Unraveling cardiolipin-induced conformational change of cytochrome c through H/D exchange mass spectrometry and quartz crystal microbalance

Cardiolipin (CL), a crucial component in inner mitochondrial membranes, interacts with cytochrome c (cyt c) to form a peroxidase complex for the catalysis of CL oxidation. Such interaction is pivotal to the mitochondrial regulation of apoptosis and is affected by the redox state of cyt c. In the present study, the redox-dependent interaction of cyt c with CL was investigated through amide hydrogen/deuterium exchange coupled with mass spectrometry (HDXMS) and quartz crystal microbalance with dissipation monitoring (QCM-D). Ferrous cyt c exhibited a more compact conformation compared with its ferric form, which was supported by the lower number of deuterons accumulated and the greater amplitude reduction on dissipation. Upon association with CL, ferrous cyt c resulted in a moderate increase in deuteration, whereas the ferric form caused a drastic increase of deuteration, which indicated that CL-bound ferric cyt c formed an extended conformation. These results were consistent with those of the frequency (f) − dissipation (D) experiments, which revealed that ferric cyt c yielded greater values of |ΔD/Δf| within the first minute. Further fragmentation analysis based on HDXMS indicated that the effect of CL binding was considerably different on ferric and ferrous cyt c in the C-helix and the Loop 9–24. In ferric cyt c, CL binding affected Met80 and destabilized His18 interaction with heme, which was not observed with ferrous cyt c. An interaction model was proposed to explain the aforementioned results.

Global hydrogen/deuterium exchange. The global H/D exchange of ferric and ferrous cyt c and their interaction with CL was analyzed using MALDI-TOF. Hypothetically, CL binding may induce conformational change on cyt c to yield varying levels of deuteron incorporation. As expected, ferric and ferrous cyt c exhibited similar masses of m/z = 12,344.0 and 12,344.1 Da, respectively ( Supplementary Fig. S3). After 50 min of deuteration, ferric cyt c retained 46.0 deuterons and the mass shifted to 12,390.0 Da (Fig. 1A). However, ferrous cyt c accumulated 29.1 deuterons and the mass shifted to 12,373.2 Da (Fig. 1B). The greater extent of deuteration indicated that ferric cyt c held a less compact structure. These results implied the redox state of hemes was relevant to a perturbation on the Fe coordination with surrounding residue such as Met 80, leading to a loosen structure. Furthermore, the CL-bound cyt c exhibited an enhanced extent of deuteration, in which 56.7 and 40.6 deuterons were incorporated in ferric and ferrous cyt c, respectively (Fig. 1C,D). This result indicated that CL binding loos- Dynamic interaction monitored using QCM-D. Compared with the end-point analysis of the global deuteration of cyt c, a dynamic study is capable of offering insight into kinetics of the interaction between cyt c and CL. We performed QCM-D to analyze the in situ binding behavior. A schematic summarizing the analytical procedures (Sections I-IV) is depicted in Supplementary Fig. S4. The corresponding frequency (f) and dissipation (D) on the SiO 2 sensor surface were recorded ( Fig. 2A,B). As depicted in Section I of Fig. 2, f shifted to approximately − 56.52 Hz ( Fig. 2A) and − 56.22 (Fig. 2B) with a D of 4.07 × 10 −6 (at the 7th overtone), which indicated that the assembly of the CL/DPPC lipid layer on the SiO 2 surface was reproducible. The subsequent exposure to the Tris-HCl buffer altered both Δf and ΔD minimally (Section II). Crucially, a sequential exposure of the supported CL/DPPC membrane to cyt c substantially increased and decreased f and D, respectively. With the equilibration to the Tris-HCl buffer, the f of ferric cyt c shifted to − 82.50 Hz and D to 5.85 × 10 −6 (Section IV, Fig. 2A). These shifts were considerably greater than those yielded by ferrous cyt c (− 72.50 Hz on f and 4.93 × 10 -6 of D), which indicated that ferric, rather than ferrous, cyt c exhibited stronger affinity with CL. Furthermore, the greater dissipation shift (ΔD) was presumably attributed to an extensional structure of the ferric cyt c-bound CL/DPPC complex. This result was consistent with that observed in global H/D exchange experiments, in which 56 and 40 deuterons were retained in ferric and ferrous cyt c, respectively (Fig. 1). On the basis of the change in f and D, the hydrated mass and shear modulus characteristic of the CL/DPPCsupported lipid membrane, cyt c-bound CL/DPPC complex, and difference representing net cyt c were estimated based on the Kelvin-Voigt model (Table 1). Accordingly, the masses of the ferric and ferrous cyt c bound with CL/ DPPC were estimated to be 455 and 185 ng/cm 2 , respectively, which suggested that ferric-rather than ferrouscyt c exhibited high affinity with CL (Fig. 2C). In addition, dissipation change data were used to estimate the shear modulus of species in the layered format. Although the CL/DPPC lipid membrane exhibited a shear modulus of approximately 6.334 × 10 3 Pa, the interaction of cyt c with such lipid layers increased the shear modulus. Because the value of the shear modulus was regarded as a measure of stiffness, the result revealed that cyt c rendered the cyt c-CL/DPPC complex more rigid compared with bare CL/DPPC. In particular, ferrous cyt c enhanced the stiffness more (shear modulus increase of 4.325 × 10 3 Pa, which was greater than 2.883 × 10 3 Pa of ferric cyt c).
Section III was further analyzed to understand the dynamic interaction between cyt c and CL/DPPC. The D values, recorded during Section III of Fig. 2A,B, were plotted against the corresponding f (Fig. 3). Accordingly, the |ΔD/Δf| ratio was calculated to determine the structural relaxation of the cyt c-bound CL/DPPC complex caused by the per unit mass increase of cyt c. In both the ferric (Fig. 3A) and ferrous (Fig. 3B) cyt c cases, four |ΔD/ Δf| values were observed throughout the increasing f curve, which were represented in phases of a → b, b → c, c → d, and beyond d (Fig. 3). The point a denotes the onset at which the CL/DPPC lipid of supported layer was exposed to cyt c. D decreased with the increase in f until point b, the state with the smallest D. The a → b period was short (15 and 28 s for ferric and ferrous cyt c, respectively). However, the |ΔD/Δf| of ferrous cyt c was greater than that of ferric cyt c, which indicated that ferrous cyt c altered the CL/DPPC lipid layer at greater amplitude,  www.nature.com/scientificreports/ presumably because of the higher shear modulus (rigidity) of ferrous cyt c (Table 1) HDXMS of pepsin-digested cyt c. In addition to using HDXMS and QCM-D, we aimed to retrieve  www.nature.com/scientificreports/ observed low deuteration levels in the loops 10-24, 25-32, and 35-46, which contained the L-site and a CL binding site composed of Lys22, Lys27, and His33, and were close to heme coordination residue His18. By contrast, the Met80-containing loop and 70 s Helix region (67-82) were deuterated to a relatively high level. Another CL binding region (A-site), containing Lys72, Lys73, Lys86, and Lys87, was located in this region. In addition, the contact regions of n-terminal and c-terminal helices, as well as the 50 s helix, were deuterated to high levels ( Supplementary Fig. S6).

Redox-dependent deuteration of cyt c.
To investigate the conformational change of cyt c induced by heme redox, an HDXMS analysis of ferrous cyt c was conducted. Compared with ferric cyt c, ferrous cyt c exhibited a lower (0-20% reduction) level of deuteration in most fragments (Fig. 4), which indicated a stable structure of the ferrous cyt c. The results were consistent with the higher shear modulus obtained through QCM-D. A detailed examination of the heme-proximity region revealed that the heme reduction decreased deuteration of regions 65-82, containing Met80. However, minimal changes occurred in regions 9-24, containing His18. The contact area between C-terminal regions 95-104 and N-terminal regions 1-10 exhibited considerable decreases in deuteration (approx. 50%). The disturbance of Fe-Met80 coordination had a limited effect on 60 s and 70 s helices. Given that no changes were observed in His18, the effects on the N-terminal helix (i.e., the deuteration decrease in this region) were probably affected by contact with the C-terminal helix. These results were consistent with those obtained using NMR 23 , in which two redox states were compared. The results proved that oxidation of the heme changed Fe-Met80 coordination and triggered a structural change to expose the heme molecule outward. Approximately 15% of the surface area of heme (III) became solvent accessible, which was higher than 7% of heme (II). This behavior highlighted the higher deuteration of ferric cyt c caused by the high solvent accessibility.
HDXMS study of CL-bound cyt c. HDXMS analyses were conducted to investigate the structural change caused by CL binding. Figure   Because the two significantly affected regions were not physically in contact with each other, CL binding presumably distorted both regions simultaneously. The increase of cyt c deuteration indicated that CL binding did not block the solvent accessibility in the binding regions, but increased the flexibility of cyt c with an extended conformation. By contrast, the deuteration increase of ferrous cyt c caused by CL binding was limited (Fig. 6), which could be attributed to a compact structure. A minimal change occurred in the deuteration level of the N-terminus and the His18-containing loop upon CL binding. Regions 9-24 in ferrous cyt c were stable and therefore CL binding did not affect this heme coordination region. The deuteration level changed substantially in the regions 95-104 (28.9%) at C-helix and Loop 22-32 (30.1%). Regions 47-64 and 65-82 exhibited minor changes at 16.5% and 21.6%, respectively. Overall, the deuteration-enhanced regions and amplitude were similar to those of ferric cyt c. Crucially, regions 9-24 exhibited a 30.6% change in ferric cyt c, compared with only 10.6% in ferrous cyt c, which indicated that CL binding disturbed heme coordination with His18 in ferric cyt c, but not in ferrous form. This also suggested that CL interacted with the A-site and L-site in ferric cyt c, but only with the A-site in ferrous form. An alternating current voltammetric study on a cyt c-immobilized electrode also supported this conclusion ( Supplementary Fig. S7), in which the potential shift (24.1 mV) of ferric cyt c induced by CL binding was greater than 9.20 mV of ferrous cyt c. Furthermore, the redox potential of cyt c, principally acted by heme, shifted cathodically, which indicated a relatively facile electron transfer occurred upon CL binding.

Discussion
The results are summarized as follows: (i) CL binding had a different effect on ferric and ferrous cyt c in C-helix and Loop 9-24. (ii) Loop 9-24, containing Fe coordination residue His18 and the thioether covalent bonding residues Cys14 and Cys17, stabilized the relative position of heme and the protein structure. (iii) CL interaction with ferric cyt c not only affected Met80 but also destabilized His18 interaction with heme, which was not observed in ferrous cyt c. (iv) The C-helix on ferric cyt c was close to full deuteration after 30 min of deuteration, which indicated a high exposure of this region; however, such behavior was not observed in ferrous cyt c. The www.nature.com/scientificreports/ structured extension of C-helix also highlighted the conclusion revealed by FRET 11 . (v) The contacted N-helix did not exceed a 10% difference, which indicated the interaction between the N-helix and C-helix remained intact in this extended conformation. Although studies on the activation mechanisms of phospholipase A 2 on membranes have indicated that membrane association and penetration mostly cause decreased deuteration 31,33,34 , Skinner et al. proved that H/D exchange depends primarily on the hydrogen bonding status of the amide hydrogen; only a minor contribution is from solvent accessibility 50,51 . Thus, the increasing deuteration revealed in the current study was probably because of the results of conformational changes and because of blocking solvent accessibility. Although either the insertion of cyt c into CL/DOPC 48,49 or the extended CL anchorage accommodated into cyt c 52 was proposed, our results suggested that cyt c did not, if any, only occurred in the very beginning association (within a couple seconds), substantially penetrate into the hydrophobic motif of liposome.
On the basis of the aforementioned results, a model was proposed. In a healthy cell, the mitochondrial intermembrane space full of protons is in an oxidizing state 53 . Therefore, protons compete with cyt c to associate with CL 54 . Free ferric cyt c can therefore transfer electrons from ETC III to ETC IV. Under oxidative stress or mitochondrial damage, the loss of proton gradient and electron transfer results in a low concentration of proton and ferrous cyt c. Without competition from proton and ferrous cyt c, ferric cyt c tightly associates with CL. CL binds to both the A-and L-sites of flexible ferric cyt c and triggers the structural extension of ferric cyt c, which further activates cyt c. Less compact and extended cyt c exposes the active site on the CL-liposomal surface to oxidize CL 55 . The oxidized CL inhibits the cyt c interacting with the mitochondrial inner membrane, leading to the release of cyt c from mitochondria and the subsequent apoptosis process.
Preparation of CL/DPPC liposome. We adopted 20 mol% of (18:1) 4 CL/DPPC as the experimental system. Considering the critical micelle concentration (CMC) difference between DOPC and DPPC [56][57][58] , the DPPC liposome can be formed readily. Plus, lipids containing lower saturation of the fatty acyl chains were considered to exist in inner membrane contact sites 59 . Furthermore, the literature also showed that 20 mol% of DPPC/CL formed the most thermodynamically stable binary monolayer 60 . CL-containing liposome was prepared by dissolving DPPC (  Data analysis. An intensity threshold (5000-count) of the peptides was first filtered using Data Analysis 3.4 (Bruker Corporation). The sequence of the peptides was identified through MS/MS analysis in triplicate. Peptide identification was performed on an X!Tandem parser 1.7.7 followed by further manual examination of product ions. The peptides that were recognized more than twice and with an excellent ion match in the product ion spectrum were selected into the pool. HDexaminer 1.3 (Sierra Analytics) was used for the mass spectra analysis, which was similar to a previously described version 29 . The results were incorporated into HDexaminer, which retrieved the charges, sequences, and retention times from the mass data. The software evaluated the match www.nature.com/scientificreports/ between the experimental data and theoretical mass envelopes, and provided a score for each peptide fragment. Every mass envelope was further manually examined to ensure the mass envelope was identified correctly. The mass shifts at different time points of 0, 10, 30, 100, 300, 1000, and 3000 s were calculated individually. The deuteration level of each peptide was determined using the ratio of the incorporated deuteron number to the maximum possible deuteration number. Because of the fast off-exchange rate of the two N-terminal residues, those residues could not retain any deuterons after liquid chromatography and therefore were not included in the calculation.

Conclusions
The differential conformational change of cyt c during CL-association was confirmed by the results from the QCM-D and HDXMS analyses. The data indicated that ferric cyt c, not ferrous cyt c, exhibited an extended and partially unfolded structure on the lipid surface. The domains of cyt c were not inserted into the liposome; if they did, it would likely occur at the initial stage for a few seconds. This was highlighted by the increasing deuteration and dissipation shift in the initial interaction. The CL molecule was proposed to be extracted out of the lipid aggregates to act as the substrate for a cyt c-induced peroxidation, which presumably caused the decreasing dissipation observed in the beginning stage of interaction (Section III).