DNA modifications impact natural transformation of Acinetobacter baumannii

Abstract Acinetobacter baumannii is a dangerous nosocomial pathogen, especially due to its ability to rapidly acquire new genetic traits, including antibiotic resistance genes (ARG). In A. baumannii, natural competence for transformation, one of the primary modes of horizontal gene transfer (HGT), is thought to contribute to ARG acquisition and has therefore been intensively studied. However, knowledge regarding the potential role of epigenetic DNA modification(s) on this process remains lacking. Here, we demonstrate that the methylome pattern of diverse A. baumannii strains differs substantially and that these epigenetic marks influence the fate of transforming DNA. Specifically, we describe a methylome-dependent phenomenon that impacts intra- and inter-species DNA exchange by the competent A. baumannii strain A118. We go on to identify and characterize an A118-specific restriction-modification (RM) system that impairs transformation when the incoming DNA lacks a specific methylation signature. Collectively, our work contributes towards a more holistic understanding of HGT in this organism and may also aid future endeavors towards tackling the spread of novel ARGs. In particular, our results suggest that DNA exchanges between bacteria that share similar epigenomes are favored and could therefore guide future research into identifying the reservoir(s) of dangerous genetic traits for this multi-drug resistant pathogen.


INTRODUCTION
The ability of bacteria to exchange genetic features by horizontal gene transfer (HGT), both within and between species, can dri v e rapid bacterial evolution. Alarmingly, this can result in the accumulation of dangerous genetic traits and the emergence of extremely problematic pathogens. Indeed, this is the case for the bacterium Acinetobacter baumannii , an ESKAPE pathogen, which is often able to escape treatment with multiple clinically-important antibiotics, and has ther efor e prov en e xtremely prob lematic in hospital settings ( 1 ). A. baumannii was shown to exchange genetic information through HGT, but the relati v e contribution of conjugation, transduction, natur al tr ansformation, and other modes of HGT to resistance towards diverse antibiotics is not well understood ( 2 ). Ne v ertheless, since most of the antimicrobial (AMR) resistance genes in A. baumannii are located on the chromosome (3)(4)(5)(6), natural transformation could potentially explain how most of these genes were acquired. Transformation is one of the primary modes of HGT that can occur in naturally competent organisms ( 7 ). Precisely, natural competence is a physiological state in which the bacterium is able to take up DNA from the environment and either integrate it into its chromosome by homologous recombination or reconstitute it into a plasmid ( 8 , 9 ). The DNA uptake machinery is mostly conserved in naturally competent bacteria ( 8 , 9 ). In Gram-negati v es, the first step, namely the uptake of extracellular DNA into the periplasm, involves the retraction of a type IV pilus (T4P) ( 9 , 10 ). The incoming DNA is then bound by the DNA-binding protein ComEA, which acts as a Brownian ra tchet, facilita ting the entry of long pieces of DNA into the periplasmic space (11)(12)(13). In the next step, ComEC is thought to degrade one strand of the incoming DN A and, to gether with ComF, transfer the single stranded DN A (ssDN A) into the cytoplasm ( 9 , 14 ). There, ssDNA is protected by Ssb and DprA, which also facilitates the loading of RecA ( 15 ). Finally, if there is sufficient homolo gy, ssDN A is integrated into the host chromosome via the action of RecA and ComM, or, in case of plasmid DNA, the plasmid is reconstituted either by annealing of two complementary imported ssDNA strands or by the synthesis of the second DNA strand prior to circularization ( 9 , 16-21 ).
To da te, na tur al tr ansformation has been experimentally demonstrated in se v eral strains of A. baumannii (22)(23)(24)(25). Howe v er, these studies reported considerable variation in the transformation frequencies obtained in the different str ains. Some str ains were non-tr ansformable, while others were transformable with frequencies ranging from 10 −8 to 10 −2 (23)(24)(25). Similar observations have also been reported for the naturally competent bacterial species Streptococcus pneumoniae and Haemophilus influenzae ( 26 , 27 ). W hile such varia tion in transformability might be a result of strain-specific competence regulation, it is also possible that DNA-specific factors could be involved. Furthermore, some studies hav e e xamined the transformability of A. baumannii using different kinds of donor DNA, including genomic DN A (gDN A), integrati v e or replicati v e plasmids, and PCR-amplified DNA fragments. While one study suggested that natural transformation occurs independently of the type of donor DNA used ( 28 ), another reported tha t transforma tion occurred most ef ficiently with purified gDNA, followed by integrati v e plasmids, and was the least efficient with PCR fragments ( 23 ). The increased efficiency of transformation with gDNA could be explained by longer regions of marker-flanking homology, yet the differences observed between the integrative plasmid and PCR product hint towards the involvement of DNA-specific features. Such features might be dependent on the source of the transforming DNA and should ther efor e be consider ed as a possible contributor to the observed variability.
Se v eral DNA features hav e pre viously been shown to play a role in the transformability of certain naturally competent bacteria. Firstly, the sequence di v ergence between the transforming DNA and the acceptor's chromosome was demonstrated to significantly diminish the bacterium's natur al tr ansformability ( 29 , 30 ). Secondly, it was reported that the presence of specific nucleotide sequences within the transforming DNA positi v ely influences DNA uptake in the genera Neisseria and Haemophilus . These so-called DNA uptake sequences (DUSs) are conserved 9-12 bp long DNA motifs, which were shown to enhance the exchange of DNA between bacterial relati v es belonging to the same species (31)(32)(33)(34). Thirdly, the presence of DNA modifications was shown to affect transformability in di v erse bacteria, such as Helicobacter p ylori , Camp ylobacter jejuni and others (35)(36)(37)(38)(39)(40). For instance, it was reported that a specific methylation motif, the result of a Campylobacter -specific orphan methylase, promotes DNA uptake in this organism ( 35 ). Importantl y, DN A methyla tion pa tterns can also be a result of restriction modification (RM) systems. RM systems typically consist of a methylase and a cognate restriction enzyme that both recognize the same DNA sequence motif ( 41 ). Methylation protects the self-DNA from degradation by the restriction enzyme, while non-methylated DNA is efficiently degraded (42)(43)(44). Notably, RM-linked restriction enzymes solely cut unmodified (e.g. unmethylated) dsDNA, which does not naturally occur during the cell cycle of RMharboring bacteria. Indeed, after replication of self-DNA, the neosynthesized strand is ra pidl y methylated by the respecti v e RM methylase enzymes, which renders the DNA resistant a gainst cleava ge by the cognate restriction enzyme ( 42 ).
In the context of natural competence, the transforming material enters the cell as a single strand. As RM-linked restriction enzymes usually target dsDNA, they are unlikely to cleave foreign DNA at this stage of the transformation process. Instead, Johnston and colleagues proposed that foreign DNA would be targeted after it has been replicated ( 42 ). Precisely, natur al tr ansformation depends on homolo gous DN A regions that flank heterolo gous stretches and f oster the f ormation of heteroduplexes. During this stage, homolo gous DN A is thought to be modified by the acceptor strain's methylase due to its complementarity to the bacterium's genomic DNA ( 42 ). Howe v er, non-complementary heterolo gous DN A will r emain single-stranded and ther efore unmethylated. Hence, after passage of the replication fork, which was recently shown to resolve transformationlinked heteroduplexes ( 45 , 46 ), this region will be converted to fully unmethylated dsDNA, which serves as target for restriction enzyme(s) ( 42 ). Interestingl y, w hile studies in Bacillus subtilis and S. pneumoniae reported only a minimal or no effect of RM systems on transformation ( 47 , 48 ), the co-occurrence of RM systems and natural competence was demonstrated to affect transformability in H. pylori , Pseudomonas stutzeri , C. jejuni, and Neisseria meningitis ( 36 , 37 , 39 , 40 ). Nonetheless, little is known about the impact of specific RM systems or DNA modifications on HGT in the genus Acinetobacter.
In this study, we aimed to understand how the source of the transforming DNA affects transformability in A. baumannii . This is especially relevant, since novel genetic traits such as antibiotic resistance genes (ARGs) are usually acquired from non-self-organisms. To unravel the involvement of specific DNA modifications in transformation, we used bacterial genetics and whole genome PacBio SMRT sequencing to identify specific genetic determinants responsible for the observed source DNA-dependent transformation phenotype. Additionally, we went on to investigate how the observ ed DNA selecti vity is regulated, to understand whether its impact could be avoided under specific conditions. Collecti v ely, our wor k highlights the importance of epigenetic modifications for natural transformability in the opportunistic pathogen A. baumannii .

Bacterial strains, plasmids and growth conditions
The bacterial strains used in this study are listed in Table  S1. The strains were cultured in l yso geny broth medium (LB; 1% tryptone, 0.5% yeast extract, 1% sodium chloride; Carl Roth, Germany) under shaking conditions (180 rpm) at 37 • C or on solid LB agar plates (1.5% agar) at 37 • C. To select for A. baumannii after bi-or triparental mating with E. coli donor strains, CHROMagar Acinetobacter plates served as medium. These plates were prepared following the protocol provided by the manufacturer (CHRO-Magar, France). Gentamicin was added to the plates to select the colonies that acquired the respecti v e transposon and apramycin was added to CHROMagar Acinetobacter plates to select for the bacteria carrying the envisioned (suicide) plasmids.
Transformation medium was pr epar ed as described previousl y ( 49 ). Briefly, medium containing 5 g / l of tryptone, 2.5 g / l NaCl, and 2% agarose D3 (Euromedex, France; agar ose fr om other manufacturers was non-functional for these assays) was autoclaved and kept at 60 • C in liquid form. Prior to the transformation experiment(s), 1ml of this transformation medium was poured into 1.5 ml Eppendorf tube(s) and kept at room temperature (RT) to solidify.

Analysis of DNA modification motifs
DNA modifications were determined based on the SMRT sequencing results of this study plus the output obtained for strain A118 for which we previously reported its genome assembly ( 50 ). The raw data of strains A118, A TCC17978, and A TCC19606 were analyzed using the smrtlink-r elease 7.0.1.66975 softwar e suite, while the raw data from all other sequenced strains were analyzed using the smrtlink-r elease 10.2.0.133434 softwar e suite. The DNA modification raw data are available from Zenodo under doi: 10.5281 / zenodo.7759910). Both software suites were obtained directly from the vendor's w e bpage: https://www.pacb.com/support/software-downloads/ . The tool named 'Base Modification Analysis Application' was used to detect modified DNA bases in the bacterial genomes, and the corresponding DNA motifs surrounding modified bases, as outlined in this White Paper: https://www.pacb.com/wp-content/uploads/2015/09/ WP Detecting DNA Base Modifications Using SMRT Sequencing.pdf. The sequencing reads of the different data sets were mapped onto the corresponding reference genome sequences of each strain and analyzed using the default parameters of the tool (95% confidence interval bounds, P-value cutoff = 0.001, minimum methylation fraction = 0.3 and minimum Qmod score = 100) and activating its 'Find Modified Base Motifs' option. For each analyzed sample, the tool produced a table listing the genomic positions at which a base modification was detected, as evidenced by the deviation of the inter-pulse duration from the expected base incorporation time. The tool's output listed motifs that are present in > 30% of the occurrence of that motif around detected modified bases. All motifs identified in at least one sample were collected and a perl script was written to count the number of occurrences of each motif in the whole genome of each sample, and to compute the fraction of each motif that contained the detected modified bases (available on Zenodo under doi: 10.5281 / zenodo.7760218). The modified positions wer e r epr esented on Cir cos plots using the R package circlize ( 52 ), and a summary heatmap comparing the fraction of modified motifs found in each sample was produced using the R function heatmap.2 from the library gplots package 3.1.3 ( https://github.com/talgalili/gplots ).

Identification of potential methylases in different A. baumannii isolates
To investigate the potential methylases across A. baumannii isolates, we performed a pangenome reconstruction using the PPanGGOLiN (v. 1.2.74) pipeline ( 53 ) (all options with default parameters; both nucleotide sequences and annotation files were used as inputs to the pipeline). Briefly, PPanGGOLiN uses genome sequences and annotations to reconstruct pangenomes and subsequently build genefamily based partitions. The progr am integr ates information from the annotation (protein coding genes, CDS, genomic neighborhood) to define gene families (for more details on the method, see ( 53 )). Next, genes are split into persistent and cloud families. PPanGGOLiN identified 5181 genes (3108 persistent and 2073 cloud genes) across the 6 strains. Out of the obtained gene r epertoir e, we identified 111 genes which contained 'methyl' in their annota tion, and, after individual cura tion, we singled out 29 specific DNA methylase / methyltr ansfer ase genes. Based on their presence / absence in each strain, we reconstructed a heatmap in R environment (v 4.2.1) ( 54 ) using the pheatmap package v 1.0.12 at https://cran.r-project. org/w e b/packages/pheatmap/index.html (default parameters with no clustering).

NCBI BLAST analyses
To assess the conservation of the RMC cluster and identifiy putati v e homologs, NCBI's Basic Local Alignment Search Tool (BLAST) was used. Briefly, a blastp analysis was performed using standard parameters and the translated H0N27 10830 coding sequence as input (accession date as of 20 April 2021). Moreover, a blastn search of the A118 RMC region was performed against NCBI's Ref-Seq Genome Database (refseq genomes; accession date as of 30 November 2022). All the parameters were set to default and the organism option was set to Acinetobacter (taxid:469).

Comparison of RMC-containing A118 genomic region with other A. baumannii strains.
Genome alignment of strains A118, ACC17978, ATCC19606, AB5075, 29D2 and 86II / 2C was performed in Geneious Prime (v 2022.2.2) with the progressi v eMauv e algorithm ( 55 ). The genomic region carrying the methylase H0N27 10820 within the RMC in strain A118 was compared with the same genomic region in other strains to identify conserved upstream and downstream flanking genes. After obtaining the specific sequences and annotations for this genomic region in the six different strains, we aligned, compared, and visualized the region using Clinker (v 0.0.25) ( 56 ) with default options.

Construction of mutants in A. baumannii
Most of A. baumannii mutants (see Table S1) were constructed following a natural transformation protocol as described previously ( 49 ) with few modifications. To genetically engineer strains, PCR amplification products served as transforming material. These PCR fragments were amplified using PWO polymerase (Roche) or Q5 polymerase (New England BioLabs). Each fragment consisted of an antibiotic resistance cassette (kanamycin or gentamicin resistance genes encoded by aph and aacC1, respecti v ely) flanked by ∼1-kb long regions homologous to the desired chromosomal insertion locus. The final PCR fr agments were gener ated by over lapping PCR using the three individual fragments as template, and were added to respecti v e bacterial strains, as described below for the natural transf ormation assa y. The transf ormed cells were resuspended in PBS and plated on selecti v e LB agar plates (containing kanamycin or gentamicin). Selected transformants were screened by colony PCR using GoTaq polymerase (Promega) and the correct DNA sequences were confirmed by Sanger sequencing (Microsynth, Switzerland).
To obtain a mutant in non-transformable strain ATCC19606, a standard suicide plasmid-based mating approach was used, as described previously ( 50 ). Briefly, a deri vati v e of the counter-selectable plasmid pGP704-Sac28 ( 57 ), carrying the desired deletion construct was engineered by standard molecular methods. The two flanking regions of the chosen genomic exchange region were amplified and joined with PCR fragment carrying antibiotic resistance cassette by overlapping PCR using Q5 polymerase (New England BioLabs). Next, this final PCR fragment was cloned into plasmid pGP704-Sac28. The cloned construct was transferred into E. coli strain S17-1 pir and confirmed by colony PCR (GoTaq polymerase, Promega) and Sanger sequencing (Microsynth, Switzerland). The correct plasmid was transferred to A. baumannii by biparental mating. Subsequent sucrose-based counter-selection was used to select plasmid-cured clones after a putati v e allelic e xchange e v ent as pre viously described ( 50 ). Allelic e xchanges were confirmed by Sanger sequencing (Microsynth, Switzerland).
To electroporate the plasmid pABA, a 2 ml of overnight culture of the respecti v e strain was centrifuged for 10 min at 10 000 ×g at room temperature (RT). The pelleted bacteria were washed twice with 1 ml of sterile 10% glycerol and resuspended in 100 l of 10% glycerol. 2 l of plasmid (400 ng) was added to 50 l of such electrocompetent cells and electroporated inside an electroporation cuvette at 1.8 kV. Immediately after the electroshock, 950 l of 2 ×YT was added to the bacteria to foster their recovery. The suspension was car efully transferr ed into 14 ml round-bottom tubes and grown aerobically for 1 h at 37 • C. After incubation, the cells were spread onto LB agar plates containing apramycin and incuba ted a t 37 • C overnight. The presence of the plasmid was confirmed by colony PCR (GoTaq polymerase, Promega). For construction of ATCC17978 hcp::aprR, a PCR product with an amplified AprR resistance cassette flanked by 1-kb long regions homologous to the desired chromosomal insertion locus was used instead of the plasmid during electroporation.
To construct mutants with inducible genes, the respective genes were cloned behind the arabinose-inducible promoter P BAD located on the miniTn7 transposon ( = TnAraC), which itself was carried on the suicide plasmid pGP704 (e.g. plasmid pGP704-TnAraC; ( 51 , 58 )). The engineered plasmids were transferred into A. baumannii through standard triparental mating ( 59 ), which, assisted by a T7 transposition machinery-encoding helper plasmid, resulted in the transposition of the TnAraC-XX (abbreviated as TnXX) construct downstream of the glmS gene on the A. baumannii chromosome.
Nucleic Acids Research, 2023, Vol. 51, No. 11 5665 Natur al tr ansf ormation assay Natur al tr ansf ormation assa ys were perf ormed as previously described ( 49 ) with minor modifications. Briefly, overnight cultures of A. baumannii were back-diluted 1:100 in 2 ml LB medium and gr own aer obically at 37 • C until they reached an OD 600 of a pproximatel y 1. Next, the cultur es wer e diluted 1:100 in sterile PBS and 2 l of this bacterial suspension was mixed with the same volume of transforming DNA at a concentration of 50 ng / l (for pABA) or 100 ng / l (for gDNA or PCR products). 3 l of the mixture was subsequently spotted onto 1 ml of solidified transformation medium inside a 1.5 ml Eppendorf tube and incuba ted a t 37 • C f or 24 h. The next da y, 300 l of PBS was added and the Eppendorf tube vortexed at max. speed to resuspend the cells. The bacterial suspension was serially diluted and spotted in duplicate on agar plates containing the appropriate antibiotic (apramycin or rifampicin) to select for transformants. Dilutions were also spotted onto plain LB agar plates to estimate the total number of bacteria. After overnight incubation at 37 • C, transformation frequencies were calculated as the number of colony forming units (CFU) of the transformants divided by total number of CFUs. Data in graphs r epr esent the average of three biolo gicall y independent replica tes. W hen no transformants were detected, the values were set to the detection limit to allow calculation of the average and for statistical analysis using the software Prism (GraphPad; San Diego USA). For sta tistical analysis, transforma tion frequencies were log-transformed ( 60 ). Detailed specifications of statistical analyses are provided in the figure legends.
The gDNA samples that were used as transforming material in the transformation assays were purified using the Qiagen Genomic DNA buffer kit and Qiagen 100 / G columns. Purification was based on 2 ml of overnight-grown cultures, as described above, and the precipitated gDNA was dissolved in 10 mM Tris-HCl (pH 8.0) buffer.
The pABA plasmid was purified using PureYield ™ Plasmid Miniprep System (Promega), starting with 5 ml of an overnight grown culture. When indicated for gene induction, overnight cultures were grown with 2% arabinose. When specified, the pABA plasmid was methylated with Dam methylase (New England BioLabs), following the manufacturer's instructions. To test the transformation of nati v e (isolated from strains A118, E. coli TOP10, or E. coli INV110) versus in vitro methylated plasmids, 35 ng / l of pABA were used during the transformation assay.
PCR products that served as transforming material were obtained by amplification with PWO polymerase (Roche) or Q5 polymerase (New England BioLabs).

Quantitativ e rev erse transcription-PCR (qRT-PCR)
Overnight cultur es wer e back-diluted 1:100 into 2 ml LB medium (in duplicate for the 2 h samples) and grown aerobically for 2 or 6 h (as specified) at 37 • C. Next, the cultures were centrifuged for 3 min at max speed and 4 • C. Bacterial pellets were washed with 1 ml of sterile and cold PBS and resuspended in 1 ml of TRI Reagent (Sigma-Aldrich). After vortexing for 10 s, the samples were snap-frozen in a dry ice-ethanol bath and stored at -80 • C. Subsequent steps included RNA extraction, DNase tr eatment, r e v erse transcription and quantitati v e PCR (qPCR), which were performed as described pr eviously ( 61 ). Expr ession le v els of individual genes were normalized to the transcript levels of the housekeeping gene gyrA ( = relative expression) and the average was calculated from three biolo gicall y independent e xperiments. Relati v e e xpr ession values wer e logtransformed ( 60 ) for statistical analysis using the Prism softwar e (GraphP ad; San Diego, USA). The parameters of the statistical analyses are described in the figure legends.

Sour ce of tr ansf orming DNA affects tr ansf ormation in A. baumannii
To better understand how efficiently different A. baumannii isolates can acquire DNA by natural transformation, we compared the transformation frequency of four A. baumannii strains, using gDNA originating from six different A. baumannii strains (Table S4). Specifically, strains A118, AB5075, 29D2 and 86II / 2C served as DNA acceptors, since they were previously shown to be naturally competent ( 22 , 23 , 49 ). The donor gDNA, used as the transforming material, was purified from an apramycin cassette-containing (AprR) deri vati v e of each of the four transformable strains mentioned above, as well as two non-transformable strains, ATCC17978 and ATCC19606 (Table S4). To obtain AprRmarked donor strains, the apramycin resistance cassette was inserted into the hcp gene of each strain. We chose hcp , which encodes a part of the type VI secretion system (T6SS), as a neutral locus, gi v en that the T6SS is not involved in competence de v elopment or DN A uptake. Importantl y, the T6SS gene cluster is highly conserved in the different A. baumannii strains, which should facilitate homologous recombination during transformation. Indeed, the pairwise identity of the genomic region 10 kb upstream and downstream of hcp is higher than 95% for most of the compared strains (Fig. S1A). In contr ast, str ain ATCC17978 carries se v eral insertions upstream of hcp , resulting in a significantly decreased pairwise identity of less than 47% when compared to the other tested strains ( Figure S1A). Ne v ertheless, the region of a pproximatel y 1 kb immediatel y upstream and downstream of hcp remains highly conserved in all strains, with only a few single nucleotide polymorphisms (SNPs) when compared to strain A118, which served as the reference strain in this study ( Figure S1B).
Since the hcp chromosomal region is highly homologous between the tested strains --at least in close proximity to the AprR cassette --one might expect that the transformation frequencies should be similar when using gDNA from differ ent sour ces. Inter estingly, howe v er, we observ ed se veral significant differences in transformation frequencies in all tested acceptor strains when the donor gDNA had distinct origins (Figure 1 A). These differences appear to be mostly independent of sequence homology between the acceptor strain and the donor DNA. Strikingly, in strain A118 (Figure 1 B) transformation frequencies were 10-to 40-fold lower when the DNA originated from any of the non-self strains. Precisely, e v en though transformation le v els were the lowest ( ∼40-fold reduction compared to self-DNA) with donor gDNA deri v ed from strain ATCC17978, the transformation frequency was in the same range ( ∼10to The mean values ( ± standard deviation, SD) of three independent biological replicates are shown for panels B and D with each individual experiment being r epr esented as a dot. < d.l., below detection limit. #, below detection limit in at least one experiments. Statistics were performed on log-transformed data, using two-way analysis of variance (ANOVA) with Dunnett's (A, C) or Tukey's (B, D) correction for multiple comparisons. Only statistically significant differences are shown. * P < 0.05, ** P < 0.01, **** P < 0.0001. 30-fold reduction compared to self-DNA) using the other non-self gDNA sources, suggesting that the reduced pairwise identity between the two strains has only a minor negati v e effect. A slight decrease in transformation le v els when using ATCC17978-deri v ed gDNA in comparison to self gDNA was also observed when strains AB5075 and 29D2 served as DN A acceptors, w hilst transformation frequencies were comparable for strain 86II / 2C (Figure 1 A). A minor or no effect of sequence di v ergence on transformation frequency was somewhat expected, since recombination can occur between highly similar DNA segments of only 25-200 bp in length ( 22 ). Ther efor e, the highly similar regions either proximal or distant to the hcp gene might be sufficient to reach high le v el of transformation w hen gDN A from ATCC17978 was used as transforming material.
To exclude any potential effects of sequence homology when investigating the impact of the source of DNA on transforma tion ef ficiency, we repea ted the experiment us-ing the small (6.8 kb) replicati v e plasmid pABA as the donor DNA. pABA is a deri vati v e of the conventional plasmid pBAD18-Kan, wherein an Acinetobacter -specific origin of replication has been added to enable replication in A. baumannii ( 62 ) and the KanR resistance cassette ( aph ) has been replaced with an AprR resistance cassette ( aac(3)IV ). This plasmid was electroporated into strains A118, A TCC17978, A TCC19606, AB5075, 29D2, and 86II / 2C and subsequently repurified using standard plasmid mini-preparations. Next, we used these strainpassaged plasmids in a transformation assay with all the tr ansformable str ains as acceptors. Interestingly, we again observed significant DNA origin-dependent transformation differences in some acceptor strains (Figure 1 C). In strain A118 the DNA-source dependent selectivity was conserv ed irrespecti v el y of the DN A type. Indeed, transformation frequencies with non-self-originating plasmids were barely above detection limit, while transformation with the plasmid originating from the strain A118 itself was around three orders of magnitude higher (Figure 1 D). For the other natur ally tr ansformable str ains, the difference between the self-or non-self-originating plasmid was less pronounced, with the notable exception of strain 29D2 transformed with the plasmid isolated from ATCC19606 (Figure 1 C). Gi v en the strong DNA source-dependent effect on transformation w hen using gDN A and plasmids, we hypothesized that epigenetic DNA modifications, might either enhance and / or limit natur al tr ansformability in some A. baumannii str ains, such as A118. This could also be the reason for the discrepancies between DNA source-dependent transformation patterns w hen gDN A or plasmids were used as transformation material and could be explained by the absence of some specific DN A modification reco gnition sites on the pABA sequence (Table S5). This hypothesis is in line with previous work on H. pylori , Pseudomonas stutzeri , C. jejuni and Neisseria meningitidis where it was shown that methylated DNA motifs could affect transformation by either impacting the specificity of DNA uptake or the degradation of incoming DNA ( 35-37 , 39 , 40 ).

SMRT sequencing rev eals div erse DNA modification patterns among different A. baumannii strains
To understand whether the DNA source-dependent differences in transformation efficiency could be explained by strain-specific epigenomes, we sequenced the genomes of all tested A. baumannii strains, except previously sequenced strain A118 ( 50 ), using PacBio SMRT sequencing (Table S2). This technique not only determines the genome sequence, but also provides information regarding various DNA modifications such as N 6 -methyl-adenine (m6A), N 4 -methyl-cytosine (m4C) and N 5 -methyl-cytosine (m5C) (e v en though the latter modification r equir es additional conversion steps pre-sequencing or very deep sequencing efforts), with single-nucleotide resolution ( 63 ). Interestingly, as shown in Figure 2 A, all the sequenced strains exhibited a specific DNA modification pattern, mostly comprising se v er al methylation motifs. For all str ains, the identified DNA modifications belonged to 6-15 nucleotide long motifs, mostly carrying methylation marks on either adenine residues (8x) or on cytosine residues (1 ×). For 6 out of the 15 identified DNA modification motifs, only the modified residue could be determined, but not the type of the modification (labeled 'unk' for unknown in Figure 2 A). Se v eral of the identified DNA modification motifs were present in more than one strain, for example m6A:GAAAGC:4 (meaning a m6A modification on the fourth nucleotide of this motif) was common to strains AB5075 and 29D2. Other modifications were strainspecific, such as m6A:RGATCY:3 in strain A118. We observed different proportions of modified recognition sites relati v e to their abundance within each genome, which varied from a pproximatel y 20-100% (shown on a scale from 0 to 1 in the legend of Figure 2 A and in Table S6).
To date little is known about the epigenome of the genus Acinetobacter . A study identified a methylase in A. baumannii strain ATCC17978 --AamA, tha t methyla tes adenine at GAATTC motifs and was suggested to be involved in surface-associated motility ( 64 ). Even though aamA homologs were present in all the tested strains (Figure S2), we did not detect m6A:GAATTC:3 modification specifically in any of them. Howe v er, two similar motifs -m6A:TTTRAATTNA:6 and m6A:TNTNAATTNAAA:6, were partially methylated in all the tested strains (Figure 2 A  and Table S6) and could ther efor e be a result of the AamA methylase. Alternati v ely, AamA might not be produced under the tested conditions.
The sole report of a complete Acinetobacter methylome was performed in Acinetobacter calcoaceticus 65, for which se v eral DNA modification motifs were identified, belonging to either type I or type II RM systems ( 65 ). Ther e wer e no common DNA modification motifs between the reported methylome of A. calcoaceticus 65 and the methylomes of the A. baumannii strains determined in this study. This is not surprising gi v en the high variability that we already observed in the modification profiles between the different isolates. Furthermor e, as RM systems ar e dynamic and their genes are often acquired through HGT ( 29 ), such variability between members of the same genus or species is to be expected.
We considered two distinct hypotheses to explain how the variability of DNA modifications could impact the transformability of strain A118, whereby transformation frequency with self-DNA was higher than with non-self DNA. First, the presence of DNA modification on the donor DNA could lead to reduced transformation in A118 if, for instance, this DN A modification was reco gnized by a A118-specific type IV restriction enzyme. Type IV restriction enzymes, unlike the other types, degrade modified DNA ( 43 ). Howe v er, analysis of our modifica tion da ta showed that no specific DNA modification was shared between all the tested strains, but absent in strain A118 (Figure 2 A). We therefore considered a second possible scenario whereby an A118-specific DNA modification could either promote DNA uptake or pre v ent DNA degradation. The latter could be explained by the presence of an A118-specific RM system that would result in methylation and ther efor e protection of self-DNA, while promoting the degradation of the non-self DNA lacking these specific methylation marks by the corresponding restriction enzyme. Indeed, as mentioned in the introduction, postreplication transformants that incorporated heterologous non-self DNA are expected to contain stretches of unmethylated dsDN A, w hich are targeted by the respecti v e restriction enzymes ( 42 ), as ar e r econstituted dsDNA plasmids. In contrast, A118-deri v ed self DNA is not e xpected to occur in a fully unmethylated double-stranded form and is ther efor e r esistant to cleavage by such restrictases. Interestingly, we identified two A118-specific DNA modification motifs --m6A:RGATCY:3 and unk:TGGCCA:4 (which we characterized as m4C-TGGCCA:4 in further sequencing runs; see below) --which, as depicted in Figure  2 B and C, are randomly distributed throughout the entire genome. Since the recognition site TGGCCA is absent from the plasmid pABA (Table S5), we hypothesized that the m6A:RGATCY:3 DNA modification was likely responsible for the observed phenotype. As shown in Figure 2 B, nearly 100% of the RGATCY recognition sites were methylated on both strands of the A118 chromosome, suggesting that the responsible methylase is highly acti v e and could ther efor e be involved in efficient DNA protection.

Comparing methylase r epertoir e betw een differ ent A. baumannii isolates
To determine the possible candidates for m6A:RGATCY:3 DNA modification, we compared the annotated genomes of the six PacBio-sequenced A. baumannii strains. As shown in Figure S2, we identified a r epertoir e of genes encod-ing for proteins with DNA methylase-like activity in each tested strain. This enabled us to assess two A118-specific DNA methylases: H0N27 12600 and H0N27 10820 (protein names are based on their locus tags according to NCBI accession number CP059039), which could be responsible for the m6A:RGATCY:3 DNA modification. The H0N27 12600, is annotated as a DNA adenine methylase, but lacks an obvious cognate restriction enzyme(s) within its gene neighborhood. Thus, while we cannot exclude that a partner restriction enzyme might be encoded elsewhere Nucleic Acids Research, 2023, Vol. 51, No. 11 5669  on the genome, we consider H0N27 12600 to be an orphan methylase (r eferr ed to as MT2 throughout the text), which is unlikely to be responsible for the observed transformation phenotype. The second methylase H0N27 10820 is also annotated as DNA adenine methylase and is located close to a restriction endonuclease-encoding gene (H0N27 10830; Figure 3 A). This suggested that the methylase H0N27 10820 might be a part of type II restriction modification (RM) system, which are often involved in defense against foreign DNA ( 43 ). We therefore aimed to further characterize this putati v e RM system and the genes located in its neighborhood.
Interestingly, the potential RM system is encoded in a region of the genome that exhibits high genetic variability between different strains (Figure 3 A), suggesting that it might be a hot-spot for genomic integration. Gi v en the involvement of RM systems in bacterial defense against foreign DNA ( 43 ) we wondered whether other A. baumannii strains also carried genes encoding DNA defense systems in the same region, potentially forming a small defense island. We ther efor e inspected this r egion in our tested strains using the Prokaryotic Antiviral Defence LOCator (PAD-LOC) ( 66 ) and DefenseFinder tools ( 67 , 68 ). This approach identified se v eral genes encoding known anti-phage defense systems such as: Gabija / CRISPR-cas complex in strain AB5075, psyrT --belonging to a to xin-antito xin system in strain 29D2, and a retron defense unit in strain 86II / 2C (Figure 3 A). Strains ATCC17978 and ATCC19606 also harbored genes encoding for a re v erse-transcriptase and an AAA family ATP ase, r especti v ely, which are often associated with bacterial DNA defense (69)(70)(71). The genomic region specific to strain A118 encompasses the four genes H0N27 10820 to H0N27 10835, coding for the aforementioned RM system (DNA adenine methylase H0N27 10820 [MT] & restriction endonuclease H0N27 10830 [RE]) interspersed by a helix-turn-helix transcriptional regulator [TR]-encoding gene H0N27 10825 and followed by gene H0N27 10835, which was predicted by DefenseFinder to encode for AbiD, involved in abortive infection during antibacteriophage defense (Figure 3 A). Protein BLAST analysis of H0N27 10830 suggests that the putati v e RE belongs to the Endonuclease Bgl II / Bst YI superfamil y, w hich typicall y reco gnizes AGA TCT or RGA TCY. This strongly suggested that this RM system is likely responsible for A118-specific DNA modification m6A:RGATCY:3 and ultimately could explain the DNA source-dependent transformation phenotype.

The A118-specific RM system impacts the str ain's natur al tr ansf ormability
To determine whether the H0N27 10820-30 RM system (Figure 3 B) (henceforth referred to as RMC = restriction modification cluster) is indeed responsible for the DNA origin-dependent differences in transformability of strain A118, we designed a RE-deficient strain A118 H0N27 10830 ( RE) and compared transformation frequencies in the WT strain and the RE-deficient strain, using various types of donor DN A. Firstl y, we used gDNA deri v ed from different strains. As expected, no significant difference was observed when using self-DNA from A118 (Figure 4 A). Importantly, howe v er, when comparing the transformation frequencies of the RE-minus mutant to those of the WT strain, we observed a significant 20 to 40fold increase in transformation with the gDNA samples that were deri v ed from the for eign strains r epr esenting non-self DN A. Secondl y, w hen we used a PCR amplified fragment of hcp ::aprR as the donor DNA, which contains 1 kb flanking regions on either side and lacks any DNA modifications, we reproducib ly observ ed a slight, but significant increase in transformation in the RE-deficient mutant (Figure  4 B). Thirdl y, w hen plasmid pABA origina ting from dif ferent strains was used as the donor DNA, the DNA-origin effect was almost completely abolished in the RE-deficient strain (Figure 4 C). The DNA motif RGATCY, carrying methylation mar ks, likely responsib le for DNA selecti vity, was present in all the donor DNA types (Table S5). These results strongly suggest that the RE limits acquisition of foreign DN A independentl y of transforming DN A type. Interestingly, the effect of the RE on transformation was the least drastic for the PCR pr oduct, more pr onounced for the diverse genomic DNA samples, and strongest when plasmids were used as transforming material. These differences might be partially due to the different copy number of the resistance gene when using distinct DNA types, which was the highest for the PCR product and the lowest for the gDNA. While this could explain the modest RE-dependent effect observed with the PCR product, the severity of the RE impact on transformation with plasmid DNA might be rather due to the different mechanisms that are required to establish plasmid or linear homologous DNA in the recipient cell.
All types of transforming DNA enter the cell in a single stranded form, which is belie v ed to be inaccessib le to most restriction endonucleases, which instead act on ds-DNA ( 72 ). Furthermore, the incoming ssDNA is thought to be pr otected fr om degradation through the tight binding of the Ssb and DprA proteins ( 15 , 73 ). Upon entry into the cell, gDNA and PCR products are integrated into the genome via homologous recombination. The reconstitution of replicati v e plasmids, on the other hand, r equir es either import and annealing of two complementary ssDNA strands or the synthesis of the second strand, prior to circularization ( 20 , 21 ). Ther efor e, plasmid cleavage most probabl y occurs immediatel y after dsDN A is formed / synthesized and before it is protected by methylation. In contrast, it is not well understood at which integration step the linear transforming DNA is cleaved by restriction endonuclease, e v en though a postreplication targeting model was proposed as outlined above ( 42 ). This model suggested that after replication, heterolo gous DN A stretches remain unmethylated and ther efor e serve as ideal substrates for REmedia ted degrada tion, resulting in the dea th of the transformed cell ( 42 ). To test whether this could explain the effect of the RE on integrati v e DNA in A. baumannii , we used a homologous rpoB PCR product carrying a SNP that confers rifampicin resistance (Rif R ) as the donor DNA. As shown in Figure S3A, the RE had no effect on the transformation frequency under these conditions. Howe v er, it should be noted that compared to the hcp::aprR PCR fragment mentioned above (Figure 4 B), which contains four RGAT CY r ecognition sites, the rpoB PCR product only carried a single and distantly located RGATCY site relati v e to the Rif R -conferring SNP (Table S5). Hence, while our rpoB transformation assay does not allow a direct comparison between homologous vs heterologous sequences, the result is consistent with the idea that the presence of heterologous region on the integrati v e DNA is r equir ed for the RE-dependent effect on transformation, which was also observ ed in S . pneumoniae and N. meningitidis ( 40 , 74 ). Moreover, our data suggest that RM systems reduce acquisition of novel genes, yet it is reasonable to specula te tha t the transformation-dri v en loss of certain genetic regions might not be impacted.
Since the RE e xclusi v ely impaired transformation with non-self-deri v ed DNA, the logical assumption is that the MT protects self-DNA from the RE-mediated degradation by adding specific methylation marks. To test this assumption, we attempted to delete both the entire RMC gene cluster and the MT gene alone. Howe v er, in further support of a functional connection between the MT and RE, it was impossible to delete the MT alone, except in the presence of a spontaneous suppressor mutation in the RE (further r eferr ed to as MT RE* strain). This is likely because in the absence of the methylation marks, the host genome would become susceptible to degradation by the RE. Next, we purified plasmids from the RMC-or MT-minus strains and used them as donor DNA for the WT strain of A118. As shown in Figures 4 D and S3B, the transformation frequency decreased significantly when the plasmid originated from the strains lacking either the RMC ( RMC) or the MT ( MT RE*), confirming the protecti v e effect of this specific methylation on the transforming material if deri v ed from WT cells. In contrast, plasmids originating from the methylase-deficient strains that were complemented by an arabinose-inducible MT gene in cis (TnMT) were able to support transforma tion a t le v els comparab le to the WT control (Figures 4 D and S3B).
Additionally, to determine whether the expression of the A118-specific MT in a different strain background could increase the transformation potential of non-self-deri v ed DNA, we introduced the arabinose-inducible TnMT construct into the strain ATCC17978. By doing so, the transformation frequency was significantl y increased w hen we used plasmids deri v ed fr om this strain (gr own under MT-inducing conditions) as compared to either the  Table S8). Details as described for Figure 2. The average values ( ± SD) of three independent biological replicates are shown (A-E). < d.l., below detection limit. #, below detection limit in at least one replica te. Sta tistical analyses were performed on log-transformed data using a two-way ANOVA with Š id ák's correction for multiple comparisons (A, C), an unpaired t test (B), or a one-way ANOVA with Dunnett's correction for multiple comparisons (D, E). * P < 0.05, *** P < 0.001, **** P < 0.0001; n.s., not significant. parental strain or an empty transposon-carrying control (Figure 4 E). Interestingl y, w hile increased by se v eral or ders of magnitude compared to the controls, the transformation efficiency did not reach the same le v el as with A118-deri v ed plasmid DNA, which could be due to the presence of additional DNA modifications in this strain.
In contrast to the results above, no transformants were obtained with plasmid DNA purified from ATCC17978 e xpressing the arabinose-inducib le A118-specifc orphan methylase H0N27 12600 (TnMT2; Figure 4 E). This result suggests that MT2 does not play a DNA-protecti v e role during natural transformation. Howe v er, the absence of protection could also be linked to the fact that the gene's induction did not work properly in our experiments or that the methylase was non-functional. To address these possibilities, we SMRT sequenced strain ATCC17978-TnMT2 after arabinose-dependent induction of the MT2 gene. The results showed that the induction of MT2 resulted in the additional DNA modification --m4C:TGGCCA:4 ( Fig. S3C and Table S7). Importantly, this DNA modification was also present in strain A118 (Figure 2 A), yet the exact modifica tion (methyla tion) onl y became a pparent in additional sequencing runs. This ther efor e showed that MT2 is responsible for the m4C:TGGCCA:4 DNA modification but that it does not have a protective role against the RE. Collecti v ely, our results suggest that the RMC cluster-encoded MT is the only methylase that is responsible for efficient DNA protection in the RE-producing strain A118.
Finally, to confirm that the m6A:RGATCY:3 methylation is a result of the identified RMC, we performed PacBio SMRT sequencing to assess the DNA methylome of the RMC-related mutants of strain A118 (Table S3). As a control, we re-sequenced the WT A118 strain, to which we refer as WT(r eSEQ) (pr eviously sequenced version of strain A118 r eferr ed as WT). The DNA methylomes of both sequencing versions of the WT were slightl y different, w hich might be explained by the differences in gDNA preparation and sequencing machines (see Materials and Methods). Howe v er, as shown in Figure 4 F and listed in Table S8, the deletion of the RE did not affect DNA modifications, as they were almost identical to those of the WT(reSEQ). In contrast, the DNA modification m6A:RGATCY:3 was no longer identified in the strains lacking either the entire RMC or the MT plus the RE* ( Figure 4 F and Table S8). This confirmed that the identified DNA modification was indeed the result of the MT within the RMC. Interestingly, Nucleotide BLAST analysis of the A118 RMC identified similar or identical RM systems in se v eral A. baumannii strains, including Ab42, ARLG-1772, UH6507, etc. as well as in other members of the Acinetobacter genus (Table S9). Futur e studies ar e ther efor e needed to test the impact of the respecti v e RM systems on these strains' transformability.

A helix-turn-helix regulator controls methylase expression
Regulation of RM systems frequently relies on conserved helix-turn-helix transcriptional regulators, which control the homeostasis between the restriction endonuclease and the methylase (75)(76)(77)(78)(79). Since the RMC also carried the same type of transcriptional regulator (H0N27 10825; TR), we aimed to understand its role in the identified system. Gi v en its putati v e function, we first checked the effect of deleting the TR gene on the transcript le v els of the RE and the MT during exponential growth (2h growth post-dilution). Interestingly, deletion of the TR gene resulted in a ∼50-fold increase of the MT gene expression (Figure 5 A). No significant differences in RE transcript le v els were detected in the TR strain compared to its parental WT strain (Figure 5 A). These results indicated that the TR likely serves as a r epr essor of the MT gene, suggesting that the absence of this regulator could result in enhanced le v els of methylation. To test this, we assessed the DNA methylome of the TR-deficient strain by SMRT sequencing. Interestingly, the DN A methylome changed significantl y in the TR-deficient strain as compared to the WT(reSEQ). Se v eral ne w DNA modifications appeared, all of which included a 'GATC' cor e sequence (Figur e 5 B and Table S10). These results suggest that the TR controls both the abundance and as a result, the specificity of the methylase.
Gi v en these results, we hypothesized that over-abundance of the methylase might lead to inappr opriate pr otection of incoming non-self DNA and ther efor e r educe the selectivity for self-DNA. We ther efor e compar ed the transformability of the TR-minus strain with the WT and se v eral REdeficient deri vati v es. As transforming material, we used either plasmid DNA or gDNA originating from strain A118 ('self') or ATCC17978 ('non-self'). As expected, all strains were transformed equally well by self-DNA ( Figures 5 C and  S4A). Howe v er, for the non-self plasmid DNA, we observ ed dif ferent transforma tion pa tterns depending on the acceptor strain. Indeed, the transformation frequency was under the detection limit in the WT strain but increased dramatically in all other strains, irrespecti v e of those being REproficient or RE-deficient (Figure 5 C). Strikingly, transformation by the non-self plasmid was also clearly detectable for the TR-minus strain (Figure 5 C). A similar pattern was also observed when non-self gDNA from ATCC17978 was used as transforming material, with intermediate transformation le v els for the TR-deficient strain ( Figure S4A). Taken together, these results suggest that overabundance of the methylase can result in (partial) protection of foreign DNA. Moreover, our data are in line with the idea proposed by Johnston et al. ( 42 ) who suggested that a race between the restrictase and the methylase for access to unmethylated DNA occurs after chromosome replication (or after reconstitution of the dsDNA of plasmids).
To confirm that the observed phenotype is caused by the TR, we performed transformation assays with the WT, TR strain, and a complemented version, wherein the TRencoding gene (H0N27 10825) was placed under the control of its nati v e promoter at the hcp locus, along with a gentamicin resistance cassette (GentR) (for selection during strain construction). The resulting strains were tested for their transformability using foreign plasmid DNA deri v ed from strain ATCC17978. Consistent with the results above, transformation fr equencies wer e close to the detection limit for the WT control strain that carried the GentR cassette within hcp ( Figure S4B). Howe v er, in the absence of the TR, transformants were readily detectable. This effect was re v ersed in the strain containing the complemented TR gene ( Figure S4B). These results confirm that overproduction of the methylase in the TR strain was  also Table S10). Details as described for Figure 2. ( C ) Transformability of strain A118 and its RMC-related deri vati v es using self and non-self plasmid DNA (isolated from A118 or ATCC17978) as transforming ma terial. Transforma tion frequencies of RMC-related deri vati v es ar e compar ed to the WT within each set that tested the same plasmid source. Only statistically significant differences are shown. ( D ) Relati v e e xpression of the MT , RE , TR, and pilA genes in strain A118 after 2 h and 6 h of growth. ( E ) Transformation frequency of strain A118 with plasmid pABA originating from strains A118, E. coli TOP10 and E. coli INV110. Dam methylase-dependent in vitro methylation was performed where indicated (+methyl). The mean values ( ±SD) were calculated from three independent biological replicates (A, C-E). < d.l., below detection limit. The data was log-transformed and statistical differences were calculated using a two-way (A, C-D) or one-way (E) ANOVA corrected for multiple comparisons with Š id ák's method. **** P < 0.0001; n.s., not significant. indeed caused by the absence of the TR and was not a result of a polar effect on the MT gene.
Gi v en the observation that the MT transcription is tightly controlled, we examined whether the MT, RE, and TR le v els vary during different growth phases. Indeed, since we previously observed that some of the essential competence genes --e.g. type IV pili (T4P) genes --are only produced during exponential growth ( 50 ), we wondered whether methylation le v els are also increased specifically during this 'transformation window', potentially to facilitate the acquisition of new genetic traits from bacteria lacking 'GATC'-like methylation mar ks. Howe v er, in contrast to the major T4P gene pilA , our results indica te tha t all the RMC genes are stably expressed in exponential and stationary growth phase (Figure 5 D). This suggests that A118 does not adapt its defense mechanism against foreign DNA during the transformation process. This in contrast to S. pneumoniae , for which the ssDNA methylase DpnA was shown to be specifically expressed during competence so that incoming transforming DNA was pr otected fr om degradation by an endogenous RM system ( 74 , 80 ). Ne v ertheless, we cannot exclude that external factors might result in the increased MT le v els in strain A118 under so far unknown conditions, which could facilitate the uptake of foreign DNA during high demand for new genetic traits, for instance, under harsh environmental conditions.

Escherichia coli dam methylation protects DNA from RMCdependent degradation
Our data so far are consistent with the DNA modification m6A:RGATCY:3 being responsible for the DNA-origindependent transformability phenotype of strain A118 (Figur e 2 A). Furthermor e, the RE we have identified within the RMC belongs to the Bgl II / Bst YI endonuclease protein famil y, w hich generall y degrades AGATCT or RGATCY. Interestingly, a common feature of these motifs is the 'GAT C' cor e, a DNA sequence that is also recognized by the orphan methylase Dam ( 81 ). The Dam methylase was originally discovered in E. coli ( 81 ) and is present in many other Gamma-proteobacteria. It is, howe v er, absent in the genus Acinetobacter . Dam r ecognizes GAT C and adds a methyl group onto the adenine. We ther efor e hypothesized that Dam methylation could serve to protect non-self DNA during transformation of strain A118. To test this idea, we purified pABA from E. coli strains that were either dam + (TOP10) or dam -(INV110) and used them as donor DNA in transf ormation assa ys. Notably, pABA purified from the dam + E. coli TOP10 supported efficient transf ormation, whereas transf ormation was below the detection limit ( ∼10 −7 ) when the plasmid was purified from the damstrain INV110 (Figure 5 E). To confirm that the methylation status was causing this transformation difference, we performed an in vitro assay whereby we methylated plasmid DNA deri v ed from E. coli strain INV110 (or A118 and E. coli TOP10 as controls) with commercially available Dam methylase (New England BioLabs) and tested whether it affected plasmid transformation in strain A118. Strikingly, the in vitr o methyla tion r ender ed the plasmid from E. coli strain INV110 fully transformation-competent leading to transformation frequencies comparable to those obtained with A118-deri v ed plasmid DNA (Figure 5 E). In contrast, when the plasmid was deri v ed from either A. baumannii strain A118 of the dam -proficient E. coli strain TOP10, in vitr o methyla tion did not further increase the plasmid's ability to serve as transforming material (Figure 5 E). These results demonstra ted tha t Dam methyla tion can ef ficiently pr otect DNA fr om degradation by the RMC-associated RE. Interestingly, according to REBASE restriction enzyme database (re base.ne b.com / re base / ) this is not an efficient protection against the commercially available restriction enzymes Bgl II and Bst YI, as Dam methylation does not have an inhibitory effect on their DNA digestion activity.
Taken together, these results strongly suggest that the RMC could play an important role in blocking DNA exchange from different bacteria lacking 'GAT C'-r elated methylation mar ks. Howe v er, our results also demonstrate that Dam methylation can bypass the RMC-dependent effect. Interestingly, according to the REBASE database, many ESKAPE pathogens, such as Pseudomonas aeruginosa , Klebsiella pneumoniae and Enterobacter spp., carry Dam or Dam-related methylation motifs, making them a possible reservoir of resistance genes for A. baumannii strains with modules similar to the described RMC.

CONCLUSION
In this study, we described a RM-dependent mechanism that significantly affects intra-and interspecific DNA exchange with competent A. baumannii strain A118. We showed that the DNA methylome is variable between the tested A. baumannii strains, which was partially reflected in the ability of the strains' DNA to serve as efficient transfor ming material. Further more, we identified a RM system in strain A118, which fostered the degradation of foreign DNA that lacked specific epigenetic marks. This resulted in indirect enhancement of transformation with kin-released DNA, suggesting that the presence of the RM systems in competent cells can reduce the le v el of transformationdri v en genome di v ersification. Indeed, for strain A118, transforming DNA deri v ed from foreign strains will be less efficiently incorporated compared to self-DNA thereby likely reducing the acquisition of novel genetic traits. While DNA acquisition is considered one of the primary roles of natur al tr ansformation, another role of transformation, namely genome curing from mobile genetic elements ( 82 ), might remain highly efficient. The primary role of RM systems is considered to be the protection against invasion of plasmids and bacteriophages ( 83 ). Interestingly, the studied RM system was found in a genomic region that varied significantly between different A. baumannii strains, but ubiquitously carried genes involved in defense including antiphage systems. The effect of the identified RM system on transformation might ther efor e be a consequence of antiphage or anti-plasmid strategies, originally aiming to prevent phage infection and preserve genome content.
Howe v er, our data suggest that the RM-dependent effect on transformation could be avoided to some extent when the le v el of the methylase was increased in the cell. While purely speculati v ely at this point, such der epr ession of the methylase gene and ther efor e facilitated acquisition of non-self DNA, as we witnessed in the TR mutant, could occur under stressful environmental conditions when new genetic features could serve as a bet-hedging strategy enab ling bacterial survi val. Additionall y, transforming DN A carrying overlapping Dam methylation marks was not impacted by the presence of the RM system suggesting a pr efer ence for DNA originating from bacterial species that carry similar epigenetic marks.
These findings could foster future research that aims at understanding the reservoir of horizontally-acquired DN A fragments , including the frequentl y observed chromosomally-located AMR genes of di v erse A. baumannii isolates. Since A. baumannii is one of the most threatening multidrug resistant pathogens, it is crucial to understand how it acquires dangerous genetic traits.