Characterization of Cyanobacterial Biliverdin Reductase CONVERSION OF BILIVERDIN TO BILIRUBIN IS IMPORTANT FOR NORMAL PHYCOBILIPROTEIN BIOSYNTHESIS*

The Synechocystis sp. PCC 6803 gene ( bvdR ) encoding biliverdin reductase was amplified by the polymerase chain reaction, cloned, and overexpressed in Esche- richia coli as the native form and as a 6-histidine-tagged amino-terminal fusion. The latter form of the enzyme was purified by affinity chromatography and shown to have the appropriate molecular weight by electrospray mass spectrometry. Both forms of the enzyme reduced biliverdin IX (cid:97) using NADPH or NADH, with NADPH as the preferred reductant. The His-tagged enzyme has a K m for biliverdin of 1.3 (cid:109) M . The pH optimum for the NADPH-dependent activity is 5.8, whereas that for rat biliverdin reductase is at pH 8.7. Absorbance spectra and high performance liquid chromatography retention times of the reaction product reaction match those of authentic bilirubin, the product of the reduction of biliverdin by the mammalian enzymes. These results provide the first evidence for the formation of bilirubin in bacteria. Fully segregated Synechocystis sp. PCC 6803 bvdR interposon mutants produce (cid:59) 85% of the normal amount of phycobilisome cores containing allophyco- cyanin and other phycocyanobilin-bearing core polypeptides, but no detectable phycocyanin. Thus, sur-prisingly, the blockage of the conversion of biliverdin to bilirubin interferes with normal phycobiliprotein biosynthesis in cyanobacteria. Possible interpretations of this finding are presented. centrifugation 12,000

Phycobiliproteins are the major light-harvesting proteins of cyanobacteria, red algae, and the cryptomonads (1-3). These proteins carry linear tetrapyrrole prosthetic groups (phycobilins) attached to the polypeptide through thioether bonds to cysteinyl residues (4). Biliverdin IX␣ (5), believed to be the common precursor to all of the phycobilins (6,7), is produced by the cleavage of heme at the IX␣ position by the ubiquitous enzyme heme oxygenase (8). The pathway of phycobilin biosynthesis beyond biliverdin has been characterized biochemically only in the atypical thermoacidophilic red alga, Cyanidium caldarium, by Beale and co-workers (9 -12), who used partially purified preparations of various enzyme activities. In Cyanidium, biliverdin is converted to 15,16-dihydrobiliverdin (11). This contrasts with the fate of biliverdin in mammals, where the enzyme biliverdin reductase uses electrons from NADPH to reduce the C 10 bridge in biliverdin to form bilirubin IX␣ (13). There are some indications that the pathway of phycobilin biosynthesis might be the same in cyanobacteria as in Cyanidium (14).
Recently, Tabata and co-workers have determined the sequence of the complete genome of the unicellular cyanobacterium Synechocystis sp. PCC 6803 (15). This presented the opportunity to ask if there were open reading frames (ORFs) 1 homologous to the mammalian biliverdin reductases from rat and human whose sequences are known. Such open reading frames might encode enzymes that catalyze reactions in the phycobilin biosynthetic pathway. At our request, Satoshi Tabata (Kazusa DNA Research Institute) generously examined the sequence of the Synechocystis sp. PCC 6803 genome for such ORFs and supplied us with both the protein and DNA sequences in advance of publication. One such ORF (slr1784; bvdR) bore significant resemblance to mammalian biliverdin reductases.
We cloned Synechocystis sp. PCC 6803 bvdR, overproduced the protein in Escherichia coli, and showed that BvdR it catalyzed the reduction of biliverdin to bilirubin under appropriate conditions. To our knowledge, this is the first demonstration of bilirubin formation in a bacterium. One might anticipate that inactivation of bvdR might lead to an increase in phycobiliprotein synthesis. Surprisingly, a fully segregated bvdR interposon mutant contains approximately 6-fold less protein-bound PCB than wild-type cells. This mutant produces smaller than normal amounts of phycobilisomes consisting only of core subassemblies that contain PCB-bearing polypeptides such as allophycocyanin and the core-membrane linker (L CM ) (16,17). These results suggest that in cyanobacteria the conversion of biliverdin to bilirubin may have a regulatory as well as a degradative function.

EXPERIMENTAL PROCEDURES
Materials-Restriction endonucleases and other enzymes for molecular biology applications were purchased from New England Biolabs, Inc. (Beverly, MA), Life Technologies, Inc., or Boehringer Mannheim. Biliverdin IX␣ was purchased from Porphyrin Products (Logan, UT). Other chemicals and reagents were purchased from Sigma.
Molecular Cloning of Biliverdin Reductase-Three oligonucleotides were used for the PCR amplification of the biliverdin reductase gene, bvdR, from Synechocystis sp. PCC 6803 chromosomal DNA using PCR with Taq polymerase (Life Technologies, Inc.). BvdR.1 (5Ј-CCCAAGGT-CATATGTCTGAAAATTTTGC-3Ј) was used along with bvdR.2 (5Ј-CGGGCTTCTGGCATGCTGGA-3Ј) for amplifying bvdR for expression purposes. The resulting 1.3-kb PCR product was digested with the restriction endonucleases NdeI and SphI and cloned into two vectors: pAED4 (a pET derivative with a T7 promoter that is a high copy number plasmid), creating pBR1, and pBS152v (histidine-tagged over-* This work was supported in part by National Institutes of Health Grant GM 28994 and by grants from the Lucille P. Markey Charitable Trust and the W. M. Keck Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  expression vector, derived from pPRO-EX-1 from Life Technologies, Inc.), creating pBR2 as shown. Oligonucleotides BvdR3 (5Ј-GAGGT-CAAGGGTACCGATGC-3Ј) and BvdR2 were used to amplify bvdR for the purposes of generating an interposon insertion mutant. This 2.6-kb PCR product was digested with KpnI and SphI prior to cloning in pUC19 digested with the same restriction enzymes creating the plasmid pBR3. This clone was digested with HpaI and ligated to the bluntended aphII interposon called C.K3 (18). Both orientations of the interposon insertion were isolated and named pBR3.1 and pBR3.2 as shown in Fig. 2 (B and C).
Cell Growth Conditions-E. coli was cultured in LB medium as described previously (19). Strain DH5␣ was used for all molecular cloning and for overexpression of bvdR using pBR2 (His-tagged form of BvdR). Plasmid pBR1 was transformed into E. coli strain BL21 DE3 cells for overexpression purposes. All overexpression of bvdR was done at 30°C. Overnight cultures were diluted 1:100 in 1-liter flasks of LB. These cultures were grown for 3 h prior to induction with 0.5 mM isopropyl-␤-D-thiogalactoside. After 3 h of induction, cells were harvested by centrifugation and frozen at Ϫ20°C.
Purification of Recombinant Biliverdin Reductase-For purification of BvdR from E. coli, cells were thawed, resuspended in 50 mM Tris-HCl, pH 8, 10 mM ␤-mercaptoethanol, and broken by passage through a French pressure cell at 20,000 p.s.i. After ultracentrifugation of the cell extract at 100,000 ϫ g in a Ti60 rotor for 45 min, the His-tagged form of BvdR was purified by passage of the 100,000 ϫ g supernatant over a nickel-nitrilotriacetic acid-Superflow-affinity column (Qiagen, Inc., Chatsworth, CA) containing 5 ml of resin. The column was washed with buffer A (20 mM Tris-HCl, pH 8, 100 mM KCl, 20 mM imidazole, 10 mM ␤-mercaptoethanol), followed by 2 volumes of buffer B (20 mM Tris-Cl, pH 8, 1 M KCl) and then by 2 volumes of buffer A. BvdR was eluted with buffer C (20 mM Tris-Cl, pH 8, 100 mM KCl, 200 mM imidazole, 10 mM ␤-mercaptoethanol). The protein was dialyzed against 100 mM KPO 4 buffer, pH 7.4, 10% glycerol and stored in aliquots at Ϫ70°C until further use. Protein concentration was calculated using an ⑀ 280 of 42,440 M Ϫ1 cm Ϫ1 , calculated from the Trp and Tyr content.
The native form of BvdR was partially purified as follows. Cells from a 2-liter induction experiment were broken as described immediately above in 100 mM KPO 4 , pH 7.4. Inclusion bodies and unbroken cells were separated by centrifugation at 12,000 ϫ g for 20 min. Glycerol was added to 10% (w/v) to the supernatant. This was loaded onto a 3-ml NADP-agarose affinity column (Sigma) equilibrated in 90 mM KPO 4 , pH 7.4, 10% glycerol. After washing with 10 column volumes of the same buffer, proteins were eluted with this buffer plus 1 mM NADPH. Fractions were collected and tested for BvdR activity. BvdR was approximately 50% pure after this step.
Mass Spectrometry-The His-tagged form of BvdR was subjected to electrospray ionization mass spectrometry. The protein (0.5 mg) was dialyzed extensively against 10 mM ammonium acetate, pH 7, prior to lyophilization. The protein was redissolved in 50% acetonitrile in 0.2% aqueous formic acid before being injected into a VG Bio-Q mass spectrometer (21).
DNA Methods-Chromosomal DNA was purified from Synechocystis sp. PCC 6803 cells grown in 25-50-ml cultures according to Ref. 19 (and references therein) with the following addition. Polysaccharides were extracted by the addition of 600 l of chloroform, 100 l of 5 M NaCl, 100 l of 10% mixed alkyltrimethylammonium bromide, 0.7 M NaCl heated at 65°C to dissolve. The DNA from the aqueous phase was precipitated by the addition of an equal volume of isopropanol.
For Southern blot analyses, chromosomal DNAs were subjected to restriction endonuclease digestion, separated by agarose gel electrophoresis, and transferred to a nylon membrane (Boehringer Mannheim). The bvdR probe was labeled using digoxigenin-dUTP in a random priming kit purchased from Boehringer Mannheim. Nonradioactive DNA detection was performed following the protocol from Boehringer Mannheim.
Generation of Synechocystis sp. PCC 6803 bvdR Mutants-Transformation of Synechocystis sp. PCC 6803 using pBR3.1 and pBR3.2 was performed according to Ref. 22. Transformants were selected on Nylon filters (Nucleopore, Inc.) on BG11 plates supplemented with Hepes and kanamycin and transferred to liquid medium containing kanamycin and glucose to promote segregation of alleles. Individual colonies were selected and tested by whole cell PCR analysis to determine if segrega-tion had occurred (23). Two transformants denoted br3.2A and br3.2G were selected for further study. Complete segregation of these two transformants was verified by Southern blot analyses.
Biliverdin Reductase Assays-Biliverdin or bilirubin were dissolved in a small amount of 0.1 N NaOH, and diluted into 0.1 M KPO 4 , pH 7.4. The concentration of biliverdin was determined in 36% HCl in methanol using the ⑀ 696 value of 30,800 M Ϫ1 cm Ϫ1 (5). The ⑀ 450 value of 53,000 M Ϫ1 cm Ϫ1 was used to determine the bilirubin concentration (24). Stock enzyme solution was prepared in a 0.1 M KPO 4 buffer, pH 7.4, 10% glycerol. The addition of the enzyme raised the pH of the assay mixture (see below) to 5.8. Optimal reaction conditions were as follows: 10% glycerol, 0.2 mg ml Ϫ1 bovine serum albumin, 0.1 M citrate buffer, pH 5.1 (final pH of the reaction mixture after enzyme and biliverdin addition was 5.8), 12.25 g of BvdR (0.31 M), 100 M NADPH, 10 -20 M biliverdin. The total volume was 1 ml, and the reaction was initiated by the addition of NADPH. The reaction was monitored spectrophotometrically at 450 nm. NADH-mediated BvdR activity was measured using 1.0 mM NADH in place of the NADPH. Lineweaver-Burk reciprocal plots were used to calculate K m values for BvdR.
For PCB reduction assays, 1-30 M PCB was used with 100 M NADPH under conditions similar to the ones for biliverdin described above. PCB was purified as described previously (25), resuspended in 100% dimethyl sulfoxide, and its concentration was determined using the ⑀ 680 value of 37,900 M Ϫ1 cm Ϫ1 (26). The disappearance of PCB was followed at 649 nm, and the spectral change was used as a measure of the amount of product formed based upon the assumption that there was only one product of this reaction. The ⑀ 649 for PCB at pH 5.8 was calculated to be 19,650 M Ϫ1 cm Ϫ1 .
HPLC Analyses-Analysis of products formed upon the reduction of biliverdin catalyzed by recombinant Synechocystis sp. PCC 6803 Histagged BvdR was performed by HPLC on a reverse phase C 18 column (4.6 ϫ 250 mm). A 1-ml reaction mixture containing 12.25 g of Histagged BvdR, 40 M biliverdin, 0.1 M citrate buffer, pH 5.1 (final pH 5.8; see above), 10% glycerol, 0.2 mg ml Ϫ1 bovine serum albumin, 100 M NADPH was incubated at room temperature for ϳ2 min. This mixture was then immediately diluted with 5 ml of 0.1% trifluoroacetic acid prior to loading on a reverse-phase C 18 cartridge. For controls, 20 M biliverdin or bilirubin in 1 ml of 0.1 M citrate buffer, pH 5.8, was diluted 5-fold with 0.1% trifluoroacetic acid, loaded onto a C 18 cartridge and treated in the same way as the BvdR mixture. The column was washed with 0.1% trifluoroacetic acid and eluted with solvent B acetone/ethanol/water/acetic acid, 50:38:11:1 (v/v). 800 l of the eluate (2 ml) was mixed with 800 l of water and then injected onto an analytical C 18 column (250 ϫ 10 mm). The solvent systems used were water (solvent A) and solvent B. The mixture was loaded at a concentration of 50% solvent B, 50% water for 2 min. A linear gradient was then applied from 50 to 100% solvent B over 30 min followed by 100% solvent B for 4 min.
Phycobilisome Preparation-Phycobilisomes were isolated from 1.6 liters of cultures grown under high light and bubbled with 5% CO 2 in air. Cells were harvested and phycobilisomes were prepared according to Ref. 27. The concentration of protein-bound PCB in 8 M urea, pH 1.9, was calculated using ⑀ 660 35.4 mM Ϫ1 cm Ϫ1 (28). The number of nanomoles of protein-bound PCB was divided by the wet weight of the cells used for the phycobilisome preparation. Fluorescence emission spectra of phycobilisome samples were determined in 0.65 M K/NaPO 4 , pH 8, with a Perkin-Elmer MPF-44B fluorescence spectrophotometer. For phycobilisomes from wild-type cells excitation was at 590 nm and at 600 nm for the core fraction from mutant cells. All data were collected with excitation and emission with slit widths set at 6 nm.
Flow Cytometric Studies-Synechocystis sp. PCC 6803 wild-type, br3.2G, and pseudorevertant br3.2A.1 were cultured in BG11 with 5 mM glucose under high light (80 microeinstein m Ϫ2 s Ϫ1 ) with shaking. Exponentially growing cells were used immediately for flow cytometry experiments using a Coulter Epics Elite flow cytometer. A 633-nm laser was used for excitation, and the emission was collected through a 665-nm bandpass filter. Dead cells were excluded by forward and side scattering. Data from 50,000 cells were collected and used for statistical analyses. Fig. 1 compares the amino acid sequences of the three known BvdRs. The enzyme from Synechocystis sp. PCC 6803 is 21.8% identical to the rat enzyme and 20.7% identical to the human enzyme with the region of residues 92-134 being the most highly conserved. This region contains three conserved His residues that may be impor-tant in substrate binding. Site-directed mutagenesis studies implicated three cysteine residues in substrate binding and catalysis for rat BvdR (32). However, the cyanobacterial enzyme contains only one of the three cysteines present in the mammalian proteins, Cys-323 in the Synechocystis sp. PCC 6803 sequence.

Comparison of the Sequences of the Cyanobacterial and Mammalian Biliverdin Reductases-
Description of Genes Adjacent to bvdR- Fig. 2A shows a diagram of genes found near bvdR in Synechocystis sp. PCC 6803 (15). An ORF (designated Reg) showing similarity to various regulatory components of sensory transduction systems is directly upstream of bvdR. It is very likely that these two genes are co-transcribed because of their proximity. Directly downstream of bvdR and on the opposite strand of DNA is a putative operon containing open reading frames with similarity to spore maturation proteins A and B and MalQ (4-␣-glucanotransferase).
Cloning of Biliverdin Reductase-The sequence of the bvdR gene from Synechocystis sp. PCC 6803 was kindly provided by Satoshi Tabata's group at the Kazusa DNA Research Institute (Chiba, Japan). We used this information to design three oligonucleotides for use in PCR reactions with Synechocystis sp. PCC 6803 chromosomal DNA. A 1.3-kb fragment was amplified and cloned into two vectors for overexpression: pAED4 and pBS152v, creating pBR1 and pBR2, respectively (see "Experimental Procedures" for details).
A 2.6-kb fragment containing sequences flanking bvdR was amplified from chromosomal DNA, digested with KpnI and SphI, and cloned into pUC19 to create pBR3. This construction was used to generate an interposon insertion mutant, described later.
Overexpression of Biliverdin Reductase in E. coli-pBR1 was used to transform the expression cell line BL21 DE3. Both the native form (from pBR1) and the His-tagged form (from pBR2) of BvdR were successfully overproduced in E. coli. The Histagged form was purified using a nickel affinity column. The protein was approximately 90 -95% pure after this one step, as judged by SDS-polyacrylamide gel electrophoresis, and was subjected to electrospray mass spectrometry to verify its mass. The mass of the major peak was 39,392.6 mass units, 129 mass units less than the calculated mass of 39,521.6 for His-tagged BvdR. The most likely explanation for this difference in mass is that the NH 2 -terminal methionine (131 mass units) was re-moved in E. coli, a very common occurrence. The loss of the NH 2 -terminal methionine would bring the measured mass within two mass units of the calculated mass.
Biliverdin Reductase Activity-Both the recombinant native and His-tagged BvdR were active in E. coli whole cell extracts. The absorption spectrum of the product of the biliverdin reduction reaction has a peak at approximately 450 nm and corresponds to that of bilirubin IX␣ (data not shown). Additions of 10% glycerol and 0.2 mg ml Ϫ1 of bovine serum albumin increased the activity of purified His-tagged BvdR and were used in all subsequent reactions. The pH activity profile for Histagged BvdR shows an optimum at pH 5.8 for both the NADPHdependent and NADH-dependent activities (Fig. 3). The native form of BvdR had a similar pH activity profile (data not shown). BvdR shows a marked preference for NADPH as reductant over NADH. The K m value for biliverdin was 1.3 M. The enzyme used PCB as a substrate, but the K m for PCB was at least 10-fold higher than that for biliverdin (data not shown).
Identification of the Reaction Product by HPLC-A reaction mixture containing 20 M biliverdin, 100 M NADPH, 0.2 mg ml Ϫ1 bovine serum albumin, 10% glycerol, 0.1 M citrate buffer, pH 5.8, and 12.25 g of His-tagged BvdR was incubated for 2 min at room temperature. Biliverdin IX␣ and bilirubin IX␣ control mixtures (lacking only His-tagged BvdR) were treated similarly prior to loading onto a C 18 reverse phase column. The elution profiles were monitored at 370, 680, and 450 nm (Fig.  4). Biliverdin elutes at ϳ9.8 min (Fig. 4A), whereas bilirubin elutes at 29.5 min (Fig. 4C). The reaction mixture elution profile is shown in Fig. 4B. The retention time of the bilirubin control and of the biliverdin reduction product are identical.
Generation of bvdR Interposon Mutant-An interposon cartridge encoding the aphII gene that confers kanamycin resistance was inserted into the HpaI site of the bvdR gene in pBR3. Both orientations of the cartridge were isolated as shown in Fig. 2 (B and C). Synechocystis sp. PCC 6803 cells were trans- formed with both interposon constructions. Because cyanobacteria are known to carry multiple copies of their genome, kanamycin-resistant transformants were repeatedly subcultured and plated to select for clones that had segregated. The clones that had been transformed with pBR3.1 did not segregate. This may be because the aphII cartridge is interfering with the transcription of the upstream ORF that resembles regulatory components of sensory transduction systems, which may serve a vital function for the cell. However, two transformants generated using pBR3.2 did segregate. This construction contains the aphII cartridge cloned such that the direction of transcription of aphII is the same as that of bvdR and should allow for normal transcription/translation of the regulatory component upstream. Complete segregation was verified by Southern blot analysis as shown in Fig. 5. Chromosomal DNA that had been purified from several of these transformants was digested with HindIII, separated by agarose gel electrophoresis, transferred to a nylon membrane, and probed with a 1.3-kb fragment containing the entire bvdR gene including sequences downstream of the SphI site. The wild-type DNA (Fig. 5, lanes 3 and  6) contains two fragments that hybridize to the probe, 1.6 (corresponds to the HindIII fragment containing most of the bvdR gene and some upstream sequence) and 4.6 kb (corresponds to the 3Ј end of the gene and downstream sequence; see Fig. 2C). The DNA isolated from two transformants denoted br3.1A and br3.1B (generated with pBR3.1; see Fig. 2B) contains three fragments that hybridize to the probe: 4.6, 2.7 (corresponds to the mutant copy of the bvdR gene that contains the 1.1-kb interposon inserted at the HpaI site), and 1.6 kb (Fig. 5, lanes 1 and 2). These transformants have retained a wild-type copy of the bvdR gene (1.6 kb). However, transformants br3.2G and br3.2A (Fig. 5, lanes 4 and 5) contain no 1.6-kb wild-type copy of bvdR and were used for further study. Pseudorevertants of br3.2A were isolated as colonies that appeared more blue. There is no 1.6-kb wild-type fragment in lanes 7 and 8 of Fig. 5, indicating that the pseudorevertants had not recovered wild-type copies of bvdR.
Characterization of bvdR Interposon Mutants-Both isolates br3.2A and br3.2G were yellow-green in color, indicating a much lower abundance of phycobiliprotein in these cells. Phycobilisomes (and phycobilisome subassemblies) were purified from the br3.2G mutant and from the wild-type strain. The sucrose gradient of the wild-type phycobilisome preparation contained a single band at the position expected for a complete phycobilisome. The absorption and fluorescence spectra of this fraction are shown in Fig. 6A. However, the sucrose gradient with the extract from the mutant contained one faint band at the position where core subassemblies normally migrate (16). The absorption and fluorescence spectra of the latter fraction (Fig. 6B) are very similar to those of core subassemblies (16,17). The amount of phycobiliprotein (measured as total phycobiliprotein-bound PCB; see "Experimental Procedures") isolated from the mutant cells was 5.6-fold less than that isolated from wild-type cells. SDS-polyacrylamide gel electrophoresis analyses of this fraction show that the ␣ and ␤ subunits of allophycocyanin and the large core-membrane linker polypeptide carry PCB and have the same mobilities as the corresponding components from phycobilisomes of wild type (data not shown). No phycocyanin was detected in the sample from the mutant cells. Both bvdR mutant isolates showed high frequency of pseudoreversion over time. These pseudorevertant colonies appeared more blue and were chosen for comparison by flow cytometry to the br3.2G mutant and to the wild-type cells. Approximately 50,000 cells of the br3.2G mutant, the wild-type strain, and of a pseudorevertant (denoted br3.2A.1) were subjected to fluorescence analyses (excitation at 633 nm, emission at 670 nm) to determine if there was a uniform population of cells with equivalent amounts of fluorescence for each cell type and to compare the total fluorescence from wild-type cells to the total fluorescence from cells of the br3.2G mutant and the br3.2A.1 pseudorevertant. Fig. 7 shows the data from this experiment as events (cells examined) versus amount of fluorescence at 670 nm. The wild-type strain (Fig. 7A) and the br3.2G mutant (Fig. 7B) show uniform populations with cells from within each type exhibiting similar levels of fluorescence when compared with other cells from that same type. However, the br3.2A.1 revertant fluorescence data (Fig. 7C) do contain a broader peak with a tail, indicating that the population is not entirely uniform. When the br3.2A.1 fluorescence levels are compared with those of cells from the other two groups, the wild-type cells are approximately 10-fold more fluorescent than the br3.2G mutant cells. The pseudorevertant br3.2A.1 cells are approximately 3.7 times as fluorescent as the br3.2G mutant cells but 2.7-fold less fluorescent than the wild type. DISCUSSION Cyanobacterial phycobiliproteins, depending on the type of chromoprotein, carry one or more of four different covalently attached isomeric phycobilins (phycocyanobilin, phycoerythrobilin, phycobiliviolin, and phycourobilin) (28). Heme is the biosynthetic precursor of the phycobilins (33-36), with biliverdin IX␣ as the initial product of heme cleavage in the pathway (6,7,36). Phycobilins are bound to phycobiliproteins through thioether bonds at specific cysteine residues, but only two phycobiliprotein phycobilin lyases have been characterized: C-phycocyanin ␣ subunit phycocyanobilin lyase (37)(38)(39)(40) and phycoerythrocyanin ␣ subunit phycobiliviolin lyase (41). The cyanobacterial genes encoding the enzymes of the phycobilin biosynthetic pathway are unknown. The determination of the complete sequence of the genome of Synechocystis sp. PCC 6803 (15) offers opportunities to explore in a directed way, by exploiting DNA sequence similarities as clues, the functions of candidate genes for enzymes involved in phycobilin metabolism. Because the phycobiliproteins of Synechocystis sp. PCC 6803 contain only PCB, this organism is one of the more "simple" ones in which to study bilin biosynthesis. Here, we describe the characterization of the function of Synechocystis sp. PCC 6803 gene bvdR, which shows 20 -22% identity to rat and human biliverdin reductases. We investigated this gene primarily because of the possibility that it encoded a reductase that catalyzes the conversion of biliverdin IX␣ to 15,16-dihydrobiliverdin, an intermediate in phycobilin biosynthesis in the red alga C. caldarium (11). The data presented here establish that Synechocystis sp. PCC 6803 bvdR encodes a biliverdin reductase that like its mammalian counterparts converts biliverdin IX␣ to bilirubin. As far as we are aware this is the first report of the characterization of a cyanobacterial biliverdin reductase and of the formation of bilirubin in a bacterium.
The cyanobacterial enzyme contains only one (Cys-323) of the three conserved cysteine residues thought to be important in catalysis and substrate binding in mammalian BvdRs (32). A mechanism has been proposed for the action of BvdR on biliverdin (42). A lysine residue near a cysteine that is essential for catalytic activity was proposed to protonate N 23 (near the C 10 bridge) as part of this mechanism. There are only two lysines that appear to be conserved among all three proteins (Lys-121 and Lys-146 of the cyanobacterial enzyme). If the proposed mechanism of action for BvdR proves correct, then it would seem probable that one of these lysines is near Cys-323 in the tertiary structure of the enzyme.
BvdR was successfully overproduced in E. coli, both as the native form of the enzyme and as the His-tagged form. Both forms of the enzyme were active, reducing biliverdin in the presence of NADPH or NADH to form a species with an absorbance spectrum and retention time on reverse phase HPLC identical to those of bilirubin. The cyanobacterial enzyme exhibited a distinctive pH activity profile. The pH optimum for both the NADPH-and NADH-dependent enzyme activities was at 5.8. The rat enzyme shows pH optima at 8.7 for NADPH-and 6.75 for NADH-dependent activities (24). This marked difference in pH activity profiles between the rat and the cyanobacterial BvdRs suggests that the latter enzyme may be in a more acidic cell compartment such as the intrathylakoidal space. Rat BvdR has been shown to reduce bilin substrates other than FIG. 8. Proposed pathway of phycobilin biosynthesis in the red alga C. caldarium (4, 7-10) believed to be present also in cyanobacteria (14) with the addition of the conversion of biliverdin to bilirubin catalyzed by the cyanobacterial BvdR (this study). The dotted arrows indicate reactions that have not been conclusively demonstrated in C. caldarium whole cell extracts. biliverdin, such as PCB, phytochromobilin, and phycoerythrobilin (43). The recombinant cyanobacterial BvdR reduced PCB to produce a product with a max at 429 nm (data not shown), similar to the bilirubinoid products produced by the rat BvdR enzyme (43). The biological significance of the latter transformation is unknown. However, the affinity of the enzyme for PCB is at least 10-fold lower than the its affinity for biliverdin (data not shown), implying that the primary role of BvdR in cyanobacterial cells is in reducing biliverdin. The rat enzyme has similar K m values for biliverdin, PCB, and phytochromobilin (43).
In mammals, the first step in the heme catabolic pathway is the heme oxygenase-catalyzed conversion of heme to biliverdin; the second is the BvdR catalyzed conversion of biliverdin to bilirubin. It seems likely that BvdR in cyanobacteria plays a catabolic role in cyanobacteria also, but our data indicate that it may play a regulatory role as well. We explored the role of BvdR in cyanobacteria by generating an interposon insertion mutant. If the role of the enzyme is purely catabolic, such a mutant would be expected to show either normal or perhaps somewhat elevated amounts of phycobiliproteins and/or free biliverdin. If the cyanobacterial enzyme did play an essential role in phycobilin biosynthesis, then the mutant would exhibit a phycobilin-minus phenotype. The actual phenotype differed profoundly from either of these alternatives.
Generation of a fully segregated bvdR mutant was successful with one of the constructions used. These mutants produced only one-sixth of the amount of protein-bound PCB relative to wild-type cells. Moreover, no complete phycobilisomes were produced. The bvdR mutant produced phycobilisome core subassemblies in about 84% of the amount produced by wild-type cells grown under high light (estimated from the allophycocyanin content) but did not produce detectable amounts of the phycocyanin-containing rod substructures. Almost no phycocyanin could be detected in whole cell extracts from the bvdR mutant using antibodies against Synechococcus sp. PCC 7002 C-phycocyanin ␤-subunit (data not shown). The cores produced by the bvdR mutant contain normal PCB-bearing polypeptides, such as allophycocyanin ␣ and ␤ subunits, and have the absorbance and fluorescence properties characteristic of such subassemblies (16,17). Moreover, the flow cytometry results (Fig. 7B) show that the residual level of expression of phycobiliproteins is uniform in the interposon mutant cell population.
Thus, the inactivation of bvdR suppresses but does not prevent the synthesis of PCB and phycobiliproteins. How can one explain the apparently selective blockage of phycocyanin synthesis? It is reasonable to suggest that elimination of the conversion of biliverdin to bilirubin would lead to an increase in the cellular level of biliverdin. Biliverdin has been shown to form a high affinity complex with apophycocyanin (25) and apophycoerythrin (44) subunits in vitro. However, in such complexes, the biliverdin is not covalently bound to cysteine residues (25,44). The relative affinities of PCB and biliverdin for apophycocyanin subunits have not been quantitated. If PCB is bound much more strongly to apo-allophycocyanin subunits and other PCB-bearing core polypeptides than it is to the apophycocyanin subunits, then this would account for the observed results. In cpcEF mutants (mutants in phycocyanin ␣ subunit PCB lyase), apophycocyanin subunits are rapidly degraded (37,38,40).
Biliverdin is a branch point between heme catabolism and phycobilin biosynthesis (Fig. 8), undergoing either transformation to 15,16-dihydrobiliverdin and thence to PCB or conversion to bilirubin. Control of the fate of biliverdin is a plausible site for metabolic regulation. Consequently, an alternative possibility to be explored is that bilirubin directly modulates the flux of biliverdin into the phycobilin biosynthetic pathway and/or influences the transcription of phycocyanin structural genes. Heme and heme precursors but not biliverdin have been shown to be regulatory factors for the transcription of phycobiliprotein structural genes in the chloroplasts of the red alga C. caldarium (45). Recently, it was suggested that human BvdR contains a zinc finger and therefore may be involved in transcriptional regulation as well as its well documented role in heme catabolism (46). However, the conserved sequence for a zinc finger is not present in the rat or the cyanobacterial enzyme.
We have isolated pseudorevertants able to produce more phycobiliprotein than the mutant but less than wild type. Study of these pseudorevertants may offer insights into the role of the BvdR-catalyzed reaction. If a phycobilin biosynthetic enzyme, normally inhibited by high biliverdin concentrations, becomes less sensitive to biliverdin by a mutation in the gene that encodes it, then this might explain how the pseudorevertants have regained some of their ability to produce phycocyanin.
Finally, in mammals, in addition to being an excretable end product of heme catabolism, bilirubin is a physiologically relevant scavenger of reactive oxygen species (47)(48)(49). It is possible that this role has ancient evolutionary origins and that bilirubin plays a significant protective role in cyanobacteria. Perhaps the lack of phycocyanin in these mutants, in response to the absence of bilirubin, is an attempt to lower their absorbance cross-section for photosynthesis to slow electron transport and therefore produce less reduced ferredoxin, a known source of reactive oxygen species. Pseudorevertants may be producing large amounts of superoxide dismutase to compensate for the loss of bilirubin. The bvdR mutants and pseudorevertants we have isolated allow the exploration of this possibility.