Bisulfite sequencing is a family of DNA sequencing methods used to measure DNA methylation at individual cytosine positions. Treatment with bisulfite converts most unmodified cytosines in DNA into uracil, which is subsequently read as thymine, while 5-methylcytosine largely retains a cytosine-like sequencing signal. Comparing the resulting sequence with a reference or untreated sequence therefore reveals the distribution of protected cytosines. The method is widely used in epigenetics, although standard bisulfite sequencing cannot distinguish 5-methylcytosine from 5-hydroxymethylcytosine.
Chemical principle
Cytosine is one of the four principal bases in DNA. Its methylated derivative, 5-methylcytosine (5mC), carries a methyl group at the fifth carbon of the base ring. This modification can influence interactions between DNA and regulatory proteins without changing the underlying nucleotide sequence.
Under appropriately controlled conditions, bisulfite reacts with cytosine in single-stranded DNA. The conversion proceeds through sulfonation, deamination, and subsequent desulfonation, producing uracil. Because uracil pairs with adenine, amplification and conventional sequencing represent the converted position as thymine. The corresponding conversion of 5mC is much slower under standard assay conditions, so most 5mC positions remain represented as cytosine.
The operational distinction is:
| Original base | Typical standard bisulfite-sequencing readout |
|---|---|
| Unmodified cytosine | Thymine |
| 5-Methylcytosine | Cytosine |
| 5-Hydroxymethylcytosine | Cytosine |
5-Hydroxymethylcytosine (5hmC) reacts with bisulfite to form a derivative that also retains a cytosine readout. Consequently, a retained C is not uniquely diagnostic of 5mC. Incomplete conversion of unmodified cytosine can produce the same apparent signal.
DNA must be sufficiently denatured for efficient conversion. Reaction conditions balance conversion efficiency against DNA damage, because bisulfite treatment can fragment DNA and reduce the number of molecules available for analysis.
Biological information measured
In mammalian genomes, methylation occurs predominantly at CpG sites, where a cytosine is followed by guanine along the same DNA strand. Many studies therefore focus on CpG methylation. Non-CpG methylation is also biologically important in plants and in particular mammalian cell types, including some neurons and pluripotent cells.
Bisulfite sequencing measures the modification state of cytosines in the DNA molecules that survive processing and are successfully sequenced. It does not directly measure gene expression, chromatin accessibility, or transcription-factor binding. Relationships between methylation and these processes depend on genomic context.
For example, methylation at some promoters is associated with reduced transcription, whereas methylation within gene bodies has different and more context-dependent relationships with gene activity. A methylation difference alone does not establish that methylation caused a change in expression.
Experimental workflow
Protocols vary, especially in the order of library construction and conversion, but most share several stages.
DNA preparation and conversion
DNA is extracted from the biological sample and assessed for quantity, purity, and integrity. Sample composition is important: a tissue containing several cell types produces a mixture of their methylation patterns.
The DNA is denatured and treated with bisulfite, followed by desulfonation and purification. An unmethylated control sequence may be included to estimate conversion efficiency. Known methylated controls can help evaluate inappropriate conversion of protected cytosines and other assay biases.
Amplification and library construction
For targeted assays, selected regions are commonly amplified using the polymerase chain reaction after conversion. Primers must account for the altered sequence and, when the aim is unbiased methylation measurement, should avoid preferentially amplifying methylated or unmethylated templates.
For high-throughput sequencing, DNA fragments receive adapters and are prepared as a sequencing library. Some protocols attach adapters before bisulfite treatment; others construct libraries from DNA after conversion. Pre-conversion designs must preserve suitable adapter sequences through treatment, whereas post-conversion designs are intended to recover converted fragments without exposing an already completed library to the full conversion process.
Many workflows require a DNA polymerase capable of copying uracil-containing templates. Amplification can introduce sequence-dependent biases, so protocol choice and amplification extent influence the resulting measurements.
Sequencing and computational processing
Reads are generated by a sequencing platform and processed to remove adapters and low-quality sequence. Specialized alignment software maps them to a reference genome while allowing for conversion-associated C-to-T changes and the corresponding G-to-A changes when considering the opposite strand.
After alignment, software identifies informative cytosine positions and counts retained C and converted T observations. Quality assessment commonly examines conversion efficiency, mapping rate, coverage, duplicate reads, and position-dependent methylation bias along reads.
Major assay formats
Targeted bisulfite sequencing
Targeted methods examine selected loci using PCR amplicons or sequence capture. They provide high coverage at predefined regions and are useful for validating findings or examining particular regulatory elements.
Traditional cloning followed by Sanger sequencing can reveal methylation patterns across individual amplified molecules. High-throughput amplicon sequencing provides many more observations, although PCR bias and duplicate amplification can limit the independence of those observations.
Whole-genome bisulfite sequencing
Whole-genome bisulfite sequencing (WGBS) aims to examine cytosines throughout the genome without restricting analysis to predefined regions. It can measure both CpG and non-CpG methylation and supports broad comparisons across genomic features.
Its principal demands are sequencing depth, cost, and computational processing. Genome-wide scope does not imply complete coverage: repetitive sequences, sequence composition, DNA damage, and mapping ambiguity can leave regions poorly measured or inaccessible.
Reduced-representation bisulfite sequencing
Reduced-representation bisulfite sequencing (RRBS) enriches a reproducible subset of the genome, usually through restriction digestion and fragment-size selection. A commonly used enzyme, MspI, recognizes CCGG sites, producing a representation enriched for many CpG-rich regions.
RRBS can provide substantial coverage at selected CpG sites with less sequencing than WGBS. However, the sampled regions are determined by the digestion and size-selection scheme; they are not an unbiased representation of the entire genome.
Single-cell bisulfite sequencing
Single-cell approaches adapt conversion and library preparation to the DNA content of an individual cell. They can reveal methylation heterogeneity that is obscured by bulk measurements.
Coverage per cell is typically sparse, and amplification or molecular loss can cause substantial missing data. Computational aggregation or statistical modeling may therefore be necessary to compare cell populations. A site not covered in a cell cannot be classified as unmethylated simply because no retained C was observed.
Quantification and statistical interpretation
At a covered cytosine position, a basic estimate of the retained-cytosine fraction is
where is the number of informative reads reporting cytosine and is the number reporting thymine. Under standard bisulfite chemistry, this is generally interpreted as the combined protected signal from 5mC and 5hmC, subject to conversion and sequencing errors.
Coverage determines how precisely the fraction is estimated. A value derived from two reads is much less informative than the same value supported by many independent molecules. Confidence intervals and statistical models can represent sampling uncertainty, although technical bias and biological variability require additional consideration.
A bulk methylation fraction of 50% does not identify a unique biological state. It may reflect:
- Different methylation states among cell types or individual cells.
- Differences between parental alleles.
- Heterogeneous modification among DNA molecules within a population.
- A combination of biological variation and technical error.
Comparative analyses often seek differentially methylated sites or regions. Biological replication, coverage differences, multiple testing, and cell-type composition are important considerations. Methods frequently use binomial or beta-binomial models, smoothing, or regional aggregation to account for sampling and variation between samples.
Limitations and sources of error
Incomplete conversion. Unmodified cytosines that fail to convert are counted as protected, inflating apparent methylation. Conversion controls are particularly important when measuring low methylation levels or non-CpG methylation.
Inappropriate conversion of modified bases. Loss of the protected signal can underestimate modification levels. Reaction conditions must therefore be evaluated for both conversion efficiency and preservation of modified cytosines.
DNA damage and selective recovery. Bisulfite treatment can degrade DNA. Recovery, library construction, and amplification can favor some sequences or methylation states over others, so sequenced molecules may not perfectly represent the original sample.
Reduced sequence complexity. Converting many cytosines to thymines makes reads less distinctive, complicating alignment, especially in repetitive regions.
Sequence variation. A genuine C-to-T sequence variant can resemble a converted cytosine. Genetic variation can also change CpG sites or affect mapping. Matched genetic information or variant-aware analysis can help separate sequence differences from modification differences.
Duplicate molecules. PCR copies of one original molecule are not independent observations. Duplicate handling depends on the assay: identical coordinates may indicate amplification duplicates in randomly fragmented libraries but are expected in targeted amplicon data.
Ambiguity of the protected signal. Standard bisulfite sequencing cannot separate 5mC from 5hmC. This is especially consequential in samples containing appreciable hydroxymethylation.
Limited causal interpretation. Associations between methylation and development, exposure, or cancer do not by themselves establish a causal mechanism.
Related methods and alternatives
Oxidative bisulfite sequencing introduces an oxidation step that converts 5hmC to a form subsequently read as thymine after bisulfite treatment, while 5mC retains a cytosine signal. Comparing oxidative and conventional bisulfite measurements can therefore estimate the contributions of 5mC and 5hmC, although uncertainty from both measurements affects the difference.
TET-assisted bisulfite sequencing uses selective protection and enzymatic oxidation to distinguish hydroxymethylated cytosines. These methods modify the chemical logic of the assay rather than merely changing the sequencing platform.
Enzymatic methylation-sequencing methods use enzymes to create a methylation-dependent sequence distinction without relying on conventional bisulfite conversion. They can reduce some forms of DNA damage, but the modifications detected depend on the particular chemistry.
Other approaches infer DNA modifications from native-molecule sequencing signals, including nanopore electrical signals and polymerase kinetics. Such methods avoid bisulfite conversion and can retain long-range sequence context, but their accuracy and modification specificity depend on the platform, coverage, and computational models.
Historical development and applications
A foundational bisulfite-based genomic sequencing method was described by Marianne Frommer and colleagues in 1992. Early studies combined bisulfite conversion with locus-specific amplification, cloning, and Sanger sequencing. The later adoption of high-throughput sequencing enabled genome-wide and reduced-representation methylation profiling.
Bisulfite sequencing is used to study embryonic development, cell differentiation, genomic imprinting, transposable-element repression, and disease-associated methylation changes. It also supports investigations of methylation-based biomarkers and the regulatory relationships among DNA modification, chromatin, and transcription. The biological resolution of these applications depends on assay coverage, sample composition, and whether the experimental chemistry distinguishes the relevant cytosine modifications.