Major concerns with the integrity of the mitochondrial ADP/ATP carrier in dodecyl-phosphocholine used for solution NMR studies

Hereby, we wish to note our objections to a paper called Substrate-modulated ADP/ATP-transporter dynamics revealed by NMR relaxation dispersion by Bruschweiler et al., published in NSMB in 2015. The subject is the yeast mitochondrial ADP/ATP carrier AAC3, which we have studied in great detail ourselves. In particular, we have solved its structure by electron and x-ray crystallography and have studied its interactions with the specific inhibitors atractyloside (ATR) and carboxyatractyloside (CATR) by single-molecule force spectroscopy. In this paper, the authors claim that AAC3 can be refolded to homogeneity from inclusion bodies produced in Escherichia coli by using the detergent dodecyl-phosphocholine (DPC), better known as Foscholine-12 (Anatrace), and that AAC3 is maintained in a folded and active state for the duration of isothermal titration calorimetry (ITC) and NMR experiments. However, in our hands the presence of DPC leads to immediate loss of tertiary structure and inactivation of AAC3 when isolated from the inner membrane of mitochondria, where it was folded and active as shown by functional complementation.

However, in our hands the presence of DPC leads to immediate loss of tertiary structure and inactivation of AAC3 when isolated from the inner membrane of mitochondria, where it was folded and active as shown by functional complementation 2,3 .
To validate the integrity of their refolded protein, the authors first used ITC to measure a Kd of 15 µM for the binding of CATR to DPC--refolded AAC3. They then determined CATR--dependent chemical shift perturbations by NMR, giving a Kd of ~150 µM. No other functional assays were done to verify the activity of the refolded protein, such as transport assays. The inhibitors CATR and ATR differ by one carboxylate group and their affinities have been described in 24 different binding studies (Supplementary  Table  1). Yet, the authors only compare their values to a Kd of 192 µM for the binding of ATR taken from one particular study by Babot et al., 2012 6 . Importantly, the µM unit used in this reference was a typographical error and the Kd value has recently been corrected to 192 nM in an erratum 7 . In fact, published Kd values for CATR binding to AAC in the membrane are all in the low nanomolar range (Supplementary Table 1), consistent with our own measurements that show that ADP transport by AAC3 is half--maximally inhibited by 1.2 nM CATR (Fig.  1a). Thus, the consensus is at least three to four orders of magnitude lower than the Kd values reported for refolded AAC3 in DPC 1 . Our ITC measurements using native AAC3 from yeast mitochondria purified in dodecyl-maltoside/tetraoleoyl cardiolipin gave an average Kd of 72 nM (Fig. 1b), which is 200 to 2000--fold lower than those reported in Brüschweiler et al.. These experiments were very difficult to do (only 2 out of 10 trials succeeded), as the enthalpic change is low and the apo--state is very unstable in detergent 5 . There are very few polar side chain interactions in AAC3 that stabilize the structure and they are mainly found on the matrix side, where they form intra--domain rather than inter--domain interactions 3 . The ring of transmembrane a-helices is held together largely by the lateral pressure of the membrane and counter pressure from the water--filled cavity (Supplementary Fig. 1a) 3 , explaining why unliganded AAC3 in detergent micelles is prone to unfolding. Binding of CATR introduces a large number of polar and van der Waals interactions, which cross--link most of the transmembrane α--helices of AAC3 together 3,17 , explaining the high affinity of CATR for AAC3 as well as the improved stability in detergents ( Supplementary  Fig.  1b). The experiments reported here and elsewhere clearly show that the Kd of CATR binding to the folded mitochondrial ADP/ATP carrier is in the low nanomolar range, consistent with its role as a powerful toxin. We further note that to obtain the crystal structure, AAC3 was first inhibited with CATR in the mitochondrial membrane, but then solubilized, purified and crystallized in maltoside detergents in the absence of CATR for 5--7 days 3 . Yet, a clear interpretable density for CATR could be observed 3 , demonstrating that the inhibitor remained bound, consistent with extremely low rates of dissociation. The extremely low affinity of CATR to refolded AAC3 in DPC begs the question whether the binding is specific at all. Only a limited number of chemical shift perturbations for CATR binding to refolded AAC3 are observed 1 and all of them are dynamic, which is inconsistent with tight binding. Moreover, with few exceptions these residues are not near the known CATR binding site nor are they on structural elements that are involved in CATR binding.
We also note that the Kd of ADP binding determined by NMR (500 μM) 1 is ~85--fold higher than the published consensus values of the carrier in the mitochondrial membrane and 25--fold higher than for the solubilized carrier (Supplementary Table 2). Residues that were assigned to have chemical shift perturbations induced by ADP are largely on the matrix side of the carrier, far away from the consensus binding site in the central cavity 8--13 . AAC3 has an isoelectric point of 9.82, meaning that it is highly positively charged, whereas CATR and the substrates ADP and ATP are negatively charged molecules at neutral pH. The chemical shift perturbations could represent non--specific interactions of CATR or ADP with AAC3 promoted by the high concentrations and temperatures used in these NMR experiments.
We are also concerned about the validity of the dynamic studies measured by NMR relaxation dispersion, as CATR binding should lock the protein in a non--dynamic aborted state, which is why would we could solve its structure by crystallography 3 . We also note that the observed dynamics in the presence of substrate do not provide a plausible structural mechanism for transport.
We have previously demonstrated that DPC is harsh enough to solubilise unfolded mitochondrial carrier protein from E. coli inclusion bodies and is able to denature functional well--folded carrier protein prepared in mild non--ionic detergents 14 . In thermostability assays 15 , AAC3 purified from yeast mitochondrial membranes displayed a typical protein melt curve when diluted into the mild detergent dodecyl--maltoside, consistent with thermal denaturation of a folded protein (Fig.  1c). When CATR was added at a molar ratio of one or above, a marked shift in stability of AAC3 to higher temperatures is observed, consistent with earlier observations 14,16 . When the same AAC3 preparation was diluted into 3 mM DPC, a high signal was observed with no transition, showing that AAC3 in DPC is in a non--native state (Fig.  1d). In this case, CATR addition had no effect, demonstrating that there was no functional binding site. Consistent with these findings, dilution of AAC3 into DPC before reconstitution resulted in a complete loss of CATR--sensitive ADP uptake in liposomes, in contrast to control tests where the protein was diluted into dodecyl--maltoside prior to reconstitution ( Supplementary Fig. 2). These observations clearly demonstrate that AAC3 is soluble but in a non--native state in DPC. In conclusion, we believe that the data presented by Brüschweiler et al. have no biological relevance. (a) Inhibition of ADP transport by the yeast mitochondrial ADP/ATP carrier AAC3 by carboxyatractyloside. AAC3 purified from yeast mitochondria in dodecyl-maltoside was reconstituted into liposomes. The CATR concentration required for half--maximal inhibition was determined by measuring the initial ADP uptake rate at different concentrations of CATR (n=3). (b) Isothermal titration calorimetry: enthalpy changes associated with the titration of CATR into AAC3 at 10 °C (left panel), and corresponding isotherms fitted to a one site binding model with ΔH, Kd, stoichiometry and ΔS as fitting parameters (right panel). (Inset) Average data (± SD) from two titration experiments. Assays were carried out in buffer containing 20 mM PIPES pH 7.0, 100 mM NaCl, 0.1% dodecyl--maltoside, 0.1 mg mL --1 tetraoleoyl cardiolipin. (c) Thermostability of the yeast mitochondrial ADP/ATP carrier AAC3 diluted into dodecyl--maltoside in the presence of different amounts of CATR (left panel) and derivatives (right panel). (d) same as (c) but with AAC3 diluted into dodecyl--phosphocholine. The temperature of the protein sample is increased from 25 to 90 °C while protein unfolding is monitored with the fluorophore N--[4--(7--diethylamino--4--methyl--3--coumarinyl)phenyl] maleimide (CPM) 15 . CPM reacts with protein thiols as they become solvent--exposed due to denaturation of the protein to give a fluorescent adduct. The CATR:AAC3 molar ratios were 0 (black), 0.001 (dark blue), 0.01 (light blue), 0.1 (green), 1 (orange), and 10 (red), and 100 (dark red). The protein concentration was approximately 1 μM AAC3.

Preparation of lipid for protein purification
Tetraoleoyl cardiolipin (18:1) dissolved in chloroform was purchased from Avanti Polar Lipids (Alabaster, Alabama). Typically, 100 mg of lipid was dispensed into a glass vial, and chloroform was removed by evaporation under a stream of nitrogen gas. Lipids were solubilized in 10% (w/v) dodecyl-maltoside by vortexing for 4 h at room temperature to give a 10 mg mL --1 lipid in 10% detergent stock.
The stocks were snap--frozen and stored in liquid nitrogen.

Purification of AAC3
A 5--L pre--culture was used to inoculate 50 L of YPG medium in an Applikon 140 Pilot System with an eZ controller. Cells were grown at 30 °C for 72 h, and harvested by centrifugation (4,000 g, 20 min, 4 °C).
Mitochondria were prepared with established methods 18

Thermostability analysis
Thermostability data were obtained by using the thiol--reactive fluorophore N--[4--(7--diethylamino--4-methyl--3--coumarinyl)phenyl] maleimide (CPM), which undergoes an increase in fluorescence emission following reaction with cysteine residues 20 . A modified procedure using a rotary qPCR machine was used, as described previously 21 . For this purpose, a 5 mg mL --1 stock of CPM dissolved in DMSO was diluted 50--fold into buffer containing 20 mM PIPES pH 7.0, 100 mM NaCl, 0.1% dodecyl--maltoside and 0.1 mg mL --1 tetraoleoyl cardiolipin, vortexed and the solution was allowed to equilibrate in the dark at room temperature for 10 min. Approximately 1.5 μg of purified protein was added into a final volume of 45 μL buffer containing either 20 mM PIPES pH 7.0, 100 mM NaCl, 0.1% dodecyl--maltoside, 0.1 mg mL --1 tetraoleoyl cardiolipin (for the dodecyl--maltoside assays) or 20 mM PIPES pH 7.0, 100 mM NaCl, 3 mM dodecyl--phosphocholine (for the dodecyl--phosphocholine assays), and 5 μL CPM working solution was added, and the solution was vortexed and allowed to equilibrate in the dark for 10 min at room temperature in the presence of increasing concentrations of carboxyatractyloside. Fluorescence of the CPM dye was measured on a Qiagen Rotorgene Q using the HRM channel, which provides excitation light at 440--480 nm with emission detected at 505--515 nm. Measurements were made in 1 °C intervals from 25 - 90 °C with a 'wait between reading' set to 4 s, which equated to a ramp rate of 5.6 °C/min, following an initial pre--incubation step of 90 seconds. Data were analyzed and melting temperatures, the peak in the derivative of the melting curve, were determined with the software supplied with the instrument.

Isothermal titration calorimetry
The enthalpy changes associated with carboxyatractyloside binding to purified AAC3 were recorded with a NanoITC--LV isothermal titration calorimeter (TA Instruments) at 10 °C. Both carboxyatractyloside titrant and protein samples were degassed under vacuum at 10 °C for at least 20 min before the titration experiment. Carboxyatractyloside (500 μM stock prepared in a buffer containing 20 mM PIPES pH 7.0, 100 mM NaCl, 0.1% dodecyl--maltoside and 0.1 mg mL --1 tetraoleoyl cardiolipin) was titrated into purified 25 μM AAC3 (0.83 mg mL --1 in the same buffer) in 2--μL injections in an initial volume of 170 μL at 3--min intervals with a stirrer speed of 350 rpm. Isotherms were analysed by using the instrument software (NanoAnalyze) and fitted to a one--site binding model with ΔH, Kd, and stoichiometry as fitting parameters.

Reconstitution
L--α--phosphatidylcholine and tetraoleoyl cardiolipin were mixed in 20:1 (w/w) ratio (10 mg total lipid) in chloroform and dried under a nitrogen stream, dissolved in methanol, and re--dried to a smear in a