•
A unanimous Supreme Court invalidated Amgen's PCSK9 antibody claims in May 2023 for failing to enable their full scope. The decision changed how broadly any biotech patent can be written.
Amgen v. Sanofi is a patent case that reads like a chemistry problem and lands like an earthquake. On May 18, 2023 a unanimous Supreme Court held that Amgen's patents claiming an entire class of antibodies were invalid because the specification did not enable a skilled scientist to make and use the full class.
Justice Gorsuch wrote the opinion. The Federal Circuit was affirmed. Every biotech patent written since has been drafted with this case in mind.
| Item | Detail |
|---|---|
| Caption | Amgen Inc. v. Sanofi |
| Court | Supreme Court of the United States |
| Docket | 21-757 |
| Decided | May 18, 2023 |
| Vote | 9-0 |
| Opinion | Justice Gorsuch |
| Statute | 35 U.S.C. section 112(a), the enablement requirement |
| Disposition | Affirmed |
PCSK9 is a protein that reduces the liver's ability to clear LDL cholesterol from the blood. Block it and cholesterol drops. Both Amgen and Sanofi developed antibody drugs that do exactly that, Repatha and Praluent respectively.
Amgen's patents did not merely claim its own antibody. They claimed the genus: every antibody that binds to specified amino acid residues on PCSK9 and blocks PCSK9 from binding to LDL receptors. That is a functional claim. It defines the invention by what the molecule does, not by what it is.
The specification disclosed the amino acid sequences of 26 antibodies. The class it claimed potentially runs to millions.
Section 112(a) requires the specification to describe the invention "in such full, clear, concise, and exact terms as to enable any person skilled in the art" to make and use it. The requirement is old. What Amgen tested is whether it applies to the whole scope of a functional claim or only to a representative sample.
Amgen offered two ways to reach the rest of the class. A "roadmap," which told scientists to generate antibodies, test them for binding, and keep the ones that work. And "conservative substitution," which told them to swap amino acids in a known antibody and test the results.
The Court's answer was that both amount to assigning research rather than teaching. If getting to the rest of the claimed class requires trial and error, the specification has not enabled it. The opinion put the principle plainly: the more a patent claims, the more it must enable.
Before Amgen, functional genus claims were a standard tool in biotechnology. You identified a target, found antibodies that hit it, and claimed all antibodies that hit it that way. That claim blocked competitors from designing around your specific molecule.
After Amgen, that strategy is very hard to sustain. Claims have narrowed toward specific sequences and specific structural features. The practical effect is more room for follow-on biologics and less protection for the first company to find a target.
There is a real tension here worth naming. Narrow claims mean competitors can enter faster, which is good for patients and prices. They also mean the company that spent a decade finding the target captures less of the value, which is the incentive patents exist to create. The Court chose the statutory text over that policy debate, which is the correct judicial answer and does not make the tradeoff disappear.
Enablement is now a routine invalidity defense rather than a long shot. Expect to see it raised at the pleading stage, in inter partes review petitions, and at summary judgment in any case involving a functional claim.
The related written description requirement, which asks whether the inventor possessed the full claimed scope, travels alongside it. The two are separate doctrines under section 112 and defendants plead both.
District court patent dockets are where this plays out. Claim construction orders and summary judgment rulings on enablement are the documents that show how the standard is being applied in practice, and most of them never get written about.
Supreme Court filings are free on supremecourt.gov. Everything downstream, the enablement fights in the District of Delaware, the Northern District of California, and the Eastern District of Texas, is on PACER.
Search current filings here: antibody enablement litigation. Because patent dockets go quiet for months and then produce a claim construction order without warning, case alerts are the sensible way to watch them.
Patent outcomes also vary sharply by district and by judge, more than in most areas of federal practice. Pulling a judge's claim construction history before filing is standard practice, and the judge lookup does it.
The practical response in biotech patent practice has been to move away from claims defined purely by function.
Claim specific sequences, and claim them in dependent form so that a challenge to one does not take the rest. Where a genus claim is necessary, disclose enough representative species across the claimed range that the specification is not asking a reader to run experiments. Describe structural features shared by the class, not just the binding behavior, so that the claim has a physical boundary and not only a functional one.
None of this is free. Narrower claims are easier to design around, which is the tradeoff the Court accepted when it read section 112(a) the way it did.
Three venues, and the practice differs in each.
District courts hear enablement at summary judgment and at trial, usually with competing expert testimony about how much experimentation the specification actually requires.
The Patent Trial and Appeal Board hears related challenges in inter partes review, though the statute limits IPR grounds and enablement arguments often arrive there indirectly.
The Federal Circuit reviews all of it, and its post-Amgen decisions are where the standard is actually being calibrated. Those opinions, not the Supreme Court's, are what a practitioner reads before drafting.
For another unanimous Supreme Court decision that narrowed a doctrine courts had applied loosely for years, read Cox Communications v. Sony Music. For a case where the Court took an appeal and then walked away without deciding it, see the Nvidia securities case. And for the intellectual property fight now consuming the AI industry, see Andersen v. Stability AI.