In this study, we isolated and characterized a human EndoV gene and its resulting protein, which shared homology with the nfi gene. This protein could cleave a specific site in inosine-containing ssDNA substrates, acting with deoxyinosine 3′ endonuclease activity similar to that of eEndoV. Surprisingly, this enzyme could also cleave a specific site in inosine-containing ssRNA substrates. Analysis of hEndoV nuclease activity and DNA/RNA binding activity revealed that the enzyme preferred inosine-containing ssRNA substrates to inosine-containing ssDNA substrates. Immunochemistry and cell fractionation analyses showed that hEndoV was localized to the cytoplasm rather than the nuclei in cells. In addition, a local distorted single-strand structure of inosine-containing dsRNA and a second ribose 3′ to the lesion were required for nuclease activity of hEndoV. TSN, which stimulates I-RNase activity in cell extracts, promoted cleavage of hEndoV activity. Inosine formed by the catalytic activity of the A-to-I editing enzyme ADAR2 was a substrate for hEndoV. Taken together, these data indicated that hEndoV had endonuclease activity, as an I-RNase, specific to a certain structure of inosine-containing ssRNA and dsRNA. Although we cannot exclude the possibility that post-translational modifications of hEndoV, such as phosphorylation or ubiquitination, enhance its nuclease activity towards DNA containing deoxyinosine, we propose that the enzyme has a major role in removing inosine-containing RNA, which may adversely affect living cells.
Two fundamental mechanisms contribute to the generation of inosine in RNA. One mechanism is through A-to-I RNA editing by ADARs20
. Two major isoforms of ADAR1 exist in mammalian cells: the p110 form is constitutively expressed, while the p150 form is induced by interferon27
. p110 localization in the nucleus and p150 localization in both the nucleus and cytoplasm have been reported previously27
, and more recent studies have suggested that both isoforms shuttle between the nucleus and cytoplasm28
. RNA-regulated interaction of transportin-1 and exportin-5 with the dsRNA-binding domain regulates the nucleocytoplasmic shuttling of ADAR1 (ref. 28
). ADAR2 is highly expressed in the brain and localizes to the nucleoli29
in cells. Although we have shown that hEndoV was localized to the cytoplasm rather than the nuclei in cells under our experimental conditions, the localization of GFP-fusion hEndoV in the nucleoli and cytoplasm has been recently reported16
. In both cases, hEndoV functions in locations where RNAs abound and are edited by ADARs, and not where DNA is found. Regarding the localization of the enzyme in cells, we realized the importance of detecting endogenous hEndoV by using anti-hEndoV antibodies as opposed to that by FLAG-tag or GFP fusion. However, because anti-hEndoV antibodies are not currently available for immunofluorescence microscopy, the precise subcellular localization of hEndoV is still unclear.
Editing occurs selectively within RNA and can result in codon changes, as inosine is interpreted as guanosine by the translation machinery. These ADARs predominantly catalyse RNA editing at specific sites in the dsRNA structure17
and are thought to hyperedit long dsRNA, which can result in up to 50% of the A residues being changed to I residues31
. As hyperedited inosine-containing dsRNA is likely to have localized changes in its RNA structure due to the relative instability of IU pairs, ADARs generally diminish the double-strandedness of dsRNA. Although we have shown that hEndoV could cleave dsRNA treated by ADAR2, hEndoV can also cleave locally distorted ssRNA structures containing inosine. It may also cleave the dsRNA hyperedited by ADAR1. The functions of hyperedited long dsRNA are not fully understood. Interferons induce p150 to form ADAR1 and are cytokines with antiviral activity. Furthermore, viral infections in ADAR1-deficient HeLa cells result in enhanced apoptosis. Thus, hEndoV, together with TSN, may have a role in the antiviral response by removing the hyperedited long viral dsRNA genome that has undergone A-to-I editing ()18
. Moreover, it is possible that hEndoV may be involved in interdicting the RNA-silencing pathway. Long dsRNA is processed by the dsRNA-specific ribonuclease Dicer to produce small interference RNA, which is a key step in the RNA-silencing pathway. As dsRNA that is extensively edited by ADARs is known to be completely resistant to Dicer, hEndoV and/or TSN might participate in the degradation of this editing dsRNA, consequently reducing the expression of small interference RNA ()20
The other mechanism that produces inosine in RNA is spontaneous and incidental modification by hydrolysis, nitrosative chemistry and incorporation of ITP1
. Elevated temperatures at pH 7.4 induced the conversion of adenine to hypoxanthine in ssDNA, and this condition will also generate hypoxanthine in mRNA. Nitrosative stress caused by increases in nitric oxide-derived nitrous anhydride during inflammation produces hypoxanthine from adenine. Moreover, unexpected incorporation of ITP into mRNA is brought about by RNA polymerase during transcription. Although ITP is removed by an inosine triphosphatase from the cellular nucleotide pool, ITP is formed because of defects in purine nucleotide metabolism. Inosines generated in mRNA via this process are thought to be potentially mutagenic, as inosine is recognized as guanine during translation, thereby generating mutant proteins that potentially inhibit cell viability36
. Moreover, A-to-I editing of mRNA by ADARs is a highly regulated mechanism controlling gene expression and gene integrity17
. Inosines in mRNA could be interpreted as inappropriate editing. Additionally, inosines in rRNA and tRNA could affect translational fidelity and efficiency. Therefore, removing unexpected inosines in RNA is an important cellular mechanism; inosine-containing RNA may be removed from living cells by hEndoV.
Deoxyinosine in DNA is also mutagenic because deoxyinosine is recognized as deoxyguanosine by replicative DNA polymerases; the lesions should be removed from DNA before replication. According to the results of E. coli
genetic studies, the deoxyinosine 3′ endonuclease eEndoV is a repair enzyme that removes deoxyinosine in AER. The absence of eEndoV induces A:T to G:C transitions in response to nitrous acid treatment, generating deoxyinosine in DNA. Previous studies have implicated hEndoV in the AER mechanism8
. Additionally, mammalian EndoV could complement the mutagenesis phenotype in the E. coli nfi
mutant. However, we think that these proteins may be highly overexpressed in that system. Regarding the biochemical results in mammalians, we and other group could not show significant deoxyinosine 3′ endonuclease activity of hEndoV in dsDNA substrates16
. Using micromolar concentrations of mEndoV and nanomolar concentrations of the annealed dsDNA substrate for hEndoV, these enzymes exhibited cleavage of the dsDNA substrate containing deoxyinosine. Even under the same experimental conditions in which mammalian EndoV cleaved dsDNA, the enzyme does not exhibit more efficient removal of deoxyinosine in dsDNA than in ssDNA8
. Although we cannot exclude the possibility that post-translational modifications of hEndoV enhance its nuclease activity towards DNA, human alkyladenine DNA glycosylase may have a major role in this repair process, as the enzyme catalyses the hydrolysis of N-glycosidic bonds to release damaged bases, including hypoxanthine38
. Together, we conclude that hEndoV is a novel RNA-modifying enzyme that has a central role in removing inosine-containing RNA.