Making the leap from structure to mechanism: are the open states of mammalian complex I identified by cryoEM resting states or catalytic intermediates?

Respiratory complex I (NADH:ubiquinone oxidoreductase) is a multi-subunit, energy-transducing mitochondrial enzyme that is essential for oxidative phosphorylation and regulating NAD+/NADH pools. Despite recent advances in structural knowledge and a long history of biochemical analyses, the mechanism of redox-coupled proton translocation by complex I remains unknown. Due to its ability to separate molecules in a mixed population into distinct classes, single-particle electron cryomicroscopy has enabled identification and characterisation of different complex I conformations. However, deciding on their catalytic and/or regulatory properties to underpin mechanistic hypotheses, especially without detailed biochemical characterisation of the structural samples, has proven challenging. In this review we explore different mechanistic interpretations of the closed and open states identified in cryoEM analyses of mammalian complex I.


Introduction to complex I
Respiratory complex I is a key metabolic enzyme in mammalian mitochondria [1,2]. Due to its defining roles in NADH homeostasis, respiration and oxidative phosphorylation, defects in complex I are associated with a wide range of clinical mitochondrial diseases [3e5]. Complex I catalyses oxidation of NADH and reduction of ubiquinone-10, coupled to transfer of four protons across the inner-mitochondrial membrane to generate the proton-motive force (Dp) that powers ATP synthesis and transport processes. Furthermore, mammalian complex I is a thermodynamically reversible catalyst, switching cleanly into reverse (Dp-driven ubiquinol-10:NAD þ oxidoreduction or 'reverse electron transfer') when Dp is high enough [6,7].
Recent developments in single-particle electron cryomicroscopy (cryoEM) have led to an explosion of complex I structures from mammals, plants, single-cell eukaryotes and bacteria [8e22], providing unprecedented opportunities for understanding its mechanisms of catalysis and regulation. They have revealed the conserved architecture of complex I and key elements of its catalytic machinery [ Figure 1a]. Redox catalysis occurs by fast and reversible NADH oxidation by a flavin mononucleotide at the top of the hydrophilic domain; electron transfer by a chain of ironesulphur clusters; and reduction of ubiquinone-10 in a long, amphipathic channel at the interface of the hydrophilic and membrane domains.
Proton transfer occurs against Dp in a series of four modules; it is most likely powered by energy transfer along a chain of charged residues from the ubiquinone-binding site. The coupling mechanism, the least understood component, is intimately linked to ubiquinone reduction, but there is no consensus on the molecular coupling mechanisms that use redox catalysis to trigger and drive proton transfer and conserve the energy released. Notably, two recent publications [11,12] have both been entitled "the coupling mechanism of mammalian [respiratory/mitochondrial] complex I" d but the two mechanisms proposed have little in common and neither has been substantiated by coherent biochemical and biophysical analyses. Nevertheless, contrasting them highlights a key issue at the heart of the debate: are the conformational states of mammalian complex I identified by cryoEM (in both these and earlier studies) off-cycle resting states or on-cycle catalytic intermediates?
Observation of multiple states in cryoEM analyses of mammalian complex I As soon as cryoEM particle classification was employed to investigate the homogeneity of 'as-prepared' samples of mammalian complex I (grouping protein molecules according predominantly to their global conformations), it was obvious that more than one class is typically present. We described three major classes in our preparations of bovine (Bos taurus) complex I [8,23] but only one in the mouse (Mus musculus) enzyme [24]. The ovine (Ovis aries) and porcine (Sus scrofa) enzymes were resolved into four [11,25] and two [12] major classes, respectively.
One of the major classes, described as the 'closed' state as it exhibits the smallest apparent angle between the hydrophilic and membrane domains that form the complex I L-shape, is common to all the analyses. As expected for a catalytically relevant state, the loops that compose the ubiquinone-binding channel at the domain interface in the closed state are well ordered [ Figure 1b]. Common to work on the bovine and porcine enzymes is also a well-defined 'open' state characterised by disorder in the ND3-transmembrane helix (TMH) 1e2, ND1-TMH5e6 and NDUFS2-b1eb2 loops that form the ubiquinone-binding channel, changes to the NDUFS7-Arg77 loop, as well as associated changes in Architecture of complex I and key elements of the active/closed-deactive/open transition. a) A schematic overview of mammalian complex I. Key subunits involved in the active/closed-deactive/open transition are shown in colour, with the remaining core and supernumerary subunits outlined in black or shaded in grey, respectively. The ubiquinone-binding channel is indicated with a box and four proton transfer routes are shown schematically with dotted lines. b) A cartoon representation of the ubiquinone-binding channel in the active/closed state [PDB: 7QSL (protein), 7QSK (Q 10 )], outlining ND1-TMH4 and the ND1-TMH5-6, ND3-TMH1-2, NDUFS2-b1-b2, and NDUFS7-Arg77 loops. the membrane domain involving the ND6-TMH3 and ND1-TMH4 helices [ Figure 1c and d]. As a result of the disordered ubiquinone-binding channel in the open state, the domain interface has relaxed and the apparent hinge angle has opened, in a movement most clearly visualised by the relative positions of subunits NDUFA5 and NDUFA10 on the hydrophilic and hydrophobic domains, respectively [ Figure 1c].
Due to variations between species, preparations, and classification strategies, different numbers of open states have been observed. An additional open class identified for the bovine enzyme has been named the 'slack' state because disorder in specific elements of the membrane domain opens the ND2eND4 interface and further relaxes the global structure [8,23]. Similar characteristics were observed in the major class from a cryoEM analysis of complex I from macaque (Macaca mulatta), which displayed only very low activity [26]. The functional competence of the slack state is thus uncertain, and we do not consider it further here. For simplicity, we focus only on the two (closed and open) states described above. Finally, we note that multiple open classes of ovine complex I have been described [11,25]. They have much in common with the open states described for the bovine and porcine enzymes, but have not been distinguished structurally in the same way. Instead, they have been described as a distribution that can be further divided by more detailed classification [25], consistent with a progressive relaxation of structural restraints (that extends to include slack-like states). The status of the distinguishing elements in 'as-prepared' samples of the four mammalian species discussed are summarised in Table 1. Next, we consider how the closed and open states of mammalian complex I observed in cryoEM may be interpreted and reconciled with what is known about its activity and behaviour. The active/deactive transition of complex I was first described by Vinogradov and co-workers [27], and later proposed to be prominent in ischaemiaereperfusion injury [28e30]. When mammalian complex I stops catalysing, it adopts the so-called 'active' resting state, a 'ready-to-catalyse' resting state. The active resting state gradually converts to the 'deactive' resting state, a pronounced resting state that requires reduction by NADH and reactivation by ubiquinone to return to catalysis [Scheme 1a] [27,29,31]. The two resting states can be differentiated biochemically by their sensitivity to Nethylmaleimide (NEM), which prevents reactivation of the deactive state by derivatising ND3-Cys39 on the ND3-TMH1e2 loop [ Figure 1d] [32e34]. On this basis, preparations of mammalian complex I typically comprise a mixture of the active and deactive resting states, so we proposed [23] that the closed structure corresponds to the active resting state, and the open structure to the deactive resting state. The tightly defined closed structure and its well-ordered ubiquinone-binding channel suggests it is ready to begin catalysing immediately, whereas the disordered elements of the open state are consistent with its need for restructuring and reactivation. Furthermore, ND3-Cys39 is occluded in the closed structure but in the open structure the ND3-TMH1e2 loop is disordered, consistent with exposure of Cys39 [ Figure 1d]. Our initial assignment was substantiated by cryoEM analysis of a sample of bovine complex I prepared from purposefully deactivated membranes, which displayed deactive biochemical characteristics (NEM sensitivity and a catalytic lag phase during reactivation [27,33]) and which was highly active following reactivation [35]. CryoEM revealed the open structure described above as the dominant state in the deactivated sample, and a matching structure was determined similarly for purposefully deactivated mouse complex I [24].
We propose that the structural elements that change during deactivation [ Figure 1 and Table 1] are unstable in the active resting state, perhaps because they are conformationally mobile during catalysis and/or because they are destabilised when the resting binding channel is occupied by water molecules or fatty acids [36] instead of ubiquinone. Therefore, they slowly relax in the resting enzyme, in a coordinated transition to the open/deactive resting state. Their closed/active conformations are recovered when substrates stimulate and template their restructuring [Scheme 1a]. Therefore, the open/deactive resting state does not feature on the catalytic cycle and sustained catalysis, including ubiquinone binding and ubiquinol release, occurs within a set of closed intermediates. Several mechanistic proposals are consistent with this interpretation [12,15,37e40], which provides a broad structural rationale for long-standing biochemical observations on the deactive state, including its sensitivity to NEM, catalytic lag phase during reactivation, and slow but spontaneous formation in ischaemic tissue that is protective against ischaemiaereperfusion injury [10,28e30].
Interpretation 2: opening and closing as an intrinsic feature of catalysis Kampjut and Sazanov recently proposed an alternative interpretation of the mixture of open and closed states that they observe in their 'as-prepared' resting ovine complex I [11,41]. Based on the fact that it is catalytically competent, they proposed that all the states observed by cryoEM in the preparation (open and closed alike) are catalytic intermediates. Consequently, they proposed that opening and closing is an intrinsic and essential part of the catalytic cycle, in which ubiquinone-binding is initiated in the open state, the enzyme closes for ubiquinone reduction, and then reopens again as ubiquinol is released [Scheme 1b]. The closed enzyme is predicted not to exist without Table 1 The status of structural features identified to differ between the mammalian closed/active and open/deactive states in preparations of complex I from different species.

Species
Given All the structures listed are 'as-prepared' enzymes that have not been treated with substrates, inhibitors or to activate or deactivate them. In S. scrofa and T. thermophila the enzyme is contained in a supercomplex. Each structural feature is compared to its structure defined in this laboratory in the 'active' Similarly, opening either occurs slowly during deactivation d or rapidly, during every turnover cycle.
In Interpretation 1, a mixture of active and deactive (and inactive) enzymes will immediately begin to turnover upon substrate addition due to the ready-to-catalyse active molecules, then increase its rate upon reactivation of the deactive molecules (inactive molecules will remain inactive). Therefore, the active/deactive model can explain the catalytic competence of a mixed population. To capture known biochemical active/deactive behaviour for ovine complex I in their proposal, Kampjut 24,35], and these changes to ND6, together with loss of density for nearby subunit NDUFA11, may instead reflect the known instability of ovine complex I in the absence of complex III [25], exacerbated by the incubation in detergent. It is possible that ovine ND6 recovers its native structure, together with the structural elements highlighted in Figure 1, during global reactivation of the deactive enzyme. Alternatively, its altered structure may not affect catalysis, as suggested by the different positions/disorder of ND6-TMH4 in structures of all the non-mammalian species listed in Table 1.

The challenges of combining catalysis and cryoEM
Experiments to freeze complex I onto cryoEM grids while it is catalysing have been carried out in a quest to observe the structures of the intermediates present directly d perhaps to catch the elements discussed above in different conformations. For complex I, the experiment is technically demanding due to challenges in providing sufficient electron acceptor (ubiquinone or O 2 ) to sustain catalysis in a high concentration sample undergoing rapid turnover for long enough for grid preparation and freezing.  Figure 1 and Table 1] are conformationally mobile during catalysis, and note that movement of the ND3-TMH1e2 loop has been suggested on the basis that using cross linking to restrict it was observed to decouple catalysis [44]. In the future, using the same method to restrict opening and closing movements may provide an alternate approach to evaluate their catalytic relevance.

Insights from cryoEM analyses of nonmammalian complexes I
The fourteen 'core' subunits of complex I are conserved in all species and considered sufficient for catalysis. Therefore, we expect the mechanism to be conserved, and all species to catalyse via a matching set of catalytic intermediates and transition states. In contrast, different enzymes, with different thermal stabilities, supernumerary subunits and physiological environments, and isolated using different procedures, may relax differently in the absence of substrates and rest in different conformations. Therefore, we surveyed structures of complex I from plants, single-celled eukaryotes, bacteria and archaea to evaluate the status of the key structural elements that change in the mammalian complex during deactivation/opening [see Table 1].
It is clear that the simple binary nature of the active/closed and deactive/open resting mammalian enzymes does not extend to the other organisms, with many exhibiting mixed characteristics. Only one, from the ciliate Tetrahymena thermophila, is observed by cryoEM in a homogeneous active/closed state under resting conditions. This structure argues against opening and closing during catalysis as it contains species-specific supernumerary subunits that lock the substrate-binding site in the active/closed conformation [17]. Common to all others is the deactive-type p bulge in ND6-TMH3 that disrupts the proposed connection between the ubiquinone-binding site and protontranslocating modules [8,11,38]. Nearby is ND1-TMH4, in the deactive-type straighter form in most species. The active-type bent form correlates partially with ordered active/active-like conformations of the ND3-TMH1e2, ND1-TMH5e6, NDUFS2 and NDUFS7 loops that form the ubiquinone-binding channel [ Fig. 1b]. These varied combinations of active-and deactive-type elements may suggest each species of enzyme relaxes differently and to a different extent, from the same initial resting state when catalysis stops. For example, Y. lipolytica complex I has a lower energy barrier for its (limited) active-to-deactive transition than the mammalian enzyme [45,46]. In contrast, other enzymes show no evidence of undergoing a transition: T. thermophila complex I has been proposed to be structurally trapped in the closed state [17], and Paracoccus denitrificans complex I (which contains a Cys39 equivalent) displays no sensitivity to NEM [47]. Between the mammalian enzymes, the relative proportions of open and closed states observed suggest that the mouse enzyme (predominantly closed) has the highest barrier for deactivation/opening and the ovine enzyme (mostly open) the lowest. Alternatively, the different combinations of conformations may result from different species adopting different initial resting states (at different stages around the cycle) when catalysis stops, so they might provide insights into how different mobile elements change their conformations individually during catalysis.
The ND6-P25L mouse model: rapid deactivation and unidirectional catalysis