Application of phage display to high throughput antibody generation and characterization

Abstract

We have created a high quality phage display library containing over 1010 human antibodies and describe its use in the generation of antibodies on an unprecedented scale. We have selected, screened and sequenced over 38,000 recombinant antibodies to 292 antigens, yielding over 7,200 unique clones. 4,400 antibodies were characterized by specificity testing and detailed sequence analysis and the data/clones are available online. Sensitive detection was demonstrated in a bead based flow cytometry assay. Furthermore, positive staining by immunohistochemistry on tissue microarrays was found for 37% (143/381) of antibodies. Thus, we have demonstrated the potential of and illuminated the issues associated with genome-wide monoclonal antibody generation.

Figure 1

Schematic representation of the process for the high throughput generation and screening of recombinant antibodies.

Figure 2

High throughput generation and characterization of recombinant antibodies. (a) For each target antigen, 94 clones were screened in ELISA and sequenced to identify unique clones. The plot shows the number of antigens selected (x-axis) versus the number of unique positive antibodies generated (y-axis) for each antigen. Separate plots are presented for antigens produced in either bacterial or mammalian expression systems, illustrating the improved success rate for mammalian antigens. (b) Example specificity data for antibodies selected against Slam f9 (produced in the HEK293 mammalian expression system). All antibodies are screened against target antigen, the relevant fusion partner that was used in selection, keyhole limpet hemocyanin (KLH), thyroglobulin, myoglobin, cytochrome c, human IgG, laminin, fibronectin, α-glycerol phosphate dehydrogenase, and total protein lysates from zebra fish (D. rerio) and yeast (S. pombe). Results are shown for 22 different antibodies as well as our routine anti-desmin control (des-D7) and a no antibody control. Detection was via time resolved fluorescence and values are shown on a logarithmic scale. (c) Global summary of secondary ELISA data for all antibodies in secondary screening. Signal generated on specific antigen is shown for all 4,437 samples (solid block). Signal achieved on one of the ten irrelevant antigens (cytochrome c) is also shown.

Figure 3

Frequency of VH and V kappa/V lambda germline gene combinations selected from the library. High quality sequence of all 4,437 clones undergoing secondary screening was generated by sequencing with six primers covering the VH and VL gene segments in both forward and reverse orientations. A consensus sequence was generated and the most closely related antibody germ line genes were identified. (a) Frequency of different combinations of VH and V kappa/V lambda germ-line genes occurring in the selected antibodies, represented both numerically and by color coding. (b) Frequency of different combinations of VH and V kappa/V lambda germline genes among the selected antibodies.

Figure 4

Assessing performance of a panel of anti-Jagged-1 antibodies. (a) Flow cytometry calibration beads with varying number of anti-human Fc antibodies were coated with Jagged-1-Fc fusion to yield antigen display levels of 29,000, 83,000, 204,000 and 619,000 Jagged-1 molecules/bead. These were labeled with a panel of different recombinant antibodies raised against Jagged-1 and binding was detected with labeled anti-FLAG antibodies. The resulting histograms are shown, giving different levels of sensitivity. In Jag1_D5 for example, five peaks are visible corresponding to uncoated beads and each of the four antigen coated beads. In the case of Jag1_C5, where there is lower sensitivity, only the two beads with highest density are resolved while the others merge with that of the uncoated bead. Where all four beads are clearly resolved, we have calculated the theoretical limit of detection of receptors per bead (Rec. #). (b) 46C mouse embryonic stem cells were stained with the panel of Jagged-1 antibodies and analyzed by flow cytometry. Staining for three antibodies (Jag1_D5, Jag1_D7 and Jag1_A1.2) is shown. The sensitivity of each corresponds to that seen in the bead assay above. (c) A plot of normalized ELISA scores with performance in the bead-based flow cytometry performance assay for nine Jagged-1 specific antibodies. ELISA was carried out using 1 μg/ml of purified antibody. The mean time resolved fluorescence score (expressed as Eu counts) was plotted against median fluorescent intensity determined for an antigen density of 618,888 antigens/bead. The line represents a linear regression analysis of the data (R² value = 0.8323). (d) Immunohistochemical (Jag1_A6 scFv antibody) and (e) in situ hybridization staining of Jagged-1 in developing kidney of E14.5 mouse embryo, demonstrating localization of staining (dark purple) in developing renal tubules.

Figure 5

Detection sensitivity in a bead-based performance assay for a panel of antigens. The sensitivity limits of 90 antibodies to 9 different antigens were tested using a mix of antigen coated beads of different antigen densities. To summarize these data, the relative median fluorescent intensity on the bead comprising 459,000 antigen copies per bead was calculated (actual values and clone IDs given in Additional data file 2) and the ratio relative to uncoated beads plotted. To further illustrate sensitivity achieved, the identity of the lowest density bead that could be detected is indicated by the data label according to the following guide: 459,000 antigen copies per bead only (diamonds), down to 57,000 antigen copies per bead (square), and down to 18,000 antigen copies per bead (triangles). The antigens were: 1, Efna2; 2, Efna4; 3, Plaur; 4, Alcam; 5, Il13ra1; 6, Sigrr; 7, Ngfr; 8, Cd22; and 9, Vcam1.

Figure 6

High throughput immunohistochemistry on tissue microarrays. Example data demonstrating cell surface staining of (a-c) CD5 (antibody ant165_155_E04) on lymphocytes in human lymphoid tissue, (d-e) nerve growth factor receptor (antibody ant54_71_C07) in nerve bundles attached to the murine esophagus and in developing nerve tracts in E14.5 embryo and (f-h) nuclear staining of the transcription factor ELF1 (antibody ant46_62_F12) in cerebrum, liver and uterus. Tissues are: (a) lymph node; (b) skin; (c) Payer's patch in small intestine; (d) murine nerve bundles in esophagus; (e) nerve tracts in 14.5 day embryo with myoblast sheet above; (f) cerebrum; (g) liver; and (h) uterus.