Two different kinds of interaction modes of deaminase APOBEC3A with single‐stranded DNA in solution detected by nuclear magnetic resonance

Abstract

APOBEC3A (A3A) deaminates deoxycytidine in target motif TC in a single‐stranded DNA (we termed it as TC DNA), which mortally mutates viral pathogens and immunoglobulins, and leads to the diversification and lethality of cancers. The crystal structure of A3A‐DNA revealed a unique U‐shaped recognition mode of target base dC₀. However, when TC DNA was titrated into ¹⁵N‐labeled A3A solution, we observed two sets of ¹H‐¹⁵N cross‐peaks of A3A in HSQC spectra, and two sets of ¹H‐¹H cross‐peaks of DNA in two‐dimensional ¹³C,¹⁵N‐filtered TOCSY spectra, indicating two different kinds of conformers of either A3A or TC DNA existing in solution. Here, mainly by NMR, we demonstrated that one DNA conformer interacted with one A3A conformer, forming a specific complex A3A^S‐DNA^S in a way almost similar to that observed in the reported crystal A3A‐DNA structure, where dC₀ inserted into zinc ion binding center. While the other DNA conformer bound with another A3A conformer, but dC₀ did not extend into the zinc‐binding pocket, forming a nonspecific A3A^NS‐DNA^NS complex. The NMR solution structure implied three sites Asn⁶¹, His¹⁸² and Arg¹⁸⁹ were necessary to DNA recognition. These observations indicate a distinctive way from that reported in X‐ray crystal structure, suggesting an unexpected mode of deaminase APOBEC3A to identify target motif TC in DNA in solution.

Short abstract

PDB Code(s): 7D3X, 3D3W, 5W2M, 5ZVA, 5ZVB, 5K83, 6K3J, 6K3K, 6bux, 5KEG, 5SWW and 5TD5

1. INTRODUCTION

APOBEC3 (A3), a family of activation‐induced cytidine deaminases, contains seven members APOBEC3A (A3A), APOBEC3B (A3B), APOBEC3C (A3C), APOBEC3DE (A3DE), APOBEC3F (A3F), APOBEC3G (A3G) and APOBEC3H (A3H). They deaminate deoxycytidines into uridines in single‐stranded DNA (ssDNA) during reverse transcription, which produces G to A hypermutations in a viral genome, generates defective proteins and proviruses, thus decreases the possibility of further viral replication. ¹ , ² , ³ , ⁴ A3s generally display differential substrate binding specificities (CC or TC), leading to altered frequencies of mutation of deoxycytidines. ¹ , ² , ³ , ⁵ , ⁶ , ⁷ , ⁸ Among them, A3DE, A3F, A3G and A3H restrict replication of HIV‐1 virus in strains lacking the virus infectivity factor (Vif) by deaminating cytidine in virus cDNA. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ Vif facilitates polyubiquitination of A3 members, leading to proteasomal degradation. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ Specifically, A3A suppresses HIV‐1 primary infection in macrophages, ²⁰ , ²¹ as well as other viruses (including hepatitis B virus, human papillomavirus, Epstein–Barr virus, cytomegalovirus, and herpes simplex virus type 1) by inducing the lethal mutation in viral genomic DNA through cytidine deamination. ²¹ , ²² , ²³ , ²⁴

To study how A3s specifically recognize the target motifs TC or CC in DNA, many structures of A3‐DNA complexes had been reported. ²⁵ , ²⁶ The structure of human A3F‐CD2 (i.e., hA3F‐CD2) in complex with a 10‐dT ssDNA (PDB code 5W2M) ²⁷ shows that DNA is far away from catalytic zinc‐binding motif, while the structures of chimeric hA3F‐CD2 (i.e., hA3Fc‐CD2) in complex with DNA strands (PDB codes 5ZVA and 5ZVB) demonstrate two DNA binding sites on hA3Fc‐CD2. ²⁸ One is formed by Tyr³³³, Lys³⁵⁸ and Tyr³⁵⁹ in one monomer of hA3Fc‐CD2, the other is composed by conserved residues Trp²⁷⁷, Y₃₀₇YFW₃₁₀ near to Zn²⁺ binding region in another monomer hA3Fc‐CD2. The structure of inactive pseudo‐catalytic rhesus macaque A3G‐NTD (i.e., rA3G‐NTD) in complex with poly‐dT ssDNA (PDB code 5K83) ²⁹ reveals that only one deoxythymidine is in a shallow cleft close to the pseudo‐catalytic zinc‐binding motif. The structures of A3G‐CD2 in complex with two deamination product DNA strands containing 5‐iodinated CUC^I and CCU^I motifs (PDB codes 6K3J and 6K3K) ³⁰ imply the structural basis of the two DNA binding modes responsible for cytidine deamination in TCC and CCC motifs catalyzed by A3G, respectively. The 10‐site A3G‐CD2 variant (i.e., P200A, L234K, N236A, C243A, P247K, Q318K, F310K, C321A, Q322A, and C356A, termed as A3G‐CTD2) was reported to have much stronger binding affinity to DNA and higher catalytic efficiency than wild‐type A3G‐CD2. ³¹ The structure of the inactive form A3G‐CTD2 E259A variant (i.e., A3G‐CTD2*) complexed with ssDNA (PDB code 6bux) ³¹ demonstrates that only base dC₀ in the target motif 5′‐C₋₁C₀C₁ inserts into the pocket of the zinc‐binding active site region. Residues Asp³¹⁶ and Asp³¹⁷ of A3G‐CTD2* form hydrogen‐bonds with base dC₋₁, accounting for their roles in identifying base dC₋₁. The structures of inactive variant A3A E72A or E72A/C171A ³² , ³³ (Glu⁷² is conserved in A3 family, Figure S1, Supporting Information) and chimeric A3B variant E255A‐CD2 ³³ (i.e., A3Bctd‐QMloop3A3Aloop1 or A3B‐CTD*, in which loop 3 of A3B was replaced by loop 1 of A3A) in complexes with DNA (PDB codes 5KEG, 5SWW and 5TD5) also show that only cytidine in hotspot 5′‐TC inserts into zinc‐binding active site pocket. One molecular DNA binds to one molecular A3A or chimeric A3B‐CD2 in a U‐shaped mode, similar to that observed in A3G‐CTD2*‐DNA complex. These three structures obviously visualize the active sites poised for catalysis of A3A and A3B, and pinpoints the residues conferring the specificity of TC motif.

All those reported results indicated that A3 proteins (including A3A) identified the target base dC₀ in one mode, where the base dC₀ must insert into the active site of A3A. From this point, the specific identification of base dC₀ is in a unique way. It is well known that, by nuclear magnetic resonance (NMR) technique, the dynamic process can be sometimes observed. Thus, we decided to investigate the interactions of A3A with DNA in solution by NMR, since the assignment of the atoms ¹³C, ¹⁵N and ¹H of inactive A3A variant L63N/C64S/E72Q/C171Q (i.e., A3A_4M in the current report) was accessible. ³⁴ The residues L63, C64 and C171 are located in the surface of A3A, the mutation from L63 to N63 improves protein solubility, while the mutations C64S and C171Q are helpful to avoid A3A aggregation by forming disulfide bond between two A3A monomers, thus resolving ambiguous assignments of free A3A. ³⁴ We made ¹⁵N‐labled NMR sample, and acquired ¹H‐¹⁵N HSQC spectrum, which displayed resolved cross‐peaks (Figure S2), consistent with the previous observation. ³⁴ However, during our NMR studies, we found that, upon TC DNA (with a sequence of 5′‐ATTTT₋₁C₀A₊₁ATT‐3′, Table 1) being titrated into the solution of ¹⁵N labeled inactive A3A_4M, A3A_4M variant two sets of ¹H‐¹⁵N cross‐peak signals belonging to A3A_4M in two‐dimensional (2D) ¹H‐¹⁵N HSQC spectrum (Figure 1a) were observed even at a molar ratio less than 1:1, and two sets of ¹H‐¹H correlation signals belonging to TC DNA were either found in 2D ¹³C,¹⁵N‐filtered ¹H‐¹H TOCSY (Figure 1b,c). After we finished assignment of the backbone atoms of A3A_4M, we selected several cross‐peaks belonging to residues L12, V150, S187 and G198 in ¹H‐¹⁵N HSQC spectra acquired at different molar ratio, and measured the intensity ratios of these cross‐peaks, the results were shown in Figure S3. From Figure S3, we can conclude that, after A3A_4M was mixed with DNA at different molar ratios (TC DNA vs. A3A_4M), the intensity ratio of the two complexes (specific vs. nonspecific binding states) is almost unchanged. Obviously, these observations indicated that A3A_4M interacted with TC DNA in solution in a manner, not completely identical to those observed in the reported crystal structures of A3‐DNA complexes mentioned above. To account for these observations, we here conducted extensive NMR studies on the interactions between A3A_4M and TC DNA.

TABLE 1.

5′‐FAM‐labeled DNA strands used in this manuscript

ssDNA	K d (μM)	Sequence
TC	0.086 ± 0.010	5′‐ATT TTC AAT T‐3′
TTC‐1	5.683 ± 0.236	5′‐TTT CTT T‐3′
TTC‐2	3.486 ± 0.172	5′‐TTT TCT TTT‐3′
TTC‐3	2.624 ± 0.070	5′‐TTT TCT TTT TT‐3′
TTC‐4	4.280 ± 0.273	5′‐TTT TCT TTT TTT T‐3′
TTC‐5	3.852 ± 0.294	5′‐TTT TCT TTT TTT TTT‐3′
TTC‐6	4.193 ± 0.306	5′‐TTT TCT TTT TTT TTT TT‐3′
TTC‐7	3.610 ± 0.091	5′‐TTT TTT CTT TT‐3′
TTC‐8	19.925 ± 2.091	5′‐TTT TTT TTC TTT T‐3′
TTC‐9	10.490 ± 1.104	5′‐TTT TTT TTT TCT TTT‐3′
TTC‐10	9.188 ± 0.846	5′‐TTT TTT TTT TTT CTT TT‐3′
TTC‐11	39.184 ± 2.524	5′‐TTT TTT TTT TTT TTT TCT TTT‐3′

FIGURE 1.

Two kinds of conformers of either A3A_4M or TC DNA in solution upon their interaction with each other were observed by NMR technique. (a) ¹H‐¹⁵N HSQC spectra superimposition at different molar ratios of A3A_4M versus DNA. (b) ¹³C /¹⁵N filtered two‐dimensional (2D) ¹H‐¹H TOCSY spectrum was overlapped with the 2D TOCSY spectrum of free TC DNA. In (b, c), the ¹H‐¹H cross‐peaks of free TC DNA were displayed in blue. The assignment of the signals in (a–c) indicated the residues in A3A^S or DNA^S (in green) and A3A^NS or DNA^NS (in pink)

2. RESULTS AND DISCUSSION

2.1. Two kinds of A3A_4M‐DNA complexes observed in solution

To study how A3A interacted with TC DNA in solution, we first conducted extensive NMR studies on the interactions between A3A_4M and TC DNA. We titrated TC DNA into ¹⁵N‐labeled A3A_4M solution. In the absence of DNA, there is only one set of ¹H‐¹⁵N cross peaks in the HSQC spectra, although some peaks have weak intensity due to the mixed forms of monomer and dimer A3A_4M in solution, determined by running ultracentrifugation (Figure S4). However, the addition of TC DNA into A3A_4M solution led to two sets of ¹H‐¹⁵N cross peaks of A3A_4M shown up in ¹H‐¹⁵N HSQC spectra, indicating that two kinds of A3A_4M conformers existing in solution. Moreover, this observation was independent on the molar ratio of A3A_4M versus TC DNA. When the molar ratio was increased up to 1:2, the cross‐peaks did not shift any longer. So, to do NMR studies on the interactions between A3A_4M and TC DNA, we made NMR samples of double isotope ¹³C, ¹⁵N labeled or triple isotope ¹³C, ¹⁵N, ²H labeled A3A_4M in complex with TC DNA at a molar ratio 1:2 for further NMR experiments.

To clearly identify the signals belonging to the bound TC DNA, we acquired 2D ¹H‐³¹P HETCOR spectrum, ¹³C/¹⁵N‐filtered ¹H‐¹H TOCSY and ¹H‐¹H NOESY spectra. Interestingly, in the 2D filtered TOCSY spectrum (Figure 1b,c), two sets of ¹H‐¹H cross‐peaks of DNA were also observed, indicating that two kinds of DNA conformers existing in solution, either. Compared the chemical shifts of A3A‐bound DNA with those of free DNA, we found that, in one A3A‐bound TC DNA conformer, which displayed more protons displaying chemical shift changes (Figure S5a), its protons demonstrated different intra‐DNA ¹H‐¹H NOE patterns from those observed in NOESY spectrum of free TC DNA. So, we tentatively named this kind of DNA conformation as A3A_4M specifically bound conformer (i.e., DNA^S). In contrast, in another A3A‐bound TC DNA conformer, which had fewer protons demonstrating chemical shifts changes (Figure S5b), its protons had intra‐DNA ¹H‐¹H NOE patterns almost similar to those observed in NOESY spectrum of free TC DNA. We thus termed this kind of TC DNA conformation as A3A_4M nonspecifically bound conformer (i.e., DNA^NS). To assign NMR signals of A3A in different states, a series of 2D and three‐dimensional (3D) NMR experiments were then performed to assign NMR signals belonging to backbone and side‐chain atoms of A3A_4M, and intramolecular NOEs in A3A_4M (described in section 4.1). Compared to the free A3A_4M, the chemical shifts changes of the backbone nitrogen atoms and amide protons of A3A_4M (Figure 2) implied that the TC DNA binding regions in two kinds of A3A_4M conformers were almost similar to each other, mainly locating in the loops 1, 3, 5 and 7. Subsequently, by analyzing 3D ¹³C‐F1 edited, ¹³C/¹⁵N‐F3 filtered NOESY spectra, we got intermolecular NOEs between A3A_4M and TC DNA. We found that one A3A_4M conformer individually displayed intermolecular NOEs with only one TC DNA conformer. We thus termed two different kinds of A3A_4M conformers as DNA^S bound A3A_4M (i.e., A3A^S) and DNA^NS bound A3A_4M (i.e., A3A^NS). To explore how DNA^S and DNA^NS interacted with A3A^S and A3A^NS, respectively, we further decided to determine the solution structures of the A3A^NS‐DNA^NS and A3A^S‐DNA^S complexes. So, we assigned all intramolecular NOEs in DNA^NS, DNA^S, A3A^S and A3A^NS, intermolecular NOEs between A3A^S and DNA^S, and between A3A^NS and DNA^NS.

FIGURE 2.

The chemical shift changes of backbone amide nitrogen and protons in two DNA‐bound states of A3A_4M compared to free A3A_4M. (a) DNA specifically bound A3A_4M. (b) DNA nonspecifically bound A3A_4M. In the left of (a, b), the secondary regions were labeled and drawn above the bar charts. The dashed lines indicated the lowest values, below which the chemical shift changes were not considered. In the right of (a, b), the residues in the surface of A3A_4M structure with different scales (0.2–0.4 ppm, 0.4–0.6 ppm and >0.6 ppm) of the chemical shift changes were displayed in yellow, green and red, respectively

2.2. DNA^S binds A3A_4M in a manner almost similar to that observed in the reported crystal structures of A3A‐DNA or A3B‐DNA complexes

To calculate NMR structures of A3A^NS‐DNA^NS and A3A^S‐DNA^S complexes, we generated 3,346 and 3,409 distance constraints derived from NOE intensities in total (including 50 and 95 intermolecular NOEs, which were listed in Tables S1 and S2), 184 phi and 185 psi dihedral angel constrains based on the chemical shifts of backbone atoms of A3A_4M, 174 hydrogen‐bond constraints based on the secondary structure of A3A_4M, as shown in Table S3. Constraints between the ligated residues (His⁷⁰, Cys¹⁰¹ and Cys¹⁰⁶) in the protein and the zinc ion were added using the procedure of Neuhaus et al as described previously. ³⁵ , ³⁶ The final two ensembles, each of which contained 20 structures of A3A_4M‐DNA complexes with the lowest energies, were displayed in Figure 3a,b, with RMSD values of 0.55 ± 0.11 Å and 0.54 ± 0.07 Å for backbone atoms in the secondary structural regions, respectively, upon overlapping the backbone atoms at the secondary regions of A3A_4M. The Ramachandran plot displayed 89.0 and 86.6% of the residues in the most‐favored regions, and no residues (0%) located in the disallowed regions, indicating that all structures were reasonable. On the whole, as shown in Figure 3c,d, the TC DNA mainly bound in a groove between loop 1 or 3 and loop 5 or 7, consistent with the NMR titration results, indicating the roles of these loops in the interactions between A3A_4M and DNA. The global folding of A3A_4M is almost identical in two bound states with an RMSD value of 1.27 Å upon 162 Cα atoms of the residues in the secondary structures. The loops 1, 3, 5 and 7 display plasticity to accommodate the conformational changes of the DNA upon its binding to A3A_4M.

FIGURE 3.

NMR structures of A3A‐TC DNA complexes. (a) The bundle of 20 structures of A3A^S‐DNA^S complex with the lowest energies. (b) The bundle of 20 structures of A3A^NS‐DNA^NS complex with the lowest energies. (c) One structure of A3A^S‐DNA^S complex. (d) One structure of A3A^NS‐DNA^NS complex. In (a, b) and (c, d), DNA (in orange) and A3A_4M (in cyan or pink) were displayed in ribbon and cartoon modes, respectively. The N‐ and C‐termini, the loops 1, 3, 5 and 7, the helices α1, α2, α3, α4, α5 and α6 of A3A_4M, and 5′‐ and 3′‐ ends of DNA were labeled, respectively. The grey balls represented zinc ions

To verify the effects of these loops on enhancing DNA binding, we varied the length of ssDNA (containing only one target motif TC) by cutting off or adding bases at either 5′‐ end or 3′‐ end, respectively, from TTC‐1 to TTC‐11 (Table 1 and Figure S6). Without adenine adjacent to TC motif at its 3′‐ end (i.e., TCA motif in TC DNA), all these DNA strands had dramatically decreased binding affinities to A3A_4M, compared to TC DNA. Compared to TTC‐2 (containing 9 bases in total), A3A_4M demonstrated no significant changes in binding affinities (all K _D < 5 μM) to TTC‐3 (11 bases in total, adding 2 dT bases at 3′‐ end), TTC‐4 (13 bases in total, adding 4 dT bases at 3′‐ end), TTC‐5 (15 bases in total, adding 6 dT bases at 3′‐ end) and TTC‐6 (17 bases in total, adding 8 dT bases at 3′‐ end). Since 3′‐ end bases of DNA mainly interacted with loops 1, 3, 5 and 7 of A3A_4m, as shown in Figure 3, this observation implied the indispensable roles of these loops in stabilizing the conformation of A3A‐DNA complex. In contrast, compared to TTC‐2, when the length of DNA was extended by adding different numbers of dT into 5′‐ end, the binding affinities (20 μM < K _D < 40 μM) of TTC‐8 (13 bases in total, adding 4 dT bases at 5′‐ end), TTC‐9 (15 bases in total, adding 6 dT bases at 5′‐ end), TTC‐10 (17 bases in total, adding 8 dT bases at 5′‐ end) and TTC‐11 (21 bases in total, adding 12 bases at 5′‐ end) were remarkably impaired by 5–10 fold. This finding indicated that, the DNA strands, with larger number of flanking nucleotides at 5′‐ end than TTC‐2, obviously weakened the interactions between A3A_4M and DNA, and the conformation of DNA became more flexible in A3A‐DNA complex. Meanwhile, TTC‐7 DNA (11 bases in total, adding 2 dT bases at 5′‐ end) displayed binding affinity to A3A_4M stronger binding affinities than TTC‐1 DNA (7 bases in total) (Figure S6a), and in an almost similar order of magnitude to those of TTC‐2 DNA (9 bases in total) and TTC‐3 DNA (11 bases in total) (Figure S6b), further implying that DNA with a length of 9–11 bases was most suitable for A3A_4M interaction, consistent with the previous report. ³⁴

In A3A^S‐DNA^S complex structure, the zinc ion coordinated with residues His⁷⁰, Cys¹⁰¹, Cys¹⁰⁶, the whole DNA interacted with A3A_4M also in a U‐shaped manner. The target base dC₀ extended into the pocket of zinc ion binding site (Figure 4a,b), and had parallel or T‐shaped π–π stacking interactions with the aromatic side‐chains of the conserved residues His⁷⁰ and Tyr¹³⁰ (Figure 4a,c), respectively, similar to those observed in the reported structures of A3‐DNA complexes (Figure 4d–f). ³¹ , ³² , ³³ The conformation of dC₀ was stabilized by several hydrogen‐bonds between Ala⁷¹ backbone NH and dC₀ O₂ atoms, between the carbonyl oxygen atoms of Trp⁹⁸ and Ser⁹⁹ and NH₂ of dC₀, and between the hydroxyl group of the side‐chain of Thr³¹ and O₃ atom of the sugar ring of dC₀, which also addressed the specificity for dC₀ over dT₋₁ (Figure 4a). The base dT₋₁ was identified by the conserved residue Asp¹³¹ in loop 7 through hydrogen‐bond between amide nitrogen of Asp¹³¹ and oxygen of dT₋₁ (Figure S7a,e). Identical to those observed in the crystal structures of A3A‐DNA complex, ³² , ³³ A3Bctd*‐DNA complex ³³ and A3G‐CTD2*‐DNA complex ³¹ (Figure S7b–d), the Asp¹³¹ side chain had a salt bridge to the side‐chain of the conserved residue Arg¹⁸⁹ in helix α6, which might enhance all hydrogen‐bond interactions between loop 7 and base dT₋₁. Similarly, as shown in Figure S8, the conserved residue His ²⁹ in loop 1 had π–π stacking interactions with dA₊₁, as reported in the crystal structures of A3A‐DNA, ³² A3Bctd*‐DNA ³³ and A3G‐CTD2*‐DNA ³¹ complexes, which once again confirmed that residue His ²⁹ might work as “latch” in the process of A3A recognizing the TCA motif in the substrate DNA.

FIGURE 4.

Structural analysis of A3A^S‐DNA^S complex. (a) The target dC₀ identified by A3 members in structures of A3A^S‐DNA^S complex. (b) The specific binding complex displayed in a cartoon mode, where dC₀ inserted into zinc‐ion binding center. (c) Conserved residues among A3A, A3B and A3G interact with dC₀. (d) The target dC₀ identified by A3 members in structures of A3A‐DNA complex (5KEG). (e) The target dC₀ identified by A3 members in structures of A3Bctd*‐DNA complex (5TD5), and (f) A3G‐CTD2* complex (6BUX). The dashed lines with stars in both ends indicate stacking interactions. The dash lines without stars in both ends imply hydrogen‐bonds

2.3. A3A^NS‐DNA^NS structure implies three new sites important to DNA binding

In A3A^NS‐DNA^NS complex structure (Figure 5a), DNA^NS located in a region almost similar to that observed in the structure of A3A^S‐DNA^S complex. However, the target base dC₀ did not extend into the zinc ion binding pocket. The averaged distances between zinc ion and N4 atom of dC₀ in all final 20 structures were measured as 12.0 Å, which was much farther than that (5.0 Å) in all final 20 structures of A3A^S‐DNA^S complex. Therefore, we thought that this kind of interaction between A3A^NS and DNA^NS was nonspecific, however, which was not single one case reported between DNA and A3 family members. For examples, as displayed in Figure S9, poly‐dT DNA interactions with rA3G‐CD1 only led to the conformational change of the loops and residues surrounding the Zn²⁺‐coordinated center. ²⁹ The hA3F‐CD2 bound to poly‐dT DNA through a positively charged site distal to the active zinc ion binding center. ²⁷ DNA (containing two TC motifs at both ends) interacted with hA3Fc‐CD2 dimer at two sites existing in the hA3Fc‐CD2 monomers. ²⁸ In all these structures, as well as that of A3A^NS‐DNA^NS complex, the target dC₀ was not in the zinc ion binding pocket. However, obviously, DNA bound to different regions of A3 proteins, suggesting that nonspecific interactions between DNA and A3 members were in varied manners.

FIGURE 5.

The NMR structure of A3A^NS‐DNA^NS complex indicated three new sites important to DNA binding. (a) The positions of residues N61, H182 and R189. DNA was displayed in stick mode, A3A_4M was shown in transparent, pink cartoon mode. (b, c) Residues H182 and R189 formed H‐bonds with DNA base dT₋₄. (d, e) Residue N61 formed H‐bond with dT₊₃ . In (b, d), the residues N61, D131, H182 and R189 in A3A_4M and bases dT₊₃ and dT₋₄ in DNA were shown in line mode in all final 20 structures with the lowest energies, indicating the convergent orientations of their side‐chains. H182 located in regid α6 helix, while N61 was in flexible loop 3, which resulted in convergent conformation of H182, and shifted‐conformation of N61. In (c, e), residues N61, D131, H182 and R189 in A3A and bases dT₋₄ and dT₊₃ in DNA were shown in stick mode. The dashed lines represented H‐bonds. (f) The binding affinities of A3A_4M and its variants to TC DNA were measured by fluorescent polarization assay

Structural analysis on A3A^NS‐DNA^NS complex showed that the backbone phosphate ions of base dT₋₄ at 5′‐ end and of base dT₊₃ at 3′‐ end formed hydrogen‐bonds with the side‐chains of the residues Asn61, His182 and Arg189 (Figure 5b–e), respectively. This observation explained why the chemical shifts of backbone atoms of residues His182 and Arg189 did not change (Figure 2b) upon DNA nonspecifically interacting with A3A_4M. At the same time, the side‐chain of Asp131 did not have a salt‐bridge with the side‐chain of residue Arg189, different from the observation in structure of A3A^S‐DNA^S complex (Figure S7a,b). To confirm whether these three H‐bonds stabilized the conformation of A3A^NS‐DNA^NS complex, we replaced them into alanine, and measured the binding affinities of these A3A_4M variants to TC DNA (Figure 5f). Their binding affinities were decreased by 2–4.5 folds, compared to A3A_4M. Thus, A3A^NS‐DNA^NS structure explored three key sites useful for A3A interaction with DNA.

3. CONCLUSIONS

In this report, we determined two solution structures of A3A_4M in complexes with TC DNA in specific and nonspecific binding ways, respectively. The nonspecific binding mode of A3A_4M to ssDNA is not unique. Previously, the catalytic domain of APOBEC3B (i.e., A3B‐CTD) was suggested to specifically identify target base dC₀ by sliding along ssDNA (i.e., nonspecific interaction with ssDNA). ³⁷ A3B slides only for a relatively short distance and tends to dissociate from the ssDNA before reaching the target sequence. The full‐length A3G (i.e., FL‐A3G, containing inactive A3G‐NTD and active A3G‐CTD domains) has higher deaminase activity than A3G‐CTD. FL‐A3G interacts with ssDNA in specific and nonspecific modes, in which A3G displayed elevated affinity for specific sequence ssDNA than for nonspecific sequence ssDNA. ³⁸ A3G‐NTD interacts with ssDNA in a nonspecific mode, while A3G‐CTD interacts with ssDNA in a specific mode. Both domains of A3G contribute to the sequence specific binding of ssDNA. To identify target motif 5′‐CC‐3′, A3G first binds to ssDNA in a sequence nonspecific manner and slides along the ssDNA without directional preference. ³⁹ Similar to A3B and A3G, before the target base dC₀ in TC motif was specifically identified, A3A must slide along ssDNA through nonspecific interaction with ssDNA. Thus, the nonspecific interaction between A3A and ssDNA plays a role in guiding substrate DNA to be specifically recognized by the active center of A3A. This not only revealed that A3s had multiple nonspecific recognition manners for substrate ssDNA, but also provide implications of how A3s deaminate ssDNA in solution.

4. MATERIALS AND METHODS

4.1. Supplemental methods and materials

4.1.1. Expression and purification of A3A_4M

The gene of A3A_4M or its variants was cloned into the region between BamHI and XhoI cleavage sites of a recombined pGEX‐6p‐1 plasmid which contained an His₆‐tag adjacent to the N‐terminal GST tag, and HRV 3c cleavage site. A3A_4M and its variants were expressed in BL‐21(DE3) Escherichia coli cells. Cell cultures were grown to OD₆₀₀ value equal to about 0.8 and induced with a final concentration of 0.4 mM IPTG for 20 hr at 16°C. Cells were re‐suspended in Ni²⁺‐binding buffer A (50 mM Tris–HCl pH 7.5, 300 mM NaCl, 50 μM ZnCl₂ and 10 mM β‐mercaptoethanol [β‐ME]) with protease inhibitor (PMSF) and lysed at 15 kpsi using a hydraulic cell disruption system (Constant System JINBO Benchtop) (Guangzhou Juneng Biology and Technology Co., Ltd., Guangzhou, China). The lysate was centrifuged at 12,000 rpm and 4°C for 55 min to remove cellular debris prior to loading into a Ni‐NTA resin (GE Health). The column was washed with 10 column volumes buffer A (50 mM Tris–HCl pH 7.5, 300 mM NaCl, 50 μM ZnCl₂ and 10 mM β‐ME) followed by 10 volumes buffer B (50 mM Tris–HCl pH 7.5, 300 mM NaCl, 20 mM imidazole, 50 μM ZnCl₂ and 10 mM β‐ME). The bound protein was eluted with 10 column volumes of buffer C (50 mM Tris–HCl pH 7.5, 300 mM NaCl, 20 mM imidazole, 50 μM ZnCl₂ and 10 mM β‐ME). The elute was then dialyzed with buffer D (25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂ and 5 mM DTT). Fractions containing A3A_4M were then concentrated and purified on a Superdex75 16/300 GL column (GE Health) previously equilibrated with buffer D (25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂ and 5 mM DTT).

4.2. DNA synthesis and purification

All DNA strands with or without FAM label were commercially synthesized from Shanghai Biosune Biotech Co., Ltd., China at a PAGE grade. The molecular weight of all strands was confirmed by running PAGE gel and MALDI TOF mass spectroscopy.

4.3. Fluorescence polarization

To determine the binding affinities of A3A_4M and its variants to different DNA strands, or 5′‐fluoresceinated DNA strands were commercially synthesized at a HPLC grade (Shanghai Sangon Biotech Co., Ltd., China) with the corresponding sequences. The fluorescence polarization (FP) was performed, where all proteins were diluted in buffer D (25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂ and 5 mM DTT) for FP and incubated with a 10 nM 5′‐ end 6‐FAM‐labeled ssDNA at room temperature in a total reaction volume of 200 μL. FP assay was measured at 25°C using SpectraMax i3x Platform (Molecular Devices, Inc.) with 490 nm excitation and 535 nm emission wavelengths, respectively. The dissociation constants (K _D) were determined by a nonlinear least‐squares analysis using the program Prism 5 (GraphPad, Inc.). Data shown are averaged values of three repeated measurements.

4.4. NMR data collection and spectra analysis

To correctly assign NMR signals of DNA used in the structural determination of A3A‐DNA complex, we first performed NMR experiments on DNA samples of TC DNA, CC DNA, TU DNA and TT DNA. All DNA samples were about 1 mM dissolved in buffer containing 25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂, 5 mM DTT, 100% D₂O. All NMR samples were placed in 5‐mm Shigemi NMR tubes. All NMR experiments were performed on a Varian Unity Inova 600 NMR spectrometer (with cryo‐probe) equipped with triple resonances and pulsed field gradients. Two‐dimensional (2D) ¹H‐¹H NOESY (with a mixing time of 250 ms), TOCSY (with a mixing time of 80 ms), and DQF‐COSY spectra were acquired at 20°C using a spectral width of 6,100 Hz in both dimensions. The acquisition data points were set to 2048 × 512 (complex points). The watergate sequence was used for water suppression. During data processing, the 45° or 60° shifted sine‐squared functions were applied to NOESY and TOCSY spectra. The fifth‐order polynomial functions were employed for the baseline corrections. The final spectral sizes are 2,048 × 1,024. The ³¹P NMR spectra were collected at about 1.5 mM concentration in D₂O (25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂ and 5 mM dithiothreitol‐d₁₀ [d₁₀‐DTT]) at 20°C at 600 MHz Varian spectrometer including the one dimensional proton‐decoupled phosphorus spectrum, and 2D heteronuclear ³¹P‐¹H Correlation Spectroscopy (³¹P‐¹H HETCOR). Assignments of the individual ³¹P resonance were accomplished by a combination of 2D ¹H‐¹H NOESY, COSY, TOCSY and ³¹P‐¹H HETCOR spectra.

To determine NMR structure of A3A_4M in complex with TC DNA, 0.25 mM uniformly ¹⁵N‐/¹³C double isotope labeled or ¹⁵N‐/¹³C‐/70% ²H triple labeled A3A_4M plus 0.5 mM unlabeled TC DNA was prepared in NMR buffer (25 mM Na₂HPO₄ pH 7.3, 200 mM NaCl, 50 μM ZnCl₂ and 5 mM d₁₀‐DTT) containing 10% or 100% D₂O. All NMR experiments were performed at 20°C on a Varian Unity Inova or Agilent 600 NMR spectrometer (with cryo‐probe) equipped with triple resonances and pulsed field gradients, or on Bruker Avance or Agilent III‐800MHz, 850MHz, 900MHz NMR spectrometers (with cryo‐probe) equipped with four channels and z‐axis pulsed‐field gradient.

The standard suite of experiments were acquired for assigning the ¹H, ¹³C and ¹⁵N backbone and side‐chain chemical shifts of ¹³C and ¹⁵N double labeled A3A_4M in complex with unlabeled DNA, and for the collection of NOE‐based distance restraints were measured, ⁴⁰ , ⁴¹ including the 2D ¹³C‐edited HSQC in both aliphatic and aromatic regions, and ¹⁵N‐edited HSQC; the three‐dimensional (3D) HNCA, HNCO, HN(CO)CA, HNCACB, CBCA(CO)NH, ¹⁵N‐resolved HSQC‐TOCSY, HCCH‐TOCSY in both aliphatic and aromatic regions, ¹⁵N‐resolved HSQC‐NOESY, ¹³C‐resolved HSQC‐NOESY for both aliphatic and aromatic resonances, 2D (H_β)C_β(C_γC_δ)H_δ and (H_β)C_β(C_γC_δC_ε)H_ε spectra for correlation of C_β and H_δ or H_ε in aromatic rings used in aromatic protons assignment. ⁴² To obtain NMR signals of bound TC DNA, 2D ¹³C/¹⁵N filtered ¹H‐¹H TOCSY, NOESY spectra and 2D ³¹P‐¹H HETCOR spectra were collected. The intermolecular NOEs between ¹³C‐/¹⁵N‐labeled A3A_4M and unlabeled TC DNA were obtained by analyzing 3D ¹³C‐F1 edited, ¹³C/¹⁵N‐F3 filtered NOESY spectra.

All NMR spectra were processed with the program NMRPipe ⁴³ and analyzed with Sparky 3 software. ⁴⁴ The ¹H chemical shifts were referenced to 2, 2‐dimethylsilapentane‐5‐sulfonic acid (DSS), the ¹³C‐ and ¹⁵N‐resonances were indirectly referenced to DSS, and the ³¹P chemical shifts were referenced to an external standard of 85% H₃PO₄.

4.5. Analysis of ultracentrifugation experiments

Sedimentation velocity experiments were performed on a Beckman XL‐I analytical ultracentrifuge equipped with an eight‐cell rotor under 42,000 rpm at 20°C. The partial specific volume of different samples and the buffer density were calculated using the program SEDNTERP (http://www.rasmb.bbri.org/). The final sedimentation velocity data were analyzed and fitted to a continuous sedimentation coefficient distribution model using the program SEDFIT. The fitting results were further output to the Origin 9.0 software and aligned with each other.

4.6. NMR structure determination

The structural calculations of the A3A^S‐DNA^S and A3A^NS‐DNA^NS complexes were carried out using a standard simulated annealing protocol implemented in the program XPLOR‐2.37 (NIH version). ⁴⁵ 3,409 and 3,346 distance constraints derived from NOE intensities were classified into 1.8–2.9 Å, 1.8–3.5 Å, 1.8–5.0 Å groups, corresponding to strong, medium and weak NOEs, respectively, while intermolecular distance restrains were sorted into weak (1.8–5.0 Å) and very weak (1.8–6.0 Å) groups. 184 phi and 185 psi dihedral angles were derived from the chemical shifts of the backbone atoms (HN, HA, CO, CA) by the program TALOS. ⁴³ , ⁴⁶ One hundred and seventy‐four hydrogen‐bond restraints for secondary structures were generated by analyzing secondary structural regions in the reported free NMR and X‐ray apo‐A3A structures ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ for the final structure calculation. Constraints between the protein ligands and the zinc ion were added using the procedure of Neuhaus et al as described previously. ³⁵ , ³⁶ A total of 10 iterations (50 structures in the initial 10 iterations) were performed. One hundred structures were computed in the last five iterations, 20 conformers with the lowest energy are used to represent the 3D structures. In the ensemble of the simulated annealing 20 structures, there was no distance constraint violation more than 0.3 Å and no torsion angle violation more than 5°. The final 20 structures of the complex A3A with different DNA with lowest energy were evaluated with the program PROCHECK‐NMR and PROCHECK. ⁵¹ All figures related to structures were generated using the program PyMOL.

AUTHOR CONTRIBUTIONS

Yaping Liu: Data curation (equal); formal analysis (equal); investigation (supporting); validation (equal). Wenxian Lan: Methodology (equal); software (equal); validation (equal). Chunxi Wang: Data curation (equal); software (equal); supervision (equal); validation (equal). Chunyang Cao: Conceptualization (equal); funding acquisition (lead); project administration (lead); supervision (lead).

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.

Supporting information

Appendix S1: Supporting information

Liu Y, Lan W, Wang C, Cao C. Two different kinds of interaction modes of deaminase APOBEC3A with single‐stranded DNA in solution detected by nuclear magnetic resonance. Protein Science. 2022;31:443–453. 10.1002/pro.4242