ABH expression is elevated in prostate carcinomas and
ABH3 expression is elevated in prostate carcinomas and has been referred to as prostate cancer antigen-1 or PCA-1. ABH3/PCA-1 was proposed as a diagnostic marker for prostate cancers, and suggested as a possible therapeutic target . However, the recent studies of ABH2 and ABH3-null mice indicated that ABH2 was more important than ABH3 in the defence against toxic alkylation damage . Hence ABH2 may be the more relevant therapeutic target.
AlkB homologs and other putative DNA/protein demethylases Sequence homologs of the E. coli AlkB protein have been described in many bacterial species, in eukaryotes including S. pombe, Drosophila melanogaster, mice and man and even in plant RNA viruses . Some bacterial species appear to have two AlkB-like proteins whereas eukaryotes often have several , . Nevertheless, these sequences are not ubiquitous and have not been found in all bacteria. Species lacking an apparent AlkB homolog, such as Bacillus subtilis, Haemophilus influenzae, ACET mg and S. cerevisiae, may have proteins with a related function that display little or no sequence similarity to AlkB. Single-stranded RNA plant viruses translate a large polyprotein containing methyltransferase, helicase and RNA replicase domains, and in some cases, an AlkB-like domain. It was initially proposed that these AlkB-like domains may reverse the host-mediated post-transcriptional silencing of RNA viral genes by plant RNA or protein methylases . However, a bioinformatic analysis of the various polyprotein domains suggests that the AlkB-like regions are more likely to be conventional DNA/RNA repair domains that have been integrated into viral genomes relatively recently, possibly from plant bacterial pathogens . DNA/RNA repair by these viral proteins remains to be demonstrated experimentally. Although many proteins are reported in the databases as AlkB homologs, only three, E. coli AlkB and human and mouse ABH2 and ABH3 are known to repair DNA , , . Six other human homologs, ABH1 and ABH4 to 8  are of unknown function (Fig. 7). An initial report suggested that ABH1 could complement the phenotype of an E. coli AlkB mutant , but this result was not reproduced by other laboratories, and neither DNA-binding nor DNA repair activities were detected in in vitro assays of purified ABH1 , , . It is also of note that two apparent S. pombe homologs (accession CAA18657 and NP_594941) overexpressed in E. coli did not complement an AlkB mutant, and the purified proteins were inactive in repair assays of 1meA in DNA (A. Orr, G. Daly and B.S., unpublished data). Hence, sequence homology to AlkB may not necessarily predict a role in repair of SN2 methylated single-stranded DNA, and the large number of AlkB homologs in eukaryotes may suggest recruitment into other roles. Constructs expressing ABH4-8 fused to green fluorescent protein have been transfected into HeLa cells and showed that all these proteins localise to the cell nucleus suggesting roles in nuclear processes (S.J., in preparation). Whilst AlkB can dealkylate large alkylated lesions in DNA and free 1-medATP, ABH2 and ABH3 are unable to catalyze these reactions , . It is possible that in human cells the role of ABH1 or ABH4-8 is to sanitize damaged precursors of DNA synthesis, remove large alkyl groups from DNA or repair minor alkylated DNA lesions, such as N2-methylguanine. Alternatively, they could have roles in the demethylation of methylated proteins, such as histones  or perhaps, less likely, be involved in relieving epigenetic silencing by 5-methylcytosine (5meC) residues. In mammalian promoter sequences, 5meC is involved in gene silencing and appears in some cases to be actively demethylated . Since the methyl group is firmly anchored by a stable C–C bond, free-radical chemistry as used by AlkB offers a chemically credible mechanism to attack 5meC, and one of the ABH proteins of unknown function could conceivably serve this role. The reaction product would be 5-hydroxymethylcytosine which is chemically stable, so the oxidised methyl group would not be released as formaldehyde. Indeed, this non-mutagenic form of cytosine occurs naturally in phage T4 DNA instead of cytosine. Oxidation of 5meC to 5-hydroxymethylcytosine could be an effective strategy of “demethylation”. In this case, expression of the relevant ABH gene might be elevated at the time in early development when 5meC residues are removed from DNA. Although mouse primordial germ cells isolated from 11.5-day-old embryos and neighbouring somatic cells expressed ABH4-8 mRNAs at similar levels (Diana Lucifero, Wolf Reik, B.S. and T.L., unpublished data), this warrants further investigation.