I. INTRODUCTION: OVERVIEW OF THE ANNEXIN FAMILY

Top Previous Next References

Nature has achieved the means to tightly control intracellular Ca²⁺ concentrations, thereby enabling the ion to serve second messenger functions in a variety of processes which couple extracellular signals to cellular responses. Systems regulating intracellular Ca²⁺ levels thus are considered part of the intricate Ca²⁺ signaling network. They include gated Ca²⁺ channels and energy-dependent pumps, which are located in organelle membranes and the plasma membrane, as well as intracellular Ca²⁺ binding proteins serving as regulated Ca²⁺ buffers. Other classes of Ca²⁺ binding proteins participate more directly in Ca²⁺ signaling as they display altered properties in response to Ca²⁺ binding. Annexins can be considered a subgroup of the latter, although their precise position within Ca²⁺ signaling chains remains elusive. Moreover, a growing body of evidence suggests that annexins can also function in their Ca²⁺-free conformation in a hitherto unknown fashion, thereby increasing the functional diversity among these proteins.

The name annexin is derived from the Greek annex meaning "bring/hold together" and was chosen to describe the principal property of all or at least nearly all annexins, i.e., the binding to and possibly holding together of certain biological structures, in particular membranes. The name also has a somewhat historical flavor as it takes into account the point that a number of the groups who independently of one another discovered annexins were in search for such scaffolding or bridging proteins. However, initially, i.e., at the date of their discoveries in the late 1970s and early 1980s, annexins received diverse and unrelated names referring to their biochemical properties. These included synexin (for granule aggregating protein, Ref. 52), chromobindins (proteins binding to chromaffin granules, Ref. 54), calcimedins (proteins mediating Ca²⁺ signals, Ref. 199), lipocortins (steroid-inducible lipase inhibitors, Ref. 85), and calpactins (proteins binding Ca²⁺, phospholipid, and actin, Ref. 101). Intensive biochemical work, protein and cDNA sequencing, as well as gene cloning led to the realization that all such proteins identified shared key biochemical properties as well as gene structure and sequence features. Hence, the concept of a novel multigene family arisen by gene duplication was developed and the common name annexin was introduced to solve the terminology tangle (55).

By definition, an annexin protein has to fulfill two major criteria. First, it must be capable of binding in a Ca²⁺-dependent manner to negatively charged phospholipids. Second, it has to contain as a conserved structural element the so-called annexin repeat, a segment of some 70 amino acid residues. Molecular structures obtained for a number of annexins over the past decade helped to extend the similarities to the three-dimensional level. Moreover, they defined a hitherto unknown structural fold, the conserved annexin domain, which is built of four annexin repeats packed into a highly -helical disk, and which now is considered to be a general membrane binding module. Once clearly defined and advanced by genome sequencing work, the annexin family has grown steadily in the 1990s, and with the turn of the century, now amounts to more than 160 unique annexin proteins present in more than 65 different species ranging from fungi and protists to plants and higher vertebrates (Fig. 1) (202, 204). In this review we summarize the biochemical and structural properties of annexins, putting a particular emphasis on novel aspects of annexin interactions with lipids and other biological ligands. For a detailed discussion of the canonical annexin properties, their structural organization, and intracellular as well as tissue distribution, the interested reader is referred to previous reviews (51, 97, 244).

View larger version (62K): [in this window] [in a new window]   Fig. 1. The new annexin nomenclature. The five major annexin groups (A-E) are shown, with details of the most extensively studied family members. The nomenclature is that proposed by Reg Morgan and Pilar Fernandez and endorsed by participants at the 50th Harden Conference on Annexins held at Wye College, UK, September 1-5, 1999. A more extensive list of annexin subfamilies and species is posted at the European annexin web site (http://www24.brinkster.com/annexins/). There are several important points to note. The vertebrate annexins (A1-A13) are unlikely to be widely represented in invertebrate species. The oldest of this group, namely, annexins A7, A11, and A13, are possible exceptions, and an annexin A11 ortholog has been described in the mollusk Aplysia. Within the B group, the Caenorhabditis elegans annexins have yet to be assigned numbers. In the C group, the Dictyostelium annexin, originally described incorrectly as annexin VII (synexin), is now established as being orthologous to the Neurospora annexin.

Having accumulated a wealth of biochemical and structural knowledge, we are still in need of assigning a physiological function to the annexin family as a whole, or better, because they are likely to differ, to individual annexins. Recent knock-out models, both at the cellular and the animal level, as well as the development and use of dominant-negative mutant proteins have introduced the first direct approaches for analyzing annexin function. They underscore the concept of functional diversity within the family. Moreover, it has recently become clear that certain dysregulations in annexin expression and activity can be correlated with human diseases and that this has led to the introduction of the term annexinopathies. Although we still have to await final proof of a direct correlation, we decided to concentrate our review on such recent developments leading to the proposal of some models concerning annexin function.

	II. BIOCHEMICAL PROPERTIES OF ANNEXINS AND THEIR THREE-DIMENSIONAL STRUCTURE

Top Previous Next References

A.  Molecular Structures

1.  Structures of annexin protein cores: the conserved membrane binding modules

Each annexin is composed of two principal domains: the divergent NH₂-terminal "head" and the conserved COOH-terminal protein core. The latter harbors the Ca²⁺ and membrane binding sites and is responsible for mediating the canonical membrane binding properties. An annexin core comprises four (in annexin A6 eight) segments of internal and interannexin homology that are easily identified in a linear sequence alignment (for review, see Ref. 244). It forms a highly -helical and tightly packed disk with a slight curvature and two principle sides. The more convex side contains novel types of Ca²⁺ binding sites, the so-called type II and type III sites (335), and faces the membrane when an annexin is associated peripherally with phospholipids. The more concave side points away from the membrane and thus appears accessible for interactions with the NH₂-terminal domain and/or possibly cytoplasmic binding partners (Fig. 2). The first structure known for an annexin core was that solved by Huber et al. (134) for annexin A5 in 1990. In the meantime, more than 10 crystal structures for annexin cores have been described showing a remarkable conservation of the overall three-dimensional fold. Aspects of molecular annexin structures have been reviewed in detail previously (see, for example, Refs. 133, 178, 305), and the purpose of this review is to discuss only recent and novel developments in this area.

View larger version (46K): [in this window] [in a new window]   Fig. 2. Crystal structure of human annexin A5. The ribbon drawing illustrates the highly -helical folding of the protein core that forms a slightly curved disk. Different colors were chosen to highlight the four annexin repeats that are given in green (repeat I), blue (repeat II), red (repeat III), and violet/cyan (repeat IV). The NH2-terminal domain appears unstructured and extends along the concave side of the molecule (green). The high and low Ca2+ forms are shown in a superposition revealing the conformational change in repeat III, which leads to an exposure of Trp-187 (violet for the low and cyan for the high Ca2+ form). Bound Ca2+ are depicted as yellow spheres. [Image kindly provided by R. Huber, S. Liemann, and A. Lewit-Bentley, as modified from Ref. 178.]

Recent findings include the elucidation of the first structures of annexins from lower eukaryotes and plants. Liemann et al. (177) crystallized the core of annexin C1 from Dictyostelium discoideum, whereas Hofmann et al. (127) elucidated the structure of a plant annexin, annexin D11 from Capsicum annuum, which revealed not only the typical annexin fold but also differences to nonplant annexins in annexin repeats I and III and in the membrane binding loops. Another recent advance is the introduction of benzodiazepine and benzothiazepine derivatives as annexin ligands and their cocrystallization with annexins. The benzothiazepine K201 was first described to bind to annexin A5 and inhibit its Ca²⁺ channel activity, most likely by restraining a hinge movement of the two annexin A5 modules formed by annexin repeats I/IV and II/III, respectively (149, 150). Other structurally related benzodiazepine compounds have subsequently been identified as ligands for various annexins with the interaction being based on similar structural principles (126). However, a possible pharmacological role of these interactions remains to be shown. Crystal structure determination and biochemical characterization in combination with site-directed mutagenesis have also proven powerful in recent years in characterizing the contribution of certain residues to the overall fold of annexin cores and/or their biochemical properties. Conserved arginine residues present in the so-called endonexin fold of each homology segment, for example, were shown to be crucial for stabilizing the tertiary structure of annexin A5. On the other hand, substitution by alanine of different serine and threonine residues and the unique tryptophan in the same annexin results in altered membrane binding underscoring the importance of these residues in mediating intermolecular, i.e., annexin-phospholipid, contacts (31, 32). Moreover, mutational analysis revealed that the aspartate residue at position 226 of annexin A5 participates as a molecular switch in a Ca²⁺- and pH-dependent conformational change (294). Like many other annexins, annexin A4 is a substrate of protein kinase C (PKC), at least in in vitro reactions, and Kaetzel et al. (147) have attempted to monitor structural changes resulting from this phosphorylation. They show that replacement by glutamate of the PKC acceptor site, threonine-6, causes a release of the NH₂-terminal domain from the protein core indicative of a regulatory role in the membrane aggregation displayed by this annexin. Annexin B12 has also been subjected to detailed scrutiny by mutagenesis involving cysteine substitutions for spin labeling purposes (see below) and the glutamate at position 105. This residue has been found to participate in the formation of intermolecular Ca²⁺ binding sites in a hexameric form of the molecule (185), and replacement of Glu-105 by lysine stabilizes this hexamer by favoring extensive hydrogen bonding (35).

Recently, techniques other than crystallization of the soluble proteins have been introduced to study in detail structural properties of annexins, in particular when bound to membrane or phospholipid surfaces. They include cryoelectron microscopy, which led to the identification of highly structured junctions formed by different annexins between opposing membranes (167), and atomic force microscopy (AFM), which enabled the high-resolution analysis of two-dimensional crystals of annexin A5 formed on planar lipid bilayers (248, 249). Two-dimensional crystals of annexin A6 formed on artificial lipid monolayers were also obtained and characterized recently, revealing an intrinsic flexibility of this eight-annexin repeat-containing molecule. Here the two lobes of annexin A6, i.e., repeats I-IV and V-VIII, respectively, were found to bind to the phospholipid in both parallel or antiparallel orientation, with the latter providing a structural basis for membrane cross-linking (6). Evidence for conformational changes occurring in annexins upon membrane binding was obtained by analyzing membrane-bound annexin A5 with transmission and internal reflection infrared spectroscopy. Interestingly, it was inferred from these studies that a new -structure with interstrand hydrogen bonds oriented parallel to the membrane surface is formed upon interaction with a lipid monolayer. On the other hand, analyses of two-dimensional crystals of the same annexin on membrane surfaces by high-resolution electron microscopy and AFM, and crystal structure analysis of a cross-linked form of annexin A2 capable of binding membranes, do not provide evidence for substantial conformational alterations accompanying the canonical Ca²⁺-dependent membrane binding (26, 29, 229, 248). Relatively subtle changes, however, might occur. These include the exposure of the unique tryptophan in repeat 3 of annexin A5 observed in high Ca²⁺ (47, 174). Thus, despite the wealth of structural information on soluble as well as membrane-bound annexins, it is not clear whether the peripheral and Ca²⁺-dependent membrane binding of annexins as a whole, or individual annexins, requires or is accompanied by conformational changes. Moreover, the structural basis of (possible) membrane insertions of annexins triggered by certain environmental changes like hydrogen ion concentration (see below) need to be described in more detail, possibly also by integrating into such analyses the characterization of folding properties of individual annexin repeats such as the first repeat of annexin A1 (49, 94).

Molecular details of the three-dimensional folds of annexin molecules are mostly restricted to the protein core domains (see above) and unique NH₂-terminal regions of the smaller annexins containing NH₂-terminal sequences of 16 or fewer residues. In these structures, the NH₂-terminal sequences extend along the concave side of the molecule partially engaged in hydrophobic interactions with the protein core. In annexin A3, a direct effect of the NH₂-terminal domain on properties displayed by the core has been shown by replacing Trp-5 (in the unique NH₂-terminal sequence) by alanine. The W5A mutant protein shows a much stronger phospholipid binding, and although having a similar overall structure has a more disordered NH₂-terminal domain. Interestingly, through urea-induced denaturation analysis, it became apparent that the NH₂-terminal domain, even though comprising only 16 residues, unfolds separately from the protein core (128). Thus it appears that the short NH₂-terminal domains of the smaller annexins, located on the concave side of the folded molecule, affect the Ca²⁺-dependent phospholipid binding executed by the convex, or opposite, side possibly through stabilizing or destabilizing slightly different conformations of the molecule. This underscores the regulatory importance of even the small NH₂-terminal domains, a notion that had previously been postulated due to the presence of sites for posttranslational modifications in these regions (see below). Moreover, the finding that subtle differences in the NH₂-terminal sequence, which do not affect the overall structure, result in significantly altered properties could at least in part explain functional diversity among otherwise highly conserved annexins.

Recently, the first complete structure of a longer annexin, annexin A1, has been determined at high resolution (255). Annexin A1 has an NH₂-terminal domain of 40 residues, the first 10-14 of which represent the binding site for a protein ligand of the S100 family, S100A11 (188, 273). Interestingly, in the Ca²⁺-free crystals of annexin A1, this NH₂-terminal sequence forms an amphipathic -helix and replaces a helix in the tightly packed core domain (helix D of repeat III), which in turn is unwound and partially extrudes from the protein surface (255). Such a structure could have interesting mechanistic and regulatory consequences. Given the tight internal packing of the NH₂-terminal helix, one would assume that this sequence is not available for S100A11 binding in Ca²⁺-free annexin A1. However, upon Ca²⁺-dependent membrane binding, the D helix of repeat III could be forced back into the position described for the Ca²⁺-loaded annexin A1 core (335), thereby freeing the NH₂-terminal helix and enabling this sequence to interact with S100A11. Moreover, such movement could be the prerequisite for the membrane aggregation activity described for annexin A1, e.g., by making accessible a second membrane binding site or a site for homophilic annexin 1 interaction (for a hypothetical model see Fig. 3). Thus, in the case of annexin A1, Ca²⁺ could have a dual regulatory function. First, it could trigger membrane attachment through the convex side of the molecule, and second, by inducing the switch of helix D, it would enable the membrane-bound protein to interact with cellular protein ligands (S100A11) and/or a second membrane surface. In this respect, it is interesting to note that thermodynamic analyses revealed cooperativity in the binding of Ca²⁺ to annexin A1 (257). Finally, the conformational switch postulated by Rosengarth et al. (255) could also modulate the accessibility of phosphorylatable residues in the NH₂-terminal domain of annexin A1 for their respective kinases (see below), thereby guaranteeing a spatially restricted, probably membrane-dependent, regulation of annexin A1 activities.

View larger version (32K): [in this window] [in a new window]   Fig. 3. Model describing the switch of helix D in the annexin A1 structure and its implications for membrane aggregation. In the crystal structure of Ca2+-free annexin A1 (red), the NH2-terminal -helix, which contains the S100A11 binding site (brown), is replacing helix D of the third repeat (255). Ca2+-dependent membrane binding could be accompanied by a conformational change establishing the Ca2+-bound crystal structure of the annexin A1 core (335) and, most likely, a more accessible NH2-terminal domain. As a result, the NH2-terminal domain can interact with a second membrane surface or the S100A11 dimer, which itself requires Ca2+ binding to establish an interaction-competent conformation. An as of yet hypothetical annexin A1/S100A11 heterotetramer would represent an entity capable of linking membrane surfaces (see text and Ref. 255 for details).

The structure of the very same NH₂-terminal domain of annexin A1 comprising residues 1-14 has also been solved in complex with its S100A11 ligand. Cocrystals of S100A11, a homodimeric protein containing two EF hand-type Ca²⁺ binding sites, with the NH₂-terminal annexin A1 peptide revealed a 1:1 stoichiometry, with the two peptides occupying hydrophobic pockets on two opposite sides of the S100A11 dimer (246). The structure of the complex proved to be very similar to that of the NH₂-terminal sequence of annexin A2 bound to a related S100 protein, S100A10 (247). In both annexins (A1 and A2), the first 14 residues form amphipathic -helices providing the binding sites for two ligands of the S100 protein family (15, 142, 188, 273). At least in the case of annexin A2, it has been shown that the formation of a heterotetrameric complex containing the S100A10 dimer and two annexin A2 chains significantly alters the properties of this annexin in vitro and also within cells (for reviews, see Refs. 97, 330). Importantly, the annexin A2-S100A10 complex can aggregate membrane vesicles at micromolar Ca²⁺ levels, a property not shared with monomeric annexin A2 or, as a matter of fact, any other annexin. The structure of the NH₂-terminal annexin A2 peptide in complex with S100A10, in combination with high-resolution images of junctions formed between adjacent membranes by the annexin A2-S100A10 complex, now provides the first detailed structural explanation of this aggregation activity. It appears that due to the highly symmetric nature of the structures, the complex links annexin A2-bound membrane surfaces through the dimerization of S100A10, i.e., the two annexin A2 subunits of the membrane-linking complex are bound to two separate bilayers with the S100A10 dimer connecting them through binding to the NH₂-terminal domains (167, 175, 247). A similar scenario could hold true for the annexin A1-S100A11 complex, which in contrast to annexin A2-S100A10 requires Ca²⁺ binding to the S100 protein and probably Ca²⁺/membrane-bound annexin (see above) for complex formation. Nevertheless, we are still in need of high-resolution structures of complete annexin A2-S100A10 and annexin A1-S100A11 complexes to prove or disprove this attractive model, as well as some evidence that the annexin A1-S100A11 exists in vivo.

B.  Annexins as Membrane Binding Proteins: Canonical and Atypical Properties

1.  Ca²⁺-dependent phospholipid binding and vesicle aggregation

Biochemically, annexins are defined as soluble, hydrophilic proteins that bind to negatively charged phospholipids in a Ca²⁺-dependent manner (they are Ca²⁺/phospholipid binding proteins). This binding is reversible, and removal of Ca²⁺ by Ca²⁺ chelating agents will lead to a liberation of annexins from the phospholipid matrix. The interaction of annexins with negatively charged phospholipids observed in vitro is thought to reflect in a more physiological scenario the binding to cellular membranes, in particular, the cytosolic leaflets of the plasma membrane and various organelle membranes. This canonical annexin property is retained within the annexin cores, the conserved annexin modules most likely representing building blocks designed for peripheral membrane association. However, although Ca²⁺-dependent phospholipid binding is shared by all annexins, individual members differ significantly in their Ca²⁺ sensitivity and phospholipid headgroup specificity. A large number of reports analyzing the Ca²⁺-regulated phospholipid binding of annexins in vitro have been published, and a comprehensive overview has been given by Raynal and Pollard (244).

Although differences in the binding to phospholipids with different headgroups (e.g., phosphatidic acid, phosphatidylserine, phosphatidylinositol) have long been recognized in in vitro studies, it has only recently become clear that annexin cores also show specificity with respect to their membrane binding in living cells. Through expression of chimeric proteins containing different annexin cores fused to the green fluorescent protein (GFP), it was possible to visualize the distribution of such annexin cores in living cells. Strikingly different distributions were observed within a given cell type showing, e.g., an endosomal localization for the annexin A1 core, an association with certain plasma membrane structures for the annexin A4 core, and a nonmembranous, cytosolic distribution for the annexin A2 core (245). Such live cell experiments have to be extended to reveal annexin dynamics and to circumvent the potential problem that in fixed cells annexin distributions could be subjected to artifacts due to the presence or absence of Ca²⁺ in the fixation/permeabilization buffers. Although the annexin cores carry specificity with respect to membrane binding, an additional layer of such specificity is most likely added by the unique NH₂-terminal domains of the annexins as in live cells full-length proteins show distributions often differing from the respective cores (73, 196, 245). Moreover, it remains to be seen how interactions with other protein ligands and posttranslational modifications (see below) affect the specific localizations of annexins to certain cellular sites.

Although well described in vitro, the physiological importance of Ca²⁺-dependent phospholipid (and membrane) binding is not understood. However, interesting models have been put forward to assign functions to a peripherally associated and abundant membrane binding protein like, e.g., annexin A5. In situ, annexin A5 can form two-dimensional crystals on planar lipid bilayers containing negatively charged phospholipids (26, 229, 248, 249). Such crystalline or semi-crystalline arrangement will most likely affect membrane properties including rigidity, fluidity, and lipid segregation and can therefore participate in the regulation and/or stabilization of membrane domains. Indeed, electron paramagnetic resonance (EPR) spectroscopy reveals that Ca²⁺-dependent binding of annexin A5 to phospholipid vesicles parallels a rigidification of the membrane (193). Moreover, it was shown that binding of this annexin to the surface of T cells (by an as yet unknown mechanism) delays programmed cell death most likely by generating a certain (in this case extracellular) membrane constraint which in turn interferes with the release of CD4⁺ membrane particles (99). On the other hand, membrane binding also affects the annexin protein, as annexin A5, thermodynamically a marginally stable protein (like annexin A1; Refs. 256, 328) is protected to a significant degree from thermal denaturation by Ca²⁺/phospholipid binding (338).

Annexins are not only capable of binding phospholipid-containing membranes but at least in some cases, e.g., annexins A1, A2, A4, A6 and A7, also mediate membrane vesicle aggregation. Again, phospholipid composition and Ca²⁺ sensitivity for this aggregation activity differ for individual members (for review, see Ref. 244). As molecular structures of annexins reveal one Ca²⁺/lipid-binding surface (see above), several models have been put forward to explain an aggregation activity based on the linking of membrane surfaces (see also Ref. 97). One proposal, based on the self-association properties of several annexins, is a protein-protein interaction of annexin molecules bound to two separate membranes (for reviews, see Refs. 51, 244; recent example in Ref. 180). A second explanation is based on the identification of a second membrane binding site in annexin A1 (for review, see Ref. 97 and also discussion in Ref. 23). A sequence in the unique NH₂-terminal domain of annexin A1, residues 24-35, constitutes a crucial part of this second binding domain and when fused to the core of annexin A5 can confer membrane aggregation activity to this otherwise inactive annexin (40). In contrast to the Ca²⁺-dependent primary membrane binding via the annexin A1 core, the secondary binding is mainly hydrophobic in nature, and it appears that lateral aggregation of annexin A1 molecules bound to one membrane surface precedes the aggregation mediated through the secondary binding site (24). Interestingly, recent crystal structure determination of full-length annexin A1 suggests that the NH₂-terminal domain of this annexin only becomes fully accessible when the protein core is linked to a membrane surface via its Ca²⁺/phospholipid binding sites (255). This could indicate that the second (NH₂-terminal) membrane binding site in annexin A1 is dormant in the cytosolic protein and only becomes activated when the protein associates with membranes. A third alternative of aggregation activity is probably realized in annexin A6, the only member of the family identified so far with eight instead of four annexin repeats. Here a duplication of the core domain has generated a second Ca²⁺-dependent phospholipid-binding module, thus allowing for two spatially separated membrane interactions (6). Yet another route is taken by annexin A2 and possibly also other annexins capable of interacting with dimeric protein ligands of the S100 family (see below). The NH₂-terminal domain of annexin A2 harbors a highly specific binding site for the small dimeric S100 protein S100A10, with protein-protein interaction leading to the formation of a heterotetrameric complex. In this complex, two annexin A2 molecules are noncovalently linked via a S100A10 dimer bound to their NH₂-terminal domains, thereby generating an entity capable of binding simultaneously to two membrane surfaces through the two annexin A2 cores (175). Thus it appears that although several annexins mediate membrane-membrane contacts, the way this is achieved differs from member to member. This could explain the different Ca²⁺ concentrations required by different annexins for half-maximal vesicle aggregation (for review, see Ref. 244) and also the differing dimensions of annexin-dependent junctions observed in high-resolution cryoelectron microscopy of lipid vesicles aggregated by different annexins in the presence of Ca²⁺ (167).

Although Ca²⁺-dependent phospholipid binding remains the criterion of choice for defining an annexin protein biochemically, additional "atypical" lipid binding properties have begun to emerge in recent years. These properties again vary between the different annexins analyzed so far, but it appears that the single most important parameter regulating Ca²⁺-independent membrane binding is the pH value chosen to analyze the interaction. Annexin A5, for example, binds to and apparently penetrates the bilayer of phosphatidylserine (PS) vesicles at pH 4 (158), and at pH 5 was shown to induce a leakage of PS vesicles (124). Both activities are observed in the absence of Ca²⁺, whereas at neutral pH Ca²⁺ binding to the protein appears to be a prerequisite for the lipid interaction (158). Most likely, this switch in properties is accompanied by a conformational change in the annexin A5 molecule, which has been shown to occur between pH 4.6 and 4 when the acid-induced unfolding of the protein was analyzed (16). This change is characterized by solvent exposure of a unique tryptophan residue in annexin A5 (Trp-187), and thus is reminiscent of a Ca²⁺-induced exposure of the same tryptophan at neutral pH (192, 293-296). Conformational changes leading to Ca²⁺-independent phospholipid binding in vitro have also been proposed to occur when Ca²⁺ sites in annexin 2 were inactivated by mutagenesis (79), although such mutations interfere with the intracellular membrane localization of this and other annexins (145, 245).

Considerable progress in analyzing Ca²⁺-independent annexin-membrane interactions occurring at lower pH has come recently through the introduction of site-directed spin labeling. By engineering protein mutants with unique cysteines and specifically derivatizing these cysteines with a paramagnetic nitroxide side chain, the groups of Haigler, Langen, and Hubbell (168, 169) were able to probe the structure of annexin B12 bound to membranes at lower pH. Combined with the use of reagents that selectively and photoactivatably label amino acid side chains exposed to the hydrophobic domain of the bilayer, they could show that annexin B12 inserts into the bilayer of PS/phosphatidylcholine (PC)-containing vesicles. This insertion is likely to be accompanied by the formation of a continuous transmembrane -helix. In the solution structure of the molecule, this part forms a helix-loop-helix motif, and it is tempting to speculate that the switch from the helix-loop-helix motif to the transmembrane helix drives a reversible membrane insertion (138, 168, 169). Based on these observations, Langen et al. (168) propose concerted conformational changes in all four annexin repeats of annexin B12, which are triggered by low pH and involve the formation of several elongated transmembrane helices from helix-loop(turn)-helix structures found in solution (Fig. 4). As a consequence, the entire molecule can assume a transmembrane topology as defined by accessibility to proteases present on either side of the membrane (289). The pH-dependent switch in conformation could be induced by the protonation of certain carboxylate residues found in or close to the loop of the helix-loop-helix motif, which upon deprotonation could drive the protein back to the solution conformation (168). Such a model could also hold true for other annexins, as all have similar solution structures, and its reversibility could perhaps explain why and how certain annexins under certain circumstances can span a lipid bilayer. The latter could be of particular importance in the case of annexins A1 and A2, which also appear to have extracellular activities and for which cell surface receptors have been described (see below). At least in the case of annexins A1 and A6, pH-driven membrane insertion has been identified (103, 258), although it is not clear whether this could lead to membrane translocation.

View larger version (46K): [in this window] [in a new window]   Fig. 4. Peripheral membrane binding and insertion by an annexin. Two potential interaction states for a monomeric annexin molecule with the cytoplasmic leaflet of a hypothetical membrane are shown. The peripherally bound annexin on the left assumes the tertiary structure depicted in Figure 2 and has been postulated to increase membrane permeability by apposition of its convex upper surface with the lipid bilayer, which in turn has been suggested to lead to ion flow. The fully membrane integrated structure on the right is based on that proposed by Langen and co-workers (168, 169), after protonation at acidic pH, destabilization of the native -helical structure, and refolding into the seven-transmembrane spanning configuration. Although the proposed structure is obliged to have NH2 and COOH termini on opposing sides of the bilayer, the orientation shown in the figure is arbitrary.

In addition to the points discussed above, Ca²⁺-independent membrane associations have also been observed for several annexins at neutral pH. Examples for these types of Ca²⁺-independent interactions are the association of annexins A2 and A6 with endosomal membranes (120, 145, 160, 275), the binding of annexin A2 to A549 cell membranes (182), and the interaction of annexin A5 with the plasma membrane of platelets (for review, see Ref. 322). At least in part it appears that such interactions are mediated through a binding of the respective annexin to a protein ligand that is itself associated with or embedded in the cellular membrane.

C.  Nonlipid Annexin Ligands

1.  Annexin complexes with EF hand-type Ca²⁺ binding proteins

The EF hand denotes a helix-loop-helix Ca²⁺ binding motif that is present in a large number of proteins comprising the EF hand superfamily with its distinct subfamilies (for review, see Ref. 154). Several EF hand proteins, in particular those of the S100 subfamily, form complexes with members of the annexin family. S100 proteins are small (~10 kDa) proteins characterized by two consecutive EF hands connected by a flexible linker region and flanked by unique NH₂- and COOH-terminal extensions. Similar to calmodulin, they are thought to interact with and thereby regulate cellular target proteins in a Ca²⁺-dependent manner (for reviews, see Refs. 68, 267). Three S100 proteins, S100A6, S100A10, and S100A11, were shown to bind specifically to three different annexins, annexins A11, A2, and A1, respectively. The best characterized of these annexin-S100 complexes is the annexin A2-S100A10 (p11) heterotetramer. Here it was clearly established that complex formation is highly specific, occurs in vivo, can be regulated by posttranslational modifications in the annexin, and modulates properties displayed by the isolated subunits (for review, see Ref. 97). S100A10 is the only member of the S100 family that has suffered deletions and mutations in its two EF hand loops, rendering the Ca²⁺ sites nonfunctional. However, it appears that the resulting conformation of the protein represents a permanently active state with respect to its capacity to bind the annexin A2 target (143, 246, 247). The S100A10 binding site on annexin A2 is restricted to the NH₂-terminal 14 residues and peptides corresponding to this sequence bind to S100A10 with high specificity and affinity. Moreover, such peptides disrupt by competition preformed annexin A2-S100A10 complexes and therefore can be used as tools for studying complex function (161). At least in studies with synthetic peptides, an important feature of this binding site is the NH₂-terminal acetylation of the NH₂-terminal serine residue of annexin A2, a posttranslational modification occurring with high efficiency in eukaryotic cells (15). On the other hand, annexin A2, expressed recombinantly in bacteria and lacking the N-acetyl group, is also capable of binding p11, although the affinity of this interaction has not been compared with that of the acetylated protein (151). The apparent discrepancy in these results remains to be resolved, in particular since only an acetylated NH₂-terminal annexin peptide is capable of disrupting the annexin A2-S100A10 heterotetramer (161).

Complex formation between annexin A1 and S100A11 is based on very similar principles, although in this case Ca²⁺ binding to the S100 protein is required to establish the interaction-competent form of S100A11 (188, 246, 273). Moreover, it remains to be established if and when this interaction occurs in vivo. In contrast to the heterotetrameric annexin A2-S100A10 complexes, standard isolation protocols do not yield annexin A1-S100A11 complexes but only the separated subunits. Likewise, a strong colocalization has not been reported so far, although ectopic expression studies using mutant proteins indicate that annexin A1 can target S100A11 to endosomal membranes in baby hamster kidney (BHK) cells (274). It appears likely that the strict Ca²⁺ dependence of the annexin A1-S100A11 interaction interferes with the visualization or isolation of complexes once Ca²⁺ drops below a certain threshold within the cells or during isolation. On the basis of the high structural similarity of the annexin A2-S100A10 and annexin A1-S100A11 complexes, it is likely that both are heterotetrameric entities with the capacity of linking membrane surfaces in a symmetric manner (see Fig. 3 for annexin A1). While the former complex is known to exist in resting cells irrespective of cellular Ca²⁺ transients (but perhaps regulated by PKC phosphorylation in the NH₂-terminal sequence of annexin A2; Ref. 144), the latter is probably dependent on (perhaps locally restricted) Ca²⁺ rises and could be of importance during Ca²⁺-regulated membrane transport events.

The third annexin-S100 protein interaction described to date is that between annexin A11 and S100A6 (312). Although the S100 binding site in annexin A11 is also located in the NH₂-terminal domain (303, 311), the mode of complex formation is likely to be different. Annexin A11 contains a long NH₂-terminal domain of almost 200 residues rich in glycine, tyrosine, and proline residues, which resembles that of annexin A7 and possibly lacks a well-ordered three-dimensional fold (177), or contains segments with pro--helices (190). However, sequences within the NH₂-terminal domain of annexin A11 do not resemble the amphipathic helices found in annexins A1 and A2 and thus are unlikely to fit in a homologous manner in a binding pocket formed by the S100 dimer. The physiological consequences of the annexin A11-S100A6 interaction remain to be established, although it is interesting to note that only one NH₂-terminal splice form of annexin A11 can interact with S100A6 at least in vitro (302). Other annexin-S100 interactions have been described, e.g., that of annexin A6 with S100A1 and S100B (96), but the structural basis and physiological significance of such complexes is less well defined. The reader is referred to the S100 literature for further detail (e.g., Ref. 68).

In contrast to S100 proteins, sorcin is a member of the EF hand superfamily containing four and not two of the helix-loop-helix motifs. It binds in a Ca²⁺-dependent manner to the GYP-rich NH₂-terminal domain of annexin A7 (28) with the NH₂-terminal domain of sorcin being required for the interaction (327). Complex formation can recruit sorcin to the membrane of chromaffin granules, which are a prime site of annexin A7 localization. Moreover, binding of sorcin inhibits the chromaffin granule aggregation mediated by annexin A7 (28), thus underscoring the regulatory importance of complex formation. The common picture emerging from the interaction analyses is that it is most likely to be the annexins that are affected in their properties by EF hand protein binding, rather than the other way round. Although the structural basis of the interaction probably differs between the different complexes, a common feature is the importance of the NH₂-terminal annexin domain for binding. Protein binding to this unique domain in the respective annexin can have a number of consequences ranging from the establishment of a different physical entity capable of interconnecting membranes (see, for example, Fig. 3) to a protein complex with altered biochemical properties.

A number of annexins have been described as cytoskeleton, in particular F-actin, binding proteins, and it has been suggested that at least some members of the family could participate in regulating membrane-cytoskeleton dynamics. Here we do not survey the entire literature in this area but only focus on recent developments. For an overview of the earlier literature, the reader is referred to previous reviews (97, 244).

Annexin A1 binds to F-actin and also interacts with profilin, a G-actin binding protein and regulator of actin polymerization. Complex formation between annexin A1 and profilin modifies the profilin effect on actin polymerization. Because of the partially overlapping intracellular localization of the two proteins, it is tempting to speculate that the annexin A1-profilin interaction participates in regulating the membrane-associated cytoskeleton (2). In addition to this interaction with a protein involved in actin cytoskeleton dynamics, some colocalization of annexin A1 with tubulin and cytokeratin-8 has also been reported. In A549 human lung adenocarcinoma cells, striking patches of annexin A1 immunolabeling are found at the plasma membrane, which are also positive for these two cytoskeletal proteins (313). It is not clear, however, whether this colocalization reflects a functional interaction.

Annexin A2 is another F-actin binding annexin that also has a Ca²⁺-dependent filament bundling activity. This bundling activity is particularly pronounced in the case of the heterotetrameric annexin A2-S100A10 complex (for review, see Refs. 97, 330). Recently, the F-actin binding site has been mapped to the COOH terminus of annexin A2, underscoring the specificity of the interaction (80). Annexin A2 is not associated with stress fibers or cytoplasmic actin filaments but appears to play a role in the organization of membrane-associated actin at sites of cholesterol-rich membrane domains. Evidence for this view is severalfold. In the presence of Ca²⁺, annexin A2 binds to and possibly promotes the lateral association of glycosphingolipid- and cholesterol-rich lipid microdomains (rafts) (8, 119). It has been proposed that in smooth muscle cells this association promotes the binding of annexin A6, which itself mediates the formation of a reversible membrane contact with the actin cytoskeleton (8, 9). On the other hand, annexin A2 could also carry out this task by itself or in conjunction with other actin binding proteins, since cholesterol-sequestering agents specifically release annexin A2 together with the cortical cytoskeletal proteins -actinin, ezrin, and actin from membranes of BHK and endothelial cells (120; J. König and V. Gerke, unpublished observations). Moreover, expression in epithelial cells of a mutant annexin A2 protein causing the submembraneous aggregation of annexin A2 and its ligand S100A10 results in the simultaneous aggregation of a transmembrane raft protein (CD44) and a redirection of actin bundles toward these clusters (216). Thus, due to its Ca²⁺-dependent membrane and F-actin binding and its intracellular location at sites of membrane rafts, annexin A2 could serve as an organizer of these membrane microdomains and their connection to the actin cytoskeleton.

Annexin A5 has been observed to relocate to the cortical membrane cytoskeleton after activation of platelets. This relocation appears to involve both binding to the plasma membrane and to a specific actin isoform, -actin, and is paralleled by an association with the platelet membrane of cytosolic phospholipase A₂, suggesting an interaction between this phospholipase and annexin A5 (319-321). Annexin A6, another actin-binding annexin, has been implicated in mediating in a Ca²⁺-dependent manner membrane-cytoskeleton contacts in smooth muscle cells (9). Spectrin is another binding partner of annexin A6 in the cortical cytoskeleton. Because annexin A6 promotes a cysteine protease-dependent type of budding of clathrin-coated vesicles at the plasma membrane, it has been proposed that the protein participates in disconnecting the clathrin lattice from the spectrin membrane cytoskeleton during the final stages of coated pit budding (148). With the exception of annexin A2 (98), it is not known whether other annexins share this spectrin binding property. In addition to the family members mentioned above, F-actin binding annexins have recently also been identified in the killifish medaka (297) and in plants (131), although their functional roles in these organisms have not been addressed so far.

In addition to Ca²⁺, phospholipid, EF hand type proteins, and cytoskeleton-associated proteins, a number of other annexin ligands ranging from proteins to RNA and smaller molecules have been described. An account of such binding partners is given in previous reviews (97, 244), and only the most recent findings are summarized here. Moreover, this section primarily focuses on intracellular binding partners, whereas extracellular protein ligands are discussed when we review extracellular activities of annexins (see sect. IVC).

Annexin protein ligands other than the ones summarized above include the cytosolic phospholipase A₂, which interacts with annexin A1 (156) and the p120 Ras GTPase activating protein (GAP), which through its C2 domain binds to annexin A6 (60). Within annexin A6, the binding site has been mapped to the unique linker region connecting the two four-repeat lobes of the protein (44). This region is not found in other annexins, thus emphasizing the specificity of the interaction. Recently, two protein kinases, Fyn (a src kinase family member) and Pyk2 (a member of the focal adhesion kinase family), have also been found in the annexin A6-p120GAP complex, indicating that annexins could also participate in certain signaling events (43). A link between annexin A6 and signaling has also been inferred from its association with activated PKC-, which was described in skeletal muscle (272). Another binding partner for annexin A6 was identified in clathrin-coated vesicles isolated from adrenocortical tissue. In a subpopulation of these vesicles, which also contain the transferrin receptor, annexin A6 tightly associates with the GTPase dynamin known to participate in the pinching off of clathrin-coated endocytic vesicles (318). In the vesicle preparations, annexin A2 was shown to bind to a yet unidentified 200-kDa protein, suggesting that these two annexins could participate in defining specific protein-lipid interaction domains during endocytosis (see also below). A similar function has been suggested for annexin A13, which interacts with the C2 domain of the Nedd4 ubiquitin protein ligase, thereby participating in the apical membrane targeting of Nedd4 in polarized epithelial cells (231). Annexin A13 exists in two NH₂-terminal splice variants, a and b, with 13b being specifically targeted to the apical transport vesicles also containing raft components (see below). The binding of annexin A13 to Nedd4 was initially identified in a yeast two-hybrid screen and a number of annexin-protein interactions, e.g., that of annexin A5 with the intracellular domain of the vascular endothelial growth factor (VEGF) receptor Flk-1 (334), have been reported using similar approaches. Interestingly, in all cases reported it has been the protein ligand and never the annexin that was used as the bait in the initial screen.

Some annexins have also been shown to bind to other cellular macromolecules. Annexins A2, A4, A5, A6, and the Caenorhabditis elegans protein annexin B7 interact with carbohydrates, in particular glycosaminoglycans, and in some cases the binding sites have been mapped to certain regions within the respective annexin molecule (83, 139, 152, 159, 265). These interactions are likely to come into play when annexins are present extracellularly, but their functional significance remains to be proven. Nucleic acids comprise yet another class of macromolecules reported to bind to annexins in a Ca²⁺-dependent manner. Whereas annexin A1 interacts with purine-rich RNA and pyrimidine-rich DNA, annexin A2 has been found associated with mRNA of a distinct polysomal subpopulation (123, 326). It is not known whether and how annexin binding affects stability or functional state of the mRNAs, e.g., in terms of translation efficiency. However, because of the sequence specific binding (A. Vedeler, personal communication) and the fact that annexins A1 and A2 are also actin binding proteins, it has been speculated that the annexin proteins participate in the intracellular positioning of certain mRNAs via an interaction with both the mRNA and the actin cytoskeleton. Single nucleotides binding to certain annexin proteins have also been described in recent reports. While ATP binds to annexins A1 (118) and A6 (for review, see Ref. 11), annexin A7 not only interacts with but also catalyzes the hydrolysis of GTP (34). This latter observation led to the suggestion that annexin A7 acts as an atypical G protein involved in mediating the Ca²⁺/GTP signal during exocytotic membrane fusion (34). However, although GTP and GDP were present in immunoprecipitates of annexin A7 from permeabilized chromaffin cells, the ratio of GTP to GDP was apparently not influenced by Ca²⁺, raising questions as to how the GTPase activity might be regulated in vivo. Clearly, further work is required to improve our understanding of how annexins interact with nucleotides, especially as annexins lack a consensus nucleotide binding site. In this context, the three-dimensional structures of annexins complexed to nucleotides will be particularly informative, together with the identification of proteins that might modulate any catalytic activity ascribed to annexins, such as the activator proteins, dissociation inhibitors, and exchange factors that collectively regulate other GTPases.

Annexins are long known to be targets for posttranslational modifications. In fact, annexin A2 was initially isolated as a major v-src protein kinase substrate, and the tyrosine kinase activity of the epidermal growth factor (EGF) receptor has long been known to phosphorylate annexin A1 (see previous reviews, Refs. 51, 97, 111 and also Ref. 259 for an overview). More recently, additional phosphorylations by signal transducing kinases of these and other annexins have been reported with at least some of them affecting annexin properties. Other tyrosine kinases recognizing annexins A1 and A2 as substrates are those associated with the platelet-derived growth factor (PDGF) receptor, the hepatocyte growth factor/scatter factor, and the insulin receptor (for review, see Ref. 259). In the latter case, annexin A2 only undergoes insulin-triggered tyrosine phosphorylation when receptor internalization is occurring (22). To some extent, this mimics the tyrosine phosphorylation of annexin A1 upon activation and internalization of the EGF receptor and indicates that both annexins and their phosphorylation are mechanistically linked to the internalization/endocytic sorting of certain ligand bound receptors (see also below). Although these phosphorylations are known to occur in vivo, their physiological consequences have not been established. In vitro or in situ studies, however, have revealed alterations in the Ca²⁺/membrane binding of annexins A1 and A2 phosphorylated at Tyr-20 (by the EGF receptor kinase, Ref. 61) and Tyr-23 (by the src kinase, Ref. 102), respectively. Tyrosine phosphorylated annexin A1 is more susceptible to NH₂-terminal proteolysis, thus showing altered phospholipid vesicle binding and aggregation activities (for review, see Ref. 111). Moreover, in contrast to the nonphosphorylated form, it requires Ca²⁺ for the association with the membrane of multivesicular endosomes (92). In the case of annexin A2, tyrosine phosphorylation decreases its affinity for phospholipids and interferes with capability of the annexin A2-S100A10 complex to aggregate chromaffin granules at micromolar Ca²⁺ concentrations (132, 236). A mutual influence of Tyr-23 phosphorylation and phospholipid binding is also corroborated by the finding that phosphorylation of annexin A2 by pp60^src is significantly enhanced when the protein is bound to PS containing vesicle (17). Recently, annexins A7 and A11 were also described to be phosphorylated on tyrosine residues, in this case in rat vascular smooth muscle cells in response to PDGF. In vitro both annexins are also phosphorylated by the Ca²⁺-dependent tyrosine kinase Pyk-2, the src tyrosine kinase, and the EGF receptor kinase, but the physiological consequences of these phosphorylations have not yet been described (91).

A number of serine/threonine kinases that phosphorylate annexins have also been described. Phosphorylation sites again reside in the unique NH₂-terminal domains of the annexins, and the modifications in some cases have been shown to affect biochemical properties of the annexins, in particular their affinity for Ca²⁺/phospholipid (already reviewed in Refs. 97, 244). PKC, for example, has long been known to phosphorylate a number of annexins, with annexin A5 being a remarkable exception as it can serve as a PKC inhibitor (261). The strongest evidence for PKC phosphorylation regulating annexin activities in cells has accumulated in the case of annexin A2 and its involvement in Ca²⁺-regulated exocytosis in adrenal chromaffin cells. Here, nicotine stimulation leads to annexin A2 phosphorylation by PKC with activation of PKC being a prerequisite for regulated exocytosis (63, 263). A link between secretion and PKC phosphorylation has also been obtained recently in the case of annexin A7, with PKC phosphorylation activating the Ca²⁺-dependent membrane fusion displayed by this annexin (33). Other kinases acting on annexins are casein kinase I, which phosphorylates annexin A2, and a yet to be defined histidine-specific kinase which phosphorylates annexin A1 (95, 208). Recent evidence for a participation in intracellular signaling has been obtained for annexin A1, whose expression levels have been coupled to regulation of the extracellular signal-regulated kinase (ERK) pathway in RAW macrophages (1).

As pointed out above, annexin phosphorylation often results in an altered susceptibility toward proteolysis. Although this could be considered a mere indication of a conformational change, it could also reflect an important intracellular consequence of the posttranslational modification directly linked to altered properties displayed by the modified annexin. Cleavage generally occurs in the unique NH₂-terminal annexin domain with the resulting NH₂-terminally truncated molecule showing, as revealed in particular for annexins A1 and A2, an altered sensitivity toward Ca²⁺/phospholipid (for review see Refs. 111, 244) and a different intracellular location (245, 275). In addition, it appears at least in the case of annexin A1 that NH₂-terminal cleavage by an intracellular protease can occur without prior phosphorylation and can thus itself be considered the regulatory event. This has been shown in human neutrophils where removal by a membrane localized metalloprotease of the NH₂-terminal eight residues of annexin A1 results in a protein species with a decreased Ca²⁺ requirement for binding to secretory vesicles and the plasma membrane, an event possibly linked to the exocytosis of different vesicle populations (206, 207). In annexin A2, another recently observed modification is a S-glutathiolation of Cys-8 in the NH₂-terminal domain, which is observed after oxidative stress, e.g., in tumor necrosis factor (TNF)--treated cells (304). Modification of this cysteine, which is located in the S100A10 binding sequence, does not affect the interaction with S100A10 (143). On the other hand, a general cysteine modification of the annexin A2-S100A10 complex by N-ethylmaleimide (NEM), which most likely also affects Cys-8, strongly inhibits the ability of the complex to aggregate lipid vesicles (282). It is not clear whether cysteine residues participate directly in the aggregation activity or whether their derivatization interferes with certain conformational changes in the molecule required for the activity. However, as annexin A2 has been implicated in membrane trafficking events possibly requiring its aggregation activity (see below), NEM, which is frequently used as an inhibitor of membrane fusion, could also affect annexin A2.

Hence, a variety of posttranslational modifications on annexins which also include the N-myristoylation of annexin A13a and -b (77, 336) have been described, with most of them affecting the Ca²⁺ and/or membrane binding properties of the molecules and thus their most probable intracellular activity. Future analyses have to reveal how these modifications are mechanistically linked to the different annexin functions.

	III. MOLECULAR EVOLUTION OF THE ANNEXIN FAMILY AND REGULATION OF ANNEXIN GENE EXPRESSION

Top Previous Next References

Ever since annexins were first reported in the literature they have been categorized as a structurally conserved family of Ca²⁺ binding proteins. The structural conservation remains a defining characteristic, but the discovery of human annexins A9 and A10 (201, 204) provides what appear to be exceptions to the unifying ability of annexins to bind Ca²⁺. Nevertheless, the conservation of annexin primary structures extends throughout multicellular eukaryotic species, and the abundance of annexin sequences provides unique insights not only into the evolution of the annexin gene family, but also genetic molecular evolution in a broader sense. For readers seeking detailed accounts of annexin evolution, there are several excellent recent articles (203, 200, 204); here we focus on the major features of annexin phylogeny and ask whether or not functional insight can be gained by examination of molecular relationships between annexins.

The simplest organisms known to express annexins are the protist Giardia lamblia and the fungus Neurospora crassa. The existence of at least three annexins in the protist is surprising given the simplicity of the organism and that more sophisticated multicellular eukaryotes such as Hydra vulgaris and Dictyostelium discoideum have at most one or two annexins. A long-running question in annexin evolution is whether any of the annexins discovered in these primitive organisms represents the ancestor of the modern vertebrate annexins. Early studies describing the D. discoideum annexin as a direct ortholog of vertebrate annexin A7 (106) now appear to be incorrect. However, the D. discoideum annexin does occupy an interesting niche in annexin evolution. The D. discoideum and N. crassa annexins share ~40% amino acid sequence identity, which given their evolutionary distance suggests they may be orthologs. Indeed, this annexin has now been discovered in the oyster mushroom and potato fungus (R. Morgan and M. Fernandez, personal communication), suggesting the evolutionary segregation of this annexin to this group of organisms.

A second major group of annexins distinct from the vertebrate cohort have been described in plants (64). Plant annexins are characterized by their lack of variable NH₂-terminal domains and, at least in modern flowering plants, by the absence of type II Ca²⁺ binding sites in repeats 2 and 3. Thus, from an evolutionary viewpoint, plant annexins have evolved in quite distinct ways to those in the animal phyla. The position of the fungal and mold annexins relative to either the animal or plant kingdoms is unclear, but sequence identity of ~40% between D. discoideum annexin C1 and human annexin A11 raises the possibility that the former is a direct ancestor of the latter. Analysis of the structure of the annexin A11 gene (10) revealed it to be the common ancestor of up to nine descendent annexins (A1, A2, A3, A4, A5, A6, A8, A9, and A10), indicating that at the time of the early chordate radiation 500-600 million years ago the first vertebrate genomes probably contained the genes for only three annexins, namely, annexins A13, A7, and A11. Exon splicing patterns within the core tetrads of annexins A13, A7, and A11 support the idea that A11 is a descendent of A7 and that this in turn evolved from A13. However, orthologs of annexins A7, A11, and A13 have not been formally identified in any nonvertebrate species, and any direct lineage between annexins in organisms such as D. discoideum and vertebrates remains conjectural.

A final point of interest to emerge from studies on the molecular evolution of the annexins concerns the origins of annexin A6. This annexin is unique within the family in that it comprises two of the tetrad repeats found in all other annexins. The two tetrads are joined by a short linking sequence, and it was previously hypothesized that annexin A6 was formed by tandem duplication and fusion of a single tetrad (285, 286). Because the 5'-tetrad of annexin A6 is most closely related to annexin A5, it was proposed that the progenitor of this duplication event was the 5'-tetrad. However, the recent discovery and analysis of the annexin A10 gene provides an alternative and much more persuasive explanation for the origins of annexin A6 (204). First, annexin A10 has greater similarity to the 3'-tetrad of annexin A6 than the two halves of annexin A6 have to one another, and significantly, an unusual single codon deletion near the start of repeat three is present in both annexin A10 and the 3'-tetrad of annexin A6. These and other phylogenetic data suggest that the two four-repeat annexins A5 and A10, which are located on human chromosome 4q26 and 4q33, respectively, may have duplicated and fused to form the 5'- and 3'-lobes of annexin A6 early in chordate evolution.

Collectively, these phylogenetic studies enable us to put the annexins into a meaningful evolutionary context, but they tell us little about annexin function. Because the invertebrate and plant annexins do not have mammalian orthologs, analysis of annexin function in these simpler organisms may yield little information about the functions of the vertebrate family. Despite the difficulty in extracting functional insight from phylogenetic analysis, the fact that the family of 12 mammalian annexins have been tightly conserved over several 100 million years suggests that these proteins do indeed have important physiological roles.

B.  Gene Structures

1.  Conservation of genomic structure in the annexins

The structural organization of annexin genes is highly conserved, at least with regard to the positions of intron-exon boundaries (286). Most four-repeat annexins comprise 12-15 exons, the variation depending in large part on the length of the NH₂-terminal domains. Thus annexins A7 and A11 have long NH₂ termini encoded by up to six exons, whereas annexin A5 has a short NH₂ terminus encoded by two exons. For several annexins, particularly those with long NH₂ termini, alternative splicing adds to the diversity of annexin isoforms, which may in turn amplify functional variability within the family as a whole. Annexin A6, which has a duplicated tetrad core and therefore 8 conserved repeats, comprises 26 exons and is the largest annexin gene extending over ~60 kb (285). Within the conserved repeats, the tendency is for intron sizes to be considerably smaller than for those introns that lie between the first two or three exons. In many mammalian annexin genes, the first two or three introns are frequently 10 kb or more, whereas introns within the tetrad core are often <1 kb. Almost all alternative splicing of annexin RNA transcripts occurs within exons that encode the variable NH₂ termini. Given that annexin NH₂ termini contain binding motifs for protein partners and sites for posttranslational modifications, alternative splicing in these domains may contribute to the regulation of annexin function. Perhaps the best-characterized exception to this general rule is the alternative splicing of exon 21 in the seventh repeat of ANXA6. Exclusion of this 18-nucleotide exon gives rise to the characteristic appearance of annexin A6 on gel electrophoresis or Western blotting as a closely spaced polypeptide doublet (205).

Cladistic analysis of the mammalian annexin gene family reveals that annexins fall into three major groups. One group comprises the earliest vertebrate annexins, these being A7, A11, and A13. A second group includes annexins A4, A5, and A8, and the third group comprises annexins A1, A2, and A3, with annexins A9 and A10 as somewhat distant members. Annexin A6 is more difficult to categorize, because the 5'-tetrad is most closely related to the A4,A5,A8 group and the 3'-tetrad to the A1,A2,A3 group. Nevertheless, the cladistic demarcation of these groups raises the question of whether or not they correspond to functional groupings. Despite the lack of clear functional data for most annexins, it is certainly possible to identify some cohesion within these groups. For example, annexins A1 and A2 both bind proteins of the S100 family, both are physiological substrates for protein serine/threonine and tyrosine kinases, and both are suggested to function in the endocytic pathway. In contrast, annexins A4 and A5 are more closely linked with regulation of ion flow (see sect. IVB), and annexin A6, which arguably belongs in both groups, has been proposed to have roles that impinge on both the endocytic pathway and regulation of Ca²⁺ signaling. Although such notions are purely speculative, the possibility that annexin clades may represent functional groupings might be relevant to the issue of functional redundancy and therefore the design of gene knock-out experiments.

The completion of the human genome sequencing project, together with increasingly sophisticated algorithms for detecting and analyzing DNA sequences, has led to the identification of unusual and interesting elements within certain annexin genes. The most detailed analyses have been conducted for the annexin A5 and A11 genes (10, 137, 251). The rat and mouse annexin A5 genes are unusual in having two promoters. In both species, the promoter proximal to the gene has a high GC content and lacks a TATA box; this is also true for the human and chick annexin A5 genes (48, 76, 227), and all have an abundance of binding sites for the ubiquitous SP1 transcription factor. In contrast, the distal promoter in the rat and mouse annexin A5 genes has a TATA box and conserved binding sites for transcription factors such as AP1, the glucocorticoid receptor, and MyoD. The significance of these observations is not clear, but the possibility exists that under certain conditions, perhaps during cell differentiation, proliferation, or transformation, transcription from the distal promoter results in an annexin A5 transcript that omits exon 2 in which the start methionine is located. Such a transcript would initiate translation within the first conserved repeat, and the protein thus generated would be predicted to lack the NH₂ terminus and have a molecular mass ~3 kDa smaller than the full-length protein. Although there are no reports of the natural occurrence of such an annexin A5 splice variant, in vitro studies of recombinant annexin A5 showed that a mutant lacking the NH₂ terminus was unable to mediate a Ca²⁺ influx into phospholipid vesicles (20). Further investigation of these annexin A5 splice forms supported by a clearer understanding of annexin A5 function could reveal the significance of the two promoters for this gene.

The mouse annexin A5 gene also contains an endogenous retrovirus (251) located in intron 4. The MuERV-L sequence is believed to exist in only 100-200 copies in the mouse genome, although there is no evidence that its presence has any impact on the regulation of annexin A5 expression. The same gene also contains a region of Z-DNA (alternating purine-pyrimidine tract) in intron 6, and other Z-DNA sequences have been identified in the annexin A6 (287) and A11 genes (10). Given the abundance of repetitive elements in mammalian genomes, it is not surprising that Alu sequences, long interspersed nuclear elements (LINEs), mammalian-wide interspersed repeats (MIRs), and other less common elements have all been described in various annexin genes. For the most part, these appear to be no more than genomic landmarks, but in the case of annexin A6, a LINE-2 element named ALF (for annexin A6 LINE-2 fragment) was shown to function as a potent and highly specific T-cell silencer that may play a role in the downregulation of annexin A6 in T cells exposed to phorbol ester and calcium ionophore (69). This sequence was also shown to be present in other genes including interleukin-4 and PKC-, both of which are similarly downregulated by this combination of agonists in T cells.

Relatively few vertebrate annexins have been subjected to detailed promoter analysis. Two annexin A1 genes have been investigated in pigeons, one of which is strongly inducible by prolactin, and both of which bind Y-box factors (237, 323). The promoters for these genes have been partially characterized, but the most detailed analyses have been performed on the human annexin A1 (70, 290), A6 (70), and A7 (301) gene promoters. The annexin A1 gene promoter contains CAAT and TATA boxes that were shown in deletion studies to be essential for minimal promoter activity. Analysis of the annexin A1 promoter also permitted investigation of the sensitivity to dexamethasone, a glucocorticoid analog. Although one study found the promoter to be unresponsive to treatment with dexamethasone (70), the other reported some level of induction (290). The different results may correspond to the use of different cell lines in each report or may reflect the exposure times used which in the former case extended to 8 h, and in the latter to 24 h. Despite these differences, both studies support the notion that annexin A1 is not a glucocorticoid primary response gene. Interestingly, studies on the cytokine responsiveness of the annexin A1 promoter showed the gene to be induced by interleukin-6 (290). This result is consistent with a role for annexin A1 in the acute phase response to inflammation.

The human annexin A6 gene promoter also contains CAAT and TATA boxes, although these are somewhat distal to the transcription start site and in this case the minimal promoter lies downstream of and does not include these elements (70). The most unusual feature of the annexin A6 promoter is a potent T cell-specific silencer located ~600 bases 5' to the transcription start site (69). This element was discussed in section IIIB2. The human annexin A7 gene promoter has also been serially dissected, and it lacks CAAT and TATA boxes but is GC rich and contains many SP1 binding sites (301). The phylogenetically related annexin A11 gene promoter also lacks CAAT and TATA boxes, and it too is GC rich (10). The presence of SP1 binding sites in most, if not all, annexin promoters so far examined, is consistent with their broad patterns of expression, but the existence of other regulatory elements, or in the case of annexin A5 an alternative promoter, suggests that under certain circumstances tight transcriptional control may be exerted.

Hydra vulgaris expresses at least two annexins, of which annexin B12 (formerly annexin XII) is the best characterized. Annexin B12 was discovered first (270) and is clearly the major annexin in Hydra, being expressed at an estimated 100-fold excess of a second as yet uncharacterized annexin. Immunofluorescence analysis of whole Hydra revealed the staining pattern of the two annexins to be segregated, with annexin B12 being largely confined to epithelial battery cells throughout the tentacles, with the second Hydra annexin being maximally located in the cytoplasm of nematocytes (269). The epithelial battery cells differentiate from gastric ectodermal epithelial stem cells, whereas nematocytes differentiate from interstitial cells. The battery epithelial cells and nematocytes are closely aligned in Hydra tentacles; both are motile and both are actively turned over. The presence of annexins in these cells therefore fits with current models of annexin function in cell matrix adhesion and cell membrane plasticity and remodeling. The