Protein ScienceVolume 18, Issue 1 p. 121-133 Article Free Access A green fluorescent protein screen for identification of well-expressed membrane proteins from a cohort of extremophilic organisms Justus Hammon, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330 Justus Hammon and Dinesh V. Palanivelu contributed equally to this work.Search for more papers by this authorDinesh V. Palanivelu, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330 Justus Hammon and Dinesh V. Palanivelu contributed equally to this work.Search for more papers by this authorJoy Chen, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this authorChintan Patel, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this authorDaniel L. Minor Jr., Corresponding Author daniel.minor@ucsf.edu Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this author Justus Hammon, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330 Justus Hammon and Dinesh V. Palanivelu contributed equally to this work.Search for more papers by this authorDinesh V. Palanivelu, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330 Justus Hammon and Dinesh V. Palanivelu contributed equally to this work.Search for more papers by this authorJoy Chen, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this authorChintan Patel, Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this authorDaniel L. Minor Jr., Corresponding Author daniel.minor@ucsf.edu Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Cardiovascular Research Institute, Department of Biochemistry and Biophysics, Department of Cellular and Molecular Pharmacology, California Institute for Quantitative Biosciences, University of California, San Francisco, California 94158-2330Search for more papers by this author Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URLShare a linkShare onEmailFacebookTwitterLinked InRedditWechat Abstract Green fluorescent protein (GFP) fusion proteins provide a potentially facile tool for identification of well expressed, properly behaved membrane proteins for biochemical and structural study. Here, we present a GFP-expression survey of 300 membrane proteins from 18 bacterial and archaeal extremophiles, organisms expected to be rich sources of membrane proteins having robust biophysical properties. We find that GFP-fusion fluorescence intensity is an excellent indicator of over-expression potential. By employing a follow-up optimization protocol using a suite of non-GFP constructs and different expression temperatures, we obtain 0.5–15 mg L−1 expression levels for 90% of the tested candidate proteins that pass the GFP screen. Evaluation of the results suggests that certain organisms may serve as better sources of well-expressed membrane proteins than others, that the degree to which codon usage matches the expression host is uncorrelated with success rate, and that the combination of GFP screening and expression optimization is essential for producing biochemically tractable quantities of material. Introduction Membrane proteins are the gateways for signals that pass between a cell and its surroundings and are central to ionic signaling mechanisms and the material transport required for nutrient uptake and toxin disposal. Understanding membrane protein molecular architecture remains one of the major challenges of modern structural biology efforts. One of the main obstacles is the difficulty associated with obtaining sufficient quantities of pure material for biochemical and structural analysis.1, 2 Because most membrane proteins are not found in readily accessible over-enriched native sources, it is necessary to develop ways to overproduce membrane protein constructs at will. The ability to overproduce proteins in heterologous systems has revolutionized structural biology studies of soluble proteins but similar scale successes have not yet been had for membrane proteins. Thus, pursuit of robust methods to express and characterize membrane proteins remains an area of vigorous research efforts. The most rigorous test of whether a protein can be crystallized is to overexpress, purify, and place the protein of interest in crystallization trials. However, taking any given protein through this procedure requires a large investment of time and resources and has a high probability of a negative outcome for any given candidate. There have been ongoing efforts to develop methods that serve as readily accessible proxies to identify well-expressed, well-behaved protein samples that should have a higher probability of leading to crystallizable material without the need for an initial complete characterization of each candidate. Such approaches attempt to leverage the power of molecular biology to provide a rapid way to screen large numbers of constructs in parallel and identify the most promising for further, more intensive characterization. A leading tactic in this regard has been the use of fusion proteins bearing Green Fluorescent Protein (GFP).3-9 Because of its spectral characteristics, GFP fluorescence provides a unique and sensitive means for detecting tagged proteins within a cellular environment. Maturation of the GFP chromophore and development of the unique GFP spectral signal are directly coupled to proper GFP folding.10, 11 Because of this property, there should be a correlation between the fate of the target protein and the fate of the GFP chromophore.3 Thus, the level of fluorescence signal from a fused GFP molecule is expected to provide a reasonable measure of the ability of the candidate protein to be expressed at high levels in a properly folded form. Here, we report the implementation of a E. coli-based whole cell, GFP fluorescence assay to survey the expression properties of a large cohort of membrane proteins (314 candidates) from extremophilic organisms. We show that measurement of the fluorescence produced by a C-terminal GFP fusion of the candidate protein relative to a set of benchmark GFP-fusion proteins provides a reliable way to identify membrane proteins that can be expressed in the 0.5–15 mg L−1 range in the membrane fraction following expression optimization. The production of such quantities facilitates further characterization by detergent extraction screens and size exclusion chromatography. Importantly, our results demonstrate that the use of the whole-cell fluorescence detection method provides a rapid way to assess the overexpression potential of many membrane protein with a limited investment of labor. This stands in sharp contrast to traditional, labor-intensive methods for expression screening using SDS-PAGE and Western blot. Analysis of candidate protein expression behavior with respect to protein properties suggests that although there appear to be no particular limits on protein size or number of transmembrane segments, proteins in the 20–50 kDa range and having 7 transmembrane segments have a higher success rate. Interestingly, we find no correlation between expression level and the degree to which the codon usage of the target gene matches the expression host. Fluorescence-based screening for identifying membrane proteins having overexpression potential Following previous reports using GFP as a reporter for membrane protein expression,4, 7-9, 12 we established an E. coli GFP-based expression system (Supporting Fig. 1) to evaluate the expression potential of a large cohort of membrane proteins from organisms found in extreme and unusual environments. To allow rapid clone generation, we constructed a ligase independent cloning vector13 to facilitate direct cloning of PCR products of the gene of interest (Supporting Fig. 1). Target protein fusions contained a C-terminal tag that bore in sequence, a Tobacco Etch Virus (TEV) protease site,14 eGFP,15 and an octahistidine (His8) tag [Fig. 1(a)]. On the basis of previous studies,4, 7-9, 12 we anticipated that there was likely to be a correlation between proper folding and membrane insertion of the target protein and GFP fluorescence. One limit to this assay is that the C-terminal GFP moiety will only mature if the C-terminus of the target protein is intracellular.7 As a consequence, any well-expressed membrane proteins having a topology that places the C-terminus extracellularly will be missed. An estimated 70% of integral membrane proteins have their C-terminus in the cytoplasm.16 Given the expected high proportion of candidates with cytosolic C-termini, and the possibility that a single error in transmembrane protein prediction would eliminate potential candidates by inadvertently switching the predicted topology, we did not use topology prediction algorithms to prescreen candidates. GFP-based integral membrane protein expression screen. (a) Fluorescence measurements of 314 GFP fusions assayed in E. coli. Dashed line indicates the cutoff threshold level based on the values obtained three benchmark proteins, KvAP, AQPz, and YhjX. Inset shows schematic of GFP constructs. (b) SDS-PAGE analysis of the C41 E. coli membrane fractions from several extremophile membrane protein candidates following expression optimization analyzed by coomassie-blue staining (left) and anti-polyhistidine tag western blot (right). Lanes are: M,
Marker; 1, Af27 (22 kDa); 2, Rm29 (30 kDa); 3, Rm8 (30 kDa); 4, Ph11 (29 kDa); 5, Sp2 (49 kDa); 6, Rm4 (25 kDa). (c) Tables showing the numbers of proteins having optimal expression as H3C, CH, and HM3C affinity tagged constructs and showing the numbers of proteins having the optimal expression at different expression temperatures. (d) Percentage of optimized candidates with expression 0.5 mg L−1 of culture, grouped according to their fluorescence values from the GFP screen. (e) Comparison of GFP fluorescence versus final amount of protein following optimization. Expression levels were determined by following single-step metal affinity purification in FC-12 and measurement of the amount of protein purified by absorbance (circles) or by comparison with a protein standard by western blot (triangles). We chose candidate membrane proteins from organisms that live in extreme pH, high salt, the presence of high concentrations of heavy metals, high temperatures, or some combination of these extremes, based on the rationale that proteins found in such organisms, and in particular those in the membrane, which should have direct contact with the harsh environment, are likely to be very stable and robust subjects for biochemical and crystallographic studies.17 The criteria for selection of candidate organisms were that in addition to living in an unusual environment the selected organisms had to have a completely sequenced genome and have genomic DNA that could be obtained from the ATCC collection to be used as template for PCR. Membrane proteins were selected for the expression survey using a combination of annotation searches for ion channel, transporter, exchanger, antiporter, carrier, and major facilitator superfamily members, and by searching the genome of each candidate organism with a generic transmembrane helix profile. From an initial list of ∼400 selected candidates, we succeeded in making 314 GFP fusion constructs and tested their expression levels by measuring in cell GFP fluorescence following induction and expression at 25°C (Supporting Table I). Comparison of expression using a simple fluorescence assay that measured the GFP fluorescence signal from a standardized density of cells of the 314 GFP fusions yielded a wide distribution of fluorescence values, 20,000–180,000 F.A.U. (Fluorescence Arbitrary Units) [Fig. 1(a), Supporting Table I]. Uninduced cells grown to similar density yielded a background value of ∼20,000 F.A.U. To set a criterion for the level of expression that would indicate the most promising candidates, we examined the performance of three benchmark membrane protein-GFP fusions. Two of the benchmark proteins were chosen because they have had X-ray crystallographic structures determined from E. coli expressed material, AQPz18, 19 and KvAP.20, 21 The third benchmark protein, YhjX, was the best expressed membrane protein in a GFP-based expression survey of E coli membrane proteins.6 Our rationale was that expression of any test protein-GFP fusion at levels equivalent to or better than these benchmarks should indicate the production of acceptable levels of properly folded material for further characterization. On the basis of the benchmark protein-GFP fusion values, AQPz (73,000 F.A.U.), KvAP (118,000 F.A.U.), and YhjX (97,000 F.A.U.), we defined a positive threshold of 60,000 F.A.U. This value was chosen to allow a generous lower bound for candidates that should be as good or better than the benchmark proteins. Sixty-four of the screened candidates (20%) yielded fluorescence values at or above the threshold (60,000 A.U.). Interestingly, we obtained vast differences in success among the extremophilic organisms (Table I). C. tepidium, R. metallidurans, and S. pomeroyi were good sources of well-expressed candidates, having success rates as high as 40–50%, whereas O. oeni, S. solfataricus, and T. volcanium proved to be very poor sources having very low success rates, 0–6%. Further analysis revealed that certain membrane protein classes express better than others (Table II). For example, candidates from the mechanosensitive ion channel and from the flagellar motor MotA proton channel families had high success rates (53.5% and 71.4%, respectively), whereas candidates from the antiporter, exchanger, and transporter families were less likely to express well (12.0%, 16.6%, and 15.2%, respectively). The origins of the differences are unclear but may relate to a combination of protein molecular weight, total number of transmembrane segments, toxic consequences of protein activity from the overexpressed protein, and the potential requirement for auxiliary subunits that were not included in the expression study. Table I. List of Extremophilic Organisms Selected, Number of Candidates Screened, Number Positive, and Hit Rates in Each Organism Initial attempts at direct purification of a number of GFP constructs having high F.A.U. scores (Af11, Bh8, Ct6, and Dr35) proved problematic and did not directly yield sufficient quantities of purified material for further characterization. In consideration of the possibility that the GFP itself was proving a burden to production of the candidate protein, we cloned 52 candidates from three different pools; 32 high-expressing proteins ( 60,000 F.A.U.), 12 medium to low expressing proteins (60,000–20,000 F.A.U.), and 8 low expressing proteins ( 20,000 F.A.U.) into three affinity tag formats for protein expression optimization experiments. These 156 constructs were subjected to overexpression optimization trials in which each construct was grown as a small scale culture (7 mL) and tested for expression at three different induction temperatures (18°C, 25°C, and 37°C). Expression levels for each sample were assayed evaluating SDS-PAGE by coomassie blue staining and western blot using an anti-polyhistidine antibody. The optimization experiments with non-GFP constructs caused clear improvements compared to the temperature-optimized GFP fusion construct of the same target (Supporting Fig. 2). Figure 1(b) shows examples of membrane fractions for six of the candidate proteins following expression optimization. Thirty-two candidates of the 52 showed expression by both coomassie-staining and Western blot analysis. Both the affinity tag and expression temperature were important variables for expression optimization [Fig. 1(c)]. No particular tag predominated as the best choice. In general, a low induction temperature, 18°C, proved most beneficial. Comparison of the amount of protein produced under optimized expression conditions by using single step using metal affinity purification in Fos-choline-12 (FC-12), which was found as the most robust detergent for membrane protein extraction (see below), followed by quantification of the amount of protein expressed in the membrane fraction showed clear differences between the ability to convert proteins from different GFP score categories into expressed material [Fig. 1(d,e)]. Strikingly, we were able to produce 90% of the candidates from the high-expressing category at levels of ≥0.5 mg L−1 [Fig. 1(d)] with 50% of these made at ≥5 mg L−1. The success rate for producing ≥0.5 mg L−1 of protein from the medium to low expressing category was also good, 25%. None of the candidates from the category having GFP values near the background fluorescence value gave expression levels 0.5 mg L−1 following optimization. Comparison of the amounts of expressed proteins following optimization with the GFP screen fluorescence values shows that the fluorescence value served as a reliable indicator of candidate proteins that could be expressed in the membrane fraction at a level that is useful for further biochemical characterization [Fig. 1(e)]. On the basis of these success rates, we estimate that 90% of the candidates that passed the GFP screen could be produced at ≥0.5 mg L−1 culture if subjected to optimization experiments. These results highlight the value of including benchmark standards in execution of GFP-based protein expression surveys and indicate that a simple GFP screen should be sufficient for initial identification of proteins with a high-likelihood of overexpression potential. Detergents are a necessary component of any membrane protein purification effort and constitute a complex variable that requires extensive investigation.22, 23 Detergent choice has significant implications for protein oligomeric state, stability, homogeneity, and crystallization. Unfortunately, it is impossible to predict which detergent will be suitable for extraction, purification, and crystallization of a given membrane protein. Having sufficient expressed material for a variety of candidate proteins from different membrane protein classes following our expression optimization experiments, we screened candidates for detergent solubilization and size exclusion profile behavior using a panel of several different detergents. These experiments were aimed at assessing whether the overexpressed material could yield protein that would be of suitable quality for further biophysical characterization and crystallization trails. We chose 24 well-expressed membrane proteins and used a panel of six commonly used detergents decyl-β-D-maltoside (DM), octyl-β-D-glucopyranoside (OG), dodecyl-β-D-maltoside (DDM), Fos-choline-12 (FC-12), n-dodecyl-N,N-dimethylamine-N-oxide (LDAO), and 3-[(3-cholamidopropyl) dimethylammonio]-2-hydroxy-1-propanesulfonate (CHAPSO). Following extraction from the membrane fraction and a single metal affinity purification step, 21 of the 24 candidates gave an included peak in the gel filtration profiles in at least one detergent (Fig. 2, Table III). Ten of the 24 were found to be monodisperse based on the presence of a single, symmetrical peak (Fig. 2, Supporting Table II). The presence of a monodisperse size exclusion profile has been shown to be a good indicator of membrane protein samples that are of suitable quality for further biochemical, biophysical, and crystallization studies.8, 9, 24-27 It is difficult to estimate the oligomeric state of a membrane protein from size exclusion results, particularly as the standards are soluble, globular proteins, and the exact biological assembly of the particular proteins we examined is unknown. Nevertheless, we found that all of the monodisperse samples ran with apparent molecular weights that are larger than that expected for monomers. This property was maintained regardless of the detergent type in the cases where more than one detergent yielded a monodisperse peak (Fig. 2, Supporting Table II). Superdex200 gel filtration profiles and SDS-PAGE analysis of exemplar candidates following batch Ni-NTA purification. The detergent used is indicated in parentheses. Elution volume and coomassie-blue stained SDS-PAGE of the main gel filtration peak fraction are shown. Arrow indicates the void volume. Sample identity, annotation, monomer molecular weight, and approximate size of the observed protein detergent complex size are: (a) Sp2, MscS channel, 49 kDa, ∼120 kDa; (b) Dr10, MscS channel, 40 kDa, ∼95 kDa; (c) Sp8, MscL channel, 15 kDa, ∼180 kDa; (d) St34, ABC transporter, 65 kDa, ∼320 kDa; (e) Ct6, proton channel, 27 kDa, ∼110 kDa (also see Supporting Table II). Table III. Detergent Extraction and Gel Filtration Behavior of Overexpressed Extremophile Membrane Proteins List of purified candidates showing several parameters including type of construct, induction temperature, yield/L, and gel-filtration behavior in selected detergents. ++, mondisperse peak; +, good gel-filtration profile; −, poor gel-filtration profile; n.p., no gel-filtration profile; n.a, not tested. Interestingly, we found that FC-12 was the most successful detergent in terms of both solubilization and behavior in gel filtration. DDM was the second most effective followed by LDAO, OG, and DM. CHAPSO proved to be least effective at both solubilization and in yielding protein samples having good size exclusion chromatography behavior. The rank order of effectiveness in terms of the ability of specific detergents to produce samples having monodisperse profiles is FC12, DDM, LDAO and is similar to that reported by Drew and colleagues for a recent survey of eukaryotic membrane proteins.9 Fos-choline detergents have been used successfully to purify, crystallize, and determine membrane protein structures,28, 29 for membrane protein NMR studies,30 and for electron microscopy studies of transporter oligomers.31 FC12 has further been shown to be efficient at membrane protein solubilization and at yielding monodisperse samples as judged by size exclusion chromatography.7 Nevertheless, concerns have been raised that even though Fos-choline detergents are efficient at extracting proteins from membrane, they may be harsh enough to need to be treated with caution27 and may not always yield a stable sample.32 It is notable that 11 of the 19 proteins that gave good gel filtration profiles in FC12 also gave similar gel filtration profiles in at least one other more mild detergent. This result suggests that, at least in these cases, the samples were not likely to be simply solubilized and denatured by FC12. Our detergent extraction and gel filtration experiments identify many candidates that were able to produce protein that gave good gel filtration behavior in at least one detergent following a very simple purification protocol (21/24 candidates) (see Fig. 4). This observation provides further support for the idea that the original GFP screen was a good proxy for identifying membrane proteins that have the potential to yield material of suitable quality for biophysical experiments. Although there are now a number of reports in which GFP-based screens have been used to identify proteins or constructs that have appropriate biophysical characteristics for structural biology experiments,3, 5-9 there have remained some questions regarding the fidelity of the assay. Our results establish that there is an excellent correspondence between the behavior of membrane protein GFP-fusions and the potential of a given candidate to be overproduced. Of the candidates that pass the screen and were subjected to further characterization, 90% can be produced at ≥0.5 mg L−1 of culture, with half of these made at ≥5 mg L−1. These expression scales are suitable for investing further effort into purification, biophysical characterization, and crystallization trials. An important feature of this work is that the screening can be done manually and requires no specialized robotic equipment. Thus, such screens should be applicable in a wide range of settings. Membrane protein expression: influence of molecular weight, transmembrane segment number, and codon usage Our survey of the expression behavior of a large number of membrane proteins from diverse sources provides a dataset that yields insight into the factors that influence the ability to overproduce a candidate membrane protein [Fig. 3(a,b)]. There are some clear general lessons regarding the likelihood of success for a candidate based on a few simple properties. In general, the total molecular weight of the candidate appears to be an important factor. We found many well expressed membrane proteins in the 20–50 kDa range, (59/262, 22.5%), but very few in the 50 kDa range (5/52, 9.6%). Similarly, there is a clear difference in the success rates with respect to the number of predicted transmembrane segments; ∼35% of the 123 candidate proteins having 1–7 transmembrane express better than the cutoff whereas only 11% of the 191 proteins having 8–17 transmembrane segments expressed well. It is particularly striking that 67% of the proteins having three predicted transmembrane segments expressed as well as or better than the benchmark proteins. Distribution of MW, TMs, and codon usage among the GFP-screened candidates with respect to their fluorescence scores. (a) Molecular weight, (b) Predicted number of transmembrane helices. Fraction expressed per class: 1 TM (25%), 2 TM (42%), 3 TM (67%), 4 TM (20%), 5 TM (21%), 6 TM (31%), 7TM (27%), 8 TM (0%), 9 TM (9%), 10 TM (11%), 11 TM (10%), 12 TM (18%), 13 TM (17%), 14 TM (0%), 15 TM (0%), 17 TM (0%). (c) CAI for each gene, calculated using RSCU values from the class II highly-expressed E. coli gene reference set.33 Higher CAI values correspond to better matches with the host codon usage. There is a general notion that one of the parameters that may have an impact on the ability to overexpress a candidate protein in E. coli is how suited the codon usage of the candidate gene is to well-expressed proteins from host organism.34, 35 However, data supporting this belief are equivocal. There are anecdotal reports of successes,36, 37 failures,38 and variable outcomes39 for particular cases of codon optimization. Large scale study of a set of 1000 soluble proteins from Plasmodium falciparum, an organism with proteins that are known to be particularly challenging to express in E. coli, found no significant correlation with codon usage.40 As we had a diverse set of proteins from many organisms that have a range codon usages, we asked whether there was any correlation between how well the candidate proteins expressed and the similarity of their codon usage to that of well-expressed E. coli genes as measured by the \"codon adaptation index,” CAI. This parameter gives a length independent assessment of how well each test gene matches the codon usage of well-expressed proteins from the host.33 Strikingly, for those candidates that pass our threshold GFP value, we find no correlation between the CAI for a given gene and expression level [Fig. 3(c)]. Indeed, a number of the proteins having the highest GFP values can be produced at very high levels following optimization in spite of having very low CAI scores (Ph11, CAI = 0.123, 10 mg L−1; So15, CAI = 0.153, 10 mg L−1; Bh8 CAI = 0.154, 12 mg L−1). Even though individual rare codons may certainly cause expression difficulties,34, 35 our data support the conclusion that codon optimization is not a key parameter for heterologous membrane protein expression in E. coli. Given the challenges associated with membrane protein biochemistry, there have been a number of reports describing the implementation of GFP-based screens as a means to identify potentially tractable candidate proteins. Although large scale GFP screens in E. coli,6 S. cereviseae9 have reported success in identification of well-expressed candidates, multi-milligram expression levels of non-host derived proteins per liter of culture have not yet been generally demonstrated. A key advance reported here is the use of benchmark proteins to set a threshold level for well-expressed candidates paired with construct and expression temperature optimization protocol. Together, this approach identifies a large number of candidate proteins that can be made at ≥0.5 mg L−1 with some candidates expressing better than 10 mg L−1. Similar to the recently reported S. cerevisiae survey,9 we find that Fos-choline 12 (FC-12) has the best success at both solubilizing membrane expressed material and producing material having a good gel filtration profile (Table III). The high success with fos-choline detergents in membrane protein solubilization has raised some of the concern regarding the potential destabilizing effects of this detergent class.27, 32 Importantly, a large number of the expressed extremophile membrane proteins that can be extracted from the membrane using FC-12 are also extracted from the membrane and give good gel filtration profiles in other, more conventionally used detergents. Annotation searches for membrane proteins denoted as ion channels, transporters, exchangers, antiporters, carriers and major facilitator superfamily in the genomes of Aeropyrum pernix (Ap), Archaeoglobus fulgidus (Af), Bacillus halodurans (Bh), Chlorobium tepidium (Ct), Deinococcus radiodurans (Dr), Desulfovibrio vulgaris (Dv), Geobacter sulfurreducens (Gs), Oenococcus oeni (Oo), Pyrococcus furiosus (Pf), Pyrococcus horikoshii (Ph), Ralstonia metallidurans (Rm), Shewanella oeneidensis (So), Silicibacter pomeroyi (Sp), Streptococcus thermophilus (St), Sulfolobus solfataricus (Ss), Thermotoga maritima (Tm), Thermus thermophilus (Tt), and Thermoplasma volcanium (Tv) were performed using the Integr8 protein database (http://www.ebi.ac.uk/integr8). To find unannotated the membrane proteins, each genome was searched with a sequence representing a generic transmembrane helix using BLASTP41 at the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov). This search yielded a list of membrane proteins in the chosen organism from which we selected several proteins with no homologues of known structure. All transmembrane helix predictions were performed with TMHMM 2.016 (http://www.cbs.dtu.dk/services/TMHMM-2.0). A complete list of screened proteins and values from the GFP screen can be found in Supporting Table I. All genes were cloned using ligation-independent cloning (LIC)13 of PCR products containing LIC overhangs and directly cloned into vector containing a complementary LIC cassette (Supporting Fig. 1). For the GFP screening, genes encoding candidate membrane proteins were cloned into a pET-based LIC vector that harbors a Tobacco etch virus protease cleavage site followed by a C-terminally His8 tagged GFP. The three vectors used in overexpression optimization experiments were H3C (N-terminal His8 tag-3C protease cleavage site42 followed by the gene of interest), HM3C (N-terminal His6 tag-maltose binding protein-3C protease cleavage site followed by the gene of interest), and CH (gene of interest followed by a TEV protease cleavage site and a His8 tag). The GFP and CH vectors share the same LIC cassette, whereas, the H3C and HM3C vectors share a different LIC cassette (Supporting Fig. 1). PCR reactions were done using Phusion high-fidelity DNA polymerase (Finnzymes) and genomic DNA for each parent organism. Genomic DNA was obtained from ATCC (www.atcc.org). Two different cell types of E. coli were used. BL21 (DE3) pLysS were used for screening of GFP-fusion candidates and C41 cells43 were used for expression optimization and protein production. BL21 (DE3) pLysS cells were used for the GFP screen owing to the observation that they are reported to be less sensitive to variations in OD600 nm of induction than the C41 strain.7 Individual E. coli BL21 (DE3) pLysS cultures for each of the candidate GFP-fusions were grown in 7 mL Luria-Bertani (LB) media at 37°C to OD600 nm ∼0.5. Induction was done at 25°C using 0.4 mM IPTG. After 20 h of growth post-induction, the cell pellet was recovered from 1 mL of culture (2 min at 13,000 rpm in a bench top 5415D centrifuge (Eppendorf)) and resuspended in 200 μL of 200 mM KCl, 20 mM Tris pH 8.0. Samples were subjected to the whole-cell fluorescence assay (465–485 nm excitation, 515–575 nm emission) using a 20/20n luminometer with a Blue fluorescent module (Turner Biosystems, U.S.A.). In each experiment uninduced cells and cells expressing the GFP fusion of BH8, a well-expressed candidate protein, were run as controls. These controls have reproducible fluorescence measurements over a number of trials, F.A.U. of 23,667 ± 1021, n = 6, and 174,964 ± 5397, n = 6, for uninduced and BH8 expressing cells, respectively. Proteins for overexpression optimization were selected based on the fluorescence values from the GFP screen. Genes for the selected proteins were cloned into H3C, CH, and HM3C tag formats and transformed into C41 cells.43 Cultures were grown in 7 mL LB media at 37°C. Induction was performed at OD600 nm ∼0.5 using 0.4 mM IPTG. Immediately following induction, cultures were moved to one of three different expression temperatures, 18°, 25°, or 37°C, and were grown for 48, 20, and 3 h, respectively. Following expression, 500 μL of cells were harvested by centrifugation (2 min at 13,000 rpm in a bench top 5415D centrifuge (Eppendorf), resuspended in 50 μL of 200 mM KCl, 20 mM Tris, pH 8.0, and lysed by the addition of 50 μL of 2% SDS, 20% glycerol, 0.05% bromophenol blue, 125 mM Tris pH 6.8. Expression was assayed by SDS-PAGE analysis with Coomassie staining and Western blot using an anti-His5 HRP conjugate antibody (Qiagen). The best-optimized condition from this assay was used for a scale-up. C41 E. coli cells harboring the expression construct of interest were grown in 1 L 2xYT (5 g NaCl, 16 g tryptone, 10 g yeast extract) under conditions that were determined to give the best expression for each individual construct under investigation. Cell pellets obtained from the 1 L culture were suspended in 90 mL lysis buffer (200 mM KCl, 20 mM Tris pH 8.0, 8% glycerol, 2 mM EDTA, 1 mM PMSF, 1 mM DTT) and disrupted by sonication for 4 min on ice using a Sonic Dismembrator model 500 (Fisher). Cell lysate was centrifuged in a JA 25.50 rotor (Beckman Coulter) at 7000g for 35 min at 4°C to remove unbroken cells and debris. The supernatant was then centrifuged in a Ti45 rotor (Beckman Coulter) at 40,000 rpm for 2 h at 4°C to separate the membrane pellet. Pellets were homogenized in a storage buffer (200 mM KCl, 20 mM Tris pH 8.0, 8% glycerol) with a Dounce Tissue Grinder (Kimble Kontes LLC) and frozen at −80°C. All extraction, purification, and characterization were done at 4°C. Membrane extractions and purifications were performed with six detergents (Anatrace, USA): DM (40 mM), OG (200 mM), LDAO (2% v/v), DDM (20 mM), CHAPSO (160 mM), and FC12 (40 mM), added and incubated for 2 h at 4°C. Detergent-solubilized membrane proteins were separated from insoluble material by ultracentrifugation for 45 min at 55,000 rpm at 4°C in a TLA 110 rotor on a bench-top ultracentrifuge (Beckman Coulter). Following recovery, the supernatant was incubated with a bed volume of 0.5 mL of pre-equilibrated Ni-NTA superflow beads (Qiagen) at 4°C. Protein-bound beads were washed with the appropriate detergent buffer consisting of 200 mM KCl, 30 mM imidazole, 8% glycerol (v/v), detergent (3× critical micelle concentration (CMC), except for LDAO which was used at 10× CMC), 20 mM Tris, pH 8.0 and eluted with the same detergent buffer containing 300 mM imidazole. The eluted protein samples were 80–90% pure based on SDS-PAGE analysis. Protein concentration was determined by absorbance at 280 nm.44 Purified samples were subjected to size exclusion chromatography on a Superdex200 column (GE Healthcare) to assess the homogeneity and stability in a running buffer of 200 mM KCl, 8% glycerol (v/v), detergent (3× CMC, except for LDAO which was used at 10× CMC), 20 mM Tris, pH 8.0. Codon adaptation index (CAI) values for each candidate gene were calculated according to Sharp and Li33 using the webserver (http://www.evolvingcode.net/codon/cai/cais.php).45 CAI = CAIobs/CAImax where the observed CAI for a given gene sequenceand the maximal CAI score is . RSCUk is the \"relative synonymous codon usage” value for the kth codon in the gene, RSCUkmax is the maximum RSCU value for the amino acid encoded by the kth codon in the gene, and L is the number of codons in the gene.where xij is the number of occurrences of the jth codon for the ith amino acid and ni is the number (from 1–6) of alternative codons for the ith amino acid. RSCUkmax values were calculated from the class II highly-expressed E. coli gene reference set.33 Producing membrane proteins and membrane protein complexes for biochemical and structural study remains a major challenge. The development of methods that provide a rapid way to identify membrane proteins that can be expressed in sufficient amounts and of sufficient quality is an area of intensive effort.4, 6-9, 46 This report provides strong evidence that screening GFP-fusions provides a fruitful approach for identifying well-expressed candidate membrane proteins that display favorable biophysical behaviors suitable for further functional and structural investigation, and that a key factor in success is the investment in optimization of construct tags and induction temperature. Structural and biochemical studies of integral membrane proteins remains challenging because identifying a candidate protein that has good expression, that can be readily purified, and that has suitable behavior in detergent solution can require a substantial investment in time and resources. By combining a GFP-screen with a simple optimization protocol, we have been able to achieve a high success rate identifying candidate proteins that can be well expressed and that display desirable biophysical characteristics. A key implementation was the use of benchmark proteins to set a level of GFP fluorescence that will identify a candidate with a high likelihood of success. We anticipate that the large number of diverse membrane proteins we identify here as having good expression profiles will provide excellent starting points for both functional and structural studies. A summary of the outcome of the screen is shown in Figure 4. As with other expression surveys of panels of membrane proteins in which multiple levels of behavior are evaluated,27 the approach we report here functions as a funnel in which the there is an attrition of candidates as the stringency imposed at each step increases. It may be possible to salvage some failed candidates by additional exploration of alternate constructs that are designed to eliminate regions of predicted polypeptide disorder. Further, just as combination of mutagenesis and GFP screens has proved useful for defining appropriate constructs of soluble proteins for crystallization and structure determination,47, 48 it is likely similar combination mutagenesis/GFP screens should prove fruitful for enhancing the production of crystal grade membrane protein samples or for rescuing failed candidates. Similarly, GFP-based studies using other expression hosts and as a means to identify well-expressed complexes through co-expression strategies49-51 should enable the structural investigation of an array of membrane proteins and membrane protein complexes. Summary of extremophile membrane protein GFP screen. Results are presented as a funnel. At each stage some candidates are eliminated due to a failure to pass progressively more stringent tests. The exception is the step from GFP-positive to expression optimization in which 32 GFP-positive proteins along with 20 representatives from categories having lower GFP signals were chosen for expression optimization and further testing. Candidates passing the final stage of being monodisperse and stable are suitable for immediate evaluation by crystallization trials, small-angle X-ray scattering (SAXS), or other biophysical and functional analysis. We thank K. Brejc, J.M. Berger, and M. Pioletti for comments on the manuscript and members of the Minor lab for input at all stages of this work. DLM is an AHA Established Investigator.Additional Supporting Information may be found in the online version of this article. PRO_18_sm_SuppFiguresandTables.pdf272.1 KB Supporting Information Figure 1. Diagram of the LIC vectors used in this study. In each case the restriction site used to prepare the LIC site in the vector is shown in light blue. Supporting Information Figure 2. Examples of construct and temperature optimization for two candidates, Sp1 and Sp2. GFP indicates the GFP-fusion construct for each protein. H3C, HM3C, and CH indicate N-terminal His8, N-terminal His6-maltose binding protein, and C-terminal His8 versions, respectively. Western blots are probed with an anti-polyhistidine antibody. Expression temperatures are indicated. Supporting Information Table 1. Candidate membrane proteins. Supporting Information Table 2. Size exclusion apparent molecular weights for monodisperse samples. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.(2001) Green fluorescent protein as an indicator to monitor membrane protein overexpression in Escherichia coli.FEBS Lett 507: 220– 224. Wiley Online Library(2005) Recent advances in GFP folding reporter and split-GFP solubility reporter technologies. Application to improving the folding and solubility of recalcitrant proteins from Mycobacterium tuberculosis.J Struct Funct Genomics 113– 119. Drew D,Slotboom DJ,Friso G,Reda T,Genevaux P,Rapp M,Meindl-Beinker NM,Lambert W,Lerch M,Daley DO,Van Wijk KJ,Hirst J,Kunji E,De Gier JW(2005) A scalable, GFP-based pipeline for membrane protein overexpression screening and purification.Protein Sci 2011– 2017. Wiley Online Library(2006) Optimization of membrane protein overexpression and purification using GFP fusions.Nat Methods 303– 313.(2006) Fluorescence-detection size-exclusion chromatography for precrystallization screening of integral membrane proteins.Structure 673– 681.(2007) High-throughput fluorescent-based optimization of eukaryotic membrane protein overexpression and purification in Saccharomyces cerevisiae.Proc Natl Acad Sci USA 104: 13936– 13941.(2002) Rapid topology mapping of Escherichia coli inner-membrane proteins by prediction and PhoA/GFP fusion analysis.Proc Natl Acad Sci USA 2690– 2695. Controlled intracellular processing of fusion proteins by TEV protease.Protein Expr Purif 312– 318.(1996) An enhanced green fluorescent protein allows sensitive detection of gene transfer in mammalian cells.Biochem Biophys Res Commun 227: 707– 711.(2001) Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes.J Mol Biol 305: 567– 580.(2003) Architecture and selectivity in aquaporins: 2.5 a X-ray structure of aquaporin Z.PLoS Biol E72.(2006) Crystal structure of AqpZ tetramer reveals two distinct Arg-189 conformations associated with water permeation through the narrowest constriction of the water-conducting channel.J Biol Chem 281: 454– 460.(2005) Structure of the KvAP voltage-dependent K+ channel and its dependence on the lipid membrane.Proc Natl Acad Sci USA 102: 15441– 15446.(1982) Three-dimensional crystals of a membrane protein complex. The photosynthetic reaction centre from Rhodopseudomonas viridis.J Mol Biol 158: 567– 572.(2002) The structure of Escherichia coli BtuF and binding to its cognate ATP binding cassette transporter.Proc Natl Acad Sci USA 16642– 16647.(2004) Crystallization and preliminary crystallographic analysis of the Escherichia coli water channel AqpZ.Acta Crystallogr D Biol Crystallogr 561– 563. Wiley Online Library The funnel approach to the precrystallization production of membrane proteins.J Mol Biol 377: 62– 73.(2002) Crystal structure of Escherichia coli MscS, a voltage-modulated and mechanosensitive channel.Science 298: 1582– 1587. Wang W,Black SS,Edwards MD,Miller S,Morrison EL,Bartlett W,Dong C,Naismith JH,Booth IR(2008) The structure of an open form of an E. coli mechanosensitive channel at 3.45 Å resolution.Science 321: 1179– 1183.(2007) Conformational dynamics of the KcsA potassium channel governs gating properties.Nat Struct Mol Biol 1089– 1095.(2006) Purification and 3D structural analysis of oligomeric human multidrug transporter ABCG2.Structure 1623– 1632.(1987) The codon Adaptation Index–a measure of directional synonymous codon usage bias, and its potential applications.Nucleic Acids Res 1281– 1295.(2005) Strategies for efficient production of heterologous proteins in Escherichia coli.Appl Microbiol Biotechnol 289– 298.(1992) Human hemoglobin expression in Escherichia coli: importance of optimal codon usage.Biochemistry 8619– 8628. Optimized gene synthesis and high expression of human interleukin-18.Protein Expr Purif 110– 118.(2008) Production, characterization and crystallization of the Plasmodium falciparum aquaporin.Protein Expr Purif 69– 78.(2008) Codon optimization can improve expression of human genes in Escherichia coli: a multi-gene study.Protein Expr Purif 94– 102. Mehlin C,Boni E,Buckner FS,Engel L,Feist T,Gelb MH,Haji L,Kim D,Liu C,Mueller N,Myler PJ,Reddy JT,Sampson JN,Subramanian E,Van Voorhis WC,Worthey E,Zucker F,Hol WG(2006) Heterologous expression of proteins from Plasmodium falciparum: results from 1000 genes.Mol Biochem Parasitol 148: 144– 160.(1990) Substrate requirements of human rhinovirus 3C protease for peptide cleavage in vitro.J Biol Chem 265: 9062– 9065.(1996) Over-production of proteins in Escherichia coli: mutant hosts that allow synthesis of some membrane proteins and globular proteins at high levels.J Mol Biol 260: 289– 298.(2005) Predicted highly expressed genes in the genomes of Streptomyces coelicolor and Streptomyces avermitilis and the implications for their metabolism.Microbiology 151: 2175– 2187.(2006) Crystal structure of the essential N-terminal domain of telomerase reverse transcriptase.Nat Struct Mol Biol 218– 225.(2005) Soluble domains of telomerase reverse transcriptase identified by high-throughput screening.Protein Sci 2051– 2058. Wiley Online Library Fitzgerald DJ,Schaffitzel C,Berger P,Wellinger R,Bieniossek C,Richmond TJ,Berger I(2007) Multiprotein expression strategy for structural biology of eukaryotic complexes.Structure 275– 279. MultiBac: multigene baculovirus-based eukaryotic protein complex production.Curr Protoc Protein Sci 5:Unit 5 The full text of this article hosted at iucr.org is unavailable due to technical difficulties. Please check your email for instructions on resetting your password. If you do not receive an email within 10 minutes, your email address may not be registered, and you may need to create a new Wiley Online Library account. Can\'t sign in? Forgot your username? Enter your email address below and we will send you your username If the address matches an existing account you will receive an email with instructions to retrieve your username
