A proteome-wide quantitative platform for nanoscale spatially resolved extraction of membrane proteins into native nanodiscs

Abstract

The native membrane environment profoundly influences every aspect of membrane protein (MP) biology. Despite this, the most prevalent method of studying MPs uses detergents to disrupt and remove this vital membrane context, impeding our ability to decipher the local molecular context and its effect. Here we develop a membrane proteome-wide platform that enables rapid spatially resolved extraction of target MPs directly from cellular membranes into native nanodiscs that maintain the local membrane context, using a library of membrane-active polymers. We accompany this with an open-access database that quantifies the polymer-specific extraction efficiency for 2,065 unique mammalian MPs and provides the most optimized extraction condition for each. To validate, we demonstrate how this resource can enable rapid extraction and purification of target MPs from different organellar membranes with high efficiency and purity. Further, we show how the database can be extended to capture overexpressed multiprotein complexes by taking two synaptic vesicle MPs. We expect these publicly available resources to empower researchers across disciplines to efficiently capture membrane 'nano-scoops' containing a target MP and interface with structural, functional and bioanalytical approaches.

Fig. 1. Generation of MAP library and development of high-throughput bulk solubilization assay.

a, Representative chemical structures of commercially available and in-house synthesized MAPs. The diversity of MAP structures is manipulated by varying hydrophobic and hydrophilic moieties, their ratios, functionalization of R groups and polydispersity. b, The general synthesis scheme for RAFT-based polymerization of MAPs. PADTC, poly-ammonium dithiocarbamate; DMF, dimethyl formate; ACHN, 1,1'-azobis(cyanocyclohexane). c, A schematic of the fluorescence bulk solubilization assay. Fusogenic liposomes are used to deliver fluorescent lipids to living cells labeling the membranes with a fluorophore. The membranes are then collected, MAP solubilized and subjected to a range of quality control experiments to determine bulk membrane solubilization through fluorescence measurements and size distribution and homogeneity through dynamic light scattering and negative-stain EM. DLS, dynamic light scattering; Fl, fluorescence. d, A schematic of bulk solubilization quantitation. The membranes labeled with fluorescent lipids are solubilized with MAPs, and a fluorescence reading (fl1) is taken before subsequent quenching with dithionite. Post quenching, a second fluorescence reading (fl2) is taken. The precise nanodisc extraction efficiency is calculated using equation (1). e,f, A bulk solubilization screen of polymer library against HEK293 cells (e) and HeLa cells (f), where before solubilization fluorescent lipids were delivered to the cell through fusogenic liposomes. Subsequently, the cells were solubilized using respective MAPS and the percent solubilizations were calculated using the procedure detailed in c and d. Here, 100% solubilization would indicate that the entire membrane was extracted into MAPdiscs, as 100% corresponds to the total starting fluorescence of the membrane. The error bars indicate the s.d. across the average value of three solubilization replicates. Panels c and d created with BioRender.com.

Fig. 2. Integrated extraction workflow and proteome-wide qualitative assessment of MAP extraction.

a, A schematic of experimental workflow for MAP extraction and proteomics analysis. The cellular membranes are solubilized with the MAP library. From the MAPdiscs isolated through ultracentrifugation, the proteins are precipitated using MTBE extraction and subsequently reduced, alkylated and digested overnight with trypsin. The tryptic peptides are subjected to label-free quantitative proteomics through liquid chromatography–tandem mass spectrometry. b, The number of MPs detected in each extraction condition, providing an overview of efficacy of MAPs in extracting MPs. c, Overlap of protein coverage from different extraction conditions. The polymers are divided into three broad classes (SMA: SMA200 and SMA300; AASTY: AASTY645, AASTY650, AASTY1145 and AASTY1150; and ChloroSMA: ChloroSMA40, ChloroSMA60 and ChloroSMA80). As depicted, while there is an overall robust overlap between different conditions, 83 MPs are only detected under the MAP conditions. d, MPs are detected from all membrane-bounded cellular organelles yielding robust organellar coverage. The MPs detected in the proteomic screen were filtered through organellar databases obtained from Uniprot; only proteins with specified organellar residence were annotated, and the proteins with more than one organelle of residence were annotated for each organelle. Organellar analysis reveals that MAP solubilization is not limited to the surface proteome, as proteins were robustly detected from all endogenous intracellular organelles. e, The variation in label-free quantitation was measured for each protein across each biological replicate. The biological replicates were plotted against each other to capture reproducibility of solubilization. Here, an R² value closer to 1 indicates better reproducibility. As shown, most MAPs show a very high degree of reproducibility. Interestingly, AASTY1150, which showed poor bulk solubilization, also showed poor reproducibility. Panels a and d created with BioRender.com.

Fig. 3. Quantitative analysis of MAP extraction efficiencies reveals solubilization preferences and polymer-specific trends.

a, A solubilization plot from the open-source web application taking EGFR as a representative MP. The y axis shows the relative solubilization efficiency of EGFR under each MAPs used, as designated on the x axis. On the basis of the normalized solubilization efficiency, SMA200 is the optimal extraction condition for the protein. b, The number of unique MPs best solubilized under each MAP and detergent condition. For example, for CS80, the number 357 denotes that CS80 is the best polymer for extraction for 357 unique MPs. As seen, three in-house synthesized polymers far outperformed commercial polymers in quantitative extraction efficiency. c, Stratification of the best-solubilized proteins by molecular weight reveals distinct population distributions. The detergents exhibit superior performance in extracting small molecular weight proteins (<40 kDa), while MAPs demonstrate efficacy in solubilizing medium to large molecular weight proteins (>40 kDa). Pop, population. d, Stratification of the best-solubilized proteins by number of transmembrane domains reveals distinct population distributions. The detergents preferentially solubilize proteins with small numbers of transmembrane helices (TMD 1–5), while MAPs excel at extracting proteins with larger numbers of TMDs (>5). e, A detailed breakdown of organellar solubilization preferences across the MAP library, revealing distinct polymer-specific trends in extraction. f, A hierarchical clustering analysis provides grouping of MAPs by extraction efficiencies, yielding a deeper understanding of polymer-specific solubilization trends.

Fig. 4. Database-driven extraction of intraorganellar MPs and determination of oligomeric state through NativeNanoBleach analysis.

a, The database provided optimal extraction conditions for five organellar MPs used as a benchmark. The red asterisk indicates the specific polymer that yields the best extraction. The corresponding best-performing MAPs (indicated with red star) are chosen for extracting the target MPs. b, The experimental extraction efficiency of each of the five GFP-tagged MPs extracted under the optimal MAPs, directly from their native organellar membranes to preserve native membrane environment (OMP25: SMA200; TGN46: CS60; VAPA: AASTY650; TMEM192: AASTY650; KRas: CS80). The error bars indicate the s.d. across the average of three solubilization replicates. c, A schematic of the NativeNanoBleach workflow. The glass slides are coasted with a PEG:poly-ʟ-lysine (PLL) PEG:biotin conjugate. The surface is sequentially coated with streptavidin and a biotinylated GFP–nanobody, which immobilizes the GFP-tagged disc on the surface. This permits rapid isolation and immobilization of GFP-tagged target MP-containing discs directly from the MAP-solubilized lysate membrane, enabling downstream TIRF microscopy and step-photobleaching analysis. d, An oligomeric distribution for GFP–TGN46 derived through NativeNanoBleach (NNB). The error bars indicate the standard deviation across the average of three biological replicates for NNB analysis. The protein was extracted using CS60. The polymers from the database are 1, SMA200; 2, SMA300; 3, ChloroSMA20; 4, ChloroSMA40; 5, ChloroSMA60; 6, ChloroSMA80; 7, AASTY645; 8, AASTY650; 9, AASTY1150; 10, AASTY1145; 11, AASTY80. Panels a and c created with BioRender.com.

Fig. 5. Native nanodisc extraction and purification of the Syp–vamp2 cocomplex.

a, A schematic of FLAG–SYP–VAMP2 complex purification. FLAG–SYP and VAMP2 are coexpressed in EXP293T cells and solubilized using ChloroSMA80 (CS80) as guided by the database solubilization index. The MAPs containing FLAG–SYP are affinity purified using affinity pulldown. b, The calculated solubilization index for the SYP–VAMP2 complex. The CholorSMA (CS) series of polymers, boxed in pink, performed well as a group, with CS80 being chosen for downstream purification because of its propensity to form larger discs. c, A negative-stain EM analysis of discs formed by CS80. Based on the population size distribution, CS80 forms discs of larger diameters than any other polymer in the CS series (Extended Data Fig. 5). For the best chance of capturing the oligomeric complex of SYP–VAMP2, CS80 was chosen due to this propensity to form larger discs. d, A Coomassie-stained gel shows the presence of two bands, one corresponding to the molecular weight of SYP and the second corresponding to the molecular weight of VAMP2. This purification as repeated twice showing presence of VAMP2, which does not have the Flag tag, in the pulldown indicates successful cocomplex purification. Panel a created with BioRender.com.

Extended Data Fig. 1. Comparative extraction efficiency of AqpZ and MAP extracted membrane protein structures.

a, Schematic detailing GFP-based calculation for protein-specific solubilization efficiency. Membranes expressing a GFP-tagged target protein are harvested and an initial fluorescence measurement (fl1) is taken. Membranes are MAP solubilized and subjected to a 200,000 g ultracentrifugation to remove insoluble material, and a second fluorescence measurement (fl2) is taken. Quantitative protein solubilization is calculated as the ratio of fl2:fl1. b, Extraction efficiency of model membrane protein Aquaporin Z into a traditional detergent system using widely cited polymer SMA200. Using standard conditions from the literature, SMA200 solubilizes poorly at 29% while standard detergent DDM solubilizes at 91%. B, Analysis of the last 150 deposited membrane protein structures to the PDB (9/7/23) reveals that only 2 structures were solved using polymer extraction. This highlights the current extremely limited utility of these native nanodisc-forming polymers which serve as a last resort for structural determination when other methods have failed. Panel a created with BioRender.com.

Extended Data Fig. 2. General polymer structure.

Table to show commercial (Com.) and homemade (HM) polymers. RAFT is noted for polymers that are synthesized using random addition fragmentation technology and CSTR is noted for polymers that are synthesized using continuous stirring tank reactor technology.

Extended Data Fig. 3. In vitro characterization of CS polymer series and quality control of samples through the proteomic pipeline.

a, Bulk solubilization of fluorescent liposomes in vitro to characterize solubilization efficiency of in-house CS series of polymers. b, Percent solubilization of CS polymer series as determined by liposome solubilization assay. The error bars indicate the standard deviation across the average value of two solubilization replicates. c, Representative negative stain EM image of discs in the preparation pipeline for proteomic analysis. d, Representative dynamic light scattering analysis of discs in the preparation pipeline for proteomic analysis. Panel a created with BioRender.com.

Extended Data Fig. 4. Confirmation of organellar localization and KRas purification.

a-d, Deltavision images showing expression of organellar marker proteins is tuned to stay within their respective organelles. a, VAPA: endoplasmic reticulum b, TGN46: Golgi apparatus c, OMP25: mitochondria. This confirms a low level of expression for each of these organellar markers ensuring they remain localized within the target organelle. d, TMEM192: lysosome Images for each organellar marker were acquired once for confirmation of localization. e, Coomassie stained gel of purified Kras in native nanodiscs. This gel shows a pure protein band corresponding to the molecular weight of peripherally associated membrane protein Kras showing a successful database-guided purification and was repeated twice.

Extended Data Fig. 5. Characterization of purified SYP-VAMP2 co-complex.

a, Population distribution of disc size across CS series. b, Fluorescence size exclusion chromatography of purified SYP-VAMP2 complex shows the purified product is in a stable nanodisc. c-e, Western blots confirm the identity and purity of SYP-VAMP2 complex. All Western blots were repeated twice for confirmation of protein identity. f, Coomassie gel from control experiment showing SYP purified in CS80. This gel was run twice for experimental confirmation.

Extended Data Fig. 6. Characterization of Cl-SMA and AASTY 80.

a, 1H NMR of Cl-SMAnh in DMSO-D6 and AASTY 80 in CDCl₃. b, Size exclusion chromatography profile of homemade polymer AASTY 80, CS40, CS60, and CS80 ran through Superdex 75 column. c, Graph and d, Table stating the value of the weight average (Mw) and number average (Mn) molar mass and dispersity Mw/Mn of each homemade polymer. Dextran molecules of different molecular weight of were run as a standard to generate the calibration curve (represented as black dots in c).