Human Cu,Zn superoxide dismutase (SOD) is a single domain all β‐sheet protein with its eight β‐strands arranged as a Greek key β‐barrel or immunoglobulin fold. Three circularly permuted variants of SOD were made by joining the native amino‐ and carboxy‐termini, and introducing new termini at sites originally within connections between β‐strands. The locations of the new termini were chosen to interrupt β‐turns between the two N‐terminal β‐hairpins and the short cross‐barrel Greek key connection. Expression levels in the Escherichia coli periplasm were indistinguishable from that of native SOD. Reaction rates for the purified proteins were similar to those of the native enzyme, indicating that the permutants are correctly folded. Interrupting the covalent cross‐bracing provided by the Greek key connection reduced the stability of the protein by ∼1.0 kcal/mol, indicating only a slight contribution to conformational stability. The experiments test and eliminate two hypotheses for folding pathways for Greek key β‐barrels that require N‐terminal β‐hairpins or covalent attachment across the short Greek key connection.
Interest in the function of the Greek key motif and Greek key connections arose from the discovery that they occur in many β‐strand proteins (Richardson, 1977). Early studies on protein taxonomy classified superoxide dismutase (SOD) as a Greek key β‐barrel, a protein with two interlocking Greek key motifs and antiparallel β‐strands arranged like the staves of a barrel (Richardson, 1977, 1981). More recent analysis has extended the classification and shown that this double Greek key occurs much more frequently than the single Greek key motif (Hutchinson and Thornton, 1993).
Functional versatility and stability are hallmarks of proteins with the Greek key β‐barrel structural motif. It is used in recognition and binding [immunoglobulin superfamily, prealbumin (transthyretin)], structural assemblies (icosahedral virus capsid proteins, lens crystallins) and catalysis (serine proteases, SOD) (Getzoff et al., 1989). Stability is exemplified by the long‐lived lens crystallins (Blundell et al., 1981) and by SOD, one of the most stable characterized human globular proteins (Malinowski and Fridovich, 1979; Lepock et al., 1990a). The stability of SOD is also of current interest because of the more than 40 point mutations in human SOD that are known to cause the fatal neurodegenerative disease, familial amyotrophic lateral sclerosis (FALS) (Rosen et al., 1993). Decreased conformational stability of the SOD Greek key β‐barrel appears to be a unifying property of SOD FALS mutants (Deng et al., 1993).
The flattened Greek key β‐barrel of wild‐type human SOD (Figure 1) is typical of the family: its eight antiparallel β‐strands can be viewed as a twisted β‐sandwich, with the two halves of the sandwich each containing four β‐strands (strands 1, 2, 3 and 6 and 5, 4, 7 and 8). The two Greek key loops are connections between β‐strands that cross the barrel between the two halves of the sandwich (between strands 3, 4 at the top and between strands 6, 7 at the bottom), whereas all the other connections are between β‐strands adjacent in the structure. Thus, the Greek key connections form a covalent cross‐bracing across the ends of the barrel and are thought to be important in Greek key β‐barrel stability (Tainer et al., 1982). The Greek key structural motif was named after a design commonly found on Greek vases, which resembles schematic diagrams showing the connections between the β‐strands (Richardson, 1977; also see Figure 2A and B). The two SOD cross‐barrel Greek key connections differ structurally (see Figure 1), and here we have focused on the shorter, simpler connection between strands 3 and 4.
Three folding pathways for Greek key β‐barrels have been proposed, all of which initiate with a folding nucleus involving β‐strands that are adjacent in the primary structure. All the proposed pathways assume that rapid local couplings between such β‐strand pairs precede formation of a folding nucleus: it is therefore essential to the proposed models that there is a short covalent connection between the β‐strands involved and that they are not separated by other β‐strands. One proposal involves the initial formation of a long, antiparallel β‐hairpin, which always involves the four β‐strands of the Greek key (Ptitsyn et al., 1979; Richardson, 1981; see Figure 2C). Another proposal is that N‐terminal β‐hairpins nucleate the structure: they are an almost universal feature of Greek key β‐barrels and might be involved in an in vivo vectorial folding process that is coupled with translation (Getzoff et al., 1989). A third proposal involves hydrophobic interactions between two β‐strands (β‐zipper model; Hazes and Hol, 1992) that are closely connected via the short Greek key connection. These β‐strands do not form β‐sheet hydrogen bonds with each other in the folded structure (strands 3, 4 of native SOD in Figure 1).
To quantitate the contribution to structural stability provided by the covalent cross‐bracing in Greek key loops and to test hypotheses about folding pathways, we made circularly permuted (Goldenberg and Creighton, 1983) derivatives of native SOD by genetic engineering (Luger et al., 1989). Circular permutation of proteins involves joining native N‐ and C‐termini and breaking the polypeptide chain elsewhere to create new termini, which will generally be close to each other at a new site in the structure.
Design of the permuted SODs.
The predicted structures and amino acid changes of the permutants compared with native SOD are shown in Figure 1. The three circularly permuted proteins have one, two, or three β‐strands transferred from the amino to the carboxy‐terminus of SOD, and are therefore named SOD N−1, C+1, SOD N−2, C+2 and SOD N−3, C+3, respectively. SOD N−2, C+2 transfers the N‐terminal β‐hairpin to the C‐terminus to produce a protein lacking the N‐terminal β‐hairpin that is required for one of the proposed folding schemes (Getzoff et al., 1989). The SOD N−3, C+3 permutant interrupts the structurally simple, short Greek key connection and was therefore chosen to test the importance of covalent cross‐bracing for β‐barrel stability. An intact short Greek key connection is required in one proposed folding pathway (Hazes and Hol, 1992) and is important in another (Ptitsyn et al., 1979; also see Discussion). The N‐ and C‐termini of the native structure are sufficiently flexible and close together (Parge et al., 1992) that joining them does not require the insertion of additional amino acids; hence a GlyProGly linker (Boissinot et al., 1993) was substituted for residues Gln153Ala1Thr2. Proline, a β‐strand terminator (Chou and Fasman, 1978), was included to help ensure that β‐strands 1 and 8 terminated at the new connection. Two structurally superfluous residues (Asn26, Gly27) were deleted from the turn linking β‐strands 2 and 3 in SOD N−2, C+2. To ensure correct processing by the signal peptidase, all the permutants were given an alanine at their new N‐termini because this residue occurs in the correctly processed native enzyme (Getzoff et al., 1992). In SOD N−3, C+3, substitution of Thr39 with Ala to produce the new N‐terminus might have resulted in distortion of this part of the structure due to the more bulky terminal amino and carboxyl groups which would replace the peptide backbone linkage. Therefore, the Ala residue was placed in front of Thr39 so that the peptide backbone could turn out towards solvent and minimize distortion of the structurally important Leu38 side chain which forms a hydrophobic ‘plug’ at this end of the barrel (Getzoff et al., 1989; Deng et al., 1993). The enzyme used as native control and as the background for all the permutants is the double mutant Cys6Ala, Cys111Ser, which has the same structure, activity and conformational stability as the wild‐type enzyme, but greatly improved reversibility of unfolding in vitro, which is useful in renaturation experiments and purification (Lepock et al., 1990a; Hallewell et al., 1991; Parge et al., 1992).
Protein expression and purification.
Native SOD and the permutants were expressed periplasmically in Escherichia coli by overnight induction of the tacI promoter (Getzoff et al., 1992). Expression levels for the permutants were indistinguishable from that of native SOD, representing ∼10% of total protein, suggesting that large alterations in folding, structure or stability had not occurred. A variety of destabilizing mutants made previously did produce large decreases in periplasmic expression levels (Banci et al., 1991; also unpublished results). The permutant SODs were fully soluble, were recovered in high yield after osmotic shock, and were purified to ∼95% homogeneity as described in Materials and methods. Amino‐terminal protein sequence analysis (Hallewell et al., 1985) gave the expected amino acid sequence for the first 20 residues in all the proteins. The sequence analysis also showed that the pre‐SOD was correctly processed with no amino acids from alternate processing or proteolysis detectable.
Rate constants for permuted SODs.
To determine the effects of circular permutation on enzyme activity, reaction rates were measured by generating a pulse of superoxide by radiolysis and monitoring superoxide decay spectrophotometrically at 245–270 nm (Getzoff et al., 1992). SOD N−1, C+1 and SOD N−2, C+2 have rate constant versus pH profiles very similar to that of the native enzyme, indicating that these molecules are fully functional and correctly folded (Figure 3). SOD N−3, C+3 has ∼75% of the activity of native SOD throughout the pH range tested, suggesting that a change in structure or stability has slightly affected the active site region. However, an incorrectly folded SOD N−3, C+3 would be expected to cause a very much greater loss of enzyme activity because correct placement of Cu‐liganding residues from the β‐barrel (Carri et al., 1994; Banci et al., 1995) and of charged residues in the active site channel (Getzoff et al., 1992; Fisher et al., 1994) are essential for maintaining the fast reaction rate of SOD. Therefore, the relatively small change in reaction rate of SOD N−3, C+3 indicates that this permutant is correctly folded.
Conformational stability of the permuted SODs.
Differential scanning calorimetry was used to compare the thermal stabilities of the permutants with native enzyme (Lepock et al., 1990a). The calorimetry profiles (Figure 4) are similar in general shape, with relatively small differences in the average unfolding temperature (Tm). We have previously demonstrated that the profiles for native SOD, the Cys6Ala, Cys111Ser SOD and several mutant SODs are best fit with two components, one representing the reversible denaturation of the oxidized Cu2+ form and the other the reduced Cu+ form (Lepock et al., 1990a). The two‐component, best fit, thermal denaturation curves for the permutant SODs, assuming reversible two‐state unfolding transitions, are shown in Figure 5. The one‐component fits were very poor (not shown), and other two‐component, non‐two‐state fits required the introduction of two additional fitting parameters and were not a significant improvement over the two‐state fit presented here. Thus, the two‐component, two‐state fit was selected as most appropriate, supporting reversible folding in vitro.
The fitting parameters (ΔH, ΔS, Tm) for each component are given in Table I. The ΔTms of the two components were 2.3 and 1.9°C for SOD N−1, C+1, −0.3 and −0.8°C for SOD N−2, C+2, and −3.7 and −1.6°C for SOD N−3, C+3. These correspond to ΔΔG values of 0.7 to 0.9 kcal/mol dimer for SOD N−1, C+1, −0.1 to −0.4 kcal/mol for SOD N−2, C+2 and −0.6 to −1.0 kcal/mol for SOD N−3, C+3. Thus, interruption of the Greek key connection (SOD N−3, C+3) shows the greatest destabilization. Interruption of the first β‐hairpin (SOD N−1, C+1) actually increases stability slightly, and interruption of the second β‐hairpin does not significantly alter stability (SOD N−2, C+2).
What is the function of the Greek key motif in β‐barrels and why is it such a commonly occurring arrangement of β‐strands? The striking structural feature of covalent cross‐bracing across each end of the β‐barrel led to the proposal that Greek key connections are involved in conformational stability (Tainer et al., 1982), whereas the unusual topological arrangement of the β‐strands led to some long‐standing proposals for their involvement in protein folding (Ptitsyn et al., 1979; Richardson, 1981). Until recently, it has not been possible to investigate the function of the Greek key connection experimentally. By using the technique of protein circular permutation (Goldenberg and Creighton, 1983; Luger et al., 1989) to break the Greek key connection, we have demonstrated that this covalent linkage does not contribute significantly to the conformational stability of the Greek key β‐barrel structural motif. In addition, we have shown that the short Greek key connection is not a requirement for correct folding in vitro or in vivo. By elimination, these findings suggest that one function of Greek key connections is to allow correct placement of structurally important hydrophobic residues at the ends of the β‐barrel. However, as discussed below, our results do not eliminate the possibility that the Greek key has advantages for folding pathways.
One inherent problem with circular permutation is the potentially destabilizing effect of the changes at the sites where the original termini were joined, and where the new termini have been created. The finding that all three permutants described here had approximately the same Tm of unfolding as native SOD suggests that neither alteration caused a significant increase or decrease in stability. The formation of new termini in SOD N−3, C+3 seemed most likely to decrease stability because the hydrophobic side chain of the C‐terminal Leu38 residue forms a structurally important hydrophobic ‘plug’ at one end of the barrel (Getzoff et al., 1989; Deng et al., 1993). Simply leaving Leu38 and Thr39 as the new C‐ and N‐termini would replace the peptide bond with more bulky carboxyl and amino groups, which might displace the Leu38 side chain and destabilize the barrel. We therefore tried to minimize this effect by inserting an additional N‐terminal alanine residue which could turn the new N‐terminal peptide bond outwards from the structure towards solvent (see Figure 1). This approach appears to have been successful, since the decrease in the stability of SOD N−3, C+3 was <0.5 kcal/mol per SOD subunit, which is much less than the 4 kcal/mol cost for displacing three buried ‐CH2‐ groups (Kellis et al., 1988).
When considering the folding mechanism of Greek key β‐barrels, there are some clear differences with α‐helical proteins. Whereas α‐helices are relatively stable structures with contiguous primary and secondary structure, isolated β‐hairpins are not particularly stable and structurally adjacent β‐strands in proteins may not be contiguous in the sequence (see Figure 2). Moreover, it is not immediately obvious how the non‐adjacent β‐strands of the Greek key are paired during folding, when there are many non‐productive β‐pairings that can occur with apparently equal probability. In this view, a mechanism must be found for ordering β‐strand interactions and preventing non‐productive ones which might lead to inclusion bodies (Mitraki et al., 1991) or amyloid structures (Hamilton et al., 1993). Amyloid structures may be of particular relevance here, because they seem to require β‐strand interactions, occur in the Greek key β‐barrel proteins prealbumin (transthyretin) (Colon and Kelly, 1992; Jarvis et al., 1994) and immunoglobulin (Hurle et al., 1994), and are responsible for several diseases (Kisilevsky et al., 1992). As discussed below, amyloidic SOD structures might be favoured by the SOD mutations that cause the neurodegenerative disease, amyotrophic lateral sclerosis.
All the proposed Greek key β‐barrel folding schemes incorporate the principle that the earliest interactions will tend to occur between folding elements that are close together in the primary structure. Thus, one feature of the folding schemes is that initiator events involve adjacent β‐strands, preferably connected by a short loop(s). One of the earliest folding schemes, proposed by Ptitsyn et al. (Ptitsyn et al., 1979; Ptitsyn and Finkelstein, 1980), solved both the problem of ordering β‐strand interactions and β‐hairpin stability by suggesting that the initiator structure is a long β‐hairpin involving the Greek key β‐strands, as shown in Figure 2c. One such hairpin structure (strands 3, 4, 5 and 6) is lost in SOD N−3, C+3, by breaking the Greek key connection, while the other (strands 4, 5, 6 and 7) remains intact (Figure 2D). It will be necessary to interrupt both Greek key connections to test definitively this folding scheme. However, our results demonstrate that the first Greek key is not necessary for folding.
A second folding scheme suggested that the N‐terminal β‐hairpin may initiate SOD folding (Getzoff et al., 1989). This proposal was based on evolutionary considerations and the observation that such N‐terminal β‐hairpins occur in all Greek key β‐barrels except prealbumin (transthyretin), which is a more complex case, because the transthyretin dimer contains paired β‐strands from different polypeptide chains and under certain circumstances folds aberrantly to form fibrils of amyloid (Hamilton et al., 1993). This hypothesis is invalidated by the correct folding of SOD N−2, C+2, which lacks an N‐terminal β‐hairpin (see Figure 1).
A third folding pathway (Hazes and Hol, 1992) based on sequence comparisons and structural considerations involves an initiator formed from interacting hydrophobic residues in β‐strands 3 and 4. These β‐strands are joined by the short Greek key connection that has been broken in SOD N−3, C+3, and one of the criteria for selecting these two β‐strands was that the connection between them is always rather short, favouring their rapid interaction. This hypothesis also appears to be invalidated by the correct folding of SOD N−3, C+3.
In addition to the three circular permutants described here, Viguera et al. (1995) have described three permutants which interrupt β‐hairpins in the small all‐β‐strand protein α‐spectrin (SH3 domain). Together, our findings indicate that close linkage of a particular pair of β‐strands as a folding determinant is not required for β‐strand proteins and suggests that the folding pathway is overdetermined. This would increase the resilience of a protein to amino acid changes or the insertion of new subdomains (Getzoff et al., 1989), and such proteins would be selected for during evolution. The structurally related immunoglobulin variable domain Greek key β‐barrel (Richardson et al., 1976) may have been selected to have a particularly resilient folding mechanism.
Do these studies provide any insights into the molecular determinants or mechanism of SOD‐linked familial amyotrophic lateral sclerosis (SOD FALS) (Rosen et al., 1993), a fatal neurodegenerative disease caused by the mutational gain of a toxic function by SOD (Gurney et al., 1994)? There are now over 40 different SOD FALS mutants and many of these occur in the Greek key loops, including a mutation of the structurally important Leu38 residue to valine (Rosen et al., 1993), which is located in the short Greek key connection and forms a hydrophobic ‘plug’ at this end of the barrel. Structural analysis suggests that destabilization of the β‐barrel is a common property of SOD FALS mutants (Deng et al., 1993). This suggests two possible toxic mechanisms for the mutant SOD FALS proteins: increased exposure of the active site Cu ion, which is known to catalyse toxic peroxidation reactions in vitro (Hodgson and Fridovich, 1975); or the formation of amyloid‐like structures. Amyloid fibrils are formed from the related Greek key β‐barrels of prealbumin (transthyretin) (Colon and Kelly, 1992; Jarvis et al., 1994) and immunoglobulin (Hurle et al., 1994) by aberrant β‐strand interactions promoted by destabilizing mutations to produce characteristic disease states. Preliminary experimental evidence supports a Cu‐based mechanism of toxicity (Wiedau‐Pazos et al., 1996). Nonetheless, an amlyoid‐like form of SOD could have a role in the disease process, possibly by causing even greater exposure of the Cu ion.
Materials and methods
Constructs were made using pPHSODllacIq, a pBR322 derivative encoding periplasmic expression of a synthetic human SOD gene (Hallewell et al., 1989), regulated by a tacI promoter and a plasmid‐encoded lacIqgene (Getzoff et al., 1992). DNA manipulations were as described in Sambrook et al. (1989). PCR primers were used to produce modified SOD gene segments for re‐insertion into the native SOD gene using existing unique restriction sites (Hallewell et al., 1989; Getzoff et al., 1992). All amino acid numbers and β‐strand numbers given below refer to the native structure (see Figure 1 legend). In the final plasmid constructs used for protein purification, the entire SOD gene was sequenced to ensure that no unintended changes were present. The genes encoding the three permutants were constructed sequentially from each other, starting with SOD N−1, C+1. To make SOD N−1, C+l, β‐strand 1 was first deleted by PCR amplification of β‐strands 2–8 as a HindIII–SalI cassette with the HindIII site located in the DNA encoding the signal peptide three amino acids from the signal peptidase processing site, and the SalI site located immediately after the stop codon (Hallewell et al., 1989; Getzoff et al., 1992). A new N‐terminus was created by substituting Pro13 with Ala and placing this immediately after the processing site (see Figure 1). This HindIII–SalI fragment was cloned by substitution into pPHSODllacIq to make pPHSOD2–8. To join β‐strand 1 to the C‐terminus, a PCR primer located at the BamHI site in human SOD (at residues Gly141Ser142; Hallewell et al., 1989) and a PCR primer base pairing to the C‐terminal coding sequence and extending the coding sequence by encoding a GlyProGly linker, β‐strand 1, a stop codon and a SalI site were used to make a BamHI–SalI PCR fragment for BamHI–SalI substitution into pPHSOD2–8. To make SOD N−2, C+2, the SOD N−1, C+l gene was used as a template from which the gene sequence encoding β‐strand 2 was deleted by PCR amplification of the other seven strands with creation of a new N‐terminus by substituting Pro28 with Ala (see Figure 1). The PCR fragment was also given HindIII–SalI ends and cloned by substitution into pPHSODllacIq. To join β‐strand 2 to the C‐terminus, a PCR primer encoding this strand with a termination codon after Ser25 was used to make an ApaI–SalI PCR fragment (ApaI site encodes residues Gly12Pro13; see Hallewell et al., 1989). This fragment was cloned by substitution into the pPHSODllacIq derivative. To make SOD N−3, C+3, the SOD N−2, C+2 gene was used as a template from which the gene sequence encoding β‐strand 3 was first deleted by amplifying the other seven strands as a HindIII–SalI fragment. In this PCR reaction, a new N‐terminal Ala residue was placed in front of Thr39. This HindIII–SalI fragment was cloned by substitution into pPHSODllacIq. To join β‐strand 3 to the C‐terminus, a PCR primer encoding this strand, with a stop codon after Leu38, was used to make an ApaI–SalI PCR fragment as described for SOD N−2, C+2. This fragment was cloned by substitution into the pPHSODllacIq derivative.
Protein expression and purification
Recombinant SOD proteins were produced in E.coli MC1061 (Huynh et al., 1985) at ∼10% of total cellular protein, after overnight induction of the tacI promoter with 0.2 mM isopropyl β‐d‐thiogalactopyranoside in 10 l of L‐broth containing 100 μg/ml ampicillin and 0.1 mM CuSO4. The mutant SODs were released from the periplasm by osmotic shock and purified to ∼95% homogeneity by DEAE–Sepharose chromatography as previously described (Getzoff et al., 1992), except that the heating step was omitted. To ensure full charging with metals, the enzyme was re‐charged with Cu and Zn in vitro (McCord and Fridovich, 1969; Nishida et al., 1994). Both metals were first removed by sequential dialysis at 4°C against 50 mM sodium acetate/10 mM EDTA at pH 3.8, 50 mM sodium acetate/100 mM NaCl at pH 3.8, 100 mM sodium acetate, pH 3.8, 100 mM sodium acetate, pH 5.5. The proteins were then charged with metals by dialysis against 100 mM sodium acetate buffer, pH 5.5, in the presence of a 40‐fold molar excess of Zn, followed by addition of a 4‐fold molar excess of Cu. At pH 5.5, the affinity of Cu for its own site is much greater than the affinity of Zn for the Cu site and when Zn occupies the Zn site, Cu can occupy its site more rapidly. Unbound metals were subsequently removed by dialysis in 20 mM NaCl, followed by dialysis against water. Protein concentrations were determined spectrophotometrically at 265 nm (Beyer et al., 1987).
Rate data were obtained from pulse radiolysis experiments (average of six measurements per data point: error <5%) using the 2 MeV van der Graaff accelerator at Brookhaven National Laboratory. Superoxide radicals were generated in air‐saturated aqueous solutions (Schwarz, 1981) containing 10 mM sodium formate, 5 mM monobasic sodium phosphate. Decay of superoxide was monitored spectrophotometrically at 245–270 nm in the presence of 36–72 μM EDTA, 1–10 μM superoxide and 1–10 μM SOD. Catalytic rates were calculated using active enzyme concentrations determined by measuring the concentration of the activity essential Cu ion by atomic absorption spectroscopy (in triplicate on a Pye‐Unicam or GBC instrument, with an error of <5% for Cu and Zn).
Differential scanning calorimetry
The DSC scans were obtained with a Microcal‐2 differential scanning calorimeter (DSC) with 1.21 ml sample pans at a protein concentration of 2–4 mg/ml (0.06–0.12 mM dimer) in 100 mM potassium phosphate, pH 7.8. The reference solution was 100 mM potassium phosphate without SOD. Both sample and reference were de‐aerated at room temperature before scanning from 25–105°C at a rate of 1°C/min. Three to five re‐scans from 25–105°C were made to check for reversibility. The baseline and change in specific heat (ΔCp) upon denaturation were corrected as previously described (Lepock et al., 1990b). The profiles were deconvoluted and the best fit curves obtained assuming two state, reversible denaturation using the program Origin (Microcal, Inc). Since the reversibility was only ∼25% after scanning to 105°C, which is of borderline suitability for a reversible fit, the fitting range was set from the beginning of the first transition to the temperature corresponding to 25% of the maximum peak height on the high temperature side of the last transition. This minimizes any distortion induced by irreversible denaturation (Lepock et al., 1992). The fitting parameters (ΔH, ΔS, Tm) and the free energy differences (ΔΔG) between control SOD and the circularly permuted SODs were calculated as previously described (Lepock et al., 1990b).
Note added in proof
It has recently been shown that astrocytes with large numbers of SOD‐containing inclusion bodies are associated with pathogenesis in an SOD FALS transgenic mouse model of SOD‐linked familial amyotrophic lateral sclerosis [Bruijn,L.I. et al. (1977) ALS‐linked SOD1 mutant G85R mediates damage to astrocytes and promotes rapidly progressing disease with SOD1‐containing inclusions. ].
We thank M.E.Pique (TSRI) for assistance with computer graphics and F.Masiarz (Chiron) for amino acid sequencing. These studies were supported by the Wellcome Trust Grant 038968 (R.A.H.) and NIH grants R01 GM 39345 (J.A.T.) and R01 GM 37684 (E.D.G.). Pulse radiolysis studies were supported by NIH grant RO1 GM 23658 (D.E.C.) and carried out at Brookhaven National Laboratory, operated under contract DE‐AC02‐76CH000016 to the US Department of Energy. Differential calorimetry studies were supported by NSERC Canada. M.B. was supported by fellowships from FRSQ and MRC Canada.
- Copyright © 1997 European Molecular Biology Organization