PpsR, a Regulator of Heme and Bacteriochlorophyll Biosynthesis, Is a Heme-sensing Protein*

Background: PpsR controls synthesis of heme and bacteriochlorophyll in purple photosynthetic bacteria. Results: PpsR binds heme as a co-factor and changes its DNA binding pattern. Conclusion: Heme affects the ability of PpsR to regulate tetrapyrrole gene expression. Significance: A cysteine axial ligand to heme is in the helix-turn-helix domain providing a way for heme to affect DNA binding properties of PpsR. Heme-mediated regulation, presented in many biological processes, is achieved in part with proteins containing heme regulatory motif. In this study, we demonstrate that FLAG-tagged PpsR isolated from Rhodobacter sphaeroides cells contains bound heme. In vitro heme binding studies with tagless apo-PpsR show that PpsR binds heme at a near one-to-one ratio with a micromolar binding constant. Mutational and spectral assays suggest that both the second Per-Arnt-Sim (PAS) and DNA binding domains of PpsR are involved in the heme binding. Furthermore, we show that heme changes the DNA binding patterns of PpsR and induces different responses of photosystem genes expression. Thus, PpsR functions as both a redox and heme sensor to coordinate the amount of heme, bacteriochlorophyll, and photosystem apoprotein synthesis thereby providing fine tune control to avoid excess free tetrapyrrole accumulation.

Heme-binding proteins are involved in a wide range of biological functions including electron transfer (1), oxygen metabolism (1), iron metabolism (2), nitric oxide synthesis (3) and neuron-degenerated disease (4,5). As a co-factor, heme can modulate transcription (6), translation (7), protein targeting (8), and degradation (2,9,10) as well as functioning as a sensor and carrier of a variety of gas molecules (11)(12)(13)(14)(15). Among photosynthetic organisms, the heme and bacteriochlorophyll biosynthetic pathways share common intermediates from ␦-aminolevulinic acid to protoporphyrin IX. Unbound heme and bacteriochlorophyll are both toxic; thus, the vast majority of these tetrapyrroles are sequestered within proteins. However, it is not clear how cells coordinate the synthesis of defined amounts of tetrapyrrole end products that utilize common intermediates with the synthesis of apoproteins that bind tetrapyrroles. In recent years, several heme-sensing regulatory proteins have been identified in prokaryotic (2) and eukaryotic organisms (9,16) that are able to sense free heme. To date, identified heme sensors contain one or more heme regulatory motifs (HRMs) 2 comprising a conserved -Cys-Pro- (8,17).
In Rhodobacter sphaeroides, heme and bacteriochlorophyll biosynthesis are co-regulated by the redox and light controlled PpsR-AppA system (18). PpsR is a DNA-binding transcription factor that recognizes conserved palindromes present in a number of tetrapyrrole and photosynthesis promoters (19,20). Under aerobic conditions, a pair of redox active Cys in PpsR undergo oxidation to stimulate binding of PpsR to target promoters to block transcription (18). PpsR is also regulated by the flavin-containing antirepressor AppA that responds to changes in both redox and light intensity. AppA inactivates PpsR by forming an inactive PpsR 2 -AppA complex under anaerobic dark conditions (18). Recently, it was reported that AppA also binds heme as a co-factor (21,22), although the role of heme bound to AppA remains ambiguous.
In this study, we present evidence that PpsR is a heme-binding protein and that the redox active Cys 424 present in the DNA binding domain of PpsR is critical for heme interaction. The binding of heme to PpsR was found to change its DNA binding pattern and to induce increased transcription of several PpsRregulated genes. These results are consistent with the hypothesis that excess heme can quickly change the state of photosynthetic gene expression from inhibition to activation, thus providing a mechanism for bacteria to react to the toxic unbound tetrapyrrole products. It is also worth noting that PpsR does not contain a typical HRM, suggesting that hemesensing regulators may be more widespread than was previously understood.
agar-solidified plates and liquid cultures for E. coli and R. sphaeroides. Kanamycin was used at 50 g/ml for agar-solidified plates and at 25 g/ml for liquid cultures as needed.
Construction of Expression Vectors-All proteins were expressed in E. coli using a modified SUMO-I (LifeSensors) overexpression system. For full-length PpsR, the coding region was cloned from pETPpsR (18) into the NdeI-SacI restriction sites of pSUMO, resulting in the recombinant plasmid pSUMO-PpsR. Plasmid pSUMO-PpsR was subjected to sitedirected mutagenesis (Herculase II; Stratagene) to construct stop codons to form truncated PpsR variants as well as PpsR point mutations.
Purification of PpsR from R. sphaeroides-PpsR-FLAG strain culture was collected and resuspended in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mg/ml lysozyme, and 20 l/per sample of protease inhibitor mixture (Sigma)). Cells were kept on ice, disrupted by sonication of 10 s with 10 s interval, for 10 min in total. Cracked cells were centrifuged at 15,000 ϫ g for 30 min, and the supernatant was incubated with 50 l of EZview TM red anti-FLAG M2 affinity gel (Sigma) at 4°C overnight. The anti-FLAG gel was then spun down and washed with TBS buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl). PpsR-FLAG was eluted with 300 l of 3ϫ FLAG peptide (Sigma).
Purification of PpsR from E. coli-SUMO-tagged PpsR was overexpressed by induction of T7 polymerase with 1 mM isopropyl-␤-D-thiogalactopyranoside at 16°C overnight in LB. Cells were disrupted by three passages through a cell cracker and clarified by centrifugation at 15,000 ϫ g for 30 min. PpsR-SUMO was purified by binding to HisTrap TM HP columns (GE Healthcare), washed with ϳ10 column volumes of 50 mM imidazole, and then eluted with a 50 -500 mM linear gradient of imidazole in a buffer comprising 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, and 5% glycerol. Eluted PpsR-SUMO was then cleaved from the SUMO tag with SUMO-protease-I (LifeSensors) for 1 h at room temperature. Following cleavage, tagless PpsR was then isolated by gel filtration chromatography using a Sephacryl S-200 HP resin (GE Healthcare) equilibrated in 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 5% glycerol.
Protein-Co-factor Interactions-Heme stock solutions were freshly dissolved in 0.01 M NaOH. The final concentration was determined spectrophotometrically using a coefficient of 58.4 mM Ϫ1 cm Ϫ1 (23). To quantify heme binding, purified protein was incubated with a 2-fold excess of heme at room temperature for Ͼ20 min. The unbound heme was then removed by passing through MidiTrap TM G-25 column (GE Healthcare) or HiTrap TM Q HP column (GE Healthcare). Heme bound to protein was determined by using the pyridine hemochromagen assay as described previously (24). All the protein-heme interactions, including the reactions used for later analysis, were carried out under oxidizing conditions unless noted otherwise.
Spectroscopy-Electronic absorption UV-visible spectra were recorded using a Beckman DU 640 spectrophotometer. Purified protein was incubated with various concentrations of heme at room temperature for more than 20 min before the measurements were taken. To measure the reduced spectrum of protein-heme complex, 1 M sodium dithionite stock was made freshly with the buffer of 2 M Tris-HCl, pH 8.0, with a final concentration of 5 mM sodium dithionite added to the sample. Carbon monoxide was added by briefly bubbling CO into the protein sample. The deoxy-PpsR-heme spectrum was obtained using the same method for reduced spectrum, except that all buffer and samples were completely degassed, and the whole procedure was done in an anaerobic hood and with flow-stop cuvette.
Tryptophan Fluorescence Quenching Assay-Fluorescence measurements were performed at room temperature using a LS50B luminescence spectrometer (PerkinElmer Life Sciences) in a 2-ml sample volume with protein in 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 5% glycerol. Heme was added into the sample to final concentration of 0.1-2.5 M. After incubation for 2 min, the sample was excited at 272 nm. Three emissions from 290 to 400 nm were then recorded and averaged. Data were processed using nonlinear curve fitting with one-to-one binding model using Origin software (OriginLab). The experiments were carried out in the PpsR-heme concentration range that exhibits linear relationship to protein fluorescence (supplemental Fig.  1).
Gel Mobility Shift Analysis-Probes for gel mobility shift assays were prepared by colony PCR amplification of the puc (262 bp) and the hemE (344 bp) promoter regions (supplemental Table 1). 10 nM DNA probes were incubated with different concentrations of purified protein for 30 min at room temperature, in 40 mM Tris-HCl, pH 8.0, 250 mM NaCl, 50 mM KCl, 5 mM MgSO 4 , 5% glycerol, 1 mM EDTA, and 20 g/ml heparin. The mixtures were then subjected to 6% PAGE at 4°C in a buffer containing 50 mM Tris-HCl, pH 8.0, 380 mM glycine, and 2 mM EDTA. After the electrophoresis, the gel was stained with SYBR Green dye (Invitrogen) as described by the manufacturer and scanned with a Typhoon 9210 variable mode imager (GE Healthcare). When reducing condition was required, fresh dithionite of 5 M final concentration was added into all of the buffers and the reactions.
DNase I Footprinting-Promoter region of puc (262 bp) was amplified with a 6-FAM-labeled primer on the forward strand and 5-HEX-labeled primer on the reverse strand. A DNA binding reaction similar to that described for the gel mobility shift assays was set up in a 20-l volume to which 5 l of DNase I was added for 30 min at room temperature to digest DNA. DNA digestion was then stopped by the addition of 20 l of 0.5 M EDTA. DNA digestion products were recovered using a Min Elute PCR purification kit (Qiagen), with sample eluted with 15 l of elution buffer containing 0.75 l of 500 LIZ TM Size Standard (Applied Biosystems). The samples were separated and detected with a 3730 DNA Analyzer (Applied Biosystems) and analyzed with Peak Scanner Software v1.0 (Applied Biosystems).
Quantitative PCR (qPCR) Analysis of Gene Expression-Different R. sphaeroides strains were grown in LB medium under aerobic conditions until the absorbance at 660 nm reached 0.2-0.3. Cells were then harvested by centrifugation, washed, and resuspended with Sistrom's minimal medium, then continued growth under aerobic conditions. After 20 min of starvation, freshly made heme (in 50% ethanol, 0.05 M NaHCO 3 ) was added into the cultures to a final concentration of 25 M, and 1-ml samples were then extracted and stored on ice at different time points. The RNA was purified with ISOLATE RNA mini kit (Bioline) and Turbo DNA-free (Ambion). Reverse transcription and real-time PCR were performed with Brilliant II SYBR Green QPCR Master Mix kit (Agilent Technologies) on either Quantitative PCR Stratagene MX3000P (Agilent Technologies) or StepOnePlus Real-Time PCR system (Applied Biosystems). The transcription level of rpoZ gene (encoding the RNA polymerase omega subunit) was used to normalize the levels of all other genes. At least three biological replications were measured on each condition, and qPCR was performed for each sample at least twice. For each biological replication, a mixture of all RNA samples was used to determine the efficiency of the primers with serial dilution methods. Primer sequences for qPCR analysis are shown in supplemental Table 2.

PpsR Binds Heme in Vivo and in Vitro-
We constructed a chromosomally encoded carboxyl-terminal FLAG-tagged PpsR construct in R. sphaeroides at its native chromosomal location by recombination. Affinity purification of FLAG-tagged PpsR yielded a colored preparation that was judged to be Ͼ95% pure based on SDS-PAGE chromatography. An absorption spectrum of the purified FLAG-tagged PpsR showed the presence of a distinct heme peak at 372 nm as well as additional broad peaks in the 440 -600-nm region ( Fig. 1A and supplemental Fig. 2A). We subsequently tested the ability of tagless apo-PpsR, purified from an E. coli overexpression system, to bind to heme-agarose using a pulldown assay. The results demonstrate that tagless PpsR is indeed capable of binding to heme-agarose in vitro (supplemental Fig. 2B).
We further investigated the ability of tagless PpsR to bind heme by monitoring changes in the absorption spectrum of heme upon addition of purified tagless apo-PpsR under oxidizing conditions (oxidized spectrum). The spectrum in Fig. 1B shows that upon the addition of apo-PpsR, the broad doublet Soret peak of free heme intensified and was converted to a single peak at 372 nm concomitant with peaks in the 500 -700-nm region of the spectrum that is virtually indistinguishable from absorbance characteristics of FLAG-tagged PpsR that was isolated from R. sphaeroides cells (Fig. 1A). Reduction of the heme-bound iron from Fe 3ϩ to Fe 2ϩ by the addition of dithionate resulted in a red shift of Soret peak from 372 nm to 426 nm as well as the appearance of ␣ and ␤ peaks in the 500 -600-nm region (Fig. 1C, reduced spectrum). Analysis under same reducing conditions but with no oxygen present (deoxy-PpsR-heme spectrum) shows that the 426-nm Soret peak and the ␣ and ␤ peaks in the 500 -600-nm region, all disappear (supplemental Fig. 2C). Adding carbon monoxide to reduced PpsR-heme species also resulted in a characteristic carbon monoxy heme spectrum with a Soret peak at 418 nm (supplemental Fig. 2D), typical for ferrous-heme-CO complex. Interestingly, myoglobin under low pH conditions demonstrates a UV-visible spectrum almost identical to that observed with PpsR-heme (25). A well studied heme-binding protein, myoglobin, under low pH conditions is pentacoordinated under oxidizing conditions and hexacoordinated when reduced (25). The similarity of the UV-visible spectrum suggests that PpsR-heme binding could utilize the same mechanism. The stoichiometry of heme binding to PpsR was addressed by reconstituting PpsR with heme in vitro and then removing the PpsR-heme complex from free heme either by size exclusion chromatography or by ion exchange chromatography. With these approaches, we could obtain PpsR-heme complex with a molar ratio of 0.85 Ϯ 0.08 PpsR to 1 heme. To confirm spectrally the chromatography results we titrated heme into PpsR and measured the absorbance change that occurs at 370 nm upon binding. The plot of the absorbance change versus the PpsR:heme concentration ratio indicates that a PpsR-heme complex is formed with a stoichiometry of 0.84 Ϯ 0.05 [PpsR]: [heme] (Fig. 1D). Tryptophan quenching data of heme titrated into PpsR can also be fitted with a one-to-one binding model with a K d for heme binding of 1.9 M (Fig. 1E). This micromolar binding affinity is similar to a dissociation constant that has been reported for heme binding by R-transferase (9).
His 275 and Cys 424 Are Critical for PpsR-Heme Interaction-Prior inspection of the PpsR sequence indicates the presence of two PAS domains followed by a carboxyl-terminal helix-turnhelix (HTH) DNA binding domain ( Fig. 2A) (26). It is well known that several PAS domains use heme as a co-factor to sense environmental signals (11,13), so several deletion constructs were made to test their involvement. Unlike full-length PpsR, spectral analysis of truncated constructs PpsR 256 and PpsR 376 did not show the formation of a single Soret peak upon addition of heme under oxidizing conditions (Fig. 2B). This type of spectrum change, the loss of the single 370-nm Soret peak and the appearance of two overlapping peaks, has been observed for regulatory proteins that have a HRM and assigned to the lost Cys-heme coordination (17). Under reducing conditions the larger construct that contained both PAS domains but no HTH domain (PpsR 376 in Fig. 2A) showed spectroscopic changes similar to those shown by wild-type PpsR with the exception that the intensity was not as increased (Fig. 2C). Thus, both the carboxyl terminus HTH domain and the second PAS domain appear to be involved in the interaction with heme.
The spectrum of the PpsR-heme complex is similar to that of several heme sensors that characteristically have a single Soret peak at ϳ370 nm under oxidizing conditions that also red shifts to ϳ420 nm under reducing conditions (2,9). For example, the heme-sensing protein Irr has a Soret peak at 377 nm under oxidizing conditions that red shifts to 413 nm under reduced conditions (2,27). The appearance of single Soret peak has been attributed to the coordination between axial residue and the iron in heme (17). In the case of Irr, His and Cys are utilized as axial ligands to iron of heme which are two of the most common axial residues found in heme-binding proteins (2,9,17,28,29). Inspection of the second PAS domain and carboxyl-terminus region indicates that there is only one His present at position 275 (His 275 ), and only one Cys (Cys 424 ) located within the HTH DNA binding domain. Consequently, we used site-directed mutagenesis to convert His 275 and Cys 424 to Ala (H275A and C424A, respectively) and then assayed what effect these mutations had on heme interaction. When PpsR H275A was assayed under oxidizing conditions the spectrum was similar to that observed with oxidized wild-type PpsR (Fig. 2, D and B, respectively). This was contrasted by spectrally assaying heme binding by PpsR H275A under reducing conditions where ferrous heme did not show a sharp and distinguish Soret peak at 426 nm and also exhibited a reduced ␤ peak at 530 nm (Fig. 2D). Analysis of heme binding with PpsR C424A under reducing conditions shows a similar spectrum as reduced PpsR H275A (Fig. 2E). However, under oxidizing conditions, the single Soret peak of heme observed with wild-type PpsR is converted to a double peak with PpsR C424A (Fig. 2E), very similar to the spectrum observed with PpsR 376 (Fig. 2B). In addition, the reconstitution of PpsR C424A with heme only achieved a molar ratio of ϳ0.38 ([heme]:[PpsR C424A ]), which is considerably less than that observed with wild-type PpsR, further confirming that Cys 424 is critical for PpsR-heme interaction. These observations, combined with the results presented earlier, can be explained with a model in which Cys 424 forms an axial ligand under oxidizing conditions where heme is pentacoordinated, whereas both Cys 424 and His 275 form axial ligands under reducing conditions where heme is hexacoordinated.
Heme Affects DNA Binding Activity of PpsR-Previous mutational and biochemical analyses demonstrated that the redox state of Cys 424 affects the DNA binding activity of PpsR (18,30). One could envision, therefore, that binding of heme to Cys 424 may also affect the DNA binding activity of PpsR. To explore this hypothesis, we performed gel mobility shift assays to the puc promoter that contains two PpsR binding sites located 8 bp apart with and without heme bound to PpsR. Incubation of PpsR with a 262-bp puc promoter probe resulted in a distinct shift as reported previously (18,20). The PpsR-DNA complex exhibited progressively slower electrophoretic mobility (supershift) as the concentration of PpsR increased (Fig. 3A, right  panel). Similar analysis with PpsR-heme shows the presence of the PpsR-DNA complexes. However, unlike PpsR alone, PpsRheme does not form a supershift as protein concentration increased (Fig. 3A, left panel). Analysis of the mobility of permuted DNA segments indicates no significant alteration in PpsR-mediated DNA bending with or without heme present. Thus, heme appears to inhibit the ability of PpsR to form a higher ordered PpsR-DNA complex (supplemental Fig. 3).
We also addressed the effect of increasing heme concentration to PpsR-puc complex. A ratio of [PpsR]:[heme] from 1:1 up to ϳ1:2 results in the disappearance of PpsR-DNA supershift, whereas a higher heme level completely releases PpsR from the puc DNA probe (supplemental Fig. 4A). Similar analysis with a hemE promoter probe, which contains two PpsR binding sites separated by 252 bp, showed a similar pattern except that the binding of PpsR was more sensitive to disruption by addition of heme than it was with the puc probe (supplemental Fig. 4B). Additional control assays with hemE also showed that neither protoporphyrin IX, zinc porphyrin, nor ferric ion affects PpsR-DNA complex mobility (supplemental Fig. 4C).
Changes in the mobility shift patterns of the PpsR-DNA complex observed with and without heme suggest that PpsR may interact differently with target promoter upon binding heme. To test this possibility, we performed DNase I footprint analysis of PpsR binding to puc promoter (Fig. 4). The protec- tion pattern exhibited by PpsR alone showed complete protection of the two PpsR binding sites (TGT-N 11 -ACA) that overlaps the Ϫ35 and Ϫ10 promoter recognition sequences, as well as two nucleic acids remaining DNase I-accessible (marked with * in Fig. 4) between the two binding sites. In the presence of heme, two previously protected nucleic acids were now DNase I-accessible (shaded peaks, Fig. 4). Another small but consistent change was also observed where the DNase I sensitivity of two adjacent nucleic acids is switched from "low-high" to "highlow" (marked with a dashed box in Fig. 4). In summary, the footprint assay confirms the different conformations of the PpsR-puc complexes with and without bound heme.
His 275 and Cys 424 Are Critical for Heme Regulation of PpsR-DNA Interaction-Because His 275 and Cys 424 appear to be critical for the interaction with heme, we examined the ability of isolated PpsR H275A and PpsR C424A to bind to a puc promoter segment containing a PpsR binding site. Under oxidizing conditions without heme, wild-type PpsR and PpsR H275A both form a similar supershifted complex, as protein is increased (Fig. 3, A and C, respectively). Similar results occur under reducing conditions without heme where a supershifted complex is observed with both wild-type PpsR and PpsR H275A as protein concentration is increased (Fig. 3, B and D, respectively). However, very different results are observed in the presence of heme. Under oxidizing conditions with heme we observed that wild-type PpsR and PpsR H275A are both incapable of forming the supershifted complex as protein is increased ( Fig. 3A and 3C, respectively). This is contrasted by reducing conditions with heme where wild-type PpsR is incapable of forming the supershifted complex (Fig. 3B, left panel), whereas the PpsR H275A mutant is still capable of forming the supershift complex (Fig. 3D, left panel).
Under oxidizing conditions, larger amounts of PpsR C424A were required to bind to the puc promoter compared with wildtype PpsR, which is in agreement with that of previous studies (18). Specifically, whereas 10 nM puc probe was completely shifted by the addition of 2.5 M wild-type PpsR (Fig. 3A), more than 20 M PpsR C424A was needed to bind to the same amount of puc probe (Fig. 3E). Furthermore, as observed with wild-type PpsR, a supershift with the puc probe was formed as the concentration of PpsR C424A was increased (Fig. 3E). When the same experiments were performed in the presence of heme, PpsR C424A was directly released from the puc promoter probe instead of exhibiting the relaxation of supershifted PpsR-puc complex as observed with wild-type PpsR (Fig. 3E). Similar DNA binding studies with PpsR C424A under reducing conditions were not feasible as this mutant is not stable under such conditions in vitro. In summary, these observations are consistent with the model we proposed earlier, where Cys 424 is critical for PpsR-heme interaction under oxidizing conditions, and both Cys 424 and His 275 are critical for PpsR-heme interaction under reducing conditions.
Heme Induces in Vivo Increase of Transcription of Subset of PpsR-controlled Genes-To explore how PpsR responds to the presence of heme in vivo, we used qPCR to measure the level of several genes known to be regulated by PpsR. Specifically, we tested expression of light harvesting represented by puc, heme/ bacteriochlorophyll common trunk represented by hemE, bacteriochlorophyll branch represented by bchC, and carotenoid biosynthesis represented by crtA ( Fig. 5A and supplemental Fig.  5). Interestingly, addition of heme to growing cultures of R. sphaeroides at 25 M induced different responses on the tested promoters (supplemental Fig. 6). After 20-min incubation with heme, the level of puc and bchC expression increased whereas the level of hemE and crtA did not show significant changes (Fig. 5B). A control experiment with the PpsR knock-out strain PPSR1 showed essentially the same level of all four genes tested, with and without heme (Fig. 5C), indicating that the change in the original strain observed by addition of heme was mediated through PpsR.

DISCUSSION
This study demonstrates that PpsR binds heme with an oxidized spectrum that is similar to several heme-sensing proteins such as Irr (2), HAP1 (17), and R-transferase (9). Indeed, the binding constant of PpsR-heme is close to that of the regulator R-transferase (9), but significantly weaker than hemoproteins that tightly bind heme to perform functions such as electron transport or transporting/sensing gas molecules (29). This is consistent with the fact that only a small fraction of bound heme is present when PpsR is harvested from R. sphaeroides (Fig. 1A). The oxidized PpsR-heme spectrum is almost identical to that of oxidized myoglobin (met-myoglobin) at low pH conditions, which is in a pentacoordinated state (25). Under reducing conditions, the PpsR-heme spectrum shares several characteristics with reduced myoglobin with CO (myoglobin-CO) at low pH conditions including the ␣ and ␤ peaks which is typical for hexacoordinated heme (25). Thus, we propose that PpsRheme has coordination states similar to those of myoglobin under low pH conditions. Myoglobin is in a U state under such conditions, representing a stable intermediate on the myoglobin unfolding pathway (25,31). This mechanism and the relatively low binding constant would suggest a flexible heme binding pocket, consistent with the proposed PpsR role as a heme-sensing protein. A flexible heme binding pocket has also been observed in IsdC (32) which is part of the heme uptake system in Staphylococcus aureus.
His 275 and Cys 424 are clearly important for PpsR to interact with heme. Our mutational analysis of His 275 shows a spectral alteration only under reducing conditions. This indicates that His 275 likely constitutes an axial ligand only to Fe 2ϩ and not when the iron is oxidized to Fe 3ϩ . This result is contrasted by an Ala substitution of Cys 424 that leads to a loss of the Soret peak under both oxidizing and reducing conditions. Because this type of spectrum change has been observed for regulatory proteins that have a HRM (17) and assigned to a Cys-heme coordination, we deduce that Cys 424 likely forms an axial ligand under both oxidizing and reducing conditions. This model is also supported by our gel mobility shift assays with the PpsR mutants. Further studies would be needed to examine this model on the structural level in the future.
Heme clearly affects the DNA binding activity of PpsR as demonstrated by the ability of heme to inhibit the formation of a higher ordered PpsR-puc supercomplex. At even higher concentrations of heme, PpsR is completely released from the puc promoter. Interestingly, Cys 424 is located within a well conserved HTH DNA binding domain (33). As show in previous study, oxidation of Cys 424 in the absence of heme stimulates the DNA binding activity of PpsR (18). The same effect of redox regulation was observed while heme was present (Fig. 3, A and  B). Thus, it appears that Cys 424 could be a versatile target for different types of modification, such as forming/breaking disulfide bridge (18), coordinating to heme as presented in this study, and potentially being oxidized into different oxidation states much like the case of OxyR (34). A three-dimensional model of the PpsR HTH domain based on homology to the Fis HTH domain structure (33) indicates that heme could potentially be accommodated by a cleft near Cys 424 (Fig. 6). If this is FIGURE 5. Heme effect on the transcription levels of PpsR-regulated genes. A, PpsR-regulated photosynthesis and tetrapyrrole biosynthesis genes. The detailed diagram is described in supplemental Fig. 5. B, changes in relative transcription level of various PpsR-regulated genes in wild-type R. sphaeroides culture after addition of heme to the culture. After 20 min of starvation, 25 M heme was added into the culture, and the samples were taken after another 20 min of incubation and normalized (supplemental Fig. 6). Data shown are mean Ϯ S.E. (error bars). C, similar to B with the exception that mRNA levels were analyzed in the ppsR deletion strain, PPSR1.
indeed the location of heme binding then inclusion of heme could affect the pitch of the protein helix resulting in the altered footprint pattern that was observed. Other types of modification of Cys 424 could also change the PpsR-DNA interaction correspondingly.
One common feature of heme-sensing proteins is that they use Cys as a heme axial ligand (29). One notable difference with PpsR is that Cys 424 is followed by an Ile whereas other heme sensors have a -Cys-Pro-as the HRM. The discovery of PpsR as a heme sensor with Cys-Ile instead of the Cys-Pro suggests that heme-sensing-based regulation may be more widespread. Nonetheless, PpsR utilizes cysteine as the critical axial residue as is the case for other HRM-containing heme sensors. Thus, it is feasible that Cys coordinated to Fe 2ϩ/3ϩ in heme is a conserved strategy for heme sensing.
Heme has a more obvious affect on puc expression than on hemE expression via PpsR (Fig. 5). However, in vitro gel shift assays show that heme affects PpsR binding to both puc and hemE promoters (supplemental Fig. 4). This suggests that different promoters respond to the presence of heme with different sensitivities. As shown in supplemental Fig. 4C, a PpsR: heme ratio of 1:1 caused complete release of PpsR from hemE promoter, whereas a ratio of 1:2 PpsR:heme can still produces a stable PpsR-puc complex (Fig. 3A). Meanwhile, puc and bchC genes which exhibit greater effects upon the addition of heme in vivo (Fig. 5) also have two PpsR recognition sites close to each other (7 and 8 bp apart, respectively) (supplemental Fig. 5). This is contrasted by crtA and hemE genes that have two PpsR recognition sites relatively far from each other (55 and 252 bp apart, respectively) that have no obvious effect of heme on expression in vivo (supplemental Fig. 5). This difference in the type of PpsR binding sites may be a reason for the observed differing heme sensitivity at different PpsR-regulated genes.
Why would PpsR have evolved as a heme-sensing transcription factor? In purple bacteria, all three branches of the tetrapyrrole biosynthetic pathway are essential for photosynthesis with bacteriochlorophyll needed for light absorption and electron donation, heme needed for electron transfer, and cobalamin (vitamin B 12 ) needed for synthesis of bacteriochlorophyll (35). The heme and bacteriochlorophyll branches share com-mon intermediates from ␦-aminolevulinic acid to protoporphyrin IX with both branches regulated by PpsR (19,20,36) (Fig. 5A and supplemental Fig. 5). In the case of the puc and bchC promoters, our in vivo results suggest that an excess of free heme can induce an increased synthesis of light harvesting II peptides and bacteriochlorophyll to which it binds. Interestingly, there is no significant increase in the level of hemE expression when heme is in excess. Thus, the interaction of PpsR with heme could potentially divert tetrapyrrole biosynthesis away from the production of excess toxic heme toward bacteriochlorophyll. This could be an advantage for photosynthetic bacteria that must coordinate the synthesis of heme with that of bacteriochlorophyll as well as with various tetrapyrrolebinding apoproteins. A similar strategy has been found in animal cells, which use heme as feedback signals between the circadian clock network and clock-controlled metabolic pathways to maintain a balance homeostasis (37). Finally, it has been shown that AppA, an antirepressor of PpsR, is also a hemebinding protein. Evidence suggests that heme bound to AppA can increase its affinity to PpsR (21), further underscoring the important role that heme has in directly and indirectly regulating the DNA binding activity of PpsR.