Interrogating membrane protein conformational dynamics within native lipid compositionsEamonn Reading*[a], Zoe Hall[b], Chloe Martens[a], Tabasom Haghighi[a], Heather Findlay[a], Zainab Ahdash[a], Argyris Politis[a], and Paula J. Booth[a]

[a] Dr. E. Reading, Dr. C. Martens, T. Haghighi, Dr. H. Findlay, Z. Ahdash, Dr. A. Politis, Prof. P. J. BoothDepartment of ChemistryKing’s College LondonBrintannia House, 7 Trinity Street, London, SE1 1DB, UKE-mail: eamonn.reading@kcl.ac.uk

[b] Dr. Z. HallDepartment of BiochemistryUniversity of Cambridge80 Tennis Court Road, Cambridge, CB2 1GA, UK

Abstract: Membrane protein-lipid interplay is important for cellular function, however, tools enabling

the interrogation of protein dynamics within native lipid environments are scarce and often invasive. We

establish that the styrene-maleic acid anhydride lipid particle (SMALP) technology can be coupled with

hydrogen-deuterium exchange mass spectrometry (HDX-MS) to investigate membrane protein

conformational dynamics within native lipid bilayers. We demonstrate changes in accessibility and

dynamics of the rhomboid protease, GlpG, captured within three different native lipid compositions, and

identify protein regions sensitive to changes in the native lipid environment. Our results illuminate the

value of this approach for distinguishing the putative role(s) of the native lipid composition in

modulating membrane protein conformational dynamics.

Cell membranes are highly complex and dynamic organizations of lipids and membrane

proteins responsible for many vital cellular functions. Previous work has demonstrated that lipid-

protein interactions influence membrane protein folding, structure, dynamics and function.[1]

However, the influence of the actual native lipid composition encountered in the membrane has

largely eluded investigation. It is imperative, therefore, to be able to explore the interplay

between native lipid environments and membrane proteins to fully understand cellular

membrane protein function.

To investigate their structure and dynamics, membrane proteins are typically solubilized and

purified from their membrane using detergent micelles. This approach has been successful for

studying membrane proteins, however, these systems lack the context of their native lipid

constituents. Recently the SMALP technology has been developed, which uses SMA to directly

solubilize membrane proteins from their native membranes into nanodiscs (Fig. 1a and Fig. S1)

– these have been termed ‘native nanodiscs’ to distinguish them from reconstituted nanodisc

technologies.[2] Excitingly, this enables membrane proteins to be captured in nanodiscs

containing a lipid make-up akin to its native membrane. This technology has been successful for

studying a variety of bacterial and human membrane proteins.[2] Here, we establish that the

SMALP technology can be coupled with HDX-MS to afford a generic method for studying

membrane protein conformational dynamics within native lipid compositions.

HDX-MS measures the rate of exchange of deuterium with backbone amide hydrogens, upon

incubation in D2O-containing buffer, which depends on solvent accessibility and hydrogen

bonding. HDX from solvent-exposed protein regions likely dominates at the earliest times of

incubation, providing information on protein structure. Over time, slow (local and global)

structural transitions, concomitant with protein motions, result in transient exchange events of

otherwise protected amide hydrogens, providing a direct measurement of protein conformational

dynamics.[3] This method differs from other approaches as it enables dynamic structural

information to be captured rapidly on small amounts of material, does not require chemical

modification or sequence alterations, and possesses the ability to analyze complex mixtures.

These advantages have contributed to its wide adoption by both academic and pharmaceutical

sectors[4], nevertheless, its application to membrane proteins is still in its infancy. Although HDX-

MS of membrane proteins within reconstituted nanodiscs has been successful [5], it has not yet

been exploited to study membrane proteins within native nanodiscs (which consist of a native

lipid composition).

We used the Escherichia coli rhomboid protease, GlpG, as a model system since membrane

immersion has been observed to be essential for functional specificity. [6] Rhomboid proteases

are a nearly ubiquitous family of intramembrane serine proteases that cleave peptide bonds

within the lipid bilayer and are regularly implicated in diseases.[7] GlpG consists of a six α-helical

transmembrane (TM) domain, which contains the catalytic dyad, and an α-helical/β-structured

cytoplasmic domain (CytD) (Fig. 1a). High resolution structures have been solved for both

domains separately but not for the complete GlpG protein[8] - with structural information on the

functionally important amino acid linker region (residues 68-90, Ln) currently unavailable.[9]

We aimed to capture GlpG in native nanodiscs consisting of different native lipid compositions

by using cell lines with subtly different lipid compositions (BL21(DE3) (BL) and C43(DE3) (C))[10],

and by altering the temperature upon induction (37 or 16 °C) to further modulate the lipid

content[11]. The three native nanodiscs attained are termed BL37, C37 and C16. We found that

GlpG was pure, folded, thermally stable, and functional within these homogenous native

nanodiscs (Fig. S2 and Table S1). We also measured the inorganic phosphate content of the

purified nanodiscs to confirm and quantify the presence of phospholipids (see Methods),

revealing ~140 phospholipids/GlpG monomer. A GlpG crystal structure within a lipid

environment revealed 14 fully or partially ordered lipid molecules forming a partial annulus

around the protein[12], therefore the native lipid encapsulated by the SMA polymer is likely

sufficient to capture a complete lipid annulus around GlpG.

Using lipidomics, we measured the composition and relative abundances of the lipids within the

GlpG native nanodiscs (Fig. 1b). We identified ~50 individual lipids from three major classes:

phosphatidylglycerols (PG), phosphatidylethanolamines (PE), and cardiolipins (CDL) (Fig. S3

and Table S2), consistent with the known composition of Escherichia coli cell membranes.[11a]

Furthermore, our analysis revealed that BL37 native nanodiscs had relatively decreased PG

content, and increased PE and CDL content, compared to C37 and C16 native nanodiscs. Whilst

the PE and PG content of C37 and C16 were similar, there was a relative decrease in CDLs within

C16 compared to C37. In all three cases, there were no differences observed in the lipid

headgroup composition for membranes compared to their native nanodiscs (Fig. 1c),

highlighting their similar lipid environment.

Next, we used tandem MS to fragment the intact lipids, to identify the fatty acid chains attached

to the different headgroups (Fig. S4). Overall, the saturations and chain lengths were similar

between C37 and BL37, however increased chain unsaturation was observed for C16. Small

changes in chain length were also observed at the lower temperature; with a modest increase in

abundance of longer chain length lipids. These observations are in line with previous reports

which showed Escherichia coli bacteria increase the chain unsaturation and proportion of cis-

vaccenoyl chains (18:1c11), with a decrease in palmitoyl chains (16:1), in response to cold-

shock.[11] An increase in chain unsaturation leads to looser lipid packing, thereby increasing

membrane fluidity at lower temperatures.

We then established a HDX-MS approach compatible with the SMALP technology. We

optimized the quenching and digestion conditions, and overall workflow, using the C37 native

nanodiscs (Fig. S5 and Table S3). Applying the same conditions to the other two native

nanodiscs (C16 and BL37) we could achieve peptide coverage sufficient for investigating the

impact that alterations in native lipid composition has on GlpG (Fig. 2a and Fig. S6).

By observing the relative fractional uptake of deuterium by GlpG across all the native lipid

compositions, we could make some general inferences (Fig. 2b and Fig. S7-8). Peptides within

TM2-6 of GlpG were relatively protected from HDX, demonstrated by a low relative fractional

uptake, probably because these regions are buried within the hydrophobic core of the

membrane. The CytD and Ln regions of GlpG displayed much higher relative fractional uptake

of deuterium in general. Surprisingly, the TM region linked to the cytoplasmic domain of GlpG

(TM1, peptide 94-108) was relatively unprotected from HDX and possessed dynamic behavior.

We then directly compared relative deuterium uptake between the three native lipid

compositions (Fig. 3). This enabled us to pinpoint regions of GlpG which are influenced by

subtle alterations in their lipid environment – informed by changes in either their protection to

HDX and/or HDX dynamics. We found that GlpG possessed similar HDX behavior between the

C37 and BL37 nanodiscs despite significant differences in PE and PG compositions (PE/PG ratio:

C37 = 1.7 ± 0.2, BL37 = 4.4 ± 1.3); however, at the maximum deuteration time (2.5 hours) the Ln

region (peptide 68-78) was deprotected from HDX within the BL37 nanodisc. The most striking

differences in HDX of GlpG were found for the C16 nanodisc environment compared to both the

BL37 and C37 nanodiscs. Here, we observed peptides within the CytD, Ln, and TM1 regions that

were typically deprotected within the C16 environment. This is likely due the C16 nanodisc

possessing a more fluid bilayer which facilitates GlpG structural fluctuations.

Overall, the HDX-MS analysis proved powerful in revealing regions that are strongly influenced

by alterations in the native lipid environment – these were the CtyD, Ln and TM1 regions.

Interestingly, the TM1 region has been found to be important for structural stability[13] and a

preferred interaction site (close to TM1/TM3) for the rhomboid protease Spitz substrate. [14] The

CtyD domain can undergo domain swapping in isolation from the GlpG domain[15] but currently

the functional importance of this region is unknown – interestingly, cytosolic extensions in other

rhomboid proteases have been found to be important for modulating substrate gating. [16] The Ln

region has been shown to be important for maintaining maximum GlpG activity in some assays.

[9a] Due to its position within the protein (connecting the GlpG and CtyD domains) it is likely to be

in direct, or close, contact with the membrane. We therefore propose that this lipid sensitive

region may play a role in the function of GlpG through interactions with the membrane. Changes

in the PE/PG ratio between C37 and BL37 did not seem to affect HDX significantly, whereas

changes in chain length and saturation within C16 did. This may be related to previous

suggestions from detergent micelle and bicelle systems, that hydrophobic mismatch could exert

an inhibitory effect on GlpG activity.[9b]

The potential of this method in examining lipid, ligand and drug interactions with membrane

proteins, in well-defined native lipid environments, is an exciting prospect which we anticipate

will have significant impact on the membrane protein structural biology field, as well as on drug

discovery pipelines.

Acknowledgements

E.R. is funded by a BBSRC Future Leader Fellowship BB/N011201/1. We acknowledge funding

from the European Research Council, ERC Advanced grant 294342 to P.J.B., The Royal

Society and the Wellcome Trust (109854/Z/15/Z) to A.P.. We thank Dr. Malcolm Anderson

(Waters), Dr. Antoni Borysik, Matthew Harris, and Kjetil Hansen for their assistance with

hydrogen-deuterium exchange mass spectrometry. We thank Andy M. Lau for the provision of

the MatLab script used for the creation of the Woods plots. We thank Prof. Julian Griffin for use

of Orbitrap MS instrumentation.

Keywords: lipids • mass spectrometry • membrane mimetics • membrane proteins • structural

biology

References

[1] a) A. Laganowsky, E. Reading, T. M. Allison, M. B. Ulmschneider, M. T. Degiacomi, A. J.

Baldwin, C. V. Robinson, Nature 2014, 510, 172-175; b) H. Hong, Adv. Exp. Med. Biol. 2015,

855, 1-31; c) C. Martens, R. A. Stein, M. Masureel, A. Roth, S. Mishra, R. Dawaliby, A.

Konijnenberg, F. Sobott, C. Govaerts, H. S. McHaourab, Nat. Struct. Mol. Biol. 2016, 23, 744-

751.

[2] J. M. Dörr, S. Scheidelaar, M. C. Koorengevel, J. J. Dominguez, M. Schäfer, C. A. van Walree, J.

A. Killian, Eur. Biophys. J. 2016, 45, 3-21.

[3] T. E. Wales, J. R. Engen, Mass Spectrom. Rev. 2006, 25, 158-170.

[4] B. Deng, C. Lento, D. J. Wilson, Anal. Chim. Acta 2016, 940, 8-20.

[5] a) C. M. Hebling, C. R. Morgan, D. W. Stafford, J. W. Jorgenson, K. D. Rand, J. R. Engen, Anal.

Chem. 2010, 82, 5415-5419; b) S. Adhikary, D. J. Deredge, A. Nagarajan, L. R. Forrest, P. L.

Wintrode, S. K. Singh, Proc. Natl. Acad. Sci. U.S.A. 2017, 114, E1786-e1795.

[6] S. M. Moin, S. Urban, eLife 2012, 1, e00173.

[7] S. Urban, Semin. Cell Dev. Biol. 2016, 60, 1-4.

[8] C. L. Brooks, M. J. Lemieux, Biochim. Biophys. Acta - Biomembranes 2013, 1828, 2862-2872.

[9] a) A. R. Sherratt, D. R. Blais, H. Ghasriani, J. P. Pezacki, N. K. Goto, Biochemistry 2012, 51,

7794-7803; b) A. C. Foo, B. G. Harvey, J. J. Metz, N. K. Goto, Protein Sci. 2015, 24, 464-473.

[10] I. Arechaga, B. Miroux, S. Karrasch, R. Huijbregts, B. de Kruijff, M. J. Runswick, J. E. Walker,

FEBS Lett. 2000, 482, 215-219.

[11] a) S. Morein, A.-S. Andersson, L. Rilfors, G. Lindblom, J. Biol. Chem. 1996, 271, 6801-6809; b)

M. C. Mansilla, L. E. Cybulski, D. Albanesi, D. de Mendoza, J. Bacteriol. 2004, 186, 6681-6688.

[12] K. R. Vinothkumar, J. Mol. Biol. 2011, 407, 232-247.

[13] R. P. Baker, S. Urban, Nat. Chem. Biol. 2012, 8, 759-768.

[14] T. Reddy, J. K. Rainey, J. Phys. Chem. B 2012, 116, 8942-8954.

[15] C. Lazareno-Saez, E. Arutyunova, N. Coquelle, M. J. Lemieux, J. Mol. Biol. 2013, 425, 1127-

1142.

[16] R. P. Baker, S. Urban, Nature 2015, 523, 101-105.

Figure 1. GlpG within native nanodiscs (SMALPs) of varying native lipid composition. (a) The GlpG

membrane domain (PDB: 2XTV) is connected to its cytoplasmic N-terminal domain (CytD, PDB: 2LEP) by

a linker region (residues 68-90, Ln) which currently has no structural information available. As an

alternative to solubilization with n-dodecyl β-D-maltoside (DDM) detergent, forming a protein-detergent

micelle complex, GlpG is solubilized from its native bilayer with the SMA polymer to form SMALPs (native

nanodiscs), which contain a native lipid bilayer. (b) Lipidomics analysis of native nanodiscs containing

GlpG. Summary of phospholipid head group compositions of the three native nanodisc systems (C 16, C37,

and BL37), and their differences in chain length and chain saturation. (c) Summary of differences in

phospholipid head group composition, chain length, and chain saturation between native cell

membranes and native nanodiscs. Black arrows represent the direction of comparison and the red ramps

represent the direction and degree (small or large) of change. The phospholipid head group, and chain

length and saturation compositions, for both the native nanodiscs and the native cell membrane are

reported in Table S2 and Fig. S4.

Figure 2. HDX-MS of GlpG in native nanodiscs (SMALPs). (a) GlpG peptide coverage across the three

native nanodisc compositions. Details on the high resolution structural information available for GlpG are

included. (b) Heat maps representing the relative fractional uptake of deuterium for peptides of GlpG in

the three native nanodiscs. The secondary structure of GlpG is shown on the right (black); β-strands (β)

are depicted as arrows, loop (L) regions and linker (Ln) regions as lines, the purification tag as dotted

lines, and α-helices (α) and transmembrane helices (TM) as bars. The degree of relative fractional uptake

of deuterium at each incubation time (from left to right: 9 s, 90 s, 900 s, and 9000 s) is displayed

according to the color code shown. Uncolored regions indicate areas with no peptide coverage.

Figure 3. Differences in relative deuterium uptake (ΔHDX) of GlpG in different native nanodiscs. HDX

behavior is mapped onto the GlpG (PDB: 2XTV) and CytD (PDB: 2LEP) crystal structures for each

deuteration incubation time. Red and blue colored regions indicate GlpG segments that contain a

peptide that became HDX deprotected or protected respectively; grey regions represent where no

significant ΔHDX is observed for any peptide and green regions represent regions of GlpG where

peptides were not obtained. A significant change in ΔHDX was determined as > 0.9 Da (99% confidence

interval). The ΔHDX data is presented in a Woods plot in Fig. S7.

Supporting Information

Experimental Section

Preparation of SMA

Styrene maleic acid co-polymer 2000 (SMA) was prepared from styrene maleic anhydride co-polymer 2000 (SMAn, poly(styrene/maleic anhydride) [67:33], MW 7,500, Polysciences Europe) as described by Lee et al.[1] Fourier transform infrared spectrophotometry (IRAffinity-1S, Shimadzu) was used to assess the hydrolysis of SMAn to SMA (Fig. S1).

Preparation of GlpG in SMALPs and detergent micelles

A pET28a plasmid containing C-terminally Myc-His6 tagged GlpG was transformed into either BL21(DE3) or C43(DE3) cells. 7 ml of an overnight LB culture was added to 1L of LB culture and grown at 37 °C until an OD of 0.6-0.8 was achieved. The cultures were then induced with 1 mM IPTG and either cooled to 16 °C and left overnight before harvesting the cells, or grown for 3 hours at 37 °C and then harvested. The cells were washed with PBS and either stored at –80 °C or used immediately. Cell pellets were thawed and resuspended in Buffer A (300 mM sodium chloride, and 20 mM 2-amino-2-hydroxymethyl-propane-1,3-diol (Tris), pH 7.4 at room temperature) supplemented with a complete protease inhibitor tablet (Roche), PMSF, Benzonase, and 5 mM beta-mercaptoethanol (β-ME). The cell suspension was passed twice through a microfluidizer (Microfluidics) at 25,000 psi. Insoluble material was pelleted by centrifugation twice at 20,000 x g for 25 min at 4 °C. Membranes were pelleted by centrifugation at 150,000 x g for 1.5 h at 4 °C.

Membranes were resuspended to 40 mg ml-1 in ice-cold buffer B (500 mM sodium chloride, 10% glycerol and 50 mM Tris, pH 7.4 at room temperature) supplemented with a complete protease inhibitor tablet (Roche) and PMSF, homogenized using a Potter-Elvehjem Teflon pestle and glass tube. GlpG was then solubilized and purified from membranes with either SMA 2000 or n-dodecyl β-D-maltoside (DDM) detergent (Anatrace).

To form GlpG native nanodiscs the protocols of Lee et al[1] were followed with slight modifications. Briefly, SMA 2000 co-polymer powder was added to the suspension at a

final concentration of 2.5 % (w/v) to solubilize the membranes. The membrane suspension with SMA was incubated for 2 hours with gentle agitation at room temperature, followed by 1 hour of centrifugation at 100,000 × g at 4 °C. 1 ml super nickel NTA agarose affinity resin (Generon), equilibrated in Buffer B, was added to the solubilized membrane suspension, and equilibrated overnight at 4 °C. The beads were then transferred to a gravity-flow column and washed with 30 column volumes (CVs) of buffer B supplemented with 20 mM Imidazole. The protein as then eluted with Buffer B containing 500 mM imidazole. The pooled samples were then buffer exchanged into buffer C (50 mM Tris, pH 7.4, 150 mM NaCl, 10 % glycerol) using a PD-10 desalting column (GE Healthcare). The samples can be stored at –20 °C or –80 °C; however, we found that samples stored at –20 °C maintained higher activity than those stored at –80 °C. SDS-PAGE electrophoresis assessed GlpG purification and protein concentration was calculated using a Markwell-Lowry assay.[2]

To obtain GlpG in DDM detergent micelles, DDM was added to the membrane suspension at a final concentration of 1 % (w/v) to solubilize the membranes. The membrane suspension with DDM was incubated overnight with gentle agitation at 4 °C, followed by 1 hour of centrifugation at 100,000 × g at 4 °C. The supernatant was then filtered before loading onto a 1 ml HiTrap (GE Healthcare) equilibrated in buffer D (500 mM sodium chloride, 20 mM imidazole, 10% (v/v) glycerol and 50 mM Tris, pH 7.4 at room temperature, 0.025 % (w/v) DDM). The column was washed with 10 CVs of buffer D, then 20 CVs buffer D containing 50 mM imidazole. GlpG was eluted with buffer D containing 500 mM imidazole and injected directly onto a Superdex 75 10/600 GL size exclusion chromatography (SEC) column (GE Healthcare) equilibrated in buffer E (50 mM Tris, pH 7.4 at room temperature, 150 mM NaCl, 10 % glycerol and 0.025 % (w/v) DDM). Peak fractions eluted from the SEC column containing GlpG were pooled and spin filtered before being flash frozen and stored at -80 °C or –20 °C (activity was similar for both). SDS-PAGE electrophoresis assessed GlpG purification and protein concentration was calculated using a Markwell-Lowry assay.[2]

Dynamic light scattering

All DLS measurements were carried out on a Zetazizer Nano ZS (Malvern).

GlpG FL-caesin proteolysis assay

Protease activity of GlpG was assayed using the EnzChek® Protease Assay Kit (green fluorescence, Life Technologies). The proteolysis assay was based on previously

described protocols.[3] GlpG in DDM detergent micelles or in native nanodiscs were incubated at 37 °C, shaking at 300 rpm, with varying concentrations of BODIPY-labeled caesin (FL-caesin) (Invitrogen) in buffer C or E. Fluorescence excitation of a 200 μL volume was induced at 485 nm and emission monitored at 535 nm using a Hidex Sense multimode microplate reader (Hidex) in the top-reading mode. All samples were assayed at least in triplicate against a buffer blank. Initial proteolysis rates (V 0) were calculated using the slope of relative fluorescence units (rfu) over 30 min after a 10 min equilibration period.

Circular dichroism spectroscopy (CD)

All CD spectra were measured in an Aviv Circular Dichroism Spectrophotometer, Model 410 (Biomedical Inc., Lakewood, NJ, USA), with specially adapted sample detection to eliminate scattering artefacts. A final protein concentration of 0.02-2 mg ml -1 was used in quartz rectangular or circular Suprasil demountable cells of pathlengths (Hellma Analytics). All CD spectra were processed using CDTool.[4] First, the multiple scans were averaged and the buffer background was subtracted. These subtracted spectra were zeroed and smoothed, which set the baseline at zero between 253 and 260 nm. For thermal protein unfolding the mean residue ellipticity at 222 nm was monitored with increasing temperature.

Lipid extraction and phospholipid concentration determination

Lipids were extracted as previously described by Dorr et al.[5] A phosphorus assay devised by Rouser et al[6] was used to assess phospholipid concentration - phospholipids in lipid extracts are estimated by phosphorus determination through an acidic digestion. The released inorganic phosphate is reacted with ammonium molybdate, the complex giving a strong blue color. Lipid samples were transferred into clean glass tubes and the solvent is completely evaporated. 0.65 ml of concentrated perchloric acid was then added and the tubes heated to 180 °C for 30 min. When cool, 3.3 ml distilled water was added, then 0.5 ml of molybdate solution (25 mg.ml -1

ammonium molybdate (Sigma) in distilled water), and then 0.5 ml of ascorbic acid solution (100 mg.ml-1 ascorbic acid (Sigma) in distilled water) with vortex agitation after each addition. The samples were then placed in a boiling water bath for 5 min, then left to cool to room temperature, and finally their absorbance was measured at 800 nm. Potassium dihydrogen phosphate (Sigma) solutions were used as phosphate standards for phosphate concentration determination. The number of phospholipids is calculated directly on a molar basis from the amount of phosphate.

Lipidomics experiments

Lipid extracts were dried down under nitrogen and reconstituted in 2:1:1 isopropanol:acetonitrile:water. Internal standard mix (14:0 cardiolipin; 16:0/18:1 D5-phosphatidylglycerol; 16:0/18:1 D31 phosphatidylethanolmine) was added at final concentration of 10 μg/mL. Samples were analyzed by liquid chromatography-mass spectrometry (LC-MS) using an Accela Autosampler (Thermo Scientific, Hemel Hempstead, UK) coupled to a LTQ Orbitrap EliteTM (Thermo Scientific). Sample was injected onto an Acuity C18 BEH column (Waters Ltd., Warrington, UK; 50 × 2.1 mm, 1.7 µm) at 55 °C. Mobile phase A was acetonitrile:water (60:40) and mobile phase B was isopropanol:acetonitrile (90:10). Both contained 10 mM ammonium acetate. A gradient run was used (Table S4) at a flow rate of 0.5 mL/min. The electrospray ionisation source temperature was 375 °C, the desolvation temperature was 380 °C and desolvation gas flow 40 arbitrary units. Spectra were acquired in negative ion mode in the range of 100 - 1600 m/z at 100,000 mass resolution. Peaks were integrated using Xcalibur 3.0 software. Lipid identity was performed by accurate mass using the LipidMaps database.[7] Tandem MS, using collision-induced dissociation, was used to determine the predominant fatty acid composition.

Hydrogen deuterium mass spectrometry

GlpG native nanodiscs were prepared at a concentration of 14 μM. Optimized peptide identification and peptide coverage for GlpG was performed from undeuterated controls. The sample workflow involved diluting 5 μl of GlpG-SMALP into 95 μl of buffer C at 25 °C followed by quenching, nanodisc solubilization, lipid removal (with ZrO2 HybridSPE beads (Supelco)) and polymer removal, and pepsin digestion (Fig. S5). Quenching, nanodisc and pepsin digestion conditions were optimized to achieve sufficient peptide digestion (Table S3) – this included exploring different acidic quench solutions (0.8-1.6 % formic acid versus 0.1 M NaH2PO4 pH 2.4), inclusion of guanidine HCl and acetonitrile, inclusion of different DDM detergent and magnesium acetate concentrations, pepsin digestion conditions (digestion in solution (Pepsin from porcine gastric mucose (Sigma)), on beads (Poroszyme immobilized pepsin (Thermo Scientific)) or on-line using a Enzymate online digestion column (Waters)), and digestion temperature.

The optimal sample scheme for HDX purposes was as follows: 5 μl of GlpG-SMALP (14 μM) was diluted into 95 μl of either buffer C or with deuterated buffer C at 25 °C. The deuterated samples were then left for 9, 90, 900, and 9000 seconds at 25 °C. All samples were then quenched with 100 μl of quench buffer (0.8 % formic acid and 0.1 %

DDM) to provide a solution at pH 2.3. When performing offline pepsin digestion, pepsin was added immediately after the quench solution. The sample was vortexed for 3 s then placed immediately on ice for 1 minute. 10 μl of ZrO2 beads (300 mg/ml equilibrated in 0.8 % formic acid) was then added and the sample vortexed for 3 s, placed on ice for 30 s, vortexed for a further 3 s, and then placed on ice for a further 30 s. The samples were then filtered through pre-chilled 0.22 μm spin filtration devices (Corning Costar Spin-X) in a pre-chilled microcentrifuge at 1,000 x g for 30 s. Digested or undigested samples were then immediately flash frozen in liquid nitrogen and stored at –80 °C before analysis.

Samples were rapidly defrosted and then injected into a Waters HDX nanoAcquity ultra-performance liquid chromatography (UPLC) system (Waters) using a pre-chilled syringe. When performing on-line digestion, samples were digested by an EnzymateTM

digestion column (Waters) (flow rate of 200 μl/min) at 20 °C in 0.1 % formic acid in water (digestion efficiency, tested using a PhosB standard, remained for hundreds of uses). For both on-line and off-line pepsin digestion, the peptic fragments were then trapped using a chilled Acquity BEH C18 1.7 μM VANGUARD pre-column for 3 min. The peptic fragments were then eluted using an 8-40% gradient of 0.1 % formic acid in acetonitrile at 40 μl/min into a chilled Acquity UPLC BEH C18 1.7 μM 1.0 x 100 mm column. The peptides were then ionized by electrospray into a Synapt G2-Si mass spectrometer (Waters). MSE data were acquired with a 20 to 30 V trap collision energy ramp for high-energy acquisition of product ions. Leucine Enkephalin (LeuEnk) was used as a lock mass for mass accuracy correction and the mass spectrometry was calibrated with sodium iodide. The on-line Enzymate pepsin column was washed with pepsin wash (1.5 M Gu-HCl, 4 % MeOH, 0.8 % formic acid) recommended by the manufacturer and a blank run was performed between each sample to prevent significant peptide carry-over from the pepsin column.

All deuterium time points and controls were performed in triplicate. Sequence identification was made from MSE data from the undeuterated samples of GlpG in native nanodiscs using the Waters ProteinLynx Global Server 2.5.1 (PLGS). The output peptides were filtered using DynamX (v. 3.0) using the following filtering parameters: minimum intensity of 1000, minimum and maximum peptide sequence length of 5 and 25 respectively, minimum MS/MS products of 2, minimum products per amino acid of 0.15, and a maximum MH+ error threshold of 15 ppm. Additionally, all the spectra were visually examined and only those with high signal to noise ratios were used for HDX-MS analysis. The amount of relative deuterium uptake for each peptide was determined using DynamX (v. 3.0) and are not corrected for back exchange. The relative fractional

uptake (RFU) was calculated from RFUa = [Ya,t/(MaxUptakea x D)], where Y is the deuterium uptake for peptide a at incubation time t, and D is the percentage of deuterium in the final labeling solution. Back exchange (29 ± 12%) was calculated from highly deuterated peptides. The standard deviation for the HDX data was 0.16 Da. Confidence intervals for the ΔHDX of any individual time point were then determined according to Houde et al.[8] Specifically, the 98 % and 99 % confidence intervals for the ΔHDX at any single time point was determined to be ±0.6 Da and ±0.9 Da, respectively. There was no correlation found between ΔHDX values and their standard deviations (R2

= 0.09). The snake diagram of GlpG was made by Protter (http://wlab.ethz.ch/protter/) using the UniProt accession: GLPG_ECOLI.

Supporting Results

Figure S1. Preparation of styrene maleic acid co-polymer 2000 (SMA) from styrene maleic

anhydride co-polymer 2000 (SMAn). (a) SMA was formed by hydrolysis of SMAn. The infrared

spectra of SMAn (Top) shows the presence of maleic anhydride carbonyl groups (1775 cm-1

signal) which are subsequently converted to carboxylate carbonyl groups (1555 cm-1 signal)

upon hydrolysis (Bottom) to form SMA. (b) DMPC liposomes without (-) and with (+) SMA

addition at 2.5 % w/v demonstrates the solubilizing ability of SMA. (c) Dynamic light scattering

(DLS) measurements of the samples in (b) show that SMA solubilizes DMPC liposomes into

homogenous SMALPs.

Figure S2. Characterization of GlpG in native nanodiscs and comparison to detergent

solubilized GlpG. (a) SDS-PAGE gel of GlpG purified in DDM detergent micelles or different

native nanodiscs (SMALPs). (b) Circular dichroism (Top) and thermal stability (Bottom) of GlpG

reveals a similar α-helical secondary structure in all conditions. GlpG in native nanodiscs are

markedly more thermally stable than when in DDM detergent micelles, as judged by the loss of

the circular dichroism helical signature at 222 nm with increasing temperature. (c) GlpG

protease function using a soluble fluorescence caesin substrate (FL-caesin). Although GlpG is

an intramembrane protease it has been shown to cleave peptide bonds in the FL-caesin

substrate in a non-competitive manner[10], unquenching the dye, this enables its fluorescence

emission to be measured with time, providing a measure of GlpG activity. As the natural

substrate of GlpG is currently unknown, and due to the difficulty of capturing cleavable TM

substrates within native nanodiscs, the FL-casein provides a useful measure of GlpG activity in

vitro. Using the initial rate of protease activity (V0) a Michaelis–Menten kinetic model (d) could

be fit revealing that GlpG has both a higher Vmax and Km when in a DDM micelle compared to

native nanodiscs (Table S1) – therefore when in a detergent micelle GlpG has a reduced affinity

to the FL-caesin substrate but a much higher proteolysis rate when compared to a native

nanodisc environment. We observed that GlpG has a similar FL-caesin protease activity within

all three of the native nanodiscs. Native nanodiscs have been found to improve the activity or

substrate binding for a range of membrane proteins when compared to solubilisation in

detergent micelles[11], these observations arguing the ability of native nanodiscs to capture an in

vitro environment more akin to a cellular one. However, we observe a significant decrease in

function of GlpG within native nanodiscs. This reduced function actually demonstrates that GlpG

has been captured in an environment more akin to the cellular one, as GlpG has been shown to

have a much-reduced activity when in a membrane, or indeed in vivo, than when in DDM

micelles.[12] Therefore, GlpG in native nanodiscs provides a more native-like environment in

which to study its structure and function in vitro. Reported are the average and s.e.m. from

repeated measurements (n = 3). (e) Dynamic light scattering of GlpG in native nanodiscs shows

that all the nanodisc particles are of similar size and homogeneity. Reported are the average

and s.e.m. from repeated measurements (n = 3).

Figure S3. Representative raw MS, MS/MS, and HPLC data for the three most abundant lipid

classes in E. coli. (a) L-α-phosphatidylglycerol (PG), (b) L-α-phosphatidylethanolamine (PE),

and (c) cardiolipin (CDL).

Figure S4. Summary of phospholipid chain length and saturation lipidomics analysis for the

extracted membrane fractions (red) and for the native nanodiscs (blue). The compositions of

fatty acid chain length (a) and levels of unsaturation (b) for each lipid species are shown.

Reported are the average and s.e.m. from repeated measurements (n = 3), except for the C16

membrane fraction which was only performed twice. The summed chain isomers 35a and 35b

correspond to fatty acids with 17/18 and 16/19 carbon chain length respectively. There was little

difference between the native nanodisc and membrane fraction lipid compositions, except for a

slight decrease in chain saturation in the native nanodiscs samples. This could be due to the

phenomenon that the SMA polymer has a stronger preference for solubilisation of lipids from

fluid phases as compared to those in either a gel phase or a liquid-ordered phase.[13]

Figure S5. Workflow protocols (a and b) for HDX-MS interrogation of membrane proteins within

native nanodiscs. The sample is exposed to deuterated buffer for four different time points (9,

90, 900 and 9000 seconds). The sample is then quenched to pH 2.3 and placed on ice – the

sample is kept at this pH and temperature from now on to reduce HDX back exchange.

Reducing the pH to 2.3 will also act to dissociate the SMA polymer from the SMALP discs as

the polymer becomes neutral at this pH and the self-assembly process is no longer favorable [11].

Addition of the divalent cation Mg2+ also aids in this process, as its chelation to the SMA polymer

aids in making the polymer insoluble, however, addition of Mg2+ did not enhance peptide

coverage (see Table S3). Addition of DDM detergent (at approximately 5.5 times its critical

micelle concentration) during this process maintains the solubility of the membrane protein

during this disc dissociation process, enhancing digestion (see Table S3) – it may also aid in

delipidation. The protein can be digested during this process by the addition of pepsin in

solution or immobilized to beads (protocol b), or digested later (protocol a). To remove lipids

(which are common column fouling agents for liquid chromatography) ZrO2 beads are added,

the beads can bind the phospholipid headgroup of lipids[14]. The ZrO2 beads and insoluble SMA

polymer are then removed by filtration at 0 °C. If the protein has been already digested, then it

can directly be analyzed by UPLC-MS (protocol b). If not, then the protein can be digested on-

line using a pepsin column before UPLC-MS analysis (protocol a).

Figure S6. Combined peptide coverage (80 %) of GlpG for all three native nanodiscs after

further manual processing of peptide data.

Figure S7. Woods plot of differences in relative deuterium uptake (ΔHDX) of GlpG within the

BL37, C37 and C16 native nanodiscs. The length of the lines represents the length of the peptide.

Figure S8. Plots of relative deuterium uptake, with deuteration time, for each GlpG peptide

within native nanodiscs; BL37 (purple), C37 (red) and C16 (blue) native nanodiscs. Reported are

the average and s.d. from triplicate measurements.

Table S1. Michaelis–Menten kinetics for GlpG proteolysis activity towards the soluble FL-caesin

substrate. Reported are the values and standard errors from Michaelis–Menten model fits.

Condition R2 Vmax (RFU/s) Km (μM)

DDM 0.98 98.7 ± 14.3 0.44 ± 0.12

C16 0.95 16.1 ± 1.7 0.14 ± 0.03

C37 0.96 14.2 ± 2.7 0.24 ± 0.15

BL16 0.95 22.1 ± 6.4 0.24 ± 0.18

Table S2. Phospholipid head group compositions (L-α-phosphatidylglycerol (PG), L-α-

phosphatidylethanolamine (PE), and cardiolipin (CDL)) of the native nanodisc and E. coli

derivative extracted membranes. Reported are the average and s.e.m. from repeated

measurements (n = 3), except for the C16 membrane fraction which was only performed twice.

Condition PE PG CDLC16 – Native nanodisc 63.9 ± 3.4 35.6 ± 3.2 0.3 ± 0.1

C16 – Membrane 66.3 ± 3.0 33.1 ± 2.9 0.5 ± 0.2C37 – Native nanodisc 62.1 ± 4.0 35.7 ± 4.2 2.0 ± 0.7

C37 – Membrane 68.7 ± 4.8 28.6 ± 4.9 2.7 ± 1.0BL37 – Native

nanodisc 79.2 ± 3.8 17.7 ± 5.1 3.1 ± 1.6

BL37 – Membrane 77.1 ± 2.8 18.0 ± 3.8 4.8 ± 2.0

Table S3. Optimization of peptide number and protein coverage of GlpG in C37.

Condition Detergent % (w/v)

Mg2+

(mM)Pepsin Digestion

temperature (°C)

# Peptides

Redundancy Coverage (%)

# TM peptides

TM coverage

(%)0.8 % formic

acid 0 0 Solution 4 45 3.0 67 4 18

1.2 % formic acid " 5 " " 33 3.2 45 2 12

0.8 % formic acid 0.05 0 " " 53 4.2 56 8 34

1.2 % formic acid " 5 " " 64 4.0 66 7 37

0.8 % formic acid 0.1 0 " " 74 4.8 66 8 38

1.2 % formic acid " 5 " " 73 4.4 69 11 48

0.8 % formic acid 0.2 0 " " 53 4.2 56 8 36

1.2 % formic acid " 5 " " 64 4.0 66 7 32

1.2 % formic acid 0.1 " " 20 81 6.2 54 5 28

1.2 % formic acid " " Beads 4 73 4.5 61 16 41

0.1M NaH2PO4,

pH 2.4" " Solution " 52 3.7 56 8 36

0.8 % formic acid,

1M Gu-HCl" 0 " " 68 4.3 66 15 58

1.2 % formic acid,

1M Gu-HCl" 5 " " 60 3.5 76 9 61

1.2 % formic acid, 1.5M

Gu-HCl" " " " 57 5.0 51 4 18

0.8 % formic acid,

1M Gu-HCl, 10 % ACN

" 0 " " 45 3.1 60 6 35

1.2 % formic acid,

1M Gu-HCl, 10 % ACN

" 5 " " 35 2.6 60 7 45

1.2 % formic acid " " On-line 20 57 2.7 84 13 85

1.6 % formic acid " 10 " " 77 3.9 73 16 52

Table S4. Chromatography gradient composition for lipidomics open profiling by LC-MS.

Time (min) Mobile Phase A (%) Mobile Phase B (%)0 60 40

0.8 57 430.9 50 504.8 46 544.9 30 705.8 19 818 1 99

8.5 1 998.6 60 4010 60 40

Supporting References

[1] S. C. Lee, T. J. Knowles, V. L. G. Postis, M. Jamshad, R. A. Parslow, Y.-p. Lin, A. Goldman, P. Sridhar, M. Overduin, S. P. Muench, T. R. Dafforn, Nat. Protocols 2016, 11, 1149-1162.

[2] M. A. Markwell, S. M. Haas, L. L. Bieber, N. E. Tolbert, Anal. Biochem. 1978, 87, 206-210.

[3] a) Y. Wang, Y. Zhang, Y. Ha, Nature 2006, 444, 179-180; b) A. C. Foo, B. G. Harvey, J. J. Metz, N. K. Goto, Protein Sci. 2015, 24, 464-473; c) A. R. Sherratt, D. R. Blais, H. Ghasriani, J. P. Pezacki, N. K. Goto, Biochemistry 2012, 51, 7794-7803.

[4] J. G. Lees, B. R. Smith, F. Wien, A. J. Miles, B. A. Wallace, Anal. Biochem. 2004, 332, 285-289.

[5] J. M. Dorr, M. C. Koorengevel, M. Schafer, A. V. Prokofyev, S. Scheidelaar, E. A. van der Cruijsen, T. R. Dafforn, M. Baldus, J. A. Killian, Proc. Natl. Acad. Sci. U.S.A.2014, 111, 18607-18612.

[6] G. Rouser, S. Fleischer, A. Yamamoto, Lipids 1970, 5, 494-496.

[7] M. Sud, E. Fahy, D. Cotter, A. Brown, E. A. Dennis, C. K. Glass, A. H. Merrill, Jr., R. C. Murphy, C. R. Raetz, D. W. Russell, S. Subramaniam, Nucleic Acids Res. 2007, 35, D527-532.

[8] D. Houde, S. A. Berkowitz, J. R. Engen, J. Pharm. Sci. 2011, 100, 2071-2086.

[9] U. Omasits, C. H. Ahrens, S. Muller, B. Wollscheid, Bioinformatics 2014, 30, 884-886.

[10] E. Arutyunova, P. Panwar, P. M. Skiba, N. Gale, M. W. Mak, M. J. Lemieux, EMBO J. 2014, 33, 1869-1881.

[11] J. M. Dörr, S. Scheidelaar, M. C. Koorengevel, J. J. Dominguez, M. Schäfer, C. A. van Walree, J. A. Killian, Eur. Biophys. J 2016, 45, 3-21.

[12] a) Seth W. Dickey, Rosanna P. Baker, S. Cho, S. Urban, Cell, 155, 1270-1281; b) S. M. Moin, S. Urban, eLife 2012, 1, e00173.

[13] J. J. Dominguez Pardo, J. M. Dörr, A. Iyer, R. C. Cox, S. Scheidelaar, M. C. Koorengevel, V. Subramaniam, J. A. Killian, Eur. Biophys. J 2017, 46, 91-101.

[14] C. M. Hebling, C. R. Morgan, D. W. Stafford, J. W. Jorgenson, K. D. Rand, J. R. Engen, Anal. Chem. 2010, 82, 5415-5419.