Frame-shifted APOBEC3A encodes two alternative proapoptotic proteins that target the mitochondrial network

The human APOBEC3A (A3A) cytidine deaminase is a powerful DNA mutator enzyme recognized as a major source of somatic mutations in tumor cell genomes. However, there is a discrepancy between APOBEC3A mRNA levels after interferon stimulation in myeloid cells and A3A detection at the protein level. To understand this difference, we investigated the expression of two novel alternative “A3Alt” proteins encoded in the +1-shifted reading frame of the APOBEC3A gene. A3Alt-L and its shorter isoform A3Alt-S appear to be transmembrane proteins targeted to the mitochondrial compartment that induce membrane depolarization and apoptosis. Thus, the APOBEC3A gene represents a new example wherein a single gene encodes two proapoptotic proteins, A3A cytidine deaminases that target the genome and A3Alt proteins that target mitochondria.

The human APOBEC3A (A3A) cytidine deaminase is a powerful DNA mutator enzyme recognized as a major source of somatic mutations in tumor cell genomes. However, there is a discrepancy between APOBEC3A mRNA levels after interferon stimulation in myeloid cells and A3A detection at the protein level. To understand this difference, we investigated the expression of two novel alternative "A3Alt" proteins encoded in the +1-shifted reading frame of the APOBEC3A gene. A3Alt-L and its shorter isoform A3Alt-S appear to be transmembrane proteins targeted to the mitochondrial compartment that induce membrane depolarization and apoptosis. Thus, the APOBEC3A gene represents a new example wherein a single gene encodes two proapoptotic proteins, A3A cytidine deaminases that target the genome and A3Alt proteins that target mitochondria.
The human APOBEC3 (A3) locus encodes six functional polynucleotide cytidine deaminases (A3A-C and A3F-H) (1), originally described as innate cellular defenses against retroviruses (2)(3)(4)(5), DNA viruses (6)(7)(8), and retroelements (9)(10)(11) mostly through cytidine deamination of single-stranded DNA (ssDNA). The mutagenic activity of APOBEC3A (A3A) and APOBEC3B (A3B) has recently been demonstrated to introduce somatic mutations in genomic DNA (12,13), and those two enzymes are now recognized as endogenous causal agents responsible for the accumulation of CG to TA transitions in cancer genomes (14). As a result of cytidine deamination, inappropriate uracil bases in DNA can be removed by uracil N-glycosylase enzyme (UNG) and generate abasic sites, further processed by apurinic/apyrimidinic endonucleases, APE1 and APE2 resulting in double-strand break (DSB) formation and apoptosis (15,16). It is now accepted that both A3A and A3B are intrinsic mutators of chromosomal DNA, although some debate persists regarding the contribution of each enzyme to the accumulation of mutations paving the way to cancer formation (17)(18)(19)(20)(21)(22). However, recent data revealing episodic waves of APOBEC3 (A3) mutations in various cancer genomes (21,23) favor a scenario in which A3A plays the main role in oncogenesis, since this enzyme is known to be upregulated through interferon (IFN) signaling in response to many cellular stress (12,(24)(25)(26), further emphasizing the longstanding observation that cancer emerges on a background of chronic inflammation (27).
To limit nuclear DNA damage, A3A expression is controlled at multiple levels. If A3A promoter regulation remains so far elusive, with the exception of demonstrated IFN responsiveness in hematopoietic cells (24), A3A mRNA expression appears almost undetectable in other tissues. A3A-3 0 UTR also contributes to reducing A3A expression, since its substitution by A3B-3 0 UTR in the context of a prevalent A3B deletion allele (ΔA3B) (28) results in increased A3A levels and nuclear DNA (nuDNA) damages (13), in keeping with the overrepresentation of APOBEC mutations in the cancer genomes of ΔA3B patients (18). Ultimately, at the posttranslational level, the nuclear fraction of A3A has been demonstrated to be degraded by TRIB3 enzyme, reducing nuDNA editing and chromosomic DNA damage (29).
In the present work, we report the identification of a novel ORF of 96 residues in the +1-shifted reading frame that completely overlaps that of A3A. Initiation from two methionine codons gives rise to two proteins, A3Alt-L (10.5 kDa) and the smaller A3Alt-S (8.6 kDa) that differ only at their Ntermini by 18 residues. Functional characterization reveals that they specifically target mitochondria, resulting in mitochondrial membrane potential depolarization and apoptosis.
Thus A3A gene appears to be a new example of completely overlapping "dual coding genes," which are rare in the human genome (35)(36)(37)(38)(39). At high levels A3A cytidine deaminases are genotoxic while the A3Alt proteins are toxic for the mitochondria. Individually both are proapoptotic. As the gene is stimulated by IFN and other forms of cellular stress, A3A appears to be a bifunctional stress-sensitive cytocidal gene.

Results
APOBEC3A gene encodes two novel alternative proteins, A3Alt-L and A3Alt-S The AUGs initiation codons S1 and S2 respectively governing the expression of A3A-L and A3A-S cytidine deaminases in A3A mRNA are only present in an "adequate" Kozak context, compatible with leaky scanning. We therefore sought for other downstream AUG codons in A3A classical transcripts (NM_145699.4, ENST00000249116, ENST00000618553). We identified two AUG (S3 and S4) in the +1 shifted reading frame of A3A, both of which being in a strong Kozak context (Figs. 1A and S1A). S3 defines an ORF of 96 residues encoding a small putative protein A3Alt-L (10.5 kDa, pI = 12.2), while the S4 methionine can be used in some alternatively spliced transcripts of A3A (NM_001270406.1), giving rise to an even smaller protein, A3Alt-S, of 78 residues (8.6 kDa, pI = 12.5) (Fig. 1, A and B and S1A). A3Alt-L and A3Alt-S only differ by 18 N-terminal residues. The observation was further confirmed by in silico analysis using the OpenProt tool, designed to unravel cryptic proteins encoded within transcripts (40), predicting A3Alt proteins translation from A3A mRNAs (Fig. S1, B and C).
To experimentally evaluate the differential translation from those four initiation codons, reporter constructs were generated fusing Firefly luciferase cDNA downstream of A3A 5 0 sequence, either in A3A or in A3Alt + 1-shifted reading frames (Fig. 1C). AUG initiation codons were substituted by ACG by site-directed mutagenesis to estimate their relative contribution to protein expression (Fig. 1C). Reporter constructs were then transfected in HeLa cells along with a Renilla luciferase coding plasmid used as transfection control. After normalization to Renilla luciferase luminescence, quantitative Firefly detection revealed that Firefly luciferase expression driven by the combination of A3A initiation codons (S1 and S2, S1 + S2 + Luc) was on a par with the positive pcDNA3.1 Luc control (Fig. 1D). Unexpectedly in A3A S2 AUG (S1 − S2 + Luc) initiated a stronger expression than S1 alone (S1 + S2 − Luc) (Fig. 1D).
Analysis of initiation of A3Alt translation showed expression of 20% from A3Alt S3 and S4 (S3 + S4 + Luc) compared with A3A (S1 + S2 + Luc, Fig. 1D). Of the two potential A3Alt A E F B C D Figure 1. A3A mRNA encodes novel A3Alt proteins. A, schematic representation of differential translation from A3A mRNA. A3A-L and A3A-S cytidine deaminases are translated from S1 and S2 "adequate" Kozak initiation codons (in orange). Catalytic residues involved in zinc coordination are depicted in yellow. A3Alt-L and A3Alt-S are translated from S3 and S4 "strong" Kozak initiation codons (in green). Mitochondrial addressing signal RIFS is highlighted in red and predicted transmembrane domain is shown in blue. B, A3Alt-L and A3Alt-S protein sequences. Mitochondrial addressing motif is shown in red, transmembrane domain is represented in blue. C, schematic representation of reporter constructs harboring 5 0 sequence of A3A transcripts (in white) upstream of Firefly luciferase coding sequence cloned in A3A open reading frame (in orange) or in A3Alt +1 reading frame (in blue) to assess the differential translation initiation from start codons, S1, S2, S3, and S4. Start codons invalidated by site-directed mutagenesis are crossed out in black. D, firefly luciferase expression from S1, S2, S3, and S4 initiation codons, normalized on Renilla luciferase activity in HeLa cells 36 h after transfection. Error bars represent standard deviation from three independent transfections, measured in duplicates. Differences compared with S1 + S2 + Luc are represented in orange, and differences compared with S3 + S4 + are represented in blue. Differences were calculated using one-way ANOVA multiple comparison with Tukey post hoc test (ns: nonsignificant, **p < 0.002, ****p < 0.0001). E, Western-blot analysis of V5-tagged A3A and A3Alt proteins in HeLa cells 24 h post transfection. β-Actin was used as loading control. Presented Western blot is representative of three independent experiments. F, Western blot analysis of A3Alt expression from A3A transcript in HeLa cells 24 h after A3Alt natural plasmid transfection where V5 tag is cloned in +1 reading frame at the end of A3Alt orf, in a plasmid containing A3A natural 5' sequence. DR: Densitometric ratio (DR) representing A3Alt natural expression relative to A3Alt-L expression after normalization on β-Actin. Presented results correspond to two independent transfections.
codons, S4 appears to be dominant. This indicates that both A3Alt proteins, A3Alt-L and A3Alt-S, can be expressed from A3A mRNA. In vivo, expression may be more complex, involving mRNA secondary structures as well as cis regulation by UTRs.
To explore the functionality of A3Alt proteins, A3Alt-L and A3Alt-S coding sequences were cloned in a V5-tag expression vector and transfected along with A3A-L and A3A-S cytidine deaminases expression plasmids in HeLa cells (schematic representation of V5 tagged constructs with Kozak context and stop codon positions is available in Fig. S2). In keeping with previous studies, A3A cytidine deaminase coding sequence driven by its natural "adequate" Kozak sequence (A3A-L adequate) resulted in the translation of the two functionally active isoforms A3A-L and A3A-S, while only A3A-L was detectable when cloned downstream of the strong Kozak initiation context of pcDNA3.1 expression plasmid (A3A-L strong) (Fig. 1E). Accordingly, the A3A-S coding plasmid, only containing the coding sequence of the shorter A3A-S isoform, allowed the detection of A3A-S isoform independently of the AUG Kozak context (A3A-S adequate or A3A-S strong) (Fig. 1E).
Transfection of A3Alt-L and A3Alt-S coding plasmids revealed suitable expression of both proteins, although A3Alt proteins levels were consistently lower than those obtained with A3A-L and A3A-S coding plasmids (Fig. 1E). Unexpectedly, although A3Alt-L translation in A3Alt-L expression plasmid was driven by a strong initiation codon, Western blot detection after A3Alt-L transfection revealed the unexpected expression of a shorter isoform migrating as A3Alt-S, confirming the finding obtained with luciferase reporter constructs that A3Alt-S S4 AUG is associated with a strong translational activity (Fig. 1D). Accordingly, S4 initiation codon was removed by mutagenesis from A3Alt-L expression plasmid to generate the A3A-L ΔS4 construct. Western blot probing of A3Alt-L ΔS4 transfection only revealed a single band at A3Alt-L expected size and therefore allowed the analysis of "full-length" A3Alt-L protein for functional study.
To validate A3Alt-L translation from A3A transcripts (Fig. 1D), an "A3Alt natural" construct was also generated corresponding to the A3A adequate construct, driven by its natural Kozak initiation context with terminal V5 tag cloned in the +1-shifted reading frame at the end of A3Alt coding sequence. Consistently with luciferase reporter transfection experiments, V5 probing confirmed A3Alt proteins expression from A3A transcript, albeit with reduced expression (25%) compared with A3Alt-L expression plasmid transfection (Fig. 1F).

A3Alt and A3Alt-S are mitochondrial proteins
To help determine the functionality of those proteins, in silico analysis was conducted. Topology prediction was first performed using PSIPRED program (41) and both sequences were predicted to contain a central 21 amino acid transmembrane domain (Fig. 1, A and B in blue and Fig. S3A in pink). Subcellular localization prediction was next conducted using DeepLoc deep learning algorithms (42), locating A3Alt proteins to the mitochondria (Fig. S3B). Mitochondrial addressing was further explored by motif search revealing that the A3Alt N-terminus may act as a mitochondrial targeting peptide, potentially cleaved by Mitochondrial Processing Peptidase (MPP) through an R3a/R3b mitochondrial cleavage site motif (RIFS) identified using TPpred2 software (Fig. S3C) (43,44). The identification of this putative MPP cleavage site that is present in both A3Alt-L and A3Alt-S proteins may be responsible for the smaller peptide detection evidenced in Western blot after A3Alt plasmids transfections (Fig. 1E, annotated with asterisks), potentially corresponding to the mature form of A3Alt proteins. Ultimately, this suggests that after cleavage both proteins may generate the same mature peptide and therefore exhibit the same functionality.
To experimentally address mitochondrial localization of A3Alt proteins, HeLa cells were transfected with A3A and A3Alt expression plasmids and analyzed by confocal microscopy using TOM20 (Translocase of Outer Membrane protein 20) staining as mitochondrial marker ( Fig. 2A). As previously reported, both A3A-L and A3A-S exhibited diffuse nucleocytoplasmic distribution, with no colocalization with mitochondrial compartment (12). By contrast, A3Alt-L, A3Alt-S, and full-length A3Alt-L ΔS4 consistently demonstrated a strong mitochondrial addressing, evidenced by TOM20 colocalization (Figs. 2A and S4 for lower magnification). Subcellular localization was further validated by cryo-electron microscopy and immunogold staining of Flag tagged A3Alt constructs, confirming mitochondrial addressing (Figs. 2B and S5 for Flag tagged A3Alt protein expression analysis). The transmembrane domain is somewhat unusual in that it encodes an arginine residue (asterisk Fig. 1B). However, alanine substitution in A3Alt-L R48A mutant construct did not alter addressing to the mitochondrial network (Figs. S6 and S4).
Subcellular fractionation was performed in transfected HeLa cells, confirming the localization of A3Alt proteins in apoptosis-inducing factor (AIF) positive mitochondrial fraction (Fig. 2C). In keeping with this, no V5 detection was found in the MEK1/2 positive cytosolic fraction, or the histone H3 nuclear fraction, confirming the specificity of A3Alt mitochondrial localization (Fig. 2C).
To further determine the submitochondrial compartmentalization of A3Alt proteins, crude mitochondria isolated from A3Alt-L transfected HeLa cells were isolated and submitted to proteinase K digestion with or without hypotonic treatment used to disrupt the mitochondrial outer membranes. While intact mitochondria treatment with proteinase K treatment resulted in complete TOM20 digestion, A3Alt-L protein detection using V5 probing was not impacted, suggesting that A3Alt proteins are not exposed on the outer face of the outer membrane the of mitochondria ( Fig. 2D, left panel). Following hypotonic treatment, AIF protein localized in the intermembrane space was further degraded while the matrix associated protein TFAM was preserved from proteinase K digestion. In hypotonic condition, A3Alt-L detection was lost following proteinase K digestion suggesting that A3Alt proteins are located to the intermembrane space of transfected cells (Fig. 2D, right panel). The efficiency of proteinase K digestion was validated pretreating mitochondria with Triton X-100 detergent, resulting in the total loss of TOM20, A3Alt-L, AIF, and TFAM detection (Fig. 2D).

A3Alt proteins induce mitochondrial outer membrane depolarization and apoptosis
Mitochondria play a central role in the life and death of eukaryotic cells. We thus investigated whether A3Alt expression had an effect on mitochondrial respiration and ATP production. Therefore, HeLa cells cultivated in galactose supplemented glucose-free media were transfected with A3Alt expression plasmids and ATP production was evaluated measuring luciferase activity. Quantitative luminescence revealed that none of A3Alt construct significantly impacted ATP production compared with mock transfected cells (Fig. 3A). Treatment with carbonyl cyanide 3-chlorophenylhydrazone (CCCP), a mitochondrial oxidative phosphorylation uncoupler, was used as positive control and resulted a significative drop in luciferase activity compared with DMSO-treated HeLa cells (Fig. 3A). As A3A cytidine deaminases genotoxicity has recently been associated with reactive oxygen species (ROS) accumulation (45) in an experimental setting compatible with A3Alt proteins expression, we next explored whether they could dysregulate ROS production at the mitochondrial level. Accordingly, A3A and A3Alt expression plasmids were transfected in HeLa cells and incubated with 2 0 ,7 0 -dichlorodihydrofluorescein diacetate (DCFDA) dye, commonly used to measure ROS production as its fluorescence is proportional to DCFDA oxidation. None of the constructs were associated with significant ROS accumulation, which was only found using tertiary-butyl hydroperoxide TBHP oxidative stress inducer (Figs. 3B, and S7). However, although A3A expression been previously described to induce ROS production (45), only a modest increase of A3A induced ROS production (8%) that did not score as significant compared with mock-transfected cells was observed in our experimental setup, which was not recapitulated with A3A C106S catalytic mutant (2.4%) (Figs. 3B and S7).
We next sought to investigate if A3Alt proteins could impact mitochondrial outer membrane potential, which can reflect the initiating event of mitochondrial outer membrane permeabilization (MOMP) and intrinsic mitochondrial apoptosis pathway. The loss of inner mitochondrial membrane potential (Δψm) was evaluated using MitoStatus Red specific dye in transfected HeLa cells. Since mitochondrial membrane potential staining is not compatible with fixation and permeabilization, functional C-terminally GFP-tagged constructs APOBEC3A mRNA encodes new proapoptotic proteins were generated to allow gating on A3A and A3Alt expressing cells (Fig. S8). If A3A-GFP and A3A C106S -GFP failed to modify mitochondrial potential compared with GFP control transfection, A3Alt-L-GFP, A3Alt-S-GFP, and A3Alt-L ΔS4-GFP consistently elicited mitochondrial membrane potential depolarization, arguing for mitochondrial outer membrane permeabilization (Figs. 3C and S9). Cell death was therefore assessed using Annexin V probing and viability dye staining to detect early and late apoptosis on transfected V5 positive cells. Etoposide-treated HeLa cells were used as positive apoptosis staining control (Fig. S10). As previously described, A3A transfection induced significative levels of apoptosis (10%) compared with the catalytically inactive A3A C106S mutant (4%) (Fig. 3D). Strikingly A3Alt-L, A3Alt-S as well as A3Alt-L ΔS4 transfections resulted in strong and consistent cellular death (22%) (Figs. 3D and S10). In keeping with what was inferred from in silico analysis, all A3Alt proteins exhibited analogous proapoptotic activity. However, cytochrome C release analysis by Western blot probing of cytosolic extracts failed to evidence cytochrome C protein in the cytoplasm of A3A and A3Alt transfected cells (Fig. 3E). In keeping with this observation, caspase activation pathway analysis using Western blot probing of cleaved PARP and cleaved caspase 3, 7, and 9 after A3A and A3Alt transfection failed to demonstrate a clear activation of caspasedependent apoptosis pathway beyond the levels observed APOBEC3A mRNA encodes new proapoptotic proteins with mock-transfected cells (Fig. 3F) as it was previously observed following A3A expression (16). This finding indicates that although A3Alt expression perturbs inner mitochondrial membrane potential (Fig. 3C), it might not result in outer membrane permeabilization associated with cytochrome C release and canonical caspase activation pathway. Accordingly, A3Alt-L transfected HeLa cells treatment with Z-VAD-FKM pan-caspase inhibitor did not significantly reduce A3Altinduced apoptosis (Fig. S11), while etoposide-induced cell death was totally abrogated (Fig. S11). Altogether, these data indicate that A3Alt proteins are addressed to the mitochondrial network and elicit cellular death, in a caspaseindependent process.

Discussion
In the present work, we report that human A3A gene encodes two distinct proteins. A3A cytidine deaminase, which can exist as two functional isoforms A3A-L and A3A-S (16), is a highly versatile enzyme involved in innate immune responses against many viruses (2-8) and retroelements (9-11), as well as in nucleic acid catabolism to evade self-DNA sponsored inflammation (25). However, the price to pay for this mutator enzyme activity is off-target editing associated with somatic mutations and chromosomal rearrangements in many cancer genomes (14,21,23,46,47). To date, no deletion or inactivation allele has been reported for A3A unlike other APOBEC3 family members (28,48). Accordingly, understanding the regulation of A3A expression is now of major concern, as its expression has also been reported to exert additional effects on cellular physiology, causing cell cycle arrest, apoptosis, as well as tumorigenic ROS production (15,16,45). In trying to understand the observed discrepancy between A3A mRNA levels after stimulation in myeloid cells and A3A detection at the protein level (25), we identified an out-offrame overlapping reading frame in A3A transcript, predicted to encode novel small alternative A3Alt proteins, A3Alt-L, and the shorter isoform A3Alt-S. If we failed to generate A3Alt specific antibodies immunizing rabbits with peptides corresponding to A3Alt-L N-terminal et C-terminal fragments, which would have allowed in vivo probing of A3Alt peptides in cells such as THP1 or isolated PBMCs known to express high level of A3A mRNA upon stimulation, the biological relevance of A3Alt is grounded in several observations. Analysis of orthologous A3A genes in primates revealed that the atypical organization AUG initiation codons in human A3A gene is shared among great apes, with A3A cytidine deaminases expression driven by "weak" AUGs while A3Alt orfs are present in a favored expression Kozak context (Fig. S13). Thus, the A3Alt reading frame is conserved in many primates, arguing for functionality. A3Alt expression in vivo is also further supported by mass spectrometry analysis, identifying A3Alt-L-derived peptide MAL-EGIRPTCATKWSAWTMAPR in the Bioplex 2.0 wide scale interactome study (49), which corresponds to the 22 N-terminal residues of A3Alt-L (Fig. S1B). This singular configuration, a single gene encoding two distinct proteins is not unprecedented with a handful of notable examples such as ATXN1 (39), GNAS complex locus (36,37), prion-related PRNP (38) as well as INK4α tumor suppressor gene (35).

A B C D
In silico analysis validated by functional study revealed that those new A3Alt proteins were targeted to the mitochondrial compartment, resulting in inner mitochondrial membrane potential depolarization as well as apoptosis. However, if Annexin V staining was demonstrated in V5 transfected cells using flow cytometry, we failed to demonstrate global cytochrome C release as well as caspase activation upon A3Alt coding plasmid transfection. This may suggest that either the cell death pathway involved is independent of canonical apoptosis program or that our experimental setup, relying on plasmid transfection, may be suboptimal to study into details the mechanism involved in A3Alt-mediated cell death. Unfortunately, every attempt to generate stable cell lines expressing A3Alt, either by lentiviral transduction or plasmidbased expression system (Tetracycline-Regulated Expression T-Rex System, Thermo Fisher Scientific), proved unsuccessful, probably given the proapoptotic nature of our proteins. The same is true for high-level inducible A3A expressing cell lines.
Be that as it may, the presented results culminate in a singular situation where two proapoptotic proteins, A3A cytidine deaminases targeting the genome, and A3Alt proteins targeting the mitochondrion, are encoded by a single gene. As A3A genotoxicity is reported to mainly affect replicating cells where nuclear genome is exposed as ssDNA and vulnerable to A3A mutational activity (50), A3Alt-mediated apoptosis may represent an alternative way to induce cellular death in overstimulated quiescent cells.
Interestingly, this study also tightens the emerging links between A3A expression and mitochondrial network homeostasis, as A3A has recently been reported to be central in cytoplasmic mitochondrial DNA (mtDNA) catabolism (25,26), to prevent danger signal accumulation and inflammation (51). The impact of A3Alt on mtDNA release into the cytoplasm will therefore need to be evaluated to better understand the contribution of A3A and A3Alt in balancing the cytoplasmic mtDNA level in cells following cellular stress. One can imagine a scenario in which modest stimulation will only result in A3A cytidine deaminase expression and agonist mtDNA clearance (25,26), while excessive stress may result in A3A and A3Alt expression, where A3Alt sponsored mitochondrial alterations may fuel mtDNA release to the cytosol and in return overstimulate A3A gene expression, leading to cell death. This observation emphasizes the complexity of A3A gene regulation, which stands at the crossroads of prosurvival interferon signaling pathways, and two distinct mechanisms leading to cell death.
In conclusion, it is intriguing that among the few examples of "dual-coding genes" A3A, an innate immunity related gene shaped by a long-standing coevolution process with pathogens, appears to have appropriated a mechanism that is common for many viruses and bacteria. More intriguingly, some proteins from overlapping reading frames from Chicken Anemia Virus, Influenza A Virus, and some Norovirus are involved in pathogenicity by targeting the mitochondria, culminating in cell death (52)(53)(54).

Cells
Human HeLa cells were maintained in DMEM glutamax medium (Thermo Fisher Scientific) supplemented with 10% FCS, 50 U/ml penicillin, and 50 mg/ml streptomycin.

Luciferase activity
For luciferase activity assay, 8 × 10 5 of HeLa cells grown in six wells plates were cotransfected with 1.8 μg of firefly luciferase constructs along with 0.2 μg of Renilla luciferase control plasmid (Promega) using Fugene HD (Roche) following manufacturer's recommendations. Transfected cells were split in 96-well at a density of 5 × 10 4 cells/well, and luciferase activity was measured 24 h later using Dual-Glo luciferase assay (Promega) with 30 min incubation times. Presented results represent luciferase activity from three independent transfections measured in triplicates.

Subcellular fractionation
At 24 h, six million HeLa transfected cells were recovered and washed in cold PBS. One million cells were collected and lysed in RIPA lysis buffer for western blotting the whole-cell extract (WCE). Subcellular fraction was then performed on the rest of the cells adapting a previously published protocol (55). Briefly, HeLa cells were resuspended in digitonin buffer (150 mM NaCl, 50 mM HEPES pH7.4 and 100 μg/ml of digitonin) for 10 min under gentle agitation. Lysates were then centrifuged for 10 min at 2000g to pellet nuclei and organelles. Supernatants were centrifuged three times at 20,000g for 20 min to purify cytosolic fraction. Nuclei and organelles pellets were washed twice in cold PBS and resuspended in NP40 lysis Buffer (150 mM NaCl, 50 mM HEPES pH7.4, 1% NP-40, protease, and phosphatase inhibitors) and incubated for 30 min on ice. Lysates were then centrifuged for 10 min at 7000g. Supernatants corresponding to the mitochondrial fraction were collected. Pellets were washed twice in PBS, resuspended in RIPA lysis buffer, and sonicated.

Subcellular fractionation and proteinase K treatment
Transfected HeLa cells were collected, washed two times in PBS, and washed one time in isolation buffer (10 mM HEPES-KOH, pH 7.4, 0.25 M sucrose, and 1 mM EDTA). Cells were then homogenized on ice using Potter-Elvehjem homogenizer (20 strokes) and centrifuged at 900g for 10 min to pellet nuclear fraction and intact cells. The supernatant was collected and centrifuged at 10,000g for 10 min. The second pellet corresponding to the crude mitochondrial fraction was resuspended either in isolation buffer or hypotonic buffer (10 mM HEPES-KOH, pH 7.4, and 1 mM EDTA) and incubated on ice for 10 min to allow mitochondrial outer membranes disruption. Mitochondria were then digested with proteinase K at a final concentration of 100 μg/ml, with or without 1% Triton X-100 on ice for 15 min. Reaction stopped by adding the protease inhibitor PMSF at a final concentration of 1 mM for 15 min on ice, and samples were resuspended in reducing 4× LDS sample buffer (ThermoFisher Scientific).

ATP production analysis
HeLa cells were cultivated in glucose-free DMEM medium (Thermo Fisher Scientific) supplemented with 10% FCS, 50 U/ ml penicillin, 50 mg/ml streptomycin, and 10 mM galactose (Sigma). Plasmid transfection was performed as previously described and culture media was replaced with serum-free glucose-free DMEM medium 6 h after transfection. Transfected cells were harvested 24 h after transfection and split in 96-well plate at a density of 5 × 10 4 cells/well, and luciferase activity was measured 24 h later using Mitochondrial ToxGlo Assay (Promega) following manufacturer's recommendations. Carbonyl cyanide 3-chlorophenylhydrazone (CCCP) treatment at 20 μM for 90 min was used as positive control.

FACS analysis ROS production
At 24 h post transfection, HeLa cells were incubated with 20 μM of the cell permeant 2 0 ,7 0 -dichlorofluorescine diacetate (DCFDA) reagent for 1 h at 37 C, following manufacturer's recommendations (DCFDA Cellular ROS Detection Assay Kit, Abcam). Cells were the washed twice in PBS, harvested, and analyzed by flowcytometry. Oxidative stress inducer tertiary-butyl hydroperoxide (TBHP) treatment was performed at a concentration of 100 μM for 30 min before DCFDA staining as positive control. Samples were analyzed on a MACSQuant Analyzer (Miltenyi Biotec) and data were processed with FlowJo software (Tree Star Inc, version 8.7.1).

FACS analysis of inner mitochondrial membrane potential (Δψm)
Twenty-four hours posttransfection, the loss of inner mitochondrial membrane potential (Δψm) was evaluated in A3A-GFP transfected HeLa cells, using MitoStatus red (BD Pharmingen) Δψm-sensitive fluorescent dye. 70% confluent transfected HeLa cells were incubated with 100 nM MitoStatus Red in cell media for 20 min in the dark. FCCP (Carbonyl cyanide 4-(trifluoromethoxy)-phenylhydrazone) a mitochondrial oxidative phosphorylation uncoupler was used a positive control and incubated at a final concentration of 10 μM in culture media for 10 min at 37 C before staining. After staining, cells were washed two times with PBS and harvested using trypsin. Cell pellet were washed two times in PBS and analyzed on a MACSQuant Analyzer (Miltenyi Biotec) and data were processed with FlowJo software (Tree Star Inc, version 8.7.1).

FACS analysis of apoptosis
Transfected HeLa cells were harvested, counted, washed twice in PBS and 1 × 10 6 cells were stained for 30 min with 1 μl 780eFluor of Fixable Viability Dye (eBioscience) in 1 ml PBS. For Z-VAD-FKM experiments, culture media was removed 6 h after transfection and replaced with 20 μM Z-VAD-FKM (Sigma) containing complete media. After PBS washing, cells were stained with 10 μl Annexin V-V450 (BD Biosciences, 560506) in 200 μl 1× binding buffer for 15 min.

APOBEC3A mRNA encodes new proapoptotic proteins
After washing, cells were fixed in 2% ice-cold paraformaldehyde (Electron Microscopy Sciences) for 15 min and permeabilized in 90% ice-cold methanol (Sigma-Aldrich) for 30 min. V5 staining was then performed using mouse anti-V5 Alexa-Fluor 647 conjugated antibody (clone SV5-Pk1, Biorad) and stained samples were acquired on a MACSQuant Analyzer (Miltenyi Biotec). Data were analyzed with FlowJo software (Tree Star Inc). Treatment with 100 μM etoposide (Sigma) in dimethylsulfoxide for 12 h was used as positive control.

Data availability
Data sharing is not applicable to this article as no data libraries were generated. The communication author will accommodate requests of relevant materials.
Supporting information-This article contains supporting information.
Funding and additional information-P. K. was supported by La Ligue contre le Cancer. The funding bodies had no role in the design of the study or collection, analysis, and interpretation of the data, or the writing of the manuscript.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.