An evolved Mxe GyrA intein for enhanced production of fusion proteins

Abstract

Expressing antibodies as fusions to the non-self-cleaving Mxe GyrA intein enables site-specific, carboxy-terminal chemical modification of the antibodies by expressed protein ligation (EPL). Bacterial antibody-intein fusion protein expression platforms typically yield insoluble inclusion bodies that require refolding to obtain active antibody-intein fusion proteins. Previously, we demonstrated that it was possible to employ yeast surface display to express properly folded single-chain antibody (scFv)-intein fusions, therefore permitting the direct small-scale chemical functionalization of scFvs. Here, directed evolution of the Mxe GyrA intein was performed to improve both the display and secretion levels of scFv-intein fusion proteins from yeast. The engineered intein was shown to increase the yeast display levels of eight different scFvs by up to 3-fold. Additionally, scFv- and green fluorescent protein (GFP)-intein fusion proteins can be secreted from yeast, and while fusion of the scFvs to the wild-type intein resulted in low expression levels, the engineered intein increased scFv-intein production levels by up to 30-fold. The secreted scFv- and GFP-intein fusion proteins retained their respective binding and fluorescent activities, and upon intein release, EPL resulted in carboxy-terminal azide functionalization of the target proteins. The azide-functionalized scFvs and GFP were subsequently employed in a copper-free, strain-promoted click reaction to site-specifically immobilize the proteins on surfaces, and it was demonstrated that the functionalized, immobilized scFvs retained their antigen binding specificity. Taken together, the evolved yeast intein platform provides a robust alternative to bacterial intein expression systems.

Figure 1

Surface display and secretion constructs. (a) In display construct pCT4Re, Aga2p is expressed at the carboxy-terminus to anchor the fusion protein to the yeast surface, while a FLAG epitope tag is expressed on the amino-terminus of the scFv or GFP to indicate full-length construct expression on the yeast surface. In the intein-containing display constructs, the non-self-cleaving Mxe GyrA intein is inserted between the carboxy-terminus of the scFv or GFP and the Aga2p surface anchor. (b) Secretion construct pRS316-FLAG is similar to the surface display construct, with a synthetic prepro leader sequence directing secretion and a six histidine epitope for purification.

Figure 2

Directed evolution of the Mxe GyrA intein. (a) For directed evolution round 1, the Mxe GyrA intein library was created by random mutagenesis and recombined into the pCT4Re-4420 construct. The library was screened in four rounds of enrichment for improved FLAG tag expression via FACS. A fifth round of enrichment selected for both improved FLAG tag expression and comensurate increases in fluorescein binding. Individual clones were isolated and screened for intein activity by the addition of MESNA, which releases the scFv from the display construct when an active intein is present. For directed evolution round 2, the round 1 clones were shuffled and mutagenized prior to screening for increased display levels. (b) Flow cytometry dot plots depicting expression and binding activity of scFv-intein clones and pools on the yeast surface. Geometric mean flurescence intensity (MFI) of the FLAG signal for the entire displaying population is shown to allow comparison. In addition, a sample sort gate is shown to illustrate the enrichment. Panel i, wild-type intein fusion; panel ii, round 1 final selected pool; panel iii, round 2 final selected pool, panel iv, unfused 4–4–20 scFv; panel v, round 1 202–08 intein mutant. (c) The MFI of the displaying population was quantified and normalized to the wild-type 4–4–20-intein contruct to compare the relative expression levels (FLAG) and activity (fluorescein binding) of the unfused 4–4–20 construct, wild-type intein construct, and the 202–08 intein mutant. Activity per molecule is expressed as the ratio of fluorescein binding to FLAG expression level. Plotted are the means ± SD from three independent yeast transformants. Statistically significant improvements over the wild-type intein construct were determined by an unpaired Student’s t-test (*p < 0.05; **p < 0.01; NS, not significant p > 0.05). Display data for other individual intein mutants are compiled in Table 1. (d) Quantitative anti-FLAG Western blotting was performed to determine the relative amount of 4–4–20 released from the yeast surface in the MESNA reaction. Plotted are means ± SD for three independent reactions originating from three independent yeast surface display transformants. Next to the bar graph are the triplicate Western blot data at the cleaved scFv size of ∼30 kDa. A small amount of the uncleaved, scFv-intein product appears at a size of ∼90 kDa due to its fusion to glycosylated Aga2p. The double asterisk represents a statistically significant increase in 4–4–20 release for clone 202–08 (p < 0.01) as determined by an unpaired Student’s t-test. (e) The crystal structure of the Mxe GyrA intein (pdb ID: 1AM2(64)) is shown with the mutations found in the 202–08 intein highlighted. A flexible loop missing from the crystal structure is denoted by a dotted line and the structure on the right was rotated 90°.

Figure 3

Analysis of surface displayed scFv- and GFP-202–08 fusions. (a) Surface display levels of unfused, wild-type intein fused or 202–08 intein fused scFvs and GFP were analyzed by flow cytometry. The MFI of the FLAG-positive yeast populations was quantified, and all were normalized to the 4–4–20 construct containing the wild-type intein. Reported are the means ± SD of three independent yeast transformants. Statistical analysis was performed by an unpaired Student’s t-test (*p < 0.05; **p < 0.01; NS, not significant p > 0.05). (b) ScFv and GFP per molecule activity was evaluated by detecting binding to the scFv antigens at saturating ligand concentrations or by measuring GFP fluorescence. Activity per molecule was determined by calculating the ratio of the geometric means for activity (binding or fluorescence) to FLAG expression levels and normalizing to the unfused construct lacking intein. Plotted are the means ± SD from three independent yeast transformants, with statistical significance determined by an unpaired Student’s t-test (*p < 0.05; **p < 0.01; NS, not significant p > 0.05) (c) For intein-mediated protein release, MESNA reacts to release the scFv or GFP from the display construct and append a carboxy-terminal thioester. For EPL functionalization, the carboxy-terminal thioester reacts with a biotinylated peptide containing an amino-terminal cysteine to covalently link the scFv or GFP to the biotin by an amide bond. (d) Products of the reaction depicted in panel c resolved and analyzed by Western blotting to detect release of the scFv or GFP (∼30 kDa) from the 202–08 intein construct using an anti-FLAG antibody (F) or biotin functionalization via EPL with an anti-biotin antibody (B). A small amount of uncleaved scFv-intein-Aga2p product can be seen in the anti-FLAG Western blot between ∼80 kDa and 100 kDa due to the glycosylation of Aga2p.

Figure 4

Secretion of scFv and GFP intein fusion proteins. (a) Yeast supernatants containing scFv or GFP fused to the wild-type intein or 202–08 intein were subjected to anti-FLAG quantitative Western blotting and compared to the unfused target protein. Values are normalized to the level of the 4–4–20–202–08 fusion to determine relative amounts. The absolute secretion titer of the 4–4–20–202–08 fusion protein is 3.1 mg/L as determined in panel b. Reported are the means ± SD from three independent yeast transformants. Statistical significance was determined by an unpaired Student’s t-test (*p < 0.05; **p < 0.01; NS, not significant p > 0.05). Western blot of supernatant samples used for the quantitation of relative 4–4–20 protein secretion is shown below the bar graph. (b) An equilibrium binding curve was generated by fluorescein quenching to compare the K_d of unfused 4–4–20 and 4–4–20 fused to 202–08. A sample curve for each of the proteins is shown, and the mean ± SD for the fitted parameters of K_d value and 4–4–20 concentration were obtained by fitting quench curves generated from supernatants resulting from three independent yeast transformants. From the molar concentrations of 4–4–20, the average mass concentration of the 4–4–20 component was calculated to be 1.6 mg/L of yeast culture for both the unfused and the intein-fused 4–4–20 (corresponding to 3.1 mg/L for the full 4–4–20–202–08 fusion protein) The K_d and 4–4–20 concentrations were statistically indistinguishable, as determined by an unpaired Student’s t-test (p > 0.05). (c) GFP activity was determined by calculating the ratio of fluorescence to FLAG expression levels and normalizing to the unfused construct lacking intein. The mean ± SD results from three independent yeast transformants. The fluorescence per molecule of unfused GFP and 202–08 fused GFP was statistically indistinguishable, as determined by an unpaired Student’s t-test (**p > 0.05). (d) The catalytic activity of 202–08 was examined by reacting secreted and purified proteins with MESNA and evaluating cleaved yield after standard 20 h reaction. Anti-FLAG Western blotting demonstrates between 70% (2224) and 99% (MR1) release of the target protein from the 202–08 intein in the presence of MESNA.

Figure 5

Strain-promoted click chemistry immobilization. (a) Secreted and purified scFv and GFP proteins fused to the 202–08 intein were released with MESNA to form scFv- and GFP-thioesters. The carboxy-terminal thioesters were subsequently reacted with a cysteine azide via EPL to install an azido group onto the protein. To immobilize the proteins on surfaces, the scFv- and GFP-azide proteins were reacted with DBCO-functionalized agarose beads in a strain promoted click chemistry reaction. (b) Fluorescent microscope images of GFP fluorescence associated with beads reacted with GFP-azide or nonazido GFP (GFP-thioester). Relative protein immobilization was quantified by measuring total bead fluorescence and normalizing to the azide-GFP loaded beads. The mean ± SD of three independent immobilization reactions is plotted. Statistical significance was determined by an unpaired Student’s t-test (**p < 0.01) (c) Binding of fluorescein to beads reacted with azide functionalized 4–4–20 was analyzed and compared to beads reacted azide-linked scFv2. FITC-dextran binding was quantified by measuring the fluorescence intensity of the beads, and the fluorescence was normalized to the 4–4–20-linked sample. Three independent immobilization reactions were carried out to obtain the mean ± SD values. An unpaired Student’s t-test was performed to determine statistical significance (**p < 0.01) (d) Immobilized EGFR scFv activity was assessed by EGFR capture from cell lysates. Fluorescent microscopy images were employed to demonstrate EGFR capture and EGFR isoform specificity. A431 cells express wild-type EGFR while U87 cells are transfected to express the EGFR vIII isoform. ScFv activity was quantified by measuring the resulting fluorescence intensity of the beads, and the fluorescence value was normalized to the signal originating from the U87-EGFRvIII lysate binding to the respective scFv. The fluorescence value for the negative control, 4–4–20, was normalized to the signal originating from the U87-EGFRvIII binding to MR1. The mean ± SD of three independent immobilization reactions is plotted. Statistical significance was determined by an unpaired Student’s t-test (**p < 0.01).