Electrosome assembly: Structural insights from high voltage-activated calcium channel (Ca_V)–chaperone interactions

Ion channels are multisubunit protein complexes that create the spark of life by producing the electrical signals that drive our thoughts, feelings, and actions. Thanks to more than two decades of effort, there are now representative structures for most of the major ion channel classes that provide a framework for functional studies and modulator development. Nevertheless, how these multi-component signaling molecules, which we term “electrosomes”, are put together and undergo quality control to ensure that all of the correct parts are assembled remains largely unaddressed [1]. Among the ion channel types, voltage-gated ion channels (VGICs) form the largest ion channel superfamily [2] encompassing various classes of voltage-gated calcium (Ca_Vs), sodium (Na_Vs), and potassium (K_Vs) channels, the TRP and TPC channel families, inwardly rectifying potassium channels, and K_2P leak potassium channels. All VGIC superfamily members share a common four-fold, barrel stave body plan in which four pore domain (PD) subunits each bearing part of the selectivity filter (SF) assemble around the central axis of the channel, making the hole through which the ions pass. The four PDs can come from four individual polypeptides (K_Vs, Kirs, TRPs), or from polypeptides that carry two (K_2Ps and TPCs) or all four PDs (Ca_Vs and Na_Vs). In addition to the PD, many channel classes have voltage sensor domains (VSDs) or voltage sensor-like domains fused to each PD. Adding further complexity, numerous channel types, such as high-voltage-activated Ca_V1 and Ca_V2s, also have auxiliary subunits that are required to make the final, functioning multiprotein complex [3] (Figure 1A). This multitude of parts required to form a VGIC superfamily channel means that during the initial phase of its life, there must be some set of partially assembled intermediates. Until recently, there was no structural information about any ion channel assembly intermediate, or clear indication about the extent to which chaperone proteins might play a role in channel assembly. Here, we briefly outline key structural issues regarding ion channel biogenesis and focus on studies uncovering the interaction of a class of paradigmatic electrosomes, Ca_Vs, with a ubiquitous endoplasmic reticulum (ER) chaperone known as the endoplasmic reticulum membrane protein complex (EMC) that provides the first structural insight into the ion channel assembly process.

To make a channel, the first thing that must happen is that the pore-forming subunit has to be synthesized by the ribosome and inserted into the ER membrane [6]. The average speed of eukaryotic protein synthesis of ~5.6 amino acids/sec [7,8] sets a biogenic “speed limit” for how quickly a protein of interest can be made that has direct consequences for the folding and assembly of multicomponent proteins such as ion channels. It takes ~3 minutes to make a thousand amino acid protein (for reference, many Ca_V and Na_Vs pore-forming subunits are ~2000–3000 amino acids long) [5]. Regardless of whether all the pore-forming components are in one or multiple polypeptide chains, the fact that the protein is made in a linear fashion means that there must be some period during the synthesis when the various VSDs and PDs are incorporated into the membrane but lack their partners for final assembly. For example, for a channel such as a Ca_V or Na_V in which all four VSDs and PDs are contained in one polypeptide, the time between completing the synthesis of the first and fourth PDs is at least 3.4 minutes [5], leaving the first few PDs without their complete complement of partners for some interval. What happens with these components during this time, particularly the PDs where four are required to make the final pore structure? Can VSDs and PDs fold before assembling into the final quaternary structure? Are partially assembled states protected by chaperone proteins from making aberrant assemblies? How does the cell identify channel parts on the way to being assembled versus leftover pieces from incomplete assembly that need to be disposed? Although the importance of these questions has been appreciated [1,6], there has been scant experimental data to address most of these issues, leaving the ion channel biogenesis as one of the largest blind spots in our understanding of how these critical components of electrical signaling function and how disease mutations or drugs might affect this crucial part of the life of a channel.

High voltage-activated Ca_Vs (Ca_V1 and Ca_V2) [3] are ubiquitous elements of electrically excitable tissues such as muscle, brain, and heart and exemplify the idea of a multisubunit electrosome, comprising three subunits, the pore-forming Ca_Vα₁, cytoplasmic Ca_Vβ [9], and extracellular Ca_Vα₂δ [10] subunits, and the calcium sensor calmodulin [11] (Figure 1A). It has been known for >30 years that Ca_Vβ and Ca_Vα₂δ association with Ca_Vα₁ profoundly shapes Ca_V biophysical properties and plasma membrane expression [9,12-16]. The domain organization of the 24 transmembrane segment Ca_V pore-forming subunit (Figure 1B) into VSDs and PDs highlights the “channel by parts” architecture found throughout the VGIC superfamily in which distinct VSD and PD elements are readily identified [2]. Accordingly, structural data show that both VSDs [4,17-20] and PDs [5] are autonomously folded subdomains (Figure 1B) that are capable of adopting native or native-like tertiary structure in the absence of the quaternary interactions that make the final, folded channel. VSD folding autonomy is not surprising, as their four-helix bundle topology places their domain components near each other in the primary sequence making them self-contained subdomains. In contrast, the ion-conducing pore is only formed only when four PDs assemble, raising the possibility that PD folding could depend on the interdomain interactions formed in the final pore structure. Recent structural studies defining noncanonical quaternary assemblies of bacterial voltage-gated sodium channel (BacNa_V) PDs [5] has shown that the two-transmembrane/P-helix PD architecture folds independently of the extensive quaternary interactions found in the final tetrameric pore (Figure 1B), in line with prior biochemical studies indicating that native-like topology can develop within a single PD subunit [21-23]. Hence, it appears that the initial assembly of the channel pore can proceed by an “assembly by parts” path in which autonomously folded VSDs and PDs come together following their synthesis to make the stereotypic quaternary structures found throughout the VGIC superfamily.

Because four PDs need to be assembled around the channel central axis to create the pore during channel biogenesis, there must be partially assembled pore intermediates comprising two and three PDs (Figure 1C). Such forms have not yet been observed structurally and determining how they form, how long lived they are, and whether they interact with chaperone proteins that could prevent spurious interactions with other membrane components remains an interesting and unaddressed question. Further, in cases like Ca_Vs where the pore is an obligate hetero-tetramer, whether there is a preferred order for how the PDs assemble is not known. Developing strategies to trap such intermediates to probe their properties is an important goal to understand how channel pores are put together and undergo quality control. Finally, for channels such as Ca_Vs that require subunits beyond the pore-forming one, the various parts need to be put together. In the case of Ca_Vs, this means forming a complex in which the intracellular Ca_Vβ subunit [9] and extracellular Ca_Vα₂δ subunit [10] are added to the pore-forming Ca_Vα₁ (Figure 1A). Both interactions control the functional expression channel and understanding how such assembly happens offers not only insight into a fundamental step in the life of a channel but also may open opportunities to target such partially assembled states to control channel function. For this reason, the recent report of the first view of a Ca_V assembly intermediate [24] establishes a new direction for addressing the multifaceted means by which cells control their electrical excitability.

In the course of purifying Ca_V1.2/Ca_Vβ complexes for structural studies, our lab succeeded in isolating the first example of an ion channel:chaperone complex and channel assembly intermediate [24]. Cryo-electron microscopy studies (cryo-EM) of purified, recombinant human Ca_V1.2 and rabbit Ca_Vβ₃ identified a large (~0.6 MDa), stable complex in which the two-channel components were bound to a nine protein ER chaperone complex known as the EMC, a chaperone thought to aid the insertion of tail-anchored proteins [25-28] and transmembrane segments having mixed hydrophobic/hydrophilic character, such as those in ion channels [29-31], receptors [32-34], and transporters [30,32,35] (Figure 2A–B). Despite much study and suggestion that the EMC could function as a holdase for partly folded membrane proteins [29,34], there have been no structural examples showing how the EMC might engage a client. Hence, the EMC:Ca_V1.2/Ca_Vβ complex not only provides the first view of an ion channel:chaperone complex but also the first view of an EMC:client complex.

The two channel proteins each have a binding site on the opposite face of the EMC from where membrane insertion activity occurs (Figure 2B). The transmembrane component of the channel, the Ca_V1.2 Ca_Vα₁ subunit, binds to a site termed the “TM dock” in which the first Ca_V1.2 voltage sensor domain (VSD I) binds the EMC1 subunit through a large (~1000 Ų) interface that spans the membrane. The TM dock has two elements: an interaction site comprising the EMC1 transmembrane helix and S1 and S2 helices from Ca_V1.2 VSD I framed by a set of hydrophilic interactions at each membrane boundary (Lumenal site and Inner Leaflet site; Figure 2C) and an interaction between the EMC1 brace/crossbar helix that lies on the lumenal side of the bilayer and the Ca_V1.2 VSD I/PD II interface. The cytoplasmic Ca_Vβ subunit binds to a site on the intracellular bulb of the EMC termed the “Cyto dock”. This protein–protein interaction has a similar sized footprint to the TM dock (~1000 Ų) and comprises two elements: a smaller site where the Ca_Vβ₃ α5-α6 helices contact EMC2 and a larger site where the Ca_Vβ₃ α3-α4 loop interacts with EMC8 (Figure 2D). Notably, the interactions made by Ca_V1.2 and Ca_Vβ are conserved throughout the seven Ca_V1 and Ca_V2 and four Ca_Vβ isoforms [24]. Biochemical purification and mass spectrometry show that in the absence of Ca_Vβ₃, Ca_V1.2 binds the EMC, and that Ca_Vβ ₃ does not bind the EMC on its own [24], indicating that Ca_Vβ is a critical element for stabilizing the EMC:Ca_Vα₁/Ca_Vβ complex. Importantly, mutations that disrupt the TM and Cyto dock sites reduce channel cell surface expression [24]. Further, the most disruptive Ca_Vβ₃ mutations cause a loss of EMC binding as measured by quantitative mass spectrometry [24]. Together, this functional and biochemical evidence indicates that the EMC:Ca_V1.2/Ca_Vβ₃ complex is a productive step in Ca_V assembly.

Expression of all three channel subunits Ca_V1.2, Ca_Vβ₃, and Ca_Vα₂δ-1 yielded purified samples in which the EMC: Ca_V1.2/Ca_Vβ₃ complex and fully assembled Ca_V1.2/Ca_Vβ₃/Ca_Vα₂δ-1 channels were found in similar proportions [24], enabling the first structural determination of Ca_V1.2 [24]. Comparison of the structures of the EMC-bound state with the assembled Ca_V1.2 channel [24] revealed two key points: (1) interaction with the EMC drives a set of rigid body movements that affect nearly every part of the Ca_V1.2/Ca_Vβ₃ complex (Figure 2E–F), and (2) the EMC and Ca_Vα₂δ bind to the same Ca_Vα₁ face in a mutually exclusive manner. Notably, key changes in the EMC-bound channel occur in the three elements that form the Ca_Vα₂δ contact points: the external part of VSD I and the PD II and PD III extracellular loops. Binding of Ca_Vβ to the EMC Cyto dock drives dramatic conformational changes in Ca_Vα₁ that involve a rotation of Ca_Vβ away from the membrane plane (Figure 2F). This change orchestrates a coordinated displacement of the Ca_V1.2 α-interaction domain (AID) that forms the high-affinity binding site for all Ca_Vβ isoforms [37-39], VSD II, and PD III. The AID/Ca_Vβ₃ unit tilts ~15° away from the membrane plane, displacing the AID C-terminal end by~11 Å relative to its Ca_V1.2/Ca_Vβ₃/Ca_Vα₂δ-1 position (Figure 2F) and enabling four AID acidic residues (D439, E445, D446, and D448) to make a new set of interactions with five basic residues (R507, R511, R514, R515, and R518) from the VSD II IIS0 helix. This AID segment is disordered in Ca_V1.2/Ca_Vβ₃/Ca_Vα₂δ-1 and other assembled Ca_V structures [40-48] and the interaction forms because IIS0 becomes more helical in the EMC complex. Hence, it appears that these electrostatic interactions provide a link that pulls VSD II and PD III away from the channel central axis as a consequence of the AID tilt, allowing the VSD III/VSD II/AID/Ca_Vβ assembly to act as a unit. The most striking outcome of these movements is the partial extraction of PD III from the core of the Ca_V1.2 pore. This change widens the pore diameter by~3 Å [24], renders PD III unable to make its native quaternary contacts with the other PDs, and leads to the destabilization of VSD III (Figure 2E). Remarkably, the PD III tertiary structure is preserved, underscoring the ability of PDs to fold autonomously [5]. Notably, the VSD III loop that interacts with Ca_Vα₂δ is unfolded. In addition to these changes, VSD I is rotated ~20° away from its position in the assembled Ca_V1.2 complex. Overall, these alterations dramatically reshape the channel in a way that splays open the Ca_Vα₂δ binding site, suggesting that the EMC acts as a holdase for the partially assembled channel and prepares it for handoff to Ca_Vα₂δ to complete assembly.

Based on the ability of the Ca_Vα₁ subunit to bind the EMC, the strong affinity of the Ca_Vα₁/Ca_Vβ complex for the EMC, structural differences between the EMC-bound and assembled states, and functional consequences of disruption of the EMC:Ca_V interface [24], we propose the following working model for Ca_V biogenesis and assembly. Following Ca_Vα₁ synthesis and membrane insertion by pathways that are not well understood and that may involve steps as outlined in (Figure 1C), Ca_Vα₁ binds to the EMC TM dock site. The stability of this complex is enhanced by the addition of Ca_Vβ forming the EMC:Ca_Vα₁/Ca_Vβ holdase complex, a step that could explain the ability of Ca_Vβs to protect Ca_Vα₁ pore-forming subunits from proteasome and ERAD pathway [49] degradation in a variety of cell types, including neurons [49,50]. It should be noted that it is also possible that pre-formed Ca_Vα₁/Ca_Vβ complexes directly bind the EMC providing a second route to the holdase complex. Designing experiments to probe these early steps of Ca_V biogenesis is an important goal for clarifying these pathways.

Holdase complex formation is followed by a handoff step in which the partially assembled Ca_Vα₁/Ca_Vβ channel is passed to Ca_Vα₂δ in a process that has two key components: the completion of a divalent ion binding site between Ca_Vα₂δ and VSD I and ordering of a set of Ca_V elements that comprise the Ca_Vα₂δ binding sites on PD II and PD III [24]. Whether this process occurs in an ordered way or via a “bind and release” mechanism is not known. However, it is interesting to speculate that there could be a concerted way in which the extracted PD III and availability of its extracellular loop, which is important for Ca_Vα₂δ binding [51] could serve as an initiation point for the handoff. The large PD III extracellular loops are disordered in the EMC:Ca_V1.2/Ca_Vβ₃ complex but make extensive interactions with Ca_Vα₂δ, indicating that folding of these elements is Ca_Vα₂δ-dependent. Movement of PD II and PD III to their native positions would seem to require the release of Ca_Vβ from the Cyto dock and disruption of the VSD II:AID interactions that are formed in the EMC complex. Rigid body rotation of VSD I between its EMC-bound and Ca_Vα₂δ bound positions completes the coordination sphere of the divalent ion shared by the Ca_Vα₂δ metal ion-dependent adhesion (MIDAS) site and VSD I (a site we term the “divalent staple”) [24] and completes the consolidation of Ca_Vα₂δ interactions with its three contact points on Ca_Vα₁. Ca_Vα₂δ coordination of divalent ions by the MIDAS site is critical for Ca_Vα₂δ binding to Ca_V1.2 [52] and Ca_V2.2 [53] and for the ability of Ca_Vα₂δ to promote Ca_V1.2, Ca_V2.1, and Ca_V2.2 plasma membrane trafficking [52-54]. This VSD I residue is conserved in Ca_V1s and Ca_V2s [24] and its mutation affects Ca_Vα₂δ-dependent Ca_V1 [52] and Ca_V2 [53] trafficking, supporting the idea that the EMC effects on VSD I conformation influence a key maturation step that is thought to occur in the ER lumen [54]. The result of the handoff would be formation of the native Ca_Vα₁/Ca_Vβ/Ca_Vα₂δ assembly and license of the channel to leave the ER to continue to the plasma membrane [12]. Currently, there is structural data for only two of the many possible intermediates in this proposed pathway (assembled Ca_V1.2 and the EMC: Ca_V1.2/Ca_Vβ complex) [24] (Figure 3). Hence, it will be important to probe this potential Ca_V assembly route to understand if other intermediates are involved, how key steps such as the association of the channel with the EMC and handoff to Ca_Vα₂δ take place, and whether there are other factors that participate in the assembly of the channel.

As the first example of an ion channel:chaperone complex and channel assembly intermediate the EMC: Ca_V complex naturally leaves more outstanding questions than it answers. Do other Ca_Vs and Ca_Vβs interact with the EMC similarly? Given the strong conservation of the Ca_Vα₁ and Ca_Vβ TM dock and Cyto dock interface residues, this possibility seems likely [24], but experimental validation will be important as it may reveal variations on the overall theme. Understanding the roles of key channel elements previously implicated in trafficking also poses important questions. The Ca_V1.2 C-terminal tail is implicated in trafficking [55] but both the full-length channel and Ca_V1.2 truncated just after the calmodulin-binding domain bound to the EMC similarly [24]. Deeper studies of the role of this channel segment as well as calmodulin, which also affects trafficking [56], are warranted to understand if these elements act at the EMC bound or later stage of channel maturation. Do some disease mutations and drugs cause functional effects by interfering with the formation of the EMC complex? Ca_Vα₂δ is the receptor for widely-used gabapentinoid anti-nociceptive and anti-anxiety drugs [16,57-60] thought to act by affecting the Ca_V numbers on the cell surface expression [57,61,62]. This type of mechanism demonstrates the power of controlling channel number to influence excitability and highlights the possibility that agents that perturb the EMC-bound step might offer a similarly potent means to affect the channel function. Developing agents to probe EMC:Ca_V interactions will be an important tool for addressing such issues. EMC:Ca_V interactions affect channel expression in heterologous cell systems routinely used to assess channel mechanism, mutations, and drug action [24]. Given the ubiquitous expression of the EMC, one would like to know whether EMC: Ca_V complexes form in native cells.

How general is the example set by the EMC: Ca_V complex? Many VGIC superfamily members share the core VSD architecture that interacts with the TM dock site, raising the possibility that other VGIC classes are EMC-dependent. Defining whether other channels interact with the EMC in ways that are similar to Ca_Vs is a key question. Although Ca_Vβ is unique to the Ca_V subfamily, it is not difficult to imagine that the large intracellular domains found in many VGICs could interact with the Cyto dock. In this regard, it will be interesting to address whether the Ca_V3 subfamily [3] that does not depend on either Ca_Vβ or Ca_Vα₂δ is EMC-dependent. Finally, the EMC has a multitude of diverse clients [27]. The EMC:Ca_V structure provides the first example of an EMC:client complex. Understanding whether other types of membrane proteins use the TM dock and Cyto dock sites or use other sites, such as the EMC3/4/7 interface found to bind the outer mitochondrial membrane protein VDAC [63], is an important avenue of study.

Do other chaperones act as holdases for channel subunits? The recent report of the structure of a GABA_A receptor subunit with a chaperone, NACHO, that acts on a subset of pentameric ligand-gated ion channels shows that this ER-resident transmembrane protein sequesters a key intermembrane interface in the absence of the other subunits required to make the functional channel [64]. This exciting finding has parallels with the EMC:Ca_V complex that highlights the role of chaperones in protecting partially assembled states of ion channels and underscores the need to define both general and specific mechanisms by which chaperones aid ion channel assembly.

Assembly from parts is the first step in the life of every ion channel. The door is now open to explore the structural principles that underlie how these marvelous entities at the core of the bioelectric signals that are required for life are put together from their various components. Probing the principles that govern ion channel assembly will not only fill a largely blank space in our knowledge of ion channel structure and function, but will offer new opportunities to understand how channel function goes wrong in disease states and inspire the development of new ways to control channel function.

The authors declare no competing interests.

This work was supported by the NIH grants (R01 HL080050 and NIH R01 DC007664 to D.L.M.).

Z.C. and D.L.M. wrote the paper.

We thank K. Brejc for comments on the manuscript.

AID

α-interaction domain

Cav

voltage-gated calcium channel

EMC

endoplasmic membrane protein complex

ER

endoplasmic reticulum

MIDAS

metal ion-dependent adhesion site

PD

pore domain

SF

selectivity filter

VGIC

voltage-gated ion channel

VSD

voltage sensor domain