On July 16, 2026, the USPTO published US20260199503A1, an application assigned to Regeneron Pharmaceuticals, Inc. and titled “GLP1R Agonist-Tethered GDF8 Antibody Conjugates and Uses Thereof.” The kind code is A1: this is a published application, pending before the examiner. Nothing in it has been granted, and the claims discussed below are the claims as published, not as they may eventually issue — if they issue at all. This record is a useful teaching case for a rule this publication applies without exception: a filing is defined by its claims, never by its title and never by its abstract. Here, the abstract and the claims do not agree, and a reader who stopped at the abstract would come away with the invention inverted. The abstract states that the disclosure provides “GLP1 peptidomimetics, antibody-drug conjugates, and compositions which comprise anti-GLP1R antibodies and GLP1 peptidomimetic payloads.” Read literally, that describes an antibody whose target is the GLP-1 receptor. The claims describe no such thing. Claim 1 — the lead independent claim — recites a composition comprising “an GLP1R agonist-tethered GDF8 antibody conjugate” [the article reads as printed in the source] in which element (a) is an antibody, or antigen-binding fragment, that specifically binds to and/or blocks the biological activity of Growth and Differentiation Factor-8 (GDF8, identified in the claim as SEQ ID NO: 1); element (b) is at least one GLP-1 receptor agonist; and element (c) is at least one linker that covalently connects the agonist to the GDF8 antibody. The antibody's target is GDF8 — the protein also known as myostatin. The GLP1R agonist is cargo, attached to that antibody, not the thing the antibody binds. The title agrees with the claims. Every independent claim in the record agrees with the claims. The abstract is the outlier, and its wording is consistent with boilerplate carried forward from a related filing in the same conjugate family — a common enough artifact in large portfolios. We report the divergence as a fact of the document and stop there: what an examiner will do about it, and what it means for the application, are not questions this record answers.

What the independent claims actually cover

The record contains 34 populated claim entries across a numbering range that runs to 147, with independent claims surviving at 1, 25, 81, 87, 101, 114, and 140. They divide into two families: what the molecule is, and how it is made. On the composition side, claim 87 gives the tightest statement — a conjugate comprising an antigen-binding protein that specifically binds and/or blocks GDF8, with a GLP1R agonist conjugated to it through a linker. Claim 25 formalizes the same architecture as a Formula (I) in which BA is the GDF8-binding antibody or fragment, P is the GLP1R agonist, L is the linker, and n — the drug-to-antibody ratio — “ranges from about 1 to about 16.” The formula itself is a structure drawing rather than text in the published record, and the DAR range is a claim boundary: it recites the span the applicant seeks to fence, and says nothing about any value that was made or measured. Claim 114 reaches the tethered GLP-1 peptidomimetic payload as a standalone compound.

The process claims are where this filing gets specific, and they are the reason the CPC classification lands where it does. Claim 101 is directed to conjugating a drug, ligand, or handle site-specifically to an isolated antibody or fragment “wherein the site of conjugation is at Gln/Q55 of the light chain of the antibody,” with the reaction assisted by microbial transglutaminase (mTG). A single named residue on the light chain, with an enzyme directing chemistry to it, is a claim about manufacturing control — producing a defined conjugate rather than a stochastic mixture. Claim 81 covers the full manufacturing sequence:

A process for manufacturing a conjugate of GLP1R agonist tethered GDF8 antibody or antigen binding fragment thereof comprising a) covalently attaching a handle comprising a first reactive moiety for Click or Diels-Alder reaction, in the presence of microbial transglutaminase; b) exposing a GLP1R agonist comprising a second reactive moiety for Click or Diels-Alder reaction, wherein the first and the second reactive moieties are complimentary to each other and form a stable conjugate; and c) isolating or purifying the conjugate of GLP1R agonist tethered GDF8 antibody or antigen binding fragment thereof.— GLP1R Agonist-Tethered GDF8 Antibody Conjugates and Uses Thereof, US20260199503A1

That claim is quoted as printed, including its spelling of “complimentary.” The two-step logic is clear enough: mTG installs a reactive handle on the antibody, the payload arrives carrying the matching partner, and Click or Diels-Alder chemistry joins them. Regeneron has filed on that chemistry as a platform in its own right — US20260053938A1, “Diels-Alder Conjugation Methods,” published February 26, 2026, claims protein-payload conjugates made by combining transglutaminase with Diels-Alder chemistry. It is the closest sibling to this record and predates it by nearly five months.

The claims are not settled, and the record says so

Eight blocks of the numbering are marked canceled: claims 2–24, 26–56, 73–80, 82–86, 88–100, 102–113, 124–139, and 141–147. Cancellation on that scale is the visible trace of an application being actively prosecuted — scope is under negotiation, and what remains on the page today is a snapshot, not an outcome. Among the survivors is a block worth flagging precisely because of how it reads. Claims 66–72 are method-of-treatment claims, and their language sounds like results: claim 67, for instance, is directed to treating Type 2 diabetes by improving glycemic control and maintaining or increasing lean body mass while reducing fat mass. That is what the application seeks to claim as a boundary. A method-of-treatment claim is a request for legal scope; it is not data, not a finding, and not an assertion that anything was achieved. This record, as published, offers no efficacy evidence, no trial status, and no regulatory posture, and none should be read into those claims. The classification places the filing squarely in conjugate chemistry. The CPC codes are A61K 47/6845, A61K 47/6811, and A61K 47/6889 — the carrier-linked-payload and antibody-drug-conjugate group — alongside C07K 16/22 (antibodies against growth factors), C07K 14/001 (peptides), and A61P 3/04. The application names eleven inventors, including Andrew J. Murphy and George D. Yancopoulos as the final two on the list. That is a checkable fact on the face of the record; it is not evidence of anything beyond itself. The record does not state a priority date, and this brief does not supply one.

Only one other Regeneron application published in the July 16 drop: US20260201410A1, “Transcription Modulation in Animals Using CRISPR/Cas Systems,” covering non-human animal cells and animals carrying CRISPR/Cas synergistic activation mediator components. Beyond that day, the 2026 file wrapper shows steady filing across adjacent engineering surfaces — US20260139044A1 (Tumor-Targeted Split IL2 Receptor Agonists, May 21, 2026), an antibody-tethered agonist architecture in a different target space; US20260092123A1 (Engineered IgG Molecules, April 2, 2026); US20260152545A1 (Metabolically Optimized Cell Culture, June 4, 2026); and US20260118314A1 (Microchip Capillary Electrophoresis Assays and Reagents, April 30, 2026), a QC-analytics filing. Each published on its own date, well before this week.

Read strictly from its claims, US20260199503A1 is an application on a conjugate that carries a GLP1R agonist on a GDF8-binding antibody chassis, and on the enzyme-directed, site-specific chemistry for building it. The abstract will tell you otherwise. The claims govern.