pK, determined for the pHdependent spectral transition (pK, = 5.9) are very similar to those observed in the wild-type spectrum. Thus, the spectral alterations that accompany replacement of aspartate 59 with glutamate do not appear to be nonspecific effects resulting from indiscriminate substitution of residues near the CDbinding domain.
Affinities of Recombinant Oncomodulin and D59E for Ca2+ and Mg2+-The binding of Ca" to aporecombinant oncomodulin was studied in 0.15 M NaCI, 0.025 M HEPES, pH 7.4, by flow dialysis, as described under "Experimental Procedures." When subjected to Scatchard analysis, the data afforded distinctly biphasic plots (Fig. 9 left, 0). This behavior suggests that the two binding sites of oncomodulin have very different affinities for Ca", consistent with their distinct ion binding specificities (18,19). Values for Kca at the CD and EF sites were extracted by fitting the binding data to a two-site Scatchard model (Equation 1). The results for three separate determinations are listed in Table I  calcium-binding measurements were performed in the presence of 1.0 mM M g + (Fig. 9 left, 0). Results from three experiments are also presented in Table I. Under these conditions, the observed dissociation constants, KC.,, are related to the true dissociation constants, KC, and K M~, as indicated in Equation 2 under "Experimental Procedures." The mean values for Kca and Kc,, were used in the calculation of Khlg Values for KMg of 3 and 0.16 mM were extracted for the CD and EF sites, respectively, of wild-type recombinant oncomodulin.
When apoD59E is titrated with Ca", the resulting Scatchard plot (Fig. 9B, 0 ) is qualitatively similar to that obtained with the wild-type protein. Binding constants determined from three separate experiments are displayed in Table I

DISCUSSION
Careful inspection of the calcium-binding loops found in intracellular calcium-binding proteins reveals certain commonalities. Each site contains six ligands arranged in a pseudo-octahedral configuration; four or five of the six are side chain oxygen atoms, while a peptide oxygen is coordinated at the -Y position (2}. In general, three or four of the liganding atoms are ionic oxygens from carboxylate groups, the remainder being neutral oxygen species furnished by serine, threonine, asparagine, or water. Certain domains have affinity for both Ca" and M 2 + under physiological conditions, while others have affinity only for Ca2+. However, despite a knowledge of the primary structures of numerous calcium-binding proteins, the factors which determine the binding specificity of calcium-binding sites have yet to be pinpointed. Apparently, neither the number of coordinating carboxylates nor the net charge in the binding loop is the sole determinant of specificity (1). It is entirely possible that residues outside of the binding loop per se, in the flanking helical regions, are critical determinants of specificity. Reid et al. (58), for example, have found that calcium-binding loops themselves have marginal affinity for calcium in aqueous solution and that the affinity is increased markedly by inclusion of both flanking helical elements. Moreover, the large decrease in affinity for Ca2+ which accompanies separation of the parvalbumin calcium-binding domains by limited proteolysis argues that domain-domain interactions might even be critical determinants of ion-binding properties (59).
The parvalbumins contain two high affinity calcium-binding domains, both of which fall into the CaZ+/Mg2+ category (1). Oncomodulin, on the other hand, has been diagnosed as containing a single Ca*+/Mg+ site (the EF site) and a single Ca"-specific site (the CD site), despite extensive sequence homology with parvalbumin (18-20). We reasoned that systematic replacement of the parvalbumin codon for the oncomodulin codon at points of nonidentity in the CD binding domain might suggest which residues are critical for calcium specificity. In assessing the consequences of these substitu-tions, we intend to use spectroscopic measurements with lanthanide ions as well as traditional metal ion-binding studies.
Lanthanide ions have proven to be useful probes of calciumbinding systems (60). Like Ca2+, all have a preference for oxygen-containing ligands, high coordination numbers, and a tolerance for irregular coordination geometry. Additionally, those in the middle of the series have ionic radii similar to Ca2+. Consequently, lanthanide ions readily substitute for Ca2+ in many calcium-binding proteins. In contrast to calcium ion, however, the lanthanides are amenable to study by a variety of spectroscopic techniques (60). We have found the luminescent lanthanide ions, Eu3+ and Tb3+, to be particularly useful in studying several calcium-binding proteins, including parvalbumin and oncomodulin.
Generalizing from our studies of pike and rat parvalbumins (27, 37), we would conclude that the Eu3+ 7Fo+5Do spectrum of a parvalbumin has three characteristic properties. 1) At low pH, the spectrup consists of a partially resolved doublet centered near 5795 A; 2) at high pH, the spectrum consists of a broad intense peak centered at 5784 A; and 3) the pKa governing the interconversion of the two spectra is 4 . 2 . We believe these features are diagnostic for a parvalbumin-like Ca2+/Mg2+ site. Although there may be exceptions to these rules, they nonetheless provide a foundation for comparison of parvalbumin and oncomodulin from rat.
Until recently, our investigation of lanthanide-oncomodulin complexes had been hampered by the difficulty and expense involved in obtaining sufficient material for experimentation. Fortunately, the cloning and expression of the protein in E. coli has eliminated that obstacle; it is now possible to isolate 50-100 mg of highly purified recombinant oncomodulin with little difficulty, which has facilitated an examination of the recombinant protein in some detail by Eu3+ 7Fo-+5Do excitation spectroscopy.
In most respects, the spectral behavior of the recombinant protein mirrors that of the tumor-derived protein. The spectra observed with the fully bound proteins are both pH-dependent. Furthermore, the spectra obtained at the pH extremes are highly similar. However, the proteins appear to differ slightly with regard to the pH at which they undergo the pHdependent spectral transition. Tumor-derived oncomodulin exhibits a pK, for the transition of 6.3, whereas the wild-type recombinant protein exhibits a pK,, of 6.0. Furthermore, the intensity of the high-pH spectrum of recombinant protein is somewhat greater than that observed with tumor-derived oncomodulin. These spectroscopic differences may stem from structural differences at the NHZ termini. The NH2 terminus of the tumor-derived protein is N-acetylserine, whereas the NH2-terminal residue of the recombinant preparations is either a serine, methionine, N-formylmethionine, or possibly a mixture of the three species.
Although superficially similar, the 7Fo-+6Do spectra arising from the CD-binding domains of rat oncomodulin and rat parvalbumin differ significantly in three respects. Whereas the features arising from the CD and EF sites are resolved in the parvalbumin spectrum at pH 5.0, the spectrum of fully bound oncomodulin exhibits a single peak. Moreover, the CD domain spectrum of parvalbumin undergoes a pH-dependent spectral transition with a pK, = 8.2, whereas the CD site of oncomodulin undergoes an analogous transition at substantially lower pH values: pK, = 6.3 for the tumor-derived oncomodulin, 6.0 for the recombinant protein. Additionally, the high pH spectrum assigned to the CD site of parvalbumin is centered at 5784 A, whilt that assigned to the CD site of oncomodulin falls at 5781 A. It was our hope in embarking upon a program of site-directed mutagenesis that we would be able to identify critical nonidentities in the two proteins responsible for these spectral differences and, furthermore, that these nonidentities would underlie the differing specificities of the two sites.
Oligonucleotide-directed mutagenesis, employing the strategy of Kunkel (42, 43) and screening method of Norris et al. (44), enabled us to produce two site-specific variants of oncomodulin with relatively little difficulty. We chose the aspartate to glutamate substitution at position 59 of oncomodulin as a starting point because cogent arguments by several investigators had indicated that the identity of the residue at this position was likely to be an important determinant of specificity (1, 20, 21). The consequences of this particular mutation were quite striking. To reiterate, substitution of glutamate for aspartate at position 59 of oncomodulin; 1) shifts the low-pH spectrum of the CD site to lower wavelength; 2) raises the pKa for pH-dependent spectral transition undergone by the CD site; and 3) shifts the center of the CD high-pH spectrum from 5781 to 5784 A. Each of these changes is in the direction expected for a more parvalbumin-like binding domain.
In contrast to the D59E mutant, replacement of tyrosine 65 by phenylalanine has very minor effects on the 7FO-+5D0 spectral characteristics of the oncomodulin CD site. This result gives us some confidence that the consequences of substituting glutamate for aspartate at position 59 are, in fact, highly specific effects and not merely the result of a general perturbation of the CD site structure. Although we anticipated that neither of these substitutions would cause disruption of the global structure of the protein, several observations support this hypothesis. The three recombinant proteins that we have examined, wild-type and mutant proteins alike, behave identically during purification, withstanding heating at 80 "C for 5 min. In addition. they exhibit virtually identical electrophoretic mobilities under denaturing and nondenaturing conditions. Moreover, all three proteins exhibit the same midpoint for the denaturation, 3.5 f 0.1 M, upon urea-gradient gel electrophoresis, suggesting that the structure of the protein has been perturbed minimally by the single amino acid substitutions.
Based on results obtained with a series of model Eu3+ complexes, Albin and Horrocks (.61) have proposed that the net charge of the Eu3+ complex is a primary determinant of peak position, with the transition being shifted to higher wavelength by ligands of increasingly negative charge. We would predict, then, that substitution of glutamate for aspartate at position 59 would bring the negatively charged carboxylate group into closer proximity to the bound Eu3+, shifting the spectrum to higher wavelength. In fact, the observtd shift in the position of the CD peak from ~5 7 9 5 to 5791 A upon substitution of glutamate for aspartate is in the opposite direction. Interestingly, although the peak position for the low-pH CD spectrum is shifted to lower wavelength by the mutation, the center of the high-pH !pectrum is shifted to higher wavelength (from 5781 to 5784 A). At present, we lack an explanation for the opposing behaviors of the low-and high-pH spectra. However, the results certainly suggest that, at least with respect to protein complexes of Eu3+, factors besides ligand charge are important determinants of peak position.
The pH-dependent spectral transition observed with oncomodulin and parvalbumin has been attributed to hydrolysis of a water molecule coordinated to the protein-bound Eu3+ ion (27, 37, 55). If this explanation is correct, then the pK, for the transition should be sensitive to the net charge of the metal-protein complex; the more positive the charge, the more highly polarized the bound water molecule and the lower the pH required for hydrolysis. Assuming that the coordination of the trivalent lanthanide precisely mimics that of Ca", the effect of the D59E mutation on the pK, for this transition is consistent with this explanation. The glutamyl carboxyl in the mutant at position 59 is presumably coordinated directly to the Eu3+ ion, whereas the aspartyl carboxyl in the wildtype protein is probably bound through a bridging water molecule (21). Thus, the net charge of the Eu3+-CD site complex should be fractionally less positive for the D59E mutant than for the wild-type protein. Accordingly, we observe a higher pK, for the putative hydrolysis.
We have also conducted a preliminary examination of the Ca2+ and Mg2+ binding properties of recombinant oncomodulin and the D59E site-specific variant. The results obtained with the wild-type protein offer further testimony to the nonequivalence of parvalbumin and oncomodulin. In Ca2+ binding studies conducted on various parvalbumins, the CD and EF binding sites are indistinguishable. The reported dissociation constants for ca2+ range from 0.37 nM for carp (PI = 4.25) parvalbumin (57) to between 2 and 8 nM for frog, hake, and rabbit parvalbumins (8). The binding constants for Ca" reported for rat parvalbumin are 1-2 nM (62). By contrast, the two binding sites in oncomodulin exhibit distinctly different affinities for Ca2+. Kc. for the EF site is estimated at 42 nM, while K M~ = 0. 16 (18,19) for the calcium-specific nature of the oncomodulin CD site. The alterations in binding affinity that accompany the substitution of glutamate for aspartate at position 59 are not striking. There appears to be a perceptible increase in the affinity of the CD site for Ca2+; Kc, decreases from 0.78 to 0.55 pM. The replacement also appears to increase the affinity of the site for M$+. The K M~ calculated for the D59E sitespecific variant is 1 mM, as compared with 3 mM for the wildtype protein. It should be noted, however, that these values O f KM, are accompanied by a significant degree of uncertainty.
Judging from our binding data, the consequences of the The minor alterations in affinity for Ca2+ and M$+ at the CD site wrought by the D59E mutation might seem paradoxical in view of the marked sensitivity of the Eu3+ luminescence properties to the same substitution. It should be noted, however, that the two experimental techniques are not reporting on identical thermodynamic quantities. The Eu3+ 7 F e 5 D o transition monitors interactions between the bound ion and the coordinating ligands and, thus, is primarily a barometer of the enthalpic contribution to the free energy of the ionprotein complex. Since the substitution of glutamate for aspartate at position 59 replaces indirect coordination via a bridging water molecule by direct coordination through a carboxylate, the mutation is expected to have a significant effect on the 7FO+5D0 spectral properties. The relative affinity of the CD domain for Ca2+ and M$+, on the other hand, will be influenced by entropic considerations as well, such as how completely solvent is excluded from the site when each ion is bound. Noncoordinating residues are likely to be important determinants of solvent exclusion, since the size of the binding pocket and the coordination geometry will be dictated by tertiary structural interactions involving the binding loop, adjacent helical segments, and the rest of the polypeptide chain. Thus, the results afforded by the two lines of inquiry are not incompatible and, in fact, illustrate why both types of measurement are necessary for a complete analysis of the ionbinding properties.
MacManus et al. (64) have also examined the consequences of the D59E mutation, employing UV absorbance, fluorescence, and NMR measurements. These investigators find that, in contrast to the wild-type protein, the spectral properties of the site-specific variant are sensitive to the binding of M e at the CD domain. In light of our binding measurements, their result suggests that the D59E substitution facilitates M$'-provoked conformational changes in the protein but does not profoundly alter the observed affinity of the CD site for the ion.
The changes in the Ca" and M F dissociation constants resulting from the D59E mutation are small, and their significance might have gone unrecognized had we not already seen the Eu3+ data. The relative ease of performing the lanthanide luminescence experiments, in comparison with conventional Ca2+ binding measurements, and their sensitivity to subtle changes in ligation make the technique well suited for screening our bank of site-specific variants for significant alterations in metal ion-binding characteristics.
Despite the aforementioned changes, the Eu3+ luminescence properties of D59E still differ substantially from those of rat parvalbumin. Moreover, the dissociation constants for Ca2+ and Mg2+ in D59E are little changed from wild-type oncomodulin. These observations are significant since the coordinating residues are identical in D59E and rat parvalbumin. The clear-cut implication is that residues besides those directly involved in ion binding are important in determining the binding characteristics of a given calcium-binding domain. We are currently investigating the significance of other sequence nonidentities in the CD binding loop and in the adjacent helical segments.