Design and production of an asymmetric anti-CD20 antibody (SBU-CD20)
CDC of an antibody is affected by the isotype and structural features of the IgG molecule [21, 22]. To examine how breaking the mirror symmetry of an IgG antibody changes its CDC for tumor cell clearance, we designed an asymmetric heterodimeric antibody that we named the Specific Bifunctional Unit (SBU). For efficient heterodimerization, CDC effector function, and prolonged circulating half-life, the SBU (scFv-FcKnob–scFab-FcHole) consists of an Fc region with a Knob mutation (T366W) and hole mutations (T366S, L368A, and Y407) [37] in each Fc polypeptide. In addition, the antigen-binding region was composed of a single-chain variable fragment (scFv) and a single-chain fragment of antigen binding (scFab) to remove undesired side products, such as Knob-Knob homodimers (scFv-FcKnob–scFv-FcKnob) and Knob monomers (scFv–FcKnob) using the IgG Ck region-binding KappaSelect resin [38, 39], while preventing mispairing between the H-chain and L-chain of the antibody (Fig. 1A).
After expressing SBU-CD20 (scFv-FcKnob–scFab-FcHole) in Expi293F cells and performing simple, single-step affinity chromatography with KappaSelect resin, we obtained highly purified SBU-CD20. SEC and RP-HPLC analyses showed that the purity of SBU-CD20 was similar to that of symmetric anti-CD20 IgG antibodies, a commercial anti-CD20 IgG antibody (rituximab: Rituxan®), and an IgG antibody (rituximab-KiH) containing a Knob mutation (T366W) and hole mutations (T366S, L368A, and Y407), which were prepared using the same steps (100% and 99.85% for rituximab, 100% and 99.86% for rituximab, and 98.92% and 97.18% for SBU-CD20 in the SEC and RP-HPLC analyses, respectively) (Supplementary Fig. 2). Interestingly, no band corresponding to undesired monomeric scFab–FcHole was detected in the SBU-CD20 sample purified with KappaSelect resin (Fig. 1C; Supplementary Fig. 1A), suggesting that almost all of the expressed monomeric scFab–FcHole molecules might have assembled into heterodimers with scFv–FcKnob to form SBU-CD20 because of the low homodimerization tendency of scFab–FcHole, which might have a slightly lower expression level than scFv–FcKnob. In contrast, when SBU-CD20 was purified using the Fc region–binding Protein A resin instead of the KappaSelect resin, undesired side products, such as Knob-Knob homodimers (scFv-FcKnob–scFv-FcKnob) and Knob monomers (scFv–FcKnob) were detected in the eluent (Supplementary Fig. 1A), and the purity of the SBU-CD20 antibody decreased to 84% in the RP-HPLC analysis (Supplementary Fig. 1B).
SBU-CD20 with broken mirror-symmetry elicits higher CDC activity than anti-CD20 antibodies with symmetric Fab arms (rituximab or rituximab-KiH)
Before comparing the CDC activities of the three anti-CD20 antibodies (SBU-CD20, rituximab, and rituximab-KiH), we analyzed their binding to the antigen and C1q. Since rituximab is known to bind to the large loop region (168–EPANPSEK–175) of CD20 on the surface of B cells [40], we fused a peptide from the 163rd to the 187th amino acids to streptavidin and prepared mammalian cells (Supplementary Fig. 3). In ELISA analysis using the purified streptavidin-fused CD20 epitope peptide, SBU-CD20 exhibited a binding affinity similar to that of rituximab and rituximab-KiH (Supplementary Fig. 4A). In addition, SBU-CD20 showed C1q binding affinity that was almost identical to that of rituximab and rituximab-KiH (Supplementary Fig. 4B), indicating that the Knob-into-Hole mutations (T366W/T366S, L368A, and Y407) introduced into Fc for heterodimeric formation and the linkers used in scFv/scFab had a negligible effect on binding to the antigen (CD20) and C1q.
On the other hand, SBU-CD20 exhibited approximately three times higher production of C4d molecules (intermediate products in the complement activation cascade) than rituximab (3.2 µg/ml vs. 1.1 µg/ml) (Fig. 1D). Next, we examined the CDC of the prepared anti-CD20 antibodies using three lymphoma B-cell lines (Ramos, WSU-NHL, and BJAB). As shown in Fig. 1E, treatment with human complement sera alone produced 11–21% lysis of lymphoma B cells, and samples treated with human complement sera and rituximab at a concentration of 20 µg/ml exhibited 47%, 38%, and 16% lysis of Ramos, WSU-NHL, and BJAB cells, respectively. These results are consistent with the report by Golay et al., which indicated that BJAB cells were rituximab-resistant to complement-mediated lysis [36]. When treated with rituximab-KiH under the same conditions, the activity of CDC against the three CD20-expressing target cell lines (Ramos, WSU-NHL, and BJAB) was almost identical to that of rituximab (40%, 32%, and 20%, respectively). In sharp contrast, SBU-CD20 with an asymmetric structure exhibited CDC of 80%, 75%, and 30%, respectively, to Ramos, WSU-NHL, and BJAB, indicating dramatically improved tumor cell clearance activity.
Increased Cd55 Expression On The Surfaces Of Rituximab-resistant Cells
Motivated by the results that i) antibodies with an asymmetric structure could have dramatically increased tumor cell killing effects via CDC effector function and ii) the heterodimeric SBU-CD20 antibody could be efficiently produced by simple expression and purification, we attempted to improve the low tumor cell-killing potency of existing anticancer antibodies by developing an asymmetric bispecific antibody with new antigen-binding capability and increased CDC that bypassed tumor resistance. First, to find an effective target antigen to combine with CD20 for the clearance of rituximab-resistant tumors, we constructed in-house rituximab-resistant cells (Ramos-RR) by continuously treating a rituximab-sensitive cell line (Ramos) with rituximab. As expected, rituximab treatment in the presence of complements induced cell lysis in 46.98% of Ramos cells but only 16.91% of Ramos-RR cells (Fig. 2). Next, we focused on CD55, a membrane-bound complement regulatory protein, in tumor cells that inhibits CDC (Supplementary Fig. 5). We analyzed the changes in CD55 and CD20 expression resulting from the acquisition of rituximab resistance by examining Ramos and Ramos-RR cells. When the expression levels of CD20 and CD55 on the surfaces of tumor cells were analyzed by FACS using anti-CD20 IgG antibody (rituximab) and anti-CD55 IgG antibody (4-1H) [41], we found that CD20 expression was significantly lower in Ramos-RR cells (MFIRamos, CD20 = 413.5; MFIRamos−RR, CD20 = 278.0). In contrast, CD55 expression was markedly increased in rituximab-resistant cells (MFIRamos, CD55 = 605.5; MFIRamos−RR, CD55 = 1 409.5). These results indicate that CD55 overexpression is highly correlated with rituximab resistance, suggesting that the CDC of anti-CD20 antibodies could be improved by simultaneously targeting CD55.
Sbu-cd55×cd20 Exhibits Excellent Physicochemical Properties And Developability
In our previous study, we discovered a monoclonal antibody against human CD55 (4-1H: chimeric anti-CD55) using a naïve chicken phage-display scFv library. When an antibody exhibiting high binding affinity to CD55 (EC50: 0.22 nM) was conjugated with 177Lu, the radioisotope-labeled anti-CD55 antibody significantly reduced tumor growth and increased the median survival time in a pleural metastatic lung cancer mouse model, demonstrating its potential as a therapeutic agent [41]. To confirm that the anti-CD55 antibody (4-1H) could be effectively introduced into our asymmetric bispecific antibody (SBU) for simultaneous targeting of both CD55 and a tumor-associated antigen (e.g., CD20, HER2, or EGFR), the expression plasmid for CD55-scFv-FcKnob was co-transfected into Expi293F cells with a plasmid encoding single-chain Fabs (CD20-scFab-FcHole, HER2-scFab-FcHole, or EGFR-scFab-FcHole derived from the variable regions of rituximab, trastuzumab, or cetuximab, respectively). Using the same expression and purification steps described for the production of the SBU-CD20 antibody, we obtained highly purified asymmetric bispecific antibodies (SBU-CD55×CD20, SBU-CD55×HER2, and SBU-CD55×EGFR) that simultaneously bound two different antigens (Fig. 3C; Supplementary Fig. 6A–E). In particular, SBU-CD55×CD20, which has a monovalent CD55 binding site, showed an apparent binding affinity similar to that of the 4-1H IgG antibody with bivalent CD55-binding paratopes (Fig. 3E). On the other hand, SBU-CD55×CD20 with a monovalent CD20 binding site showed a slightly lower apparent CD20 antigen binding ELISA signal than rituximab, which has bivalent antigen-binding capability, due to decreased avidity (Fig. 3D).
SBU-CD55×CD20, which is smaller than a natural IgG antibody (~ 125 kDa vs. ~150 kDa), has flexible linkers and Knob-into-Hole mutations in the antigen-binding and Fc regions, respectively. To examine the effects of these structural differences from native IgG molecules on the function of the Fc domain, we analyzed the binding affinity of SBU-CD55×CD20 to the Fc binding ligands human FcγRs, human FcRn, and human C1q. ELISA results indicated that SBU-CD55×CD20 exhibited binding to all human FcγRs and pH-dependent binding to human FcRn, almost identical to that of rituximab and 4-1H (Supplementary Fig. 7). In terms of C1q binding, SBU-CD55×CD20 showed slightly lower ELISA binding signals than rituximab and 4-1H (Fig. 4A). Because heterogeneous glycosylation profiles may cause differences in the structure and function of antibodies, we conducted N-linked glycan profile analysis using mass spectrometry. The results revealed that the glycan profile of the SBU-CD55xCD20 antibody was highly similar to that of rituximab in terms of the main glycan forms such as GOF, G1F, and G2F (Supplementary Fig. 8A). In a thermostability analysis using a DSC instrument, SBU-CD55×CD20 showed a high melting temperature (Tm1 = 68.77°C) that was comparable to that of rituximab IgG (Tm1 = 72.61°C) (Supplementary Fig. 8B). Taken together, the results of physicochemical analyses suggest that the SBU-CD55×CD20 antibody has excellent developability as a therapeutic bispecific antibody.
SBU-CD55×CD20 elicits higher CDC than an anti-CD20 IgG antibody, anti-CD55 IgG antibody, or the combination of the two monospecific IgG antibodies
Although rituximab has been used to treat non-Hodgkin’s lymphoma and chronic lymphocytic leukemia since its FDA approval in 1997, relapsed or refractory cancers resistant to rituximab have been reported [42, 43]. When resistance develops and recurs, the available drugs are very limited; therefore, new treatments are needed to overcome the limitations of existing treatments.
CD55 inhibits CDC by accelerating C3 convertase degradation [44]. To analyze the CDC of SBU-CD55×CD20, which has an asymmetric structure capable of binding to CD55 and CD20 simultaneously, we used two rituximab-sensitive cell lines (Ramos and WSU-NHL) and two rituximab-resistant cell lines (in-house developed Ramos-RR and BJAB). Rituximab-sensitive cell lines (Ramos and WSU-NHL) showed CDC of 47% and 38%, respectively, and 4-1H (anti-CD55 antibody) exhibited CDC-induced tumor cell clearance of 10% and 21%, respectively. The combination of rituximab and 4-1H improved CDC to 60% and 54.72% in Ramos and WSU-NHL cells, respectively. In sharp contrast, the SBU-CD55×CD20 antibody exhibited dramatically improved tumor cell clearance of 91.42% and 83.21% in the Ramos and WSU-NHL cells, respectively (Fig. 4C, D). In the rituximab-resistant cell lines (Ramos-RR and BJAB cells), the CDC of the individual monospecific antibodies (rituximab or 4-1H antibody) was very low (8.89–16.91%), and the combination of the two monospecific antibodies (rituximab and 4-1H) also had low tumor cell-killing activity of 22.24% and 18.34% in Ramos-RR and BJAB cells, respectively. In sharp contrast, the SBU-CD55×CD20 antibody exhibited a remarkably increased CDC of 83% and 50.49% in Ramos-RR and BJAB, respectively (Fig. 4E, F), indicating its superior tumor cell killing effect in resistant tumor cell lines.
Comparison Of Cdc Between The Sbu And Bispecific Antibodies With Different Forms
Several different forms of bispecific antibodies have been developed to simultaneously control two different antigens [6]. We prepared different forms of bispecific antibodies and compared their efficacies with that of SBU-CD55×CD20. Unlike SBU-CD55×CD20 (scFv-FcKnob–scFab-FcHole), where the CD55 and CD20 antigen-binding sites of the antibody were scFv and scFab, respectively, we produced two different forms of bispecific antibodies: 1) scFv-Fc-CD55×CD20 (scFv-FcKnob–scFv-FcHole), comprising two scFvs against CD55 and CD20, respectively, and an Fc region with a Knob mutation (T366W) and hole mutations (T366S, L368A, and Y407), and 2) IgG-scFv-CD55×CD20, which had an anti-CD55 scFv fused to the C-terminus of the heavy chain of rituximab IgG (Fig. 5A). When the three forms of bispecific antibodies (SBU-CD55×CD20, scFv-Fc-CD55×CD20, and IgG-scFv-CD55×CD20) were expressed in Expi293F cells, purified with KappaSelect resin or Protein A resin, and analyzed by SDS-PAGE, bands corresponding to the expected molecular weights of the three formats of bispecific antibodies (rituximab IgG,150 kDa; SBU-CD55×CD20,125 kDa; scFv-Fc-CD55×CD20,100 kDa; IgG-scFv-CD55×CD20,200 kDa) were detected (Fig. 5B). In contrast to the IgG-scFv-CD55×CD20, which displayed two bands (~ 200 kDa), and the scFv-Fc-CD55×CD20, which displayed an impurity band (~ 50 kDa) in the SDS-PAGE analysis, possibly due to glycosylation heterogeneity and non-heterodimerized products (scFv-Fc), respectively, we confirmed that SBU-CD55×CD20 was clearly purified (Fig. 5B).
The CDC of rituximab IgG and three fo rms of anti-CD55×CD20 bispecific antibodies were then analyzed in the two rituximab-resistant cell lines (Fig. 5C, D). Compared with the CDC of rituximab, IgG-scFv-CD55×CD20 showed 0.72- and 0.75-fold CDC in BJAB and Ramos-RR cells, respectively, indicating lowered tumor cell killing activity; scFv-bsAb showed 1.27- and 1.15-fold CDC in BJAB and Ramos-RR cells, respectively, indicating similar or slightly enhanced CDC. In contrast, SBU-CD55×CD20 showed 2.7- and 3.8-fold CDC in BJAB and Ramos-RR cells, respectively, which was a significant improvement that led to tumor cell lysis of 50% and 76%, respectively. Taken together, these results show that CDC is highly variable depending on the form of the bispecific antibody, and the asymmetric bispecific antibody (SBU) demonstrated efficacy that was highly superior to that of the other forms in tumor cell killing, as well as having advantages from a bioprocessing standpoint.