Cytoglobin is a recently discovered hexa-co-ordinate haemoglobin that does not appear to function as a classical oxygen-binding protein. Its function is unknown and studies on the effects of changes in its expression have not decisively determined its role within the cell. In the present paper, we report that the protein is transformed from hexa-co-ordinate to penta-co-ordinate on binding a lipid molecule. This transformation occurs with the ferric oxidation state of the protein, but not the ferrous state, indicating that this process only occurs under an oxidative environment and may thus be related to redox-linked cell signalling mechanisms. Oleate binds to the protein in a 1:1 stoichiometry and with high affinity (Kd=0.7 μM); however, stopped-flow kinetic measurements yield a Kd value of 110 μM. The discrepancy between these Kd values may be rationalized by recognizing that cytoglobin is a disulfide-linked dimer and invoking co-operativity in oleate binding. The lipid-induced transformation of cytoglobin from hexa-co-ordinate to penta-co-ordinate does not occur with similar hexa-co-ordinate haemoglobins such as neuroglobin, and therefore appears to be a unique property of cytoglobin among the haemoglobin superfamily. The lipid-derived transformation may explain why cytoglobin has enhanced peroxidatic activity, converting lipids into various oxidized products, a property virtually absent from neuroglobin and much decreased in myoglobin. We propose that the binding of ferric cytoglobin to lipids and their subsequent transformation may be integral to the physiological function of cytoglobin, generating cell signalling lipid molecules under an oxidative environment.
Cygb (cytoglobin) is a recent addition to the haemoglobin superfamily. This low-abundance protein is ubiquitously expressed in vertebrate tissues, but found in higher concentrations in the brain, eyes, liver, heart and skeletal muscle [1,2]. Unlike Hb (haemoglobin) and Mb (myoglobin) where the haem iron is penta-co-ordinate, the ferrous and ferric oxidation states of Cygb are hexa-co-ordinate with both proximal and distal histidine residues occupying the fifth and sixth co-ordination sites of the haem iron in the absence of exogenous ligands . This hexa-co-ordinate configuration of the haem iron is also observed with Ngb (neuroglobin) and non-symbiotic plant haemoglobins [4,5]. Previously known as histoglobin and stellate cell activation-associated protein, Cygb is capable of reversibly binding oxygen and has an affinity for oxygen comparable with that of Mb . Thus the function of Cygb has been assumed by some to facilitate oxygen transport to the mitochondrial respiratory chain [1,7,8]. However, Cygb is present at low concentrations in cells that are not generally associated with particularly high metabolic rates and thus oxygen consumption (unlike Mb). Therefore the proposed role of Cygb in facilitating oxygen diffusion to the respiratory chain of the mitochondria seems unlikely . There have been many alternative suggestions for the in vivo function of this protein. It has been proposed that Cygb may function as an NO dioxgenase, due to its rapid reaction with NO and its up-regulation in Mb-free amphibians, suggesting that it may function as an important NO sink in the vascular wall . However, like the oxygen diffusion hypothesis, the low abundance of the protein in most cells casts doubt on Cygb acting as a cell protectant against compounds such as NO or hydrogen peroxide.
The main function of Cygb thus remains unclear. It has been observed, however, that hydrogen peroxide induces Cygb up-regulation . This change in expression was not observed with other stresses, including heat stress, high osmolarity and UV radiation. Cygb has been reported to have no appreciable catalase activity , but has considerable peroxidatic activity, consuming both hydrogen peroxide and lipid peroxides [7,12]. Cygb-knockdown neuroblastoma cells exacerbated cell death upon treatment with hydrogen peroxide, suggesting a role for protection against oxidative stress , perhaps as reactive oxygen scavengers . Cygb may play a role in fibrotic organ disorder as it is up-regulated in fibrotic lesions of the pancreas and kidney with overexpression of Cygb in NIH 3T3 fibroblast cells inducing a decrease in migratory activities and increasing the expression of collagen mRNA . Cygb expression has also been reported to be involved with tumour suppression, with hypoxic cancers significantly up-regulating Cygb . Thus expression of Cygb may also provide a new target for therapy of cancer .
The gene and protein expression studies have correlated Cygb up-regulation with functions that may be linked to protection of cells against oxidative stress, perhaps through the redox chemistry of Cygb. However, the biochemical mechanism of Cygb cellular protection remains unknown. Our investigations of Cygb reveal a plausible mechanism through which such protection may be afforded. In the present paper, we report that ferric, but not ferrous, Cygb undergoes a conformational transition from hexa-co-ordinate to penta-co-ordinate haem on binding, with high affinity, one lipid molecule per Cygb. This conformational change induced in Cygb at low lipid concentration appears to be a unique function among the haemoglobin superfamily. We propose, under an oxidative environment, that the protein will be oxidized to the ferric form, allowing interactions with lipid and the opening of the haem pocket to facilitate redox chemistry that catalyses the oxidization of the lipid. Many products of lipid oxidization have known cell signalling functions, such as isoprostanes [16–19] or electrophilic lipids [20–23]. Therefore, even though the function of this protein appears to be substantially pro-oxidant, the low concentration of this protein facilitates cell signalling processes that may allow the cell to up-regulate antioxidant defences before extensive oxidative damage occurs, or potentially to induce apoptosis if the balance of oxidative chemistry cannot be maintained.
Cytoglobin engineering and expression
The gene for human CYGB was purchased from OriGene (via Cambridge Bioscience), and was contained in a pCMV6-AC cloning vector. To facilitate protein expression and purification the gene was subcloned from the cloning vector into a pET28a expression vector (Merck) following insertion of a restriction site (FaqI). The protein lead sequence contained a thrombin-cleavable His6 tag (Met-Gly-Ser-Ser-6×His-Ser-Ser-Gly-Leu-Val-Pro-Arg-Gly-Ser, with the thrombin-recognition site underlined). At each stage the plasmid DNA sequence was confirmed by GATC Biotech. Vector was transformed into XL1B cells or BL21 DE3 cells (Agilent, Stratagene) for cloning or protein expression respectively. Transformation was achieved using the heat-shock method.
For protein expression Escherichia coli was added to Luria–Bertani medium (10 g of tryptone, 10 g of sodium chloride and 5 g of yeast extract per litre) containing kanamycin (50 μg/ml) and shaken at 120 rev./min at 37 °C. When the attenuance was 0.8 at 600 nm, 1 mM IPTG (isopropyl β-D-thiogalactopyranoside), 500 μM 5-aminolaevulinic acid and 100 μM ferric citrate were added to the broth. CO gas was bubbled through the solution for 2 min and the flasks were sealed and incubated for a further 18–24 h . Cells were isolated by centrifugation (15000 g) and cells were lysed using an Avestin Emulsiflex C5 homogenizer. His-tagged Cygb was purified using a His-tag nickel-affinity column (GE Healthcare), according to the manufacturer's instructions. Imidazole was removed through dialysis and the His-tag was cleaved through incubation with bovine thrombin (Sigma–Aldrich; 10 units/mg of protein) at room temperature (22 °C) overnight with gentle mixing. His-tag-free protein was purified using the nickel-affinity column, and the thrombin was separated from the Cygb by precipitating the thrombin at pH 5.0 (using a 0.1 M sodium acetate buffer). Thrombin was discarded after centrifugation (4000 g for 10 min) and the Cygb was filtered through a Vivascience Vivaspin filter [100 kDa MWCO (molecular mass cut-off)] to remove trace bacterial catalases, dialysed and concentrated using a Vivaspin or Vivacell filter (5 kDa MWCO).
Optical and stopped-flow spectroscopy
Optical spectra were taken using a Varian Cary 5E UV-Vis-NIR spectrophotometer. Sodium oleate (Sigma–Aldrich) was dissolved in water to a concentration of 50 mM with gentle warming to ensure complete dissolution. Oleate was diluted in water and titrated (5 μl aliquots, giving a final concentration of ~0.5 μM) into Cygb (5 μM) in 50 mM sodium phosphate buffer (pH 7.4) and optical readings were taken following each addition. Cardiolipin (Sigma–Aldrich) was dissolved in ethanol and added to Cygb in 20 mM Hepes buffer (pH 7.4). Stopped-flow absorbance spectroscopy was used to measure the kinetics of change in haem iron co-ordination using an Applied Photophysics SX20 stopped-flow spectrometre fitted with both a single wavelength photomultiplier and a diode array spectrometer. Cygb [10 μM in 50 mM sodium phosphate buffer (pH 7.4)] was rapidly mixed with oleate [0–2000 μM in 50 mM sodium phosphate buffer (pH 7.4)] to give final concentrations of one-half of their original concentration.
Liposome preparation and measurement of lipid oxidation
Soya bean phospholipids (50 mg, lecithin ‘asolectin’, 19% phosphatidylcholine, Sigma–Aldrich, type II-S P-5638) was suspended in 10 ml of 25 mM sodium phosphate buffer (pH 7.4) containing 25 μM DTPA (diethylenetriaminepenta-acetic acid) and the solution was sonicated in a water bath for 10 min. Further sonication was used if lumps of lecithin were still present. Small unilamellar liposomes were formed from the sonicated solution using a Northern Lipids stainless steel extruder fitted with a Whatman nucleopore drain disc (25 mm diameter) and two layers of Whatman nucleopore polycarbonate membranes (25 mm diameter with a 0.1 μm pore size) as described previously . Liposomes were stored at 4 °C and used within 2 h of preparation.
An Agilent 8453 diode array spectrophotometer was used to measure the oxidation of liposomes. Ferric Mb or Cygb was incubated with the liposomes (200 μg/ml) at 25 °C at pH 7.4 (25 mM sodium phosphate containing 25 μM DTPA). The oxidation of liposomes was measured by the appearance of lipid-based conjugated dienes measured optically at 234 nm. To correct for changes in light scattering, a three-point baseline drop was calculated from optical measurements taken at 220, 234 and 250 nm. The lipid hydroperoxide HPODE [hydroperoxyoctadeca-(9Z,11E)-dienoic acid; ϵ234=2.5×104 M−1·cm−1]  was used to calculate the molar absorption coefficient using the above method (ϵ234=1.1×104 M−1·cm−1).
Samples were analysed on an Agilent HP1100 HPLC fitted with a diode array spectrophotometer. The column used was a Zorbax StableBond 300 C3 250 mm×4.6 mm column fitted with a 12 mm×4.6 mm guard column. Solvents were (A) 0.1% TFA (trifluoroacetic acid) and (B) acetonitrile containing 0.1% TFA. The gradient was initially 35% solvent B, stable for 10 min, increasing to 37% solvent B over 5 min. This was increased to 40% solvent B over 1 min and then to 43% solvent B over 10 min. The flow rate was 1 ml/min, and the temperature was 25 °C.
Standard and concerted model of lipid binding to cytoglobin
Addition of lipids such as oleate and cardiolipin appear to follow a standard classic fractional saturation binding curve that can be described by eqn (1) below:
where [PT] is the total protein concentration, [S] is the substrate concentration (oleate) and K is the binding constant. Data were fitted to equations using the least-squares method by either the Microsoft Excel solver program or Originlab Origin.
However, the lipid-binding data may also be fitted to the concerted model, previously devolved for the allosteric binding of oxygen to Hb . The equation to fit the binding curve of a co-operative protein is given below (eqn 2):
where n=number of subunits,http://www.BiochemJ.org/bj/434/bj4340483add.htm) was generated: n=2, c=0.005882 and L=5340.
Cygb (80 μM) with or without sodium oleate (160 μM) was transferred (250 μl) to Wilmad SQ EPR tubes (Wilmad Glass). Tubes were flash-frozen in dry-ice-cooled methanol. Frozen samples were transferred to liquid nitrogen (77 K) where they were stored prior to measurements.
All EPR spectra were measured using a Bruker EMX EPR spectrometer (X-band) equipped with a spherical high-quality resonator SP9703 and an Oxford Instruments liquid helium system as described previously . The modulation frequency was 100kHz. Accurate g values were obtained using the built-in microwave frequency counter and a 2,2-diphenyl-1-picrylhydrazyl powder standard, the g value for which is g=2.0037±0.0002 . The EPR spectra of the blank samples (frozen water) were subtracted from the EPR spectra of the protein samples to eliminate the baseline caused by the walls of the resonator, quartz insert or quartz EPR tube. Relative concentrations of the HS (high-spin) and LS (low-spin) ferric haem forms in Cygb were estimated from the intensity of the HS ferric haem signal which has been compared wih a horse metMb standard sample stored at 77 K. The concentration of the HS ferric haem form in the standard sample has been accurately determined as described previously . The comparison was possible owing to close lineshapes of the EPR signals of the HS ferric haem forms in Cygb and horse Mb.
RESULTS AND DISCUSSION
Cygb expression in E. coli induces catalase expression
It has been reported that under aerobic conditions the expression of Cygb in E. coli produces a ‘pine-green’ coloured protein . Under the hypoxic conditions we employed (bubbling argon gas to partially deoxygenate the culture) the protein produced was also green in colour (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/434/bj4340483add.htm). Analysis by HPLC showed that this colour is due to significant incorporation of green d-type ‘chlorin’ haem into Cygb in place of the red b-type haem (Fe–protoporphyrin IX, see Supplementary Figure S3 at http://www.BiochemJ.org/bj/434/bj4340483add.htm). In addition, there was significant contamination of the Cygb preparation with bacterial catalase (which contains d-haem) that necessitated further purification of the protein to remove catalase activity. This catalase production did not occur to any significant extent under identical conditions with the same expression vector expressing Mb instead of Cygb. Thus it appears that the expression of Cygb by E. coli is accompanied by extensive catalase production, resulting in the presence of d-type haems that are acquired by the apo-form of Cygb to generate the ‘green’ or ‘pine-green’ coloration. It is known that adding CO to the culture prior to induction of protein expression prevents this problem . We show that adding CO results in Cygb essentially only incorporating b-type red haem. Our interpretation is that CO strongly stabilizes the protein as the ferrous-CO form and prevents its redox activity. This conclusion suggests that redox activity plays a role in the cell signalling pathways, inducing up-regulation of catalase production by the bacteria. This is in keeping with our proposal that Cygb functions as a cell signalling protein.
Optical changes and kinetics of lipid binding to cytoglobin
The optical spectrum for Cygb in the ferric oxidation state is essentially identical with that described previously  and has a Soret (γ) band at 416 nm, and β and α bands in the visible region of the spectrum at 535 and 565 nm respectively, showing the haem iron in a LS HisF8–Fe(III)–HisE7 co-ordination. Figures 1(A) and 1(B) show the optical changes induced by oleate binding to ferric Cygb. The binding of oleate creates a hypsochromic shift of the Soret peak from 416 nm to 412 nm accompanied with a slight decrease in intensity. The visible peaks also decreased in intensity as the oleate concentration increases, with the appearance of a peak at approx. 620 nm. The latter is consistent with the formation of a HS penta-co-ordinate ferric iron, similar to Mb or Hb, and indicative of water in the sixth co-ordination site of the iron .
Lipid-induced optical changes to Cygb
The optical changes of ferric Cygb are plotted as a function of oleate concentration in Figure 2(A). Addition of oleate appears to follow a classic single-site-binding curve that can be described by eqn (1) given in the Experimental section. The best fit to the data points yields a binding dissociation constant (Kd) of 0.70 μM and the total protein concentration [PT] of 4.5 μM, close to the calculated protein concentration of 5 μM as measured by optical spectroscopy. Thus the stoichiometry of oleate binding to Cygb is essentially 1:1. Figure 2(B) shows the dependence of the pseudo-first-order rate constant kobs as a function of oleate concentration, measured by stopped-flow spectrophotometry. This linear relationship is consistent with a simple mechanism and can be described by eqn (3):
where kon and koff are the association and dissociation rate constants respectively. Thus the slope of the graph gives the kon value, calculated to be 4.53×104 (±8.58×102) M−1·s−1 and the intercept at [oleate]=0 gives the value of koff, determined as 5.0±0.32 s−1, giving a Kd value of approx. 110 μM.
Oleate binding to Cygb
Oleate is not the only lipid to induce optical changes to Cygb upon binding; cardiolipin also binds to Cygb, resulting in a transformation of the haem-iron co-ordination (see Supplementary Figure S3). The titration also shows identical optical changes that were observed in the presence of oleate (results not shown) and follows classical binding as described in eqn (1) in the Experimental section. Also, like oleate, the binding stoichiometry of cardiolipin to Cygb is almost 1:1 with a calculated total protein concentration of 6.8 μM, compared with the actual protein concentration of 5 μM (thus giving a 1:1.4 stoichiometry of protein to cardiolipin). The binding dissociation constant was calculated as 0.64 μM, virtually identical with the Kd for oleate binding (Figure 2A).
The binding of oleate to Cygb appears, from Figures 1 and 2, to occur by a simple mechanism. However, the kinetics of oleate binding to Cygb (Figure 2B) is not compatible with the results of the oleate titration, giving Kd values of 110 and 0.7 μM from the kinetics and titration data respectively. Thus the kon and koff values for oleate binding indicate that with 5 μM of protein and 10 μM oleate, the occupancy of the lipid should be approx. 20% (kon=0.9 s−1, koff=5 s−1). However, titration of Cygb with oleate clearly shows that over 80% of the protein is in the penta-co-ordinate conformation with a 1:2 protein/oleate ratio. Furthermore, although the kinetics predict low occupancy, the spectral changes seen in these stopped-flow experiments are typical of full occupancy and do not change on increasing the lipid concentration, indicating that the protein is saturated (see Supplementary Figure S5 at http://www.BiochemJ.org/bj/434/bj4340483add.htm). Clearly a single-site mechanism of the form ‘lipid+Cygb (hexa-co-ordinate)↔complex (penta-co-ordinate)’ is too simple and the data demand a more complex model of lipid binding and change in haem iron co-ordination. As Cygb is a homodimer, a possibility arises that the subunits act co-operatively as lipid binds. Thus the binding of a lipid to one subunit could drive both subunits into the penta-co-ordinate state, with the second subunit binding lipid rapidly and with higher affinity. The kinetics would thus reflect the initial binding event, but report the full optical transition. The second lipid binding would not be monitored in such transient experiments, but would influence the equilibrium titration experiments, thus allowing a discrepancy in the value of Kd between the measurements. Cygb co-operativity has previously been reported for oxygen binding to the ferrous protein, with both positive and negative co-operatively observed depending on temperature and pH . The titration curve from Figure 2(A) can also be fitted to a co-operative ‘concerted model’ (with a Hill coefficient of 1.4) of protein–ligand binding for a dimeric protein, with a low-affinity tense or ‘T’ state and a high-affinity relaxed or ‘R’ state of the concerted model . The resulting residuals between the data and the fit are not significantly different from those in the fit using the fractional saturation binding curve (see Supplementary Figure S1). Therefore co-operativity between the subunits of the dimer appears plausible.
Support for this proposal may be found in the structure of Cygb. Although the quaternary structure of Cygb in situ is currently unknown, in the crystal Cygb exists as a cysteine-linked homodimer with each haem pocket facing opposite sides of the dimer surface . In a double cysteine mutant of Cygb (C38S, C83S) the disulfide bridge linking the homodimer is absent. Under these conditions one subunit of the homodimer has the haem iron in a hexa-co-ordinate state, whereas the other subunit has partial haem penta-co-ordination, resulting from a shift of the E1–E10 segment . A cysteine disulfide bridge (Cys46–Cys55) is also present in Ngb and has been found to influence the affinity of the iron atom for the distal histidine . Reduction of the disulfide bond, or mutation of the cysteine residues, leads to an increase in the affinity of the iron atom for the distal histidine residue . Thus removing the disulfide bridge appears to weaken the hexa-co-ordinate nature of Cygb, but strengthen it in Ngb. It appears, therefore, that intersubunit interactions influence the penta-co-ordinate–hexa-co-ordinate equilibrium of the haem. This conclusion is in keeping with our proposal that binding of lipid to one subunit will affect the dissociation of the distal histidine residue from the iron in the other subunit, driving it HS, and enhance the binding of lipid to this second subunit.
Distinct to the Cygb family are extended N- (1–17) and C- (172–190) termini; both segments are highly disordered, hence these segments do not appear in the crystal structures. Neither of the two segments falls close to crystal contacts, indicating that their disorder should not be ascribed to crystal packing effects . The function of these disordered end segments is currently unknown. However, in the light of the fact that the protein appears to specifically bind lipids, it is feasible that these segments may be involved in lipid binding.
Effect of oleate on the acid/alkaline spectral transitions
The optical transitions observed in Cygb in response to changes in pH are shown in Figure 3(A). Ferric Cygb in the absence of oleate shows multiple acid–alkaline transitions with four distinct optical species between pH 5 and pH 13 linked by three pKa values. The highest pKa value at 13.3 is accompanied by an increase in absorbance at 390 nm and is assigned to haem release on protein denaturation. The lower two pKa values represent optical changes between three distinct species, the spectra of which are shown in Figure 3(B). The transition with a pKa of 8.2 accompanied changes in the spectrum from 416, 532 and 543 nm at low pH to 413, 565 and 578 nm at high pH with an additional band at 620 nm. These changes are similar to that observed upon the addition of oleate (Figure 1) and are thus likely to also represent a LS-to-HS spin change in haem iron co-ordination. Therefore at physiological pH the ferric Cygb shows partial penta-co-ordination (14%). A second acid–alkaline phase is observed with a pKa of 11.0, accompanied by a shift of the Soret peak maximum to 418 nm and the disappearance of the 620 nm band in the more alkaline species (Figure 3B, dashed line). This high pH species is likely to be a hydroxide ligand bound to the haem iron, an acid–alkaline transition also observed in other penta-co-ordinate haemoglobins, although usually at lower pH values (human Hb pKa=8.0, horse Mb pKa=8.9 ). Thus this transition for Cygb is far outside the normal physiological pH range. Thus from acid to alkali the three species are (i) His–Fe(III)–His, (ii) His–Fe(III)–H2O and (iii) His–Fe(III)–OH−.
Effect of oleate on the acid–alkaline transition of ferric Cygb
The pH titration of ferric Cygb in the presence of 2-fold excess oleate is also shown in Figure 3(A). In the presence of oleate, there is an acid–alkaline transition with a pKa of ~13.6, indicative of protein denaturation. Additionally, a transition with a pKa of 10.6 can be seen, essentially identical with that observed in the absence of oleate. The most acidic transition, however, shows small changes in the optical spectra with a pKa of 7.0, significantly different from the transition with pKa of 8.2 observed in the absence of oleate. The optical spectra of the three optical species in the presence of oleate are shown in Figure 3(C). The spectrum at pH 6.5 (Figure 3C, solid line) is considerably different from the acidic spectrum observed in the absence of oleate (Figure 3B, solid line), with a Soret peak at 409 nm and three peaks in the visible region at 535 nm, 568 nm and 620 nm. The initial transition with pKa of 7.0 shows only minor changes in the positions and intensity of the peaks (pH 9.8, Figure 3C, dotted line). Thus, for the lipid bound Cygb, the distal histidine residue is already displaced; therefore the most acid form of the protein is already penta-co-ordinate. The small transition observed with a pKa of 7 may represent a slight disturbance of the optical spectrum due to protonation of a group in the proximity of the iron, probably the distal histidine residue itself.
The results of the present study suggests that the hexa-co-ordinate conformation of Cygb can easily be disrupted in the ferric oxidation state, much more so than other similar hexa-co-ordinate haemoglobins, such as Ngb, despite their common ability to exchange the distal histidine residue for ligands such as O2 and CO.
Effect of oleate on the EPR spectrum of Cygb
The EPR spectrum of Cygb in the presence and absence of oleate is shown in Figure 4. Signals at g=3.18 and g=2.05 are observed, indicative of a LS haem iron. These are components of a previously reported LS signal of ferric Cygb (g1=1.20, g2=2.06 and g3=3.20), and assigned to an LS3 signal typical of a HisF8–Fe(III)–HisE7 co-ordination [33,34]. Furthermore, a HS signal at g=5.87 is present. This axial HS ferric haem signal is virtually identical with that observed in other globins, such as Mb and Hb, and denotes a water molecule co-ordinated to the haem iron [30,35,36]. This HS signal has also previously been observed in ferric Cygb . Addition of oleate at pH 7.4 decreases the LS signal, with a concurrent increase in the g=5.87 HS signal. Although this signal is broader than in the absence of oleate, it is entirely consistent with water co-ordinated to the iron as suggested by the optical spectra (Figure 1). The LS form of the protein decreases from 70% to 25% on addition of oleate, showing that addition of lipid destabilizes the LS form of the protein, shifting the spin-state equilibrium strongly towards the HS state, as concluded from the optical data.
EPR spectra of cytoglobin in the presence and absence of oleate
Although in general agreement, the EPR and optical experiments show some quantitative discrepancies: the EPR spectra report more HS protein in the absence of and less in the presence of oleate than expected from the optical data. An explanation for these differences may be found in our observation that the colour of the protein is temperature-dependent. At room temperature or when frozen at >−10 °C, the protein solution at neutral pH is red, typical of a bis-His ferric haem protein. However, at temperatures of −20 °C or below, the frozen sample takes a brownish hue, typical of HS water-co-ordinated ferric haem. Thus it appears that low temperature alters the equilibrium between HS and LS forms, destabilizing the LS form. Partial dissociation of the histidine from the iron on lowering the temperature is unexpected as such co-ordination is usually strengthened under these conditions. This indicates that the conformation of the protein must be poised in such a way that in lowering the temperature, interactions elsewhere in the protein are strengthened, leading to a conformational transition that strains the Fe–His bond. In the presence of oleate such strain between the parts of the molecule is relieved. Under these conditions, lowering the temperature favours Fe–His bond formation and stabilizes the LS form. This can also be seen through colour changes on freezing.
A C38S/C83S double mutation of Cygb completely removed the HS signal, stabilizing the LS form of the protein . Again, this supports the hypothesis that a degree of co-operativity exists between the protein dimers.
Comparison between lipid-induced spectral changes in Cygb and Ngb
Ngb is very similar to Cygb in that it contains a hexa-co-ordinate haem, is present at low concentrations in the cell and there is a possible signalling role for the oxidized form of the protein . Thus we tested whether lipids such as oleate induced similar conformational changes in ferric Ngb. The spectral comparison between Cygb and Ngb on oleate addition is shown in Figures 5 and 6. As seen in Figure 1, the addition of oleate at stoichiometric concentrations induced the conformational change of Cygb, showing an appearance of a peak at 620 nm (Figures 5A and 5B). The addition of oleate to Ngb at these concentrations does not induce any significant optical changes and therefore does not induce any changes in the iron co-ordination of the haem moiety. At concentrations of oleate in large excess, the gradual appearance of a band at 640 nm can be observed (Figures 5C and 5D). This implies that the haem iron is also transformed from hexa-co-ordinate to HS penta-co-ordinate; however, the effect is only observed at very high oleate concentrations and the process is not complete even at concentrations of 1 mM oleate, 200-fold excess in relation to the protein concentration. This transformation is also observed with Cygb, with a loss of the 620 nm band formed at low oleate concentration, and the appearance of a 640 nm band at high oleate concentration, identical with that observed in Ngb (Figures 5A and 5B). These optical changes are accompanied with a major change in the Soret spectrum with a partial hypsochromic shift to a band 378 nm, indicative of the haem in free solution. Thus, at high oleate concentrations, both proteins are partially denatured by the excess lipid, although Cygb seems to be more resistant to denaturation by lipids compared with Ngb (Figure 6).
Effect of low and high oleate concentration on the optical spectrum of ferric Cygb and Ngb
Effect of oleate on the optical spectra of Cygb and Ngb
Mb is known to participate in lipid oxidation (peroxidation) reactions that can be detected in vivo in some disease conditions, such as acute renal failure following rhabdomyolysis [18,38,39]. Thus Mb is a useful protein with which to compare Cygb with regards to the capacity to catalyse lipid oxidation reactions. Figures 7 and 8 show the oxidation of lecithin liposomes catalysed by ferric Mb and ferric Cygb at pH 7.4. The optical changes for the reaction of both Cygb and Mb with liposomes show an increase in the absorbance at 234 nm, signifying the oxidation of the liposome lipids (Figure 7). The increase in absorbance is driven by the free-radical-mediated oxidation of lipids, causing the re-arrangement of dienes, such as the 1,4 cis-cis-pentadiene systems found in arachidonate and linoleate, to a 1,3-cis-trans-conjugated diene, showing absorbance maxima at 234 nm or a 1,3-conjugated diene further conjugated to another diene or keto group showing absorbance maxima at 280 nm [40,41]. Free radical damage can also result in the depletion of conjugated diene, as shown previously . However, under these experimental conditions, the extent of conjugated diene formation outweighs the extent of conjugated diene destruction and hence can be used to follow lipid oxidation, but should not be used to quantify the overall extent of lipid oxidation. Accompanying the increase in conjugated dienes are optical changes of the Cygb Soret haem spectrum, which shows a decrease in absorbance associated with haem bleaching from redox-mediated free radical damage (Figure 7B). The Soret peak following addition of Cygb to the liposomes is at 412 nm, not at 416 nm. This shows that the normal LS hexa-co-ordinate ferric protein is not present immediately after addition of the protein to the liposomes, with the Cygb spectrum resembling the optical spectrum of Cygb following oleate addition.
Oxidation of liposomes by Cygb and Mb
Comparison of the lipid oxidation capability of Mb and Cygb
Mb-catalysed oxidation of liposomes is shown in Figures 7(C) and 7(D). Unlike Cygb, however, the optical changes in the Soret spectrum of the Mb haem shows both increases in intensity at 425 nm as well as decreases in haem intensity at 408 nm, with an isosbestic point at 420 nm. These optical changes are indicative of the formation of ferryl Mb from ferric Mb, with no appreciable loss of haem optical intensity due to Mb degradation.
Following a lag period, both Mb and Cygb induce a cascade of lipid oxidation. The lag periods were 7.6 min and 42.2 min for Cygb and Mb respectively (Figure 8A). During the lag period, small amounts of lipid peroxides, present in the liposome preparation, react with the haem protein to generate ferryl haem iron. This ferryl in turn reacts with lipids to cause free radical damage and the generation of lipid oxidation products, including more lipid peroxides. The level of haem protein redox cycling eventually generates a critical level of lipid-based radicals which causes the cascade of lipid oxidation, shown by the rapid increase in the signal at 234 nm. Cygb is much more pro-oxidant leading to a cascade of lipid oxidation 5-fold faster than Mb. Ngb showed no appreciable lipid oxidation reaction during the time course of the reaction beyond autoxidation of the lipids in an aerobic buffer (Figure 8). Measurement of the maximum rate of lipid oxidation can also be used as a measure of the pro-oxidant capacity of the proteins. The maximum rates of conjugated-diene formation for Cygb, Mb and Ngb are shown in Figure 8(B). Cygb oxidizes lipids during the cascade 5-fold faster compared with Mb, with rates of conjugated-diene formation reaching a maximum of 26.7 nM·s−1 and 134.3 nM·s−1 for Mb and Cygb respectively. The extent of conjugated-diene formation with Mb was greater than with Cygb; 65.1 μM for Mb and 55.3 μM for Cygb after lipid oxidation was complete. This does not necessarily mean that more lipids were oxidized by Mb as conjugated-diene formation does not represent an unambiguous quantitative measure of overall lipid oxidation. Bleaching of Cygb could result in incomplete oxidation of lipid; however, addition of 3 μM Cygb to liposomes instead of 1 μM did not change the overall extent of conjugated-diene formation, indicating that lipid oxidation was complete with the lower concentration of Cygb (results not shown).
The results of the present study show that the addition of lipids to Cygb in the ferric oxidation state, but not the ferrous state, induces optical changes that are consistent with a protein conformational change resulting in a transition of the haem iron from hexa-co-ordinate to penta-co-ordinate. This lipid-induced transformation occurs with a range of lipids and does not occur under comparable conditions to similar hexa-co-ordinate haemoglobins such as Ngb. Therefore this lipid-induced transformation of Cygb iron co-ordination may be a unique property of this haemoglobin and thus may be related to its physiological function. The capacity of Cygb to induce lipid oxidation reactions in vitro, together with the lipid-binding data, lead us to propose that the interaction of Cygb with lipids explains their potent pseudo-peroxidase activity and is likely to play an important role in the physiological function of the protein. Our results show that Cygb is more potent at generating lipid oxidation products compared with Mb. As Ngb appears to have little capacity to oxidize lipid, this suggests that the physiological functions of Cygb and Ngb are very different. This high level of Cygb pro-oxidant activity at first appears to contradict many studies that have proposed that Cygb functions to protect the cell against oxidative stress and ROS (reactive oxygen species). However, with the low concentration of the protein in cells, together with the degradation of the haem under advanced redox activity means that this pathway of lipid oxidation is unlikely to cause extensive cellular damage. This mechanism is similar to that proposed by Kagan et al.  with the interaction of cytochrome c and cardiolipin. In that study lipid oxidation is catalysed by a peroxidase activity of cardiolipin-bound cytochrome c, inducing pro-apoptotic pathways. Modified lipids formed by the redox activity of cytochrome c, as well as other haemoglobins, are known potent cell signalling molecules. Free-radical-catalysed lipid oxidation is known to induce the generation of biologically active molecules, such as the vasoactive isoprostanes, which participate as pathophysiological mediators in oxidant injury , or electrophilic lipids, reacting with thiols and cysteine residues that affect various cell signalling pathways [20,22]. Thus under an oxidative environment, Cygb will be oxidized to the ferric form, allowing interaction between the protein and lipid to form biologically active cell signalling molecules that may induce cell signalling pathways and an antioxidant response to the oxidative environment.
Brandon Reeder conceived the study, performed the protein engineering, and expression, optical and kinetic studies, HPLC and liposome preparation. Brandon Reeder and Michael Wilson analysed data and wrote the paper. Dimitri Svistunenko generated and interpreted EPR data.
We thank Badri Rajagopal and Jonathan Worrall (Department of Biological Sciences, University of Essex, Colchester, Essex, U.K.) for their kind donation of neuroglobin and Tom Brittain (University of Auckland, Auckland, New Zealand) for the kind donation of the plasmid containing the neuroglobin gene.
This work was supported by the Biotechnology and Biological Sciences Research Council [grant number BBF0076631].