The deaminase ADAL-pivoted catabolism checkpoint suppresses aberrant DNA N6-methyladenine incorporation


 Abundant RNA N6-methyladenine (m6A) is degraded in RNA decay and potentially induces aberrant DNA N6-methyladenine (6mA) misincorporation. Biophysically, like truly methylated product DNA 6mA, misincorporated 6mA also destabilizes the DNA double helix and thus ditto affects DNA replication and transcription. By heavy stable isotope tracing, we demonstrate that intracellular degradation of RNA m6A cannot induce any misincorporated DNA 6mA, unveiling the existence of a catabolism checkpoint that blocks DNA 6mA misincorporation. We further show that the deaminase ADAL preferentially catabolizes N6-methyl-2’-deoxyadenosine monophosphate (6mdAMP) in vitro and in vivo, and adenylate kinase 1 restricts the phosphorylation rate of 6mdAMP, together contributing to the identified checkpoint. Noteworthy, low ADAL expression reduces dramatically the patient survival in four cancers. Collectively, our data strongly support a pivotal role of ADAL in the suppression of 6mA misincorporation and implicate that both ADAL and misincorporated 6mA may mark cancer abnormalities.

Meanwhile, epigenetic mRNA N 6 -methyladenosine (m6A), which shares the same N 6 -methyladenine base with DNA 6mA, is one of the most abundant internal posttranscriptional modi cations in mammals. mRNA m6A is highly enriched in 3 untranslated regions, stop codon anking regions, and long internal exons of mRNA [25][26][27] and regulates transcription and processing in the nucleus and translation and decay in the cytoplasm [28][29][30][31] . Notably, as a prevalent activity RNA degrades, and the half-life of mRNA is generally as short as minutes to hours [32][33][34][35] . Introns and spacer sequences in mRNA during processing; defective mRNA fragments during transcription, processing, and functioning; and scrapped mature mRNAs are degraded as signaling by a surveillance system [36][37][38] . Along with mRNA degradation, unmodi ed adenosine phosphates are released into nucleotide pools and then partly converted into reusable 2'-deoxyadenosine triphosphate (dATP) through purine salvaging 39 . Similarly, substantial m6A-related species (e.g., N 6 -methyladenosine monophosphate (m6AMP) 40 ) are released to the nucleotide pools. It is expected that following purine salvage, the released m6A-related species might be ultimately converted into N 6 -methyl-2'deoxyadenosine triphosphate (6mdATP), which can be misincorporated into genomic DNA by DNA polymerases 13,41 and result in the generation of DNA 6mA in a replication-dependent but epigenetically independent manner.
To avoid possible confusion, here, misincorporated 6mA is named i6mA, postreplicative and methylase-deposited 6mA is named pr6mA, and 6mA is designated the collection of DNA i6mA and pr6mA. Chemically, i6mA has the same chemical structure as methylase-deposited pr6mA; thus, they are indistinguishable. Biophysically, like the truly methylated product pr6mA, misincorporated i6mA should also destabilize the DNA double helix,and thus ditto affects DNA replication and transcription and potentially falsify the epigenetic landscape of DNA pr6mA, which might be present at extremely low abundance 13,42 . Collectively, i6mA, which is unintentionally set on DNA, should be considered a form of DNA damage.
In this work, by taking advantage of unique heavy stable isotope tracing, we examined the misincorporation of DNA i6mA in response to intracellular RNA m6A degradation. We found that intracellular RNA m6A degradation cannot induce any i6mA. Furthermore, we discovered a deaminase ADAL-pivoted and adenylate kinase 1 (AK1)-assisted catabolism checkpoint. Mechanistically, the identi ed checkpoint blocks the reformulation of RNA decay-derived m6A nucleotides into 6mdATP and thus eliminates misincorporated i6mA in the process of intracellular RNA m6A degradation followed by purine salvaging.

Results
Labeling of RNA m6A by heavy stable isotope-labeled adenine nucleoside We rst exploited heavy stable isotope-labeled nucleoside [  Recently, it was reported that intracellular RNA m6A degradation could induce misincorporated DNA i6mA in a number of cells, including mES cells and HEK293T cells 41 . As noted, the misincorporation of i6mA showed a delayed phase (~ 5 days delay) compared to de novo RNA synthesis 41 . Following this clue, we treated mES cells with the tracer [ 15 N 5 ]-rA over 7 days. With a high labeling e ciency (60%), the labeled mRNA m6A reached a level of 2.26 per 10 3 rC (Fig. 1C), but no labeled 6mA ([ 15 N 4 ]-6mA and [ 15 N 5 ]-6mA) was detected, except trace amounts of the unlabeled 6mA in genomic DNA (8.8 per 10 8 dC) (Fig. 1C). Even after 50 days of labeling for tracing RNA m6A degradation, we still failed to observe any labeled 6mA in the genomic DNA of mES cells (Fig. 1D, upper panel). After performing similar treatments (7-50 days), we did not observe any labeled 6mA in HEK293T cells yet (Fig. 1D, lower panel). Notably, our UHPLC-MS/MS method could detect 6mA with a sensitivity of less than one 6mA per 10 8 total dA. These results consistently support that intracellular RNA m6A degradation cannot induce any misincorporated DNA 6mA.
We also treated mES cells with a second heavy stable isotope tracer, [D 3 ]-L-methionine, for 7 days. In vivo, the tracer The above data consistently supported the absence of labeled DNA 6mA no matter intracellular RNA m6A was labeled with [ 15 N 4 ] or [D 3 ]. These results drove us to propose the existence of a catabolic RNA m6A checkpoint that blocks the reformulation of intracellular RNA m6A degradation products to form incorporable 6mdATP and thus eliminates DNA 6mA misincorporation.
The deaminase ADAL preferentially catabolize 6mdAMP in vitro Previous studies have shown that the adenine deaminase-like protein ADAL can protect RNA by reducing m6A misincorporation 44 . However, it is not known whether ADAL can suppress DNA misincorporation. We speculated that ADAL might catabolize N 6 -methyldeoxyadenosine monophosphate (6mdAMP) to nontoxic deoxyinosine monophosphate (dIMP) ( Fig. 2A). If so, ADAL should play a critical role in the catabolism checkpoint blocking i6mA misincorporation. This drove us to test the catalytic activity of ADAL on 6mdAMP. We rst investigated the activity of mouse recombinant ADAL protein on m6AMP. As detected by UHPLC-MS/MS analysis, the recombinant ADAL protein catabolized m6AMP to inosine monophosphate (rIMP) (Fig. 2B). The result is in agreement with previous report 44,45 . Then, we investigated the catalytic activity of mouse recombinant ADAL protein on 6mdAMP in vitro. Interestingly, the recombinant ADAL protein also catabolized 6mdAMP to dIMP (Fig. 2C). Along with reductions in 6mdAMP (detected in the form of 6mA by UHPLC-MS/MS), a dramatic increase in the level of the nontoxic product dIMP (detected in the form of dI) was observed (Fig. 2C). Evidently, ADAL displays a novel ability to catalyze the deamination of 6mAMP nucleotides in vitro.
To further explore the relative catalytic activity, we next measured the catalytic e ciency (K cat /K m ) of ADAL for both 6mdAMP and m6AMP. As shown in Fig. 2D, in our in vitro reaction system, the catalytic e ciency of ADAL for 6mdAMP was 33.4 ± 0.83 mM − 1 s − 1 . This value is 2.14 fold as high as that of ADAL for m6AMP (15.6 ± 0.71 mM − 1 s − 1 ) (Fig. 2E). In addition, the ADAL protein showed greatly reduced activity toward 6mA deoxynucleoside and 6mdATP ( Fig. S2A & S2B). These data support that ADAL preferentially catabolizes 6mdAMP. The m6A demethylases FTO and ALKBH5 can erase the methyl group of mRNA m6A 46,47 and downregulate the level of mRNA m6A. Surprisingly, the depletion of these enzymes (by CRISPR/Cas9-based knockout) also failed to induce any labeled 6mA in genomic DNA ( Fig. S3E and S3F).
These results support that only the deaminase ADAL is critically involved in the checkpoint for suppressing 6mA misincorporation and that other known adenine-related deaminases and m6A demethylases do not have any direct role in this checkpoint. Adenylate kinase 1 is an important accessary factor maintaining the catabolic checkpoint To further identify the factors involved in the checkpoint, we constructed a series of overexpression plasmids for nucleotide kinases functioning in the purine salvage synthesis pathway, including adenosine kinase (ADK), adenine phosphoribosyltransferase (APRT), adenylate kinase 1 (AK1), and adenosine diphosphate kinase (DNPK). These proteins are hypothesized to facilitate the formation of 6mdATP by increasing phosphorylation at a certain step in the purine salvage pathway. However, no labeled 6mA was detected in any of the gene overexpression groups or control groups of wild-type mES cells (Fig. S4). These results may suggest that such a simple increase in the expression of nucleotide kinases cannot break the misincorporation-suppressing checkpoint. Then, we replaced wild-type cells with Adal −/− cells to explore whether nucleotide kinases are regulatory factors rather than determining factors for the checkpoint. As shown in Fig. 4, notably, the level of labeled 6mA in the AK1-overexpressing group (7.80 per 10 7 dC) is approximately 3.25-fold higher than that in the control group (EV, 2.40 per 10 7 dC). In contrast, the labeled 6mA level in the ADK-, APRT-, and NDPK-overexpressing groups (1.94-2.29 per 10 7 dC) is similar to that in the control group. These results suggest that AK1 participates in the catabolism checkpoint of RNA m6A. Preferential misincorporation of m6A in RNA revealed by extracellular m6A treatment Our unpublished data showed that m6A nucleosides could be detected in human urine at a high abundance, indicating the presence of nonnegligible extracellular m6A-related species in humans. Therefore, in addition to intracellular RNA m6A degradation, humans may be exposed to extracellular m6A-related nucleosides or nucleotides.
To test the effects of such exposure, we exploited the heavy stable isotope tracer [  Fig. 5B), respectively. These values suggested that RNA m6A misincorporation is 9.1-fold as high as DNA 6mA misincorporation, supporting a preference on RNA misincorporation.

Differential ADAL expression in cancers
We also analyzed the potential clinical signi cance of ADAL in four human cancers based on the TCGA data sets. In cervical squamous cell carcinoma, kidney renal papillary carcinoma, pancreatic ductal adenocarcinoma, and rectum adenocarcinoma, patients with low ADAL expression in tumor tissues had worse survival rates than those with high ADAL expression (Fig. 6). This hints that the ADAL aberrance may play a critical role in cancer development. lost, and we did not observe any labeled m6A in mRNA (data not shown). Taken together, intracellular RNA m6A can be traced using heavy stable isotope labeled rA.

Discussion
Intracellular RNA m6A degradation cannot induce misincorporated DNA i6mA. By the use of the e cient [ 15 N 5 ]-rA labeling strategy, intracellular RNA m6A can be e ciently labeled. If the degraded RNA m6A could be reformulated to generate 6mdATP via purine salvage pathway, we should observe labeled DNA 6mA. However, we did not observe any labeled DNA 6mA over 7-50 days of treatment, indicating the absence of misincorporated DNA i6mA. Although it was reported that DNA misincorporation occurred at a delayed phase (approximately 5-day delay) in comparison to RNA labeling 41 , our observation on the absence of labeled DNA 6mA (or misincorporated i6mA) over 50 days of treatment proves that the m6A nucleotides generated from RNA m6A decay are not reutilized in DNA replication. By the use of the second strategy involved [D 3 ]-L-methionine labeling, we also observed the labeled RNA m6A ([D 3 ]-m6A), but consistently did not observe any labeled DNA 6mA after 7 days of treatment. Notably, both labeling strategies showed high labeling e ciency (> 60%). Collectively, our data strongly support that intracellular RNA m6A decay cannot induce any misincorporated DNA i6mA at least in the tested cells.
Checkpoint, ADAL, and AK1. Given the capacity of DNA polymerases to incorporate 6mdATP 13 and the absence of the labeled DNA 6mA (also misincorporated i6mA herein), we infer that the degradation products of RNA m6A cannot be ultimately converted into 6mdATP via the purine salvage pathway. This inference drives us to propose that a checkpoint functioning similar to that in the cell cycle 49,50 and in the immunological response 51-53 exists for suppressing DNA misincorporation along with RNA m6A degradation. Essentially, we identi ed two players in this proposed checkpoint, ADAL and AK1.
The depletion of the deaminase Adal by either knockdown or knockout resulted in the presence of the labeled DNA 6mA. In contrast, the depletion of the mRNA m6A demethylase FTO or ALKBH5 did not break the checkpoint, nor did overexpression of the kinases involved in the purine salvage pathway alone. All these ndings support the idea that ADAL plays a pivotal role in the checkpoint. Mechanistically, ADAL simultaneously catabolizes the methylated nucleotides m6AMP and 6mdAMP to the nontoxic nucleotides IMP and dIMP, respectively, reducing the misincorporation of N 6 -methyladenine via DNA polymerase; thus, the hydrolysis of both m6AMP and 6mdAMP by ADAL contributes critically to the blockade of replication-dependent 6mA incorporation. Moreover, the catabolizing activity of ADAL toward 6mdAMP is 2.14-fold as high as that toward m6AMP. Of note, here we showed, for the rst time, the activity of ADAL in catabolizing 6mdAMP. Collectively, our data indicate that ADAL functions as a key checkpoint enzyme to block the catabolic conversion of intracellular RNA m6A to genomic DNA 6mA by catabolizing both m6AMP and 6mdAMP, particularly the latter.
In addition to the deaminase ADAL, adenylate kinase 1 (AK1) was identi ed as a protein involved in the checkpoint.
Overexpression of AK1 did not induce any detectable labeling of DNA i6mA in the presence of su cient ADAL but increased the labeled DNA i6mA in the absence of ADAL. This nding suggests that the phosphorylation of 6mdAMP into 6mdADP is a much slower process than the hydrolysis of 6mdAMP mediated by ADAL. In other words, the preferential substrate for AK1 is not 6mdAMP but dAMP, which is required for DNA synthesis. Overexpression of AK1 can partially compensate for the slow phosphorylation of 6mdAMP by AK1. The slow phosphorylation process would provide a time window long enough for ADAL to catabolize the generated 6mdAMP completely.
Source of the unlabeled DNA 6mA. As shown in this work, in mES cells treated with heavy stable isotope tracers, we found large proportions of the labeled m6A in both mRNA and total RNA, but no labeled genomic 6mA was detected.
However, trace amounts of the unlabeled 6mA were detected. This nding is consistent with our recent work 13 . The observed unlabeled DNA 6mA was formed through polymerase-dependent misincorporation 13,41 . However, the source for generating the unlabeled 6mA was not known yet. In this work, we clearly showed that the observed DNA 6mA did not originate from intracellular RNA m6A, as described above. Because more than three-quarters of the m6A modi cations were labeled, it is impossible that only unlabeled RNA m6A was incorporated into genomic DNA, and the labeled m6A could not be incorporated via the purine salvage pathway. The absence of the labeled DNA 6mA also excludes an origin from methylase-deposited methylation. We speculate that the observed DNA i6mA originated from catabolism of extracellular RNA m6A. Indeed, we detected urinary m6A nucleosides at high abundance (data not published), suggesting the existence of free m6A in extracellular spaces in humans. Meanwhile, by the extracellular exposure of heavy stable isotope-labeled m6A nucleoside, we clearly showed that exogenous m6A nucleoside can induce misincorporated DNA 6mA and such a process is also restricted by ADAL-pivoted checkpoint. It is reasonable that, under physiological conditions, both intracellular and extracellular RNA m6A are tightly regulated by the catabolism checkpoint. Biochemically, the N 6 -methyladenine modi cation on the DNA template disturbs base pairing 54,55 , hinders new chain extension, affects replication 24 and transcription 23 , and regulates related gene expression 22,54 . Therefore, a strict catabolism checkpoint of N 6 -methyladenine is required to maintain genome integrity.
Biological implications. The expression of ADAL varies in different cells. It is expected that misincorporated i6mA could be observed in cells that express ADAL at low levels. Musheev et al 41 reported that C2C12 and especially NIH 3T3 cells had exceptionally high levels of misincorporated i6mA, 13-500 i6mA per million dA. However, by our experiments on C2C12 and NIH 3T3 cells, we observed i6mA at a level of 100 folds lower (Fig. S6). We did not detect misincorporated i6mA in 293T and mES cells. Collectively, the misincorporated i6mA must be strictly controlled.
Our observation on the blockade of DNA 6mA misincorporation strongly suggests that the epigenetic DNA 6mA landscape should be tightly maintained in the tested cells. Essentially, the ndings on epigenetic mark DNA 6mA remain elusive due to the neck-of-bottle of analytical technologies for 6mA detection, the contamination of coexisting bacteria carrying abundant 6mA, poor LC-MS/MS skill, and a lack of reliable 6mA sequencing technology 56 . However, it is still very surprising that misincorporated i6mA must be strictly controlled, while epigenetic pr6mA is rarely found in human cells. These con icting observations may hint that true epigenetic 6mA (methylase-deposited 6mA) should have extremely important functions and appear in every cell in certain scenarios, e.g., responses to certain stimuli. Otherwise, the mammalian cells should not set such a strict bar for misincorporation. Interestingly, as the most important and abundant epigenetic mark, misincorporation of 5mC must also be strictly controlled [57][58][59] . In contrast, DNA 5-hydroxymethylcytosine (5hmC) is found only in limited cells at moderate abundance and can be misincorporated into genomic DNA 60 . On the other hand, low expression of Adal may shorten the survival of cancer patients (Fig. 6). These data and underlying logics allow us to infer that the ADAL-pivoted checkpoint and the suppression of i6mA misincorporation have biological importance, at least in tumor development and therapy.
RNA m6A checkpoint in the purine salvage pathway. In the free nucleotide pool, free m6AMP can be recycled by rephosphorylation via nucleotide kinases and further reduced to N 6 -methyldeoxyadenosine diphosphate (6mdADP) by ribonucleotide reductase. 6mdADP is next transformed into 6mdATP, which is utilized by DNA polymerases, forming DNA i6mA. Our results indicated that the substantially labeled RNA m6A in mES cells was not transformed and incorporated into genomic DNA. Taken together, our ndings led us to propose a catabolic checkpoint of N 6methyladenine along with the purine pathway (Fig. S7). This checkpoint functions to block potential DNA 6mA misincorporation via ADAL-mediated hydrolysis of both m6AMP and 6mdAMP, AK1 mediated restriction on the phosphorylation of 6mdAMP, and other unknown steps.
In this work, by exploiting unique heavy stable isotope labeling technology coupled with sensitive UHPLC-MS/MS analysis, we demonstrated that ADAL catabolizes m6AMP and, more e ciently, 6mdAMP to nontoxic nucleotides and blocks DNA 6mA misincorporation. In addition, adenylate kinase 1 also contributes to the suppression of DNA 6mA misincorporation by limiting the rate of 6mdAMP phosphorylation, which may provide adequate time for ADAL to catabolize 6mdAMP. Therefore, ADAL and AK1 together suppress DNA 6mA misincorporation and maintain genomic epigenetic integrity.
To be labeled with stable isotope tracers [ 15 N 5 -rA]  Total RNA was extracted with TRIzol Reagent (Thermo) according to the manufacturer's instructions. mES cells were harvested and collected into 1.5 ml aseptic centrifuge tubes on ice. Then, the cell pellets were lysed and homogenized in TRIzol reagent and immediately subjected to chloroform extraction and ethanol precipitation. The extracted total RNA was dissolved in DEPC-treated water and quantitated using a NanoDrop 2000 (Thermo). Then, mRNA was extracted from the total RNA using a Dynabeads™ mRNA Puri cation Kit (Thermo) according to the manufacturer's instructions.
Enzymatic digestion of RNA (5.0 µg) was conducted with a mixture of 1 U of benzonase, 0.25 U of NP1, 0.02 U of SVP, and 1 U of CIP in Tris-HCl buffer (10 mM Tris-HCl, pH 8.0, plus 1 mM Mg 2+ ) at 37°C for 12 h. Then, the enzymes were removed by ultra ltration (molecular weight cutoff: 3 kDa; Pall Corporation, USA). The ltered solution containing nucleosides was subjected to UHPLC-MS/MS analysis. siRNA transfection mES cells were seeded into 6-well cell culture clusters at a density of 10 5 per well the night before transfection with 30 pmol of siRNA against Adal using Lipofectamine RNAiMAX Transfection Reagent (Thermo) according to the manufacturer's protocols. Control cells were transfected with 30 pmol of negative control sequence. The transfected mES cells were cultured for another 36 h before qPCR analysis or 48 h before western blot analysis. UHPLC-MS/MS analysis UHPLC-MS/MS analysis was performed as described previously 13 . Enzymatic digestion products of DNA and RNA were injected into an Agilent 1290 II UHPLC system coupled with an electrospray ionization (ESI)-triple quadrupole mass spectrometer (6470, Agilent Technologies, Santa Clara, CA), and a Zorbax SB-Aq column (2.1 × 100 mm, 1.8 µm particle size, Agilent, USA) was used for separation. The mass spectrometer was operated under positive ionization using multiple reaction monitoring (MRM) mode. and the nebulization gas pressure was set at 40 psi. The other conditions were as described previously 13 . Treatment of mES cells with extracellular m6A nucleotides mES cells of wild-type or Ada1 −/− were seeded into 6-well cell culture clusters at a density of 2×10 5 per well, and fresh cell culture medium was supplemented with [D 3 ]-m6A nucleoside at concentrations of 0, 0.1, 0.2, 0.5, or 1 µM. The treated cells were harvested after incubation for 2 days, and DNA and RNA were extracted for further analysis.

Activity analysis of ADAL in vitro
An ADAL fusion protein whose N-terminus was tagged with maltose-binding protein (MBP) was expressed in E. coli and then puri ed via a prepacked hydrophobic interaction chromatography column and a prepacked MBP-Trap HP 5 ml column (GE Healthcare, Uppsala, Sweden). Then, the purity was analyzed with a BCA Protein Assay Kit (Beyotime) and via SDS polyacrylamide gel electrophoresis (PAGE).
Next, we prepared 6mdAMP by digesting plasmid DNA that carries abundant DNA 6mA. The plasmid DNA was extracted from E. coli. The plasmid of 10 µg was digested in a 100 µl reaction mixture containing 10 mM Tris-HCl (pH 8.0), 2 mM Mg 2+ , 1 U of benzonase, and 0.02 U of SVP. After 6 h of incubation, the mixture was ltrated by centrifugation through an ultra ltration tube (MW cutoff: 3 kDa; Pall Corporation, USA). Then, the ltrated solution was mixed with 400 nM m6AMP. By this protocol, solutions containing two ADAL substrates, 6mdAMP and m6AMP, were prepared. For characterization of the preparations, 1.5 µg product that had been treated with 1.0 U CIP for 1 h, and an equivalent amount of untreated product was analyzed by UHPLC-MS/MS for quanti cation of 6mA or m6A.
The difference in 6mA or m6A between the CIP-treated product and the untreated product represented the content of monophosphorylated N 6 -methyladenine.
An ADAL activity assay was conducted in 50 µl reaction buffer (20 mM Tris-HCl (pH 7.0), 0.1 µg/µl BSA, 2 mM DTT) supplemented with 1.5 µg of the above enzymatic products containing m6AMP andr 6mAMP and 0 or 0.5 µg of ADAL protein at 37°C for 1 h. After the reactions, the solutions were treated with proteinase K at 55°C for 30 min and heated for enzyme inactivation at 95°C for 10 min and then incubated with 5.0 U CIP for another 1 h. The reaction substrates m6AMP and 6mAMP and the products rIMP and dIMP were converted into m6A, 6mA, rI, and dI, respectively, for UHPLC-MS/MS analysis. After passing through ultra ltration tubes (MW cutoff: 3 kDa), the nucleoside solution was subjected to UHPLC-MS/MS analysis.
The catalytic e ciency (K cat /K m ) of ADAL for m6AMP and 6mAMP was analyzed in a 50 µl reaction mixture of 20 mM Tris-HCl (pH 7.0), 0.1 µg/µl BSA, 2 mM DTT, 0.2 µg of ADAL protein, and varying concentrations of the test compound. After incubation at 37°C for 10 min, the reaction mixture was incubated with 5.0 U of CIP for another 1 h to convert all of the nucleotides in the mixture into nucleosides.   Overexpression of adenylate kinase 1 (AK1) elevates DNA 6mA misincorporation as ADAL is depleted. Quanti cation of genomic 6mA in [15N5]-rA-treated Adal-/-cells overexpressing ADK, APRT, AK1 or NDPK gene for 7 days.