The crystal structure of A3G191-384-2K3A has a core α-β-α fold consistent with other cytidine deaminases (Chen et al., 2008
; Holden et al., 2008
; Prochnow et al., 2007
). As seen with the other A3G-CTD structures (Harjes et al., 2009
; Holden et al., 2008
), this structure has a five-stranded β-sheet surrounded on both sides by six α-helices (). Secondary structural elements are numbered after the A3G191-384-2K3A NMR structure (PDB ID: 2KEM) (Harjes et al., 2009
) so that consistent comparisons can be made between the structures. The second β-strand is discontinuous, as also observed in the wild-type A3G-CTD (Furukawa et al., 2009
) and A3G-CTD-2K3A (Chen et al., 2008
; Harjes et al., 2009
) NMR structures. As previously observed, the catalytic zinc is coordinated directly by H257, C288, and C291, and indirectly by the catalytic residue E259 via a water molecule. Thus, in terms of the overall fold, this structure is similar to the previously published structures but many key specific differences exist.
Crystal structure comparison
Comparison of A3G191-384-2K3A with the 3E1U crystal structure reveals numerous differences (highlighted in green and blue in ). Some of these differences are in loop regions but other variations include large regions of secondary structure, notably the β1-β2 region (M227-Q237), the β2′-α2 region (H248-G255), and the α2-β3 region (P267-D274). In the A3G191-384-2K3A crystal structure, the second β-strand is discontinuous, with residues L235-R239 forming a prominent “bulge”. This has been observed in both mutant and wild-type A3G-CTD NMR structures (Furukawa et al., 2009
; Harjes et al., 2009
; Chen et al., 2008
Figure 2 (A) Major regions of structural differences between A3G191-384-2K3A and 3E1U highlighted in blue and green. Residues F268-D272, which were mis-traced in the 3E1U structure are highlighted in red. (B) NOE violations of more than 6 Å (red lines (more ...)
Crystallography and NMR spectroscopy are complementary techniques with well-established methods for verifying the structural integrity of protein structures. PDB_REDO recently organized some of the verification of crystal structures in the Protein Data Bank (Joosten et al., 2009b
; Joosten et al., 2009a
; Sanderson, 2009
). Unfortunately, the only other A3G-CTD crystal structure in the database was flagged as having discrepancies in the PDB_REDO database (PDB ID: 3E1U with R-factor 25.2%, R-free 26.7%, Resolution 2.3 Å). Many of the regions in the 3E1U structure (Holden et al., 2008
) that differ from the NMR structures, are regions where the 3E1U structure’s experimental data are the weakest (α-carbon B-factors shown in ). The regions differing between these structures include β1-β2, β2′-α2 and α2-β3.
Evidence for ambiguity in the 3E1U structure is apparent in four ways: (1)
36 NOE restraints are violated in the 3E1U structure, by more than 6 Å, with respect to NMR structural data; including in the very well-ordered α-helical and β-sheet regions that are well-established by NMR ( and Experimental Procedures). (2)
In addition to being flagged in PDB-REDO, analyses of the structure and experimental structure-factors of the 3E1U structure, downloaded from the Protein Data Bank (www.rcsb.org
), revealed seven residues that were in unfavorable Ramachandran space (3 in disallowed and 4 in generously allowed regions). Most significantly, residues F268-D272 are completely mis-traced in the electron density, burying K270 away from the surface (). These residues, located near the end of α2-helix, may be essential for the stability of the A3G-CTD active site because mutating W269 or L271 abrogates deaminase activity (Chen et al., 2008
). When we re-fit this region in the 3E1U data, the final turn of α2-helix fit better within the electron density (). Thus, the structure in this region converged towards the conformation observed in the NMR structures not the one originally modeled in the 3E1U structure. The electron density in two other regions, β1-β2 and β2′-α2, remained ambiguous, making refitting unattainable. (3)
In contrast to these regions in the 3E1U structure, our crystal structure of A3G-191-384-2K3A (R-factor 16.57%, R-free 20.82%, Resolution 2.25 Å) is well-ordered. Our structure of A3G191-384-2K3A has only 9 NOE restraint violations to the NMR data; the α2-helix is well ordered; and the structure has lower α-carbon B-factors () (). (4)
Mass spectrometry analysis validates that the β1-β2 structure is conserved in full-length A3G. In contrast to the 3E1U structure, all NMR structures and the A3G191-384-2K3A crystal structure show that residues M227 and W232 are adjacent to each other in the two neighboring beta-strands (β1-β2). This was verified by introducing two cysteine mutations: M227C in β1 and W232C in β2, into both full-length A3G-2K3A (Fig. S1
) and A3G (data not shown). Mass spectrometry determined that these two residues, M227C and W232C, form a disulfide bond and no peptides containing free M227C or W232C are detected in either construct. In the 3E1U structure, M227 and W232 are not adjacent, rather R226 and L235 (part of the putative “continuous β2” strand) face each other. Applying similar mass spectrometry analysis to full-length A3G, with mutations R226C and L235C, did not result in a detectable peptide containing a disulfide bond (data not shown).
Authors of the 3E1U structure write: “Therefore, an intact full-length β2 strand and the five-stranded β-sheet core is probably the feature of wild-type APOBEC3G-CD2 and all other APOBEC proteins.” (Holden et al., 2008
). However, the A3G-191-384-2K3A crystal structure and the additional experimental data described above do not support this statement. These data suggest that the “bulge” observed between the β2-β2′ strands in NMR experiments and the A3G-191-384-2K3A crystal structure, is not merely an experimental artifact but an intrinsic feature of A3G-CTD structure.
Analysis of A3G191-384-2K3A crystal packing interfaces
The two molecules in the asymmetric unit in the crystal structure of A3G191-384-2K3A pack in such a way as to produce four major interfaces, all of which are the result of non-crystallographic symmetry, with surface areas of: 901 Å2, 604 Å2, 427 Å2 and 246 Å2, respectively (–). Multiple sequence alignment of the A3G-CTD to ten other homologs: A3G-NTD, A3A, A3B-CTD/NTD, A3C, A3DE-NTD/CTD, A3F- NTD/CTD and A3H reveals that many of the residues contributing more than 30 Å2 to these interfaces are unique to A3G-CTD ().
Interface 1 between two A3G-CTD molecules in the asymmetric unit (A) Surface representation and (B) details of the largest interaction interface. Molecule A is shown in dark green, molecule B is shown in light green.
Interface 4 between two A3G-CTD molecules in the asymmetric unit (A) Surface representation and (B) detailed interactions. Same coloring scheme as in .
Figure 7 Multiple sequence alignment of all human APOBEC3 protein sequences (Refseq) by their respective NTDs and CTDs aligned to the A3G-CTD sequence. Residues contributing more than 30 Å2 to the buried surface area on an inter-molecular interface are (more ...)
The largest interaction interface (901 Å2
) displays excellent shape complementarity as observed from the surface representation of interfacing molecules (). Extensive surface contacts are observed, primarily between identical residues at the α1-loop-β1 from both molecules in the asymmetric unit. Twelve residues in each molecule contribute at least 30 Å2
to the interface, however, residues W211, R213 and Q318 together contribute over 450 Å2
of the interfacial area (Table S1
) (). None of these three residues are conserved across the other APOBEC3’s (). Three direct hydrogen bonds, nine water-mediated hydrogen bonds and one ionic interaction, occur across the entire interface (Table S2
). All of these contacts verify the intimacy of this extensive interface. As the residues forming this extensive interface are not conserved among the other APOBEC3’s or the NTD of A3G (), this interface may be unique to A3G-CTD.
To assess the functional significance of this interface, a variant of A3G was made with the three amino acid substitutions W211A, R213A and R374E designed to profoundly disrupt the observed packing. This variant shows near undetectable levels (background) of DNA deaminase activity in vitro
and abrogated anti-viral activity in the Vif-deficient HIV-1 reporter virus assay (). In contrast, the active site A3G-E259Q variant is catalytically dead and unable to inhibit Vif-deficient HIV-1, in agreement with prior studies (Schumacher et al., 2008
; Hache et al., 2008
) (). All of the variants studied have near normal cellular expression levels and incorporate into viral particles (). Thus the residues that are integral to the first interface in the A3G191-384-2K3A crystal structure are essential to both A3G deaminase and anti-viral activity.
Figure 8 DNA deaminase activity and HIV-1 restriction data. (A) A graph showing the results of a DNA oligonucleotide-based deamination assay with the indicated A3G-GFP constructs (MOCK: GFP only; WT: A3G-GFP; E259Q: A3G-E259Q-GFP; INT1: interface 1 A3G-GFP mutant (more ...)
The second largest interface (604 Å2
) involves the β2′-loop-α2 residues 247–254 in each of the two molecules of the asymmetric unit (). At this interface, eight residues of each molecule bury extensive surface area, with the largest surface area being buried by residues F252 and F268 (Table S1
). This loop also coordinates an intermolecular zinc-binding site. H248 and H250 of one molecule in the asymmetric unit and the second molecule’s C261 and D264 (through a water-mediated hydrogen bond) coordinate an intermolecular zinc ion (Table S2
). The combination of these four residues occurring simultaneously is unique to A3G-CTD amongst the APOBEC3’s (). An additional four hydrogen/water mediated bonds are formed within this interface. This zinc-coordinating interface may provide insights into how a metal mediated switch could specifically modulate A3G oligomerization.
Figure 4 Interface 2 between two A3G-CTD molecules in the asymmetric unit (A) Surface representation and (B) detailed interactions. Same coloring scheme as in . (C) Interface 2 overview with coloring based on data from HSQC spectra. Blue residues: signal (more ...)
NMR experiments support Zn2+
mediated oligomerization at this interface. Through titration of Zn2+
into isotope labeled A3G191-384-2K3A, loss of signal intensities and chemical shift perturbations were observed at 1mM Zn2+
(labeled blue, Fig. S2
). The residues most affected are located in the β2′-loop-α2 and α2 regions. In fact, these residues in the A3G191-384-2K3A crystal structure (blue, ) map predominantly to the zinc-coordinating second interface, described above, indicating that this interface may contribute to zinc mediated oligomerization in solution.
To test the ability of zinc to modulate oligomerization a further series of NMR experiments was performed. A reference HSQC spectrum was taken at 50μM of Zn2+ concentration () then, Zn2+ concentration was increased to 1.25mM causing the disappearance of amide proton NMR signals (). Next, 0.4 mM ethylenediaminetetraacetic acid (EDTA) was added to chelate the free Zn2+. This caused the signals within the NMR spectra to reappear (). The similarity of chemical shifts, signal intensities and lineshapes of HSQC signals in combine to suggest that zinc mediates the equilibrium of A3G-CTD between oligomeric () and monomeric states.
Zn2+ dependent A3G191-384-2K3A aggregation is reversible. HSQC spectra of A3G191-384-2K3A with (A) 50μM Zn2+, (B)1.25mM Zn2+, and (C)1.25mM Zn2+ / 0.4mM EDTA. Protein concentration was 300μM for all spectra
A variant designed to disrupt this zinc binding site in the second interface, H248A, H250A and C261A, shows normal expression levels and anti-viral activity (). These observations demonstrate that zinc-mediated oligomerization may not be essential for A3G’s HIV-1 restriction activity. This variant is partly defective for DNA deaminase activity (). This alteration in DNA deaminase activity, may be attributable to C261A alone (Chen et al., 2008
), and was not significant enough to compromise A3G anti-viral activity. Thus the residues that are integral to the second interface in the A3G191-384-2K3A crystal structure appear to impact A3G deaminase activity.
The third interface (427 Å2
) involves residues at N-terminal (β1-β2 strands) and C-terminal ends of A3G191-384-2K3A () (Table S1
). Nine residues contribute extensively to the interface with Q354 burying the largest surface area at this interface followed by Q237, H228 and G355. In six of the eleven aligned APOBEC3 sequences (Q or E354)/G355 are conserved () but not H228 or Q237. Four direct and three water mediated hydrogen bonds are formed at this interface. Specifically, Q237 and R238 at the β2-β2′ “bulge” of one molecule (molecule A), form intermolecular hydrogen bonds to N-terminal residues R194 (side chain disordered) and S196 with the other molecule (molecule B). In addition, Q354 in the α5–α6 loop (molecule B) forms a hydrogen bond with R226 in β1 (molecule A), this interaction is potentially conserved in APOBEC3A and APOBEC3B-CTD. This interface would correspond closest to the β2-β2 interface observed in APOBEC2, but is clearly not formed and would entail extensive rearrangement of the α5 and α6 helices and their connecting loop. Of the four major interfaces, the packing of the third interface is least complementary and mutations at this interface (H228A, V233R and L235R) do not disrupt A3G’s DNA deaminase or antiviral activities ().
Interface 3 between two A3G-CTD molecules in the asymmetric unit (A) Surface representation and (B) detailed interactions. Same coloring scheme as in .
The final of the four major interfaces buries a total of 246 Å2
and involves substantial burial of six residues from each molecule and the formation of two hydrogen bonds and a salt bridge () (Tables S1, S2
). This interface is primarily made up of the tops of the α5 and α6 helices. Interestingly, residues in α5 and α6 are also affected by the addition of Zn2+
. These residues showed small chemical shift changes in HSQC spectra, suggesting that this region is involved in intermolecular interactions under a fast exchange (labeled yellow, , Fig. S2
). At this interface, residues Y340 and D365 contribute most to the buried interface area. These two residues are well conserved among the various APOBEC3’s () Y340 is conserved in all but one and (D/E/Q)365 is conserved in eight of the eleven APOBEC3 sequences. Both of these residues and the less conserved S341 are involved in hydrogen bonding; a salt-bridge is observed between the moderately conserved K344 and D365. Although extensive sequence conservation is present, this interface is not as large as the others and a variant of A3G with mutations Y340A, S341A, K344A and D365A has normal DNA deaminase and HIV-1 restriction activities ().