In this study, the Streptomyces coelicolor wild-type (wt) strain M145 (Kieser et al., 2000) and its RNase III-deletion strain derivative [rnc, M145 rnc::aac(3)IV (Sello and Buttner, 2008)] were used. 10⁸ spores were inoculated on solid R2YE medium (Kieser et al., 2000) covered by cellophane at 29°C. Bacterial samples were collected after 24, 48, and 72 h of cultivation, where each time point represented a different developmental stage, i.e., vegetative mycelium, aerial mycelium, and spores.

Total RNA was isolated using a TRIzol method (Van Dessel et al., 2004). Harvested cells were immediately submerged in TRIzol reagent (Ambion) on ice (1 ml of TRIzol per 50 cm² of culture dish surface area). Five glass beads (3 mm in diameter) were added to the cell suspension. The cells were disrupted using a Minilys homogenizer (Precellys) twice for 2 min at 3,000 rpm and twice for 2 min at 4,000 rpm, cooled on ice between the cycles. The samples were subsequently centrifuged for 2 min at 10,000 g and purified in TRIzol/chloroform (5:1) and chloroform. For RNA precipitation, the samples were incubated in isopropanol at −20°C overnight and centrifuged for 30 min at 10,000 g. RNA samples were washed in ethanol and resuspended in 30 μl of RNase-free water. Residual DNA in the RNA samples was removed by DNase I treatment (Ambion). Typically, a concentration between 1 and 3.5 μg/μl was obtained. RNA quality was checked on a 1% agarose gel.

RNA samples were isolated after 48 h of cultivation of both wt and rnc strains. Antisense RNA expressions were tested by means of the FirstChoice RLM-RACE Kit (Ambion) following the manufacturer's protocol with the several exceptions:

Because the uncapped 5′ ends of bacterial RNAs are sensitive to the CIP (calf intestinal phosphatase) enzyme dephosphorylation, the treatment was omitted from the 5′ RACE procedure.

Final PCR products were separated on a 1.2% agarose gel. Products that were found in samples but absent in negative controls were excised and purified using the Qiagen MinElute PCR purification kit. The purified products were cloned into the TOPO vector using the TOPO TA Cloning (Invitrogen) and transformed into E. coli One Shot TOP10F' competent cells (Invitrogen). Plasmids containing the cDNA inserts were extracted using the QIAprep Miniprep kit, and sequenced to map 5′ and 3′ ends of RNAs reverse-transcribed.

Experiments were performed according to the reverse transcription and PCR protocols described in the FirstChoice RLM-RACE Kit using gene specific DNA probes (details in Results and discussion). The PrimeScript (Takara, 100 units per 10 μl of reaction mixture) was used as a reverse transcriptase. A negative control PCR reaction used the original RNA sample as a template.

Here, to probe if the antisense RNA expression occurs in the vicinity of the RNase III-binding mRNAs (or alternatively, the binding is not complexed with asRNA), we firstly analyzed the genes whose transcripts are affected by RNase III (i.e., those mRNAs that are increased by more than two-fold in the JSE1880 rnc-mutant or detected in BARD as listed in the Table S2 in Gatewood; Gatewood et al., 2012). We asked whether in the proximity of these genes lies a gene encoding for any of 1713 small RNAs found or predicted in S. coelicolor up-to-date (Pánek et al., 2008; Swiercz et al., 2008; D'Alia et al., 2010; Vockenhuber et al., 2011; Moody et al., 2013) that could act on the messenger by the antisense mechanism. We found out that from the 153 mRNAs increased in JSE1880 (at a single experimental time point), 45 neighbor with one of the 1713 small RNAs that could thus theoretically act as an asRNA. From these 45 mRNAs, 21 mRNAs were listed in the BARD, i.e., they bind RNase III. These data, summarized in Table 1, encouraged us to search for novel asRNAs expressed in the opposite direction to other RNase III-binding mRNAs.

Genes for the experimental analyses had been selected independently of the in silico search above. We altogether tested 30 genes for the possible novel antisense expression. The initial experimental set consisted of 17 exemplar genes which we selected out of a total of 37 genes whose mRNAs were enriched in the BARD, i.e., they co-precipitated with RNase III (listed in Table 2, see also Gatewood's Table 2 and Table S3 in Gatewood et al., 2012). Within this set, only as1625, as2198, and as5112 genes have been predicted before (see Table 1). The control set consisted of three other genes—SCO5737, adpA, and rnc (the gene of RNase III) that had been previously shown to be directly targeted by RNase III in vitro (Xu et al., 2008, 2010; Gatewood et al., 2012). Additionally, to determine whether the existence of antisense transcripts involves only those messengers bound by RNase III or is even more widespread, we decided to add into the control set the genes whose transcripts have not been shown to bind RNase III. For these tests, we selected 10 genes encoding sigma factors (HrdA, HrdB, HrdC, HrdD, SigB, SigD, SigE, SigH, SigR, and WhiG) as important transcriptional regulators that govern gene expression, controlling cell development and/or responses to various stresses (Bobek et al., 2014).

We reasoned that many asRNAs overlap the ribosome binding site and the start codon of their target, possibly leading to negative translational control. Following this, all of the DNA primers used here to find antisense transcripts have been designed accordingly (Figure 1). To demonstrate expression of antisense transcripts, 5′ RACE analyses were performed using the primers. Sequence(s) extended from the primer was/were PCR amplified and sequenced. To avoid false positive results, each reverse transcription was accompanied with a negative control sample lacking the reverse transcriptase in the reaction. The resulting electrophoretograms of both the experimental and the negative control samples were compared to exclude non-unspecific fragments in the experimental sample from sequencing.

From the total of 30 samples tested (involving both the experimental and the control set), 22 revealed a novel antisense transcript in cis (5′ end(s) detected; see Table 2).

From the 17 samples of the experimental set (i.e., our selection from those mRNAs enriched in the BARD), 17 cognate asRNAs were detected (three of them have been predicted previously as could be seen in Table 1), signifying a 100% outcome within this group (Figure 1A). The 3′ RACE revealed 16 3′ ends within the experimental set (scr0168 unsuccessful, northern blot was not performed).

In the control set (Figure 1B), our RACE data detected transcripts antisense to adpA (as-adpA), rnc (as-rnc), and three sigma factor genes (as-sigB, as-sigH, as-sigR). Seven sigma factor genes and the transcript SCO5737, which is bound by RNase III in vitro, did not reveal antisense expression.

Due to our primer design, we did not detect such antisense transcripts acting on messenger's 3′ end [in Moody (Moody et al., 2013) termed as cutoRNAs]. Also the experimental approach did not allow identification of trans-encoded sRNAs, although similar complementary sequence on the genome could be found (as examples, novel RNAs as0772, as0600, as3983, and as5216 could theoretically act on other targets throughout the genome).

Sometimes the RACE analysis may produce false positive results due to the RNA self-priming during the reverse transcription which could be theoretically caused by stem-loop structures formation. In order to confirm the existence of the newly found cis-antisense transcripts and to compare their expression during Streptomyces cell development (24, 48, and 72 h at standard growth conditions, see methods), we performed northern blot analyses. From the total of 22 analyzed, 18 asRNAs revealed an apparent signal (Figure 2; for the raw northern blot images see Presentation 1 in Supplementary Material). The signals of the remaining four were too weak or undetectable.

Details describing expression profiles of asRNAs are summarized in Table S1 and signal quantification is presented in Table S2 online, where the listed relative values proportional to the 5S RNA controls correspond with intracellular levels of the analyzed transcripts.

In Gatewood's work, expression of two small RNAs, scr6925 and scr2101, was enriched in the rnc strain (scr = small S. coelicolor RNA). These data imply their antisense role and subsequent processing by the RNase III enzyme. The expression of scr2101 was previously revealed by Swiercz (Swiercz et al., 2008). Gatewood et al. (2012) showed that the molecule is up-regulated in the JSE1880 strain (lacking the RNase III), where its level was 7-fold higher compared to that in the wild type strain. scr2101 was also shown to be bound by the enzyme (presented in the BARD, although its enrichment was only 0.57). As the RNA's gene lies on the opposite strand between SCO2100 and SCO2101, we raised the question whether these two protein-coding genes form an operon, producing one common transcript targeted by the scr2101 antisense RNA. The potential SCO2100-SCO2101 common transcript was used here as a template for reverse transcription using a DNA probe complementary to the 5′ end of the SCO2101 mRNA (TGTCCCGGCTGCTCCAGGGA). The second DNA primer, used for the following PCR amplification, was identical to the 3′ end of the SCO2100 mRNA sequence (CGTAGGTCCCCGCCCGCT), thus forming a 635-nt product, which was indeed produced. Use of a negative control PCR reaction, where the original RNA was used instead of a template, eliminated the risk of false positive results (Figure S1). This finding suggests that the antisense function of scr2101 (which thus should be termed as as2101) targets the region between two open reading frames. Although the binding experiments are outside the scope of this paper, our findings raise demand to expand the RNase III-binding analyses to the unannotated sRNA transcriptome.

The double stranded stems of stem-loop mRNA structures are thought to be too short to be digested by the RNase III enzyme (Robertson, 1982). On the other hand, one may argue that the stem-loop structures on several mRNAs are the only possible targets for RNase III activity. Indeed, the enzyme is involved in pre-rRNA, tRNA, and polycistronic RNA processing (Conrad and Rauhut, 2002; Drider and Condon, 2004), where such stem-loop structures are present and possibly long enough to be cleaved by the enzyme. Gatewood clearly showed that some lone mRNAs are targeted and cleaved by RNase III in vitro (SCO5737, or the SCO3982 to SCO3988 mRNA region). In accordance to the first example, our experiments did not reveal any antisense transcript to the SCO5737 (which encodes a polynucleotide phosphorylase). Its mRNA transcript is 2,220 nt in length, and thus might be capable of forming several longer double-stranded regions observed by RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi; fig. not shown), which are most likely targeted by the RNase III enzyme in vitro.

On the other hand, we show that at least one member of the SCO3982-SCO3988 operon – SCO3983 possesses an antisense gene. Its transcript, as3983, is 103 nt long and exhibited a strongest expression signal in vegetative cells (24 h) of the rnc strain. The mRNA also had a strong signal at 24 h and 48 h, even elevated in the rnc strain. Moreover, we observed that the as3983 sequence is nearly identical to the region adjacent to SCO3268 and theoretically may thus also act on this transcript in trans.

In addition to the identified as3983 RNA, we have found antisense transcripts to two other mRNAs targeted by RNase III in vitro. These two mRNAs code important developmental regulators that influence antibiotic production—AdpA and the RNase III itself.

According to Gatewood's results, the adpA mRNA was enriched 22-fold in the BARD. Therefore, we asked whether AdpA expression is controlled by an antisense mechanism. Indeed, our RACE analysis revealed a 223-nt long antisense transcript termed as-adpA (Figure 1B) and although subsequent northern blot weakly detected the transcript in the wt strain (faint signal at 72 h stage of growth, Figure 2B), it was clearly found in the rnc strain at later stages, peaking at the 48 h old mycelium.

The rnc gene (SCO5572) that encodes the RNase III enzyme is the last member of a three gene operon, about 2,000 nt in length, that also encodes a hypothetical protein SCO5570 and a ribosomal protein L32 (SCO5571). A point mutation at amino acid position 120 of the RNase III protein causes deficiency in the ribonucleolytic activity of the enzyme, whereas its ability to bind double stranded RNAs remains intact (Huang et al., 2005). Observations that the rnc mRNA abundance is increased in these mutants led to discovery that RNase III cleaves among others its own transcript (Xu et al., 2008). Consistently, our northern blot analyses did not detect the rnc transcript in the wt strain, though we clearly confirmed the as-rnc RNA, which was detectable during the time course. Hence the antisense mechanism is probably involved in gene expression control of regulatory proteins such as the RNase III enzyme or the AdpA transcription regulator. Our future effort will be focused on more detailed characterization of the involvement of these two novel asRNAs in Streptomyces development.

as2198 is an asRNA to the glnA gene (SCO2198, encoding a glutamine synthase) and was previously independently shown to be expressed, termed cnc2198.1 (D'Alia et al., 2010). The authors performed a detailed functional analysis and revealed that overexpression of cnc2198.1 affects growth rate and antibiotic production. In the overexpression strain, the intracellular level of the targeted GlnA protein was decreased by 40%. The authors speculated that the glnA-cnc2198.1 RNA complex blocks glnA translation, which may lead to its subsequent degradation. We further hypothesize that the complex is degraded by the RNase III enzyme, as our northern blot revealed that the asRNA level is increased in the rnc mutant. Our RACE mapping estimated the size of the antisense transcript to be 432 nt, nearly four times longer than the 121 nt transcript described by D'Alia et al. (2010). Moreover, the detected 3′ end overlapped the adjacent SCO2197, revealing that as2198 is in fact the 5′UTR of the SCO2197 mRNA. The fact that our northern blot showed a second, weaker fragment of 425 nt in size, could mean that the SCO2197 gene possesses another promoter or its transcript is further processed. Both of the detected fragments, as well as the cognate SCO2198 mRNA, had the strongest expression signals in samples from vegetative 24-h old cells, even strengthened in the rnc strain.

Because the sense mRNAs of the antisense transcripts found here are targeted by RNase III (or at least, their expression was negatively affected in the presence of the enzyme), we speculated that the RNase III enzyme is the most likely candidate for the paired transcript degradations. The 18 northern blot positive results proved the existence of novel asRNAs. However, only five of those asRNAs were increased in the rnc strain when compared to the expression in the wild type strain. These included as0494, as0864, as2198, as3983, and the as-adpA. On the other hand, in two other cases, our northern blots suggest even a positive effect of the presence of RNase III enzyme on the cellular level of asRNAs (as0323 and as5112, see below). Consistently, Gatewood (Gatewood et al., 2012) showed that the majority of known sRNAs, detected in their RNA-seq analysis, do not exhibit significant expression differences between the wild-type and the rnc-deletion mutant. As another example of sense-antisense RNA pair with an undistinguishable or even a positive effect of the enzyme on its stability in Streptomyces is the scr4677-SCO4676 complex (Hindra et al., 2014). The possibility that the RNase III enzyme does not always post-transcriptionally degrade sense-antisense complexes is also inferred from a study on sRNA degradation by three different RNases (RNase Y, J1, and III) in another Gram-positive bacterial model, Bacillus subtilis (Durand et al., 2012). In Bacillus, RNase III depletion has little effect on antisense RNAs observed by high-resolution tiling arrays. Although several RNAs showed increased abundance in the RNase III mutant, their half-lives were not affected by the enzyme, as observed by northern blot analysis. The authors concluded that the role of RNase III in Bacillus subtilis lies more likely in indirect transcriptional control rather than post-transcriptional RNA turnover. One may argue that the function of the enzyme might be substituted by other ribonucleases in rnc mutants. As the RNase III-binding mRNAs may serve as a fruitful source for novel antisense transcript discoveries, we can assume that the mRNA-asRNA pairs are targeted by the enzyme but not always degraded. An involvement of some other RNA-binding proteins, such as RNA helicase or Hfq-like protein which has not been found in Streptomyces yet, could be expected (Gerhart Wagner, Uppsala University, Sweden, personal communication). On the other hand, the possibility that in other cases the antisense transcripts might protect their mRNA targets against the RNase III cleavage should be taken in mind. Clarification of the exact role of RNase III enzyme in Streptomyces remains to be established.

The genus Streptomyces can be presented as a model bacterial group lying on the top of prokaryotic cellular complexity. Their 8–10 Mbp long genome encodes all the developmental stages, including morphological changes (spore formation and germination, vegetative branching hyphae, aerial twisting mycelium), secondary metabolite production (antibiotics and a variety of other bioactive compounds, siderophores, pigments, etc.) and a capacity to respond to all possible environmental changes (diverse stresses) that are encountered. Developmental transitions and environmental intricacy require advanced regulatory networks that involve a concerted action of more than 900 transcription regulatory proteins known thus far. Here we unveiled 22 novel antisense transcripts, out of which 18 were confirmed by northern blot analyses. These results suggest an equivalent role and possibly even bigger number of non-protein-coding RNA regulators. The cis-antisense transcripts are efficient gene expression modulators with minimal space requirements on their genome (Georg and Hess, 2011), with a theoretical capability to act on nearly all genes. Moreover, the mode of their action (whether based on co-transcriptional collision or post-transcriptional 100% complementarity) is simple and effective. Our work suggests that antisense transcription is widespread in Streptomyces and somehow connected with the function of RNase III, although the absence of the rnc gene did not greatly influence the majority of the transcripts. It is noteworthy that the majority of antisense transcripts found here escaped previous sRNA predictions and/or whole genome searches, suggesting that their expression level is often low and may be lost during statistical background subtraction. Nevertheless, the results in this work confirm the extensiveness of the antisense transcripts and raise the demand for elucidation of their role in gene expression control in Streptomyces.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.