The limitation that matters in base-editor claims is the deaminase, not the word "CRISPR." Beam Therapeutics' US11155803B2, "Adenosine deaminase base editors and methods of using same to modify a nucleobase in a target sequence" (issued October 26, 2021; CPC C12N 9/78), and Harvard's US11999947B2, "Adenosine nucleobase editors and uses thereof" (issued June 4, 2024; CPC C12N 9/22), both claim a fusion protein: an adenosine deaminase joined to a programmable DNA-targeting protein. The targeting domain points; the deaminase does the chemistry.
What the specification must enable is the engineered deaminase. Wild-type adenosine deaminases do not efficiently act on DNA; the inventive step in this lineage (the David R. Liu and Nicole Gaudelli line of work appears across these grants) was evolving a deaminase that does. So the claim limitation reciting the deaminase — its activity on a target adenine within a programmable complex — is the genus-defining element. Construe that narrowly and the claim covers specific engineered enzymes; construe it broadly and it reaches a class of A-to-G editors.
Distinguish this from ordinary CRISPR-Cas cutting. A nuclease like Cas9 makes a double-strand break; base editors deliberately avoid that, nicking at most one strand and chemically converting the base. The claims reflect that distinction — they recite deamination and base conversion, not cleavage. A practitioner reading a base-editor claim should look for the conversion language (A·T to G·C) as the signature that separates it from the cut-and-repair CRISPR genus.
The continuation activity signals where the field is tightening. Beam's later grant US12454694B2, "Compositions and methods for improving base editing" (issued October 28, 2025), claims improvements to the editing system — the kind of follow-on filing that builds a thicket of refinements around the foundational deaminase-fusion concept. Multiple grants from the same assignee on incremental improvements is portfolio behavior, not a single dispositive claim.
For freedom-to-operate, the question is layered. Does your editor use an adenosine deaminase fused to a programmable targeting protein? If yes, you are in the neighborhood of these claims and must read the deaminase limitation against your specific enzyme. The assignee landscape here — Harvard, Broad, Beam, and the University of California all hold base-editing IP — means clearing one family does not clear the field; the foundational and improvement claims are distributed across institutions.
The takeaway, in this column's house style: a base-editor patent is a deaminase-fusion claim, and you analyze it by construing the deaminase and the conversion it performs, not by waving at "gene editing." The targeting domain makes it programmable; the engineered deaminase makes it a base editor — and that engineered enzyme is the limitation that decides how far the claim reaches.