Introduction

tRNA is an adaptor molecule that enables protein synthesis by converting the triplet genetic code in mRNA into amino acids. The fidelity of base paring between mRNA codons and tRNA anticodons is monitored within ribosomes and is critical for properly incorporating the amino acids bound to the 3’ ends of tRNAs into growing polypeptides. For optimal translation, the abundances, and properties of tRNA isoacceptors are fine-tuned by diverse mechanisms, including chemical modifications[14] Dysregulation of tRNA abundance and/or structure leads to defective decoding and results in ribosome pausing and collisions, protein misfolding, stress responses and can have detrimental or lethal effects on the cell[510].

Chemical modifications of tRNA (tRNA modifications) are found in all kingdoms of life and fine-tune tRNA properties including mRNA decoding efficiency, recognition by aminoacyl-tRNA synthetases, half-life and structural stability[2, 1012]. Modifications are prevalent in the anticodon loop, particularly at the first letter of the anticodon. Modifications of the anticodon loop directly modulate codon recognition, whereas modifications in the tRNA body region primarily stabilize tRNA tertiary structure, protecting them from degradation in the cell. tRNA modifications are generated by dedicated site-specific enzymes referred to as tRNA modifying enzymes. tRNA modifications have been extensively characterized in a few model organisms[2, 13], but their profiles, regulation, and functions in non-model organisms, including bacterial pathogens, are understudied[13]

tRNA sequencing (tRNA-seq) allows for the rapid and systematic prediction of many tRNA modification sites[14, 15]. We recently developed a comparative tRNA-seq protocol to profile tRNA modifications in organisms with uncharted tRNA modification profiles; in Vibrio cholerae, this approach led to the discovery of a new RNA modification and RNA editing process[16]. tRNA-seq enables rapid prediction of modified sites through detection of reverse transcription-derived signatures, such as nucleotide misincorporation and early termination, both of which occur more frequently at modified sites. Furthermore, several chemical treatments of tRNA can convert modifications that are not recognizable as reverse transcription-derived signatures into detectable signals, expanding the repertoire of modifications that can be distinguished by tRNA-seq[1720].

Mycobacterium tuberculosis (Mtb), the agent of tuberculosis (TB), is a global pathogen that caused >10.5 million cases and over 1.5 million deaths worldwide in 2020[21]. Several studies have uncovered roles for non-Mtb mycobacterial tRNA modifications in stress responses, adaptation to environmental changes, and persister cell formation[22]. Mycobacterium bovis BCG, an organism closely related to Mtb, responds to hypoxia by reprogramming 40 ribonucleoside modifications in tRNA to facilitate translation of a subset of proteins that promote survival in hypoxic conditions[22]. To date, studies of the profiles and functions of Mtb tRNA modifications have been limited.

Here, we conducted tRNA-seq in Mtb. We assigned modifications to many of the reverse transcription-derived signatures identified, using information on the presence of homologs of known modifying enzymes in Mtb. Chemical treatments of tRNAs carried out prior to tRNA-seq increased the detectability of certain modifications. We constructed two Mtb deletion mutant strains, with deletions of mnmA and truB, and confirmed that the absence of these modification enzymes eliminated the predicted signals in tRNA-seq data. Furthermore, while deletion of mnmA in Mtb did not affect the pathogen’s growth in in vitro laboratory growth conditions, the mnmA knockout strain’s growth was attenuated in a macrophage infection model. Our findings suggest that tRNA modifications warrant further study as we unravel the complexity of Mtb infections, as they may serve as targets for new therapeutics.

Results

In silico prediction of Mtb tRNA modifying enzymes

To predict tRNA modifications in Mtb, we used Basic Local Alignment Search Tool (BLAST)[23] to identify homologues of all tRNA modification enzymes registered in Modomics in the Mtb genome[24] (Supplementary Table 1). With a stringent threshold (E-value <1×10−10), 31 Mtb genes homologous to genes encoding known RNA modification enzymes were identified (Supplementary Table 2). 23 of these genes are predicted to synthesize 16 tRNA modifications in Mtb (Supplementary Table 2 and 3), including miaA and miaB for 2-methylthio-6-isopentenyl-adenosine (ms2i6A), tsaD, tsaB, tsaE, and sua5 for N6-threonylcarbamoyladenosine (t6A), mnmA for 2-thiouridine (s2U), truB, truA, rluA, and pus9 for pseudouridine (Ψ), trm2 for 5-methyluridine (m5U), trmD for 1-methylguanosine (m1G), trmI for 1-methyladenosine (m1A), trmL for 2’-O-methylcytidine (Cm) or 2’-O-methyluridine (Um), trmB for 7-methylguanosine (m7G), two trmH for 2’-O-methylguanosine (Gm), trmR for 5-methoxyuridine (mo5U), trcM for 5-methylcytidine (m5C), dusB for dihydrouridine (D), tadA for inosine (I), and tilS for lysidine (k2C). While eight additional genes met the threshold for homology to modification enzymes, they exhibited greater similarity to a Glutamyl-tRNA synthase; Rv2992c, ribosome associated GTPases (Der; Rv1713 and Era; Rv2364c), molybdopterin biosynthesis proteins (MoeW; Rv2338c and MoeB2; Rv3116), thiosulfate sulfur transferases (CysA; Rv3117 and SseA; Rv3283) and a riboflavin biosynthesis protein (RibG; Rv1409), and likely do not correspond to tRNA modification enzymes (Supplementary Table 2).

We mined data from genome-wide Tn-seq and CRISPRi screens[25, 26] to assess the impacts of Mtb tRNA modifications on its growth. Five genes encoding Mtb tRNA modifying enzymes, trmD, tilS, tadA, tsaC2, and miaA, were reported to be essential for Mtb growth in both Tn-seq[25] and CRISPRi screens[26] (Supplementary Table 3), suggesting that the modifications they produce are critical for tRNA functions. Indeed, E. coli trmD, tilS, tadA, and tsaC2 are also essential for growth and critical for codon decoding, aminoacylation, and reading frame maintenance[2730]. The Mtb modifications synthesized by these enzymes likely have similar impacts on tRNA functions. Unexpectedly, one modifying enzyme, MiaA, is non-essential in E. coli grown in nutrient rich medium, but apparently essential in Mtb, suggesting that i6A, the modification introduced by MiaA, may have more profound roles in Mtb translation than in E. coli.

Several of the putative Mtb tRNA modifying enzymes are conserved across all three domains of life (e.g., TruB and TsaD) (Fig. 1). By contrast, some enzymes were limited to bacterial species closely related to Mtb, possibly suggesting their species-specific physiological roles, including pathogenesis. For example, TrmI (Rv2118c) homologs are widely present in Archaea and Eukaryotes, but are sparsely distributed in bacteria, where they are primarily limited to Actinomycetia, including Mtb, and several thermophilic bacterial species (e.g., Thermus thermophilus). TrmI synthesizes m1A at position 58 in eukaryotes[31] and at position 57/58 in archaea[32]; indeed, TrmI in Mtb has been proven to generate m1A at position 58[33]. Pus9 (Rv3300c) has many homologs exhibiting weak similarity across eukaryotes and bacteria. However, some species, including Pseudomonas aeruginosa, Acinetobacter baumannii, Neisseria gonorrhoeae, and several species of Actinomycetia, encode proteins showing strong homology to Mtb Pus9, suggesting that this set of pseudouridylases have a distinctive property such as substrate specificity.

Phylogenetic distribution of Mtb tRNA modifying enzyme homologs.

Heat map of log10 E-values from BLAST search results. BLAST searches were conducted against 120 manually picked organisms using Mtb tRNA modifying enzymes as queries. When one organism has multiple hits, the lowest log10(Eval) values among hits is shown. iTol[34] was used to depict the results.

tRNA modifying enzymes predicted in Mtb were observed in mycobacteria species, including Mycolicibacterium smegmatis and Mycobacteroides abscessus (Fig. 1), suggesting that tRNA modification patterns are similar among mycobacterium species.

Profiling Mtb tRNA modification sites by tRNA sequencing

To begin profiling detectable Mtb tRNA modifications, we sequenced tRNAs isolated from wild type Mtb strain H37Rv grown in 7H9 medium. In this protocol, tRNAs are first reversed transcribed to cDNA[16]. During cDNA synthesis, chemical modifications on tRNA nucleotides disrupt Watson-Crick base pairing and increase the frequency of reverse transcriptase errors, leading to incorporation of the incorrect nucleotide or early termination of cDNA synthesis[35]. These reverse transcription derived ‘signatures’ typically correspond to modified sites[14, 15] and are depicted in the heat map in Fig. 2.

Heat map of misincorporation and early termination frequency in sequencing of tRNAs from wild type Mtb.

(A, B) Heatmaps show misincorporation (A) and termination (B) frequencies at all positions across tRNAs (read 5’ to 3’). Predicted modifications are labeled based on similarity to known modifications in other organisms and the presence of the tRNA modifying enzyme homologs (Supplementary Table 3). The positions with more than 10% misincorporation in Mtb but not in E. coli are depicted in white in A. (C) M. tuberculosis tRNA modifications predicted in this study. Schematic tRNA secondary structure with sites of modifications identified either by the presence of modifying enzymes and/or tRNA-seq. Modifications and tRNA species that are not observed in E. coli are shown in red. Modifications and positions that are predicted by both RT-derived signature and the presence of the homologs of tRNA modifying enzymes are shown in yellow (without chemical treatment) and green (with chemical treatment), whereas modifications that are only predicted by the presence of the homologs are shown in light blue. Genes reported to be essential in Mtb are shown in bold.

Comparison of the reverse transcription derived signatures observed in Mtb to E. coli, where tRNA modifications have been well characterized[16] enables the prediction of the presence of common modifications, including ms2i6A, m1G, I, and k2C (Fig 2 and Supplementary Fig. 1). These predictions are strongly supported by the set of tRNA modification enzymes identified in the Mtb genome (Fig. 1 and Supplementary Table 3), including miaA and miaB (ms2i6A), trmD (m1G), tadA (I), and tilS (k2C). Some tRNA modifications were observed in Mtb but are not present in E. coli. In other actinobacteria, A58 and A59 are likely modified to m1A[36], and since trmI, the methylase that generates this modification is present in Mtb[33], most Mtb tRNAs likely contain this modification as well. Nucleoside variation in the sequence of tRNA genes can account for some of the variations in modified sites between E. coli and Mtb. For example, in Mtb, termination signatures derived from G at position 37 were detected in tRNA-Arg2, -Gln1, and -Gln2, whereas this position in these tRNAs in E. coli are modified A, such as m2A and m6A, which are silent in tRNA-seq. Since m1G induces strong termination during reverse transcription, these positions are likely modified to m1G, as observed in the Bacillus subtilis tRNA-Arg2 gene position 37 G[37].

tRNA samples were also treated with several chemical treatments prior to sequencing, to expand the set of tRNA modifications detectable by tRNAseq. These treatments included iodoacetamide (IAA), for detection of sulfur modifications,1-cyclohexyl-(2-morpholinoethyl) carbodiimide (CMC) for detection of Ψ, and alkali for detection of D and m7G. The chemical treatment protocols were first carried out with E. coli, to validate the methods.

IAA is a thiol-reactive compound that covalently attaches carboxyamidomethyl to thiolated uridines via nucleophilic substitution[38] and modified s4U is detected as C instead of U. In IAA-treated samples, position 8 and 9, corresponding to s4U in many tRNA species, had high misincorporation frequencies, confirming that IAA treatment modifies s4U, leading to elevated misincorporation (Supplementary Fig. 2). Furthermore, we observed higher misincorporation and termination signals at the positions corresponding to other sulfur modification, s2C, s2U and their derivatives, such as position 32 in tRNA-Arg3, -Arg5, -Ser3, and Arg4, and 34 in tRNA-Glu, -Gln1, and -Lys (Fig. 3 and Supplementary Fig. 2), revealing that IAA treatment facilitates the detection of not only s4U but also additional sulfur modifications, which are weakly detected without the IAA treatment.

IAA treatment promotes detection of sulfur modifications on tRNAs by enhancing termination signals.

(A, B) Heatmaps of the termination signals of E. coli tRNAs treated with (A) or without (B) IAA. Known modification sites, including sulfur modifications (s4U, s2C, s2U in white) are shown. (C) Termination frequency at s4U, s2C, and s2U sites of tRNAs treated with or without IAA.

Next, we applied IAA treatment to Mtb tRNA-seq. IAA treatment increased termination signals from position 34 in tRNA-Glu1, -Gln1, and -Lys, which contain s2U derivatives in E. coli (Fig. 4A). The 2-thiouridine modification is carried out by MnmA in E. coli[39], and a homolog, Rv3024c, of this enzyme was identified in the Mtb genome (Supplementary Table 3)[40]. We used double stranded DNA-based recombineering to delete Rv3024c in Mtb, yielding strain MtbΔmnmA[41]. Sequencing of tRNA isolated from MtbΔmnmA with prior IAA treatment showed reduced termination signals from position 34 in tRNA-Glu1, -Gln1, and -Lys in treated samples, indicating that Rv3024c plays a critical role in the modification responsible for increased termination frequency derived from position 34 in these tRNAs (Fig. 4B and 4C). Together, these observations strongly suggest that Rv3024c encodes an MnmA-like enzyme that sulfurates position 34 uridines in three Mtb tRNA isoacceptors.

Heat map and plot of early termination frequency from sequencing of tRNAs from wild type and MtbΔmnmA with and without RNA alkylation.

(A, B) Heat map of early termination frequencies across tRNA molecules and positions for WT (A) and MtbΔmnmA (B). Sulfur modification is shown in white. (C) Plot of termination frequencies at position 37 in wild type (WT) Mtb and MtbΔmnmA for lysine_UUU, glutamate_UUG, and glutamine_UUC isoacceptors. IAA; iodoacetamide.

As shown previously[42], CMC treatment increased both misincorporation and termination signatures at a subset of Ψs in E. coli tRNAs (Fig 5, and Supplementary Fig. 3 and 4). In addition, CMC-treated samples showed increased frequencies of both misincorporation and termination at position 16, 17, 20, and 20A corresponding to D, and m7G at position 46. Both modifications are known to undergo base elimination in mild alkali conditions[43]. Since these signals were also observed in the reaction condition in which CMC was not added, these signals are likely attributable to the alkali treatment that is common to both conditions.

CMC and alkali treatment facilitate detection of Ψ, D, and m7G modifications in E. coli.

(A, B) Heatmaps of the misincorporation signals of E. coli tRNAs treated with (A) or without (B) CMC. In both conditions, tRNAs are incubated in alkali condition. Known Ψ, D, and m7G sites are shown. Ψ is shown in white. (C) Misincorporation frequency at known Ψ, D, m7G sites of tRNAs treated with CMC+/alkali+ or CMC-/alkali+, and tRNAs without treatment.

Reverse transcription-derived signatures derived from alkali-treated D showed a distinctive pattern. With this treatment, when two Ds are at consecutive positions, e.g., D16 and D17, termination signals were elevated at the following position, i.e., position 18 (Supplementary Fig. 3, 5 and 6). Furthermore, alkali treatment also led to higher misincorporation frequencies at singlet Ds (Fig. 5, Supplementary Fig. 5 and 6). Thus, termination and misincorporation signatures enabled the prediction of known E. coli tRNA sites modified to D.

CMC/alkali treatment facilitated the identification of additional modifications in Mtb tRNAs. Regardless of CMC treatment, alkali treated samples showed increased misincorporation and termination frequencies derived from U located at position 16, 17, 20, and 20A (Fig. 6 and Supplementary Fig. 7). As observed in E. coli, termination signals at position 18 and 21 likely correspond to consecutive Ds at position 16 and 17, and 20 and 20A, respectively. Mtb Rv0823c is a homolog of dihydrouridylase DusB (Supplementary Table 3), which likely accounts for the synthesis of D at these positions. Furthermore, alkali treatment increased the misincorporation frequencies at G at position 46 (Supplementary Fig. 8). Since Rv0208c is a homolog of TrmB, which synthesizes m7G at position 46 in E. coli, multiple Mtb tRNA species likely contain m7G at position 46 (Fig. 6 and Supplementary Fig. 8).

Heat map of early termination frequency from sequencing of tRNAs isolated from WT and MtbΔtruB following CMC treatment.

Heat map of early termination frequencies across tRNA molecules and positions for WT (left) and MtbΔtruB (right). Termination signals derived from position 55 are shown in white.

CMC treatment also increased the termination frequency at several sites. Termination signatures derived from position 55, which is exclusively uridine in all tRNA species, increased in most tRNA species, suggesting that Mtb tRNAs contain pseudouridines at this position[37]. Rv2793c is an Mtb homologue of E. coli TruB and deletion of Rv2793c reduced the termination frequencies at this position in tRNAs isolated from MtbΔtruB (Fig. 6 and Supplementary Fig. 9)[41]. Together, these observations suggest that Rv2793c encodes a TruB-like enzyme that modifies position 55 uridines to Ψ across tRNA species. Furthermore, the presence of a TruA homolog in Mtb (Rv3455c) suggests that in multiple tRNA species U at positions 38-40 can be modified to Ψ. Indeed, the termination signatures derived from positions 38 and 39 increased depending on CMC treatment, strongly suggesting that these positions are modified to Ψ (Fig. 6 and Supplementary Fig. 9).

In total, among 16 tRNA Mtb tRNA modifications predicted by the presence of tRNA modifying enzymes (Fig 1 and Supplementary Table 3), 9 species of modifications were detected based on reverse-transcription derived signatures (Fig. 2C).

Growth of MtbΔmnmA is attenuated in a macrophage infection model

To address whether Mtb tRNA modifications impact the pathogen’s growth in the host environment, we used the MtbTnDB transposon insertion sequencing (Tn-seq) database[44, 45] to determine if transposon insertions in genes encoding tRNA modifying enzymes have been associated with in vivo growth defects. Transposon insertions in mnmA were reported to attenuate Mtb growth in mice infected with a library of Mtb transposon mutants, suggesting that mnmA facilitates Mtb growth in vivo. We found that the growth of WT and MtbΔmnmA were similar in 7H9 medium (Fig. 7), suggesting the absence of s2U modification at position 34 does not impair Mtb growth in culture. In contrast, the MtbΔmnmA mutant was significantly impaired for growth in a macrophage infection model[45] (Fig. 7). Defective growth of the MtbΔmnmA mutant was also observed in macrophages treated with all-trans retinoic acid (ATRA), which promotes macrophage control of Mtb infection[46]. These observations strongly suggest that modification of U to s2U by MnmA facilitates Mtb growth in macrophages.

MtbΔmnmA is attenuated in a macrophage infection model.

Top: Wild type and MtbΔmnmA do not display growth differences in 7H9 medium. Bottom: Auto-luminescent wild type and ΔmnmA Mtb strains were diluted to a multiplicity of infection of 2 bacteria per mouse bone marrow-derived macrophage with or without all-trans retinoic acid (ATRA). Survival was measured by luminescence and normalized to luminescence at time 0.

Discussion

Here, we profiled Mtb tRNA modifications with tRNA sequencing to provide the first maps of the tRNA modification landscape in this global pathogen. In total, nine modifications, including six modifications without chemical treatment, were identified based on reverse transcription derived signatures. CMC/alkali treatment and IAA treatment further identified Ψ and m7G and sulfur modifications, respectively. Although we did not chemically validate the modifications predicted by tRNA-seq with mass spectrometry, the identification of Mtb homologs of tRNA modifying enzymes strongly bolsters the RT-signature-based predictions. Furthermore, the deletion of truB and mnmA genes in Mtb eliminated the respective modification signatures of pseudouridine and s2U, validating that the enzymes encoded by these genes synthesize these modifications. Finally, the growth defect of the ΔmnmA strain within macrophages but not in a nutrient-rich medium suggests that s2U tRNA modification facilitates Mtb adaptation to the host intracellular environment.

IAA treatment was developed for detecting s4U modification in pulse-chase experiments to measure RNA turnover[38]. We found that this treatment enhances the RT-signatures of additional sulfur modifications as well, including s2C and s2U, indicating that IAA should have general utility in tRNA-seq-based profiling of sulfur modifications, including in studies tracking changes in tRNA sulfuration[47, 48].

Multiple modifications do not cause reverse transcriptase errors. Although tRNA-seq provides a simple and rapid method for profiling the landscape of modifications in all tRNA species, this approach does not enable comprehensive identification of modifications. Here, several modifications, including t6A, m5U, mo5U, m5C, Cm, Um, and Gm, predicted by the presence of Mtb homolog of tRNA modifying enzymes, such as TsaC2, TsaB, TsaD, TsaE, Trm2, TrcM, TrmR, TrmL, and TrmH, were not detected in tRNA-seq. Additionally, CMC treatment may not determine all Ψ positions, thus, the target of Rv3300 and Rv1540 remain unclear. Since these genes are similar to E. coli rluA, which also targets rRNA, these genes may target rRNAs instead of tRNAs. Mapping and further analysis of these modifications will require different approaches such as tRNA purification and RNA mass spectrometric analysis.

Several strong RT signatures were detected in Mtb tRNAs but not in E. coli (Fig. 2A), including G45 in tRNA-Gly2. Most of these signals are strict misincorporations of C or T at A or G position, respectively. Since modifications at a purine position, e.g., m1G and m1A, generally cause random misincorporation of the other three nucleosides, these Mtb-specific misincorporation signatures are not likely to be derived from modification. Further mass spectrometric analysis of purified tRNAs will be necessary to assess whether these signals are indeed derived from Mtb-specific modifications; such Mtb-specific modifications could be related to Mtb pathogenesis and/or survival in the host.

The role of s2U in Mtb appears to be unusual. s2U is a universally conserved modification observed in all three domains of life at the wobble position in the anticodon. This modification enhances the stacking of s2U with U35 to stabilize the anticodon structure, facilitating codon-anticodon interactions. s2U is also recognized by multiple aminoacyl-tRNA synthetases for efficient amino acylation[12]. Although elimination of this sulfur modification causes severe growth phenotypes in most organisms[39, 49], unexpectedly, the deletion of mnmA in Mtb did not attenuate growth of the pathogen in vitro, suggesting that the Mtb requirement for s2U modification differs from other organisms. The marked growth retardation of ΔmnmA strain within macrophages indicates the specific requirement of this modification within host cells. This modification may be necessary for maintaining general translation efficiency inside host cells and/or facilitate the expression of specific genes that are necessary for survival within macrophages. In fact, lack of mnmA is reported to sensitizeE. coli to oxidative stress, raising the possibility that s2U modification promotes Mtb growth under oxidative stress elicited by the host.

Differential codon usage between house-keeping genes and virulence genes could contribute to the differential growth phenotypes observed in vitro and in vivo in the ΔmnmA mutant. Among multiple codons encoding Glu, Gln, and Lys, s2U modification dependent codon usage might be preferentially distributed in genes associated with intracellular growth. For example, Mtb has two tRNA isoacceptors, tRNA-Glu1(s2UUC) and tRNA-Glu2(CUC), to decipher two Glu codons, GAA and GAG. According to the wobble paring rule, GAA is only decoded by tRNA-Glu1(s2UUC), whereas GAG is decoded by both tRNA-Glu1(s2UUC) and tRNA-Glu2(CUC); i.e., GAG can be deciphered by an s2U-independent tRNA. Thus, genes required for intracellular growth might be enriched with GAA, an s2U-dependent codon. Similar codon usage differences could be present in Gln and Lys codons deciphered by s2U-containing tRNAs.

In most organisms, s2U is further modified into derivatives containing an additional chemical moiety at position 5[2, 50, 51]. However, Mtb does not contain apparent homologs of the tRNA modifying enzymes that introduce the additional modifications to s2U. Thus, Mtb may contain s2U or s2U derivatives synthesized by other types of enzymes. Additional analyses to elucidate the structures of modified s2U in Mtb are warranted.

Another unexpected finding is the presence of a TrmR homolog in Mtb. TrmR in Bacillus subtilis is a methylase that converts a 5-hydroxyuridine (ho5U) into a 5-methoxyuridine (mo5U) at the tRNA wobble position, suggesting that Mtb possesses mo5U. However, Mtb does not encode an apparent homolog for the enzyme that mediates the RNA hydroxylation reaction that yields ho5U, the substrate for TrmR. Although we cannot rule out the possibility that Mtb TrmR methylates a substrate other than RNA, it is possible that Mtb utilizes an unknown hydroxylation pathway for synthesizing ho5U.

Our findings serve as a valuable starting point for the research community to continue characterizing the physiological roles and mechanisms of Mtb tRNA modifications. Since tRNA-seq offers an efficient and scalable platform for surveying changes in tRNA modifications, this approach will be valuable across growth conditions, and may be extended to growth inside host cells. Finally, further studies elucidating the mechanisms by which tRNA modifications facilitate Mtb growth in host cells should be valuable for designing new therapeutics for tuberculosis.

Acknowledgements

We appreciate all the members of Waldor lab for fruitful discussion and comments on the manuscript. We also thank Gregory Babunovic for his valuable assistance setting up the macrophage infections used here. This work is supported by NIH/NIAID grants to M.K.W. (R01AI-042347) and E.J.R. (P01AI-095208), a Dean’s Innovation Award from Harvard Medical School to E.J.R. and the Howard Hughes Medical Institute (HHMI) to M.K.W.

Materials and Methods

Bacterial strains and growth conditions

Mtb strains were grown from frozen stocks into Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.05% Tween-80, and ADC (5 g/L bovine serum albumin, 2 g/L dextrose, 3 μg/ml catalase). Cultures were incubated at 37 °C. Strains were grown to mid log-phase for all experiments (OD600 0.4-0.6). Growth was measured on a BioTek plate reader for in vitro growth by measuring OD600 every 24 hours.

Bacterial strain construction

Supplementary Table 4 depicts the strains, plasmids, primers, and recombinant DNA used for this study. Plasmids were built by restriction digest of a parental vector and inserts were prepared by Gibson assembly[52]. Plasmids were isolated from E. coli and confirmed via Sanger sequencing carried out by Genewiz, LLC (Massachusetts, USA).

Deletion mutants

The knockout strain MtbΔmnmA::zeo (zeocin) was built using double-stranded recombineering in the parental Mtb strain H37Rv. A linear dsDNA fragment was constructed using stitch PCR with the primers listed in Supplementary Table 4 which consisted of a 500bp region upstream of mnmA (Rv3024c)), 500 bp downstream region, and a lox-zeo-lox fragment. This cassette was transformed into an H37Rv recombineering strain as described[53] and plated on 7H10 + zeocin plates.

Homology search

Local BLAST was performed to search for Mtb homologs of tRNA modifying enzymes. First, the uniport IDs of tRNA modifying enzymes were obtained from Modomics[24], and 12 proteins were manually added to the list [16,54−58], including Q47319/TapT, P24188/TrhO, P76403/TrhP, O32034/TrhP1, O32035/TrhP2, P36566/CmoM, and Q87K36/TrcP, O34614/MnmM, Q8N5C7/DTWD1, Q8NBA8/DTWD2, O32036/TrmR, Q9KV41/AcpA. Uniprot ID provides a fasta file of tRNA modifying enzymes from the Uniprot database[59]. A blast database file was generated by ‘makeblastdb’ script using a fasta file of Mtb proteins (H37Rv strain) retrieved from NCBI. Then, the homologs of tRNA modifying enzymes were searched against the Mtb protein database using the fasta file of tRNA modifying enzymes as a query. Output format is defined by the following script: -outfmt “7 qacc sacc stitle score qcovs evalue pident” -evalue 1e-10. The output file was modified by excel (Supplementary Table 4).

Phylogenetic analysis of Mtb tRNA modifying enzyme homologs

Local BLAST was conducted to search for homologs of Mtb tRNA modifying enzymes in 120 manually picked organisms across three domains of life. A custom database was generated by combining the fasta files of organisms’ proteins retrieved from NCBI. Homologs of eighteen Mtb tRNA modifying enzymes were searched against the custom protein database using local blast. Log10Evalues were visualized by iTol[34] with a phylogenetic tree generated by phyloT[60].

tRNA sequencing

Extraction of total RNA

Strains were grown to mid-log phase with the appropriate antibiotics and inducing agents described above. RNA was collected at the same OD600 for each strain (between 0.4-0.6). Cells were left on ice for 20 minutes, then pelleted by centrifuging at 4,000 rpm for 10 minutes at 4 °C. Pellets were resuspended in 0.5-1 mL of TriZol (Life Technologies) and lysed using a BeadBug microtube homogenizer (Millipore Sigma). 200 μL of chloroform was added to each tube, after which samples obtained from Mtb strains were removed from biosafety level 3 precautions. Samples were centrifuged at 15,000 rpm for 15 minutes at 4 °C and the aqueous layer was collected into a fresh tube. To the original tube, 250 μL of sodium acetate buffer (300 mM sodium acetate pH 5.2 and 10 mM EDTA pH 8.0) was added, and samples were vortexed at 4 °C for 5 minutes then centrifuged at 15,000 g for 15 minutes at 4 °C. The aqueous layer was added to the fresh sample-containing tubes. 400 μL chloroform was added, and tubes were briefly vortexed and then centrifuged at 15,000 rpm for 1 minute at 4 °C. The aqueous phase was collected into a fresh tube and RNA recovered by ethanol precipitation. RNA pellets were resuspended in 10 mM sodium acetate pH 5.2 and stored at −80 °C until processed for sequencing. Total RNA samples were alkali-treated prior to tRNA extraction to deacylate all tRNAs (1 hour at 37 °C in 100 mM Tris-HCl pH 9.0).

Isolation of tRNA fraction

1−2 μg of total RNA was run on a 10% TBE-UREA gel (ThermoFisher Scientific) at 250 V for 1 hour. Gels were stained with SYBR Gold (ThermoFisher Scientific), and tRNA was excised. Excised gels containing tRNA fractions were mashed in RNAse-free tubes, and 300 μL elution buffer (300 mM NaOAc pH 5.5, 1 mM EDTA pH 8.0, 0.10% SDS) was added to each tube. Samples were shaken on a thermoshaker (Eppendorf) for 1−4 hours at 37 °C and supernatant was collected using an Ultrafree filter column (Millipore Sigma). tRNA was recovered by isopropanol precipitation.

tRNA dephosphorylation

tRNA was dephosphorylated using QuickCIP (New England BioLabs) according to manufacturer instructions, and tRNA was collected by phenol-chloroform extraction followed by isopropanol precipitation.

Iodoacetamide (IAA) treatment

IAA treatment was performed as described[38]. Briefly, 500 ng of total RNA is combined with 10 mM of iodoacetamide, 50 mM NaPO4 pH 8.0, and 50% DMSO in a final volume of 50 μL. Reactions were incubated at 50 °C for 15 minutes and quenched with DTT.

CMC treatment

CMC treatment was carried out as described in ref. Briefly, 2.5 μg tRNA fraction in 0.5 μl was mixed with 15 μl CMC-BEU buffer with or without CMC (0.34 M or 0 M CMC, 7 M urea, 4 mM EDTA, and 50 mM bicine pH 7.9) and incubated at 37 °C for 20 min. Adding 100 μl CMC stop solution (0.3 M NaOAc pH 5.2 and 100 mM EDTA) quenched the reaction. RNA was desalted with PD-10 desalting column (Cytiva) and recovered by ethanol precipitation. RNA was dissolved in 40 μl of 50 mM sodium carbonate buffer (pH 10.4) and incubated at 37 °C for 4 hours, followed by ethanol precipitation.

Adapter ligation

0.5 μL RNase inhibitor was added to 3.5μL dephosphorylated tRNA (200−250 ng tRNA) and samples were boiled at 80 °C for 2 minutes. Boiled tRNA was mixed with 12 μL PEG buffer mix (10 μL 50% PEG8000, 2 μL 10 × buffer B0216S; New England Biolabs). 3 μL of 5’ adenylated linkers (Supplementary Table 4) were added (33 pmol/μL) along with 1 μL T4 RNA ligase 2 truncated (New England BioLabs) and incubated at 25 °C for 2.5 hours. Samples were recovered by isopropanol precipitation and run on a 10% TBE-Urea PAGE gel for 40 minutes at 250 V. Ligated products were recovered by gel excision as described above.

Reverse transcription

Identical quantities of samples with different adapter sequences were pooled for reverse transcription for a total of 200−250 ng tRNA. Reverse transcription was performed by combining 2.1 μL dephosphorylated tRNA with 100 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 1.25 μM RT primer (Supplementary Table 4), 450 mM NaCl, 5 mM MgCl2, 5 mM DTT, 500 nM TGIRT (InGex), and 15% PEG8000 in a final volume of 9 μL. Samples were incubated at 25 °C for 30 minutes, after which 1 μL 10 mM dNTPs (New England BioLabs) were added and reactions incubated at 60 °C for 1 hour. 1.15 μL NaOH was added, and samples were boiled for 15 minutes and run on a 10% TBE Urea PAGE gel at 250V for 1 hour. Reverse transcription products were excised from gels and cDNA recovered by isopropanol precipitation. Linear single-stranded cDNA was circularized using CircLigase II (Lucigen) in accordance with manufacturer instructions.

PCR of tRNA libraries

PCR reactions were set up using HF Phusion according to the manufacturer’s instructions using a universal reverse primer (Supplementary Table 4) and a different index primer for each pool of samples. PCR reactions were aliquoted into 4 tubes and collected after 6, 8, 10, and 12 cycles. Samples were run on a Native TBE PAGE gel (ThermoFisher Scientific) at 180 V for 50 minutes, and amplified products were cut from the same cycle for each sequencing run. Samples were recovered by gel excision and isopropanol precipitation.

Sequencing

Sequencing was performed on a MiSeq instrument (Illumina) using 150 bp single end reads with a version 3, 150 cycle kit.

Analysis

3’ linker sequences and two nucleotides at the 5’ end were trimmed using cutadapt and fastx-trimmer. Bowtie v1.2.2 was used with default settings to map reads to reference Mtb tRNA sequences retrieved from Mycobrowser[40] (Supplementary Table 5). Mpileup files were generated using samtools (samtools mpileup -I -A --ff 4 -x -B -q 0 -d 10000000). For analysis of termination frequencies, 5’ end termini of mapped reads were piled up using bedtools genomecov (option, -d -5 - ibam). The number of 5’ termini at each tRNA position was divided by the total number of mapped termini at that position plus all upstream (5’) positions.

Macrophage infection

Auto-luminescent wild type and ΔmnmA Mtb strains grown to the same OD600 were pelleted by centrifugation and prepared in RPMI media by soft spinning as described[61]. Briefly, cells were washed, pelleted, resuspended, and centrifuged at 121 g, with the top half of the centrifuged supernatant used. Suspensions were diluted to a multiplicity of infection of 2 bacteria per mouse bone marrow-derived macrophage by determining the OD600. Macrophages were infected for 6 hours, followed by a PBS wash and addition of RPMI with or without all-trans retinoic acid (ATRA). ATRA promotes macrophage control of Mtb infection[46] and was used to assess strain survival in an increasingly restricted macrophage environment. Survival was measured by luminescence in a BioTek plate reader and normalized to luminescence reads at time 0.