Annexin XXI (ANX21) of Giardia lamblia has sequence motifs uniquely sdhared by giardial annexins and is specifically localized in the flagella.

We have identified a novel annexin, ANX21, in trophozoites of Giardia lamblia. The nucleotide sequence encoding this protein deviated from a published sequence in predicting an additional endonexin fold in the fourth annexin domain. In addition, several motifs exclusively shared by other annexins of G. lamblia in their predicted fourth repeat and predicted to be localized on the opposite (concave) surface of the molecule became apparent. Western blots of trophozoite fractions probed with antiserum against the recombinant protein indicated that this annexin, like the other giardial annexins ANX19 and ANX20, associates with phospholipids in the presence of Ca(2+). Finally, confocal laser scanning of trophozoites showed that the protein, apart from the median body, was exclusively localized in the eight flagella. Together, these data suggest that ANX21 may function as a Ca(2+)-regulated structural element linking phospholipid bilayer and underlying axoneme.

Giardia lamblia (syn. Giardia intestinalis, Giardia duodenalis), a group of diplomonadid parasitic protists, is classified as early branching eukaryotes (1). G. lamblia occurs throughout the world and triggers a form of diarrhea called giardiasis (2). Its life cycle consists of two stages: the infective, immobile cyst form that by virtue of its tough cell wall is able to survive the inhospitable conditions of the host's stomach, and the vegetative, mobile trophozoite form that attaches to the epithelial cells of the gut. It does so with the help of a cytoskeletal structure called ventral disk, which probably functions as a suction cup. The ventral disk consists of spiraling microtubuli with flat structural elements called microribbons protruding from them into the cytoplasm (3). The edges of these microribbons contain two proteins that were originally designated as ␣-giardins (4,5) and shown to be associated with the cytoskeletal fraction by detergent extraction of the insoluble cell pellet (6). Based on their predicted amino acid sequence, they were later identified as members (ANX19 and ANX20) of the annexin (ANX) 1 family (7). Annexins are eukary-otic proteins that usually bind to phospholipid bilayers in a Ca 2ϩ -dependent manner (for an exception, see Ref. 8) and supposedly play a role in Ca 2ϩ -dependent membrane dynamics (9). They consist of four homologous, mainly ␣-helical domains folded into a concave/convex shape. In annexins of higher eukaryotes, each of these four domains possesses the canonical repeat GXGTD followed by an aspartate or glutamate residue 38 positions downstream that forms a loop (AB loop) at the convex surface and functions as high affinity ("type II") Ca 2ϩ -binding site (10). Other Ca 2ϩ -binding sites with lower affinity ("type III") provided by spatially clustered carboxylate groups are likewise positioned at the convex surface of the molecule. Most annexins exhibit an ion channel activity with ion transport assumed to occur through a central pore lined by charged residues, particularly glutamate and arginine (9). Although the giardial annexins ANX19 (11) and ANX20, 2 in addition to interacting with the cytoskeleton fraction (6,12), bind to phospholipids in a Ca 2ϩ -dependent manner, they lack an endonexin fold and do not exhibit ion channel activity.
In a report on the structure of a gene encoding a pyruvateferredoxin oxidoreductase from G. lamblia, a flanking complementary DNA sequence putatively encoding an additional annexin has been identified (13). To extend the open reading frame and to increase the overall similarity of the predicted ANX21 to other annexins, the authors postulated an insert of an undefined nucleotide at position 9846 of the complementary sequence (14). We here present a corrected nucleotide sequence predicting a protein with, in its fourth domain, a bona fide endonexin fold on its convex surface and conserved giardin motifs on its concave surface. In trophozoites, ANX21 was exclusively localized in the flagella. These data are consistent with a model in which ANX21 functions as a Ca 2ϩ -regulated structural element linking the flagellar membrane and the axoneme.
Nucleic Acid Manipulations-Genomic DNA was isolated from fresh trophozoites using an Elu-Quick kit (Schleicher and Schuell). To amplify the open reading frame flanking that of the pyruvate-ferredoxin oxidoreductase (13), which has been predicted to code for an annexin homologue (14), we constructed the following synthetic oligonucleotide primers (bold, restriction sites for BamH1): 5Ј-GTTTTTGTGACACTC-GAGAGTAAAATGGC-3Ј (C1anx21) and 5Ј-GAATTATTTACACGAC-TACAACTCGAGAG-3Ј (C2anx21). This primer pair frames the full * This work was supported by the Deutsche Forschungsgemeinschaft by a grant (to A. S.) within the framework of the graduate college "Molecular Physiology." 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) L17221.
¶ To whom correspondence should be addressed: Faculty of Biology/ Chemistry, Barbarastrasse  open reading frame from position 318 to 1369 of the DNA sequence deposited in the sequence data base (GenBank TM L17221). PCR on genomic DNA as template was performed using Pfu-polymerase (Stratagene, Heidelberg). The PCR profile was as follows: 5 min at 94°C, 45 cycles with 2 min at 55°C, 2 min at 72°C, and 1 min at 94°C, then 2 min at 55°C, and 5 min at 72°C. The amplification product was purified using a QIAEx-II-Gel-Extraction-Kit (Qiagen, Hilden Germany) according to the instructions of the manufacturer, subcloned into pBSK, and sequenced by the automated dideoxy chain termination method (MWG Biotech and in house). The sequence was determined twice in both directions. Expression of Recombinant Protein and Antibody Production-The amplification product was digested with BamH1 and ligated in frame into the multiple cloning site of the expression vector pET16b, which contains a 5Ј-extension sequence coding for 10 histidine residues ("Histag"). After transfection of the construct into Escherichia coli BL21 and induction with isopropyl-1-thio-␤-D-galactopyranoside, the overproduced protein product was partially present in the soluble fraction as checked by SDS-PAGE. For purification, the soluble E. coli extract was adjusted to 25 mM imidazole, the mixture applied onto a Ni-NTA column, and the recombinant protein eluted with 20 mM Tris-HCl, pH 7.9, containing 250 mM imidazole. This yielded a Ͼ95% pure recombinant protein judging from SDS-PAGE (data not shown), which was sent out for the immunization of rabbits (Eurogentec, Belgium). On a Western blot of an extract of E. coli containing the recombinant protein and of the purified recombinant protein, the antiserum (1:1000) reacted with a single protein band at an apparent M r 40,500; the same band was detected with antipenta-His antiserum (Qiagen, 1:2000; data not shown).
Isolation of Annexins and Binding Studies-Annexins were isolated from crude trophozoite homogenates by EGTA extraction according to Ref. 16. Pellet and supernatant fractions were analyzed by SDS-PAGE and Western blotting. For phospholipid binding, 400 l of 20 mM Hepes brought to pH 7.4 with NaOH containing 100 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, and 0.5 mg multilamellar liposomes (brain extract; Sigma) was added to 100 l of soluble EGTA extract containing 30 g of protein.
The mixture was incubated for 40 min at room temperature under shaking and then centrifuged for 10 min at 15,000 ϫ g. In a parallel experiment, the free Ca 2ϩ concentration in the mixture was adjusted to 1 mM, and in control incubations the brain extract was omitted. For isolation of detergent-insoluble cytoskeletal proteins cells were homogenized in the presence of 4 mM Ca 2ϩ , the homogenate was centrifuged for 10 min at 15,000 ϫ g, and the pellet fraction subsequently was extracted with 0.5% (w/v) Triton X-100. The extract was mixed with an equal volume of ice-cold 10% (w/v) trichloroacetic acid, and equivalent fractions of the precipitated protein (Triton X-100 extract) and the Triton-insoluble pellet were analyzed by SDS-PAGE and Western blotting.
Isolation of Flagella-Trophozoite flagella were isolated by the method of Clark and Holberton (17). Briefly, the cells were washed twice by centrifugation in 0.25 M sucrose and resuspended in TMSK buffer consisting of 30 mM Tris, 2.5 mM MgSO 4 , 0.2 M sucrose, 25 mM KCl, 1 M E-64, and 0.005% (w/v) phenylmethylsulfonyl fluoride, pH 7.4. The suspension was homogenized for 1 min with an Ultra-Turrax (Braun Melsungen, Germany). Cell bodies were removed by centrifuging for 10 min at 220 ϫ g in a swingout rotor. The flagella, which were contained in the supernatant, were pelleted at 13,000 ϫ g for 20 min and purified by density gradient centrifugation through a self-forming Percoll gradient. To this end, the crude flagellar fraction was mixed with 10 ml of 40% (v/v) Percoll (Amersham Biosciences) in TMSK buffer and centrifuged for 60 min at 48,000 ϫ g. For calibration, a parallel gradient was run with Percoll density marker beads. Fractions containing flagella (density range between 1.09 and 1.11 g/ml) were diluted with TMSK, and the flagella were pelleted for 5 min at 13,000 ϫ g and analyzed by Western blotting.
Gel Electrophoresis and Immunoblotting-SDS-PAGE was performed in the Tris/glycine system of Douglas et al. (18) using 10% gels. Proteins were visualized by dispersion staining with Coomassie Brilliant Blue G-250 (19). Western blotting onto nitrocellulose membranes was performed according to Ref. 20 in a CAPS/NaOH-buffer, pH 11.0, containing 10% methanol. For immune decoration of the blots, rabbit antiserum against recombinant ANX21 was used in a dilution of 1:1000. As secondary antibodies we used goat anti-rabbit IgG (Pierce; 1:2,000) coupled to peroxidase, and as chromogenic substrate, we used 4-chloro-1-naphthol.
Immunofluorescence-Trophozoites were allowed to attach to coverslips at 37°C, fixed for 7 min with methanol, and permeabilized for 5 min with acetone, both at Ϫ20°C (21). After rehydrating with PBS for 10 min at room temperature, the cells were incubated with blocking buffer (3% fetal porcine serum in PBS) for 10 min and then were reacted for 1 h with the rabbit anti-ANX21 antiserum (1:10,000 in PBS). After washing three times with PBS the cells were incubated for 1 h in the dark with CY3 TM -conjugated anti-rabbit F(ab) 2 fragment from sheep (Sigma; 1:100 in PBS) as secondary antibody. As a positive control cells were labeled with a monoclonal antibody against antiacetylated tubulin (mouse IgG2b isotype from Sigma, 1:2000 in PBS) followed by antimouse IgG CY3 conjugate F(ab) 2 fragment from sheep (Sigma; 1:400 in PBS). After another three washes with PBS the cells were analyzed in a confocal laser scanning microscope (Zeiss cLSM 410) equipped with an Ar/Kr excitation laser at 568 nm and an emission filter for 590 nm.

RESULTS AND DISCUSSION
Nucleotide Sequence of anx21 and Its Implications for the Predicted Protein-To check for the postulated insertion of an undefined nucleotide (14) in the published sequence of anx21 (13) and to pinpoint its nature and exact position we amplified the complete open reading frame using genomic DNA as template and sequenced the amplified product (sequence available from the EMBL data base under the accession number AJ271737). While confirming an insert of one base (identified as G), we found its exact position to occur at nucleotide number 9836 rather than 9846, as proposed by Morgan and Fernandez (14), of the complementary sequence. This correction of the published nucleotide sequence increases the identity of the second half of the sequence to that of giardial ANX19 -20 from 14 to 19% and to that of human ANX5 (22) from 11 to 19% and reveals an additional endonexin fold (GSGSD{38}E) at positions 257-261 and 299 ( Fig. 1; numbering adapted to ANX5 sequence). Moreover, the corrected predicted C-terminal sequence now shares several sequence motifs with ANX19 and ANX20, namely IT(G/A)M at 268 -270 and 272, KXXYK (X stands for a variable residue) at 281-285, DXER at 293-296 and Trp at 311 (Fig. 1). (Some of these residues have been boxed as "protist-specific" in Ref. 13, Fig. 6). Meanwhile, the genome data base of G. lamblia (23) has yielded Ն13 other independent sequences encoding putative annexins, and strikingly, all of these are predicted to exhibit the four motifs common to ANX19 -21 (but not the endonexin fold shared by ANX21 and ANX5; data not shown), whereas these motifs fail in all known annexins from other organisms. In most of these sequences a positively charged residue (usually Arg) follows the tryptophan residue at 311. In a molecular model the indole ring of this tryptophan together with the KXXYK motif ends up on the concave surface of the molecule, opposite the endonexin  (30). White letters on a gray background, motifs conserved in annexins of G. lamblia; black bold letters on a gray background, complete type II Ca 2ϩ -binding sites (endonexin fold). fold (Fig. 2), suggesting that these residues could constitute a binding site for a cytoplasmic interaction partner.
Expression of anx21 in Trophozoites and in E. coli-Both Southern blot analysis and data base searching indicated that the G. lamblia genome possesses just one nucleotide sequence encoding ANX21 (data not shown). Northern blot analysis confirmed expression of this anx21 gene in trophozoites with a transcript size (ϳ1050 nt) that leaves ϳ40 nt for the 5Ј-and 3Ј-untranslated regions (Fig. 3). To raise anti-ANX21 antibodies we overproduced the recombinant, His-tagged protein heterologously in E. coli (for details, see "Experimental Procedures"). Probing of Western blots with the antiserum raised against the recombinant protein confirmed that ANX21 is present in trophozoites (Fig. 4).
Association of ANX21 with Phospholipids and with the Detergent-insoluble Cytoskeletal Fraction-To investigate the association of ANX21 with negatively charged phospholipids, we extracted trophozoite homogenate with EGTA and incubated the soluble supernatant with multilamellar liposomes prepared from brain extract. Fig. 4A shows that the extracted ANX21, just like ANX19 (11) and ANX 20 (data not shown), ended up in the pellet fraction in the presence of excess Ca 2ϩ (lane 4) but remained in the supernatant in its absence (lane 1).
This effect was strictly dependent on the presence of the liposomes (lanes 5-6). These data confirm that ANX21 behaves as a classical annexin in associating with negatively charged phospholipids in a Ca 2ϩ -dependent way. In a complementary set of experiments, we homogenized the cells in the presence of free Ca 2ϩ . As expected, in the presence of endogenous phospholipids, ANX21 remained in the insoluble fraction (lane 2 of Fig.  4B). Likewise, when the pellet fraction was subsequently extracted with detergent, ANX21 remained in the pellet fraction (Fig. 4B, lane 4). However, when this pellet was treated with an 1 mM excess of EGTA, ANX21 went into solution (lane 5, Fig.  4B). As any phospholipid membranes that had been precipitated from the homogenate should be dissolved into floating micellar structures by the detergent, we interpret these data to mean that ANX21, apart from binding to phospholipids, also associates with detergent-insoluble cytoskeletal elements in the presence of Ca 2ϩ . An association with the detergent-extracted cytoskeletal fraction has also been reported (17) for ANX19 and ANX20 (at that time denoted as ␣-giardins). Combining the model of Fig. 2 with the fractionation behavior  Fig. 4, we speculate that the G. lamblia-specific motifs of ANX19 -21 are responsible for interaction (directly or indirectly) with detergent-insoluble cytoskeletal elements.
Immunolocalization of ANX21 in Trophozoites-In fixed and permeabilized trophozoites, ANX21 was exclusively localized in the eight flagella (and in those cells where it was observable, in the median body) (Fig. 5A). Control incubations with antisera against tubulin (Fig. 5B) and a proteasome subunit (not shown) indicated that this restricted distribution was not due to a lack of permeability of the cells. In agreement with these data, Western blot analysis of isolated flagella revealed cross-reaction with both the tubulin and ANX21 antibodies (Fig. 5C) but not with anti-ANX19 antibodies (data not shown). Together with our observation that ANX21 interacts with both phospholipids and the detergent-insoluble cytoskeletal fraction in a Ca 2ϩ -dependent way, we interpret the flagellar localization to mean that ANX21 may play a Ca 2ϩ -regulated structural role in trophozoite motility. Specifically, our data would fit in with a model in which binding of Ca 2ϩ to the convex surface of ANX21, apart from increasing the affinity of this surface to the flagellar membrane, induces a conformational change at the concave surface that enables the latter to interact with a cytoskeletal element of the axoneme (possibly an adapter protein). In the literature, there are precedents for both Ca 2ϩ -dependent binding of annexins to cytoskeletal proteins and for the localization of annexins in flagella or cilia. Thus, many vertebrate annexins bind F-actin in a Ca 2ϩ -dependent way (24). For ANX2, this binding has been shown to depend upon the nine C-terminal amino acid residues that, just like the motifs specifically shared by the annexins of G. lamblia (Fig. 2), are predicted to be located on the concave surface of the protein (25). (The F-actin-binding ANX2 residues are not conserved in the G. lamblia annexins; data not shown). As to axonemal localization, the ciliated cells of lung epithelium specifically contain ANX1 in their cilia (26), and both ANX1 and ANX2 have been found to move to the sperm flagellum during ram sperm maturation (27). An interesting variant on the theme of association of Ca 2ϩ -regulated proteins with axonemes is provided by the paraflagellar rod complex from Trypanosoma brucei, which instead of annexins, contains EF-hand Ca 2ϩ -binding proteins (28,29).
As mentioned above, we hypothesize that for Ca 2ϩ -dependent interaction of ANX21 with the G. lamblia axoneme the four motifs conserved in the last domains of giardial annexins ( Fig.   1) may be important; consequently, our future experiments will be directed at identifying the corresponding protein binding partner(s) of these annexins.