BciD Is a Radical S-Adenosyl-l-methionine (SAM) Enzyme That Completes Bacteriochlorophyllide e Biosynthesis by Oxidizing a Methyl Group into a Formyl Group at C-7*

Green bacteria are chlorophotorophs that synthesize bacteriochlorophyll (BChl) c, d, or e, which assemble into supramolecular, nanotubular structures in large light-harvesting structures called chlorosomes. The biosynthetic pathways of these chlorophylls are known except for one reaction. Null mutants of bciD, which encodes a putative radical S-adenosyl-l-methionine (SAM) protein, are unable to synthesize BChl e but accumulate BChl c; however, it is unknown whether BciD is sufficient to convert BChl c (or its precursor, bacteriochlorophyllide (BChlide) c) into BChl e (or BChlide e). To determine the function of BciD, we expressed the bciD gene of Chlorobaculum limnaeum strain DSMZ 1677T in Escherichia coli and purified the enzyme under anoxic conditions. Electron paramagnetic resonance spectroscopy of BciD indicated that it contains a single [4Fe-4S] cluster. In assays containing SAM, BChlide c or d, and sodium dithionite, BciD catalyzed the conversion of SAM into 5′-deoxyadenosine and BChlide c or d into BChlide e or f, respectively. Our analyses also identified intermediates that are proposed to be 71-OH-BChlide c and d. Thus, BciD is a radical SAM enzyme that converts the methyl group of BChlide c or d into the formyl group of BChlide e or f. This probably occurs by a mechanism involving consecutive hydroxylation reactions of the C-7 methyl group to form a geminal diol intermediate, which spontaneously dehydrates to produce the final products, BChlide e or BChlide f. The demonstration that BciD is sufficient to catalyze the conversion of BChlide c into BChlide e completes the biosynthetic pathways for all “Chlorobium chlorophylls.”

GSB that synthesize either BChl d or BChl c are green in color, whereas those containing BChl e or the unnatural BChl f are brown. BChl e only differs from BChl c by the presence of a formyl group instead of a methyl group at the C-7 position.
Similarly, BChl f is identical to BChl d except for the C-7 formyl group. The available genome sequences of brown-colored GSB contain homologs of all of the genes required for BChl c biosynthesis; thus, the BChl e biosynthetic pathway is predicted to proceed via the same reactions as that for BChl c until the oxidation of the methyl group at the C-7 position (18,24,25). A similar oxidation occurs at the C-7 position during the biosynthesis of Chl b from Chl a; Chl b differs from Chl a by the presence of a formyl group instead of a methyl group at the C-7 position (36). In the case of Chl b, this reaction is catalyzed by an oxygen-dependent enzyme, chlorophyllide a oxygenase (37,38). However, because GSB are strictly anaerobic organisms and their genomes do not encode a homolog of this oxygenase, an as yet uncharacterized enzyme (or enzymes) capable of oxidizing the C-7 methyl group must be responsible for this reaction in the biosynthesis of BChl e.
Comparative genomics of brown-colored and green-colored GSB previously identified a conserved gene cluster that could potentially encode an enzyme involved in the conversion of BChlide c to BChlide e (3,24,25,39). Similar versions of this gene cluster occur in all BChl e-producing strains (supplemental Fig. S1), and recent evidence suggests that this conserved gene cluster can be horizontally transferred among GSB by the action of bacteriophages (40). The smallest of these clusters contains about 11 genes extending approximately from a putative radical S-adenosyl-L-methionine (SAM) methyltransferase (bchQ2) similar to bchQ on one border to a gene encoding a homolog of isoprenylcysteine carboxyl methyltransferase, an enzyme that methylates the C-terminal carboxyl group of prenylated proteins (41), near the other border. Previous studies showed that one gene in this cluster, cruB, encodes a ␥-carotene cyclase that is essential for the synthesis of ␤-carotene and isorenieratene, carotenoids that are uniquely and nearly universally produced by brown-colored GSB (42). This gene cluster also includes a third homolog of bchF and bchV, genes that encode C-3 vinyl hydratases (21,(27)(28)(29).
The conserved gene cluster in brown-colored GSB notably contains an open reading frame, designated as bciD, which encodes a putative radical SAM enzyme. Radical SAM enzymes are capable of performing a variety of radical-mediated reactions under anoxic conditions (43), and they often act as substitutes for oxygen-dependent enzymes in anaerobic organisms. For example, BchE is a radical SAM enzyme that catalyzes the 6-electron oxidation and cyclization of the isocylic E ring in chlorins in anaerobic chlorophototrophs (3,18,24,25). BchE catalyzes the same reaction as an oxygen-dependent monooxygenase, AcsF, during chlorophyll biosynthesis in aerobic and microaerophilic chlorophototrophs (3,24,25,44,45).
When the bciD gene was inactivated by natural transformation in Chlorobaculum limnaeum strain RK-j-1, the resulting strain could no longer synthesize BChl e but instead produced BChl c (46). Although this result shows that BciD is necessary for the synthesis of BChlide e, it does not show that BciD alone is sufficient to catalyze this transformation. Repeated attempts to express the bciD gene, as well as some of the surrounding genes from the conserved gene cluster shown in Fig. 2, in Chlorobaculum tepidum uniformly failed to produce a strain that could synthesize BChl e. 3 Because the phenotype of the bciD mutant reported by Harada et al. (46) was not verified by complementation, these failures led us to question whether BciD was directly involved in the synthesis of BChl e.
In this study, we verified that inactivation of bciD in a second strain of C. limnaeum, strain DSMZ 1677 T , also led to the synthesis of BChl c in the resulting mutant. The bciD gene was heterologously expressed in Escherichia coli, and the resultant protein was purified and characterized. The purified enzyme contains a single [4Fe-4S] cluster, and when incubated with S-adenosyl-L-methionine (SAM) and sodium dithionite, it catalyzed the conversion of BChlide c into BChlide e and BChlide d into BChlide f, and it produced 5Ј-deoxyadenosine as a by-product. These results demonstrate that BciD is a radical SAM enzyme and that it is sufficient to catalyze these two transformations. Thus, the biosynthetic pathway for all Chlorobium Chls is now complete.

Results
Inactivation of bciD in C. limnaeum-To confirm the phenotype of the bciD mutant described by Harada et al. (46), we insertionally inactivated the bciD gene in C. limnaeum 1677 T . The mutant was made by conjugation from E. coli strain S17-1 carrying the plasmid pCLCON::bciD with C. limnaeum 1677 T . This plasmid contained a single homologous region completely internal to the bciD gene. After a single homologous recombination event, this should disrupt the bciD gene as shown in FIGURE 1. Scheme showing the biosynthetic pathway leading from Chlide a to BChl c. The side chain at C-8 (R-8) can be ethyl, n-propyl, isobutyl, or neo-pentyl. The side chain at C-12 (R12) can be methyl or ethyl. The C-8 ethyl group of 3-vinyl-BChlide d is methylated by BchQ, and the C-12 methyl group of 3-vinyl-BChlide d is methylated by BchR (22). The asterisk at C-3 1 indicates a chiral carbon center that can have either R-or S-stereochemistry. For further details, see the Introduction (3,18,24,25).
supplemental Fig. S2A. Transconjugants were screened by PCR with primers bciDtestF and aadAtestR flanking one border of the insertion site (Table 1). Positive transconjugants produced a PCR amplicon of ϳ1.1 kb as expected (supplemental Fig. S2B), whereas wild-type cells did not produce an amplicon. The resulting 1.1-kb PCR product was verified by DNA sequencing.
Cultures of the bciD mutant of C. limnaeum were green in color, whereas wild-type cultures are brown. To confirm that this color change was due to a change in the BChl content of the bciD mutant strain, we compared the UV-visible absorption spectra of whole cells and of the methanol-extracted pigments of wild-type C. limnaeum, the bciD mutant of C. limnaeum, and wild-type C. tepidum, a closely related BChl c-containing GSB (Fig. 2) (47). Whole-cell spectra ( Fig. 2A) of the bciD mutant very closely matched the spectrum of wild-type C. tepidum. The only notable difference was observed around 515 nm, where C. tepidum appears to have greater absorbance due to its greater carotenoid content. Similar to C. tepidum, the Q y absorption peak of the green-colored, bciD mutant strain shifted to longer wavelength, whereas the Soret absorption peak of the mutant shifted to shorter wavelength. The bciD mutant also lacked the secondary absorption peak at around 525 nm that is typical of brown-colored, BChl e-containing strains. The UV-visible absorption spectra of the extracted pigments of the bciD mutant again matched that of C. tepidum except for the carotenoid region (450 -500 nm), where C. tepidum again exhibited greater absorbance (Fig. 2B). These obser-vations agree with those reported by Harada et al. (46) and confirm that inactivation of bciD in the C. limnaeum strain DSMZ 1677 T results in loss of the ability to synthesize BChl e.
Purification and Characterization of BciD-His 6 -To clarify the role of BciD in the synthesis of BChl e, BciD from C. limnaeum strain DSMZ 1677 T was overproduced as a C-terminally His 6 -tagged protein in E. coli. The isc operon from plasmid pDB1282 was co-expressed with bciD to promote [Fe-S] cluster assembly (48). Protein purification was monitored by SDS-PAGE (Fig. 3). After induction, cells contained an abundant protein with an apparent mass of about 44 kDa (Fig. 3, lane 1), which is similar to the expected mass of BciD-His 6 (47.3 kDa). Cells were lysed under anoxic conditions, and the lysate was subjected to ultracentrifugation to separate the soluble fraction from unbroken cells, membranes, and other debris. Although some of the protein formed inclusion bodies (data not shown), a sufficient amount of the protein of interest was observed in the soluble fraction (Fig. 3, lane 2) to justify continuing with the purification. The protein was purified by nickel affinity chromatography under strictly anoxic conditions. After concentration and buffer exchange, a relatively pure protein of roughly the expected size was obtained (Fig. 3, lane 7). This protein was subjected to in-gel trypsin digestion, and the resulting peptides were analyzed by mass spectrometry, and the analysis confirmed that the protein was BciD-His 6 .
As expected for a radical SAM protein, which by precedent is predicted to contain at least one [4Fe-4S] cluster, the purified BciD-His 6 protein was brown in color. Fig. 4 shows the UVvisible absorption spectrum of the as-isolated protein, which had broad, weak absorbance characteristics of an S 3 Fe charge transfer band with a maximum at about 400 nm. After reconstitution with iron and sulfide as described under "Experimental Procedures," this absorbance band increased substantially (roughly 2-fold), indicating that the Fe/S cluster binding site in the as-isolated protein was not completely occupied. To confirm the presence of the predicted [4Fe-4S] cluster, we measured the EPR spectra of as-isolated and reconstituted BciD. Fig. 5 shows the difference spectra of protein samples reduced with 5 mM dithionite minus the spectra of the untreated samples. In the untreated (oxidized) samples, no EPR resonances were observed; however, after reduction, as seen in the difference spectra, both the as-isolated and the reconstituted protein showed spectral features with axial symmetry and g values of 2.038 and 1.945. These spectra correspond to those typical of proteins with a single, reduced [4Fe-4S] cluster as predicted for this putative radical SAM enzyme.
BciD-His 6 Activity Assay with BChlide c-Activity assays for BciD-His 6 were performed under strictly anoxic conditions at FIGURE 2. Absorbance spectra of whole cells and pigment extracts. A, whole-cell absorbance spectra of the wild-type (solid line) and bciD mutant (dotted line) strains of C. limnaeum compared with wild-type C. tepidum (dashed line). B, absorbance spectra of pigments extracted with 100% methanol from the wild-type (solid line) and bciD mutant (dotted line) strains of C. limnaeum compared with wild-type C. tepidum (dashed line). To facilitate the comparison, the spectra were normalized at the Q y absorbance maxima for A and at the Soret maxima for B. AU, absorbance units. room temperature. Purified BciD-His 6 was incubated with a mixture of methylation homologs of BChlide c or BChlide d, SAM as a cofactor, and sodium dithionite as reductant. Control reactions without added BciD-His 6 were always prepared in parallel. The pigments present in the mock control reaction were compared with those present after incubation with the enzyme by reversed-phase HPLC, and the masses of the products and reactants were determined by mass spectrometry in parallel reactions. Fig. 6A shows the HPLC elution profile of the pigments present in the samples containing BChlide c as the substrate. The dotted line contains a series of peaks that make up the substrate; the peaks at 22.3, 26.1, 29.0, 32.3, and 35.0 min account for the various methylation homologs of BChlide c, whereas the peaks at 36.9, 39.6, 40.3, and 43.0 min represent the methylation homologs of BPheide c, a breakdown product of BChlide c from which the central magnesium atom has been lost. As shown in Fig. 6B, the UV-visible spectra of BChl c in methanol and the in-line absorption spectrum of the major BChlide c substrate peak eluting at 26 min are essentially identical. The solid line in the HPLC elution profiles in Fig. 6A shows a series of peaks that are derived from reaction products and unreacted substrate molecules. The peaks at 12.1, 15.3, 16.6, 18.7, 20.0, and 21.6 min uniquely occur in the elution profile of the products ( Fig. 6A and Table 2). The earlier elution times of the product peaks are consistent with the increased hydrophilicity expected for BChlide e due to replacement of the methyl group by the formyl group at the C-7 position. In Fig.  6C, the UV-visible spectra of the major 20-min product peak and the 26-min substrate peak are compared. Compared with the substrate BChlide c, the Q y absorption peak of the product is blue-shifted, and its Soret peak is red-shifted, as expected for BChlide e. In Fig. 6D, the spectra of the 20-min product and BChl e in methanol are compared, which confirms that the reaction product is BChlide e. Mass spectrometry showed that the masses of the BChlide c substrate peaks increase in 14-Da increments as they become more hydrophobic because of the additional methyl groups of the homologs ( Table 2). Correspondingly, the product peaks are shifted toward earlier elution times due to greater hydrophilicity, and these compounds had masses that were 14 Da larger than the corresponding substrate peak, which diminished in intensity during the reaction ( Table  2). This is consistent with the results expected if the methyl group of BChlide c was transformed into a formyl group to yield BChlide e.
In a separate analysis by HPLC, SAM was consumed, and 5Ј-deoxyadenosine was formed in the complete reaction containing BciD-His 6 , but these changes were not observed in a control reaction lacking the enzyme (Fig. 7). All of these observations are consistent with the expected functioning of a radical SAM enzyme that converts BChlide c into BChlide e.
The minor product peak at 15.2 min is particularly noteworthy. This product is obviously more hydrophilic than the major BChlide e product peak at 20 min. The in-line absorption spectrum of this product, with maxima at 441 and 666 nm ( Fig. 6E and Table 2), more closely resembled the spectrum of the BChlide c substrate than the BChlide e product but was nevertheless distinct. The mass of this compound was 2 Da larger than that of the BChlide e product at 20 min. These properties are consistent with those expected for 7 1 -hydroxy-BChlide c. These observations strongly imply that BciD is a radical SAM enzyme that acts by catalyzing two consecutive hydroxylation reactions of the C-7 methyl group.
BciD-His Activity Assay with BChlide d-A similar activity analysis is shown in Fig. 8 for the complete reaction and control with BChlide d as the substrate. Similar to the results shown for BChlide c in Fig. 6, the substrate contained methylation homo-  The spectra for as-isolated (solid line) and reconstituted BciD-His 6 are shown. To facilitate comparison, the spectra were normalized at the UV absorbance maxima. AU, absorbance units.

FIGURE 5. EPR spectra of as-isolated and reconstituted recombinant
BciD-His 6 . EPR spectra of as-isolated (solid line) and reconstituted recombinant BciD-His 6 (dotted line). Each spectrum represents the difference spectrum between the reduced sample and the untreated (oxidized) sample. The samples were reduced by the addition of 5 mM sodium dithionite. The concentration of reconstituted BciD was 25 M, and the concentration of as-isolated BciD was 55 M. Spectrometer conditions are provided under "Experimental Procedures." The g values of the axial signal are indicated above the spectra. AU, absorbance units; mT, millitesla.
logs of BChlide d and small amounts of BPheide d (Fig. 8A), and as expected, the UV-visible spectrum of the major substrate peak at 21.6 min matched that of BChl d in methanol (Fig. 8B). The major product peak at 18.9 min differs from BChlide d (Fig.   8C) but matches the spectrum of BChl f, as expected for the conversion of BChlide d to BChlide f (Fig. 8D). The main product peak appears at 18.9 min and appears to come from the second most abundant substrate peak at 25.4 min. The shift to  Table 2.  (Table 3). These data establish that BciD is not only necessary for the conversion of BChlide c and d into BChlide e and f, respectively, but it is sufficient to catalyze these transformations. As observed with BChlide c as substrate, a product peak at 14 min was probably due to 7 1 -hydroxy-BChlide d. This product was more hydrophilic than the BChlide f products and had an absorption spectrum with maxima at 433 and 653 nm (Table  3 and Fig. 8E) that was more similar to BChlide d than to BChlide f. These data strongly imply that BciD-His 6 sequentially hydroxylates the C-7 methyl group twice during the synthesis of BChlide f.

Discussion
The goal of this study was to clarify the role of BciD in the biosynthetic pathway of BChl e. Previous studies (46) had shown that inactivation of the bciD gene in C. limnaeum prevented the biosynthesis of BChl e, resulting in production of BChl c in the mutant. However, repeated attempts to express bciD in C. tepidum failed to produce strains capable of BChl e synthesis or related compounds (e.g. 7 1 -hydroxy-BChl c). Therefore, we first verified the phenotype reported for a bciD mutant of C. limnaeum strain RK-j-1 (46). The results presented in Fig. 2 agree well with those from the previous study and show that inactivation of bciD of C. limnaeum DSMZ 1677 T also produced a BChl c-producing mutant strain.
The EPR spectra of recombinant BciD-His 6 had features that are consistent with the presence of a single [4Fe-4S] cluster in the enzyme. In radical SAM enzymes, the SAM-activating [4Fe-4S] cluster has a binding motif consisting of three cysteines, and the canonical motif is CXXXCXXC. Supplemental Fig. S3 shows a multiple sequence alignment of BciD from four brown-colored, BChl e-producing GSB. These proteins contain five conserved cysteine residues at positions 78, 124, 133, 136, and 261; we propose that the cysteines at positions 124, 133, and 136 are ligands to the observed [4Fe-4S] cluster. The proposed roles of these three cysteine residues in ligating the [4Fe-4S] cluster in BciD will be tested by site-specific mutagenesis in future studies. In the presence of SAM and dithionite, BciD-His 6 converted BChlide c and BChlide d to BChlide e and BChlide f, respectively. We additionally observed the formation of more hydrophilic products. Based upon the elution properties, absorption spectra, and masses of these compounds, we propose that they are methylation homologs of 7 1 -OH BChlide c and 7 1 -OH BChlide d, respectively (Tables 2 and 3). Finally, we observed that BciD consumed SAM as a cofactor and produced 5Ј-deoxyadenosine as a by-product of the reaction. Taken together, these observations lead us to conclude that BciD is a radical SAM enzyme with a non-canonical cysteine motif of three cysteine residues that ligates a SAM-activating [4Fe-4S] cluster and that catalyzes the conversion of the C-7 methyl group of BChlide c or d to the formyl group of BChlide e or BChlide f via consecutive hydroxylation reactions (Fig. 9).
Given that BciD appears to be sufficient to convert BChlide c into BChlide e in vitro, it is still a mystery why expression of bciD in the BChl c-containing GSB C. tepidum has repeatedly failed to produce BChls containing either hydroxymethyl or formyl side chains at C-7. One possible explanation is that C. limnaeum and C. tepidum have significantly different optimal growth temperatures, and BciD might not be active at higher temperature. C. tepidum is routinely grown at 40 -48°C in the laboratory, whereas C. limnaeum is typically grown at 25-30°C. However, we were still able to observe BciD-His 6 activity in vitro at 40°C (data not shown), which seemingly excludes this explanation. Alternatively, although it seems unlikely, one could postulate that there is no appropriate electron donor for this reaction in C. tepidum. Obviously, sodium dithionite is not the biological electron donor for this reaction in vivo in C. limnaeum. GSB, including C. tepidum, contain small, highly abundant soluble ferredoxins that typically serve this function (3,49,50). A counterargument to this point is that there is no gene encoding a ferredoxin or other potential electron donor protein encoded in the gene cluster that is specifically found in brown-colored, BChl e-producing GSB strains. Although it again seems unlikely because C. tepidum produces many Fe/S proteins, it is possible that the [4Fe-4S] cluster of BciD is unable to assemble, or that BciD is unable to fold properly, in the context of the C. tepidum cytoplasm.
An alternative explanation is that there is a biosynthetic problem downstream from BciD and BChlide e that prevents the accumulation of BChl e in C. tepidum. Specifically, the enzymes responsible for adding the esterifying alcohols to Chls and BChls show considerable specificity for the (B)Chlide substrate but usually have broader specificity for the esterifying alcohol (20,(51)(52)(53). In C. tepidum BchK is responsible for the addition of the farnesyl tail to BChl c, and ChlG and BchG add phytadienol and phytol to Chl a and BChl a, respectively (20). A comparative analysis of the genomes of GSB shows that all have at least three esterifying enzymes: BchK, ChlG, and BchG. Interestingly, C. limnaeum has two paralogs of the bchK gene; one paralog encodes BchK1, which is very similar to the enzyme found in C. tepidum, but paralog BchK2 is much more divergent (supplemental Fig. S4). Moreover, all brown-colored, BChl e-producing GSB species contain a BchK enzyme that is phylogenetically distant from the BchK of C. tepidum. These obser-vations suggest that a possible problem in producing BChl e in C. tepidum might result from an incompatibility between the esterifying enzyme (BchK) in C. tepidum and the BChlide e substrate. If this is correct, then it might be possible to circumvent this problem by expressing the bchK2 gene of C. limnaeum together with bciD in C. tepidum. The construction of a brown-  Table 3 for identities of compounds and additional information.

TABLE 3 Summary of properties and assignments of peaks in BciD reaction with BChlide d (see Fig. 8)
Masses in parentheses are inferred from those identified for the control reaction. ND, not determined; Me, methyl; Et, ethyl, Pr, n-propyl; Ib, isobutyl.  (4,6). In aquatic systems, organisms with chlorosomes containing BChl d typically occur at shallower depths than organisms containing BChl c (12, 54 -56). A similar pattern is observed in benthic mat systems in chlorophototrophic mats associated with alkaline siliceous hot springs in Yellowstone National Park (57,58). When BChl e assembles into supramolecular aggregates in chlorosomes, an emergent absorption band appears at about 525 nm (see Fig. 4) that allows these organisms to very efficiently harvest blue light (6). Such organisms are also typically able to grow at greater depths and at remarkably low light intensities (10 -13, 54 -56). Thus, the ability to synthesize BChl e is associated with the ability to inhabit a light niche that is inaccessible to other chlorophototrophs. Should the bciD gene be horizontally transferred to a recipient capable of BChl c synthesis, the ability to synthesize BChl e would allow that organism to modify its light-harvesting chlorosomes and invade niches with lower light intensities and fewer competitors.
Evidence exists for horizontal gene transfer for the entire gene cluster that includes bciD and cruB (40). In Lake Cisó in Spain, the resident population of GSB recently shifted from green to brown, and a metagenomic analysis showed that the genome of the brown-colored organism was extremely similar to the type strain of Chlorobium luteolum DSM 273 T , a greencolored, BChl c-producing strain (3). Some evidence pointed to a DNA phage as a possible vector for this transfer (40). Interestingly, Chlorobium luteolum DSM 273 T has two bchK genes, one of which is similar to the bchK gene of C. tepidum and one of which (bchK2) belongs to the clade of enzymes that are found in brown-colored, BChl e-producing GSB strains (supplemen-tal Fig. S4). The presence of a bchK2 gene, the product of which can presumably esterify BChlide e with farnesol, might "predispose" Chlorobium luteolum to transformation from a BChl c-producing into a BChl e-producing strain by the acquisition of bciD, although the horizontally transferred, conserved gene cluster does not include a gene for a chlorophyll synthase. These ideas can be tested in future studies by inactivating the bchK1 and bchK2 genes in C. limnaeum as well as by introducing the bciD gene into C. tepidum in combination with each of the two bchK genes encoding the two paralogous esterification enzymes of C. limnaeum.
Questions remain about the detailed mechanism of hydroxylation, the source of oxygen for this reaction (presumably water), and the structure of the BciD enzyme. BciD is the first radical SAM enzyme of its subclass to be characterized. A blastp comparison of BciD against the UCSF structure-function linkage database showed that BciD is most closely related to uncharacterized subgroup 20 and uncharacterized subgroup 11 of the radical SAM superfamily (59). A blastp search of the NCBI protein database with the BciD query returned 928 sequences with an e value Ͻ0.001; however, none of these proteins have been characterized in vitro. The blastp search also showed that BciD has similarity to the following conserved domains: Elp3, rSAM_MSMEG_0568, Radical_SAM, and BioB. These conserved domains correspond to the superfamily denoted Elongator protein 3/MiaB/radical SAM, which contains several radical SAM enzyme families, the protein family radical SAM MSMEG_0568, which has no characterized function, the radical SAM superfamily, and a domain found in biotin synthase and related enzymes. The absence of obviously close relatives makes the BciD enzyme interesting for future study, not just from the perspective of BChl synthesis, but also from the perspective of radical SAM biochemistry more broadly.

Experimental Procedures
Strains Used in This Study-C. limnaeum strain DSMZ 1677 T (47) was obtained from the culture collection of Dr. Johannes Imhoff (University of Kiel, Germany). C. limnaeum is capable of thiosulfate utilization (13), and it was routinely grown on medium CL used for C. tepidum (60,61) at irradiances of 10 -100 mol of photons m Ϫ2 s Ϫ1 provided by either tungsten or cool white fluorescent lamps. C. limnaeum cells were grown at room temperature in CL medium in liquid and the same medium solidified with 1.5% (w/v) Bacto TM Agar (BD Biosciences) in a Coy anoxic chamber under an atmosphere of N 2 /CO 2 /H 2 (80:10:10, v/v/v). Plates were further enclosed in Gas-Pak jars containing thioacetamide as described previously (61). The complete genome of this organism has been sequenced and is available in GenBank TM as accession number CP017305. 4 Routine cloning to produce constructions for gene inactivation were performed using chemically competent ␣-select Escherichia coli cells (Bioline, Taunton, MA). The E. coli strain used for conjugation was S17-1 (62), and E. coli strain BL21 (DE3) was used for expression of the bciD gene. E. coli strains were grown in Luria-Bertani (LB) medium containing antibiotics as required. For overproduction of BciD, the medium was supplemented with ferric ammonium citrate (100 M) and L-cysteine (100 M). Antibiotic concentrations for E. coli were as follows: spectinomycin, 100 g ml Ϫ1 ; ampicillin, 100 g ml Ϫ1 streptomycin, 50 g ml Ϫ1 ; and kanamycin, 50 g ml Ϫ1 .
Cloning and Inactivation of bciD in C. limnaeum-Inactivation of the bciD gene of C. limnaeum strain DSMZ 1677 T was performed by conjugative gene transfer essentially as described (13). A gene-internal region of the bciD gene was amplified by PCR using oligonucleotide primers CLbciDconSphIF and CLb-ciDconSalIR (Table 1) and cloned into the SphI and SalI sites of the conjugation vector pCLCON (13) to create vector pCLCON::bciD (supplemental Fig. S2A) The resulting plasmid was transformed into E. coli strain S17-1. Conjugation between the S17-1 strain and wild-type C. limnaeum was performed as described previously (13), and transconjugants were selected on solid medium containing spectinomycin, streptomycin, and kanamycin. Transconjugant colonies were screened by PCR using oligonucleotide primers CLbciDtestF and aadAtestR ( Table 1) that flank one border of the gene insertion site; amplicons were analyzed by agarose gel electrophoresis (supplemental Fig. S2 B), and positive PCR amplicons were confirmed by DNA sequencing at the Genomics Core Facility (Pennsylvania State University, University Park, PA).
Cloning and Heterologous Expression of bciD in E. coli-The bciD gene from C. limnaeum was amplified via PCR using oligonucleotide primers CLbciDNdeIF and CLbciDEcoRIR ( Table  1). The resulting PCR amplicon and plasmid pET26b were digested with NdeI and EcoRI, and the products were ligated to insert the bciD gene upstream of the sequence for a C-terminal His 6 tag. Site-directed mutagenesis was used to change the bciD stop codon to a leucine codon using the QuikChange II mutagenesis kit (Agilent Technologies, Santa Clara, CA) and oligonucleotide primers CLbciDstopLeuF and CLbciDstop-LeuR (Table 1). The complete coding sequence and the downstream flanking region encoding the His 6 tag in the resulting plasmid, pET26b(bciD-His 6 ), were confirmed by DNA sequencing at the Genomics Core Facility.
The expression and purification of BciD-His 6 was modified from protocols used for other radical SAM enzymes (48). The expression plasmid pET26b(bciD-His 6 ) was transformed into E. coli strain BL21 (DE3) together with plasmid pDB1282, which contains part of the Azotobacter vinelandii isc operon for [Fe-S] cluster assembly under the control of an arabinose repressor and operator (48,63,64). A starter culture of the expression strain was grown overnight at 37°C in LB medium supplemented with ampicillin and kanamycin. A 2.5-liter culture in minimal medium containing 2% (v/v) glycerol, ampicillin, and kanamycin was inoculated with a 25-ml overnight culture. The culture was grown at 37°C with gentle rotation to an OD 600 of 0.2-0.3, and expression of the isc operon of pDB1282 was induced with 0.2% L-arabinose after supplementing the medium with ferric ammonium citrate (100 M) and cysteine (100 M). After the addition of arabinose, cells were grown at 37°C to an OD 600 of 0.7-0.8 and then cooled to ϳ18°C. Finally, BciD-His 6 production was induced by adding isopropyl-␤-D-1thiogalactopyranoside (50 M). After further incubation at 18°C for 16 -20 h, cells were harvested by centrifugation and either immediately transferred to an anoxic chamber or frozen and placed at Ϫ80°C for short term storage before lysis.
BciD Purification-Buffers used for purification of BciD were made in an anoxic chamber by dissolving the solid materials in anoxic water. The cell pellet was thawed on ice in the anoxic chamber, and cells were resuspended in lysis buffer containing 50 mM HEPES, pH 7.5, 300 mM NaCl, 20 mM imidazole, and 10 mM 2-mercaptoethanol. Lysozyme was added to 1 mg ml Ϫ1 , and the suspension was incubated with rocking for 30 min at room temperature. Cells were cooled on ice and then lysed by sonication. Lysed cells were transferred to ultracentrifuge tubes and sealed inside the anoxic chamber before pelleting the cell debris by ultracentrifugation. The tubes were returned to the anoxic chamber, and the soluble fraction was loaded onto a nickel affinity column pre-equilibrated with lysis buffer. The column was washed three times with three volumes of wash buffer. Wash buffer contained 50 mM HEPES, pH 7.5, 300 mM NaCl, 40 mM imidazole, 10 mM 2-mercaptoethanol, and 20% glycerol. BciD was eluted with a minimal volume of elution buffer, which was the same as the wash buffer except for the imidazole concentration (250 mM). The dark brown-colored fractions were pooled and concentrated using an Amicon stirred ultrafiltration cell with a YM-10 filter (10,000 molecular weight cut-off; Millipore, Bedford, MA). The protein was exchanged into storage buffer containing 50 mM HEPES, pH 7.5, 100 mM NaCl, 10 mM dithiothreitol, and 20% glycerol using PD-10 gel filtration columns (GE Healthcare) and stored in aliquots in liquid nitrogen.
Protein Analyses and Verification-The purity of the BciD-His 6 protein was assessed by electrophoresis on 12% (w/v) polyacrylamide gels in the presence of SDS; proteins were stained with Coomassie Blue R-250 (65). The identity of purified BciD-His 6 was verified by in-gel trypsin digestion (66) accompanied by mass spectrometry at the Proteomics and Mass Spectrometry Core Facility (The Pennsylvania State University, University Park, PA). Protein concentrations were determined by the Bradford assay using a BSA standard (67). Quantitative amino acid analysis performed by the University of California-Davis Proteomics Core Facility established that a correction factor of 0.87 was necessary for protein concentrations determined by the Bradford assay.
Reconstitution of Fe/S Cluster-The reconstitution procedure (48,68,69) was carried out on ice with gentle stirring in an anoxic chamber. A final concentration of dithiothreitol (2 mg ml Ϫ1 ) was added to reconstitution buffer (100 mM HEPES, pH 7.5, 300 mM NaCl, 10% (v/v) glycerol), and the solution was incubated for 10 min before adding 300 M FeCl 3 . Purified BciD-His 6 was added to a final concentration of 30 M, and the solution was incubated for 1 h. A Na 2 S solution (100 mM) was slowly added to the protein solution over a 3-h period to produce a final concentration of 300 M, and the resulting reaction mixture was incubated on ice for 12 h. The protein solution was centrifuged at 10,000 ϫ g for 10 min, and the supernatant was concentrated and exchanged into storage buffer. To remove adventitiously bound iron, BciD-His 6 was chromatographed on a Superdex 75 10/300 GL size exclusion column (GE Healthcare) maintained in an anoxic chamber.
Spectroscopic Measurements-UV-visible absorbance spectra of protein and pigment samples were measured on a GENESYS 10 spectrophotometer (ThermoFisher Scientific Corp., Waltham, MA). Spectra of protein samples were measured in quartz cuvettes that were sealed inside an anoxic chamber.
For EPR measurements, samples of as-isolated and reconstituted BciD-His 6 were transferred to EPR tubes in storage buffer. To characterize the presence of Fe/S cluster(s), each protein preparation was measured in both the untreated and reduced states. To reduce the Fe/S cluster(s) fully, dithionite in storage buffer was added to a final concentration of 5 mM to the appropriate protein samples, and the samples were then frozen in liquid nitrogen. EPR spectra were collected using a cylindrical TE 011 mode resonator and a Bruker E500 spectrometer (Bruker Biospin Corp., Billerica, MA) at X-band (9.39 GHz). The temperature was maintained using an ESR 900 liquid helium cryostat and an ITC-4 temperature controller (Oxford Instruments, Concord, MA). EPR spectra were recorded at 15 K at 0.2-milliwatt microwave power with modulation amplitude of 20 G (2 millitesla) at a modulation frequency of 100 kHz. Sixteen scans were collected and averaged for each sample; the difference spectrum was calculated by subtracting the spectrum for the untreated protein from that for the dithionitereduced protein.
Pigment Separation and Preparation of Substrate Compounds-Pigments were separated and analyzed by reversedphase HPLC using a 25 cm ϫ 4.6-mm Discovery 5-m C-18 column (Supelco, Bellefonte, PA) and an Agilent series 1100 HPLC system equipped with a diode array detector (Agilent Technologies, Palo Alto, CA) (70). The data were analyzed using Agilent ChemStation software (revision B.02.01-SR1 6100 series). Pigment samples or reaction products were filtered through a 0.2-m syringe filter, and 10 mM ammonium acetate was added before injection onto the column. Previously described separation methods (71)(72)(73) were modified slightly to separate the BChlide and bacteriopheophorbide (BPheide) components of interest. The column was pre-equilibrated with 12% solvent B (methanol, ethyl acetate, acetonitrile, 50:30:20 (v/v/v)) and 88% solvent A (methanol, 1 M ammonium acetate, 70:30 (v/v)) at a flow rate of 0.75 ml min Ϫ1 at injection and maintained for 5 min. After 5 min, solvent B was increased linearly to 43%, and the flow rate was increased linearly to 1 ml min Ϫ1 at 44 min. Solvent B was linearly increased to 100% from 44 to 55 min and then held constant for 10 min. The column was then returned to 12% solvent B and washed extensively before the next injection.
BChlide c was prepared from a 2-liter culture of the bciD mutant of C. limnaeum (see "Results"), and BChlide d was prepared from a 2-liter culture of the bchU mutant of C. tepidum (12). Cells were harvested by centrifugation, and pigments were extracted with 7:2 acetone/methanol (40 ml) for 1 h at room temperature in the dark with stirring. Cell debris was removed by centrifugation, and the solubilized pigments were purified using a previously described, reversed-phase HPLC method (13). The BChl fractions were collected, pooled, dried under liquid nitrogen, washed with acetone, and finally dried again. To hydrolyze the esterifying alcohol groups of the BChls, the dried pigments were suspended in 4:1 solution of acetone/water containing 0.1 M NaOH, and the solution was incubated in the dark for 3 h (53). One volume of hexane was added to extract unreacted BChls, and the acetone layer was dried under liquid nitrogen and washed three times with acetone. Dried samples were stored at Ϫ20°C until required.
Mass Spectrometry-Mass spectrometric analysis was performed on a Waters Q-TOF Premier quadrupole/time-of-flight mass spectrometer (Waters Corp. (Micromass Ltd.), Manchester, UK). MassLynx TM software version 4.1 was used to operate the mass spectrometer. Samples were introduced into the mass spectrometer using a Waters 2695 high performance liquid chromatograph. The separation was performed using the same HPLC column and solvent gradient described above for analysis of the reaction products. The samples were resuspended in acetonitrile (100 l) and vortexed until dissolved. The injection volume was 25 l. The nitrogen drying gas temperature was set to 300°C at a flow of 7 liters min Ϫ1 , and the capillary voltage was 2.8 kV. The mass spectrometer was set to scan from 400 to 700 m/z in positive ion mode, using electrospray ionization. Data acquisition was performed in the middle of the run, dependent upon the anticipated elution time of the analytes.
Enzyme Activity Assay-Enzyme assays were performed inside an anoxic chamber in the dark at room temperature. The reaction buffer consisted of 100 mM MOPS, pH 8.0, 100 mM NaCl. Tryptophan (200 M final concentration) was added as an internal standard. In addition, a final concentration of 700 M SAM and 1 mM dithionite were added as cofactor and reductant, respectively. Reactions contained a final concentration of 0.2 mg ml Ϫ1 purified BciD; an equal volume of enzyme storage buffer was added to control reactions. BChlide c (final concentration, 0.05 mg ml Ϫ1 ) or BChlide d (final concentration, 0.18 mg ml Ϫ1 ) in acetone was added to the enzyme solution to start the reaction. At appropriate times, aliquots were removed from each reaction for analysis. An equal volume of cold acetone was added to stop the reaction, and precipitated protein was pelleted by centrifugation. Samples were stored on ice or at Ϫ20°C until the pigments were analyzed by reversedphase HPLC.
To analyze SAM-related products, separate aliquots were taken at each time point, and an equal volume of 100 mM H 2 SO 4 was added to stop the reaction. SAM products were separated by a reversed-phase HPLC procedure modified from Lanz et al. (48), using a 150 ϫ 4.6-mm Kinetex 5-m C-18 column (Phenomenex Inc., Torrance, CA) and a Shimadzu UFLC system (Shimadzu Scientific Instruments, Columbia, MD). Reaction products were filtered through a syringe filter (0.2 m) before being injected onto the column. The column was pre-equilibrated with 2% solvent B (HPLC grade acetonitrile) and 98% solvent A (5% methanol, 40 mM ammonium acetate, pH 6.2) at a flow rate of 0.5 ml min Ϫ1 at injection and maintained for 6.5 min. Solvent B was increased linearly to 12% from 6.5 to 25 min, then to 24% from 25 to 33 min, and finally to 50% from 33 to 40 min. Solvent B was maintained at 50% for 10 min and then returned to the starting conditions. Products were monitored by UV-visible detection at 260 nm and were compared with authentic standards for SAM and 5Ј-deoxyadenosine.
Phylogenetic Analyses-Phylogenetic relationships were inferred by the maximum likelihood method based on the JTT matrix-based model (74). The tree with the highest log likelihood value (Ϫ10,582.4753) is shown. Based upon 100 bootstrap resamplings of the data, the percentage of trees in which the associated taxa clustered together is indicated next to the branches. Initial tree(s) for the heuristic search were obtained automatically by applying neighbor-joining and BioNJ algorithms to a matrix of pairwise distances that were estimated using a JTT model and then by selecting the topology with the highest log likelihood value. The tree is drawn to scale, with branch lengths scaled to the number of substitutions per site. The analysis employed 52 amino acid sequences. All positions containing gaps due to insertions/deletions were eliminated from the alignment. The final alignment data set contained 244 positions. Evolutionary analyses were conducted in MEGA7 (75).
Author Contributions-J. L. T. designed and performed the experiments, interpreted the data, and wrote the initial draft of the manuscript; B. H. F. performed the electron paramagnetic resonance measurements; and J. H. G. assisted in the interpretation of the EPR data and edited the manuscript. D. A. B. conceived the project, directed the research, assisted in the interpretation of the data, and wrote and edited the manuscript. All authors approved the final version of the manuscript.