The danger of a T-cell engager is that it works everywhere, including healthy tissue. US11242390B2 - "Protease-activated T cell bispecific molecules," issued February 8, 2022 to Hoffmann-La Roche Inc. - claims an answer: a bispecific that stays switched off until it reaches the tumor.

The load-bearing limitation is the conditional-activation mechanism, and claim 1 builds it out of a precise set of parts. The claim recites an idiotype-specific polypeptide covalently attached through a linker to a T cell activating bispecific molecule comprising an anti-CD3 antigen-binding site. That idiotype-specific polypeptide binds the idiotype of the anti-CD3 site itself - it physically caps the CD3 arm - to reversibly conceal it. The linker holding the cap in place is not a generic tether: it is a protease-cleavable linker comprising a protease recognition site of SEQ ID NO: 106. The molecule is therefore born masked; only when a protease cuts at that defined site does the cap fall away and the CD3 arm become competent to engage a T cell.

"The present invention generally relates to novel protease-activatable T cell activating bispecific molecules and idiotype-specific polypeptides."- U.S. Patent No. 11,242,390 source

The claim set is short - five claims - and that compactness is itself informative. The masking element is specified to be an anti-idiotype scFv (claim 2): a single-chain antibody fragment that recognizes the unique binding loops of the anti-CD3 arm and sits over them. Claim 3 specifies that the linker is a peptide linker; claim 5 reiterates the protease recognition sequence of SEQ ID NO: 106; claim 4 covers a pharmaceutical composition built around the polypeptide. The architecture, then, is: a known anti-CD3 engager, a single-chain anti-idiotype cap, and a defined protease-cleavable peptide bridge between them. The binding arms (anti-CD3 and, in the broader engager format, an anti-tumor-antigen arm) are conventional engager parts - the novelty is the protease-cleavable switch and the idiotype-specific cap that the switch releases.

Why is the switch the invention and not the arms? Because engagers that bind CD3 and a tumor antigen are well-trodden prior art. What is new and claimable here is the conditional, protease-gated unmasking - and, more narrowly still, doing the masking by anti-idiotype concealment of the CD3 site rather than by some other steric block. The disclosed mechanism is elegant: instead of masking the antigen-binding paratope with a generic peptide that might or might not occlude it, the claim uses a binder raised against the engager's own idiotype, which guarantees a tight, specific cap. The CPC tags reflect the layered design: C07K 16/4208 (the cleavable/masking element) alongside the standard C07K 16/2809 (anti-CD3) and C07K 16/2863 binding components, with A61P 35/00 as the oncology classification.

The claim-scope question turns on the protease-cleavable linker and the masking architecture - specifically the protease recognition site of SEQ ID NO: 106 and the anti-idiotype concealment strategy. A competitor's masked bispecific infringes only if its switch and cap fall within this design: a different protease recognition sequence, or a masking moiety that is not an idiotype-specific polypeptide against the CD3 arm, sits outside the literal language. The binding targets are almost beside the point for infringement; the switch and how it is concealed are everything.

The structure of claim 1 repays a second look, because it is drafted around the mask rather than the engager. The independent claim's subject is "an idiotype-specific polypeptide" - the cap - that is covalently attached through the protease-cleavable linker to a bispecific bearing an anti-CD3 site. By making the masking polypeptide the claimed entity and the engager a recited environment, the drafter centers the patent's enforceable weight on the concealment technology. That is a deliberate choice: it means the claim reads on the mask-plus-linker assembly wherever it is deployed, provided the protease recognition site of SEQ ID NO: 106 is present.

The reversibility language - the cap "reversibly conceal[s]" the anti-CD3 site - is also load-bearing. It signals that the molecule is designed to be inert in circulation and competent only after cleavage, which is the whole safety rationale of conditional activation: localize activity to where the triggering protease is enriched. The disclosed mechanism is a conditional-activity switch built from two specific molecular pieces, an anti-idiotype scFv (claim 2) and a defined protease recognition peptide (claims 1 and 5), rather than from a generic steric mask.

For claim construction, the compactness of the set - five claims, two of which restate the protease recognition site - tells you the patentee chose depth over breadth. There is no sprawling Markush here and no long menu of alternative masks or proteases; the claim commits to anti-idiotype concealment and to SEQ ID NO: 106. A competitor designing a masked engager has a clear map: a different masking modality (for example, a coiled-coil or a steric peptide that is not idiotype-specific) or a different protease recognition sequence is, on the face of these claims, a different invention. That clarity cuts both ways - it makes the claim easy to design around, and equally hard to invalidate as overbroad.

For the landscape, conditional-activation grants represent the safety-engineering frontier of the bispecific estate. As engagers moved from blood cancers toward solid tumors, on-target-off-tumor activity in healthy tissue became the limiting problem, and the patents that solve it - masking, protease activation, logic-gating - form a distinct and increasingly valuable sub-estate layered over the basic engager format. A freedom-to-operate map for any masked or pro-drug engager has to read this claim narrowly and precisely: it is not a patent on "masked bispecifics" in general but on a specific anti-idiotype, protease-cleavable concealment of an anti-CD3 arm.