Comparison of CcrM-dependent methylation in Caulobacter crescentus and Brucella abortus by nanopore sequencing

Abstract

Bacteria rely on DNA methylation for restriction-modification systems and epigenetic control of gene expression. Here, we use direct detection of methylated bases by nanopore sequencing to monitor global DNA methylation in Alphaproteobacteria, where use of this technique has not yet been reported. One representative of this order, Caulobacter crescentus, relies on DNA methylation to control cell cycle progression, but it is unclear whether other members of this order, such as Brucella abortus, depend on the same systems. We addressed these questions by first measuring CcrM-dependent DNA methylation in Caulobacter and show excellent correlation between nanopore-based detection and previously published results. We then directly measure the impact of Lon-mediated CcrM degradation on the epigenome, verifying that loss of Lon results in pervasive methylation. We also show that the AlkB demethylase has no global impact on DNA methylation during normal growth. Next, we report on the global DNA methylation in Brucella abortus for the first time and find that CcrM-dependent methylation is reliant on Lon but impacts the two chromosomes differently. Finally, we explore the impact of the MucR transcription factor, known to compete with CcrM methylation, on the Brucella methylome and share the results with a publicly available visualization package. Our work demonstrates the utility of nanopore-based sequencing for epigenome measurements in Alphaproteobacteria and reveals new features of CcrM-dependent methylation in a zoonotic pathogen.

Introduction:

DNA methylation is a critical epigenetic modification where motif-specific addition of a methyl group to adenine or cytosine can affect gene regulation, structure, or properties (¹). In bacteria, the methyltransferases responsible for this process are grouped into two main types: restriction-modification systems and orphan methyltransferases. While restriction-modification systems harbor DNA methyltransferases along with associated restriction enzymes that serve to protect the bacteria’s own DNA while degrading foreign DNA (^2,3), orphan methyltransferases lack associated restriction endonucleases. Epigenetic marks laid down by orphan enzymes play crucial roles in controlling transcription in cellular processes such as initiation of DNA replication and development (^4,5 ).

In the dimorphic alphaproteobacterium Caulobacter crescentus (Caulobacter), the ccrM gene encodes an orphan cell cycle-regulated methyltransferase. CcrM generates N6-methyladenine (m6A) marks in GANTC motifs enriched in intergenic regions that often lead to changes in transcription of downstream genes. For instance, methylated GANTC sites in the dnaA promoter (⁶) or hemi-methylated GANTC sites in the ctrA promoter (⁷) increase the expression of these genes. Notably, CcrM is cell cycle regulated, with activity peaking at end of the S-phase of the cell cycle and then lost through rapid degradation by the Lon protease (^6,8 ). Thus, newly synthesized DNA remains hemi-methylated during replication and full methylation occurs once the chromosome has fully replicated as CcrM activity peaks, resulting in a steady increase in GANTC methylation measuring from origin to terminus in asynchronous populations (^9,10). Upregulation of CcrM activity due to overexpression (¹⁰) or stabilization (⁸) results in loss of the characteristic cell cycle GANTC methylation and an overall increase in methylation at CcrM-dependent sites. More recently, additional regulation of methylation has been proposed from findings that regions of hypomethylation seem to arise from loss of CcrM access to GANTC sites due to competitive binding of factors such as MucR (¹¹).

Several outstanding questions emerge from these observations. For example, regulated removal of methylation marks by demethylases alters epigenetic status in eukaryotic genomes (¹²); however, demethylation of CcrM deposited methylation marks has not been shown. One particularly intriguing candidate for a possible demethylase is AlkB, a hydroxylase responsible for repair of alkylated DNA during damaging conditions (¹³ ). In Caulobacter, alkB expression is under cell cycle control, suggesting a possible role in this process (¹⁴). In vitro, E. coli AlkB is capable of removing m6A marks from DNA templates (¹⁵), leading to a tempting speculation that AlkB may contribute to Caulobacter genome methylation, perhaps during the cell cycle. In addition, although CcrM and Lon are highly conserved among Alphaproteobacteria, whether the model that Lon degradation of CcrM controls its activity as seen in Caulobacter extends to all Alphaproteobacteria remains unclear. One important example is in the zoonotic pathogen Brucella abortus (Brucella), where both CcrM and Lon were found to be modulators of physiology and virulence (^16,17), but to our knowledge no studies on methylation in this bacteria have been reported.

In this paper, we address some of these questions using nanopore-based sequencing to measure global methylation. We first establish that machine-based learning models and nanopore reads can accurately measure cell cycle dependent methylation in Caulobacter. We show that loss of Lon results in pervasive elevated methylation across all sites, as would be expected from the known elevated levels of CcrM in ∆lon strains and correlate these effects with known transcriptional changes. We test if AlkB controls methylation and while we find no global change, a handful of sites exhibit minor methylation marks that in at least one case alters gene expression level. Next, we profiled the global methylome of Brucella and found that both chromosomes show signatures of cell cycle and CcrM-dependent methylation. Interestingly, loss of Lon results in chromosome-specific loss of cell cycle methylation and global methylation of both chromosomes is lower, in contrast to the Caulobacter case. Finally, we address the impact of the MucR transcription factor on global methylation in Brucella.

RESULTS

In Caulobacter, cell cycle regulated CcrM-dependent methylation controls DNA replication and gene expression (^18,19). The prevailing model (Figure 1B) is that replication of the fully methylated chromosome initiates during the swarmer-stalk transition with newly synthesized DNA strands remaining unmethylated throughout replication (^9,10). This process results in the genome transitioning from a fully methylated state to a hemi-methylated state as the replication fork progresses. Once replication is complete, CcrM activity surges due to increased expression and stabilization from Lon degradation resulting in full methylation of the chromosome (⁹). As a consequence, GANTC sites located nearer to the origin of replication are likely to remain hemi-methylated for a more extended period than those closer to the terminus.

Figure 1:

(A) CcrM facilitates the transfer of a methyl group from S-Adenosyl methionine (SAM) to adenine bases resulting in the formation of N-6 methyladenine and S-adenosyl-L-homocysteine (SAH) (⁴⁷). Lon protease degrades CcrM, thereby restricting its activity to the predivisional cell state. The dealkylase AlkB could play a role in the removal of the m6A epigenetic mark as described in vitro (¹⁵). (B) The current model of chromosomal m6A CcrM-dependent methylation during replication. Newly synthesized DNA is unmethylated, resulting in a progressive transition into a fully to hemi-methylated state comparing origin to terminus until the expression of CcrM near the end of S-Phase. (C, D) Probability of CcrM-dependent methylation at GANTC motifs across the chromosome of Caulobacter in both wild-type (C) and ∆lon (D) strains. Lines are linear fits to methylation probability assuming symmetry over the terminus. (E) Scatter plot of probability of methylation at intergenic sites in ∆lon and wild type strains, with red dots marking sites with both significant transcriptional changes and deviating from the mean by 2.5σ (CCNA designation in legend; highlighted motifs listed in Table S4). Histogram inset, displays the distribution of differences between methylation of ∆lon and wild-type with the dashed line indicating the line of identity and bold line displaying mean difference in methylation (∆ = 0.043).

Past global methylation studies of Caulobacter have used PacBio-based SMRT-Seq (^9,10) and methylation-specific restriction enzymes (REC-Seq¹¹). In this current work, we take advantage of nanopore sequencing combined with machine learning pipelines developed to call modifications of bases, including methylation. We first sequenced and called the methylome for wild-type Caulobacter using ONT v9.5 pores and mCaller (²⁰). Our findings corroborated the expected methylation pattern, as GANTC methylation is low at the origin and gradually increases to maximum at the terminus, consistent with cell cycle dependent methylation and an asynchronous population of cells (Figure 1C). Our results align well with prior studies that employed SMRT sequencing (¹⁰), with strong correlations comparing average methylation across sections of the chromosome and somewhat weaker correlation between individual sites (SFig 1). Importantly, m6A marks that are not GANTC motifs, and therefore not subject to CcrM control, exhibit an even and consistently high methylation across the genome, supporting that CcrM is the only cell cycle regulated adenine methyltransferase (SFig 4). We conclude that methylome detection using nanopore sequencing can be used effectively for Caulobacter genomes.

Loss of Lon globally affects CcrM-dependent methylation in Caulobacter

CcrM was the first substrate known to be degraded by the Lon protease in Caulobacter (⁸) and loss of Lon results in increased CcrM levels with increased methylation observed at individual sites (^8,21). However, there has been no global measurement of methylation in ∆lon Caulobacter. We now address this gap by using nanopore sequencing. Consistent with a critical role of Lon degradation in controlling CcrM levels and timing, we find that the cell cycle-dependent variation in GANTC methylation is lost in ∆lon and methylation across the entire genome is elevated (Fig 1D). Again, this was limited to m6A methylation of GANTC as non-GANTC m6A marks remain the same as wild-type (SFigure S4). Based on this, Lon-dependent degradation of CcrM clearly plays a significant role in controlling CcrM activity globally.

Methylation of promoter regions is thought to affect transcription of the neighboring genes (^10,11,19,22 ). We tested if this increase in methylation of promoter regions in ∆lon is correlated with changes in transcription in this strain. Using a cutoff of 300 base-pairs upstream of annotated genes as the promoter region, we performed a Fisher exact test comparing significant changes in gene expression between wild-type and ∆lon (²³) to changes in methylation of those promoter regions. This reveals no significant enrichment (p= 0.206).

However, among the sites exhibiting methylation probabilities that diverged by more than 2.5 standard deviations from the adjusted mean in the ∆lon strain, five genes showed significant transcriptional alterations (CCNA_00169, CCNA_00681, CCNA_03826, CCNA_00157, and CCNA_00084; labeled in order as a-e in the figure) with differing numbers of CcrM-dependent methylation sites in the promoter regions (Figure 1E). We conclude that although there is not a global change in transcription directly attributable to changes in promoter methylation, some genes may be more susceptible than others.

Loss of AlkB does not globally affect methylation

There is no in vivo evidence of demethylase activity to counter m6A modifications in bacteria. However, given that AlkB homologs have been shown to be biochemically proficient for removal of these modifications in vitro (¹⁵) we addressed whether AlkB affected CcrM-dependent methylation in Caulobacter using the same nanopore approach as before. There was a high correlation between methylation of ∆alkB and wild-type strains at all GANTC sites (Pearson coefficient 0.95). Differences in GANTC methylation across all the sites exhibited a normal distribution centered at zero (Anderson-Darling test), suggesting no significant global variation of methylation. We also tested whether AlkB might be affecting ‘spurious’ methylation of non-GANTC m6A sites, but found no significant difference in those comparisons either (SFigure 4).

Interestingly, one GANTC site that shows a higher degree of methylation in ∆alkB than expected (>2.5σ) is in the ccrM promoter region. This motif has been shown in past studies to have a minor influence on expression of ccrM (²⁴). Based on western blots, we found that CcrM levels are approximately two-fold higher in ∆alkB strains compared to wild-type in early exponential phase (Figure 2) but reach the same final levels as cells enter stationary phase. We conclude that AlkB does not play a global role in genome methylation under standard lab conditions but may have a gene-specific impact as shown with ccrM.

Figure 2:

(A) Probability of methylation at GANTC motifs across the chromosome of ∆alkB Caulobacter. (B) Scatter plot comparing difference in probability of methylation at intergenic motifs between ∆alkB and wild-type Caulobacter, with points highlighted in red marking motifs that deviate from the mean by 2.5 standard deviations (List of highlighted motifs are available in Table S5). Histogram inset, displays the distribution of differences between all ∆alkB and wild-type motifs with the dashed line indicating the line of identity and mean. (C) Western blot of CcrM protein levels in wild-type (black/top), ∆alkB (purple/ middle), and ∆lon (blue/bottom) during growth. (D) Quantification of CcrM protein levels after 3 hours of growth (OD600 0.2), in wild-type (black) and ∆alkB (purple) normalized to ClpP protein levels. (**) P-value of 0.002 by Student’s t-test.

CcrM-dependent methylation in Brucella is cell cycle regulated

We next used a nanopore platform to determine the methylation status of the Brucella genome. In this case, newly released technology and pipelines from Oxford Nanopore Technology were used (²⁵) for methylation detection due to changes by the sequencing vendor. We used wild-type, CcrM overexpressing (¹⁷), ∆lon (¹⁶), and ∆mucR (¹⁸) strains of Brucella to explore the molecular drivers of the methylome for this zoonotic pathogen.

Methylome measurements of wild-type Brucella reveal an increase in m6A at GANTC sites progressing from origin-proximal to terminal regions for both chromosomes (Fig 3 A), suggesting cell cycle dependent methylation. We reasoned that this was due to CcrM activity similar to Caulobacter as only m6A marks associated with GANTC motifs show this progressive change (SFig 5) and overexpression of CcrM results in increased methylation across the genome only at these cell cycle regulated GANTC sites (Fig 3A, B). Both chromosomes in Brucella showed the same degree of change in GANTC methylation suggesting that even though Chromosome II initiates replication after Chromosome I (²⁶) both chromosomes are subject to the same cell cycle regulation by CcrM under normal conditions.

Figure 3:

(A) Probability of CcrM-dependent methylation at GANTC motifs on Chromosome I (upper graphs) and Chromosome II (lower graphs) of Brucella in different genetic contexts: wild-type (black), with CcrM overexpression (green), and deletion of lon (blue). (B) Distribution of differences in strains where CcrM is overexpressed (green) or ∆lon (blue) verse wild-type, across all GANTC sites and among both chromosomes (left: Chromosome I, right: Chromosome II)

We next explored the role of the Lon protease in CcrM dependent methylation in Brucella. Given the high conservation of both proteins (65–70% identity), we envisioned similar regulation. Surprisingly, we found substantial Lon-dependent methylation differences between these species. Specifically, Chromosome I showed the expected loss of cell cycle dependent methylation when lon was deleted, but Chromosome II did not (Figure 3A). In addition, upon loss of Lon both chromosomes exhibited lower methylation overall compared to the wild-type strain (Figure 3B). This contrasts the Caulobacter case where elevated CcrM-dependent methylation is seen when lon is deleted (Fig 1). The interpretation of these results is explored in the discussion, but it is clear that the link between Lon and CcrM is more complex in Brucella.

MucR modulates local methylation in Brucella

In Caulobacter, the MucR transcription factor regulates developmental transitions (²⁷) and occludes GANTC motifs methylated by CcrM (¹¹). MucR also has important roles for virulence in Brucella (²⁸) and recent ChIP-Seq/Hi-C experiments reveal that MucR may also play a role in structural organization of chromosomes (¹⁸). We investigated the role of MucR in methylation of Brucella chromosomes by measuring the methylome of a ∆mucR strains. Overall, there was no global difference in methylation between wild-type and ∆mucR genomes across either chromosome (Figure 4A). However, when we specifically focused on MucR binding sites (¹⁸) we found that some overlapping sites had unusually high methylation in the ∆mucR strain compared to wild-type (Figure 4A). Interestingly these sites are part of a cluster of 18 hypomethylated sites in wild-type genomes (Figure 4B). Of those hypomethylated sites, those that overlap with MucR binding sites show high methylation in ∆mucR while those that do not overlap with MucR binding sites are not affected (Figure 4C; Table S6). We conclude that MucR shields a subset of hypomethylated sites from methylation by CcrM, similar to the Caulobacter system, but does not universally protect binding sites from methylation.

Figure 4:

(A) Density plot (gray) depicting the distribution of differences in methylation probability between ∆mucR and wild-type strains at all GANTC motifs. Green histogram shows sites that overlap with MucR binding sites, with inset showing highest changing sites. (B) Scatterplot of methylation probability at GANTC sites across both chromosomes of ∆mucR and wild-type Brucella. Hypomethylated sites in the wild-type overlapping with MucR binding sites are highlighted in green, while those not overlapping are in red. (C) Box plots illustrating changes in methylation probability between ∆mucR and wild-type for sites that are hypomethylated in wild-type with (green) and without (red) overlap with MucR binding regions, along with all sites (gray). Student t-test shows p-value as 6.0 e-6 (***), 1.7e-10(***) and no significant difference (N.S).

Discussion

Epigenetic modification of bacterial chromosomes was once considered only for restriction-modification systems used to protect cells from phage attack. The identification of orphan methyltransferases that modify chromosomes during development suggested that these modifications could also serve regulatory roles, much like CpG methylation in eukaryotes. Bisulfite-sequencing is a common method for detecting cytosine methylation in eukaryotic genomes; however, the primary regulatory modification in bacteria is adenine methylation, which cannot be converted in the same manner as methyl-cytosine. Use of single-molecule techniques such as Pacific Biosciences SMRT sequencing platform has resulted in direct observation of DNA methylation and application of this to bacterial genomes has made inroads into this regulatory mechanism. However, the costs associated with this approach make it challenging to do routinely. In this work, we take advantage of machine learning pipelines and increasingly inexpensive nanopore sequencing to measure the methylome of Alphaproteobacteria.

Our initial validation with Caulobacter showed the characteristic cell cycle dependent methylation of GANTC sites due to restricted activity of the CcrM methyltransferase only to the end of the replication cycle. The correlation with previous measurement with SMRT sequencing approaches was very high (SFig 1), supporting this approach. We then tested the assumption that loss of the Lon protease results in dysregulation of CcrM activity because Lon normally degrades CcrM. Consistent with this model, ∆lon strains fail to show the same cell cycle regulation of GANTC methylation and their genomes are generally more methylated. Next, we examined if AlkB, a dealkylase capable of removing m6A marks from DNA in vitro (¹³), affected methylation in vivo. We found no evidence for a global role of AlkB in DNA methylation in vivo under normal growth conditions, but identified some sites that were differentially methylated in an ∆alkB strain that could impact gene expression, as shown with CcrM expression (Figure 2).

For Caulobacter ∆lon, we found a handful of sites with particularly substantial changes in methylation and were also altered in transcription in this strain. The gene CCNA_00169 has multiple GANTC sites in its promoter region all of which were hypermethylated in ∆lon and showed an increased transcription in ∆lon. Although this gene is annotated as a hypothetical lipoprotein, it is adjacent to a TonB receptor and next to a cluster of Type 2 secretion genes, suggesting that CCNA_00169 may be a Type 2 secreted substrate (²⁹). Loss of this gene has a general fitness defect across multiple conditions based on bar-seq results and is highly conserved (³⁰). Interestingly, this same region is normally kept hypomethylated due to protection by MucR ( ¹¹), supporting a possible link between Lon and MucR in Caulobacter either directly or due to stabilization of CcrM.

Using this nanopore approach, we measured the methylome of Brucella abortus for the first time that we know of. As would be predicted from the conservation of CcrM, we find that GANTC methylation shows the same cell cycle dependency as Caulobacter. This dependency is found in both chromosomes of Brucella. Interestingly, loss of Lon protease in Brucella has species-specific effects on CcrM-dependent methylation. Cell cycle dependency of methylation was lost in Chromosome I in ∆lon, much like that seen in Caulobacter; however, Chromosome II showed the same methylation pattern as in wild-type. Additionally global GANTC methylation was lower in ∆lon than in wild-type for both chromosomes, in contrast to our expected findings in Caulobacter (Figure 3), leading to the possibility that steady state levels of CcrM may be overall lower in a ∆lon strain. We speculate that CcrM cell cycle regulation must be altered for Chromosome I to be affected, consistent with a role of Lon in regulating CcrM (^8,21). The Lon-independence of this methylation in Chromosome II suggests an additional layer of control where another methyltransferase may be present to enforce this mark or that the pool of CcrM associated with Chromosome II is still degraded even in the absence of Lon. Given that Lon is important for Brucella virulence (^16,31) this complex connection with CcrM methylation will be a future topic of interest.

The transcription factor MucR binds to Brucella’s chromosomes in an H-NS like manner, is recruited to AT-rich regions, and may contribute to nucleoid structure (¹⁸). Our findings suggest that MucR also protects specific hypomethylated sites from methylation by CcrM. One of these sites on Chromosome II is adjacent to BAB2_1107, a known autotransporter important for adhesion to host cells (³²). Of note, is that loss of MucR affects Chromosome II only in short-range interactions, versus its more global effect on Chromosome I (¹⁸). Similarly, loss of Lon affects Chromosome I methylation more dramatically (Figure 3) suggesting a possible link between MucR and Lon. Finally, an equal number of hypomethylated sites remain protected in ∆mucR and are not bound by MucR (Figure 4), suggesting that other MucR-like proteins may shield those sites. However, overall, loss of MucR does not dramatically affect the global methylation patterns of either of the chromosomes.

Thanks to new machine learning pipelines such as mCaller (²⁰) or Dorado (²⁵), together with the ever decreasing costs of nanopore sequencing, detection of epigenetic modifications can become more standardized. In our work presented here, we focused on two Alphaproteobacteria genomes and monitor how known or predicted regulators of methylation affect their epigenome. However, we note that the increasing adoption of nanopore sequencing of plasmids by contract companies (Plasmidsaurus, etc.) will make monitoring epigenetic changes in these genetic elements an unexpected bonus.

Methods

Bacterial growth conditions and sequencing

Refer to Supplementary Table 1 for all strains used in this paper.

Caulobacter crescentus NA1000, along with its derivative strains, were initially cultured on Peptone Yeast Extract (PYE) agar plates. These plates contain a specific nutrient mix: 2 g/L of peptone, 1 g/L of yeast extract, 0.5 mM of MgSO4, 0.5 mM of CaCl2, and 1.5% agar (produced by Difco), and were incubated overnight at 30°C. Following incubation, isolated colonies were picked and further cultured in PYE liquid media overnight at 30°C within shaken flasks. The cultures were then diluted to achieve a starting optical density (OD600) of 0.1, and allowed to grow to a mid-exponential phase (achieving an OD600 of 0.6) prior to the isolation of chromosomal DNA.

Brucella abortus 2308 and derivative strains were grown on Schaedler agar (Becton, Dickinson, and Co., Sparks, MD, USA) supplemented with 5% defibrinated bovine blood (SBA) incubated at 37°C under 5% CO2 or in Brucella Broth (Becton, Dickinson, and Co., Sparks, MD, USA) incubated at 37°C with shaking. Growth media were supplemented with kanamycin (45 ug/mL) or chloramphenicol (10 ug/mL) when appropriate. For isolation of chromosomal DNA, Brucella abortus 2308 and derivative strains were grown overnight in Brucella Broth, then subcultured to an OD 600 of 0.1 and then grown to mid-exponential phase (OD600 of 0.6) before isolation of chromosomal DNA by guanidine thiocyanate as described previously (³³).

Genomic DNA was sequenced by Seqcenter, LLC. Caulobacter strains were sequenced using nanopore R.9.5 pores, and Brucella strains were sequenced using nanopore R.10.4.1 pores.

Western Blot Analysis

For western blot analysis, Caulobacter strains were grown overnight in PYE media, and back-diluted to an OD600 of 0.05. Then grown in shaking flasks at 30°C in PYE, samples were collected at 1.5, 3, 6, and 9 hours and centrifuged at 26,000 rpm for 10 minutes to pellet the cells. The pellets were resuspended in 2x SDS-PAGE dye (12% glycerol, 4% SDS, 100 mM Tris-HCl pH 6.8, trace bromophenol blue, 40 mM DTT). 10µl of each sample was loaded into a 16-well 10% Bis-Tris plate for electrophoresis. Post-separation the membrane was incubated overnight with CcrM antibodies at a 1:5,000 dilution in 5% milk. Following overnight primary probing at 4 °C, the membranes were washed three times with Tris Buffered Saline with 1% Tween (TBST). Proteins were then visualized using IRdye-labeled mouse anti-rabbit antibody (LI-COR Biosciences) at 1:10,000 dilution and an Odyssey Scanning system (LI-COR). As ClpP levels remain constant in Caulobacter, it was used as a loading control for quantification. Membranes were incubated for 1 hour with anti-ClpP (1:5000 dilution) and visualized using IRdye-labeled mouse anti-rabbit antibody (LI-COR Biosciences) at 1:10,000 dilution and an Odyssey Scanning system (LI-COR).

Pipeline for nanopore R.9.5 through mCaller.

Raw nanopore sequencing outputs (Fast5 format) were quality-controlled and trimmed using Guppy (https://community.nanoporetech.com). High-quality reads were aligned to the reference genome NC_011916.1 (³⁴) with Minimap2 (^35,36), using default parameters. Fast5 files were converted to slow5 files (^37,38) and aligned to the mapped reads using the ‘eventalign’ function within Nanopolish (³⁹). Methylated bases were quantified using mCaller (²⁰), employing the precompiled ‘r95_twobase_model_NN_6_m6A.pkl’ model for assessing modified adenines. The methylation results were then annotated using Bedtools (⁴⁰) and then analyzed with R (⁴¹).

Pipeline for nanopore R.10.4.1 through Dorado

Raw nanopore sequencing outputs (Fast5 format) from nanopore sequencing, were converted into pod5 files for optimization. Dorado’s (²⁵) standard basecalling command was employed, utilizing the model ‘dna_r10.4.1_e8.2_400bps_sup@ v4.2.0’. In addition, modified base calling was executed using the specialized model ‘dna_r10.4.1_e8.2_400bps_sup @v4.2.0_6mA@v2’. The produced bed files were then aligned to NCBI reference genomes NC_007618.1 and NC_007624.1 (⁴²) with the mapped reads using the align and sort functions of Samtools (⁴³). The resulting modified bam files were run with modbam2bed (⁴⁴) with default settings and then analyzed with R (⁴¹).

Data Processing in R

Definitions in relation to methylation

In our study, motifs are defined as individual occurrences of the ‘GANTC’ sequence. Given the palindromic nature of ‘GANTC’, we pair motifs within a 2-base-pair distance to define “sites”. When comparing the probability of methylation among different strains, we considered motifs that deviated greater than or less than 2.5 standard deviations from the mean of a dataset as significantly different. Expanding upon this, sites in which both motifs fell below 2.5 standard deviations below the mean methylation probability for that particular strain were classified as hypomethylated.

Identification of and filtration of m6A motifs

Data was imported from both mCaller and Dorado pipelines into RStudios (⁴⁵) for analysis. Probability of methylation of GANTC motifs obtained through obtaining positional index through regexes of reference files and filtration of bed files. Additional motifs were identified through filtering all adenines with above a 80% score, coverage greater than 50, and homology with previously identified methyltransferase motifs in the Rebase database. All identified motifs in this study are available in Supplementary Table 3.

Slopes of probability of methylation as function of genomic position for each strain’s probability of methylation was calculated by finding the linear regression between genomic position from both ends of the origin to the terminus under the assumption that there is symmetry over the terminus. Pearson correlation coefficient between Nanopore and SMRT sequencing data was performed using data obtained from (¹⁰), full methods described with Supplementary Figure 1. Additional methods, and scripts are available on GitHub (https://github.com/mbcampbell-work/Alphaproteobacteria-Methylation).

Identification of genes in relation to gene expression

In order to assess whether changes in methylation had association with transcriptional changes, we conducted Fisher’s exact tests comparing the change of GANTC methylation upstream of genes against transcriptional change in mutants. Focusing upon GANTC sites located in intergenic regions, particularly those within 300 base pairs of a gene’s start codon or within 200 base pairs of a known transcription start site (TSS). TSS data for Caulobacter were retrieved from prior study available at (⁴⁶).

Changes in gene expression between the ∆lon mutant and wild-type Caulobacter were analyzed using data from the Gene Expression Omnibus (GEO) database, under accession number GSE152426, as reported in (²³). Additionally, microarray and ChIP-seq data in relation to the ∆mucR mutant of Brucella abortus were sourced from (¹⁸).

Importance:

DNA methylation plays an important role in bacteria to maintain genome integrity and regulate of gene expression. We used nanopore sequencing to directly measure methylated bases in Caulobacter crescentus and Brucella abortus. In Caulobacter, we showed that stabilization of the CcrM methyltransferase upon loss of the Lon protease results in prolific methylation and discovered that the putative methylase AlkB is unlikely to have a global physiological effect. We measured genome-wide methylation in Brucella for the first time, revealing a similar role for CcrM in cell-cycle methylation but a more complex regulation by the Lon protease than in Caulobacter. Finally, we show how the virulence factor MucR impacts DNA methylation patterns in Brucella.