Overlapping embedded genes, such as htgA/yaaW, are assumed to be rare in prokaryotes. In Escherichia coli O157:H7, gfp fusions of both promoter regions revealed activity and transcription start sites could be determined for both genes. Both htgA and yaaW were inactivated strand specifically by introducing a stop codon. Both mutants exhibited differential phenotypes in biofilm formation and metabolite levels in a nontargeted analysis, suggesting that both are functional despite YaaW but not HtgA could be expressed. While yaaW is distributed all over the Gammaproteobacteria, an overlapping htgA-like sequence is restricted to the Escherichia-Klebsiella clade. Full-length htgA is only present in Escherichia and Shigella, and htgA showed evidence for purifying selection. Thus, htgA is an interesting case of a lineage-specific, nonessential and young orphan gene.
Overlapping embedded genes are considered to be rare in prokaryotes, and only very few have been described (e.g. Silby & Levy, 2008; Tunca et al., 2009; Cheregi et al., 2012). However, the length distribution of overlapping open reading frames in bacteria suggest more of such genes exist (Mir et al., 2012).
The gene htgA (high-temperature growth, Dean & James, 1991) is located upstream of dnaK (James et al., 1993), completely embedded antisense in the hypothetical gene yaaW (Fig. 1) and only found in Escherichia and Shigella (Delaye et al., 2008). Despite its name, a heat shock induction of htgA could not be confirmed (Nonaka et al., 2006), and thus, its annotation has been questioned (see Supporting Information, Data S1 for an extended introduction).
We present functional information on both htgA and yaaW, based on promoter-fusions, strand-specific single-gene knockouts, 5′-RACE and protein expression. Furthermore, the phylogeny of htgA is reexamined.
Materials and methods
Three-hundred base pairs (bp) upstream of htgA (Z0012), yaaW (Z0011) and yaaI (Z0013) were PCR-amplified (for primers, see Table S1) using E. coli O157:H7 EDL933 (EHEC, NC_002655, CIP 106327). The amplicons were cloned upstream gfp in pProbe-NT (Miller et al., 2000). EHECs with plasmids (verified by sequencing) were grown in LB (Sambrook & Russel 2001) with 25 μg mL−1 kanamycin. GFP was measured for 1 s of cultures grown in the dark to OD600 nm = 1, washed once with PBS, and using 200 μL of 1 : 5 and 1 : 10 dilutions (Victor3, Perkin-Elmer). Empty vector control values were measured, and fluorescence was normalized to OD600 nm. The mean of four wells was calculated from three independent experiments.
Transcriptional start sites
5′-RACE was performed using the 5′RACE System for Rapid Amplification of cDNA Ends Version 2.0 (Invitrogen) according to the manufacturer. For htgA, the pProbe-NT plasmid with an inserted putative promoter region was used, and transformed cells were grown in LB. For yaaW, the bacteria were grown in 1 : 10 diluted LB medium at pH6 with 200 mg L−1 Na-nitrite (R. Landstorfer, S. Simon, S. Schober, D. Keim, S. Scherer & K. Neuhaus, unpublished data) to induce yaaW. After gel electrophoresis, the most intense bands were purified (Invisorb® Fragment CleanUp, STRATEC, Berlin), used as template for subsequent amplification and sequenced using nested primers (LGC Genomics, Berlin).
For ΔhtgA and ΔyaaW, two DNA-fragments were amplified, up and downstream of the site to be mutated, enclosing the mutated site. Both amplicons are used in the subsequent reaction, using the two nonoverlapping primers, to recreate the gene with the mutation. The final product was cloned into pMRS101 (Sarker & Cornelis, 1997). The high-copy ori was removed, and the plasmids transferred to E. coli CC118λpir (Manoil & Beckwith, 1985). After verification by sequencing, they were transferred to E. coli SM10λpir (Miller & Mekalanos, 1988) for mating. EDL933 NalR (spontaneous mutation) and the respective SM10λpir plus the modified pMRS101 were mixed and plated on LB-agar (24 h, 30 °C). Cells were resuspended again and plated on LB-agar with 30 μg mL−1 streptomycin and 20 μg mL−1 nalidic acid. Correct plasmid integration after a first cross-over was checked by PCR. Second cross-over events, resulting in plasmid loss, either restore wild type or create the mutation. Thus, bacteria were grown without selection to OD600 nm = 0.8 and plated on LB-agar without NaCl plus 10% sucrose for sacB-counter selection. Desired mutants were identified using PCR.
Biofilm experiments were conducted according to Domka et al. (2007). A culture grown in M9 minimal medium (Sambrook & Russel 2001) was diluted to OD600 nm = 0.05. Flat-bottom wells of a microtiter plate (Greiner Bio One, Germany) were filled with 100 μL and incubated 24 or 48 h without shaking at 30 °C or 37 °C. OD600 nm was measured (Victor3). The planktonic cells were removed, and each well was carefully washed with water. Staining was achieved using 135 μL 0.1% crystal violet (20 min, RT). After washing thrice with water and air drying, the stain was solubilized in 95% ethanol, transferred to a new plate and the absorbance at 600 nm was measured (Victor3). The mean was calculated (10 wells, three biological replicates) after subtracting zero controls (medium only).
Amplicons of htgA and yaaW were cloned into pBAD/Myc-His C (Invitrogen). EHEC with plasmids (sequenced for verification) were grown in LB with 100 μg mL−1 ampicillin and induced with 0.2% arabinose. Proteins were purified according to QIAexpress® Ni-NTA Fast-Start kit under denaturing conditions (Qiagen). For this, the bacteria were sonicated in the provided lysis buffer. For SDS-PAGE (15%), Laemmli-buffer was added, and the sample denatured for 5 min at 95 °C. PageRuler Protein Ladder (Fermentas) was used as marker. After electrophoresis, the proteins were electroblotted (20 min, 120 mA) to an activated PVDF membrane (Amersham). Subsequently, the membrane was blocked, incubated with mouse-anti-human c-myc-antibodies (BD Biosciences), washed, incubated with alkaline phosphatase anti-mouse chimera antibodies (Dianova, Hamburg), washed again, equilibrated and incubated in buffer supplemented with BCIP/NBT.
Metabolites were profiled using Ion cyclotron resonance Fourier transform Mass spectrometry (ICR-FT/MS) on a Bruker solariX with a 12-T magnet (Bruker Daltonics, Bremen). Three biological replicate cultures of wild type, ΔhtgA, and ΔyaaW were grown shaking in 1 : 2-diluted LB to OD600 nm = 1. Cultures were vacuum filtered using HVLP filters (0.45 μm; Millipore). The bacteria and the filter were flash frozen in liquid nitrogen and extracted with 50% methanol using a FastPrep (MP Biomedicals) with zirconia beads (0.1 mm, and a few beads of 2 mm diameter) three times for 45 s at 6.5 m s−1. Samples were centrifuged, filtered (0.22 μm), diluted 1 : 20 with 70% MeOH, and infused at 120 μL h−1. ICR-FT/MS was externally calibrated on clusters of arginine (10 ppm in 70% MeOH). A time domain transient of 2 megawords was used, and 300 scans were accumulated for one spectrum. Spectra were internally calibrated with an error of ≤ 0.1 ppm, exported with a signal-to-noise ratio of 3, and aligned within a 1 ppm window. Putative metabolites were annotated using MassTRIX (Wägele et al., 2012). Only masses found in all replicates were considered and analyzed in Genedata Expressionist for MS 7.6 (Genedata, Martinsried).
Promoters were searched by bprom (Softberry Inc., New York) and terminators by webgester db (Mitra et al., 2011). Microarray data were accessed from the Gene Expression Database (genexpdb, http://genexpdb.ou.edu/index.php, see Table 1).
Table 1. Differential expression of htgA or yaaW detected in microarray experiments
Data are from genexpdb. Upregulation is shown with positive numbers, downregulation with negative numbers; WT is wild type; n.v., no value given. References are listed under the respective accession number. Note that yaaW and htgA are treated as synonyms in genexpdb, despite differential expression values.
Sequences were searched with blastp or tblastn (NCBI, http://blast.ncbi.nlm.nih.gov/Blast.cgi, default parameters) using YaaW (Z0011) as query (Table S2). The evolutionary history of all species was inferred using the software package mega5 with a concatemer of 16s rRNA gene, atpD, adk, gyrB, purA, and recA by Minimum Evolution using p-distance. The bootstrap consensus was inferred from 1000 replicates (Tamura et al., 2011). For some strains, not all sequences were available, and thus close relatives were used as surrogate, for example, some genes of Comamonas testosteroni CNB-2 were used for the yaaW-bearing strain ATCC 11996. The presence of htgA was detected using pairwise blastp alignments with htgA (Z0012) as query (starting from the first GTG).
htgA/yaaW sequences were examined for their nonsynonymous over synonymous rate ratio ω as described (Sabath et al., 2008; Sabath & Graur, 2010) including correction for multiple testing according to Benjamini & Hochberg (1995), after omitting alignment gaps (Tamura et al., 2011).
Results and discussion
Transcription of htgA and yaaW
5′-RACE determined the major 5′-end of the + 1 transcription start of htgA to be 135 bp upstream. However, minor sites might be present, since Missiakas et al. (1993) found a site 82 bp upstream; others were predicted 98 (BProm) or 114 bp (Tutukina et al., 2007) upstream of the CTG-start codon of htgA.
The upstream region of htgA was successfully tested for promoter activity using a promoterless gfp reporter. No terminator could be detected directly downstream of htgA but was detected downstream of dnaK (Fig. 1). Recently, strand-specific transcriptome sequencing showed that htgA is transcribed, albeit weakly, at some nonlaboratory growth conditions only (R. Landstorfer, S. Simon, S. Schober, D. Keim, S. Scherer & K. Neuhaus, unpublished data).
The 5′-RACE major transcription start site of yaaW is 32 bp upstream of yaaI, but a minor site, 107 bp upstream of yaaW, was also detected. Testing both putative promoter regions, we found yaaI to show promoter activity, but yaaW was similar to the empty vector control (Fig. 1b). A terminator was predicted by webgester downstream of yaaH or, in antisense at the same position, downstream of mog. This suggests that yaaW is most likely organized as operon yaaIWH in EHEC and transcribed from the yaaI-promoter and terminated downstream of yaaH.
Interestingly, data from genexpdb indicate that htgA and yaaW are expressed differentially in E. coli strains under certain experimental conditions (see Table 1), clearly prohibiting htgA synonymizing with yaaW, which has been performed in some databases.
Translation of htgA and yaaW
HtgA and YaaW were expressed in EDL933 using a plasmid that generates concomitant myc and His-tag fusions. Proteins were prepurified using the his-tag and detected on Western blots using the myc-tag. YaaW (30 kDa) was detectable, but no band for HtgA was found (Fig. 2), which is in accordance with Narra et al. (2008). Thus, the protein might be unstable and difficult to discover. Missiakas et al. (1993) presented a 21-kDa gene product by 35S-labeling, which is a more sensitive approach.
Phenotypes of ΔhtgA and ΔyaaW mutants
Previous work always used a double knockout mutant. We created strand-specific deletion mutants for the first time, in which only htgA or yaaW was interrupted (Fig. 3). The annotated htgA-start codon is CTG, which is quite rare for bacteria. The next GTG is more likely to be the start codon. Counting from there, htgA has 525 bp (or 174 amino acids); our htgA-knock out terminates either product. By introducing a single-point mutation to create a stop in one frame, we minimized the disturbance of the other, as the mutations are synonymous in the latter (Tunca et al., 2009). For the first time, it was possible to distinguish effects of ΔhtgA from ΔyaaW.
Both mutants showed no difference in their growth compared with wild type at 37 °C or after temperature shift from 30 °C to 45 °C (Fig. 4a). As no heat shock phenotype of ΔhtgA could be confirmed (as found before, Nonaka et al., 2006), htgA should no longer be annotated as heat shock gene. In minimal medium, biofilm formation of ΔhtgA or ΔyaaW was reliably increased when incubated for 48 h at 37 °C (Fig. 4b). This is in accordance with Domka et al. (2007), who found a threefold increase in biofilm formation for E. coli K12 in a htgA/yaaW double mutant. We speculate that the higher increase compared with our experiments might be due to additive effects of both genes in the double mutant compared with each single one. We therefore suggest to rename htgA to mbiA (modifier of biofilm).
Metabotypes of ΔhtgA and ΔyaaW mutants
As no difference in growth could be found, we measured the metabotypes. Metabolite changes could still be detectable even though they may not manifest in growth (Raamsdonk et al., 2001). ΔhtgA, ΔyaaW, and wild type were subjected to nontargeted metabolomics using ICR-FT/MS. Indeed, twenty-two different metabolites (putatively annotated, see Table S3) between the strains were found significantly changed (P ≤ 0.01). When comparing ΔhtgA to wild type, we found four differences, comparing ΔyaaW to wild type, 14 differences, and comparing ΔhtgA to ΔyaaW, four differences. In both mutants, all metabolites were decreased compared with wild type. The differential changes provide evidence that both reading frames are functional. The majority of changes were associated with fatty or amino acid metabolism. Neither htgA nor yaaW appear to be directly involved in the cellular metabolism and any functional explanation is as yet highly speculative.
Is htgA an RNA only?
Instead of being protein coding, htgA could produce a regulatory (metabolite-binding) or antisense RNA. This is considered unlikely as several metabolites are affected. More importantly, antisense-RNA regulation is achieved by base pairing of longer stretches between the antisense and target RNA (Lasa et al., 2012), but we engineered single-base substitutions, which should not cause any detectable differences in pairing.
Taxonomical distribution and evolution of htgA
yaaW homologs are present in a variety of bacteria (Fig. 5, Table S2), but a complete htgA-frame is present only in Escherichia and Shigella. A minority of Salmonella contains yaaW, but htgA is always a pseudogene in those species and interestingly in each case disrupted at the same positions.
Evolution of yaaW is restricted when it contains an overlapping htgA-frame (Delaye et al., 2008). The rate between synonymous and nonsynonymous mutations in a gene is used to infer selection. However, embedded genes influence each other, invalidating models used for nonoverlapping genes. Sabath et al. (2008) designed a model to estimate the nonsynonymous over synonymous substitution rate of overlapping genes to infer selection, comparing two scenarios: The first makes no assumptions on any selection intensity, the second assumes ‘no selection’ for the overlap, here htgA. In strains in which htgA was interrupted, indeed no selection was found. However, the estimation of selection intensities is limited in case of low sequence diversity, which is the case for yaaW (max. 2.6% on amino acid level). htgA is encoded in frame-2 in relation to yaaW, which provides the least flexibility for amino acid changes of both (Rogozin et al., 2002). This may partly explain the comparatively low degree of divergence. Despite these limitations, htgA is expected to be under (purifying) selection, and hence functional, in at least 24 strains of Escherichia and Shigella (Table S4).
We suggest that htgA is a young orphan (taxonomically restricted gene), as full-length htgA is restricted to Escherichia and Shigella, originating probably before Citrobacter or Klebsiella have separated. Orphans seem to be responsible for lineage-specific adaptations and most of these are assumed to be evolutionary ‘young’ genes, showing higher divergence rates, lower expression rates and encode shorter proteins compared to older genes (Tautz & Domazet-Loso, 2011). Despite that such genes most likely have no essential function and, therefore, may be prone to be lost again (e.g. in Salmonella), htgA should be added once again to the genome annotation of E. coli as an interesting case of an overlapping gene which emerged recently.
This study was funded by the DFG (SCHE316/3-1, KE740/13-1). We would like to thank Luke Tyler for assisting with the language. The authors declare that they have no conflict of interests.