Designer nanodiscs to probe and reprogram membrane biology in synapses

Abstract

Signal transduction at the synapse is mediated by a variety of protein-lipid interactions, which are vital for the spatial and temporal regulation of synaptic vesicle biogenesis, neurotransmitter release, and postsynaptic receptor activation. Therefore, our understanding of synaptic transmission cannot be completed until the elucidation of these critical protein-lipid interactions. On this front, recent advances in nanodiscs have vastly expanded our ability to probe and reprogram membrane biology in synapses. Here, we summarize the progress of the nanodisc toolbox and discuss future directions in this exciting field.

Development of the nanodisc technology

Synapses are highly polarized intercellular junctions, mediating the flow of information from a pre- to a postsynaptic neuron¹. As they are compartmentalized by membranes, the transfer and computation of synaptic information rely on the orchestration of numerous membrane proteins embedded in or peripherally associated with the lipid bilayer. The exterior of a membrane is hydrophilic, but the interior is hydrophobic as it consists of aliphatic chains. Accordingly, membrane proteins exhibit complex biochemical properties, containing both hydrophobic domains to interact with the lipids and hydrophilic regions to contact water-soluble molecules, presenting a formidable challenge for structural and functional studies. Traditionally, detergents are utilized to extract these proteins from the membrane for biophysical characterizations and thus often inevitably disrupt their native conformational states due to the removal of lipids. Alternatively, membrane proteins could be reconstituted into liposomes for functional studies. However, the resulting proteoliposomes are insoluble and not amenable to many biophysical approaches. This dilemma was addressed using an elegant design of lipid nanoparticles by Dr. Sligar and coworkers in 2002, known as the nanodisc (ND; Fig. 1A)^{2, 3}, in which membrane proteins could be incorporated in nanoscale lipid bilayers yet remain soluble in an aqueous solution without the need of detergent.

Figure 1. Illustration of the nanodisc technology.

(A) Architecture of the most used nanodisc enclosed by MSP1, which has a diameter of ~10 nm (Ref. 2). Coordinates of the nanodisc structure were from Mazhab-Jafari et al.¹⁰⁴. Upper, side view; Bottom, top view. (B) Construct designs to increase the diameter of nanodiscs. MSP1 is composed of ten amphipathic helixes from ApoA1. By genetic fusion of three (NW30) or five (NW50) copies of MSP1 together^{8, 9}, the diameter of nanodiscs can be increased to 30 and 50 nm, respectively.

The nanodisc technology is inspired by a family of amphipathic proteins called apolipoproteins that mediate lipid transport in the form of high-density lipoprotein (HDL) in all mammals. One of the best-characterized apolipoproteins is Apo-A1, which has been shown to form 10-30 nm lipid nanoparticles in a variety of structures through the interaction between lipids and a set of amphipathic helices of the protein⁴. In addition, early studies have also demonstrated that the discoidal form of HDL particles could provide a soluble supported lipid bilayer for the incorporation of membrane proteins⁵. However, the reconstitution of membrane proteins in HDL is difficult due to the existence of a large degree of heterogeneity. Through extensive protein engineering to carefully delete its N-terminal region and rearrange individual amphipathic helices⁶, the engineered protein, now known as membrane scaffold proteins (MSPs, Fig. 1B) could enclose homogenous 6-50 nm nanodiscs^7-9. The principle of the nanodisc framework is quite simple: the lipid bilayer is usually 5 nm in thickness and thus requires two copies of MSPs to encircle each leaflet of the lipid bilayer (Fig. 1A). As such, the diameter of nanodisc is defined by the size of MSPs. For example, the initial design of MSP1D1 generates ~10 nm nanodiscs. By fusing three (NW30) or five (NW50) copies of MSP1D1 together (Fig. 1B), 30 or 50 nm nanodiscs could be produced^{8, 9}. However, these large nanodiscs are not stable and often quite heterogeneous. To overcome these difficulties, Dr. Wagner and coworkers first engineered circularized nanodiscs by covalently connecting the N- and C-termini of MSPs via the sortase-mediated protein ligation⁸, thereby vastly improving the stability and homogeneity of nanodiscs.

Because of the enhanced stability and solubility in a native lipid environment, nanodiscs have become a powerful platform for structural and functional studies of membrane proteins. The remarkable performance of this technology has motivated several groups to identify other proteins¹⁰, peptides^{11, 12} and polymers^{13, 14} that could also form the nanoscale discoidal structure, greatly broadening the spectrum of the nanodisc toolbox. In this review, we mainly focus on the work using MSP-based nanodiscs for the characterization of protein-membrane interactions in synapses. For those with interest in nanodiscs encased by peptides and polymers, we refer them to several other excellent recent reviews^{15, 16}.

Characterization of integral membrane proteins in synapses using nanodiscs

Synaptic transmission in neuronal circuits requires information processing on a millisecond timescale, which is enabled by the coordinated action of a variety of integral membrane proteins (iMPs). Mechanistic dissection of iMPs faces several challenges that can be addressed using nanodiscs¹⁷. First, the normal function of many iMPs often relies on their native lipid environment, thus necessitating a need to accommodate them in membranes. By virtue of the enclosed lipids, nanodiscs are excellent tools to study the impact of membranes on the conformational dynamics of iMPs, as demonstrated in a diverse array of protein families³, ranging from transporters to ion channels. In conjunction with single-particle cryoEM, individual lipid molecules and their engaging transmembrane segments could be well-resolved, allowing for the direct comparison of iMP structures in nanodiscs made with different lipids. Not surprisingly, previous studies have shown that structural dynamics of several iMPs are profoundly regulated by the interacting lipids. For example, an elegant work from Gao et al. was the first to demonstrate the power of integrating nanodisc technology with single-particle cryoEM¹⁸, which showed that the sensing of the TRPV1 channel to ligands or thermal stimuli entails the release of the associated phosphatidylinositol lipids. More recently, Zhang et al. showed that the membrane tension in nanodiscs could be specifically manipulated, revealing a critical function of lipids in coupling mechanic forces in lipid bilayers to the opening and closing of the mechanosensitive ion channel MscS¹⁹. In another interesting study, Nadezhdin et al. discovered that the temperature-induced opening of TRPV3 was only observed using the circularized nanodisc, presumably due to the difference in its lipid environment as compared to the non-circularized one²⁰. Therefore, circularized large nanodiscs might enable the characterization of conformational states of iMPs unattainable in previous studies. Over the past few years, structures of many synaptic ion channels and GPCRs have also been resolved in nanodiscs^{7, 21-23}, corroborating the importance of lipids in the regulation of iMPs. In line with these structural studies, Marty et al. utilized high-resolution native mass spectrometry to dissect the physical and chemical principles of iMP-lipid interactions reconstituted in nanodiscs²⁴, enabling precise quantification of the stoichiometry and the accompanied energetics of these interactions. Recent progress in native mass spectrometry can further reveal how the lipid microenvironment regulates the association and dissociation of membrane protein complexes involved in a rhodopsin receptor signaling cascade²⁵. As the functional assembly of many membrane protein complexes depends on lipids, nanodiscs can serve as powerful tools to investigate the underlying molecular mechanism.

Second, a large fraction of membrane protein-protein and protein-ligand complexes are only stable in lipid bilayers and are thus often disrupted by detergent solubilization, excluding them from biophysical studies. On this front, nanodiscs provide a perfect solution to characterize these detergent-sensitive protein complexes. Pioneering studies from the Sligar group demonstrated that transducin binding to rhodopsin is ~10 fold more potent than that measured in detergents, similar to the estimated values from cell-based assays²⁶. In addition, the membrane proteome can be reconstituted in nanodiscs, from which transient membrane protein complexes were captured^27-29, facilitating drug discovery against specific cell types. Moreover, the superior stability of nanodiscs is essential for the isolation and structural characterizations of the mGlu5 receptor heterodimer in the apo state²⁹, enabling the establishment of a mechanistic framework for receptor activation through the conformational propagation from the Venus flytrap to the 7TM domains, which reorient into close proximity that may be a hallmark of activation in this protein family. Consistently, it is also found that lipid bilayers in the circularized nanodisc are critical to stabilize the formation of the NTS–NTSR1–Gαi1β1γ1 complex for cryoEM studies³⁰.

Finally, nanodiscs can readily isolate individual membrane protein complexes for biophysical interrogations. Over the past two decades, advances in single-molecule biophysics have significantly improved our ability to elucidate the molecular mechanism of numerous cellular machineries³¹, from DNA replication to protein translation. However, the application of these single-molecule approaches for studying iMPs in a lipid environment is challenging, largely due to their tendency to oligomerize and cause intermolecular interference that confounds data analysis³². These issues can be alleviated by the physical separation of single iMPs in nanodiscs to improve data acquisition and processing. Furthermore, nanodiscs confer access to both sides of the membrane for the study of transmembrane signaling mechanisms through single-molecule imaging. As such, it was found that the β2-adrenergic receptor (β2AR) spontaneously samples both active and inactive conformations in the lipid bilayers. Binding of different ligands on the extracellular surface of β2AR shifts the conformational distribution of the cytoplasmic surface to either conformation, thereby eliciting distinct functional outcomes through the same receptor³³. With the continuous improvement in biophysical methods to capture high-energy and transient states in protein dynamics³⁴, we believe that nanodiscs will tremendously help to delineate the conformational transitions of iMPs at a molecular level.

Mechanistic dissection of fusion pores using nanodiscs

An essential pillar to the precision of synaptic transmission is the release of neurotransmitters into the synaptic cleft through the exocytic fusion pore, a critical intermediate state formed between the plasma membrane and secretory vesicles. Using electrophysiology and microscopy approaches^35-40, an omega-shape structure of fusion pores has been captured in vivo for quantitative analysis. Although these pores are very transient states and only exhibit a diameter of 1-2 nm in the initial open state, they are the central hub in mediating the transfer and processing of information along neuronal circuits⁴¹. Therefore, alterations of pore properties have a substantial impact on the outcome of synaptic transmission, causing differential activation of postsynaptic receptors^{42, 43}. Despite a large body of cell-based studies, the underlying molecular mechanism that determines the size and lifetime of fusion pores remains unclear.

A simple possibility is that the components forming fusion pores dictate their properties, which can be directly interrogated using bottom-up reconstitution approaches. In the early 1990s, seminal work from the Rothman group identified the soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein receptor (SNARE) proteins that directly catalyze the fusion between vesicles and the plasma membrane in the brain^44-47. Using purified proteins reconstituted in liposomes, it is well established that vesicle SNAREs (v-SNAREs) and their cognate target membrane SNAREs (t-SNAREs) form trans-SNARE complexes that drive membrane fusion^47-49. The assembly specificity of trans-SNARE complexes is encoded in their inherent physicochemical properties; only complementary v- and t-SNAREs would pair between lipid bilayers, maintaining the identity of cellular compartments. Further structural and bioinformatic analysis revealed that all SNARE proteins have a typical heptad repeat motif of ~60-70 residues^{50, 51}, termed the SNARE motif. Extensive biochemical and structural studies have reached a consensus that four cognate SNARE motifs assemble into extremely stable four-helix bundles⁵², which are composed of mostly leucine-zipper-like layers, providing the energy to drive membrane fusion.

For synaptic vesicle exocytosis, membrane fusion is mediated by synaptobrevin2 (syb2) on the vesicles and syntaxin-1A (syx1A) and SNAP25 on the plasma membrane^{53, 54}. syb2 and syx1A are tail-anchored transmembrane proteins, while SNAP25 is localized on the plasma membrane through four palmitoylated cysteines. Membrane fusion can readily occur by incubation of one set of liposomes harboring syb2 and another set of liposomes bearing syx1A and SNAP25⁴⁷. The high fusion efficiency of this minimized system prompted considerable efforts from many groups to reconstitute nascent fusion pores for mechanistic dissections. Most of these studies employed reconstitution approaches to monitor the content exchange between two separate populations of liposomes harboring v- and t-SNAREs^55-57. In these assays, pores tend to expand quickly, and only limited information can be obtained regarding nascent fusion pores. Nevertheless, these early efforts showed that the fusion pore in the initial open state is indeed a short-live intermediate state with a diameter of 1-2 nm, in agreement with the findings from cell-based assays.

Because of the ease of changing the components in reconstituted assays, it is quickly found that the transmembrane domains (TMDs) of SNAREs are not essential to drive the fusion of lipid bilayers⁵⁸. Moreover, reconstitution of membrane fusion could even take place using complementary DNA molecules anchored on the surface of two opposing liposomes⁵⁹, or using highly fusogenic lipids alone without additional proteins⁶⁰. Based on these observations, the hemifusion hypothesis is proposed that SNARE-mediated membrane fusion occurs through a nascent fusion pore formed solely by lipids - the lipidic fusion pore model⁶¹. In this model, the pairing of trans-SNAREs brings two opposing membranes closer and provides the energy for fusion through the zippering of the SNARE motif. This model was later further supported by theoretic modeling and calculations^{62, 63}. However, amperometry and electrophysiology measurements suggest that currents of fusion pores were altered by mutations specific on one side of the TMDs of syb2 and syx1A^64-67, similar to the gap junction ion channel, evoking the possibility of a proteinaceous pore model. Due to the technical difficulty of isolating the nascent fusion pore, its structure is still a subject of debate.

A breakthrough came in 2012 again by the Rothman group, establishing the use of nanodiscs to isolate nascent fusion pores⁶⁸. By virtue of the rigid framework of nanodiscs, fusion pore dilation is prevented and thus trapped in the initial early stage, allowing for in-depth investigations. In addition, the well-defined size of nanodiscs allows for systematic investigations in the composition of the reconstituted fusion pore. For example, the lipidic fusion pore will not be able to form using 6 nm nanodiscs, as it will at least require a 10 nm nanodisc to accommodate two lipid bilayers. However, we observed pore formation in 6 nm nanodiscs⁶⁹, indicating that the nascent fusion pore is not purely lipidic in this reconstituted system. Interestingly, we further found that two copies of SNAREs were sufficient to form a pore large enough to allow the release of a bona fide neurotransmitter, glutamate. Therefore, the pore is neither solely formed by the TMDs of SNARE proteins, because two TMDs are too few to form a proteinaceous gap-junction-like structure. In conjunction with chemical probing, we showed that reconstituted fusion pores are most likely composed of lipids and the TMDs of SNARE proteins, thereby prompting the hypothesis of a hybrid fusion pore model (Fig. 2A). This model is also independently validated using molecular dynamics simulation by Sharma et al.^{70, 71}.

Figure 2. Probing the structure and dynamics of fusion pores using nanodiscs.

(A) A hybrid fusion pore reconstituted in nanodiscs. Left, the fusion between nanodiscs and vesicles results in pore opening and glutamate release, monitored using iGluSnFR¹⁰⁵. Right, residues of SNARE TMDs lining the fusion pore formed between nanodiscs and vesicles are highlighted in red. (B) Illustration of the single fusion pore assay. Fusion pores formed between nanodiscs and planar lipid bilayers could be dissected using electric recordings with high temporal resolutions. Figures were adapted from Ref. 66 and 69 with permission.

The chemically well-defined nascent fusion pores trapped in nanodiscs opened many possibilities, as they can be readily interrogated by a panel of biophysical approaches. Combining nanodisc reconstitution with single-molecule electrical recording^{72, 73}, we captured the dynamics of single fusion pores, showing that the size and dynamics of fusion pores are a result of the intrinsic characteristics of the SNARE proteins (Fig. 2B). Further, we find that the zippering of SNAREs in the early stage is unstable and even reversible, causing the flickering behavior of fusion pores. Increasing the copy number of SNAREs in nanodiscs stabilizes the pore structure and expands its size. In addition, the open accessibility of this simple assay allows for the characterization of several regulatory proteins involved in synaptic transmission. Not surprisingly, regulatory proteins can stabilize and expand the fusion pore by promoting SNARE zippering or destabilizing the lipid bilayer to overcome the energy barrier of pore formation^{73, 74}. Similar observations have also been reported in reconstituted assays that characterized fusion pores formed between nanodiscs harboring v-SNAREs and cells displaying t-SNAREs^{75, 76}. Finally, it is also feasible to isolate a partially assembled trans-SNARE complex between two nanodiscs^{73, 77}, enabling the investigation of regulatory proteins on the zippering of SNAREs through single-molecule imaging. Since many new approaches are underway in single-particle cryoEM studies of membrane proteins, we expect to see the structure of the fusion pore reconstituted in nanodiscs at atomic resolution in the coming years.

Nanodiscs for structural and functional characterizations of peripheral membrane proteins

In many cases, transmembrane signaling reactions are relayed through peripheral membrane proteins (pMPs), and synaptic transmission is no exception. Previous studies have identified a large number of pMPs essential for the speed and amplitude of synaptic transmission^{51, 53, 54, 78}. Disturbances in many of them (e.g., α-synuclein) are implicated in brain diseases⁷⁹. Although nanodiscs have been widely used for the characterization of iMPs, the potential of nanodiscs for the study of pMPs has not been much explored and only started to emerge as a viable option. Because of the small size of 10 nm nanodiscs, the interaction of α-synuclein (α-syn) with lipids could be dissected using single-molecule imaging⁸⁰ and NMR⁸¹. The results revealed the molecular detail of how lipids contribute to the folding and aggregation of α-syn on membranes. These molecular understandings of α-syn and lipid interactions might help to promote the development of targeted therapeutics for Alzheimer’s diseases (AD) and related dementia, as these interactions are involved in the phase transition and membrane damage of the pathogenic α-syn aggregates^82,83.

However, many pMP-lipid interactions are difficult to capture using 10 nm nanodiscs, perhaps due to the limited membrane surface. In theory, these problems could be resolved using larger nanodiscs with increased diameters by engineering MSPs. Unfortunately, it is challenging to make larger ones due to the dynamics of MSPs. Recently, we expanded the size and function of nanodiscs to tackle this challenge. First, we grafted proximity labeling (PL) enzymes onto nanodiscs⁸⁴, so that transient protein-membrane interactions could be readily detected (Fig. 3A). PL generates reactive biotins to label proteins within a distance of 10-20 nm. The labeled proteins can then be identified using an array of different biochemical approaches. Second, we increased the size range of nanodiscs from 10 to over 100 nm through circularization of large MSPs via SpyCatcher-SpyTag⁹ or split GFP⁸⁵, thereby enlarging their surface areas to capture the action of pMPs on lipid bilayers (Fig. 3B). The design of these nanodiscs was inspired by Nasr et al. showing that the stability and monodispersity of large nanodiscs can be vastly enhanced using circularized MSP by sortase⁸. However, circularization of MSPs via the sortase-mediated protein ligation method is time-consuming and suffers from low yields. Using SpyCatcher-SpyTag, we greatly simplified the process of building circularized large nanodiscs with much-improved yields. These circularized nanodiscs exhibited remarkable stabilities^{8, 9}, and might become useful tools for the characterization of temperature-dependent conformational changes of membrane proteins²⁰. Moreover, these nanodiscs enabled facile characterizations of the interaction between several synaptic proteins with membranes, including synaptotagmin and α-syn. Using the GFP circularized nanodiscs⁸⁵, we further created a robust fluorescent probe to detect the structural transition of lipid bilayers caused by a panel of pMPs such as α-syn, sar1p, synaptotagmin-1 and pore-forming proteins (Fig. 3C). This sensor also allowed us to identify an unknown membrane remodeling activity of complexin, which is contributed by its C-terminal amphipathic helix. In conjunction with our single fusion pore assay (Fig. 2B)⁷², we showed that this membrane remodeling activity of complexin could locally destabilize the structure of lipid bilayers and thus promote the formation of fusion pores⁷⁴. Interestingly, we note that amphipathic helices are found in many other fusogenic proteins^86-88, but are conspicuously absent in SNARE proteins. The amphipathic helix in complexin plays a key role in synaptic vesicle exocytosis and may correspond to a fusion peptide that facilitates SNARE-mediated membrane fusion. As many more pMPs are involved in synaptic transmission, we believe that recent advances in the nanodisc toolbox could markedly accelerate mechanistic understandings of this exciting yet untapped field.

Figure 3. Expanding the structure and function of nanodiscs.

(A) Schematic of nanodiscs functionalized with proximity labeling (PL) enzymes for probing protein-membrane interactions. (B) Circularization of nanodiscs via SpyCatcher-SpyTag (spND). Left, schematic of the spMSP construct (top) and circularized nanodiscs (bottom). Right, negative stain EM micrographs of the spNDs with diameters ranging from 11 to 100 nm. Scale bar, 50 nm. (C) Schematic of split-GFP circularized nanodiscs for probing protein-lipid interactions. Top, illustration of the iGlu-MSP construct. Bottom, illustration of iGlu-MSP nanodiscs for reporting membrane bending by synaptotagmin-1 (grey). These nanodiscs were circularized by the complementation of the split circular permutated GFP (cpGFP) and further stabilized by the two halves of the glutamate binding protein (Glu). Figures were adapted from Refs. 9, 81, and 82 with permission.

Translational applications of nanodiscs for neurological disorders

In addition to its excellent performance in probing basic research of membrane biology, nanodiscs are also useful in steering the immune response. This is largely attributed to its nanoscale size, within the range of 10-100 nm, which can easily penetrate different tissues without being quickly removed by circulation, exhibiting the so-called enhanced permeability and retention effect of lipid nanoparticles¹⁵. Furthermore, nanodiscs can boost immune responses by promoting the clustering of antigens and the co-delivery of different adjuvants. Thus, it is not surprising that enormous efforts have been made to test the efficiencies of nanodiscs for drug delivery. For example, it has been shown that nanodiscs harboring neoantigens could potently potentiate the efficacy of PD-1/PD-L1 antibodies⁸⁹, especially in solid tumors where current immunotherapy regimes still pale. Moreover, immunogenic molecules presented by ApoA1-encased nanodiscs⁹⁰, termed nanobiologics, can manipulate the trained immunity that might halt neurodegenerative diseases. Through the interaction between ApoA1 and the ATP-binding cassette transporter A1/G1, nanobiologics bind myeloid cells and their progenitors with high affinity, and can metabolically or epigenetically reprogram the trained immunity by activating the nucleotide-binding oligomerization domain-containing protein 2 (NOD2) receptor using muramyl dipeptide (MDP) derivatives⁹¹. These nanobiologics will bring about a paradigm shift in treating neurological disorders because they do not need to cross the blood-brain barrier.

Besides immunotherapy, nanodiscs made using apolipoproteins, also known as HDL-mimics, can be useful for treating neurodegeneration through several other mechanisms. First, HDL suppresses the pathological accumulation of Aβ amyloids by promoting their clearance from the brain⁹². Consistently, a growing body of evidence in humans and mice have also found that circulating HDL levels are associated with AD risk^{93, 94}. In particular, ApoE4, among the three isoforms of ApoE, is the dominant genetic risk for AD and is inefficient in clearing Aβ amyloids⁹⁵. Second, the structure and lipid composition of HDL has a critical role in controlling neuroinflammation by blocking the action of pro-inflammatory cytokines⁹⁶. Finally, HDL stimulates the production of nitric oxide from brain endothelial cells^97-99, thereby preventing oxidative stress in neurons. Therefore, several groups have reported the use of engineered HDL-mimics from ApoA or ApoE to reduce brain damage in animal models of AD^{100, 101}. Although few have entered clinical trials for AD, two HDL formulations are in advanced phase clinical trials for cardiovascular diseases^{102, 103} and the outcome of these trials will shed new light on the therapeutic development for neurological disorders.

Concluding remarks and future directions

In the past two decades, nanodiscs and many other lipid nanoparticles have been tailored to study membrane proteins of different sizes and properties. The rich resources in this toolbox should be able to further extend our ability to probe the function of lipids in the regulation of membrane proteins involved in synaptic transmission. Combining with recent advances in single-particle cryoEM, we expect to see structural elucidations of membrane protein complexes that were unable to be resolved in previous studies. In conjunction with the emerging breakthroughs of biophysical approaches, ‘hard-to-see’ conformational dynamics of membrane proteins will be captured in greater details. Moreover, it seems that the nanodisc framework has remarkable plasticity for further engineering to stretch its geometry and function, which will be useful for both basic and translational studies of protein-lipid interactions. Together, a far more exciting golden age of nanodiscs is ahead of us to better understand and steer membrane biology in synapses.