High-Throughput Antibody Generation Using Multiplexed Immunization and Immunogen Array Analysis

Abstract

In this work, the authors developed a new screening approach using multiplexed immunization and immunogen array analysis to improve the efficiency of antibody screening for high-throughput antibody generation. The immunogen array is based on a 96-well format in which different immunogens and negative as well as positive controls are immobilized in each well, thus making it possible to screen hundreds of antibody candidates simultaneously. To demonstrate this approach, a model of 4 mixed immunogens immunization was employed. In total, 675 antibody candidates were screened before and after established antibody hybridomas in parallel with immunogen arrays and enzyme-linked immunosorbent assay. The signal intensity, specificity, and cross-reactivity of produced antibody candidates were analyzed using a hierarchical cluster algorithm to track the characteristics of antibody candidates during antibody generation, which might reduce the number of false-positive and false-negative binding of antibodies. Moreover, 4 monoclonal antibodies that were produced successfully recognized their corresponding target antigens.

Introduction

The rapid development of proteomics research and technology has made it possible to identify thousands of proteins simultaneously, which means that a large amount of affinity molecules is required in the subsequent functional studies. ^1-3 Unfortunately, the creation of those affinity molecules is low throughput and costly. New ways to improve antibody generation such as phage display and multiplexed immunization techniques have attracted attention in recent years as they enable rapid generation of antibody fragments, protein scaffolds, or nucleic acid scaffolds. ^4-7 However, one common problem is that the number of antibody candidates produced is much higher than the traditional single-antigen/single-antibody generation approach. For example, a pilot study was performed by Schofield et al. ⁸ using a phage display technology in which thousands of recombinant antibodies were screened with enzyme-linked immunosorbent assay (ELISA), and sensitive detection was demonstrated using a bead-based system. The screening of hundreds of antibody clones and identification of their corresponding antigens through ELISA is cumbersome and time-consuming. Therefore, the development of new high-throughput antibody screening approaches is necessary to resolve this bottleneck in the antibody generation process.

Protein array technology that is high throughput, is highly sensitive, and requires a low amount of reagents is a promising solution. ^9-14 For example, De Masi et al. ¹⁵ decreased antibody screening time using multiplexed immunization and antigen microarrays. In their work, the antigen microarrays were fabricated by coating the aminosilane-modified slide with a layer of antigen. High-throughput screening (HTS) was performed by spotting hundreds of hybridoma supernatants on the surface of the antigen-coated slide, which significantly improved the efficiency of antibody screening. However, the specificity and cross-reactivity of an antibody cannot be not well characterized because only 1 antigen is used in the multiplexed assay. Furthermore, the long spotting process may impair protein activity.

To address these limitations, we developed a new antibody screening approach by using multiplexed immunization and immunogen array analysis. The difference from the approach by De Masi et al. ¹⁵ is that the immunogen array described here is based on a 96-well format where immunogens and negative and positive controls are spotted in each well. This type of multiplexing allows the specificity and cross-reactivity of antibody candidates to be characterized during the antibody generation. This is especially useful for HTS of hundreds of samples within a short time. Moreover, the amount of work saved is proportional to the number of immunogens. Finally, only minute amounts of immunogens and antibodies are employed. In this work, a model of 4 mixed immunogens immunization was employed. In total, 675 antibody candidates produced before and after established antibody hybridomas were screened in parallel with immunogen arrays and ELISA. The signal intensity and specificity as well as cross-reactivity of these antibody candidates were characterized using a hierarchical cluster algorithm.

Materials and Methods

Generation of immunogens

Three proteins (PRDX2, PRDX3, and PRDX4) sharing 72% amino acids and an unrelated ABCC2 were selected as immunogens ( Table 1 ). To avoid the selection of monoclonal antibodies (mAbs) against the protein purification tags, all selected proteins were expressed in 2 prokaryotic vectors with a HIS or GST tag that were used for immunization and screening, respectively. The cloning, expression, and affinity purification of the proteins used in this work were performed as previously described. ^16,17

Table 1.

Description of the Immunogens Used for Multiplexed Immunization

		Size of Immunogen
Immunogen	Full Name	aa	kDa	Expression System
ABCC2	Adenosine triphosphate binding cassette, subfamily C, member 2	180	20	Escherichia coli
PRDX2	Peroxiredoxin 2	192	21	E. coli
PRDX3	Peroxiredoxin 3	177	20	E. coli
PRDX4	Peroxiredoxin 4	169	19	E. coli

Immunization

BALB/c mice were immunized with 4 mixed immunogens (ABCC2, PRDX2, PRDX3, and PRDX4) with the HIS tag at the N-terminus. BALB/c mice were boosted at 14-day intervals with 20 µg of each protein, with a total of 80 µg each time. The serum titer levels were monitored by ELISA every 10 days after each boost to 1:8000 dilutions. Then the antibody hybridomas were generated by fusing the spleen cells with myeloma partner SP2/0 cells based on standard protocols. ⁴

Fabrication of immunogen arrays

To perform HTS of the produced antibody candidates, an immunogen array based on a 96-well format was developed using 6 aldehyde-modified slides (CEL Associates, Pearland, TX) assembled into a microarray holder. Each slide consisted of 3 × 6 arrays with the array spacing identical to that of a 96-well plate. This makes this technique compatible with multichannel pipettes or robotics. For each individual array, 4 immunogens (ABCC2, PRDX2, PRDX3, and PRDX4) with a GST tag were spotted in triplicate with the Smart Arrayer-48 microarrayer (CapitalBio, Beijing, China). Bovine serum albumin (BSA) and biotinylated goat antimouse IgG antibody (Sino-Am Biotech Co., Luoyang, China) served as negative and positive controls, respectively. The biotin groups on the surface of goat antimouse IgG could be recognized by streptavidin-Cy3 (Sigma, St. Louis, MO) to produce fluorescent signal. Humidity and temperature were maintained at 60% and 20°C throughout the spotting process. All prepared immunogen arrays were blocked in 10% BSA solution at 4°C overnight.

Detection and image analysis

Before screening, the immunogen arrays were rinsed with 1× PBST (phosphate-buffered saline with 0.5% Tween-20, pH 7.4) 3 times and ddH₂O once for 5 min per rinse. Then, 25 µL of culture supernatants containing hybridoma fusion cells was transferred to each array surface and incubated for 1 h at room temperature. The arrays were rinsed with PBST again following the procedure described above, and then the immunogen arrays were further incubated with biotinylated goat antimouse IgG solution (1:100) (Sino-Am Biotech Co.) and streptavidin-Cy3 solution (1:800) for 1 h, respectively. Finally, the immunogen arrays were rinsed in PBST 3 times and air-dried.

After the reaction, each slide was scanned with the Luxscan-10K/A microarray scanner (CapitalBio, Beijing, China) at 532 nm. The fluorescent image analysis was performed with Spot Data pro 3.0 software (CapitalBio). The local background intensity was subtracted from the median fluorescent intensity of each spot.

Hierarchical cluster analysis

The specificity and signal intensity of antibody candidates based on immunogen array screening was analyzed, and a heat map was drawn with MultiExperiment Viewer 4.1 software (www.tigr.org). ¹⁸ To perform hierarchical cluster analysis, the signal-to-noise (S/N) ratio was first calculated by dividing the signal of each antibody candidate by its negative control. Then the data were transformed by using the following equation:

$${\text{Y}}_{\text{i}}={\mathrm{Log}}_2\left({\text{X}}_{\text{i}}/\text{M}\right),$$

where X is the S/N ratio, i is the sequence of antibody candidate, and M is the median S/N ratio of all antibody candidates.

Western blot detection

Total protein concentrations were determined with a bicinchoninic acid (BCA) assay (Pierce, Rockford, Illinois) using BSA to create the standard linear curve. Then, 1 µg of recombinant protein with a HIS or GST tag was separated on a 10% reduced sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a nitrocellulose membrane. The membrane was then blocked with PBST containing 5% nonfat dry milk powder at 4°C overnight. The following morning, the membrane was incubated with 5 mL of culture supernatant containing the hybridomas at room temperature for 1 h followed by an addition of horseradish peroxidase (HRP)–conjugated antimouse IgG (Amersham Pharmacia Biotech, Buckinghamshire, UK). Immunoreactive blots were identified by using a chemiluminescence detection kit (GE Healthcare Biosciences, Piscataway, NJ).

Indirect immunofluorescence staining detection

Immunofluorecent staining of liver carcinoma cells (HepG2) was performed. Monolayer cells were washed with 1× PBS (pH 7.2) twice at room temperature and fixed with 1% Triton in 4% polyformaldehyde/PBS for 15 min. The fixed cells were washed 3 times with PBS and blocked with 1% BSA for 30 min and incubated with the primary antibody overnight at 4°C. The cells were then incubated with the secondary antibody, FITC-conjugated goat antimouse IgG, for 30 min at room temperature. Finally, the cells were incubated with Hoechst (Zhongshan Chemical Co., Beijing, China) in PBS (1:5000), and the slides were observed under fluorescence microscopy (Olympus, Tokyo, Japan).

Results And Discussion

Multiplexed immunization

With the development of hybridoma technology, cell fusion-based hybridoma technology has been the most important resource for the generation of mAbs. However, this traditional 1-immunogen/1-antibody approach has become the bottleneck in current proteomics research. ¹⁹ The development of a high-throughput antibody generation methodology will promote advancements in proteomics and associated biological research. One solution is to produce mAbs in a multiplex format, such that a single mouse can be immunized with several mixed immunogens so that their corresponding mAbs are produced at one time. The antibodies produced would specifically recognize their corresponding target or several targets. ^4,15,20 Although antibody generation has improved, the accompanying high-throughput antibody screening methods are still not well established. In this report, we describe an antibody screening approach based on an immunogen array and hierarchical cluster analysis. To demonstrate our HTS approach, we immunized mice with 4 immunogens. Moreover, to see whether the screened mAbs could tell the difference between proteins sharing similar sequences, 3 proteins (PRDX2, PRDX3, and PRDX4) sharing 72% amino acids and an unrelated ABCC2 were selected. Mice were immunized with these 4 mixed immunogens, and the hybridomas secreting antibodies were isolated. Cultures of all of the 96-well sterile plates plated with fusion cells were subjected to immunogen array and ELISA analyses in parallel. To prevent potential cross-reaction between antibody and tag epitope, the recombinant proteins were expressed with 2 expression tags such that the HIS tag protein was used for BAlb/C mice immunization, and the GST tag protein was used for antibody screening.

Screening of mAbs using immunogen array and hierarchical cluster analysis

Comparison of assay performance between immunogen array and ELISA

To demonstrate the ability of our immunogen array for the screening of antibody candidates, an albumin-antialbumin antibody pair was employed. Albumin was immobilized on aldehyde-modified slides and in ELISA plates, as well as incubated with mouse antialbumin antibody in parallel. Antialbumin antibody concentrations from 1 to 30 ng/mL were tested, as shown in Figure 1 . It is clear that there is a good correlation for the immunogen array and ELISA assay, with a correlation factor of 0.94. This result confirms the previous reports that the results of protein array and ELISA methods are comparable if using the same antibodies and antigens. ²¹ However, the signal of the immunogen array assay decreased faster than ELISA when the concentration of antialbumin antibody was lower than 4 ng/mL. This reveals that the ELISA assay might have higher sensitivity. But the sensitivity of immunogen array assays can be further improved using signal amplification methods, such as tyramide signal amplification or the planar waveguide technology. ²²

Fig. 1.

Comparison of the assay performance between immunogen array and enzyme-linked immunosorbent assay (ELISA). The concentration of antialbumin antibody tested is from 1 to 30 ng/mL.

Specificity and affinity analysis of antibody candidates

To screen the produced antibody candidates in high throughput, an immunogen array based on 96-well format was developed. A 5 × 5 array was fabricated in each well consisting of 4 GST tag recombinant proteins (ABCC2, PRDX2, PRDX3, and PRDX4) in triplicate, 7 negative controls (BSA), and 6 positive controls (biotinylated goat antimouse IgG; Fig. 2C ). Also, a 10 × 10 array can be fabricated in each well, which is able to detect up to 40 immunogens simultaneously. In this experiment, 675 culture hybridoma supernatants were screened. A representative fluorescent image is shown in Figure 2A . The antibody candidates show various signals to the immobilized immunogens, which are dependent on their affinity, specificity, cross-reactivity, and abundance. Figure 2B is an example in which the left antibody is highly specific to ABCC2 and the right antibody has strong cross-reactivity to all immunogens. To select positive antibody candidates, the median fluorescent intensity of negative controls plus 2 folds of standard deviation was used as the cutoff.

Fig. 2.

The fluorescent image of antibody screening using an immunogen array based on the 96-well format. (A) A representative image of antibody screening. (B) An antibody candidate with high-specificity and low cross-reactivity (left) and an antibody candidate with strong cross-reactivity (right). (C) Scheme of spotted protein array; 500 µg/mL of bovine serum albumin (BSA) and a 1:10 dilution of biotinylated goat antimouse IgG were used as negative and positive controls, respectively. a to d indicate the immunogens used in multiplexed immunization.

To characterize these antibodies, 22 candidates and their subclones produced before and after established hybridomas were selected to perform specificity and cross-reactivity analysis. The S/N of each antibody candidate was calculated, normalized, and used to draw a heat map. In Figure 3 , the row is the immunogens, and the column is the number of clones ( Fig. 3A ) and subclones ( Fig. 3B ). This bioinformatics tool permits us to characterize the signal intensity and specificity of the antibody candidates using a pseudo-color scheme from green to red that corresponds to the signal intensity from low to high. Figure 3A reveals that 7 antibody candidates (AF11, EA09, EC01, HF06, HH04, HA02, IE04) specifically recognize their corresponding proteins with an S/N ratio of cross-reactivity below 1. However, 3 antibody candidates (IG11, JE04, and KB07) recognize 3 common proteins (PRDX2, PRDX3, and PRDX4) with an S/N ratio greater than 17. These 3 proteins belong to 1 peroxiredoxin family of antioxidant enzymes, and they share 72% amino acids. Two antibody candidates (AB04 and BA04) recognized all 4 proteins with an S/N ratio greater than 4. The hybridoma of HA04 was not tested before established hybridomas, but it showed a quantify signal in previous ELISA assays with 4 mixed immunogens (data not shown), so it was included in the subclone screening.

Fig. 3.

The heat map shows the specificity and signal intensity of antibody candidates based on immunogen array analysis (A) before and (B) after established antibody hybridomas. The row is the immunogens, and the column is the number of (A) clones and corresponding (B) subclones. The rainbow from green to red corresponds to the fluorescent intensity from low to high. The gray color means that this antibody hybridoma was not detected. The data normalization is shown in the Materials and Methods section.

Moreover, a comparison of these antibody candidates and their subclones ( Fig. 3A, 3B ) was performed. HH04 has high specificity to PRDX3, whereas its subclone, HH041, shows strong cross-reactivity with PPRDX4. Compared to their subclones, the affinity of antibody candidates AB01 and BB01 is low, and they show weak cross-reactivity to PPDX3 with an S/N ratio from 2 to 6. However, their subclones, AB016 and BB013, have high specificity, and the S/N ratios of cross-reactivity were decreased to below 1. These results highlight the necessity of characterizing the antibody specificity and cross-reactivity during antibody generation to prevent the potential false-positive or false-negative mAbs being produced.

Validation of antibodies using ELISA, Western blot, and Immunofluorescence staining

To validate the antibodies produced, 16 candidates were examined using immunogen arrays and ELISA, 7 antibodies were evaluated via Western blot, and 4 antibodies were tested with immunofluorescence staining ( Fig. 4 ). The results show a correlation between immunogen array and ELISA, Western blot, and immunofluorescence staining of 91%, 89.3%, and 75%, respectively. Finally, 4 mAbs (UHA041, UHH041, UJE042, and UKB071) were obtained that could be used in Western blot analysis, and 3 antibodies (UHA041, UHH041, and UKB071) could be used in the immunofluorescence staining analysis ( Table 2 ). The test results of these antibodies are shown in Figure 5 . Figure 5A shows that UHA041 is specific to ABCC2 and UHH041 is specific to PRDX3 on Western blot. However, these data contradict the results of the immunogen array analysis, which showed that UHHP41 recognized both PDRX3 and PDRX4 antigens ( Fig. 4 ). It is possible that the microarray signal arose from conformational epitopes displayed on the array but were absent from the denatured immunoblots. The results obtained indicate that it is possible to produce mAbs for proteins sharing similar sequences by using a multiplexed immunization approach. In addition, UJE042 and UKB071 may be used for the detection of these 3 proteins (PDRX2, PDRX3, and PDRX4). Figure 5B shows that UHA041, UHH041, and UKB071 recognize their targets in HepG2 cells using immunofluorescence staining. All of these results show that multiplexed immunization and the immunogen array analysis approach can be used for high-throughput mAbs generation.

Fig. 4.

Validation of produced antibody candidates using enzyme-linked immunosorbent assay (ELISA), Western blot (WB), and immunofluorescence staining (IF) techniques. The red color is positive. The black color is negative. The gray color is no value, in which case the antibody candidate was not tested. IA, immunogen array.

Fig. 5.

Representative images of produced mAbs from (A) Western blot and (B) immunofluorescence staining analysis. The Western blot was performed using the recombinant proteins as antigens. The immunofluorescence staining of mAbs was performed in HepG2 cells, in which a mouse IgG was used as the negative control, and a mAb (AEB018) specific to human mitochondria protein was used as the positive control.

Table 2.

List of Antibodies Produced Using a Multiplexed Immunization and Immunogen Array Screening Approach

				Immunogen
Subclone	IA	IF	WB	ABCC2	PRDX2	PRDX3	PRDX4
UHA041	+	+	+	+	−	−	−
UHH041	+	+	+	−	−	+	+
UJE042	+	−	+	−	+	+	+
UKB071	+	+	+	−	+	+	+

WB, Western blot; IF, immunofluorescence staining; IA, immunogen array.

Conclusion

In this work, we developed an antibody screening approach based on immunogen array and bioinformatics analysis for high-throughput mAbs generation. The immunogen array based on the 96-well format is compatible with standard laboratory equipment, which enables the rapid screening of hundreds of samples. Furthermore, using hierarchical cluster analysis, the signal intensity and specificity as well as cross-reactivity of candidate antibodies corresponding to their immunogens can be simply characterized simultaneously. This enables us to track the mAb characteristics during the mAb generation process and might prevent the potential false-positive or false-negative antibodies being produced. These results show that this new antibody screening approach has promise to push antibody generation in high throughput for proteomics and related biological research.