|Home | About | Journals | Submit | Contact Us | Français|
It is often an immense challenge to overexpress human membrane proteins at levels sufficient for structural studies. The use of Human Embryonic Kidney 293 (HEK 293) cells to express full-length human membrane proteins is becoming increasingly common, since these cells provide a near-native protein folding and lipid environment. Nevertheless, the labour intensiveness and low yields of HEK 293 cells and other mammalian cell expression systems necessitate the screening for suitable expression as early as possible. Here we present our methodology used to generate constructs of human membrane proteins and to rapidly assess their suitability for overexpression using transiently transfected, glycosylation-deficient GnT I −/− HEK 293 cells (HEK 293S). Constructs, in the presence or absence of a C-terminal Enhanced Green Fluorescence Protein (EGFP) molecule, are made in a modular manner, allowing for the rapid generation of several combinations of fusion tags and gene paralogues/orthologues. Solubilization of HEK 293S cells, using a range of detergents, followed by Western blotting is performed to assess relative expression levels and to detect possible degradation products. Fluorescence-detection size exclusion chromatography (FSEC) is employed to assess expression levels and overall homogeneity of the membrane proteins, to rank different constructs for further downstream expression trials. Constructs identified as having high expression are instantly suitable for further downstream large scale transient expression trials and stable cell line generation. The method described is accessible to all laboratory scales and can be completed in approximately three weeks.
Human embryonic kidney cells have been successfully utilized to overexpress a wide variety of mammalian membrane proteins (1-3), including that of human Rh family, C Glycoprotein (RhCG) whose x-ray crystal structure was recently determined in our laboratory (4). With respect to the overexpression of human membrane proteins for structural studies, the advantages that HEK 293 cells, as well as other mammalian cells such as Chinese Hamster Ovary (CHO), possess over lower order expression systems such as Escherichia coli, yeast, and insect cells are clear; the human membrane protein of interest is presented with near native translocation machinery, post-translational modifications, and lipid milieu that can be of critical importance in the biosynthesis of functional human membrane proteins (5, 6). Despite these inherent advantages, it is often an immense challenge to overexpress human membrane proteins in HEK 293 cells. While examples of robust expression levels of human membrane proteins in HEK 293 cells have been described (3), it might be reasonably expected that these examples represent the exception and that modest yields are to be typically expected (4, 7, 8).
To this end, it can be of critical importance to screen a variety of human membrane protein expression constructs for both expression levels and homogeneity, and to assess whether or not these expression constructs are tractable for structural studies. Several variables have been shown to have a pronounced effect on expression level, homogeneity, and crystallizability of membrane proteins, including affinity tag type and location (9), codon usage (10), detergent type (11), truncations (12), paralogues or orthologues (13), ligands (14), and lipids (15, 16). Fortunately, many of these variables can be initially assessed in a facile manner in transiently transfected HEK 293 cells, and do not require the lengthy time scales associated with stable cell line generation. For instance, fluorescence-detection size exclusion chromatography (FSEC) has been shown to possess sufficient sensitivity to screen a number of human P2X receptor paralogues from small-scale, transiently transfected HEK 293 cells (13).
In this manuscript, we describe in detail the methods that we employ to both generate human membrane protein expression constructs and to rapidly assess their suitability for expression by using transiently transfected, glycosylation-deficient GnT I −/− HEK cells (HEK 293S) (17). Our criteria for suitability is that there would be a reasonable expectation that the human membrane protein of interest would be, subsequent to stable cell line generation and protein purification, pure and homogeneous with a final yield of greater than ~0.05 mg of protein per liter of cell culture medium. While these yields are modest, the continued miniaturization of the crystallization screening process allows us to typically screen human membrane proteins purified from medium-scale (~3-6 L) spinner flask HEK 293S cell cultures, circumventing the requirement for larger-scale (~10-20 L) HEK 293S cell cultures until bona fide crystal hits are generated. Our application for these methods are ultimately aimed towards the structure determination of human membrane proteins via x-ray crystallography, nevertheless, the methodologies described in this manuscript are of general interest, and are entirely accessible, to those that are interested in assessing human membrane proteins expressed in HEK 293S for expression and/or homogeneity.
The expression constructs described in this manuscript are all derived from the tetracycline inducible mammalian cell expression vector pACMV-tetO (18). Inducible expression vectors offer the advantage of delaying transgene expression until high density cell cultures are established, therefore offsetting the toxicity effects associated with the overexpression of certain human membrane proteins constitutively (19). In our experience, a wide variety of human membrane proteins can be subcloned into vectors derived from pACMV-tetO and transfected into HEK 293S cells with little adverse effect on cell viability.
Various paralogues/orthologues and N- and C-terminal truncations can be subcloned into pACMV-tetO, whereby a two-step PCR protocol introduces N- and/or C-terminal affinity tags to the expression construct via PCR primers (Figure 1). In our experience, we have typically utilized an N-terminal FLAG and/or a C-terminal His10 affinity tag for the purification of human or mammalian membrane proteins from HEK 293S cell cultures, nevertheless, the cloning protocol described below can in principle be applied to any affinity tag that is short enough to be introduced via a PCR primer (Table 1). This may be of benefit given the pronounced effect that both affinity tag type and location can have on membrane protein expression and homogeneity (9). In those cases which possess an N-terminal signal sequence, either cleavable or non-cleavable, we typically utilize a sole C-terminal His10 affinity tag.
We have introduced a C-terminal EGFP molecule into the pACMV-tetO vector, termed pACMV-tetO-EGFP (Figure 3), for use in FSEC based screening of membrane proteins. This construct is typically only utilized in those cases where the C-terminus of the membrane protein of interest is directed towards the cytoplasm, given that GFP is prone to misfolding and aggregation in the presence of oxidizing environments such as the extracellular space or ER lumen (20). A recent report that a “superfolder” GFP molecule is well behaved when targeted to oxidizing environments via the Sec translocon (21) may thus be of interest in the FSEC screening of membrane proteins in the future, and can be introduced into pACMV-tetO using the same methodology as described here.
The sequence for EGFP was PCR amplified using a forward N-terminal primer consisting of NNNNNN-NotI-spacer-thrombin-GFP(N), and a C-terminal primer consisting of GFP(C)-His8-XhoI-NNNNNN, where the spacer encodes for a 5 amino acid Gly/Ala repeat, N represents any nucleotide, and GFP(N or C) represent complementary sequence to the N- and C-termini of EGFP. Following PCR amplification and gel purification of the PCR product, the EGFP insert and pACMV-tetO were doubly digested with NotI and XhoI, and gel purified following standard protocols. Ligation was performed using 25 ng of pACMV-tetO, 75 ng of EGFP insert, 0.5 μL of T4 ligase (New England Biolabs, M0202S) and 1X ligase buffer in a total volume of 10 μL, at room temperature for 2 hours. Ligation products were transformed into electrocompetent XL-1 Blue cells following the manufacturer’s protocols.
Following sequence confirmation that pACMV-tetO-EGFP is generated, the transgene of interest is subcloned into pACMV-tetO in the same manner as described above for cloning into pACMV-tetO, with the exception that the C-terminal primer incorporates a NotI site and is tagless, given that a C-terminal EGFP and His8 are provided for by the vector.
All expression trials described in this manuscript were performed using GnTI−/− HEK 293S cells (kindly provided by Dr. H. Gobind Khorana). As mentioned above, the inducibility of these cells using tetracycline make it amenable to the overexpression of human membrane proteins, as high density cell cultures can be generated prior to protein expression, thereby alleviating toxicity effects. In addition, the lack of GnTI results in the production of shorter, homogeneous N-linked glycans (Man5-GlcNAc2) which can be more easily removed via glycosidases compared to wild-type N-linked glycans. Nevertheless, enzymatic removal of the Man5-GlcNAc2 moieties are not necessary for the crystallization of human membrane proteins, as suggested by the structure of human RhCG which was solved in the glycosylated state (4).
Transient transfection makes quantification difficult since the amount of DNA transfected can vary considerably and thus affect the final yield. To make an initial assessment of protein expression levels, a small scale detergent solubilization of the transiently transfected cells, using a panel of four detergents, followed by Western blotting is performed. The rationale behind screening a panel of detergents at this stage, as opposed to utilizing a single detergent, is that in those cases where solubilization is observed, the homogeneity of the membrane protein in that particular detergent can be immediately assessed via FSEC.
The magnitude of the solubilized Western blot signal is a critical factor in determining if a particular construct is suitable for stable cell line generation. In our experience, in those cases where no Western signal is observed upon film exposure for 5 minutes, or in those cases where significant degradation is observed, the construct should be redesigned to test for different orthologues/paralogues, affinity tags, or codon optimization, as it is highly unlikely that this construct will produce sufficient amounts of the target membrane protein for structural studies. In cases of low expression levels, the quality of the membrane protein as determined via FSEC (see section 2.4) will determine if the particular construct should be utilized for stable cell line generation. For instance, low expressing constructs that possess poor FSEC profiles (see Figure 5A - “full-length”, Figure 6A - “DDM”) are poor candidates for stable cell line generation, as the membrane protein expressed is unlikely to be suitable for structural studies. Given the modest yields associated with expressing human membrane proteins in HEK-293S cells, any construct that expresses at a very high level as determined via Western blotting is typically utilized for stable cell line generation.
Solubilization conditions that increase the ratio of the solubilized material (after spin, see Figure 4) relative to the total protein produced (before spin, see Figure 4) are often beneficial, as this may result in more useable membrane protein produced per liter of HEK-293S cell culture. Near 100% solubilization of human membrane proteins expressed in HEK-293S cells can be achieved (Figure 4--2,2, “FC-14”), and for the case of human RhCG, the x-ray crystal structure was determined using HEK-293S expressed material that was nearly 100% solubilized using both OG and DDM (4). Nevertheless, the degree of solubilization itself should not be used to determine the suitability of a particular detergent for solubilization, as certain detergents can more readily solubilize misfolded or aggregated membrane proteins (23). Therefore, the ratio of solubilization, in conjunction with FSEC (see sections 2.4 and 2.5) should be used to determine suitable solubilization conditions.
Fluorescence-detection size exclusion chromatography (FSEC) is traditionally used to assess the quality of the target gene (13). We find that FSEC can be used in a quantitative manner (Figure 5A) to assist in the determination of expression levels along with Western blots (see section 2.3). We normally prefer to test our constructs a number of times to assess reproducibility between the transient transfections. By stringent protocol adherence, we observe that reproducibility in terms of expression levels is surprisingly good when gauged by FSEC. For most cases the standard deviation of expression assessed by FSEC is 30% or lower (Figure 5B). We find that expression levels are best evaluated by normalizing the area of the included FSEC peak to the area of the absorbance at 280 nm (A280) chromatogram (Figure 5A, 5B). This normalization is performed to account for differences in cell count among the separate transient transfections.
In FSEC, as is the case for standard size exclusion chromatography, the suitability of the sample for overexpression trials and structural studies is assessed in two ways. Firstly, proteins that migrate in the void volume of the size exclusion column are likely to be higher order aggregates, indicative of misfolding and/or insolubility, whereas proteins that migrate in the included volume may be properly folded (see Figure 5A). Therefore, constructs, detergents, ligands, etc., that reduce the magnitude of the void volume peak, and increase the magnitude of the included peak are beneficial for structural studies. Secondly, the shape of the included peak is an excellent indicator of protein quality. A homogeneous protein, which is the ideal sample for crystallization trials, will possess a monodisperse, included peak (e.g. Figure 6C - “DDM”), whereas a protein that exist in multiple conformations and/or oligomeric states will possess a polydisperse, included volume peak (e.g. Figure 6B - “DDM”). Constructs, detergents, ligands, etc. that increase the monodispersity of the included volume peak are beneficial for structural studies.
Fluorescence-detection size exclusion chromatography can also be used to assess the homogeneity of GFP-fusion proteins (13). In the methodology described here, the effect that detergent type has on the homogeneity of the membrane protein of interest is assessed by performing FSEC on all solubilized material that scores positively via Western blotting (Figure 4). In principle, additional variables known to impact membrane protein homogeneity can also be assessed via FSEC, simply by performing the small scale whole cell membrane solubilization at different pH, glycerol, salt, or in the presence of ligands.
Perform the FSEC experiment as described in sections 2.4.1 to 2.4.3, using material solubilized in a panel of four detergents as discussed in section 2.3. The relative magnitude of the fluorescence peak correlates with the intensity of the Western blot band (see Figure 4), providing a second estimate for the amount of solubilizable membrane protein that is produced. In those cases where single, monodisperse peaks are not observed (Figure 6A,B), the magnitude of the included volume peak will dictate whether or not effort should be placed into trying to improve the profile, via pH, ligands, etc., or if additional constructs should be tested. As mentioned above, given the modest expression levels associated with expressing human membrane proteins in HEK 293S cells, those constructs that express at high levels, albeit with poor FSEC profiles (Figure 6B - “DDM”), should still be considered for further large-scale expression trials, whereas those constructs that both express poorly and possess poor FSEC profiles (Figure 6A) are unlikely to be suitable for structural studies.
The purpose of the methodology described in this manuscript is to generate and identify expression constructs that may be suitable for structural studies. At the transient transfection level, one can assess, in a facile manner, protein expression levels (via Western blotting and FSEC) and quality (via FSEC), both of which are key variables that determine if an expression construct is tractable for structural studies. Any expression construct generated using the methodologies described in this manuscript is immediately suitable for large scale protein expression trials, either via large-scale transient transfections or via stable cell line generation followed by large scale cell cultures.
It is important to consider expression trials in a variety of systems to assess which system is most suitable for overexpression. Other expression systems, namely insect cells, have been used with success to heterologously express human membrane proteins for structural studies (e.g. 24) Nevertheless, the near-native translocation machinery and lipid milieu encountered when expressing human membrane proteins in human cells can lead to less misfolding and degradation (Figure 7). Furthermore, the fact that we have successfully utilized HEK293S cells in the structure determination of human RhCG (4) suggests that human cell expression systems could be of increased prominence in the structure determination of human membrane proteins in the future.
The work described in this manuscript was supported by NIH/NIGMS grants P50 GM73210, U54 GM094625 and R37 GM24485. BPP is supported by the Danish Cancer Society.