A rapid expression and purification condition screening protocol for membrane protein structural biology

Abstract

Membrane proteins control a large number of vital biological processes and are often medically important—not least as drug targets. However, membrane proteins are generally more difficult to work with than their globular counterparts, and as a consequence comparatively few high‐resolution structures are available. In any membrane protein structure project, a lot of effort is usually spent on obtaining a pure and stable protein preparation. The process commonly involves the expression of several constructs and homologs, followed by extraction in various detergents. This is normally a time‐consuming and highly iterative process since only one or a few conditions can be tested at a time. In this article, we describe a rapid screening protocol in a 96‐well format that largely mimics standard membrane protein purification procedures, but eliminates the ultracentrifugation and membrane preparation steps. Moreover, we show that the results are robustly translatable to large‐scale production of detergent‐solubilized protein for structural studies. We have applied this protocol to 60 proteins from an E. coli membrane protein library, in order to find the optimal expression, solubilization and purification conditions for each protein. With guidance from the obtained screening data, we have also performed successful large‐scale purifications of several of the proteins. The protocol provides a rapid, low cost solution to one of the major bottlenecks in structural biology, making membrane protein structures attainable even for the small laboratory.

Abbreviations

Cmc

critical micelle concentration

EM

electron microscopy

FSEC

fluorescence‐detection size‐exclusion chromatography

GFP

green fluorescent protein

IMAC

immobilized metal affinity chromatography

LCP

lipidic cubic phase

TEV

tobacco etch virus

NM

n‐nonyl‐β‐d‐maltoside

DM

n‐decyl‐β‐d‐maltoside

UDM

n‐undecyl‐β‐d‐maltoside

DDM

n‐dodecyl‐β‐d‐maltoside

CYMAL‐4/CM‐4

4‐cyclohexyl‐1‐butyl‐β‐d‐maltoside

CYMAL‐5/CM‐5

5‐cyclohexyl‐1‐pentyl‐β‐d‐maltoside

CYMAL‐6/CM‐6

6‐cyclohexyl‐1‐hexyl‐β‐d‐maltoside

CYMAL‐7/CM‐7

7‐cyclohexyl‐1‐heptyl‐β‐d‐maltoside

OG

n‐octyl‐β‐d‐glucoside

NG

n‐octyl‐β‐d‐glucoside

DDAO

n‐decyl‐N,N‐dimethylamine‐N‐oxide

UDAO

n‐undecyl‐N,N‐dimethylamine‐N‐oxide

LDAO

n‐dodecyl‐N,N‐dimethylamine‐N‐oxide

C₈E₄

octyl tetraethylene glycol ether

C₁₀E₅

decyl pentaethylene glycol ether

C₁₂E₈

dodecyl octaethylene glycol ether

Introduction

Membrane proteins provide the contact points between the cell and the outside world, and are key players in a remarkable number of cellular processes; including transport of small molecules and ions, energy transduction, interaction of cells with other cells or pathogens, intra‐ and extracellular signaling and enzymatic catalysis.1 Membrane proteins are involved in many different human diseases, and are consequentially also targets for a substantial part of the drugs currently on the market.2, 3 Approximately one third of all genes encode membrane proteins, but their abundance and importance is not reflected in the current body of structural information. There are comparatively very few high‐resolution structures known for membrane proteins, relative to soluble proteins (approximately a 1:100 ratio).4 One reason behind this severe imbalance is that membrane proteins are generally much harder to produce than their soluble counterparts. In the field of membrane protein structural biology, a large amount of the labor time and consumables is commonly spent on identifying protein variants and production conditions that give adequate expression levels and pure, stable, properly folded protein.5, 6 Because of this, membrane protein structure projects are often considered too expensive and time consuming to pursue, especially for smaller laboratories.

The choice of detergent is very important for membrane protein preparation. Many detergents are available for solubilization of membrane proteins, but the best detergent for solubilizing a membrane protein is not necessarily the one that is best for keeping it monodisperse and stable, or allows the protein to form crystals.7, 8 To screen for the right detergent for a particular protein, a medium‐ or large‐scale protein preparation is commonly made in one detergent at a time. The protein is then run through a number of tests to assay homogeneity, multimeric state and stability, including size‐exclusion chromatography (SEC) and polyacrylamide gel electrophoresis before being subjected to crystallization trials, cryo‐EM sample preparation or biochemical studies.9 If the protein does not crystallize in the solubilization detergent, the detergent can also be exchanged, usually for one with smaller micellar size, in order to expose more of the protein and thus increase the chances of forming crystal contacts. For lipidic cubic phase (LCP) crystallization, the detergent used for crystallization becomes less important. During in meso crystallization trials, the protein is stripped of (most of) its detergents upon entering the cubic phase. However, some of the most common detergents have been shown to adversely affect the LCP, causing swelling of the phase structure and the formation of lipidic sponge phase.9, 10, 11 Screening for detergents that solubilize a protein well and can still be kept at low concentrations with little adverse effects on the LCP itself, would therefore be of utmost importance also for in meso crystallization.

With the advent of the direct electron detector and the “resolution revolution”, electron microscopy is now a viable route for obtaining structures of proteins and protein complexes at high resolution.12, 13 Here the choice of detergent is crucial for obtaining a homogenous sample suitable for data collection. Moreover, the ability to rapidly screen for the appropriate expression and purification conditions for a membrane protein or a membrane protein complex will greatly increase the chances of success. To speed up the tedious, iterative procedure of getting a reliable protocol for preparation of a particular target protein, various strategies have been employed to monitor membrane protein quality. In one study, a modified GFP molecule; folding reporter GFP, was C‐terminally fused to a set of globular proteins to enable monitoring of protein folding.14 The amount of GFP fluorescence was shown to be strongly correlated to if the N‐terminally attached protein was properly folded or not. In a subsequent paper focusing on membrane proteins, the GFP moiety would only be fluorescent when the fusion protein was properly inserted in the plasma membrane.15 Another study demonstrated that a membrane protein‐GFP fusion that runs as a single symmetric peak in a fluorescence‐detection size‐exclusion chromatography (FSEC) trace is likely to behave well in subsequent crystallization experiments.16 In the same paper it was also shown that FSEC analysis of a protein can be performed on detergent‐solubilized membranes, without the extensive purification that is needed for other types of detection. Folding reporter GFP has since been employed to screen a number of membrane protein targets for good candidates to move into crystallization trials,17, 18, 19, 20 as well as to determine two‐dimensional topologies for the (almost) entire E. coli and S. cerevisiae membrane proteomes.21, 22

In this article we present a screening protocol specifically designed to accommodate a large number of proteins, or protein variants, and purification conditions simultaneously in one simple, rapid, small‐scale assay. The protocol closely mimics the procedures commonly used in large‐scale protein production, allowing facile upscaling for biochemical and structural studies, while avoiding labor‐intensive ultracentrifugation and membrane preparation steps. The E. coli membrane protein library originally created for the topology determination study21 provides an excellent testing ground for this assay, and as a proof of principle, we applied the protocol to 60 selected membrane proteins. Moreover, guided by the obtained screening data we have performed large‐scale purifications for several of the proteins, highlighting the potential of our protocol to rapidly generate high quality protein samples for structure determination by X‐ray crystallography or cryo‐EM.

Results

We set out to design a protocol for rapid microscale solubilization and affinity purification in a 96‐well format, in order to be able to quickly evaluate whether a particular membrane protein is amenable to production for structural studies or not. For comparative purposes, we selected a battery of 16 commonly used detergents of varying size and head group chemistry, divided into five detergent families (Supporting Information Table SI). The screening protocol was engineered to be as fast and small‐scale as possible, while still simulating large‐scale preparation procedures closely; the latter to be able to identify conditions that allow for upscaling.

For the screening trials, 60 proteins were selected from a folding reporter GFP fusion membrane protein library where the membrane protein is followed by a TEV protease cleavage site, a C‐terminal folding reporter GFP molecule and a C‐terminal His‐tag [Fig. 1(A)].21 The proteins were chosen to cover a wide range of molecular weights and globular/transmembrane domain ratios (Table 1).

Figure 1.

Fusion protein constructs. (A) For the previously published E. coli membrane protein library used in this study, a His‐tagged folding reporter GFP was added C‐terminally of each protein, with a TEV protease site separating the two moieties. (B) For large‐scale production, all target constructs were modified to contain only a C‐terminal 8× His‐tag.

Table 1.

Our Target List of 60 E. coli Membrane Proteins

Target #	Protein name	Size (kD)	Size (aa)	TM helices
1	AroP	50	458	12
2	BglF	66	626	10
3	CcmB	24	221	6
4	CcmC	28	246	6
5	Cmr	44	411	12
6	CodB	44	420	12
7	CreC	52	475	2
8	CybB	20	177	4
9	DcuS	61	544	2
10	FeoB	84	774	10
11	FtsW	46	415	10
12	FtsX	39	353	4
13	GlpT	50	453	12
14	HemY	45	399	2
15	HrsA	70	659	9
16	HtpX	32	294	4
17	Kch	46	418	6
18	KdpD	99	895	4
19	Lgt	33	292	5
20	LspA	18	165	4
21	MdoH	97	848	6
22	MotA	32	296	4
23	MrdB	40	371	9
24	PgpB	29	255	6
25	PheP	51	459	12
26	PhoR	50	432	2
27	PntA	55	511	5
28	PntB	49	463	9
29	PnuC	27	240	7
30	ProP	55	501	12
31	PutP	54	503	12
32	Rbn	33	291	6
33	RfaL	47	420	12
34	Rfe	41	368	11
35	Tap	58	534	2
36	Tar	60	554	2
37	UhpT	51	464	12
38	WcaJ	52	465	6
39	WzxC	54	493	12
40	XasA	55	512	12
41	YbcZ	54	481	2
42	YcdP	16	138	2
43	YcdQ	51	442	6
44	YceJ	21	189	4
45	YdaN	37	328	2
46	YfbF	36	323	2
47	YfeV	50	475	10
48	YgeD	42	398	10
49	YgiY	50	450	2
50	YiaD	22	219	2
51	YihG	36	311	2
52	YiiP	33	301	6
53	YodB	20	177	4
54	ZntA	77	733	6
55	AtpI	14	127	4
56	MreD	19	163	5
57	PgsA	21	183	4
58	CpxA	52	458	2
59	YdgR	54	501	14
60	AmpG	53	492	14

Protocol design

The general flow chart for the protocol is depicted in Figure 2. The procedure was designed for parallel processing of a large number of samples, in order to keep time consumption down to a minimum. The solubilization and purification steps were also miniaturized in order to cut down on consumables, while still keeping the protocol as similar as possible to the large‐scale procedure. The screening assay is performed in a 96‐well format, where the user is free to select what number of proteins versus detergents that should be tested. Here, we applied the assay to 12 proteins at a time (see detailed description below). Briefly, the 12 lysates were dispensed in 96‐well plates and solubilized with 16 of the most common detergents used for X‐ray crystallography of membrane proteins (Supporting Information Table SI), rendering 2 × 96 protein‐detergent combinations. These were subsequently subjected to fluorescence‐detection size‐exclusion chromatography (FSEC). Additionally, the solubilized samples were purified in a microwell format, in order to assess their compatibility with standard IMAC affinity purification procedures.

Figure 2.

Experimental layout. (A) For one experiment, 12 target proteins were solubilized in 16 detergents each. (B) The flow chart briefly describes the different steps of the protocol.

As a starting point for the screening assay, we used the following standard expression conditions (largely adapted from the previous topology mapping study): 150 mL LB media, 25 μM kanamycin, 37°C. The cultures were grown in a large‐scale LEX expression system (LEX; Harbinger Biotech). At OD₆₀₀ ≈ 0.8, expression was induced by adding 0.2 mM IPTG, and after an additional 3 h of growth at 37°C the cells were harvested by centrifugation. The protocol can be miniaturized further; with the expression cultures grown in a multi‐well format (see materials and methods for details). However, we advocate that this step should preferably be performed in the same type of equipment that will later be used for upscaling, in order to minimize discrepancies between screening and large‐scale purification results caused by differences in culture conditions. All cultures were adjusted to the same volume and culture density. After harvesting, the pellets were resuspended in buffer, split in three aliquots, flash‐frozen and stored at −80°C.

The cells were lysed by adding an equal volume of 2× lysis buffer, followed by gentle thawing on ice. After three cycles of freeze‐thawing and mixing by pipetting, the samples were aliquoted for detergent solubilization. This lysis method was chosen as it allows for quick parallel processing of the samples under identical conditions, making it possible to perform the entire solubilization and purification procedure in one day. Lysing the cells by other methods, e.g., sonication or high‐pressure homogenization, is possible but would take considerably longer time and increase the risk for variation between samples. The lysates were aliquoted in a 96‐well format and detergents were added (1:5 of a 5× stock solution). From the solubilization step and onwards all steps were performed in a 96‐well format, allowing for fast processing of up to 2 × 96 protein/detergent combinations at a time.

The proteins were solubilized at 4°C with gentle shaking and large debris was collected by centrifugation. To clean up the samples for FSEC, the samples were filtered by passing them through a 96‐well filter plate, avoiding the need for high‐speed centrifugal clearing of lysates. 10% of the cleared, solubilized lysate was removed and saved for FSEC injection. The remaining sample was mixed with a small volume of IMAC media, added directly to each well. Batch binding was chosen as a strategy as we have found that it is generally more efficient for membrane proteins than loading the sample directly onto an IMAC column. It is also readily compatible with the 96‐well format without the cost of pre‐packed 96‐well IMAC plates. The samples were incubated for 2 h at 4°C with gentle shaking, transferred to a new 96‐well filter plate and centrifuged briefly. After three washes, the proteins were eluted from the IMAC resin using imidazole. The eluates were diluted 5× immediately upon elution, in order to dilute the imidazole and avoid protein precipitation.

Both the filtered, solubilized lysates and the IMAC purified proteins were analyzed by FSEC using a silica‐based size exclusion column. These columns give good separation over a wide molecular weight range and can be run at high pressures (and thus high flow rates). In our setup, the silica column is connected to a HPLC system with autosampler. The latter makes it possible to run FSEC on 192 protein samples without manual intervention (we typically run 96 samples overnight or 192 samples over two nights). We find that the small injection volumes required (typically ≤10 μL) in combination with pre‐filtering of the solubilized lysates, eliminates the need for a labor‐ and time‐consuming ultracentrifugation step. We routinely run 500–1000 samples on the column before an increase in back pressure indicates that cleaning is required. The use of a guard column is still recommended though, as it protects the column and increases its lifespan considerably. Guard columns containing disposable cartridges are preferable, as regular replacement of the cartridges often makes column cleaning unnecessary.

Detergent screening data interpretation

The general aim for the FSEC data analysis is to find protein‐detergent combinations that render strong, sharp fluorescence peaks with minimal trailing and no large molecular weight aggregation products. Preferably, the samples should also retain these properties after being put through the IMAC purification procedure. We have selected a few representative proteins, highlighting “good” and “bad” examples, in order to illustrate the results generated by the protocol (Figs. 3 and 4). The results are also summarized in Table 2, and a complete set of FSEC chromatograms is included as Supporting Information (three of our targets rendered incomplete data due to technical errors and were excluded from the analysis).

Figure 3.

Solubilized lysates. FSEC chromatograms for six representative target proteins from our list of 60. Each graph shows one protein in one family of detergents. All graphs for the same protein are set to the same scale. Solid arrow; free GFP. Dashed arrow; endogenous fluorescent species. Y‐axis shows fluorescence intensity (arbitrary units).

Figure 4.

IMAC purified samples. FSEC chromatograms for the six targets in Figure 3 after IMAC purification. Each graph shows one IMAC purified protein in one family of detergents. All graphs for the same protein are set to the same scale. Solid arrow; free GFP. Y‐axis shows fluorescence intensity (arbitrary units).

Table 2.

Qualitative summary of the FSEC chromatograms

	Lysates					IMAC purified samples
Target	LA	LG	LP	LC	LM	IA	IG	IP	IC	IM
1	4	3	2	5	5	3	2	1	5	5
2	4	1	4	4	4	3	1	3	4	4
3	5	1	2	5	5	3	1	2	4	4
4	4	1	3	5	5	4	1	2	5	5
5	4	1	2	5	5	4	1	2	5	5
6	4	1	2	5	5	3	1	2	5	5
7	5	1	3	5	4	3	1	1	5	5
8	5	5	5	5	5	4	3	4	5	5
9	4	1	4	4	4	3	1	2	4	4
10	4	2	4	4	4	3	2	2	4	4
11	3	2	1	5	5	2	2	1	5	5
12	4	3	2	3	3	4	4	2	5	4
13	4	2	3	5	5	4	1	2	5	5
14	2	2	1	2	2	4	4	4	4	4
15	5	1	3	5	5	5	1	3	5	5
16	1	1	1	1	1	1	1	1	1	1
17	4	4	4	4	4	4	5	3	4	4
18	4	1	1	4	4	4	1	1	4	4
20	5	3	5	5	5	4	2	3	5	5
21	5	2	5	5	5	4	5	2	4	4
22	3	1	2	3	3	4	1	2	4	4
23	3	2	1	5	5	3	2	1	5	5
24	3	1	3	5	5	4	1	1	5	4
25	4	2	4	5	5	4	2	3	5	5
26	4	1	1	5	4	4	1	1	4	4
27	5	4	4	5	5	4	3	3	5	5
28	4	2	2	4	4	4	1	2	4	4
29	4	3	4	5	5	4	2	3	5	5
30	4	1	3	5	5	4	1	2	5	5
31	4	1	2	5	5	4	2	3	5	5
32	3	2	4	5	5	3	2	2	5	5
33	4	1	1	5	5	4	1	1	4	4
34	4	2	2	5	4	4	2	2	4	4
35	1	1	1	1	1	3	1	1	2	2
36	5	5	4	5	5	5	5	3	5	5
37	4	1	2	5	5	4	1	1	5	5
38	1	1	1	1	1	2	1	2	2	2
39	5	2	3	5	5	5	1	3	4	4
40	3	4	4	5	5	3	4	3	5	5
42	1	1	1	1	1	2	1	1	1	1
43	4	1	4	4	4	4	1	4	4	4
44	5	5	4	5	5	5	4	3	5	5
45	3	1	2	3	3	5	2	2	5	4
46	2	1	4	3	3	4	1	4	3	4
47	5	2	3	5	5	5	2	3	5	5
48	4	1	3	4	4	5	2	3	5	5
49	4	1	2	4	3	5	2	1	5	4
50	4	4	4	4	4	5	4	3	5	5
51	3	1	2	3	3	4	3	2	4	4
52	5	4	3	5	5	5	4	2	5	5
53	5	5	5	5	5	4	1	1	4	4
54	4	2	2	4	4	4	1	3	4	4
55	5	3	3	5	5	4	2	2	5	5
56	5	5	4	5	5	5	4	3	5	4
57	5	4	4	5	5	4	2	2	4	4
59	4	1	1	4	4	4	2	1	4	4
60	5	1	4	5	5	4	2	3	5	5

The values (1–5) represent our assessment of the quality of the MP‐GFP fusion protein peak.

L = lysate; I = IMAC; A = amine oxides; G = glucosides; P = PEGs; C = CYMALs; M = maltosides; 5 (red) = very good; 4 (orange) = good; 3 (yellow) = fair; 2 (green) = poor; 1 (blue) = very poor.

FSEC analysis was performed both on the solubilized cell lysates (Fig. 3) and the eluates from IMAC purification (Fig. 4). The results for the solubilized cell lysates (Fig. 3 and Table 2; columns LA‐LM) show that while 6/60 proteins are solubilized by all detergents (e.g. target #44), and 14/60 by members of all detergent families, 28/60 proteins perform very poorly in at least one group of detergents (e.g., targets #24 and #59 in the alkyl glucosides). For most proteins, detergents from the same family (with the exception of the polyethylene glycols) render FSEC chromatograms of similar shape. Overall, Maltosides and CYMALs seem to work well throughout the solubilization trials, while comparatively few of the proteins on our list were adequately solubilized in the polyethylene glycols and the alkyl glucosides (Table 2; columns LG‐LP and Supporting Information Fig. S1). Proteins solubilized in the longer‐chain versions (DDM/UDM, CYMAL‐7/CYMAL‐6) often, but not always, display higher and sharper peaks with less shoulders, trailing and additional aggregate peaks. 9/60 proteins display multiple peaks (e.g., targets #9 and #54), likely representing different multimeric states. For multi‐peak proteins, the distribution between peaks varies between detergent families, and sometimes within the same family.

If proteins with certain properties would consistently solubilize better in a certain detergent family, this could provide rough guidelines for detergent selection. To identify such trends, we made a comparative analysis between the detergent families (Supporting Information Figs. S1–S3). We would however like to caution that this data represents a limited set of proteins that was not randomly selected.

Comparing the best solubilizing detergent from each detergent family, polypeptide length does not seem to affect how efficiently a protein is solubilized, but for all five detergent families there is a weak trend suggesting that an increasing number of transmembrane helices renders an FSEC peak of higher amplitude, indicating that more protein was solubilized (Supporting Information Fig. S1). Normalized values (i.e., relative to the alkyl Maltoside value for the same protein) show that this trend is less pronounced for the non‐Maltosides, especially for the amine oxides which appear to perform better for proteins with few helices than the other detergent families. However, since the variation between the individual proteins is so large we would not advocate selecting a detergent for solubilization solely based on this.

As has been noted before in the literature, the GFP moiety is cleaved off to a certain extent during the protein production process.15, 16, 23 This can also be seen in our chromatograms as a separate peak, or sometimes as a shoulder peak, of lower molecular weight (Figs. 3 and 4, solid arrow). The mechanism for this cleavage is not known. However, in light of the fact that the fusion constructs used in all the three previous studies were engineered so that the target protein and the GFP moiety were separated by a protease cleavage site, the GFP moiety may very well be cleaved in vivo by an unknown proteolytic activity.15 In the cell lysate chromatograms, at least one low molecular weight fluorescent species (dashed arrow) is also visible. However, small endogenous fluorescent species are also present in FSEC traces of GFP‐less His‐only proteins [Fig. 5(A)].

Figure 5.

GFP cleavage. The free GFP peak (Solid arrow) present in the FSEC chromatogram of the GFP fusion of target #47 (B) is absent in the chromatogram of the same protein in its His‐only version (A). The GFP moiety is completely cleaved off when the lysate is prepared in the absence of protease inhibitors (C). Dashed arrows; endogenous fluorescent species.

The IMAC purification step was included in our screening protocol in order to see if a protein‐detergent combination would be compatible with IMAC purification. If the FSEC data after IMAC purification closely resembles the data for the respective lysate, this indicates that a protein can withstand the procedure in a particular detergent (Fig. 4, e.g., targets #24, #59). Most proteins seem to tolerate IMAC purification well in the alkyl Maltoside, CYMAL, and amine oxide families, while PEGs and especially glucosides generally render lower relative yields (Table 2; columns IA‐IM and Supporting Information Fig. S2). However, there is still a lot of variation between proteins and the differences between detergent families is so small that we would not suggest excluding a detergent based on this. Protein size does not appear to impact purification efficiency in any of the detergent families but, intriguingly, there appears to be a weak trend that the relative yields are lower from proteins with more TM helices.

The endogenous fluorescent species is clearly missing after IMAC purification. Surprisingly, the GFP peak (solid arrow) is also reduced for most proteins (i.e., targets #11 and #44). As the constructs are His‐tagged at the C‐terminal end of the GFP molecule, the free GFP would be expected to bind to the IMAC resin and remain in the sample also after being cleaved off. One plausible explanation is that the proteolytic activity is removed at an early stage during the affinity purification process, while the solubilized lysates are exposed during the entire protocol (for practical reasons, the lysates were kept at 4°C for ∼6 h during the IMAC purification procedure, before the FSEC program was started for all samples). Another possible explanation is that the His‐tag on the free GFP is somehow more susceptible to proteolysis than the His‐tag on the membrane‐attached fusion protein. Interestingly, during large‐scale purification of target #24 as a GFP fusion protein, the free GFP is removed during the membrane preparation step, and no subsequent cleavage appears to occur after that. On the other hand, extended incubation of the total lysate makes for almost complete cleavage of the GFP moiety (Fig. 6). Excluding protease inhibitors completely in the lysis buffer during sample preparation for FSEC renders similar results for target #47 [Fig. 5(C)]. This data would suggest that the unknown proteolytic activity is not membrane‐attached and rather resides in the bacterial cytoplasm or periplasm. However, the presence of free GFP in the solubilized lysates generally does not significantly impede the qualitative assessment of the chromatogram for a given protein‐detergent combination. When we analyze the amount of cleaved GFP in the different lysates; neither protein size, nor helix number, nor detergent family appears to affect cleavage very much (Supporting Information Fig. S3).

Figure 6.

GFP cleavage in large‐scale preparations. During large‐scale purification of target #24 as a fusion protein, GFP is cleaved after extended incubation of the total lysate (A), but not in the membrane preparation (B). (the chromatogram of GFP is included as a grey dashed line for reference.).

Effect of the detergent used for FSEC

The fluorescence‐detection size exclusion chromatography (FSEC) was performed in running buffer containing 0.03% DDM and thus, for the majority of our samples, the solubilized protein‐detergent complex was exposed to a new detergent. What effect this would have on the FSEC results is previously not very well described in the literature. In an earlier study, it was noted that running the gel filtration step in DDM will not affect the chromatogram of the protein, but no data was provided to support that statement.16 To thoroughly assess whether this was true for our screening protocol, a subset of proteins were solubilized in five detergents; one from each detergent family. Every solubilized protein was then analyzed by FSEC twice; first with DDM in the running buffer and then with the detergent used for solubilization (Fig. 7). For the most part, the two chromatograms are strikingly similar. There are some protein‐detergent combinations where small differences can be detected (e.g., target #20/C12E8 and target #23/LDAO), but in no case did running the FSEC step in DDM affect the results to an extent that would change the assessment of whether a detergent was suitable or not. The reason for this could be that the gel filtration step is too rapid to change the micellar detergent composition and size, or that the initial separation on the column largely defines the qualitative outcome. It is also possible that the high concentration of the solubilizing detergent in the sample could play a role; with empty micelles co‐migrating and effectively competing with DDM for access to the protein‐containing micelles. This suggests that in order to determine if a detergent can solubilize and keep a protein in a homogenous state, it is not necessary to run the analytical gel filtration in the solubilization detergent.

Figure 7.

Effect of performing all FSEC runs in DDM (solid grey line) instead of the detergent used for solubilization (dashed line). All graphs for the same protein are set to the same scale. Solid arrow; free GFP. Dashed arrow; endogenous fluorescent species. Y‐axis numbers show fluorescence intensity (arbitrary units).

Upscaling

The obvious next step after finding the best expression and purification conditions was to make large‐scale preparations of the protein in one or a few detergents, selected according to the FSEC chromatograms. Because some of the proteins gave rise to quite low overall peak FSEC amplitudes, we first set up a small‐scale expression screening trial for a subset of our targets. Two commonly used growth media (LB and TB) and two expression temperatures (37°C and room temperature) were compared (Supporting Information Fig. S4). Most proteins show stronger expression in TB than LB, while the effect of temperature varies between proteins (for example, target #35 is readily expressed at R.T. but hardly at all at 37°C, while target #54 clearly prefers 37°C). It should be noted though that the expression setups were different from the detergent screening (multiwell plate cultures vs. the LEX system).

For the large‐scale experiments, we made new expression constructs for all 60 proteins, where the C‐terminal TEV‐GFP‐His moiety was removed and exchanged for a C‐terminal 8× His‐tag [Fig. 1(B)]. We decided to modify the DNA constructs in order to keep the large‐scale protocol straightforward, with as few steps as possible, but also to avoid concerns about steric hindrance of the TEV protease cleavage site. We then set out to determine whether the FSEC chromatogram for a given GFP‐tagged protein would be a good representation of how a His‐tagged version of the protein behaves in the final SEC step of our standard large‐scale purification procedure. Eight proteins, both “good” and “bad” ones according to the data from our screening protocol, were initially selected for upscaling. All proteins were purified from membrane preparations using IMAC purification, followed by a size exclusion chromatography step using a preparative scale silica‐based column of similar type as the analytical scale column used for FSEC screening.

Comparing the FSEC data to the 280 nm absorbance SEC data from the corresponding large‐scale purifications, a good yield in the large‐scale preparation clearly correlates with a strong peak in the FSEC chromatogram for the microscale purified protein [Fig. 8(A)]. Correspondingly, little protein is obtained for proteins displaying weak FSEC data [Fig. 8(B)]. As the protein concentration is substantially higher (at least 10–20×) in the large‐scale preparation, proteins could also be more prone to aggregate. This is likely the reason why two of the targets, #25 and (especially) #49, show markedly reduced performance in the large‐scale preparations [Fig. 8(C)]. An alternative explanation for this phenomenon is that the change of C‐terminal tag (from GFP‐His to only His) would somehow have a destabilizing effect on these two proteins. Despite the very different purification scales and the fact that one construct has GFP while the other one does not, the upscaling data fits remarkably well with the screening data, emphasizing the robustness of our protocol.

Figure 8.

Comparison of chromatograms from FSEC screening (top panels; all set to the same scale) and large‐scale protein preparations (bottom panels; all set to the same scale). All eight proteins were solubilized and prepared in DDM. Y‐axis represents fluorescence intensity or 280‐nm absorbance, respectively. X‐axis shows SEC volume. Generally, large‐scale preparations were quite consistent with the FSEC data (A: promising targets and B: difficult targets). For targets #25 and #49 (C.), however, the large‐scale preparations show markedly reduced main peaks (solid arrows), most likely due to aggregation (dashed arrows).

Discussion

Overall, the results from our rapid screening assay correlate well with the large‐scale preparation yields, making our protocol a reliable tool for separating the promising membrane protein candidates from the more difficult ones. Because detergents from the same family (same head group) often render similar peak profiles, a more streamlined version of the assay could include a single detergent from each group. This more limited array of detergents would be useful for screening very large numbers of proteins or protein variants, with the possibility of adding a second, fine‐tuning step by running a more extensive set of detergents for the most promising proteins. Newer classes of detergents could also be included (perhaps at the expense of the PEG family), especially the Neopentyl Glycols which have lately proven to be very good for membrane protein crystallography.24, 25, 26 In addition to detergent screening, other purification conditions could obviously be evaluated using the same rapid procedure; e.g., salt concentration, pH and the effects of various additives. With this assay, protein stability under different conditions can also be efficiently assessed by exposing the samples in the 96‐well plate to various temperatures and/or additives, followed by running the FSEC anew and comparing the results to the initial chromatograms obtained before exposure.

With new, powerful methodology becoming available for structure determination, such as fourth‐generation synchrotrons and high‐resolution cryo‐EM, starting with a high‐quality protein sample is the key to success. Our protocol can significantly reduce the time and investments required for production of high‐quality membrane proteins for structural genomics projects. Moreover, it has the potential to make membrane protein projects attainable also for the small structural biology laboratory.

Materials and Methods

Plasmids and strains

Sixty selected proteins were picked from the E. coli membrane protein‐GFP fusion library.21 The pGFPe and pGFPd vectors used to construct the library have been described previously.14 Plasmid minipreps were made and transferred to the expression strain using standard chemical transformation. For large‐scale purifications, the plasmid vectors were modified by the insertion of two annealed oligonucleotides (GATCCCACCATCACCACCATCATCACCATTAAA and AGCTTTTAATGGTGATGATGGTGGTGATGGTGG) encoding a C‐terminal 8× His‐tag and stop codon between the BamHI and HindIII restriction sites. The insert replaced the TEV protease site, folding reporter GFP and original 8× His‐tag (Fig. 1). All clonings were performed using NEB5‐alpha and all proteins were expressed in T7express‐pLysY (both strains from New England Biolabs, Ipswich, MA).

Expression of proteins for FSEC

All targets were grown at 37°C in standard 250 mL reagent bottles in a LEX large‐scale growth system (Harbinger Biotech, Markham, ON, Canada). Overnight cultures (LB, 25 μg/mL Kanamycin, 0.2% glucose) were diluted 1:100 in 150 mL expression media (LB, 25 μg/mL kanamycin, 0.1% Antifoam 204). Expression was induced with the addition of 0.2 mM IPTG at OD₆₀₀ ∼0.6. Cultures were grown for 4 more hours at 37°C and harvested by centrifugation for 15 min at 4000 g in a benchtop centrifuge. The cultures were adjusted so that each sample corresponded to 200 mL culture with an OD₆₀₀ of 1.0. The cell pellets were resuspended in 15 mL P7 buffer (50 mM phosphate pH 7.0, 200 mM NaCl), split in 3× 5 mL aliquots, flash‐frozen in liquid nitrogen and stored at −80°C.

For the experiments shown in Figure 5, cultures were grown in a multiwell format. Overnight cultures (LB, 25 μg/mL Kanamycin, 0.2% glucose) were diluted 1:100 in 3 mL expression media (LB with 25 μg/mL kanamycin) in 24 × 10 mL deep well plates (Axygen Scientific, CA) and incubated with orbital shaking at 37°C. 0.2 mM IPTG was added at OD₆₀₀ ∼0.8. Cells were harvested after 4 h induction and frozen (−20°C).

Sample preparation for FSEC

5 mL lysis buffer (50 mM phosphate buffer pH 7.0, 200 mM NaCl, 2 mg/mL lysozyme, 100 μg/mL DNase, EDTA‐free cocktail protease inhibitors [Roche Molecular Systems, CA]) was added to 5 mL frozen cell suspension and the cells were thawed on ice. After end over end rotation mixing, the cells were lysed using four cycles of freeze‐thawing in liquid nitrogen/room temperature water, followed by mixing (pipetting up and down). 0.6 mL crude lysate was mixed with 0.15 mL 5× detergent stock (for detergent concentrations; see Supporting Information Table SI) in a 96 × 2 mL deep‐well plate (Axygen Scientific, CA). The mixture was incubated on a plate shaker with orbital shaking for 2 h at 4°C (all detergents from Anatrace, OH). Large debris was pelleted by centrifugation for 15 min at 4000 g in a benchtop centrifuge with plate inserts. The supernatant was filtered by centrifugation for 15 min at 1000 g through a 350 μL Acroprep 0.2 μm bio‐inert membrane 96‐well filter plate (Pall Corp., NY). 50 μL of each sample was transferred to a regular, round‐bottomed 96‐well plate and saved for FSEC.

For the experiments shown in Figure 5, cell pellets were freeze‐thawed directly in the 24‐well growth plate before proceeding to solubilization. These samples were also centrifuged to remove debris, but not filtered, before injection onto the column.

Microplate format metal affinity chromatography (IMAC)

500 μL of each sample was transferred to a fresh deep‐well plate. 10 μL (bed volume) Ni‐NTA superflow (Qiagen, Hilden, Germany) was added (20 μL of 50% stock washed 3× with 50 mM phosphate buffer pH 7.0, 200 mM NaCl), and the plate was incubated 2 h at 4°C with orbital shaking. Samples were loaded onto a fresh 350 μL Acroprep 0.2 μm bio‐inert membrane 96‐well filter plate and centrifuged for 2 min at 1000 g. The plate was washed 3× with 0.2 mL wash buffer (50 mM phosphate buffer pH 7.0, 200 mM NaCl, 25 mM Imidazole and the respective detergent) and proteins eluted with 25 μL elution buffer (50 mM phosphate buffer pH 7.0, 200 mM NaCl, 250 mM Imidazole, EDTA‐free cocktail protease inhibitors and the respective detergent). 100 μL buffer (50 mM phosphate buffer pH 7.0, 200 mM NaCl) without Imidazole was added into the collection plate before elution (to dilute the imidazole and avoid precipitation), and the eluates were subjected to FSEC.

Fluorescence‐detection size‐exclusion chromatography (FSEC)

10 μL of each sample (before and after IMAC purification) was run on a 5 mL BIOSEP‐SEC‐S3000 column (Phenomenex, CA) connected to a Shimadzu Prominence system with 487/512 nm fluorescence detection (Shimadzu Corp., Kyoto, Japan). The flow rate was 0.5 mL/min.

The samples in Figure 5 were run on a 7 mL Yarra‐S3000 column (Phenomenex, CA). The FSEC was performed in running buffer H7‐DDM (25 mM HEPES pH 7.0, 200 mM NaCl, 0.03% DDM). The flow rate was 1 mL/min. For the detergent comparison data in Figure 7, all detergents were added to the FSEC running buffer at a concentration of ∼3× their respective c.m.c. (See Table SI for critical micelle concentrations).

Large‐scale protein expression

The targets were grown at 37°C in standard 2.0 L reagent bottles in a LEX large‐scale growth system. Overnight cultures (LB, 25 μg/mL Kanamycin, 0.2% glucose) were diluted 1:100 in 1.5 L expression media (TB, 25 μg/mL kanamycin, 0.1% Antifoam 204). Expression was induced with 0.2 mM IPTG at OD₆₀₀ ∼1.0. Cultures were grown for 4 more hours at 37°C and harvested by centrifugation for 15 min at 7500 g in a JLA 8.1000 rotor (Beckman Coulter, CA). After harvesting, pellets were transferred to 50 mL tubes, centrifuged briefly, overlayed with 1 mL buffer (50 mM HEPES pH 7.0, 150 mM NaCl) to prevent drying, and frozen at −20°C. For membrane preparations, pellets were resuspended in lysis buffer (50 mM HEPES pH 7.0, 150 mM NaCl, 0.5 mM CaCl₂, 2.5 mM MgCl₂, 1 mg/mL lysozyme, 50 μg/mL DNase, EDTA‐free cocktail protease inhibitors) and lysed in an Emulsiflex C3 system (Avestin, ON, Canada). All lysis and membrane preparation steps were performed at 4°C. Lysates were cleared by 30 min centrifugation at 20,000 g in a Beckman JA 25.50 rotor. The supernatants were transferred to a Beckman Ti45 ultracentrifuge rotor and centrifuged at 40,000 rpm for 1 h. The membrane pellets were resuspended in washing buffer (50 mM HEPES pH 7.0, 150 mM NaCl). After a second round of ultracentrifugation, pellets were resuspended in 10 ml wash buffer, flash‐frozen in liquid nitrogen and stored at −80°C.

Large‐scale protein purification

Membranes were thawed and solubilized in 1% DDM at 4°C. Insoluble material was cleared by ultracentrifugation for 30 min at 40,000 rpm in a Beckman Ti70.1 rotor. The supernatant was incubated at 4°C with Ni²⁺‐charged Profinity IMAC resin (0.5 mL bed volume) for 1 h with gentle rotation and then loaded on a gravity flow column (Bio‐Rad). After washing with 25 ml (∼50 CV) IMAC washing buffer (50 mM HEPES pH 7.0, 150 mM NaCl, 10 mM Imidazole), samples were eluted with 2.5 mL (∼5 CV) IMAC elution buffer (50 mM HEPES pH 7.0, 150 mM NaCl, 250 mM Imidazole). Eluates were run on a 105 mL Yarra 3000 PREP column (Phenomenex, CA) connected to a Shimadzu prominence system with UV/VIS detection (280nm). Fractions were collected in 96 × 2 mL deep‐well plates. Fractions representing the major peaks were pooled, concentrated to 5–10 mg/mL using Vivaspin concentration devices (Sartorius, Göttingen, Germany), flash‐frozen and saved for subsequent structural and functional studies.

Supporting information

Supporting Information

Supporting Information

Supporting Information