Inhibiting APOBEC3 Activity with Single-Stranded DNA Containing 2'-Deoxyzebularine Analogues.

APOBEC3 enzymes form part of the innate immune system by deaminating cytosine to uracil in single-stranded DNA (ssDNA) and thereby preventing the spread of pathogenic genetic information. However, APOBEC mutagenesis is also exploited by viruses and cancer cells to increase rates of evolution, escape adaptive immune responses, and resist drugs. This raises the possibility of APOBEC3 inhibition as a strategy for augmenting existing antiviral and anticancer therapies. Here we show that, upon incorporation into short ssDNAs, the cytidine nucleoside analogue 2'-deoxyzebularine (dZ) becomes capable of inhibiting the catalytic activity of selected APOBEC variants derived from APOBEC3A, APOBEC3B, and APOBEC3G, supporting a mechanism in which ssDNA delivers dZ to the active site. Multiple experimental approaches, including isothermal titration calorimetry, fluorescence polarization, protein thermal shift, and nuclear magnetic resonance spectroscopy assays, demonstrate nanomolar dissociation constants and low micromolar inhibition constants. These dZ-containing ssDNAs constitute the first substrate-like APOBEC3 inhibitors and, together, comprise a platform for developing nucleic acid-based inhibitors with cellular activity.

Enzymes of the human APOBEC3 (A3A-H) family normally combat retroviruses and other pathogenic elements by deaminating 2′-deoxycytidine to 2′-deoxyuridine in single-stranded DNA (ssDNA) (Figure A). The combination of this deamination-dependent mechanism and a deamination-independent mechanism,[1] most likely dependent on nucleic acid binding, constitutes a potent block to parasite replication. Not surprisingly, viral pathogens have developed A3 counteraction strategies that range from active degradation (HIV-1 and related lentiviruses)[1−4] to apparently passive avoidance (papilloma viruses and polyomaviruses).[5,6] Moreover, the fact that many immune-escape and drug-resistance mutations occur within A3-preferred di- and trinucleotide motifs[7−10] strongly suggests that viruses have evolved mechanisms for both regulating and benefiting from A3 mutagenesis.

Figure 1

(A) Deamination of dC in ssDNA by A3 enzymes. (B) TSAs used in this work: zebularine, its 2′-deoxy analogue (dZ), 5-methyl-2′-deoxyzebularine (dZMe), and tetrahydrouridine (THU).

A3 enzymes have intrinsic preferences for deaminating cytosine bases preceded by thymine (5′-TC, A3A-D, A3F, and A3H) or by another cytosine (5′-CC, A3G).[10−14] The genomes of many different tumor types, including bladder, breast, cervix, head/neck, and lung, often have large fractions of mutations in 5′-TC motifs.[15−17] These 5′-TC-to-TT and 5′-TC-to-TG mutations are typically followed on the 3′-side by bases other than cytosine, that is, adenine, guanine, or thymine, thereby constituting an APOBEC mutation signature. A range of genetic, biochemical, and structural studies has combined to implicate A3B as the primary source of these mutations and A3A and A3H as potential secondary sources (depending on patient genotype and tumor type). APOBEC mutagenesis has been shown to contribute to both clonal and subclonal mutational events,[17,18] and its frequency often increases from primary to metastatic disease.[16] A3B expression levels and APOBEC signature mutations also correlate with poor clinical outcomes, including disease recurrence, metastasis, and drug resistance.[15,19,20] These observations support a model in which APOBEC mutagenesis promotes tumor evolution and strongly influences disease trajectories. Therefore, chemical modulators of APOBEC activity may yield useful chemical probes for mechanistic studies and, possibly, therapeutic compounds to harness APOBEC mutagenesis.[21]

The mechanism of cytosine deamination for APOBECs is thought to be similar to that for cytidine deaminase (CDA), an enzyme that processes individual nucleosides.[22] The cytidine analogues zebularine [Z (Figure B)], 2′-deoxyzebularine (dZ), and tetrahydrouridine (THU) are known transition-state analogues (TSAs) of cytidine deaminase (CDA).[23−25] These competitive inhibitors bind tightly to the active site of CDA, as indicated by crystal structures.[23−28] Here we show that these TSAs as free nucleosides do not alter the activity of A3 enzymes (Figure S1), but micromolar-potent A3 inhibitors are obtained upon introduction of dZ in place of the target 2′-deoxycytidine in DNA substrates (dZ-ssDNA). These findings open new avenues for further investigations of interactions between active A3 enzymes and ssDNA and, importantly, for the rational design of competitive A3 inhibitors for use with living cells.

Materials and Methods

Detailed methods are provided in the Supporting Information.

Synthesis of 2′-Deoxyzebularine (dZ), Its Phosphoramidite, and Oligonucleotides Containing dZ and dZMe

Synthetic procedures are provided in the Supporting Information.

Protein Expression and Purification

Human APO-BEC3A (residues 1–199, Uniprot entry P31941) was cloned as the inactive E72A mutant with a His6 C-terminal fusion tag into an expression vector (pETite, Lucigen), expressed in Escherichia coli BL21 DE3 cells (Hi-Control, Lucigen), and purified as described previously.[29]

The A3B C-terminal domain (residues 187–378) was cloned into the pET24a vector (Novagen) to produce A3BCTD proteins with a noncleavable C-terminal His6 tag (LEHHHHHH) that were derived as previously described.[30] Several derivative constructs previously reported[31] were used in this study. A3BCTD-QM-ΔL3 and A3BCTD-QM-ΔL3-E255A were expressed in E. coli strain BL21(DE3) (Lucigen), and A3BCTD-QM-ΔL3-AL1swap was expressed in E. coli strain C41(DE3)pLysS (Lucigen). The E. coli culture was grown at 37 °C in LB medium; once the mid log growth phase had been reached, the culture was supplemented with 100 μM zinc chloride, before protein expression was induced by the addition of isopropyl β-d-1-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM and overnight incubation at 18 °C.

A3BCTD-DM was expressed and purified as reported in ref [31]. A3GCTD (residues 191–384, wt) was purified as described previously.[32] The glutathione S-transferase (GST)-fused A3GCTD was expressed in E. coli BL21(DE3) cells overnight at 17 °C. After being harvested, the cells were resuspended in 50 mM sodium phosphate buffer (pH 7.4) and lysed by sonication. After ultracentrifugation at 25000g for 10 min, the supernatant was added to glutathione (GSH)-Sepharose, which was subsequently washed. For kinetic analysis, the GST fusion protein was eluted from the Sepharose matrix with 100 mM GSH in phosphate buffer. By using filtration at 4000g, the buffer was changed to a solution containing 75 mM sodium phosphate and 75 mM citrate (pH 5.5).

Fluorescence Polarization Assay

Fluorescence polarization assays were performed with recombinant [purified from E. coli strain BL21(DE3)] A3A (amino acids 1–195, expressed using the pGEX vector as a GST fusion protein)[30] with the catalytic glutamic acid mutated to alanine (E72A) to render the enzyme unable to deaminate the substrate. The assay buffer consisted of 2-(N-morpholino)ethanesulfonic acid (MES; 50 mM, aqueous), NaCl (100 mM, aqueous), tris(2-carboxyethyl)phosphine (TCEP; 2 mM, aqueous), and 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS; 4 mM, aqueous) at pH 6.0. The 15 μM stock solutions of fluorescent tracer 5′-(6-FAM)TTTTCAT (Integrated DNA Technologies; molecular weight of 2598.3 g/mol) in molecular biology grade water were diluted to 15 nM in assay buffer. All FP experiments were performed with a 10 μL assay volume in black round-bottom low-volume 384-well plates (Corning 4514). A direct binding experiment (Chart S3) was first performed to determine the Kd of 5′-(6-FAM)TTTTCAT-3′ with A3A-E72A by serially diluting the protein (1:1, 10 μM starting concentration) and incubating with a constant concentration (15 nM) of the fluorescent tracer. Plates were incubated at room temperature for 30 min, gently shaken for 1 min, and then analyzed for fluorescence polarization on a BioTek Synergy 2 instrument (using standard instrument settings) with an excitation wavelength of 485 (20) nm, an emission wavelength of 528 (20) nm, and the top optics position at 510 nm. The resulting anisotropy values were fit using the one-site binding (hyperbola) function in GraphPad Prism 7.0 to obtain the Kd of the fluorescent tracer. The directly measured Kd was 18.2 ± 1.0 nM (Chart S3). This value was used for all further calculations. Competition binding experiments were then performed with the prebound fluorescent tracer and various test ligands to quantify their binding affinities, as described in the Supporting Information.

Isothermal Titration Calorimetry (ITC)

Desalted unmodified DNA oligonucleotides were purchased (Integrated DNA Technologies) at 1 or 5 μmol synthesis scale and dissolved in one of the buffers described below to give 10 mM solutions. ITC experiments were conducted at 25 °C using a Micro-Cal ITC200 (now Malvern Instruments) isothermal titration calorimeter. A3A-E72A (130 μM in high-salt or medium-salt buffer) or A3BCTD-QM-ΔL3-AL1swap (100 μM, activity assay buffer) was titrated in the corresponding buffer. DNA oligonucleotides at 1.6 mM (for A3A-E72A) or 300 μM (for A3BCTD-QM-ΔL3-AL1swap) were added in 18 steps of 2.0 μL each (plus a first addition with decreased volume of 0.4 μL to prevent dilution of the DNA in the syringe due to the long wait before the start of the experiment). Oligos and the enzymes were dialyzed against the appropriate buffer. For A3A-E72A, the high-salt buffer consisted of 25 mM sodium phosphate, 500 mM NaCl, 300 mM choline acetate, 5 mM β-mercaptoethanol, and 0.2 mM Na2-EDTA (pH 6.0) and the medium-salt buffer consisted of 50 mM MES, 100 mM NaCl, and 2.0 mM tris(2-carboxyethyl)phosphine (pH 6.0). For A3BCTD-QM-ΔL3-AL1swap, the activity assay buffer consisted of 50 mM citrate-phosphate buffer, 200 mM NaCl, and 2 mM β-mercaptoethanol (pH 5.5).

Thermal Shift Assay

A fluorescence-based thermal shift assay was used to assess the capability of ssDNA oligonucleotides to bind A3BCTD proteins, through examination of changes in the thermal stability of the proteins. Binding assays were conducted using A3BCTD protein constructs, A3BCTD-QM-ΔL3 and A3BCTD-QM-ΔL3-E255A, where the over the time of the experiment there was insignificant conversion of dC to dU by the very weakly active A3BCTD-QM-ΔL3 and none by the inactive mutant A3BCTD-QM-ΔL3-E255A to determine if differences in binding affinity occur due to a single amino acid change (E255A) in the protein. Purified A3BCTD protein was appropriately diluted in buffer [50 mM citrate-phosphate (pH 5.5), 200 mM NaCl, 2 mM β-mercaptoethanol (pH 5.5), and 200 μM 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS)]. Assay experiments were set up in a total volume of 25 μL containing 20 μM A3BCTD protein and 100 μM ssDNA oligonucleotide, mixed with SYPRO Orange dye (Bio-Rad) at a final concentration of 10×. Assays were dispensed into wells of a white low-profile 96-well plate (Roche), sealed with an optical seal, shaken, and centrifuged. Thermal scanning (from 20 to 95 °C at a rate of 0.6 °C/min) was performed using a real-time polymerase chain reaction setup on a LightCycler 480 instrument II (Roche) with fluorescence emission spectra recorded with combinations of excitation and emission filters (483–610 and 483–568 nm, respectively).

Evaluation of Nucleosides by a Fluorescence-Based Deamination Assay Using hA3A and hA3BCTD Expressed in HEK293T Cells

A3-deaminase activity assays were performed according to the method described by Li and colleagues[33] with the following modifications. Stock solutions of nucleosides in molecular biology grade water were serially diluted in protein dilution buffer (10 μL). Recombinant (purified from HEK293T cells) full length A3A-mycHis[33] (10 ng) or C-terminal domain (amino acids 195–382) A3BCTD-mycHis[34,35] (25 ng) proteins were delivered to each well in protein dilution buffer (10 μL). The deamination dual-fluorophoric substrate oligo 5′-(6-FAM)-AAA-TAT-CCC-AAA-GAG-AGA-(TAMRA) (2 pmol) and UDG (25 units) were delivered in TE buffer (10 μL). Cleavage of oligonucleotides containing an abasic site using NaOH (4 M, 3 μL) occurred after incubation of the reaction mixture at 37 °C for 30 min. Plates were read using a BioTek Synergy H1 plate reader with an excitation wavelength at 490 nm and emission at 520 nm. Each experiment was performed in biological duplicates with three technical replicates per condition. The resulting total fluorescence values were reported together with the no-protein low control and protein only high control (no inhibitors). Nonspecific inhibitor MN-1 was used as a positive control (Figure S1).[33]

Evaluation of Inhibitors in a Nuclear Magnetic Resonance (NMR)-Based Assay

The Supporting Information provides details about the experiments and how inhibition constants were calculated.

Results

Methodology of This Investigation

First, we confirmed that pyrimidine-based TSA nucleosides (Figure B) do not inhibit A3 enzymes (Figure S1). Because ssDNA is the preferred substrate of A3 proteins, we decided to evaluate the inhibitory effect of chemically modified ssDNAs in which the target 2′-deoxycytidine in the recognition sequence[32,36] is substituted with TSAs. We focused on 2′-deoxyzebularine (dZ), as it is a known, stable, moderately potent TSA inhibitor of CDA with an apparent Ki of 2.9 μM[37] and its incorporation into DNA by automated synthesis is feasible.[38] The quantity of modified oligos needed for the planned experiments warranted development of a new synthesis procedure for dZ. Then, we investigated the binding affinity and inhibitory activity of selected modified oligos. Binding data were recorded using a combination of fluorescence polarization (FP), ITC, and thermal shift assays. We confirmed that 9–10 nucleotides is a good compromise between the length of the oligo and binding affinity and concluded that binding of TSA-containing oligos must be investigated with catalytically active proteins. Finally, our activity data show clearly the inhibition of selected A3 variants by dZ-containing oligos. The list of proteins used can be found in Figure S3, and the sequences of oligos in Tables and 2. One should note that the preferred A3G substrate motif is CCCA (underlined C preferentially deaminated), whereas A3A and A3B prefer the TCA motif but can also readily deaminate the CCCA motif. The rationale for the design of each oligo can be found in Table S3.

Table 1

Oligo Concentrations when the Fluorescence Anisotropy Decreased to the Half-Maximum Level (IC50) and Indirectly Measured Dissociation Constants (KdFP) of ssDNA from A3A-E72A Obtained during Fluorescence Polarization (FP) Experiments in a Medium-Salt Buffera

	DNA sequence (5′–3′)	IC50 (μM)	KdFP (μM)
Oligo-3*c	ATTCCCAATT	0.0550 ± 0.0011	0.007 ± 0.001
Oligo-4*	TTCCC	5.6 ± 0.8	0.68 ± 0.11
Oligo-5*	CCCAA	1.8 ± 0.10	0.22 ± 0.02
Oligo-10*	TTTTCAT	0.2800 ± 0.0016	0.034 ± 0.002
Oligo-11	TTTTdUATb	15.0 ± 1.9	1.8 ± 0.3
Oligo-12	TTTTCATTTT	0.0960 ± 0.0012	0.012 ± 0.001
Oligo-13	AAAAATTCAAAGA	0.120 ± 0.004	0.015 ± 0.001
Oligo-14	TTCAAAAA	0.1400 ± 0.0022	0.017 ± 0.001
Oligo-15	TTTTdZAT	2.40 ± 0.10	0.29 ± 0.02

Values shown are means ± SEM (standard error of the mean). IC50 values were converted to KdFP values using the Kenakin equation (see the Supporting Information) with appropriate error propagation.

dU is 2′-deoxyuridine.

An asterisk denotes an oligo evaluated in both ITC and FP experiments.

Table 2

Dissociation Constants (KdITC) of ssDNA from A3 Enzymes Obtained by Isothermal Titration Calorimetry in Different Buffersa

	DNA sequence (5′–3′)	buffer	enzyme	KdITC (μM)
Oligo-1	AAAAAAAATTCAAAAAAAAA	high-salt	A3A-E72A	24 ± 7
		medium-salt	A3A-E72A	0.11 ± 0.05
Oligo-2	ATTTCATTT	high-salt	A3A-E72A	25.3 ± 0.9
		medium-salt	A3A-E72A	0.20 ± 0.04
Oligo-3*b	ATTCCCAATT	medium-salt	A3A-E72A	0.24 ± 0.10
Oligo-4*	TTCCC	medium-salt	A3A-E72A	5.0 ± 0.4
Oligo-5*	CCCAA	medium-salt	A3A-E72A	3.1 ± 0.4
Oligo-6	TTCAT	medium-salt	A3A-E72A	0.48 ± 0.10
Oligo-7	ATTCCdZAATT	medium-salt	A3A-E72A	0.97 ± 0.15
Oligo-8	ATTCCdZMeAATT	medium-salt	A3A-E72A	1.7 ± 0.3
Oligo-9	ATTTdZATTT	activity assay	A3Bc-QM-ΔL3-AL1swap	5.5 ± 0.6
Oligo-10*	TTTTCAT	med. salt	A3A-E72A	0.27 ± 0.04
Oligo-16	5′-(6-FAM)TTT TCAT	med. salt	A3A-E72A	0.41 ± 0.04

The high-salt buffer consisted of 25 mM sodium phosphate, 500 mM NaCl, 300 mM choline acetate, 5 mM β-mercaptoethanol, and 0.2 mM Na2-EDTA (pH 6.0). The medium-salt buffer consisted of 50 mM MES, 100 mM NaCl, and 2.0 mM tris(2-carboxyethyl)phosphine (pH 6.0). The activity assay buffer consisted of 50 mM citrate-phosphate buffer, 200 mM NaCl, and 2 mM β-mercaptoethanol (pH 5.5). Means ± the standard deviation (SD) are shown. Uncertainties (SD) in KdITC were calculated using standard error propagation methods from partial derivatives.

An asterisk denotes an oligo evaluated in both ITC and FP experiments. A3A prefers to deaminate the third C in the CCC motif.[36]

A3 Activity Is Unaffected by Free Nucleoside Transition-State Analogues

Human A3A and the catalytic C-terminal domain of A3B (A3BCTD) were purified from human 293T cells (hA3A, hA3BCTD)[33] and tested using a fluorescence-based activity assay[33] (for details, see Materials and Methods and the Supporting Information) for inhibition by nucleosides THU, Z, and dZ. The nonspecific small molecule inhibitor MN-1[33] almost completely abolished the activity of enzymes used in the assay (hA3A, hA3BCTD, and UDG). In contrast, THU, Z, and dZ did not affect the deamination of ssDNA by these enzymes, even at the high concentration of 2 mM (Figure S1). These results are consistent with earlier observations of no or very weak binding of A3 proteins to individual nucleotides and barely detectable deamination of 2′-deoxycytidine.[39,40]

Synthesis of dZ-Containing ssDNA

We developed a new straightforward procedure for preparing dZ phosphoramidite for incorporation into ssDNA based on the classical silyl modification of the Hilbert–Johnson reaction (Scheme and the Supporting Information). This synthesis, which is free of Lewis acids and requires no workup, leads straightforwardly to product 3a, which was purified by flash chromatography. We obtained a ratio and yield of the desired β anomer (88:12 β:α) much higher than those determined by the previously published procedure.[38,41] dZ was further used in the synthesis of modified 2′-oligodeoxynucleotides (Oligos) as described in the Supporting Information. Commercially available phosphoramidite of 5-methyl-2′-deoxyzebularine was used in the synthesis of dZMe-containing oligos.

Scheme 1

Reagents and conditions: (i) CHCl3, distillation, 10 min, 54%, 12:88 α:β (3a); (ii) 28% aqueous ammonia, MeOH, 48 h (3b); (iii) 4,4′-dimethoxytrityl chloride, pyridine, 0 °C → room temperature, overnight, 54% (4a); (iv) N,N-diisopropylamino-2-cyanoethoxychlorophosphine, Et3N, CH2Cl2, 30 min, 88% (4b).

Binding of Modified and Unmodified ssDNAs to A3 Proteins

We tested how the introduction of dZ and its 5-methyl derivative, dZMe, affected the affinity of ssDNA for the A3 enzymes, in particular A3A and the catalytically competent A3BCTD and selected mutants, to probe the generality of zebularine analogues as inhibitors of A3 proteins. Given that different A3 proteins[13,36] have varying preferences for nucleobases surrounding the target cytosine, the sequences surrounding the zebularine moiety were chosen to match the preferred ssDNA recognition motif. To study A3–ssDNA binding, we used the inactive A3A-E72A mutant, where the active-site glutamic acid is replaced with alanine, to avoid complications of deamination reactions occurring when the ligand contained cytosine. The E72A mutation does not change the geometry of the active site, as water was observed in the crystal structure instead of the glutamate side chain.[42] As A3A is a highly active enzyme with the highest affinity for DNA among human A3 catalytically active domains,[29,36] it is best suited for binding assays. Two methods widely applied for binding characterization were used: fluorescence polarization (FP) and isothermal titration calorimetry (ITC), which yield the indirectly determined dissociation constant KdFP [sometimes termed Ki (see the Supporting Information)][43−45] and the directly measured dissociation constant KdITC, respectively.

FP measurements were taken in competitive mode, where unlabeled ssDNA competed with the fluorescently labeled ssDNA for the protein (Figure S2). This results in dissociation constants KdFP (Table ) that should be directly comparable with KdITC, assuming that the binding of the unlabeled ssDNA and the binding of fluorescently labeled ssDNA to the protein are identical. As a positive control, KdFP for the nonlabeled ssDNA (Oligo-10) was 34 ± 2 nM compared to the directly measured value of 19 ± 2 nM for titration of the fluorescently tagged Oligo-10. The differences between the two values are likely attributable to the experimental conditions (competitive displacement with a nonlabeled oligo vs direct binding with a fluorescently labeled oligo). Weaker binding of ssDNA containing dU (Oligo-11; KdFP = 1800 ± 300 nM), as a negative control, compared to dC is in accord with published data[29,46,47] and highlights the binding specificity. The dZ-modified DNA had an affinity for A3A-E72A lower than that of the dC-containing substrate [for Oligo-15, KdFP = 290 ± 20 nM; for Oligo-10, KdFP = 34 ± 2 nM (Table and vide subra)]. The affinity of the dZ-modified DNA was still substantially higher than that of the dU-containing ssDNA that mimics the product of deamination [Oligo-15 vs Oligo-11 (Table )]. The comparison of Oligo-10, -12, -13, and -14 shows some increase in binding affinity with a length from seven to ten nucleotides and no further increase in binding affinity up to 13 nucleotides (Oligo-13).

ITC experiments[29] (Table ) were initially performed in a high-salt buffer (800 mM ionic strength), which was necessary for sample stability over the multiday, multidimensional NMR measurements.[29] For Oligo-1 (20-mer, A8T2CA9) and the much shorter Oligo-2 (9-mer, AT3CAT3) with a different sequence, very similar dissociation constants (KdITC) of 24 ± 7 and 25 ± 1 μM, respectively, were obtained (Table ), the former value being the same as that reported previously.[29] Given the very marked differences from the FP dissociation constants, further studies proceeded in buffer with a lower salt concentration (medium buffer, 150 mM ionic strength), comparable to that for FP measurements. The new KdITC values obtained for Oligo-1 and Oligo-2 were more than 100 times lower than values measured in the high-salt buffer (Table ). Moreover, differences in binding were now evident, such that the 20-mer bound nearly twice as strongly (KdITC = 0.11 ± 0.05 μM) as the 9-mer (KdITC = 0.20 ± 0.04 μM). Further shortening of the DNA sequence to five nucleotides resulted in a decreased affinity for the protein (Oligo-4, -5, and -6 compared to Oligo-3 in Table ), showing that binding affinity increases with oligo length and confirming that a length of ∼9–10 nucleotides is a good compromise between the length of the oligo and binding affinity. This length is used here and was used previously in NMR activity assays.[32,48]

In general, KdFP values were somewhat lower than KdITC values (Tables and 2), which can likely be attributed to differences in experimental conditions: competition binding studies for KdFP versus direct binding experiments for KdITC. Additionally, the different concentrations of assay components used may contribute to different oligomeric states of the protein, thereby affecting the measured binding affinities. The oligomeric state of A3A-E72A under these conditions warrants further investigation. Nonetheless, consistent trends were observed between these two biophysical methods. Shortening of the oligo resulted in a reduced level of binding to A3A-E72A in both methods [Oligo-4 and -5 vs Oligo-3 (Tables and 2)], consistent with the published data.[47] In parallel with FP results and despite the different oligo sequences that were used, introduction of dZ into the ssDNA sequence (for Oligo-7, KdITC = 0.97 ± 0.15 μM) resulted in weaker binding, relative to that of the corresponding cytidine-containing ssDNA (for Oligo-3, KdITC = 0.24 ± 0.10 μM); specifically, 8.5- and 4-fold increases in KdFP and KdITC, respectively, were observed (Table ). The addition of a 5-methyl substituent to dZ (dZMe), Oligo-8, had a negligible effect on binding relative to that of unmethylated Oligo-7 (Table ). This impaired binding of transition-state analogues, dZ- or dZMe-ssDNA, to the E72A mutant, compared to substrate, is consistent with Glu72 being a critical residue in the formation of the transition state for active A3 enzymes.

To test the importance of the catalytic glutamate (Glu72 in A3A), we evaluated binding of ssDNAs to two protein constructs of the catalytically active C-terminal domain of mutant A3B, A3BCTD-QM-ΔL3 [quadruple mutant with loop 3 removed (Figure S3)] and its catalytically inactive derivative, A3BCTD-QM-ΔL3-E255A.[30] For this experiment, we cannot use active A3A, the most potent A3 deaminase,[36] as it will fully deaminate the substrate during the experiment. The A3BCTD-QM-ΔL3 enzyme has the essential glutamic acid in the active site but because of the removal of loop 3 is only very weakly active in vitro:[31] low conversion of the substrate to the product was detected by 1H NMR spectroscopy only after incubation overnight (Figure S4). The interaction of TSA with the active site is not likely to be affected by the deletion of loop 3 from these proteins, as the loss of deaminase activity and associated binding affinity is due to effects outside of the active site. Deaminase activity can be restored by swapping loop 1 of A3BCTD-QM-ΔL3 with loop 1 of A3A, yielding the active A3BCTD-QM-ΔL3-AL1swap enzyme.[31] On the other hand, in the A3BCTD-QM-ΔL3-E255A mutant the active-site glutamic acid (equivalent to Glu72 in A3A) is replaced with alanine. Therefore, these proteins provide a unique pair for evaluating the importance of the catalytic glutamate for binding of dZ-containing oligos.

No binding of ssDNA to A3BCTD-QM-ΔL3 was seen by ITC under our standard conditions [50 mM citrate-phosphate buffer, 200 mM NaCl, and 2 mM β-mercaptoethanol (pH 5.5)], consistent with this mutant’s very low deaminase activity. Moreover, for a weak binding event such as this to be studied by ITC, the required concentration of the protein would be in the high micromolar range, which is prohibitive because A3 proteins tend to precipitate from solution at this concentration.[49] Consequently, we decided to use our previously published thermal shift assay,[29] in which high ligand concentrations are used to enhance complex formation (Figure ).[50] Protein unfolding as the temperature increases is reported by the increased fluorescence of the nonpolar dye SYPRO Orange, which binds to hydrophobic patches exposed upon thermal denaturation. Usually, a higher thermal stability is associated with the strong binding of the ligand, in our case ssDNA, to the protein. We observed that the level of binding of oligos to inactive A3BCTD-QM-ΔL3-E255A is significantly lower than to weakly active A3BCTD-QM-ΔL3, which retains the catalytic glutamic acid. Moreover, for A3BCTD-QM-ΔL3, the ΔTm value for dZ-containing Oligo-9 is 0.72 ± 0.10 °C higher than that for the substrate containing dC (in Figure , see the 95% confidence interval), which is the expected result for an ssDNA that bears a TSA instead of dC. For the binding of inactivated A3BCTD-QM-ΔL3-E255A to different DNA oligos, the binding of dZ oligo cannot be distinguished from binding of the substrate, illustrating the importance of the active-site glutamate for binding of transition-state analogues.

Figure 2

Results of thermal shift assay for A3BCTD-QM-ΔL3 (green) and A3BCTD-QM-ΔL3-E255A (blue) in the presence of the product of deamination (5′-dATTTdUATTT), substrate (5′-dATTTCATTT), and dZ-containing Oligo-9 (dATTTdZATTT). The concentration of proteins was 20 μM, and the concentration of oligos was 100 μM in the buffer: 50 mM citrate-phosphate (pH 5.5), 200 mM NaCl, 2 mM β-mercaptoethanol (pH 5.5), and 200 μM 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS). Confidence intervals (95%) are shown as error bars.

Altogether, these results demonstrate that the choice of buffer, especially the ionic strength, affects strongly, by several orders of magnitude, the affinity of ssDNA for the A3 proteins. Moreover, binding of inactive protein to ssDNA-containing TSA is not a reliable predictor of inhibitory potential of these oligonucleotides. Therefore, we decided to use our previously described NMR-based activity assay[32] to assess directly the inhibition of substrate deamination in fully active A3 enzymes by ssDNA-containing TSAs.

2′-Deoxyzebularine Incorporated into ssDNA Is a Micromolar Inhibitor of A3 Enzymes

The Michaelis–Menten constants for the enzymes and their mutants used in this study (described in the Supporting Information), along with their ssDNA substrates, are summarized in Table . The behavior of A3BCTD-QM-ΔL3-AL1swap[30] (see a description of the enzyme in Figure S3), which has loop 1 of A3A transplanted into A3BCTD, was intermediate between previously reported activities for A3A[40] and A3GCTD,[32] consistent with the previously reported data.[36] However, the kcat of A3BCTD-DM [DM, double mutant (see Figure S3)] was anomalously low [0.008 ± 0.001 s–1 compared to 0.10 ± 0.04 s–1 for A3GCTD[32] (Table )]. Residual enzyme activity in the presence of dZ-ssDNA inhibitors was monitored directly by our NMR method.[32,48]

Table 3

Michaelis–Menten Constants for Selected A3 Enzymes and Mutants Measured by the NMR Assay[32,40]

protein	substrate	kcat (s–1)	Km (μM)	kcat/Km (s–1 μM–1)
A3GCTDa	Oligo-3	0.10 ± 0.04	570 ± 90	0.00018
A3Ab	Oligo-2	1.2 ± 0.1	66 ± 7	0.018
A3BCTD-QM-ΔL3-AL1swapc	Oligo-2	0.28 ± 0.04	200 ± 30	0.0014
A3BCTD-DMc	Oligo-2	0.008 ± 0.001	320 ± 60	0.000025
A3BCTD-QM-ΔL3c	barely active

See ref (32).

See also ref (40).

See the Supporting Information.

Only the use of dZ instead of dC in the preferred A3GCTD substrate [Oligo-7 versus Oligo-9 (Figure A,B)] conferred significant inhibition of A3GCTD deaminase activity. dZMe-ssDNA (Oligo-8) had a marginal effect on the conversion of the preferred A3GCTD substrate (AT2C3A2T2, Oligo-3) (Figure A). These data are consistent with 5-Me-cytidine being a much poorer substrate than cytidine for A3 proteins, with the exception of A3H.[51−53] Therefore, for further investigations, we focused on the evaluation of dZ-containing oligos. For both A3BCTD-QM-ΔL3-AL1swap and A3BCTD-DM (Figure and the Supporting Information), using preferred substrate 5′-AT3CAT3 (Oligo-2), significant inhibition was observed with dZ-ssDNA [Oligo-9 (Figure C,D)]. The linear dependence of inverse deamination speed on inhibitor concentration was analyzed as described in the Supporting Information, assuming competitive inhibition, yielding micromolar inhibition constants (Ki) (Figure B–D). In particular, for A3BCTD-QM-ΔL3-AL1swap, the Ki and KdITC for Oligo-9 are very similar, 7.5 ± 1.7 and 5.5 ± 0.6 μM, respectively (Figure C and Table ). No binding of Oligo-7 (or any other oligo) to A3GCTD (or catalytic mutant) was seen by ITC [75 mM sodium phosphate and 75 mM citrate (pH 5.5)] due to the low DNA binding affinity for A3GCTD [Km(Oligo-3) ≈ 600 μM, and Ki(Oligo-7) ≈ 50 μM] and low protein solubility.

Figure 3

Inhibition of A3 proteins by modified oligos in the NMR deaminase assay. (A) Effects of dZ and dZMe-containing ssDNAs on deamination of Oligo-3 (dAT2C3A2T2, 500 μM) by A3GCTD (850 μM) at 298 K. The speed of deamination in the absence of inhibitor (No inhibitor) and in the presence of 100 μM ssDNA inhibitors (dZ-containing Oligo-7 and Oligo-9 as well as dZMe-containing Oligo-8) is shown. (B) Inhibition of A3GCTD-catalyzed deamination of Oligo-3 by dZ-containing Oligo-7. (C) Inhibition of A3BCTD-QM-ΔL3-AL1swap-catalyzed deamination of Oligo-2 by dZ-containing Oligo-9. (D) Inhibition of A3BCTD-DM-catalyzed deamination of Oligo-2 by dZ-containing Oligo-9. The bold C is the target 2′-deoxycytidine deaminated by the enzyme. In all cases, the DNA substrate and inhibitor have the same sequence except that the cytidine of the substrate has been changed to dZ.

To confirm the competitive nature of inhibition, the substrate concentration was varied while the inhibitor (Oligo-9) and enzyme (A3BCTD-QM-ΔL3-AL1swap) concentrations were kept constant. In the double-reciprocal plot shown in Figure , the lines from the least-squares fits to the data in the absence and presence of inhibitor cross the y-axis at essentially the same point (1/vmax = 71 ± 12 s μM–1 and 1/vmax = 83 ± 32 s μM–1, respectively). This result validates our initial prediction that dZ-ssDNAs are competitive inhibitors of A3 enzymes.

Figure 4

Double-reciprocal plot (1/V vs 1/[S]) of deamination of the DNA substrate (Oligo-2) by A3BCTD-QM-ΔL3-AL1swap at different substrate concentrations in the absence and presence of dZ-containing Oligo-9 at 20 μM.

Discussion

By means of several complementary binding and activity assays, we have characterized inhibition of selected A3 variants by chemically modified ssDNAs possessing 2′-deoxyzebularine analogues. One finding was that binding of dZ-ssDNA to inactivated proteins in which the active-site glutamic acid is replaced with alanine (Glu72 in A3A and Glu255 in A3BCTD) was significantly reduced in comparison to the substrate. This indicates that the glutamate in the active site is strongly involved in interaction with the transition state of deamination. These data are consistent with a model in which the Glu modifies the TSA by protonation of N3 of the nucleobase causing the addition of a water molecule to the C4 of dZ, as established for CDA.[54] In addition, our data clearly show an influence of buffer on dissociation constants, with a high ionic strength (800 mM) depressing the affinity by more than 2 orders of magnitude relative to that with a medium ionic strength (∼100 mM). Our observation that dZMe in the structure of ssDNA did not provide significant inhibition of A3GCTD is consistent with earlier observations that 5-Me-cytidine was not deaminated by A3GCTD.[51−53] On the other hand, dZ instead of dC in ssDNA produced competitive, micromolar inhibitors of A3GCTD, A3BCTD-QM-ΔL-AL1swap, and A3BCTD-DM. The fact that dZ and dZMe used in the same oligo sequence had different inhibitory effects on A3GCTD under identical conditions means that structure of the TSA determines the inhibitory potential of the TSA-containing oligo. Moreover, adjacent nucleotides proximal to the TSA define the selectivity for binding and concomitant inhibition of a particular A3 family member as in the case of selective inhibition of A3GCTD by Oligo-7 and not by Oligo-9 (Figure A). These sequence-dependent molecular recognition properties provide opportunities for the design of specific inhibitors of A3 family members.

Through high-throughput screening efforts, covalent small molecule A3G inhibitors with low micromolar potencies have been identified.[33,55] To date, no small molecule A3A or A3B inhibitors have been reported. Chemically modified ssDNAs have been utilized in previous studies to understand the deamination of cytosine in the absence of structural information about the ssDNA/A3 complex.[56] Here, we evaluated chemically modified ssDNAs as potential A3 inhibitors and found similar (micromolar) inhibition constants as found for small molecule A3G inhibitors. An approximately 10–30-fold decrease in the apparent inhibition constants (Ki) of dZ oligos over Km of the corresponding dC-containing substrates was observed in our study. In comparison, zebularine as an individual nucleoside shows a 100-fold lower Ki in comparison to the Km of cytidine for CDA.[54] This indicates that the preferences for TSA vary between A3 and CDAs. In addition, the strength of binding of the dZ-ssDNA to the A3 enzyme may be mediated by the interactions of surrounding nucleotides with amino acids close to the active site. The latter possibility is supported by the fact that introduction of modified nucleotides at positions −2 to +1 substantially affected the rate of deamination (target dC has number 0, nucleotides in the 5′-direction have a minus sign, and nucleotides in the 3′-direction have a plus sign).[56] Recent work has revealed that ssDNA adopts a U-shape for the target dC to enter the active site of the protein,[29,30,42] such that nucleotides at positions −2, −1, and +1 have strong interactions with the proteins, as well. Our previously published NMR-based method of small changes shows additional transient interactions with nucleotides farther from the active site.[29] This also correlates with our observations that short DNA sequences (5-mers) bind much more weakly than longer ssDNAs (>9-mers) to A3A-E72A (Tables and 2) and with the results of other groups.[39,40,47]

The observed inhibition of A3 enzymes by dZ-ssDNAs highlights the importance of Glu72 for A3A (Glu255 for A3BCTD and Glu259 for A3GCTD) in controlling substrate and dZ-ssDNA binding. Therefore, development of A3 inhibitors based on TSAs that specifically react with the active-site water requires evaluation on active enzymes in contrast to the common practice of evaluation of binding to inactive active-site mutants.

Overall, our work shows that transition-state analogues potently inhibit selected A3 variants when incorporated into ssDNA. The inhibition of wild type A3A and full length A3B and A3G is a subject of ongoing investigations. Nevertheless, our results provide a starting point to rationally create new A3 inhibitors, which have a potential to be further developed into adjuvants to be used in antiviral and anticancer treatments. Because cytidine-containing ssDNA species are not substrates for CDA,[57] which accepts only individual nucleotides, our inhibitors for A3 enzymes will not affect primary metabolic functions of CDA. Taking advantage of different preferences of A3 for nucleotides surrounding the reactive 2′-deoxycytidine may also allow development of A3-specific inhibitors, targeting individual family members.