Background
Testing candidate monoclonal antibody therapeutics in preclinical models is a critical step in drug development. Identifying antibody therapeutic candidates that bind to their human targets and cross-react with their mouse homologs is often challenging, particularly for targets with low sequence homology. In such cases, surrogate antibodies that bind to the mouse homolog are necessary. 9D9, an antibody that binds to mouse CTLA-4, is a commonly used surrogate for CTLA-4 checkpoint blockade studies in mouse cancer models.
On January 24, 2025, Brett Robison's team published an article titled "Engineered ipilimumab variants that bind human and mouse CTLA-4" in MAbs. In this work, phage and yeast display technologies were used to design ipilimumab variants that bind to mCTLA-4. The engineered variants exhibited pH-independent cross-reactivity with both mouse and human CTLA-4. The crystal structure of a variant in complex with mouse and human CTLA-4 confirmed that it targets the same epitope as ipilimumab. These cross-reactive ipilimumab variants may help improve translatability and facilitate future studies of the mechanism of action of anti-CTLA-4 targeting in mouse models.
9D9 and ipi have different affinities and epitopes for their respective CTLA-4 targets
9D9 exhibits strong pH-dependent binding to mCTLA-4, with binding weakening with decreasing pH. 9D9 exhibits poor binding to mCTLA-4 in the lower pH environment of lysosomes and in the tumor microenvironment (TME) of murine cancer models. Ipilimumab's binding to hCTLA-4 is unaffected by pH within the pH range of 6.0 to 7.4, demonstrating pH independence. The crystal structure solved at 3.1 Å resolution reveals that 9D9 targets a membrane-distal epitope on mCTLA-4, primarily composed of the BC and FG loops of mCTLA-4, with additional contacts observed within the C'C'' loop and within the A strand near the N-terminus. The epitope bound by ipilimumab is primarily located on the C, G, and F strands of hCTLA-4, similar to the binding epitope of tremelimumab. This epitope is distinct from the binding site of the B7 ligand, suggesting that ipilimumab exerts its effects by blocking the binding of CTLA-4 to the B7 ligand.
Construction of an iPi that binds to mCTLA-4 using phage and yeast display technologies
The mCTLA-4-binding iPi was constructed by a two-step method using phage and yeast display systems. First, a structure-guided phage library of the ipilimumab single-chain variable fragment (scFv) was constructed. Sanger sequencing of individual enriched clones identified mutations that increased affinity for mCTLA-4. These mutations were then used for subsequent yeast display optimization.
By constructing two yeast surface display libraries (one targeting the complementarity-determining region (CDR) of the heavy and light chains, respectively), these libraries were designed to contain a single degenerate NNK codon at each CDR position, so that each library member contained only one mutation. Variants with higher affinity for mCTLA-4 were ultimately obtained by fluorescence-activated cell sorting (FACS) selection.
Binding characterization of engineered mipi variants
Engineered mipi variants (especially mipi.4) successfully achieved high-affinity binding to hCTLA-4 and mCTLA-4, and maintained stable binding ability under different pH conditions.
As bivalent IgGs, the mipi variants showed improved binding to hCTLA-4 cells compared to the parental ipi at pH 6.0 and 7.4. As Fabs, the mipi variants bound to hCTLA-4 cells with much higher affinity than the ipi at both pH 6.0 and 7.4. As IgGs, the mipi variants bound to mCTLA-4 cells with greater than or equivalent affinity. This indicates that the engineered mipi variants bind tightly to mCTLA-4 and also have improved binding to hCTLA-4.
Blockade of CTLA-4 enabled CD28-dependent co-stimulation, as reflected by higher luciferase expression. Compared to ipi, mipi variants in IgG and Fab formats exhibited more potent, dose-dependent stimulation, consistent with their binding activity.
Crystal structures of the mipi.4–hCTLA-4 and mipi.4–mCTLA-4 complexes
The crystal structures of mipi.4 Fab in complex with hCTLA-4 and mCTLA-4 were determined at resolutions of 1.64 Å and 1.57 Å, respectively. Superposition of these complexes aligned on hCTLA-4 revealed a 9° shift in the Fv region of mipi.4 compared to the parental antibody, ipilimumab. Despite this shift, the Fv domains of mipi.4 and ipilimumab display a high degree of structural similarity (Cα RMSD of 0.35 Å) and recognize similar epitopes on hCTLA-4.
Both antibodies use six CDRs to bind to hCTLA-4, with the heavy and light chains contributing 60% and 40% of the Fab's buried surface area, respectively. Mipi.4 has a larger buried surface area on hCTLA-4 (approximately 836 Ų) than ipilimumab (763 Ų).
Mipi.4 and ipilimumab adopt similar conformations in terms of main-chain and side-chain rotations of the CDR loops. The LCDR1 of mipi.4 adopts a different conformation from that of ipilimumab (only the LCDR1 alignment Cα RMSD is 1.93 Å), which may be caused by the S30W and Y32P substitutions.
These changes shift the LCDR1 contact residues from S31 and Y32 in ipilimumab to V28, G29, and S30W in mipi.4, and increase the buried surface area of the light chain from 260 Ų to 337 Ų.
Mipi.4's hCDR1 and HCDR2 mutations introduce new polar interactions at the hCTLA-4 binding interface. The hydrogen bond between N56 in ipilimumab's HCDR2 and E33 of hCTLA-4 is replaced by a salt bridge involving the mutated residue N55R and a hydrogen bond to N56H in mipi.4. Furthermore, the S31K substitution in mipi.4's HCDR1 forms a salt bridge with E48 of hCTLA-4. These acquired electrostatic interactions provide a structural rationale for the cross-reactive binding of the engineered mipi.4 variant to mCTLA-4 while also enhancing its binding to hCTLA-4.
Summarize
Through engineering approaches, researchers have successfully developed ipilimumab variants capable of binding both human and mouse CTLA-4, specifically the mipi.4 variant. mipi.4 maintains high affinity across pH conditions and binds to an epitope similar to ipilimumab, but with enhanced binding through specific mutations. Pairing mipi variants with various Fc domains (with or without FcγR binding) will help reveal the impact of potent checkpoint blockade in the absence of Fc effector function. It is believed that mipi variants will become versatile tools in the future, complementing existing surrogate and transgenic models for studying therapeutic strategies targeting CTLA-4 in mouse cancer models.
