Structural snapshots of the cellular folded protein translocation machinery Bcs1

The protein Bcs1 (Bcs1) facilitates the transport of iron‐sulfur protein or ISP in a folded form across the inner membranes of mitochondria. ISP is needed for the assembly of Complex III, a key element of the respiratory chain. Structures of Bcs1 in three nucleotide states offer insights into the mechanisms of ISP translocation and Complex III assembly and shed light on the molecular basis of diseases caused by mutations in Bcs1.

Abstract

Proteins destined to various intra‐ and extra‐cellular locations must traverse membranes most frequently in an unfolded form. When the proteins being translocated need to remain in a folded state, specialized cellular transport machinery is used. One such machine is the membrane‐bound AAA protein Bcs1 (Bcs1), which assists the iron‐sulfur protein, an essential subunit of the respiratory Complex III, across the mitochondrial inner membrane. Recent structure determinations of mouse and yeast Bcs1 in three different nucleotide states reveal its homo‐heptameric association and at least two dramatically different conformations. The apo and ADP‐bound structures are similar, both containing a large substrate‐binding cavity accessible to the mitochondrial matrix space, which contracts by concerted motion of the ATPase domains upon ATP binding, suggesting that bound substrate could then be pushed across the membrane. ATP hydrolysis drives substrate release and resets Bcs1 conformation back to the apo/ADP form. These structures shed new light on the mechanism of folded protein translocation across a membrane, provide better understanding on the assembly process of the respiratory Complex III, and correlate clinical presentations of disease‐associated mutations with their locations in the 3D structure.

Abbreviations

AAA

ATPases associated with various cellular activities

Bcs1

bc1 synthesis 1

bcs1

yeast bcs1 gene

Bcs1

human bcs1‐like protein

BCS1L

human bcs1‐like gene

cyt bc₁

cytochrome bc1 complex or Complex III

cyt

cytochrome

IMS

mitochondrial intermembrane space

ISP

Rieske iron‐sulfur protein

ISP‐ED

extrinsic domain of ISP

mBcs1

mouse Bcs1 protein

pmf

proton motive force

Rip1

yeast iron‐sulfur protein

RIP1

yeast iron‐sulfur protein gene

ROS

reactive oxygen species

TM

transmembrane

yBcs1

yeast bcs1 protein

Introduction

Protein translocation occurs when proteins are moved across membranes to various destinations inside and outside the cell, whereas protein dislocation takes place in the process of protein quality control and homeostasis. These two very important processes, despite requiring different cellular machinery, occur through a ubiquitous threading mechanism in which protein substrates are unfolded while being translocated. Examples of protein translocation apparatus are the SecY channel on the prokaryotic cytoplasmic membranes, the Sec61 channel of the endoplasmic reticulum (ER) membranes of eukaryotes [1], and the TOM and TIM23 complexes of mitochondria [2]. On the other hand, the human AAA unfoldase Cdc48/p97 is a well‐established example of cellular protein dislocation machinery needed for protein quality control and homeostasis, including ER‐associated degradation [2, 3, 4].

Not all proteins are transported in the same unfolded fashion. The transport of the mitochondrial Rieske iron‐sulfur protein (ISP) across the inner membrane (IM) is a well‐established example. The ISP is one of the three essential subunits of the mitochondrial respiratory Complex III, also known as the cytochrome bc ₁ complex (cyt bc ₁). As a mid‐segment of the respiratory chain, Complex III couples the reaction of electron transfer from ubiquinol to cyt c to the proton translocation across the membrane, contributing to the mitochondrial proton motive force (pmf) [5, 6]. Inhibition of Complex III has evolved as an effective survival strategy in the microbial world over the course of evolution and has been widely used for disease treatment in agriculture and in medicine [7]. The functional importance of Complex III is also exhibited by the fact that mutations in its components lead to various mitochondrial diseases [8]. Complex III of eukaryotic organisms is dimeric, each monomer consisting of 10 (Saccharomyces cerevisiae) or 11 (Bos taurus) different subunits [5, 6]. All subunits of Complex III, except for cyt b, are nucleus‐encoded, brought into the mitochondrial matrix and refolded before being integrated into Complex III [9, 10]. The ISP subunit in particular requires additional assistance to cross the mitochondrial IM again, going into the intermembrane space (IMS) as folded protein. Bcs1 is a protein that plays a key role in assisting the ISP subunit to cross the mitochondrial IM [11].

A mature ISP subunit, as shown in the structure of Complex III, consists of two domains: an N‐terminal transmembrane (TM) helix and a C‐terminal, functional globular domain, also known as the ISP extrinsic domain (ISP‐ED), which contains a 2Fe‐2S cluster and is located in the IMS in the final assembled Complex III. A flexible yet highly conserved linker connects the two domains [12, 13]. The integration process of ISP into Complex III of high eukaryotes such as the cow is slightly different from that of lower eukaryotes as in yeast. The maturation of the ISP precursor takes place in the mitochondrial matrix because of the matrix localization of the iron‐sulfur cluster synthesis machinery [14]. Following its entry into the mitochondrial matrix as an unfolded polypeptide chain, the ISP precursor folds into an apoprotein amenable to 2Fe‐2S cluster installation [15, 16]. For the fully assembled ISP precursor, its functional C‐terminal domain, ISP‐ED, must be translocated again across the IM into the IMS, while its TM helix becomes associated with the membrane section of Complex III and the N‐terminal long signal sequence remains in the matrix. After the integration is complete, the precursor of the mammalian ISP is processed in a single step, resulting in a mature ISP with the signal peptide retained as subunit 9 [16]. Yeast Rip1 (yeast ISP), however, requires a two‐step process by mitochondrial processing peptidase and mitochondrial intermediate protease [16].

Identified initially in baker's yeast (S. cerevisiae) as a protein factor required for incorporation of Rip1 into the yeast Complex III [17, 18], Bcs1 is a mitochondrial membrane protein consisting of a Bcs1‐specific domain at its N terminus and an ATPase domain at its C‐terminus, the latter bears sequence similarity to the ATPases associated with various cellular activities family proteins (AAA). AAA family members are involved in a wide range of essential cellular processes including protein quality control and sorting, DNA replication and repair, and membrane fusion [19]. Morphologically, they most frequently occur as ring‐shaped homo‐hexameric ATPases, forming a central pore for peptide threading or moving along DNA strand. Recent advances in structural studies of AAA proteins have unveiled a hand‐over‐hand or split washer threading mechanism common to these hexameric AAA proteins, in which coexisting, sequentially arranged nucleotide states in subunits of the hexamer are found coordinating with channel‐forming substrate‐engaging pore loops that are arranged in a spiral staircase‐like fashion. Presumably, a cycle of sequential, around‐the‐ring ATP hydrolysis performed by the ATPase domains results in stepwise substrate translocation [3, 20, 21, 22, 23, 24]. Whether the operation of Bcs1 follows the established principle of AAA protein function remains to be seen, given the fact that some of the critical residues in the pore loops of AAA proteins are not conserved in Bcs1 [11]. As an integral membrane protein, Bcs1 is anchored to the mitochondrial IM by single N‐terminal TM helices of its subunits with most of the molecule remaining in the matrix [25]. Linkage of Complex III deficiencies to mutations in DNA identified the culprit to be human bcs1‐like gene (BCS1L), the human ortholog of yeast bcs1 [26, 27, 28]. In addition to its requirement for Complex III assembly, Bcs1 was shown to also regulate the amount of Complex III in mitochondria in response to availability of fermentable carbon sources in yeast [29]. Despite the clear evidence of involvement of Bcs1 in Complex III assembly shown by genetic and biochemical studies, exactly how Bcs1 performs its function remains unclear. The recent reported structures of mouse (mBcs1) and yeast (yBcs1) Bcs1 shed new lights on the assembly of Complex III, mutations linked to Complex III deficiency, and the mechanism of function of a distinct family of AAA proteins [30, 31].

Structure of apo Bcs1

Structures of the mBcs1 and yBcs1 in apo form were recently determined by Cryo‐EM [30, 31]. Because the structures are similar in most aspects, the following discussion will be based largely on the mBcs1. References will be made if certain structural features are only observed in the yBcs1. Apo mBcs1 takes the form of a heptamer and, when viewed perpendicular to the 7‐fold axis with the mitochondrial matrix at the bottom, has an overall shape of an upside‐down mushroom (Fig. 1A,B). The heptameric arrangement of Bcs1 was also confirmed by crystal diffraction data, displaying a 7‐fold symmetry axis in self‐rotation function searches from both full‐length and a AAA domain fragment crystals [30], indicating a deviation from the classic hexameric structure found in most AAA proteins [19]. The structure of apo mBcs1 can be divided into three regions: the AAA region, the Bcs1‐specific region (also called the seal‐forming region), and the TM region; each region is ~ 30–40 Å in length. Unique to the apo structure is the ~ 15 Å gap, called interstitial gap, between the Bcs1‐specific region and the AAA region. The apo structure features two large cavities: One is referred to as the substrate‐binding cavity, which is encircled by AAA and part of the Bcs1‐specific domains of the subunits and the other is called the TM basket, which is formed by TM helices. The two cavities are separated by a protein layer made of residues from the Bcs1‐specific region (Fig. 1C). At the center of the protein layer, there is a small pore connecting the two cavities, which is plugged by a piece of density of unknown identity.

Fig. 1.

Structure of mBcs1 in apo state. (A) Ribbon diagram of mBcs1 determined by cryo‐EM in the absence of nucleotide viewed perpendicular to the 7‐fold axis. The model is orientated with the IMS of the mitochondria at the top, and the matrix is at the bottom. The mitochondrial IM is indicated by the two parallel black lines. The TM helices are only partially visible from the EM density. The model is derived from PDB: 6UKP. All ribbon diagrams were rendered using the molecular graphics program pymol (http://pymol.org). (B) A 90‐degree rotated view of mBcs1 model looking down the 7‐fold axis. (C) Surface rendition of mBcs1 model with two of its subunits removed showing the large substrate‐binding cavity, the TM cavity, or TM basket and the protein layer separating these two cavities. (D) Ribbon diagram of yBcs1 in the ADP‐bound state as a comparison where models for TM helices are complete.

The structure of the apo mBcs1 has only half of the TM helix modeled (PDB: 6UKP), whereas the yeast structure in the ADP state contains the entire TM helix (PDB: 6SH4), which measures ~ 50 Å in length and forms an angle of about 35° with the plane of the membrane (Fig. 1D). They contact each other on the IMS side of the membrane, but on the matrix side at the beginning of the Bcs1‐specific domain two neighboring helices are separated by a 25 Å gap.

Interestingly, two apo forms have been observed for the yBcs1 based on cryo‐EM 3D classification and refinement, which were referred to as Apo1 and Apo2 [31]. A notable difference between the Apo1 and Apo2 state was found in the Bcs1‐specific domain. While the pore separating the substrate‐binding cavity and TM basket domain in the Apo1 form is tightly sealed, that for Apo2 appears to be open [31].

Domain structures of Bcs1

Sequence alignment of Bcs1 from different eukaryotic organisms indicates identical domain organization of their subunits, consisting of an N‐terminal mitochondrial targeting and sorting domain that includes a TM segment [25, 26], a middle Bcs1‐specific or seal‐forming domain, and a C‐terminal AAA ATPase domain (Fig. 2A).

Fig. 2.

Structures of Bcs1 domains: (A) Each Bcs1 subunit has three domains: the TM (purple), Bcs1‐specific (brown), and AAA domains. The AAA domain is further divided into the AAA‐large (AAA‐L) in green and AAA‐small (AAA‐S) in magenta. The mitochondrial IM is indicated by the two parallel lines. (B) The Bcs1‐specific domain has a novel fold, formed by two β‐sheets (Sheet I and Sheet II) and two α‐helices (Helix I and Helix II). (C) The heptameric association of the Bcs1‐specific domain, viewed along the 7‐fold axis from the matrix, allows formation of seven 6‐stranded β‐sheets, which is the structural basis for the seal. (D) The structure of the AAA domain consists of two subdomains: a large RecA‐like domain, also called AAA‐L, and a small helical domain, referred to as AAA‐S. The AAA‐L domain features a unique 7‐stranded β‐sheet. The ATPγS is shown as a stick model with carbon in black, oxygen in red, and sulfur in yellow. The Mg²⁺ ion is shown as a metallic ball.

The tertiary structure of the Bcs1‐specific domain (Residues 49–165 of mBcs1) appears to be unique. A search of the protein structure database revealed no similar protein folds [32]. The Bcs1‐specific domain includes two β‐sheets (Fig. 2B): Four antiparallel strands A, F, E, and D contribute to Sheet I. Sheet II, formed by two antiparallel stands (B and C), is also called connector hairpin because it reaches down and interacts with a neighboring AAA domain (Fig. 1A). In addition, there are two helices (H1 and H2) that provide support for the two β‐sheets. When associated in the heptamer, the Bcs1‐specific region forms two layers when viewed along the 7‐fold axis from the membrane (Fig. 2C). The Sheet I forms the top layer and Sheets II and the two helices are in the bottom layer. Furthermore, Sheet I joins Sheet II of its neighboring subunit to form a continuous six‐stranded β‐sheet, giving the entire Bcs1‐specific region the appearance of a 7‐bladed propeller. In addition, a loop between strands E and F, named the seal loop, protrudes toward the central axis of the heptamer, forming a small constriction ~ 5 Å in diameter (Fig. 2C).

The AAA ATPase domain has 254 residues (Residues 165–418 for mBcs1). In addition to the EM structure, the structure of the AAA domain alone was determined by X‐ray crystallography at 2.2 Å resolution (PDB: 6U1Y). These structures contain features that are consistent with the analysis of yeast mutants and their revertants [33]. As in typical AAA domains, there are two subdomains: a large RecA‐like domain and a small characteristic helical domain (Fig. 2D). At the interface between the two subdomains lies the nucleotide‐binding site. Most interestingly, the structure reveals that the AAA domain of Bcs1 is unique in that it includes a 7‐stranded β‐sheet in the RecA‐like subdomain, which is the first of its kind reported for a AAA protein (see below for more details about unique features in Bcs1).

For apo Bcs1, the heptameric association of Bcs1 subunits leads to creation of a large barrel‐like cavity or substrate‐binding cavity encircled by the AAA domains and part (Sheet II or connector hairpin and Helix 2) of the Bcs1‐specific domains from the subunits. The barrel has a diameter of ~ 40 Å at its widest part and is capped by the rest (Sheet I and Helix I or seal‐forming domain) of the Bcs1‐specific region (Fig. 1C). A large entrance accessible to the matrix allows protein substrate to enter the cavity.

Conformational change revealed by the structure of mBcs1 with bound ATP

Isolated Bcs1 is enzymatically active. mBcs1 possesses the kinetic constants for ATP binding (K _m) of 0.04 mm and V _max of 5.8 nmol Pi/nmol protein/min [30]. The presence of ATP or its analogs in solution results in Bcs1 undergoing conformational change that can be detected by Blue‐Native PAGE (BN‐PAGE), causing a shift in the migration of the protein band. Interestingly, conformation of mBcs1 seems not to be affected by presence of ADP in solution, as no change was detectable in the migration of the protein in BN‐PAGE. A similar observation was also previously reported for yeast Bcs1 [34].

When the structure of mBcs1 in the presence of ATPγS was first determined, a piece of density was found in the nucleotide‐binding pocket of the AAA domain, which matches the shape and size of ATPγS (Fig. 2D). Both the EM density and the resulting AAA domain model were successful in phasing the diffraction data of a ^ΔNmBcs1 crystal that grew in the presence of AMP‐PNP, yielding a high‐resolution heptameric ^ΔNmBcs1 structure that shows clearly bound AMP‐PNP (PDB: 6U1Y), indicating conformational similarity between the ATPγS‐ and AMP‐PNP‐bound Bcs1 [30].

Unlike the apo mBcs1, the TM region in the ATPγS state was not visible. Therefore, the first residue seen for this structure is I49 at the beginning of the Bcs1‐specific domain. ATPγS binding drives mBcs1 to undergo a conformational transition when compared to the apo mBcs1. It resembles the shape of a bullet head morphologically, when viewed perpendicular to the 7‐fold axis, and becomes more compact with a reduced barrel diameter (Fig. 3A,B). Although the binding of ATPγS seems to have minimal impact on the AAA domain as well as the Bcs1‐specific domain in terms of structural changes, the binding does induce significant shifts in relative positions of the domains, thus altering how a subunit interacts with its neighboring subunits. Such changes can be clearly illustrated by superposing an ATPγS‐bound subunit to an apo subunit and observing the domain movement of the same subunit and relative changes in positions of the neighboring subunits (Fig. 3C). For example, by superposing the two AAA ATPase domains in different nucleotide states, one can observe a very large movement by the Bcs1‐specific domain, where such a movement can be as large as 45 Å measured from the tip of the Helix II of the Bcs1‐specific domain.

Fig. 3.

Structure of mBcs1 in ATP state and conformational changes. (A) Ribbon diagram of the structure of heptameric mBcs1 with bound ATPγS viewed perpendicular to the 7‐fold axis. The two horizontal parallel lines delineate the boundary of mitochondrial IM from the matrix and IMS as indicated. Seven ATPγS molecules are drawn as stick models and one of them is labeled. (B) 90‐degree rotated view from the orientation in (A). Again, one ATPγS is labeled. (C) Superposition of AAA domains of Bcs1 subunits between ATPγS‐bound (green) and apo (magenta) conformations. The positional movement of Helix II expressed as the distance change in residue K145 from the tip of the Helix II is indicated. (D) Collapsing of the substrate‐binding cavity as Bcs1 transitions from the apo to the ATPγS‐bound forms. On the left is the Apo‐Bcs1 model with two subunits deleted to more easily view the Bcs1 interior. Similarly, cartoon rendition of the ATPγS‐bound mBcs1 is given on the right, which has two of its subunits removed.

The largest conformational change by far that can be visualized is when mBcs1 subunits in ATPγS state are associated as heptamers, as the interstitial gap between the Bcs1‐specific region and the AAA region disappears and the substrate‐binding cavity collapses. For the apo mBcs1, the prominent feature is the barrel‐shaped, very large substrate‐binding cavity, which is accessible from the mitochondrial matrix. The barrel is ~ 40 Å in diameter at its widest point and runs as deep as 40 Å along the 7‐fold axis (Fig. 3D). It has an estimated volume of ~ 28 000 Å³, sufficient to contain a folded protein such as the C‐terminal functional domain of ISP (ISP‐ED, 126 residues, 21 000 Å³). This cavity contracts dramatically when ATPγS binds, shrinking the widest part of the barrel by half to 20 Å and reducing its length to 25 Å. As a result, the volume of the cavity decreases to ~ 9000 Å³ and therefore is no longer able to accommodate the folded ISP‐ED (Fig. 3D).

The nucleotide dependence of the conformational change in the TM region is poorly defined, because structures in this region under both apo and ATP states were not well determined for mouse Bcs1. Nevertheless, a glimpse of the TM helix conformation of apo Bcs1 was obtained on the basis of the partially determined structure in that region of mBcs1 and the fully determined TM helices for the yBcs1 in ADP state (Fig. 1A,D). The direction of the partially determined TM helices in mBcs1 suggests close association at the IMS side of the TM helices in the apo structure, which is supported by the observation in the yBcs1 structure of fully ordered TM helices. On the other hand, TM helix conformation for the ATPγS‐bound Bcs1 could not be clearly determined due to poor density and could only be guessed from the starting position of the residues in the Bcs1‐specific domain, which was residue I49. The fact that the TM helices are disordered suggests a lack of interactions among the TM helices when ATP is bound, which seems consistent with the post‐translocation conformation.

Structures of Bcs1 in ADP state

We began our structure determination of mBcs1 by crystallizing purified mBcs1 in the presence of ATPγS·Mg²⁺, because biochemical studies suggested a more compact protein when mBcs1 and ATPγS·Mg²⁺ were present together in solution. We obtained crystals and a complete diffraction data set to 4.4 Å resolution. Surprisingly, we were able to determine and refine the structure by the molecular replacement method only when we used the coordinates of the apo mBcs1 as a phasing model. In contrast to the apo mBcs1, the nucleotide‐binding pockets in the AAA domains of this structure were occupied by large pieces of density that could be modeled as bound ADP in standard geometry [30]. ADP binding induces only local changes in the vicinity of bound nucleotides, while the overall conformation of mBcs1 resembles that of apo mBcs1.

One of the reported yeast structures has ADP molecules that uniformly occupy the nucleotide‐binding pockets of the AAA domains [31], which was the result of incubating the protein with ATP during sample preparation and likely represents an ADP‐bound posthydrolysis ground state of the protomers (PDB: 6SH3). Consistent with the crystal structure of the mBcs1 in ADP state, the ADP‐bound yBcs1 adopts the conformation of apo Bcs1.

Mechanistic insight concerning folded protein translocation across membranes by Bcs1

As early as 1992, the Bcs1 gene was identified in yeast, its product localized to the mitochondrial IM, and recognized as a requirement for ISP incorporation into Complex III, suggesting its ability to move the 15 kDa functional domain (ISP‐ED) of the ISP subunit across the IM of mitochondria [17, 35]. This hypothesis gained support by experiments showing that mutations causing partial unfolding of ISP failed to be transported by Bcs1 [34]. Because installation of the 2Fe‐2S cluster into apo ISP occurs in the mitochondrial matrix, translocation of the ISP‐ED going into the IMS in its folded form becomes mandatory. However, it is by no means trivial to move a 126‐residue folded protein domain across the membrane barrier, giving its 25 Å diameter and the need to maintain a large pmf across the IM. Until the structure determination of Bcs1, cellular machinery facilitating passage of substrates of this size through the channel of a AAA protein had not been visualized [11]. By obtaining the mBcs1 and yBcs1 structures in three different nucleotide states and many conformations, we are now in a position to derive important insights into the mechanics of the transport process (Fig. 4). First, in order to transport a substrate, the size of the ISP‐ED, the Bcs1 channel must be sufficiently large. Therefore, Bcs1 occurs in the form of a heptamer instead of a hexamer (the latter form is more frequently seen in AAA proteins). Indeed, the heptameric Bcs1 affords it the ability to create a larger entrance and substrate‐binding cavity to accommodate the folded protein substrate.

Fig. 4.

Current model for the translocation of ISP by Bcs1. Two conformations of Bcs1 are presented schematically: Apo and ADP (D) conformation and ATP (T) conformation. Here, TM helices are colored in purple‐blue, the Bcs1‐specific domains in two different tones of orange, the AAA‐L domain in green and the AAA‐S domain in blue. The IM of mitochondria is shown as two black horizontal lines. The ISP is rendered with two domains: ISP‐ED as a yellow oval and TM helix as a yellow rod. The 2Fe‐2S cluster is depicted as a red dot. In the apo state, Bcs1 has a large entrance and cavity to hold the entire ISP‐ED. Upon ATP binding, Bcs1 undergoes a very large conformational change that contracts the substrate‐binding cavity and allows translocation of the ISP‐ED across the mitochondrial IM. ATP hydrolysis releases ISP to the membrane section of Complex III and resets Bcs1 back to the apo conformation. The dashed diagonal line indicates that substrate binding could take place in the ADP state.

Secondly, it seems that the binding cavity can only accommodate the substrate when all subunits of Bcs1 are in the apo or ADP‐bound state, because in this conformation the substrate‐binding cavity features an average diameter of 35 Å and a volume of 28 000 Å³, which is large enough to contain the substrate. Our finding is in agreement with a prior observation showing that a strong interaction of ISP with Bcs1 only occurs in the absence of ATP [34]. The substrate‐binding barrel is encircled by the AAA domains and by the seal‐forming, Bcs1‐specific domains that sharply narrow the barrel to a constriction of <5 Å in diameter, which is made of four hydrophobic residues (M121, M123, V124, and L128 of mouse sequence) arranged symmetrically from the seal loop of each subunit, separating the barrel from the TM basket. It is worth pointing out that the physiological relevance of the apo state is not clear because of the amount of ATP and ADP in the mitochondrial matrix. It is possible that the ADP form allows loading of substrate, which triggers subsequent nucleotide exchange. The observation that the presence of two different Apo conformations (Apo1 and Apo2) with different constriction sizes in the Bcs1‐specific region may indicate that the pore opens and closes in a sort of breathing motion for Bcs1 in the absence of substrate. This could be possible as long as the membrane integrity is not compromised [31].

Thirdly, upon ATP binding, as seen from the ATPγS‐bound structure, all Bcs1 subunits undergo, as proposed, a concerted conformational transition that shrinks the substrate‐binding cavity to a volume of barely 9000 Å³, reducing the cavity size by ~ 70%, so that it can no longer accommodate the substrate. Simultaneously, the opening to the cavity is reduced in diameter from 40 to 20 Å, preventing substrate from slipping back out of the cavity. Therefore, the ATPγS‐bound structure represents the conformation of postsubstrate translocation. It should also be mentioned that after the translocation, the substrate presumably remains bound to the Bcs1 via the TM helix and, possibly, the N‐terminal segment [34], which is consistent with the model for Complex III assembly. Furthermore, in our ATPγS‐bound structure, the pore dilates to a diameter of ~ 14 Å and is no longer plugged with the density of unknown identity. This 14 Å‐diameter pore appears sufficient to accommodate the TM helix of the ISP.

Finally, hydrolysis of ATP by the Bcs1 subunits may achieve two objectives: one is rapid release of the TM helix of the substrate into the membrane section of Complex III, which suggests the possibility of opening the Bcs1 ring asymmetrically, and another is resetting the system back to the apo/ADP Bcs1 conformation, as represented by the postsubstrate release ADP‐bound mBcs1 and yBcs1 structures and the apo mBcs1 structure.

Bcs1 represents a distinct class of AAA proteins

The assembly of Complex III is not the only process that requires transport of a folded protein across the membrane. In photosynthetic bacteria and in plant chloroplasts, the task of transporting folded proteins is designated to the Twin arginine translocation (Tat) pathway, a multi‐subunit membrane protein complex coupling its function to membrane pmf [36]. Most interestingly, one substrate of the Tat system is the ISP subunit of the cyt b ₆ f complex, a homolog of the cyt bc ₁ complex (Complex III) in plant chloroplasts [37]. Thus, researchers have speculated that over the course of evolution Tat has been replaced by Bcs1 transporting folded proteins in mitochondria. It is worth pointing out that other systems exist in bacteria for transporting folded proteins. One well characterized example is the chaperone‐user pathway, in which folded periplasmic pilin subunits are translocated across the outer membrane to form pili that are essential for bacterial pathogenesis [38].

Identical to other type I AAA proteins, Bcs1 has a functional, specialized domain at its N terminus and features a characteristic single AAA ATPase domain C‐terminus to its subunit, which bears signature sequence motifs common to all AAA proteins. However, outside the core conserved AAA motifs such as Walker A and Walker B motifs, Bcs1 is quite different from other AAA proteins, displaying a low overall sequence identity (~ 30%). The difference is also shown in a phylogenetic analysis of AAA domain sequences, in which Bcs1 was grouped into its own clade [39]. The structures of mBcs1 and yBcs1 exhibit the following unique features that may be important to the function of this class of AAA proteins. First, the RecA‐like domain of Bcs1 contains a 7‐stranded β‐sheet, in contrast to a 5‐stranded β‐sheet common to most AAA proteins. The extra two N‐terminal β‐strands connect to the Bcs1‐specific domain. Second, the β‐sheet II, or the connector hairpin, of the Bcs1‐specific domain makes contact with the AAA domain of its neighboring subunit, indicating a communication pathway between neighboring subunits. Finally, ATP binding induces a sliding movement between neighboring AAA subunits such that the Arg‐finger residues (R343 and R346) in the apo mBcs1 structure, which are 10 and 9 Å, respectively, away (Fig. 5A) from the superposed ATP in the ATP‐bound structure, snap into position to make contact with the γ‐phosphate of the bound ATPγS. For comparison, other AAA domains such as the bacterial σ54 activator NtrC1 and the D1 domain of protein unfoldase Cdc48/p97 show relatively small movements between neighboring subunits going from ATP to ADP states [40].

Fig. 5.

Unique structural features of Bcs1. (A) Sliding movement at the interface of neighboring AAA domains. Upon ATP binding, AAA domains of mBcs1 undergo a sliding movement relative to its neighbor. Shown as cartoon in orange is the structure in the vicinity of the ATP‐binding site for the apo structure (both subunits A and B), in which the Arg‐finger residues R343 and R346 of subunit B of the apo mBcs1 are shown as stick models also in orange. Shown in magenta is the subunit A of the ATPγS‐bound mBcs1 and in green is its neighboring subunit B. The Arg‐finger residues R343 and R346 of subunit B of the ATPγS‐bound structure are presented as stick models. The two structures are superimposed on the basis of subunit A. The distances from the side chain atoms NH2 of R343 and R346 of apo or ATPγS‐bound Bcs1 to the γ‐S atom of ATPγS are given. (B) One D2 subunit in red of p97 is superposed on the corresponding Bcs1 subunit in the heptameric context. Insect shows a portion of the superposition, demonstrating that the substrate‐binding loop in p97 is missing in Bcs1.

We propose that a unique, concerted transport mechanism may govern the function of Bcs1 based on the features revealed by the structures, which is in contrast to the more ubiquitous threading mechanism revealed by recent EM studies on a number of hexameric ring‐shaped AAA proteins including the protein unfoldases p97 (Cdc48 in yeast) and Hsp104 [3, 4, 20], the DNA helicase [24], the proteasomal ATPase [21, 22], the membrane fission Vps4 [41], and Rix 7 in ribosome assembly [42]. In those cases, protein substrates or DNA strands are dragged through a narrow channel formed by six AAA subunits. Mechanistically, in these AAA proteins, different nucleotide states simultaneously present in the six subunits of the hexamer coordinate with distinct spiral staircase‐like arrangement of substrate‐engaging pore loops in the central channel. The ATPase domains undergo a sequential, around‐the‐ring ATP hydrolysis cycle that leads to stepwise substrate translocation/dislocation.

Given the fact that Bcs1 belongs to the family of AAA proteins, we cannot rule out the possibility that Bcs1 uses a mechanism similar to the asymmetric hand‐over‐hand mechanism to transport the ISP‐ED, especially considering the absence of substrate in current studies. That said, it is equally important to understand the roles of the distinct features in the AAA domain of Bcs1 in substrate translocation, which include Bcs1 forming its own clade in the phylogenetic tree, featuring a distinct 7‐stranded β‐sheet, and undergoing relatively large movements between subunits transitioning from apo to ATP states. Additionally, sequence alignment demonstrates that the AAA domain of Bcs1 is actually more similar to the D1 domain of p97/Cdc48, which does not appear to actively participate in substrate translocation. The D2 domain of p97/Cdc48, on the other hand, plays a major role in the hand‐over‐hand mechanism [3]. Moreover, the Bcs1 subunit lacks the critical substrate‐contacting pore loop (Fig. 5B), one of the conserved structural motifs of those AAA ATPases that employ the hand‐over‐hand mechanism for threading. Also missing in Bcs1 is the sensor‐II region in the small helical domain that is known to involve in binding of ATP.

Structural comparison between mBcs1 and yBcs1

In the two recently published papers on the structures of Bcs1, a total of five EM structures and two crystal structures were presented (Table 1). Among these, three are EM structures for yBcs1, and two EM and two X‐ray structures are for mBcs1, representing two dramatically different conformations and three nucleotide states. Because there is no equivalent structure of mBcs1 in the ATPγS‐bound state for yBcs1, structural comparison was only conducted between apo‐mBcs1 (PDB: 6UKP) and ADP‐bound yBcs1 (PDB: 6SH3). Both structures were well determined at 3.8 and 3.4 Å resolution, respectively. Structure superposition for the entire heptamer supports the notion that apo mBcs1 and ADP‐bound yBcs1 are in the same conformation due to a rms deviation of 3.5 Å for all aligned 2126 CA atoms using the molecular modeling program coot [43]. Indeed, superposed yBcs1 in ADP‐bound state and apo mBcs1 show the same pore size, the same volume of the substrate‐binding cavity, and the same dimension for the entrance to the substrate‐binding cavity. Structure‐based sequence alignment of Bcs1 proteins from yeast and mouse gave a more accurate alignment with a sequence identity of 44% and similarity of 59%, respectively (Fig. 6A). The sequence similarity for the TM helix is limited compared to the Bcs1‐specific and AAA domains.

Table 1.

Structural models of Bcs1 in the PDB.

PDB	Organism	Construct (no. residues)	Method	Nucleotide state	Overall map. resolution (Å)	Model	Ref
6UKP	Mouse	Full‐length (1–418)	EM	Apo	3.8	29–417	[30]
6UKS	Mouse	Full–length (1–418)	EM	ATPγS	3.2	49–418	[30]
6U1Y	Mouse	AAA domain (151–418)	X‐ray	AMP‐PNP	2.2	164–418	[30]
6UKO	Mouse	Full‐length (1–418)	X‐ray	ADP	4.4	49–417	[30]
6SH3	Yeast	Full‐length (1–456)	EM	ADP	3.4	49–449	[31]
6SH4	Yeast	Full‐length (1–456)	EM	Apo1	4.4	49–449 (No side chain)	[31]
6SH5	Yeast	Full‐length (1–456)	EM	Apo2	4.6	84–449 (No side chain)	[31]

Fig. 6.

Comparison of mouse and yeast Bcs1. (A) Structure‐based sequence alignment between mBcs1 and yBcs1. In the alignment, conserved residues are shown as boldfaced characters and similar residues are colored green. (B) Superposition of the mouse Bcs1 (red) in the apo state and yeast Bcs1 (gray) in the ADP‐bound state. Superposed structures are shown in cartoons in two orthogonal orientations.

Despite the overall good structural alignment, the TM helices do not superpose well, as TM helices of the apo mBcs1 are rotated counter‐clockwise relative to those of ADP‐bound yBcs1, when viewed along the 7‐fold axis toward the matrix (Fig. 6B). Similar counter‐clockwise rotation is also found for the seal‐forming loops, as Bcs1 goes from the ADP‐bound to the apo state, suggesting the possibility that these loops work like a camera diaphragm shutter, controlling the size of the pore. In addition to the rotation movement, there is an upward movement of the Bcs1 specific domain as the protein undergoes nucleotide state transition. Such an upward movement is also observed for the AAA domain. Thus, there is a coupling of the rotational and translational movement of each subunit during ADP to apo transition. The scale of this transition is likely magnified when Bcs1 is bound to ATP, leading to substrate translocation.

Structural interpretation of Bcs1 mutations that cause human diseases

Clinical investigations of patients with Complex III deficiencies have linked some cases to mutations in the gene BCS1L, which is the human bcs1 ortholog gene (bcs1‐like gene). Presumably, mutants of Bcs1 fail the task to install the ISP subunit into the Complex III [27], leading to Complex III deficiency (MIM# 124000, www.omim.org) in neonates or infants with clinical presentation of encephalopathy alone or in combination with visceral involvement. However, it is puzzling that different mutations in Bcs1l result in a wide variety of clinical phenotypes that range from the relatively mild Björnstad syndrome to the severe GRACILE syndrome [44]. The GRACILE syndrome (MIM# 603358, Growth Retardation, Aminoaciduria, Cholestasis, Iron overload, Lactacidosis, and Early death) is a severe form of the disease that results in liver and kidney failure in young children. By contrast, the Björnstad syndrome (MIM# 262000) has a much milder clinical presentation with patients showing highly restricted pili torti (twisted hair) and sensorineural hearing loss [45]. Curiously, Björnstad syndrome‐causing mutations are accompanied by increased level of reactive oxygen species (ROS) [46]. In real life, clinical presentations of BCS1L mutations can fall anywhere between the two extremes [47].

Diseases caused by BCS1L mutations are autosomal recessive, requiring both copies to be mutated. Positively charged residues, mostly arginine residues, are involved in 60% of missense mutations. Furthermore, many of the mutated residues are highly conserved. Studies on the phenotype‐genotype relationship have primarily focused on mapping BCS1L mutations to structural models of Bcs1l [46, 47]. These studies faced difficulties in the use of structures created by homology modeling, which we are now able to overcome by using experimental structures of mBcs1. The structure of mouse Bcs1 is an excellent representation of human Bcs1 structure, because they share 94% sequence identity. By mapping 2D sequence locations of the mutations onto the 3D structure of mBcs1, we were able to establish a correlation between the 3D positions of the mutations and the reported severity of symptoms displayed by patients. For example, we found that GRACILE syndrome is often seen in patients with mutations in the Bcs1‐specific domains, whereas the milder Björnstad syndrome links more frequently to mutations in the AAA domain. Thus, structural analysis suggests a relationship between the severity in clinical presentation or prognosis of BCS1L mutations and the structural integrity of various regions of the protein. Because the Bcs1‐specific region may work as a seal or shutter to control substrate passage through the membrane and may play an essential role in preventing harmful dissipation of membrane pmf during the translocation process, it seems reasonable that mutations in this region result in more severe consequence. By contrast, the AAA domain mutations may only cause problem in the assembly of Complex III, which seems consistent with the observation that these mutations often lead to increased production of ROS due to increased pmf [46].

Conclusions and Future directions

Having determined the structures of mBcs1 and yBcs1 in different nucleotide states and conformations, we now have acquired a structural framework from which more detailed mechanistic insights into the transport function of Bcs1 can be expected. Future studies will likely reveal how Bcs1 recognizes and binds the folded ISP‐ED, capture its action in transporting the ISP substrate across the membrane, and witness how substrate is released into the membrane. To achieve these goals, a combination of various research approaches will have to be used.

From a structural point of view, it is necessary to obtain the structure of Bcs1 in complex with the substrate ISP in order to address the questions on how substrate binding trigger changes in Bcs1 and whether binding of substrate is sufficient to induce nucleotide exchange. Structures are also needed to determine whether subunits of Bcs1 functions in a sequential fashion or in a concerted manner. The former is the hallmark of the hand‐over‐hand or split wash mechanism of translocation displayed by many hexameric AAA proteins. In the apo and ADP‐bound structures, the unknown density plugging the small pore in the center of the Bcs1‐specific domains should also be investigated.

Biochemically, kinetic study of the life span of different nucleotide states will provide clues on the rate limiting steps in the reaction landscape. Coupling these studies with mutagenesis will likely play a major role in verifying various mechanistic hypotheses. For example, to prevent proton leakage during translocation, an airlock‐like mechanism was proposed [31]. However, how the opening and closure of the seal pore is controlled requires further elucidation. Mutagenesis studies will allow functional and structural characterizations of many documented disease‐related mutants. The structures should also facilitate development of drugs to modulate function of Bcs1.

Conflict of interest

The authors declare no conflict of interest.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1111/febs.15576.