The APOBEC-2 crystal structure and functional
implications for the deaminase AID
Courtney Prochnow1*, Ronda Bransteitter1*, Michael G. Klein1, Myron F. Goodman1& Xiaojiang S. Chen1
senger RNA-editing enzyme catalytic (APOBEC) polypeptides,
which deaminates mRNA and single-stranded DNA1,2. Different
APOBEC members use the same deamination activity to achieve
diverse human biological functions. Deamination by an APOBEC
ical for generating high-affinity antibodies3, and deamination by
APOBEC-3 proteins can inhibit retrotransposons and the replica-
tion of retroviruses such as human immunodeficiency virus and
hepatitis B virus4–7. Here we report the crystal structure of APO2.
APO2forms arod-shaped tetramerthatdiffers markedlyfrom the
square-shaped tetramer of thefree nucleotide cytidine deaminase,
with which APOBEC proteins share considerable sequence homo-
logy. In APO2, two long a-helices of a monomer structure prevent
the formation of a square-shaped tetramer and facilitate forma-
tion of the rod-shaped tetramer via head-to-head interactions of
two APO2 dimers. Extensive sequence homology among APOBEC
family members allows us to test APO2 structure-based predic-
tions using AID. We show that AID deamination activity is
and substrate access. The structure suggests how mutations in
patients with hyper-IgM-2 syndrome inactivate AID, resulting in
defective antibody maturation.
We crystallized APO2, which contains amino acid residues
41–224, with four monomers in each asymmetric unit that form a
tetramer with an atypical elongated shape (Fig. 1a). This tetramer
*These authors contributed equally to this work.
1Molecular and Computational Biology, University of Southern California Los Angeles, California 90089, USA.
127 Å52 Å
β β1′ ′
Figure 1 | ThestructureofAPO2. a,TheAPO2tetramerstructure.Ithasan
accession number: 1MQ0), a fntCDA. c, d, The APO2 monomer structure
rotated by 90 degrees, showing the unique features of APO2: the short b19
strand and helices h4 and h6. h4 and h6 dictate how APO2 oligomerizes.
e, The APO2 dimer formed by two monomers (in purple and yellow). Each
has a different conformation for the h1/b1-turn (in red): a loop (L1) and a
hairpin. f, The tetrameric interface, showing the extensive interactions
charged amino acids in h4, h6 and L1) of the interactions at the tetramer
Vol 445|25 January 2007|doi:10.1038/nature05492
assembles through two different monomer–monomer interfaces, in
contrast to the canonical square shape of the free nucleotide cytidine
deaminase (fntCDA) tetramer (Fig. 1b), in which all four monomers
of a butterfly (Fig. 1a) with an end-to-end span of approximately
The APO2 monomer appears to adopt the typical core fold of the
(Figs 1c, d). However, one new attribute is the additional a-helices
surrounding the core b-sheet (Figs 1c, d); six long helices are present
fntCDA monomer (excluding the shorter 310helices)8–11. Helices h3
helices within the monomer subunit. On the basis of the close
sequence homology of APO2 with other APOBEC proteins, the long
helix (h4) probably serves as a structural signature of this family
(Figs 1c, d).
The APO2 dimer is formed by pairing two long b-strands (b2)
resembles a ribcage (Fig. 1a, e). Twelve residues (residues 82–93) on
each b2 strand form 12 hydrogen bonds through main-chain atoms,
providing the principal bonding force between the two monomers.
The dimer interface is reinforced by the side-chain interactions
occurring through the loops and helices located on both sides of
the b-sheet. Ordered water molecules also help to stabilize this
The dimer is nearly symmetrical (Fig. 1e) with six helices (h2, h3
and h4 of both molecules) located on one side of the augmented
b-sheet and four helices (h1 and h5 of both monomers) on the other
side. Capped on both edges of the b-sheet are h4 and h6. However,
one part of the dimer shows obvious asymmetry at the turn between
h1 and strand b1 (h1/b1-turn). This h1/b1-turn (residues 57–68)
assumes a hairpin structure (b19-hairpin) in one monomer, and a
loop conformation (L1) in the other monomer (Fig. 1e, f).
The APO2 tetramer is formed by two dimers joining through
head-to-head interactions. The two dimers make extensive contacts
via the residues from h4 and h6, as well as the loop L1 at the h1/b1-
turn (Fig. 1f). Residues Y61, F155, M156, W157, P160, Y214 and
Y215 from each side of the interface form extensive hydrophobic
packing interactions, and residues R57, S62, S63, R153, E158, E159
and E161 establish salt bridges and hydrogen bonds (Fig. 1g). Some
charged residues even use their aliphatic side chains to interact with
hydrophobic residues. Thus, hydrophobic, polar and charged amino
acid side chains are all involved in the tetramerization interactions.
The total buried area is 1,745A˚2within the tetramer interface, where
h4 and h6 play amajor role forming the interface (Fig. 1f). h4 and h6
also sterically hinder the formation of the square-shaped fntCDA-
need to be. Therefore, h4 and h6 appear to determine directly the
elongated tetramer formation.
A prominent feature of the APO2 tetramer distinctive from the
Figure 2 | The APO2 active site. a, The APO2 active sites are accessible to
only to free nucleotides. c, The outer APO2 active sites show Zn
coordination(yellowdashedlines)bythree residues(H98,C128, C131)and
hydrophobic ring of Y61 interacts with the guanidine group of R65,
stabilizing the conformation. f, In the h1/b1 loop, the E60 coordinates with
Zn. Y61 now rotates away from R65 and interacts with R57, facilitating the
disruption of the b19-hairpin and stabilizing the loop conformation.
g, Superimposed monomers show that the h1/b1 loop (purple) is pulled
down ,8.5A˚towards the active site owing to the E60–Zn bond formation.
NATURE|Vol 445|25 January 2007
DNA substrates (Fig. 2a). In the square-shaped fntCDA tetramer,
loops from two neighbouring monomers cover the active sites so
that only small free nucleotides can bind to the buried sites
(Fig. 2b). Although the yeast fntCDA, CDD1, has been reported to
deaminate the apolipoprotein B mRNA in vitro, its known biological
substrate in vivo is a free nucleotide and the CDD1 structure is a
canonical square-shaped tetramer9.
In fntCDAs, the active centre Zn atom is coordinated by three
residues (either three cysteines, or two cysteines1one histidine)
and forms a fourth bond with a water molecule with a bond distance
of ,3.0A˚(ref. 10). This type of Zn coordination is also present in
APO2 (Fig. 2c), but only in the twoouter monomers ofthe tetramer.
Surprisingly, the active sites for the other two monomers in the
molecule and makes the fourth coordination bond with the Zn
(Fig. 2d). This coordination of Zn by four residues is unexpected
given that all known fntCDA structures have only three amino acid
residues participating in Zn coordination8–13.
A closer examination of the structure reveals a ‘built-in’ mech-
anism for a conformational switch between the two types of Zn
coordination. The switch is mediated by sequences contained in
the h1/b1-turn, in which E60 is located. The h1/b1-turn can adopt
either a hairpin (b19-hairpin, Fig. 2e) or a loop (L1) conformation
E60 is located 6A˚from the Zn when the h1/b1-turn is a b19-hairpin
(Fig. 2e). The b19-hairpin is stabilized by main-chain hydrogen
bondswithin theb19-hairpinandreinforcedbyinteractions between
an APO2 tetramer, the h1/b1-turn folds into a loop (L1, Fig. 1f). In
this conformation, Y61 rotates its side chain to interact with R57
instead of the R65 (Fig. 2f, g). The new pairing of Y61 with R57
destabilizes the b19-hairpin while stabilizing the loop. In the loop
conformation, the E60 is 2.2A˚from the Zn (Fig. 2f).
The hairpin–loop switch may have two important consequences.
First, switching to the loop and forming the fourth Zn coordination
by E60 prevents coordination by water and subsequent hydroxyla-
tion of Zn necessary for deamination (Fig. 2d, f). Second, Zn coor-
dination by E60 pulls the h1/b1-turn approximately 8.5A˚towards
theactivecentre (Fig.2g),which couldrestrict substrateaccess tothe
of E60 may allow the loop to move away from the active centre to
form the b19-hairpin as observed in the outer monomers. The E60
would no longer prevent the Zn hydroxylation and nucleic acid sub-
strate access to those active sites. Thus, the hairpin–loop switch can
be a molecular mechanism for regulating substrate access and
enzyme activity mediated through Zn coordination.
The APO2 fragment in the structure shares a 33.3% amino acid
identity (44.6% homology) with AID, and the buried residues in
APO2 share a 75% identity (96% homology) with AID (Fig. 3a).
The highly conserved residues buried inside the structure and
those located at the dimer/tetramer interfaces strongly suggest a
structural conservation of AID with APO2. Thus, the APO2 crystal
structure should provide functional insights for AID and other
activity of APO2. For this reason, we use AID as a surrogate to test
how mutations guided by APO2 structure affect AID deamination
We generated glutathione S-transferase (GST)–AID mutants with
amino acid substitutions located atthe tetrameric interface (Fig. 3b),
and showed that the mutants either had no detectable or signifi-
cantly reduced deaminase activity compared to wild-type GST–
AID (Fig. 3d–f). Mutants R112C and Y114A/F115A were inactive
(Fig. 3e, f), while mutants K16A and C116A had a 3.3-fold reduction
in activity (Figs 3e, f).
AID mutations within the predicted dimerization domain, F46A/
Y48A (Fig. 3c), displayed a fourfold decrease in deamination activity
K16A R112C Y114A
Single-stranded DNA deamination
(pmol µg–1 h–1)
3′ + AID
3′ + NaOH
3′ + UDG
Figure 3 | Structurally guided mutagenesis of
AIDimpairs deamination activity. a, A sequence
alignment of APO2 and AID, showing significant
homology. Red, identical residues; grey shading,
buried residues; red squares, active centre
residues; green dots, tetrameric interface
residues; blue diamonds, dimeric interface
residues; purple stars, HIGM mutated residues;
and black triangles, mutated AID residues.
b, Mutated AID residues (in green) at the
tetramer interface as modelled based on the
APO2 structure. c, Mutated AID residues (in
green) in the dimer interface as modelled based
cytidine deamination assay. F, fluorescein; UDG,
is uracil DNA glycosylase. e, Denaturing PAGE
analysis of the deamination activity for wild-type
band indicates deamination activity. f, Bar
representation of the specific activities for wild-
type and mutant AID proteins.
NATURE|Vol 445|25 January 2007
(Fig.3e,f).Thedimerinterface isextensive, sotwomutationsshould
not completely disrupt dimeric AID. This explains why weak deami-
nation activity was observed with this double AID mutant. These
mutational results suggest that the residues within the predicted
dimeric and tetrameric interfaces are important for deamination
activity. One caveat is that the residues on the tetramer interface
are also present on the exposed surface of the outer ends of the
tetramer and thus could also be involved in an additional role beside
tetramerization (Figs 3b, 4d). We noticed that in gel filtration assays
the tetramer was a minor species when compared with the dimer,
suggesting a stronger dimeric interaction.
AID has an arginine (R19) at the equivalent position of the APO2
E60 residue (Fig 3a) that may have a negative regulatory role for
APO2 activity by blocking Zn hydroxylation and substrate access
(Fig. 2f). We showed that an AID R19E mutant mimicking APO2
E60 had a significantly decreased deamination activity (about 4.6-
fold lessthanthewildtype,Figs 3e,f). Similarly,theAID R24residue
is equivalent to APO2 R65, which interacts with Y61 of APO2 to
stabilize the open b19-hairpin conformation. We predicted that the
into the closed loop conformation to block substrate access and
pletely inactive on single-stranded DNA (Fig. 3e, f).
Mutations in human AID cause hyper-IgM-2 (HIGM-2) syn-
drome, characterized by an impaired production of high-affinity
antibodies14,15. The mutated AID residues of HIGM-2 patients are
highly conserved in APO2 (Fig. 4a). A plausible explanation for why
and how HIGM-2 mutations disrupt AID function is given by the
structure of APO2 (Fig. 4b–e). On the basis of the crystal structure,
HIGM-2 AID mutations can be divided into four classes. The first
mutant class (A111, R112, L113 and N168) occurs at the tetramer-
ization interface (Fig. 4b). The second mutant class includes residues
in and near the active centre (Fig. 4c), H56, E58, S83, S85 and C87,
which are conserved among all APOBEC enzymes. The AID R24
residue is also mutated inHIGM-2 patients. As previously discussed,
R24 may stabilize the b19-hairpin, which keeps the active site open
for DNA/RNA access. The third class consists of residues located on
the enzyme surface (Fig. 4d), including those residues located at the
tetramer interface (A111, R112, L113, N168; see Fig. 4d). A fourth
class of HIGM-2 AID mutations are those with large hydrophobic
side chains buried within the core (Fig. 4e), including W80, L106,
M139 and F151. Three of these residues are located near the active
centre. Mutating these residues should disrupt the folding and
stability of AID.
Since many of the APOBEC enzymes are reported to form dimers
and multimers, the APO2 structure may shed light on how these
enzymes oligomerize16–23. The elucidation of the APO2 structure,
fortified by the structure-guided predictions for the activity of spe-
cific AID mutants, provides astructural basis to pursue further func-
tional studies of APOBEC proteins with an eye towards developing
therapeutic strategies to deal with deficiency in deaminating cytidine
and to restrict retroviral replication.
Crystallography statistics can be found in the Supplementary Information.
224 was cloned and expressed in Escherichia coli as a recombinant GST fusion
protein.Following GSTcleavage by thrombin, furtherpurification of APO2was
achieved using Superdex-75 gel filtration chromatography. Native and sel-
enium-methionine labelled protein was concentrated to 15mgml21in a buffer
containing 25mM Hepes, pH7.0, 50mM NaCl and 10mM dithiothreitol.
Crystalsweregrownat 18uC by hanging-drop vapourdiffusionfrom a reservoir
R24W H56Y E58K W80R S83P S85N C87R L106P
H98E100 W121 S124
L113P M119V F151S N168S
Figure 4 | AIDHIGM-2mutations. a,AlignmentofmutatedresiduesofAID
from HIGM-2 patients with the corresponding residues in APO2, showing
high sequence conservation. b, Mapping the residues in AID HIGM-2
mutations (R112, L113, N168) to the tetramer interface as modelled from
the APO2 structure. c, Mapping the AID HIGM-2 mutations, S83 and S85,
near the active site. d, Mapping the AID mutations, K16, Y114/F115 and
C116 (in green), to the exposed surface of an outer monomer. The HIGM-2
AID residues (R112, L113, N168, in yellow), which are at the tetramer
interface (b), are also located on this exposed surface. e, Mapping of AID
HIGM-2 mutations, W80, L106, M139 and F151, to the interior core
NATURE|Vol 445|25 January 2007
solution of 85mM Na-citrate, pH5.6, 160mM LiSO4, 24% (weight/volume)
polyethylene glycol monomethyl ether and 15% glycerol.
Structure determination and refinement. Native and selenium-multiwave-
length anomalous diffraction (MAD) data were collected at the synchrotron
and processed using HKL200024(Supplementary Table 1). An initial solution
was obtained at 3.5A˚using the program Solve25and located eight selenium
atoms using the peak wavelength data set. A search with SHELXD26found four
additional selenium atoms (totalling twelve) and, subsequently, the program
SHARP identified four additional weaker anomalous scattering atoms, which
and fourfold non-crystallographic symmetry (NCS) averaging were applied
using the program RESOLVE27. Additionally, phase extension in RESOLVE
was performed with the native data set to 2.5A˚resolution using the two-wave-
length MAD phases calculated in SHARP. The molecular model was built based
on this experimental map using the program ‘‘O’’ and was refined with the
Crystallography and NMRSystem (CNS).A twofoldNCS constrainwas applied
during the initial simulated annealing, but the final refinement was carried out
without NCSconstrain. Theproteingeometryis excellentwhenexaminedusing
the program PROCHECK.
Construction of AID mutants. Mutant AID proteins were contructed by site-
directed mutagenesis using the pGEX-KG-AID vector as the PCR template and
CTC TAC TTC TGT GCG GCC CGC AAG GCT GAG CCC GAG-39 (E117/
E118A), 59-AAG TTT CTT TAC CAA TTC GCA AAT GTC CGC TGG GCT
AAG-39 (K16A),59-ACCGCGCGCCTCTAC TTCGCTGAG GACCGCAAG
GGT CGG CGT-39 (R19E), and 59-ACA TCC TTT TCA CTG GAC GCT GGT
GCT CTT CGC AAT AAG AAC GGC-39 (F46A/Y48A). Mutant constructs were
verified by DNA sequencing.
described28with the exception that the substrate used was a fluorescein-dT
incorporated single-stranded DNA substrate (59-taa agg fluorescein-dTga
aga gag gag aga gaa gta agC tga aga gag aga agg aag aga gtg aag gag-39).
Reaction products were visualized on a BioRad FX scanner.
Received 10 October; accepted 28 November 2006.
Published online 24 December 2006.
1.Pham, P., Bransteitter, R. & Goodman, M. F. Reward versus risk: DNA cytidine
Conticello, S. G., Thomas, C. J. F., Petersen-Mahrt, S. K. & Neuberger, M. S.
Evolution of the AID/APOBEC family of polynucleotide (deoxy)cytidine
deaminases. Mol. Biol. Evol. 22, 367–377 (2004).
Bransteitter, R., Sneeden, J. L., Allen, S., Pham, P. & Goodman, O. M. F. First AID
(activation-induced cytidine deaminase) is needed to produce high affinity
isotype-switched antibodies. J. Biol. Chem. 281, 16833–16836 (2006).
Chiu, Y. L. & Greene, W. C. Multifaceted antiviral actions of APOBEC3 cytidine
deaminases. Trends Immunol. 27, 291–297 (2006).
resistance factors. J. Virol. 80, 1067–1076 (2006).
Franca, R., Spadari, S. & Maga, G. APOBEC deaminases as cellular antiviral
Bonvin, M. et al. Interferon-inducible expression of APOBEC3 editing enzymes in
human hepatocytes and inhibition of hepatitis B virus replication. Hepatology 43,
Johansson, E., Mejlhede, N., Neuhard, J. & Larsen, S. Crystal structure of the
tetrameric cytidine deaminase from Bacillus subtilis at 2.0A˚resolution.
Biochemistry 41, 2563–2570 (2002).
9. Xie,K. etal. The structure ofayeastRNA-editingdeaminase provides insight into
the fold and function of activation-induced deaminase and APOBEC-1. Proc. Natl
Acad. Sci. USA 101, 8114–8119 (2004).
10. Teh, A. et al. The 1.48A˚resolution crystal structure of the homotetrameric
cytidine deaminase from mouse. Biochemistry 45, 7825–7833 (2006).
bound to a potent inhibitor. J. Med. Chem. 48, 658–660 (2005).
The 2.3A˚crystal structure of an enzyme: transition-state analog complex. Curr.
Biol. 235, 635–656 (1994).
13. Smith, A. A., Carlow, D. C., Wolfenden, R. & Short, S. A. Mutations affecting
transition-state stabilization by residues coordinating zinc at the active site of
cytidine deaminase. Biochemistry 33, 6468–6474 (1994).
14. Durandy, A., Peron, S. &Fischer, A. Hyper-IgM syndromes. Curr. Opin. Rheumatol.
18, 369–376 (2006).
with hyper IgM syndrome. Clin. Immunol. 97, 203–210 (2000).
16. Anant, S. et al. ARCD-1, an apobec-1-related cytidine deaminase, exerts a
dominant negative effect on C to U RNA editing. Am. J. Cell Physiol. 281,
17. Jarmuz, A. et al. An anthropoid-specific locus of orphan C to U RNA-editing
enzymes on chromosome 22. Genomics 79, 285–296 (2002).
18. Shindo, K. et al. The enzymatic activity of CEM15/Apobec-3G is essential for the
regulation of the infectivity of HIV-1 virion but not a sole determinant of its
antiviral activity. J. Biol. Chem. 278, 44412–44416 (2003).
19. Wiegand, H. L., Doehle, B. P., Bogerd, H. P. & Cullen, B. R. A second human
antiretroviral factor, APOBEC3F, is suppressed by the HIV-1 and HIV-2 Vif
proteins. EMBO J. 23, 2451–2458 (2004).
20. Opi, S. et al. Monomeric APOBEC3G is catalytically active and has antiviral
activity. J. Virol. 80, 4673–4682 (2006).
21. Navarro, F. et al. Complementary function of the two catalytic domains of
APOBEC3G. Virology 333, 374–386 (2005).
22. Wang, J. et al. Identification of a specific domain required for dimerization of
activation-induced cytidine deaminase. J. Biol. Chem. (in the press).
23. Teng, B. et al. Mutational analysis of apolipoprotein B mRNA editing enzyme
(APOBEC1): structure-function relationships of RNA editing and dimerization. J.
Lipid Res. 40, 623–635 (1999).
24. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol. 276, 307–326 (1997).
25. Terwilliger, T. C. & Berendzen, J. Automated MAD and MIR structure solution.
Acta Crystallogr. 55, 849–861 (1999).
26. Schneider, T. R. & Sheldrick, G. M. Substructure solution with SHELXD. Acta
Crystallogr. 58, 1772–1779 (2002).
27. Terwilliger, T. C. Maximum-likelihood density modification. Acta Crystallogr. 56,
28. Bransteitter, R., Pham, P., Scharff, M. D. & Goodman, M. F. Activation-induced
cytidine deaminase deaminates deoxycytidine on single-stranded DNA but
requires the action of RNase. Proc. Natl Acad. Sci. USA 100, 4102–4107 (2003).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank L. Chen for comments on the manuscript. We also
the Argonne National Laboratory for assistance in data collection. This work was
supported in part by National Institutes of Health grants to M.F.G. and X.S.C. and
an NIH-NIA Predoctoral Traineeship to R.B.
Author Information The structure of APO2 has been uploaded to the Protein Data
Bank under accession number 2NYT and to the Research Collaboratory for
Structural Bioinformatics (RCSB) under accession number RCSB040471. Reprints
declare no competing financial interests. Correspondence and requests for
materials should be addressed to X.S.C. (firstname.lastname@example.org).
NATURE|Vol 445|25 January 2007