To accommodate perpendicular heme groups in a simple protein structure, an all-parallel four-α-helix motif was selected. The design, shown in Figure 1, incorporates four helical segments of identical sequence with each of the two pairs of helices linked by a disulfide bond. The resulting noncovalent dimer of helix–link–helix peptides contains a pseudofourfold symmetry axis parallel to the helix dipoles. The design accommodates four heme groups oriented perpendicular to one another.
The amphiphilic α-helical units are 27 amino acids long with a C-terminal Gly-Cys sequence that, when oxidized, forms the flexible disulfide link between two parallel helices. Each helix contains two histidine residues at positions 6 and 7, yielding four pairs of ligands that are designed to bind the four low-spin heme groups via bis-histidine axial ligation. The design requires that a histidine at position 6 on one helix and a histidine at position 7 from an adjacent helix serve as the pair of ligands that binds each heme.
The backbone structure of the protein, called 6H7H, was modeled by starting with four identical 27 amino acid helices with histidines at positions 6 and 7. The backbone dihedral angles of the helices were adjusted to yield 3.6 residues/turn (see below). Four heme groups were then attached to the four pairs of histidine ligands, fixing each NεHis6-FeHeme-NεHis7 angle to 180° and each FeHeme-NεHis bond distance to 2.0 Å. While maintaining these parameters and the necessary fourfold symmetry axis, the histidine side chain dihedral angles (κ1 and κ2) were adjusted until the C-terminal end of the helices reached an appropriate distance to allow formation of the disulfide bond between C-terminal cysteine residues. The result is the V-shaped structure evident in Figure 1.
Positioning of hydrophobic and hydrophilic residues that constitute the helical domain was determined using a method that expands on the repeating heptapeptide method often used to design two-helix coiled coils and four-helix peptides. The heptad method assumes 3.5 residues/turn and, therefore, seven unique positions. Typically, hydrophobic residues are assigned to positions a and d, while positions b, c, e, f, and g are assigned as hydrophilic residues (DeGrado et al. 1989; Myszka and Chaiken 1994). It is common to assign opposite charges to residues at positions e and g to maximize favorable interchain electrostatic interactions (Zhou et al. 1994). Although the heptad method has been somewhat successful in yielding α-helical proteins of predefined structure, it is more realistic to characterize α-helices in a four-helix protein by 3.6 residues/turn. A simple yet relevant survey of the α-helices in three highly helical natural proteins (ROP, Mb, cyt b562) reveals that the average value for the sum of the backbone dihedral angles (Φ and Ψ) is −104°. This value translates to 3.57 residues/turn yielding 25 unique positions. For simplicity, this was rounded to 3.6 residues/turn, yielding 18 (a–r) unique positions that define an α-helix, instead of the seven in the heptad method. As indicated in Figure 2, the a–r method assigns positions a, d, e, h, k, l, and o to hydrophobic interior residues defining nearly 40% of the helix face. Positions i, b, and p and g, n, and r define opposite sides of the interhelical interface and are assigned as hydrophilic residues of opposite charge to assure optimal interhelical electrostatic interactions. The remaining positions (c, f, j, m, and q) are opposite the hydrophobic face and are assigned to hydrophilic residues with the potential to only form intrahelical contacts. The a–r method will, of course, assign hydrophobic residues to positions different from what would have been assigned by the heptad method; however, it yields sequences with hydrophobic positions that match more closely with that of natural four-helix proteins (Z. Xu, R.S. Farid, unpubl.).
The sequence of the 27-amino-acid peptide is presented below:
As described previously, positions a, d, e, h, k, l, and o are hydrophobic residues (with the exception of the histidine ligands at positions d and e). The identity of the hydrophobic residues was determined using CORE, our newly developed protein-design program (Jiang et al. 2000). Three separate runs were initiated starting with Ala at each core position. This yielded three distinct families of sequences; however, the top-ranked sequence from each run was identical. Therefore, this sequence was chosen as the one most likely to produce a uniquely folded protein with high thermal stability.
Slices through the helices (Fig. 3) of the resulting structure show the computer-designed hydrophobic core residues. Also highlighted are potential inter- and intrahelical electrostatic interactions among interface and exterior hydrophilic residues. Four tryptophan residues at positions 3 and eight phenylalanine residues at positions 10 and 13 effectively fill the interior space of the V-shaped backbone structure while defining the top and bottom of the heme binding sites, respectively. Lysine residues at positions 2 and 4 are designed to form favorable electrostatic interactions with the heme propionate side chains.
The relative geometry of adjacent perpendicular heme groups in the model of 6H7H is similar to that found for the two perpendicular heme a groups in cytochrome c oxidase (CcO). The angle between planes that define heme a and a3 in native CcO is 108°; this value is similar to the 90° angle between adjacent hemes in 6H7H. The shortest edge-to-edge distance between heme groups in CcO is 7 Å as compared to 6 Å in the model of 6H7H.
Detailed titration studies were undertaken to determine whether 6H7H binds four heme groups via bis-histidine axial ligation as designed. Hemin from DMSO or 100 mM NaOH solution was incrementally added to a 23.5 μM protein solution. An absorption spectrum was recorded following each addition. Binding is considered complete when difference spectra match that of free hemin. At the protein concentration used, this required addition of ∼215 μM of hemin or ∼9 equivalents.
To determine the concentration of bound and free heme at all concentrations of added heme, each of the 36 spectra collected during the heme titration were fit to a linear combination of component spectra using a program written in Mathematica (Wolfram Research; see Supplemental Material). The same three component spectra were sufficient to fit each of the 36 spectra. The first component spectrum is simply that of free heme obtained from difference spectra at the end of the titration, where the concentration of added heme far exceeds the concentration of protein. The second component spectrum corresponds to difference spectra obtained early in the titration, below 1.8 equivalents of added heme. These difference spectra are nearly identical to one another and, therefore, represent a basis set spectrum. A third component spectrum was needed to fit the spectra at heme concentrations above 2 equivalents. The three component spectra corresponding to free heme, bound heme below 2 equivalents, and bound heme above 2 equivalents are shown in Figure 4. The two component spectra that correspond to bound heme are clearly similar to one another and are consistent with low-spin bis-histidine bound-heme groups. Indeed, the component spectrum corresponding to the first hemes that bind to 6H7H is nearly superimposable with that of native cytochrome b561, a natural protein that binds a single heme via bis-histidine ligation. Figure 5 compares the synthetic and natural hemoprotein spectra revealing remarkable similarity, including spectral widths and extinction coefficients for both the oxidized and reduced states. The component spectrum that corresponds to hemes that bind above 2 equivalents, although distinct from that of the first hemes, is also clearly consistent with bis-histidine ligation. The Soret band maximum is centered at 410 nm, compared to 416 nm for the first hemes. This 6 nm difference is consistent with differences in λmax among native cytochrome b proteins.
Having established that two distinct types of hemes bind to 6H7H, it remains to determine heme binding stoichiometry and dissociation constants. To do this, the precise concentration of bound heme and free heme at each point in the titration must be known. This was made possible by noting that at low initial hemin concentrations, only one of the component spectra was needed to fit the difference spectra and that at very high initial hemin concentrations, the difference spectra correspond to free hemin. Therefore, it was possible to determine the extinction coefficients of the three component spectra presented in Figure 4. Accordingly, at all heme concentrations, the fit of the difference spectra yields the concentration of heme bound to the protein or free in solution. A plot of these concentrations as a function of the initial hemin concentration is shown in Figure 6. The plot shows that below 2 equivalents of added heme, two hemes bind tightly, as evidenced by the minimal concentration of free heme observed. The first component spectrum in Figure 4 corresponds to these two heme groups. Above 2 equivalents, the concentration of the two tightly bound hemes remains constant, while the concentration of the other two spectrally distinct bound hemes increases. These two hemes bind more loosely than the first two hemes; an additional 7 equivalents of heme is needed to fully fill the third and forth binding sites. These results clearly demonstrate that, as designed, 6H7H binds four bis-histidine ligated heme groups.
Binding of the first two hemes is clearly associated with submicromolar dissociation constants, as evidenced by the nearly linear increase in bound heme concentration with addition of heme (Fig. 6). To gain greater insight into the nature of the binding of these hemes, careful analysis of the difference spectra below 2 equivalents of added heme was conducted. This revealed a slight red shift of the heme Soret band on heme addition (Fig. 7). This result implies that the λmax of the monoheme and diheme proteins differ slightly. The data were fit to two KD values, revealing positive cooperativity in heme binding with a KD2/KD1 ratio of 0.22, a λmax of the monoheme protein of 414.3 nm, and a λmax of the diheme protein of 415.5 nm. It is not possible from these data to determine individual values for KD1 and KD2; only the ratio of KD values can be determined. To obtain accurate values for the individual dissociation constants, heme titration at lower protein concentration (0.95 μM) was conducted (Fig. 8). At this protein concentration, complications from binding of the third and forth heme are essentially eliminated. Fixing KD2/KD1 to a value of 0.22 and fitting the data in Figure 8 to a model that incorporates two KD values yields KD1 = 80 ± 10 nM and KD2 = 18 ± 2 nM.
Analysis of the shift in λmax above 2 equivalents was not possible because of the presence of relatively high concentration of free heme, precluding accurate determination of any λmax shift that may exist. However, as the first two heme binding sites are nearly 100% filled at 2 equivalents of added heme, it is possible to determine the dissociation constants for the third and forth hemes (KD3, KD4) by fitting the data in Figure 6 above 2 equivalents of added heme. KD1 and KD2 were simply assumed to be zero, which is reasonable given that their values are approximately three orders of magnitude lower than the initial protein concentration. Fitting the data to a model that incorporates two KD values cannot yield individual values for KD3 and KD4; however, the fit does yield KD3 × KD4 = 1700 ± 50 and KD3 ≥ 3 mM. The quality of the fit is maintained for values of KD3 ≥ 3 mM as long as the product of KD3 and KD4 is 1700. For example, if KD3 is assumed to be 3 mM (the lower limit), KD4 must be 570 nM (the upper limit for KD4), and if KD3 is set to 10 mM, KD4 is 170 nM, and so forth. Although only a lower limit for KD3 and an upper limit for KD4 could be determined, it is clear that substantial positive cooperativity in heme binding exists between the third and fourth hemes.
As a consequence of the substantial negative cooperativity in heme binding between the second and third heme and the low (nM) KD values for the first and second heme, it is possible to isolate nearly pure diheme-6H7H. If the protein concentration is as low as 1.5 μM and 2 equivalents of hemin are added, 90% of the protein species in solution is diheme-6H7H. The remaining 10% exists as monoheme-6H7H (7%) and apo-6H7H (3%). At a higher initial protein concentration of 20 μM, addition of 2 equivalents of hemin gives diheme-6H7H in 97% yield. At the same concentration of protein and assuming the lower limit for KD3 (3 mM), addition of ∼10 equivalents of hemin is required to give a 90% yield of tetraheme-6H7H. Because of the positive cooperativity between the third and forth heme, the remaining 10% exists as diheme-6H7H, not fourth triheme-6H7H. The yield of tetraheme-6H7H is not dependent on the value chosen for KD3, as the product of KD3 and KD4 is constant.
Mono-histidine variant of 6H7H
The above results suggest that, as designed, holo-6H7H adopts an all-parallel topology in which His6/His7 pairs bind heme. To further support the designed structure, a variant that replaces His 7 with Ser was synthesized, yielding a four-helix protein (6H7S) with only four His residues. If the precise designed backbone structure and His side chain orientation of 6H7H is maintained in the 6H7S variant, the four His7 residues would be orthogonal to one another, thereby precluding incorporation of heme groups via axial bis-histidine ligation.
Surprisingly, 6H7S does bind heme. Figure 9 shows the heme titration data that indicates binding of a single heme group with a KD of 12 ± 5 μM. The absorption spectrum of holo-6H7S (not shown) is superimposable with that of diheme-6H7H, showing that the heme is bound via bis-histidine ligation. The KD is, however, significantly larger than that of the first two hemes that bind to 6H7H. Size exclusion chromatography reveals a molecular weight for holo-6H7S that is consistent with a four-helix structure (data not shown). Given these unexpected results, it is necessary to address possible structures for 6H7S that could lead to binding of single low-spin heme group. An antiparallel four-helix structure is ruled out because the two-fold symmetry axis would yield two identical heme binding pockets at opposite ends of the protein. It is very unlikely that in such a structure only one heme group would bind. Therefore, the most plausible explanation for the heme binding data is that 6H7S maintains an all-parallel topology with slight changes in the structure that allow a pair of His 7 residues to orient in such a way as to allow bis-histidine ligation of a single heme. Furthermore, the incorporation of a single heme group must somehow preclude binding of a second heme. This idea was tested by attempting to generate a model for 6H7S in which a single heme group is bound to two His7 residues from diagonally disposed helices. In such a structure, the bound heme would entirely block incorporation of a second heme. It was possible to model such a structure by a slight rotation of the helices and modification of the His side-chain torsion angles (κ1 and κ2). Although a reasonable model could be generated, several unfavorable van der Waals interactions were identified between the bis-histidine ligated heme group and hydrophobic amino acid side chains, thus providing a plausible explanation for the relatively large KD value for heme binding to 6H7S.
ANS is a hydrophobic probe molecule that has been used extensively to probe the “molten globule” and native state of synthetic and natural proteins (Kuwajima 1989; Semisotnov et al. 1991; Quinn et al. 1994; Poklar et al. 1997). ANS has also been used to determine the hydrophobicity of the heme cavity of apocytochrome b5 (Falzone et al. 1996). Since it was observed that apocyt b5 has no apparent affinity for ANS, it was concluded that the heme pocket is not especially hydrophobic.
Addition of 10 μM apo-6H7H to a solution containing 0.5 μM ANS results in a significant increase and blue shift of ANS fluorescence (Fig. 12), indicating that the probe molecule binds to the hydrophobic interior of apo-6H7H, presumably in the designed heme binding pocket. A plot of the fluorescence intensity as a function of protein concentration (not shown) yields a KD of 8.2 μM. In contrast to apo-6H7H, diheme-6H7H does not exhibit any appreciable binding of ANS (Fig. 12), consistent with a compact structure that precludes incorporation of the hydrophobic probe molecule.
Redox potentials for holo-6H7H were measured using standard spectroelectrochemical methods. The data, shown in Figure 13A, was initially fit to a single Nernst (see Supplemental Material); however, the residuals indicated the strong possibility that an additional redox couple was present. Therefore, the data for diheme-6H7H was fit to a double Nernst yielding two potentials of equal amplitude: E1o = −91 and E2o = −133 mV. These potentials presumably correspond to the two distinct bound heme groups. This is confirmed by the observation that λmax of the α-band corresponding to reduced heme blue shifts with decreasing potential (Fig. 13A). If the potentials of the two hemes were identical, no shift in λmax would have been observed. Indeed, the monoheme variant 6H7S does not exhibit any detectable shift in λmax, suggesting that the shift observed for 6H7H is not an artifact of the experiment. The λmax data for diheme-6H7H can in principle be fit to four parameters: two electrochemical potentials, (E1o and E2o) and two λmax values (λ1 and λ2), corresponding to the two heme groups. However, the level of noise in the data precludes a four-parameter fit. Instead, the data was fit fixing E1o and E2o to the values determined above while allowing λ1 and λ2 to vary. This fit yielded the curve shown in Figure 13A, and values for λ1 and λ2 of 560.6 and 556.6 nm, respectively. The excellent fit to the data strongly suggests that indeed the hemes in diheme-6H7H have distinct electrochemical potentials with a ΔEo of −42 mV.
It is tempting, but not entirely correct, to assign the origin of this ΔEo to electrostatic interaction between the two bound hemes. In a molecule with two noninteracting redox centers, the difference in redox potentials, ΔEo, is −35.6 mV at 25°C (Bard and Faulkner 1980). The difference is calculated from −(2RT/F)ln2 and arises from statistical considerations; the Eo of the second heme is more negative (harder to reduce) than the first because there is only one heme rather than two available for reduction. The experimental ΔEo value of −42 mV is close to the theoretical value of −35.6 mV, suggesting that there is little, if any, electronic interaction between the hemes in diheme-6H7H.
Characterization of the redox properties of tetraheme-6H7H was also conducted. Because large concentrations of free heme interfere with the spectroelectrochemical experiment, only 4 equivalents of hemin was used. At this initial concentration of hemin and given the initial protein concentration of 20 μM, the distribution of species in solution is the following: concentration of free heme = 28 μM (35%), diheme-6H7H = 13.7 μM (34%), and tetraheme-6H7H = 6.1 μM (31%). Note that nearly one-third of the added hemin is incorporated into tetraheme-6H7H. Essentially no triheme-6H7H exists in solution as a result of the substantial positive cooperativity in heme binding between the third and forth heme. The redox data for this mixture of free heme, diheme-, and tetraheme-6H7H are presented in Figure 13B. The data fit well to the sum of two Nernst equations yielding potentials of −110 mV and −195 mV. The high potential reduction is nearly identical to the average of the two redox potentials for diheme-6H7H. The lower potential matches that of free heme (Fig. 13C). Attempts to fit the data to more than two Nernst equations were unsuccessful because of the complexity of the mixture of holoproteins and free heme. However, it is clear that the reduction potential of the hemes in tetraheme-6H7H is very similar to that measured for diheme-6H7H. The apparent similarity in heme redox potentials is consistent with the pseudo fourfold symmetry of the designed protein. It suggests that the hemes are bound in pockets with similar local dielectric constants and solvent exposure (Stellwagen 1978).
The four tryptophan residues at position 3 on each helix of 6H7H define one edge of the heme binding site. To probe the local protein environment in the vicinity of the Trp residues for apo- and holo-6H7H, fluorescence spectroscopy was employed. Tryptophan fluorescence has been used extensively as a probe of local protein environment in natural proteins (Chetverin et al. 1980; Lee et al. 1989; Viguera et al. 1992; Genov et al. 1993; Gilardi et al. 1994; Chen and Sanyal 1999). In particular, the tryptophan fluorescence maximum, λF, and full width at half maximum (FWHM) are used as a probe for the extent of solvent exposure: λF for a buried Trp is 330–332 nm, for a Trp with limited exposure λF is 340–342 nm, and λF is 350–353 nm for a Trp exposed to water (Burstein et al. 1973). The tryptophan fluorescence maximum of apo-6H7H is 342.7 nm, and the FWHM is 56.7 nm. These values are indicative of a side chain immobilized at the surface of the protein with limited solvent exposure (Burstein et al. 1973), consistent with the model presented in Figures 1 and 3, Fig. 3.. Titration with heme results in concomitant blue shift in λF as well as significant quenching. Addition of 1.5 equivalents of heme shifts λF to 340.1 nm (FWHM = 56.1 nm). On addition of 2.0 equivalents of heme, the tryptophan fluorescence is no longer observable. The λF for diheme-6H7H, calculated by extrapolation, is 339.4 ± 0.2 nm. This value is similar to that of rabbit pyruvate kinase (λF = 339 nm), which has three partially exposed Trp residues near the surface of the protein (Burstein et al. 1973). These results indicate that, as designed, the bound heme groups are proximal to the tryptophan residues, allowing efficient quenching via energy transfer and that binding of the hydrophobic heme groups results in diminished exposure of the Trp residues to water. This may occur as the result of either direct shielding by heme or through a more closely packed protein structure in which solvent is more effectively excluded from the interior of the protein. Indeed, the calculated solvent exposure of Trp in 6H7H decreases from 85% to 62% on heme binding assuming no gross change in the backbone structure. (Solvent exposure was calculated using the program DSSP [Kabsch and Sander 1983].)
Visible circular dichroism
Hemin is an achiral molecule and, therefore, in solution does not exhibit a CD signal. However, when hemin is incorporated into the highly asymmetric environment of a protein, pronounced dichroism in the Soret region is often observed. Heme CD has been attributed to coupled oscillator interactions between heme transitions and allowed ππ* transition on nearby aromatic residues (Hsu and Woody 1971). It has also been shown that CD can be the result of coupled oscillator interactions with peptide ππ* transitions and high-energy peptide and thioether sulfur transitions (Blauer et al. 1993). In addition, inherent heme chirality arising from nonplanar distortions can contribute to the Soret CD (Blauer et al. 1993). More recently, it has been shown in studies with hemoglobin that heme–heme excitonic interactions also induce CD in the Soret band (Goldbeck et al. 1997). This has also been supported by theoretical work (Woody 1985).
To investigate any potential heme–heme excitonic coupling in holo-6H7H, CD spectra in the Soret region were collected for solutions in which 1, 2, and 3 equivalents of hemin were added to 55.6 μM of apo-6H7H. Figure 15 shows these CD spectra and for comparison the CD spectrum of the monoheme 6H7S variant. Holo-6H7S displays classic negative and positive Cotton effects at 403 and 424 nm, respectively. Furthermore, the normalized CD spectra for holo-6H7S are identical at all heme concentrations up to 1 equivalent (data not shown). The CD spectra for holo-6H7H are more complex, and in contrast to 6H7S, display dramatic spectral modulation as a function of initial hemin concentration. To better understand the spectra, it is important to recognize the distribution of holo-species at each initial hemin concentration. Using the previously determined KD values, 1 equivalent of added hemin yields ∼20% monoheme-6H7H, 40% diheme-6H7H, and 20% apo-6H7H. At 2 equivalents of hemin, 98% is diheme-6H7H. At 3 equivalents, 72% is diheme-6H7H and 27% is tetraheme-6H7H. The change in the CD spectra on going from 1 to 2 equivalents is presumably the result of a higher concentration of diheme-6H7H relative to monoheme-6H7H. As major changes in the protein structure have been ruled out from far-UV CD studies, the most likely explanation for the spectral differences is heme–heme excitonic coupling in the diheme protein that is necessarily absent in monoheme-6H7H. Because of the positive cooperativity in heme binding, it was not possible to obtain a spectrum corresponding to pure monoheme-6H7H. Differences between the spectrum corresponding to 3 equivalents of added heme and that of diheme-6H7H are tentatively assigned to the presence of tetraheme-6H7H in solution. Again, the spectral differences most likely arise from heme–heme excitonic coupling in tetraheme-6H7H distinct from that present in diheme-6H7H. Although not electrostatically coupled, these results strongly suggest that the hemes in holo-6H7H are electronically coupled.