Binding of Yeast and Human Cytochrome c to Cardiolipin Nanodiscs at Physiological Ionic Strength

Abstract

Binding of cytochrome c (Cytc) to membranes containing cardiolipin (CL) is of considerable interest because of the importance of this interaction in the early stages of apoptosis. The molecular-level determinants of this interaction are still not well defined and there appear to be species-specific differences in Cytc affinity for CL-containing membranes. Many studies are carried out at low ionic strength far from the 100 – 150 mM ionic strength within mitochondria. Similarly, most binding studies are done at Cytc concentrations of 10 μM or less, much lower that the estimated range of 0.1 to 5 mM Cytc present in mitochondria. In this study, we evaluate binding of human and yeast Cytc to CL nanodiscs using size exclusion chromatography at 25 μM Cytc concentration and 100 mM ionic strength. We find that yeast Cytc affinity for CL nanodiscs is much stronger than that of human Cytc. Mutational analysis of the site A binding surface shows that lysines 86 and 87 are more important for yeast Cytc binding to CL nanodiscs than lysines 72 and 73, counter to results at lower ionic strength. Analysis of the electrostatic surface potential of human versus yeast Cytc shows that the positive potential due to lysines 86 and 87 and other nearby lysines (4, 5, 11, 89) is stronger than that due to lysines 72 and 73. In the case of human Cytc the positive potential around site A is less uniform and likely weakens electrostatic binding to CL membranes through site A.

Graphical Abstract:

Yeast iso-1-cytochrome c binds more strongly to cardiolipin nanodiscs than human cytochrome c at physiological ionic strength because the charge distribution near site A (lysines 72, 73, 86, 87) is more uniformly positive. Mutagenesis studies on site A lysines suggest that N- and C-terminal helix lysines contribute to electrostatic binding.

1. Introduction

Cytochrome c (Cytc) is a protein best known for its role in the electron transport chain [1–3]. However, at the onset of apoptosis, it interacts with cardiolipin (CL) [4–6], a lipid making up about 25% of the inner mitochondrial membrane [7]. It is thought that one of the important modes of interaction between Cytc and CL is mediated by a conserved binding site on Cytc known as Site A [8–10]. The interaction between Cytc and CL induces Cytc to act as a peroxidase which, in the presence of reactive oxygen species, leads to oxygenation of CL at the onset of apoptosis [11,12]. Ultimately, oxygenation of CL facilitates pore formation in the outer mitochondrial membrane leading to release of Cytc into the cytoplasm [11–14]. The fundamental importance of this interaction to apoptosis has motivated significant interest and study [8–10,15–34]. Because of the difference in peroxidase activity of yeast iso-1-Cytc and human Cytc [35,36], the interaction of iso-1-Cytc with cardiolipin has also generated interest [16,17,25]. Electrostatic binding of Cytc to CL liposomes through site A (lysines 72, 73 86 and 87) [9,10] has been well studied in yeast, horse and human Cytc [16,17,24,26]. While studies on horse Cytc indicate a prime role for Lys72 [26], variants of yeast iso-1-Cytc that replace site A lysines with alanine either individually or pairwise (K72A/K73A or K86A/K87A) showed relatively modest effects on iso-1-Cytc binding to CL liposomes [16,17]. In fact, yeast iso-1-Cytc with all four site A lysines removed could still bind to CL liposomes [17]. These results indicate significant species-dependent effects on the binding of Cytc to CL-containing membrane bilayers.

In healthy living cells, Cytc is primarily associated with the inner mitochondrial membrane cristae structures [37,38]. The ionic strength of the intermembrane space is known to be 100 to 150 mM [39]. The concentration of Cytc in mitochondria has been estimated to be 0.5 to 5 mM [40]. However, other reports indicate values of 100 μM to 700 μM depending on the state of the mitochondria [41]. Direct measurement of the concentration of Cytc in respiring yeast mitochondria yields a value of 100 μM [42]. However, Cytc is primarily confined to the cristae, so this latter value represents a lower limit.

There has been increasing interest in studying the interaction of Cytc with CL under physiologically relevant conditions [43]. While some binding studies have been carried out at higher ionic strength [20,21,29–32], many studies done on binding of mammalian Cytc and yeast iso-1-Cytc have been done under low salt conditions [16–19,24–26] and most all studies use Cytc concentrations of 10 μM or less. For horse and bovine Cytc, binding to CL liposomes persists in the presence of 50 to 300 mM NaCl concentration [20,21,29–32,37]. However, increasing NaCl concentration weakens binding and appears to switch horse Cytc from an extended to a compact conformer on the surface of CL liposomes [20,21,29–32]. On the other hand, addition of 150 mM NaCl completely dissociated human Cytc from CL nanodiscs at pH 8, although only partially at pH 7.4 [19]. To gain more insight into the effects of Cytc concentration and ionic strength, as well as to assess species differences in Cytc binding to CL-containing membrane mimics through site A, we have used size exclusion chromatography (SEC) [37] to evaluate the binding of wild type (WT) yeast iso-1-Cytc, WT human Cytc, and site A variants of yeast iso-1-Cytc to CL nanodiscs at 100 mM ionic strength and at somewhat higher Cytc concentration (25 μM). We find that yeast iso-1-Cytc binds more strongly to CL nanodiscs than human Cytc and that lysines 86 and 87 are more important components of site A binding than lysines 72 and 73 at this ionic strength.

2. Experimental section

2.1. Expression of membrane scaffold protein 1D1

Growth of membrane scaffold protein 1D1, MSP1D1, was modified from a procedure described previously [44]. DNA from a pET28a vector carrying the His-tagged MSP1D1 gene was transformed into competent Escherichia coli BL21 CodonPlus cells (New England BioLabs) and grown on LB agar plates carrying 50 μg/mL kanamycin. The transformed cells were then suspended in 3 mL of sterile LB media (10 g/L tryptone, 5 g/L yeast extract, 6 g/L NaCl) and used to inoculate 2.5 L of Terrific Broth media (24 g/L yeast extract, 20 g/L tryptone, 4 mL/L glycerol, 0.017 M potassium phosphate monobasic, 0.072 M potassium phosphate dibasic). Kanamycin was added to a final concentration of 50 mg/L and a few drops of Antifoam C Emulsion (Sigma Aldrich) was also added. Growth was performed in a BioFlo IIc fermenter (New Brunswick Scientific) at 37 °C with stirring at 500 rpm and air flow at 3.0-4.0 standard liters per minute. The OD at 600 nm was monitored until it reached 0.9-1.0, and then expression was induced with IPTG at a final concentration of 1 mM. After 3 hours of expression, the cells were spun down at 5,000 rpm for 15 minutes (ThermoFisher Lynx 6000, F12 rotor). Cells were lysed in 10 mM Tris, pH 7.5, 500 mM NaCl, 2 mM PMSF, and small amounts of DNase and RNase by passing 3 times through an Avestin C5 Cell Homogenizer. Lysate was spun down at 10,000 rpm for 30 minutes (ThermoFisher Lynx 6000, F14 rotor) and then passed over a Ni²⁺ charged HisTrap FF crude column (Cytiva). Once the sample was loaded onto the column, it was washed with 2 to 3 column volumes each of the following buffers, in order: 40 mM Tris, 300 mM NaCl, 1% Triton-X, pH 8.0; 40 mM Tris, 300 mM NaCl, 50 mM sodium cholate, pH 8.0; 40 mM Tris, 300 mM NaCl, 20 mM imidazole, pH 8.0. After washing, the His-tagged MSP1D1 was eluted with 40 mM Tris, 300 mM NaCl, 400 mM imidazole, pH 8.0 and dialyzed overnight against 50 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 1 mM DTT. Dialyzed MSP1D1 was then condensed down to 1.0-1.5 mL by ultrafiltration. The concentration of MSP1D1 was determined by absorbance at 280 nm (extinction coefficient: 21,430 M⁻¹ cm⁻¹) [44]. TEV protease was added at a ratio of 1:200 TEV:MSP1D1 and left overnight in a 30 °C water bath to remove the His-tag. The solution was then passed over the same HisTrap FF crude column and the flow-through containing MSP1D1 without the His-tag, MSP1D1(−), was collected and exchanged into 20 mM NaH₂PO₄, 25 mM Na₂HPO₄ using Amicon Ultra-15 ultracentrifugation units with a molecular weight cut-off (MWCO) of 10,000 Da (MilliporeSigma) using a Sorvall Legend X1R Centrifuge (4,000 rpm). The MSP1D1(−) was then separated into 250 μL aliquots, flash frozen in liquid nitrogen and stored at −80 °C until needed for nanodisc formation.

2.2. Formation of 100% CL nanodiscs

CL nanodiscs were formed based on a modification of a published procedure [44]. Aliquots of MSP1D1(−) stored at −80 °C were thawed and, just prior to nanodisc formation, further purified by a single round of size exclusion chromatography (SEC) using a Superdex 10/300 200 Increase size exclusion column (GE Healthcare) attached to an Agilent 1200 series HPLC. 1,1 ’2,2’-Tetraoleoyl cardiolipin (sodium salt) dissolved in chloroform (Avanti Polar Lipids) was dried in 1 mg quantities into a thin film in a 15 mL round-bottom culture tube under a stream of argon gas. Then, 200 μL of solubilization buffer (50 mM Tris, 300 mM NaCl, 48 mM sodium cholate, pH 7.5) was added to the thin film followed by sonication for 30 minutes at 30 °C to solubilize the lipid and form vesicles. An appropriate amount of the “belt” protein, MSP1D1(−), was added to the solution such that there was 1 MSP1D1(−) for every 25 cardiolipin molecules (or 0.58 mg MSP1D1(−) per 1 mg cardiolipin). The mixture was then refrigerated at 4 °C for 30 minutes and subsequently subjected to two one-hour rounds of detergent removal via SM-2 biobeads from BioRad. The resulting solution was then separated from the biobeads by centrifugation and loaded on a Superdex 10/300 200 Increase size exclusion column (GE Healthcare) in 50 mM Tris, 300 mM NaCl, pH 7.5 to purify the nanodiscs. SEC was then repeated for an additional 1-3 rounds until sufficient nanodisc purity was achieved (Fig. S1). Purified nanodiscs were flash frozen in liquid nitrogen and stored at −80 °C. Nanodiscs were thawed and exchanged into the desired buffer using Amicon Ultra-15 ultracentrifugation units (10,000 Da MWCO; MilliporeSigma) in a Sorvall Legend X1R Centrifuge running at 4,000 rpm and then used without further purification.

2.3. Purification of cytochrome c variants

Previously-expressed and purified WT Human Cytc [16], wild type yeast iso-1-Cytc (yWT) [16] and the following variants [17]: K72A/K73A iso-1- Cytc (K72|73A), K86A/K87A iso-1-Cytc (K86|87A), and K72A/K73A/K86A/K87A iso-1- Cytc (K72|73|86|87A) were thawed from storage at −80 °C and purified using an AKTA PrimePlus FPLC and a HiTrap HP SP 5 mL cation exchange column (GE Healthcare Life Sciences), as described previously [36]. The yeast variants were all expressed from E. coli, so the lysine at position 72 is not trimethylated as it is for protein expressed from yeast [45].

Prior to SEC experiments, freshly purified Cytc was oxidized with potassium ferricyanide at room temperature for 20-30 minutes. Oxidized Cytc was separated from the potassium ferricyanide using a G25 size exclusion column (GE Healthcare) with 25 mM HEPES, pH 7.4, 75 mM NaCl as the running buffer. UV-Vis spectroscopy was used to confirm full oxidation of Cytc and to determine Cytc concentration [46]. Oxidized Cytc was then diluted to 50 μM with 25 mM HEPES, pH 7.4, 75 mM NaCl.

2.4. Cation exchange chromatography of oxidized cytochrome c

WT human Cytc and yWT and the K72|73A and K86|87A variants of iso-1-Cytc stored at −80 °C were thawed and oxidized with ferricyanide. Each protein was then purified using an AKTA PrimePlus FPLC and a HiTrap HP SP 5 mL cation exchange column at a flow rate of 1.0 mL/min. Buffer A was 50 mM sodium phosphate pH 7.0 and Buffer B was 50 mM sodium phosphate, pH 7.0, 1.0 M NaCl. The gradient program used was 8 mL at 0 %B, followed by a 60 mL gradient to 60 %B, the %B was then raised to 100% during 1 mL of buffer flow and held at 100 %B for 10 mL of buffer flow, then returned to 0 %B over 1 mL of buffer flow, followed by 10 mL of buffer flow at 0 %B. The oxidized Cytc peak was collected, followed by concentration and exchange into Buffer A using ultrafiltration (Millipore Ultra15 centrifuge ultrafiltration device, with MWCO of 10,000 Da). The concentration of the fully oxidized Cytc was evaluated using UV-Vis spectroscopy [46]. Using the measured concentration, 1.2 mg of the purified oxidized Cytc was then re-run over the HiTrap HP SP 5 mL cation exchange column using the same gradient program. Experiments were done at 4.0 °C. Chromatograms were exported as .csv files and an overlay plot of the WT human Cytc and yWT and the K72|73A and K86|87A variants of iso-1-Cytc was produced using SigmaPlot 13. Data from the conductivity detector of the AKTA PrimePlus FPLC demonstrated that the NaCl gradients of each chromatography experiment were highly consistent with each other. A mass spectrum was obtained for each purified protein at the Mass Spectrometry Core Facility of the University of Montana using an Agilent 6520 Accurate Mass Q-TOF LC/MS interfaced to an Agilent 1260 UPLC. Observed molecular weights were within 1 amu of the expected mass for each protein.

2.5. Size exclusion chromatography detection of cytochrome c binding to CL nanodiscs

All SEC was performed using a Superdex 10/300 200 Increase size exclusion column (GE Healthcare) at a flow rate of 0.75 mL/min. The column was run on an Agilent 1200 series HPLC with detection at 220 nm, 280 nm and 410 nm. The column was equilibrated to 25 mM HEPES, 75 mM NaCl, pH 7.4, which was also used as the running buffer. An injection volume of 40 to 50 μL was used for each binding experiment. Cardiolipin nanodisc and Cytc variant concentrations were set at 25 μM each. Stock solutions of 50 μM of each were mixed 1:1 and the sample was incubated at room temperature for 15 to 20 minutes to allow for binding [22].

2.5. Electrostatic surface calculations

Electrostatic surface potential was calculated using the plugin for the Adaptive Poisson– Boltzmann Solver (APBS) [47] in PyMol v. 2.52 [48]. For all calculations, temperature was set to 298.15 K, the ionic strength was set to 0.1 M and a grid spacing of 0.25 Å was used. Otherwise, the default parameters in the ABPS plugin were used. For human Cytc, molecule A of the PDB file 3ZCF was used. For yeast iso-1-Cytc, a modified form of the PDB file 2YCC that converted trimethyllysine 72 to a lysine was used. This PDB file was further modified in PyMol using the Mutagenesis wizard to convert lysines 72 and 73 to alanine (K72|73A variant) or lysines 86 and 87 to alanine (K86|87A variant) prior to APBS calculations. All surfaces shown are Connolly solvent-excluded surfaces shown as heat maps in units of kT/e.

3. Results

3.1. Comparison of binding of yeast iso-1-Cytc and human Cytc to CL nanodiscs

In previous work, we used fluorescence correlation spectroscopy (FCS) to measure the binding of Zn-substituted human Cytc to CL nanodiscs [19]. We initially planned to apply the same method to measure the binding of WT yeast iso-1-Cytc (yWT) and site A variants of iso-1-Cytc to CL nanodiscs. Although Zn-substituted iso-1-Cytc has been prepared and its low temperature spectra reported [49], we found that Zn-substituted yWT Cytc was unstable and tended to aggregate under the standard conditions used to make FCS measurements [50].

Previous studies have shown that binding of bovine heart Cytc to CL nanodiscs produced with apolipophorin III (apoLp-III) [51] can be monitored by size exclusion chromatography (SEC) in 20 mM HEPES, pH 7.4, 75 mM NaCl [37]. In our work using FCS [19], we showed that human Cytc binds strongly to CL nanodiscs made from MSP1D1(−) at low ionic strength (25 mM TES pH 7.4 or 8), but does not bind to nanodiscs prepared with 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and MSP1D1(−). These results are consistent with data for binding bovine heart Cytc to CL versus DMPC nanodiscs prepared with apoLp-III [37]. Because the basic characteristics of Cytc binding to these two forms of nanodiscs are similar and previous work from our lab was done using CL nanodiscs prepared with the MSP1D1(−) belt protein, we produced CL nanodiscs using MSP1D1(−) (Fig. S1) for SEC monitored binding studies on human and yeast Cytc. The CL nanodiscs elute from the SEC column as a uniform peak at a retention time near 14 min (Fig. 1A). The CL nanodiscs absorb strongly at 280 nm, but do not absorb near 410 nm, where the heme Soret band of Cytc absorbs strongly. Thus, association of Cytc (12 – 13 kDa) with the high molecular weight CL nanodiscs (~146 kDa for MSP1D1(−) as belt protein) can be detected without interference from nanodisc absorbance.

Fig. 1.

(A) Size exclusion chromatography (SEC) chromatogram of CL nanodiscs after loading 50 μL of a 25 μM solution of CL nanodiscs prepared with MSP1D1(−) in 25 mM HEPES pH 7.4, 75 mM NaCl. (B) SEC chromatogram after loading 50 μL of a solution containing 25 μM CL nanodiscs and either 25 μM WT Human Cytc or 25 μM yWT Cytc. All samples were incubated at room temperature (22 ± 1 °C) for 15 – 20 min. The traces at 280 nm are offset from the traces at 410 nm for clarity of presentation. The running buffer for the Superdex 10/300 200 Increase column was 25 mM HEPES pH 7.4, 75 mM NaCl at a flow rate of 0.75 mL/min.

We evaluated the binding of WT Human Cytc to CL nanodiscs at 100 mM ionic strength (25 mM HEPES, pH 7.4, 75 mM NaCl) using SEC chromatography following an incubation time of 15 to 20 minutes at room temperature. The buffer conditions were chosen to be similar to the previous study of binding of bovine Cytc to CL nanodiscs [37] using SEC and to approximate the pH in the cristae of mitochondria (pH 6.9 to 7.3) [52]. The chromatography peak corresponding to the CL nanodisc at 14.1 minutes absorbs strongly at 280 nm (dashed blue trace, Fig. 1B), however the absorbance at 410 nm (dashed red trace, Fig. 1B) is modest at 14.1 minutes, indicating little binding of WT human Cytc to the CL nanodiscs under these conditions. The large A₄₁₀ peak in the chromatogram has a retention time of 25.0 minutes corresponding to unbound Cytc.

We also measured the binding of yWT Cytc to CL nanodiscs under the same conditions. In the A410 trace for yWT Cytc (solid red line, Fig. 1B) the largest peak occurs at 13.9 minutes, indicating that yWT Cytc is co-eluting with the CL nanodiscs. There is no A₄₁₀ peak near 25 minutes (solid red line, Fig. 1B) indicating that all yWT is bound to the CL nanodiscs at the time of loading. However, a broad peak with a maximum at 17.4 minutes follows the peak at 13.9 minutes indicating that yWT dissociates from the CL nanodiscs during SEC chromatography. The A₄₁₀/A₂₈₀ ratio at 13.9 minutes is 1.22, which is less than the ratio of about 1.77 that would be expected for a 1:1 ratio of yWT Cytc to CL nanodiscs made using MSP1D1(−), consistent with some loss of Cytc from the CL nanodiscs during chromatography. At the center of the broad peak at 17.4 minutes, the A₄₁₀/A₂₈₀ ratio rises to 3.1. This ratio is lower than expected for pure oxidized Cytc (~4.6) [46], which is likely due to the bandwidth setting at 410 nm (16 nm) used to improve signal-to-noise for the low protein concentrations used here.

3.2. Role of site A lysines in binding of iso-1-Cytc to CL nanodiscs at 100 mM ionic strength

Having shown that yWT Cytc binds to CL nanodiscs at 100 mM ionic strength, we carried out experiments to determine the relative importance of lysines 72, 73, 86 and 87 from site A [8–10] in the binding of yeast iso-1-Cytc to CL nanodiscs under these conditions.

Fig. 2A shows the SEC trace of the K72|73A variant of iso-1-Cytc compared to yWT after binding to CL nanodiscs for 15 – 20 minutes at room temperature. The behavior of the K72|73A variant is similar to yWT. There is no evidence of free iso-1-Cytc at a retention time near 24 minutes in either case. The only significant peak in the SEC trace at 410 nm for the K72|73A variant is at a retention time of 13.75 minutes coinciding with the peak in the A₂₈₀ trace for the CL nanodiscs. However, unlike with yWT, it appears that there is minimal dissociation of the K72|73A variant during the time period of the SEC experiment. The A₄₁₀/A₂₈₀ ratio at 13.75 minutes is 1.76, very close to the theoretical value of 1.77 for a 1:1 ratio of iso-1-Cytc to CL nanodiscs made using MSP1D1(−).

Fig. 2.

SEC chromatogram after loading 50 μL of a solution containing 25 μM CL nanodiscs and (A) 25 μM of yWT or K72|73A iso-1-Cytc and (B) 25 μM K86|87A or K72|73|86|87A iso-1-Cytc. Traces are representative trials of each variant. All samples were incubated at room temperature (22 ± 1 °C) for 15 – 20 min before loading onto the column. The traces at 280 nm are offset from the traces at 410 nm for clarity of presentation. The running buffer for the Superdex 10/300 200 Increase column was 25 mM HEPES pH 7.4, 75 mM NaCl at a flow rate of 0.75 mL/min.

SEC chromatograms for the K86|87A and K72|73|86|87A show no peaks in the A₄₁₀ trace (Fig. 2B, red solid and dashed traces) at the retention time of the CL nanodiscs at 14.0 to 14.1 minutes as observed in the A₂₈₀ trace (Fig. 2B, blue solid and dashed traces). Strong peaks are observed in the A₄₁₀ trace only at a retention time of 24.0 minutes for the K86|87A variant and at 23.6 minutes for the K72|73|86|87A variant, consistent with unbound iso-1-Cytc. Thus, removal of lysines 86 and 87 is sufficient to eliminate binding of yeast iso-1-Cytc to CL nanodiscs at 100 mM ionic strength.

3.3. Cation-exchange chromatography of WT and variant cytochromes c

As a second approach to comparing the ionic strength dependence of the binding of Cytc to a negative charge surface, we compare the elution times of Cytc from a cation exchange resin using the same salt gradient. Fig. 3 compares the elution times for WT human Cytc, yWT iso-1-Cytc and the K72|73A and K86|87A variants of iso-1-Cytc. The time of the elution peak of the oxidized WT human protein occurs when %B in the gradient is 30.1% (~0.30 M NaCl) much earlier in the gradient than the elution peak for yWT iso-1-Cytc (%B = 51.8%, ~ 0.52 M NaCl), consistent with the weaker binding of human Cytc to CL nanodiscs (Fig. 1B) in the presence of NaCl. In Fig. 3, the elution time for the K86|87A variant (%B = 34.7%, ~ 0.35 M NaCl) is just after that of the WT human Cytc consistent with the lack of significant binding to CL nanodiscs (Fig. 2B), as observed for WT human Cytc. The K72|73A variant, elutes at a %B of 44.6% (~0.45 M NaCl), just slightly earlier than yWT iso-1-Cytc. This is consistent with the observed binding of the K72|73A variant to CL nanodiscs at an ionic strength of 0.1 M. These data indicate that the interaction of the yeast protein with a negative charge surface is more resistant to ionic strength than the human protein. Similarly, for the yeast protein lysines 86 and 87 confer a greater resistance to ionic strength than lysines 72 and 73 for binding to the negatively-charged sulphopropyl groups of the cation exchange resin.

Fig. 3.

Overlay of ion exchange chromatography profiles monitored at 280 nm of WT human Cytc, yWT iso-1-Cytc and the K86|87A and K72|73A variants of yeast iso-1-Cytc. The %B is the percent of Buffer B, which is 50 mM sodium phosphate pH 7, 1.0 M NaCl, in the gradient. A 5 mL HiTrap SP cation exchange column was used. Chromatography was carried out at 4 °C. The peak from each chromatogram was collected and its UV-Vis spectrum measured to show that the Cytc was in the oxidized state (Fig. S2).

4. Discussion

4.1. Yeast iso-1-Cytc versus human Cytc binding to CL nanodiscs

In our previous work comparing binding of yeast iso-1-Cytc and human Cytc to cardiolipin vesicles at low ionic strength (20 mM TES pH 8) [16], two binding phases were observed. Cooperativity parameters obtained for the first phase of binding monitored by heme Soret circular dichroism (CD) indicated that yeast iso-1-Cytc accommodates roughly two lipids in its binding site, while human cytochrome c only has one lipid bound at Site A. However, the second phase of binding, observed with Trp59 fluorescence intensity at higher lipid-to-protein ratios (LPR), indicated that two lipids bind to both human and yeast Cytc. An increase in the Trp59 fluorescence of Cytc is generally associated with disruption of the tertiary structure of Cytc [53], consistent with partial unfolding on the surface of the vesicle. Finally, the fluorescence-detected phase of binding to 100% cardiolipin vesicles, occurs at a higher LPR for human Cytc (LPR of 36.0) versus yWT iso-1-Cytc (LPR of 23.4) consistent with the weaker binding we see for human Cytc by SEC at higher salt concentration (Fig. 1).

Yeast iso-1-Cytc is also known to unfold more cooperatively than human Cytc [36,54], and is less stable than human Cytc [55]. In our previous study of the binding of yeast versus human Cytc to CL vesicles [16], we attributed the weaker binding (higher LPR for binding monitored by Trp59 fluorescence) of the human protein to the higher stability of the human Cytc. This factor may play a role, but we also evaluated the electrostatic surface potential of site A using APBS (Fig. 4) [47]. It is evident that site A for yWT iso-1-Cytc provides a larger and more uniformly electropositive surface to interact with a CL nanodisc than human Cytc despite the fact that the pI values calculated based on sequence are essentially identical (pI: yeast, 9.54; human, 9.59; calculated using the EXPASY ProtParam tool [56]). The T69E and N62D substitutions in 60s helix of human Cytc relative to yWT Cytc (Fig. S3) weaken the electropositive potential in between the K72/K73 and K86/K87 lysine pairs of site A on the surface of the human protein. Thus, it is likely that the more electropositive binding surface at site A for yWT iso-1-Cytc contributes to the stronger binding of the yeast protein to CL nanodiscs compared to the human protein at 0.1 M ionic strength. This behavior is mirrored in the lower salt concentration needed to elute human Cytc from a cation exchange column relative to that needed to elute yWT iso-1-Cytc (Fig. 3).

Fig. 4.

Electrostatic potential surface calculated for WT human Cytc and WT Yeast iso-1-Cytc using the APBS plugin in PyMol. The scale bar is shown at the bottom of the figure in units of kT/e. The calculations were carried out as described in the Experimental section at 25 °C and an ionic strength of 0.1 M. APBS does not recognize heme, so the exposed heme surface, near the top of each protein, is represented as a surface colored by element (C, white; O, red, N, blue).

Previous studies of the NaCl concentration dependence of Cytc binding to CL containing liposomes at pH 7.4 indicate that the initial binding step (intrinsic binding, K₀) is independent of salt concentration for 20% and 50% CL liposomes, whereas the partial unfolding at higher LPR that leads to a more extended conformer does depend on NaCl concentration [31,33]. These studies indicate that higher salt concentration shifts the equilibrium from the extended conformer back toward the more compact conformer. Notably, for 100% CL liposomes, K₀ is also decreased significantly in the presence of salt [31,33]. Since partial unfolding increases the apparent binding constant, increased salt concentration weakens overall binding, and because of the effect on K₀ at 100% CL [31,33], we would expect the affinity of Cytc binding to 100% CL nanodiscs in both the compact and extended conformers would be decreased. The effect of NaCl on K₀ for 100% CL liposomes was suggested to be due to a uniform Stern layer of Na⁺ ions that would form on the consistently negative surface of 100% CL liposomes. However, treatment of the 100% CL liposome binding data with theory that accounts for formation of a Stern layer produced physically unreasonable results (K₀ increased with increasing salt concentration) [31,33]. Therefore, we focus here on the possible effects of NaCl on the equilibrium between compact and extended conformers of Cytc on the surface of CL nanodiscs.

Based on fluorescence resonance energy transfer (FRET) experiments with horse Cytc site-specifically labeled with a fluorophore, partial unfolding involves dissociation of the C-terminal helix from the N-terminal helix induced by binding of the C-terminal helix to the negatively-charged CL surface of the membrane. The more uniformly electropositive surface of the N- and C-terminal helices of yWT iso-1-Cytc compared to human Cytc (see Fig. S6) would strengthen the interaction of the C-terminal helix with the CL headgroups on the surface of the nanodiscs, increasing the population of the extended conformer, even in the presence of NaCl. The higher population of the extended conformer would allow the apparent binding constant of yWT iso-1-Cytc to remain sufficiently high that binding remains essentially complete in 25 mM HEPES, pH 7.4, 75 mM NaCl, whereas that of human Cytc does not (Fig. 1B).

4.2. Comparison of binding of bovine Cytc and human Cytc to nanodiscs

Fox et al. [37] showed that bovine Cytc was fully associated with CL nanodiscs prepared with apoLp-III under similar buffer conditions (20 mM HEPES, pH 7.4, 75 mM NaCl). An alignment [57] of the sequences of bovine and human Cytc shows ten substitutions (Fig. S4). None involve Lys or Arg residues. There is a repositioning of a Glu from position 89 in human to position 92 in bovine Cytc and a conversion of an Ala at position 50 in human Cytc to an Asp in bovine Cytc. The effect on the surface electrostatic potential around site A, as calculated by APBS is modest (Fig. S5). Therefore, surface electrostatics do not appear to be the primary difference in the apparently better affinity of bovine Cytc for CL nanodiscs. The stability of human Cytc is 9.46 kcal/mol [58], compared to 8.38 kcal/mol for bovine Cytc [59]. The lower stability of bovine Cytc, which may facilitate partial unfolding of the protein on the surface of a CL nanodisc, could provide an explanation for the apparently higher affinity of bovine Cytc for CL nanodiscs compared to human Cytc at an ionic strength near 0.1 M.

There are differences between the two CL nanodisc systems that may also be factors in the apparently higher affinity of bovine Cytc for CL nanodiscs. The CL nanodiscs used to study binding of bovine Cytc were made with apoLp-III, which forms nanodiscs with a diameter of 18.5 Å [51] compared to the 9.6 Å diameter of the MSP1D1(−) nanodiscs [44]. The surface area of MSP1D1(−) nanodiscs is about 4400 Ų [60], whereas that of the apoLp-III nanodisc should be about 21,400 Ų based on its diameter. We prepared nanodiscs using tetraoleoyl cardiolipin (TOCL), whereas Fox et al. [37] used tetramyristoyl cardiolipin (TMCL). The headgroup surface area of TOCL is 129.8 Ų [61], whereas that of TMCL is 83 Ų [62], because TMCL is in a gel phase at room temperature whereas TOCL is in a liquid crystalline phase [63]. Thus, the negative charge density of the apoLp-III CL nanodiscs will be ~1.6-fold higher than that of the MSP1D1(−) CL nanodiscs. As a result, the higher affinity observed for bovine Cytc for the apoLp-III CL nanodiscs may be caused by the higher negative charge density expected for these nanodiscs compared to the MSP1D1(−) CL nanodiscs. Other factors that could affect observed binding to the CL nanodiscs are similar. The Cytc concentration in the samples used for SEC chromatography is 25 μM in this study compared to about 20 μM in Fox el al. [37]. Similarly, the LPR in our samples was 68:1 compared to 76:1 in the study of Fox et al. [37].

In our case, each nanodisc surface will have about 34 CL headgroups whereas each surface of the apoLp-III CL nanodiscs will hold about 258 TMCL headgroups. Given the K_d in units of LPR of 36 observed for the second phase of binding human Cytc to TOCL vesicles [16], the surface of the MSP1D1(−) nanodisc will be just large enough to bind the partially unfolded form of human Cytc. Therefore, we cannot rule out the possibility that the size of the MSP1D1(−) nanodisc could affect binding affinity for human Cytc. However, our previous measurements of the K_d of the human Cytc/MSP1D1(−) CL nanodisc interaction by FCS produced robust binding curves, indicating that the size of the Cytc binding site on these nanodiscs is sufficient to accommodate human Cytc [19].

4.3. Constituents of the site A binding site of Cytc

The lysine pairs, K72/K73, and K86/K87, which are canonically considered to mediate site A binding [8–10], produce dramatically different effects on binding to CL nanodiscs (Fig. 2), when each pair is mutated to alanines. Since each variant resulted in the removal of two of the charged lysines that control electrostatic binding, it might be expected that they would have similar binding profiles. Thus, the arrangement of surface charges appears to be important, not just the total number at higher ionic strength.

Previous results as monitored by CD and Trp fluorescence indicated that the binding of both of these variants to cardiolipin vesicles is maintained in a low salt buffer (20 mM HEPES, 0.1 mM EDTA, pH 8) [17]. However, the low ionic strength binding titrations showed interesting differences for the K72|73A variant compared to the K86|87A variant that provide insight into the relative resistance of these variants to the effect of NaCl on their binding to CL nanodiscs. In particular, the increase of Trp59 fluorescence upon binding to 100% CL vesicles is larger for the K86|87A variant and smaller for the K72|73A variant compared to yWT iso-1-Cytc. Similarly, the red shift in the Trp59 fluorescence maximum upon binding to CL nanodiscs was more for the K86|87A variant and less for the K72|73A variant than for yWT. These data indicate, respectively, that relative to yWT iso-1-Cytc, the Trp59 is more distant from the heme and is more solvent exposed [64] for the K86|87A variant (i.e. more unfolded) and that the opposite is true for the K72|73A variant. The apparent number of lipids that bind to iso-1-Cytc in this partial unfolding step increased for the K86|87A variant (n = 2.88 ± 0.15) and decreased for the K72|73A variant (n = 1.77 ± 0.13) relative to WT iso-1-Cytc (n = 2.3 ± 0.2) [17]. Thus, the energetic cost of unfolding is likely higher for the K86|87A variant and the number of Na⁺ displaced is likely higher compared to yWT iso-1-Cytc, with the opposite being true for the K72|73A variant. The fact that yWT iso-1-Cytc is intermediate in these properties relative to the K72|73A and K86|87 variants may explain why WT iso-1-Cytc partially dissociates from the CL nanodiscs and the K72|73A variant does not (Fig. 2A) during SEC chromatography.

The difference in surface electrostatics caused by these mutations provides additional support for the effect of NaCl on the partial unfolding of yeast iso-1-Cytc when bound to CL nanodiscs. In Fig. 5 (right), the K72|73A variant is rotated to show the electrostatic surface on the other side of the K86/K87 pair. A large positively-charged patch on the protein surface consisting of K4, K5, K11 on the N-terminal helix and K99 and K100 on the C-terminal helix is evident. These additional lysines, when combined with lysines 86, 87 and 89, appear to provide an alternate binding interface that replaces site A in the absence of lysines 72 and 73 and allows strong binding at 0.1 M ionic strength. This focused electrostatic interface may lead to a lesser degree of unfolding that is energetically more favorable and requires fewer CL and thus less displacement of Na⁺ from CL headgroups at higher ionic strength. This alternate electrostatic binding interface is much less uniformly electropositive for human Cytc (Fig. S6), which may be another reason for the weaker binding of human Cytc to CL nanodiscs at 0.1 M NaCl. A E89G substitution for bovine Cytc relative to human Cytc (Fig. S5), somewhat improves the uniformly electropositive nature of this alternate electrostatic binding interface for bovine Cytc. Although as discussed above, there may be other reasons for the stronger binding of bovine Cytc to apoLp- III CL nanodiscs at 0.1 M ionic strength [37] than that of human Cytc to MSP1D1(−) CL nanodiscs observed here.

Fig. 5.

Electrostatic potential surface calculated for the K86|87A and K72|73A variants using the APBS plugin in PyMol. The K72|72A variant is rotated to show the electrostatic potential surface due to lysines 4, 5, 11, 99 and 100 on the N- and C-terminal helices adjacent to lysines 86, 87 and 89. The scale bar is shown at the bottom of the figure in units of kT/e. The calculations were carried out as described in the Experimental section at 25 °C and an ionic strength of 0.1 M. APBS does not recognize heme, so the exposed heme surface, near the top of each protein, is represented as a surface colored by element (C, white; O, red, N, blue).

The loss of the K86/K87 pair in the K86|87A variant significantly decreases the electrostatic potential of site A (Fig. 5, left) and places a larger distance between the positive patch around K72/K73 and the one near the N- and C-terminal helices compared to yWT iso-1-Cytc (Figure 4, right). This change in the electrostatic surface would require a larger degree of partial unfolding and binding of more CL compared to yWT iso-1-Cytc. Thus, at higher ionic strength this would require displacement of more Na⁺ ions from the surface of the CL nanodiscs when iso-1-Cytc partially unfolds.

5. Conclusion

Previous work has suggested that electrostatic binding involves more residues than the four lysines commonly referred to as site A and that other positively charged residues may also contribute to binding [17,65]. Our current results provide further support for site A being a larger electrostatic binding surface than previously supposed, particularly at the ionic strength present in mitochondria. Our results also suggest that the nature of this binding interface has evolved from yeast to mammals, perhaps to optimize this interaction for the role of Cytc in apoptosis. We note that there is some overlap between site A and the Cytc binding site for apoptosis protease activation factor 1 (Apaf-1). In particular, Lys72 is critical for Apaf-1 binding. Trimethylation of Lys72, as in yeast iso-1-Cytc, is sufficient to eliminate binding and caspase activation [66,67]. Residues in the 60s helix and the N-terminal helix differ between yeast and human Cytc and are likely important in enhancing the affinity of yeast iso-1-Cytc relative to human Cytc for CL-containing membranes. However, these same residues strongly reduce Apaf-1 activation by yeast iso-1-Cytc relative to horse Cytc [67]. Notably, Saccharomyces cerevisiae lacks Apaf-1 [68]. With the advent of Apaf-1, there appears to have been co-evolution of the surface of Cytc, to properly balance CL binding and Apaf-1 binding for the needs of apoptosis. With regard to CL binding, future experiments at physiological salt concentrations are necessary to elucidate exactly which positively charged residues outside of the canonical site A are involved in electrostatic binding of Cytc to CL.

Highlights.

• Human cytochrome c does not bind to cardiolipin nanodiscs at 100 mM ionic strength

• Yeast iso-1-cytochrome c binds to cardiolipin nanodiscs at 100 mM ionic strength

• The electrostatic surface potential is stronger for yeast versus human cytochrome c

• Site A lysines 86 and 87 are most important for binding at 100 mM ionic strength

Abbreviations

APBS

Adaptive Poisson–Boltzmann Solver

apoLp-III

apolipophorin III

CD

circular dichroism

CL

cardiolipin

Cytc

cytochrome c

FCS

fluorescence correlation spectroscopy

LPR

lipid-to-protein ratio

MSP1D1

membrane scaffold protein 1D1

MSP1D1(−)

MSP1D1 without the His-tag

MWCO

Molecular weight cut-off

TMCL

tetramyristoyl cardiolipin

TOCL

tetraoleoyl cardiolipin

yWT

yeast iso-1-Cytc expressed from E. coli containing Lys72 that is not trimethylated