The lipid-free structure of apolipoprotein A-I: effects of amino-terminal deletions

Deletion mutants of human apolipoprotein A-I (apo hA-I) have been produced from a bacterial expression system to explore the function of the specific domains comprising residues 1-43, 1-65, 88-98, and 187-243, respectively, in the lipid-free conformation and in the lipid-binding mechanism of apo hA-I. Initial studies on apo Delta(1-43)A-I and apo Delta(187-243)A-I have already been reported. To aid purification of these mutants, a histidine-containing N-terminal extension was incorporated (+his); in cases where comparison with the (-his) construct was possible, little effect on the physical properties due to the (+his) extension was found. All mutants have folded structures in their lipid-free state, however these structures differ widely in their relative thermodynamic stability and extent of secondary structure. The mutant with the fewest residues deleted, apo Delta(88-98)A-I(+his), has the least secondary structure (only 34% helix) and is also the least stable (DeltaG = 2.9 kcal/mol). Determined from sedimentation velocity measurements on the lipid-free proteins, all but apo Delta(1-65)A-I(+his) exhibited a range of conformers in solution, which fluctuated around a highly elongated species (dimensions equal to approximately (14-16) x approximately 2.3 nm). Apo Delta(1-65)A-I(+his) exhibited a discrete species which was less asymmetric (dimensions equal to 9 x 2.9 nm). Apo Delta(88-98)A-I(+his) showed extreme heterogeneity with no predominating conformer. Spectroscopic studies (ANS binding and circular dichroism) indicate that there is little difference in the lipid-free structure of the carboxy-terminal deletion mutant, apo Delta(187-243)A-I(+/-his) compared to wild-type (wt) apo wtA-I(+/-his), but substantial differences are observed between wt and the amino-terminal deletion mutants, apo Delta(1-43)A-I, apo Delta(1-65)A-I(+his), and apo Delta(88-98)A-I(+his). In contrast, the lipid-binding properties are impaired for apo Delta(187-243)A-I(+/-his), as measured by dimyristoyl phosphatidylcholine (DMPC) liposome turbidity clearance kinetics and palmitoyloleoyl phosphatidylcholine (POPC) equilibrium binding. Apo Delta(1-43)A-I, apo Delta(1-65)A-I(+his), and apo Delta(88-98)A-I(+his) show lipid affinities statistically similar to apo wtA-I(+his), but significantly defective DMPC clearance kinetics. Interestingly, lecithin:cholesterol acyltransferase (LCAT) activation results correlate qualitatively with the lipid-binding affinity for all mutants but apo Delta(88-98)A-I(+his), suggesting that this mutant has an altered and possibly noncooperative lipid-bound structure as well as an altered lipid-free structure. These results suggest helix 1 (residues 44-65) and helix 10 (residues 220-240) are both required for native lipid-binding properties, while the presence of internal residues, at least helix 3 (residues 88-98), is essential for proper folding of both the lipid-free and lipid-bound conformations. Importantly, studies on apo Delta(88-98)A-I(+his) provide the first experimental evidence that a native-like structure is not necessary for native-like lipid affinity, but apparently is necessary for both DMPC solubilization and LCAT activation. These results provide support for a hypothetical, multistep structure-based mechanism for apo hA-I lipid binding.

Free ricin A chain, proricin, and native toxin have different cellular fates when expressed in tobacco protoplasts

The catalytic A subunit of ricin can inactivate eukaryotic ribosomes, including those of Ricinus communis where the toxin is naturally produced. How such plant cells avoid intoxication has remained an open question. Here we report the transient expression of a number of ricin A chain-encoding cDNA constructs in tobacco protoplasts. Ricin A chain entered the endoplasmic reticulum lumen, where it was efficiently glycosylated, but it was toxic to the cells and disappeared with time in a brefeldin A-insensitive manner, suggesting reverse translocation to the cytosol and eventual degradation. Proricin (the natural precursor form containing A and B chains joined together by a linker sequence) was glycosylated, transported to the vacuole, and processed to its mature form, but was not toxic. Free ricin A chain and proricin were not secreted, whereas free ricin B chain was found entirely in the extracellular medium. The coexpression of ricin A and B chains resulted in the formation of disulfide-linked, transport-competent heterodimers, which were secreted, with a concomitant reduction in the observed cytotoxicity. These results suggest that the production of ricin as a precursor is essential for its routing to the vacuole and for protection of ricin-producing cells.

Expression of mutant dynamin protects cells against diphtheria toxin but not against ricin

Diphtheria toxin is believed to enter sensitive mammalian cells via receptor-mediated endocytosis from clathrin-coated pits, while ricin can enter via both clathrin-dependent and clathrin-independent endocytosis. The present study has confirmed this by determining the toxin sensitivity of COS-7y cells which were transiently overexpressing a trans dominant negative mutant of dynamin, a GTPase required for the budding of clathrin-coated vesicles from the plasma membrane. Cells overexpressing wild-type dynamin showed normal receptor-mediated endocytosis of transferrin and remained sensitive to both diphtheria toxin and ricin. Cells overexpressing a mutant dynamin defective in GTP binding and hydrolysis were unable to endocytose transferrin and were protected against diphtheria toxin, but they remained completely sensitive to ricin intoxication. Treating non-transfected cells or cells overexpressing mutant dynamin with nystatin caused a redistribution of the caveolae membrane marker protein VIP21-caveolin from the cell surface to intracellular locations, but did not affect their sensitivity to ricin. The redistribution of caveolin seen after nystatin treatment may reflect the disappearance of caveolae. If this is the case, caveolae are not responsible for the endocytosis of ricin. An alternative clathrin-independent route may operate for ricin, since cellular uptake, intracellular transport, and translocation into the cytosol remain unaffected when clathrin-dependent endocytosis is effectively blocked.

Structural analysis of apolipoprotein A-I: effects of amino- and carboxy-terminal deletions on the lipid-free structure

An amino-terminal deletion mutant (residues 1-43) and a carboxy-terminal deletion mutant (residues 187-243) of human apoliprotein A-I (apo hA-I) have been produced from a bacterial expression system to explore the importance of the missing residues for the conformation of apo hA-I. Our focus has been to study the lipid-free structure of apo hA-I to understand how discrete domains influence the conformational plasticity of the protein and, by inference, the mechanism of lipid binding. All spectral and physical measurements indicate that both apo delta(1-43)A-I and apo delta(187-243)A-I have folded, tertiary structures. These structures differ in the specific arrangement of helical domains based, in part, on their relative thermodynamic stability, near- and far-UV CD, limited proteolysis, and the accessibility of tryptophans to fluorescence quenchers. In addition, all data indicate that the folded domains of apo hA-I and apo delta(187-243)A-I are very similar. Results from analytical ultracentrifugation suggest that lipid-free apo hA-I and the deletion mutants each exist in a dynamic equilibrium between a loosely folded, helical bundle and an elongated monomeric helical hairpin. The conformational heterogeneity is consistent with significant ANS binding exhibited by all three proteins and could help to explain the facile lipid binding properties of apo hA-I.

Retrograde transport: going against the flow

Certain protein toxins act by catalytically modifying substrates in the cytosol of mammalian cells. To reach this compartment, these proteins undergo retrograde transport from the cell surface, via the Golgi complex, to the endoplasmic reticulum.

Proteolytic processing of ricin A chain is not required for cytotoxicity

[125I]-labeled ricin A chain was endocytosed by macrophages and Vero cells either on its own or as part of the ricin holotoxin. Subsequently [125I]-ricin A chain was recovered from intoxicated cells by immunoprecipitation or acetone precipitation. The recovered protein was shown to have the same electrophoretic mobility as the material supplied to the cells. These data challenge an earlier suggestion that limited intracellular processing of ricin A chain by proteolytic enzymes is required for maximal toxicity.

Ricin A chain can transport unfolded dihydrofolate reductase into the cytosol

Ricin is a heterodimeric protein toxin. The ricin A chain is able to cross the membrane of intracellular compartments to reach the cytosol where it catalytically inactivates protein synthesis. It is linked via a disulfide bond to the B chain, a galactose-specific lectin, which allows ricin binding at the cell surface and endocytosis. To examine the potential of ricin A to carry proteins into the cytosol and the requirement for unfolding of the passenger protein, we connected mouse dihydrofolate reductase (DHFR) to ricin A by gene fusion via a spacer peptide. DHFR-ricin A expressed in Escherichia coli displayed the biological activities of the parent proteins and associated quantitatively with ricin B to form DHFR-ricin. The resulting toxin was highly cytotoxic to cells (4-8-fold less than recombinant ricin). DHFR-ricin cytotoxicity was inhibited by methotrexate, a DHFR inhibitor stabilizing DHFR-ricin A in a folded conformation. The DHFR moiety of DHFR ricin bound to the plasma membrane. Although methotrexate prevented this binding, it did not significantly affect DHFR-ricin endocytosis, which proceeded via ricin B chain. Intoxication kinetics data and a cell-free translocation assay demonstrated that protection of cells from DHFR-ricin cytotoxicity resulted from a selective inhibition by methotrexate of DHFR-ricin A translocation. We conclude that ricin A is a potential carrier of proteins to the cytosol, provided that the passenger protein is able to unfold for transmembrane transport.

Characterization of prokaryotic recombinant Aspergillus ribotoxin alpha-sarcin

The Aspergillus ribonuclease alpha-sarcin is toxic to intact mammalian cells but the mechanism by which it enters the cells to reach its ribosomal RNA substrate is unclear. Here we have compared the cytotoxicity of alpha-sarcin to that of ricin, another catalytic toxin that targets the same rRNA sequence but whose mechanism of cell entry is better understood. Intact ricin binds to cell surface components and enters the cells by receptor-mediated endocytosis, whereas the catalytic polypeptide of ricin (the A chain or RTA) which, like alpha-sarcin, is unable to bind to surface components directly and enters cells by fluid phase uptake. Recombinant alpha-sarcin was produced in Escherichia coli and purified to homogeneity. The protein was soluble, stable and its ability to inhibit in vitro protein synthesis was indistinguishable from that of native alpha-sarcin. Further, recombinant alpha-sarcin had the same in vitro protein synthesis inhibition activity as ricin A chain. The cytotoxicity of alpha-sarcin and ricin A chain to HeLa cells was also the same. The cytotoxicity of alpha-sarcin was due to its RNAase activity rather than to specific membrane effects at the cell surface, since a mutant containing a single substitution at a putative key catalytic residue had reduced ribonuclease activity and an equivalent reduction in cytotoxicity. One interpretation of the data is that a-sarcin enters mammalian cells in the same way as free ricin A chain.

Structural analysis of apolipoprotein A-I: limited proteolysis of methionine-reduced and -oxidized lipid-free and lipid-bound human apo A-I

The domain structures of lipid-free and lipid-bound apolipoprotein A-I (apo A-I) containing reduced and oxidized methionines were analyzed by limited proteolysis. Lipid-free apo A-I is cleaved primarily in the extreme carboxy-terminus and, to a much lesser extent, in the central region of the protein between residues 115 and 136. Oxidation of methionines 112 and 148 to the corresponding sulfoxides in putative amphipathic helices 4 (P99-E120) and 6 (P143-A164), respectively, causes helices 1 (L44-G65), 2 (P66-S87), and 7 (P165-G186) to become susceptible to protease digestion. These results are consistent with a discrete, globular tertiary structure for the lipid-free protein minimally formed from amphipathic helices 1, 2, 4, 6, and 7. In distinct contrast to lipid-free apo A-I, lipid-bound apo A-I is most susceptible to cleavage in the extreme amino-terminus and, to a lesser extent, in both the central and carboxy-terminal regions. The observed cleavage pattern for the reduced lipid-bound protein supports the existence of many of the turns between helices predicted by sequence analysis of the lipid-bound protein. Methionine oxidation of lipid-bound protein results in a decreased protease susceptibility in the extreme amino-terminus and a concomitant increase in protease susceptibility in the central and carboxy-terminal regions. The results from methionine oxidation indicate the oxidation state of the protein is an important determinant in defining the conformation of both lipid-free and lipid-bound apo A-I.

Restoration of lectin activity to a non-glycosylated ricin B chain mutant by the introduction of a novel N-glycosylation site

Ricin B chain (RTB) is an N-glycosylated, galactose-specific lectin. Removal of the two native N-glycosylation sites at Asn95 and Asn135 by site-directed mutagenesis generated a recombinant protein devoid of lectin activity. Two novel N-glycosylation sites were introduced into RTB at Asn42 and Asn123, either singly or in combination. Microinjection of pre-RTB transcripts into Xenopus oocytes showed that these novel sites became glycosylated in vivo. The single oligosaccharide site chain at Asn42 restored lectin activity to RTB, whereas glycosylation at Asn123 or simultaneous glycosylation at Asn42 and Asn123 failed to do so.

Comparison of ribosome-inactivating proteins in the induction of apoptosis

The aim of this study was to evaluate the ability of verocytotoxin-1 (VT1), VT1 B chain alone, ricin and a hybrid toxin (RASTA2) consisting of ricin A chain linked to VT1 B chain to inhibit protein synthesis and to induce apoptosis. The lethal effects of the toxins were compared using vero cells (originating from green African monkey kidney tissue). As previously described cell death occurred through apoptosis which was quantified using the diphenylamine assay. DNA fragmentation was seen with VT1 at 10 pg/ml but there was no effect with B chain alone. Fragmentation with ricin was seen at 10 ng/ml and with RASTA2 at 1 ng/ml. Protein synthesis inhibition was measured by [(35)S]methionine incorporation. VT1 had an IC50 of 0.0024 ng/ml, B chain alone was ineffective at inhibiting protein synthesis. Ricin had an IC50 of 0.39 ng/ml and RASTA2 of 1.7 ng/ml. In vero cells the B chain of these toxins does not participate in cell killing.

Construction of a Bayesian network for mammographic diagnosis of breast cancer

Bayesian networks use the techniques of probability theory to reason under uncertainty, and have become an important formalism for medical decision support systems. We describe the development and validation of a Bayesian network (MammoNet) to assist in mammographic diagnosis of breast cancer. MammoNet integrates five patient-history features, two physical findings, and 15 mammographic features extracted by experienced radiologists to determine the probability of malignancy. We outline the methods and issues in the system's design, implementation, and evaluation. Bayesian networks provide a potentially useful tool for mammographic decision support.

Free ricin A chain reaches an early compartment of the secretory pathway before it enters the cytosol

During the intoxication of mammalian cells by ricin, the catalytically active A chain must cross the membrane of an intracellular compartment in order to reach its ribosomal substrates in the cytosol. The actual site of ricin A chain translocation is unclear, and conflicting views hold that it enters the cytosol from endosomes or from an early compartment of the secretory pathway, possibly the lumen of the endoplasmic reticulum. Here we show that treating cells with brefeldin A, or transiently overexpressing mutant GTPases known to inhibit biochemical complexes mediating anterograde and retrograde transport between the endoplasmic reticulum and the Golgi complex, protected cells from intoxication by free ricin A chain. These data indicate that ricin A chain, either free or as part of intact ricin, reaches an early compartment of the secretory pathway before translocation into the cytosol occurs.

Ricin A chain fused to a chloroplast-targeting signal is unfolded on the chloroplast surface prior to import across the envelope membranes

The initial stages of chloroplast protein import involve the binding of precursor proteins to surface-bound receptors prior to translocation across the envelope membranes in a partially folded conformation. We have analyzed the unfolding process by examining the conformation of a construct, comprising the presequence of a chloroplast protein linked to ricin A chain, before and after binding to the chloroplast surface. We show that the presequence is highly susceptible to proteolysis in solution, probably reflecting a lack of tertiary structure, whereas the A chain passenger protein is resistant to extremely high concentrations of protease, unless deliberately unfolded using denaturant. The A chain moiety is furthermore active, indicating that the presence of the presequence does not prevent formation of a tightly folded, native state. In contrast, receptor-bound p33KRA (fusion protein comprising the 33-kDa presequence plus 22 residues of mature protein, linked to the A chain of ricin) is quantitatively digested by protease concentrations that have little effect on the A chain in solution. We conclude that protein unfolding can take place on the chloroplast surface in the absence of translocation and without the aid of soluble factors.

Major structural differences between pokeweed antiviral protein and ricin A-chain do not account for their differing ribosome specificity

Pokeweed antiviral protein (PAP) and the A-chain of ricin (RTA) are two members of a family of ribosome-inactivating proteins (RIPS) that are characterised by their ability to catalytically depurinate eukaryotic ribosomes, a modification that makes the ribosomes incapable of protein synthesis. In contrast to RTA, PAP can also inactivate prokaryotic ribosomes. In order to investigate the reason for this differing ribosome specificity, a series of PAP/RTA hybrid proteins was prepared to test for their ability to depurinate prokaryotic and eukaryotic ribosomes. Information from the X-ray structures of RTA and PAP was used to design gross polypeptide switches and specific peptide insertions. Initial gross polypeptide swaps created hybrids that had altered ribosome inactivation properties. Preliminary results suggest that the carboxy-terminus of the RIPs (PAP 219-262) does not contribute to ribosome recognition, whereas polypeptide swaps in the amino-terminal half of the proteins did affect ribosome inactivation. Structural examination identified three loop regions that were different in both structure and composition within the amino-terminal region. Directed substitution of RTA sequences into PAP at these sites, however, had little effect on the ribosome inactivation characteristics of the mutant PAPs, suggesting that the loops were not crucial for prokaryotic ribosome recognition. On the basis of these results we have identified regions of RIP primary sequence that may be important in ribosome recognition. The implications of this work are discussed.

A hydrophobic region of ricin A chain which may have a role in membrane translocation can function as an efficient noncleaved signal peptide

Ricin A chain is a polypeptide of 267 amino acids containing a hydrophobic region near its carboxyl-terminus (residues 245-256) which has been implicated in the membrane translocation step necessary for this catalytically active toxin to reach its intracellular substrate. DNA fusions were constructed that encoded hybrid proteins consisting of carboxyl-terminal residues 233-267 or residues 238-267 of ricin A chain preceding mouse dihydrofolate reductase. When in vitro transcripts prepared from these constructs were translated in cell-free systems, the ricin A chain-derived sequences functioned as efficient signal peptides which directed dihydrofolate reductase into microsomes or into proteoliposomes containing microsomal membrane components.

Catalytic and cytotoxic activities of recombinant ricin A chain mutants with charged residues added at the carboxyl terminus

Ricin A chain (RTA) mutants which had been modified by the addition of three lysine residues, three lysines and an alanine, or six histidine residues at the carboxyl terminus were expressed in Escherichia coli. The recombinant proteins were purified to homogeneity by ion-exchange chromatography on CM-Sepharose CL-6B. The 28S ribosomal RNA N-glycosidase activities of the three RTA mutants were indistinguishable from each other and from the activity of wild-type recombinant RTA. The RTA mutants were not impaired, compared with wild-type RTA, in their ability to reassociate with ricin B chain to form ricin holotoxin. Holotoxins containing mutant RTAs were as readily dissociated into subunits under reducing conditions as native holotoxin, and the RTA mutants were indistinguishable from wild-type RTA in the extent of their interaction with biological membranes. Ricin holotoxins containing the RTA mutants were, however, less cytotoxic to Vero cells than ricin containing wild-type RTA. At equivalent concentrations, a time course assay showed that holotoxin containing the mutant RTAs took longer to kill target cells than that containing wild-type recombinant RTA, suggesting that the mutant forms of RTA are less efficiently processed or translocated across an intracellular membrane than is wild-type RTA.

Point mutations in the hydrophobic C-terminal region of ricin A chain indicate that Pro250 plays a key role in membrane translocation

A series of mutations have been made in the carboxyl terminus of ricin A chain, centred on the hydrophobic region between amino acid residues Val245 and Val256. The mutant ricin A chains were expressed to a high level in an Escherichia coli system and the proteins purified to homogeneity. The enzymic activity of each of these A chain molecules was tested on rabbit reticulocyte ribosomes; in all cases, the activities were found to be comparable to wild-type recombinant ricin A chain. Following reassociation of these A chains to ricin B chain, Vero cells were challenged with these holotoxins and the cytotoxicities determined. Mutant ricin A chain with Ile247-->Ala was unable to reassociate and form holotoxin, indicating the importance of this residue in the interaction with ricin B chain. Mutant ricin A chain with Pro250-->Ala readily reassociated with ricin B chain, forming holotoxin with a 170-fold reduction in cytotoxicity to Vero cells. Other mutations in this region also produced A chain proteins which gave marked reductions in holotoxin cytotoxicity. We propose therefore that the C-terminal hydrophobic region of ricin A chain may be involved in membrane interactions prior to the translocation of this subunit into the cytosol, and that Pro250 plays a key role in one or both of these steps.

Mutational analysis of the Ricinus lectin B-chains. Galactose-binding ability of the 2 gamma subdomain of Ricinus communis agglutinin B-chain

Ricin B-chain (RTB) is a galactose-specific lectin that folds into two globular domains, each of which binds a single galactoside. The two binding sites are structurally similar and both contain a conserved tripeptide kink and an aromatic residue that comprises a sugar-binding platform. Whereas the critical RTB residues implicated in lectin activity are conserved in domain 1 of Ricinus communis agglutinin (RCA) B-chain, the sugar platform aromatic residue Tyr-248 present in domain 2 of RTB is replaced by His in RCA B-chain. In this study, key residues in the vicinity of the binding sites of the Ricinus lectin B-chains were altered by site-directed mutagenesis. The recombinant B-chains were produced in Xenopus oocytes in soluble, stable, and core-glycosylated forms. Both sites of RCA B-chain must be simultaneously modified in order to abolish lectin activity, indicating the presence of two independent, functional binding sites/molecule. Activity associated with the domain 2 site of RCA B-chain is abrogated by the conversion of Trp-258 to Ser. Moreover, the domain 2 site appears responsible for a weak binding interaction recombinant RCA B-chain with GalNAc, not observed with native tetrameric RCA. Finally, the introduction of His at position 248 of RTB severely disrupts but does not abolish GalNAc binding.

Ricin cytotoxicity is sensitive to recycling between the endoplasmic reticulum and the Golgi complex

Cytotoxic proteins that kill mammalian cells by catalytically inhibiting protein synthesis must enter the cytosol in order to reach their substrates. With the exception of diphtheria toxin, which enters the cytosol from acidified endosomes, the intracellular site of translocation of other toxins including ricin, Escherichia coli Shiga-like toxin-1, and Pseudomonas exotoxin A is likely to involve early compartments of the secretory pathway. We have used a molecular approach to identify the site and mechanism of toxin delivery to the cytosol by transiently expressing mutant GTPases that inhibit the assembly of biochemical complexes mediating anterograde and retrograde transport in the exocytic and endocytic pathways. The results provide evidence to suggest that receptors actively recycling between the endoplasmic reticulum and terminal Golgi compartments are essential for toxin translocation to the cytosol from the endoplasmic reticulum. The rapid kinetics of intoxication demonstrate a substantial level of bidirectional membrane flow and sorting through the early secretory pathway.

Preliminary investigation of a Bayesian network for mammographic diagnosis of breast cancer

Bayesian networks use the techniques of probability theory to reason under conditions of uncertainty. We investigated the use of Bayesian networks for radiological decision support. A Bayesian network for the interpretation of mammograms (MammoNet) was developed based on five patient-history features, two physical findings, and 15 mammographic features extracted by experienced radiologists. Conditional-probability data, such as sensitivity and specificity, were derived from peer-reviewed journal articles and from expert opinion. In testing with a set of 77 cases from a mammography atlas and a clinical teaching file, MammoNet performed well in distinguishing between benign and malignant lesions, and yielded a value of 0.881 (+/- 0.045) for the area under the receiver operating characteristic curve. We conclude that Bayesian networks provide a potentially useful tool for mammographic decision support.

Analysis of oxidative titrations of Desulfovibrio gigas hydrogenase; implications for the catalytic mechanism

The oxidative titrations of hydrogenase (Hase) from Desulfovibrio gigas [Barondeau, D. P., Roberts, L. M., & Lindahl, P. A. (1994) J. Am. Chem. Soc. 116, 3442] were simulated using model descriptions of the redox reactions in the enzyme. The data fit best to a model that assumed Hase contains one [Fe3S4]1+/0 cluster, two [Fe4S4]2+/1+ clusters, and a Ni center stable in four redox states (Ni-B, Ni-SI, Ni-C, and Ni-R), each separated by one electron. A model in which Ni-SI, Ni-C, and Ni-R correspond to Nickel(2+) dithiolate, nickel(1+) dithiol, and nickel(2+) dithiol hydride, respectively, is compatible with all established relevant properties of the Ni center. This model and the concept of redox microstates were employed to define electronic states of the enzyme and to reformulate the catalytic mechanism initially proposed by Cammack et al. [Cammack, R., Patil, D. S., Hatchikian, E. C., & Fernandez, V. M. (1987) Biochim. Biophys. Acta 912, 98] into three interconnected catalytic cycles. These cycles differ in the average oxidation level of the Fe4S4 clusters. The cycle with the most reduced clusters appears to operate reversibly (catalyzing both H2 oxidation and H+ reduction), while those with more oxidized clusters function only to oxidize H2. The difference in reversibility is explained by assuming that Ni-R prefers to reduce an [Fe4S4]2+ cluster instead of H+ and that H+ is reduced only when that Fe4S4 cluster is in its reduced state.