Evaluation of multiplexed liquid glycan Array (LiGA) for serological detection of glycan-binding antibodies

Abstract

We test the performance of the multiplexed liquid glycan array (LiGA) technology in serological assays. Specifically, we use LiGA to detect ABO blood group antibodies in human serum. This LiGA, which we name ABO-LiGA, contains ABO blood group trisaccharide glycans with an ethylazido aglycone conjugated to groups of ten multi-barcoded M13 particles carrying dibenzocyclooctyne (DBCO) on p8 proteins. ELISA clonal binding assays to anti-A/B antibodies confirmed the functional performance of ABO-clones and aligned with next-generation sequencing (NGS) of the mixed clones. Multiple DNA-barcoded technical replicates in LiGA allow for quantification of reproducibility and robustness as determined by the Z'-score using NGS. We then tested ABO-LiGA for specific detection of IgG and IgM anti-A and anti-B IgG and IgM antibodies in human serum samples. Comparison of antibody binding responses in sera from 31 healthy donors to ABO-LiGA with an ABO-Luminex-based method revealed consistent responses to LiGA-ABO but also minor deficiencies of ABO-LiGA such as low robustness of the current assay format and a limited ability to detect low intensity antibody responses. Some results point to undesired interactions of serum antibodies with small-footprint glycans conjugated to phage via the bulky DBCO moiety. This report illuminates the path for future development of LiGA-based serological assays and suggests the need to develop alternative methods for conjugating glycans to phage to avoid liabilities of the hydrophobic DBCO moiety.

Keywords: anti-glycan antibodies; blood groups; liquid glycan array; serological assay.

Fig. 1

Synthesis and characterization of ABO phage glycoconjugates a) representation of the two-step chemical glycosylation of phage with the H disaccharide antigen. b) MALDI mass spectrometry characterization of the starting material (protein pVIII), the alkyne-functionalized product (DBCO–pVIII, P1) and the glycoconjugate product (P2). c) Representation of the synthesis of MSDB glycan conjugates. d) MALDI-TOF spectra of MSDB conjugated to ABO glycans with varying densities. This modification density was achieved by controlling the reaction conditions and concentration of DBCO-NHS ester. For example, by changing the concentration of DBCO-NHS from 1.5 mM to 1 mM and 0.5 mM, we could produce conjugates tri-AN3-[1500], tri-AN3-[600], tri-AN3-[90], tri-BN3-[1800], tri-BN3-[600], tri-BN3-[120], Di-N3-[1500], Di-N3-[750], Di-N3-[120]. Numbers in square brackets indicate the number of glycans per phage particle. These numbers have been determined by measuring the ratio of pVIII and DBCO-pVIII in the alkyne-functionalized intermediate. We note that at the highest modification density, a minor amount of the double-modified pVIII is also visible in MALDI spectra. Based on the MS and pVIII fragmentation studies in the previous report, we attribute major modification to amide bond formed with N-terminua of Alanine1 of pVIII whereas a minor second modification likely taking place at the side chain of Lysine8 of the same pVIII protein.

Fig. 2

Functional validation of ABO-phage conjugates using ELISA a) representation of the phage-ELISA assay. b) Dilution studies to determine the optimum coating concentration of anti-a (top) and anti-B (bottom) antibodies: Antibodies were coated at five different dilutions and tested with phage glycosylated with either the a or B antigen. Indicated p-values are calculated using two tailed t test. c) EC₅₀ of ABO phage particles as determined using dose titrations. d) Graphical illustration of functional validation of MSDB phage-ABO glycoconjugates using a plaque forming assay. e) Recovery of phage clones glycosylated with tri-AN3 and tri-BN3 on anti-a, anti-B antibodies and BSA coated wells. f) Recovery of MSDB-ABO LiGA library on anti-a, anti-B antibodies, isotype mouse IgM and BSA coated wells.

Fig. 3

Density scan of MSDB ABO₃-LiGA anti-a and anti-B antibodies and Z’ score analysis of MSDB ABO₃-LiGA screening on anti-a anti-B antibodies. Error bars designate variance propagated from the variance of the NGS data for input and output populations. Asterisks indicate significance of enrichment over median population.

Fig. 4

Comparison of the ABO-Luminex and ABO _M -LiGA for 31 serum samples. A) Heatmap with the mean MFI of IgG and IgM binding to Luminex beads bearing a or B type tri- and tetrasacharides. B) FC enrichment of ABO_M-LiGA glycans on IgG and IgM-coated surfaces. ABO_M-LiGA employs three densities of AN3, BN3, and diN3. LiGA data is a median of measurements from 8–10 DNA-encoded MSDB replicates for each glycan. For unaveraged data and dispersion in the data due to PCR amplification or NGS analysis, see the supporting information. Data acquisition failed for 8/31 IgG and 9/31 IgM samples (grey bars). c) Comparison of the AN3 and BN3 data for IgG and IgM repertoires, color intensities are identical to those in panels a or B. In A–C, data is clustered by blood type in the rows and by glycans in the columns. D) Scatter plot of MFI from Luminex vs. log2(FC) from LiGA; numbers denote the number of datapoints in each quadrant.