High-throughput Identification of DNA-Encoded IgG Ligands that Distinguish Active and Latent Mycobacterium tuberculosis Infections

Abstract

The circulating antibody repertoire encodes a patient's health status and pathogen exposure history, but identifying antibodies with diagnostic potential usually requires knowledge of the antigen(s). We previously circumvented this problem by screening libraries of bead-displayed small molecules against case and control serum samples to discover "epitope surrogates" (ligands of IgGs enriched in the case sample). Here, we describe an improved version of this technology that employs DNA-encoded libraries and high-throughput FACS-based screening to discover epitope surrogates that differentiate noninfectious/latent (LTB) patients from infectious/active TB (ATB) patients, which is imperative for proper treatment selection and antibiotic stewardship. Normal control/LTB (10 patients each, NCL) and ATB (10 patients) serum pools were screened against a library (5 × 10⁶ beads, 448 000 unique compounds) using fluorescent antihuman IgG to label hit compound beads for FACS. Deep sequencing decoded all hit structures and each hit's occurrence frequencies. ATB hits were pruned of NCL hits and prioritized for resynthesis based on occurrence and homology. Several structurally homologous families were identified and 16/21 resynthesized representative hits validated as selective ligands of ATB serum IgGs (p < 0.005). The native secreted TB protein Ag85B (though not the E. coli recombinant form) competed with one of the validated ligands for binding to antibodies, suggesting that it mimics a native Ag85B epitope. The use of DNA-encoded libraries and FACS-based screening in epitope surrogate discovery reveals thousands of potential hit structures. Distilling this list down to several consensus chemical structures yielded a diagnostic panel for ATB composed of thermally stable and economically produced small molecule ligands in place of protein antigens.

Figure 1. DNA-encoded combinatorial library plan

The library synthesis reaction sequence consisted of three acylations, enumerated as diversification positions Pos₁ (green), Pos₂ (orange), and Pos₃ (purple). The BB set included carboxylic acids (X_i, amino acids and haloacids) for diversifying the main chain scaffold and amines (A_i) for diversifying the intervening amide via displacement of the haloacids. Pos₁ diversity included 10 amino acids, 3 haloacids, and 20 amines. Pos₂ and Pos₃ diversity included 4 haloacids and 20 amines. If a BB was used at a given position, its indicator (upper right) is filled with the appropriate color.

Figure 2. Serum IgG binding assay schematic and FACS-based high-throughput screening data

(a) In single-color screens, library beads are incubated with serum-containing IgGs (gray IgGs), some of which bind specific beads. Probing with Alexa Fluor 647 anti-human IgG conjugate (red IgG, λ_em = 660 nm) labels serum IgG-bound beads for collection in FACS. (b) In two-color screens, serum pools are pre-labeled with anti-human mFab. The NCL pool is labeled with anti-human mFab488 (green Fab, λ_em = 530 nm), the ATB pool is labeled with anti-human mFab647 (red Fab, λ_em = 660 nm), and the pools are mixed. Probing the library with this mixed serum generates three populations of IgG-bound beads in multiplexed FACS analysis: NCL-specific IgG binding correlates with 530-nm fluorescence (upper left quadrant), ATB-specific IgG binding correlates with 660-nm fluorescence (lower right quadrant), and non-specific IgG and mFAB binding correlates with both channels (diagonal). FACS analysis of a single-color of pooled NCL serum screen (c) and pooled ATB serum screen (d) was gated for collection of hits > 30,000 RFU (λ_em = 660 nm, red). Side scatter (SSC) was used to gate single beads. Two-color FACS analysis of a mixed, pre-labelled NCL/ATB serum pool screen (e) wastwo-dimensionally gated (black lines) for hit collection (red).

Figure 3. DNA-encoded solid-phase synthesis and bead-specific barcoding

(a) The DNA-encoded solid-phase synthesis bifunctional resin linker displays amine sites for compound synthesis and DNA headpiece sites (HDNA, a tether that covalently joins the two DNA strands) for enzymatic ligation of encoding oligonucleotides. The encoding tag contains a synthesis encoding region and bead barcoding region flanked by forward and reverse primer binding modules (gray). After ligation of the forward primer sequence, each monomer coupling step accompanies an enzymatic cohesive end ligation that installs a dsDNA encoding module. A submonomer approach includes various main chain scaffold structures (purple) and amine side chains (orange). Corresponding encoding modules appear in the same color. After encoded synthesis, ligation of two additional encoding modules assigns a bead-specific barcode, and reverse primer ligation completes the encoding tag. (b) Bead-specific barcodes distinguish beads that harbor identical compounds, which would otherwise display identical DNA sequences. (c) Combinatorial ligation of i sequence modules in the first bead-specific barcoding position (cyan hues) and j sequence modules in the second position (green hues) yields i × j possible unique bead-specific barcodes.

Figure 4. Pan-library structure-activity relationship profile and high-priority hit validation

Left, the position-dependent occurrence frequency of each monomer structure (% observed) was plotted for decoded structures observed in the single-color screening hit collection (blue), two-color screening hit collection (cyan), and a random sample of the library (black). Library values are shown with the standard error of three random library samples and the gray line indicates the theoretical monomer frequency as calculated from the library plan. Several monomers occurred significantly more frequently in hit structures. The monomer index is the 8-digit sequence identifier that specifies the monomer structure and position. Center, high-priority exemplar hit structures 1 – 8 are shown with colored structures containing the identically colored high-frequency monomer indices. Right, 1 – 8 were purified, re-immobilized on beads, and their ATB serum IgG binding (blue bars) compared to NCL serum IgG binding (gray bars) at two serum concentrations (0.25 and 1 mg/mL; 0.25 mg/mL binding data are magnified 4 fold). All compounds statistically significantly bound more ATB serum IgG than NLC serum IgG (p = 0.005).

Figure 5. Patient serum IgG binding profiles

Each hit ligand's serum IgG binding was evaluated for individual patient serum samples classified as TB negative control, latent TB, or active TB. Samples of each classification were either a component of “discovery” serum pools used for library screening or additional “test” samples. The binding behavior of each serum sample is displayed as side-by-side color-coded bars. The left bar indicates whether IgG binding exceeded the statistical significance threshold for the ligand, the right bar indicates the fraction of serum IgG bound in the presence of soluble ligand as the competitor (10 μM).

Figure 6. Hit compound validation and native antigen identification

(a) Beads displaying 2-B bound statistically significantly more ATB discovery serum pool IgG compared to the NCL discovery serum pool IgG over a wide range of [serum]. Competition binding analysis of 2-B revealed competitive binding of hypervirulent culture filtrate proteins (CFP, 250 μg/mL) derived from several hypervirulent Mtb strains (HN878, CDC1551, H37Rv), while E. coli and Mtb lysates weakly competed (b). Purified Mtb proteins Ag85A and Ag85B competed (the latter strongly so) though the recombinantly expressed forms were unreactive. (c) Competition titration analysis of native Ag85A and Ag85B with beads displaying 2-B revealed selective reactivity with Ag85B. (d) ELISA analysis of all serum samples using non-specifically immobilized native Ag85B as the antigen yielded 22% diagnostic sensitivity and 100% specificity.