Mammalian ALKBH8 possesses tRNA methyltransferase activity required for the biogenesis of multiple wobble uridine modifications implicated in translational decoding

Abstract

Uridines in the wobble position of tRNA are almost invariably modified. Modifications can increase the efficiency of codon reading, but they also prevent mistranslation by limiting wobbling. In mammals, several tRNAs have 5-methoxycarbonylmethyluridine (mcm5U) or derivatives thereof in the wobble position. Through analysis of tRNA from Alkbh8-/- mice, we show here that ALKBH8 is a tRNA methyltransferase required for the final step in the biogenesis of mcm5U. We also demonstrate that the interaction of ALKBH8 with a small accessory protein, TRM112, is required to form a functional tRNA methyltransferase. Furthermore, prior ALKBH8-mediated methylation is a prerequisite for the thiolation and 2'-O-ribose methylation that form 5-methoxycarbonylmethyl-2-thiouridine (mcm5s2U) and 5-methoxycarbonylmethyl-2'-O-methyluridine (mcm5Um), respectively. Despite the complete loss of all of these uridine modifications, Alkbh8-/- mice appear normal. However, the selenocysteine-specific tRNA (tRNASec) is aberrantly modified in the Alkbh8-/- mice, and for the selenoprotein Gpx1, we indeed observed reduced recoding of the UGA stop codon to selenocysteine.

FIG. 1.

ALKBH8 domain architecture and outline of the gene-targeting strategy. (A) Exons are numbered 1 to 12. The blue dashed line represents noncoding exonic regions, whereas the red dashed line represents exons deleted in the Alkbh8^−/− mice. The predicted RNA recognition motif (RRM) is indicated in green, and the region of ALKBH8 that has sequence similarity to E. coli AlkB is shown in orange. The regions of ALKBH8 that display sequence similarity to the human protein KIAA1456 and to the S. cerevisiae tRNA methyltransferase Trm9 are indicated in brown. The localization of the RTSFTFR and GCGNG motifs predicted to be essential for oxygenase and MT activities, respectively, is indicated. KO, knockout. Nonconserved amino acids are indicated by small capitals. (B) Schematic organization of the genomic Alkbh8 locus and the gene-targeting strategy. Black rectangles represent Alkbh8 coding sequences, red rectangles represent noncoding exon portions, and the solid line represents the chromosome. Dashed lines point out the chromosomal region targeted by homologous recombination. The Neo cassette is indicated by an open rectangle, and the LoxP sites by blue triangles. Restriction sites, probes used for Southern blotting, and PCR primers required for verification of homologous recombination are indicated in blue. All other details of gene targeting are available upon request. Diagram is not depicted to scale. (C) Results of Southern blot analysis for verification of accurate 5′ (left panel) and 3′ (right panel) homologous recombination in ES cells (see panel B for details). The genomic DNA of 10 PCR-tested (not shown) ES cell clones (2A11, 2B11, 2C11, 2C12, 2D1, 3A11, 3B11, 4B3, 4C11, and 5B3) was extracted from expanded ES cells and compared with wild-type DNA (E14ES). Digested DNA was blotted on a nylon membrane and hybridized with either the external 5′ probe to screen for 5′ homologous recombination events or the internal 3′ probe to screen for 3′ homologous recombination events. As shown by the results in the left panel, the 5′ external probe detected the expected wild-type and recombined allele in 5 clones. Three of these clones (2C11, 3A11, and 4C11) revealed wild-type and targeted signals of equal intensity by the 3′ internal probe and were particularly well suited for blastocyst injections. Clones in bold (2C11 and 3A11) were injected into blastocysts for the generation of gene-targeted mice.

FIG. 2.

Alignment of the ALKBH8-MT (methyltransferase) domain with TRM9 and KIAA1456. The amino acid sequence alignment shown was extracted from a slightly more extensive alignment that was generated with MAFFT (25) from 13 sequences, including, in addition to the sequences shown, ALKBH8 from Caenorhabditis elegans and Drosophila melanogaster, ALKBH8 and KIAA1456 from Gallus gallus and Xenopus tropicalis, and Trm9 from Arabidopsis thaliana. Arrows indicate a GXGXG motif expected to be critical for methyltransferase activity. Hs, Homo sapiens; Mm, Mus musculus; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe.

FIG. 3.

Wobble uridine modifications in total tRNA from wild-type and Alkbh8^−/− mice. (A) Chemical structures of the nucleosides investigated. (B to D) LC-MS-MS chromatograms for mcm⁵U, mcm⁵Um, mcm⁵s²U, and cm⁵U of tRNAs purified from wild-type (upper inset chromatogram), and Alkbh8^−/− (lower inset chromatogram) liver (B), testis (C), or brain (D). Relative quantity and retention time are indicated on the y and the x axis, respectively. Blue curves in the chromatograms correspond to mass transitions from loss of ribose neutrals (−132 Da), and red curves correspond to qualifier mass transitions from additional loss of the base methylester groups as methanol neutrals (−32 Da). The relevant peak is identified by its retention time (when detected). An arrow points to the expected position for the modified uridine nucleoside when absent or only just detectable.

FIG. 4.

Wobble uridine modifications in tRNA^Sec, tRNA^Glu(UUC), and tRNA^Arg(UCU) from wild-type and Alkbh8^−/− mice. (A to C) LC-MS-MS analysis of modified uridine nucleosides in isoacceptors tRNA^Sec (A), tRNA^Glu(UUC) (B), and tRNA^Arg(UCU) (C) from wild-type and Alkbh8^−/− mice. Blue and red curves are as described in the Fig. 3 legend. Chromatograms for nucleosides only present in trace amounts in a given isoacceptor (e.g., mcm⁵s²U in tRNA^Sec) are not shown. Retention times and arrows are as in Fig. 3.

FIG. 5.

MALDI-TOF mass spectrometry of RNase T1 fragments of tRNA^Sec (A) and tRNA^Glu(UUC) (B) purified from wild-type (black) and Alkbh8^−/− (red) mice. Only one fragment (inset; representing the anticodon loop) displayed a difference in atomic mass when wild-type and Alkbh8^−/− samples were compared, and the relevant m/z interval of this fragment is shown in greater detail. The 2′-3′-cyclic phosphate versions of the RNase T1 fragments represent the major peaks, and 1* indicates minor signals from the 3′-phosphate versions. The peaks corresponding to the modifications identified by LC-MS-MS analysis are indicated.

FIG. 6.

Enzymatic activity of the human ALKBH8/TRM112 methyltransferase complex. (A) Copurification of human TRM112 (arrows) with the MT domain of human ALKBH8. 6×His-tagged ALKBH8 or its individual MT (amino acids 352 to 664) or AlkB (amino acids 1 to 354) domain (indicated by boxes) was coexpressed with untagged TRM112 in E. coli as indicated and affinity purified on Talon beads. Asterisks indicate background bands representing E. coli proteins (not visible for ALKBH8-AlkB, which gave higher expression levels, and hence, less eluate was loaded). (B) MT activity of 100 pmol ALKBH8-MT/TRM112 on 10 μg of total tRNA from different organisms. tRNA was saponified where indicated. (C) The MT activity of ALKBH8 requires TRM112. The methyltransferase activity on 10 μg of saponified calf liver tRNA using different amounts of purified recombinant protein was measured. MT-mut, mutant ALKBH8-MT with G416A, G418A, and G431A amino acid substitutions; TRM112^#, 6×His-tagged TRM112. (D) ALKBH8-MT/TRM112-mediated remethylation of saponified tRNA^Glu(UUC) from calf liver, as demonstrated by MALDI-TOF mass spectrometry of RNase T1 fragments. In this experiment, the RNase T1 digestion primarily yielded the 3′-phosphate version of the fragments (18.0 Da heavier than the 2′-3′-cyclic phosphate version). (E) LC-MS-MS analysis of nucleosides from total tRNA (3.2 μg) from livers of Alkbh8^−/− mice, not treated or incubated with 100 pmol of ALKBH8-MT/TRM112. Retention times and arrows are as in Fig. 3. (F) Titration of ALKBH8-MT/TRM112 activity on Alkbh8^−/− tRNA (3.2 μg). LC-MS-MS analysis was performed as described for panel E, and the peaks were quantified. The maximum levels of mcm⁵U and cm⁵U were estimated to be approximately 1 modification per 10,000 unmodified nucleosides. No peaks corresponding to mcm⁵Um or mcm⁵s²U were detected.

FIG. 7.

Selenoprotein expression in wild-type versus Alkbh8^−/− mice. (A) ⁷⁵Se labeling and selenoprotein analysis. Mice were labeled with ⁷⁵Se for 48 h, and liver protein was extracted from wild-type and Alkbh8^−/− mice prior to electrophoresis to detect ⁷⁵Se-labeled proteins. Gpx1 and Txnrd1 were detected by gel electrophoresis and phosphorimaging, and the corresponding band intensities are shown in the bar graphs. Identification of Gpx1 and Txnrd1 was based on the method of a previous study (6). A Coomassie blue-stained gel (right) was used to correct for loading differences. (B) Measurement of glutathione peroxidase activity in liver extracts from Alkbh8^−/− and wild-type mice. Error bars represent standard deviations in the results of triplicate experiments. The y axis represents nanomoles of NADPH/min/mg protein. *, Gpx1 is responsible for the majority of the glutathione peroxidase activity in liver and thus is indicated on the y axis. (C) Reduced efficiency of stop codon recoding by Alkbh8^−/− relative to that of wild-type tRNA. DualLuc-Gpx1 reporter RNA (schematically represented in the left panel) was translated in tRNA-depleted rabbit reticulocyte lysates in the presence of calf liver tRNA or tRNA from wild-type or Alkbh8^−/− mice, luciferase activities were measured, and the Fluc/Rluc ratios were calculated. Error bars represent ranges between results for duplicate samples in a typical experiment. nt, nucleotide.

FIG. 8.

The role of ALKBH8 in the biogenesis of wobble uridine modifications in higher eukaryotes. The indicated scheme integrates the results of the present study with current knowledge of wobble uridine modifications. Red print indicates modifications directly formed by ALKBH8, and blue print indicates modifications that rely on ALKBH8-generated substrates for their formation. The isoacceptors addressed in the present study and their modification status in the wild-type (WT) and Alkbh8^−/− mice are indicated.