Detection of Very Long Antisense Transcripts by Whole Transcriptome RNA-Seq Analysis of Listeria monocytogenes by Semiconductor Sequencing Technology

The Gram-positive bacterium Listeria monocytogenes is the causative agent of listeriosis, a severe food-borne infection characterised by abortion, septicaemia, or meningoencephalitis. L. monocytogenes causes outbreaks of febrile gastroenteritis and accounts for community-acquired bacterial meningitis in humans. Listeriosis has one of the highest mortality rates (up to 30%) of all food-borne infections. This human pathogenic bacterium is an important model organism for biomedical research to investigate cell-mediated immunity. L. monocytogenes is also one of the best characterised bacterial systems for the molecular analysis of intracellular parasitism. Recently several transcriptomic studies have also made the ubiquitous distributed bacterium as a model to understand mechanisms of gene regulation from the environment to the infected host on the level of mRNA and non-coding RNAs (ncRNAs). We have used semiconductor sequencing technology for RNA-seq to investigate the repertoire of listerial ncRNAs under extra- and intracellular growth conditions. Furthermore, we applied a new bioinformatic analysis pipeline for detection, comparative genomics and structural conservation to identify ncRNAs. With this work, in total, 741 ncRNA locations of potential ncRNA candidates are now known for L. monocytogenes, of which 611 ncRNA candidates were identified by RNA-seq. 441 transcribed ncRNAs have never been described before. Among these, we identified novel long non-coding antisense RNAs with a length of up to 5,400 nt e.g. opposite to genes coding for internalins, methylases or a high-affinity potassium uptake system, namely the kdpABC operon, which were confirmed by qRT-PCR analysis. RNA-seq, comparative genomics and structural conservation of L. monocytogenes ncRNAs illustrate that this human pathogen uses a large number and repertoire of ncRNA including novel long antisense RNAs, which could be important for intracellular survival within the infected eukaryotic host.


Introduction
Listeria monocytogenes is a non-sporulating, Gram-positive soil bacterium which has a low GC content. The ubiquitous nature of the bacterium enables it to enter the human food chain via foodprocessing environments. In addition, the ability of L. monocytogenes to grow at low temperatures and to resist harsh preservation techniques increases the risk of food contamination. By uptake via contaminated food products, L. monocytogenes can cause listerial infection known as listeriosis. Listeriosis often manifests with clinical symptoms such as meningitis, meningoencephalitis, septicaemia, abortion, prenatal infection and also gastroenteritis. Furthermore, high mortality rates of up to 20-30% in humans which are diseased with listeriosis (especially pregnant women, elderly and immunocompromised persons) makes L. monocytogenes a serious life-threatening human pathogen [1,2].
Comparative whole genome sequencing of representative strains comprising the entire species of L. monocytogenes was performed by Kuenne et al. [9]. In the genus Listeria, genome reduction has led to the generation of non-pathogenic species from pathogenic progenitor strains [10]. Indeed, many of the genomic regions specific for pathogenic species (such as L. monocytogenes) represent genes which are absent in non-pathogenic species (such as L. innocua and L. welshimeri) [10]. This also effects the number of conserved non-coding RNAs (ncRNAs) within the genus Listeria [9,11]. Recently genome sequencing of different L. monocytogenes serotypes has been accompanied by transcriptional profiling using whole genome microarrays and RNA-seq. This has been done to examine the adaptive changes of L. monocytogenes to grow in different natural environments and to identify responsible genes and ncRNAs mediating transcriptional responses [9,[11][12][13][14][15]. For L. monocytogenes, 262 ncRNAs have been identified yet including 134 putative sRNAs, 86 antisense RNAs (asRNAs) and 42 riboswitches [16]. Also in other bacteria, asRNA transcripts could be observed for 10% up to 50% of protein-coding genes, e.g. in Escherichia coli, Synechocystis sp. PCC6803, Helicobacter pylori [17], Bacillus subtilis [18] and Mycobaterium tuberculosis [19].
In this study we present information on transcriptomic profiling using RNA-seq, comparative genomics and structural conservation of L. monocytogenes ncRNAs. The bacterial strains have been grown in BHI broth (extracellular conditions) and in the cytosolic environment of the host cell (intracellular condition). To our best knowledge, this is the first time that Ion Torrents Personal Genome Machine (PGM) (Life Technologies) was used for RNAseq analysis of a bacterial human pathogen by next generation semiconductor sequencing technology to detect novel small and long ncRNAs. Using this technology, we found antisense transcripts in Listeria with a length up to 5,400 nt.

Bacterial strains and growth conditions
The strains L. monocytogenes EGD-e [20], L. monocytogenes 1043S [21] and L. monocytogenes EGD-e DprfA [22] were grown in BHI broth (VWR) overnight at 37uC with shaking at 180 rpm (Unitron, Infors). Overnight cultures were diluted 1:50 in 20 ml fresh BHI broth using a 100 ml Erlenmeyer flask and were incubated at the same conditions mentioned above until OD 600 nm 1.0.
Cell culture and infection model P388D1 murine macrophage cells (ATCC CCL-46) were cultured in RPMI1640 (Gibco) supplemented with 10% fetal calf serum (PAA Laboratories) in 85-mm-diameter tissue culture plates. For intracellular growth assays bacteria were added to P388D1 murine macrophages monolayer at a multiplicity of infection (MOI) of 10 Listeria per eukaryotic cell. The intracellular growth assays were performed as described in [23].

RNA isolation
For RNA extraction from L. monocytogenes grown extracellularly in BHI, we applied aliquots of 0.5 ml from the same Listeria culture grown until mid-exponential phase used to infect P388D1 macrophages. The bacterial cells were treated with 1.0 ml RNA protect (Qiagen) for 5 min and were collected by centrifugation for 10 min (8000 g). The bacterial pellets were stored at 280uC until use. RNA extraction from intracellularly grown L. monocytogenes in macrophages, 4 h post infection, was performed as described previously [23]. Briefly, infected host cells (see above: Cell culture and infection model part) were lysed using cold mix of 0.1% (wt/ vol) sodium dodecyl sulfate, 1.0% (vol/vol) acidic phenol and 19% (vol/vol) ethanol in water. The bacterial pellets were collected by centrifugation for 3 min (16,000 g). Total RNA was extracted using miRNeasy kit (Qiagen) with some modifications [11]. The collected pellets were washed with SET buffer (50 mM NaCl, 5 mM EDTA and 30 mM Tris-HCl (pH 7.0)). After centrifuga-tion at 16000 g for 3 min pellets were resuspended into 0.1 ml Tris-HCl (pH 6.5) containing 50 mg/ml lysozyme (Sigma), 25 U of mutanolysin (Sigma), 40 U of SUPERase (Ambion), 0.2 mg of proteinase K (Ambion). The incubation for 30 min was carried out on a thermo mixer at 37uC and with shaking (350 rpm). QIAzol (Qiagen) was added, mixed gently and incubated for 3 min at room temperature. An additional incubation for 2 min at room temperature was done after adding 0.2 volume chloroform followed by centrifugation at 16000 g at 4uC for 15 min. The upper aqueous phase, containing RNA, was transferred to a new collection tube and 1.5 volumes of 100% ethanol was added and mixed thoroughly. The probes containing RNA were transferred into columns supplied with the miRNeasy Kit (Qiagen) and treated according to the manual including an on-column DNase digestion (RNase-Free DNase, Qiagen). RNA was eluted by RNase-free water and stored at 280uC until needed. The quantity of the isolated total RNA was determined by absorbance at 260 nm and 280 nm, and the quality was assessed using Nanochips for Agilents 2100 Bioanalyzer.

RNA sequencing
To deplete bacterial rRNA we applied the Ribo-Zero Magnetic Kit (Bacteria) (Epicentre) and treated the depleted RNA with tobaco acid pyrophosphatase (Epicentre) as recommended by the manufacturer.
Afterwards, the RNA was fragmented by RNase III (Applied Biosystems) at 37uC for 4 min. The yield and size distribution of the fragmented RNA was assessed using Quant-iT RNA assay kit with Qubit Fluorometer (Invitrogen) and the Agilent RNA 6000 Pico Chip kit with Agilent 2100 Bioanalyzer instrument. Size distribution of RNase III fragmented RNA delivered median size of 200 nt. For the cDNA library preparation, Ion Total RNA-seq kit v2 (Ion Torrent, Life Technologies) was used as recommended by the manufacturer. The libraries were purified by AMPure XP Reagent (Beckman Coulter). The yield and size distribution of the amplified cDNA were assessed by Qubit Fluorometer (Invitrogen) and DNA 1000 kit (Agilent). In the next step, clonally amplified Ion Sphere Particles (ISPs) containing the amplified cDNA were prepared using the Ion OneTouch System (Life Technologies). The amplified libraries were diluted to 8.3 nM and loaded on 316 Chip of the Ion Torrent semiconductor sequencing instrument personal genome machine (PGM) (Life Technologies).

Real-time-RT-PCR
Reverse transcription to produce cDNA was performed by SuperScript II Reverse Transcriptase (Invitrogen) using 1 mg RNA. The probes were subjected to quantitative real-time PCR in a final volume of 25 ml using QuantiTect SYBR Green PCR kit (Qiagen) according to the manufacturers instruction. A standard curve was generated for the used primer pairs (see supplemental material) using different copy numbers of genomic DNA from L. monocytogenes EGD-e. For each primer pair a negative control (water), RNA sample without reverse transcriptase (to determine genomic DNA contamination) and a sample with known amount of copy numbers (to test the efficiency of the reaction) were included as controls during cDNA quantification. After real-time PCR all samples were run on a 1.5% agarose gel to verify that only a single band was produced. The expression level of each gene was calculated by normalizing its mRNA quantity to the quantity of the mRNA of gyrB encoding gyrase B [24] for the same sample using a mathematical model for relative quantification in real-time PCR published by Pfaffl [25].

In silico Genome Data Analysis
In order to analyze the genome of L. monocytogenes (NC_003210) with RNA-seq data and to detect potential novel ncRNAs, we investigated the genome searching for: (a) proteins, (b) known ncRNAs, (c) conserved regions, (d) locally stable structures, (e) possible de novo ncRNAs, and (f) positions of known potential small RNAs from literature [13,15,26].
Annotation of known proteins. Protein annotation from NCBI (NC_003210) was extended by a de novo protein prediction with BacProt [27] based on homologous proteins of other firmicutes. Furthermore, BacProt predicts species specific novel proteins based on Listeria specific information on Shine-Dalgarno sequences and TATA boxes gained from the homology search.
Locally stable secondary structures. Locally stable secondary structures are indicating positions for small RNAs. Those structures were calculated with RNALfold (v.2.0.7) [35] using parameters -d 2 -L 120. Hits with a total length less than 50 nt were discarded. A dinucluotide shuffling of each sequence with shuffle -d -n 1000 was performed to predict thermodynamically stable RNA structures. For further analyses only extraordinarily stable structures with a Z-score cut-off #23.0 (top 5% of stable structures) were taken into account.
Detection of possible de novo non-coding RNAs. For the detection of potential novel non-coding RNAs, all intergenic regions with a minimum length of 10 nt and a minimum coverage of ten reads were defined as 'seeds'. For the analysis of long (antisense) non-coding RNAs, we merged seed regions, with a distance less than 100 nt. All candidates were scored according to For evaluation of the ncRNA candidates, a scoring system retrieved from known ncRNAs (Rfam, [13,15,26], see supplemental material) was developed. For increasing length, number of reads and GC content, scores are summed up along the column; for example, an ncRNA candidate of length 100 nt receives a score of +1. The higher the score of a candidate, the higher its probability to be an ncRNA. doi:10.1371/journal.pone.0108639.t001 Table 2. Overview of RNA-seq libraries. For further analyses, we took only candidates with a score of 2.5 or higher into account. Additionally, we checked our candidates for possible overlaps with the 59UTR predicted by Wurtzel et al. [15].

Full ncRNA candidate set
In this study we analyzed the transciptomes of L. monocytogenes grown extracellularly in BHI broth and L. monocytogenes grown intracellularly in murine macrophages. Our analysis was based on three independent biological replicates for each condition resulting in six RNA-seq libraries produced by the Ion torrent (PGM) next generation sequencing platform. We obtained 3.1-3.7 million reads up to a length of 385 nt (see Tab. 2).
The experimental approach was combined with comprehensive in silico studies. To detect novel ncRNAs, we investigated various characteristic features of ncRNAs in the L. monocytogenes genome and transcriptome: seeds, GC-content, secondary structure, conservation and multiple genome-wide alignment.
(1) A seed is defined by an intergenic region covered by $10 reads for $10 nt. We searched for seeds and merged them to one candidate if they were at most 100 nt apart. We received 2074 candidate ncRNA locations. Locations longer than 50 nt, 75 nt and 100 nt were rewarded by +0.25, +0.5 and +1 respectively (see Tab. 1). If the number of reads was at least ten, the score of the ncRNA candidate was increased by 1. If the number of reads even exceeded 100, the score was again increased by 1.
(2) We analyzed the GC-content. The whole genome of L.
monocytogenes EGD-e has an GC content of 38%. The ncRNAs of Rfam identified in L. monocytogenes EGD-e were found to have an GC content of 52% and 44% (with and without rRNAs/tRNAs). We decided to reward ncRNAs with GC content above 40% with 0.25, and another 0.25 points for GC content above 50%. However, previously reported ncRNAs [13,15,26] showed a lower GC content (on average 37%, 37.8% and 37.6% respectively). (3) Using RNALfold we searched for locally stable secondary structures. For 87/143 ncRNAs described in Rfam and 118/ 260 ncRNA candidates previously described in the literature [13,15,26], we found a region which was identified by RNALfold as locally stable secondary structure. If a candidate was predicted to contain a locally stable secondary structure region, we rewarded this candidate by adding +0.25 to its score. (4) Another hint for an (ncRNA) gene is its conservation among closely related species. Therefore, we computed a genomewide multiple sequence alignment comparing L. monocytogenes EGD-e with 12 other L. monocytogenes serotypes. If the candidate region was present in all other serotypes, the candidate was rewarded by adding another +0.25 to its score. (5) The multiple genome-wide alignment was used as input for RNAz to predict novel ncRNAs. If a candidate was identified to be a novel ncRNA with probability above 0.9, we added another +0.25 to its score.
For the further analysis we took only those novel ncRNA candidates into account that exceeded a given threshold. We chose this threshold by checking how many of the previously described ncRNAs would have been selected. For a threshold of 2.5, 132/ 143 of the ncRNAs described in Rfam and 137/260 of the previously putative ncRNAs described in the literature, would have been selected. Using this threshold, we present a set of 441 potential novel ncRNA candidates. To get a full set of ncRNA locations, we added the previously described ncRNAs to our set of novel ncRNA candidates. This results in 741 ncRNA locations (since both sets are overlapping), ranging from to 10-5,347 nt (mean: 239 nt) length for L. monocytogenes. If we use our threshold also for the previously described ncRNA locations, we get a set of 611 ncRNA candidates. The list of all candidates, their genomic locations and features as described above, as well as overlaps to previously described ncRNAs and adjacent proteins is given in the supplemental material.

Comparison to previous studies
As mentioned above, 260 locations of ncRNA candidates (including start-and stop positions) were previously described in the literature [11,13,15]. We compared our 611 ncRNA candidates with the results of these previous studies (see Fig. 1).
In 2009, Toledo-Arana et al. [13] used tiling arrays and RNAs from wild type and mutants grown in vitro, ex vivo, and in vivo, to present a complete operon map of L. monocytogenes. In this study, 100 ncRNA candidates were suggested. Of this 100 putative sRNAs, 77 locations were also confirmed by our observations, whereas 23 locations had a score #2.5 or were not even identified as seeds.
Mraheil et al. [11] reported 150 putative regulatory RNAs identified by deep sequencing with cDNA obtained from extracellularly grown bacteria and from L. monocytogenes isolated from infected macrophages using 454 pyrosequencing. From these 150 putative regulatory RNAs, we identified 102 using our method and a score threshold of 2.5. More than half of the remaining 48 ncRNAs were covered with less than 10 reads and were not part of our seeds.  [15]. Note that whenever an ncRNA prediction of this study overlaps with multiple previously described candidates, it is a single hit in the diagram. Altogether, including previous literature, Rfam and this work, now 741 putative ncRNAs are described. In this work we defined 611 to be putative ncRNAs, of which 474 ncRNAs are not part of previous literature, 33  Wurtzel et al. [15] performed a comparative study of L. monocytogenes and the non-pathogenic L. innocua using strandspecific cDNA sequencing. This resulted in genome-wide transcription start site maps and the identification of 183 ncRNAs. From the 183 reported ncRNAs, 100 were identified by our method, whereas half of the remaining ncRNAs were lacking expression.
Interestingly, there were a few examples where Wurtzel et al. [15] described a long candidate, which was covered by two or more candidates from our putative ncRNA set. These regions were discovered as several candidates by our method, since the expression pattern dropped down in between the candidates. The most noticeable example is anti1846 with a described length of 1371 nt, which overlaps with four of our candidates (216 nt, 141 nt, 23 nt and 227 nt).
In general, our method rather predicted longer ncRNAs which overlap with two or more previously described ncRNAs. For example, LhrC-1-LhrC-4 were reported earlier as four ncRNA candidates [15] and have been merged by our approach to a single putative ncRNA, which conforms to the first description of this ncRNA by Christiansen et al. [38] in 2006. But even though the complete region was covered, the expression was not continuously on the same level.
Nevertheless, we missed a few of the ncRNA candidates described in previous studies (see Fig. 1). This can be attributed to the differences in the experimental setup: we used a different sequencing technology, different organisms at different expression time points, and a different subsequent in silico scoring. From the previously reported ncRNA candidates that were actually covered by reads, only a small fraction was rejected by our filtering steps.
From the 611 ncRNAs detected by our method, 474 were identified here by RNA-seq for the first time. From these, 33 candidates were already known from Rfam and 441 have, as far as we know, never been reported before.
In our set of predicted ncRNAs we found some highly interesting (long-)antisense ncRNAs (lasRNAs) with up to 5,400 nt, which were induced under intracellular conditions. Most of the lasRNAs described below were validated by qRT-PCR (Fig. 2).

Internalins are very likely controlled by our detected lasRNAs
Two long ncRNA candidates were detected as antisense transcripts of two genes coding for the proteins lmo0333 and lmo1136 (see Tab. 3, and Fig. 3A,B). Both proteins lmo0333 and lmo1136 are similar to internalin proteins (according to NCBI annotation) and contain an LRR-LPXTG-motif.
Internalins (Inls) are a large group of proteins containing leucine-rich-repeats (LRR) and are known to play an important role in host-pathogen-interactions. The bacterial cell-surface anchored proteins InlA and InlB are required for cell-, tissueand organ-specific invasion of L. monocytogenes. InlA engages the cell-junction protein E-Cadherin as its cellular receptor and InlB uses the hepatocyte growth factor receptor (HGFR, c-Met) for internalization [39]. Another cell-surface bound internalin is InlK, which binds to the Major Vault Protein (MVP) and thereby shields the bacterium from autophagy [40]. The secreted internalin InlC interacts directly with IKKa, a subunit of the IkB kinase complex, which is critical for the phosphorylation of IkB and activation of NF-kB, to suppress the inflammatory response [41].
The regulation of internalins is relevant to understand the virulence of L. monocytogenes. Previous studies showed that the master virulence regulatory protein PrfA regulates several internalins, e.g., inlAB and inlC [42]. Moreover, transcriptional regulation by the alternative sigma factor SigB was reported for several internalins, e.g., inlA, inlB, lmo0263 and lmo0610 [43,44].
Using RNA-seq, we showed in this study that internalins encoded by lmo0333 (inII) and lmo1136 are subject of antisense transcriptional regulation by long non-coding antisense RNAs (lasRNAs) las0333 and las1136. Lmo1136 is presumed to encode an internalin [20] which has not been studied so far. InlI was recently described and investigated by Sabet et al. [45] in the mouse infection model, but a knockout mutant for the inlI gene  Table 3. Selected candidates.  did not exhibit any difference in virulence when compared to the wild type [45]. The long antisense transcripts of internalin have a length of 163 nt and 493 nt (Tab. 3). According to the expression levels those transcripts are presumably even longer, 1214 nt and 1617 nt respectively (see Fig. 3A,B). For lmo0333 another antisense transcript of only 15 nt length, which is covered by 121 uniquely mapped reads, was detected. The number of reads mapping to the proposed lasRNAs varies between 28 and 121 reads. Interestingly transcription seems to be specific for Listeria grown in macrophages (intracellular) as for the extracellular condition no expression was observed.
We quantified the extra-and intracellular expression levels by qRT-PCR for all five selected lasRNAs (see Fig. 2A) and their corresponding mRNA transcripts (see Fig. 2B). All lasRNAs were up-regulated in the intracellular compartment. mRNA targets of las0333 and las0936 were repressed, whereas transcription of lmo0996, lmo1136 and lmo2677 was induced under intracellular conditions. This might indicate that these newly identified lasRNAs are involved in depression of target mRNAs (lmo0333 and lmo0936) and stabilization of mRNA transcripts (lmo0996, lmo1136 and lmo2677), what has been also reported for other lasRNA transcripts, e.g. from Prochlorococcus [46].
Novel long antisense transcript (2,400 nt-3,800 nt) An extremely long antisense transcript, spanning at least 2,400 nt (see Tab. 3), was observed antisense to lmo0537 and lmo0538. Gene lmo0537 codes for an amidohydrolase including a dimerization domain. The transcript contains four asRNA candidate loci, which might be also a single long antisense transcript. It is likely that the detected lasRNA influences its antisense genes lmo0538 and lmo0537. However, this cannot be proven yet. Nevertheless, a rough inverse transcript pattern of the proteins and their expected antisense regulators is observable (see Fig. 3C). The antisense transcript of lmo0537 seems to be specific for intracellular conditions.
The antisense transcript of lmo0581 was mainly observed for the intracellular condition (see Fig. 3D). Even though the expression is very low in some parts, it is spanning lmo0581 (1161 nt) completely. Gene lmo0581 itself is transcribed under extracellular and intracellular growth conditions.
The second putative lasRNA spans three genes (see Fig. 3E): it was detected antisense to lmo0936 (similar to nitroflavin-reductase), lmo0935 (SpoU, rRNA methylase) and lmo0934 (uncharacterized Fe-S protein, energy production and conversion). One striking feature of this candidate is its length of 2,500 nt. Even though the transcription rate is very low in some regions, an antisense transcript of this length is remarkable. Whereas the transcription of the lasRNA is specific for intracellular grown Listeria, the genes are covered with reads originating from both growth conditions. The third methyltransferase having putative asRNA transcripts is lmo0996 (see Fig. 3F), which is similar to methylated DNAprotein-cystein methyltransferase. This asRNA is an intergenic transcript and appears to be transcribed continuously with its syntenic genes lmo0997 (clpE, ATP-dependent protease) and lmo0995 (predicted acetyltransferase). The intergenic transcription is observed only in intracellularly grown Listeria. This indicates that the reads cannot be simply attributed to extended 59 or 39 UTRs, but are rather a putative specific intracellular ncRNA. We observed only very low transcription for the protein gene lmo0996, neither for extracellular nor for intracellular conditions. All of the above mentioned antisense transcripts are short (91-221 nt) and covered by 16-1750 reads (see Tab. 3). The read pattern of the ncRNA candidates is rather unsteady. A direct influence of the lasRNAs to the methylases can be only hypothesized.
The kdpEDABC operon is controlled by an extremely long non-coding antisense RNA Among the newly detected lasRNAs we have identified a very long antisense RNA of about 5,400 nt which completely covers the region from lmo2677 up to lmo2680 and partially the gene kdpB (see Tab. 3 and Fig. 4). This lasRNA is strongly activated during the intracellular growth phase of the pathogen and was confirmed by qRT-PCR (see Fig. 2) analysis. Previously Wurtzel et al. [15] described an asRNA for lmo2678, which is transcribed under exponential growth at 37uC and is controlled by SigB. The gene lmo2678 encodes the response regulator (KdpE) of a two component system (TCS) together with a cognate histidine kinase (KdpD) encoded by lmo2679 [47]. Under high-osmolarity conditions the KdpED TCS regulates the adjacent kdpABC operon which is responsible for high-affinity potassium uptake as previously reported for Escherichia coli [48]. Several different reports described KdpED to be involved in intracellular survival of pathogenic bacteria, for example Staphylococcus aureus, enterohaemorrhagic E. coli, Salmonella typhimurium and Yersinia pestis [49]. In L. monocytogenes, however, it does not seem to play an important role in virulence [50]. This is supported by the observation that the entire locus lmo2677-lmo2681(kdpB) is down-regulated by massive antisense transcription. This suggests that alternative uptake systems exist to ensure potassium uptake. Such systems have been already reported for B. subtilis [51]. It is, however, unclear why this long asRNAs is necessary to block the kdpED TCS and kdpABC operon under intracellular conditions. Why is a short asRNA, as described by Wurtzel et al. [15], produced during extracellular growth conditions, not sufficient to stop transcription of lmo2678 and the kdpED TCS/kdpABC operon? We speculate that these asRNAs do not only stringently regulate transcription in cis, but also in trans.
Recently Mellin et al. [16] reported that in the presence of vitamin B 12 , the corresponding riboswitch induces transcriptional termination. This causes an antisense RNA aspocR to be transcribed as a short transcript. In the absence of vitamin B 12 , aspocR is transcribed as a long antisense RNA, inhibiting pocR expression [16]. A similar non-classical function could be also assumed for the kdpEDABC interfering las2677/las2678 RNAs.
Furthermore, there seems to be a correlation between the asRNA read pattern and the start and stop sites of the operon genes. For example, for lmo2678/kdpE there is an increase and decrease correlating with the start and stop positions of this (see POMAGO); cyan blue -potential new ncRNAs predicted by RNAz; pink -annotated ncRNAs. A better resolution of the figure can be found in the supplement. doi:10.1371/journal.pone.0108639.g003 Fig. 4). It is tempting to speculate whether this lasRNA is originating from lmo2676 or not. In case it is originating from lmo2676, the transcript might resemble an excludon. Interestingly another ncRNA candidate was detected directly downstream to lmo2677 (see Fig. 4). Nevertheless, this seems to be a separate transcript and not an extended 39UTR, since there is an obvious decrease of reads at the end of lmo2677. This ,300 nt RNA antisense to the 59part of lmo2676 is stronger expressed under extracellular conditions.
To confirm our newly identified asRNAs in another L. monocytogenes serotype 1/2a strain, we have preformed additional RNA-seq experiments (unpublished RNA-seq data, online supplementary material) with the commonly used L. monocytogenes strain 10403S grown under extra-and intracellular conditions. Comparison of presence/absence of the las0333, las0936, las0996, las1136 and las2677 showed a similar occurrence of these asRNAs between L. monocytogenes strain 10403S and EGD-e. This implicates a conserved expression mechanism for L. monocytogenes serotype 1/2a strains for these selected asRNA candidates.
In addition, we have also tested the transcription regulator mutant of L. monocytogenes EGD-e DprfA under the same experimental conditions described above. Our RNA-seq analysis (unpublished RNA-seq data, online supplementary material) showed that all above mentioned asRNAs were independently controlled by the master virulence regulator PrfA. Furthermore, these new RNA-seq data warrant detailed investigation in future.

Conclusion
We systematically used the semiconductor sequencing technology for RNA-seq to identify ncRNAs and determine the difference of extra-and intracellular growth conditions. We reported bacterial antisense transcripts with a size up to 5,400 nt. It would be interesting to use our pipeline to examine whether similar transcripts can be observed in other bacteria. Further work has to be done to fully understand the functional role of these long noncoding antisense RNAs in bacterial physiology. Particularly in the case of the kdpABCD operon, the regulation of K + by long noncoding antisense RNAs now deserves further attention.  20. Dark purpledetected ncRNA candidates; lightgreen -NCBI annotation; darkgreen -BacProt annotation; black -reads of the extracellular library; dark blue -reads of the intracellular library; violet -locally stable secondary structure (analyzed with RNALfold); blue -conserved region among other L. monocytogenes serotypes (analyzed with POMAGO); cyan blue -potential new ncRNAs predicted by RNAz; pink -annotated ncRNAs; teal green -ncRNA candidates of previous studies. doi:10.1371/journal.pone.0108639.g004