Insights into Phycoerythrobilin Biosynthesis Point toward Metabolic Channeling*

Phycoerythrobilin is a linear tetrapyrrole molecule found in cyanobacteria, red algae, and cryptomonads. Together with other bilins such as phycocyanobilin it serves as a light-harvesting pigment in the photosynthetic light-harvesting structures of cyanobacteria called phycobilisomes. The biosynthesis of both pigments starts with the cleavage of heme by heme oxygenases to yield biliverdin IXα, which is further reduced at specific positions by ferredoxin-dependent bilin reductases (FDBRs), a new family of radical enzymes. The biosynthesis of phycoerythrobilin requires two subsequent two-electron reductions, each step being catalyzed by one FDBR. This is in contrast to the biosynthesis of phycocyanobilin, where the FDBR phycocyanobilin: ferredoxin oxidoreductase (PcyA) catalyzes a four-electron reduction. The first reaction in phycoerythrobilin biosynthesis is the reduction of the 15,16-double bond of biliverdin IXα by 15,16-dihydrobiliverdin:ferredoxin oxidoreductase (PebA). This reaction reduces the conjugated π -electron system thereby blue-shifting the absorbance properties of the linear tetrapyrrole. The second FDBR, phycoerythrobilin:ferredoxin oxidoreductase (PebB), then reduces the A-ring 2,3,31,32-diene structure of 15,16-dihydrobiliverdin to yield phycoerythrobilin. Both FDBRs from the limnic filamentous cyanobacterium Fremyella diplosiphon and the marine cyanobacterium Synechococcus sp. WH8020 were recombinantly produced in Escherichia coli and purified, and their enzymatic activities were determined. By using various natural bilins, the substrate specificity of each FDBR was established, revealing conformational preconditions for their unique specificity. Preparation of the semi-reduced intermediate, 15,16-dihydrobiliverdin, enabled us to perform steady state binding experiments indicating distinct spectroscopic and fluorescent properties of enzyme·bilin complexes. A combination of substrate/product binding analyses and gel permeation chromatography revealed evidence for metabolic channeling.

Phycobilins are linear tetrapyrrole molecules that are important cofactors of the photoreceptor phytochrome and the cyanobacterial light-harvesting phycobiliproteins. One of the major pigments found in the phycobilisomes of certain cyanobacteria, red algae, and cryptomonads is phycoerythrobilin (PEB). 3 In these organisms PEB is covalently linked to the phycobiliprotein phycoerythrin (PE), a major constituent of the light-harvesting structures called phycobilisomes. These structures allow the organisms to efficiently absorb light in regions of the visible spectrum that are poorly covered by chlorophylls. Through resonance energy transfer the absorbed light energy is transferred to the photosynthetic reaction centers in the membrane. Freshwater cyanobacteria of the genus Calothrix (Fremyella) are able to adapt their phycobiliprotein composition within the phycobilisome in response to different light conditions. In a process called complementary chromatic adaptation the organisms are able to adjust the quantities of phycocyanin and PE for maximal light harvesting efficiency. Not only is the synthesis of apophycobiliproteins and linker proteins regulated by light, but also the biosynthesis of the enzymes required for PEB synthesis. It has been demonstrated that the expression of the genes pebA and pebB encoding ferredoxin-dependent bilin reductases (FDBRs) in Fremyella diplosiphon (Calothrix or Tolypothrix sp. PCC 7601) is up-regulated by green light, as is the expression of the cpeBA genes encoding the ␣and ␤-subunits of PE (1).
In a similar manner, marine cyanobacteria of the Synechococcus group are able to regulate phycourobilin to PEB ratios by adjusting the expression of phycoerythrins with different phycourobilin content, PE(I) and PE(II) (2), or, as recently suggested, by lyases that mediate PEB isomerization on the phycobiliproteins (3). Because these organisms retain a fixed phycocyanin:PE ratio (4), they are not considered as classical chromatic adapters.
The biosynthesis of phycobilins proceeds via the heme biosynthetic pathway. The final product, heme, is cleaved by heme oxygenases to yield biliverdin IX␣ (BV), which is subsequently further reduced by a family of FDBRs (Fig. 1). These enzymes are characterized by a distinct double bond regiospecificity resulting in bilins with a wide variety of spectroscopic properties. Synthesis of phytochromobilin (P⌽B), the chromophore of plant phytochromes, is catalyzed by phytochromobilin syn-thase (HY2) through a formal two-electron reduction at the A-ring 2,3,3 1 ,3 2 -diene structure. Phycocyanobilin:ferredoxin oxidoreductase (PcyA) catalyzes the four-electron reduction of BV to phycocyanobilin (PCB), the chromophore of certain cyanobacterial phytochromes and one of the major light-harvesting pigments in cyanobacterial phycobilisomes. PcyA is the best described member of the FDBR family. It mediates two subsequent electron reductions at both vinyl groups of BV (5). In this reaction 18 1 ,18 2 -dihydrobiliverdin is a visible semi-reduced intermediate. Because no metal or organic cofactors could be detected in the FDBR family, a radical mechanism for PcyA was postulated (5). Evidence for the appearance of an intermediate substrate radical was recently demonstrated by absorbance and EPR spectroscopy (6). Structural information to this new family of enzymes has recently been added through the solved crystal structure of the Synechocystis sp. PCC 6803 PcyA (7).
In contrast to the PcyA-catalyzed reaction the biosynthesis of PEB (which is an isomer of PCB) requires two independent enzymes, each catalyzing a two-electron reduction. 15,16-Dihydrobiliverdin:ferredoxin oxidoreductase (PebA) reduces the C-15 methine bridge of BV and phycoerythrobilin:ferredoxin oxidoreductase (PebB) the A-ring diene structure of 15,16-dihydrobiliverdin (DHBV), respectively. The biosynthesis of phycourobilin still remains unknown, but it might proceed analo-gously to the PecE/F isomerase/lyase activity of Mastigocladus laminosus, which covalently attaches and isomerizes PCB to yield bound phycoviolobilin (8).
Here we present the biochemical characterization of recombinant PebA and PebB from the filamentous freshwater cyanobacterium, F. diplosiphon, and the unicellular marine cyanobacterium Synechococcus sp. WH8020. From our results, the involvement of PebA and PebB in metabolic channeling is postulated.

EXPERIMENTAL PROCEDURES
Reagents-Unless otherwise specified, all chemical reagents were ACS grade or better. Spinacia oleracea ferredoxin, Clostridium pasteurianum ferredoxin, Porphyra umbilicales ferredoxin, ferredoxin:NADP ϩ oxidoreductase, glucose-6-phosphate dehydrogenase, and NADP ϩ were purchased from Sigma-Aldrich. Restriction enzymes and T4 DNA ligase were obtained from New England Biolabs and Phusion TM DNA polymerase from Finnzymes. HPLC grade acetone, chloroform, and formic acid were purchased from Acros, Sigma-Aldrich, and J. T. Baker, respectively.
Glutathione-Sepharose TM 4FF, PreScission TM protease, and expression vector pGEX-6P-3 were obtained from GE Healthcare. Expression vector pASK-IBA45 ϩ and Strep-Tactin-FIGURE 1. Biosynthetic pathway of bilin pigment synthesis. In plants, red algae, cyanobacteria, and cryptomonads BV is reduced by FDBRs with high double bond regiospecificity to produce phytobilins. PEB synthesis is catalyzed by PebA and PebB and proceeds via the semi-reduced intermediate DHBV.
Sepharose were purchased from IBA. Stirred ultrafiltration cell and filters were obtained from Millipore.
Construction of pebA and pebB Expression Vectors-The sequences for pebA and pebB (GenBank TM accession number AY363679) were amplified from chromosomal DNA of F. diplosiphon (Fredi) strain Fd33 (9) obtained from the laboratory of D. Kehoe. PCRs were set up using Phusion TM DNA polymerase. The forward and reverse primers were: 5Ј-GGAATTCGATCTATA-AGTGCTTCCTTGAGC-3Ј and 5Ј-CCGCTCGAGCTATTTG-GCTACAACAGTTGCTAATG-3Ј for pebA; and 5Ј-GGAAT-TCGATCCGGAGCGAAGCGAAGTTG-3Ј and 5Ј-AACTGCA-GTTATTTGATAGCTGATGTGAGCTTTC-3Ј for pebB (the underlined bases indicate the EcoRI, XhoI, or PstI sites). The PCR product pebAFredi was ligated into pGEX-6P-3 vector for N-terminal fusion with glutathione S-transferase (GST). The plasmid was transformed in Escherichia coli BL21(DE3) cells. The pebBFredi construct was ligated into pASK-IBA45 ϩ for N-terminal fusion with Strep-tag II and transformed in E. coli DH10B cells. The integrity of the plasmid constructs was confirmed by DNA sequencing. Cloning strategies for the Synechococcus sp. WH8020 (Synpy) pebA und pebB were described previously (10).
Production and Purification of Recombinant PebA and PebB-For production of recombinant PebAFredi, PebASynpy, and PebBSynpy, 2 liters of LB medium containing 100 g/ml ampicillin was inoculated at 1:100 from an overnight culture of BL21(DE3) carrying the respective plasmid construct and cultivated at 37°C to an A 578 nm ϳ 0.6 -0.8. After a temperature shift to 17°C, protein expression was induced by adding 100 M isopropyl-␤-D-thiogalactopyranoside, and cells were cultured for a further 15 h at 180 rpm. Cells were harvested by centrifugation and stored at Ϫ20°C. Frozen cells were thawed, resuspended in 20 ml of lysis buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 5 mM MgCl 2 , 0,05% Triton X-100), and disrupted by passage through a French press cell at 20,000 p.s.i. After ultracentrifugation for 50 min at 170,000 ϫ g the supernatant was loaded on a glutathione-Sepharose TM 4FF column. Washing and elution were done according to instructions supplied by the manufacturer. Protein-containing fractions were cleaved with 2 units of PreScission TM protease/mg of protein in the recommended cleavage buffer. Cleavage led to an additional eight amino acid residues on the N termini of the proteins, and the first amino acid was changed from Met to Ile. A second glutathione-Sepharose column was used to separate the GST tag from the protein. The protein solutions were dialyzed against reaction buffer (25 mM TES-KOH buffer, pH 7.5, 100 mM KCl). If indicated, an additional purification step was performed using gel permeation chromatography on a High Load TM 26/60 Superdex TM 75 prep grade column (GE Healthcare). The proteins were concentrated using a stirred ultrafiltration cell with a molecular weight cut-off of 10,000 and stored up to 3 days on ice.
Production of recombinant PebBFredi was induced using anhydrotetracycline (200 g/ml). Cultivation conditions (i.e. medium, temperature) were identical to those described above. Purification was done on a Strep-Tactin-Sepharose column as recommended by the manufacturer. Strep-tagged PebBFredi is N-terminally extended by 20 additional amino acids (Strep-tag II).
Purification of Recombinant Reductants-The DNA sequence of Synechococcus sp. PCC 7002 ferredoxin (petF) was amplified from the plasmid pSe280fd (obtained from D. Bryant) using the following forward and reverse primers: 5Ј-GGAATTC-GATCGCTACATATAAGGTTAC-3Ј and 5Ј-CCGCTCGA-GCTAGTAGAGTTCTTCCTCTTT-3Ј (the underlined sequences indicate the EcoRI or XhoI sites). The PCR product was ligated to pGEX-6P-3. Expression was done in NZCYM medium as described elsewhere (11), and protein production was induced using 1 mM isopropyl-␤-D-thiogalactopyranoside at an A 578 nm of 0.6 -0.8. Cells were harvested 4 h after induction. Purification was done with two sequential glutathione-Sepharose columns as described for the bilin reductases. The employed cleavage buffer was free of dithiothreitol and EDTA.
Determination of Protein and Bilin Concentrations-Concentrations of the bilin reductases were determined using the calculated molar extinction coefficient (⑀ 280 nm ) (12).
The concentration of recombinantly produced ferredoxin from Synechococcus sp. PCC 7002 was determined using an ⑀ 420 nm of 9.7 mM Ϫ1 cm Ϫ1 (13). Concentration of BV IX␣ was calculated using an ⑀ 376 nm of 68.6 mM Ϫ1 cm Ϫ1 and an ⑀ 698 nm of 32.6 mM Ϫ1 cm Ϫ1 in 2.5% HCl-MeOH (14). The concentration of 3E-PEB was determined using ⑀ 326 nm 15.8 mM Ϫ1 cm Ϫ1 and ⑀ 591 nm 25.2 mM Ϫ1 cm Ϫ1 in 5% HCl-MeOH (15). The concentration of DHBV was determined by measuring the absorbance at the 561-nm maximum in 5% HCl-MeOH and using the long wavelength extinction coefficient of 3E-PEB for calculation. All concentrations were determined using a Ultrospec 2000 UVvisible spectrophotometer (GE Healthcare).
Bilin Reductase Activity Assay-Assays for bilin reductase activity were done as described previously with small modifications (10). The standard assays contained 1.5 M bilin reductase, 5 M bilin substrate, and 4.8 M recombinantly produced Synechococcus sp. PCC 7002 ferredoxin or the alternative ferredoxins in reaction buffer. The assays were incubated for 15-30 min at 30°C in the dark. Bilins were isolated using C18 Sep-Pak columns (Waters) and evaporated to dryness in vacuo. For spectrometric detection of electron transfer activity, the assay was performed using 10 M PebAFredi, 10 M BV, 40 M NADP ϩ , and 0.0125 units/ml ferredoxin: NADP ϩ oxidoreductase.
Preparative Production of 15,16-Dihydrobiliverdin-Larger quantities of DHBV were produced enzymatically by setting up a 10-ml bilin reductase activity assay containing 20 M BV, 5 M PebAFredi, and 4.8 M Synechococcus sp. PCC 7002 ferredoxin in reaction buffer at 30°C. The reaction progress was monitored by measuring absorbance spectra at different time points during the reaction. If no further substrate conversion was observed, the reaction was stopped immediately by adding 40 ml of 0.1% (v/v) trifluoroacetic acid and cooling on ice. A C18 Sep-Pak light column was preconditioned with sequential washes of CH 3 CN, H 2 O, 0.1% (v/v) trifluoroacetic acid, and 10% (v/v) MeOH in 0.1% trifluoroacetic acid. The bilin was loaded on the column washed with 6 ml 0.1% (v/v) trifluoroacetic acid and 6 ml of 20% MeOH in trifluoroacetic acid, eluted with CH 3 CN, and dried in vacuo. The purity of produced DHBV was controlled by HPLC for contamination by other bilins.
Absorption and Fluorescence Spectroscopic Analysis-All protein solutions used for binding site saturation experiments were checked for homogeneity with analytical gel permeation chromatography on a Superdex 75 HR10/30 column. Protein solutions were adjusted to concentrations ranging from 0.5 to 18 M; substrate and product complexes were formed by incubating the protein solution with a 4 M final concentration of the bilin (5-10 l of stock solution) for 20 min on ice in the dark. All fluorescence measurements were performed under physiological conditions in reaction buffer using an Aminco Bowman AB2 spectrofluorimeter at a constant temperature of 20°C. The excitation/emission wavelengths used were 590 nm/610 nm for PebA⅐DHBV, 605 nm/645 nm for PebB⅐DHBV, and 545 nm/630 nm for PebB⅐3E-PEB. Both excitation and emission slit widths were set at 4 nm, and the scan speed was 2.5 nm/s. Determination of binding constants of substrate/product to the enzymes was done according to Clarke (16). Binding curves were measured by the increase of fluorescence intensities, as the bilin⅐enzyme complex is formed at equilibrium. Obtained data were analyzed using Sigma Plot 9.0 (Systat Software Inc.), and data were fitted against Equation 1.
Time-dependent absorbance measurements (Fig. 4) were performed in a stirred cell tempered to 30°C on an Agilent Technologies 8453 spectrophotometer with ChemStation biochemical analysis software. Absorbance spectra (Fig. 5) were measured using a Lambda 2 UV-visible spectrophotometer (PerkinElmer Life Sciences).
Gel Permeation Chromatography of Enzyme⅐Bilin Complexes-Enzyme⅐bilin complexes were formed by incubating protein solution with approximately double the molar concentration of bilin for 3 min at 30°C. The complex was purified by passing it through a NAP TM -5 column (GE Healthcare) prior to analytical gel permeation chromatography on a Superdex 75 HR10/30 column. During isocratic elution, absorbance was simultaneously detected at 280, 585, and 605 nm using the UV-900 detector of the ÄKTApurifier system (GE Healthcare).

Recombinant Production and Purification of FDBRs-PebA
of the filamentous freshwater cyanobacterium F. diplosiphon and also PebA and PebB of the coccoid marine cyanobacterium Synechococcus sp. WH8020 were expressed using a tac promoter-driven N-terminal GST fusion protein. A protocol using overnight proteolytic cleavage of the affinity-purified GST fusion protein followed by a second affinity chromatography to remove GST tag and protease led to the best results. This purification strategy led to less than 10% impurity (Fig. 2). Prior to all quantitative experiments, a third purification step using gel permeation chromatography was performed to remove possibly aggregated enzyme. N-terminal sequencing of PebAFredi by Edman degradation revealed no GST or other protein contamination after this purification step, and the yields of this method varied, depending on the enzyme, between 3 and 7 mg/liter cell culture. The best results for PebBFredi purification were achieved with tet promoter-driven expression followed by single-step purification of the Strep-tagged enzyme (Fig. 2). This procedure yielded about 1 mg/liter cell culture.
Activity of the Recombinant Enzymes-To verify the activity of the purified bilin reductases, we used an in vitro assay system as described previously with an excess of reductant (10). The optimal pH value for PebAFredi activity was determined to be pH 7.5, and therefore all further assays were performed at this pH. As expected, both PebAs were found to convert BV to DHBV (Fig. 3). The reduction of the 15,16-double bond of BV is accompanied by obvious blue-shifts of the absorption that enabled us to monitor the in vitro reaction progress spectroscopically (shown in Fig. 4). The analyzed PebBs catalyzed the reduction of the A-ring diene system of DHBV to PEB, which may likely be a 2,3-reduction, followed by isomerization to 3Z-PEB, which is the supposed natural chromophore of PE. The overall reaction can be followed by HPLC in an assay mixture containing BV, PebA, and PebB at once (Fig. 3) or individually using BV as a substrate for PebA or purified DHBV as substrate for PebB (data not shown) to confirm the specific catalytic activity for both enzymes. The appearance of the energetically stable 3E-PEB may be a result of the bilin extraction procedure as described previously (10). In our assay system both reactions were found to be most efficient using plant type [2Fe-2S] ferredoxins of Synechococcus sp. PCC7002 or S. oleacera as redox partners followed by the [2Fe-2S] ferredoxin from P. umbilicalis; only marginal activity could be detected using [4Fe-4S] ferredoxin from C. pasteurianum. These results are in good agreement with results obtained for PcyA of Anabaena sp. PCC7120 (5) and for phytochromobilin synthase of Avena sativa (17). Consistent with PcyA is the insensitivity of PebAFredi toward metal chelators like EDTA (10 M), o-phenanthroline (5 M), or 2,2Ј-dipyridyl (5 M), indicating no involvement of protein-associated metal ion cofactors during catalysis (data not shown) (5).  (1) and PebASynpy (3) after affinity and gel filtration chromatography. One-step affinity-purified Strep-tagged PebBFredi (2) and PebBSynpy after two affinity chromatography columns (4). M is the protein molecular mass marker.

Reoxidation of 15,16-DHBV-The
PebA-catalyzed reduction of BV to DHBV was found to be reversible. Incubation of a preformed complex of PebA and DHBV for 3-6 days on ice in the dark led to a visible transformation of the color from pink to green, the result of a back-oxidation of DHBV to BV, which was confirmed by HPLC (data not shown). This back-oxidation was slower under low oxygen conditions; a control experiment with carbonic anhydrase instead of PebA in the solution resulted in much lower DHBV reoxidation, indicating that this reaction is accelerated in the presence of PebA.
Substrate Specificity of FDBRs-To analyze the substrate specificities of the various FDBRs we examined different natural bilins (Table 1). In our standard HPLC assay system, we did not find 3E-PCB to be converted by PebA, indicating that a lack of A-and D-ring vinyl moieties, together with a changed geometry of the A-ring ethylidene group, prevents recognition of 3E-PCB as substrate. DHBV was not converted by PcyA, demonstrating that the reduction of the 15,16-double bond causes structural difference in the bilin, which likely prevents proper placement of the bilin in the active site pocket. Interestingly, we found that PebA was able to reduce the plant bilin P⌽B to PEB, indicating that an A-ring ethylidene group instead of an A-ring vinyl group is not critical for substrate recognition by PebA.
Spectroscopic Properties of FDBR Complexes with Their Substrates or Products-All tested bilins incubated with PebA or PebB displayed distinct spectroscopic properties differing from those of the free pigments (Fig. 5). Not only were the absorbance maxima shifted, but also the peak intensities and the ratio of the long wavelength absorbance peak and the near UV absorbance peak intensities ( max2 / max1 ) changed (Table 2). Interestingly, the protein environment of PebA and PebB influences the spectral properties of one and the same bilin, indicating differences in the bilin binding pocket of both FDBRs. BV binding to PebA causes an increase in absorbance compared with free BV, with a shifted long wavelength absorbance maximum from 681 to 691 nm. Bound to PebB, the absorbance maximum is less intensely increased, but the long wavelength absorbance maximum is shifted from 681 to 706 nm. Consequently, the ratio of max2 / max1 did not change notably, and lies between 0.3 for BV and 0.52 for PebA⅐BV ( Fig. 5A and Table  2). Spectral analyses of complexes of the semi-reduced intermediate DHBV with PebA (enzyme⅐product complex) and PebB (enzyme⅐substrate complex) revealed noticeable differences. DHBV binding to PebA led to an absorbance increase at both maxima and to a shift from 565 to 590 nm for max2 . Binding to PebB shifted the max2 to 605 nm and decreased the max1 absorbance, thereby changing the color of the complex from pink to blue. The max2 / max1 ratio was changed from 0.63 (free DHBV) to 0.97 (PebA⅐DHBV) and 1.36 (PebB⅐DHBV) (Fig. 5B    and Table 2). The binding of 3E-PEB to PebB led to an increased absorbance at the long wavelength absorbance maximum of 535 nm, and shifted the absorbance maximum to 545 nm. Binding to PebA shifted the absorbance maximum to 586 nm ( Fig.  5C and Table 2).
Fluorescent Properties of FDBR⅐Bilin Complexes-During the bilin binding experiments with PebA and PebB, we observed that the intermediate enzyme⅐bilin complexes in PEB biosynthesis are fluorescent, although their attachment to the protein is not covalent. Excitation at their max2 absorbance maxima led to intense fluorescence emission, and the respective excitation/emission pairs were determined to be 590/610 nm for PebA⅐DHBV, 605/645 nm for PebB⅐DHBV, and 545/630 nm for PebB⅐3E-PEB (Fig. 6).

Comparative Binding Affinities of Bilin Substrates and
Products-The specific fluorescent properties of enzyme⅐bilin complexes provide the opportunity to perform steady state binding analyses to determine binding constants for the respective enzyme⅐bilin complexes that occur during PEB biosynthesis in the absence of reduced ferredoxin. Because fluorescence signals of enzyme-bound bilins were intensely increased, we were able to determine binding constants by saturation titration of the bilin with increasing amounts of protein (16). Binding could only be determined at equilibrium because the binding was so rapid (probably in the ms range) that the kinetics were not measurable with our instruments (data not shown). At equilibrium we observed hyperbolic saturation for all tested enzyme⅐bilin complexes, indicating one binding site per enzyme. Using the fluorescent properties of the intermediate enzyme⅐bilin complexes allowed determination of bilin binding because almost no signal was derived from unbound bilins (Fig. 7). Experiments were done in triplicate, and binding affinities with K d values around 1 M for the binding of the product DHBV to PebA (K d ϭ 1.09 M) and the binding of DHBV as substrate for PebB (K d ϭ 1.48 M) were determined. An unexpected result was the tight binding of the product DHBV to PebA, which was in the   same range as the substrate binding to PebA. The K d for binding of PebB to the product 3E-PEB was determined to be about 5-fold higher (K d ϭ 5.8 M).
Transfer of DHBV Intermediate from PebA to PebB-The binding of DHBV to PebA was tight enough to enable the purification of a preformed complex by passing it through a NAP TM 5 column. Because of the high affinity of the Sephadex G25 material toward the free bilins, we were able to generate enzyme solutions that contained only marginal concentrations of unbound bilin. The purified complexes were subjected to gel permeation chromatography analysis, and we are able to detect protein-specific absorbance as well as specific absorbance for the PebA⅐DHBV complex (Fig. 8A). All analyzed FDBRs were determined to elute as monomers under the employed conditions, with a relative molecular weight of about 30,000. In contrast, the GST fusion protein of PebB eluted as a dimer of about 115,000 (Fig. 8B). The latter result is not surprising because GST is known to form dimers (18). The spectroscopic properties of the GST fusion proteins do not differ from those of GSTfree protein (data not shown), indicating that the N-terminal GST fusion does not alter the substrate binding pocket that is structurally located between the central ␤-sheet and the C-terminal ␣-helices (7). Gel permeation chromatography analysis of a molar 1:1 mixture of the purified PebA⅐DHBV complex and the GST⅐PebB fusion revealed approximately equal quantities of DHBV bound to PebA and PebB (Fig. 8C).

PebA and PebB Belong to the FDBR Family of Radical
Enzymes-In this current work we have presented the first detailed biochemical characterization of the two enzymes involved in PEB biosynthesis. They belong to the newly defined family of FDBRs. The best described member of this family is PcyA, which catalyzes the four-electron reduction of BV IX␣ to PCB. Because of the lack of bound metal cofactors and the detection of a bilin radical intermediate, the family of FDBR constitutes a novel family of radical enzymes (6). Although we have not presented any EPR data, PebA lacks metal ion cofactors, and the reaction most likely proceeds via radical intermediates. Interestingly, we observed that DHBV bound to PebA can be reoxidized to BV by molecular oxygen. Reactive oxygen species such as peroxyradicals are known to reoxidize albumin bound bilirubin (BR) to BV (19). Although it seems unlikely that reactive oxygen species are produced under our experimental conditions, reactive oxygen species or molecular oxygen itself could serve as the oxidant. These observations fit the results in which a higher yield of a recombinant, in vivo produced PEB adduct of phytochrome (Cph1) was achieved under low aeration as reported recently (20). In this regard, PcyA has been described to reduce BV much more efficiently under anaerobic conditions (6). However, in our assay system removal of oxygen did not significantly enhance the rate of BV  . Gel permeation chromatography elution profiles of purified PebA and PebB co-eluting with DHBV. A, purified PebA was incubated with its product, DHBV, and the complex was purified using a NAP TM 5 desalting column prior to analytical gel permeation chromatography on a Superdex 75 HR10/30 column. The protein was detected by absorbance at 280 nm (solid thick line), and DHBV was detected at 585 nm (dotted line), corresponding to the absorbance maxima of PebA⅐DHBV, or 605 nm (solid thin line), corresponding to the absorbance maxima for PebB⅐DHBV. Binding was tight enough for coelution of enzyme and bilin (monomer). B, elution profile of the subsequent enzyme PebB (as a GST fusion dimer) without bound bilin. C, elution profile of purified PebA⅐DHBV complex incubated with an equimolar amount of GST-PebB shows that approximately half of the bilin was detected as coeluting with PebB after incubation.
reduction, but the intermediate DHBV is more stable under anaerobic conditions (data not shown).
All Members of the FDBR Family Have Distinct Substrate Specificities-All tested bilins that occur in one organism cannot be converted by the other bilin reductases. But we find that PebA can convert P⌽B to PEB indicating that an A-ring ethyl group instead of a vinyl group is not critical for substrate recognition by PebA (Table 1). Because we used a mixture of the 3E-and 3Z-isomer, of P⌽B we cannot precisely claim which isomer is a substrate for PebA. Results from PcyA would infer that only the 3Z-isomer can be turned over. This is an interesting result insofar as P⌽B had originally been reported to be an intermediate in PEB biosynthesis in Cyanidium caldarium (21). The inability of PcyA to convert DHBV and of PebB to convert 18 1 ,18 2 -DHBV is a result of the changed planarity in DHBV. These high substrate specificities of the FDBRs demonstrate that bilin biosynthetic pathways are evolutionary designed to enable strict regulation and to avoid cross-reactions between different bilins occurring in the same organism.
PebA and PebB Form Distinct Complexes-Binding of substrate or product to the enzymes leads in all cases to a red shift of the max2 . Red shifts like that were reported previously to occur upon protonation of the basic pyrrolenic nitrogen atom within the tetrapyrrole ring structure (22,23) or as a result of an enhanced conjugated -electron system of the bilin (23,24). Because protonation is presumably the first step in the catalytic bilin reduction by PcyA (6), it is quite possible that during enzyme binding the positioning of the bilin close to a proton donating amino acid or the coordination by an electropositive ligand causes these red shifts. Interestingly, DHBV in acidic methanol displays a rather blue-shifted spectrum (data not shown). Therefore, the observed spectral effects cannot easily be assigned as due to protonation or geometrical changes. The massive changes in the max2 / max1 ratio that occur during binding of DHBV to PebA or PebB suggest a more linearly stretched conformation of the DHBV bound to the enzyme compared with free DHBV (22). This fact is interesting in regard to the conformation of BV in the PcyA crystal structure, which is cyclic (7). The spectroscopic differences between PebA⅐DHBV and PebB⅐DHBV indicate that the protein environment, possibly the conformation or orientation of the tetrapyrrole, is different in both complexes.
Fluorescent Properties of PebA⅐Bilin and PebB⅐Bilin Complexes-Free bilins exhibit very low fluorescence, but it has long been known that the covalent adducts of bilins to proteins lead to fluorescent proteins with high fluorescence quantum yields (25,26). Because of their fluorescent properties, phycobiliproteins, especially PEs, are frequently used for fluorescence labeling in biotechnological applications as well as for quantification of phytoplankton composition (27,28). Phytochromes become intensely fluorescent when the bilin chromophore is hindered from undergoing photoisomerization either by steric hindrance caused by the protein environment (29) or because the chromophore lacks the 15,16-double bond, as in the case of plant apophytochrome reconstituted with PEB (25) or bacterial apophytochrome reconstituted with DHBV (30). DHBV bound to PebA or PebB shows distinct fluorescent properties. The occurrence of this fluorescence indicates that the bilins are held in a sterically tightly fixed conformation as is the case in PE. Furthermore, bridging and restriction of substituent mobilities could account for the high fluorescence (23). This observation implies that the binding mechanism to both enzymes is different as is the position of the 15,16-double bond and, accordingly, the A-ring vinyl moiety of the substrate to be reduced.
PebA and PebB May Be Involved in Metabolic Channeling-Most parts of the tetrapyrrole metabolism involve rather labile intermediates, which in aerobic solution would last only for short periods of time. A tight control is further necessary, as many tetrapyrroles are highly toxic. Metabolic channeling has been demonstrated as a way to avoid unwanted or harmful sidereactions, e.g. for biosynthesis of the early precursor, 5-aminolevulinic acid (31,32). Here we present some indications that this also might be relevant for the stages of bilin biosynthesis.
The involvement of two enzymes in the biosynthesis of PEB was already inferred from the genomic localization of the encoding genes in an operon when the bilin reductases were first identified in 2001 (10). In F. diplosiphon this operon structure was confirmed and, it was shown that pebA and pebB are co-transcribed from the same promoter generating a polycistronic mRNA (1). The fact that two enzymes are necessary to yield one product (PEB) and the operon structure led to the postulation of potential metabolic channeling (10). Metabolic channeling would ensure direct transfer of the intermediate DHBV from PebA to PebB without releasing it into the bulk solvent. In this regard, it has been demonstrated previously that the conversion of DHBV to PEB is not the rate-limiting step in PEB biosynthesis (1). Furthermore, it has already been postulated that P⌽B synthase might be involved in metabolic channeling with a heme oxygenase in plants (17). Our current results provide more evidence that the intermediate DHBV is not released from PebA but rather is transferred directly to PebB. The results of the transfer of DHBV from PebA to PebB, together with the tight binding of the product DHBV to PebA, led us to the suggestion that under the expected low substrate concentration in vivo, PebB may play an important role in releasing and taking over the intermediate from PebA. The lower binding affinity of PEB to PebB may demonstrate the release of the final product PEB from the enzyme; but it may also be result of differences in the ethylidene side chain geometry of the 3Z-compared with the 3E-isomer that we used for titration experiments.
Conclusion and Outlook-Our results presented here have provided new insights into phycoerythrobilin biosynthesis with hints that it may involve metabolic channeling. Current efforts in our laboratory are focused on combining co-immunoprecipitation with surface plasmon resonance spectroscopy and other biophysical methods to ultimately prove the existence of such complexes.