The terms, per the grant. On March 25, 2025, Amgen was issued US12258404B2, “Bispecific antibody construct directed to MUC17 and CD3.” T-cell engagers are a contested oncology category, and a construct grant on a specific target pair is the kind of asset that anchors dealmaking in the race.

The grant's value to a deals desk comes from how narrowly the construct is defined. Independent claim 1 does not cover MUC17xCD3 bispecifics generally; it covers a construct with "a first domain comprising an antibody which binds to MUC17" defined by six specific complementarity-determining-region sequences (CDR-H1 through CDR-L3, set out as named SEQ ID NOs), paired with "a second domain comprising an antibody which binds to an extracellular epitope of the human CD3ε chain." That is a particular antibody, specified down to its CDR sequences — the molecule itself, not the target concept.

Why a business desk values a construct grant: in the T-cell-engager field the differentiated asset is the specific molecule — the binding domains, the format, the Fc engineering — and composition-of-matter claims over that construct are exactly what an oncology licensing or co-development deal is written around. The grant fixes the molecule a partner would be paying for.

“The present invention provides bispecific antibody constructs characterized by comprising a first domain binding to MUC17, a second domain binding to an extracellular epitope of the human and the Macaca CD3ε chain and optionally a third domain, which is a specific Fc modality.”— U.S. Patent No. 12,258,404 source

The dependent claims fill in the format and the engineering, which is where bispecific value is contested. The construct can add "a third domain which comprises two polypeptide monomers, each monomer comprising a hinge, a CH2 domain and a CH3 domain" (claim 2), arranged "hinge-CH2-CH3-linker-hinge-CH2-CH3" (claim 3) — the Fc modality referenced in the abstract, which governs half-life and manufacturability. The claims reach a "bispecific single chain antibody construct" format (claim 6), fix the domain order "first domain — peptide linker — second domain" with named linker sequences (claim 8), and tie the CD3-binding domain to a defined set of sequences (claim 12). For a deals desk, the format and Fc claims are what separate this asset from a competitor's MUC17xCD3 engager.

The grant also claims cross-reactivity and the supporting commercial scaffold. The second domain "further binds to an extracellular epitope of the Macaca CD3ε chain" (claim 17) — cynomolgus cross-reactivity, the feature that makes a construct usable in standard nonclinical work and therefore more straightforward to develop under a partnership. And the grant reaches "a pharmaceutical composition comprising the bispecific antibody construct…and a carrier" (claim 15) and "a kit" with "a means for reconstituting or diluting the antibody construct" (claim 16). Composition and kit claims layered on the construct give a licensing desk a fuller estate around the single molecule.

The structure point: a construct grant supports composition-of-matter licensing — the strongest, longest-lived form of antibody IP — and the milestone-and-royalty deal structures that attach to a specific molecule, rather than a method royalty tied to a use. For a model, a sequence-defined construct with a claimed Fc format and species cross-reactivity is a clean, ownable asset whose boundaries a counterparty can read directly.

What the grant does not promise: a clinical result, an approval, or that MUC17xCD3 is a validated target. It is an exclusivity claim on a specific construct, defined by its CDR sequences and format — valuable precisely because it is specific, and bounded to the molecule the claims recite.

The reason a sequence-defined construct is the asset a T-cell-engager deal is priced against, rather than the target pair, is that the target pair is shared and the molecule is not. Multiple developers can pursue MUC17 and CD3 simultaneously; what differentiates one program is the specific binding domains, the format that joins them, and the Fc engineering that governs half-life and manufacturability. By defining the MUC17 binder through six named CDR sequences (claim 1) and the construct's overall sequence through additional SEQ ID NOs (claims 11, 14), the grant claims the molecule itself. Composition-of-matter protection of that kind is the strongest and longest-lived form of antibody IP, which is why it is the cleanest thing to attach milestone-and-royalty economics to.

The format and Fc claims are where bispecific value is genuinely contested, so a deals desk reads them closely. The optional third domain (claim 2) supplies the Fc modality referenced in the abstract, built from "two polypeptide monomers, each monomer comprising a hinge, a CH2 domain and a CH3 domain," arranged in the order "hinge-CH2-CH3-linker-hinge-CH2-CH3" (claim 3). That architecture is what gives an otherwise small bispecific a usable half-life and a manufacturable structure, and claiming it specifically is what separates this asset from a competitor's differently-formatted engager. The domain-order and linker claims (claims 8 and 9, with named linker SEQ ID NOs) fix how the binders are strung together; claim 12 ties the CD3 arm to a defined sequence set. The cross-reactivity claim (claim 17), in which the CD3 binder "further binds to an extracellular epitope of the Macaca CD3-epsilon chain," is a development-enabling feature that makes standard nonclinical work straightforward and therefore makes the asset more attractive to a partner. Composition and kit claims (claims 15 and 16) round out the estate around the single molecule, giving a licensing desk a fuller, more defensible position to transact on.

The takeaway: in the T-cell-engager race, the construct grants are the load-bearing assets, because the differentiated molecule is what a deal is priced against. Amgen's March 2025 MUC17xCD3 construct grant is a dated, concrete example of the composition IP behind that dealmaking — specified down to the CDR sequences and the Fc arrangement.