Modified DNA bases are widespread in living organisms from viruses to mammals. In T4 phages, most of the cytosine residues are present as 5-hydroxymethylcytosine and are often completely glucosylated by virally encoded β-GT.2,22
Similarly, in the mammalian genome, a percentage of cytosine residues are 5-mC, and a percentage of that is further oxidized to 5-hmC. In 1972, formic acid hydrolysis followed by chromatographic analysis was used to detect 5-hmC in murine brain and liver DNA. Using this method, 5-hmC was estimated to comprise ~15% of the total cytosine residues.23
Recently, β-GT was used in combination with a radioactive UDP-glucose to determine global 5-hmC levels. In our previous report, we successfully utilized a glucosylation step using a recombinant β-GT in combination with restriction enzyme cleavage to quantify the percentage of C, 5-mC, and 5-hmC in mammalian genomes.24
To understand the detailed biochemical properties of recombinant β-GT, we determined a series of steady-state kinetic properties and the reaction mechanisms of the enzyme.
Recombinant β-GT glucosylates double-stranded 5-hmC DNA very efficiently, as shown by its ability to completely glucosylate ~100-nucleotide substrates and T4-gt
DNA when the enzyme is given sufficient time. However, clusters of six or more 5-hmC residues on the 100 bp duplex substrate led to reduced turnover numbers and Km
values, indicating that β-GT binds tightly to clustered 5-hmC on DNA. Such behavior stands in contrast to that of other transferases such as DNA methyltransferase 3A (Dnmt3A), where increases in the density of methylation target sites resulted in higher rates of DNA methylation.25
Crystallographic studies of β-GT were not able to distinguish between a processive or distributive mode of glucosylation between adjacent 5-hmC residues. A processive mechanism seems unlikely because the active site of the β-GT is completely closed upon substrate DNA binding, requiring unfolding after catalysis to release the UDP product and bind new UDP-glucose substrate.26,27
This hypothesis was experimentally confirmed in our studies, as the chase nonbiotinylated substrate was able to remove the enzyme and UDP-glucose from the prebound biotinylated substrate DNA. This reaction mechanism was confirmed by product inhibition studies in which β-GT appears to follow a bi-bi random reaction mechanism. Similar reaction mechanisms are common for other histone and DNA methyltransferases, such as lysine methyltransferases (G9a28
) and DNA cytosine 5-methyltransferases (DNMT1,29
). However, in vivo, most of the enzymes may reside in complexes, and the reaction pathways and product formation may be different.
Cell culture-based studies have identified a strong correlation between DNA hypomethylation (reduction of 5-mC) and tumorigenesis.32,33
In a recent study, 5-mC was purposely converted to 5-hmC in adult mouse brain, and this promoted demethylation through a process requiring base excision repair. Furthermore, the 5-hmC is actively removed by the activation-induced deaminase/APOBEC family of cytidine deaminases in mammalian cells.10
This evidence would indicate that late stage tumorigenic samples should have lower levels of 5-hmC, when compared to match normal tissue, because 5-hmC is the precursor to the hypomethylation. To test this hypothesis, we measured 5-hmC levels in matched pairs of human genomic DNA from brain, lung, breast, colon, and liver tissues that were either normal or tumorgenic. On average, all the tumor samples had significantly (p
< 0.001) lower levels of genomic 5-hmC than their matched normal tissue (Figure B). This evidence lends credibility to the possibility that 5-hmC is acting as a precursor to hypomethylation or 5-hmC reductions are a reflection of reduced precursor 5-mC. Indeed, 5-mC levels in a large series of tumor DNA samples are significantly lower than what was observed in their matched normal tissue.34
Comparison between 5-hmC levels in control HCT116 cell and DNMT1 null background also demonstrated reduced levels of 5-hmC due to a genetic knockout of the methyltransferase gene, suggesting that a decrease in the level of the 5-mC precursor may lead to less oxidative product.18
There is one area of concern when utilizing β-GT to measure global 5-hmC levels, the presence of 5-hydroxymethyluracil (5-hmU) in the tested genomic DNA samples. 5-hmU, which is also a substrate for β-GT,35
arises from oxidation of the methyl group on thymine in genomic DNA.36
When utilizing β-GT for genomic 5-hmC measurements, the presence of 5-hmU can give false measurements of elevated 5-hmC levels. This was not a concern for our tested mammalian genomic DNA samples because in these biological systems 5-hydroxymethyluracil-DNA glycosylase (hmUDG) will glucosylate 5-hmU, leading to its subsequent removal by nucleotide excision repair,37
yet no homologue for hmUDG has been discovered in either plants or sea urchins; therefore, care needs to be taken when interpreting the genomic 5-hmC levels with β-GT in lower organisms.
Our β-GT-based measurements of global 5-hmC levels during ES cell differentiation were further validated for their accuracy by comparing the data with HPLC-based measurement of the genomic DNA that was hydrolyzed as nucleobases.8
Indeed, the patterns of 5-hmC changes were very similar in both assay systems. A study by Szwagierczak et al. used a comparable radioactive technique to measure total 5-hmC content, and even though they utilized different DNA standards, they also identified a similar linear relationship between the β-GT transfer of [3
H]glucose and total 5-hmC content (r
However, in this earlier study, the 5-hmC levels were measured only in different mouse tissues at different stages of development. We expanded upon this study by establishing a more comprehensive measurement of 5-hmC that was then applied to more than 40 different genomic DNA samples, spanning 13 different prokaryotic and eukaryotic species. In conclusion, β-GT-mediated global determination of 5-hmC levels is consistent with other analytical methods, and because of the simplicity of glucose incorporation measurement, smaller amounts of DNA can be used in the high-throughput format.