Engineering Escherichia coli for Functional Expression of Membrane Proteins

Franz Y. Ho; Bert Poolman1    Department of Biochemistry, Groningen Biomolecular Sciences and Biotechnology Institute & Zernike Institute for Advanced Materials, University of Groningen, Groningen, The Netherlands
1 Corresponding author: email address: b.poolman@rug.nl

1 Introduction

Integral membrane proteins contribute 15–30% of the open-reading frames found in the genomes of organisms from all domains of life (Bendtsen, Binnewies, Hallin, & Ussery, 2005; Wallin & von Heijne, 1998). However, our understanding of function–structure relationships of membrane proteins is low compared to water-soluble proteins (Granseth, Seppälä, Rapp, Daley, & Von Heijne, 2007). There are many difficulties in studying membrane proteins, and the challenge already starts with the expression of the genes and production of the proteins in a functional state. Recombinant expression in heterologous hosts is the most versatile strategy, but difficult to produce proteins often cause toxicity to the cells and end up themselves in insoluble aggregates (Bill et al., 2011). Expression conditions can be optimized but this is a laborious process (Francis & Page, 2010).

Large-scale production of proteins is commonly achieved using microorganisms, because of the ease of growth and many genetic and biochemical tools are available. Several factors determine the production level and functionality of the proteins, such as (i) the toxicity the proteins evoke to the host cells (Wagner et al., 2007), (ii) the posttranslational modifications required for structure and/or functionality (Grisshammer & Tate, 1995), (iii) the difference in codon usage between the recombinant gene and the expression host (Angov, Hillier, Kincaid, & Lyon, 2008), (iv) limitations of protein synthesis precursors such as tRNA and amino acids (Marreddy et al., 2010; Puri et al., 2014), (v) overloading of foldase and chaperone activities (Tate, Whiteley, & Betenbaugh, 1999), (vi) saturation of the membrane protein insertion machinery (Loll, 2003; Wagner, Bader, Drew, & de Gier, 2006), and (vii) uncoordinated protein biosynthesis kinetics (Bill et al., 2011; De Marco, 2013). Membrane proteins have a complex biogenesis pathway, requiring chaperones and machineries for membrane insertion, protein folding, and oligomerization, but often the limiting factor(s) leading to low or nonfunctional expression is not understood. Rather than screening numbers of expression conditions, it is easier to improve the production by evolving the host.

In this method, we describe a simple procedure for engineering Escherichia coli by forcing the cells to produce more desired protein by fusing a folding indicator and antibiotic resistance marker to its C-terminus (Fig. 1A). The same strategy has been proven successful for the overexpression of membrane proteins in both the Gram-positive bacterium Lactococcus lactis and the Gram-negative bacterium E. coli (Gul, Linares, Ho, & Poolman, 2014; Linares, Geertsma, & Poolman, 2010).

1.1 Green fluorescent protein as a folding reporter

Protein folding can start right after the nascent chain is formed, but in case of membrane proteins, it is delayed as translation and membrane insertion are coupled and folding is completed after the last transmembrane segments have left the translocon (Luirink, von Heijne, Houben, & de Gier, 2005). In multidomain proteins, domains are tethered by either structured or disordered linkers and the rate of folding of individual domains can differ (Arviv & Levy, 2012; George & Heringa, 2002; Jappelli, Luzzago, Tataseo, Pernice, & Cesareni, 1992; Zhang & Ignatova, 2009). In the present method, green fluorescent protein (GFP) is used as a global folding reporter, which requires fusion of the fluorescent protein to the C-terminus of the target membrane protein. Thus, the synthesis of the membrane protein precedes that of the fluorescent protein. The folding of a membrane protein is typically faster than the folding and maturation of GFP, which takes about 30–90 min (Evdokimov et al., 2006; Reid & Flynn, 1997; Waldo, Standish, Berendzen, & Terwilliger, 1999). Therefore, when the target membrane protein is misfolded, it is likely to drag the not yet fully synthesized, folded, and/or matured GFP into an aggregated state. In contrast, when the target membrane protein is properly folded, the GFP β-barrel is formed and the chromophore will mature. Proper folding of the protein protects the chromophore from quenching by water dipoles, paramagnetic oxygen, or cis–trans isomerization (Tsien, 1998). The maturation process involves a series of covalent rearrangements of the amino acids that form the tripeptide chromophore (Ser/Thr65, Tyr66, and Gly67) within an α-helix that is buried inside the hollow β-barrel. When the chromophore matures, GFP becomes SDS resistant, and the protein migrates faster in SDS-PAA than the fully denatured polypeptide. The apparent difference in migration is about 10 kDa, and this difference is also observed when GFP is fused to another protein (Geertsma, Groeneveld, et al., 2008). Thus, when a membrane protein is well folded, the C-terminal GFP will fully mature, and the whole fusion protein migrates faster on SDS-PAA than the misfolded membrane protein–GFP fusion. By analyzing protein expression on immunoblots, one can easily discriminate the well-folded and misfolded proteins (Drew, Lerch, Kunji, Slotboom, & de Gier, 2006; Geertsma, Groeneveld, et al., 2008). Moreover, the fluorescence reports the absolute amount of folded fusion protein, which can be observed not only in gel but also in vivo (Drew et al., 2006; Linares et al., 2010).

We and others have observed that the in vivo activity of transport proteins correlates with the corresponding in-cell and in-gel GFP fluorescence of the fusion proteins (Geertsma, Groeneveld, et al., 2008; Hibi et al., 2008; Schlegel et al., 2012). In brief, the glutamate...