Taming of a Poison: Biosynthesis of the NiFe-Hydrogenase Cyanide Ligands

See allHide authors and affiliations

Science  14 Feb 2003:
Vol. 299, Issue 5609, pp. 1067-1070
DOI: 10.1126/science.1080972


NiFe-hydrogenases have an Ni-Fe site in which the iron has one CO and two CN groups as ligands. Synthesis of the CN ligands requires the activity of two hydrogenase maturation proteins: HypF and HypE. HypF is a carbamoyltransferase that transfers the carbamoyl moiety of carbamoyladenylate to the COOH-terminal cysteine of HypE and thus forms an enzyme-thiocarbamate. HypE dehydrates theS-carbamoyl moiety in an adenosine triphosphate–dependent process to yield the enzyme thiocyanate. Chemical model reactions corroborate the feasibility of this unprecedented biosynthetic route and show that thiocyanates can donate CN to iron. This finding underscores a striking parallel between biochemistry and organometallic chemistry in the formation of an iron-cyano complex.

Hydrogenases catalyze the reversible oxidation of molecular hydrogen into protons and electrons. They are widely distributed among microorganisms, and they provide them with the capacity either to use hydrogen as an energy source or to dissipate excess reducing equivalents in the form of molecular hydrogen. These enzymes have attracted considerable attention, not only because of the distinctive chemical nature of their substrate and the reaction mechanism but also because of their potential biotechnological applications (1).

Two major classes of hydrogenases can be differentiated according to the metal content of the active site cofactor: Fe-hydrogenases and NiFe-hydrogenases. Although the overall structure of their metal centers differs, they share one unusual feature: diatomic, nonproteinaceous iron ligands, namely, carbon monoxide and cyanide. In NiFe-hydrogenases, the iron of the center carries two cyanide and one carbon monoxide moieties (2). The presence of these ligands stabilizes iron in a low oxidation and spin state.

In metal center synthesis and incorporation into proteins, important issues are control of the fidelity of insertion of the correct metal into the cognate target protein, maintenance of a folding state of the protein competent for metal addition, and conformational change to internalize the assembled metal center (3). In this context, the synthesis of the CN and CO ligands of NiFe-hydrogenases poses intriguing questions because of their toxicity in the free state and, in particular, because addition of CO and CN to the metal reflects organometallic chemistry unprecedented in biology.

A total of seven auxiliary proteins are required for the synthesis and incorporation of the metal center in NiFe-hydrogenases (4). One of them, the HypF protein, accepts carbamoyl phosphate (CP) as a substrate and catalyzes both a CP phosphatase reaction in the absence of any other substrate and a CP-dependent hydrolysis of ATP into AMP and inorganic pyrophosphate (PPi). Because the latter reaction is reversible, as shown by a CP-dependent pyrophosphate-ATP exchange, it was postulated that an adenylated CP derivative is an intermediate in the reaction and that the CP phosphatase activity may reflect a side reaction followed in the absence of ATP (5). The necessity of CP for the synthesis of the metal center was then clearly proven by the inability of mutants devoid of CP synthetase activity to mature NiFe-hydrogenases (6).

The formation of an adenylated CP derivative by HypF prompted studies on catalytic self-carbamoylation of the protein. To determine whether a stable carbamoyl derivative of purified HypF protein fromEscherichia coli is formed, a filter binding assay was devised (7) (Fig. 1). The radioactivity of 14C-labeled CP was not stably incorporated into the HypF protein per se (Fig. 1A, second bar). HypF and HypE fromHelicobacter pylori reportedly interact with each other (8), and HypE shares sequence similarity with the PurM protein, which catalyzes an ATP-dependent dehydration reaction during purine biosynthesis (9), a reaction that formally could also transform a carbamoyl into a cyano moiety. Consequently, transcarbamoylation of HypE catalyzed by HypF was studied. Purified HypE protein from E. coli indeed was found to hydrolyze ATP into ADP and inorganic phosphate in the absence of any other substrate (10). When HypE was incubated with the HypF protein from the same organism with different molar ratios in the presence of ATP and CP (Fig. 1A), a stable adduct between radiolabeled CP and the HypE protein was formed. Its radioactivity quantitatively correlated with the amount of HypE, but not HypF, protein used in the assay. The stoichiometry between CP-derived radioactivity and HypE protein was between 50 and 60%. These results support the conclusion that HypF is required for the transfer of the carbamoyl group of CP to HypE. This transfer depends on the cleavage of ATP into AMP and PPi (Fig. 1B). With the nonhydrolyzable analog ADP-CH2-P (in which a methylene group replaces the phosphoanhydride bond) in place of ATP, transfer still took place, but in the presence of AMP-CH2-PP, it was blocked. These results are consistent with HypF functioning as a carbamoyltransferase that requires prior adenylation of CP, which is not possible with AMP-CH2-PP as a substrate.

Figure 1

Modification of the hydrogenase maturation protein HypE by the HypF protein. (A) Binding of radioactively labeled CP to HypF and/or HypE proteins assessed by binding of the proteins to nitrocellulose filters. HypEΔC is a HypE variant lacking the COOH-terminal cysteine residue. (B) HypE (2 μM) and HypF (2 μM) were reacted with radioactive CP in the absence of ATP (bar 1), in the presence of ADP-CH2-P (bar 2) or AMP-CH2-PP (bar 3). Bar 4 presents the result of a standard assay with ATP as substrate, but the assays contained 250 μg bovine serum albumin instead of HypE and HypF. (C and D) Lane 1 contains only HypF; lane 2 contains only HypE. (Lanes 3 to 6): HypF protein (2 μM) was reacted with 6 μM HypE protein in the assay mixture given above, which contained as nucleoside triphosphate (NTP) (a) ATP, (b) AMP-CH2-PP, or (c) ADP-CH2-P. On the right, F in (C) denotes HypF; E indicates HypE, and E*, a dimer of HypE. Denaturation with SDS in the absence of dithiothreitol (DTT) allows more radioactivity to be recovered, which indicates that the adduct is labile to thiols. The altered migration possibly reflects an oxidized form of HypE.

To prove that the HypE protein indeed acts as the acceptor and to study whether binding of the putative carbamoyl residue is covalent, HypE and HypF were incubated with [14C]CP and ATP, and then the assay mixture was denatured by SDS under reducing and nonreducing conditions, separated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE), and blotted onto a nitrocellulose membrane. The blot was autoradiographed (Fig. 1D) and then stained (Fig. 1C) (7). The results show that the radioactivity is transferred to HypE during the reaction and that the binding is stable under denaturing conditions, which suggests covalent attachment.

Apart from the sequence similarity with the PurM protein, HypE carries a PRIC motif as the last four amino acids (Pro-Arg-Ile-Cys), which is invariable in all members of the HypE family. This cysteine residue at the COOH-terminus prompted speculation that its thiol group might be the target for carbamoylation. Indeed, the purified HypE variant with a deletion of this amino acid could not accept the putative carbamoyl group transferred by HypF (Fig. 1A, last bar).

To identify the amino acid residue of HypE that is modified, samples of the reaction mixtures were digested with endoproteinase Asp-N, and the peptides were analyzed by mass spectrometry (Fig. 2) (7). The peptide pattern of an assay containing HypF, HypE, and CP plus ATP differed from that of a control assay lacking CP (11) in the appearance of a single peptide with a mass of 1385.7. This peptide corresponds to the COOH-terminal peptide of the HypE protein, ranging from amino acid 311 to 322 (mass 1360.7), and the mass of the additional peptide is compatible with the presence of a cyano group replacing a hydrogen from an amino acid side chain (Fig. 2A).

Figure 2

Mass and tandem mass spectrometric analysis of HypE protein modified by HypF. (A) Spectrum of a reaction of HypE, HypF, CP, and ATP and (B) like (A) but with ADP-CH2-P instead of ATP, digested with endoproteinase Asp-N. Peptide 311-322 (m/z 1360.7) modified with (A) CN (m/z 1385.7) and (B) CONH2 (m/z 1403.7) is enlarged. (C) depicts the tandem mass spectrum of unmodified peptide 311-322 (m/z 1360.7), (D) of peptide 311-322 modified with CONH2 (m/z 1403.7), and (E) of the peptide modified with CN (m/z 1385.7). y ions modified with CONH2 and CN are labeled with (*) and (#), respectively. For clarity, only the spectrum of the m/z range from 640 to 1500 is depicted. The peptide designation follows the standard nomenclature (22) [for details see the insert in Fig. 2C and (7)].

Because HypE has intrinsic adenosine triphosphatase activity and exhibits sequence similarity to PurM, an enzyme catalyzing an ATP-dependent dehydration reaction, it seemed feasible that the CN modification derives from the dehydration of a carbamoyl precursor. This conjecture was tested by analyzing the peptide pattern of an assay in which ATP was replaced by the nonhydrolyzable ADP-CH2-P. Under this condition, the transcarbamoylation by HypF (requiring cleavage of ATP into AMP and PPi) should still be possible, but the dehydration should be blocked. The mass increment of peptide 311-322 increases from 1360.7 to 1403.7 (Fig. 2B); this proves that HypE indeed carries a carbamoyl modification. To identify the amino acid residue carrying the carbamoyl or the cyano modification, tandem mass spectrometry (MS/MS) of this peptide from assays lacking CP and ATP (Fig. 2C) was performed. The pattern was compared with those from reactions delivering carbamoylated (in the presence of CP and ADP-CH2-P) (Fig. 2D) or cyanated (in the presence of CP and ATP) (Fig. 2E) HypE protein. It was found that, in the MS/MS spectrum of the 311-322 peptide derived from a reaction in which the substrates were present, all y ions but no b ions carried the modification (Fig. 2, D and E). Moreover, one peptide, a11/b11, corresponds to a fragment of amino acids 311 to 321 (lacking the cysteine). Because it also lacks the modification, it is clear that each of the modifications is bound to the COOH-terminal cysteine. Finally, the spectrum shown inFig. 2E demonstrates that HSCN is cleaved off the precursor; therefore, it is the COOH-terminal cysteine that is either cyanated or carbamoylated.

The steps in CN formation catalyzed by the hydrogenase maturation proteins HypF and HypE are summarized in Fig. 3. HypF forms the postulated AMPOCONH2 from which the carbamoyl group is transferred to the COOH-terminal cysteine of HypE (step 1). HypE-carbamoylation (step 1) is followed by the ATP-dependent dehydration by HypE (step 2) and the transfer to the iron atom (step 3).

Figure 3

Biosynthesis of the cyanide ligand. (Step 1), formation of carbamoyladenylate by HypF (F) as delineated from the ATP-PPi exchange reaction, and carbamoyl transfer to the COOH-terminal cysteine of HypE (E). (Step 2), ATP-dependent dehydration involving phosphorylation of the carbonyl oxygen of CP followed by dephosphorylation. (Step 3), chemical transfer of the cyano group to iron.

Chemical models were studied to determine the chemical feasibility of steps 2 and 3 in Fig. 3 and related reactions. The reactions carried out are shown in Fig. 4(7). Thus, treatment ofS-(n-decyl) thiocarbamate with ethyl polyphosphate, PPE (12, 13), gave a 55% yield of the corresponding thiocyanate (14) [Fig. 4(reaction 1)]. Similarly, (η5-C5H5)Fe(CO)2CONH2(FpCONH2) (15) can be dehydrated, and sequential treatment with PPE followed by triethylamine results in a 57% yield of FpCN [Fig. 4 (reaction 2)], despite the previously reported (16) facile fragmentation of FpCONH2cleaving the iron-to-carbon bond. Nevertheless, the biochemical results show that the dehydration of the carbamoyl moiety occurs when bound to sulfur, rather than iron, followed by transfer of the CN moiety to iron. Consequently, chemical modeling of such a transfer was studied. Reduction of phenylthiocyanate by phenylthiolate in the presence of FpBr generated a 50% yield of FpCN [Fig. 4(reaction 3)]. This reaction validates the proposed model (Fig. 3, step 3) in which a nucleophilic CN is transferred to an electrophilic iron center. Because the identity of the iron species in the biological system is unknown, another chemical model system was studied in which the iron is nucleophilic and the CN electrophilic. Treating t-butylthiocyanate with Fp brought about a 20% yield of FpCN (17) [Fig. 4 (reaction 4)]. These chemical models validate the feasibility of the proposed biosynthesis of the CN ligand and its transfer from sulfur to iron (18).

Figure 4

Chemical models for the proposed biosynthesis of LnFeCN. (1) Dehydration ofS-(n-decyl) thiocarbamate to the corresponding thiocyanate with ethyl polyphosphate (PPE). (2) Dehydration of iron carboxamido complex to iron cyanide where Fp is (η5-C5H5)Fe(CO)2. (3) Transfer of a nucleophilic CN group to electrophilic iron atom. (4) Transfer of an electrophilic CN group to nucleophilic iron atom.

In sum, the cysteine residue of HypE undergoes unprecedented biotransformations in serving as the site for dehydration of a carboxamido moiety to CN, in carrying the CN residue, and in transferring the CN to iron. Previously, thiocyanates were rarely encountered in nature. Thiocyanate synthesis in the marine spongeAxinyssa n. sp. has been proposed to involve sulfuration of cyanide to give SCN followed by reaction with a sesquiterpene cation (19), although cyanation of a thiol moiety has also been suggested (20, 21).

Supporting Online Material

Materials and Methods

References and Notes

  • * To whom correspondence should be addressed. E-mail: august.boeck{at}


View Abstract

Stay Connected to Science

Navigate This Article