For the search, 10 selected Streptomyces well annotated genomes were used (GenBank assembly IDs shown in parenthesis following organism name): S. coelicolor A3(2) (GCA_000203835.1_ASM20383v1), S. ambofaciens ATCC 23877 (GCA_001267885.1_ASM126788v1), S. nodosus (GCA_000819545.1_ASM81954 v1), S. reticuli (GCA_001511815.1_TUE45), S. avermitilis MA 4680 (GCA_000009765.2_ASM976v2), S. griseus NBRC (GCA_000010605.1_ASM1060v1), S. scabies 87 22 (GCA_000091305.1_ ASM9130v1), S. cattleya NRRL8057 (GCA_000237305.1_ASM 23730v1), S. lincolnensis (GCA_001685355.1_ASM168535v1, S. pristinaespiralis ATCC 25486 (GCA_000154945.1_ASM 15494v1).

The search was performed in the following steps:

To increase reliability of the search, the candidates had to fulfill the following additional criteria derived from genomic properties of known 6S and Ms1 RNAs that were: (1) existence of similar sequences in evolutionarily close species identified by BLAST with BLAST E-values < 10^–20, and (2) conserved genomic position in the ten searched Streptomyces genomes.

The full-length sequences were predicted using the approximate length of the transcripts detected by Northern blotting and the genomic positions of oligonucleotide sequences used for sRNA verification by Northern blotting (Figure 1). The sequences of approximate length were constructed using the oligonucleotides positioned at the 5′ end, center, and 3′ end. The three constructed sequences were BLASTed against genomes of the ten Streptomyces species listed in the previous paragraph. As sRNAs in general are conserved in evolutionarily related species, here in Streptomyces species, the constructed sequences had to be identified by BLAST in multiple Streptomyces species to ensure that they were constructed correctly. Using the superposition of constructed sequences according to their conservation identified by BLAST, we reconstructed probable sRNA sequences in Streptomyces coelicolor. The probable sequences are shown in Supplementary Table 1 for expressed Streptomyces sRNAs.

The search was based on text phrases in genomic annotations of the genes flanking potential Ms1 RNA genes and sequence similarity search based on BLAST for verification. The search was possible as sRNAs including Ms1 RNA are known to be conserved in evolutionarily close bacteria (Hnilicova et al., 2014). The search has several steps:

Annotations of flanking genes were first split into single words and non-specific words were removed. The non-specific words were (allowing for grammar errors): ‘hypothetical’, ‘protein’, ‘type’, ‘family’, ‘domain’, ‘putative’, ‘precursor’, ‘component’, ‘subunit’, ‘subfamily’, ‘conserved’, ‘chain’, ‘or’, ‘superfamily’, ‘unknown’, ‘function’, ‘of’, ‘like’, ‘containing’, ‘II’, ‘and’, ‘short’, ‘dependent’, ‘probable’, ‘associated’, ‘to’, ‘that’, ‘the’, ‘predicted’, ‘uncharacterized’, ‘production’, ‘proteases’, ‘fold’, ‘in’, ‘by’, ‘universal’, ‘pathway’, ‘involved’, ‘related’, ‘general’, ‘group’, ‘sequence’, ‘class’, ‘cluster’, ‘accepting’, and ‘determining’.

We also removed the so-called questionable non-specific words whose (non)specificity depended on the analyzed context. They were: ‘box’, ‘enzyme’, ‘factor’, ‘secreted’, ‘release’, ‘neighborhood’, ‘solute’, ‘accessory’, ‘peptide’, ‘biosynthesis’, and ‘substrate’.

Specific words were left. They characterized specific aspects of gene function. An example is the ‘Fic/Doc family protein’ annotation contained two non-specific words, ‘family’ and ‘protein’, and one specific word ‘Fic/Doc’.

The first synteny phrases were obtained from annotations of homologs of M. smegmatis Ms1 RNA identified by sequence similarity by BLAST in 498 species of 8 genera—Mycobacterium, Mycolicibacterium, Rhodococcus, Nocardia, Gordonia, Mycobacteroides, Hoyosella, and Tsukamurella (Supplementary Figure 1). Specific words were selected as those occurring in annotations of flanking genes above the average (Supplementary Figure 1). The specific words were ‘oxidoreductase’, ‘hydrolase’, ‘HAD’, ‘IB’, ‘morphological’, ‘differentiation’, ‘inhibition’, ‘phosphoserine’ and ‘phosphatase’. The occurrence in genera was used rather than the occurrence in species as there were genera with many species with repeating annotations with repeating specific words causing bias of an occurrence of certain words.

Phrases were generated from the most frequent specific words according to their co-occurrence in the annotations from which they were extracted and which defined their semantic binding. For example, specific words ‘hydrolase’, ‘HAD’, ‘IB’, ‘morphological’, ‘differentiation’, ‘inhibition’, ‘phosphoserine’ and ‘phosphatase’ (identified in Supplementary Figure 1 as most occurring) formed the following phrases: ‘IB HAD’, ‘HAD hydrolase’, ‘inhibition morphological differentiation’ and ‘phosphoserine phosphatase’. A complete list of first phrases is in Table 1.

Synteny phrases were repeatedly updated after identification of Ms1-containing IGRs in new species (Supplementary Figures 2–4). New annotations either contained synonyms to old phrases, e.g., oxidoreductase, whose synonym was Fic that also has oxidoreductase activity or contained new annotation phrases. The synonyms were identified using information found on the internet, most often in Wikipedia and/or many various public protein databases/knowledgebases. New annotations were considered as Ms1-syntenous when they repeated for one flanking gene of newly identified Ms1 IGRs in many species, while annotation of the other flanking gene must have contained the known Ms1 phrase. New annotations were analyzed to get new phrases out of them that were merged with old phrases as can be seen in Tables 2, 3.

Mycobacterium smegmatis mc² 155 cells ATCC 700084 (wt, LK865) were grown at 37°C in Middlebrook 7H9 medium with 0.2% glycerol and 0.05% Tween 80, and harvested in exponential (OD₆₀₀ ∼0.5) or early stationary phase (OD₆₀₀ ∼2.5–3, 24 h of cultivation). S. coelicolor A3(2) spore stock expressing HA-tagged HrdB was thawed and inoculated to 2YT medium. Germination was carried out at 30°C for 5 h as described previously (Moody et al., 2013; Šmídová et al., 2019). The germinated spores were harvested by centrifugation (3200 × g, 25°C, 5 min), inoculated into Na-glutamate medium supplemented with trace element solution and TMS1 as described previously (Šmídová et al., 2019), cultivated at 30°C and harvested by centrifugation (3200 × g, 4°C, 5 min) at different time points after germination. Rhodococcus erythropolis CCM259 (Strnad et al., 2014) (LK1556) was cultivated at 26°C in 2x TY medium to exponential phase (OD₆₀₀ ∼2) or stationary phase (OD₆₀₀ ∼11, 24 h of cultivation), Corynebacterium glutamicum ATCC 13032 (LK1557) was grown in 2xTY at 30°C and harvested in exponential (OD₆₀₀ ∼1) or stationary phase (OD₆₀₀ ∼8, 24 h of cultivation). Bacillus subtilis 168 strain was grown in LB medium at 37°C to exponential phase (OD₆₀₀ ∼0.3) or stationary phase (OD₆₀₀ ∼4).

Each frozen cell pellet was resuspended in 240 μl TE (pH 8.0) plus 60 μl LETS buffer (50 mM Tris–HCl pH 8.0, 500 mM LiCl, 50 mM EDTA pH 8.0, 5% SDS) and 600 μl acidic phenol (pH∼3):chloroform (1:1). Lysates were sonicated in a fume hood, centrifuged, the aqueous phase extracted three more times with acidic phenol (pH∼3): chloroform and precipitated with ethanol. RNA was dissolved in water, treated with DNase (TURBO DNA-free Kit, Ambion) and visualized on a 7 M urea 7% polyacrylamide gel by staining with GelRed (Labmark).

RNAs were resolved on a 7% polyacrylamide gel and transferred onto an Amersham Hybond-N membrane according to the protocol described in Pánek et al. (2011). 5′ biotinylated oligonucleotide probes were hybridized to the membrane and detected with BrightStar BioDetect Kit (Ambion) according to manufacturer’s instructions. For Northern blot probes sequences, see the Supplementary Material.

The protocol was adopted from Martini et al. (2019). An adapter oligo CTGGAGCACGAGGACACTGACATGGACTG AAGGAGTagaaa (lower case letters are ribonucleotides, upper case letter deoxyribonucleotides) was ligated to RNA 5′ ends. 5 μg DNase-treated RNA was treated with RppH (NEB). Treated and untreated RNA samples (8 μl) were mixed with 1 μl of 1 μg/μl adapter oligo and incubated at 65°C for 10 min, then ligation reaction was set up including 10 μl 50% PEG8000, 3 μl 10X ligase buffer, 3 μl 10 mM ATP, 3 μl DMSO, 1 μl Murine RNase inhibitor (NEB), and 1 μl T4 ligase (NEB). Ligation was performed at 20°C overnight and RNA cleaned using RNA Clean and Concentrator 25 kit (Zymo). RNA was reverse-transcribed into cDNA (SuperScript III, Invitrogen) with random hexamers. PCR was done using a forward primer CTGGAGCACGAGGACACTGA and reverse (gene specific) primers (Str11, 5′-AGCCGCTCCCCTGGTCTGGG-3′, Ms1 candidate 5′-GGTGTCCATGCTCGGTCC-3′). The nucleotide position of the TSS was taken from EMBL/GenBank Accession No. AL645882.

The protocol was adopted from Sedlyarova et al. (2017). 550 pmol of 5′-phosphorylated RNA adaptor (5′P -AAUGGACUCGUAUCACACCCGACAA-3′) was ligated to 6 μg of total DNase treated RNA using T4 RNA Ligase 1 (ssRNA Ligase, New England Biolabs) according to the manufacturer’s protocol overnight at 16°C. RNA was purified with RNA Clean and Concentrator 25 kit (Zymo) and reverse-transcribed into cDNA (SuperScript III, Invitrogen) with 3′ RACE specific primer (5′-TTGTCGGGTGTGATACGAGTCCATT-3′). The same primer was used as reverse primer for PCR together with gene specific forward primer (5′- GATCACCTTAAACACGCATATGG-3′).

Streptomyces coelicolor cells expressing HA-tagged HrdB were pelleted and resuspended in lysis buffer (20 mM Tris–HCl pH 7.9, 150 mM KCl, 1 mM MgCl₂, 1 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride (PMSF), supplemented with Calbiochem Protease Inhibitor Cocktail Set III protease inhibitors), sonicated 15 × 10 s with 1 min pauses on ice and centrifuged. 500 μg (protein) of lysates were incubated for 16–18 h at 4°C with 20 μl of Protein G plus agarose beads (Santa Cruz) coated with 5 μg mouse monoclonal anti-β subunit of RNAP antibody [clone 8RB13] (BioLegend), 2.5 μg of anti-σ⁷⁰ antibody [clone 2G10] (BioLegend), 1.25 μg of anti-HA antibody [clone HA-7] (Sigma-Aldrich) or 5 μg of mouse non-specific IgG (Sigma-Aldrich) used as a negative control, respectively. The captured complexes were washed 4 times using 20 mM Tris–HCl pH 7.9, 150 mM KCl, 1 mM MgCl₂, finally resuspended in 300 μl 20 mM Tris–HCl pH 7.9, 150 mM KCl, 1 mM MgCl₂ and divided into two parts. One third of the beads were incubated in SDS sample buffer for 5 min at 95°C and eluted proteins were detected by Coomassie staining and Western blotting. The remaining two thirds of the beads were resuspended in 200 μl 1% SDS, 150 mM KCl, 20 mM Tris–HCl pH 7.9, 1 mM MgCl₂ and incubated on a rotating platform with 200 μl acidic phenol (pH∼3):chloroform (1:1) for 15 min. Eluted RNA was precipitated with ethanol, dissolved in water and DNase treated (TURBO DNA-free Kit, Ambion). RNA was reverse transcribed into cDNA (SuperScript III, Invitrogen) using random hexamers and amplified by quantitative reverse transcription PCR (RT-qPCR) in a LightCycler 480 System (Roche Applied Science) in duplicate reactions containing LightCycler 480 SYBR Green I Master and 0.5 μM primers (each). Primers were designed with Primer3 software and their sequences are in the Supplementary primer list. Negative controls (no template reactions and reactions with RNA as a template to control for contamination with genomic DNA) were run in each experiment, the quality of the PCR products was determined by dissociation curve analysis, and the efficiency of the primers determined by standard curves. The relative amounts of co-immunoprecipitated RNAs were quantified on the basis of the threshold cycle (Ct) for each PCR product that was normalized to input values according to the formula 2^∧[Ct(immunoprec)–Ct(input)].

Proteins were detected by Western blotting using a rat monoclonal antibody recognizing the HA tag conjugated with HRP [clone BMG-3F10] or a mouse monoclonal antibody recognizing β subunit of RNA polymerase [clone 8RB13] in a combination with secondary antibody conjugated with HRP.

First, we conducted a computational homology search for putative 6S RNA/Ms1 RNA in the Streptomyces genus.

Using this computational search, we identified 12 candidate genes for 6S/Ms1 RNAs in Streptomyces (see Supplementary Material). Interestingly, flanking genes of one of them (Str11, Supplementary Table 1) displayed annotations syntenous to 6S-1 RNA in Firmicutes (Wehner et al., 2014), making it a prime candidate for Streptomyces 6S RNA. Nevertheless, none of the putative identified sRNA genes had the synteny of mycobacterial Ms1 RNAs. We suspected that the search might have not identified all candidate genes (for reasons see section “Discussion”). Therefore, prior to the experimental validation, we extended the homology search with linguistic synteny analysis approach.

The linguistic approach was based on search for text phrases that were specific to Ms1 RNA synteny and could be found in Ms1 flanking genes annotations. Note that meaning of the term ‘synteny’ in this work is ‘conserved genomic context’. Synteny annotations are annotations of conserved flanking genes of Ms1 RNAs. The phrases were used to identify new IGRs containing putative Ms1 RNAs in other species.

Synteny was rarely employed previously for identification of sRNAs (Sridhar and Gunasekaran, 2013), mostly because the sRNAs and their synteny occurred only in certain phyla, e.g., in Enterobacteriaceae (Sridhar and Rafi, 2007) and were not widespread in bacteria. Nevertheless, 6S RNA is an example of a sRNA identified throughout the bacterial kingdom with synteny conserved in specific taxons—for example in Enterobacteriaceae (γ-Proteobacteria) (Wehner et al., 2014). Although no 6S RNA syntenic pattern is valid for all bacteria, some proteins frequently occur in the syntenic regions of the 6S RNA throughout the bacterial kingdom. For example, ygfA, which encodes 5-formyltetrahydrofolate cyclo-ligase, is found adjacent to 6S RNA gene in α-, γ- Proteobacteria, and some species from β-Proteobacteria, δ-Proteobacteria, or Firmicutes (Barrick et al., 2005; Wehner et al., 2014). Therefore, as Ms1 and 6S RNAs are structurally and functionally similar, we assumed that Ms1 flanking genes would be at least partially conserved in Actinobacteria similarly to 6S RNA in γ-Proteobacteria.

Annotations of flanking genes of Ms1 RNAs identified previously in Mycobacterium, Rhodococcus, and Nocardia were conserved (Hnilicova et al., 2014). We speculated that the conservation would also be kept in those Actinobacteria where Ms1 RNA had not been identified. To verify this assumption, we first identified homologs of M. smegmatis Ms1 RNA by sequence similarity using BLAST in 498 species of 8 genera and analyzed annotations of their flanking genes. As expected, we found it conserved (Supplementary Figure 1). Therefore, we extracted synteny text phrases (Table 1) from the annotations and used them to search genomic annotations for their occurrence indicating putative Ms1 RNAs.

Intergenic regions with flanking genes with Ms1 RNA-specific synteny phrases (synteny hits) were found in numerous Streptomyces species for both ‘HAD hydrolase’ and ‘inhibition morphological differentiation’ phrases paired with the ‘oxidoreductase’ phrase. For example, in the first of the Streptomyces species, S. actuosus, a total of six hits of ‘HAD hydrolase’ were obtained. Only one of them fulfilled the other criteria of the synteny search, which was Ms1 synteny phrase (‘oxidoreductase’) in annotation of the other flanking gene and the length of IGR between flanking genes larger than 300 nucleotides. This IGR was considered as an IGR containing a putative Streptomyces Ms1 RNA.

The next criterion was an evolutionary conservation of the candidate Ms1 IGR in related species, i.e., in other Streptomyces species, analogously to Ms1 RNA from M. smegmatis conserved in other Mycobacteria. Therefore, the sequence of the S. actuosus Ms1 IGR was BLASTed against the NCBI nt database, which produced ∼250 BLAST hits in 188 different Streptomyces species (Supplementary Figure 6) with E-values < 1 × 10^–24, i.e., strong sequence similarity and with a similar position in the middle of the linear Streptomyces chromosome (Figure 2A). Both sequence similarity and similar genomic loci indicated evolutionary conservation thus suggesting that the IGRs contained Streptomyces Ms1 RNAs.

Among the BLAST hits, a 419 nucleotides long IGR in S. coelicolor A3(2) (ENA ID AL645882.2) were identified at genomic locus 3934559: 3934978. To find out where within this IGR a putative Ms1 RNA was, the 419 nt sequence was sampled with 200–300 nt subsequences in 5 nt steps for which suboptimal secondary structures were predicted using UNAfold (Markham and Zuker, 2008) and compared to the secondary structure of M. smegmatis Ms1 RNA used as a structural template. This way we aimed at identification of a subsequence of the IGR able to adopt a Ms1 RNA-like secondary structure, thus identifying a local position of Ms1 RNA. The Ms1 RNA-like secondary structures were obtained with 220–235 nt subsequences at positions 126–141 downstream of 5′ end of the IGR sequence. Note that this sequence of the putative S. coelicolor Ms1 RNA had no similarity detectable by cross-species exploration BLAST to the sequence Ms1 RNA from M. smegmatis.

To summarize this part, we identified a total of 13 (12 + 1) potential 6S/Ms1 RNA candidate genes in S. coelicolor.

Next, we used Northern analysis to determine expression of these putative sRNA, probing their expression from both strands. The analysis revealed that several of them were expressed, to various degrees, in exponential and stationary phases in S. coelicolor (Figures 1A–F, 2C). Expression of the remaining six sRNAs was not detected. Genomic loci of the new sRNAs are depicted in Figure 1 and their basic characteristics are described in detail in the next two sections and summarized in Supplementary Table 1 (including predicted sequence, position in the S. coelicolor genome, annotations of flanking genes, and location at the chromosome in the Streptomyces genus).

Str1 sRNA is an antisense RNA (as1791, Figure 1A) to the tetratricopeptide repeat protein gene (sco1791) and it was found in two forms (90 and 120 nt). Additionally, for Str1 we detected a short transcript (100 nt) from the opposite strand, a fragment of the sco1791 (1260 nt) mRNA.

Str3 localizes to the sco1761 gene and similarly to Str1, Str3, is also an antisense RNA (130 nt, AS1711, Figure 1B). Furthermore, we also detected two short transcripts (110 and 80 nt, respectively) from the opposite strand, fragments of the 1086 nt long sco1761 mRNA. The antisense nature of both Str1 and Str3 sRNAs suggests that these sRNAs might be cis-acting antisense RNAs and the detected transcripts are fragments of the respective regulated mRNAs.

Str5 (scr5401, Figure 1C), Str8 (scr3567, Figure 1D), and Str10 (scr5145, Figure 1E) were expressed from intergenic regions and their lengths ranged from 120 to 130 nt.

We concluded that the identified genes encode bona fide sRNAs that are expressed in S. coelicolor but their short length (<150 nt) excluded them as potential 6S/Ms1 RNAs. 6S RNA/Ms1 RNA must adopt specific secondary structures and one of the shortest known 6S RNAs is from Aquifex aeolicus, which is ∼160 nt long (Köhler et al., 2015).

Str11 (scr1506), the candidate with the same synteny as 6S-1 RNA in B. subtilis, was long enough (∼220 nt long, Figure 1F) to be considered as a 6S RNA candidate. We identified the exact Str11 5′ end by 5′ RACE (Figures 1G,H). Str11 also had a 6S-like predicted consensus secondary structure (Supplementary Figure 5).

Str13, the Ms1 candidate identified by the linguistics search, partially overlaps with the previously discovered ssrS gene (Panek et al., 2008), which was proposed to encode a 192 nt long 6S RNA (Mikulík et al., 2014). The Str13 sequence also overlaps with the scr3559 sRNA identified by RNA-seq (Vockenhuber et al., 2011; Moody et al., 2013). In S. coelicolor, both scr3559 and ssrS are located between the sco3558 and sco3559 genes (Vockenhuber et al., 2011; Moody et al., 2013). The flanking genes and positions of Str13 in S. coelicolor and Ms1 in M. smegmatis and MTS2823 in M. tuberculosis are shown in Figure 2B; Arnvig et al. (2011), Hnilicova et al. (2014).

To start deciphering whether Str13 (Ms1 candidate) or ssrS (putative 6S RNA) is expressed from the sco3558 - sco3559 intergenic region, we performed Northern blot analysis. We used a probe that could hybridize to both Str13 and ssrS/6S RNA and we detected a signal that corresponded to a ∼230 nt RNA (Figure 2C). This could represent the previously reported 244 nt long ssrS/6S RNA unprocessed transcript (Mikulík et al., 2014). Although we also detected a shorter transcript (Figure 2C), the major isoform was the ∼230 nt long RNA and not the 192 nt ssrS/6S RNA. The size of the RNA detected by the Northern blot (∼230 nt) corresponded to the previously published lengths of scr3559: 235 bp (Vockenhuber et al., 2011) or 227 bp, respectively (Moody et al., 2013).

Next, we mapped the 5′ and 3′ ends of Str13 by 5′ RACE. The 5′ end was identified at position 3,934,693 (Figure 2D) that is 52 nucleotides downstream from the 5′end of the previously annotated ssrS/6S RNA (Mikulík et al., 2014; Figure 2F) and 134 nt downstream from 5′ end of IGR. This position agrees with the computationally predicted genomic locus of the putative S. coelicolor Ms1 (Str13) RNA - predicted Ms1 starts 126–141 nt downstream from 5′ end of IGR. This 5′ RACE result matches the 5′ end of scr3559 as determined by RNA-seq (Vockenhuber et al., 2011; Moody et al., 2013) and dRNA-seq (Romero et al., 2014; Jeong et al., 2016; Kim et al., 2020) in S. coelicolor. We also searched for additional transcription start sites in previously published data and found position 3,933,713 (Kim et al., 2020) that was 20 nucleotides upstream of the 5′ end of scr3559 but also did not correspond to the 5′ end of ssrS/6S RNA that is 52 nucleotides upstream. The 3′ end of Str13 was then determined by 3′ RACE (Figure 2E) at the position 3,934,920, which corresponds to the 3′ end of scr3559 (Vockenhuber et al., 2011; Moody et al., 2013) but not ssrS/6S RNA. The 3′ end in the same position (3,934,920) was identified also by Term-seq (Lee et al., 2020).

To determine whether scr3559, along with ssrS/6S RNA is perhaps expressed, we used three different probes for Northern blot analysis. Probe “A” should specifically hybridize to ssrS/6S RNA, probe “B” to both sr3559 and ssrS/6S RNA, and probe “C” only to the sr3559 (Figure 2F). We detected the ∼230 nt band only with probes B and C that hybridized to sr3559 (Figure 2G). No signal specific for ssrS/6S RNA was detected at 35 and 65 h after germination, indicating that the putative 6S RNA (ssrS gene) is not expressed in detectable amounts in these growth phases.

We cannot exclude that ssrS/6S RNA is expressed under unknown conditions but the main sRNA transcript derived from the sco3558-sco3559 genomic locus starts at position 3,934,693 and differs from the ssrS/6S RNA sequence that was used to experimentally test the interaction of the putative 6S RNA with the RNAP-HrdB holoenzyme in vitro (Mikulík et al., 2014).

As we identified the full sequence of the Ms1 candidate by 5′ and 3′ RACE, we used suboptimal structure folding to search for Ms1-like secondary structure motifs. We folded the Ms1 candidate sequence by UNAfold (with parameters P = 5000, W = 2 and X = 100) that predicted 78 suboptimal structures. The structures were clustered into 5 clusters based on their mutual structure similarity to find structurally representative folds. The clusters represented structural variations of a typical Ms1 fold. The most representative fold was identified in a cluster with most mutually similar suboptimal structures that contained typical Ms1-like structures which resembled Ms1 from M. smegmatis (Figure 2I), revealing its potential to interact with RNAP. An example of the secondary structure from that cluster is shown in Figure 2H.

We concluded that both Str11 (scr1506) and Str13 (scr3559) satisfied the criteria for potential 6S/Ms1 candidates and we selected them for further analysis.

To answer whether Str11 (scr1505) and/or Str13 (scr3559) are homologs of 6S RNAs or Ms1, we wanted to immunoprecipitate the primary σ factor, HrdB, and RNAP from S. coelicolor. As the commercially available antibody against the primary σ⁷⁰ (clone 2G10) interacted with HrdB only weakly (Figure 3A, lanes 2 and 6), we used a strain with an HA-tagged S. coelicolor hrdB gene (Šmídová et al., 2019) and immunoprecipitated HA-HrdB with an anti-HA antibody from exponentially growing (42 h post-germination) and stationary (66 h post-germination) cells. The anti-HA antibody pulled down HA-HrdB and α, β, β′ subunits of RNAP, especially at 42 h post-germination (Figure 3A, lane 3, protein band identities were verified by mass spectrometry). Thus, the anti-HA antibody interacted both with HA-HrdB alone and also with the RNAP-HrdB complex, which binds 6S RNA in many bacterial species. Then, we immunoprecipitated RNAP with the antibody against the RNAP β subunit (Figure 3A lanes 1 and 5). This antibody preferentially recognizes the RNAP core without the primary σ factor (Figure 3A). We subsequently isolated co-immunoprecipitated RNAs and measured their relative amounts by RT-qPCR (Figures 3B,C).

Str11 (scr1505) associated neither with the RNAP core nor with the RNAP-HrdB holoenzyme, similar to four control RNAs that also did not interact with RNAP: 16S rRNA, sco3552, sco3710 encoding membrane proteins, and sco2013 encoding response regulator PdtaR.

Importantly, ∼2% of Str13 (scr3559) was bound to the RNAP core at 42 h post-germination in S. coelicolor (Figure 3B, the input represents the total amount of scr3559 isolated from the cell lysates), and it increased to ∼5% at 66 h post-germination (Figure 3C). scr3559 bound neither HrdB alone nor the HrdB-RNAP complex. As a control, we performed immunoprecipitation with the same antibody from stationary phase M. smegmatis cells and ∼6% of Ms1 co-immunoprecipitated with the RNAP core (Figure 3D). As we noticed that only a low amount of HrdB-HA was immunoprecipitated at 66 h post-germination (Figure 3A, lane 7) compared to 42 h (Figure 3A, lane 3), we compared the relative levels of HrdB-HA and RNAP in S. coelicolor. The amount of HrdB-HA significantly decreased at 66 h post-germination compared to 42 h (Figure 3E) while the level of the RNAP β subunit was almost unchanged. This suggests that the level of HrdB and subsequently, the level of the RNAP-HrdB complex is low in the late phase of growth in S. coelicolor, similar to previous observations in M. smegmatis (Hnilicova et al., 2014). Str13 (scr3559) was associated in vivo with the RNAP core but not with the HrdB-RNAP holoenzyme.

Therefore, we concluded that Str11 (scr1505) is neither 6S RNA nor Ms1 but a sRNA of unknown function. To the contrary, Str13 (scr3559) is a bona fide homolog of Ms1 in S. coelicolor and we propose to rename Str13 (scr3559) as Ms1.

In Mycobacterium smegmatis and Rhodococcus erythropolis, Ms1 is an abundant RNA in stationary phase, prominently visible in the gel, similarly to 6S RNA in Bacillus subtilis (Figure 4). In Streptomyces coelicolor, we detected a weak ∼230 nt long RNA visible in stationary phase RNA, which might be Ms1 RNA identified in this study. However, in Corynebacterium glutamicum, a species that is relatively evolutionarily close to Mycobacteria (both are in one order - Corynebacteriales), there are no prominent bands in the ∼200–300 nt range, suggesting that it might not contain an Ms1 RNA or its expression is below the detection limit of the staining. Therefore, we decided to extend the linguistic search to the whole group of Actinobacteria to reveal how widespread Ms1 RNA is within other Actinobacteria species.

After we identified S. coelicolor Ms1 RNA using the linguistic search, we applied it to other Actinobacteria. A flowchart of the search procedure is shown in Figure 5.

We started with the synteny phrase identified in Mycobacterium, Rhodococcus, Nocardia, Gordonia, Mycobacteroides, Hoyosella, and Tsukamurella (Table 1, step ii. in Figure 5) which had been used to discover Str13 (Ms1 homolog) in Streptomyces coelicolor.

For the search, Actinobacteria genera with more than four annotated species available in GenBank, according to NCBI Taxonomy (Schoch et al., 2020), were used. In total, there were 40 including Streptomyces. Synteny hits (step iv. in Figure 5) were obtained in five of them, namely in Cellulomonas, Williamsia, Actinospicaceae, Actinopolyspora, and Streptomyces (Str13). Note that sequences of the synteny hits, i.e., IGRs containing putative Ms1 RNAs, may be dissimilar to each other and therefore could not be identified by sequence similarity searches.

Based on sequence similarity to synteny hit from Streptomyces actuosus, we identified IGRs containing Ms1 in 158 species from 46 Actinobacteria genera other than Streptomyces (step v. in Figure 5). Based on occurrence of specific words from the annotations of the identified IGRs (Supplementary Figure 2 step vi. in Figure 5), we generated the new synteny phrases (step i. in Figure 5) and added them to the original phrases (this returned us to the step ii. in Figure 5 and new Table 2 was generated with updated synteny phrases).

Using the updated synteny phrases and the second iterative synteny search, a synteny hit in Cellulomonas gilvus was found. Based on sequence similarity, Ms1 IGRs in 708 species from 109 Actinobacteria genera were identified. Specific words and a histogram of their occurrence in synteny annotations of the 708 Ms1 IGRs are shown in Supplementary Figure 3. The subsequently updated synteny phrases are shown in Table 3.

The phrases in Table 3 yielded synteny hits into another three genera: Williamsia, Actinopolyspora, and Actinospica whose synteny annotations did not produce any new synteny phrases. Sequence similarity of the Williamsia synteny hit was limited to the Williamsia genus and sequence similarity of the both Actinopolyspora and Actinospicaceae synteny hits identified with mostly already known Ms1 RNA candidates in species of evolutionarily closed genera.

Interestingly, in Catenulispora (Catenulisporales), Brevibacterium (Micrococcales), Corynebacterium (Corynebacteriales), Actinomyces (Actinomycetales), and Bifidobacterium (Bifidobacteriales) we found no IGRs that could contain Ms1 RNA (Figure 5A, labeled in gray). We found IGRs with the Ms1 RNA synteny in the species of these taxonomic groups but they were too short to accommodate Ms1 RNA. For example, in Corynebacterium bovis, the Ms1-syntenous IGR was only 6 bps long and in species of the other groups there were Ms1 RNA-syntenous IGRs between only 20 and 100 nucleotides long.

In summary, the linguistics gene synteny search identified Ms1 RNA homologs in 824 Actinobacteria species (Supplementary Table 2 and Supplementary Figure 5 for specific words in their synteny annotations) belonging to 146 genera and 14 Actinobacteria orders (Figure 6A, labeled in red and Table 4). Thus, Ms1 RNA is widespread among Actinobacteria and Ms1 RNA interaction with the RNA polymerase core is conserved both in Mycobacterium smegmatis and Streptomyces coelicolor.

We bioinformatically identified Str13 as the Ms1 RNA candidate in S. coelicolor. Str13 overlaps with scr3559 sRNA (Figure 2B), an sRNA with unknown function. We showed that Str13/scr3559 is an Ms1 homolog in S. coelicolor. Str13/scr3559 had a similar predicted secondary structure with the mycobacterial Ms1 (cf. Figures 2H,I) and both RNAs bind the RNAP core (Figure 3). Ms1 RNA is thus conserved in evolutionarily distant and morphologically divergent Mycobacterium and Streptomyces. Mycobacteria are unicellular rod-shaped bacteria, while Streptomyces have a complex life cycle, which starts with the germination of a spore that prolongs into filamentous tubes of highly branched vegetative (primary) mycelium, then secondary mycelium is formed and eventually spores (Trotochaud and Wassarman, 2004, 2006; Faucher et al., 2010; Cavanagh and Wassarman, 2013; Hoch et al., 2015). Despite the completely different life cycles, both bacterial species have maintained Ms1 sRNA.

In circular genomes of Mycobacterium, Nocardia, and Rhodococcus, Ms1 RNA is located close to the ori (replication start site) with the direction of transcription oriented toward it (Hnilicova et al., 2014). This is similar to Ms1 RNAs in Streptomyces that is positioned in the middle of the linear genome, close to ori (Figure 2A). The Streptomyces genomes have a core region of about 4.9 Mb containing essential genes and left and right arms with 1.5 Mb and 2.3 Mb, respectively (Hopwood, 2006), carrying mostly non-essential and species-specific genes. The position of Ms1 RNA genes in the core region indicates that this sRNA belongs among conserved genes in Streptomyces, consistent with our findings that identified Ms1 RNA candidates in 188 Streptomyces species (see Supplementary Figure 6). We further found Ms1 RNA in 145 other Actinobacteria genera (Table 4). Note that the number of newly identified Ms1 RNAs was limited by the availability of annotated Actinobacteria genomes as there were relatively many genera with only one or two genomes available or/and with a single species classified per genus.

In some genera, such as Corynebacterium, Bifidobacterium, and Actinomyces, we detected the Ms1 synteny but the IGR was too short to accommodate Ms1. The missing Ms1 RNA-containing IGR in Corynebacterium was consistent with the absence of a prominent band in the 200–300 nt range in RNA gels from Corynebacterium (Figure 4). The Corynebacterium genus belongs to the Corynebacteriales order, which also includes Mycobacterium, Nocardia, Rhodococcus [where Ms1 RNA has been already described (Hnilicova et al., 2014)], Williamsia, Gordonia, and Tsukamurella, where we identified Ms1 RNAs using the linguistic gene synteny search (Figure 6A). Within the Corynebacteriales order, Corynebacteria have the smallest genome (Figure 6B). Ms1 thus might have been lost from Corynebacterium due to the evolutionary pressure to maintain a reduced genome. Alternatively, Ms1 RNA could be essential for Actinobacteria with the larger genomes.

A comparison of genome sizes of the main Actinobacteria orders with the occurrence of predicted Ms1 RNAs (Figures 6A,C) reveals a trend where Ms1 is lost in bacteria with smaller genomes while the Ms1 synteny is still present. Examples are Bifidobacteriales and Actinomycetales where the respective IGRs were too short to accommodate Ms1 RNA; these orders have the smallest genomes within Actinobacteria (mostly <3.0 Mbp).

However, in Frankiales, Geodermatophilales, and Glycomycetales, where neither the Ms1 synteny nor Ms1 RNA were detected, genome sizes were comparable to Actinobacteria orders with identified Ms1 RNAs. Frankia (Frankiales) genome sizes vary between 4.3 and 10 Mb and this variability is due to the degree of host dependence. Frankia are N₂-fixing filamentous plant symbiotic bacteria that can either survive independently in the soil or be entirely dependent on their host plants (Benson et al., 2011). The high diversity of Frankia genomes might be a reason why Ms1 RNA was not detected in our search (the gene synteny on which our search depends, might be too low in Frankia). Glycomycetales are aerobic bacteria that produce branched vegetative mycelia and aerial mycelia with chains of square-ended conidia (Labeda and Kroppenstedt, 2004). Glycomyces were isolated from soil, hypersaline habitats, and seawater (Han et al., 2014; Xing et al., 2014; Nikitina et al., 2020). Geodermatophilales create pigmented (very often black) colonies with the individual cubic cells and extracellular matrix forming cauliflower-like aggregates and have been reported to be highly resistant to stresses such as gamma-radiation, UV, and desiccation (Hezbri et al., 2016). Species from both Glycomycetales and Geodermatophilales can adapt to extreme stress conditions and thus might have evolved specific regulatory pathways independent of Ms1 RNA.

To summarize this section, the genome size is not the only indicator of the Ms1 RNA presence. Alternatively, Ms1 RNA may be present in these species but the Ms1 RNA gene synteny was lost and therefore we were unable to detect Ms1 RNA using our gene-synteny based approach.

The Ms1 RNA synteny itself, especially the “HAD hydrolase” gene is of interest. HAD hydrolase is annotated also as “inhibition of morphological differentiation protein” or “phosphoserine phosphatase” (see Tables 1–3) and it was found in the Ms1 RNA synteny of most species. In Mycobacterium smegmatis, HAD hydrolase is MSMEG_6173, a 293 amino acid long protein with a predicted transmembrane domain at its C-terminus. SCO3558, a 5′ end flanking gene of scr3559, that shares 60% identity with MSMEG_6173, is homologous to CicA (Bellier et al., 2006), which encodes a phosphotransferase in Caulobacter crescentus. An increased concentration of CicA in Caulobacter crescentus causes a loss of the normal rod shape, an almost 10-fold increase of the bacteria’s cell volume and a cell division block (Fuchs et al., 2001). As Ms1 regulates the RNAP amount in stationary phase (Sikova et al., 2019) it is tempting to speculate that the conserved association of the HAD hydrolase gene with the Ms1 RNA gene indicates a link between transcription regulation and cell size and shape.

From the perspective of computational biology, our work demonstrated limits of the use of secondary structure in computational homology searches for homologs of known sRNAs in bacterial genomes. Homology searches use similarity of secondary structure between potential homologs and known sRNAs either solely [e.g., (Pánek et al., 2011)] or in combination with sequence similarity [e.g., (Barquist et al., 2016)]. Structure similarity increases the efficiency and capability of these searches to find homologs as the sRNA secondary structures are more evolutionarily conserved than sRNA sequences.

Nevertheless, still the efficiency of homology searches decreases substantially with the increasing evolutionary distance. This was demonstrated here in the extremely diverse group of Actinobacteria by the search for Ms1 RNA homologs, which was not successful. Also Rfam (Kalvari et al., 2017), an RNA database that employs the infernal software (Barquist et al., 2016) for computational search for homologs of known RNAs, provides Ms1 RNA candidates only from species closely related to Mycobacteria.

A weak point of homology searches is the limited reliability of secondary RNA structure prediction, which decreases especially for sequences >100 nt. Furthermore, a correct RNA sequence, i.e., a sequence with both a correct position in the genome and a correct length, is required for the prediction. But it is not always available in the homology search as (1) the length of the sequence can vary substantially between species or genera, and (2) genomic position of the sequence can be only approximated. In the presented work, the length difference between the known M. smegmatis Ms1 RNA and the identified S. coelicolor Ms1 RNA homolog was 70 nucleotides and therefore predicted secondary structures of potential S. coelicolor Ms1 RNA homologs were wrong and structure similarity to known Ms1 RNA could not be detected. Thus, the unavailability of correct sequences of potential homologs could be another reason why homology search was not successful in the presented work.

To overcome this problem, we employed a genomic synteny search. We adopted a linguistics approach based on similarity of genome annotations rather than similarity of nucleotide or amino acid sequences of flanking genes. Ms1 RNA has conserved synteny with a highly specific phrase in one of its flanking genes, the ‘HAD IB hydrolase’. This phrase is relatively rare and occurs only a few times in well annotated genomes (e.g., 5 × in S. coelicolor). The annotation helped us to identify the IGRs that might contain Ms1 RNA even in extremely distant species as represented here by the Actinobacteria genera with predicted Ms1 RNA, in which flanking nucleotide/amino acid sequences might be dissimilar.

We designed the most parsimonious form of a linguistic search for conserved synteny based on text searches for exact words or phrases. Once the search had identified the first possible Ms1 homolog in a genus (synteny hit), it then proceeded in an iteratively progressive manner within the genus and also in evolutionarily close genera using sequence similarity of the synteny hits. The obtained information was subsequently applied to other genera, expanding the list of identified candidates. Even in this simple form, the computational text search was able to identify Ms1 homologs in Streptomyces and other distantly related Actinobacteria genera. Its versatility and ease of use make it a convenient tool that can be, in principle, applied to searches for other RNAs/genes, allowing their fast identification across a wide range of organisms.