A haem cofactor is required for redox and light signalling by the AppA protein of Rhodobacter sphaeroides


*E-mail Gabriele.Klug@mikro.bio.uni-giessen.de; Tel. (+49) 641 99 355 42; Fax (+49) 641 99 355 49.


The AppA protein of Rhodobacter sphaeroides is unique in its ability to sense and transmit redox signals as well as light signals. By functioning as antagonist to the PpsR transcriptional repressor, it regulates the expression of photosynthesis genes in response to these environmental stimuli. Here we show binding of the cofactor haem to a domain in the C-terminal part of AppA and redox activity of bound haem. This is supported by the findings that: (i) the C-terminal domain of AppA (AppAΔN) binds to haemin agarose, (ii) AppAΔN isolated from Escherichia coli shows absorbance at 411 nm and absorbances at 424 nm and 556 nm after reduction with dithionite and (iii) AppAΔN can be reconstituted with haem in vitro. Expression of AppA variants in R. sphaeroides reveals that the haem binding domain is important for normal expression levels of photosynthesis genes and for normal light regulation in the presence of oxygen. The haem cofactor affects the interaction of the C-terminal part of AppA to PpsR but also its interaction to the N-terminal light sensing AppA-BLUF domain. Based on this we present a model for the transmission of light and redox signals by AppA.


Rhodobacter sphaeroides is a facultatively photosynthetic bacterium that synthesizes pigment protein complexes in response to redox and light signals. At high oxygen tension, R. sphaeroides performs aerobic respiration and does not form photosynthetic complexes. If oxygen tension drops and the bacteria face the risk of low energy yield by aerobic respiration, the formation of photosynthetic complexes is induced, even in the dark. This process involves the syntheses of pigments, pigment binding proteins and the formation of an intracytoplasmic membrane system that organizes the photosynthetic complexes. Many proteins are involved in the redox regulation of photosynthesis genes in Rhodobacter species (Reviewed in Gregor and Klug, 1999; Zeilstra-Ryalls and Kaplan, 2004). The two-component system RegB/RegA in Rhodobacter capsulatus (Sganga and Bauer, 1992; Mosley et al., 1994) and PrrB/PrrA in R. sphaeroides (Eraso and Kaplan, 1994; 1995; Phillips-Jones and Hunter, 1994), respectively, is a central redox regulator. Oxidation of the sensor kinase RegB/PrrB results in the formation of an intracellular disulphide bond and converts the active dimer to an inactive tetramer (Swem et al., 2003). After autophosphorylation, RegB/PrrB transfers the phosphoryl group to the response regulator RegA/PrrA, which activates the transcription of most photosynthesis genes, but also of genes for nitrogen and carbon fixation and genes for components of anaerobic respiratory chains (Joshi and Tabita, 1996; Qian and Tabita, 1996; Gibson et al., 2002). Furthermore, evidence was provided that the sensor kinase PrrB senses the respiratory electron flow through the cbb3 oxidase (Oh and Kaplan, 2001; Oh et al., 2004). A recent study revealed that the binding of ubiquinone affects the autophosphorylation of RegB, thus connecting the redox state of the electron transport chain to RegB/RegA signalling (Swem et al., 2006).

While the RegB/RegA or PrrB/PrrA two-component system is a global redox regulator in Rhodobacter, the CrtJ (R. capsulatus) or PpsR (R. sphaeroides) protein, respectively, is a repressor specific for photosynthesis genes at high oxygen tension (Penfold and Pemberton, 1994; Ponnampalam et al., 1995). CrtJ undergoes a dithiol-disulphide switch when oxygen tension increases and binds its DNA targets with higher affinity (Ponnampalam and Bauer, 1997; Masuda et al., 2002). In R. sphaeroides, PpsR repression is antagonized by the AppA protein (Gomelsky and Kaplan, 1997); at low oxygen tension, AppA binds to PpsR and releases it from the DNA (Masuda and Bauer, 2002).

Besides these two important regulatory systems, other proteins like FnrL, PpaA or thioredoxins are implicated in the redox control of photosynthesis genes (Zeilstra-Ryalls et al., 1997; Zeilstra-Ryalls and Kaplan, 1998; Pasternak et al., 1999; Gomelsky et al., 2003; Li et al., 2003).

Although oxygen is the major factor that determines the formation of photosynthetic complexes, the expression of photosynthesis genes is also affected by light. In the absence of oxygen, light qualities absorbed by the photosynthetic apparatus of Rhodobacter activate the expression of photosynthesis genes. The light signal is transmitted via photosynthetic electron transport and components of the respiratory chain and finally affects transcription through the RegA/RegB, PrrA/PrrB two-component systems (Happ et al., 2005). As a consequence, higher amounts of photosynthetic complexes are synthesized in the presence of light and the absence of oxygen to allow an increased production of ATP.

At intermediate oxygen concentrations, blue light represses photosynthesis genes in R. sphaeroides (Shimada et al., 1992), presumably to avoid the production of singlet oxygen. This light response is transmitted by the PpsR/AppA system and is missing in R. capsulatus (Braatsch et al., 2002; 2004). The N-terminal BLUF domain [sensor of blue-light using FAD (Gomelsky and Klug, 2002)] (Fig. 1) of AppA was identified as a new type of blue light photoreceptor. AppA is the only known protein that simultaneously transmits light and redox signals. The C-terminal domain of AppA (Fig. 1) is sufficient for transmitting the redox signal to PpsR, but it cannot transmit the blue light signal (Han et al., 2004). The AppA-BLUF domain can transmit the light signal only in combination with the C-terminal domain, even if the two domains are not covalently linked (Han et al., 2004). Several investigations have addressed the photocycle of the AppA-BLUF domain and the structural changes within the BLUF domain during illumination are considered very small (Kraft et al., 2003; Gauden et al., 2005; Masuda et al., 2005a). It is not known how light signals are transmitted to the C-terminal part of AppA, and how light or redox-dependent changes in the C-terminal part affect the binding of AppA to PpsR.

Figure 1.

Schematic presentation of AppA and its domains. Two Pfam domains, a BLUF domain and a vitamin B12 binding domain, were predicted by SMART algorithm (http://smart.embl-heidelberg.de) in the AppA protein. The stars above the scheme indicate the sites of mutation in the strains selected after random mutagenesis. The ruler above the scheme indicates the number of amino acid.

In order to elucidate signal sensing and transmission by AppA, we performed random mutagenesis of a region of the C-terminal domain of AppA which was shown to be essential for transmission of the redox signal (Han et al., 2004). Mutations that impaired redox signalling or light signalling by AppA mapped to a putative vitamin B12 binding domain. Our data reveal that this domain does indeed bind haem and establish a role of haem binding in the function of AppA as PpsR antagonist and light regulator.


The C-terminal domain of AppA (AppAΔN) encompasses a haem binding domain

Our previous work provided evidence for an interaction of the C-terminal domain of AppA to PpsR on one hand and to the N-terminal light sensing BLUF domain on the other hand. The 54 C-terminal amino acids of AppA were not required for redox regulation (Han et al., 2004 and data not shown). In order to identify amino acids involved in light and/or redox signal transmission we randomly mutagenized the C-terminal domain of AppA (amino acids 135–378), excluding the 54 amino acids not required for signal sensing or transmission. We selected R. sphaeroides clones expressing AppA variants due to a lighter colony colour under low oxygen tension, indicative for altered redox regulation. All AppA variants we sequenced had 2–3 amino acids exchanged as indicated in Fig. 1 (domain structure) and 2 (alignment). Most of the mutations were clustered in a region, which was predicted to be a vitamin B12 binding domain by the SMART algorithm from the Pfam data base (http://smart.embl-heidelberg.de/). Figure 2 shows good similarity of the amino acid sequence of the C-terminal domain of AppA to the vitamin B12 binding domains of other proteins. Among them, glutamate mutase (PDB ID: 1ccw) and methionine synthase (PDB ID: 1bmt) have been proven to bind vitamin B12. Strikingly, however, a highly conserved histidine, predicted to be involved in vitamin B12 binding is exchanged to Pro (position 286 of AppA). We exchanged the neighbouring His residue at position 284 of AppA to Ala, but this did not result in any significant change of phenotype compared with the control strain (data not shown). A significant homology of the primary sequence of AppA to a known haem binding protein was not obtained. Nevertheless, we tested the binding of vitamin B12 or haem to the C-terminal part of AppA (AppAΔN protein, amino acids 5–190 deleted, 29.7 kDa) (Fig. 1) because the tetrapyrrole haem is used as a cofactor in redox active proteins, e.g. DcrA and EcDOS (Fu et al., 1994; Sasakura et al., 2006).

Figure 2.

Multiple sequence alignment of the C-terminal domain of AppA and selected members of B12 binding proteins. Sequence similarity search of the C-terminal domain of AppA was performed at NCBI (http://www.ncbi.nlm.nih.gov/BLAST/) using blastp with default settings. The selected sequences (except glutamate mutase and methionine synthase) shown in this figure have scores > 160 and E-value < 1e−35. The multiple sequence alignment was generated by clustal w with default parameters. The similarity identification was performed using BioEdit sequence alignment editor with the shade threshold ≥ 60%. Regions of similarity (grey) and identity (black) are highlighted. The triangles (Δ) above the sequences represent the mutation sites in the strains selected after random mutagenesis. The star (*) represents the conserved histidine which ligands the cobalt in vitamin B12 binding protein. The proteins listed from top to bottom and their respective GenBank accession numbers are AppA, Rhodobacter sphaeroides 2.4.1 YP_351608, the antirepressor of PpsR; glutamate mutase, Clostridium cochleariumCAA53484; methionine synthase, Escherichia coliCAA34601; NAP1-09377, Erythrobacter sp. ZP_01041675; SKA53-02696, Loktanella vestfoldensisSKA53 ZP_01002893; RPB_3983, Rhodopseudomonas palustris HaA2 YP_487587; SfumDRAFT_2617, Syntrophobacter fumaroxidans MPOB ZP_00665619; MXAN_0232, Myxococcus xanthus DK 1622 YP_628514.

After inducing overexpression in the Escherichia coli strain JM109(pQEAppAΔN), we applied cell extracts to either vitamin B12-agarose or haemin agarose. As a control, we applied cell extract of strain E. coli JM109(pQE30), which does not contain the appA gene. After extensive washing, the agarose was loaded on an SDS gel. As shown in Fig. 3A, the protein patterns were identical for extracts of the His-AppAΔN expressing strain and the control strain when applied to vitamin B12-agarose. Very strong protein bands around 30 kDa in size were detected when the extract of the His-AppAΔN expression strain was incubated with haemin agarose. These bands were missing when extracts of the control strain were applied. Western blot analysis confirmed that these bands indeed contain the C-terminal part of AppA (Fig. 3B). Western blot also detected very low amounts of His-AppAΔN bound to the vitamin B12-agarose. The pattern of multiple bands is most likely due to partial degradation of the protein. We also incubated cell extracts of E. coli strains expressing the His-AppAΔN variants His-AppAΔN9 (V306A/A349V/P357L) and His-AppAΔN14 (P282Q/A314V) with haemin agarose. R. sphaeroides strains expressing one of these AppA variants (RM9 and RM14) showed phenotypes similar to that of strain App11 which does not express AppA (Table S1, Supplementary material). No significant amounts of mutated His-AppAΔN were bound to the agarose, suggesting that the affinity of these His-AppAΔN mutant proteins to haem is low (Fig. S1, Supplementary material). This finding also assures that binding of wild-type His-AppAΔN was not due to unspecific interaction of the His-tag to the agarose. This was further confirmed by incubating the haemin agarose with the cell extract of E. coli JM109(pGEXAppAΔN), which expresses the GST-tagged AppAΔN. The GST-AppAΔN can also bind to haemin agarose (data not shown).

Figure 3.

The C-terminal domain of AppA (AppAΔN) encompasses a haem binding domain. One millilitre of cell extract containing 1 mg total protein was incubated with 40 μl of vitamin B12-agarose or haemin agarose (Sigma-Aldrich) at 4°C for 1 h. After extensive washing (six times with 1 ml of PBS, pH 7.3), the agarose was loaded on a 12% SDS-polyacrylamide gel, and the signals were analysed by Coomassie blue staining (A) and Western blot using an AppAΔN-specific antibody (B). The arrows indicate the expected size of AppAΔN. M, molecular mass standards (also shown in the left of figures); CL, clear lysate before incubation (10 μl out of 1 ml of cell extract containing 1 mg total protein); F, flow-through (the same volume as CL); W1, the first wash fraction (the same volume as CL); W6, the last wash fraction (the same volume as CL); B, agarose after extensive washing.

When His-AppAΔN was purified from the E. coli expression strain, it was only lightly coloured, suggesting that low amounts of the tetrapyrrole cofactor were bound. Spectral analysis revealed a peak at around 411 nm (Fig. 4A), as is typical for protein bound haem. In Fig. 4A, the dotted line represents air oxidized AppAΔN and the solid line represents the same sample reduced with sodium dithionite. The shift in the Soret peak from 411 nm to 424 nm after addition of dithionite demonstrates that the bound haem is redox active. The reduced form also showed a broad maximum at 556 nm (Fig. 4B). This spectrum is typical for pyridine haemochromes with absorbances at 420 nm (γ band), weak absorbance at 524 nm (β band) and a α band at 556 nm (Berry and Trumpower, 1987). These data demonstrate that the bound cofactor is iron containing haem.

Figure 4.

Absorption spectra of purified His-AppAΔN.
A. Spectra of His-AppAΔN (2.57 μg μl−1) directly after purification from E. coli (dotted line) and the same sample treated with sodium dithionite (solid line).
B. Spectra of oxidized pyridine hemichrome (dotted line) and reduced pyridine haemochromes (solid line) of isolated His-AppAΔN.
C. Spectra of His-AppAΔN (1.37 μg μl−1) reconstituted with haemin (dotted line) and the same sample treated with sodium dithionite (solid line).
D. Spectra of oxidized pyridine hemichrome (dotted line) and reduced pyridine haemochromes (solid line) of His-AppAΔN reconstituted with haemin. The spectra were recorded after mixing 50 μl of His-AppAΔN (1.37 μg μl−1) with 50 μl of 50% (v/v) pyridine in 0.2 M NaOH followed by the addition of potassium ferricyanide (dotted line) or sodium dithionite (solid line).

The association of iron with the C-terminal part of AppA was further confirmed by ICP-MS. Cell extracts from E. coli strains JM109(pQEAppAΔN) or JM109(pQE30) were bound to Ni-NTA agarose. The eluates were analysed for their content of iron, magnesia and manganese ions. The eluate containing the C-terminal domain of AppA contained an iron concentration which was about sevenfold higher than the concentration found in the eluates from the control, while the concentration of the other ions was comparable for extracts from both strains.

After incubating the protein with haemin and removing any unbound cofactor by dialysis, the protein fraction showed strong absorbance at 411 nm (Fig. 4C). After adding sodium dithionite, the reduced form also gave the typical pyridine haemochrome spectrum with a Soret peak at 424 nm and the absorption maxima at 524 nm and at 556 nm (Fig. 4D). The haem concentration was calculated from the differences in absorbance at 556 versus 540 nm using an absorption coefficient of 22.1 mM−1 cm−1 (Klatt et al., 1996). 1 M AppAΔN bound 0.3 M haem. Due to the low amount of haem bound to the protein after isolation from E. coli, the haem concentration could not be determined exactly. We obtained an approximate value of 0.75 mol iron per 1 mol of haem.

We also did titration experiments by incubating different amounts of haemin with AppAΔN. Lysozyme as well as bovine serum albumin (BSA) were used as controls and bound only minor amounts of haem (not shown). The spectra obtained after incubation of haemin with lysozyme were deduced from the spectra obtained for AppAΔN, and these corrected spectra are shown in Fig. S2 (Supplementary material). The same experiment was carried out with the variants AppAΔN14 and AppAΔN9, revealing that these proteins have less capacity to bind haem (Fig. S2, Supplementary material). When these protein variants were isolated from E. coli the amount of bound haem was so low that no meaningful spectra could be obtained.

Further experiments showed that AppAΔN can be reduced not only by sodium dithionite (E°′∼−420 mV at pH 7) but also by ascorbic acid (E°′∼60 mV at pH 7) (data not shown). The reduced protein was rapidly oxidized within few minutes when exposed to air (data not shown). No significant absorbance was detected after reconstitution of the protein with vitamin B12 or the haem precursor protophorphyrin IX (data not shown).

AppAΔN interferes with binding of PpsR to the puc promoter region

A previous in vivo analysis strongly suggested that neither the BLUF domain of AppA nor the 54 C-terminal amino acids encompassing a cysteine-rich domain are required for binding to PpsR (Han et al., 2004 and unpubl. data). In order to test binding of the C-terminal AppA domain to PpsR, their interaction was studied by analysing the interference of His-AppAΔΝ with DNA binding of the PpsR protein by gel retardation analysis. When a 259 bp DNA fragment spanning the puc promoter region was incubated with increasing concentrations of GST-tagged PpsR (GST-PpsR), two retarded bands were observed in a native gel (Fig. 5A). The DNA probe contains two consensus PpsR binding sequences and PpsR forms tetramers or dimers in vitro (Masuda and Bauer, 2002). When the DNA fragment (15.4 fmol) was incubated with 9 pmol of PpsR and increasing amounts of the N-terminally truncated His-tagged AppA (His-AppAΔN), a small increase in the amount of unbound DNA was observed (Fig. 5B). When His-AppAΔN was reconstituted with haemin, the amount of shift 2 clearly decreased with increasing protein amount, while the amount of DNA in shift 1 and unbound DNA increased (Fig. 5C). Neither the GST alone nor His-AppAΔN resulted in retardation of the puc promoter DNA fragment (Fig. 5D and E). These in vitro data support the conclusion based on in vivo experiments (Han et al., 2004) that PpsR interacts with the C-terminal part of AppA and show that binding of haem strengthens this interaction.

Figure 5.

Inhibition of the PpsR DNA binding activity by AppAΔN. Reduced PpsR (amount as indicated in the figure) was preincubated without AppAΔN or with AppAΔN at room temperature for 1 h and then incubated for further 40 min at 30°C with the 32P-labelled puc DNA probe. The samples were then analysed on a 4% native polyacrylamide gel.
A. PpsR can bind to the 259 bp puc promoter region.
B. PpsR was preincubated with AppAΔN which was not reconstituted with haemin.
C. PpsR was preincubated with AppAΔN which was reconstituted with haemin.
D and E. Neither GST alone (D) nor His-AppAΔN (E) can bind to the DNA probe.

The haem cofactor influences the interaction of AppAΔN to PpsR and to the BLUF domain

Our experiments showed that AppAΔN interferes with DNA binding by PpsR. However, several methods like pull-down assays or size exclusion chromatography failed to directly show the AppA–PpsR interaction (data not shown). In order to prove this interaction and to get quantitative data, we used surface plasmon resonance (SPR) measurements. The protein–protein binding was studied using an SPR spectroscope (Plasmonic®, HSS Systeme, Wallenfels, Germany) with microcuvette technology. Unlike other systems (e.g. Biacore), no flow of the buffer is required, which makes this system very sensitive. Association constants calculated as described in Experimental procedures provide a reliable value for the strength of the protein–protein interaction.

When PpsR was immobilized to a C18 matrix and increasing concentrations of AppAΔN were applied, we calculated an association constant (kass) of 18.8 s−1 mM−1 for this interaction (Table 1 and Fig. S3, Supplementary material). When AppAΔN was immobilized and PpsR was applied in solution, similar results were obtained (data not shown). In the presence of haem, the association constant of PpsR-AppAΔN increased 2.4-fold to a value of 44.6 s−1 mM−1 (Table 1 and Fig. S3, Supplementary material).

Table 1.  Association constants of the AppAΔN–PpsR interaction and the BLUF–AppAΔN interaction in the absence of haem and in the presence of haem.
Without haemWith haemWithout haemWith haem
Association constant (s−1 mM−1)18.844.69.931.2

Our previous in vivo data provided evidence for a direct interaction of the two AppA domains, even without covalent linkage (Han et al., 2004). Therefore, we also tested the interaction between the BLUF domain and AppAΔN. The same binding kinetics were observed when AppAΔN was immobilized with variation of the BLUF concentration or vice versa. An association constant of 9.9 s−1 mM−1 was detected (Table 1 and Fig. S3, Supplementary material), indicating a very weak interaction. The presence of haem increased this value by a factor of 3.1 (31.2 s−1  mM−1, Table 1 and Fig. S3, Supplementary material). Illumination of the samples in the presence of haem did not significantly change the association constant (33.3 s−1 mM−1, Fig. S3, Supplementary material). This may be due to the fact that the BLUF protein used in this experiment was not saturated by its cofactor FAD.

The determination of binding kinetics between the AppAΔN-mutants and BLUF in the presence of haem resulted in association constants of 24.5 s−1 mM−1 and 18.7 s−1 mM−1 for AppAΔN9 and AppAΔN14, respectively (Table 2). These results show that the AppAΔN-mutants in the presence of haem have a reduced binding capacity for BLUF in comparison to the wild-type protein AppAΔN (31.2 s−1 mM−1), but these association constants are still higher than the binding of AppAΔN and BLUF without the cofactor haem.

Table 2.  Association constants of the AppAΔN variants–PpsR interaction and the AppAΔN variants–BLUF interaction in the presence of haem.
Association constant (s−1 mM−1)44.620.921.431.224.518.7

For the binding of AppAΔN variants to PpsR, the kinetics showed tendencies similar to that seen for binding to BLUF. In the presence of the haem cofactor association constants of 20.9 s−1 mM−1 (AppAΔN9) and 21.4 s−1 mM−1 (AppAΔN14) were determined (Table 2). These results show that the AppAΔN-mutants have a reduced binding capacity for PpsR in comparison to the wild-type protein AppAΔN (44.6 s−1 mM−1), and these association constants are similar to the binding of AppAΔN and PpsR without the cofactor haem.

We also wanted to test whether the interaction of AppAΔN to the BLUF domain interferes with its interaction to PpsR. To this end, PpsR was coupled to the matrix and control experiments confirmed that no significant interaction of PpsR to the BLUF domain occurs. When AppAΔN alone together with haem was added to the coupled PpsR, an increase of about 26 arbitrary units (AU) was observed. When the two AppA domains were preincubated in the presence of haem and then added to the coupled PpsR, an increase of only 15.5 AU occurred. We conclude that the presence of BLUF interferes with the AppAΔN–PpsR interaction. When the two AppA domains were mixed in the presence of haem and illuminated, an increase of only 11 AU was observed, suggesting that the interference of the BLUF domain with the AppAΔN–PpsR interaction is stronger in the presence of light. In all experiments including both AppA domains, a big part of AppAΔN (about 50%) was released from PpsR during washing, indicating that the formed complexes are not very stable.

Strains expressing AppA variants with amino acid exchanges in the haem binding domain show altered expression and blue light regulation of photosynthesis genes

As already mentioned the DNA sequence encoding amino acids 135–378 of AppA was randomly mutagenized by error-prone polymerase chain reaction (PCR). After conjugation of the R. sphaeroides appA strain App11 with a pool of E. coli clones, each harbouring a plasmid with potentially mutagenized appA sequence (all named pRKappARM), colonies which showed colour clearly different from that of the control strain App11(p484-Nco5) and from the majority of the transconjugants were selected. All selected strains accumulated less bacteriochlorophyll (BChl) at low oxygen tension than the control strain having wild-type appA on a plasmid (Table S1, Supplementary material). The amino acid exchanges we detected for the individual strains are listed in Table S1 (Supplementary material) and indicated in Figs 1 and 2. These exchanges clustered in a domain spanning from amino acid 204 to amino acid 357 which we identified as haem binding domain.

All appA sequences in the selected strains harboured more than one amino acid exchange. Based on the identified amino acid exchanges, we constructed further AppA variants by site directed mutagenesis. All these mutants with only one or two amino acid exchanged had considerably higher BChl content than the strains selected after random mutagenesis (Table S1, Supplementary material).

For all mutant strains listed in Table S1 (Supplementary material), we monitored the redox-dependent expression of the puf and puc operons, which encode pigment binding proteins.

As shown for some representative strains in Fig. 6 and as summarized in Table S1 (Supplementary material), all strains that were selected showed very low puc mRNA levels, even after shifting the cells to low oxygen. Low puc mRNA levels and moderate to strong increases in the transcript level were observed after the reduction of oxygen tension for some strains (e.g. RM14 and RM17, Fig. 6). For those strains with very low puc mRNA levels, semi-quantitative reverse transcription (RT)-PCR was performed to detect puc mRNA (e.g. RM9, Fig. 6). This revealed that puc mRNA levels in these strains increase after a shift to low oxygen tension, even if the total level remains very low (Table S1, Supplementary material). At low oxygen tension, all these strains showed puf mRNA levels which were reduced by more than 50% compared with the control strain APP11(p484-Nco5) expressing the wild-type appA gene from a plasmid. The puf mRNA levels in the mutants, however, were less affected than puc mRNA levels. All mutants showed a strong increase in puf mRNA levels after the reduction of oxygen tension, which was also observed in control cells.

Figure 6.

puf and puc expression in strains expressing representative AppA variants. Kinetics of puf and puc expression in control strains (I), AppA variants screened after random mutagenesis (II). Total RNA was isolated at indicated time points, and puf and puc transcript levels were monitored by RNA gel-blot analysis. A 14S rRNA-specific probe was used as internal standard.
A. All strains were cultivated under high oxygen tension (200 μM) and then shifted to growth under low oxygen tension. Cells were harvested at the time of transition, 90 and 180 min after the oxygen shift.
B. Cells were grown at 104 ± 24 μM dissolved oxygen and shifted from the dark into blue light at time point 0 or kept in the dark. Qualitative values for the expression changes from repeated experiments are provided in Table S1 (Supplementary material).

All strains harbouring appA sequences with only a single amino acid exchange in the haem binding domain accumulated similar puf mRNA levels and only moderately reduced puc mRNA levels. These strains showed similar redox regulation as the control strain expressing the wild-type appA gene from a plasmid. The same was observed for the double mutants I (A349V/P357L), J (V306A/P357L) and K (V306A/A349V) (Table S1, Supplementary material).

At high oxygen tension, the PpsR protein represses transcription of photosynthesis genes. When the oxygen tension drops to 104 ± 24 μM or below, the AppA protein releases PpsR repression in dark grown cultures (Gomelsky and Kaplan, 1997; Braatsch et al., 2002). In the presence of blue light, however, AppA does not interfere with the repressing effect of PpsR (Braatsch et al., 2002; Happ et al., 2005; Masuda et al., 2005b). We quantified puf and puc mRNA levels in R. sphaeroides strains expressing the different AppA variants after blue light illumination of cultures grown at 104 ± 24 μM dissolved oxygen. Results for some representative strains are shown in Fig. 6B and results for all strains are summarized in Table S1 (Supplementary material). All strains that expressed AppA proteins with only one amino acid exchange in the haem binding domain and the double mutants I (A349V/P357L), J (V306A/P357L) and K (V306A/A349V) showed blue light-dependent repression of puf and puc genes similar to that observed in control strains (Table S1, Supplementary material). Some of the strains, which express AppA variants with three amino acid exchanges and show only very low puc mRNA levels, failed to show blue light inhibition of puf expression: RM3 (L292R/A349V/P357L), RM9 (V306A/A349V/P357L) (Fig. 6B) and RM12 (V307G/A349V/P357L). The same was observed for the double mutants RM14 (P282Q/A314V) (Fig. 6B) and RM21 (L204P/Q257R). Two of the AppA variants with three amino acid exchanges were able to support blue light repression of puf genes, even though puc mRNA levels were quite low: RM13 (S348P/A349V/P357L) and RM17 (I263T/A349V/P357L) (Fig. 6B). AppA variants that resulted in light-independent puf or puc expression still showed redox regulation, although total puf and puc mRNA levels were reduced.

The puf and puc expression and regulation in strains RM3 (L292R/A349V/P357L) and RM9 (V306A/A349V/P357L) were very similar to the low expression levels in strain App11, which does not express AppA. Therefore, we intended to test AppA levels in some representative strains by Western blot analysis. We were unable to detect AppA protein in strains RM3, RM9 and RM13. This could be due to the fact that (i) no protein is produced, (ii) the protein is highly unstable or (iii) the protein is not detected by the antibody. Protein levels detected by antibodies in strain RM14 (P282Q/A314V) were moderately reduced when compared with the control strain. We checked the stability of AppA for strains APP11(p484-Nco5) and RM14, after blocking translation by chloramphenicol. We did not find any significant differences in AppA turn-over (data not shown). We quantified the appA and ppsR mRNA levels in strains APP11(p484-Nco5), RM9 and RM14 by semi-quantitative RT-PCR, and did not detect any significant differences (data not shown). In order to see, whether the antibody would detect the C-terminal domain of the AppA variant in strain RM9, we expressed His-AppAΔN9 and His-AppAΔN14 in E. coli and performed Western blot analysis. After partial purification, we observed that the AppA variants were not (His-AppAΔN9) or moderately (His-AppAΔN14) detected by the AppAΔN-specific antibody, even though they were clearly visible in comparable amounts after staining and the wild-type protein was well detected (data not shown). It is conceivable that amino acid exchanges in the haem binding domain cause a local structural change in AppA that weakens detection by the antibody even after denaturing electrophoresis. Together our data suggest that the phenoptype of strain RM9 is not due to a lack of the AppA variant but to the low haem binding capacity of this protein that correlates with very weak interaction to the antibody due to local structural changes.


A haem cofactor is required for redox and light signalling by the AppA protein

The AppA domain spanning amino acids 274–393 was predicted to bind vitamin B12. Our data strongly suggest that this domain binds the tetrapyrrole haem instead. This view is supported by (i) binding of AppAΔN to haemin agarose, (ii) an absorbance typical for bound haem to isolated AppAΔN protein and (iii) the effect of haem in AppA binding studies. The AppA variant RM9 resembles the APP11 strain which lacks AppA, and binds only low amounts of haem. The AppA variant RM14 which shows higher expression of photosynthesis genes than RM9 binds more haem compared with this variant but less than the wild-type protein.

Haem proteins have been known to play important roles in biology as oxygen carriers (e.g. haemoglobin), oxygen activator (e.g. cytochrome P450 and peroxidase) and mediator of electron transfer (e.g. cytochrome c) (Antonini and Brunori, 1971; Poulos, 1988; Sono et al., 1996). More recently, a new class of haem proteins has been discovered that serve as biological sensors (Rodgers, 1999; Gilles-Gonzalez and Gonzalez, 2005). The signalling mechanisms of these haem-based sensors are based on the binding of ligands to the haem (the input signal). This input signal can induce a protein conformational change that ultimately regulates DNA binding or other enzymatic activities (the output signal). Several redox active proteins are haem-based sensors and they sense oxygen via a haem cofactor [e.g. FixL, EcDOS, AxPDEA1 and others (Reviewed in Gilles-Gonzalez and Gonzalez, 2004)]. The spectral analysis with the Soret peak shift and the different absorption at 556 nm of the oxidized and reduced AppAΔN (Fig. 4) demonstrate that the haem bound to AppAΔN is redox active. It is likely that the redox state of the haem cofactor alters the structure of AppAΔN and, consequently, its affinity in binding to PpsR. This could not be tested by plasmon resonance studies because the redox state of the proteins cannot be controlled with the existing set-up. However, our binding studies revealed that the mutations that led to a loss of haem binding resulted in reduced interactions between AppAΔN and PpsR, when compared with the interaction of the wild-type domain in the presence of haem, emphasizing the importance of haem for signal transduction. The primary sequence of AppAΔN shows no significant homology to known haem binding proteins and no conserved His residues are present. We therefore considered that AppAΔN might bind haem precursors without the central iron. However, we could not detect any binding of protoporphyrin IX, and the binding of AppAΔN to haemin agarose, its redox activity and absorbance characteristics of the bound haem strongly support the presence of iron, suggesting that AppA harbours a new type of haem binding domain. It was shown for some proteins that haem can be co-ordinated by other amino acids than His. Tyrosine residues co-ordinate the IsdA protein of Staphylococcus aureus (Vermeiren et al., 2006) and in a variant of the CcmE haem chaperone of E. coli (García-Rubio et al., 2007). In the gamma subunit of ethylbenzene dehydrogenase from Aromatoleum aromaticum methionine and lysine are the axial haem ligands (Kloer et al., 2006). The AppA protein contains a Lys next to the position where other vitamin B12 binding proteins have the highly conserved His (Fig. 2). The exact co-ordination of haem in AppA needs further analyses.

The AppA/PpsR system co-ordinately regulates photosynthesis gene expression together with the PrrB/PrrA system

The remarkable feature of the R. sphaeroides AppA protein is its ability to transmit and integrate both redox and light signals. We analysed AppA variants with amino acid exchanges in the haem binding domain with regard to their expression levels and redox and light regulation of the puf and puc genes. Some of the mutants with three amino acid exchanges in the haem binding domain resemble the APP11 strain that lacks the AppA protein. While puf mRNA levels were reduced to about 20% of those of the control strain, puc mRNA levels were reduced to a much larger extent, and puc signals were no longer detectable by Northern blot. The puc genes carry PpsR binding sites in their promoter region and are therefore under direct control of the AppA/PpsR system. Because no PpsR binding sites are present in the proximity of the puf promoter, the effect of AppA/PpsR must be indirect. All phenotypically APP11-like strains showed increased expression of puf and puc after a drop in oxygen tension, but at very low overall expression levels. These results imply that redox regulation of puf and puc at low expression level occurs independently of AppA/PpsR. However, no light regulated puf or puc expression was observed in the strains with strongly reduced puc mRNA levels. We conclude that the AppA/PpsR system is the major system that represses both the puf and the puc genes in the presence of light.

In R. sphaeroides, the two-component system PrrB/PrrA activates the expression of photosynthesis genes at low oxygen tension. PrrB functions as a sensor histidine-kinase/phosphatase and PrrA functions as a response regulator through binding DNA targets. The PrrA binding site contains two GCGNC inverted repeats with variable half-site spacing and this consensus sequence is present in the promoter regions of both puc and puf genes (Laguri et al., 2003). Moskvin et al. (2005) showed that the expression of prrA, which has no PpsR binding sites in its upstream region, indirectly depends on PpsR suggesting that PrrA-dependent regulation is also modulated by PpsR. Moreover, it was shown that the PrrB/PrrA system co-ordinately controls light-dependent regulation of photosynthesis gene expression together with the AppA/PpsR system under anaerobic (Happ et al., 2005) and semi-aerobic conditions (Jaeger et al., 2007). These results unravelled an intimate interplay between the AppA-PpsR and PrrA regulatory pathways of photosynthesis gene expression. While the PrrB/PrrA system is sufficient for some redox control of puf and puc transcription, it cannot mediate a blue light response at intermediate oxygen concentrations. If AppA is lacking, or not binding to PpsR due to mutations in the haem binding site, puf and puc genes are repressed by PpsR. Despite this repressing effect, some activation by PrrA at low oxygen concentration takes place. Blue light regulation at intermediate oxygen tension, however, completely depends on the presence of AppA and a change in its interaction with PpsR.

Integration of redox and light signals by AppA

Based on our data we suggest a model for the integration of redox and light responses by the AppA protein (Fig. 7). According to this model, AppAΔN is only responsive to light when the haem cofactor is in reduced state at low oxygen tension (low pO2). Haem affects the interaction of AppAΔN and PpsR, as revealed by SPR. The presence of haem increases the association constant of the two proteins by a factor of about 2.4 in vitro. Interestingly, haem also affects the interaction of AppAΔN and the AppA N-terminal BLUF domain. This suggests that depending on the redox status the affinity of AppAΔN to bind to PpsR increases and at the same time it gains the potential to interact with the BLUF domain. The strength of the interaction of the two AppA domains is determined by the light signal. Our interaction studies failed to prove a direct effect of light on the interaction of the two AppA domains. However, the interaction studies support the view that the BLUF domain can interfere with the binding of AppAΔN to PpsR as outlined in the model (Fig. 7, low pO2, light). This interfering effect is clearly stimulated by light. Our data suggest that the haem redox state keeps AppAΔN in a conformation that alternatively favours interaction with the BLUF domain or with PpsR at low oxygen tension (Fig. 7, low pO2). Thus, the haem cofactor is essential for redox signalling and light signalling by PpsR. The question whether the interaction of the two AppA domains is intramolecular, as implicated in Fig. 7, or intermolecular, as already considered in a previous publication (Masuda et al., 2005b) needs further investigation.

Figure 7.

Model for the integration of redox and light responses by the AppA protein. The AppA protein contains two functional domains: the BLUF domain (FAD as cofactor, sensing light signal) and the C-terminal domain (haem as cofactor, sensing redox signal). At high oxygen tension oxidized haem leads to a conformation of the C-terminal domain that does not allow interaction or only weak interaction with PpsR. At low oxygen tension reduced haem leads to a conformation of the C-terminal domain that now strengthens interaction to PpsR as well as to the BLUF domain. At low oxygen tension in the dark AppA strongly binds to PpsR and releases its repressing effect on transcription of photosynthesis genes. At low oxygen tension in the light the BLUF domain interacts with C-terminal domain and PpsR will repress photosynthesis genes.

It is conceivable that PpsR like its counterpart CrtJ in R. capsulatus undergoes a redox switch, but needs AppA for this redox switch due to its much lower redox potential compared with CrtJ (Kim et al., 2006). Reduction of the haem cofactor at low oxygen tension as well as reduction of the flavin by blue light illumination may not only influence direct interaction of the proteins but also the electron flow between AppA and PpsR.

We can not exclude the possibility that AppA senses the amount of haem in the cell instead of directly sensing redox conditions. In R. sphaeroides the expression of hem genes, which are involved in the biosynthesis of haem and other tetrapyrrole-related products (e.g. BChl), is regulated by the PrrB/PrrA and AppA/PpsR systems (Moskvin et al., 2005) and therefore depends on the redox state. Smart and Bauer (2006) found that the regulation of hem gene expression in R. capsulatus is a feedback control involving the haem binding protein HbrL. Thus, AppA might be a protein sensing the overproduced haem in response to changes in cellular redox and consequently controls the expression of hem genes and other photosynthesis genes.

In summary, our model proposes a light switch of AppA mediated by the BLUF photoreceptor domain that only affects photosynthesis gene expression when a haem-dependent redox switch affects PpsR interaction. The combination of these two events leads to the integration of the two stimuli by AppA.

Experimental procedures

Bacterial strains and growth conditions

Bacterial strains and plasmids are listed in Table 3. R. sphaeroides strains were cultivated at 32°C in a malate minimal salt medium. Light-shift and oxygen-shift experiments were performed as described previously (Braatsch et al., 2002; Han et al., 2004). E. coli strains were cultured in Luria–Bertani broth at 37°C. Conjugation was performed as described elsewhere (Klug and Drews, 1984). When required, antibiotics were used at the following concentrations: kanamycin 25 μg ml−1; tetracycline 20 μg ml−1 (E. coli) or 2 μg ml−1 (R. sphaeroides); ampicillin 200 μg ml−1 (E. coli); trimethoprim 50 μg ml−1 (R. sphaeroides). In the presence of light, no tetracycline was used.

Table 3.  Strains and plasmids used in this study.
Strain or plasmidRelevant featuresReferences/source
E. coli
 JM109Host strain for plasmid constructions and protein overexpressionNew England Biolabs
 M15 (pREP4)Host strain for His-tag protein overexpression, KmrQiagen
 S17-1Tra+ donor for diparental conjugationSimon et al. (1983)
R. sphaeroides
 2.4.1Wild typeW.R. Sistrom
 APP112.4.1 appA::TprGomelsky and Kaplan (1995)
 pGEX-4T-1Cloning vector for GST-fusion protein overexpression, AprAmersham Pharmacia
 pQE30Cloning vector for His-tag protein overexpression, AprQiagen
 pGppspGEX-2TK::ppsR, AprGomelsky et al. (2000)
 pGEXAppAΔNpGEX-4T-1::appAΔ5–190, AprThis work
 pQEBLUFpQE30::appA(1–168), AprThis work
 pQEAppAΔNpQE30::appAΔ5–190, AprThis work
 pQEAppAΔN9pQE30::appAΔ1–196 (V306A/A349V/P357L), AprThis work
 pQEAppAΔN14pQE30::appAΔ1–196 (P282Q/A314V), AprThis work
 pRK415Tcr, lacZa, IncPKeen et al. (1988)
 pRKappARMA plasmid collection with randomly mutagenized appA genes, the sites of mutation are indicated in Table S1 (Supplementary material)This work
 p484-Nco5pRK415 derivative; contains full length appA with its own promoterGomelsky and Kaplan (1995)
 p484-Nco5ΔIn frame deletion of appA in p484-Nco5; results in loss of codon 5–190Gomelsky and Kaplan (1998)

Genetic techniques

DNA cloning was performed according to standard protocols (Sambrook and Russell, 2001). Oligonucleotides leading to suitable recognition sites for cloning in PCR products were synthesized by Roth (Karlsruhe, Germany). DNA sequencing was performed in the ABI-Prism 310 genetic analyser (Applied Biosystems).


Random mutagenesis of the appA central part (nucleotides 403–1134 from the start codon of appA) was performed by error-prone PCR (using primers 5′-CATGCATGCGCCGACAACACCAACATC-3′ and 5′-TGCTCTAGAGGAGTCCTTCAGCTTCGA-3′) in a buffer containing 10 mM Tris-HCl, 50 mM KCl, 1.5 mM MgCl2, 0.24 mM of each dATP, dTTP, dGTP, dCTP, 0.2 mM dITP and 10% (v/v) DMSO. The pool of PCR products was used to in frame replace the original wild-type sequence. The resulting 1.95 kb-DNA fragments, containing the full length appA gene pool and its own promoter, were cloned into vector pRK415, resulting hybrid plasmids collectively called pRKappARM. These plasmids were conjugated into the appA strain R. sphaeroides APP11. The site-directed mutagenesis was performed by overlap PCR according to standard protocols (Sambrook and Russell, 2001) and the site of mutation was identified by DNA sequencing.

Gene expression analysis

Expression of puc, puf and rRNA genes was monitored by RNA gel blot analysis or semi-quantitative RT-PCR as described previously (Braatsch et al., 2002; Gomelsky et al., 2003).

Bacteriochlorophyll measurements

Photopigments were extracted from cell pellets with acetone-methanol (7:2, v/v). BChl concentration was calculated using an extinction coefficient of 76 mM−1 cm−1 at 770 nm (Clayton, 1966).

Construction of expression vectors

The 0.64 kb BglII–KpnI DNA fragment, containing appA codons 1–168, was amplified by PCR (using primers 5′-GGAAGATCTCAACACGACCTCGAGGC-3′ and 5′-CGGGGTACCTGTGCTGCAAGGCGATTA-3′) and cloned into the BamHI and KpnI sites of vector pQE30 (Qiagen, Hilden). The recombinant plasmid designated pQEBLUF was transformed into E. coli M15 (pREP4).

The 1.04 kb BglII–KpnI DNA fragment was amplified by PCR (using primers 5′-GGAAGATCTCAACACGACCTCGACAAG-3′ and 5′-CGGGGTACCGACGCTGCAAGAATC-3′) using p484-Nco5Δ (Gomelsky and Kaplan, 1998) as template and cloned into the BamHI and KpnI sites of vector pQE30. The recombinant plasmid designated pQEAppAΔN was transformed into E. coli JM109.

The 0.76 kb BglII–SmaI DNA fragments, containing appA codons 197–450 with mutation sites V306A/A349V/P357L or P282Q/A314V, were amplified by PCR (using primers 5′-GGAAGATCTGATCTGCTGAGCACCGAT-3′ and 5′-TCCCCCGGGTCAGGCGCTGCGGCGG-3′) and cloned into the BamHI and SmaI sites of vector pQE30. The recombinant plasmids designated pQEAppAΔN9 and pQEAppAΔN14, respectively, were transformed into E. coli JM109.

The 0.8 kb BglII–SmaI DNA fragment was amplified by PCR (using primers 5′-GGAAGATCTCAACACGACCTCGACAAG-3′ and 5′-TCCCCCGGGTCAGGCGCTGCGGCGG-3′) using p484-Nco5Δ as template and cloned into the BamHI and SmaI sites of vector pGEX-4T-1 (Amersham Biosciences, Freiburg). The recombinant plasmid designated pGEXAppAΔN was transformed into E. coli JM109.

Protein overexpression and purification

His-tagged proteins were induced in E. coli strains at 17°C overnight with 0.5 mM isopropyl-β-d-thiogalactopyranoside. The purification was performed using Ni-NTA agarose (Qiagen, Hilden) according to the manufacturer's instructions. GST-PpsR was induced in E. coli JM109 and purified as described previously (Gomelsky et al., 2000). GST-AppAΔN was induced in E. coli JM109 at 17°C overnight with 0.5 mM isopropyl-β-d-thiogalactopyranoside. The purification was performed using Glutathione Sepharose 4B according to the manufacturer's instruction (Amersham Biosciences, Freiburg). The eluted proteins were dialysed in storage buffer (250 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, 15% Glycerol). Proteins were quantified using the Bradford assay (Bradford, 1976). The antibody was produced from rabbits by BioGenes (Berlin).

Gel mobility shift analysis

The DTT (1,4-dithiothreitol, 100 nmol) treated protein or protein mixtures were incubated with a 259 bp 32P-labelled puc promoter fragment (−281 to −23 from start codon of pucB gene, amplified by using primers 5′-GGACGAGCAGCGTCAATTTC-3′ and 5′-GGACGAGCAGCGTCAATTTC-3′) containing two PpsR binding sites at 30°C for 40 min in 20 μl of binding solution [50 mM Tris/HCl (pH 7.0), 1 mM EDTA, 150 mM NaCl, 10% glycerol, 1 μg of BSA and 1 μg of salmon sperm DNA]. Afterwards, the mixtures were subjected to 4% native polyacrylamide gel electrophoresis in Tris-acetate-EDTA buffer. The signals were analysed by using a Phosphoimaging system (Molecular Imager® FX; Bio-Rad) and the appropriate software (QUANTITY ONE; Bio-Rad).

Reconstitution assay

The purified protein was dialysed in storage buffer containing 10 mg l−1 haemin or 2 mg l−1 vitamin B12 at 4°C overnight with low agitation. Then dialysis was performed in the same storage buffer without haemin or vitamin B12 at 4°C overnight and the same volume of fresh storage buffer was replaced twice during dialysis. Absorbance spectroscopy was performed on a spectrophotometer (Lambda 12; Perkin Elmer). The last dialysis buffer was used as blank. The pyridine haemochrome and pyridine hemichrome were prepared as described by Berry and Trumpower (1987).

Surface plasmon resonance determinations

Surface plasmon resonance is an optical tool used for monitoring the binding processes between proteins, antigen-antibodies or other complexes. The protein–protein binding was studied using an SPR spectroscope (Plasmonic®, HSS Systeme, Wallenfels, Germany). The spectroscope was used in normal autosampler modes. One of the main advantages of the Plasmonic® SPR system is the microcuvette technology. Up to eight measurements can be performed at the same time.

All reagents and proteins were handled and diluted in phosphate buffered saline (PBS, 0.25 M, pH 7.3). All measurements were performed at least twice at a temperature of 22.0°C. For sample incubation with light, a strong light (1500 W, white light) was placed in the SPR machine at an adequate height (30 cm) over the sample tray. Samples were illuminated for 10 min before application.

Protein immobilization was performed using a non-specific binding surface modification of the gold chip. This hydrophobic C18-surface was created using Octadecyltrimethoxysilane on a previously cleaned gold surface according to the procedure from Hartmann (2004). After protein immobilization, all further binding residues were saturated using a high-concentrated BSA solution. After blocking the remaining binding sites, the cuvette and chip surface were washed using PBS buffer. Finally, the second protein was applied to the cuvette and interactions of the second protein to the immobilized one on the chip surface were studied.

Measuring different concentrations of one protein with a constant concentration of the immobilized interacting partner allows the determination of the association constant (kass) for these two proteins. For this purpose, the method of Edwards and Leatherbarrow (1997) was used. This method is based on the assumption that this kinetic parameter can be determined from the initial binding rates and the maximum binding capacity.

kass is determined graphically from the equation Xslope = Rmax × kass, where Xslope is the slope, and Rmax is the maximum binding capacity at the highest protein concentration.

ICP-MS measurements

Samples were analysed by an inductively coupled plasma mass spectrometry system equipped with an Octopole Reaction System (ORS-ICP-MS Agilent 7500ce; Agilent Technologies, Santa Clara, CA, USA). Mg was analysed in normal (nogas) mode whereas Fe was detected on 56Fe in H2-mode to eliminate for 40Ar16O-interference. Detection limits – determined by threefold standard deviation of repeatedly measured blank solutions – were at 0.024 μg l−1 for Mg and 0.034 μg l−1 for Fe.


We thank Mark Gomelsky for providing plasmids, Karl Forchhammer for help in initial plasmon resonance measurements, Peter Friedhoff for help with bioinformatic analyses, Andreas Jäger and Sebastian Frühwirth for technical assistance, Wolfgang Reiher for ICP-MS measurements and Erik Rytting for correcting the manuscript. This work was supported by Deutsche Forschungsgemeinschaft (Kl563/15-3 and Graduiertenkolleg Biochemie von Nukleoproteinkomplexen).