Considerations for the expression strategy of SAG1 (SAG2A)
The GPI-anchored surface proteins of T. gondii tachyzoites, which include SAG1 and SAG2A, have well-known N- and C-terminal topogenic signal sequences [16, 24]. However, surprisingly little attention has been paid in the past to their potential influence on antigenicity of the recombinant proteins used for diagnostic purposes when left intact in (see e.g. [11, 25-29]). Likewise, deletions as well as N-terminal fusions with relatively large proteins like glutathione-S-transferase (GST) were introduced. However, in the case of dimeric SAG1, a previous study by Graille et al.  provided convincing evidence that a conformational epitope of the monomers, important for recognition by human antibodies from infected individuals, is found at the N-terminus of the mature protein (Fig. 1).
This conclusion was based on the 3D structure of a complex of a monoclonal antibody (mAb) bound to SAG1. This mAb competes very efficiently with the binding of human antibodies by making contact with discontinuous N-terminal residues, forming what appears to be the immunodominant epitope of SAG1 (highlighted in blue in the dimeric form; Fig. 2) . Thus, we considered it to be important to conserve the structural integrity, in particular access to the N-terminus of the protein, when expressing recombinant SAG1. Consequently, full length, non-fused and correctly folded dimeric SAG1  is considered to be the best antigen for optimal recognition by human antibodies.
Furthermore, since our main objective was to use SAG1 in BBMA where the usual immobilization of proteins to the Luminex microbeads is via chemical coupling we reasoned that this could affect SAG1’s recognition by antibodies. In this immobilization procedure lysine side chains, in particular those that are surface-exposed, are coupled in a non-selective manner via EDC (1-Ethyl-3-[3- dimethylaminopropyl]carbodiimide hydrochloride) and Sulfo-NHS (N-hydroxysulfosuccinimide) to the carboxy groups of the beads. Dimeric SAG1 contains 40 lysine residues, of which 38 have a calculated solvent accessible surface area, SAS, (as determined in the known 3D structure) ≥ 40 Å2 (highlighted in black in Fig. 2). Of those, 21 have a SAS ≥ 100 Å2 (cyan in Fig. 2), providing a rich landscape of potential attachment sites. Several lie within or very close to the dominant epitope, thereby possibly destroying or severely affecting antibody binding. Consequently, a targeted immobilization strategy that would allow SAG1 to be coupled exclusively via its C-terminal end (similar to its GPI anchor attachment in the plasma membrane; [24, 30]) could improve immune recognition by human sera.
It is long known that efficient humoral SAG1 recognition depends on correct folding of the protein. From Fig. 2 it is apparent that proper formation of three of the six disulfide bonds of SAG1 will directly affect the formation of the dominant epitope (see Additional file 1: Movie S1). Recombinant truncated SAG1 versions lacking any of these disulfide bonds (e.g. ) will therefore be suboptimal.
Taken all this into account our rationale for the expression construct for SAG1 (and also SAG2A) was as follows: the recombinant protein should
- contain the entire mature coding region to include all possible epitopes of the native protein (Fig. 1),
- allow correct S-S bonding, thereby maximizing correct folding,
- allow oriented, controllable immobilization on magnetic beads ,
- possess a cleavable fusion partner to aid in increased solubility.
Construction of a three-plasmid expression system for SAG1/SAG2A
To accomplish above aims pAviTag-MBP-SAG1 and pAviTag-MBP-SAG2A were designed (Fig. 3; Additional file 1: Figure S1) and assembled as described in the Methods section. Both proteins were fused N-terminally with maltose binding protein (MBP), which has been shown to be superior for enhanced solubility during translation and folding . At the same time MBP had finally to be cleaved off in situ so that antibody access to epitopes is not inhibited, as discussed above. Therefore, MBP is followed by a cleavage recognition site (tev) for the Tobacco Etch Virus (TEV) protease  that would lead in the case of SAG1 to mature authentic protein with Ser31 as N-terminal amino acid (see Additional file 1: Figure S1A). The putative GPI-attachment site (Gly289) at the C-terminus is followed by a 4 kDa peptide sequence (AviTag) that is recognized by E. coli biotin ligase BirA, catalyzing the attachment of biotin at the lysine within the sequence . The resulting biotinylated protein can thus be immobilized via its C-terminal end by biotin-streptavidin interaction in an oriented fashion. The AviTag is followed by a His6 tag for affinity purification by metal chelate affinity chromatography (Fig. 3; Additional file 1: Fig. S1A).
The six disulfide bonds of SAG1 pose a challenge for correct folding in a reducing environment like E. coli’s cytosol . We therefore chose the system developed by Nguyen et al.  that allows improved cytoplasmic disulfide bond formation in E. coli (called ‘CyDisCo’). It consists of the pre-expression of a sulfhydryl oxidase together with a protein disulfide isomerase (PDI) and can be combined with an E. coli strain deleted of the genes for gor and trxB . The latter two genes are involved in disulfide bond reduction. Their deletion and the additional expression of DsbC in the bacterial cytoplasm results in better disulfide bond formation in the E. coli strain SHuffle . Plasmid pMJS9 contains genes for codon-optimized sulfhydryl oxidase Erv1p from Saccharomyces cerevisiae and codon-optimized human PDI, regulated by an arabinose-inducible promoter  (Fig. 3).
BirA is present only in very small amounts in E. coli cells and therefore its overexpression is required for substantial in vivo biotinylation . A third plasmid, pBAD1031-TB, expresses TEV protease and BirA (Fig. 3; Additional file 1: Figure S1B). Although it has been shown to be active as an N-terminal fusion protein  we opted for a construct where the sequences for TEV protease and BirA are separated by a tev cleavage site. Such an arrangement has been shown to result in post-translational self-processing of the fusion protein in stoichiometric amounts of the individual protein entities .
Strain E. coli SHuffle containing the three plasmids, each possessing a different resistance gene as well as compatible replication origins, was named BioSAG1 (Fig. 3). A similar strain with pAviTag-MBP-SAG2A was constructed and termed BioSAG2A.
Expression, purification and characterization of biotinylated SAG1 and SAG2A
Recombinant protein production in the BioSAG strains starts by addition of arabinose, which induces expression of Erv1p and PDI on pMJS9 as well as TEV protease and BirA on pBAD1031-TB due to the presence of the arabinose-inducible promoter on both plasmids. Such pre-expression has been reported previously to increase correct S-S bond formation  as well as biotinylation . Then, rhamnose is added, leading to production of MBPtev-SAG1-AviTag-His6 (MBPtev-SAG2A-AviTag-His6), on which the pre-expressed proteins then can act upon (i.e. disulfide bridge formation and correct folding by Erv1p, PDI and DsbC; cleavage of MBPtev-SAG1 and TEVtev-BirA by TEV protease; biotinylation by BirA). This regimen results in soluble expression of N-terminal fusion-free SAG1bio-His6, as seen in Fig. 4A, where a cell lysate of BioSAG1 was separated into soluble and insoluble fractions and analyzed by SDS-PAGE followed by immunoblotting with a mouse mab (DG52) that recognizes a disulfide bond-dependent conformational epitope [18, 23, 41]. While the pellet still contains substantial amounts of insoluble protein, in both fractions DG52 recognizes its epitope, indicative of proper disulfide bond generation. In situ cleavage of MBP by TEV protease is rather efficient since only small amounts of DG52 reactivity is seen at a size of >70 kDa, the size of the fusion protein (calculated Mw of 74,8 kDa). As shown in Fig. 4B BirA can be detected as a single protein band of the expected size (ca. 37 kDa) upon induction only in a strain that contains pBAD1030G-TB, indicating successful self-cleavage of the TEVtev-BirA fusion protein. Endogenous BirA is undetectable in a strain lacking the plasmid, consistent with the low endogenous amount of the ligase under normal growth conditions [42, 43].
Adding a second affinity chromatography step to the purification procedure (see Fig. 3 and Methods) allowed entire MBP (cleaved or as fusion) retention on the dextrin affinity column after prior buffer exchange of the eluate on a desalting column, leading to the purification of SAG1bio-His6 as well as SAG2Abio-His6 to homogeneity (Fig. 5A). Probing a blot of both proteins with Sav showed that they were also biotinylated (Fig. 5B,C).
Using this expression system we could purify several hundred micrograms of pure SAG1bio-His6 and SAG2Abio-His6, respectively, from 1 liter of bacterial culture. It should be noted, however, that protein preparations that contain e.g. uncleaved MBPtev-SAG1bio-His6 (Fig. 4C) that had been co-purified on the metal chelate affinity column could still be used efficiently for BBMA, with a higher overall yield than the optimized 3-step protocol.
Bead-based multiplex assay with biotinylated SAG1 and SAG2A as antigens
The overall aim of this study was to establish a BBMA with biotinylated SAG1 and SAG2A as antigens for analysing seroconversion due to T. gondii infection in humans. Magnetic beads have distinct advantages over non-magnetic ones, like ease of processing/washing, higher bead recovery etc. [8, 44]. However, since at the beginning of these studies Sav-coated MagPlex® microbeads were not commercially available, we custom prepared them by chemical coupling of Sav to various bead regions (see Methods).
We determined the minimal amount of protein that would be required to obtain maximal MFI with human control sera of known anti-T. gondii IgG antibody titers (Fig. 6A). Ten ng per serum sample of a SAG1bio-His6 preparation similar to Fig. 4C were shown to be sufficient to obtain MFI of >25,000, the maximum MFI value that is usually useful. Those sera could be titrated down to more than a 1:12,000 dilution, with still positive signals above those obtained by a negative control serum (Fig. 6A). This indicates that the obtained dose-response curve allows also low amounts of antibodies to be specifically detected.
Using a panel of 27 human sera previously tested positive (11 sera) or negative (16 sera) for anti-T. gondii antibodies by a commercial ELISA (Euroimmun) both antigens allowed a clear distinction between those donors, with a good correlation between positive titers determined by the commercial test vs. our BBMA titers (Fig. 6B and 6C). To verify these data an additional set of 50 positive and 50 negative sera each was analyzed (Fig. 6D and 6E), whereby titers in these human sera had been determined previously with a commercial automated ELIFA (bioMérieux) that is in clinical use and showed a sensitivity above 99% and specificity above 98% in comparative studies . We essentially obtained similar results with SAG1bio-His6, allowing a perfect discrimination between positive and negative sera as classified by the ELIFA, whereas analysis of SAG2Abio-His6 beads showed a slightly lower sensitivity and specificity of 98% each (see also Table 1).
Finally, the high diagnostic value of our recombinant proteins in a BBMA was proven by testing a panel of 102 sera with titers slightly below or above the diagnostic cut-off of the ELIFA (8 IU/mL). In this assay, sera between 4 and 8 IU/mL are classified as equivocal, while sera above 8 and below 4 IU/mL are considered positive or negative, respectively, by the manufacturer. When those results were compared with our BBMA for SAG1bio-His6 and SAG2Abio-His6, equivocal sera could also not be discriminated, showing an almost perfect 50/50 ratio of positive and negative sera (Fig. 7). In contrast, with sera ≥8 IU/mL or <4 IU/mL we were able to classify them as either positive or negative with high confidence, showing that highly similar performance and sensitivity of our antigens can be obtained compared to commercial assays, even when sera close to the cut-off values are analyzed.
N-terminal MBP influences binding of human antibodies to SAG1
As a proof for our hypothesis that N-terminal fusions to SAG1 would influence the binding of human antibodies we coupled MBPtev-SAG1bio-His6 purified from a strain devoid of TEV but expressing BirA (from pBAD1031-B) to Sav-coated beads (Fig. 8 inlet). We then added TEV protease to one half of the beads to release MBP from SAG1 and incubated them for various times. Cleavage was very efficient even after 1h, indicated by only minute anti-MBP mab binding (Fig. 8). Probing these as well as TEV protease-untreated beads allowed us to quantitatively compare binding of anti-SAG1-directed antibodies present in human sera since the amount of bead-bound SAG1bio should be identical between both conditions. Whereas negative sera showed no binding in any case, removal of MBP lead to a higher fluorescence intensity (30-35%; Fig. 8) with the four tested sera. The effect was less pronounced (10-20%, depending on the serum) with lower amounts of initial protein (data not shown). Notably, the observed high activity of TEV protease on the fusion protein could even allow omitting the in situ cleavage by plasmid-encoded TEV protease and the dextrin affinity column step. Instead, one could just rely on the in vitro cleavage protocol of MBPtev-SAG1bio-His6, purified only by metal chelate affinity chromatography.
We conclude that N-terminal fusion proteins do influence the binding of human antibodies to SAG1 and that their removal result in less protein being required for BBMA. However, uncleaved MBPtev-SAG1bio-His6 is still a very useful antigen for this purpose.