High level expression in Escherichia coli and characterization of the EF-hand calcium-binding protein caltractin.

Caltractin is a member of the calmodulin superfamily of Ca(2+)-binding proteins that was originally cloned at the DNA level from the unicellular green alga Chlamydomonas reinhardtii. Human and mouse homologs to algal caltractin have been recently characterized. In the studies reported here, recombinant Chlamydomonas caltractin was expressed at high levels in Escherichia coli and purified to homogeneity. The use of the ompT-host BL21 proved critical for obtaining high yields of homogeneous full-length protein. Growth and purification protocols were optimized to allow reproducible and efficient production of tens of milligrams of pure protein from 1-liter cultures. Caltractin has a distinct UV spectrum which is largely dominated by the fine structure due to the 9 Phe residues. Unlike other members of the same protein family, the UV and the CD spectra do not change upon addition of Ca2+ to the apoprotein. However, the 1H NMR spectrum shows distinct changes upon Ca2+ binding, which are indicative of structural and/or dynamic changes largely reminiscent of other members of the calmodulin superfamily. Ca2+ binding measurements demonstrated the binding of four Ca2+ ions to caltractin with two higher affinity (Kd = 1.2 x 10(-6) M) and two lower affinity (Kd = 1.6 x 10(-4) M) sites. Caltractin is highly stable in both the apo- and the Ca(2+)-loaded states. The unusual stability of apocaltractin makes this protein highly suited for structural studies by multidimensional NMR aimed at understanding the structural and dynamic consequences of Ca2+ binding, and the molecular basis of Ca2+ signal transduction.

8 To whom correspondence should be addressed. %I.: 619-554-9826; representatives of a separate branch in the evolution of EFhand calcium-binding proteins (6-9). There is also evidence to suggest that these proteins are related at the level of cell function; they have been found to be localized to the major microtubule organizing center in their respective cells, i.e. the basal body complex in flagellated green algae (11, the centrosome in human and other mammalian cells (5, 61, and the spindle pole body in yeast (10). Genetic data indicate that caltractin in Chlamydomonas and the CDC3lgp in yeast are required for the normal duplication and segregation of the microtubule organizing center in the respective cells (11, 12).
In algal cells, caltractin is most prominently localized to calcium-sensitive contractile fibers, the striated fiber roots that connect the basal body complex t o the underlying nucleus in interphase cells (for reviews, see Refs. 13,14). It has been suggested that caltractin is a structural component of the cytoskeletal framework of fine filaments that compose these striated fibers (15).
One of the remarkable characteristics of the calmodulin superfamily of calcium-binding proteins is an extensive structural homology (16,17). Like calmodulin and troponin C, caltractin contains four potential calcium-binding sites of predicted helix-loop-helix or EF-hand structure (18). Furthermore, the Ca2+-binding loops in caltractin are in exact sequence register with those found in calmodulin, with the NH,-terminal domain (Ca2+ sites I and 11) separated from the COOH-terminal domain (Ca" sites I11 and IV) by an %residue central helix or linker. Algal and mammalian caltractins, as well as the yeast CDC3lgp, are distinguished from calmodulin and troponin C by the presence of NH,-terminal extensions of different lengths (from 17-24 residues) and distinctive amino acid sequences (see Refs. 4,9).
Because of caltractin's clear evolutionary relatedness to calmodulin, its unique association with a calcium-sensitive filamentous contractile system which appears to be distinct from actomyosin-based contractile systems, and its specific association with the microtubule organizing center in different organisms, we have been interested in studying its biochemical and biophysical properties. Our long term objectives are to integrate structural studies using NMR spectroscopy with genetic and molecular analyses to define at high resolution how calcium modulates caltractin's functional properties. In this report, we describe procedures to obtain high level expression of Chlamydomonas caltractin in bacteria, methods to rapidly purify the protein, and characterization of some of its biochemical and biophysical properties. We have obtained evidence that purified recombinant algal caltractin, in contrast to other related calcium-binding proteins, is highly stable in the apo state and is, therefore, uniquely suited for structural determination by the application of multidimensional NMR spectroscopy.

MATERIALS AND METHODS
Cloning-The Chlamydomonas caltractin cDNA, pCaBP-4 (2), was subcloned into the bacterial expression vector pKK 223-3 (Pharmacia 15795 Expression and Characterization of Caltractin LKB Biotechnology Inc.). The nucleotide sequence GAAAAC preceding the initiating ATG codon in pCaBP-4 was converted into an EcoRI site using a mutagenic synthetic oligonucleotide according to the method of Zoller and Smith (19). The mutagenized cDNA insert was subcloned into the unique EcoRI cloning site in the pKK 223-3 vector. The plasmid pKKSB, containing the caltractin cDNA inserted in the sense orientation was introduced into the Escherichia coli hosts Y1090 (ApMBlO9) and BL21 using standard methods (20). Inducible expression in these hosts was obtained by cointroduction of the plasmid pMS421 (kindly provided by Dr. Steve Libby), a pGB2 derivative harboring the lacP fragment inserted into the EcoRI site (21). The plasmids were maintained by growth of the transformed host cells on media containing 100 mg/liter ampicillin or carbenicillin and 25 mg/liter spectinomycin or streptomycin. Expression-For high yield expression runs, carbenicillin was used at a concentration of 500 mg/liter. Growth media used included the following: M9 medium supplemented with 0.01% thiamine and trace element solution (22); LB; Superbroth (20); and Celtone (Martek Inc., Columbia, MD) supplemented with 5 g/liter glucose. The "N-labeled samples were produced using a variation of the M9 medium with 99%-"NH,Cl(l g/liter) as the sole nitrogen source. All media were titrated to pH 7.4 a t room temperature. Host cells harboring the two plasmids pKK9B and pMS421 were grown at 37 "C in Erlenmeyer or Fernbach flasks on a rotatory shaker (250-300 revolutiondmin) to an OD,,, between 2 and 4 (typically attained in 6-8 h). IPTG' was then added to a final concentration of 1 m M and the cells were left growing for an additional 8-20 h. Alternatively, one volume of fresh growth medium including antibiotics and containing 2 m M IPTG was added to promote further growth, re-establish selection pressure, and induce caltractin expression. Another variation of the growth protocol involved usage of non-inducible strains with the same total growth period. Cells were harvested by centrifugation and were either immediately lyzed or frozen as cell pellets at -20 "C for later purification of caltractin.
Cell density was assessed by measuring the optical density at 600 nm. Cell morphology and viability was inspected in 1:lO dilutions of the culture in fresh medium or phosphate-buffered saline at x 400 magnification using phase contrast optics. Other parameters routinely monitored included titer of plasmid-bearing cells at induction and at the end of the expression r u n , pH, odor, and color of the medium and amount of foam. Expression levels were judged either by SDS-PAGE of total cell lysates or by the final yield of purified caltractin.
Purification-Cell pellets were resuspended in 2.4 M sucrose, 40 m M Tris-HC1, pH 8.0, 10 m M EDTA at 2.5 mug wet cells. 2.5 volumes of 50 m M Tris-HC1, pH 7.5, 1 m~ EDTA, 0.1 m M dithiothreitol, 0.2 m~ phenylmethylsulfonyl fluoride containing 1 mg/ml lysozyme were then added, and the suspension was left stirring on a magnetic stirrer for up to 3 h at room temperature. Insoluble material was removed by centrifugation at 50,000 x g for 30 min a t 4 "C, and the supernatant was brought to 0.5 M NaCl and 5 m M CaC1,. A trace amount of DNase I was then added, and the solution was stirred at room temperature for 10 min and centrifuged again at 50,000 x g for 30 min at 20 "C.
The supernatant was filtered through a glass prefilter and a 0.2-pm membrane, then applied to a 1.6 x 10-cm phenyl-Sepharose column (Pharmacia) equilibrated with column buffer (50 m~ Tris-HC1, pH 7.5, 0.5 M NaCl, 1 m M CaC1,). Unbound material was washed from the column with a minimum of 10 volumes of column buffer, followed by a minimum of 10 volumes of column buffer lacking sodium chloride at a flow rate of 4 mumin. Caltractin was eluted with 50 m M Tris-HC1, pH 7.5, 1 m M EGTA at a flow rate of 1 mumin. The column was then regenerated by washing with two volumes of 6 M urea and back-equilibrated into the starting buffer. The caltractin pool of the previous step was applied directly to a 1 x 10-cm Mono Q column (Pharmacia) at 4 Mmin. Caltractin was eluted with a linear gradient of 150-300 m M NaCl(15 column volumes) in 50 m M Tris-HC1, pH 8.0. The column was washed with 2.5 volumes of 1 M NaCl in the same buffer and re-equilibrated to starting conditions with five volumes of 50 m M Tris-HC1, pH 8.0.
For gel filtration experiments, 1-5 ml of concentrated caltractin solution, typically obtained by ultrafiltration using Y " 2 membranes (Amicon, Beverly M A ) were applied to a 1. Caltractin solutions were either sterile filtered, or 0.02% sodium azide was added, before storage at 4 "C or -20 "C.
Spectroscopy-UV spectra were measured on a Shimadzu 170 or a Beckman Lambda 4B spectrophotometer in semimicro quartz cells with 1-cm path length at room temperature. CD spectra were measured on an Aviv 60 DS spectrophotometer in 2-mm quartz cells at 27 "C. Experiments were run in Tris buffer at pH 7.4 in the presence of 1 m M EDTA with caltractin concentrations of 0.5-5.0 mg/ml. NMR spectra were acquired from 1 m~ samples of caltractin in 20 m~ Tris-d,,-HC1, pH 6.5,200 m~ KCl. In most cases, spectra were measured a t 303 K on a Bruker AMX 500 spectrometer. Two-dimensional 16N-lH HSQC spectra (23) were acquired with 16 scandincrement, a total of 128 complex points in t, and 2048 points in t, with spectral widths of 1115 Hz in F1 and 6580 Hz in F2. The total acquisition time was 80 min. Nonspecific Ca2+ binding was clearly observed in the Ca*+-sensitive electrode experiments at very high levels of free Ca2+ M). Instead of reaching a plateau value, the bound Ca2+ concentration increased linearly with the free Ca2+ concentration in this regime. This nonspecific binding phenomenon was also seen in control experiments on chicken gizzard calmodulin. Nonspecific binding to the protein or the electrode at such high levels of free Ca2+ is not uncommon (25).' This phenomenon was factored into the analysis of the binding curve by addition of a linear term to the binding equation (see "Results"). For display purposes, this linear nonspecific binding term is subtracted directly from the raw data fitted by non-linear regression analysis.
Equilibrium dialysis experiments were performed essentially as described by Potter et al. (26) except that, since the free Ca2' concentrations tested in these experiments were so high (1.00 x and 3.16 x M), Ca2+ buffering was not necessary. Attempts were also made to determine the binding constant using Ca'+-sensitive indicators (BAPTA, Br,-BAPTA, and calcium green coupled to dextran). However, all of the indicators tried changed their spectral characteristics in the presence of caltractin in a Ca2+-dependent manner. Hence, this method proved not to be reliable for quantitative measurements of free Ca2+, and correspondingly, for the determination of the Ca2+-binding constants of caltractin.

Cloning
With the objective of obtaining large quantities of caltractin for biophysical characterization and structurdfunction studies, a cDNA containing the entire coding sequence for Chlamydomonas caltractin was cloned into the bacterial expression vector pKK 223-3, previously shown to provide for the expression of recombinant calmodulin (27). The cDNA was engineered so that the initiating ATG codon was 10 bases downstream from a Shine-Delgarno sequence, and expression of the protein placed under the control of the hybrid trp-lac (tad promoter.

Expression
Initially, the expression vector with the caltractin cDNA was used to transform the protease-deficient (Zon-) strain Y1090 S. Linse and S. Forsen, personal communication.  (1) and the immunoreactive bands detected with an alkaline phosphatase-conjugated secondary antibody.

Expression and Characterization
(kindly provided by Dr. Fred Heffron) which had been cured of the pMC9 plasmid to allow for the selection and maintenance of cells containing the caltractin cDNA. Although induced expression could not be obtained with this strain because it no longer bears the lacIq repressor, constitutive expression of caltractin was found not to be detrimental to the bacteria. As seen in Fig.  1, bacteria transformed with the caltractin expression plasmid (only with the caltractin cDNA inserted in the sense orientation) expressed a protein which migrated in SDS gels with an apparent molecular mass of 20 kDa, and based on densitometric scans of Coomassie Blue-stained gels it represented over 10% of the total bacterial cell protein. The bacterially expressed protein was found to be specifically recognized by a rabbit antibody raised against native algal caltractin ( Fig. 1).
In initial studies on recombinant caltractin purified from Y1090 cells, two forms of the protein were detected by native PAGE and resolved from each other by Mono-Q ion-exchange chromatography (results not shown). NH,-terminal sequencing indicated that one of the species was essentially full-length, lacking only the cDNA-encoded amino-terminal methionine residue. The initiating methionine was apparently quantitatively cleaved in the bacterial host, leaving an unblocked serine-2 residue at the NH, terminus. The other protein species was found to be lacking the cDNA encoded NH,-terminal 6 amino acids (MSYKAK), with threonine 7 at the NH, terminus. The ratio of the two species was found to be dependent on the growth phase of the E. coli cells at the time of harvest and the temperature and duration of the cell lysis. These observations suggested that caltractin expressed inY1090 was susceptible to being clipped by an endogenous protease with a trypsin-like activity. This problem was alleviated when the expression plasmid pKK9B was introduced into the o m p T E. coli strain BL21. Recombinant caltractin purified from this host was found to be homogeneous and unclipped and not affected by the temperature or time of incubation during cell lysis. These results, coupled with the observation that BL21 cells grew very well in all media tested (see Table I), led us to select BL21 cells as the preferred host for caltractin expression.
With the demands of a multidimensional NMR study in mind, namely, multiple isotope-enriched protein preparations of tens of milligrams each, it was necessary to optimize caltrac- tin expression. The goal was simple and reliable production of 50-100 mg of pure caltractin from a minimal volume of a medium that can be completely enriched with either "N, 13C, or both.
To maximize both the volumetric yield of caltractin and the efficiency of nutrient utilization, caltractin was purified from BL21 cells grown to saturation. As seen in Table I, the protein yield scaled almost linearly with the cell mass. No differences were detected in the integrity of the protein isolated from cells grown to stationary phase versus the protein produced by cells harvested at mid-log phase. Protein integrity was monitored by PAGE in the presence of SDS and under native conditions, as well as UV and NMR spectroscopy. The consequences of constitutive versus induced expression of caltractin on protein yield was also investigated. Induced expression was obtained by cotransformation of BL21 cells bearing the pKK9B plasmid with pMS421, a pGB2 derivative carrying the lacIq repressor.
IFTG-induced expression for up to 16 h in high density cell cultures gave overall caltractin yields comparable to that obtained with constitutive expression of the protein. However, while the growth characteristics of uninduced BL21 pKK9B pMS421 cells resembled that seen for wild-type host cells, constitutively expressing BL21 pKK9B cells exhibited a very long lag phase of up to 8 h, and grew at a slower rate than the BL21 parental strain. As a consequence, overgrowth of cells lacking plasmid was more frequently observed for the constitutive expression system, at times resulting in lower overall caltractin yield. Inducible expression was found, therefore, to be generally more reliable and reproducible.

Purification
Recombinant caltractin expressed in E. coli was found to be completely soluble, independent of the host, growth regime, and induction time employed. Complete lysis of the host cells was achieved using lysozyme a t room temperature over 1 h or more. For optimal results in the subsequent chromatography steps, it was essential to clarify the lysate well and lower its viscosity. In this respect, we found lysis by lysozyme to be more suitable than sonication because the particles generated were easier to remove by centrifugation. Solubilized DNA could be partially pelleted with the cell debris, and the rest was digested with DNase I. Remaining particles were removed by filtration. These procedures yielded lysates with sufficiently low viscosity to avoid matrix compression during column chromatography at the relatively fast flow rates employed.
The purification procedure implemented for recombinant caltractin followed a protocol similar to that previously used for purifying native Chlamydomonas caltractin (1) and other EFhand type Ca2+-binding proteins (28). It involved, as a first step, calcium-dependent affinity chromatography on the hydrophobic resin phenyl-Sepharose and, subsequently, Mono-Q ion-exchange chromatography. The purification protocol was optimized with respect to yield and speed. Most notably, flow rates were increased to the maximum that the given matrix or column could withstand, and chromatographic procedures were developed to reduce to a minimum the number of buffer changes required.
As seen in Fig. 2, essentially all of the caltractin present in the soluble fraction of a total cell lysate bound to the phenyl-Sepharose matrix in the presence of Ca2+ and application of EGTA-containing buffer to the column effected elution of nearly pure caltractin (Fig. 2, lane 5). The contaminating species at this point were not easily visualized on gels routinely stained with Coomassie Blue, but due to the peculiar nature of caltractin's UV spectrum (see below), they contributed substantially to the background absorption at 280 nm, but much less so a t 260 nm. We, therefore used the ratio of the absorbance at 260 and 280 nm (A,,,jA,,,) as a sensitive indicator of caltractin purity. TypicallyA,,,jA,,, was 1.4-1.5 at this point. Minor protein contaminants were resolved from caltractin by applying the EGTA-eluted peak from phenyl-Sepharose directly onto a Mono-Q column. Caltractin eluted from the Mono-Q column as a relatively broad peak at approximately 250 mM NaC1. In general, this ion-exchange chromatography step removed all remaining contaminating material as judged by silver staining of SDS gels. In experiments in which silver staining revealed the presence of residual protein contaminants, these were eliminated by a second round of Ca2+-dependent affinity chromatography on phenyl-Sepharose. Preparations of pure caltractin gave an absorbance ratio A,,,jA,,, of 1.61 k 0.09. Protein integrity and purity was also assayed by NMR spectroscopy because it provides a large number of reporter sites distributed uniformly throughout the protein that are extremely sensitive to structural perturbations. Virtually any change in the protein can be detected, manifested either as an alteration of resonance frequencies or the appearance of new peaks. The sensitivity of the NMR method is high because chemical shift changes of only a few hundredths of a ppm can be readily detected and the detection limit for new peaks is in the range of 1-3%. To monitor stability, an 15N-labeled apocaltractin NMR sample was stored a t 4 "C in 20 mM Tris-HC1,200 mM KCl, 0.5 mM EDTA, 0.02% NaN, at pH 6.5. This sample gave identical 15N-lH HSQC spectra directly after preparation and 8 months later. From this and other similar experiments, we concluded that caltractin prepared in the manner described above remained stable for long periods of time in both the apoand Ca2+-loaded states. The NMR method was also used to establish that our caltractin samples were >98% pure, that the protein was pH stable in the range 6.0-8.0, and that the structure was not affected by the presence of up to &fold excess EDTA.
It has been previously documented that native caltractin isolated from Chlamydomonas (1) shows a minor, but distinct, calcium-dependent alteration in its electrophoretic mobility in the presence of SDS, a characteristic of calmodulin and other members of the EF-hand calcium-binding protein family (29).

43-
As seen in the top panel of Fig. 3 (lanes 1 and 2), purified recombinant caltractin was found to migrate with an apparent molecular mass of 20 kDa in the presence of added CaCl, and was slightly more retarded in its migration in the presence of added EGTA. In the same gel (Fig. 3, top panel, lanes 3 and 41, a more pronounced calcium-dependent alteration in electrophoretic migration was detected for chicken gizzard calmodulin, from an apparent molecular mass of 17-21 kDa. Although only a minor calcium-dependent shift in electrophoretic mobility was detected for caltractin in the presence of SDS, under non-SDS, native electrophoretic conditions, caltractin exhibited a more distinct mobility shift, comparable to that observed for calmodulin (Fig. 3, bottom panel 1. In addition to its apparent calcium-binding properties, caltractin expressed in E. coli was found to be heat stable, a property that it also shares in common with calmodulin and other members of the calcium-modulated family of proteins (30). As seen in Fig. 4, when a soluble fraction from a bacterial lysate containing caltractin was heated for 5 min to 95 "C, followed by centrifugation to pellet heat-denatured proteins from soluble heat-stable proteins, caltractin was quantitatively recovered in the heat-stable supernatant fraction.
Characterization Gel Filtration-Purified recombinant algal caltractin was found to have a molecular mass of 20 kDa as determined by amino acid analysis and electrophoretic migration in the presence of SDS. However, it typically eluted from a Superdex 200 column a t 0.70 column volumes as a symmetric peak, which corresponded to an apparent molecular mass of 30 kDa based on a calibration of the column with globular proteins (Table 11). Calmodulin was found to elute from the same gel filtration column a t nearly the same point as caltractin (Table 11). These   results suggested that caltractin was likely to be an elongated molecule with a domain architecture similar to calmodulin in which the long axis of the molecule determined its behavior on a sizing column.
The elution profile of caltractin on the gel filtration column was dependent on the Ca2+ content and ionic strength of the buffer used (Table 11). In the presence of 200 mM NaCI, the Ca2+-loaded and the Ca2+-free states of caltractin eluted a t very similar volumes. In the absence of salt, both forms behaved as markedly larger molecules, the Ca2+-free form a s a 44-kDa particle, the Ca"-loaded form as a 33-kDa particle. A similar trend was observed by NMR (see below) and has been interpreted as transient association of two or more caltractin molecules.
Spectroscopy-Due to the lack of Trp and Cys residues and the high 9:l ratio of P h e w , algal caltractin exhibited characteristic W spectrum with distinct shoulders a t 274 and 283 nm from the single Tyr and the fine structure typical of Phe, including peaks a t 252, 258, and 264 nm and a shoulder at 267 nm (Fig. 5). This is comparable to invertebrate calmodulins (311, which have very similar amino acid compositions. The extinction coefficients in the near W are given in Table 111.
They were in general very much lower than for a typical globular protein, again a characteristic associated with caltractin's specific amino acid composition.
In contrast to calmodulins (311, there was no detectable change of the W spectrum upon Ca2+ binding. The detection limit for a spectral change was 21% of the full amplitude. We attribute this observation to two factors. (i) Calmodulins usually contain several tyrosines, including a critical reporter resi- 'Absorbance at 1 mg/ml. The caltractin concentration was determined by quantitative amino acid analysis and the BCA protein assay (Pierce Chemical Co.), using bovine serum albumin as the standard. due in the third EF-hand, whereas the single tyrosine in caltractin is found in position 3 of the sequence, far away from any of the four EF-hands. Thus, little change in tyrosine absorbance is expected upon Ca2+ binding. (ii) The 9 phenylalanine residues are dispersed over the whole sequence, but due to their low extinction coefficient, a Ca2+-dependent change in absorption is expected to be fairly weak and thus difficult to detect.
The CD spectrum of algal caltractin has the typical features of a protein rich in a-helix, namely, minima a t 207 and 222 nm (results not shown). An estimate of 65% a-helix is obtained using the method of Chang et al. (32). This value is in excellent agreement with the 70% a-helix content predicted by the Robson algorithm (33) from the deduced amino acid sequence of Chlamydomonas caltractin (2). In contrast to calmodulin and troponin C, there was no change observed in the CD spectrum of caltractin upon addition of Ca2+.
The 'H NMR spectrum of caltractin contained several features typical of an EF-hand Ca2+-binding protein. The chemical shift dispersion seen in the one-dimensional 'H NMR spectrum, although not large, clearly indicated a folded globular protein (Fig. 6). The small number of low field shifted C"-proton resonances and the fairly limited chemical shift dispersion point to a high proportion of residues in either a-helical or "randomcoil" conformation and a distinctively low amount of p-sheet secondary structure.
In the Ca2+-free state, caltractin appeared to exist as a monomer in 1 mM samples containing 200 mhl KC1 and yielded NMR spectra of comparable quality to calmodulin. As an example, we show an I5N-'H HSQC spectrum of uniformly '5N-enriched caltractin (Fig. 7), which can be compared with Fig. 4

al. (34).
Upon Ca2+ loading of caltractin, significant reorganization of the spectrum and substantial increases the were observed (Fig. 6 ) . This effect was even more pronounced at low ionic strength (data not shown). The linewidths of the Ca2+free protein at 200 mM KC1 were in the range expected for a protein of its size, whereas the linewidths of the Ca2+-loaded protein were more consistent with a dimer or larger aggregate.  Only upon saturation with Ca2+ did the lines sharpen up again, although they did not return to the same linewidth as was seen for the Ca2+-free protein (Fig. 6 ) .
There were a number of distinct chemical shift changes detected in the 'H NMR spectrum of caltractin upon Ca2+ binding. Four amide proton resonances were present at the very low field edge of the spectrum of Ca2+-loaded caltractin (inset in Fig.  6 B ) , but only the signal at 11.3 ppm was detected in the apoprotein (inset in Fig. 6A). By analogy to other EF-hand Ca2+binding proteins, these signals could be tentatively assigned to the residues at the sixth position of the four Ca2+-binding loops (37,381, which are in caltractin the canonical Gly in loops I, 11, and 111, but Asn in loop N (see Table IV). In the aromatic region (6.5-7.5 ppm), the envelope of the resonance lines was shifted upon Ca2+ binding. Likewise, there were changes in the aliphatic part of the spectrum, particularly in the high field shifted resonances around 0 ppm. All of these changes were indicative of a structural and/or dynamic change due t o Ca2+ binding and involving at least a modest rearrangement of the hydrophobic core of the protein, a previously documented characteristic of calmodulin and troponin C (39, 40).
Calcium Binding Properties-Chlamydomonas caltractin contains in its deduced amino acid sequence four potential Ca2+-binding sites (2). To obtain some information on Ca2+ affinities of these sites, purified recombinant caltractin (50 p~ in 50 m~ HEPES, 75 mM KC1, pH 7.50) was titrated with a Ca2+ standard solution, and the free Ca2+ concentration was measured with a Ca2+-selective electrode following the method described by Linse et al. (41). The titration curve is presented in Fig. 8. To demonstrate that the low affinity nonspecific Ca2+ binding in the Ca2+ electrode experiments was not due to binding to caltractin, two points on the binding curve were measured independently by equilibrium dialysis. As can be seen in Fig. 8, the corrected binding curve fell well into the range obtained by the equilibrium dialysis control experiments. An inflection point was observed in the binding curve a t two bound Ca2+ ions, which suggested two discrete binding events: two Ca2+ ions bound to caltractin at approximately M free Ca", and two additional Ca2+ions bound a t approximately M free Ca2+. Thus, all four EF-hands of caltractin appeared to contain  (Table V). The binding constants of caltractin were also measured in the presence of 5 mM Mg2+, and as expected, the Ca2+ affinity of both classes of binding sites was reduced (Fig. 8, Table V). Calcium-magnesium antagonism is a characteristic feature of EF-hand calcium-binding proteins and has been well documented for calmodulin (43). The uncertainty in the caltractin-binding constants was higher for the low affinity sites than for the high affinity sites due to the onset of non-linear effects on the Ca2+-sensitive electrode. Fitting to analogous models with four independent sites or to the Hill equation did not improve the fit as judged visually and by statistical analysis, and also did not result in any substantial changes in the derived binding constants.
To ensure that we were obtaining accurate binding constants, control experiments were carried out on chicken gizzard calmodulin (Table V). The corresponding dissociation constants (1.0 x M and 1.5 x M) were in excellent agreement with the values reported by Iida and Potter (25) for bovine testis calmodulin, measured under similar conditions and fit with a two higher affinity/two lower affinity binding site model. DISCUSSION The high level expression in E. coli and biophysical characterization of recombinant Chlamydomonas caltractin has been described. The physical properties of the protein clearly confirm that caltractin is a member of the EF-hand family of Ca2+binding proteins, as was first proposed on the basis of its homology in amino acid sequence with calmodulin and troponin C (2) and from the biochemical characterization of native caltractin from Chlamydomonas (1). The gel filtration results indicate that, like other members of this protein family, caltractin is not a spherical molecule but rather elongated. The CD and NMR results show that it is largely helical and that upon Ca2+ binding there is virtually no effect on the secondary structure. However, distinct conformational and/or dynamical changes associated with reorganization of packing of the helical elements upon Ca2+ binding were evident in the NMR experiments. Finally, the binding of four Ca2+ ions was demonstrated with two higher affinity and two lower affinity sites. The two higher affinity sites fall midway between resting and maximal Ca2+ levels of most cells (44) and are, thus, likely regulatory sites. The low values of the binding constants for the other two sites indicate that they would not be fully occupied in uiuo. However, it is known that the affinity of calmodulin and other calciumbinding proteins for Ca2+ is increased by interactions with target proteins (45); a property that caltractin may also share.
In an effort to identify which of the calcium-binding sites in caltractin are likely t o be the higher and lower affinity sites, the four binding loop sequences have been analyzed for relative homology to the Strynadka and James (16) consensus sequence for EF-hand calcium-binding sites (see Table IV). Sites I and I1 of algal caltractin meet all of the criteria for classification as standard EF-hands. The only deviation from the canonical sequence in both of these sites is the presence of Ser, rather than the more common Asp or Asn, at position 5. This substitution is not likely to affect high affinity calcium binding since a number of EF-hand calcium-binding proteins, in particular, nearly all parvalbumins (16,17), have a Ser at this position and exhibit high Ca" affinities. In contrast, binding loops I11 and IV in algal caltractin deviate from the consensus sequence at one or more positions. The Ca2+ affinity of site 111 is likely to be substantially lower than a consensus site because the highly conserved Glu at position 12 is replaced by Asp. Site-directed mutagenesis experiments have demonstrated the importance of Glu at position 12 for calcium binding (31,46). In Site IV, there are several deviations from the consensus sequence. The replacement of the conserved Gly at position 6 with Asn is predicted to have the strongest consequence on calcium binding. The conserved Gly at this position allows the polypeptide chain to make a sharp turn to position the backbone carbonyl of the adjacent 7th residue for chelation of the ion (16). With Asn in position 6, the Glu in position 7 may still be able t o provide a Ca2+ ligand, but it is likely that the affinity for Ca2+ in the site would be lower than that of a consensus EF-hand. The fact that the usual four low field shifted amide proton resonances appear in the NMR spectrum of caltractin upon Ca2+ loading (Fig. 6, insets) lends support to the view that loop IV binds Ca2+ like a consensus EF-hand, even if it is substantially weaker due to conformational strain. This comparative sequence analysis strongly suggests that the two higher affinity sites in caltractin are located in the NH,-terminal domain and the two lower affinity sites in the COOH-terminal domain.

Expression and Characterization of Caltractin
Although caltractin shares a number of biophysical and biochemical properties with calmodulin and other members of the superfamily, it does display several distinguishing features.
While the apoprotein at high concentrations behaves as a monomer, the purified recombinant protein has a strong tendency to aggregate in the calcium-loaded state. This was true under conditions close to physiological in which calcium-loaded calmodulin and troponin C were found to be monomeric. Also, unlike calmodulin and troponin C , there are no detectable changes in the UV or CD spectra upon calcium binding to caltractin, although there are large changes in the 'H NMR spectrum. In addition, Ca2+-sensitive dyes which have been successfully used as quantitative indicators of free Ca2+ concentrations in studies on many EF-hand Ca2+-binding proteins (41,47) were not useful in studies on caltractin. A Ca2+-dependent change of the spectral properties of the dyes in the presence of caltractin was observed. Taken together, our studies on the in vitro characteristics of recombinant algal caltractin serve to establish caltractin as a member of the calmodulin superfamily of Ca2+-binding proteins. In many respects, Ca2+ binding to the protein generates changes in the structure that are very similar to those associated with other members of this protein family. However, as detailed above, caltractin also is, in some respects, clearly distinct from other members of this protein family. These results underscore the classification of caltractin and related proteins into a separate subgroup based on sequence comparisons (6-9).
Because of their central roles in translating calcium signals into metabolic or mechanical response, as well as in uptake, transport, and homeostasis of Ca", the EF-hand calcium-binding proteins have been the subject of intensive investigation by high resolution structural techniques (x-ray crystallography, NMR spectroscopy in solution). The available evidence to date suggests that in calmodulin and troponin C, and presumably other EF-hand proteins with regulatory functions, the binding of Ca2+ leads to exposure of a series of hydrophobic residues to solvent. This "hydrophobic patch" is proposed to be a critical feature of the Ca2+-activated states of calmodulin and troponin C (48). A number of site-directed mutagenesis experiments and related results lend strong support to this hypothesis (49, 50). Clearly, direct determination of the structural response to Ca2+ binding is required t o firmly establish that this mode of "triggering" is operative. However, this information has been very difficult to come by due to the difficulty in crystallizing a specific Ca2+-binding domain in the Ca2+-free and Ca2+-loaded states. Furthermore, neither of the well-characterized proteins of the calmodulin superfamily, calmodulin and troponin C, appear to have a suitably well-defined apo state for high resolution NMR structure determination. The unique long term stability of apocaltractin makes it an excellent candidate for multidimensional NMR studies aimed a t determining its threedimensional structure in solution. This critical apo structure will allow the consequences of Ca2+ binding on caltractin to be determined, thereby providing new and important insights into the molecular basis for Ca2+ signal transduction.