Antisense RNA asPcrL regulates expression of photosynthesis genes in Rhodobacter sphaeroides by promoting RNase III-dependent turn-over of puf mRNA

ABSTRACT

Anoxygenic photosynthesis is an important pathway for Rhodobacter sphaeroides to produce ATP under oxygen-limiting conditions. The expression of its photosynthesis genes is tightly regulated at transcriptional and post-transcriptional levels in response to light and oxygen signals, to avoid photooxidative stress by the simultaneous presence of pigments, light and oxygen. The puf operon encodes pigment-binding proteins of the light-harvesting complex I (genes pufB and pufA), of the reaction centre (genes pufL and pufM), a scaffold protein (gene pufX) and includes the gene for sRNA PcrX. Segmental differences in the stability of the pufBALMX-pcrX mRNA contribute to the stoichiometry of LHI to RC complexes. With asPcrL we identified the third sRNA and the first antisense RNA that is involved in balancing photosynthesis gene expression in R. sphaeroides. asPcrL influences the stability of the pufBALMX-pcrX mRNA but not of the pufBA mRNA and consequently the stoichiometry of photosynthetic complexes. By base pairing to the pufL region asPcrL promotes RNase III-dependent degradation of the pufBALMX-prcX mRNA. Since asPcrL is activated by the same protein regulators as the puf operon including PcrX it is part of an incoherent feed-forward loop that fine-tunes photosynthesis gene expression.

Graphical Abstract

Introduction

Rhodobacter sphaeroides is a facultative phototrophic bacterium. In the presence of oxygen, Rhodobacter performs aerobic respiration to generate ATP, while in the absence of oxygen, it can switch to anoxygenic photosynthesis to maintain ATP production. To prevent the cost-intensive formation of the photosynthetic complexes and the generation of reactive oxygen species under aerobic conditions, the expression of photosynthesis genes is tightly regulated, mainly by the availability of light and oxygen [1,2]. Under high oxygen conditions, the repressor PpsR can bind consensus sequences upstream of photosynthesis genes or gene clusters to prevent their expression [3]. When the oxygen tension decreases, the anti-repressor AppA can bind to PpsR and antagonizes its function [4–6]. Besides PpsR and AppA, the PrrA/PrrB two-component system is involved in the activation of the photosynthesis genes when oxygen levels drop [7–9]. This system is known to be the main redox regulator within R. sphaeroides and can affect the regulation of about 25% of all genes [10]. Under low oxygen tension, the membrane-bound histidine kinase/phosphatase PrrB is auto-phosphorylated and can subsequently transfer the phosphate to the response regulator PrrA [11]. The phosphorylated PrrA can then activate the transcription of several genes including photosynthesis genes like the puc and puf genes by binding directly to the DNA [8,9,12,13]. In addition, the FnrL protein is required for growth under photosynthetic conditions. FnrL activates the expression of the puc and puf genes under anaerobic conditions [14].

The photosynthetic apparatus of R. sphaeroides consists of different multimeric transmembrane protein complexes: the peripheral light harvesting (LH) antenna complex II, the LHI complex, which is associated with the reaction centre (RC) and a cytochrome bc₁ complex. The antenna complexes are collecting the light energy and pass the excitation to the reaction centre. The ratio of the LHII complex to the core complex of RC and LHI is strongly influenced by the availability of light whereas the ratio of LHI and RC is fixed and light-independent [15].

The proteins required for the assembly of the photosynthetic apparatus are encoded in different operons. R. sphaeroides possesses two polycistronic puc operons puc1BA and puc2BA which encode the proteins of the LHII complex. The genes for the pigment-binding proteins of the RC and the LHI complex are encoded in the polycistronic puf operon, adjacent to the genes required for bacteriochlorophyll and carotenoid syntheses in the photosynthesis gene cluster. The puc1BA operon localizes near the puf operon, while the puc2BA operon is located at a different part of the chromosome [16,17]. The polycistronic puf mRNA encodes six proteins of the photosynthetic apparatus and gives raise to one small non-coding RNA, PcrX, downstream of the coding region (Fig. 1) [18,19]. PufBA encode the pigment-binding proteins of the LHI complex, pufLM of the RC, pufX a scaffold protein, which is important for the correct assembly of LHI and RC. The PufQ protein regulates the porphyrin efflux and pufK encodes a putative small protein with 20 amino acids that was suggested to be essential for the translation of pufB [18,20–22].

Figure 1.

Model for the processing of the primary pufBALMX-PcrX transcript

Model for degradation of the puf mRNA based on studies in R. capsulatus and R. sphaeroides, which show the same puf processing pattern and half-lives. Decay of the polycistronic pufQBALMX-pcrX transcript is initiated by RNase E-mediated cleavage within pufQ. The 2.7 kb pufBALMX-pcrX mRNA segment is protected by three secondary structures at the 5´end against RNase E attack and by the terminator structure at the 3´end against exoribonucleases. RNase E was shown to initiate degradation in the 5´ region of pufL and is responsible for maturation of the sRNA PcrX that consequently base pairs to pufX and contributes to destabilization of the pufBALMX transcript. The higher stability of the pufBA mRNA is due to an intercistronic hairpin loop between pufA and pufL that protects against 3´to 5´exoribonucleases.

Previous studies revealed that post-transcriptional regulation of the polycistronic puf mRNA plays an important role in the regulated formation of photosynthetic complexes (see Fig. 1 for an overview). Initially, processing of the puf mRNA was intensely studied in Rhodobacter capsulatus [23]. R. capsulatus is closely related to R. sphaeroides, the organization of the puf genes and important mRNA secondary structures are highly conserved and the same puf mRNA species are detected on Northern blots. In R. capsulatus cleavage of RNase E within the pufQ mRNA initiates processing of the unstable pufQBALMX primary transcript into the 2.7 kb long pufBALMX fragment with a half-life of about 8–12 min [19,23] (Fig. 1). Secondary structures in the 5´ UTR of pufB protect against further RNase E attack at the monophosphorylated 5´ end [24]. An additional RNase E cleavage in the 5´ region of pufL initiates processing of the pufBALMX fragment into the pufBA fragment [25]. Stem-loops at the 3´ end of pufA prevent the 3´-5´ exonucleolytic degradation of this 0.5 kb long fragment [26], which has a half-life of about 33 min. This post-transcriptional regulation is required for the differential expression of the puf genes, supports thereby the ratio of LHI to RC which is 15:1 and is required for optimal growth under phototrophic conditions [27].

To date two small non-coding RNAs could be identified, which play a crucial role in the regulation of photosynthesis genes in R. sphaeroides. PcrZ regulates the expression of some photosynthesis genes as part of a mixed incoherent feed forward loop. Its expression depends on the two-component system PrrA/PrrB and on FnrL upon the decrease of oxygen tension as is also the case for the photosynthesis genes. It negatively affects the levels of its confirmed target genes puc2A and bchN that is required for bacteriochlorophyll synthesis. Thereby it counteracts the induction of the photosynthesis genes via PrrAB and FnrL [28,29].

The second sRNA, PcrX, is derived from the 3´-UTR of the puf-operon in R. sphaeroides by RNase E-dependent processing [30] (Fig. 1). The levels of PcrX are therefore controlled by the activity of the puf promoter and are regulated by the PrrAB and FnrL proteins. PcrX affects the stability of the pufBALMX fragment by binding to its target, the pufX coding region of this polycistronic mRNA. Since it does not affect stability of the pufBA mRNA segment, it consequently influences the ratio of the photosynthetic complexes [19].

Here, we analysed the function of a third sRNA, which affects the formation of the photosynthetic apparatus. Unlike the other known sRNAs acting on photosynthesis gene expression, PcrZ and PcrX, this cis-encoded RNA is antisense to the 5´ part of pufL. Plasmid-derived overexpression of the antisense RNA asPcrL affects the amounts of photosynthetic complexes by decreasing the stability of the pufBALMX mRNA of R. sphaeroides.

Materials and methods

Growth conditions

All Rhodobacter strains listed in table S1 were cultivated in malate minimal medium [31] at 32°C. Cells were incubated under continuous shaking at 140 rpm in the dark either in baffled flasks, which were filled up to 25% for aerobic growth or in normal flasks, which were filled up to 80% for microaerobic growth. For phototrophic growth, bacteria were cultivated in sealed Meplat bottles filled to the top and illuminated with 60 W*m⁻² with white light.

For pigment analysis and growth curve experiments, precultures were grown under aerobic conditions over night, diluted 1:1 with fresh media and transferred to Erlenmeyer flasks. After 120 min of microaerobic growth, the cultures were diluted to an OD₆₆₀ = 0.2 and cultivated either under microaerobic or phototrophic conditions until they reached an OD₆₆₀ = 0.5.

For half-life experiments pre-cultures, which were cultivated under microaerobic conditions overnight to an OD₆₆₀ = 0.5–0.7 were diluted to an OD₆₆₀ = 0.2 and cultivated under the tested conditions until they reached an OD₆₆₀ = 0.4. Samples for RNA isolation were harvested directly before the addition of rifampicin (t = 0 min) or at the indicated time points after the addition of rifampicin (final concentration 220 µg ml⁻¹, Serva Electrophoresis GmbH). After harvesting, all samples were immediately cooled on ice.

Strain construction

All oligonucleotides used for plasmid constructions are listed in table S2.

For the deletion of the puf-operon sequence of R. sphaeroides, the suicide plasmid pPHU281Δpuf::Gm was transferred into the wild type by diparental conjugation and screening for the insertion of the gentamycin resistance cassette into the chromosome via homologous recombination was performed based on resistance. To generate this plasmid a 548 bp fragment was amplified using the primers puf-up-for and puf-up-rev (table S1). This fragment was inserted into the vector pPHU281 with XbaI and BamHI resulting in plasmid pPHU281-puf-up. A second fragment was amplified using primers puf-down-for and puf-down-rev resulting in a 619 bp fragment. This fragment was inserted into the vector pPHU281-puf-up with XhoI and BamHI giving rise to the vector pPHU281-puf-up/down. The gentamycin resistant cassette obtained from pWKR209-CII was inserted with BamHI between the up- and down-fragment of the vector pPHU281-puf-up/down. The resulting plasmid pPHU281Δpuf::Gm was transferred into the E. coli strain S17-1 for following di-parental conjugation. To confirm insertion of the gentamycin cassette into the chromosome of R. sphaeroides by double crossover, conjugants were selected on malate minimal medium agar plates containing 10 μg/ml gentamycin. Insertion of the gentamycin cassette into the puf-operon deleted 2359 bp out of 2856 bp of its coding region.

In order to overexpress asPcrL, its sequence was cloned into a low copy number plasmid under the control of the 16S rRNA promoter. A 237 nt asPcrL fragment was amplified by PCR using primers asPcrL-for and asPcrL-rev. The terminator sequence of the phage protein 32 was inserted into the plasmid pRK4352 (table S2) with KpnI and EcoRI resulting in plasmid pRK4352Ω. The asPcrL fragment was inserted into pRK4352Ω with BamHI and XbaI. The fragment containing the 16S rRNA promoter, asPcrL and the terminator were inserted into plasmid pBBR1-MCS2 with HindIII and EcoRI resulting in plasmid pBBR:asPcrL.

For in vivo ß-galactosidase assays the complete coding sequence of pufL and 200 bp upstream were amplified using primers pufL-for and pufL-rev. The resulting 1022 bp fragment was inserted into pPHU235-4352 with XhoI and HindIII.

RNA isolation and Northern blot analysis

Cells for RNA isolation were harvested, cooled down immediately on ice and centrifuged for 10 min at 10.000 rpm at 4°C. For Northern blot analysis total RNA was isolated using the hot phenol method [32]. For the detection of the antisense RNA, a 10% urea-polyacrylamide gels was used for Northern blot analysis as described [33]. For detection of mRNA, 1% formaldehyde agarose gels were used for Northern blot analysis as previously described [5]. Oligonucleotides which were used for hybridization are listed in table S1. T4 polynucleotide kinase was used for the end-labelling reaction with [γ-³²P]-ATP (30 Ci/µl) (Hartmann Analytics). For detection of the antisense RNA and puf-mRNA, PCR-products (primer listed in table S1) were labelled with [α-³²-P]-CTP (Hartmann Analytics) using the Prime-a-gene-Labelling System (Promega) as described in the manufacturer`s manual. Membranes were exposed on phosphoimaging screens (Bio-Rad) and signals were quantified using 1-D-Quantity One software (Bio-Rad). For quantification, the signal intensity of the whole band was analysed.

Pigment analysis

For pigment analysis, Rhodobacter strains were cultivated over night under aerobic conditions. On the next day the cultures were diluted to an OD₆₆₀ = 0.2 and simultaneously shifted to microaerobic conditions until they reached an OD₆₆₀ = 0.6–0.8. For whole-cell absorbance spectra, 1 ml of the culture was transferred into a cuvette and recorded on a Specord 50 spectral photometer. The amount of bacteriochlorophyll was measured as previously described [34].

RNA-isolation and qRT-PCR

For qRT-PCR total RNA was isolated using peqGOLD TriFast™ (VWR) as described by the manufacturer. Afterwards, the RNA was treated with TURBO DNA-free™ Kit (Ambion) as described in the manufacturer´s manual to remove DNA contaminations. RNA was precipitated overnight in 2.5 volume of ice-cold ethanol and 0.1 volume 3 M sodium acetate pH = 4.5. The absence of DNA contaminations was tested by PCR using oligonucleotides targeting gloB (RSP_0799) as described [35].

For qRT-PCR the Brilliant III Ultra-Fast SYBR® Green QPCR Master Mix was used for reverse transcription and PCR as described in the manufacturer´s manual. For relative quantification of mRNAs in each of three independent experiments qRT-PCR samples containing 40 ng of total RNA were run in a BioRad RealTime System C1000™ Thermal Cycler. Oligonucleotides used for amplification are listed in table S1. Crossing points (C_p) with a fluorescent threshold of 980 were visualized with the BioRad CFX Manager 3.1 software. The expression of the target mRNAs in the strain of interest was calculated relative to the respective control strain containing and normalized to the housekeeping gene rpoZ [36].

Co-immunoprecipitation

For co-immunoprecipitation the cells were cultivated under microaerobic conditions until an OD₆₆₀ = 0.4 was reached. 200 ml of culture were harvested by centrifugation for 10 min at 10.000 rpm and 4°C. Pellets were resuspended in 2 ml lysis buffer (20 mM Tris pH 7.5; 150 mM KCl; 1 mM MgCl₂; 1 mM DTT). The co-immunoprecipitation and western blot analysis were carried out as previously described [37].

Protein purification

Expression of the recombinant RNase III from R. capsulatus in E. coli and lysis of the cells were performed as described [38]. Cell lysate was loaded onto a HisTrap column (HisTrap High performance 5 ml, GE Healthcare) and was washed with sonication buffer with a final concentration of 20 mM imidazole 40. The protein was eluted with a gradient of 0–600 mM imidazole in sonication buffer. The samples containing RNase III were pooled and loaded on a gel filtration column (Superdex™200, GE Healthcare) and were eluted with the sonication buffer. Protein was stored at −20°C.

For the purification of RNase E from E. coli the same procedure was used as described above for the RNase III. Buffers were used as described [39].

In vitro transcription

For run-off transcription, the DNA sequence for the required RNA was cloned into the pDrive cloning vector (Qiagen) behind the T7 promoter. The primer sequences are listed in table S1. The plasmids were linearized with HindIII. T7-polymerase (NEB) and [α-³²P]-UTP (30µCi/µl) (Hartmann Analytics) were used to generate internally labelled in vitro transcripts.

RNase cleavage assay

100 fmol of the radio-labelled in vitro transcripts were incubated with 1 pmol of nonradio-labelled in vitro transcript in 1x RNase III cleavage buffer (30 mM Tris-HCl pH 7.5, 10 mM Mg₂Cl, 130 mM, 5% Glycerol) or 1 x RNase E cleavage buffer (20 mM Tris-HCl pH 7.5, 500 mM NaCl, 10 mM MgCl₂, 0,5 mM EDTA, 10 mM DTT, 5% Glycerol) for 5 min at 35°C. Afterwards, RNase III or RNase E was added and incubated for 2 and 5 minutes, respectively. The reaction was stopped by adding 7 μl formaldehyde urea loading dye. Samples were incubated for 10 min at 65°C and loaded on a 7 M Urea gel containing 6% polyacrylamide. After 4 hours at 400 V, the gel was transferred to a vacuum gel dryer for 45 min at 85°C. Signals were visualized by a phosphoimager (Personal Phospho Imager FX, Biorad).

Electrophoretic mobility shift assay

100 fmol of the radio-labelled in vitro transcripts were denatured at 95°C for 2 min and cooled down on ice for additional 2 min. 2 µl structural buffer (25 mM MgCl₂, 300 mM KCl) were added and the in vitro transcript was renatured for 5 min at 32°C before 1 µl of the Hfq-protein solution with different concentrations (0, 1, 10, 50, 100, 500, 1000, 5000 nM) were added and incubated for 30 min at 32°C. For competition EMSA the different in vitro transcripts were de- and renatured separately as described above before 1 mM purified Hfq was added. After the incubation the reactions were loaded on a native 6% polyacrylamide gel and run for 3 h at 150 V and 4°C. Afterwards the gel was transferred to a vacuum gel dryer for 45 min at 85°C. Signals were visualized by a phosphoimager (Personal Phospho Imager FX, Biorad).

ß-Galactosidase assay

For measuring the ß-galactosidase activity, strains carrying the pufL-reporter plasmid were incubated in biological triplicates at 32°C under microaerobic conditions until they reached an OD₆₆₀ = 0,6. Cells were harvested and the ß-galactosidase assay was performed as previously described [40].

RACE-PCR

For RACE-PCR 7 µg total RNA were used. 5´-RACE-PCR reactions were performed as described in [42] and 3´-RACE-PCR was done as described in [28]. The used primers are listed in table S1.

Results

Expression of the antisense RNA asPcrL is oxygen-dependent

RNA-sequencing of total RNA from R. sphaeroides cultures grown under microaerobic conditions [30] (GEO accession number GSE104278) revealed the existence of an antisense sRNA, asPcrL (antisense photosynthesis control RNA L), that is partially complementary to the pufL mRNA. The antisense RNA spans the 5´ part of pufL and the intercistronic region between pufL and pufA (Fig. 2). Another short RNA was detected antisense to the pufK gene (Fig. 1). RNAseq data indicate that asPcrL is much less abundant than the other puf mRNAs under microaerobic conditions in exponential phase. dRNAseq [41] and 5´ RACE mapped the 5´end of asPcrL at genome position 1,982,339 which is identical to the end that was observed from sequencing of total RNA and is located in 80 nt distance from the start of pufL (red arrow on plus strand in Fig. 2).

Figure 2.

Organization of the puf-operon. Screen shot of the Integrated Genome Browser showing the total reads of the RNAseq corresponding to the puf-operon on the minus strand and to the antisense RNA on the plus strand. Red arrows indicate transcriptional start sites, the blue arrow marks the 3´-end of asPcrL and the green arrow the transcriptional termination site of the puf-operon. The scales for the plus and minus strand are different for better visualization of asPcrL

Promoter sequences are not very conserved in R. sphaeroides. Most promoters, but not all, have a TTG around position −35 and an A at position −11/-10 [41]. In respect to the mapped 5´end a TCG is located at position −34 and an A at position −10. The RNAseq data do not hint to a clear 3´end of asPcrL. 3´RACE mapped a single 3´end of asPcrL at genome position 1,982,525, in 106 nt distance to the start of pufL, within the intercistronic region between pufA and pufL (blue arrow in Fig. 2).

Northern blot analysis confirmed the existence of antisense RNA asPcrL with defined length of about 190 nt and revealed its oxygen-dependent expression (Fig. 3A). Fig. 3B shows the asPcrL sequence and structure as predicted by RNAfold [42]. A GC-rich secondary structure that is not followed by U residues is found at the 3´end. While asPcrL was hardly detectable under aerobic conditions (bacteriochlorophyll content 0.03 µM/OD₆₆₀), its level was strongly induced after a shift to microaerobic conditions that promoted formation of photosynthetic complexes (bacteriochlorophyll content 1.2 µM/OD₆₆₀ at OD 0.6) and reached a maximum at 60 min after the transition. A shift from microaerobic to phototrophic conditions (bacteriochlorophyll content 1.8 µM/OD₆₆₀) also transiently increased the asPcrL levels. Like the expression of the puf genes, expression of asPcrL depends on different protein regulators. In a strain lacking the response regulator PrrA, which acts as a main activator of photosynthesis genes under low oxygen, the level of the antisense RNA was under the detection limit. asPcrL levels were also strongly reduced in strains lacking FnrL or AppA (Fig. 3C). Thus, asPcrL expression is controlled by the same protein regulators as the promoter of the puf genes, which is located within the bchZ coding region (red arrow on minus strand in Fig. 2).

Figure 3.

Expression levels of asPcrL. (A) An aerobic over night culture of R. sphaeroides wild type 2.4.1 strain was diluted to OD₆₆₀ = 0.2 and incubated under aerobic conditions (A) until an OD₆₆₀ = 0.4 and then shifted to microaerobic conditions (MA) and after 180 min of growth to phototrophic conditions (PT) for a further 180 min. Cells were harvested at the indicated time points. Northern blot analysis was performed with 10 µg of total RNA for each sample. A specific PCR-probe was used to detect the antisense RNA. 5S rRNA served as loading control. As a size-marker we used the sRNA UpsM with a known full-length size of 206 nt and a main processing product of 130 nt. (B) Predicted secondary structure of the 186 nt long asPcrL based on the results of 5´- and 3´-RACE-PCR by using RNAfold. A GC-rich secondary structure that is not followed by U residues is found at the 3´end (C) R. sphaeroides wild type and strains lacking the regulatory proteins PrrA, FnrL, or AppA were cultivated under microaerobic conditions. Northern blot analysis was performed using 10 µg total RNA. An asPcrL-specific PCR-probe was used to detect the antisense RNA. 5S rRNA was used as loading control

asPcrL influences formation of photosynthetic complexes

Since deletion of asPcrL or mutations in the gene would also affect the sequence and possibly structure of the coding sense mRNA pufL we overexpressed asPcrL to investigate its biological function. The corresponding DNA sequence was cloned behind the 16S rRNA promoter on a low copy number plasmid and introduced into the wild type strain R. sphaeroides 2.4.1 via conjugation (pBBR:asPcrL). As a control, the empty vector (pBBR:16S) was also transferred into the wild type strain. Earlier results showed that the 16S rRNA promoter slightly responds to the oxygen level. The strongest activity could be observed under phototrophic conditions while the activity is reduced in aerobic (by 20%) and microaerobic (by 40%) growth conditions (Müller K., PhD thesis University Giessen, 2016). The overexpression strain showed increased levels of asPcrL under microaerobic (log₂-fold change 2.6) and phototrophic (log₂-fold change 3.9) conditions (Fig. S1) and a lighter colour under microaerobic and phototrophic conditions compared to the control strain.

To quantify the pigment content, we determined the bacteriochlorophyll levels from cultures shifted from aerobic to microaerobic conditions. The quantification of bacteriochlorophylls as shown in Fig. 4A confirms the reduced pigmentation in the overexpression strain compared to the control at 6 h after the shift to low oxygen. Since all the pigments in the cell are bound to the pigment-protein complexes LHII, LHI and RC, we analysed pigment-protein complexes by their absorption spectra in the near infrared (Fig. 4B). The LHII complex has characteristic absorbance at 800 nm and 855 nm, while maximal absorbance of LHI occurs at 875 nm. Absorbances of the RC complex at 803 nm and 870 nm are covered by the absorbances of the much more abundant LH complexes [27]. Due to the overlap of LHI and LHII absorbance, the shoulder at 875 nm represents the presence of the LHI complex. The spectral analysis revealed that the LHI-specific absorbance was reduced in the overexpression strain, while LHII absorbance was almost identical in both strains.

Figure 4.

asPcrL affects the level of photosynthetic complexes. (A) The amount of bacteriochlorophyll is reduced in the 16S:asPcrL overexpression strain in comparison to the control strain containing pBBR:16S in the wild type background. R. sphaeroides strains containing pBBR:16S or pBBR:asPcrL were grown over night under aerobic conditions. At the next morning strains were diluted to an OD₆₆₀ = 0.2 and simultaneously shifted to microaerobic conditions until they reached an OD₆₆₀ = 0.6. For the measurement 1 ml of each culture was extracted by acetone/methanol (7:2). The bacteriochlorophyll specific absorbance was measured at 770 nm and normalized to the OD₆₆₀ and the extinction coefficient of 76 mM⁻¹ cm⁻¹ at 770 nm was applied [43]. The average from biological triplicates with technical duplicates is shown, the standard deviation is indicated. T‐test was applied for statistical analysis and revealed p = 0.033 for the changed bacteriochlorophyll amounts between the overexpression strain and the empty vector control. (B) R. sphaeroides strains containing pBBR:16S or pBBR:asPcrL were grown over night under aerobic conditions. At the next morning, strains were diluted to an OD₆₆₀ = 0.2 and simultaneously shifted to microaerobic conditions until they reached an OD₆₆₀ = 0.8. For whole-cell spectral analysis the absorbance was measured between 600 and 950 nm. Peaks at 800 and 850 nm are specific to LHII, LHI absorbs at 870 nm. (C) Growth curves of the R. strains containing pBBR:16S or pBBR:asPcrL. Cultures were shifted from aerobically grown pre-culture to microaerobic conditions. The optical density was measured at 660 nm. The results show the mean of three independent experiments. Standard deviation is too low to be visible in the blot. (D) Growth curves of the R. strains containing pBBR:16S or pBBR:asPcrL. Cultures were shifted from aerobically grown pre-cultures to phototrophic conditions. The optical density was measured at 660 nm. The results show the mean of three independent experiments. Standard deviation is too low to be visible in the plot

Furthermore, we investigated the growth behaviour of the overexpression strain in comparison to the empty vector control. When cells were shifted from aerobic to microaerobic conditions, the growth of the two strains was identical during all growth phases (Fig. 4C). In contrast to this, the growth of the overexpression strain was impeded when cells were shifted from aerobic to phototrophic conditions (Fig. 4D). The overexpression strain showed an extended lag phase compared to the control. In the exponential phase the growth rate of the overexpression strain was reduced up to 13% and the doubling time is prolonged from 4.9 to 5.6 hours. After 32 hours both strains reached the same maximal OD₆₆₀ but the control strain entered the stationary phase earlier than the overexpression strain.

asPcrL influences the amount of puf-mRNA

Since asPcrL affects the amounts of photosynthetic complexes, we hypothesized that it affects the levels of puf mRNAs. We applied real-time RT-PCR to quantify the amount of pufL mRNA in the overexpression and in the control strain. Indeed, the amount of pufL mRNA was reduced by a log₂-fold change of around 1.2 in the overexpressing strain (Fig. S2A). To elucidate if these changes are due to a reduced promoter activity, we monitored activity of a transcriptional reporter fusion to the native puf-promoter. We could detect a small, but not significant difference of the puf promoter activity between the two strains (Fig. S2B).

To test whether asPcrL affects mRNA stability, we determined the half-lives of the puf-mRNA by adding rifampicin to the microaerobically grown cultures. Indeed, the half-life of the 2.7 kb long pufBALMX fragment is influenced by the antisense RNA: in comparison to the empty vector control the half-life was reduced from 12 to 7 minutes in the strain overexpressing asPcrL (Fig. 5). The half-life of the pufBA fragment was unaffected by asPcrL overexpression in agreement with our model that asPcrL only acts on degradation of the pufLMX-pcrX region. The same observation was made for overexpression of PcrX that targets the pufX coding region [19].

Figure 5.

asPcrL influences the puf-mRNA half-life. Half-lives were determined for the puf-mRNA. The R. sphaeroides strains containing pBBR:16S or pBBR:asPcrL were cultivated under microaerobic conditions, rifampicin was added and samples were taken at the indicated time points. Northern blot analysis was performed with 10 µg of total RNA for each sample. PufBA-specific PCR fragments were used as probe to visualize the pufBALMX and pufBA fragment. The signal intensities of the mRNAs were normalized to the 14S rRNA and plotted semi-logarithmically in percentages against the time in minutes. The calculated half-life of the pufBALMX and the pufBA fragment are shown underneath a representative Northern blot. The half-lives below the Northern blot are the average of 3 half-life determinations from 3 independent experiments and the standard deviation for the half-lives is shown. The plot shows the average for each time point from the three experiments and the standard deviation at each time point

To test whether the stability of the antisense RNA depends on the presence of its target pufL mRNA, we overexpressed asPcrL not only in the wild type but also in a Δpuf background. In absence of puf-mRNA, the stability of asPcrL rose from 7 min to over 20 min, indicating that both RNA species, asPcrL and pufL, are degraded together (Fig. 6 A).

Figure 6.

Stability of asPcrL is pufL- and Hfq-dependent. Cells were incubated under microaerobic conditions until they reached an OD₆₆₀ = 0.4, when rifampicin was added. Samples were harvested directly before the addition of rifampicin (0 min) and at the indicated time points. Northern blot analysis was performed with 10 µg of total RNA for each sample. We used a specific PCR-probe to detect asPcrL and 5S rRNA was used as loading control. The signal intensity of the antisense RNA was normalized to the 5S rRNA and plotted semi-logarithmically. The half-lives of the antisense RNA are shown underneath representative Northern blots. The half-lives below the Northern blots are the average of 3 half-life determinations from 3 independent experiments and the standard deviation for the half-lives is shown. The plots show the average for each time point from the three experiments and the standard deviation at each time point. (A) The overexpression plasmid was introduced into either the wild type or the Δpuf strain. After hybridization and exposure to the phosphoimaging screens (Bio-Rad) the samples of the wild type and the Δpuf mutant containing the pBBR:asPcrL plasmid were scanned individually (dashed line) to better visualize the bands for both strains. (B) Wild type and the Δhfq strain were incubated under microaerobic conditions. We used a specific PCR fragment as probe to detect the chromosome-derived asPcrL and the 5S rRNA was used as loading control

Hfq destabilizes puf-mRNA and asPcrL

The Hfq protein functions as an RNA chaperon and plays important roles in RNA-based regulation. The interaction between trans-encoded sRNAs and their targets often requires the assistance of Hfq due to their limited complementarity [44–47]. Hfq is also found together with RNase E in ribonucleoprotein complexes and promotes cleavage by RNase E [48–50]. Although complementarity between sense and anti-sense RNA is perfect, we tested a possible interaction between asPcrL and Hfq by co-immunoprecipitation (coIP) with a 3xFLAG-tagged Hfq protein. Total RNA and coIP RNA were used for semiquantitative qRT-PCR with asPcrL- and pufL-specific primers (Fig. 7A). The antisense RNA was enriched in the 3xFLAG fraction, strongly suggesting that there is a direct interaction between the antisense RNA and Hfq. A western blot confirming presence of Hfq in the coIP is shown in Fig. S3.

Figure 7.

Hfq directly interacts with asPcrL and influences the pufBALMX half-life. (A) Co-immunoprecipitation (coIP) of asPcrL and 3xFlag-tagged Hfq. Strains harbouring either the endogenous or a triple-Flag-tagged Hfq were cultivated under microaerobic conditions. Total RNA and coIP-RNA were used for semiquantitative qRT-PCR with asPcrL- or pufL-specific primers. PCR-products were loaded on a 10% polyacrylamide gel and 1 kb plus ladder (Thermo Scientific) was used as marker. (B) The amount of bacteriochlorophyll is not reduced upon 16S:asPcrL overexpression in comparison to the control strain containing pBBR:16S in the Δhfq background. For the measurement 1 ml of each culture was extracted by acetone/methanol (7:2). The bacteriochlorophyll specific absorbance was measured at 770 nm and normalized to the OD₆₆₀ and the extinction coefficient of 76 mM⁻¹ cm⁻¹ at 770 nm was applied [43]. The average from biological triplicates with technical duplicates is shown and the standard deviation is indicated. T‐test was applied for statistical analysis and revealed p = 0.20. (C) Wild type and the Δhfq strain were incubated under microaerobic conditions until they reached an OD₆₆₀ = 0.4. Rifampicin was added and cells were harvested on ice at the indicated time points. 10 μg of total RNA were used for Northern blot analyses. We used a specific PCR fragment as probe to detect only the 2.7 kb-long pufBALMX fragment and a specific pufBA probe to detect the pufBA fragment. The signal intensity of the mRNA was normalized to the 14S rRNA and plotted semi-logarithmically. The half-lives of the mRNAs are shown underneath the Northern blot. The half-lives below the Northern blot are the average of 3 half-life determinations from 3 independent experiments and the standard deviation for the half-lives is shown. The plot shows the average for each time point from the three experiments and the standard deviation at each time point

To validate the interaction between Hfq and asPcrL electrophoretic mobilty shift assays (EMSAs) were performed using purified Hfq and radioactively labelled asPcrL in vitro transcript. A shift of the in vitro transcript could be observed with increasing amounts of purified Hfq (Fig. S4A). While the unlabelled in vitro transcript of sinI from Sinorhizobium meliloti (no sequence similarity to asPcrL) could not reduce the interaction of asPcrL and Hfq, unlabelled asPcrL in vitro transcript led to a reduction of the interaction of radioactively labelled asPcrL in vitro transcript, supporting a specific interaction between asPcrL and Hfq (Fig. S4B). Furthermore, CoIP followed by qRT-PCR also unravelled the binding of pufL to Hfq (Fig. 7A).

To investigate if the influence of the antisense RNA on the puf-operon is Hfq-dependent, the overexpression plasmid and the empty vector control plasmid were introduced into the Δhfq strain. The phenotypic alteration, which was observed upon overexpression of asPcrL in the wild type background could not be detected in the Hfq mutant. The bacteriochlorophyll content was similar between control and the asPcrL overexpression strain in the hfq⁻ background (Fig. 7B). Furthermore, the reduction of the pufL-amount which was observed in the overexpression strain in the wild type background in comparison to the corresponding empty vector control could not be observed in the overexpression stain in the hfq deletion mutant. qRT-PCR experiments showed that the amount of pufL is elevated in the overexpression strain in the hfq deletion strain in comparison to the empty vector control (Fig. S2A). Northern blot analysis revealed that the half-life of asPcrL is prolonged from 7 min in the wild type to 12 min in the Hfq-deletion strain (Fig. 6B), the half-life of the pufBALMX mRNA is prolonged from 10 min in the wild type to 21 min in the Hfq-deletion strain (Fig. 7C).

asPcrL-mediated cleavage of pufL by RNase III

To investigate which RNase is involved in the processing of the puf-mRNA and the antisense RNA an in vivo reporter assay was applied. The coding sequence of pufL and 200 bp upstream sequence were translationally fused to lacZ under control of the 16S rRNA promoter on a pPHU235 plasmid backbone [51]. The plasmid was conjugated into the wild type and into different mutant strains. We could not detect altered lacZ activity when the plasmid was introduced into an RNase E mutant. In this strain the endogenous RNase E was replaced by a thermosensitive variant from E. coli [30]. Heat treatment for 20 min at 42°C did not lead to a significant change of ß-galactosidase activity in the wild type or the mutant strain (Fig. 8A). In contrast to this, the lack of RNase III (deletion and antibiotic cassette within the rnc gene; ⁵²) resulted in strongly increased ß-galactosidase activity, indicating stabilization of the fusion mRNA of pufL and lacZ (Fig. 8A).

Figure 8.

RNase III cleavage requires pufL mRNA and asPcrL. (A) ß-galactosidase activity based on a translational fusion in vivo assay. The 16S_pufL-pPHU plasmid which contains the complete coding sequence of pufL and its 5´-UTR under the control of the 16S rRNA promoter were conjugated into the wild type and different RNase mutants. The average from biological triplicates with technical duplicates is shown as well as standard deviation. Activity in the wild type was set to 100%. T‐test was applied for statistical analysis and revealed p = 0.001 for the changed ß-galactosidase activity between the wild type and the RNase III-mutant. (B) Quantification of the asPcrL and pufL RNA level by qRT-PCR in the wild type strain in comparison to the RNase III-mutant strain. R. sphaeroides strains were cultivated under microaerobic conditions. Specific asPcrL and pufL primers were used for qRT-PCR. RNA levels of the wild type were calculated relative to the RNase III mutant strain as log₂ ratios. The housekeeping gene rpoZ was used as reference gene. The average from biological triplicates with technical duplicates is shown together with the standard deviation. (C) In vitro RNase III degradation assay. 100 fmol of radio-labelled pufL transcript were incubated without or with 1 pmol of cold asPcrL and 3 nM RNase III. As a negative control RNase III was not added to the RNA (-). Samples were mixed with 7 µl of formamide-containing dye, incubated at 65°C for 10 min and analysed on a denaturing 10% polyacrylamide-7 M urea gel

A similar result was detected by using qRT-PCR with pufL- and asPcrL-specific primers. The qRT-PCR revealed higher amounts of asPcrL (log₂-fold change 0.7) and pufL (log₂-fold change 0.55) in the RNase III mutant in comparison to the wild type, which indicates an accumulation of the both RNA-species due to the lack of the degradation by RNase III (Fig. 8B). RNase III is a double-strand specific endoribonuclease that is also involved in degradation of sRNA-mRNA hybrids [53].

To confirm these results, in vitro assays with the purified RNase III from R. capsulatus and radioactively labelled RNA in vitro transcripts were performed. The pufL mRNA was only degraded by RNase III, when incubated together with non-radioactive in vitro transcript of asPcrL (Fig. 8C). The shift we observe in presence of asPcrL indicates that asPcrL binds to pufL even under denaturing conditions. By using different lengths of the antisense RNA, we could show, that only the 3´ part (bases 78–186) of asPcrL promotes cleavage by RNase III, while the presence of the 5´ part (bases 1–77) had no influence on the degradation in vitro (Fig. S5A). This indicates that the RNase III cleavage site is within 39 to 31 nt upstream of the pufL start codon.

Determination of the RNase III cleavage position by primer extension on total RNA isolated either from microaerobic grown wild type or Δrnc cultures did not detect stable 5´-ends. We assume that this is due to a rapid attack of the monophosphorylated 5´ end by RNase E after initial RNase III cleavage. A previous RNAseq study revealed multiple RNase E cleavage sites within the pufL sequence downstream of the putative RNase III cleavage region [30]. In vitro degradation assays with purified RNase E from E. coli showed a rapid degradation of the radioactively labelled pufL in vitro transcript also in the absence of asPcrL (Fig. S5B).

To identify the RNase III cleavage site, we applied 5´-RACE with isolated total RNA derived from the wild type and the RNase E mutant strain grown at 32°C and after heat inactivation for 20 min at 42°C. The nested PCR-products were applied on a 10% PAA-gel. For each sample more than one signal could be detected on the gel after the ethidium bromide staining, reflecting the rapid degradation and the emerging processing products of the pufL-fragment. The longest PCR-product of about 500 nt could be detected in the heat-inactivated RNase E-mutant samples, whereas the longest PCR-products which could be detected in the other samples were about 300 nt long (Fig S7). The sequences obtained by 5´-RACE were aligned against the pufL-sequence by using BLASTn (NCBI). A 5´-end at genome position 1.982.483 was detected in the heat-inactivated RNase E-mutant samples (position green arrow Fig. S8), which could not be detected in the other samples. This 5´-end enriched in the mutant, could also be detected in a previous RNAseq approach [30]. 5´-ends which are enriched in the RNase E-mutant strain after heat-inactivation are due to RNase E-independent cleavage [30]. Such an endonucleolytic cleavage leads to an RNA-molecule which harbours a monophosphate on its 5´-end, which is favoured by the RNase E [54]. Active RNase E will rapidly degrade this RNA intermediate, while the RNA will accumulate when the RNase E is inactivated. The results from 5´ RACE support our assumption that the 5´end generated by RNase III cleavage is rapidly removed by RNase E.

Discussion

asPcrL is the first identified antisense RNA, which is involved in the regulation of photosynthesis gene expression in Rhodobacter. Our data demonstrate that the antisense RNA is transcribed from an own promoter from the opposite strand of the pufL gene. Expression of asPcrL is affected by the protein regulators PrrA, AppA, and FnrL. Based on 170 PrrA binding sites from microarray and genomic data, a weak consensus sequence was postulated: two recognition blocks GCGNC and/or GNCGC separated by 2–9 nt [55]. The sequence GTCGC is present around the −35 region of the asPcrL 5´end and the sequence GCGAA more upstream, separated by 7 nt from the second block. AppA is an antirepressor to PpsR that directly binds to the DNA. Based on a CHIP analysis for PpsR binding sites Bruscella et al. [56] deduced the consensus sequence TGTCN-10-GACA and Imam et al. [57] the sequence TGT-12-ACA. Such a sequence is not present in the vicinity of the asPcrL 5´end, but even within the binding sites shown in [56], there is a big sequence variation, making prediction of PpsR binding sites questionable. Furthermore, binding of PpsR upstream of the prrA promoter was observed [57], so that PpsR can affect gene expression also indirectly, via direct binding to the prrA promoter region. A quite defined binding sequence was described for FnrL: TTGA-6-TCAA [55]. Such a sequence is not present around the 5´end of asPcrL. However, we observed an effect of FnrL on the expression of the appA gene (unpublished data), so that the FnrL effect may also be indirect. Most importantly, PrrA and FnrL directly and AppA indirectly through PpsR induce the puf mRNA levels, which according to our model results in increased turnover and consequently lower levels of asPcrL.

The 3´end as determined by RACE maps to a very strong secondary structure (to the basis of the 3´stem, ΔG: −26.4 kcal/mol) that is not followed by U residues and not predicted as a terminator structure. It is well possible that transcription proceeds further and that the secondary structure protects against 3´-5´exonucleolytic decay that initiates at a downstream cleavage site. This mechanism is also responsible for generation of the 3´end of the pufBA RNA segment [27,58]. Several RNase E cleavage sites were mapped to the intercistronic region between pufA and pufL [30].

It has been shown in the past that cis-encoded sRNAs have a huge potential to regulate gene expression of their target mRNAs by influencing mRNA stability, blocking translation and promoting degradation [59]. Here we demonstrate that overexpression of asPcrL affects the relative amounts of photosynthetic complexes in the wild type of R. sphaeroides but not in a mutant lacking Hfq. The deletion of Hfq leads to an increase in the half-lives of the pufBALMX fragment and of asPcrL. Co-immunoprecipitation supported a direct interaction of Hfq and asPcrL. asPcrL harbours a region of 7 adjacent A and U residues and two adjacent ARN sequences. Such motifs of sRNAs are known to interact with Hfq [60], but the exact Hfq binding sites of asPcrL cannot be predicted from sequence.

In R. sphaeroides various trans-acting sRNAs like UpsM, Pos19, and CcsR1-4 require Hfq to properly affect their target mRNA [61–63] and lack of Hfq significantly reduced the amount of photosynthetic complexes [37]. Earlier results indicated that Hfq is dispensable for the effect of the sRNAs PcrZ and PcrX on photosynthesis gene expression [19,28]. Even though most of the cis-encoded sRNAs act Hfq-independently, there are a few examples for Hfq dependence. For instance, Ross et al. [64] reported an important role of Hfq in the translational regulation of the transposase mRNA (RNA-IN) by the antisense RNA-OUT. Furthermore, the antisense RNA GadY, which regulates the expression of acid response genes in E. coli is Hfq-dependent. Co-immunoprecipitation indicated a direct interaction between Hfq and GadY and in a strain lacking Hfq GadY could not be detected. Nevertheless, the authors could not reveal whether Hfq is required for the base pairing of GadY and its complementary mRNA gadX or whether Hfq plays a role in the binding to unknown targets in trans [65]. In R. sphaeroides some antisense RNAs were found to co-immunoprecipitate together with Hfq [37]. However, the function of these antisense RNAs and the exact mechanisms by which Hfq assists asPcrL function remain unclear.

asPcrL destabilizes the pufBALMX transcript by binding to the 5´-region of pufL. The regulated processing of the puf transcript ensures a ratio of about 15:1between LHI and RC complexes [27]. The processing events within the puf-operon were primarily analysed in R. capsulatus, a near relative of R. sphaeroides [66,67]. A rate-limiting RNase E processing site is found in the 5´-part of pufL [68] and a cleavage at this site and further 3´-5´ exonucleolytic degradation generate the stable pufBA fragment which is protected by a highly stable stem-loop in the intercistronic region between pufA and pufL. In R. sphaeroides asPcrL is complementary to the pufL region around an RNase E cleavage site in a similar position (Fig. S8). Recently, we demonstrated that RNase E has a strong impact on the transcriptome of R. sphaeroides [30]. RNase E is crucial for maturation and/or processing of different sRNAs like UpsM, SorX, PcrZ and PcrX [19,28,63,69]. Moreover, reduced RNase E activity strongly impacts phototrophic growth of R. sphaeroides but not chemotrophic growth [30]. However, the in vivo reporter assay strongly suggested that RNase E cleavage is not catalysing the processing step that is influenced by asPcrL. In vivo and in vitro experiments demonstrate destabilization of the pufL fragment in presence of asPcrL and RNase III.

A role of RNase III in processing/degradation of sense-antisense RNA duplexes is well established, including duplexes of sRNAs and mRNAs [70,71]. For instance in Bacillus subtilis Type I toxin antitoxin systems the antisense sRNAs Sr5 and SR6 base pair with the bsrE and yonT mRNAs and promote degradation by RNase III [72,73]. In E. coli the antisense RNA CopA post-transcriptionally regulates the replication of the R1 plasmid by binding to the leader region of its target mRNA CopT. The duplex of the two RNAs is then cleaved by RNase III [74]. In Legionella pneumophila a cis-encoded sRNA base-pairs to the 5´ UTR of the hfq mRNA in an Hfq-dependent manner likely leading to cleavage of the duplex by RNase III. This antisense mechanism regulates Hfq expression in a life cycle-dependent manner [75]. A global study analysing the double-stranded transcriptome of E. coli identified a total of 316 potentially functional asRNAs that are primarily encoded opposite to the 5´ends of mRNAs. These asRNAs included several short RNAs and were only detected in absence of RNase III [76]. Similar results were obtained by RNAseq approaches in Staphylococcus aureus, implying the importance of antisense transcription and RNase III-dependent processing of sense-antisense duplexes in both, gram-negative and gram-positive bacteria.

Cleavages by endoribonucleases like RNase III generate monophosphorylated 5´ ends, which allow attachment of RNase E and subsequent cleavages in overall 5´ to 3´ direction [77]. TIER-Seq revealed two prominent AU-rich RNase E cleavage sites within the 5´ part of pufL and within the intercistronic region between pufA and pufL [30] indicating a contribution of RNase E in the processing of the puf-mRNA in R. sphaeroides.

BlastN search indicates that PcrX is conserved within other Rhodobacteraceae [19] while PcrZ is exclusively present in R. sphaeroides. The presence of asPcrL can only be revealed by RNAseq data, which are not available for most relatives of R. sphaeroides. RNAseq for R. capsulatus SB1003 (GSE134200) did not reveal such an antisense RNA, implying that different strategies are used within the Rhodobacteraceae to fine-tune the expression of the photosynthesis genes.

Expression of asPcrL is activated by the response regulator PrrA, the protein activator FnrL and the antirepressor protein AppA. The same activators are required for the expression of different photosynthesis genes and operons (e.g. puc, puf, bchFBN) including the sRNAs PcrZ and PcrX. While the orphan gene PcrZ is transcribed by its own promoter, PcrX is co-transcribed together with the puf-operon and processed from the 3´ UTR by RNase E-mediated cleavage. Overexpression of PcrZ leads to a down-regulation of a subset of genes, which are involved in the formation of the photosynthetic complexes. puc2A and bchN mRNAs were identified as direct targets of PcrZ [29]. The second identified sRNA affecting photosynthesis gene expression, PcrX reduces the half-life of the pufBALMX fragment by binding to its target mRNA pufX. Both trans-encoded sRNAs are part of mixed incoherent feed-forward loops [78,79] that counteract the strong induction of photosynthesis genes under low oxygen tension and thereby guarantee their balanced expression [19,28]. Incoherent feed-forward loops contain a transcription factor that by acting on different downstream factors at the same time leads to activation and repression of the same genes. A mixed feed-forward loop contains as well protein as RNA regulators. Our data demonstrate that asPcrL is also part of a mixed incoherent feed-forward loop which together with the protein regulators and PcrX controls the expression of the puf operon (Fig. 9).**** [,80]******

Figure 9.

Scheme of the regulatory loop including asPcrL and affecting the expression of the puf-operon. The response regulator PrrA and the FnrL protein activate expression of the puf-operon including PcrX and also activate expression of asPcrL. Both small RNAs are parts of a mixed incoherent feed-forward loop and negatively affect the expression of the puf-operon by destabilizing the pufBALMX mRNA

Funding Statement

This work was supported by the Deutsche Forschungsgemeinschaft [Kl563/28].

Disclosure statement

We declare no conflict of interest.

Supplementary material

Supplemental data for this article can be accessed here.