Phycocyanobilin:Ferredoxin Oxidoreductase of Anabaena sp. PCC 7120

In cyanobacteria, the biosynthesis of the phycobiliprotein and phytochrome chromophore precursor phycocyanobilin is catalyzed by the ferredoxin-dependent enzyme phycocyanobilin:ferredoxin oxidoreductase (PcyA), which mediates an atypical four-electron reduction of biliverdin IXα. Here we describe the expression, affinity purification, and biochemical characterization of recombinant PcyA from Anabaena sp. PCC 7120. A monomeric protein with a native M r of 30,400 ± 5,000, recombinant PcyA forms a tight and stable stoichiometric complex with its substrate biliverdin IXα. The enzyme exhibits a strong preference for plant type [2Fe-2S] ferredoxins; however, flavodoxin can also serve as an electron donor. HPLC analyses establish that catalysis proceeds via the two electron-reduced intermediate 181,182-dihydrobiliverdin, indicating that exovinyl reduction precedes A-ring (endovinyl) reduction. Substrate specificity studies indicate that the arrangement of the A- and D-ring substituents alters the positioning of the bilin substrate within the enzyme, profoundly influencing the course of catalysis. Based on these observations and the apparent lack of a metal or small molecule cofactor, a radical mechanism for biliverdin IXα reduction by phycocyanobilin:ferredoxin oxidoreductase is envisaged.

Phycocyanobilin (PCB) 1 is a linear tetrapyrrole (bilin) found in cyanobacteria, algae, and cryptomonads that functions as the direct precursor of the chromophores of the light-harvesting phycobiliproteins and cyanobacterial/algal phytochromes (1,2). The PCB biosynthetic pathway shares common intermediates with those of heme and chlorophyll to the level of protoporphyrin IX, whereupon the pathways diverge upon metalation with either iron or magnesium (1). All phycobilins share the common intermediacy of biliverdin IX␣ (BV IX␣), which is derived from heme (1,3). BV IX␣ is the product of heme oxygenase-mediated cleavage of the heme macrocycle, which yields equimolar quantities of BV IX␣, CO, and iron. The metabolic fate of BV IX␣ differs in mammals, cyanobacteria, and plants, where BV IX␣ is metabolized by different reductases with unique double bond specificities. In contrast with the mammalian NAD(P)H-dependent BV IX␣ reductase, cyanobacteria and red algae possess ferredoxin-dependent bilin reductases primarily for the synthesis of the linear tetrapyrrole precursors of their phycobiliprotein light-harvesting antennae complexes, whereas evolutionarily related ferredoxin-dependent bilin reductases are found in higher plants for the synthesis of the phytochrome chromophore precursor phytochromobilin (P⌽B) (3,4).
We recently documented that pcyA genes from cyanobacteria and oxyphotobacteria encode bilin reductases, which catalyze the ferredoxin-dependent reduction of BV IX␣ to (3Z)-PCB, the bilin precursor of their phycobiliprotein and phytochrome chromophores. Designated phycocyanobilin:ferredoxin oxidoreductases (EC 1.3.7.5), PcyA enzymes are atypical bilin reductases, because they catalyze a four-electron reduction; all others catalyze two-electron reductions (4). Formally two-electron reductions of vinyl substituents on the pyrrole A-and D-rings of BV IX␣, the sequence of reductions mediated by PcyA, is presently unknown (see Fig. 1). Neither of the two likely dihydrobiliverdin (DHBV) intermediates, (3Z)-P⌽B or 18 1 ,18 2 -DHBV IX␣, have yet been detected (4). Since (3Z)-P⌽B is an intermediate in the biosynthesis of PCB, the precursor of the phytochrome chromophore in the green alga Mesotaenium caldariorum (5), its intermediacy in the PcyA-mediated biosynthesis of PCB in cyanobacteria is therefore a reasonable possibility. Previous studies have shown that the biosynthesis of PCB in the red alga Cyanidium caldarium proceeds via phycoerythrobilin as an intermediate ( Fig. 1) (1). The enzyme(s) that catalyze the conversion of phycoerythrobilin to PCB have not yet been identified; hence, it is conceivable that PcyA might also mediate this conversion.
This study was undertaken to characterize the biochemical properties of a representative PcyA enzyme from the filamentous cyanobacterium Anabaena sp. PCC 7120. The specific objectives of these experiments were to identify the semireduced intermediate produced during the catalysis of BV and to probe the bilin substrate specificity of this unusual ferredoxindependent four-electron reductase. Based on these investigations, a chemical mechanism for PcyA-mediated bilin reduction is proposed.

EXPERIMENTAL PROCEDURES
Reagents-Unless otherwise specified, all chemical reagents were American Chemical Society grade or better. Glutathione-agarose, spinach ferredoxin, Clostridium pasteurianum ferredoxin, ferredoxin: NADP ϩ oxidoreductase (FNR), and size exclusion molecular weight markers (MW-GF-200) were purchased from Sigma. Restriction enzymes and Taq polymerase were obtained from Invitrogen. HPLC grade acetone, chloroform, and 80% formic acid were purchased from Fisher. The expression vector pGEX-6P-1 and PreScission™ protease were obtained from Amersham Biosciences. Centricon-10 concentrator devices were purchased from Amicon (Beverly, MA).
Expression and Purification of PcyA-Anabaena sp. PCC 7120 pcyA was cloned into the Escherichia coli expression vector pGEX-6-P1 (Amersham Biosciences) to produce pGEXpcyA (4). E. coli strain DH5␣ containing pGEXpcyA was induced to express glutathione S-transferase-PcyA, which was purified according to instructions supplied by the manufacturer and protocols described earlier (10). Proteolytic cleavage with the PreScission™ protease yielded the native protein with the N-terminal amino acid extension GPLGSPEF and with the initiator methionine residue changed to isoleucine. Purified PcyA protein concentration was estimated from the absorbance at 280 nm using the calculated ⑀ 280 nm of 29,726 M Ϫ1 cm Ϫ1 (11).
Purification of Recombinant Reductants-Synechococcus sp. PCC 7002 ferredoxin and flavodoxin clones, obtained from Dr. D. A. Bryant, were expressed and purified as described previously (12,13). Expression and purification of putidaredoxin and putidaredoxin reductase, whose clones were kindly provided by Dr. Paul Ortiz de Montellano (University of California, San Francisco), were performed as described. 2 Flavodoxin was quantified by absorption at 467 nm and an absorption coefficient of ⑀ 467 nm 9,500 M Ϫ1 cm Ϫ1 (13), whereas putidaredoxin and putidaredoxin reductase were quantified by absorption at 454/415 nm, respectively, using the absorption coefficients of ⑀ 454 nm 10,000 M Ϫ1 cm Ϫ1 and ⑀ 415 nm 11,100 M Ϫ1 cm Ϫ1 . 2 Standard Bilin Reductase Activity Assay-Assays for bilin reductase activity were performed as described previously (10,14). Standard assays contained 1.5 M PcyA, 4.8 M ferredoxin, and 5 M BV IX␣ in 25 mM TES-KOH, pH 7.5 (assay buffer), and were incubated for 30 min at 28°C under green safe light unless otherwise specified. Following catalysis, bilins were isolated using a C18 Sep-Pak column (Waters) and evaporated to dryness in vacuo (4).
Direct HPLC Analysis-Bilin reaction products were dissolved in 10 l of Me 2 SO and diluted with 200 l of the HPLC mobile phase.
Following brief centrifugation and filtration through a 0.45-m polytetrafluoroethylene syringe filter, bilins were resolved by reversed phase chromatography using an Agilent Technologies 1100 Liquid Chromatograph. The HPLC column used for all of the analyses was a 4.6 ϫ 250-mm Phenomenex Ultracarb 5-m ODS(20) analytical column with a 4.6 ϫ 30-mm guard column of the same material. The mobile phase consisted of acetone, 20 mM formic acid (50:50 by volume), and the flow rate was 0.6 ml/min. Eluates were monitored at 650, 560, and 380 nm using an Agilent Technologies 1100 series diode array detector. As needed, complete spectra were obtained for the peaks desired. Peak areas were quantitated using Agilent Technologies Chemstation software.
Glycerol Gradient Centrifugation-PcyA preparations (40 g) were sedimented through a 2.5-ml continuous 10 -25% glycerol gradient in size exclusion chromatography buffer. A detailed experimental procedure described previously was used for sedimentation coefficient determination (15).
Spectroscopic Analysis of Biliverdin Binding-Increasing amounts of PcyA were added to 5 M (final concentration) BV IX␣, BV XIII␣, or BV III␣ solutions in a final volume of 500 l of 25 mM TES-KOH, pH 7.5, buffer under green safe light. After incubation for 30 min at room temperature, absorbance spectra were recorded using an HP 8453 spectrophotometer. Normalization of the spectra and spectral deconvolution were performed using Microsoft Excel. To obtain bilin-PcyA dissociation constants, absorbance differences (⌬A) at the max of each bilin-PcyA complex were plotted as a function of PcyA concentration. Dissociation constants were obtained by fitting this data to the hyperbolic equation  combinant PcyA, obtained by "on-column" proteolytic cleavage of the glutathione S-transferase fusion protein, was purified to ϳ90% homogeneity as shown in Fig. 2A. On-column cleavage was preferable to "in-solution" proteolysis of the glutathione S-transferase-PcyA fusion protein, which led to extensive protein precipitation and poor protein recovery. One-liter bacterial cultures typically yielded 3 mg of on-column cleaved PcyA. All results presented here correspond to PcyA. However, glutathione S-transferase-PcyA fusion protein preparations showed nearly identical catalytic properties (data not shown).

Expression and Purification of Recombinant
Determination of the Native Molecular Mass of PcyA-The native molecular mass of PcyA was determined using size exclusion chromatography and glycerol gradient sedimentation (Fig. 2, B and C). A relative molecular weight of 30,400 Ϯ 5,000 for PcyA was deduced with both methods, which is in good agreement with the calculated molecular mass of 28,726 daltons. Thus, recombinant PcyA appears to be a monomeric enzyme. In order to determine whether PcyA can form a stable complex with spinach ferredoxin, PcyA was incubated with a 2-fold molar excess of spinach Fd and evaluated by both methods. Higher order complex formation between PcyA and Fd was not observed with either method under the conditions examined; nor did the addition of a 2-fold molar excess of BV IX␣ influence the result (data not shown).
PcyA Lacks Metal or Small Molecule Cofactors-Purified recombinant PcyA was analyzed using absorption spectroscopy for the presence of light-absorbing cofactors such as hemes, flavins, iron-sulfur clusters, etc. Spectroscopic evidence for any of these cofactors was not obtained for PcyA at concentrations as high as 5 mg/ml. In order to understand whether solventaccessible metal ions are critical for activity (directly or indirectly as structural components), purified PcyA was incubated with the metal chelators EDTA (10 mM), 1,10-phenanthroline, and 2,2Ј-dipyridyl (5 mM each). After removal of the chelator from the protein by passing the mixture through a G-25 desalting column (Amersham Biosciences), enzyme activity was determined. None of the chelators had any inhibitory effect on the activity of PcyA (data not shown).
Reductant Specificity of PcyA-The enzymes that mediate the reductive conversion of BV IX␣ to phycobilins are all dependent on plant type [2Fe-2S] ferredoxins (4,14). For this reason, all PcyA assays were performed in the presence of saturating levels of spinach ferredoxin and the ferredoxinreducing system, consisting of spinach FNR and NADPH. Omission of any of these components led to no PcyA activity (data not shown) (4). To test whether other reductants can serve as electron donors to PcyA, we tested recombinant Synechococcus sp. PCC 7002 ferredoxin and flavodoxin also in combination with spinach FNR. Other reductants tested included C. pasteurianum ferredoxin, a 2[4Fe-4S] ferredoxin, and a putidaredoxin/putidaredoxin reductase system from Pseudomonas putida. As shown in Table I, maximum PcyA activity was obtained with plant type ferredoxin either from spinach or Synechococcus, whereas the FMN-containing redox protein flavodoxin less effectively supported PcyA catalysis. C. pasteurianum ferredoxin and the putidaredoxin/putidaredoxin reductase system from P. putida were considerably less active.
Bilin Binding to PcyA-Bilin binding experiments demonstrated that PcyA forms a complex with its bilin substrate that is stable through ultrafiltration, size exclusion chromatography, and dialysis. Spectrophotometric titration experiments with its natural substrate BV IX␣ and the two analogs, BV XIII␣ and BV III␣, are shown in Fig. 3A. Upon binding to PcyA, significant blue shifts of the long wavelength absorption maxima were detected for all of the bilin analogs along with an increase in molar absorption coefficient and the appearance of a shoulder at longer wavelengths. The appearance of the shoul-  a PcyA assays were performed as described under "Experimental Procedures" for 20 min using the HPLC method. Since the use of spinach ferredoxin represents standard assay conditions, this activity was set to 100% based on the production of PCB (peak absorbance area). b 13.2% represents the percentage of PCB produced; the value in parenthesis includes the amount of produced intermediate.
der was evident only for BV IX␣ and BV XIII␣, but not BV III␣ (Fig. 3B), a result that was consistent with the ability of PcyA to metabolize these two bilins (see below). Fig. 3C depicts replots of these absorption changes as a function of increasing PcyA concentration. From hyperbolic curve fitting of this data, the equilibrium dissociation constants were estimated to be 12.5, 14.5, and 4.5 M for the respective IX␣, XIII␣, and III␣ isomers under these experimental conditions. As a control, BV IX␣ binding experiments were also performed using bovine carbonic anhydrase from the Sigma MW-GF-200 molecular weight kit. No spectral changes were observed with increasing carbonic anhydrase concentration, indicating that the altered spectra reflected the formation of bilin-PcyA complexes (Fig.  3C).
The PcyA Two-electron Reduced Intermediate-The conversion of BV to PCB is a four-electron reduction that formally consists of sequential two-electron reductions with the intermediacy of a DHBV. As shown in Fig. 1 Table II). This new pigment, which eluted earlier than (3Z)-PCB (labeled I in Fig. 4A), reached a maximum level within 10 min and disappeared after 30 min (Fig.  4B). Although pigment I eluted at the same retention time as (3E)-P⌽B, its absorption spectrum differed from that of (3E)-P⌽B (Table II). The time course of the disappearance of BV IX␣, the appearance/disappearance of pigment I, and the appearance of the two isomers of PCB, shown in Fig. 4B, supports the intermediacy of pigment I in the PcyA-mediated conversion of BV IX␣ to PCB.
To verify that pigment I was a bona fide intermediate in the formation of PCB, it was collected and tested for its ability to bind to apo-Cph1 using a coupled phytochrome assembly assay (10,14) and for its ability to be further metabolized by PcyA to PCB using HPLC. As shown in Fig. 4C, isolated pigment I failed to produce a photoactive phytochrome upon incubation with apo-Cph1, indicating that pigment I was not (3E)-P⌽B or (3Z)/(3E)-PCB. Further incubation of pigment I with PcyA yielded products that could assemble with apo-Cph1 to produce a photoactive bilin adduct. Fig. 4C shows that the phytochrome difference spectrum of this adduct was identical to that of the authentic PCB adduct, both of which were blue-shifted from that of the Cph1:P⌽B adduct. Together with the ability of PcyA to convert pigment I to a mixture of pigments that co-elute with (3Z)-and (3E)-PCB (data not shown), these studies support the conclusion that pigment I is a bona fide intermediate in the PcyA-mediated conversion of BV to PCB. 18 1 ,18 2 -DHBV Is the Intermediate in the PcyA-mediated Reduction of BV-Since the intermediate failed to form a photoactive bilin-adduct with apo-Cph1 but had the same retention time as (3E)-P⌽B, we tested both (3E)-and (3Z)-P⌽B as substrates for PcyA. These studies showed that (3Z)-P⌽B, but not (3E)-P⌽B, was metabolized by PcyA. (3Z)-P⌽B was converted to a mixture of (3Z)-and (3E)-isomers of PCB, a result that was confirmed by assembly with apo-Cph1 (data not shown). Since (3Z)-P⌽B elutes at a different retention time from the intermediate on the HPLC (Fig. 4A, Table II), these studies confirm that neither P⌽B isomer is the semireduced intermediate in the PcyA-mediated reduction of BV. According to Fig. 1, the other likely intermediate is 18 1  Bilin Substrate Specificity Studies-Since the substrate analogs BV XIII␣ and BV III␣ bind to PcyA (see Fig. 3), their ability to be metabolized by PcyA was also examined. As shown in Fig. 5A, BV XIII␣ could be metabolized by PcyA to yield two products. Based on the relative retention time of known bilins in our HPLC system and the absorbance spectra of the two products (Fig. 5B), we propose that BV XIII␣ is converted by PcyA to the (3E)-and (3Z)-isomers of iso-P⌽B (16). This hypothesis is also supported by the observation that both products yield identical difference spectra upon incubation with apo-Cph1 (Fig. 5C). By contrast with the other two BV isomers, BV III␣ was not metabolized by PcyA. This result is interesting in view of the observation that BV III␣ has the highest binding affinity for PcyA of the three BV isomers (Fig. 3C). The results of the bilin substrate specificity experiments are summarized in Table III.

PcyA Is a Monomeric Enzyme That Forms a Stable Porphyrin-like Complex with
Bilins-Among the family of ferredoxindependent bilin reductases, PcyA is unique in its ability to catalyze the four-electron reduction of BV IX␣ (4). Like oat phytochromobilin synthase, a ferredoxin-dependent bilin reductase that converts BV IX␣ to P⌽B in plants (14), PcyA is a monomeric enzyme. BV binding neither promoted PcyA dimerization nor oligomerization, suggesting that the distinct spectral properties of the three PcyA-BV complexes studied here reflect the unique protein environment and conformation of the bound bilin. The observed spectral features (i.e. long to short wavelength absorption ratio Ͻ1) indicate that bilins bind to PcyA in a cyclic, porphyrin-like configuration, as opposed to the more extended configurations found in phytochromes and phycobiliproteins (17). This cyclic configuration precludes simultaneous protonation of both B-and C-ring nitrogen atoms of the bilin prosthetic group due to steric crowding (see Fig. 6). This conclusion is further supported by the observed lack of fluorescence of the PcyA-BV complex as efficient proton transfer between hydrogen-bonded pyrrole rings would be expected to quench the excited state of the PcyA-BV complex.
Interestingly, a long wavelength shoulder was detected in the spectra of the PcyA complexes of BV IX␣ and BV XIII␣. Since this shoulder was not observed for BV III␣, a nonmetabolized PcyA substrate analog, we speculate that this new absorption band corresponds to a distinct bilin-PcyA interaction, which reflects the ability of bilin to be reduced (i.e. hydrogen bonding, protonation, or aromaticinteraction). Beale and Cornejo (18) has established that reduction of BV in the rhodophyte C. caldarium is Fd-mediated. This result was later confirmed with the cloning of the bilin reductase family and the demonstration that all bilin reductases are Fd-dependent enzymes (4,10). The present studies revealed that PcyA exhibits a preference for plant-type [2Fe-2S] Fds. Fldx, a two-electron acceptor that also can undergo two succes-  sive low potential single electron reductions (i.e. Ϫ413 mV at pH 7, 25°C) (19), also supported PcyA activity. Under the conditions tested here, the Fldx-dependent activity was ϳ13% of that of Fd. More recently, we have been able to increase the Fldx-mediated PcyA activity up to 50% of that of Fd by increasing the pH of the assay buffer to 8.5 and adding 100 mM KCl. 3 Based on these results, it is conceivable that Fldx can functionally substitute for Fd to drive PcyA activity under iron-limiting conditions in vivo (20).

PcyA Prefers Plant Type [2Fe-2S] Ferredoxins-Pioneering work by
The observation that the 2[4Fe-4S] ferredoxin from C. pasteurianum poorly supports PcyA-mediated BV reduction needs to be interpreted with the following caveats. The PcyA assay requires a spinach FNR-containing ferredoxin reducing sys-tem, which may be unable to effectively reduce the C. pasteurianum Fd, thereby limiting PcyA activity. Since spinach Fd and C. pasteurianum Fd have the same redox potential around Ϫ420 mV, C. pasteurianum Fd conceivably could support PcyA activity, assuming the right reducing system is present. In this regard, C. pasteurianum Fd serves as an electron acceptor in the anaerobic oxidation of pyruvate (21), the components of which may be able to drive C. pasteurianum Fd-dependent PcyA activity.
The Endogenous Reductant for PcyA in E. coli May Be Flavodoxin-Recent studies reporting the assembly of holophytochrome and holophycobiliproteins in E. coli (22)(23)(24)(25) indicate that PcyA can use naturally occurring reductants in living cells. E. coli cells possess several possible reductants. E. coli Fd is an adrenodoxin-type [2Fe-2S] ferredoxin, which genetic analyses have shown performs an essential role in the maturation of various iron-sulfur proteins (26). Indeed, E. coli Fd is more structurally related to the adrenodoxin-type ferredoxins (i.e. bovine adrenodoxin and P. putida putidaredoxin) than to planttype Fds (27). As such, E. coli Fd probably functions as a component of the complex machinery responsible for the biogenesis of Fe-S clusters. Based on the observation that the PcyA-mediated catalysis is poorly supported by the putidaredoxin system (see Table I), we hypothesize that engineered PCB biosynthesis in E. coli uses a different reducing system. Other than this adrenodoxin-type ferredoxin, the E. coli genome possesses two Fldx genes and a flavorubredoxin gene (28). In light of the data presented here, we propose that the biosynthesis of PCB in E. coli is driven by one of the two Fldxs.
PcyA-mediated 18-Vinyl Reduction Precedes A-ring Reduction-The identification of 18 1 ,18 2 -DHBV IX␣ as an intermediate in the conversion of BV to PCB has established that D-ring exovinyl reduction precedes A-ring reduction. PcyA is therefore composed of two separate activities mediated by a 18 1 ,18 2 -DHBV:ferredoxin oxidoreductase and a PCB:ferredoxin oxidoreductase. This double bond specificity of PcyA presumably ensures that P⌽Bs are never produced in PcyA-containing cyanobacteria or red algae, the production of which might lead to misincorporation of P⌽Bs into their phycobiliproteins. We speculate that P⌽B-containing phycobiliproteins would be more susceptible to photooxidative damage than the natural PCB-containing antennae of these organisms due to the presence of the reactive exovinyl group on the former. Evolution of P⌽B-producing bilin reductases, such as HY2, would therefore prove a selective disadvantage to these organisms, a selection pressure that would not apply to terrestrial plants that lack phycobiliproteins. In this regard, it will be of interest to clone the genes for these enzymes from the green alga M. caldariorum, which mediate the conversion of BV to PCB via the intermediacy of P⌽B (5).
Through examination of substrate analogs, which include the unnatural XIII␣ and III␣ isomers of BV and the A-ring reduced phytochromobilin isomers, (3Z)-and (3E)-P⌽B, our studies have provided insight into the catalytic specificity of 3 S.-L. Tu, unpublished data.
Not metabolized a The identity of this product was ascertained by analogy (relative retention times on the HPLC) and in comparison with published spectra (16).
FIG. 5. PcyA-mediated metabolism of BV-analogs. A, HPLC product profiles for BV analogs after metabolism by PcyA for 30 min as described under "Experimental Procedures." Elution position of BV IX␣, XIII␣, and III␣ are indicated by arrows. B, absorbance spectra of BV XIII␣ (dashed line) and the two iso-P⌽B reaction products (solid lines) are shown. C, phytochrome difference spectra assay following assembly of apo-Cph1 with BV XIII␣ before (solid line) and after PcyAmediated catalysis (dashed line) are shown. The difference spectra of the PCB adduct of apo-Cph1 (i.e. Cph1(PCB)) is shown for comparison.
PcyA. Of the two unnatural BV analogs, only BV XIII␣ was metabolized by PcyA, yielding the two-electron reduced iso-P⌽B product (both (3Z)-and (3E)-isomers). Since BV XIII␣ is symmetrical and lacks the exovinyl group found on BV IX␣, this result indicated that PcyA-mediated A-ring reduction can occur in the absence of exovinyl reduction. The apparent lack of the reduction of the second endovinyl group of BV XIII␣ also indicated that iso-P⌽B is a poor PcyA substrate. By contrast to the IX␣ and XIII␣ isomers, BV III␣ was not metabolized by PcyA. This result was unexpected, given the sequence of PcyAmediated vinyl reductions of BV IX␣. In this regard, BV III␣ is symmetrical, possessing two exovinyl groups, one or both of which should have been reduced by PcyA. Moreover, our results indicate that BV III␣ binds to PcyA with the highest affinity of the three BV isomers tested. These data indicate that bound BV III␣ is not properly oriented within the enzyme's bilin binding site for catalysis. Our studies show that (3Z)-P⌽B can be metabolized by PcyA, yielding a mixture of PCB isomer products, whereas (3E)-P⌽B is not a substrate for PcyA. These data indicate a strong influence of the geometry of the 3-ethylidene moiety on catalysis. Whether this is due to a positioning defect or to a lack of binding of (3E)-P⌽B to PcyA remains to be determined. Taken together, these studies suggest that proper substrate positioning/activation within the enzyme is a prerequisite for catalysis.
A Radical Mechanism for Bilin Reduction by PcyA-Four major Fd-dependent enzymes have been characterized to date: FNR, Fd:nitrite oxidoreductase, glutamate synthase, and Fd: thioredoxin reductase (29). All of these enzymes contain redoxactive cofactors including FAD (FNR), iron-sulfur clusters (glutamate synthase, nitrite reductase, and sulfite reductase), and siroheme (nitrite reductase). By contrast with these Fd-dependent enzymes, PcyA appears to lack a metal or flavin cofactor that can mediate single electron transfers. For these reasons, we propose that the PcyA-mediated reduction of BV proceeds via bilin radical intermediates as depicted in Fig. 6.
Based upon the absorption spectrum of the PcyA⅐BV complex, the bilin substrate is depicted in a cyclic conformation within the protein cavity (see Fig. 6). The lack of photochromism of the PcyA⅐BV complex can be rationalized by the binding of the terminal pyrrolinone A-and D-rings into the protein matrix with its propionate side chains extending toward the solvent. This porphyrin-like configuration not only would sterically prevent photoisomerization of the C5 and C15 double bonds but would also bury reactive radical intermediates within the protein matrix, thus minimizing side reactions with molecular oxygen. This substrate binding model is consistent with the broad substrate specificity of the extended bilin reductase family, which includes the enzyme RCCR that metabolizes a chlorophyll catabolite with monomethyl ester and isocyclic ring substituents (30). In this regard, the hypothesis that the bilin reductase family may have evolved from ancestors that metabolized (Mg)-porphyrins remains a intriguing possibility.
As shown in Fig. 6 (step 1), we envisage that bilin reduction occurs by binding of reduced Fd to the PcyA-BV complex, followed by electron transfer to the bound bilin and proton transfer from a protein residue labeled D 1 -H to generate a neutral radical shown in step 2. The benzylic position would help to stabilize this radical by resonance within the extended tetrapyrrole -system, until a second electron and proton transfer, shown in steps 2 and 3, occurs to produce the intermediate 18 1 ,18 2 -DHBV IX␣. The hypothetical proton donors, D 1 -H and D 2 -H, could either be carboxylic acids (i.e. Asp or Glu), sulfhydryls (i.e. Cys), phenolics (i.e. Tyr), or even protonated nitrogen residues such as histidine or lysine. It is also possible that protons are derived from bound water molecules that are protonated by appropriate protein residues. For all of these protein residues except for histidine or lysine, proton transfer would be accompanied by an increase in negative charge, which would be a reasonable "driving force" for the release of product. Since PcyA kinetically reduces the intermediate 18 1  Phycocyanobilin:Ferredoxin Oxidoreductase without its release, we hypothesize that D1 and D2 are protonated histidine and/or lysine residues.
We propose that the subsequent reduction of the A-ring of 18 1 ,18 2 -DHBV IX␣ proceeds in a similar fashion, generating another resonance-stabilized bilin radical intermediate shown in Fig. 6 (steps 4 -6). For this transformation, we hypothesize that the proton-donating residues are carboxylic acids, sulfhydryls, and/or phenolics, which would generate negative charge within the bilin pocket, thereby promoting release of the PCB product (Fig. 6, step 6). Experiments to detect potential radical intermediates by electron spin resonance spectroscopy and to identify putative proton-donating residues within the PcyA polypeptide by site-directed mutagenesis are in progress. With this experimental approach, we hope to elucidate the molecular basis for the unique double bond reduction specificities of the different members of the extended bilin reductase family (4) and ultimately to engineer novel specificity of this important family of enzymes.