Sometimes, you really want to stand out from the crowd. For years, 5-hydroxymethylcytosine (5hmC) has lurked in the shadows, often lumped together with 5-methylcytosine (5mC). However, these modifications have different effects—5mC generally turns down expression and 5hmC ramps it up.
Sure, several methods can tease apart 5mC and 5hmC, but they require a lot of sample or underestimate levels of the modifications. But now, two new methods shine a light on 5hmC and allow it to stand out proudly (but separately) from 5mC. Building directly on previous work, these methods reveal the amounts of both 5hmC and 5mC in single cells.
DARESOME: A Daring Restriction Enzyme Approach
Though restriction enzymes have been used to assess methylation, no method has been able to provide separate 5mC and 5hmC results at the same time. A few years ago, Lih Feng Cheow’s lab (National University of Singapore) developed DARE, a genome-wide, single-tube restriction enzyme approach that analyzes methylation in single cells. The problem is that 5hmC is still stuck in the crowd, not being fully appreciated.
Now, the team has “dared” to adapt the method to shine a light on 5mC and 5hmC simultaneously in single cells and cell-free DNA. In this technique, different barcodes are attached to the DNA, depending on which enzymes cut it. Here are the details of DARESOME (DNA analysis by restriction enzyme for simultaneous detection of multiple epigenomic states):
- HpaII cuts genomic DNA only at unmodified CCGG sites and then a U-tag adapter is added
- Remaining CCGG sites are cut with MspI and given H-tag adapters, and at the same time, NlaIII cuts fragments into shorter bits and adds a partial Illumina universal read 2 sequencing adapter called an N-tag
- 5hmCs get glucosylated, so another digestion with MspI now only cuts the 5mCs and puts an M-tag adapter where 5mCs were
- In the end, unmodified fragments have U tags, 5hmCs have H tags, and 5mCs have M tags
Tests with adult mouse brain cells showed that, as expected, high gene expression is correlated with a high 5hmC/5mC ratio, but DARESOME can do this in just a single assay. With a slight variation to the N tag to help the team pool barcoded single cells, they observed strand asymmetry in 5hmC that was consistent with semi-conservative replication—something you can’t detect in bulk samples. The group also used DARESOME on cell-free DNA, but since this is already fragmented, they didn’t have to use NlaIII. Instead, they added sequencing adapters to everything and only sequenced the pieces that had both a tag and a sequencing adapter. With this approach, they found methylation differences in samples from healthy versus colorectal cancer patients.
Joint-snhmC-seq: A Deamination Approach
In addition to restriction enzymes, there are deamination strategies like bisulfite conversion and enzymatic methods. The labs of Hao Wu and Rahul Kohli (both University of Pennsylvania) have been shining their spotlights on cytosine methylation for a while with ACE-seq that detects 5hmC and, most recently, DM-Seq that directly detects 5mC.
The teams recognized that their previous strategies require lots of separate steps, which isn’t ideal for single-cell analyses. So, this time, they collaborated to bring due genome-wide attention to both modifications in the same cell with Joint-snhmC-seq (joint single-nucleus (hydroxy)methylcytosine sequencing). Here’s how the approach evolved:
- They first developed whole-genome bisulfite-assisted ACE-seq (WG-bACE-seq) to profile 5hmC:
- First comes bisulfite treatment, which protects 5hmC by transforming it to cytosine-5-methylenesulfonate (CMS)
- Then AOPBEC3A, which doesn’t work well on cytosines with bulky groups at the 5-position like CMS, gets to work deaminating the other cytosines (C and 5mC)
- By combining WG-BS-seq (5mC and 5hmC info) with WG-bACE-seq (5hmC info), the method provides 5mC data through subtracting the 5hmC signal
To get to the single-cell or single-nucleus level:
- They adapted snmC-seq to develop snhmC-seq, which generates 5hmC data from single cells or nuclei that are sorted onto plates, treated with bisulfite, and deaminated
- Two rounds of barcoding later, the samples are pooled and sequenced
- Combining an optimized snhmC-seq called snhmC-seq2 with optimized snmC-seq called snmC-seq2, the team developed Joint-snhmC-seq
- Single-stranded bisulfite-treated DNA molecules from single nuclei are split and then snhmC-seq2 and snmC-seq2 are done separately, but side-by-side, or jointly, with a subtraction for 5mC data
The approach allowed the group to obtain 5mC and 5hmC data for single mouse brain nuclei, and they found differences across neuronal, as well as non-neuronal, cortical cells, showing that a postnatal increase in 5hmC isn’t just confined to maturing neurons, as people had thought. They could also use the data to identify cell types and cortical neuronal subtypes, as well as analyze methylation in CH contexts.
Shine On
These new methods let you determine 5mC versus 5hmC levels when using very low amounts of starting material, such as with single cells or cell-free DNA samples. This could be especially useful for future neuroscience studies or for developing diagnostic or surveillance strategies for detecting cancer signatures noninvasively in liquid biopsies. Ultimately, with these approaches, you can really let 5hmC (and 5mC) shine bright like a diamond that isn’t relegated to the shadows any longer.
Get all the details on DARESOME at Science Advances, September 2023, and on Joint-snhmC-seq at Nature Biotechnology, August 2023.