Annexin A1 Crystal Structure: Interaction of Annexins with Membranes

Abstract

Annexins are structurally divided into a highly conserved core domain and a variable N-terminal domain. The core domain mediates
the calcium-dependent phospholipid binding of annexins, whereas the N-terminal domain, which is unique in sequence and length
for each member of this protein family, is responsible for the specificity among the different members. Annexin Al has been
shown to possess membrane aggregation and fusion activity in the presence of calcium in vitro. Due to the high sequence homology
of the core domain among different annexins, the property of membrane aggregation has been attributed to the N-terminal domain.
For instance, a chimera protein comprising the core of annexin 5, which by itself does not exhibit membrane aggregation properties,
and the N-terminal domain of annexin Al is able to induce membrane aggregation. Numerous three-dimensional structures of annexins
have been solved using x-ray crystallography, however, none reveal the tertiary structure of an N-terminal domain or its interaction
with the core domain. We have solved the x-ray structure of full-length annexin Al with an N-terminal domain comprising 42
amino acids in the absence of calcium ions at 1.8 Å resolution. Residues 2–26 of the N-terminal domain exhibit a mainly a-helical
conformation with a kink at residue 17. Helix NA (residues 2 to 16) inserts into repeat III of the core domain thereby replacing
the old helix D. Helix D on the other hand unfolds into an extended loop that forms a flap over the top of helix NA of the
N-terminal domain. As a result, the type II calcium-binding site located in core repeat III is destroyed because the “capping
residue” for calcium ion coordination is not in the proper conformationllocation any more. Also presented in this article
is the structure of full-length annexin Al in the presence of 1 mM CaCl2. The structure of full-length annexin Al in the absence of calcium ions is thought to represent the “inactive” form of the
protein. We provide a model for the annexin A1-induced membrane aggregation and discuss it in light of the literature published
to date.

Full text

Page 1

Annexins
114
Annexins, edited by Joanna Bandorowicz-Pikula. ©2003 Eurekah.com.
Annexin 1 Crystal Structure:
Interaction of Annexins with Membranes
Anja Rosengarth and Hartmut Luecke
Abstract
A
length for each member of this protein family, is responsible for the specificity among the
different members. Annexin 1 has been shown to possess membrane aggregation and fusion
activity in the presence of calcium in vitro. Due to the high sequence homology of the core
domain among different annexins, the property of membrane aggregation has been attributed
to the N-terminal domain. For instance, a chimera protein comprising the core of annexin 5,
which by itself does not exhibit membrane aggregation properties, and the N-terminal domain
of annexin 1 is able to induce membrane aggregation. Numerous three-dimensional structures
of annexins have been solved using x-ray crystallography, however, none reveal the tertiary
structure of an N-terminal domain or its interaction with the core domain. We have solved the
x-ray structure of full-length annexin 1 with an N-terminal domain comprising 42 amino acids
in the absence of calcium ions at 1.8 Å resolution. Residues 2-26 of the N-terminal domain
exhibit a mainly α-helical conformation with a kink at residue 17. Helix NA (residues 2 to 16)
inserts into repeat III of the core domain thereby replacing the old helix D. Helix D on the
other hand unfolds into an extended loop that forms a flap over the top of helix NA of the
N-terminal domain. As a result, the type II calcium-binding site located in core repeat III is
destroyed because the “capping residue” for calcium ion coordination is not in the proper
conformation/location any more. Also presented in this article is the structure of full-length
annexin 1 in the presence of 1 mM CaCl2. The structure of full-length annexin 1 in the ab-
sence of calcium ions is thought to represent the “inactive” form of the protein. We provide a
model for the annexin 1-induced membrane aggregation and discuss it in light of the literature
published to date.
nnexins are structurally divided into a highly conserved core domain and a variable
N-terminal domain. The core domain mediates the calcium-dependent phospholipid
binding of annexins, whereas the N-terminal domain, which is unique in sequence and
Introduction
Over the past 20 years, annexins have been described as calcium- and phospholipid-binding
proteins. Biochemical and cell biological experiments have been conducted in order to eluci-
date the properties of different annexins in vitro to further define their biological role. Annexins
share a high sequence homology among their core domains, and the core domain is responsible
for calcium-dependent lipid binding. The unique properties of each annexin depend on the
N-terminal domain, which is different in sequence and length for each member of this protein
family.1-4
Besides biochemical work, x-ray crystallography has been used in order to reveal the struc-
tural requirements for calcium- and membrane binding of several different annexins. In 1990
and 1992, Huber and co-workers published the first annexin structure, i.e., that of annexin 5
(PDB code: 1AVH & 1AVR).5,6 The overall shape of the protein can be described as a curved
disk with the calcium ions located on the convex site of the disk. This disk is composed of four

Page 2

115Annexin 1 Crystal Structure
repeats, which contain five α-helices (named A through E) each, while the loops that connect
α-helices A/B and D/E harbor the calcium-binding sites. Two new calcium-binding motifs
have been determined in annexins—type II and type III binding sites. They are different from
the type I, or EF-hand, calcium-binding sites, with respect to the coordination of the calcium
ion. In the type II binding-site the calcium ion is coordinated by three backbone carbonyl
oxygen atoms located in the AB loop, the side chain of an acidic amino acid that resides ap-
proximately 40 residues downstream in sequence from the AB loop (Asp or Glu) and two water
molecules (pentagonal bipyramidal coordination). Type III binding sites provide a different
coordination: the calcium ion is coordinated by two backbone carbonyl oxygen atoms, one
side chain of an acidic residue and three water molecules.
The prevailing view that calcium-dependent membrane binding occurs via the type II
sites on the convex surface of the protein was supported by x-ray structures of annexin 5 in the
presence of calcium ions and lipid analogs (PDB-code: 1A8A and 1A8B).7 At this point, the
calcium-dependent phospholipid-binding of annexins could be explained, but the mechanism
of membrane aggregation and fusion still remained obscure. Annexins I and II have been re-
ported to exhibit membrane aggregation properties. Although their respective core domains
are very similar to annexin 5, these two annexin species contain N-terminal domains with
more than 30 amino acid residues (annexin 5 has an N-terminal tail of 16 residues). The
N-terminal domain of annexin 1 contains 42, and that of annexin 2 contains 32 amino acid
residues.
In 1993, the x-ray structure of an annexin 1 derivative lacking the first 32 amino acid
residues was published by Huber and co-workers (PDB-code: 1AIN).8 The structure revealed a
typical annexin core domain, very similar to the one of annexin 5.6 Calcium ions were bound
in the AB-loops of each of the four repeats, plus two extra ones in the DE-loops of repeat I and
repeat IV. Due to the unanticipated proteolytic removal of the first 32 amino acid residues, no
information was available on the structure of the N-terminal domain itself or its interaction
with the core domain. Furthermore, no explanation was apparent as to why annexin 1 is ca-
pable of membrane aggregation and annexin 5 is not. Obviously, the N-terminal domain of
annexin 1 plays an important role in the membrane binding activities described in these bio-
chemical and structural experiments.
Annexin 1, the protein of interest in our research during the last six years, has been shown
to aggregate chromaffin granules and artificial membranes in the presence of calcium ions.9-13
Membrane fusion activity has also been reported.11,13 A multitude of biochemical experiments
have been conducted during the last ten years in order to understand the mechanism of the
annexin 1 induced membrane aggregation. Unfortunately the results are quite contradictory:
Ernst and co-workers tested an annexin 1/5 chimera in which they fused the first repeat of
annexin 1 (plus eight amino acid residues of the N-terminal domain) to repeats II to IV of
annexin 5. Annexin 5 has not been shown to promote membrane aggregation, whereas this
chimera did.14 Reutelingsperger’s group chose a different approach: they designed a chimera
comprising the first 45 residues of annexin 1 fused to the core domain of annexin 5 (residues
19-320). This chimera was also able to promote membrane aggregation, leading to the hypoth-
esis that the N-terminal domain of annexin 1 alone was responsible for membrane
aggegation.15,16 However, in 1994 Creutz and co-workers studied the aggregation properties of
proteolytically cleaved isoforms of annexin 1, and interestingly they could show that a trunca-
tion at Lys26 resulted in a protein (∆1-26 annexin 1) that mediated half-maximal membrane
aggregation in the presence of 32 µM Ca2+. Full-length annexin 1 and ∆1-29 annexin 1 re-
quired 63 µM Ca2+ whereas the ∆1-12 annexin 1 derivative required 0.1 mM Ca2+ to induce
comparable aggregation.12 Cho and co-workers published similar results in the late 1990s.
This group worked with truncated annexin 1 derivatives as well as mutant proteins. From
theses studies it became evident that the residues Lys26 and Lys29 themselves play a role in

Page 3

Annexins
116
membrane aggregation.17 The N-terminal domain is obviously very important for the annexin
1-induced membrane aggregation, however, from all the above mentioned studies no mecha-
nism could be derived.
Various research groups have described another interesting feature of annexin 1: after
calcium-dependent binding of annexin 1 to lipids, a secondary membrane-binding site be-
comes accessible. Interestingly, this secondary binding site exhibits neither calcium depen-
dency nor lipid specificity.9,13,18-20 These data suggest that the N-terminal domain of annexin
1 alone might provide a second membrane-binding site. This hypothesis is supported by a
publication from Lee and co-workers in which they present data that a peptide comprising
residues 1-26 of annexin 1 alone is able to form an α-helix in 50% trifluorethanol (TFE)/water
and 10 mM sodium dodecyl sulfate (SDS), suggesting membrane binding activity.21
We are interested in the annexin 1-induced membrane aggregation and even fusion of
membranes in vitro and in vivo. As described above a lot of biochemical data is available on the
aggregation and fusion properties of annexin 1, however, the structural requirements for this
interaction have not been examined in detail.
The Three-Dimensional Structure of Full-Length Annexin 1
in the Absence of Calcium Ions
Full-length recombinant porcine annexin 1 was expressed and purified according to
Rosengarth et al (1999).22 Crystals were grown in 2.2 M ammonium sulfate, 0.1 M Tris-HCl,
pH 8.5 at 4ºC.23 Native crystals diffracted to better than 1.8 Å at the Advanced Light Source
(ALS) in Berkeley, beamline 5.0.2 and the structure was solved using a combination of the
Molecular Replacement (MR) and the Multiple Isomorphous Replacement (MIR) technique.24
After mass spectrometry showed that the protein in the crystals was the full-length pro-
tein,23 the question arose: how does the N-terminal domain fold? Secondary structure predic-
tion using the program SOPMA for the N-terminal domain suggested a highly α-helical sec-
ondary structure for residues 1–26 including a kink at positions 16 and 17. Residues 27–42
were predicted to be in a random coil conformation.25 In agreement with this prediction is the
x-ray structure of an N-terminal peptide of annexin 1 comprising residues 2-15 in complex
with the S100A11 protein in which the peptide adopts an α-helical conformation (PDB-code:
1QLS).26 Our structure of full-length annexin 1 is in excellent agreement with the predictions
and peptide conformations—the first 26 amino acid residues adopt an α-helical conformation
with a kink at residue 17. However, the N-terminal domain is not simply located at the con-
cave side of the molecule only as one might have expected from the structure of annexin 1
lacking the first 32 amino acid residues.
Shown in Figure 1 is a comparison of the x-ray structures of the annexin 1 derivative
lacking the first 32 amino acid residues from N-terminus (PDB-code: 1AIN) and the full-length
protein (PDB-code: 1HM6). The overall structure of the full-length protein still resembles a
typical annexin core domain, however no calcium ions are bound. Shown in yellow is the
N-terminal domain of annexin 1, including the two α-helices (named NA and NB) with a
kink at residue 17. As mentioned before, the N-terminal helices are not just an extension of the
N-terminal tail shown in the annexin 1 structure without the complete N-terminal domain.
Surprisingly, the helices of the N-terminal domain interact with repeat III of the core domain
by replacing the D-helix of this repeat. As a result, the former D-helix unfolds into an extended
loop that is located on top of the NA helix, like a flap.
Figure 2 illustrates the conformational change observed in the full-length structure of
annexin 1 in comparison to the derivative structure in more detail. Inspecting this region in
stereo makes it even easier to see the dramatic conformational change in repeat III. The first
four to five amino acid residues of the amphipathic helix NA insert into the same place that is
occupied by part of the D-helix in the truncated annexin 1 structure. Helix C in the full-length
structure is shorter as in the truncated structure, therefore making the extended loop formed

Page 4

117Annexin 1 Crystal Structure
by the former D-helix even longer. Also very noticeable is the fact that the extended loop of the
former core domain D helix now folds over the NA helix of the N-terminal domain like a flap,
burying parts of the hydrophilic side of the amphipathic helix (see Fig. 3). Further examination
of the electron density for residues 2-8 reveals favorable packing of the hydrophobic residues
on the other side of this amphipathic helix (Met3, Val4, Phe7) into the hydrophobic pocket
formed by residues Phe221, Leu225, Phe237 and Val268 of repeat III.
The Crystallographic Annexin 1-Dimer and Consequences
for Calcium Binding
Full-length annexin 1 crystallizes as a dimer in the absence of calcium ions. It is very likely
that the observed dimer is a crystallographic rather then a biological one. The dimer contacts
occur via the Lys250 residues, whose ε-amino groups are coordinated by three backbone carbo-
nyl oxygen atoms in the AB-loop of repeat II in the respective opposing monomer, thereby
replacing the calcium ion observed in the truncated annexin 1 derivative.8 Therefore, the lysine
ε-amino group replaces the calcium ion coordinated in repeat II that was observed in the
structure of annexin 1 lacking the first 32 amino acid residues (see Fig. 4). The replacement of
calcium ions by lysine residues has also been observed in the case of the annexin 12 hexamer
and the annexin 12 mutant protein E105K (also a hexamer).27,28 If the observed dimer were of
biological significance, the protein could not bind to the membrane via the calcium ions any-
more. It has been shown that especially the calcium-binding site in repeat II is of great impor-
tance for membrane binding.29 Additionally, the type II binding-site in repeat III is destroyed
as well, because the helix D that is now unfolded into an unstructured loop no longer supplies
the capping residue (Glu255) for calcium coordination. Another reason to believe that the
dimer we observed in the crystal structure is not a biological one is that we could not detect
dimers in solution using dynamic light scattering (data not shown).
Model for the Annexin 1-Induced Aggregation of Membranes:
Monomeric Annexin 1 Displaying Two Membrane-Binding Sites
The cartoon shown in Figure 5 illustrates a possible mechanism of the annexin 1-induced
aggregation of membranes. This model is supported by our x-ray structure of full-length annexin
1 and by results published by other groups over the last ten years.13,18,20,21,24 In this model, the
structure of full-length annexin 1 in the absence of calcium ions reflects the starting point of
the reaction cascade that eventually causes membrane aggregation. This reaction cascade would
start with calcium-dependent binding of one annexin 1 monomer to a membrane containing
negatively charged phospholipids. During this binding event, the N-terminal helix NA would
be expelled from the core of the protein, because helix D of repeat III is required to refold for
calcium-dependent lipid binding (step 1). Once the N-terminal domain is not shielded by the
annexin core any more, its amphipathic helix NA could interact with another membrane (step
2). This secondary membrane binding is not calcium-dependent and non-specific with respect
to the lipids in the membrane.13,20 This scheme proposes that the secondary lipid-binding site
in annexin 1 is provided by the N-terminal helix NA. Residues 2-26 of annexin 1 have been
shown by NMR experiments to adopt a helical conformation in 50%TFE/water and 10 mM
SDS, suggesting that this peptide interacts with membranes.21
The mechanism depicted in Figure 5 is one possibility for annexin 1 to be part of the
membrane aggregation and fusion machinery. Other possibilities include the dimerization of
two annexin 1 molecules via their N-terminal amphipathic helices (hydrophobic interaction!)
or the formation of a heterotetramer composed of two annexin 1 monomers and one S100A11
dimer.30,31 It has been observed by cryo-electron microscopy that annexin 1 bridges
dioleoylphosphatidylglycerol (DOPG)-dioleoylphosphatidylcholine (DOPC) pairs of bilayers

Page 5

Annexins
118
Figure 1. Overlay of the ribbon diagrams of one monomer of recombinant porcine annexin 1 comprising
protein core and the N-terminal domain (solid rendering, PDB-code: 1HM6)24 with human annexin 1
lacking the first 32 amino acids shown (slightly shifted transparent rendering, ∆1-32 anx1; PDB-code
1AIN).8 Repeat I is presented in red, repeat II in green, repeat III in blue, repeat IV in purple and the
N-terminal domain in yellow. The yellow N-terminal helix in the full-length structure is replacing the
two-turn blue helix in the core of the ∆1-32 annexin 1. Bound calcium ions in ∆1-32 annexin 1 are
illustrated as yellow spheres.
Figure 2. Stereo overlay of the three-dimensional structures of repeat III of full-length annexin 1 comprising
helices A, B, C and E (blue) plus residues 2 to 26 of the N-terminal domain (yellow) with repeat III of ∆1-32
annexin 1 (gray). Note that helix D of ∆1-32 annexin 1 (gray) unwinds into an extended loop (loop D, dark
blue) in the full-length structure, which forms a flap over the N-terminal helix. The yellow arrows indicate
the direction of the N-terminal α-helices, which are connected to repeat I of the core domain via an extended
linker.