Jurkat (TIB-152), Raji (CCL-86), HL-60 (CCL-240), K562 (CCL-243), HEL (TIB-180), H1975 (CRL-5908), HCC827 (CRL-2868), HEK293T (CRL-3216), Phoenix-Ampho (CRL-3213), and HeLa (CRL-2) were purchased from American Type Culture Collection (ATCC, Manassas, VA). NB4, MOLM14, CHRF-288-11 (CHRF), and PC9 cells were kindly provided by Dr. Daniel G. Tenen. SET-2 cells were kindly provided by Dr. Kimiharu Uozumi. Mononuclear cells (MNCs) and polymorphonuclear leukocytes (PMNs) were isolated from healthy volunteers as previously described [19, 20].

The normalized expression dataset as RSEM (RNA-seq.V2) and its correlation with overall survival from 200 primary acute myelogenous leukemia (AML) samples [21] were obtained from cBioPortal (http://www.cbioportal.org/) [22, 23] and 173 samples which have mRNA expression data were analyzed in GraphPad Prism 6 (GraphPad Software, La Jolla, CA).

Total RNA was isolated from leukemia and lung cancer cell lines as well as MNCs and PMNs using RNeasy Mini Kit (Qiagen, Valencia, CA). Complementary DNA was synthesized using SuperScript II (Thermo Fisher Scientific, Waltham, MA). Primers used to amplify A3D splicing variants were 5’-GGAGACTGGGACAAGCGTAT-3’ and 5’- GGATGACCAGGCAAAAGCAC -3’.

Plasmids containing human APOBEC3B or APOBEC3G coding sequence were previously described [24]. Expression vectors for HA-tagged A3Dv1, v2 and v6 were generated by sub-cloning of coding sequences into pCAG-GS vector from human cell lines. To generate expression vectors for A3Dv7, coding sequences of exon 1 and exon 5 to exon 7 were released from pCAG-GS-A3Dv1 expression vector by PCR and assembled by the Gibson assembly method [25]. Coding sequences of APOBEC3Dv1, v2, v6, v7 or APOBEC3B were released and cloned into MigR1-IRES-EGFP vector to make retroviral constructs. The expression vector for UGI, pEF-UGI was provided by Dr. Ruben S. Harris [26]. pCAG-GS-A3Dv1-E80/264Q and pCAG-GS-A3Dv6-E264Q were generated by PCR based mutagenesis using pCAG-GS-A3Dv1, pCAG-GS-A3Dv6 as templates.

The mRNA expression levels of genes were measured by Q-RT-PCR using the Rotor Gene 6000 sequence detection system (Qiagen, Valencia, CA). RNA was reverse transcribed, and the resulting cDNA was used in amplification reactions with iTaq Universal SYBR Green Supermix (BIO-RAD, Hercules, CA). The expression levels of the glyceraldehyde 3-phosohate dehydrogenase (GAPDH) were used for normalization. Reactions were performed in technical triplicates. The primers for qPCR are listed in Table S2.

HEK293T cells were transfected with pCAG-GS-A3Dv1, v2, v6, v7 or A3B using TransIT-X2 (Mirus, Madison, WI). After 48 hours of incubation, cells were lysed in cell lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 1 mM β-glycerophosphate, 1 mM Na3VO4, 1 mM NaF, protease inhibitor cocktail set III (EMD Millipore, Darmstadt, Germany) and 1 mM PMSF). Protein lysates (20 μg) were loaded on 10% SDS-polyacrylamide gels and transferred on PVDF membranes (Millipore, Bedford, MA). Anti-HA antibody (H3663) (Sigma-Aldrich, St. Louis, MO), anti- β-Actin antibody (sc-47778) (Santa-Cruz, Santa Cruz, CA) and Anti-p24 antibody (ab9071) (Abcam, Cambridge, UK) were used at 1:10000, 1:500 and 1:1000 dilutions, respectively.

HeLa cells were transfected with pCAG-GS-A3Dv1, v2, v6, v7, A3B or empty control using TransIT-X2 (Mirus, Madison, WI). After 48 hours of incubation, cells were fixed in 4% paraformaldehyde and blocked in PBS with 5% normal horse serum (Life Technologies, Palo Alto, CA) and 0.25% Triton X-100 (Sigma-Aldrich, St. Louis, MO). Cells were incubated with anti-HA-tag antibody at 1:1000 dilution (ab18181) (Abcam, Cambridge, UK), followed by anti-mouse IgG antibody conjugated with Alexa488 at 1:400 dilution (A21202) (Life Technologies, Palo Alto, CA). Slides were mounted with VECTASHIELD with DAPI (Vector Labs, Burlingame, CA), and images were captured using Zeiss LSM 510 Meta microscope (Zeiss, Oberkochen, GER).

All primers used for 3D-PCR are listed in Table S2. For the foreign DNA editing assay, HEK293T cells were co-transfected with pcDNA3.1(−)-EGFP, pEF-UGI, and a pCAG expression vector for A3Dv1, v2, v6, v7, A3B or empty control using TransIT-X2 (Mirus, Madison, WI). 4 days after transfection, DNA was isolated using DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA). Q5 High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA) was used for the first round of 3D-PCR with following reaction profile: 30 sec at 98 °C; 30 cycles of 10 sec at 98 °C; 30 sec at 65.7 °C; 30 sec at 72 °C; and followed by 2 min at 72 °C. The first round PCR products were purified using QIAquick PCR Purification Kit (Qiagen, Valencia, CA) and 25 ng of purified PCR products were used as templates for the second round of 3D-PCR, which was performed with HotStar HiFidelity Polymerase Kit (Qiagen, Valencia, CA). The reaction profile was: 5 min at 95 °C; 35 cycles of 15 sec at 86.3-87.9 °C; 1 min at 62 °C; 80 sec at 72 °C; and followed by 10 min at 72 °C. The second PCR products derived from a denaturation temperature (Td) of 86.8 °C in cells expressing A3D or A3B, and Td of 87.9 °C in mock cells were gel extracted using Gel Extraction Kit (Qiagen, Valencia, CA). The extracted second round PCR products were cloned using pGEM-T Easy Vector System (Promega, Madison, WI) and sequenced. For the genomic TP53 DNA editing assay, HEK293T cells were retrovirally infected with MigR1-IRES-EGFP expressing A3Dv1, v2, v6, v7, A3B or empty control, and EGFP positive cells were sorted using FACSAria (BD Biosciences, Franklin Lakes, NJ) 72 hours after infection. DNA was isolated 4 weeks after retroviral infections using DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA). The first round of 3D-PCR with the following reaction profile was performed: 30 sec at 98 °C; 30 cycles of 10 sec at 98 °C; 30 sec at 60 °C; 30 sec at 72 °C; and followed by 2 min at 72 °C. 25 ng of purified first round PCR products were used as templates for the second round of 3D-PCR with the following reaction profile: 5 min at 95 °C; 35 cycles of 15 sec at 83.8-85.3 °C; 1 min at 62 °C; 80 sec at 72 °C; and followed by 10 min at 72 °C. The amplifications of second round PCR products from the lowest Td in each group were cloned and sequenced. The primers for 3D-PCR are listed in Table S2.

We prepared luciferase reporter viruses without Vif by co-transfection of pNL43/ΔEnv-Luc plus pVSV-G with expression vectors for A3G or A3D variants using X-tremeGene DNA transfection kit (Roche, Basel, Switzerland). We collected viruses from the supernatant 48 hours after transfection and measured virus titer with an enzyme- linked immunosorbent assay kit for the p24 antigen (RETRO-TEK, ZeptoMetrix, Buffalo, NY). We then challenged an adjusted amount of viruses to HEK293T cells. 24 hours after infection, cells were lysed in passive lysis buffer (Promega, Madison, WI), and their luciferase activities were measured by ARVO microplate reader (Perkin-Elmer, Waltham, MA). We presented values as percent infectivity normalized to the value of empty vector sample.

Genomic DNA was extracted from HEK293T cells infected with VSV-G pseudotyped Vif-deficient HIV-1 that contained lentivirus for A3G or A3D variants after 2 days of infection using the QuickGene DNA whole blood kit S (KURABO, Osaka, Japan). We amplified HIV-1 gag sequences using regular PCR with the KOD FX Neo (ToYoBo, Osaka, Japan). The primer set is available in Table S2. We cloned amplicons in T-vector pMD20 (Takara Bio, Kyoto, Japan) and picked 5 colonies to sequence using the 3130xl Genetic Analyzer (Applied Biosystems, Waltham, MA).

HEK293T cells transiently transfected with expression constructs containing A3Dv1, v2, v6, v7, A3B or an empty control were analyzed by LSRII flow (BD Biosciences, Franklin Lakes, NJ) at 4 days after transfection. We used FlowJo (FLOWJO, LLC, Ashland, OR) to analyze data.

Differences between the experimental groups were tested with Student’s t-test. Differences between the survival curves were tested with log-rank test (Mantel-Cox). P-values of less than 0.05 were considered statistically significant.

As an initial step to understand the roles of A3 on leukemia development, we used Q-RT-PCR to survey the mRNA levels of A3 members in a small panel of leukemia cell lines, together with normal MNC isolated from healthy volunteers. While A3A and A3H were not detected, other A3 genes were expressed in all leukemia cells (Fig. 1A). Next, we analyzed previously-published The Cancer Genome Atlas (TCGA) data from patients with AML [21] to find somatic alterations in APOBEC3 genes using cBioPortal [22, 23]. We found that each APOBEC3 was upregulated by 4-7%, and 19% of patients had an upregulation of at least one APOBEC3 (Supplementary Fig. S1A). Overall survival of patients with APOBEC3 upregulation was lower than those without upregulation (p=0.0256, Hazard ratio 1.614, 95%CI; 1.072 to 2.829) (log-rank test) (Supplementary Fig. S1B), suggesting an important role of A3s on prognosis in patients with AML. As it has been shown that A3A, A3B, and A3D show capacity to inflict DNA damage [18], we decided to focus on A3D because its role in cancer development and drug resistance are unknown. Our attempt to generate A3D expression constructs revealed that there are multiple bands in PCR products (Fig. 1B) from normal human cells (peripheral blood mononuclear cells: MNCs, polymorph nuclear leukocytes: PMNs), leukemia and lung cancer cell lines. Following cloning of all transcripts, we identified 5 transcript variants (A3Dv1, v3, v4, v5, and v6) from leukemia cell lines. According to Ensembl (http://ensemblgenomes.org/), 3 transcripts have been reported so far (A3D-201, 202, and 203). A3Dv1 corresponds to either A3D -201 and A3Dv6 to A3D-202. Although we were unable to detect A3D-203 in our analysis, we included this transcript for further study (A3Dv7). We also included another variant (A3Dv2), which lacks exon 4 and exon 5, that was detected in lung cancer cell lines (Fig. 1B and manuscript in preparation). These variants were numbered in descending order of PCR products length. We found that A3Dv1 was a full-length mRNA containing all 7 exons and harbored two cytidine deaminase domains (CD1 and CD2) (Fig. 1C and 1D). Protein products from A3Dv2 and A3Dv7 harbored a single CD1 and CD2 catalytic domain, respectively (Fig. 1D). A product from A3Dv6 was predicted to harbor a single hybrid catalytic domain (Fig. 1D). Products from A3Dv3, A3Dv4, and A3Dv5 had neither domain, suggesting that these variants are not functional. Therefore, we tested 4 A3D variants (A3Dv1, v2, v6, v7) in further studies (Fig. 1D).

To assess the protein stability of these 4 variants, we generated expression constructs encoding each variant with an HA-tag and transfected them into HEK293T cells. We also generated constructs expressing A3B as control. We detected A3Dv1 (46 kDa), A3Dv2 (21 kDa), A3Dv6 (30 kDa), and A3Dv7 (23 kDa) as expected (Fig. 1E).

It has been shown that A3D, A3F, and A3G are located in the cytoplasm, A3B in the nucleus, and A3A, A3C, and A3H, each with a single deaminase domain, are present in both compartments [18]. Thus, we examined where the cellular localization will be for the full length A3D (A3Dv1) protein and truncated A3D (A3Dv2, v6, v7) proteins. HeLa cells were transfected with pCAG-GS-A3Dv1, v2, v6, v7, A3B or empty control and analyzed 2 days after transfection. Confocal microscopy analysis revealed that full-length A3Dv1 was located in the cytoplasm while A3B was located in the nucleus as previously shown [18] (Fig. 2). Although all truncated isoforms (A3Dv2, v6, 7v) showed cell-wide distribution. A3Dv2 and A3Dv6 were predominantly located in the nucleus, and A3Dv7 was primarily located in the cytoplasm (Fig. 2).

APOBEC3 proteins including A3D are originally shown to inhibit Vif-deficient HIV-1 infectivity [27]. We tested the anti-viral activity of A3D variants against VSV-g pseudotyped HIV-1 ΔVif viruses using luciferase reporter assays. All A3D variants were detected in the viral supernatant (Fig. 3A). A3Dv1 inhibited HIV-1 infectivity less effectively than A3G as previously reported [27], whereas the other A3D variants (A3Dv2, v6, v7) showed similar antiviral activity against HIV-1 (Fig 3B). These data suggest that A3D variants retain binding ability to HIV-1 gag and RNA for incorporation into the budding HIV-1 virion and show active anti-HIV-1 activity. To analyze the pattern of mutation, we cloned and sequenced HIV-1 gag which was amplified by regular PCR method using the genomic DNA of infected cells. We detected 81.9 and 11.4 mutations per 10³ base pairs (bps) in HIV-1 gag derived from A3G or A3Dv1 containing viruses, respectively. A3Dv2, v6 and v7 induced 7.2, 9.2, and 5.1 mutations per 10³ bps in HIV-1 gag, respectively (Fig. 3C). Most of these mutations were C/G to T/A transitions (Fig. 3C, 3D). A3G preferred CC dinucleotide context, whereas A3D variants targeted TC and CC dinucleotides (Fig. 3E). No mutations were detected in the PCR products of mock cells.

To confirm emergence of mutations are mediated by catalytic activity we generated constructs expressing A3Dv1-E80/264Q and A3Dv6-E264Q, which harbor missense mutations at catalytic domains (E80Q for CD1 and E264Q for CD2) and impair catalytic activity. Protein expression was similar in all extracts. (Fig. 4A). Antiviral activity was partially reversed by A3Dv1-E80/264Q or A3Dv6-E264Q compared to wild type A3Dv1 or A3Dv6 (Fig. 4B). In terms of frequency of mutations, we detected 11.0 and 1.5 mutations per 10³ bps in HIV-1 gag derived from A3Dv1 or A3Dv1-E80/264Q containing viruses, respectively. A3Dv6 and A3Dv1-E264Q induced 6.5 and 2.0 mutations per 10³ bps in HIV-1 gag, respectively (Fig. 4C). These results indicate that mutations induced by A3Ds depend on their catalytic activity.

We then asked whether full-length and truncated forms of A3D have ability to induce mutations in DNA. To examine whether A3D proteins induce mutations in foreign DNA in human cells, we co-transfected HEK293T cells with an expression vector for EGFP, an expression vector for Uracil-DNA glycosylase inhibitor (UGI), and pCAG-GS containing A3Dv1, v2, v6, v7, A3B, or empty control as previously described [24]. Interestingly, flow cytometry analysis revealed that the cells transfected with A3D variants and A3B had lower EGFP positive cell rate than cells transfected with empty pCAG-GS vector as a control after 4 days (Fig. 5A). To confirm catalytic activity is crucial for mutagenesis, we used A3Dv1-E80Q/E264Q and A3Dv6-E264Q as negative controls. The EGFP positive cell rate remained similar in cells transfected with A3Dv1-E80Q/E264Q or A3Dv6-E264Q, suggesting that catalytic activity is crucial for attenuate EGFP activity, possibly induced by mutagenesis. To test this possibility, DNA was isolated from these cells for 3D-PCR analysis as previously described [24]. PCR products were detected at a denaturation temperature (Td) of 87.9 °C in all samples. Amplification at Td of 86.8 °C was lost in control cells, whereas PCR products were still amplified in the cells transfected with all A3D variants and A3B expression vectors (Fig. 5C). To confirm that mutations were introduced in the EGFP gene, we cloned and sequenced PCR products at Td of 86.8 °C in cells expressing either A3D variants or A3B and PCR product of mock cells at Td of 87.9 °C as a control. Mutation frequencies were 6.1-10.5 per 10³ bps in cells transfected with each A3D variant, whereas there were 18.7 mutations per 10³ bps in cells transfected with A3B (Fig. 5D and 5E). No mutations were detected in the PCR products of mock cells at Td of 87.9 °C (data not shown). Only approximately 30-60% of mutations were C/G to T/A transitions in cells transfected with A3D variants compared to 90% of mutations in cells transfected with A3B (Fig. 5D). In terms of target dinucleotide sequences, A3B and A3Dv2 preferred 5’-TC, while other A3D variants did not show preference (Fig. 5F). We also observed clones with 47- to 334- bps deletions in all A3D variants or A3B expressing cells, but not in mock cells (Table 1). Interestingly, 1- to 9- bases micro homologous sequences in the broken ends before joining were detected in some clones. Existence of these micro homologous sequences suggests that DNA double strand breaks were repaired by microhomology-mediated end joining (MMEJ).

To assess whether A3D can induce mutations in genomic DNA, we generated HEK293T cells stably expressing EGFP and either A3Dv1, v2, v6, v7, or A3B. EGFP positive cells were sorted 72 hours after retroviral infection and cultured. 4 weeks after retroviral infection, genomic DNA was extracted and subjected to 3D-PCR to detect whether mutations were introduced in the TP53 gene locus. The amplifications existed at a Td of 84.3 °C in cells expressing either an A3D variant or A3B; however, no bands were detected in DNA from mock cell cultures (Fig. 6A). Sequencing of the PCR products at the lowest Td revealed that the mutation frequency in the TP53 gene was 0.4-4.5 per 10³ bps in cells expressing A3D variants and A3B (Fig. 6B). The rates of C/G to T/A transition in these mutations were 70%, 100%, 67% and 85% in the cells expressing A3Dv1, v2, v6 and A3B, respectively (Fig. 6B and 6C). However, mutations in the cell expressing A3Dv7 had no difference compared to mock cells (Fig. 6B and 6C). Unexpectedly, a preference for 5’-TC was not observed in cells expressing A3D variants and A3B (Fig. 6D). The preferred target of A3Dv1 and A3Dv6 was 5’-GC dinucleotide while A3Dv2 preferred 5-CC (Fig. 5D). Clones that harbored deletions in the TP53 gene were also observed (Table 2). Only 1- to 3- bps homologous sequences internal to the broken ends before joining were present in the TP53 gene.