Characterization of a Ca(2+)-binding site in human annexin II by site-directed mutagenesis.

Annexin II, a major cytoplasmic substrate of the src tyrosine kinase, is a member of the annexin family of Ca2+/phospholipid-binding proteins. It is composed of a short N-terminal tail (30 residues) followed by four so-called annexin repeats (each 70-80 residues in length) which share sequence homologies and are thought to form (a) new type(s) of Ca(2+)-binding site(s). We have produced wild-type and site specifically mutated annexin II molecules to compare their structure and biochemistry. The recombinant wild-type annexin II displays biochemical and spectroscopical properties resembling those of the authentic protein purified from mammalian cells. In particular, it shows the Ca(2+)-induced blue shift in fluorescence emission which is typical for this annexin. Replacement of the single tryptophan in annexin II (Trp-212) by a phenylalanine abolishes the fluorescence signal and allows the unambiguous assignment of the Ca(2+)-sensitive spectroscopic properties to Trp-212. This residue is located in the third annexin repeat in a highly conserved stretch of 17 amino acids which are also found in the other repeats and known as the endonexin fold. To study the precise architecture of the Ca2+ site which must reside in close proximity to Trp-212, we changed several residues of the endonexin fold in repeat 3 by site-directed mutagenesis. An analysis of these mutants by fluorescence spectroscopy and Ca(2+)-dependent phospholipid binding reveals that Gly-206 and Thr-207 seem indispensible for a correct folding of this Ca(2+)-binding site.

Characterization of a Ca2+-binding Site in Human Annexin I1 by Site-directed Mutagenesis* (Received for publication, January 16,1991) Carsten Thiel, Klaus Weber, and Volker GerkeS From the Deoartment of Biochemistrv. Max Planck Institute for Biophysical Chemistry, P. 0. Box 2841, 0-3400 Goettingen, " I Federal Republic of Germany Annexin 11, a major cytoplasmic substrate of the src tyrosine kinase, is a member of the annexin family of Ca2+/phospholipid-binding proteins. It is composed of a short N-terminal tail (30 residues) followed by four so-called annexin repeats (each 70-80 residues in length) which share sequence homologies and are thought to form (a) new type(s) of Ca2+-binding site(s). We have produced wild-type and site specifically mutated annexin I1 molecules to compare their structure and biochemistry. The recombinant wild-type annexin I1 displays biochemical and spectroscopical properties resembling those of the authentic protein purified from mammalian cells. In particular, it shows the Ca2+-induced blue shift in fluorescence emission which is typical for this annexin. Replacement of the single tryptophan in annexin I1 (Trp-212) by a phenylalanine abolishes the fluorescence signal and allows the unambiguous assignment of the Ca2+-sensitive spectroscopic properties to Trp-212. This residue is located in the third annexin repeat in a highly conserved stretch of 17 amino acids which are also found in the other repeats and known as the endonexin fold. To study the precise architecture of the Ca2+ site which must reside in close proximity to Trp-212, we changed several residues of the endonexin fold in repeat 3 by sitedirected mutagenesis. An analysis of these mutants by fluorescence spectroscopy and Ca2+-dependent phospholipid binding reveals that Gly-206 and Thr-207 seem indispensible for a correct folding of this Ca2+binding site.
Annexin I1 belongs to a recently defined family of Ca2+dependent membrane-and phospholipid-binding proteins (annexins) which are thought to participate in processes involving membrane fusion, membrane-cytoskeletal linkage, membrane-channel formation, and/or phospholipase Az inhibition (for recent reviews on annexins, see Refs. [1][2][3][4][5]. All annexins are built from segments of 70-80 amino acid residues. These are repeated either 4-(32-39-kDa annexins) or 8-fold (68-kDa annexin) along the polypeptide chain. The annexin repeats share sequence homologies within a particular protein and also between different members of the family. The homology is especially pronounced in a segment of 17 residues known as the "endonexin fold" (6), which is present * This study was supported in part by a grant from the Bundesrninisterium fur Forschung und Technologie and a Kekule stipend from the Fond der Chemischen Industrie (to C. T.). 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.
$ To whom correspondence should be addressed.
in each annexin repeat. Within the endonexin folds of all repeats characterized so far, a GT(D, N, or R) motif at positions 4-6 (relative counting of the 17 residues of the endonexin fold) and an arginine residue at position 17 are almost invariant. Limited proteolysis experiments revealed that binding sites for the common annexin ligands (Ca2+, phospholipid) map to a protease-resistant core domain which comprises the sum of the annexin repeats. In the primary structure this core is preceded by a protease-sensitive Nterminal domain (known as the tail) which is variable in sequence and length. In annexin 11, the N-terminal tail harbors the binding site for a unique protein ligand, p l l , and also contains the phosphorylation sites for protein kinase C and the src tyrosine kinase (for reviews see Refs. 7-9). Since both phosphorylation events occur i n vivo, it has been proposed that annexin I1 is involved in signal transduction during cellular growth and differentiation. Although 10 different annexins have been identified and sequenced so far, structural analyses have not revealed details as to the architecture of the Ca2+-and/or phospholipid-binding sites. While obvious sequence motifs known to form Ca2+binding sites, such as the helix-loop-helix structure (EF-hand) found in Ca2+-binding proteins like calmodulin and parvalbumin, are absent from annexin core sequences, it has been speculated that the endonexin fold forms a loop-helix structure involved in Ca2+-binding (6). The hypothetical assignment of a Ca2+-binding site to the endonexin fold is supported by Tb3+ fluorescence studies. Geisow et al. (6) suggested that a Tb3+-binding site (presumably identical to a Ca2+-binding site) in annexin IV (10) is in close proximity to a tryptophan residue within one endonexin fold. Similarly, Marriott et al. (11) used resonance energy transfer experiments to show that the sole tryptophan of annexin I1 (Trp-212), which is found in position 10 of th? endonexin fold in repeat 3, is located within less than 8 A of the Tb3+ and, by implication, Ca2+binding site. Fluorescence spectroscopy also revealed that the emission maximum from annexin I1 undergoes a pronounced blue shift upon Ca2+-(and Tb") binding (11)(12)(13). Although 1-2 mM Ca2+ are required to saturate this blue shift, the affinity for Ca2+ is increased more than 100-fold in the presence of phospholipid (14). For the annexin 11-pll complex, membrane and phospholipid binding is observed at micromolar (15) or even submicromolar (16, 17) Ca2+ concentrations, indicating that an annexin 11-membrane interaction can be expected at intracellular (i.e. submicromolar) Ca2+ levels.
Here we have employed the well-defined blue shift in the fluorescence emission maximum to characterize the molecular parameters of a unique Ca2+-binding site in annexin 11. Using site-directed mutagenesis, several amino acid substitutions were introduced in the endonexin fold of the third annexin repeat. Analysis of these mutants by fluorescence spectros-copy and Ca2+-dependent phospholipid binding reveals that Gly-206 and Thr-207 seem involved in the correct folding of the Ca2+-binding site present in the third repeat.

EXPERIMENTAL PROCEDURES
Annexin II cDNA Cloning and Expression in E. coli-A X@ 10 cDNA library prepared from HT29 (a human adenocarcinoma cell line) mRNA (kindly provided by Dr. D. Louvard, Pasteur Institute, Paris, France) was screened with synthetic oligonucleotides whose sequences were derived from the human annexin I1 cDNA (18). DNA from a positive phage clone was isolated and cleaved with EcoRI and XbaI. This treatment yields three annexin I1 cDNA fragments which are generated due to the internal EcoRI site (nucleotide position 915 of the protein coding region) and an XbaI site in the 3"untranslated region (nucleotide position 57 after the stop codon). The EcoRI and the EcoRI/XbaI fragments, which contained the entire protein-coding region of human annexin I1 plus 43 base pairs of 5'-and 57 base pairs of 3"untranslated sequence were gel-purified, ligated, and cloned into M13mp18 linearized with EcoRI and XbaI. M13 constructs containing the entire annexin I1 cDNA in the correct orientation were identified by restriction and sequence analysis. To facilitate cloning and expression, site-directed mutagenesis was employed to create a unique BarnHI site at the 5' end of the cDNA (position -11 with respect to the start codon) and to eliminate an internal HindIII site (position 343 of the coding region) without changing the amino acid sequence. Subsequently, a BamHI-Hind111 fragment containing the entire coding region was isolated from M13mp18 and cloned into the procaryotic expression vector pDSlO (19) which was linearized with BarnHI and HindIII. Following this strategy the cDNA insert is in reading frame with the AUG start codon of the plasmid. Transformation of Escherichia coli strain JM 101 with this construct resulted in high level synthesis of recombinant annexin I1 (10-50% of the total cellular protein). N-terminal protein sequencing revealed that the recombinant annexin I1 starts with the sequence MRGSFKMSTV. This represents the expected fusion of 6 residues (MRGSFK), which are encoded by the expression vector, to the annexin I1 N terminus.
Purification of Recombinant Annexin 11-E. coli JM 101 carrying wild-type or mutated annexin I1 expression constructs were grown for 14 h at 37 "C in 1.5 liters of LB medium containing 100 pg/ml ampicillin. Inclusion bodies were prepared following the method of Nagai and Thogersen (20) with some modifications. The lysis buffer contained 5 mM EGTA,' a Dounce homogenizer was employed to shear high molecular weight DNA, and a sonication of the lysed cells (3 X 20 s on ice, setting 3, Branson sonifier) was included. To reduce proteolysis all buffers contained 2 mM phenylmethylsulfonyl fluoride, 100 mg/liter ovumucoid (Sigma Chemical GmbH, Munich, Federal Republic of Germany (F. R. G.)), and 2.5 FM E64 ([~-3-trans-carboxyoxiran-2-carbonyl]-l-leu-agmatin; Peptide Institute, Osaka, Japan). Purified inclusion bodies containing the recombinant annexin I1 were resuspended in 80 ml of 8 M urea, 20 mM imidazole-HC1, pH 7.5, 20 mM NaC1, 1 mM Tris, pH 7.5, and the solution was clarified by sonication. Subsequently, this solution was rotated for 12 h at room temperature with 20 ml of Q-Sepharose (Pharmacia LKB Biotechnology Inc.) equilibriated in the same buffer. The unbound fraction was then incubated for 4 h at room temperature with 20 ml of CM-52 (Whatman) equilibrated in the urea buffer. Again annexin I1 remained in the unbound fraction. It was renatured by adjusting the protein concentration to 0.5 mg/ml and dialysis versus 20 mM imidazole-HC1, pH 7.5, 100 mM NaCl, 2 mM NaNs, 1 mM EGTA, 1 mM DTT. Correctly folded protein was separated from the insoluble residue by centrifugation at 100,000 X g for 30 min. The soluble fraction was adjusted to 25 mM sodium acetate, pH 5.6, and applied to a CM-52 column equilibriated in CM buffer (25 mM sodium acetate, pH 5.6, 2 mM NaN3, 1 mM DTT). The column was developed with a linear NaCl gradient (0-1 M) which led to the elution of annexin I1 at around 280 mM NaC1. Fractions containing annexin I1 were dialyzed against CM buffer and applied to a Mono-S fast protein liquid chromatography column (Pharmacia LKB Biotechnolgy Inc.) equilibrated in the same buffer. Pure annexin I1 was eluted with a salt gradient at 380 mM NaC1.
Site-specific Mutagenesis-Mutations were introduced in the annexin I1 cDNA by oligonucleotide-directed mutagenesis following the ' The abbreviations used are: EGTA, [ethylenebis(oxyethylenenitrilo)] tetracetic acid DTT, dithiothreitol; SDS, sodium dodecyl sulfate. method of Eckstein and co-workers (21). Oligonucleotides carrying the desired mutations were synthesized on an 8750 Milligen Biosearch DNA synthesizer and purified on denaturing polyacrylamide gels. In vitro mutagenesis was performed with a mutagenesis kit (Amersham Buchler, Braunschweig, F. R. G.) according to the manufacturer's protocol using the annexin I1 cDNA cloned into M13mp18 as singlestranded DNA template. DNA from recombinant plaques was analyzed by dideoxy sequencing (22) with T7-polymerase (Pharmacia-LKB, Uppsala, Sweden). Positive clones were amplified and the RF-DNA purified using Qiagen pack 500 (Qiagen Inc., Studio City, CA). After confirmation of the desired mutation by sequence analysis the replicative form DNA was cleaved with BamHI and HindIII (after the first round of mutations, which eliminated the internal HindIII site, HindIII was used to create the 3' end of the annexin I1 cDNA insert). The annexin I1 cDNA insert was purified by agarose gel electrophoresis and cloned into the pDS 10 expression vector as described above. Other cloning steps were carried out following standard procedures (23).
Spectroscopy-For UV difference and fluorescence spectroscopy, proteins were dialyzed against 1000 volumes spectra buffer (20 mM imidazole-HC1, pH 7.5, 100 mM NaCl, 2 mM NaN3, 1 mM DTT, 20 p M EGTA) and adjusted to a concentration of 0.2-2 mg/ml with a Centricon 10 microconcentrator (Amicon, Danvers, MA). All protein samples were centrifuged at 100,000 x g for 10 min prior to spectroscopical analysis. Absorption difference spectra were recorded as described (12) on a Cary 2200 spectrophotometer between 250 and 310 nm. Corrected, steady-state fluorescence emission spectra were recorded as described (11) on a SLM model 8000 spectrofluorometer (Urbana, IL) with the excitation wavelength set at 295 nm. The base line for spectra buffer alone was subtracted from each spectrum. In Ca2+-titration experiments, all spectra were corrected for the dilution obtained by the addition of CaClz solution. CD measurements were made at 20 "C on a Jobin Mark V with the protein samples in 20 mM sodium phosphate buffer, pH 7.2.
Phospholipid Binding-Liposomes were prepared by sonication of 1 mg/ml dioleoylglycerophosphoserine (kindly provided by Dr. H. J. Eibl, Max Planck Institute for Biophysical Chemistry, Gottingen, FRG) in water. Binding experiments were performed in 20 mM imidazole-HC1, pH 7.5, 100 mM NaCl at the Ca'+-concentrations indicated. Each reaction contained 5 pg of protein and 28 pg of liposomes in a total volume of 150 pl. After incubation at room temperature for 15 min, samples were centrifuged at 200,000 X g for 12 min. Bound protein was extracted from the liposome pellet at room temperature for 20 min with 150 p1 of buffer containing 10 mM EGTA instead of Ca2+. The fraction of bound and unbound proteins was determined quantitatively by SDS-polyacrylamide gel electrophoresis and subsequent densitometry of the Coomassie Blue-stained annexin I1 band.
Miscellaneous Techniques-a-Chymotrypsin treatment of annexin I1 and the different annexin I1 mutants was carried out in 20 mM imidazole-HC1, pH 7.5,lOO mM NaCl, 1 mM DTT, 20 FM EGTA with the protein concentration adjusted to 0.4 mg/ml at an enzyme/ substrate ratio of 1:lOO (by mass). After incubation at room temperature for the times indicated, the reaction was stopped by boiling the sample for 3 min in SDS sample buffer. In experiments where chymotryptic cores were assayed for liposome binding, the proteolysis was stopped by adjusting the sample to 1 mM phenylmethylsulfonyl fluoride. Trypsin digestion was performed in 20 mM imidazole-HC1, pH 7.5, 100 mM NaC1, 2 mM CaClz with recombinant annexin I1 at 0.6 mg/ml and an enzyme/substrate ratio of 1:lOO. Incubation was at room temperature and stopped after the given time intervals as described above. For N-terminal sequence analysis, tryptic fragments were separated in SDS-polyacrylamide gels and transferred electrophoretically to Immobilon membrane (Millipore). After staining with Amido Black protein bands were cut out and subjected directly to automated gas-phase sequencing. N-terminal protein sequence analysis was performed on an Applied Biosystems gas-phase sequenator (model 470A). Immunoblotting of different annexin I1 mutants was carried out as described (24) using either a polyclonal annexin I1 rabbit antiserum (25) or the mouse monoclonal antibody H28 (26). Recombi-c a n t H u m a n Annexin 11-A full-length cDNA clone for human annexin I1 was isolated from an HT 29 (an adenocarcinoma cell line) cDNA library made in Xgt 10. For cloning purposes, a unique BamHI site was introduced in the 5'nontranslated region (nucleotide position -11 with respect to the initiator methionine), whereas the internal Hind111 site (position 343 of the coding region) was destroyed by sitedirected mutagenesis without changing the amino acid sequence. Subsequently, the entire protein-coding region (contained in a BamHI-Hind11 fragment) was cloned into the procaryotic expression vector pDS 10. Transformation of E.

Expression and Biochemical Characterization of
coli with this construct leads to the efficient expression of recombinant human annexin 11, which is driven by the coliphage T5 promotor of pDS 10 (Fig. L4). When synthesized in bacteria human annexin I1 is insoluble, but can be extracted from the inclusion bodies by 8 M urea. Following renaturation (achieved by the protocol outlined under "Experimental Procedures") final purification was obtained by ion-exchange chromatography on CM-52 and Mono-S (Fig. lA). This approach routinely yields 10 mg of pure annexin I1 from 1 liter of bacterial culture. The recombinant protein is recognized by an annexin I1 rabbit antiserum (25) but not by the murine monoclonal antibody H28 (26). This antibody detects in immunoblots porcine, bovine, and chicken annexin I1 but not the murine or human protein. Based on a comparison of annexin I1 sequences from different species, it has been deduced that residue 65 represents an important contact site for the H28 monoclonal (glutamic acid in reactive annexin I1 molecules; valine or alanine in non-reactive proteins) (27). By introduction of glutamic acid in place of alanine a t position 65, which was achieved by site-directed mutagenesis, we confirmed this prediction. The A65E variant of human annexin I1 is clearly recognized by the H28 monoclonal (Fig. 1B). For subsequent studies, the A65E variant was used as the wildtype annexin I1 control; all human annexin I1 mutations discussed below also contain glutamic acid a t position 65.
The purified recombinant annexin 11, i.e. the A65E molecule, shows the same biochemical and physicochemical properties as the protein isolated from mammalian cells. It binds to phospholipids in a Ca"-dependent manner (cf. Table I), A displays a characteristic CD spectrum which reveals the CYhelical conformation (not shown) and shows Ca"-induced conformational changes as documented by fluorescence and UV difference spectroscopy (see below). In addition, the protein core of recombinant annexin I1 exhibits the characteristic stability toward mild proteolytic attack. Chymotrypsin treatment, for example, converts the bacterially synthesized protein into a 33-kDa species (Fig. 2 A ) . Direct protein sequence analysis revealed that this derivative starts at position 30 of the annexin I1 sequence (data not shown) and thus represents the typical chymotryptic core (28,29). Similarly, limited trypsin cleavage produces two major fragments of 20 and 15 kDa that have already been described for porcine annexin I1 (Fig. 2R) (30). The N-terminal protein sequence of the two tryptic fragments starts a t Ala-28 (20 kDa) and Lys-205 (15 kDa) (data not shown) again identical to the situation reported for annexin I1 purified from porcine intestinal epithelium (30).
Ca"-induced Conformational Changes in Annexin II Can Re Monitored by Fluorescence and UV Absorption Properties of the Single Tryptophan (Trp-212)"Annexin I1 contains a single tryptophan at position 212. Previous studies have interpreted Ca"-induced differences in UV absorption and fluorescence emission of annexin I1 solely in terms of this tryptophan (11,13). In particular, this chromophore seemed responsible for the blue shift in fluorescence emission and a number of negative UV difference peaks which are observed upon Ca'+ binding (for discussion of these vibronic structures, see Marriott et al., 1990). All these spectroscopical analyses point to a close proximity of a Ca2+ site and the single tryptophan which seems to reside in a highly non-polar environment (11). In order to unambiguously identify Trp-212 as the residue responsible for the effects discussed, we employed site-directed mutagenesis to introduce a phenylalanine in place of the tryptophan at position 212 (Figs. 3 and 5). The bacterially synthesized mutant protein was purified following the protocol developed for wild-type, i.e. A65E, annexin I1  (lanes 2-4) were synthesized in hacteria and purified as shown in panel A. Equivalent amounts were run in SDSpolyacrylamide gels, transferred to nitrocellulose, and analyzed by immunoblotting using an annexin I1 rabbit antiserum (lanes 1 and 2 ) or the H28 monoclonal antibody (lanes 3 and 4 ) . Note that the monoclonal only recognizes the A65E variant. Arrows mark the position of t.he annexin I1 (~3 6 ) polypeptide chain.

Ca"-dependent phospholipid binding of different annexin II mutants
Wild-type (A65E) annexin I1 and different repeat 3 mutants were assayed for binding to phosphatidylserine (PS) liposomes. Reactions were performed in buffers containing increasing Ca2+ concentrations and the fraction of bound and unbound protein was determined by SDS-polyacrylamide gel electrophoresis (see "Experimental Procedures"). The Ca'* concentrations required for binding to the PSliposomes were determined in several independent experiments for each mutant protein. To eliminate a potential influence of the Nterminal tail of recombinant annexin I1 all assays were carried out with protein cores generated by limited n-chymotrypsin treatment (7 min at room temperature, enzyme/substrate ratio of 1:lOO).

Mutants at half-maximal
Ca" conc.  (see above). Purified W212F displays biochemical properties indistinguishable from A65E, i.e. it binds phospholipids in a Ca"-dependent manner and retains a core of 33 kDa resistant to further chymotrypsin treatment (data not shown). However, spectroscopical studies reveal a fundamental difference between wild-type annexin I1 and the W212F mutant. While the wild-type molecule shows the typical negative UV difference peaks which are induced by the addition of Ca" (Fig.   4A), W212F fails to display any such structure. In the latter case, only a small negative UV difference is seen upon Ca'+ binding which stretches over almost the entire absorbing region (Fig. 4A). Thus, tyrosine residues must be responsible for this small UV difference, whereas the structured signal seen with the wild-type protein clearly reflects a Ca"-induced change in the environment of the single tryptophan. Similarly, the pronounced blue shift in fluorescence emission observed for annexin I1 upon Ca'+ binding can be assigned clearly to tryptophan 212. With excitation a t 295 nm, the emission of wild-type annexin I1 exhibits a maximum a t 321 nm, which is shifted to 311 nm upon addition of Ca'+ (Fig. 4B). When the identical experiment is performed with the W212F mutant, no significant fluorescence emission is observed, indicating that tryptophan 212 is the only chromophore in wildtype annexin I1 which is excited a t 295 nm (Fig. 4R). Thus, the single tryptophan in annexin I1 (Trp-212) resides in a rather hydrophobic environment in the absence of Ca'+ (fluorescence emission a t 321 nm) and becomes buried in an even more non-polar environment in the Ca'+-bound conformation (fluorescence emission a t 311 nm).

S 214 A ---------------A ----------
T o assess whether this phenomenon is unique for the tryptophan located in the third annexin repeat, we introduced a tryptophan nesidue in the same relative position of the endonexin folds in repeats 1, 2, and 4. All mutant proteins were constructed to contain a single tryptophan/molecule, i.e. the original tryptophan 212 was substituted by a phenylalanine (Fig. 5). The different mutant proteins were synthesized in bacteria and purified as described above. They displayed the same biochemical properties as wild-type annexin 11. Only the protein core of I56W showed a somewhat reduced resistance toward chymotrypsin attack (data not shown). Fluorescence emission spectra of the tryptophan mutants are given in Fig.  5. With excitation at 295 nm, the emission maxima are found at 320 nm (I56W), 326 nm (L127W), and 333 nm (L287W), respectively. Thus, the tryptophan residues situated in the same relative position of the endonexin fold in repeats 1, 2, 3, and 4 (position 10 of the fold) reside in different environments. Interestingly, the three tryptophan mutants show no or only a very minor Ca2+-induced alteration in the fluorescence emission spectra. While the spectra of I56W and L287W remain unchanged, the intensity of the fluorescence emission of L127W is slightly reduced in the Ca2+-bound conformation (data not shown). Thus, only the naturally occurring tryptophan of annexin I1 (Trp-212) resides in an environment that is clearly different in the Ca2+-bound and the Ca2+-free molecule.
Structural Characterization of the Ca2+-binding Site i n the Third Annexin Repeat-Since previous energy transfer studies had indicated a close proximity of a Ca2+-binding site and Trp-212 in annexin I1 ( l l ) , we chose the endonexin fold of the third repeat to study the effect of single amino acid substitutions on Ca2+ binding. The mutant proteins listed in Fig. 3 were purified from bacterial inclusion bodies and subjected to fluorescence spectroscopy with the excitation wavelength set at 295 nm. All spectra show maxima at around 321 nm in the absence of Ca2+ (Fig. 6). Thus, Trp-212 in the different repeat 3 mutants is located in a similar environment, indicating that all mutant proteins assume the correct conformation upon renaturation. This conclusion is also supported by the finding that the different mutants exhibit the same resistance toward proteolysis as wild-type annexin 11, i.e. a typical protein core is produced by mild chymotryptic €4 m treatment. However, a remarkable difference between the wild-type molecule and some of the repeat 3 mutants is seen when the fluorescence emission spectra are recorded in the presence of varying Ca2+ concentrations. While wild-type annexin I1 (Fig. 6) as well as the S214A protein (data not shown) display the typical blue shift in the fluorescence emission at free Ca2+-concentrations of 1-2 mM, the other mutant molecules (G206A, T207A, D208A, and D208N) require considerably higher Ca2+ levels for the same effect (Fig.  6). The strongest difference is seen with the T207A mutant. Here, even the addition of 20 mM Ca2+ to the protein solution is not sufficient to produce a pronounced blue shift.
The combined data on the Ca2+ titration of the fluorescence emission shift are summarized in Fig. 7. Four different types of mutations can be distinguished. 1) The S214A mutation, which has eliminated the hydroxyl function of the conserved serine (or threonine) residue usually found in position 12 of the endonexin fold, does not cause significant perturbations in Ca2+ binding. Less than 2 mM Ca2+ are required to shift the fluorescence emission maximum of the S214A protein from 321 nm (Ca2+-free conformation) to 312 nm (Ca2+-bound conformation). The Ca2+ titration curve of S214A is almost identical to that of wild-type annexin 11.2) The D208A protein is still able to bind Ca2+, albeit with reduced affinity. It requires 8 mM Ca2+ to display the blue shifted emission maximum, i.e. to assume the Ca2+-bound conformation. 3) The D208N and, in particular, the G206A mutant proteins show a markedly reduced affinity for the divalent cation. In both cases, more than 20 mM are necessary to establish a significant blue shift. (4) The T207A mutation, finally, causes the most severe effect. Even 20 mM Ca2+ is not sufficient to induce a significant blue shift in the emission spectrum of the T207A protein.
To obtain a set of independent data which allow a direct comparison of Ca2+ affinities of the different repeat 3 mutants, we performed a series of Ca2+-dependent liposome pelleting assays. In these experiments, a Ca2+ titration of the liposome binding of different annexin I1 mutants was employed to evaluate their relative Ca2+ affinities in comparison to the wild-type molecule. We chose this approach instead of Ca2+-  binding measurements by equilibrium dialysis since the latter method only revealed conflicting data for different annexins so far, e.g. annexins I and I1 were found to contain in the presence of phospholipids either four or only two Ca2+-binding sites with dissociation constants of 75 and 4.5 PM, respectively (14,33). Since the presence of phospholipid increases the affinity for the divalent cation by at least two orders of magnitude (14), Ca2+ levels in the micromolar range were employed in the liposome pelleting assay. Table I  concentrations for half-maximal phospholipid binding. However, even the T207A protein, which most likely contains an inactive binding site in the endonexin fold of the third repeat (see above), is still able to interact with the phosphatidylserine liposomes in a Ca2+-dependent manner.

DISCUSSION
Wild-type and site specifically mutated annexin I1 molecules were produced in E. coli to compare their structural and biochemical properties. The Ca2+-dependent phospholipid binding, proteolytic cleavage pattern, as well as physicochemical properties of the recombinant wild-type annexin I1 resemble those of the authentic protein purified from mammalian cells. Thus, the approach presented, i.e. a mutational analysis of annexin I1 synthesized in bacteria, is valid to study the structure of this particular annexin.
The replacement of the single tryptophan in annexin I1 (Trp-212) by a phenylalanine led to the unambiguous assignment of the Ca*+-sensitive spectroscopic properties to this tryptophan residue. Thus, our data confirm conclusions drawn in previous spectroscopical studies (11)(12)(13). With excitation at 295 nm, it is indeed the single tryptophan which absorbs energy and shows the characteristic fluorescence emission maxima at 321 nm in the absence and 311 nm in the presence of Ca2+. Using energy transfer experiments, Marriott et al. (11) located the Ca2+-binding site, whose occupation induces the describFd shift in the fluorescence emission maximum, to within -8A of Trp-212. By studying both the fluorescence properties and the Ca2+ requirements for liposome binding of different annexin I1 point mutants, we now show that the residues Gly-206, Thr-207, and Asp-208 seem involved in the formation of this Ca2+-binding site. However, with the experiments described here we are not able to distinguish whether ( a ) side-chain oxygen of Thr-207 and/or Asp-208 are coordinating the Ca2+ ion, ( b ) free carbonyl electrons of peptide bonds between amino acids 205 and 209 are involved in Ca2+ complexation, or (c) the GTD sequence (amino acids 206-208) is indispensible for the correct folding of the Ca2+ site. It seems likely, however, that at least the Asp-208 side chain is not directly involved in Ca'+ coordination since D208A shows only a mild defect in Ca2+ binding, whereas D208N (a mutant still containing a side chain with free electron pairs in position 208) is markedly impaired.
When our analysis was complete Huber et al. (31) reported the x-ray structure for human annexin V. Annexin V was shown to be an extraordinary compact molecule in which each annexin repeat is composed of five densely packed a-helices. Within the repeats each endonexin fold follows an a-helix (helix a ) and describes a short loop (residues 1/2-5/6 of the fold) followed by another a-helix (helix b, residues 5/6-17). Interestingly, the side chain of the hydrophobic residue in position 10 of the endonexin fold is surrounded by amino acids of different hydrophobicity in each of the repeats 1, 2, 3, and 4. This aspect of the annexin V structure is in line with our data on the different tryptophan mutants of annexin 11. Our fluorescence spectra reveal different emission maxima for I65W, L127W, L287W, and wild-type annexin I1 (Fig. 5), indicating that the tryptophan positioned as residue 10 of the endonexin fold clearly resides in different environments in the four annexin repeats.
Unfortunately, the conformation of the endonexin fold in repeat 3 of annexin I1 cannot be deduced by a simple extrapolation of the annexin V crystal structure. Despite a very good overall similarity, the primary and most likely also the secondary and tertiary structure of annexins I1 and V seem to diverge in the third repeat. In particular, the loop described by the first part of the endonexin fold and the two flanking helices are clearly different. While helix a, i.e. the helix preceding the endonexin fold, ends in an ELK sequence in annexin V, the corresponding region shows a cluster of basic residues (KRK) in annexin 11. In addition, the beginning of helix b, which is characterized by a row of acidic residues (DEE) in annexin V, reads DVP in annexin 11. Although these differences leave ambiguities in interpreting the annexin I1 structure in the third repeat, our data suggest that Gly-206 and Thr-207 are part of a loop likely to be involved in Ca'+ binding. This view is also supported by the finding that the peptide bond between residues Arg-204 and Lys-205, i.e. the 2 residues directly preceding Gly-206, is the only bond in the annexin I1 core which is susceptible to limited trypsin treatment (30). Similar loops which are described by the first residues of the endonexin fold are found in all other repeats of annexin V (31) and probably exist in annexin I1 as well. Thus, more than one Ca"-binding site can be expected in each annexin molecule.
Direct Ca"-binding studies revealed different numbers of binding sites for different annexins. In the absence of phospholipid, Owens and Crumpton (32) reported one high affinity site (Kd M) per annexin VI molecule which is composed of eight annexin repeats. If lipid was included in the Ca'+binding assay, two and four sites were found for the fourrepeat annexins I1 and I, respectively (14,33). Our data on the T207A mutant strongly indicate that annexin I1 contains more than one Ca2+-binding site. Although the T207A protein shows no Ca2+ binding in the endonexin fold of the third repeat as judged by fluorescence spectroscopy (cf. Fig. 7), the mutant is still able to interact with phospholipids in a Ca2+dependent manner (Table I). This phospholipid binding might well be mediated through (a) Ca2+ site(s) in the first and/or second repeat since proteolytically derived annexin I1 derivatives consisting of repeat 1 or repeats 1 plus 2 will bind to phosphatidylserine vesicles in the presence but not in the absence of Ca'+ (30). Similarly, a truncated annexin I1 molecule, which has been constructed by introducing a UAG (stop) codon at amino acid position 179, binds to liposomes in a

'-binding Site in Annexin II
Ca'+-dependent manner.' However, truncated and proteolytically shortened annexin I1 versions require considerably higher Ca'+ concentrations for phospholipid binding. Since these derivatives display an increased susceptibility toward proteases, the elevated Ca2+ requirement could reflect an altered conformation of an annexin repeat in the truncated molecules as compared to full-length annexin 11. This is actually known for the proteolytically derived fragments which display a reduced a-helix content as judged by CD spectroscopy (30). However, it also remains possible that the putative Ca'+-binding sites in the different repeats of annexin I1 either have different affinities or exhibit some sort of cooperativity. The latter interpretations draw support from our data on the repeat 3 mutants. T207A, for example, a mutant showing no Ca'+ binding in the third repeat but a Ca'+-dependent interaction with phospholipids, requires around %fold higher Ca2+ for liposome binding than wildtype annexin 11. This observation can be explained by two models. 1) The Ca2+ site(s) present in other repeats which must be responsible for mediating the lipid binding of T207A show a weaker affinity toward the divalent cation than the site in repeat 3.2) Cooperativity between different sites is the basis for higher affinity Ca'+ binding.
Cooperativity in the binding of different annexin ligands is observed when the affinity of the Ca2+ sites in annexin I1 is compared in the presence and absence of phospholipid. The relatively low affinity for the divalent cation (Kd -0.5 mM) is increased by two orders of magnitude if phosphatidylserine is included (14). Similar findings have been reported for various other annexins and most likely reflect a general property of the members of this protein family (for review, see Refs. [1][2][3][4][5]. Although the phospholipid-binding site(s) have not been mapped yet, fluorescence quenching experiments suggest that the single tryptophan of annexin V (Trp-187) is located at the protein-phospholipid interphase (34). Interestingly, this tryptophan is found in position 3 of the endonexin folds in repeat 3, i.e. in the loop between helices a and b in annexin V. In the annexin V crystal, this loop as well as the loops described by the other three endonexin folds reside in relatively close proximity on one side of the molecule (31). This configuration could explain the mutual influence of Ca'+ and phospholipid binding and might also be the basis for some cooperativity between Ca'+ sites. Future experiments have to reveal whether in the three-dimensional conformation of annexin I1 phospholipid and additional Ca2+ sites are indeed found in the vicinity of Gly-206 and Thr-207.