Engineering G protein-coupled receptor expression in bacteria.

A pproximately 800 genes in humans are dedicated to the production of a large superfamily of seven-transmembrane helical receptors called G proteincoupled receptors (GPCRs) (1). These proteins are responsible for the intracellular chemical communication and for the sensory input by chemical stimuli (smell and taste). Consistent with their central role in human physiology, GPCRs constitute the largest class of targets for drug discovery with approximately one-third of all approved therapeutics acting by modulating GPCR function (1). Until very recently, however, only one GPCR structure had been solved, that corresponding to the signaling-off state of bovine rhodopsin (2). During the past year, we have experienced a revolution in GPCR structural biology, with five additional structures published, including those of the human 2-adrenergic receptor ( 2AR) in the absence and presence of an inverse agonist (3, 4) and the turkey 1AR complexed with a high-affinity antagonist (5). Collectively, these structures have opened new vistas for the structural modeling of other receptors, and for the virtual screening of ligand libraries by in silico ligand docking (6). A new study by Sarkar et al. in this issue of PNAS (7) describes a strategy for engineering GPCRs that may help increase the number of solved structures even further. The structural analysis of GPCRs poses formidable challenges. Most of these proteins are found in very low abundance in their natural tissues (8). With the exception of bovine rhodopsin, which can be isolated from retinae, GPCRs have to be expressed and purified from heterologous hosts. For example, protein expressed in insect cells was used for the crystallization of 1AR and 2AR (3–5). However, heterologous expression typically results in low yields of membrane-integrated receptor, or in protein that is largely inactive. Further, recombinant GPCRs often accumulate in an aggregated state within inclusion bodies, which are challenging to denature, solubilize, and refold (8). Even when a reasonable expression level of active receptor can be achieved, the instability of GPCRs in detergent micelles and their conformational f lexibility make the formation of high-quality diffracting crystals exceedingly difficult (6). In the case of 1AR, useful crystals could only obtained by using a variant with enhanced thermostability in detergent micelles (5). Similarly, crystallization of 2AR was possible only after a highly flexible loop was constrained, either by the use of a specific loopbinding antibody fragment or by its complete replacement with a well folded soluble protein (T4 lysozyme) (3, 4). Sarkar et al. (7) describe an elegant strategy for GPCR engineering that promises to kill a number of hard-tocatch birds with one stone, namely, the engineering of GPCRs for: (i) high-yield expression of active protein in a convenient host organism; (ii) enhanced stability in detergent micelles; and (iii) defined ligand selectivity. The authors expressed a large library of variants of the rat neurotensin receptor-1 (NTR1) in the bacterium Escherichia coli and sought to isolate clones exhibiting enhanced ligand binding. NTR1 was expressed as a tripartite fusion to maltosebinding protein and to thioredoxin, both of which are known to prevent aggregation (9, 10). For the isolation of desired mutants, Sarkar et al. adapted a highthroughput protein-screening methodology called periplasmic expression with cytometric screening (PECS) (11). In brief, a library of NTR1 mutants was first expressed in E. coli (Fig. 1). Incubation of the cells in a high-osmolarity buffer allowed a BODIPY-labeled fluorescent conjugate of the oligopeptide neurotensin to cross the permeability barrier that is normally imposed by the bacterial outer membrane and equilibrate into the periplasmic space. Cells expressing NTR1 variants with enhanced ligand-binding activity sequester more ligand and as a result display brighter BODIPY fluorescence (green) than cells producing wild-type (wt) NTR1. These cells can be easily isolated by fluorescence-activated cell sorting (FACS). Sarkar et al. applied three rounds of mutagenesis by error-prone PCR, each followed by four sequential FACS enrichment steps, and the isolated genes were subjected to a fourth round of error-prone PCR and gene shuffling. The resulting library was screened by four additional FACS steps (a total of