Landscape of Post-Transcriptional tRNA Modifications in Streptomyces albidoflavus J1074 as Portrayed by Mass Spectrometry and Genomic Data Mining

ABSTRACT

Actinobacterial genus Streptomyces (streptomycetes) represents one of the largest cultivable group of bacteria famous for their ability to produce valuable specialized (secondary) metabolites. Regulation of secondary metabolic pathways inextricably couples the latter to essential cellular processes that determine levels of amino acids, carbohydrates, phosphate, etc. Post-transcriptional tRNA modifications remain one of the least studied aspects of streptomycete physiology, albeit a few of them were recently shown to impact antibiotic production. In this study, we describe the diversity of post-transcriptional tRNA modifications in model strain Streptomyces albus (albidoflavus) J1074 by combining mass spectrometry and genomic data. Our results show that J1074 can produce more chemically distinct tRNA modifications than previously thought. An in silico approach identified orthologs for enzymes governing most of the identified tRNA modifications. Yet, genetic control of certain modifications remained elusive, suggesting early divergence of tRNA modification pathways in Streptomyces from the better studied model bacteria, such as Escherichia coli and Bacillus subtilis. As a first point in case, our data point to the presence of a non-canonical MiaE enzyme performing hydroxylation of prenylated adenosines. A further finding concerns the methylthiotransferase MiaB, which requires previous modification of adenosines by MiaA to i6A for thiomethylation to ms²i6A. We show here that the J1074 ortholog, when overexpressed, yields ms²A in a ΔmiaA background. Our results set the working ground for and justify a more detailed studies of biological significance of tRNA modification pathways in streptomycetes.

IMPORTANCE Post-transcriptional tRNA modifications (PTTMs) play an important role in maturation and functionality of tRNAs. Little is known about tRNA modifications in the antibiotic-producing actinobacterial genus Streptomyces, even though peculiar tRNA-based regulatory mechanisms operate in this taxon. We provide a first detailed description of the chemical diversity of PTTMs in the model species, S. albidoflavus J1074, and identify most plausible genes for these PTTMs. Some of the PTTMs are described for the first time for Streptomyces. Production of certain PTTMs in J1074 appears to depend on enzymes that show no sequence similarity to known PTTM enzymes from model species. Our findings are of relevance for interrogation of genetic basis of PTTMs in pathogenic actinobacteria, such as M. tuberculosis.

INTRODUCTION

Representatives of actinobacterial genus Streptomyces (streptomycetes) constitute one of the largest groups of bacteria in terms of a number of validly described species, as well as abundance in terrestrial and aquatic ecosystems (1). Interest in this genus is fueled by their ability to produce bioactive compounds that have found widespread medical use, most notably as antibiotics. Studies of the regulation of antibiotic production in Streptomyces are an important research venue for 2 reasons. First, such studies would provide new insights and tools to generate overproducers of industrially important compounds. Second, they will reveal how streptomycetes limit the expression of most biosynthetic gene clusters (BGCs) for specialized metabolism, effectively leading to the phenomenon of silent BGCs (2). The latter refers to the fact that the majority of BGCs within a given genome are not transcribed or transcribed at a level too low to lead to accumulation of the natural product (3). Importance of certain tRNA species for robust expression of streptomycete BGCs has been documented in several studies (4, 5), although the mechanism(s) behind the observed effects are not well understood.

Since 2015, we have been working on identification and analysis of genes for post-transcriptional tRNA modifications (PTTMs) in Streptomyces (6). Our results provided ample evidence of the importance of certain PTTMs for morphogenesis and antibiotic production in streptomycetes. Particularly, even partially derailed conversion of adenosine residues in position 37 of tRNA_XXA into hypermodified 2-methylthio-N6-(cis-hydroxyisopentenyl) adenosine (ms²io⁶A) was shown to decrease moenomycin production by Streptomyces viridosporus(ghanaensis) ATCC14672 (7) and antibacterial activity of cell extracts of Streptomyces albidoflavus (albus) J1074 (8). However, the aforementioned PTTM is just one of over 100 described for bacterial tRNAs (9). The full chemical diversity of modified ribonucleosides remains uncharacterized and their biological significance for Streptomyces largely unknown. To the best of our knowledge, a single attempt to assess such a diversity was undertaken in 2015 for streptomycin producer Streptomyces griseus ATCC13350 (10). Authors of this study revealed 17 modified nucleosides, a significantly lower number than that observed for Bacillus subtilis, another Gram-positive bacterium (11). Particularly, 2- and 4-thiouridines (s²U, s⁴U) were reported to be absent, despite highly conserved nature of these modifications across different domains of life (12), and a clear presence of genetic determinants for s²U and s⁴U biosynthesis in Streptomyces genomes (6). On the other hand, both previous and our results point to the presence of certain modifications in Streptomyces, which are known from model bacteria, but which have no genetic determinants yet assigned. For example, this is true for the aforementioned oxidized version of ms²i⁶A, ms²io⁶A, as the gene for respective oxidase eluded unambiguous identification (8). These discrepancies lead us to believe that current knowledge of Streptomyces PTTMs is fragmentary and can be improved via more rigorous genome mining efforts (to identify missing PTTM genes) and mass spectrometry (MS)- based analysis of tRNA hydrolyzates. We set out to perform such studies in a single experimental model, S. albidoflavus J1074, allowing direct correlation of genomic and MS data. Strain J1074 is one of the most popular chassis for heterologous expression of antibiotic biosynthesis genes (13). Therefore, our results will be of direct relevance for a biotechnologically valuable platform. Here, we describe a set of tRNA modifications revealed by MS approaches in J1074, and the corresponding list of putative PTTM genes unearthed in the J1074 genome. We also report our ongoing efforts to deeper understand hypermodification ms²io⁶A. Particularly, failure of knockout of the XNR_1636 gene for hydroxylase MiaE homolog to block the production of modified nucleosides ms²io⁶A and io⁶A in S. albidoflavus implies that genetic control of this PTTM in J1074 (and, likely, other streptomycetes) is different from known precedents. Early evolutionary divergence of actinobacteria from other bacteria (14) may lead to the situation where certain orthologous proteins share sequence similarity too low to be detected via structural bioinformatics approaches (15). This necessitates further rigorous genome mining and experimental work on all candidate genes for tRNA modifications in Streptomyces.

RESULTS

S. albidoflavus J1074 produces a diverse set of modified nucleosides.

Our initial studies of PTTM genes in S. albidoflavus J1074 have already shown the presence of different PTTMs in this strain (8). However, analysis of the entire set of PTTMs in this strain was not the main focus of previous studies, and the liquid chromatography (LC) techniques used in the latter are less informative than modern liquid chromatography-tandem mass spectrometry (LC-MS)/MS approaches. We therefore decided to identify all possible modified nucleosides in S. albidoflavus using an extensive LC-MS data set for the wild-type strain (SAM2) grown in either replete (TSB) or defined glycerol-based minimal (NL5) media. In this study, we employed an LC-MS/MS neutral loss scan (NLS) technique as it is conceptually outlined in previous work (16). The NLS approach can detect all nucleosides containing a ribose, regardless of the nucleobase, based on the fragility of the glycosidic bond. The detection scheme was here adjusted to also scan for ribose-methylated nucleosides. Overall, it provides a wealth of information in the form of nucleoside mass, its LC retention time and fragmentation pattern. Its application showed the same qualitative pattern of PTTMs in extracts prepared from 24 h and 45 h old cultures. We therefore worked mainly with 45 h old cultures as the modified nucleoside were more abundant at this time point compared to 24 h. Our findings are summarized in the Table 1. We also identified a few modified nucleosides known to be present in rRNA (11), most likely due to rRNA contamination of our tRNA extracts. These nucleosides are listed in Table S1, electronic supplemental material (ESM). LC-MS/MS data supporting identification of the nucleosides listed in Table 1 and Table S1 are given in ESM (Fig. S1 to S30).

TABLE 1.

Nucleosides from S. albidoflavus J1074 total tRNAs detected by LC-MS/MS

Nucleoside	MH+a	NLSb	Rt –stdc	Rt –obsd	WT – TSBe	WT –NL5f
D, dihydrouridine	247	132	3.8	3.8	+	+
Ψ, pseudouridine	245	36	4.0	4.1	+	+
m1A, 1-methyladenosine	282	132	4.2	4.4	+	+
m7G, 7-methylguanosine	298	132	6.3	6.7	+	+
m5C, 5-methylcytidine	258	132	8.4	8.4	+	+
Cm, 2′-O-methylcytidine	258	146	9.3	9.3	+	+
I, inosine	269	132	9.9	9.9	+	+
m5U, 5-methyluridine	259	132	10.9	10.8	+/−g	+
Um, 2′-O-methyluridine	259	146	12.1	12.1	+	+
m5Cm, 5,2′-O-dimethylcytidine	272	146	12.5	12.6	+	+
s4U, 4-thiouridine	261	132	13.2	13.0	+	+/−
m1G, 1-methylguanosine	298	132	13.4	13.4	+	+
Gm, 2′-O-methylguanosine	298	146	13.6	13.6	+	+
t6A, N6-threonylcarbamoyladenosine	413	132	19.8	19.3	+	+
m2A, 2-methyladenosine	282	132	19.9	20.0	+	+
m6A, N6-methyladenosine	282	132	20.8	20.9	+	+
ms2A, 2-methylthioadenosine	314	132	25.3	25.4	+/−	+/−
io6A, N6-(cis hydroxyisopentenyl) adenosine	352	132	28.3	28.4	+	+
ms2io6A, 2-methylthio- N6-(cis hydroxyisopentenyl) adenosine	398	132	34.1	34.2	+	+
i6A, N6-isopentenyladenosine	336	132	34.8	34.8	+	+
ms2i6A, 2-methylthio- N6-isopentenyladenosine	382	132	37.8	38.0	+	+/−
Modified nucleosides detected in the absence of authentic standard
cmnm5s2U, 5-carboxymethylaminomethyl-2-thiouridine	348	132	–	9.0	+	+
cmnm5s2Um, 5-carboxymethylaminomethyl-2′-O-methyl-2-thiouridine	362	146	–	11.6	+/−	−
nm5s2U, 5-aminomethyl-2-thiouridine	290	132	–	15.0	+	+
s2Um, 2-thio-2′-O-methyluridine	275	146	–	21.1	+	+/−
cmnm5U, 5-carboxymethylaminomethyl uridine	332	132	–	36.7	+	+/−

^aCation m/z.

^bFragmentation pattern of the nucleoside, e.g., either loss of ribose (132 Da) or 2′-O-methylated ribose (146 Da), or pseudouridine dissociation (36 Da).

^cRetention time of the nucleoside standard, wherever available (−, absence of the standard).

^dRetention time for the presumed modified nucleoside observed in this work.

^ePresence of the modified nucleoside in total tRNA sample prepared from S. albidoflavus biomass grown in TSB for 45 h.

^fPresence of the modified nucleoside in total tRNA sample prepared from S. albidoflavus biomass grown in NL5 for 45 h.

^gAbundance of the modified nucleoside was very low (equal to or less than 10³ ion counts).

We reliably identified extracts 21 modified nucleosides in SAM2 tRNA, including 17 previously reported for S. griseus (10). Also, 5 mass-peaks were tentatively assigned to various modifications of uridine in the wobble position, e.g., (c)mnm⁵(s)²U(m). Absence of the authentic standards as well as low abundance of some of the predicted PTTMs precluded their unambiguous determination. However, we note that some of the tentatively identified PTTMs (cmnm⁵s²U, s²Um) were mapped in Mycobacterium tuberculosis BCG, and their chromatographic mobility relative to well-established modifications in published (17) and our data sets are in good agreement. Four PTTMs were identified for the first time in Streptomyces, namely, m⁵C, s⁴U, m⁵U, and ms²A; the latter is present in trace amounts in the wild type (Table 1), but its level can be increased as described below. We did not detect any of 5-hydroxyuridine-based hypermodifications in S. albidoflavus known to be present in Mycobacterium, such as mo⁵U, cmo⁵U, and mcmo⁵U (17).

Gene XNR_1636 is not involved in production of hydroxylated versions of ms²i⁶A and i⁶A.

In Streptomyces, the ms²io⁶A tRNA modification is the best studied one, both genetically and physiologically. Yet, the genome of J1047 does not encode an ortholog of the MiaE protein, which was reported to catalyze the conversion of ms²i⁶A and i⁶A into ms²io⁶A and io⁶A, respectively (8). Therefore, we attempted to address this issue experimentally before undertaking the in silico search of PTTM proteins in J1074. Our initial hypothesis was that, in streptomycetes, hydroxylation of ms²i⁶A and i⁶A might be mechanistically similar to that described in enterobacteria and involve nonheme diiron monooxygenase (18, 19); its level of amino acid sequence similarity to enterobacterial counterparts might be too low to be recognized by pairwise alignment–based methods. We, therefore, systematically screened the J1074 genome for proteins harboring a diiron domain of characteristic 4 alpha-helix bundle fold as revealed by Pfam (20) and HHPred (21) search engines. We finally zeroed in on a single protein Xnr_1636, a MiaE homolog, which fits all search parameters (Fig. S31, ESM). Orthologs of Xnr_1636 are omnipresent in Streptomyces and some of them are annotated as MiaE hydroxylase. We generated a XNR_1636 knockout strain, named Δxnr_1636, as detailed in Fig. S31. The mutant produced sparse and irregular spore chains, in contrast to vigorous sporulation of the parental strain; neither bioassays nor LC-MS revealed differences in production in the spectrum of small molecules produced by SAM2 and Δxnr_1636 (Fig. S32, ESM). LC-MS/MS analysis of total tRNA hydrolysates unequivocally showed that Δxnr_1636 still produced hydroxylated derivatives of ms²i6 and i⁶A (Fig. 1), similar to the parental strain (see Fig. S22 to 25, ESM). Therefore, we conclude that hydroxylation of ms²i⁶A is likely governed by an enzyme showing no significant similarity to enterobacterial MiaE proteins.

FIG 1.

Knockout of XNR_1636 did not affect the production of io⁶A and ms²io⁶A. Composite extracted ion chromatogram (CEIC) shows that hydrolyzates of total tRNA extracts from the Δxnr_1636 strain (grown in TSB for 48 h) contained all the expected mia gene-controlled modifications, as they are found in wild-type strain (see ESM Figs. S20-S24). Identity of ms²io⁶A eluting at 34.2 min is shown as an inset to the CEIC (MS and NLS data). Chromatograms shown on the figure represent typical result of 3 biological repeats.

J1074 genome mining reveals candidate genes for most of the identified PTTMs.

Our previous understanding of genetics behind PTTMs was limited to homology-based discovery approaches where E. coli PTTM genes were used to identify orthologs in J1074 (6). We decided to revisit this issue because we now have better knowledge of the PTTMs present in J1074, as well as a validated data set on PTTMs in B. subtilis 168 (11), a representative of the order phylogenetically more proximal to Streptomyces than enterobacteria are. Using reciprocal best BLASTP hit strategy (22), we were able to identify S. albidoflavus candidate PTTM proteins that are orthologous to respective B. subtilis PTTM proteins and are enough to enable the production of almost all firmly identified PTTMs in Streptomyces. Protein Xnr_5177 was an exception; it was identified as an ortholog of Trm61 methyltransferase from Mycobacterium tuberculosis H37Rv (23). Our findings are summarized in Table 2.

TABLE 2.

Predicted Streptomyces albidoflavus J1074 proteins involved in PTTMs

Xnr_a	Codeb	Function	PTTMs (MH+)c
1074	IVG	MiaA, tRNA A37 dimethylallyltransferase	ms2io6A (398), ms2i6A (382), io6A (352), i6A (336)
1078	IVG	MiaB, tRNA-i(6) A37 methylthiotransferase	ms2io6A (398), ms2i6A (382), ms2A (314)
3758	OEP	TruA, tRNA pseudouridine synthase A	Ψ (245), positions 38-40
1143	OEP	TruB, tRNA pseudouridine synthase B	Ψ (245), position 55
4806	OEP	TruC, tRNA pseudouridine synthase C	Ψ (245), positions 65, 67
5204	OPP	RluA, rRNA/tRNA pseudouridine synthase	Ψ (245), position 32
2881	OEP	TadA, tRNA-specific adenosine deaminase	I (269)
4421	OPP	DusB, tRNA-dihydrouridine synthase	D (247)
1471	OEP	TsaC, threonylcarbamoyl-AMP synthase	t6A (413)
3786	OEP	TsaE, tRNA threonylcarbamoyladenosine biosynthesis protein	t6A (413)
3791	OEP	TsaD, N(6)-L-threonylcarbamoyladenine synthase	t6A (413)
3789	OEP	TsaB, t(6)A37 threonylcarbamoyladenosine modification protein	t6A (413)
1347	OEP	IscS1, putative cysteine desulfurase, associated with tRNA 2/4-thiouridine synthase	s4U (261), s2U (261)
1345	OEP	MnmA, tRNA-specific 2-thiouridylase	s2U (261)
4942	OEP	IscU, Fe-S cluster assembly protein	s4U (261)
2491	OEP	ThiI, tRNA 4-thiouridine synthase	s4U (261)
3428	OEP	TilS, tRNA(Ile)-lysidine synthetase	k2C (372)
0992	OEP	RlmCD/TrmA, rRNA/tRNA uridine-C(5)-methyltransferase	m5U (259)
1195	OEP	RlmN, 23S rRNA (A(2503)-C(2))- and tRNA (A(37)-C(2))-methyltransferase	m2A (282)
1214	OEP	TrmD, tRNA (G37-N1)-methyltransferase	m1G (298)
2813	OEP	TrmB, tRNA (G(46)-N(7))-methyltransferase	m7G (298)
5177	OEP	Trm61, tRNA (A(58)-N(1))-methyltransferase	m1A (282) position 58
4495	SBE	TrmN6, tRNA(A(37)-N6)-methyltransferase	m6A (282)
4256	SBE	TrmH, tRNA (G(18)-2′-O)-methyltransferase	Gm (298)
5296	SBE	TrmL, tRNA (C/U-2′-O)-ribose methyltransferase	U/Cm (259/258), xUmd, (275, 362), Cm (272)

^aPTTM protein of S. albidoflavus J1074, GenBank accession CP004370.1.

^bEvidence code: IVG, in vivo evidence via gene knockout; OEP – orthology to experimentally verified PTTM protein of B. subtilis 168, as annotated in (11); OPP – orthology to predicted PTTM protein of E. coli K-12 str MG1655 and/or B. subtilis 168; SBE – selected by elimination procedure, see main text and Fig. S33, ESM. Exception: protein Xnr_5177 was identified as Trm61 ortholog of experimentally verified tRNA (A(58)-N(1))-methyltransferase Rv2118c of Mycobacterium tuberculosis H37Rv (23).

^cAbbreviated notation of the PTTM is used along with the m/z of its cation, in brackets.

^dx denotes complex modifications, such as nm⁵, cmnm⁵, s², and combinations thereof.

Assignment of several S. albidoflavus-encoded methyltransferases to certain tRNA methylation sites remains largely speculative. It is not supported by orthology to known methyltransferases in E. coli and B. subtilis and is complicated by the fact that the mentioned above model organisms harness unrelated enzymes to introduce some of these modifications (24). We, therefore, resorted to elimination procedure to identify the most plausible S. albidoflavus counterparts of methyltransferases TrmH, TrmL, and TrmN6. For this purpose, all S. albidoflavus J1074 proteins annotated as methyltransferase/methylase (89 in total) were inspected on an individual basis to cherry-pick those (i) involved in neither known secondary nor primary metabolic pathways, and (ii) having particular domains, such as SPOUT (TrmH, TrmL) and MTS (TrmN6). In this way, several J1074 proteins were revealed that possess SPOUT domain and show complex orthology-paralogy relationships to verified Trm proteins (Fig. S33, ESM). By elimination of orthologs of rRNA methylases and considering domain structures of the remaining methyltransferases, we picked the most plausible candidates for TrmL and TrmH counterparts, as listed in Table 4; a similar line of reasoning was used to identify TrmN6. Our proposal that Xnr_5296 corresponds to the U34/C34 O-methyltransferase (TrmL) is not consistent with a phylogenetic analysis of SPOUT family methylases of S. albidoflavus. Here, Xnr_3051 formed monophyletic groups with TrmL and CspR, irrespective of tree reconstruction approach (Fig. S34, ESM). However, these trees showed conflicting results for the other proteins, casting doubts on the reliability of phylogenetic inference for this case. Moreover, Xnr_3051 has no orthologs in many streptomycete genomes, and it appears to carry extra domain in its structure, not present in known TrmL proteins (see Fig. S33). These data collectively weighed in favor of Xnr_5296 as the most likely TrmL counterpart in S. albidoflavus.

As described in the preceding chapter, the identity of the streptomycete counterpart of MiaE hydroxylase remains enigmatic. The S. albidoflavus genome also apparently lacks homologs of MnmE and MnmG enzymes involved in complex modifications nm⁵(s²)U and cmnm5(s²)U. In contrast, S. albidoflavus encodes an ortholog of lysidine synthetase TilS, although the corresponding modification was not detected in tRNA hydrolyzates, likely due to its low abundance.

We searched the genomes of 5 model species encompassing different clades of Streptomyces phylogenetic tree for the orthologs of identified J1074 PTTM enzymes. Our results, summarized in Fig. S35 (ESM), showed that most of J1074 proteins listed in Table 2 have counterparts in these genomes. Proteins Xnr_4806 (Ψ synthase C) and Xnr_2491 (s⁴U synthase) were 2 prominent exceptions: their orthologs or even homologs could not be identified in the majority of the analyzed genomes.

Ectopic expression of XNR_1078 (miaB) in S. albidoflavus ΔmiaA mutant prompts accumulation of ms²A.

The introduction of a second copy of the miaB gene into a miaA-null background was recently reported by us to have phenotypic effects on S. albidoflavus (25). One possible explanation of this observation is that methylthiotransferase MiaB may act on unmodified A37 residues. Indeed, we detected trace amounts of ms²A in the tRNA hydrolyzates of the wild type (Table 3), and ms²A had also been reported in E. coli (16). If the effects of miaB overexpression were caused by its action on tRNA, then one should observe an increased amount of certain PTTMs. We performed qualitative analysis of tRNA hydrolyzates of the wild type, miaA mutant and miaA mutant carrying a second copy of the miaB gene under the control of strong constitutive promoter (ΔmiaA+miaB). While the ms²A mass peak was barely detectable (below 10³ au) in SAM2 tRNA hydrolyzates and absent in ΔmiaA, it was abundantly present in ΔmiaA+miaB strain (Fig. 2A and C). A few conserved tRNA modifications were quantified in the above mentioned strains, both in absolute and relative modes, and shown to be produced by the analyzed strains at a similar level (Fig. 2B and C). In contrast, the relative abundance of ms²A in the miaB-overexpressing ΔmiaA mutant was shown to increase around 40-fold compared to either SAM2 or ΔmiaA strains (Fig. 2D).

FIG 2.

Extra copy of miaB gene in S. albidoflavus markedly increased the accumulation of ms²A. (A) Composite extracted ion chromatogram (CEIC) shows that hydrolyzates of total tRNA extracts from parental strain SAM2 (grown in TSB for 48 h) contained all expected mia gene-controlled modifications, ms²A level is marginally low; ΔmiaA produced no Mia-related modifications; ΔmiaA carrying second copy of miaB (XNR_1078), labeled as ΔmiaA+miaB produced significant quantities of ms²A. Identity of ms²A eluting at 25.5 min is shown as an inset to the CEIC (MS and NLS data). Quantification of several tRNA modifications produced by the mentioned above strains using SILIS approach (B) or relative to adenosine abundance (C and D). y axis for absolute (SILIS-based) quantification data: % of amount of modified nucleoside with regard to the amount of adenosine (A); y axis for relative quantification: ratio of abundance (ion counts) of the modified nucleoside to the abundance of adenosine. See Methods for details. Mean values of ms²A abundance are noted on the diagram. Error bars: ± 2SD of the mean of 2 biological replicates. One-way ANOVA was used to reveal significant differences: *** P < 0,001. Values in Parts (B and C) were not found to be significantly different at P < 0,05.

Domain structure of MiaB from J1074 (Xnr_1078) does not differ from that of model MiaB enzymes (Fig. S36A). Also, we observed almost perfect superposition of X-ray structure of MiaB from Bacteroides uniformis (BuMiaB, PDB accession: 7MJZ) and the AlphaFold-predicted structure of Xnr_1078 (Fig. S36B). Nevertheless, while BuMiaB harbors tetralysine stretch starting at position 84, Thr272, and Lys409-Arg410 thought to be involved in recognition of prenylated A37 (26), homologous positions of Xnr_1078 are occupied by amino acids that differ in charge and/or geometry of side chains (Fig. S36C). Also, we note that Xnr_1078 is longer than BuMiaB (513 versus 457 aa) which may also impact substrate specificity of the former.

DISCUSSION

In this study, we provide the first in-depth analysis of the landscape of PTTMs that are present in S. albidoflavus J1074 and summarize the current state of our understanding of the genetic control behind the identified PTTMs. J1074 produces a set of PTTMs that is wider than previously assumed for Streptomyces. For example, thiouridines are known to be omnipresent in all studied to date taxa, but seemed to be absent in S. griseus, streptomycin producer whose PTTMs were studied in 2015 (10). For S. albidoflavus we observed reliable MS signals corresponding to s⁴U and presumed s²U derivatives (Table 1). Moreover, genes for both s⁴U and s²U production are clearly present in S. albidoflavus. The ubiquity of orthologs of J1074 PTTM proteins across several distinct clades of the Streptomyces tree (with several exceptions, see above) suggests that our findings are largely applicable, species-wide. We noted that the abundance of the aforementioned PTTMs was very low when the bacterial cultures for tRNA isolation were grown in defined medium NL5. Similarly, low sulfur concentrations in media resulted in decreased levels of s²U in B. subtilis (27). Influence of growth conditions on the other PTTMs was described for E. coli (28–30). Different culture conditions, therefore, might be one of the reasons for the discrepancies between this and previous works.

It remains a formidable challenge to correctly identify genes for many of the described modifications. Essential metabolism of Streptomyces offers many examples of deviations from the studied precedents, making them “a most unusual group of bacteria” (31). In retrospect, this may not be too unexpected, given that streptomycetes belong to the phylum Actinobacteria, which forms one of the most deeply branching clades on the phylogenetic tree (32). Distribution of MiaE proteins across the phylogenetic tree is very reticulate, and hydroxylation of i⁶A and ms²i⁶A was speculated to be controlled by remotely related enzymes (15). Thus, it is likely that mechanisms of certain PTTMs in Streptomyces are different from what we know for enterobacteria or Bacillus. Indeed, even the best studied organisms in this regard, E. coli, still remains a source of discovery of alternative pathways for seemingly well understood PTTMs (33). Therefore, care should be taken so as not to over-interpret the results of computational predictions of gene function in Streptomyces. We believe that a wealth of new information about actinobacterial PTTMs could be gained through experimental interrogation of genes for SPOUT methylases within the SAM2 genome, and its further mining in order to find candidates for MiaE and MnmEG enzymes.

Methylthiotransferase MiaB from bacterium Thermotoga maritima has been shown to accept (dimethylallyl)adenosine (within a synthetic RNA oligomer that corresponds to anticodon stem-loop of tRNA) as a substrate for thiomethylation (34). The requirement for dimethylallyl unit as a A37 recognition element was also recently observed in structural studies of MiaB from B. uniformis (26). Nevertheless, the identification of unusual nucleoside, such as ms²A and msms²i⁶A in tRNA hydrolyzates from E. coli (16, 35), raises the possibility that MiaB is capable of catalyzing the formation of additional products, including those lacking an isoprenoid moiety. Since the overexpression of the miaB gene (XNR_1078) in an S. albidoflavus ΔmiaA strain had clear phenotypic consequences (25), we investigated whether it was also associated with the formation of unconventional modified nucleosides. We show that ms²A can be reproducibly detected in the aforementioned strain. Hence, phenotypic changes observed in the ΔmiaA pTES1078 strain might be caused by accumulation of ms²A containing tRNAs. It remains to be studied whether the ability of MiaB to catalyze methylthiolation of adenosine in the absence of an isoprenoid structure is an intrinsic property of enzymes only from certain species, or if it is a result of specific experimental conditions (e.g., miaB overexpression). Experiments aiming to elucidate these questions are under way in our laboratories.

MATERIALS AND METHODS

Plasmids, microorganisms, and culture conditions.

All plasmids and strains used in this study are described in Table S2 (ESM). S. albidoflavus (albus) SAM2, a derivative of J1074 carrying single attB^φC31 site (36) was a parent strain for all mutants used throughout this study. E. coli strains were grown under standard conditions (37). For intergeneric matings, S. albidoflavus strains were grown on soy-flour mannitol (SFM) agar (38). Manipulations of E. coli DH5α were performed at 37°C, other bacteria were grown at 30°C, unless otherwise stated. For total tRNA extraction, S. albidoflavus strains were grown in the following liquid media: tryptic soy broth (TSB; Merck Millipore, cat no 1.05459), NL5 (g/L: NaCl 1, KH₂PO₄ 1, MgSO₄ × 7H₂O 0.5, trace elements solution [as described in 38]) 2 mL, glycerol 25, l-glutamine 5.84, pH 7.0). Liquid media for tRNA purification were 0.22 μm filter-sterilized, other media were autoclaved. To reveal endogenous antibacterial and antifungal activities S. albidoflavus strains were grown on R5 (38) and SG2 agar (g/L: glucose 20, yeast extract 5, soytone 10, pH 7.2 prior to autoclaving). For phenotypic examinations, S. albidoflavus were grown on SFM and SMMS agars (38). Where needed, strains were maintained in the presence of apramycin (25 μg/mL) or hygromycin (100 μg/mL); chromogenic substrate X-Gal and inducer IPTG were added to media to a final concentration of 50 and 20 μg/mL, respectively.

Plasmid constructions.

Suicide vector pKC1132 (39) was used to prepare a knockout plasmid for XNR_1636, a putative miaE homologue. The aforementioned gene was amplified with approximately 2-kb flanking arms from SAM2 chromosome with primers xnr1636_dup and xnr1636_drp (4.8-kb product). Sequences of all primers used in this study are given in Table S3, ESM. The amplicon was digested with XbaI and BamHI restriction endonucleases, and cloned into XbaI-BamHI sites of pKC1132 yielding plasmids pJS3. Hygromycin resistance cassette hyg, flanked with P-GG and B-CC sites, was amplified from patt-shyg (40) with primers xnr1636_da_up and xnr1636_da_rp. Resulting 1.4-kp PCR product was used to replace XNR_1636 gene in pJS3 using recombineering strain E. coli BW25133 pKD46⁺ (41). As a result, plasmid pXNR1636hyg was constructed. These and other constructs generated during the course of the work were verified by PCR and sequencing. To construct plasmid pTES1636, gene XNR_1636 was amplified with xnr1636_XbaIup and xnr1636EcoRIrp (752-bp product) from SAM2 chromosome, digested with XbaI and EcoRI restriction endonucleases, and cloned downstream of ermEp into pTES (42).

Generation and verification of the S. albidoflavus recombinant strains.

All constructs were transferred into S. albidoflavus conjugally from E. coli ET12567 (pUZ8002), as described elsewhere (8). The XNR_1636 knockout strain of S. albidoflavus was selected for hygromycin resistance (gene replacement with hyg) and apramycin sensitivity (loss of vector sequences). PCR was employed to confirm the presence of the plasmids and expected gene replacements in S. albidoflavus, as detailed in ESM.

Purification of total tRNA from S. albidoflavus.

S. albidoflavus strains were plated on SFM agar and spores were harvested into saline after 168 h of growth. Then, 250-mL round flasks, containing 50 glass beads (Ø 5 mm, Sigma cat. No 1.04017) and 30 mL of TSB, were inoculated with thick spore suspension [(3.0 ± 1.0)×10⁸ CFU in 100 μL] of respective strains. The flasks were shaken (185 rpm) for 24 h, then 10 mL of the resulting cell suspension were transferred into 1-L round flasks, containing 120 glass beads and 100 mL of either TSB or NL5. Biomass was grown for a specified period (24 and 46 h), spun down at 4°C and washed twice with ice-cold Buffer I (50 mM Tris-HCl, 0.9% NaCl pH 7.5). The cell pellet (approximately 1 to 2 g, wet weight) was stored at −80°C unless used immediately for tRNA isolation.

Guanidinium thiocyanate method was used to purify total tRNA, as described below. A 1.8 g aliquot of S. albidoflavus biomass was resuspended in 20 mL of TRI reagent (Sigma-Aldrich) and incubated at room temperature (22°C) for 5 min. Then, 4 mL of chloroform were added, tubes were vortexed for 15 sec and left to stand at room temperature for 15 min. The tubes were centrifuged at 12 000 g for 15 min at 4°C and the colorless upper phase was transferred into new tube. Treatment with chloroform (4 mL) was repeated, and the upper phase was transferred into a new tube. A totoal of 10 mL of 2-propanol was added to the upper phase and mixed. The mixture was left to stand for 10 min at room temperature and then centrifuged (15 000 g) for 30 min at −4°C. The supernatant was carefully discarded (so as not to dislodge the pellet on the bottom of the tube) and washed with 70% ethanol (freshly prepared from ice-cold water and absolute ethanol stored at −80°C). The pellet in 70% ethanol was centrifuged (15 000 g) for 10 min at −4°C, and then the liquid phase was carefully discarded. The pellet was air-dried for 5 min at room temperature and resuspended in 150 μL of deionized water. The quantity of tRNA in the sample and its purity was assessed spectrophotometrically (NanoDrop) and electrophoretically (Agilent TAPE station). The tRNA yields were within 900 ÷ 3600 ng/μL range, depending on the culture conditions and the strain being used.

Digestion of tRNA, liquid chromatography-tandem mass spectrometry (LC-MS), and nucleoside quantification.

Up to 5000 ng of total tRNA per sample was digested to nucleoside level using 0.6 U nuclease P1 from Penicillium citrinum (Sigma-Aldrich), 0.2 U snake venom phosphodiesterase from Crotalus adamanteus (Worthington), 0.2 U bovine intestine phosphatase (Sigma-Aldrich), 10 U benzonase (Sigma-Aldrich), 200 ng Pentostatin (Sigma-Aldrich), and 500 ng Tetrahydrouridine (Merck Millipore) in 5 mM Tris (pH 8) and 1 mM magnesium chloride for 2 h at 37°C.

A total of 2000 ng of digested total tRNA was analyzed qualitatively via LC-MS (Agilent 1260 series and Agilent 6470 Triple Quadrupole mass spectrometer equipped with an electrospray ion source [ESI]) using neutral loss scan mode (16) in positive ion mode. For absolute quantitative analysis of selected nucleosides (Ψ, m⁷G, Cm), ¹³C stable isotope-labeled internal standard (SILIS; prepared from E. coli) was applied. An amount of 100 ng of digested total tRNA and 50 ng SILIS were analyzed via LC-MS using Agilent MassHunter software in the MRM (multiple reaction monitoring) mode. The solvents consisted of 5 mM ammonium acetate buffer (pH 5.3; solvent A) and LC-MS grade acetonitrile (solvent B; Honeywell). The elution started with 100% solvent A with a flow rate of 0.35 mL/min, followed by a linear gradient to 10% solvent B at 20 min, 25% solvent B at 30 min and 80% solvent B after 40 min. Initial conditions were regenerated with 100% solvent A for 14 min. The column used was a Synergi Fusion (4 μM particle size, 80 Å pore size, 250 × 2.0 mm; Phenomenex). The UV signal at 254 nm was recorded via a diode array detector (DAD) to monitor the main nucleosides. ESI parameters were as follows: gas temperature 300°C, gas flow 7 L/min, nebulizer pressure 60 lb/in², sheath gas temperature 400°C, sheath gas flow 12 L/min, and capillary voltage 3000 V. The amount (ion counts) of target modification was first normalized with regard to SILIS and expressed in femtomol. Amount (UV signal) of the reference nucleoside (adenosine; A) after subtraction of UV-SILIS, and taking into account external UV calibration, was expressed in picomol. Absolute quantity of the target modification was finally expressed as the percentage of target modification per unit of reference nucleoside (%mod/A; (([mod]/1000)/[A])×100). Detailed protocol for preparation of SILIS and MRM measurement mode can be found in (16, 43). Relative quantification was performed for nucleosides ms²io⁶A and ms²A (for which SILIS was not available), and also for Ψ, m⁷G, and Cm to make sure that the results are comparable with the absolute amounts. Here, the amount of target modification was expressed as the ratio of abundance (ion counts) of the former to abundance of adenosine (MS mod/MS A). Results of absolute and relative quantification for Ψ, m⁷G, and Cm did not differ.

Phenotypic analysis of S. albidoflavus.

Photos of the lawns of the mutant and parental strains grown on various agar media were taken after 72 to 125 h of incubation. For scanning electron microscopy, small pieces of 76 h old lawns were cut off the OM or SFM agar plate samples, vacuum-dried, and directly analyzed on a Jeol JSM-T220A scanning microscope. Native antibiotic activity of S. albidoflavus strains was monitored using agar plug antibiotic diffusion assay. Briefly, strains were grown on SG2 and R5 agar for 5 days. Then, agar plugs (Ø 5 mm) were cut off the lawn and stacked on top of TSB agar plates with test culture Debaryomyces hansenii spread immediately prior to the experiments or B. cereus spores, as it is described above. Halos of growth inhibition around the plugs were observed after 18 h of incubation. LC-MS analysis of antibiotic extracts was carried out as described in Koshla et al. (25).

Bioinformatic methods.

Orthologs of known PTTM enzymes were searched in J1074 proteome (taxid: 1886) as the reciprocal best BLASTP hits, as described in Kuzniar et al. (22). The genome of J1074 (GenBank accession number CP004370.1) was used to retrieve subsets of proteins annotated as either putative diiron hydroxylases/monooxygenases or methyltransferases/methylases and processed to identify MiaE and SPOUT family methyltransferases, respectively, via conserved domain analysis using search tools of Pfam (http://pfam.xfam.org/) and HHPred (https://toolkit.tuebingen.mpg.de/tools/hhpred). PDB and AlphaFold databases were used to retrieve structures of known or predicted MiaE structures. Phyre2 (44) was used to model the structure of Xnr_1636 using normal modeling mode (http://www.sbg.bio.ic.ac.uk/~phyre2/html/page.cgi?id=index); see ESM. Phylogenetic analysis of streptomycete SPOUT family methyltransferases was carried out on webserver phylogeny.fr (45) using maximum likelihood, Bayesian, and neighbor-joining algorithms (http://www.phylogeny.fr/programs.cgi). The optimal evolution model parameters were estimated with the help of IQTree webserver (http://iqtree.cibiv.univie.ac.at/) as detailed in ESM. Structure inference pipeline AlphaFold2 version 2.2.3 + 49 (46) used to model Xnr_1078 was accessed via https://console.latch.bio/workflows.