Peptide-based probe for FSEC
Covalent fusion of GFP to target proteins is the conventional method for high-throughput precrystallization screening of recombinant proteins by FSEC.[4, 8] In this study, we sought to completely eliminate the need to generate GFP fusion proteins by capitalizing on the ubiquitous use of polyhistidine-tags to purify recombinant proteins by immobilized metal-affinity chromatography. A peptide-based fluorescent probe, designated as P3NTA was evaluated (Fig. 1). The chemical synthesis of P3NTA involves solid-phase peptide synthesis to introduce a fluorescein label. The three adjacent thiol groups in the reduced, purified peptide were modified by a reaction with maleimido-C3-NTA. A short ten residue affinity-tag was added to the probe to facilitate its separation on Streptactin resin. The synthetic reaction was monitored by mass spectrometry.
In contrast to GFP fusions, the central design principle for the P3NTA probe was to avoid any potential impact that a covalently attached GFP fusion partner could have on the function, cellular localization and processing of target proteins, overproduction (artificially decreased or increased yield), and then eliminate the need for proteolytic removal of the GFP fusion partner following purification. We believe that because the P3NTA probe is added to crude cell preparations following protein overproduction and that it is noncovalently linked to the target protein, these features minimize false-positive rates that have been previously observed with GFP fusion proteins.
The physical interaction between the P3NTA probe and the target protein is driven by the same principles that are routinely used to purify and capture polyhistidine-tagged proteins using NTA-based resins and matrices. Given that the P3NTA probe has three NTA groups (Fig. 1), it is possible that the probe-to-target binding stoichiometry could be greater than one. However, previous studies have thoroughly characterized the interaction between hexahistidine and decahistidine-tagged proteins with spectroscopic probes containing two-to-four NTA moieties with a single covalently linked fluorescein label by isothermal titration calorimetry and analytical size exclusion chromatography.[16, 17] These studies demonstrate that the binding affinity and stability of the probe increases with the accessibility and the number of available NTA moieties, and that the binding stoichiometry is approximately 1:1 (probe-to-target) when working at saturating probe concentrations. We observed stable fluorescence signals for replicate samples that were analyzed across 16 hour automated HPLC sequences following a four hour incubation at 4°C with excess P3NTA probe prior to FSEC. As with related spectroscopic probes, we conclude that saturatable and constant binding with the P3NTA probe has been established.
Validating GPCR expression
The utility of the P3NTA probe for FSEC was evaluated using the human adenosine 2a receptor as a model membrane protein, which has well-defined production and purification protocols.[18-21] In a series of experiments, we compared the use of the peptide-based P3NTA probe to detect the production of an A2a-eGFP fusion protein (residues 1-316 of the GPCR) (Fig. 2). Suspension-adapted HEK293-6E cells transiently expressing this A2a-eGFP fusion protein with a 10xHis-tag were analyzed by FSEC. Detergent-solubilized membranes were analyzed in the absence and presence of saturating concentrations of the P3NTA probe. In determining the amount of P3NTA probe that was needed for saturation, we found no evidence to suggest that this probe induced protein aggregation. The void volume of the column used in this study is 1.9 mL, and we observed an equivalent small peak in the absence (blue trace) and presence (red trace) of the P3NTA probe containing detergent-solubilized protein aggregates (Fig. 2). Given that the fluorescence emission wavelengths of eGFP and the P3NTA probe overlap, the observed fluorescent peak is the sum of both components (Fig. 2). By subtracting the fluorescent signal originating from the eGFP fusion protein in the absence of P3NTA, the fluorescent signal from the probe can be calculated. As shown in Figure 2, the addition of the P3NTA probe at a saturating concentration of 1 μM, results in a two-fold increase in fluorescence at an elution volume of 2.6 mL, corresponding to the expected elution volume for this A2a-eGFP fusion protein. We conclude that the fluorescence originating from the P3NTA probe is equivalent to that of the eGFP fusion, further demonstrating the utility of the P3NTA probe.
Figure 2. Validating A2a receptor production by FSEC. Suspension-adapted HEK293-6E cells expressing an A2a (residues 1-316)-eGFP fusion protein was detergent solubilized, and analyzed by FSEC (blue). Addition of the P3NTA probe to saturation resulted in a two-fold increase in the fluorescence of the 2.6 mL elution volume peak, corresponding to the expected position of the fusion protein (red). The void volume of the column is 1.9 mL.
Download figure to PowerPoint
The addition of the P3NTA probe (molecular weight of less than 3 kDa) is not expected to impact the elution profile of the detergent-solubilized membranes containing the A2a-eGFP fusion protein (expected to be greater than 100 kDa), as the difference in molecular weight is not sufficient to alter the elution volume (Fig. 2). As with all integral membrane proteins, the elution volume, and therefore, observed molecular weight, is dependent upon the physical properties of the membrane protein in complex with the detergent micelle that is used for membrane solubilisation.[22, 23] For example, the protein-detergent complex of the neurotensin receptor NTS1 shows an observed molecular weight of 210 kDa by size exclusion chromatography, whereas MALS analysis demonstrates that molecular weight of the dimeric protein to be 77 kDa.
Finally, a broad elution peak containing the free P3NTA and the P3NTA probe mixed with cellular debris is also observed at an elution volume greater than 3.3 mL, and is not shown in this figure.
For difficult to produce membrane protein targets, it is common practice to screen large libraries of protein constructs to identify suitable variants with improved production and stability properties. Efficient parallel screening of these variants is reliant upon a high-throughput miniaturized plate-based screening platform to evaluate protein production levels and to perform biophysical characterization. We validated the utility of the P3NTA probe for analysis of protein levels, in a high-throughput manner, using a deletion library of this GPCR. The construct library was transiently expressed in mammalian cells using a plate-based 24-well platform for suspension-adapted HEK293-6E cells. This platform allowed us to complete the analysis of triplicate cultures of ten constructs plus controls within 4 days. FSEC analysis only required approximately 100,000 cells per culture, corresponding to about 100 μL of suspension culture.
Figure 3(A) shows the FSEC results derived from the P3NTA probe for full-length A2a, five C-terminal truncations and a nontagged negative control. Using triplicate cultures for each construct, FSEC peak profiles were consistent with %CV for the peak area below 20%. Replicate samples were analyzed using automated HPLC sequences such that equivalent detergent-solubilized membrane samples were analyzed at six hour time intervals, and our results showed no significant differences in fluorescence signal. These observations were also consistent with the associated negative controls reflecting signal background. Taken together, we conclude that these experiments were performed under conditions where P3NTA binding was both saturated and constant.
Figure 3. High-throughput FSEC screening of the A2a receptor. A: High-throughput FSEC screening of full-length and truncated variants of the human A2a receptor transiently expressed in HEK293-6E cells. The A2a receptor was detergent solubilized from whole cells using dodecylmaltoside, and analyzed by FSEC by incubating the clarified extract with the P3NTA probe. FSEC peak profiles for full-length and five C-terminal truncation variants are shown. Cells transfected with pDEST12.2 oriP containing non-tagged protein is included as the negative control. B: Western blot analysis of transient HEK293-6E cell expression of full-length and five C-terminal truncations. Two independent transfections were analyzed in adjacent lanes for each construct. Target protein was detected using an anti-His antibody that recognizes the C-terminal His-tag. Prestained protein markers are shown in the left lane, and an empty vector control (showing the detergent-solubilized HEK293-6E whole cell membrane background) is shown in the right lane.
Download figure to PowerPoint
Relative elution times were found to correlate with the expected molecular weights for these variants [Fig. 3(A)], and the FSEC peak profiles are in agreement with the published literature. For example, the FSEC peak profiles for truncation 2 (residues 1–334) and truncation 3 (residues 1–316) show single monodisperse peaks [Fig. 3(A)]. Both constructs have previously been demonstrated to be stable in detergent, and truncation 3 with additional directed point mutations was used to determine crystallographic structures of this GPCR. Shorter variants, truncation 4 (residues 1–294) and truncation 5 (residues 1–279), showed no overproduction and had FSEC peak profiles that were identical to the nontagged negative control. Truncation 5 removes a portion of the seventh transmembrane helix, and would expectedly destabilize the GPCR and negatively impact the FSEC profile.
Although the boundaries for A2a constructs 2 (residues 1–334) and 3 (residue 1–316) are equivalent to those published previously, which have been demonstrated to be stable in detergent, the constructs used in this study are not precisely equivalent as they contain a N-terminal decahistidine-tag, a TEV cleavage site and a short four residue linker prior to the A2a coding sequence. Moreover, the literature construct that was used to determine the A2a crystal structure also contains a cluster of point mutations to facilitate crystallization. However, qualitative appearance of UV size exclusion chromatography profile for the detergent stable construct (residues 1–334) of A2a precisely matches the FSEC profile observed in this study using the P3NTA probe. This data demonstrate that when coupled with FSEC, the P3NTA probe can be used to identify appropriate construct boundaries using protein transiently expressed in suspension-adapted HEK293-6E cells with a plate-based miniaturized platform without the need for purification.
The overproduction levels of monodispersity A2a protein for these constructs were estimated by integrating the elution peak volume. Several assumptions were made in calculating these values: (1) calibrating the concentration of the free P3NTA probe under identical chromatographic conditions; (2) 1:1 binding stoichiometry between the P3NTA probe and the decahistidine-tag of the A2a protein; and (3) an average cell density of 1 × 109 cells/L. Using these measurements, the overproduction level of the full-length A2a fusion protein was estimated to be 0.7 mg/L. Constructs ending at residues V334 (0.4 mg/L), A316 (0.45 mg/L) or L267 (0.1 mg/L) had lower overproduction levels, whereas those ending at residues T279 or E294 could not be detected. Taken together, we estimate the detection limit of the P3NTA probe to be better than 0.02 mg/L.
The advantages of FSEC are retained when used in conjunction with the P3NTA probe in ranking of protein constructs as demonstrated by Western blot analysis [Fig. 3(B)]. Not only does FSEC provide information on the quality of the protein produced (monodispersity) rather than the total protein yield, but it also facilitates the ranking of proteins with complex banding patterns when resolved using SDS-PAGE. For membrane proteins, oligomeric bands detected by SDS-PAGE can often indicate SDS-induced protein aggregation. However, as demonstrated with the human M2 muscarinic acetylcholine receptor, the T4 lysozyme fusion construct used to determine the crystal structure of the receptor appeared to be aggregated in the presence of SDS under reducing conditions (accounting for higher molecular weight bands on the gel), whereas the receptor appeared to be monodisperse by size exclusion chromatography under nondenaturing conditions.
Although all truncation variants of A2a gave multiple strong bands on Western blots at the expected molecular weights, no detergent-solubilized protein was observed in the FSEC profiles for truncations 1–279 and 1–294 [Fig. 3(A)]. Moreover, none of these truncated A2a constructs produced monodisperse FSEC profiles. Further complexities arise in interpreting Western blotting data from the heterogeneous glycosylation patterns of recombinantly produced GPCRs, including the human A2a receptor. Although directed mutants, designed to eliminate specific glycosylation sites on the A2a receptor, displayed simplified banding patterns on Western blots, understanding the impact on protein yield and stability remains a significant challenge. The FSEC peak profiles of variants of truncation construct 3 (residues 1–316) with a single point mutation at residue N154 showed increased homogeneity, reduced molecular weights, and displayed small but consistent differences in the appearance of their FSEC peaks, and the FSEC profile of the N154D mutant can clearly be distinguished from the N154A, N154S, or N154Q mutants (Supporting Information Fig. S1). The accompanying Western blot is shown in Supporting Information Figure S2, which further emphasizes the complexity of interpreting Western blots data generated from crude whole cell lysates. Taken together these results further demonstrate the sensitivity and utility of the P3NTA probe when coupled to FSEC.
Bacterial ortholog screening
We next evaluated the utility of the P3NTA probe on a bacterial membrane protein target MraY, which is a member of the peptidoglycan biosynthetic pathway that catalyzes the translocation of a UDP-linked phospho-N-acetylmuramoyl-pentapeptide from the cytosol to the cytosolic leaflet of the cytoplasmic membrane through the formation of a covalent linkage to membrane-embedded undecaprenyl phosphate. Secondary structure predictions suggest that MraY consists of ten transmembrane helices that are connected by five cytoplasmic and four periplasmic loops.
The conceptual basis of ortholog screening is that primary sequence differences between functionally and structurally related membrane proteins can be used to identify an ortholog with improved recombinant production levels and stability for biophysical studies. A panel of twenty-four MraY orthologs, selected from a diverse dendogram composed of the primary sequences of 110 MraY proteins, identified by a PSI-BLAST search using the primary sequence of E. coli MraY, was generated. A high-throughput plate-based screening platform for expression in E. coli, was used to evaluate MraY recombinant protein production levels.
Although the E. coli ortholog of MraY has proven difficult to overproduce and purify, the B. subtilis ortholog has been produced recombinantly for a variety of biophysical studies.[27-29] This open reading frame was used to evaluate bacterial strains for protein production, growth media formulations, expression plasmids, and the position of the polyhistidine-tag prior to evaluating the 24 member MraY ortholog panel. Using the P3NTA probe, FSEC was used to rank the impact of these changes to the overproduction conditions. Optical density readings were used to normalize cell density prior to cell lysis for each condition tested. These results indicated that E. coli BL21 (DE3) cultured in Superbroth supplemented with glycerol was optimal for MraY overproduction from plasmid pET26b+ containing a cleavable C-terminal 10xHis-tag. For the 24 member ortholog panel, each open reading frame was then codon optimized for E. coli, and subcloned into this plasmid. The entire process, from bacterial transformation to FSEC analysis took 3 days, including triplicate clones and all relevant controls. As shown in Figure 4, only three MraY orthologs showed symmetrical FSEC peak profiles that were superior to the E. coli enzyme, they are B. subtilis (blue trace #25), C. bolteae (green trace #24) and T. thermophilus (red trace #23). The vast majority of the orthologs were either not overproduced or showed only minimal levels.
Figure 4. Ortholog screening of a bacterial membrane protein using FSEC. Orthologs were expressed in E. coli, and whole cell lysates were solubilized with dodecylmaltoside prior to addition of probe. Monodisperse protein elutes at 2.9 mL, whereas the free probe elutes at 3.9 mL. Cells transformed with empty expression vector were used as a negative control. To ensure that the full expression range was observed, detergent-solubilized samples were analyzed with increasing probe concentrations until a significant peak of free probe was observed at 3.9 mL. Samples: 1 - empty vector control; 2 - C. diphteriae; 3 - C. testosteroni; 4 - C. trachomatis; 5 - E. coli; 6 - E. faecalis; 7 - F. nodosum; 8 - F. placidus; 9 - F. prausnitzii; 10 - H. neptunium; 11 - H. pylori; 12 - M. thermophila; 13 - M. voltae; 14 - P. gingivalis; 15 - P. horikoshii; 16 - R. rickettsii; 17 - S. pneumoniae; 18 - T. mathranii; 19 - T. neutrophilus; 20 - M. thermautotrophicus; 21 - M. thermoacetica; 22 - L. gasicomitatum; 23 - T. thermophilus; 24 - C. bolteae; and 25 - B. subtilus.
Download figure to PowerPoint
On the basis of these results, MraY from C. bolteae was overproduced at a preparative scale. This MraY ortholog could be readily purified in milligram quantities (0.3 mg/L) to greater than 90% purity (Supporting Information Fig. S3). The anomalous migration pattern of MraY from C. bolteae on SDS-PAGE is consistent with previously reported results for the B. subtilis ortholog. Moreover, using a recently developed assay, purified MraY from C. bolteae demonstrated the expected enzymatic activity (unpublished results). Taken together these results exemplify how the P3NTA probe can be used in conjunction with FSEC to support the identification of well-behaving clones that can be overproduced at milligram levels without the need for subcloning.