Technical Comments

Response to Comment on “Protein assemblies ejected directly from native membranes yield complexes for mass spectrometry”

See allHide authors and affiliations

Science  08 Nov 2019:
Vol. 366, Issue 6466, eaax3102
DOI: 10.1126/science.aax3102


Hirst et al. claim that proteins ejected directly from mitochondrial membranes in our study are degraded, are incorrectly assigned, lack lipids, and show discrepancies with “native states” mostly obtained in detergent micelles. Here, we add further evidence in full support of our assignments and show that all complexes are either ejected intact or in known intermediate states, with core subunit interactions maintained. None are degraded or rearranged.

The fundamental and important point of our paper (1) is that multiple complexes are ejected simultaneously, without chemical intervention, from their native membrane environment. It is therefore imperative not to compare the properties of these proteins with those extracted in detergent, which promotes lipid association to compensate for the unnatural environment, and changes subunit stoichiometry and composition (2, 3). An important distinction is that we did not claim the complexes were in their native state, but rather that they were ejected from native membranes. However, the vast majority remain intact, particularly with respect to their membrane-embedded subunits. We also reiterate our assignment strategy in which all complexes were within ±0.3% of the calculated mass. Protein masses were taken from database entries, and posttranslational modifications were taken into account where reported.

Discussions with Hirst led us to revise our table for complex I, submitted to Science prior to receiving these concerns and published as an erratum. The assignment of complex I was not “combinatorially” made, neither for the erratum nor for the original supplementary table. As is common in native mass spectrometry (nMS), the mass difference between consecutive charge state series was used to assign complexes (4). In this case, the difference was correctly ascribed to the NDUFS3 subunit. For this subunit to dissociate, the N module must be absent. Our revised table, containing only masses from Hirst et al. (5), does not change the correct assignment of peaks to complex I. Taking into account the corrected subunit masses, the subcomplex (809 kDa) corresponds within 0.1% to loss of the N module (Fig. 1A). Because supercomplexes lacking the N module have been observed previously (6) and have been proposed as intermediates in supercomplex regeneration, this new assignment is both structurally (Fig. 1A) and biologically relevant (7).

Fig. 1 Summary of mitochondrial complexes as derived from masses measured in Chorev et al.

(A) Proposed structure of complex I missing the N module, based on PDB 5LDW. NDUFS3 is present substoichiometrically, leading to peak splitting and enabling us to assign complex I [inset (i) is the schematic used in our original manuscript]. Subunits that were not present according to mass measurement are shown in gray. (B) Complex III structures as assigned according to mass in our original report (based on PDB 1BGY) shown schematically in (i) and color-coded according to the structure in (ii). The monomer was validated using CID (figure S12 of original manuscript); the additional copy of UQCRB, which was used to identify the complex, is shown in our original schematic (pink). (C) The dimer is assembled from the two monomers (left) together with our original assignments (i), now also color-coded according to subunits in the structure (ii).

Similarly, our assignment of complex III is based on the dissociation of UQCRB and the simultaneous dissociation of CyB from a complex III monomer, with and without UQCRB subunits derived from the second monomer. All membrane-spanning proteins but one are preserved in the dimer, including UQCRFS1, and phospholipids are resolved. Both the monomer and the dimer are structurally viable, the latter assigned on the basis of the monomer (Fig. 1, B and C). For complex IV, the difference between database masses for isoforms from bovine heart (Hirst et al.) and those used in our table is 69 Da. For an average charge state of 26+, this would equate to 69/26 = 2.65 m/z units; within a peak width at half height of 5.14 m/z units, this would not be detected. A mass of 412,220 ± 38 Da is equivalent to the dimer of complex IV; an additional 7 kDa of adducts are attributed to associated lipids or cofactors. For complex V, we measured all subunit masses in-house, and our results yield discrepancies with the table of Hirst et al. but do not change the assignment of the complex. Comparing lipid binding for ATPases in micelles is not relevant. Recent structures of complex V, including porcine, which shows the c-ring occupied by the protein 6.8PL (8) (observed in our study), as well as yeast complex V and Thermus thermophilus V-type (9, 10), also show absence of lipids.

For ANT-1, the mass and peak width of this small transporter are well within our range for accurate mass measurement (66,391 ± 8 Da). No cardiolipins were observed associated; these are added during purification in detergent or prior to crystallization (11, 12). The measured mass corresponds to that of a dimer, supported by our proteomics of this preparation, which show that the protein is succinylated at least twice (Fig. 2), acetylated multiple times, and highly modified (1). That ANT-1 is dimeric when ejected from membranes is further supported here with collision cross sections consistent with the molecular dynamics simulation and by binding of two ANT-1 inhibitors added to mitochondrial membrane preparations (Fig. 3). Despite hundreds of proteins being ejected simultaneously from mitochondrial membranes, inhibitors bound only to ANT-1 dimers.

Fig. 2 Proteomics data for ANT-1 summarized.

Percent coverage, 78%; predicted mass with methionine removed, 32,836 Da (33,364 including modifications). Only modifications found with high confidence are marked.

Fig. 3 ANT-1 is a dimer when ejected from native membranes.

(A) Structure as derived from molecular dynamics simulations reported previously, and ion mobility collision cross section (CCS) values measured for three different charge states of the dimer, are compared with the calculated value obtained using IMPACT. An average value of 5108 Å2 corresponds closely to the calculated value of 5115 Å2. (B) Tandem MS (m/z = 4151) of the complex (66,391 ± 8 Da) ejected directly from bovine mitochondrial inner membranes yields a subunit mass of 33,195 ± 5 Da. (C and D) Inhibitor binding to the ANT-1 dimer. The upper spectra show the charge states of the ANT-1 dimer prior to addition of inhibitors; the lower spectra show binding of bongkrekic acid and CATr, with measured masses 66,887 ± 15 Da (calculated 66,877) and 67,165 ± 10 Da (calculated 67,161), respectively. Membranes were analyzed under conditions that allow removal of fatty acids, as reported in the original publication.

Hirst et al. ask whether we have assigned our complexes correctly and claim that many are in artifactual states. No complexes were degraded and all complexes are assigned correctly. All proposed structures are viable, intact, or represent assembly states of respiratory complexes (6, 7, 13), which have been observed using structural biology approaches, including electron microscopy, and considered “native” and not artifactual. Our complexes are not identical to those that may have been modified and selectively purified in detergent micelles. The identity of these complexes is supported by our proteomic studies, native and denaturing gels, tandem MS experiments, and detergent extraction controls. Since publication, new studies supporting many of our findings have emerged, including studies that show cardiolipin mediating binding between the BAM complex and holotranslocon (14), dimeric ANT-1 in reconstituted membranes (15), E. coli ATP synthase bound to SecYEG (16), and relatively low populations of a complex IV dimer, relative to monomer, in amphipols (17).

Although our primary objective was to report a new methodology, we disagree that there are no new insights into mitochondrial biology. We have uncovered cardiolipin binding to complex I, and have shown that ANT-1 is a dimer when ejected from membranes and binds to multiple palmitate ions. Furthermore, two inhibitors added directly to mitochondrial membranes show specific binding to dimeric ANT-1 ejected directly from membranes, included as part of this response, expanding the applications of this methodology. We acknowledge that ANT-1 can function as a monomer; so can complex V, despite being a dimer. Moreover, a single monomer of complex III is active within a dimer (6); complex IV is functional as both a monomer and a dimer, with the monomer likely the more abundant form (1, 17).

Detergents have provided valuable data for decades and will continue to do so, but they have also caused numerous controversies. Our report demonstrates a method that enables multiple complexes to be ejected simultaneously from native membranes directly into a mass spectrometer. As such, accounting for individual modifications, arguing about isoforms and cofactors for complexes of hundreds of kilodaltons, as well as capturing supercomplexes, are beyond the current capabilities of nMS technology (18). Breakthroughs like this drive MS instrumentation forward; this approach will progress, along with technological advances, to provide new insights into membrane proteins in their native environments.


Acknowledgments: Supported by ERC advanced grant ENABLE no. 695511.
View Abstract

Navigate This Article