Elephant APOBEC3A cytidine deaminase induces massive double-stranded DNA breaks and apoptosis

The incidence of developing cancer should increase with the body mass, yet is not the case, a conundrum referred to as Peto’s paradox. Elephants have a lower incidence of cancer suggesting that these animals have probably evolved different ways to protect themselves against the disease. The paradox is worth revisiting with the realization that most mammals encode an endogenous APOBEC3 cytidine deaminase capable of mutating single stranded DNA. Indeed, the mutagenic activity of some APOBEC3 enzymes has been shown to introduce somatic mutations into genomic DNA. These enzymes are now recognized as causal agent responsible for the accumulation of CG- > TA transitions and DNA breaks leading to chromosomal rearrangements in human cancer genomes. Here, we identified an elephant A3Z1 gene, related to human APOBEC3A and showed that it could efficiently deaminate cytidine, 5-methylcytidine and produce DNA breaks leading to massive apoptosis, similar to other mammalian APOBEC3A enzymes where body mass varies by up to four orders of magnitude. Consequently, it could be considered that eAZ1 might contribute to cancer in elephants in a manner similar to their proposed role in humans. If so, eAZ1 might be particularly well regulated to counter Peto’s paradox.

Interestingly, this functional attenuation was also observed for the rhesus monkey rhA3B enzyme compared to rhA3A indicating that this mutagenic dichotomy was maintained for ~38 million years 6 . Moreover, the deletion of most of the A3B gene results in a higher odds ratio of developing breast, ovarian or liver cancer [20][21][22][23] . Indeed, complete genome sequencing of ΔA3B -/breast cancer genomes revealed a higher mutation burden 24 . Finally, fine analysis of signatures mutations in cancer genomes unraveled for twice as many A3A specific mutational signature (YTCA) over A3B (RTCA) suggesting a major role of A3A in cancer mutagenesis 8,25 .
Another difference between A3A and A3B lies in their evolutionary history. A3A is present across most placental mammals, indicating that this evolutionary experiment has been running ~150 million years 26 . There are some notable exceptions -an A3A gene is absent among all members of the order Rodentia, pigs, while for Felidae the gene is inactivated but identifiable 3,26 . By contrast A3B is unique to the order Primates and arose by gene conversion involving A3A. We have previously shown that the A3A enzymes from 8 mammalian species from rabbits to cows and horses were capable of deaminating C and 5MeC in ssDNA as well as producing DSBs, even though activities varied considerably 26 .
The incidence of developing cancer was hypothesized to increase with the body size, referred to as Peto's paradox 27 . However, as large animals exist and do not invariably die of cancer this paradox fails to explain the presence of compensatory mechanisms that protect the genome. With this in mind, we were intrigued by a recent report showing that elephants appeared to have a lower-than-expected rate of cancer which might possibly be coupled to multiple copies of TP53 even though most were processed pseudogenes 28,29 . It is equally possible that the A3 enzymes of large mammals could have been attenuated by mutation. Accordingly, we decided to explore the function of the elephant A3Z1 enzyme.

Results and Discussion
Synthesis and expression of elephant APOBEC3Z1 cytidine deaminase. To explore the implication of elephant A3 enzyme in tumorigenesis, in silico data mining was performed using blast/blat analyses of the genomic for A3Z1 like sequences. We retrieved an elephant A3Z1 sequence named eA3Z1, equivalent to the p2 protein form of human A3A (hA3A), that is to say missing the first coding exon, what is also a feature of dog, horse and cow A3A genes. Furthermore, p1 and p2 forms of A3A are functionally equivalent 26,30,31 . There was 44% amino acid divergence between the human and elephant protein sequences (Fig. 1). The elephant sequence carried an 8 residue deletion in loop 3, which is not without precedent 26 and impacts little the overall structure as can be seen from Fig. 2a. All the key amino acid residues typical of an A3A enzyme were conserved 3 . A phylogenetic analysis using the neighbor-joining method revealed that eA3Z1 was closely related to those from the Primate lineage (Fig. 2b). To prove that eA3Z1 is expressed in vivo and to validate the putative eA3Z1 DNA sequence inferred from elephant genome assembly, total RNA from the liver of an African Savana elephant (Loxodonta africana) that had died from an encephalomyocarditis virus infection 32 was analyzed. Total RNA was extracted, cDNA generated and sequences corresponding to complete eA3Z1 transcripts were amplified by a semi-nested PCR procedure. As shown in Fig. 2c, strong eA3Z1 cDNA amplifications were obtained giving rise to two overlapping PCR products. Finally, eA3Z1 sequence (accession number: MK156802) was validated by direct sequencing ( Supplementary  Fig. S1) and identical to the previous BLAST/BLAT analysis search sequences from genomic data.
Accordingly, an eA3Z1 cDNA was synthesized and cloned into pcDNA3.1D/V5-His-TOPO with a strong Kozak motif (ACCATG) for functional studies. When overexpressed in transfected HeLa or HEK-293T cells, Western blot analysis revealed a strong expression of V5-tagged eA3Z1 on a par with that for its human counterpart (hA3A) or an inactive mutant (hA3A C101S ; Fig. 2d). The slightly lower molecular weight of eA3Z1 is in agreement with the calculated molecular weights (hA3A 23.0 kDa; eA3Z1 20.9 kDa). The subcellular localization was assessed in HeLa cells by confocal microscopy. Anti-V5 staining revealed that eA3Z1 exhibited the classical nuclear and cytoplasmic distribution described for human (Fig. 2e) as well as other mammalian A3A enzymes 26 . Elephant APOBEC3Z1 editing of nuclear DNA and 5-methylcytidine. To assay catalytic activity, eA3Z1 and hA3A plasmids were transfected into HEK-293T cells and cellular lysates were used in a fluorescence resonance energy transfer assay based in vitro deamination assay where C to U conversion in a TAMRA-FAM-labeled DNA oligonucleotide allows fluorescence detection following cleavage by uracil-DNA glycosylase (UNG) activity 7 . Elephant eA3Z1 activity was on a par with hA3A (Fig. 3a). To explore A3 hyperediting of chromosomal DNA, the HEK-293T-UGI cell line was transfected with eA3Z1 and hA3A plasmids. The HEK-293T-UGI cell line constitutively expresses uracil N-glycosylase (UNG) inhibitor (UGI) where UNG is the crucial enzyme involved in excising uracil from DNA. As UNG is rate limiting for the detection of hyperedited chromosomal DNA, inhibition by UGI is necessary 9 . At 48 hours post-transfection, total DNA was extracted and TP53 DNA was amplified by 3D-PCR, a technique that selectively amplifies A3-edited ssDNA molecules 33 . The lowest PCR denaturation temperature (Td) allowing amplification of unedited TP53 target DNA was 87°C (Fig. 3b). For eA3Z1 and hA3A transfections, 3D-PCR products were recovered at Tds as low as 84.1°C and 84.9°C respectively which is diagnostic for A3 editing (Fig. 3b). Nonetheless, 3D-PCR products from the 85.9°C amplification (white asterisk, Fig. 3b) were cloned and sequenced as those of the last positive amplification for plasmid vector (pv) and hA3A C101S as negative controls. Hyperedited TP53 target sequences were recovered with an average editing frequency of 10% compared to a background value of 0.6% (Fig. 3c). The monotony of editing is confirmed by the frequency of non-CG-> TA mutations which did not differ from background values. Cytidine editing was strongly associated with TpC, and to a lesser extent CpC dinucleotides (Fig. 3d) to the detriment of GpC and ApC which is typical for mammalian A3A enzymes 26,[34][35][36][37] .
To demonstrate that no endogenous activity of hA3A present in HEK-293T-UGI cells would give rise to hypermutated sequences, hA3A or eA3Z1 transfections were performed in QT6 cell lines in presence of UGI. QT6 is a quail cell line and was chosen as there is no endogenous A3 background 38 . As expected, 3D-PCR product sequence analysis demonstrated the same profile of hypermutated sequences and dinucleotide contexts (data not shown). One of the singular traits of mammalian A3A deaminases is their ability to efficiently deaminate 5MeC 26 . To demonstrate that eA3Z1 can deaminate 5MeC, QT6 cells were transfected with the eA3Z1 expression plasmid and subsequently transfected by 5MeC-substituted HIV env DNA fragments 16 . As shown in Fig. 4a, 3D-PCR products were recovered at temperatures as low as 75.7°C and 77 °C for the eA3Z1 and hA3A transfections respectively, compared to 82.1°C for the plasmid vector or hA3A C101S (Fig. 4a). The 82.1°C and 80.3°C 3D-PCR products (Fig. 4a), indicated by an asterisk were cloned, sequenced and confirmed the presence of edited 5MeC in the expected TpC dinucleotide context (Fig. 4b,c). Hyperedited 5MeC V1V2 target sequences were recovered with an average editing frequency of 5% and 6.8% respectively for hA3A and eA3Z1 compared to background value (Fig. 4c).
Elephant APOBEC3Z1 induces double strand DNA breaks and apoptosis. Human A3A editing of chromosomal DNA results in the formation of DSBs and can be readily scored by analysis of histone variant H2AX phosphorylation at serine 139 (γH2AX), a well-known marker for DSBs and the DNA damage response 39 . HeLa and QT6 cell lines were transfected with eA3Z1 and hA3A constructs ±UGI using plasmid vector as negative control. As can be seen in Fig. 5a, eA3Z1 and hA3A generated DSBs in HeLa cells ∼35 and ∼20-fold over background. While DSBs were more pronounced in QT6 with ∼70 and ∼40-fold higher over plasmid control with eA3Z1 being the more active of the two enzymes (Fig. 5b). In the presence of UGI, a decrease in A3-induced DSBs was noted for hA3A and eA3Z1 transfected cells indicating that UNG plays an important role in the formation of DSBs upon DNA editing. As A3-induced DSBs lead to apoptosis, we measured cytochrome c release (Fig. 5c). Transfection of equal amounts of plasmid DNA resulted in greater cytochrome c release for eA3Z1 using Annexin V and propidium iodide as markers for apoptosis (Fig. 5c,d). These data demonstrate that eA3Z1 is a strong editor of ssDNA and can induce DSB leading to apoptosis.
Taken together, these data demonstrate that eA3Z1 clearly exhibit an enzymatic activity similar to human and other mammalian A3A cytidine deaminases since eA3Z1 edits both C and 5MeC residues in ssDNA and can make DSBs leading to apoptosis. A side-by-side comparison of eA3Z1 to 8 mammalian A3As is shown in Fig. 5e. All constructs are well expressed by Western blotting and immunofluorescence with the signal exception of the tamarin construct 26 . While eA3Z1 is well ranked among the series, it is less efficient than the cow and horse constructs which, although they have large body mass and longevity (wild cattle 18-25 + years, up to 900 kg; horses 30-40 years, up to 600 kg), are not comparable to the elephant (median 56 years, up to 70 years, up to 7000 kg). Obviously, there are many variables that can alter A3 function: mi-and lncRNAs and transcription factor sites in the promoter as well as the role of negative interactors like TRIB3 34,40 which is part of the broad CtIP-Rb-BRCA1-ATM protein network that involves cell cycle control, cell survival, DNA repair, and genome stability. The eA3Z1 described here is clearly comparable to those of many large mammals, being able to damage chromosomal DNA and might therefore contribute to oncogenesis. If so, perhaps the A3s of elephants must be tightly regulated to lower the incidence of cancer.
For the 5MeC deamination assay, a 679 bp fragment of HIV-I LAI env gene was amplified using total substitution of dCTP by 5Me-dCTP (Trilink) using the primer pair MC1, 5′TTGATGATCTGTAGTGCTACAGCA and MC2, 5′GCCTAATTCCATGTGTACATTGTA. The 5MeC containing DNA was heat denatured and chilled on ice and 200 ng of synthesized DNA was transfected using jetPRIME 24 hours following initial transfection of A3 coding plasmids in QT6 cells as described earlier 16  Immunofluorescence. Approximately 50,000 HeLa cells were seeded in Nunc ™ Lab-Tek ™ II Chamber Slide ™ System Thermo Scientific ™ and transfected 24 hours later with 1 µg of plasmid DNA according to the Fugene ® protocol. Two days after transfection, coverslip grown transfected HeLa cells were washed three times with PBS and fixed with 4% paraformaldehyde (Electron Microscopy Sciences) for 15 min. Cells were then washed two times and permeabilized with a 50% methanol/acetone mix for 10 min. After two PBS washings, permeabilized cells were incubated for 1 hour at room temperature, first with 0.5% bovine serum albumin (BSA) PBS 1/200 mouse monoclonal anti-V5 antibody (Invitrogen) and then with 0.5% bovine serum albumin PBS 1/500 anti mouse Alexa Fluor 488 conjugated antibody (ThermoFisher). After several PBS washings, coverslips were mounted with Vectashield mounting medium for immunofluorescence (Interchim). Imaging was performed using a Leica SP5 confocal microscope. For apoptosis, transfected HeLa cells were collected and washed with PBS, then incubated with complete DMEM medium at 37 °C for 2 hours. After washing with cold PBS, cells were resuspended in 1X Binding Buffer (BD Pharmingen) and then counterstained with 1 μg/ml FITC Annexin V antibody (BD Pharmingen) and 5 μg/ml Propidium Iodide (PI) (BD Pharmingen) to distinguish between early apoptotic and late apoptotic or necrotic events. Treatment by 100 mM etoposide in dimethylsulfoxide was used as positive control. The labelled samples were analyzed on a MACSQuant ® analyzer harboring violet, blue, and either a red laser (measure of dsDNA breaks and apoptosis). The data were analyzed using the FlowJo ® software (Tree Star Inc., version 10.1r5 for Mac). Western blotting. Cells were recovered 48 hours after transfection. Protein extraction and Western blot analysis were carried out according to standard procedures. After blocking, membranes were probed with either a 1:5000 dilution of anti V5-tag horseradish peroxidase-coupled antibody (Invitrogen), or a 1:15000 dilution of anti β-actin (Sigma). The membrane was subjected to detection by SuperSignal ™ West Pico chemiluminescent substrate (ThermoFisher Scientific).