A comparative study of antibodies against human CD33: ELISA, western blot, immunofluorescence, and immunoprecipitation

Abstract

In this study we selected by phage display several new antibodies against human CD33. We then tested them in a variety of immunodetection assays, in parallel with previously described recombinant antibodies as well as commercial antibodies. The reactivity of individual antibodies varied strongly depending on the method used for immunodetection. The results demonstrate that the recombinant antibodies achieve performance comparable to the animal-derived antibodies.

Introduction

The antibody isolation and characterization described in this paper, and in the complementary paper on antibodies against human PD-1 (Dhungana et al., 2026), were undertaken to develop two fully recombinant antibodies and to evaluate their performance in comparison with established antibodies against proteins of broad biomedical relevance, contributing to ongoing efforts to assess new/alternative antibody-production technologies. Human CD33 (Siglec-3; Uniprot #P20138) is a transmembrane sialic acid-binding lectin primarily expressed on myeloid cells. It acts as an inhibitory immune checkpoint, regulating innate immune responses (McMillan and Crocker, 2008). CD33 is a key diagnostic marker for acute myeloid leukemia, making it a target for antibody-drug conjugates (Sievers et al., 1999). It is also a potential therapeutic target for Alzheimer’s disease due to its role in regulating microglial amyloid-β clearance (Zhao, 2019). Several antibodies have been reported to bind specifically CD33. Here we describe several new anti-CD33 antibodies and compare their reactivity with previously discovered anti-CD33 antibodies by ELISA, western blot, immunofluorescence and immunoprecipitation.

Materials & Methods

Antigen: For phage display selection and ELISA detection, we used a fusion protein comprising a human IL-2 signal peptide, the extracellular domain of human CD33 (amino acids 17–259) lacking its native signal peptide, and a C-terminal Twin-Strep-Tag® (IBA Lifesciences; CD33-TST). The fusion protein was produced and secreted in transiently transfected HEK293 cells and purified from cellular supernatants using MagStrep Strep-Tactin XT beads according to the manufacturer’s instructions (IBA Lifesciences #2-5090-002). For other immunodetection methods, we used HeLa cells expressing transiently a fusion protein composed of the human IL-2 signal sequence for insertion in the ER followed by the coding sequence of the human CD33 extracellular domain and fused at its C-terminus with the IL-2-RA transmembrane and cytosolic domain followed by a C-terminal HA Tag, as previously described (Marchetti et al., 2023). The resulting membrane-embedded CD33 (CD33-TM-HA) is expected to be present at the cell surface as well as in the membrane of endocytic and exocytic intracellular compartments.

Antibodies: Two CD33-binding VHH domains (ABCD_RC236 and RC241) (ABCD nomenclature, http://web.expasy.org/abcd/) as well as three CD33-binding human scFv (ABCD_RC223, ABCD_RC224 and ABCD_RC225) were selected by phage display using the CD33-TST protein adsorbed to MagStrep Strep-Tactin XT beads. ABCD_AH713, ABCD_AK555 and ABCD_AK556 (therapeutic anti-CD33) and ABCD_AF291 (anti-HA) were described previously (Sievers et al., 1999; Caron et al., 1992; Keszei & Picard, 2019). These antibodies will be referred to RC236, RC241, RC223, RC224, RC225, AH713, AK555, AK556 and AF291. To produce the recombinant antibodies described above, HEK293 suspension cells growing in HEK TF medium (Xell #861-0001, Sartorius), supplemented with 0.1% Pluronic F68 (Sigma #P1300), were transiently transfected with the vector coding for the antigen-binding scFv or VHH domain of each antibody fused to a rabbit Fc domain. The synthesized scFv sequences (GeneArt, Invitrogen) correspond to the sequences of the variable regions joined by a peptide linker (GGGGS)₃. Supernatants containing secreted antibodies (~50 mg/L) were collected after 4 days of culture. Anti-CD33 antibodies were also acquired from AbCam (rabbit IgG ab269456 and mouse IgG1 ab30371), and Thermofisher (rabbit IgG MA5-48347 and 945-RBM12-P1ABX). These antibodies are referred to here as 269456, 30371, 48347 and RBM12.

ELISA: The whole procedure was carried out at room temperature. Purified CD33-TST was immobilized on MaxiSorp ELISA plates (Thermofisher #439454) for 30 min. Each well was rinsed three times with 100 μL of washing buffer (WB: PBS + 0.5% (w/v) BSA + 0.05% (w/v) Tween20), then incubated for 1 hour with 50 µl WB containing the indicated antibody (1, 0.1 or 0.01 µg/mL) (Fig. 1). After rinsing 3 times (100 µl WB), wells were incubated with horseradish peroxidase-coupled goat anti-rabbit IgG (Sigma #A8275, 1:1000, 50 μL per well) for 30 min. After 3 rinses, Tetramethylbenzidine (TMB) substrate (Sigma #T5569) was added (50 μL per well). The reaction was stopped by the addition of 25 μL of 2 M H₂SO₄. The absorbance (OD) was measured at 450 nm, and the absorbance at 570 nm was subtracted.

Western blot: HeLa cells (5 x 10⁶) expressing the CD33-TM-HA fusion protein, or not, were lysed in 400 µL Lysis buffer (PBS + 0.5% TX100 + protease inhibitors (20 µg/mL leupeptin, 20 µg/mL aprotinin, 18 µg/mL PMSF, 1.8mg/mL iodoacetamide) for 15 min at 4°C, then centrifuged at 10’000 g to remove unlysed material. The cleared cell lysate was then mixed to an equivalent volume of 4x-concentrated non-reducing Sample buffer (20.6% (w/v) sucrose, 100 mM Tris pH 6.8, 10 mM EDTA, 0.1% (w/v) bromophenol blue, 4% (w/v) SDS, no -mercaptoethanol). The proteins were loaded on an SDS-PAGE (10 µL per lane), separated by electrophoresis (200 V, 30 min) on a 4-15% acrylamide gel (Mini-PROTEAN® TGX™ Precast Gel, Bio-Rad #456-1086), and transferred to a nitrocellulose membrane using a dry transfer system for 10 min (iBlot gel transfer device, Invitrogen #IB1001EU). The membranes were blocked overnight at 4°C in PBS containing 0.1% (v/v) Tween20 and 7% (w/v) milk, then washed three times for 15 min in PBS + 0.1% (v/v) Tween20. The membranes were incubated with the indicated anti-CD33 antibody (1 µg/mL in PBS-Tween) overnight at 4°C, then washed three times for 15 minutes. As a control, the anti-HA antibody was used. The membranes were then incubated 1 h with horseradish peroxidase-coupled goat anti-rabbit IgG (Novex, dilution 1:3000) and washed 5 times for 5 min in PBS-Tween. The signal was revealed by enhanced chemiluminescence (ECL) (Amersham Biosciences) using a PXi-4 gel imaging system (Syngene).

Immunoprecipitation: To assess the ability of antibodies to immunoprecipitate the CD33 protein, individual antibodies (1 µg) were incubated with 50 µL Protein G-Sepharose (GE Healthcare 17-0618-01) for 1 h at room temperature. The Sepharose resin with the bound antibodies was washed 3 times with 1mL PBS, then incubated for 1 h at 4°C with a cleared lysate of 10⁶ HeLa cells transiently transfected to express the CD33-TM-HA fusion protein, obtained as described above by lysing 10⁷ cells in 10 mL Lysis buffer. The resin was then washed 5 times at 4°C with PBS+0.1% TX-100, one time with PBS, then resuspended in reducing Sample buffer (containing 6% (v/v) -mercaptoethanol), migrated on an SDS-PAGE gel and transferred to a nitrocellulose membrane. The CD33-TM-HA protein was revealed using the AF291 (anti-HA) antibody fused to a mouse Fc and a secondary HRP-coupled anti-mouse Fc antibody (Novex, dilution 1:3000). For immunoprecipitation with the mouse ab30371 antibody, the CD33-TM-HA protein was revealed using the AF291 antibody fused to a rabbit Fc and a secondary HRP-coupled anti-rabbit Fc antibody (Novex, dilution 1:3000).

Immunofluorescence: The whole procedure was carried out at room temperature. Transfected HeLa cells expressing transiently the CD33-TM-HA fusion protein were rinsed once with PBS, fixed with PBS + 4% paraformaldehyde (w/v) (Applichem, #A3013) for 30 min, and blocked with PBS + 40 mM ammonium chloride (NH₄Cl) (Applichem, #A3661) for 5 min. Cells were then permeabilized in PBS + 0.2% saponin (w/v) (Sigma, #S7900) for 10 min, washed once (5 min) with PBS + 0.2% (w/v) BSA (PBS-BSA), and incubated for 30 min with the rabbit anti-CD33 antibodies (1 µg/mL) and a mouse AF291 anti-HA. After 3 washes (5 min) with PBS-BSA, cells were incubated for 30 min in PBS-BSA with a secondary anti-rabbit IgG conjugated to AlexaFluor-488 (1:400, Molecular Probes, #A11029) and with an AlexaFluor-647 conjugated anti-mouse IgG (1:400, Molecular Probes, #A11029). After 3 washes (5 min) with PBS-BSA, cells were mounted on slides (Menzel-Gläser, 76x26 mm) with Möwiol (Hoechst) + 2.5% (w/v) DABCO (Fluka, #33480). Pictures were taken with identical settings using a Zeiss LSM700 confocal microscope, with a 63x Neofluar oil immersion objective.

Results

When tested by ELISA, all antibodies bound CD33 immobilized on ELISA plates in a concentration-dependent manner, with the exception of 269456 (Fig. 1). RC223, RC236 and RC241 showed a less efficient binding at low concentrations. All other antibodies bound CD33 with similarly high efficiency (Fig. 1).

Figure 1. Specific binding of antibodies to CD33 as detected by ELISA. Therapeutic antibodies (A) newly identified anti-CD33 antibodies (B) and commercial antibodies (C) bound CD33 with variable efficiencies.

We next tested the ability of the same panel of antibodies to bind and immunoprecipitate an HA-tagged transmembrane form of CD33 (CD33-TM-HA). We used a lysate of HeLa cells expressing CD33-TM-HA. Each antibody was immobilized on Protein G-Sepharose beads, then incubated with a cell lysate containing CD33-TM-HA. After washing the beads, the CD33-TM-HA protein bound to the antibody was migrated on an SDS-PAGE gel, transferred to nitrocellulose and detected using a rabbit anti-HA antibody (Fig. 2). As expected, CD33-TM-HA was efficiently immunoprecipitated by an anti-HA antibody. It appeared as a mixture of a ≈50 kDa monomer (star) and ≈100 kDa dimer (arrowhead). The three therapeutic anti-CD33 antibodies (AH713, AK555 and AK556) immunoprecipitated CD33 efficiently. The 48347, RBM12, 30371, RC224 and RC225 also immunoprecipitated CD33, albeit less efficiently (Fig. 2).

Figure 2. Immunoprecipitation of CD33-TM-HA. CD33-TM-HA was immunoprecipitated from a lysate of HeLa cells expressing CD33-TM-HA (CD33), or not (CTRL). The immunoprecipitated protein migrated mostly as a mixture of a monomer (star) and a dimer (arrowhead) on an SDS-PAGE gel and was revealed by western blot using an anti-HA antibody (mouse AF291).

To test the ability of the same panel of antibodies to recognize CD33 by immunofluorescence, we used HeLa cells expressing CD33-TM-HA. Cells were fixed, permeabilized and incubated with individual anti-CD33 antibodies, and then with a fluorescent secondary antibody (Fig. 3; CD33; green). In order to identify transfected cells, the cells were also labelled with an anti-HA antibody (Fig. 3; Tag HA; red). These two antibodies suggest an additional localization for CD33 not mirrored by anti-HA: anti-CD33 shows a more punctate and widespread pattern, compared to the more peripheral plasma membrane-oriented localisation revealed by anti-HA. This may be due to cleavage of the HA tag, and/or more rapid degradation of the HA tag compared to the CD33 protein intracellularly. An irrelevant antibody (AX047) failed to generate any signal (Fig. 3A). On the contrary, all antibodies tested recognized efficiently CD33, with the exception of RC223 (very weak signal) and 269456 (no signal). Finally, we tested the ability of each antibody to recognize CD33 by western blot. For this, lysates from cells expressing the CD33-TM-HA fusion protein were migrated on an SDS-PAGE gel, transferred to nitrocellulose, and incubated with the indicated antibodies (Fig. 4). The control antibody AF291 (anti-HA) recognized efficiently CD33-TM-HA, as expected. Among the tested anti-CD33 antibodies, only RBM12 recognized efficiently the fusion protein. AK555, AK556, AH713, 48347 and 30371 also detected the fusion protein, but much less efficiently. Other antibodies tested did not generate any specific signal.

Figure 3. Staining of CD33-TM-HA by double immunofluorescence. HeLa cells expressing a CD33-TM-HA fusion protein were stained with an antibody against CD33 (CD33; green) and an antibody against the HA tag (Tag HA; red). Bar: 20 µm.

Figure 4. Detection of the CD33 protein by western blot. Cell lysates containing the CD33-TM-HA fusion protein (CD33) or not (Non-transfected, NT) were migrated on an SDS-PAGE gel, transferred to nitrocellulose and revealed with the indicated anti-CD33 antibody. An anti-HA antibody (AF291) was used as a positive control.

Discussion

In this study we isolated and characterized new recombinant antibodies against human CD33 and analyzed their potential to recognize CD33 in different experimental settings. In parallel we tested three previously described therapeutic antibodies as well as four commercial antibodies. As summarized in Fig. 5, different patterns of reactivity were seen for each antibody analyzed. With the exception of 269456, all antibodies tested recognized CD33 in at least two experimental settings, but with varying efficacies. No single antibody was the most reactive in all conditions tested. Therapeutic antibodies were as expected excellent binders of the protein in native conditions (ELISA, immunofluorescence, immunoprecipitation) but hardly bound to the denatured protein (western blot). This superior detection is likely due to their screening in production for higher native affinity, slower dissociation rates, more robust and stable nature, and purer formulations. However, a large panel of antibodies were seen to recognize the CD33 protein in ELISA and immunofluorescence with relatively small differences in the observed signals. From a practical point of view, our results stress strongly the need to choose adequate antibodies and to verify their reactivity every time a different experimental setting is considered. The recombinant antibodies exhibited performance comparable to that of the animal-derived antibodies. The use of sequence-defined recombinant antibodies may also improve experimental reproducibility and transparency, addressing longstanding challenges in biomedical research (Modi, 2025; Barroso et al., 2020). To this end, we review recommendations that have been made over recent years in a complementary Commentary article, which should accelerate the greater adoption and wider use in research of recombinant antibodies in place of animal-derived antibodies. All stakeholders in research have a role to play to reduce research waste, increase replicability and human relevance, and take a more ethical/Three Rs-compliant stance by embracing such initiatives (Bailey et al., 2026).

Figure 5. Summary of anti-CD33 reactivity in different experimental conditions. *: contrary to other antibodies, the 30371 antibody is a mouse antibody, and required use of different secondary reagents. Hence its reactivity can only be compared qualitatively with other antibodies.

Conflict of interest

MD is supported by funding from ABCD Antibodies SA

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.