Post-expression strategies for structural investigations of membrane proteins

Abstract

Currently, membrane proteins only comprise 1.5% of the protein data bank and, thus, still remain a challenge for structural biologists. Expression, stabilization in membrane mimics (e.g. detergent), heterogeneity (conformational and chemical), and crystallization in the presence of a membrane mimic are four major bottlenecks encountered. In response, several post-expression protein modifications have been utilized to facilitate structure determination of membrane proteins. This review highlights four approaches: limited proteolysis, deglycosylation, cysteine alkylation, and lysine methylation. Combined these approaches have facilitated the structure determination of more than 40 membrane proteins and, therefore, are a useful addition to the membrane protein structural biologist’s toolkit.

Introduction

Membrane proteins are the gatekeepers of cells regulating the flow of information and small molecules across the cell membranes. To better understand membrane proteins, structural biologists determine their high-resolution structure using X-ray crystallography or NMR spectroscopy. The structure provides a starting-point to base hypotheses about how proteins function. Most membrane proteins exist in at least two states – “open or closed” or “on and off” – and move between these states in order to mediate the movement of a signal or molecule across the membrane. A structure potentially provides a snapshot of the protein in one of these states. Obtaining these high-impact structures remains a challenge as indicated by their underrepresentation in the protein data bank.

Membrane protein structural biology has required a multi-tier, highly empirical approach to obtain samples that are amenable to the structural techniques commonly used for soluble proteins. Much attention is paid to the construct (e.g. organism, chimeras, and mutations), expression, and membrane mimic selection [1,2]. However, several structures have required post-expression protein modifications. The post-expression modifications improve structure determination by a variety of mechanisms such as creating homogenous crystal contacts, trapping a single conformational state, and removing flexible regions. This review highlights the use of limited proteolysis, deglycosylation, reductive methylation, and cysteine alkylation (Figure 1) as post-expression modification approaches to structural investigations of membrane proteins.

Figure 1.

Post-expression modifications that have contributed to membrane protein structural investigations. Limited proteolysis cleaves exposed and dynamic regions of membrane proteins and has been applied to membrane proteins for x-ray crytallogaphy structure determination and solution NMR assignments. The scissor icon represents a protease, which is commonly chymotrypsin or trypsin. Glycosidases are used to remove post-translational glycosylations (deglycosylation) to facilitate structure determination. The scissor icon represents the site of cleavage for the two types of enzymes. Reductive methylation and cysteine alkylation are covalent modifications that are used to facilitate conformational equilibria or crystallization, respectively.

Limited proteolysis

Limited proteolysis is the treatment of a protein with a protease such that only exposed and dynamic regions (not folded domains) are cleaved according to the protease selectivity. In addition, for membrane proteins, the transmembrane regions are protected from protease cleavage by the membrane or membrane mimic. The proteolytic product is typically evaluated using sodium dodecyl sulfate (SDS) polyacrylamide gel electrophoresis and mass spectrometry. The proteolytic cleavage is typically performed during the purification steps (which are mostly done in detergents) and the small cleaved segments are removed from the sample before structure determination.

In some cases, limited proteolysis is used to map out the domains of a protein. The results are then used to guide the molecular cloning of specific domains of interest. For example, the flexible periplasmic domains of the 809-amino acid P pilus usher PapC prevented crystallization and limited proteolysis was used to identify a 55 kD fragment corresponding to the outer membrane translocation domain [3]. The 55 kD fragment was then expressed, purified, and crystalized and the structure was determined with 3.2 Å resolution [3]. In some cases, the new genetic constructs may have lower expression levels, interfere with cellular targeting, or result in instability during purification. Thus, limited proteolysis during the purification may be preferred. Limited proteolysis can be used to remove affinity tags that do not have specific cleavage sites especially in the case of C-terminal tags for which it is desirable to not have additional residues (e.g. using carboxypeptidase for the µ- and δ- opioid receptors [4,5]). Beyond tag removal and domain mapping, limited proteolysis has facilitated nuclear magnetic resonance (NMR) and X-ray crystallography structure determination.

Limited proteolysis: NMR

In order to determine an NMR structure, an assignment of observed resonances needs to be completed. For large proteins or complexes, the assignment process can be challenging. Various assignment strategies exist such as amino acid specific labeling [6,7] and segmental labeling [8]. Another approach is limited proteolysis where loop regions are removed with a protease yet the fragmented protein remains folded. The removal of these flexible regions simplifies the NMR spectrum by removing intense resonances that have spectral overlap with weaker resonances from the folded membrane region. Limited proteolysis has proven to be an effective method for assigning β-barrel membrane protein resonances [9]. The β-barrel membrane proteins are highly stable once folded in detergent as indicated by their resistance to unfolding in SDS (even boiled) and, thus, are likely more amenable to this approach than α- helical membrane proteins.

To date, less than ten β-barrel membrane protein structures have been determined with solution NMR. Most of the proteins investigated do not have significant soluble portions in the periplasmic or extracellular loops. However, in the case of Opa₆₀, the four extracellular loops comprised approximately half of the protein. These loop resonances are more intense than those of the β-barrel and, in many cases, had significant line broadening (Figure 2B). The loop resonance occluded many of the resonances from amino acids within the barrel. β-barrel membrane proteins are typically very stable and remain folded in SDS; thus, the folded and unfolded protein forms migrate differently on an SDS-PAGE gel (Figure 2A). As a result, limited proteolysis of β-barrel membrane proteins in a membrane mimic with a protease may not result in denaturation. For Opa₆₀ [10], Opa₅₀, and OprH [11], the resulting five fragments after trypsin proteolysis maintained the β-barrel fold and the corresponding resonances were nearly superimposable with that of the untreated proteins (Figure 2B) [9]. The ability to conduct the assignments on the proteolysed sample and map the assignment on to the full length was essential to determining the structure of the Opa₆₀ β-barrel [9].

Figure 2.

Limited proteolysis facilitate NMR assignments. (A) SDS-PAGE of Opa₆₀ (Lanes 1–4), Opa₅₀ (Lanes 5–8), and OprH (Lanes 9–12) 8 M urea (unfolded), folded, cleaved with trypsin, and boiled after trypsin cleavage. (B) Full-length (red) and trypsin-treated (blue) ¹⁵N, ¹H-HSQC spectra of ²H, ¹⁵N-labeled Opa₆₀.

Limited proteolysis: Crystallization

Limited proteolysis has also been a useful tool for obtaining x-ray crystal structures. The strategy is different than that used for NMR structure determination. Limited proteolysis is used to (i) remove dynamic soluble regions and/or (ii) to remove soluble domains that may interfere with crystallization. The first KcsA potassium channel crystal structure was the first demonstration of the use of limited proteolysis for structure determination [12]. In this case, chymotrypsin was used to cleave 45 residues from the C-terminus and the cleaved purified product yielded well-diffracting crystals. An interesting note to this protocol was the finding that certain commercial chymotrypsin batches had a trypsin contaminant that yielded a second cleavage product that interfered with crystallization [13]. After demonstrated success of limited proteolysis in the structure determination of KcsA, the method was applied to thirteen other membrane proteins from 1998 to 2014 (Table 1). In all structures determined, either trypsin or chymotrypsin was used as the protease. The cleaved regions tend to be significant portions of the termini. For instance, the chymotrypsin treated LptD:LptE complex resulted in N-terminal residues 22-224 removed from LptD and residues 170-196 from LptE [14]. The method does not seem limited in the membrane protein fold or function investigated as limited proteolysis has been effective for the structure determination of α-helical and β-barrels and a variety of functions (e.g. channels, proteases, and transporters).

Table 1.

X-ray crystal structures determined with post-expression modification^*

Protein	Organism of origin	Expression system	PDB	Reference, year
Limited proteolysis
KcsA	Streptomyces lividans	Escherichia coli	1BL8	[12], 1998
TolC	E. coli	E. coli	1EK0	[44], 2000
GIpG	E. coli	E. coli	2IC8	[45], 2006
VirV7:VirB9:VirB10(T4S)	E. coli	E. coli	3JQO	[46], 2009
AQP4	Homo sapiens	Pichia pastoris	3DG8	[47], 2009
FimD-FimC	E. coli	E. coli	3RFZ	[48], 2011
KvLm PM	Listeria monocytogenes	E. coli	4H33	[49], 2012
UT-B urea transporter	Bos taurus	Spodoptera frugiperda	4EZC	[50], 2012
FimD-FimC-FimF-FimG-FimH	E. coli	E. coli	4J30	[51], 2013
CsgG	E. coli	E. coli	4Q79	[52], 2014
LptD:LptE	Salmonella enterica	E. coli	4N4R	[14], 2014
Serotonin 5-HT3	Mus musculus	HEK293F cells	4PIR	[20], 2014
YbgH peptide transporter	E. coli	E. coli	4Q65	[53], 2014
P-glycoprotein	Cyanidioschyzon merolae	Komagataella pastoris	3WME	[54], 2014
Deglycosylation
AQP1‡	B. taurus	B. taurus	1J4N	[55], 2001
β2 adrenergic receptor‡	H. sapiens	S. frugiperda	2R4R	[56], 2007
A2A adenosine receptor‡	H. sapiens	S. frugiperda	3EML	[57], 2008
Dopamine D3 receptor‡	H. sapiens	S. frugiperda	3PBL	[58], 2010
CXCR4 chemokine receptor‡	H. sapiens	S. frugiperda	30DU	[26], 2010
GPR40 free fatty-acid receptor 1‡	H. sapiens	S. frugiperda	4PHU	[59], 2014
P2Y12‡	H. sapiens	S. frugiperda	4NTJ	[60], 2014
NMDA receptor†	Xenopus laevis	HEK293S GnTI-cells	4TLL	[37], 2014
GluA2 Glutamate receptor (AMPA)†	Rattus norvegicus	HEK293S cells	4U2P	[61], 2014
Serotonin 5-HT3‡	M. musculus	HEK293F cells	4PIR	[20], 2014
0X2 orexin receptor‡	H. sapiens	S. frugiperda	4S0V	[62], 2015
Cysteine alkylation
β2 adrenergic receptor	H. sapiens	S. frugiperda	2RH1	[63,64], 2007
A2A adenosine receptor	H. sapiens	S. frugiperda	3EML	[57], 2008
Dopamine D3 receptor	H. sapiens	S. frugiperda	3PBL	[58], 2010
Histamine H1 receptor	H. sapiens	P. pastoris	3RZE	[65], 2011
Sphingosine 1-phosphate receptor	H. sapiens	S. frugiperda	3V2W	[66], 2012
κ-opioid receptor	H. sapiens	S. frugiperda	4DJH	[67], 2012
μ-opioid receptor	H. sapiens	S. frugiperda	4DKL	[5], 2012
δ-opioid receptor	H. sapiens	S. frugiperda	4EJ4	[4], 2012
Nociceptin/orpiianin FQ (N/OFQ) receptor	H. sapiens	S. frugiperda	4EA3	[68], 2012
Protease-activated receptor 1	H. sapiens	S. frugiperda	3VW7	[69], 2012
M3 muscarinic acetylcholine receptor	H. sapiens	S. frugiperda	4DAJ	[70], 2012
M2 muscarinic acetylciioline receptor	H. sapiens	S. frugiperda	4MQS	[71], 2013
5-HT1B serotonin receptor	H. sapiens	S. frugiperda	4IAR	[72], 2013
5-HT2B serotonin receptor	H. sapiens	S. frugiperda	4IB4	[73], 2013
Smoothened receptor	H. sapiens	S. frugiperda	4JKV	[74], 2013
P2Y12	H. sapiens	S. frugiperda	4NTJ	[60], 2014
Metabotropic Glutamate Receptor 1	H. sapiens	S. frugiperda	40R2	[75], 2014
Metabotropic Glutamate Receptor 5	H. sapiens	S. frugiperda	4009	[76], 2014
GPR40 free fatty-acid receptor 1‡	H. sapiens	S. frugiperda	4PHU	[59], 2014
US28 chemokine receptor	Human cytomegalovirus	HEK293S GnTI- cells	4XT1	[77], 2015
OX2 orexin receptor	H. sapiens	S. frugiperda	4S0V	[62], 2015
Disulfide crosslinking
Glutamate transporter (GItph)	Pyrococcus horikoshii	E. coli	3KBC, 3V8F	[41], 2009 [42], 2012
β2 adrenergic receptor	H. sapiens	S. frugiperda	3PDS	[40], 2011
ABC transporter BtuCD	E. coli	E. coli	4F13	[39], 2012
NMDA receptor	Xenopus laevis	HEK293S GnTI- cells	4TLL	[37], 2014
Heterodimeric ABC exporter TM287/288	Tiiermotoga maritima	E. coli	4Q4J	[38], 2014
CXCR4 chemokine receptor	H. sapiens	S. frugiperda	4RWS	[78], 2015

^*For proteins with multiple structures determined, the first report is provided.

^†Endoglycosidase H was used,

^‡PNGaseF was used.

Deglycosylation

Membrane proteins isolated from natural sources or expressed in eukaryotic systems can be glycosylated. In some cases, glycosylation may not interfere with crystallization (ref [15,16] for example); however, for other proteins, glycosylation can inhibit crystallization often due to heterogeneous glycosylation. One approach to avoid glycosylation in recombinant systems is to mutate the glycosylated residue so the expressed mutated protein is not glycosylated. This is especially necessary if the modification is not accessible to glycosidases [17].The mutation approach has been used for many membrane proteins (ref [18] for example); however, for ten membrane protein structures, glycosidases were used (Table 1). Three different glycosidases, Peptide-N-glycosidase F (PNGase F), endoglycosidase F and endoglycosidase H, were exploited to remove glycosylation modifcations from asparagine residues (Figure 1). Endogycosidase F and H cleave after the first N-acetylglucosamine in high mannose or hybrid glycans. PNGase F cleaves before the first N-acetlyglucosamine on the asparagine residue of high mannose, hybrid, and complex oligosaccharides.

The first demonstration of the use of glycosidases for membrane protein crystallography was with AQP1. AQP1 was isolated from bovine red blood cells and both glycosylated and non-glycosylated forms were observed [19]. PNGase F was incubated with AQP1 after solubilization from the membrane for two days and then further purified. Since 2001, nine proteins have been expressed with glycosylation and treated with glycosidases (Table 1). The most recent membrane protein structure obtained using PNGase F was the serotonin 5-HT3 receptor, which also was subject to trypsin limited proteolysis [20]. In the case of PNGase, the entire modification is removed before crystallization. Therefore, the glycosylation is important for protein production, but hinders crystallization and is removed. In contrast, when EndoH is used an N-acetlyglucosamine remains on the modified asparagine residue. The NAMDA (Fig 3) and AMPA receptors are the only demonstration of the use of EndoH and in both cases, N-acetylglucosamine is observed at crystal contacts.

Figure 3.

The structure and crystal packing of the NMDA receptor. For one hetero-tertramer, GluN1 is colored orange and GluN2B subunit is colored blue. N-acetylglucosamine and modified asparagine residues are rendered with spheres and colored according to atom type. The disulfide crosslink between GluN2B is shown as red spheres. A symmetry mate is colored gray to highlight the N-acetylglucosamine moiety at the crystal contact.

Cysteine Alkylation

Treatment of proteins with iodoacetamide (after the addition of a reducing agent) alkylates any accessible cysteine thereby preventing the cysteine residue from forming disulfide bonds. The importance of this modification was recently illuminated for the β₂ adrenergic receptor [21]. A comparison of the size exclusion chromatogram of the iodoacetamide treated and untreated protein demonstrated that alkylation prevented dimer and higher order oligomers from forming [21]. Since the application of cysteine alkylation for the β₂ adrenergic receptor in 2008, the method has been applied to the structure determination of 21 GPCRs (Table 1). The GPCR structures that did not utilize cysteine alkylation are rhodopsin, NTS1 neurotensin receptor (selected using directed evolution approaches [22] or engineered for thermostability [23]), glucagon receptor [24], corticotropin-releasing factor receptor 1 [25], CXCR4 chemokine receptor [26], CCR5 chemokine receptor [27], and β₁ adrenergic receptor [28] (the four later proteins were engineered for thermostability). Although to date cysteine alkylation has not contributed to the structure determination of other types of membrane proteins, the approach should be considered if a protein has required cysteine residues in flexible regions of the protein that could facilitate disulfide mediated dimers.

Lysine Methylation

Accessible lysine residues and amino termini can be mono- and dimethylated using sodium borohydride and ¹³C labeled formaldehyde as the methyl donor. The incorporated ¹³C-labeled methyl can be detected by NMR and, coupled with a high resolution structure, ligand-binding and ligand-induced conformational changes can be investigated. The resulting ¹³C-methylated receptor can be investigated with NMR, specifically using ¹H,¹³C-HSQC (heteronuclear single-quantum coherence) or saturation transfer differencing-filtered HMQC (heteronuclear multiple-quantum coherence) spectra. Each resonance of the ¹H, ¹³C correlation spectrum can be assigned using lysine to arginine mutations and monomethylated resonances are distinct from dimethylated.

This approach was used to investigate the conformational equilibrium of the extracellular loops of the β2 adrenergic receptor upon ligand binding [29]. For the unliganded receptor, two methyl peaks are significantly shifted upfield from the other observed methyl resonances (indicating a different chemical environment) and were assigned to the dimethylated Lys305. In the crystal structure, Lys305 (at the transition from TM7 and extracellular loop 3, which connects transmembrane helices 6 and 7) forms a salt bridge with Asp192 (in extracellular loops 2, which connects transmembrane helixes 4 and 5). Based on mutagenesis and several other experiments, the chemical shift perturbation of Lys305 was concluded to be due to the formation of the salt bridge. These peaks were monitored with different ligands to investigate the conformational equilibria of the salt bridge. When agonist carazolol is bound, the two methyl peaks shift downfield and the chemical shift difference between the two methyl peaks decreases. Thus, binding of carazolol alters the chemical environment surrounding the Lys305 methyls. Binding of neutral antagonists, such as alprenolol, do not cause chemical shift changes in the observed Lys305 methyl resonances with the peaks remaining unchanged compared to unliganded receptor. Agonists, such as formoterol, also modulate the Lys305 methyl resonances; however, rather than chemical shift changes, the peaks are no longer observed indicating a loss of the salt bridge. The combined results led to the hypothesis that there are three distinct conformations with respect to TM6 and TM7 that are modulated by the Lys305 – Asp192 salt bridge, which in turn is coupled to ligand binding. These results complement the static crystal structures by providing conformational dynamics and defining the energetic landscape modulated by diffusible ligands.

Beyond NMR investigations, it is interesting to note that reductive methylation has improved soluble protein crystallization [30-35] and the approach may be useful in membrane protein crystallization. Reductive methylation in soluble proteins improves resolution and decreases B factors compared to non-methylated native structures [35]. A more ordered crystal packing appears to be the result of an increase in the interaction radius of the lysine residue upon methylation [35]. Reductive methylation and alkylation is an inexpensive approach to modifying protein surface properties that could be used in conjunction or as an alternative to mutagenesis required for membrane protein crystallization [18,28,36].

Other post-expression considerations

There are a few other post-expression modifications and treatments that are worth mentioning in this review: engineered disulfides or crosslinking and ligand complexes. At least five membrane protein structures were determined with engineered cysteine residues for disulfide formation [37-40] or cross linking with mercury [41,42] (Table 1). The engineered cysteine residues are used to improve crystal diffraction quality and/or stabilize alternative protein conformations. In the case of β2 adrenergic receptor, a disulfide crosslink was formed between the protein and a low affinity agonist in order to trap the agonist-bound protein conformation [40]. The cysteine residues cross-linked with mercury stabilized a conformation of the glutamate transporter distinct from the outward facing conformation observed without the disulfide [41]. For the NMDA structure, the disulfide improved the diffraction quality of the crystals (compared to the non-crosslinked structure) by decreasing the conformational flexibility of the extracellular domains (Fig 3) [37].

Although not a covalent modification, protein-protein and protein-ligand complexes are often formed after protein expression. These complexes can be difficult to form and can require 24 hour incubation periods. For instance, to obtain the SecA-SecYEG complex with ADP:BeFx ligand bound, the complex was dialyzed overnight with buffer containing ADP and BeFx [43]. Simply adding the ligands directly and setting up crystallization trays did not yield complex formation. To form the β₂ adrenergic receptor–G protein complex, over 50 agonists were screened for complex stabilizaton and BI-167107 was determined to provide significant stabilization of the G-protein complex with a half-time of ~30 hours. To form the complex, the two proteins were incubated for 3 hours in the presence of agonist. The complex was further stabilized with a nanobody, which was added to the agonist stabilized complex and incubated for 1 hour before crystallization. These two examples highlight the difficulties in forming membrane protein complexes and different approaches used to obtain structures of these complexes.

Summary

In the soluble structure determination field, many techniques have been developed for recalcitrant proteins. All membrane proteins fall into this challenging category and almost every stage of preparation requires screening of conditions. Although daunting in the number of avenues to try, post-expression modifications provide additional options for the membrane protein structural biologists. The approaches described in this review are not specific to membrane proteins; however, their use for membrane proteins has been demonstrated and should be added to the toolkit of membrane protein structure determination. Additional modifications specific to the properties of membrane proteins (e.g. amino acid modifications that increase solubility) are uncharted territory and may lead to significant improvement in the quality and number of membrane protein structures.

Highlights.

Post-expression modifications have enabled membrane protein structure determination.

Limited proteolysis removes flexible regions.

Deglycosylation removes heterogeneous post-translational modifications.

Cysteine alkylation prevents disulfide dimers and higher order oligomers.

Other modifications allow the structure determination of alternative conformations.