Histidine Phosphorylation of Annexin I in Airway Epithelia*

Although [Cl−] i regulates many cellular functions including cell secretion, the mechanisms governing these actions are not known. We have previously shown that the apical membrane of airway epithelium contains a 37-kDa phosphoprotein (p37) whose phosphorylation is regulated by chloride concentration. Using metal affinity (chelating Fe3+-Sepharose) and anion exchange (POROS HQ 20) chromatography, we have purified p37 from ovine tracheal epithelia to electrophoretic homogeneity. Sequence analysis and immunoprecipitation using monoclonal and specific polyclonal antibodies identified p37 as annexin I, a member of a family of Ca2+-dependent phospholipid-binding proteins. Phosphate on [32P]annexin I, phosphorylated using both [γ-32P]ATP and [γ-32P]GTP, was labile under acidic but not alkaline conditions. Phosphoamino acid analysis showed the presence of phosphohistidine. The site of phosphorylation was localized to a carboxyl-terminal fragment of annexin I. Our data suggest that cAMP and AMP (but not cGMP) may regulate annexin I histidine phosphorylation. We propose a role for annexin I in an intracellular signaling system involving histidine phosphorylation.

In epithelia, both [Cl Ϫ ] i and cell volume are tightly regulated during fluid secretion and/or absorption (1). Epithelia transport many times their own cellular mass of chloride and other ions, but it is unclear how cells sense [Cl Ϫ ] i . It is known that [Na ϩ ] i and [Cl Ϫ ] i interact in cell volume regulation (2). The multiple cellular functions regulated by [Cl Ϫ ] i (cystic fibrosis transmembrane regulator channel conductance, epithelial sodium channels, cell volume, Na ϩ K ϩ 2Cl-cotransporter activity, etc.) argue for a common regulatory pathway.
We have previously shown that [Cl Ϫ ]/[anions]/ATP/GTP differentially determine the profile of phosphorylation of a number of apical membrane proteins in airway epithelia (3,4). We identified two of the phosphoproteins, a 19/21-kDa doublet that showed chloride concentration-dependent phosphorylation, as two isoforms of nucleoside-diphosphate kinase (3,5). We also observed that phosphorylation of a 37-kDa protein (p37) from both human nasal and sheep tracheal epithelia was differentially modulated by nucleotides (GTP, ATP, and GDP), [Cl Ϫ ], and other anion species (3,4). We found that GTP was the principal phosphate donor for p37 from both species; and in the presence of ATP/GDP, phosphorylation of p37 was enhanced 20-fold. However, this phosphorylation could not be inhibited by known serine/threonine and tyrosine protein kinase inhibitors, suggesting that phosphorylation was not due to the activity of these enzymes.
In this study, our objective was to establish the identity of the ion-sensitive phosphoprotein p37 and to characterize its phosphorylated amino acid residue. By selectively enhancing phosphorylation of p37 (3) and by applying metal affinity chromatography, we show that p37 is identical to annexin I and provide evidence for novel annexin phosphorylation on histidine residues. Our data suggest that annexin is a component of an intracellular signaling system involving histidine phosphorylation, which is regulated by chloride concentration.

EXPERIMENTAL PROCEDURES
Sample Preparation-A membrane fraction was prepared from ovine airway epithelia as described previously (3). Briefly, tracheal epithelia were scraped and dislodged into homogenization buffer (250 mM sucrose, 10 mM triethanolamine; one protease inhibitor tablet from Roche Molecular Biochemicals per 50 ml buffer). Pooled scrapings were homogenized and spun at 600 ϫ g for 15 min. The post-nuclear supernatant was re-spun at 100,000 ϫ g for 2 h. The pellet was resuspended in homogenization buffer and spun for 30 min at 16,000 ϫ g (this procedure was repeated three times). All procedures were conducted at 4°C. Use of lactate dehydrogenase as a cytosol marker showed that the pellet contained no cytosol (3). Aliquots of the cytosol and membrane pellet were stored in liquid nitrogen.
Phosphorylation and Electronic Autoradiography-Phosphorylation and detection of phosphoproteins were performed as described previously (3). Briefly, proteins were phosphorylated with 37 kBq of [␥-32 P]ATP or [␥-32 P]GTP (final nucleotide concentration of 6.8 M, without the addition of unlabeled nucleotide) in 10 mM MOPS 1 (pH 7.9) containing 5 mM dithiothreitol. The reaction was terminated by adding 5ϫ Laemmli sample buffer (6), and the proteins were separated by SDS-12.5% polyacrylamide gel electrophoresis. Prestained molecular mass markers were used to avoid staining and destaining of gels prior to imaging and quantification. The incorporation of 32 PO 4 into individual protein bands was detected by electronic autoradiography (Canberra-Packard Instant Imager) and quantified as described previously (3,4).
Western Blotting-Proteins (10 g), separated by SDS-polyacrylamide gel electrophoresis, were transferred to polyvinylidene difluoride membrane (Millipore Corp.) by semidry electrophoretic transfer (Amersham Pharmacia Biotech) at 0.8 mA/cm 2 for 1 h with 20% methanol added to standard SDS-polyacrylamide gel electrophoresis running buffer. Prestained markers were used to confirm transfer. The primary antibody (1:2000) was an affinity-purified rabbit polyclonal antibody * This work was supported by Wellcome Trust Grant 0044854/Z/95/A, by Biomed II Network Grant BM H4-CT96-0602, and by grants from the Cystic Fibrosis Trust, Tenovus Tayside, the Deutsche Forschungsgemeinschaft, the Royal Society, and the Anonymous Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Immunoprecipitation-Membranes were resuspended in 100 l of 10 mM MOPS (pH 7.9) and phosphorylated with 67 nM [␥-32 P]ATP/GDP (500 nM) for 5 min at 37°C. The reaction was terminated with 50 mM EDTA, followed by the addition of 9 volumes of immunoprecipitation buffer (10 mM Tris-HCl (pH 7.4), 2 mM EDTA, 1 mM NaF, 1 mM dithiothreitol, 1% sodium deoxycholate, 1% Nonidet P-40, 0.3 M aprotinin, and 0.2 M phenylmethylsulfonyl fluoride). The mixture was precleared with protein G-Sepharose beads (30 min at 4°C) and centrifuged at 350 ϫ g for 5 min at 4°C, and the supernatant was incubated with either a rabbit polyclonal or mouse monoclonal antibody to annexin I (1 g) for 1 h at 4°C. New beads were added, and the mixture was incubated overnight at 4°C. The incubation mixture was centrifuged at 350 ϫ g for 5 min, and the pelleted beads were washed in 1 ml of radioimmune precipitation assay buffer (50 mM Tris-HCl (pH 7.4), 1% Nonidet P-40, 0.5% sodium deoxycholate, 5 mM EDTA, and 150 mM NaCl). This wash step was repeated three times, and then 50 l of 5ϫ sample buffer containing 100 mM dithiothreitol was added to the pellet, left at room temperature for 30 min, and finally spun at 420 ϫ g for 5 min. A 20-l aliquot was run on 12.5% polyacrylamide gels for analysis by electronic autoradiography.
Purification of Annexin I-The membrane fraction from sheep tracheal epithelia was solubilized with 0.25% glucopyranoside in 10 mM Tris-HCl (pH 7.4) containing a mixture of protease inhibitors (Complete TM protease inhibitor tablets, Roche Molecular Biochemical) and then centrifuged at 100,000 ϫ g for 30 min at 4°C. The supernatant fraction (100 ml, 1434 mg of protein) was incubated for 5 min at 37°C with 50 nM ATP and 500 nM GDP to enhance phosphorylation of the 37-kDa protein 20-fold as described previously (3). Following incubation, the temperature and pH of the sample (100 ml) were simultaneously adjusted to 4°C and 5.0, respectively, by adding 400 ml of buffer A (0.1 M acetic acid/NaOH (pH 5.0), 0.5 M NaCl, 0.4 mM 4-(2aminoethyl)benzenesulfonyl fluoride, and 0.1% glucopyranoside). The sample (500 ml) was then applied to a chelating Sepharose fast flow column (50 ml) charged with ferric chloride and pre-equilibrated with buffer A. The column was washed with buffer A (15 column volumes/ stable A 280 ϳ 0.01), and the bound phosphoproteins were selectively eluted with 20 mM sodium biphosphate in buffer A. The eluate (100 ml, 1.13 mg) was dialyzed against buffer B (20 mM Tris-HCl (pH 8.5) and 0.1% glucopyranoside) and applied to a strong anion exchange column (POROS HQ 20, 1 ml) pre-equilibrated with the same buffer. Elution from the ion exchange column was carried out with a linear gradient of 0 -1 M NaCl in buffer B.
On discovering that p37 was identical to annexin I, we modified the purification procedure and included a Ca 2ϩ -chelating agent to extract membrane-associated calcium-binding proteins. The membrane fraction was incubated with 10 mM Tris-HCl (pH 7.4) containing 5 mM BAPTA for 30 min on ice. Following centrifugation, the supernatant was phosphorylated with unlabeled ATP/GDP and applied to a chelating Fe 3ϩ -Sepharose column as described above.
Protein Sequence Analysis-N-terminal sequencing and in-gel tryptic peptide digests of the purified proteins were conducted according to published procedures (7).
Phosphoamino Acid Analysis-To determine the nature of the phosphate linkage on [ 32 P]annexin I, a membrane fraction was labeled with [␥-32 P]GTP or [␥-32 P]ATP for 5 min at 37°C. The reaction was terminated with 5ϫ Laemmli sample buffer (6). Phosphorylated protein (10 g) was then incubated for 10 min at 30°C in reaction mixtures containing phosphorylation buffer (3) alone, 0.1 N HCl (pH 1), 0.1 N NaOH (pH 13), or 0.8 M hydroxylamine and 0.1 M acetate (pH 5.2). Phosphoramidates are stable under alkaline conditions and labile under acidic conditions, whereas acylphosphates are labile under both extremes. In the presence of hydroxylamine, both linkages are labile at pH Ͻ5.5 (8).
On the other hand, serine and threonine linkages are acid-stable and base-labile, whereas tyrosine is relatively stable under both acidic and basic conditions. The precise phosphorylated residue was determined by HPLC phosphoamino acid analysis as described previously (9). Briefly, proteins were phosphorylated and blotted onto polyvinylidene difluoride as described above. The phosphorylated bands were excised, digested under alkaline conditions, and analyzed in the presence of phosphoamino acid standards by anion exchange HPLC coupled with fluorescence detection. Fractions were collected, and radioactivity was detected by Cerenkov counting.

RESULTS
Purification and Characterization of p37-Phosphoproteins and phosphopeptides are known to bind to immobilized iminodiacetic acid when charged with ferric chloride (10). Because p37 is the only major phosphorylated protein when membrane proteins from ovine airway epithelia are incubated with ATP and GDP (3), we used metal chelate (iminodiacetic acid-Fe 3ϩ ) affinity chromatography (10) to isolate phosphorylated p37. The eluate from the chelating Sepharose column was further purified by anion exchange chromatography (POROS HQ 20). More than 95% of the protein (A 280 ) applied was found in the anion exchange column flow-through fraction. SDS-polyacrylamide gel electrophoresis analysis of this fraction (Coomassie Blue staining) showed a major band of 37 kDa and two minor bands of 70 and 80 kDa (Fig. 1a). Microsequence analysis of the purified products using in-gel peptide digestion and N-terminal sequencing showed that all three bands had 100% homology to bovine annexin I (Fig. 1b). Two of the sequences obtained related to the N-terminal domain of annexin I, which is unique to annexin I. Immunoprecipitation of phosphorylated ovine p37 with annexin I-specific antibodies provided additional proof that p37 and annexin I were identical (Fig. 1c).
To test whether annexin I possessed intrinsic protein kinase activity, purified annexin I was incubated with [␥-32 P]ATP or [␥-32 P]GTP. However, no phosphorylation was observed (data not shown).
Phosphoamino Acid Analysis-Annexin I is phosphorylated on serine and tyrosine residues within the N-terminal domain by protein kinase C and epidermal growth factor receptor kinase, respectively (11). However, the identity of p37 as annexin I was intriguing particularly because general and specific inhibitors of protein kinase C, tyrosine kinase, casein kinase II, and PKA failed to inhibit its phosphorylation (data not shown) (4). We had also observed loss of label from p37 on SDS gels destained with acetic acid-containing destaining solutions. In addition, phosphoamino acid analysis procedures that included  3). The 37-kDa protein was excised from the gel, and peptides were obtained by tryptic digest, purified by HPLC, and sequenced. Additionally, N-terminal sequencing was conducted using an aliquot of the purified protein solution. b, shows the amino acid sequences obtained. All matched the published sequence of bovine annexin I at the amino acid positions shown in parentheses. c, immunoprecipitation of p37 from membranes phosphorylated with either ATP or GTP confirmed the identity of the purified p37 protein as annexin I. The failure to immunoprecipitate phosphorylated ␤-subunits of G proteins (phosphohistidine proteins of similar size) with a G protein ␤-subunit (G␤ Common)-specific antibody confirmed the selectivity of annexin I phosphorylation under these conditions. acid hydrolysis (6 N HCl for 2 h at 105°C) of phosphoannexin I (labeled with [␥-32 P]GTP or [␥-32 P]ATP for 5 min) removed Ͼ95% of the radioactivity from p37 (annexin I) (data not shown). Further analysis showed that [ 32 P]annexin I, phosphorylated with either [␥-32 P]GTP or [␥-32 P]ATP, was not only acid-and hydroxylamine acetate-labile (Fig. 2a), but also basestable (Fig. 2a) and inhibitable following preincubation with diethyl pyrocarbonate (data not shown). Loss of phosphate due to rapid proteolytic degradation of annexin I was discounted because of the presence of [ 32 P]annexin I in control experiments lacking acid and hydroxylamine (Fig. 2a, Buffer lane). Taken together, the above results suggested the presence of a phosphoramidate linkage on annexin I and, in particular, phosphohistidine. To confirm the identity of the alkali-stable phosphorylated residue, an alkaline hydrolysate from [ 32 P]annexin I was analyzed by anion exchange chromatography coupled with fluorescence detection. Coelution of the 32 P-labeled material with derivatized N 3 -phosphohistidine standard established the presence of phosphohistidine (Fig. 2b).
Localization of Phosphohistidine on Annexin I-To further characterize annexin I phosphorylation, a BAPTA extract partially purified on a chelating Fe 3ϩ -Sepharose column was phosphorylated with [␥-32 P]GTP for 5 min at 37°C and then incubated for 30 min at 30°C in the presence of calcium (4 mM). An additional 21-kDa phosphorylated band was observed. This result suggested cleavage of annexin I within the core domain by a protease(s) present in the sample (Fig. 3a). Sequence analysis of in-gel peptide digests from the 21-kDa band showed that this fragment corresponds to the carboxyl-terminal region of annexin I most likely covering annexin repeats 3 and 4 (Fig.  3b). The identity of the 21-kDa fragment within the C-terminal half of the annexin core was further confirmed using an affinity-purified polyclonal antibody (C-19, Santa Cruz Biotechnol-ogy) raised against a peptide fragment from the C-terminal region of annexin I. This antibody recognizes the phosphorylated 21-kDa fragment (Fig. 3a) and thus verifies that it is part of the C-terminal core domain of annexin I. Multiple sequence alignments with sequences from four species showed that only two of the three (or four depending on the species) histidine residues (His 246 and His 293 ) within this C-terminal domain are conserved (Fig. 3c).
Regulation of Phosphorylation by cAMP-While characterizing another chloride-sensitive phosphoprotein, nucleosidediphosphate kinase (a multifunctional protein-histidine kinase), we observed inhibition of p37 phosphorylation by cAMP. Interestingly, cAMP has been shown to bind to annexin I and to abolish its ability to act as a calcium channel in artificial lipid bilayers (11). Following preincubation with N 6 -Bt-cAMP (0 -1 mM) for 30 min at 4°C, a membrane preparation was phosphorylated with [␥-32 P]ATP for 5 min at 37°C. Maximal inhibition of phosphorylation of annexin I by N 6 -Bt-cAMP occurred at 1 mM (Fig. 4). Studies conducted to characterize the mechanism of inhibition by cAMP showed that neither cGMP (1 mM) nor myristoylated PKA inhibitor-(14 -22)-amide (4 M) inhibited annexin I histidine phosphorylation (Fig. 4b). These results suggested that the inhibitory effect was specific to cAMP and, unusually, independent of PKA activity. Furthermore, no significant difference was observed between two phosphodiesterase-resistant cAMP analogues, (R p )-and (S p )-cAMP-S (100 M) (Fig. 4b), which inhibit and activate PKA, respectively. Unexpectedly, both isomers significantly elevated annexin I phosphorylation above control levels (p Ͻ 0.05, paired t test). The increased phosphorylation was dose-dependent, with higher isomer concentrations (1 mM) generating higher phosphate incorporation (data not shown). This result and our observed inhibition of annexin I phosphorylation by 5Ј-AMP, but not by adenine or adenosine (1 mM) (Fig. 4c), suggested that inhibition of annexin I phosphorylation by cAMP might result from hydrolysis of the ester bond within cAMP. However, selective and nonselective inhibitors of phosphodiesterase activity (3-isobutyl-1-xanthine, Ro 20-1724, rolipram, dipyridamole, zaprinast, trequinsin, and 8-methoxymethyl-3-isobutyl-1-methylxanthine) failed to influence the inhibitory effect of N 6 -Bt-cAMP (data not shown). Consistent with the above results, the cAMP analogues N 6 -2Ј-O-Bt 2 cAMP and 2Ј-O-Bt-cAMP, which are susceptible to phosphodiesterase activity, also failed to inhibit annexin I phosphorylation (data not shown).

DISCUSSION
The transfer of a reactive phosphoryl group from phosphorylated histidine kinase to an acceptor is a primary step in many cellular signaling responses. For example, in bacteria, yeast, and higher plants, phosphorylation of histidine is the first step in substrate sensory cascades such as the two-component phosphorylation systems (for reviews, see Refs. 8 and 12-14). In mammals, phosphohistidine has previously been identified on the C terminus of P-selectin from activated platelet cells (15) and on G protein ␤-subunits in HL-60 cell membranes (16), but its cellular role remains to be established. The work reported here identifies a previously recognized phosphoprotein (p37) from airway epithelia as annexin I and shows that the protein is phosphorylated on histidine residue(s) using either GTP or ATP as phosphate donor. This represents a novel phosphorylation of annexin I and is the first report of histidine phosphorylation of an annexin protein. The transient nature of phosphohistidine on annexin I and its susceptibility to hydrolysis under conditions used to detect phosphoesters may explain why it remained undetected in previous studies.
The failure of annexin I to autophosphorylate when incubated with nucleotides indicates that it is a substrate for a distinct histidine protein kinase whose activity is predicted to be regulated by chloride and other ions (3,4). Interestingly, annexins share sequence homology with a cAMP-activated chloride channel, the cystic fibrosis transmembrane regulator, in the region of its common mutation Phe 508 (17) and, among other functions, appear to regulate secretion and ion transport by unknown mechanisms. Annexins II and IV have been shown to modulate various chloride channel activities in epithelial and endothelial cells (18 -21).
Structurally, annexins share a conserved series of 70-amino acid repeats organized into a core structure that binds reversibly to acidic phospholipids and cellular membranes in a calciumdependent manner. Each annexin also contains an amino-terminal segment that is unique to each protein and that varies in sequence and length (for reviews, see Refs. [22][23][24][25]. Although the repeat units of the core domain contain both the Ca 2ϩ -and phospholipid (membrane)-interactive elements of the protein, the N-terminal domain regulates both the binding affinities for phospholipids and the subcellular localization (26) and is thus thought to provide functional specificity for individual annex-ins. Moreover, regulation via the N-terminal domain occurs through several post-translational modifications within the Nterminal sequences of some annexins and includes phosphorylation (27), transglutamination (28), proteolytic cleavage (29), and attachment to other proteins (30,31). The above modifications influence the annexin-induced aggregation of phospholipid vesicles and chromaffin granules, together with the calcium requirement for these reactions (32)(33)(34)(35)(36). In addition, they control the interaction with some S100 calcium-binding proteins (34 -36). The cellular consequences of these regulatory events remain to be established.
The generation of a phosphorylated fragment whose sequence corresponds to the C-terminal half of annexin I localizes the site of histidine phosphorylation to the core. To our knowledge, this is only the second report showing phosphorylation of an annexin core. In vitro phosphorylation of annexin I at threonine 216 by PKA has been reported (37), but its physiological significance is controversial. Thus, phosphorylation may regulate annexin I function via both the N-terminal and core domains. Notably, all the histidine residues of annexin I lie a, an autoradiograph is presented of a membrane preparation incubated with an increasing concentration of N 6 -Bt-cAMP (N6-mbcAMP) showing maximal inhibition of annexin I phosphorylation occurring at 1 mM (representative of n ϭ 3). b, quantitation (mean Ϯ S.E., n ϭ 6) by electronic autoradiography showed that N 6 -Bt-cAMP induced a 6-fold reduction in phosphate incorporation compared with the buffer control (p Ͻ 0.001, Student's t test). No significant difference was found between the buffer control and PKA peptide inhibitor (4 M) or cGMP (1 mM). In contrast, both (R p )-and (S p )-cAMP-S (100 M each) significantly elevated annexin I phosphorylation (p Ͻ 0.01, Student's t test). c, AMP (but not adenosine or adenine; 1 mM each) induced a reduction in annexin I phosphorylation (n ϭ 3, Ϯrange).
within the Ca 2ϩ -dependent phospholipid-binding core and, with one exception, are located on the face of the molecule that interacts with phospholipids (38). The proposed functions of the core, which include membrane organization and aggregation (39,40), may be affected by phosphorylation-induced charge alterations on histidine. Consistent with this notion, maximal changes in histidine ionization occur within the physiological range of intracellular pH. In addition, cAMP binds annexin I via a proposed cAMP-binding site, adjacent to a conserved histidine (His 103 ), and profoundly alters both the membrane aggregating and calcium channel properties of annexin I in lipid bilayers (11). Our observed inhibition of histidine phosphorylation of annexin I by N 6 -Bt-cAMP and AMP could result from an altered conformation such that the histidine residue(s) become inaccessible to the kinase. Alternatively, AMP may also bind to the histidine kinase and allosterically regulate its activity or compete with ATP for the active site. The inhibition by AMP is not observed with IMP or CMP, 2 indicating specificity for the adenosine moiety. cAMP binding and the resulting alteration of the conformation of intracellular effector molecules such as PKA are thought to be the basis of most of the cAMP effects on cellular function (41). The cAMP signal is then subsequently inactivated and terminated by phosphodiesterases whose activity may be regulated by PKA in a feedback loop (42,43). Although a number of phosphodiesterases exist in airway epithelia (44), our results suggest that PKA and phosphodiesterase activities do not directly influence the effects of cAMP on annexin I phosphorylation. Thus, cAMP may either act through a hitherto unknown pathway or bind directly to the histidine kinase and allosterically regulate its activity. The enhanced annexin I phosphorylation observed with both stereoisomers of cAMP indicates that, unlike PKA, the histidine kinase fails to distinguish between (R p )-and (S p )-cAMP-S; thus, cAMP-binding site(s) on the histidine kinase may be different from those on PKA. The failure of cGMP to influence annexin I phosphorylation suggests that there is no crossactivation between the two cyclic nucleotides.
Phosphorylation within the core domain, which is highly conserved among annexins, suggests that histidine phosphorylation may also be a feature of other members of the annexin family. Considering the limited number of post-translational modifications that occur within the N-terminal domains of annexins and the diversity of processes in which annexin function has been suggested, it is possible that other post-translational modifications of the core exist. Clearly, modification of the core domain by phosphorylation suggests that phosphatase(s) and other proteins may interact with this domain to modulate the histidine phosphorylation signal. Proteolytic cleavage of annexin I within its N-terminal domain is well documented (although the protease responsible is not known) and has been shown to modulate annexin function (29,45). Susceptibility to proteolytic cleavage of the N-terminal "tail" is increased following phosphorylation on Tyr 21 by the epidermal growth factor receptor kinase (46,47). In contrast, the core has previously been regarded to be highly resistant to proteolytic cleavage. The cleavage of [ 32 P]annexin I within the core to generate a phosphorylated C-terminal fragment requires further investigation, as it is presently unclear whether or not its cleavage is induced by phosphorylation. A similarly sized frag-ment from annexin I has been reported previously in samples from cells labeled with phosphate (48). It is possible that phosphohistidine may initiate annexin I cleavage and thus generate active peptide fragments or lead to further degradation of the protein.