https://orcid.org/0000-0002-3701-5965
Abstract
RNA acetylation is a universal post-transcriptional modification that occurs in various RNAs. Transfer RNA (tRNA) acetylation is found at position 34 (ac4C34) in bacterial tRNAMet and position 12 (ac4C12) in eukaryotic tRNASer and tRNALeu. The biochemical mechanism, structural basis and functional significance of ac4C34 are well understood; however, despite being discovered in the 1960s and identification of Kre33/NAT10 and Tan1/THUMPD1 as modifying apparatuses, ac4C12 modification activity has never been reconstituted for nearly six decades. Here, we successfully reconstituted the ac4C12 modification activity of yeast Kre33 and Tan1. Biogenesis of ac4C12 is primarily dependent on a minimal set of elements, including a canonical acceptor stem, the presence of the 11CCG13 motif and correct D-arm orientation, indicating a molecular ruler mechanism. A single A13G mutation conferred ac4C12 modification to multiple non-substrate tRNAs. Moreover, we were able to introduce ac4C modifications into small RNAs. ac4C12 modification contributed little to tRNA melting temperature and aminoacylation in vitro and in vivo. Collectively, our results realize in vitro activity reconstitution, delineate tRNA substrate selection mechanism for ac4C12 biogenesis and develop a valuable system for preparing acetylated tRNAs as well as non-tRNA RNA species, which will advance the functional interpretation of the acetylation in RNA structures and functions.
Introduction
RNAs are decorated with over 150 different chemical modifications, which are collectively termed RNA epitranscriptome (1). Among the various RNA classes, transfer RNAs (tRNAs) harbor the most extensive and diverse modifications, accounting for ∼80% of the modifications identified (2).
Acetylation is a universal post-transcriptional modification found in various types of RNA [messenger RNA (mRNA), tRNA and ribosomal RNA (rRNA)] and across all branches of life (1). Two acetylated residues, N6-acetyladenosine (ac6A) and N4-acetylcytidine (ac4C) (Supplementary Figure S1), have been reported to date (3,4). While ac6A is found in archaeal tRNAs only (3,5), ac4C is present in all domains of life (4,6). The predominant distribution of ac4C in the coding region of human mRNAs was previously observed (7). However, a recent study failed to identify ac4C sites in human and yeast mRNAs under physiological conditions (8). The ac4C modification has also been identified in mammalian and yeast 18S rRNAs, where it contributes to pre-rRNA processing and translational accuracy (9).
In bacterial tRNAs, ac4C is exclusively found in the wobble position (position 34, ac4C34) of the elongator tRNAMet (10). At this position, acetylation is mainly catalyzed by the GCN5-related acetyltransferase, TmcA (4,11). As such, ac4C34 modifications can be successfully reconstituted in vitro via TmcA. Biochemical and structural investigations into TmcA have revealed its catalytic mechanism, showing that it utilizes acetyl-CoA (as an acetyl donor) and ATP as substrates (Supplementary Figure S2A) (4,11). In addition, a recent study identified a tRNAMet cytidine acetate ligase (TmcAL) responsible for ac4C34 biogenesis in Bacillus subtilis. Reconstitution of TmcAL activity revealed that TmcAL uses acetate as an acetyl donor and that its catalytic mechanism (tRNA aminoacylation-like mechanism) and domain composition are completely different from those of TmcA (Supplementary Figure S2B) (12). ac4C34 plays a critical role in translation accuracy by ensuring the precise decoding of the AUG Met codons and concurrently preventing the misreading of the near-cognate AUA Ile codons (12,13).
Archaeal tRNAs also carry ac4C modifications, but the modified sites are more diverse, covering multiple tRNAs from various archaea (5,8). Archaeal ac4C modifying enzymes have been identified (14). Archaeal tRNA ac4C modification is stimulated at higher temperatures, and acetyltransferase-deficient strains exhibit temperature-dependent growth defects, suggesting that temperature-dependent ac4C biogenesis is a unique adaptive survival strategy for archaea (8,14). However, the loss of ac4C modification has no influence on the melting temperature (Tm) of archaeal tRNAs (14). To date, in vitro investigations into the activity of archaeal ac4C modifying enzymes have not been reported.
In eukaryotic cytoplasmic tRNAs, ac4C is found exclusively at the middle position of the 11CCG13 motif (position 12, ac4C12) in the D-arm of class II tRNASer and tRNALeu with long variable arms (6,15). The CCG motif is also the site of all eukaryotic mRNA and rRNA modifications (6). In fact, ac4C12 was the first site of RNA acetylation identified, being observed in yeast and mammalian tRNASer in the 1960s, and was later additionally identified in eukaryotic tRNALeu (16–22). In 2004, Tan1 was found to be involved in tRNA ac4C12 modification. However, the lack of an acetyltransferase domain in Tan1 suggests that it is not a catalytic enzyme but likely functions as a helper protein (23). Yeast strains with Tan1 deletion, leading to a lack of ac4C12, or with simultaneous deletion of Tan1 and tRNA (uracil-2′-O)-methyltransferase (Trm44), depriving them of both ac4C12 and 2′-O-methyluridine at position 44 (Um44), exhibited reduced levels of tRNASer(CGA) at higher temperatures due to 5′–3′ exonuclease-mediated RNA degradation. This suggests that ac4C12 contributes to tRNA stability (24). More recently, the enzyme responsible for ac4C12 biogenesis was identified as the rRNA acetyltransferase Kre33/NAT10. This enzyme needs to work alongside Tan1/THUMPD1 in both yeast and human cells in tRNA ac4C12 biogenesis (Supplementary Figure S2C) (25). In contrast, in eukaryotic 18S rRNA ac4C modification, small nucleolar RNAs are required for guiding Kre33/NAT10 to the target modification site (25–27). However, in vitro activity reconstitution of rRNA ac4C modification shows that yeast Kre33 or human NAT10 alone is capable of introducing ac4C in rRNA fragment (9,28).
Altered levels of ac4C are associated with various human diseases, including inflammation, diabetes, renal failure, pulmonary fibrosis, multiple sclerosis and various infections (6). Due to the wide distribution of ac4C across mRNA, tRNA and rRNA, ac4C-related diseases cannot be attributed solely to tRNA hypomodification. However, a series of biallelic variants in THUMPD1 impair its activity and cause a loss of ac4C12 modification. These changes lead to a syndromic form of intellectual disability, highlighting the direct association of tRNA ac4C12 with human diseases (29).
Though the field has a clear understanding of the molecular mechanism of bacterial ac4C34 biogenesis and its biological significance in Met codon decoding (4,11,12), our knowledge of eukaryotic ac4C12 modifications, including the modification mechanism, tRNA substrate selection, structural basis and its impact on tRNA structure and function, remains very limited. In addition, little is known about the structure and function of Tan1/THUMPD1 in ac4C12 biogenesis. Moreover, the mechanisms underlying the disease-causing variations of NAT10 and THUMPD1 remain unclear. These limitations are derived, at least in part, from the fact that the eukaryotic tRNA ac4C12 modification has never been reconstituted in vitro. Consequently, a recent study investigating the potential effect of ac4C12 on tRNA thermodynamic properties was dependent on chemical solid-phase RNA synthesis of an ac4C12-containing fragment of the tRNA D-arm but could not utilize intact tRNA (30). The failure to prepare ac4C12-modified tRNA has greatly impeded our understanding of the molecular mechanism and putative role of ac4C12 in tRNA structure and function.
In this work, based on the solved and predicted structures of yeast Kre33 (31–34), we constructed and purified a C-terminal flexible domain (FD)-truncated Kre33. In combination with purified Tan1, we successfully reconstituted ac4C12 modification of both tRNASer and tRNALeu via the Kre33–Tan1 activity. We elucidated the tRNA recognition mechanism involved in ac4C12 biogenesis and proposed a molecular ruler mechanism for substrate selection by Tan1 as a distance-measuring strategy. We were also able to modify various non-substrate cytoplasmic tRNAs by introducing a single A13G mutation. Moreover, we identified site-specific ac4C modifications in small RNAs according to the tRNA recognition mechanism of Kre33–Tan1. In vitro and in vivo data suggest that tRNA recognition by Tan1 is mediated by coordinated protein–RNA interactions. We determined that ac4C12 alone had no influence on the Tm values of tRNA transcripts, at least in vitro, or on the charging levels of tRNASer and tRNALeuin vitro and in vivo. Together, our results demonstrate the first in vitro reconstitution of ac4C12 modification and propose the principles underlying tRNA substrate selection in eukaryotic ac4C12 biogenesis. This work lays a solid foundation for further investigation of the molecular mechanism, structural basis, and functional significance of eukaryotic ac4C12 modifications and the molecular etiology of ac4C12-related diseases.
Materials and methods
Materials
[14C]Acetyl-CoA and [14C]Ser were purchased from PerkinElmer (Waltham, MA, USA). KOD-Plus mutagenesis kits were purchased from TOYOBO (Osaka, Japan). Yeast was transformed using the Yeastmaker Yeast Transformation System 2 kit (TaKaRa, Japan). The Ni2+-NTA Superflow resin was purchased from Qiagen (Hilden, Germany). 50× Denhardt solution (B548209-0050) and 20× saline sodium citrate (SSC) (B548109) were purchased from Sangon Biotech (Shanghai, China). Primer synthesis and DNA sequencing were performed by Tsingke (Shanghai, China). Anti-His6 (AbHO012HS01) was purchased from Shanghai HuiOu Biotechnology Co. Ltd (Shanghai, China). Anti-GAPDH (60004-1-Ig) was purchased from Proteintech (Rosemont, IL, USA). A standard ac4C sample (HY-W019670) was purchased from MedChemExpress (Monmouth Junction, NJ, USA).
Plasmid construction, expression and protein purification
Open reading frames (ORFs) encoding yeast Kre33 (Met1–Leu936) (UniProt No. P53914), Tan1 (UniProt No. P53072) and SerRS (UniProt No. P07284) were amplified from yeast genomic DNA. For gene expression in Escherichia coli, the Kre33 ORF was inserted between the SacI and NotI sites of pRSFDuet1 using an N-terminal His6 tag. The Tan1 ORF was inserted between the NdeI and XhoI sites of pET28a using an N-terminal His6 tag. The SerRS ORF was inserted between the NdeI and XhoI sites of pET30a with a C-terminal His6 tag. The Tan1 ORF was inserted between the BcuI and XhoI sites of p416TEF with a C-terminal His6 tag for yeast complementation assays. Mutagenesis of Tan1 was performed according to the protocol provided with the KOD-Plus mutagenesis kit. The primers used for cloning are listed in Supplementary Table S1. All relevant constructs were expressed in E. coli BL21(DE3) cells, except for the Kre33 ORF, which was expressed in Rosetta. The transformants were incubated overnight at 16°C with 200 μM isopropyl-β-d-1-thiogalactopyranoside when the absorbance at 600 nm of the initial cell culture reached 0.6. The process of protein purification from E. coli transformants has been described in a previous report (35). Tan1 was further purified by gel filtration on a Superdex S200 column with running buffer (150 mM NaCl, 50 mM Tris–HCl, pH 8.0).
tRNA gene cloning and transcription
Genes encoding Saccharomyces cerevisiae cytoplasmic tRNASer(CGA) [SctRNASer(CGA)], SctRNALeu(CAA), SctRNAThr(AGU), SctRNAIle(AAU) and SctRNAAsn(GUU) were integrated into the pTrc99b plasmid. SctRNAArg(ACG) and SctRNATyr(GUA) genes were inserted into the pTrc99b plasmid along with a hammerhead ribozyme to improve transcription efficiency (36). tRNAs were obtained by in vitro transcription, as previously described (37). For the two tRNAs with a ribozyme, the transcription mixture was incubated at 65°C for 1 h after DNA template digestion to promote the self-splicing of the ribozyme. Mutagenesis of the tRNA genes was performed according to the protocol provided with the KOD-Plus mutagenesis kit. The primers used for template preparation are listed in Supplementary Table S1.
Preparation of ac4C-modified tRNAs
ac4C-harboring tRNAs were prepared in a 200 μl reaction mixture containing 60 mM Tris–HCl (pH 8.0), 10 mM KCl, 10 mM MgCl2, 5 mM DTT, 4 mM ATP, 300 μM acetyl-CoA, 20 μM SctRNAs (or its variants), 5 μM Kre33 and 5 μM Tan1. The reaction was incubated at 30°C for 1 h, and ac4C-modified tRNA was purified with phenol and chloroform/isoamyl alcohol (24:1) and then precipitated with 100% ethanol and NaAc at −20°C overnight. The concentration of ac4C-modified tRNA was determined by denaturing 15% urea polyacrylamide gel electrophoresis (PAGE) based on a linear curve obtained from serial dilution transcribed tRNAs with known concentrations.
Ultra-performance liquid chromatography–tandem mass spectrometry analysis of ac4C-modified tRNA
One microgram of SctRNASer(CGA) or SctRNALeu(CAA) transcript or Kre33–Tan1-modified SctRNASer(CGA), SctRNASer(CGA)-C12U, SctRNALeu(CAA) or SctRNALeu(CAA)-C12U was completely hydrolyzed by benzonase nuclease, phosphodiesterase I and bacterial alkaline phosphatase in a 60 μl reaction solution containing 100 mM NH4Ac (pH 5.2) for 48 h at 37°C. The product was subjected to ultra-performance liquid chromatography–tandem mass spectrometry (UPLC–MS/MS) and the mass jump from m/z 286.0 to 154.0 (ac4C) was monitored and recorded. ac4C was detected using a Hypersil GOLD aQ column (Thermo Fisher Scientific, Waltham, MA, USA).
Liquid chromatography–electrospray ionization mass spectrometry for tRNA fragment analysis
For RNA fragment analysis, the isolated tRNAs, including SctRNASer(CGA) and SctRNALeu(CAA) transcripts, ac4C-SctRNASer(CGA) and ac4C-SctRNALeu(CAA) (5 μg) were digested with 1 μl RNase T1 and 44 μl SMART Digest RNase Buffer (Thermo Fisher Scientific) at 37°C for 30 min. The digests were mixed with one-tenth volume of 0.1 M triethylamine acetate (pH 7.0), and 2 μl sample was subjected to a Q Exactive™ Plus Orbitrap mass spectrometer (Thermo Fisher Scientific) equipped with an electrospray ionization (ESI) source and Thermo Vanquish LC system. The digested tRNA fragments were fractionated with a DNA Pac RP column (100 mm × 2.1 mm, 4 μm, Thermo Fisher Scientific). Solvent system consisted of 2% HFIP and 0.1% TEA in H2O (solvent A) and methanol (solvent B). The samples were separated at a flow rate of 200 μl/min using a linear gradient of 1% B in 0–2 min, 1–30% B in 2–15 min, 30–90% B in 15–16 min, 90–1% B in 16–17 min and 1% B in 17–20 min. The chromatographic eluent was ionized by an ESI source in negative polarity mode and scanned over an m/z range of 400–3000. The HCD was performed by data-dependent scan with collision energies of 17, 20 and 23. Biopharma Finder software (Thermo Fisher Scientific) was used for data analysis.
Activity determination of Kre33–Tan1
The ac4C modification reactions were performed at 30°C in a 30 μl reaction mixture containing 60 mM Tris–HCl (pH 8.0), 10 mM KCl, 10 mM MgCl2, 5 mM DTT, 4 mM ATP, 66.6 μM [14C]acetyl-CoA, 5 μM SctRNAs (or its variants), 1 μM Kre33 and 1 μM Tan1. The kinetics for both SctRNASer(CGA) and SctRNALeu(CAA) were assayed in a reaction mixture containing 60 mM Tris–HCl (pH 8.0), 10 mM KCl, 10 mM MgCl2, 5 mM DTT, 4 mM ATP, 66.6 μM [14C]acetyl-CoA, 0–20 μM SctRNASer(CGA) or SctRNALeu(CAA), 1 μM Kre33 and 1 μM Tan1. Equal amounts (9 μl) of the reaction solution were added to the filter pads at different time points and quenched using cold 5% TCA. The filter pads were rinsed three times with cold 5% TCA for 15 min each, followed by three rinses with 100% ethanol for 10 min each. Finally, the filter pads were dried under a heat lamp, and the radioactivity of the precipitates was quantified using a scintillation counter (Beckman Coulter, Atlanta, GA, USA).
Yeast total RNA extraction
Yeast cells (BY4742 and ScΔTan1) were cultured in 2 ml YPD medium for 12 h and then transferred to 200 ml new YPD medium at a ratio of 1:100 for 5 h to A600 = 2.0. The pellet was collected by centrifugation at 5000 × g for 10 min. After resuspending the precipitate in TRIzol, the yeast was ground using a tissue grinder, and the supernatant was collected by centrifugation at 5000 × g for 10 min. The supernatant was mixed with one-fifth volume of chloroform/isoamyl alcohol (24:1) and centrifuged at 5000 × g for 15 min at 4°C. The samples were then mixed with three-volume 100% ethanol and precipitated at −20°C overnight. The precipitated total RNA was centrifuged, washed with 75% ethanol, dried and then solubilized in 50 μl of 0.1 M NaAc (pH 5.2). A portion of the precipitate was also taken and dissolved using 0.1 M Tris–HCl (pH 9.0) and then treated at 37°C for 90 min as a deacylation sample.
Acidic northern blotting
Acidic northern blotting was performed as described previously (38,39). Briefly, tRNAs were isolated by acidic 10% PAGE (containing 8 M urea). Five microgram tRNA samples were loaded onto a 1.5 mm thick gel in 0.1 M sodium acetate buffer (pH 4.5) and electrophoresed at 4°C and constant 20 W for 22 h. Gel fraction containing tRNAs was transferred to a Nylon membrane (Millipore, Bedford, MA, USA) by electrophoresis with 0.5× TBE buffer at 250 mA for 45 min. After UV cross-linking (8000 × 100 J/cm2), prehybridization was performed using 10 ml of prehybridization solution [6× SSC, 50× Denhardt buffer and 0.5% sodium dodecyl sulfate (SDS)] for 1.5 h at 55°C. Hybridization was performed at 55°C for 12 h with 10 ml of the same liquid in the presence of a digoxigenin-labeled tRNA probe (Supplementary Table S1). The membranes were washed with 2× SSC solution containing 0.1% SDS for 10 min at 55°C, then washed in washing buffer (5× maleic acid buffer, 3‰ Tween 20) for 20 min, hybridized with anti-digoxigenin-AP and visualized with CDP-star. The signals were quantified by Multi Gauge v3.0.
Aminoacylation assays
Aminoacylation of SctRNASer(CGA) was measured in a reaction mixture containing 50 mM Tris–HCl (pH 7.5), 15 mM MgCl2, 60 mM KCl, 5 mM DTT, 2. 5 mM ATP, 181.8 μM [14C]Ser, 10 μM SctRNASer(CGA) or ac4C-SctRNASer(CGA), and 500 nM SerRS at 30°C. The processing procedures were the same as those described in the ‘Activity determination of Kre33–Tan1’ section.
Western blotting
Whole-cell lysates of yeast transformants were separated using 10% SDS–PAGE and then transferred to methanol-activated polyvinylidene fluoride membranes, which were then blocked with 5% skimmed milk in PBST (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4 and 0.5‰ Tween 20) for 1 h at room temperature. The membranes were immunoblotted overnight using anti-His6 or anti-GAPDH. After three washes with PBST, the membranes were incubated with anti-mouse secondary antibody diluted 1:10 000 in PBST for 1 h at room temperature, respectively. Detection was performed using SuperSignal West. Signals were quantified using Multi Gauge v3.0.
tRNA melting temperature (Tm) determination
Specific tRNAs were dissolved in a buffer containing 50 mM sodium phosphate (pH 7.0), 100 μM EDTA and 100 mM NaCl and diluted such that the initial absorbance of the tRNA at 260 nm was between 0.2 and 0.3. The tRNA was analyzed using an Agilent Cary 100 spectrophotometer at 260 nm to determine the Tm profile, with a heating rate of 1°C/min from 25 to 95°C.
Yeast complementation assays
Yeast Tan1 gene deletion strain (ScΔTan1) was kindly provided by Dr Jin-Qiu Zhou in our institute. For yeast complementation, genes of interest were cloned into p416TEF. Complementation was performed by transforming individual constructs into the ScΔTan1 strain. The transformants were cultured in liquid SD/Ura− medium. The culture was diluted to a concentration equivalent to OD600= 1.0, and a 10-fold dilution of the yeast culture was plated onto YPD plates, which were then cultured at 30 or 37°C. The growth rates of various strains were observed and compared. Western blotting analysis was performed to detect protein levels in various strains.
Results
Primary sequence and tertiary structure analyses of Kre33 and Tan1
Saccharomyces cerevisiae Kre33 is a multidomain 1056-amino acid protein. Based on primary sequence alignment and the cryo-electron microscopy (cryo-EM) structure of Kre33 (PDB No. 6LQP) (31), Kre33 was found to comprise an N-terminal head domain (Met1–Asp224), a helicase domain (Asp225–Thr495), an N-acetyltransferase domain (Leu496–Ser771), a tail domain (Asn772–Leu936) and an invisible C-terminal FD (Pro937–Asn1056) (Supplementary Figure S3A and B). The FD domain of the thermophilic fungus Chaetomium thermophilum Kre33 was also not visible in the cryo-EM structure (PDB No. 5JPQ) (32) (Supplementary Figure S3C). Indeed, the structure of full-length Kre33 predicted by AlphaFold (33,34) suggested that the C-terminal FD contains four α-helices with several disordered connecting peptides and protrudes away from the main domains, suggesting high flexibility (Supplementary Figure S3D). Sequence alignment between Tan1 and the THUMP domains of Archaeoglobus fulgidus Trm11 (m2G10 tRNA methyltransferase) (40) and Thermotoga maritima ThiI (4-thiouridine synthetase) (41) showed that the THUMP domain of Tan1 ranges from Ala170 to Asp259, but the sequences of the three THUMP domains are not conserved (Supplementary Figures S3E and S4). No structures of the Tan1/THUMPD1 family proteins have been reported. We assessed the AlphaFold-predicted structure of Tan1 (Supplementary Figure S3F) and found the THUMP domain forms three stranded β-sheets (β5–β7) and two α-helices (α4 and α5) on one side. An additional four stranded β-sheets (β1–β4) and one α-helix (α1) in the N-terminal portion of Tan1 form a subdomain highly resembling the N-terminal ferredoxin-like domain (NFLD) in A. fulgidus Trm11 (40) and T. maritima ThiI (41). However, the NFLD domain of Tan1 is much longer than that of ThiI or Trm11 due to a large insertion (Gly63–Pro112) between α1 and β2 and an α-helix (α3) inserted between β3 and β4 (Supplementary Figure S3F, Gly63–Pro112 colored sky blue and α3 colored cyan). Based on primary sequence and structural analyses, we defined two domains of Tan1: the N-terminal NFLD domain (Gly33–Asn169) and the C-terminal THUMP domain (Ala170–Asp259) (Supplementary Figure S3E). Overall, the core structure of Tan1 contains a curved seven-stranded β-sheet (β1–β7) surrounded by three α-helices (α1, α4, α5) localized on the outer surface (Supplementary Figure S3F).
Purification of a truncated form of Kre33 and full-length Tan1
We initially tried to express the gene encoding full-length Kre33 in E. coli in the absence or presence of a co-expressed Tan1 gene. This was based on the finding that S. cerevisiae Tan1 acts in collaboration with Kre33 during tRNA modification (25). However, the formation of Kre33 aggregates prevented this expression system from functioning as desired. Considering the potentially high flexibility of the FD, we constructed a Kre33 mutant lacking this domain (Met1–Leu936). The resultant truncated Kre33 was successfully expressed and purified (Supplementary Figure S5A). Tan1 was also expressed alone and was purified (Supplementary Figure S5B). Gel filtration showed that Tan1 was eluted with a single peak between bovine serum albumin (66 kDa) and carbonic anhydrase (29 kDa) (Supplementary Figure S5C), with a calculated molecular mass of 43 kDa (Supplementary Figure S5D). Given that the theoretical molecular mass of Tan1 includes a 35-kDa N-terminal as well as C-terminal His6 tags, these results indicate that Tan1 exists as a monomer in solution.
Reconstitution of eukaryotic tRNA ac4C12 modification activity
We first modified SctRNASer(CGA) using Kre33 or Tan1 alone with [14C]acetyl-CoA as the acetyl donor. However, acetylated SctRNASer(CGA) was not obtained. We then pre-incubated Kre33 and Tan1 (Kre33:Tan1, 1:1) and found that SctRNASer(CGA) was obviously modified (Figure 1A), in accordance with previous genetic data showing that Tan1 was essential for tRNA ac4C12 biogenesis (23). The modification of SctRNALeu(CAA) was also observed; however, its modification efficiency was lower than that of SctRNASer(CGA) (Figure 1A and B). To confirm that the modified product was ac4C, modified SctRNASer(CGA) and SctRNALeu(CAA) were digested with benzonase and analyzed by UPLC–MS/MS. The results showed that ac4C was readily detected in the modified tRNAs but was absent in the tRNA transcripts (Figure 1C). To further explore whether the modification occurs at position 12, C12 was mutated to U12 in SctRNASer(CGA) and SctRNALeu(CAA). The subsequent modification assay showed that SctRNASer(CGA)-C12U and SctRNALeu(CAA)-C12U were not modified by Kre33–Tan1 (Figure 1A and B), which was confirmed by UPLC–MS/MS (Figure 1C).
Reconstitution of eukaryotic tRNA ac4C12 modification activity. (A) Time-course curves of the ac4C modification of SctRNASer(CGA) transcript by Kre33–Tan1, of SctRNASer(CGA) transcript by Kre33, of SctRNASer(CGA) transcript by Tan1 and of SctRNASer(CGA)-C12U by Kre33–Tan1. (B) Time-course curves of the ac4C modification of SctRNALeu(CAA) transcript and SctRNALeu(CAA)-C12U by Kre33–Tan1. (C) UPLC–MS/MS analysis of the digested products of two SctRNA transcripts, SctRNA mutants and modification products by Kre33–Tan1. (D) Time-course curves of ac4C modification of SctRNASer(CGA) by Kre33 both without Tan1 and with increasing amounts of Tan1. The data represent mean ± standard error of the mean (SEM) (n = 3) in panels (A), (B) and (D).
We subsequently determined the optimal ratio of Tan1 to Kre33 in mediating ac4C modification of SctRNASer(CGA) by elevating the Tan1 concentration in activity measurements. Increasing the Tan1 concentration exhibited little stimulatory effect on the acetylation activity of Kre33 (Figure 1D). Thus, in the subsequent in vitro assays, the ratio of Kre33 to Tan1 was kept at 1:1.
We further determined the kinetics of Kre33–Tan1 for both SctRNASer(CGA) and SctRNALeu(CAA) in acetylation. The results showed that Kre33–Tan1 exhibited comparable Km values for both substrates; however, the kcat value for SctRNASer(CGA) was approximately twice of that for SctRNALeu(CAA) (Supplementary Table S2).
Above all, these results demonstrate that we successfully reconstituted the eukaryotic tRNA ac4C12 modification activity of Kre33 and Tan1.
Presence of a 3′-CCA terminus was important but not essential for ac4C12 modification
Biochemical and structural data have shown that the THUMP domain binds to the 3-terminal extension of tRNAs (41–43). However, it does not seem to recognize specific nucleotides at the 3′-CCA terminus (41). To understand the role of the 3′-CCA terminus (Supplementary Figure S6A) in ac4C12 modification, we first constructed two deletion mutants of SctRNASer(CGA), with deletion of CCA or GCCA (Supplementary Figure S6A). The modification level of SctRNASer(CGA)-ΔCCA or SctRNASer(CGA)-ΔGCCA reduced to ∼50% when compared with that of wild-type SctRNASer(CGA) (Supplementary Figure S6B), suggesting that GCCA terminus is important but not essential for modification. Moreover, we replaced the GCCA terminus with AUUG. The modification assay showed that wild-type SctRNASer(CGA) and SctRNASer(CGA)-AUUG were modified with comparable efficiencies (Supplementary Figure S6B). Considering that the THUMP domain of ThiI appears to contain a binding site for N73 (41,44), we mutated G73 to A73, U73 or C73. SctRNASer(CGA)-G73U displayed a similar modification efficiency with wild-type tRNA, while modification of SctRNASer(CGA)-G73A reduced to ∼50%. Meanwhile, modification of SctRNASer(CGA)-G73C exhibited a lower plateau. Given the unaltered modification efficiency of SctRNASer(CGA)-AUUG, it suggested that replacement of terminal CCA by UUG compensated the inhibitory effect of A73 in modification (Supplementary Figure S6B). Overall, these results suggested that the presence of a 3′-CCA terminus was important but not essential for ac4C12 modification.
Sequences in acceptor stem, T-arm, variable region and anticodon arm contributed little to ac4C12 modification
Reconstitution of the ac4C12 modification activity of Kre33 and Tan1 made it possible to investigate the mechanism of tRNA substrate selection during acetylation. SctRNAThr(AGU) (Supplementary Figure S7A) harbored no ac4C12 modifications in vivo (1). An in vitro acetylation assay revealed that SctRNAThr(AGU) was not modified by Kre33–Tan1 (Supplementary Figure S7B). Based on the sequence of SctRNASer(CGA) and SctRNAThr(AGU), we designed four SctRNASer(CGA) mutants by replacing various elements with their counterparts in SctRNAThr(AGU)—SctRNASer: Thr-1 (amino acid acceptor stem swapping), SctRNASer: Thr-2 (TψC stem–loop swapping), SctRNASer: Thr-3 (variable region swapping) and SctRNASer: Thr-4 (anticodon stem–loop swapping) (Supplementary Figure S7C). The anticodon stem–loop swapping (SctRNASer: Thr-4) reduced ac4C modification by ∼50%. However, swapping the amino acid acceptor stem, the TψC stem–loop or the variable region had little influence on the level of ac4C modification by Kre33–Tan1 in vitro (Supplementary Figure S7D).
These data suggest that the nucleotide sequences in the acceptor stem, T-arm, variable region and anticodon arm do not harbor the key elements for ac4C12 modification. Alternatively, element(s) shared between these regions in the two tRNAs may determine ac4C12 formation. However, the ability of Kre33–Tan1 to modify various non-substrate tRNA mutants (described below) makes this possibility unlikely.
C11–G24 base pair and G13 are indispensable for ac4C12 biogenesis
All eukaryotic ac4C modifications across various RNA species occur in the CCG motif, with the central C being the modification site. In SctRNASer(CGA) and SctRNALeu(CAA), the 11CCG13 motif was located in the D-stem/loop region. To understand the putative role of this region in ac4C12 modification, we mutated the G10–C25 base pair to A10–U25, U10–A25 or C10–G25 in SctRNASer(CGA). Simultaneously, C11–G24 was mutated to A11–U24, U11–A24 or G11–C24, respectively (Figure 2A). In comparison to wild-type SctRNASer(CGA), changes in the C11–G24 base pair completely abolished (C11G/G24C mutant) or significantly impaired (C11U/G24A and C11A/G24U mutants) acetylation of SctRNASer(CGA). Despite the impaired modification of SctRNASer(CGA)-G10C/C25G, the modification levels of the G10A/C25U and G10U/C25A mutants were not reduced, suggesting that the G10–C25 base pair is not a determinant of ac4C12 modification (Figure 2B).
The C11–G24 base pair and G13 are indispensable for ac4C12 biogenesis. (A, C, E) Design of SctRNASer(CGA) D-arm mutants. (B, D, F) Time-course curves of ac4C modification of SctRNASer(CGA) D-arm mutants by Kre33–Tan1. The data represent mean ± SEM (n = 3) in panels (B), (D) and (F).
Moreover, G23 was changed to C23 to disrupt base pairing between C12 and G23, and G23 and G24 were simultaneously changed to C23 and C24 to release the CCG motif as a single strand. Finally, A22 was substituted with C22 to form a G13–C22 base pair (Figure 2C). The single-stranded CCG motif (G23C/G24C) was found to be a poor substrate for Kre33–Tan1. However, the formation of base paired G13–C22 (A22C mutant) or unpaired C12 (G23C mutant) did not abolish ac4C12 formation (Figure 2D).
To further understand the potential role of the universally conserved G13 in ac4C12 acetylation, we assessed three further mutants: SctRNASer(CGA)-G13A, -G13C and -G13U (Figure 2E). In vitro acetylation assays showed that none of these mutants were modified by Kre33–Tan1, clearly demonstrating that G13 is an identity element for ac4C modification (Figure 2F).
These results illustrate that the C11:G24 base pair and G13 are critical elements for ac4C12 modification.
Sensitivity of tRNA acetylation to the length of acceptor stem
Cytoplasmic tRNAs undergo several processing steps before maturation, which include removal of 5′ leading and 3′ trailer sequences and addition of CCA sequence (45). Some cytoplasmic tRNAs undergo intronic splicing. A previous study reported that ac4C12 is found in the S. cerevisiae intron-containing pre-tRNASer(CGA), which lacks terminal sequences (46), raising the possibility that Kre33–Tan1 can modify pre-tRNAs. To address this possibility, we transcribed different forms of pre-tRNASer(CGA), including pre-tRNASer(CGA) containing the intron sequence (5′-UGGAAUAAAAAGUUCGGCU-3′) (pre-SctRNASer-1), the 5′-leader sequence (5′-GGAAAUCGAAAU-3′) (SctRNASer-2), the 3′-trailer sequence (5′-UUUAAUUUUU-3′, without a CCA tail) (pre-SctRNASer-3), or both the 5′-leader and 3′-trailor sequences (pre-SctRNASer-4) (Figure 3A).
Length of acceptor stem determines ac4C12 modification. (A) Design of SctRNASer(CGA) precursor mutants. (B) Modification levels of SctRNASer(CGA) precursor mutants by Kre33–Tan1. (C) Design of SctRNASer(CGA) mutants with altered acceptor stem length. (D) Modification levels of SctRNASer(CGA) acceptor stem length mutants by Kre33–Tan1. The data represent mean ± SEM (n = 3) in panels (B) and (D).
The presence of the intron (pre-SctRNASer-1) or the 3′-trailer sequence did not affect the level of ac4C12 modification by Kre33–Tan1. However, the presence of the 5′-leader sequence (pre-SctRNASer-2) substantially reduced the ac4C modification efficiency. Remarkably, modification of pre-SctRNASer-4 was nearly abolished (Figure 3B). These results suggested that a mature 5′-terminus is critical for tRNA acetylation and that, in yeast, ac4C12 modification likely occurs after the removal of the 5′-leader sequence by RNase P.
Careful inspection of pre-SctRNASer-2 and pre-SctRNASer-4 showed that the length of the acceptor stem extended from a canonical 7 bp to either 8 bp (forming a −1U–G73 base pair) or 11 bp (forming duplex between −4AAAU−1 and G73-UUU, the underlined part being the first three nucleotides of the 3′-trailor sequence) in both tRNAs (Figure 3A). Given the minor influence of the extended 3′-single strand on acetylation, stem length extension may constitute at least one of reasons for tRNA hypomodification. To explore the potential role of the 7-bp acceptor stem in ac4C12 modification, we deleted the 7A–U66 base pair to construct an SctRNASer(CGA) mutant with a 6-bp acceptor stem (SctRNASer-1). Moreover, we inserted an additional G–C base pair next to the 7A–U66 base pair to construct an SctRNASer(CGA) mutant with an 8-bp acceptor stem (SctRNASer+1) (Figure 3C). These deletions and insertions were centered on the 7A–U66 base pair since it is distant from the CCA terminal region, where the THUMP domain binds. In vitro acetylation assays showed that the modification efficiency for SctRNASer+1 was nearly abolished compared to that of the wild-type tRNA and was comparable with that of pre-SctRNASer-4 (another mutant with an elongated acceptor stem), suggesting that the addition of one base pair reduced the level of modification in both tRNA mutants. Remarkably, Kre33–Tan1 did not acetylate SctRNASer-1 (Figure 3D).
These data demonstrate that a canonical 7-bp acceptor stem is critical for ac4C12 biogenesis.
Reliance of ac4C12 biogenesis on proper D-arm orientation
The L-shaped tRNA structure is facilitated by interdomain base pairs, including the conserved 18G–U55 and 19G–C56 (47) (Figure 4A). We constructed several SctRNASer(CGA) mutants, including SctRNASer(CGA)-U55G, -C56G and -U55G/C56G, to disrupt the conformation of the L-shaped structures. Acetylation modification assays showed that disruption of a single base pair between the D-loop and the TφC loop had little influence on ac4C12 modification. Loss of interaction between both 18G–U55 and 19G–C56 due to U55G/C56G double mutations reduced the modification efficiency by 2–3-fold (Figure 4B), suggesting that a correct L-shape is important but not determinant for ac4C12 modification.
Reliance of ac4C12 biogenesis on a proper D-arm orientation. (A) Secondary structure of SctRNASer(CGA) with specific interdomain interactions indicated. (B) Modification levels of various SctRNASer(CGA) mutants by Kre33–Tan1. The data represent mean ± SEM (n = 3).
In the L-shaped structure, besides the interaction between the D-loop and the TφC loop, bases 8 and 9 (UG/UA in most cases) determine the relative orientation of the minihelix, containing acceptor stem and TφC arm, and the double helix, containing the D-arm and anticodon arm (2,48). Given the spatial localization of the CCA-terminal region (captured by the THUMP domain) and C12 in the L-shaped tRNA structure, we hypothesized that D-arm (harboring C12) orientation may be responsible for correctly positioning C12 in the active site of Kre33, which lies in its N-acetyltransferase domain. Mutation of G9 to A9 did not substantially suppress acetylation level; however, deletion of G9 caused hypomodification of SctRNASer(CGA) by Kre33–Tan1 (Figure 4B). Deletion of G9 may pose dual roles in changing the orientation of D-arm. On the one hand, it would alter the spatial angle of the D-arm relative to the minihelix (TφC acceptor stem) due to distance restriction only with a U8 base; on the other hand, base 9 frequently forms additional base pairs with other residues in the D-arm, including 9-12-23 or 9-13-22 triple (49,50). These results suggest that the correct orientation of the D-arm determines the efficiency of ac4C12 modification by Kre33–Tan1.
Converting non-substrate tRNAs to allow modification by Kre33–Tan1
The systematic analyses of the key elements in tRNASer described above showed that Kre33–Tan1 is dependent on a 7-bp acceptor stem, an 11CCG13 motif and correct D-arm orientation for efficient ac4C12 biogenesis. We surveyed all cytoplasmic tRNAs in S. cerevisiae and found that SctRNAThr(AGU), SctRNAAsn(GUU), SctRNATyr(GUA) and SctRNAIle(AAU) all contain an 11CC12 motif and a C11:G24 base pair, but no ac4C12 modification. Indeed, the first three tRNAs contained A13, whereas the last tRNA contained C13 (51) (Supplementary Figure S8A). Accordingly, in vitro acetylation assays showed that none of these tRNAs were modified by Kre33–Tan1 (Supplementary Figure S8B). Given the pivotal role of G13 in determining the ac4C12 modification, we mutated A13 or C13 in these tRNA species to G13. Consequently, all tRNA mutants with the 11CCG13 motif were successfully acetylated with various modification efficiencies (Supplementary Figure S8B). These results reinforced the idea that for a canonical tRNA, the presence of an 11CCG13 motif is sufficient to introduce ac4C12 modification, suggesting that other sequences (including the acceptor stem sequence) contributed little to tRNA acetylation.
Realization of ac4C modification of small RNAs
These findings prompted us to explore whether we can modify a non-tRNA species if they obey the requirements found to govern tRNA selection, namely the presence of a 7-bp ‘accepter’ stem, a suitable angle between the modified arm and the ‘accepter’ stem, and a CCG motif. To this end, we replaced the region from position 26 to 65 (comprising anticodon arm, variable arm, TψC arm) with six nucleotides (AAACCC) in SctRNASer(CGA), SctRNAIle(AAU)-C13G, SctRNATyr(GUA)-C13G and SctRNAThr(AGU)-C13G (all of which demonstrated substrates of Kre33–Tan1) (Figure 1A and Supplementary Figure S8). This modification maintained a proper spatial orientation of the modified CCG motif relative to the ‘acceptor’ stem while simultaneously avoiding base pairing between 8UG9 and the inserted oligonucleotides (Figure 5A). We found that these small RNAs (designated as small RNASer, RNAIle, RNATyr and RNAThr) were efficiently modified by Kre33–Tan1 (Figure 5B) but not by Kre33 alone, suggesting a crucial role of Tan1 in binding of these small RNAs (Supplementary Figure S9). The modification efficiencies of small RNAIle and small RNAThr were substantially higher than those of the small RNASer (Figure 5B). These results confirmed that the anticodon arm, variable arm and TψC arm of natural tRNA substrates contributed little to ac4C12 biogenesis. Moreover, considering that modification by Kre33–Tan1 is not influenced by the presence of an elongated 3′-terminus (Figure 3A and B), modification of even longer non-tRNA species by Kre33–Tan1 is likely achievable.
Reconstitution of small RNA ac4C12 modification activity. (A) Design of small RNAs: small RNASer, small RNAIle, small RNATyr and small RNAThr. (B) Modification levels of small RNAs by Kre33–Tan1. The data represent mean ± SEM (n = 3).
Single-point mutations in Tan1 have little effect on ac4C12 biogenesis in vitro and in vivo
After establishing a tRNA recognition mechanism using Kre33–Tan1, we aimed to identify key residues in Tan1. The THUMP domain is responsible for binding the tRNA CCA terminus, as observed in the crystal structure of ThiI-tRNA or Trm11 (41,43). Thus, based on the primary sequence alignment of Tan1 family proteins (Supplementary Figure S10) and the orientation of the side chains toward the putative tRNA-binding interface, as revealed by the AlphaFold-predicted Tan1 structure, we mutated a series of conserved residues in or in close proximity to the THUMP domain of Tan1. These mutations included five residues in the THUMP domain (R206, R213, D236, N239 and E247) and four residues in the NFLD or C-terminus (R156, K160, R267 and K268), all of which were converted to Ala (Supplementary Figures S3E and S11A). These selected residues constitute nearly all conserved residues in the core structure of the Tan1 family of proteins.
We first utilized the thermosensitive phenotype of yeast cells with Tan1 gene deletion (ScΔTan1) (24,52). We transfected plasmids containing Tan1 or its mutants into ScΔTan1 yeast and observed growth under 30 or 37°C. Complementation assays showed that the deletion of Tan1 caused growth defects at 37°C but not at 30°C (Supplementary Figure S11B), consistent with previous reports (24,52). Unexpectedly, none of the point mutations affected yeast growth at either temperature (Supplementary Figure S11B). Western blotting analyses showed that none of the mutants exhibited a lower protein level; in fact, abundance of some mutants was higher than that of wild-type Tan1 (Supplementary Figure S11C). We further purified all Tan1 mutants from the E. coli expression system and performed ac4C12 modification assays with Kre33 to understand the potential effects of the point mutations on tRNA modification in vitro. Consistent with in vivo complementation data, all the mutants showed comparable modification efficiencies (Supplementary Figure S11D and E). Together, these results suggest that the interaction between Tan1 and tRNA is coordinated by multiple amino acid residues; thus, changes at a single point are insufficient to elicit functional defects in Tan1.
Presence of ac4C modification did not elevate tRNA Tm values or charging levels
Having investigated the mechanisms underlying ac4C modification, we then turned to understanding its function. In particular, we were interested in understanding whether the ac4C modification stabilizes the tRNA structure. To this end, we transcribed SctRNASer(CGA) and SctRNALeu(CAA) and prepared ac4C12-modified SctRNASer(CGA) and SctRNALeu(CAA) using cold acetyl-CoA. RNA fragment mass spectrometry (LC–ESI-MS) analyses showed that SctRNASer(CGA) was fully modified with acetylation, while the modification efficiency of SctRNALeu(CAA) was 60.71% (Supplementary Figure S12). The concentrations of ac4C12-SctRNASer(CGA) and ac4C12-SctRNALeu(CAA) were quantified by denaturing urea-PAGE based on the linear curve obtained from serial amounts of the corresponding tRNA transcripts with known concentrations. We first determined the Tm values of these tRNAs, which showed that both SctRNASer(CGA) and SctRNALeu(CAA) exhibited similar thermostability in the absence or presence of acetylation (Supplementary Table S3), consistent with previous results obtained from archaeal tRNAs (14).
We then investigated whether ac4C12 promotes tRNA aminoacylation via the corresponding tRNA synthetases. We purified yeast cytoplasmic seryl-tRNA synthetase (SerRS) and performed tRNA aminoacylation using transcribed and modified tRNAs. The results showed that ac4C12 had little effect on stimulating tRNASer aminoacylation (Supplementary Figure S13A). To further understand the role of ac4C12 in aminoacylation in vivo, we compared the SctRNASer(CGA) and SctRNALeu(CAA) charging levels between wild-type BY4742 and ScΔTan1 yeast using northern blotting under acidic conditions. The results showed that the presence or absence of ac4C12 had little effect on the aminoacylation levels of SctRNASer(CGA) and SctRNALeu(CAA) (Supplementary Figure S13B).
Altogether, the above data demonstrated that ac4C12 alone had no influence on Tm values of tRNA transcripts, at least in vitro, or on the charging levels of tRNASer and tRNALeuin vitro and in vivo. In bacterial ribosome structures, a very tight interaction between C12 in the minor groove of D-arm and the helix 69 in 23S rRNA was observed in the P-state (53). Similar interaction between D-arm and rRNA was also observed in the eukaryotic ribosome structure (54). These observations indicate that ac4C12 modification probably substantially contributes to translocation during mRNA translation.
Discussion
Tan1 is an outlier among all THUMP-containing proteins because it lacks a catalytic domain and instead functions as a modifying enzyme partner (23,25). Using the AlphaFold algorithm (33,34), the tertiary structure of Tan1 was predicted with high confidence. We also surveyed all the crystal structures of THUMP domain-containing tRNA-modifying enzymes. The FLD domain always connects directly to the THUMP domain. For ThiI (PDB No. 4KR6) (41), Trm11 (PDB Nos. 5E71 and 6ZXW) (40,42,43), Trm14 (PDB No. 3TM4) and TrmN (PDB No. 3TMA) (55), the FLD and THUMP domains are located in the N-terminal portion of enzymes, while for CDAT8 (PDB No. 3G8Q) (56) the cytidine deaminase domain is located in the N-terminus. Pseudouridine synthase is the only protein with a THUMP domain that lacks an FLD domain, as evidenced by the human Pus10 structure (PDB No. 2V9K) (57). Considering the role of FLD in setting the proper separation between the CCA terminus and the modification site (41,43), lack of the FLD domain likely connected to the fact that Pus10 is the sole THUMP-containing member that recognizes the TφC loop and the amino acid acceptor stem but not the CCA terminus (58). Consequently, Pus10 does not use a molecular ruler mechanism (as suggested for Tan1 in the discussion that follows) to select or position the modified site.
The FLD domain in Tan1 is ∼70 amino acids longer than its counterparts in ThiI and various methyltransferases due to a large insertion between α1 and β2, leading to formation of an α2 helix within a disordered region in the predicted structure. In Thil and the methyltransferases, the FLD–THUMP domains are naturally ligated to the catalytic domain, and there is no need to mediate additional protein–protein interactions. Moreover, capture of the tRNA CCA terminus likely requires THUMP to be highly flexible, as evidenced by Trm11 structures (40,43). Even Trm11, which needs a cofactor for methylation, is the catalytic domain, not FLD or THUMP, which is responsible for the Trm11–Trm112 interaction (40). Indeed, the FLD and THUMP domains have never been shown to mediate protein–protein interactions. However, Tan1 must interact with Kre33 to acetylate tRNA (25). Given little possibility of FLD and THUMP domains in protein–protein interaction, we propose the inserted region to be a candidate region for Tan1’s interaction with Kre33. The detailed mode of interaction between Kre33 and Tan1 requires further in-depth investigation.
One of the most prominent features of the THUMP domain is the universal binding of the tRNA CCA terminus. However, several studies have observed varied effects of the CCA terminus on tRNA modification efficiency. For instance, ThiI exhibited ∼40% activity against the CCA-truncated tRNAPhe-derived minimal substrate, TPHE39A. Moreover, the replacement of ACCA with GUUA in TPHE39A led to similar modification efficiency (41). Similarly, yeast Trm11–Trm112 displays ∼60% activity against CCA-truncated tRNAIle (43). Further, archaeal Trm11 can modify CCA-truncated tRNATrp with ∼40% efficiency compared to that of wild-type tRNATrp (42), and human Pus10 can efficiently modify tRNATrp without the CCA terminus (58). However, THUMPD3–TRMT112 is unable to modify tRNALeu(CAG) without a CCA terminus (59). Despite the various contributions of the CCA terminus, nearly all THUMP domains are responsible for binding the CCA region in both solved and modeled protein–tRNA structures (41,43,56). The ability of ThiI or Kre33–Tan1 to modify the tRNA substrate with a swapped CCA-terminus suggests that the binding between THUMP and the CCA terminus is not base-specific, in accordance with the observation that THUMP binds tRNA mainly by recognizing the phosphate backbone (41). The FLD domain has been shown to participate in binding to the acceptor stem of tRNAs, accounting for the retained modification activity toward CCA-truncated tRNAs (41). We performed Ala-scanning mutagenesis targeting conserved polar amino acids to identify residues in Tan1 that are crucial for tRNA modification. However, these single-point mutations had little impact on tRNA modification both in vitro and in vivo, suggesting that the interaction between Tan1 and tRNA is mediated by a complicated network composed of a group of residues rather than being dependent on a specific residue–tRNA interaction.
Given that only class II tRNASer and tRNALeu species are modified in eukaryotes, it is easy to assume that the long variable arms of eukaryotic tRNASer and tRNALeu contribute to tRNA substrate selection. However, our extensive mutagenesis of SctRNASer(CGA) clearly showed that ac4C12 requires only a minimal set of tRNA elements, including a canonical acceptor stem, the 11CCG13 motif, and the proper orientation and angle of 11CCG13. The lack of a role for the long variable arm in modification is also echoed by the fact that introducing G13 efficiently converts several non-substrate tRNAs (all class I tRNAs lacking a long variable arm) to be susceptible to Kre33–Tan1 modification. Notably, we revealed that either lengthening or shortening of the acceptor stem was catastrophic to ac4C12 biogenesis. However, the detailed sequence of the acceptor stem seems to play little role in determining ac4C12 modification. Moreover, deletion of G9, leading to an alteration of the D-arm relative to the acceptor stem, nearly abolished the ac4C12 modification. This detrimental effect was not observed with the G9A mutation. These analyses led us to speculate that Kre33–Tan1 utilizes a molecular ruler mechanism to correctly select its substrates (Figure 6). The THUMP and FLD domains collectively capture the CCA terminus and acceptor stem and then utilize the canonical tRNA structure to precisely position the modified site (C12) into the active site of Kre33. Manipulation of the acceptor stem length or D-arm angle rendered the active site inaccessible to C12. The C11–G24 base pair and G13, both of which function as identity elements, must be recognized by Kre33 itself, as all ac4C modifications occur in the CCG, including those in mRNA or rRNA (8,25). Thus, Tan1 is unlikely to participate in 11CCG13 motif recognition because modification of mRNA or rRNA is not mediated by Tan1 (25). Most importantly, once the small RNAs meet these minimal sequence and structural requirements, they are robustly ac4C-modified, despite their lack of anticodon, variable and TψC arms. This strongly supports the rationality of our proposed model.
A model of tRNA substrate selection mechanism by Kre33–Tan1. Kre33–Tan1 relies on a 7-bp acceptor stem, 11CCG13 motif and correct D-arm orientation for efficient ac4C12 biogenesis. Recognition of the acceptor stem was not sequence-specific. The length of the acceptor stem is likely monitored by the THUMP domain in Tan1, together with the active site in Kre33 (left quadrant). Alterations in either acceptor stem length (upper right quadrant) or optimal orientation (lower right quadrant) do not precisely locate the 11CCG13 motif in the active site of Kre33.
By surveying the tRNA recognition mechanism and crystal structures of THUMP-containing tRNA-modifying enzymes, we suggested that the molecular ruler mechanism is a common strategy utilized by these enzymes (40–43,55,56), except for pseudouridine synthetase, which does not capture the CCA terminus to position the modified site (57). ThiI forms a dimer and the THUMP domain in one subunit captures the CCA terminus of one TPHE39A and the catalytic domain in this subunit accesses the modification site (U8) in another TPHE39A substate. ThiI is highly susceptible to changes in acceptor stem; combined with biochemical and structural data, this clearly demonstrates that ThiI relies on such an acceptor stem ruler for catalysis (41). Additionally, from the modeled CDAT8–tRNA structure, it was predicted that the THUMP domain of CDAT8 captures the CCA terminus of the substrate tRNA and positions the modification site (C8) at the active site (56). The molecular ruler mechanism is also evidenced by the differential spatial localization of THUMP and the catalytic domains of m2G6- (Trm14 and TrmN) (55) and m2G10-modifying enzymes (Trm11) (43). G6 is closer to the CCA terminus than G10. Thus, the shorter distance between THUMP and the methylase domains of Trm14 and TrmN enables the catalytic pocket to access G6 (55). In contrast, a longer distance between the THUMP and the methylase domains of Trm11 aids the positioning of G10 in the catalytic pocket (40,42). Very recently, another THUMP domain protein, THUMPD2, was found to modify U6 snRNA. However, whether THUMPD2 employs a molecular ruler mechanism to locate the modification site remains unclear (60). A similar distance-based ruler mechanism is also utilized by other tRNA-related proteins, probably because of the well-conserved structure of tRNAs. For example, the eukaryotic tRNA splicing endonuclease complex relies on the TSEN54 subunit to precisely locate splice sites for cleavage by TSEN2 and TSEN34 for the removal of tRNA introns (61,62). RNase P utilizes its catalytic and specific domains to bind the minihelix (TφC acceptor stem) of pre-tRNAs and measure a precise, invariant distance from the tRNA elbow region to precisely cleave its substrates (63). T-box riboswitches are bacterial gene-regulatory mRNA elements that control transcription in response to changes in the amino acid supply: in these elements, the proximal portion of stem I recognizes the tRNA anticodon and the distant portion of stem I measures the proper distance from the tRNA body to orient the tRNA CCA terminus for interaction with the anti-terminator domain (64,65). The dimeric RNase Z clamps the tRNA elbow and recognizes the first base pair and the discriminator base to measure the distance between the elbow and the 3′-region to remove the 3′-trailor sequence (66). A similar mechanism is also employed by alanyl-tRNA synthetase, in which the conserved C-Ala domain nonspecifically accommodates the tRNA elbow region, whereas the aminoacylation domain specifically recognizes the sole identity element G3–U70 to orient the CCA terminus into the aminoacylation or editing active sites (67).
The ability to reconstitute ac4C12-containing tRNAs would greatly aid investigations into their impact on tRNA structure and function. For example, it enables the study of tRNA conformation and folding kinetics in the absence or presence of acetylation at different temperatures in vitro, which is of interest given the role of ac4C12 in tRNA stability under heat stress (8,14,24). In addition, it would be possible to determine and compare the structures of unmodified and ac4C12-modified tRNAs, either free or complexed with their binding factors or ribosomes, to understand their impact on tRNA folding, various tRNA metabolic pathways and dynamic ribosomal decoding. This work not only represents the first reconstitution of ac4C12 tRNAs but also demonstrates the ability to utilize Kre33 and Tan1 to make site-specific preparations of ac4C-containing non-tRNAs, provided the RNA sequence meets the minimal requirements. Indeed, in this study, we were able to generate ac4C modifications in various small RNAs, having determined the sequence and structure components necessary for Kre33–Tan1 acetylation.
In investigating these structural features, we identified a limited role for the acceptor stem sequence, CCA terminus sequence and length of 3′-terminus of RNA substrates, leading us to suggest that Kre33–Tan1 may be able to modify longer 3′-terminus-extended RNA substrates. The ability to prepare ac4C-containing small or long RNAs will advance our understanding of roles of the acetylation in RNA life cycle events, including, but not limited to, RNA structure, folding, immunogenicity regulation and translational efficiency. Moreover, the availability of ac4C-modified RNA will facilitate the identification of potential ac4C-reader or eraser proteins. Therefore, the reconstitution of ac4C modification activity, clarification of RNA recognition mechanisms and preparation of ac4C-harboring tRNAs or non-tRNA RNA species are of great significance for future functional delineation and biogenic regulation of RNA acetylation modifications.
Data availability
All data presented in this study are available within the figures, tables and supplementary information.
Supplementary data
Supplementary Data are available at NAR Online.
Acknowledgements
We thank Dr Jin-Qiu Zhou in our institute for providing ScΔTan1 strain. We are grateful to Prof. Eric Westhof (University of Strasbourg) for insightful discussion of interaction between tRNA D-stem and the ribosome. We also thank Thermo Fisher Scientific (China) Co., Ltd for assistance in LC–ESI-MS analysis.
Author contributions: X.-L.Z. conceived the study, designed the experiments, analyzed the data, and drafted and revised the manuscript. C.-R.M. performed most of the experiments, analyzed data and drafted the manuscript. N.L. and H.L. helped in plasmid construction, Tan1 protein purification and mass spectrometry analyses. H.X. analyzed some data.
Funding
National Key Research and Development Program of China [2021YFC2700903, 2021YFA1300800]; Natural Science Foundation of China [32271300]; Strategic Priority Research Program of the Chinese Academy of Sciences [XDB0570000]; Committee of Science and Technology in Shanghai [22ZR1481300, 22JC1400503]; CAS Project for Young Scientists in Basic Research [YSBR-075]. Funding for open access charge: Strategic Priority Research Program of the Chinese Academy of Sciences.
Conflict of interest statement. None declared.
Comments