Nanopore sequencing enables direct detection of methylated cytosines (e.g., at CpG sites), without the need for bisulphite conversion. CpG sites frequently occur in high density clusters called CpG islands (CGI) and most of vertebrate genes have their promoters embedded within CGIs.

Changes in methylation patterns within promoters is associated with changes in gene expression and disease states such as cancer: exploring methylation differences between tumour samples and normal samples can help to uncover mechanisms associated with tumour formation and development.

Adaptive sampling (AS) offers a fast, flexible and precise method to enrich for regions of interest (e.g. CGIs) by depleting off-target regions during sequencing itself with no requirement for upfront sample manipulation.

To read more about how the method works, and how it compares to other techniques for analysing methylation (e.g. EPIC arrays, bisulfite), please see our Introduction to Reduced-Representation Methylation Sequencing.

RRMS can be deployed on MinION Mk1b, GridION and PromethION P2S, P24 and P48 platforms.

When running on MinION/GridION, we recommend running a single sample per flow cell - using our Ligation sequencing gDNA V14 - reduced representation methylation sequencing (RRMS) (SQK-LSK114) protocol.

Alternatively, it is possible to multiplex up to 4 samples on a single PromethION flow cell, as outlined in this protocol.

Human sample sequencing

The RRMS protocol enables users to target 310 Mb of the human genome which are highly enriched for CpGs including all annotated CpG islands, shores, shelves and >90% of promoter regions (100% of promoter with more than 4 CpGs). As well as other rich CpG regions in the genome. The total number of CpG sites in the .bed file is 7.18 million.

For benchmarking purposes, we performed RRMS on five replicates of a metastatic melanoma cell line and its normal pair for a male individual (COLO829/COLO829_BL) and a triple negative breast cancer cell-line pair (HCC1395/HCC1935_BL). Each sample was run on a single MinION flow cell. RRMS resulted in high-confidence methylation calls (>10 overlapping reads) for 7.3-8.5 million CpGs per sample.

For comparison, we also performed Reduced Representation Bisulphite Sequencing (RRBS), which typically yields 1.7–2.5 high confidence calls per sample. More information on this comparison can be accessed in our RRMS performance document and poster.

Mouse sample sequencing

The RRMS protocol and a new .bed file have also been developed to target 308 Mb of the mouse genome, covering 100% of CpG island and promoter regions; as well as other rich CpG regions in the genome.

The performance of RRMS for mouse samples was characterised on replicates of a blastocyst-derived, embryonic stem cell line (ES-E14TG2a) and a leukemia cell-line (BALB/c AMuLV A.3R.1). A non-RRMS library was also run as a control. Each sample was run on a single MinION flow cell: RRMS resulted in high-confidence methylation calls (>10X reads per site) for 5.0–5.8 million CpGs per sample in the mouse genome, compared to ~400,000 CpGs in the control library.

Alternative vertebrate genomes could be sequenced using the RRMS protocol and a bespoke .bed file.
However, please note Oxford Nanopore Technologies has only validated this method using human and mouse samples.

This protocol describes how to carry out DNA extraction and reduced representation methylation sequencing (RRMS) for up to 4 samples on a single PromethION flow cell, using the Native Barcoding Kit (SQK-NBD114.24) and the Adaptive Sampling feature in MinKNOW.

Steps in the sequencing workflow:

Prepare for your experiment

You will need to:

• Ensure you have your sequencing kit, the correct equipment and third-party reagents
• Download the software for acquiring and analysing your data
• Ensure that you have the correct .bed file for Adaptive Sampling
• Check your flow cell to ensure it has enough pores for a good sequencing run

Sample preparation

• Extract your DNA using the QIAGEN Puregene Cell Kit.
• Fragment your DNA using the Covaris g-TUBE, and check its length, quantity and purity. The quality checks performed during the protocol are essential in ensuring experimental success.

Library preparation

The table below is an overview of the steps required in the library preparation, including timings and optional stopping points.

Sequencing and analysis

You will need to:

• Start a sequencing run using the MinKNOW software, which will collect raw data from the device and convert it into basecalled reads. While configuring the run, turn on the Adaptive Sampling setting and import a pre-prepared .bed file with your regions of interest, along with a FASTA reference file.
• Sequence the sample for a total of 96 hours, with two flow cell washes when the available pore count drops to around 40% of the starting pore count (typically after ~24 hours and the second time after ~48 hours).
• Use Dorado to call modified bases, for more information please refer to the Dorado github page.
• Use the commands recommended at the end of this protocol to aggregate the modified bases and perform CpG island annotation.

Input requirements per sample for the extraction method:

• 5 x 10^6 cells per sample

Input requirements per sample for the library preparation:

• 2 µg of g-tube fragmented gDNA

How to QC your input DNA

It is important that the input DNA meets the quantity and quality requirements. Using too little or too much DNA, or DNA of poor quality (e.g. highly fragmented or containing RNA or chemical contaminants) can affect your library preparation.

For instructions on how to perform quality control of your DNA sample, please read the Input DNA/RNA QC protocol.

Chemical contaminants

Depending on how the DNA is extracted from the raw sample, certain chemical contaminants may remain in the purified DNA, which can affect library preparation efficiency and sequencing quality. Read more about contaminants on the Contaminants page of the Community.

We have validated and recommend the use of all the third-party reagents used in this protocol. Alternatives have not been tested by Oxford Nanopore Technologies.

For all third-party reagents, we recommend following the manufacturer's instructions to prepare the reagents for use.

We highly recommend that you check the number of pores in your flow cell prior to starting a sequencing experiment. This should be done within three months of purchasing for MinION/GridION/PromethION or within four weeks of purchasing Flongle Flow Cells. Oxford Nanopore Technologies will replace any flow cell with fewer than the number of pores in the table below, when the result is reported within two days of performing the flow cell check, and when the storage recommendations have been followed. To do the flow cell check, please follow the instructions in the Flow Cell Check document.

Note: We are in the process of reformatting our kits with single-use tubes into a bottle format and reducing the concentration of EDTA.

Single-use tubes format with higher EDTA concentration:

Higher concentration of EDTA with a clear cap.

Bottle format with reduced EDTA concentration:

Reduced concentration of EDTA with a blue cap.

Note: This Product Contains AMPure XP Reagent Manufactured by Beckman Coulter, Inc. and can be stored at -20°C with the kit without detriment to reagent stability.

Note: The DNA Control Sample (DCS) is a 3.6 kb standard amplicon mapping the 3' end of the Lambda genome.

Extract DNA from your sample(s) using one of our recommended extraction protocols.

For the benchmarking of this method, the Oxford Nanopore team extracted DNA from ~5 million cells using the protocol: Human cell line DNA – QIAGEN Puregene Cell Kit. The steps for this method are outlined below.
Note: this method is also suitable for mouse cell line DNA.

We also offer multiple mammalian sample extraction protocols, which you can use for other sample types.

To prepare fragmented gDNA for the library prep protocol, mechanical fragmentation is performed using a g-TUBE (Covaris) to shear DNA to a fragment length of approximately 6kb.

For this method, the flow cell is washed after ~24 hours of sequencing to restore pores to ensure efficient data acquisition. After an additional 24 hours of sequencing, the flow cell is washed and reloaded a second time. For this reason, enough library was generated for 3 flow cell loads in the adapter ligation step of the protocol.

• This washing procedure aims to remove most of the initial library and unblock the pores to prepare the flow cell for the loading of a subsequent library.
• Data acquisition in MinKNOW should be paused during the wash procedure and library loading.
• After the flow cell has been washed, the next library can be loaded.

You can navigate to the Pore Activity or the Pore Scan Results plot to see pore availability.

Below you can find example data for pore states observed on a flow cell before and after wash steps are performed. Additionally, you can observe an example for the cummulative sequencing data output, including the wash and reload steps. The red asterisks indicate the flow cell wash and reloads.

Figure 1. Channel state for a 96 hour run. The flow cell washes are incorporated into the method to restore blocked pores, to allow continuous data acquisition. Red asterisks denote when a flush was performed.

Figure 2. Cumulative sequencing data output, over a 96 hour run. Red asterisks denote when a flush was performed.

For a full overview of nanopore data analysis, which includes options for basecalling and post-basecalling analysis, please refer to the Data Analysis document.

The sequencing device control and data acquisition are carried out by the MinKNOW software. Please ensure MinKNOW is installed on your computer. Further instructions for setting up your sequencing run can be found in the MinKNOW protocol.

• Select the Native Barcoding Sequencing Kit 24 (SQK-NBD114.24) in kit selection.

• Turn basecalling OFF (this will automatically turn off barcoding).
  Note: Basecalling and barcoding will be carried out post-sequencing in the downstream analysis section of the protocol.

• Turn Adaptive Sampling ON, and select Enrich.
  Input the human reference file for alignment and the .bed file for enumerating regions (check online catalogue for the human RRMS .bed file).

• Set the run duration for a minimum of 96 hours.

• Set up your desired output parameters.
  To ensure the downstream analysis functions correctly, we recommend keeping the default options of the output file format (.POD5).

• Click “Start” begin the sequencing run.

Basecalling:

Dorado stand-alone is used for basecalling using the dorado basecaller. Open a terminal and enter the following commands:

dorado basecaller hac,5mCG_5hmCG --kit-name SQK-NBD114-24 \
--secondary “no” -Y \
--reference {reference_fasta} {input_pod5_folder} \
| samtools view -e '[qs] >= {qscore_filter}' \
--output {out_pass_bam} \
--unoutput {out_fail_bam}

Notes:

• We recommend using the high accuracy model (hac) for RRMS sequencing runs. However, if using the super accurate model (sup), ensure you are utilizing the correct model in the above command.
• Alignment can be performed while basecalling by providing a reference FASTA file, the recommended human reference file can be downloaded here.
• Secondary alignments are discarded by using “--secondary no” and -Y option is enabled, to allow soft-clipping supplementary alignments.
• By providing the option “--kit-name SQK-NBD114-24” dorado will also classify reads into the different barcodes present by adding a tag to the generated BAM file.
• We recommend setting the qscore filter to 10.
• Please note, GPU compute is needed to perform basecalling with dorado, more information on how to run dorado can be found in the github repository.

Demux

Dorado demux is used to sort reads per barcode using the following command:

dorado demux --no-classify --sort-bam --output-dir <out_folder> {out_pass_bam}

Notes:

• This step will generate sorted BAM files for each of the possible barcodes for kit used (e.g. SQK-NBD114-24).
• Trimming of adapters and barcodes will happen by default in dorado. Once barcodes are trimmed reads can not be demuxed again.
• For more information check the dorado documentation.

RRMS target bed file can be downloaded from the AS catalogue available here.

Mosdepth is used to check coverage on target regions for the barcodes of interest:

mosdepth -x -t 8 -n -b {target_bed} {out_prefix} {input_pass_bam}

Human variation pipeline is used to aggregate modifications per genomic positions using modkit.

The workflow is available in the following repository: wf-human-variation github.

The documentation can be found in the following space: wf-human-variation EPI2ME page

Modification calling:

For most RRMS runs we recommend running the following command:

nextflow run https://github.com/epi2me-labs/wf-human-variation \
-profile singularity \  
--mod \
--bam <bam> \
--bed RRMS_human_hg38.bed \
--ref GCA_000001405.15_GRCh38_no_alt_analysis_set.fasta \
--sample_name <sample> --out_dir <output_dir>

(Optional) For haplotype-specific methylation:

If haplotype-specific methylation is required, you can provide options “--snp –phased“ to aggregate modifications identified on each of the haplotypes (i.e. one bedmethyl file for each of the haplotypes will be generated):

nextflow run https://github.com/epi2me-labs/wf-human-variation \
-profile singularity \  
--mod --snp --phased \
--bam <bam> \
--bed RRMS_human_hg38.bed \
--ref GCA_000001405.15_GRCh38_no_alt_analysis_set.fasta \
--sample_name <sample> --out_dir <output_dir>

Note: For this specific analysis, a sample coverage of >30X is recommended.

For detection of differentially methylated regions across different samples “modkit dmr” can be used.

For more information check the modkit documentation available here.

The BAM file(s) generated by dorado contains canonical bases as well as per-read modifications stored in MM and ML BAM tags. To visualise the per-read modification calls, IGV can be used to load the BAM file and set "colour reads as" to “base modification 2-color (all)”.

If phasing was performed using wf-human-variation pipeline, the haplotagged BAM file can be uploaded in IGV and alignments can be grouped by haplotype using the IGV option “group by” and selecting “phase”.

Per-position methylation frequencies can also be visualised in IGV by using BIGWIG format. For this, modkit is used to generate BEDGRAPH files using the following command:

modkit pileup --cpg --combine-strands --bedgraph \ 
--threads 10 --prefix {out_prefix} \  
--ref {reference_fasta} \ 
{out_folder} {input_pass_bam}

Please note, a different bedgraph file will be created for each of the modifications present, in this case 5mC and 5hmC.

Next, bedGraphToBigWig is used to generate bigwig files which can be uploaded together with your BAM file in IGV:

bedtools sort -i {out_folder}/{prefix}_m_CG0_combined.bedgraph | cut -f 1-4 > {out_folder}/{prefix}_m_CG0_combined_sort.bedgraph

bedGraphToBigWig {out_folder}/{prefix}_m_CG0_combined_sort.bedgraph {reference_chrSize} {out_mod_bed_agg_filt_bigwig}

For information about benchmarking the performance of RRMS for human samples, please see our RRMS performance document

Due to the extended sequencing time, and the multiple flow cell washes and library reloads, we do not recommend re-using the flow cells used in this method.

Re-using these flow cells for subsequent sequencing experiments may result in insufficient data generation for analysis.

We also have an FAQ section available on the Nanopore Community Support section.

If you have tried our suggested solutions and the issue still persists, please contact Technical Support via email (support@nanoporetech.com) or via LiveChat in the Nanopore Community.

We also have an FAQ section available on the Nanopore Community Support section.

If you have tried our suggested solutions and the issue still persists, please contact Technical Support via email (support@nanoporetech.com) or via LiveChat in the Nanopore Community.