Amgen v. Sanofi: Enablement Rule, Holding, and Patent Impact

In Amgen v. Sanofi, decided May 18, 2023, the U.S. Supreme Court unanimously held that two of Amgen’s patents on cholesterol-lowering antibodies were invalid because Amgen claimed an entire class of antibodies defined by what they do, while teaching the public how to make only a handful of them. Justice Neil Gorsuch, writing for the Court, framed the rule plainly: the broader the claim, the more the patent must disclose. Amgen’s specification did not clear that bar.¹

The Drugs and the Patents at Issue

Amgen and Sanofi both developed injectable antibody drugs that target a protein called PCSK9. PCSK9 attaches to LDL receptors on liver cells and drives them to be destroyed, which reduces the liver’s ability to clear “bad” cholesterol from the blood. Antibody drugs that block PCSK9 keep those receptors intact. Amgen sold its version as Repatha; Sanofi sold its as Praluent.²

In 2011, each company obtained a patent on the specific antibody in its own drug, identified by its unique amino acid sequence. Those patents were not in dispute. The fight was over two later patents Amgen obtained in 2014. Those patents did not just cover the 26 specific antibodies Amgen had actually identified. They claimed every antibody, disclosed or not, that could perform two functions: bind to certain spots on the PCSK9 protein and block it from attaching to LDL receptors. Because antibodies are extraordinarily diverse at the molecular level, that functional definition potentially reached millions of antibodies Amgen had never made or described.³

When Sanofi launched Praluent, Amgen sued, arguing that Sanofi’s antibody fell within Amgen’s broad functional claims. Sanofi answered that the patents were invalid because Amgen had not taught the public how to produce the vast universe of antibodies it was trying to own.

The Enablement Requirement

The case turned on a rule known as enablement. Federal patent law requires a patent’s specification to describe the invention in enough detail that a skilled professional in the field can make and use it.⁴

Enablement is the trade at the heart of the patent system. The inventor gets a time-limited monopoly. In exchange, the public gets a working blueprint, so that when the patent expires anyone can pick up the specification and reproduce the invention. If the disclosure does not actually allow that, the exchange breaks down.

The rule does not demand perfection. A specification can leave room for routine hands-on work by someone who knows the field: calibrating temperatures, adjusting parameters, refining conditions. What it cannot do is force the reader into open-ended trial and error with no assurance of success. At that point the patent stops being a blueprint and becomes a research assignment.¹

The Court’s holding tied that principle directly to claim scope. If a patent claims an entire class of inventions, the specification must enable a skilled person to make and use the entire class. Some experimentation is acceptable; what counts as reasonable depends on the nature of the invention and the state of the underlying science. But identifying a desired result and leaving the rest to discovery is not enough.

What Amgen’s Specification Actually Taught

To support claims that reached across an entire functional class, Amgen’s specification offered three things: the amino acid sequences of 26 antibodies that performed the claimed functions, the three-dimensional structures of two of those 26, and two methods that others could theoretically use to find additional antibodies.¹

The first method, which the Court called the “roadmap,” told scientists to generate a large pool of antibodies, screen them for binding to PCSK9, screen the survivors for binding at the right region, and then screen those for the ability to block PCSK9 from attaching to LDL receptors. In substance: make candidates and keep testing until something works.

The second method, “conservative substitution,” told scientists to start with an antibody already known to work, swap out amino acids for chemically similar ones, and test whether the new version still performed both functions. This was more targeted, but the outcome still could not be predicted. Even small changes to an antibody’s structure can destroy its function, and current science offers no reliable way to know in advance which changes will.

The Court saw both methods for what they were. Neither told a skilled scientist which antibodies within the claimed class would work; both required building and testing each candidate individually. Justice Gorsuch called them “little more than two research assignments” and compared the disclosure to a hunting license rather than a map to the quarry.¹

Before the case reached the Supreme Court, the Federal Circuit had made the factual picture concrete. It found the claims “far broader in functional diversity than the disclosed examples” and the antibody field too unpredictable for Amgen’s limited disclosure to bridge the gap. Experts acknowledged that translating an amino acid sequence into a predicted three-dimensional structure was not yet possible, which meant there was no shortcut around the screening process Amgen described. The Federal Circuit held that claims using broad functional language face high hurdles in unpredictable scientific fields, and the Supreme Court affirmed that analysis without reservation.⁵

Older Cases the Court Relied On

One of the notable features of the opinion is how directly it connected Amgen’s claims to nineteenth-century patent disputes. The Court treated the legal principle as long-settled and reached back for two examples.

The first was O’Reilly v. Morse, decided in 1854. Samuel Morse invented the telegraph and received valid patents on his specific device. His eighth claim, however, tried to reach every use of electromagnetism to print characters at a distance, regardless of the machine. The Court struck that claim down. Morse could not claim every possible method of achieving a result when he had invented only one.⁶

The second was The Incandescent Lamp Patent, decided in 1895. Two inventors, Sawyer and Man, discovered that carbonized paper worked as a light bulb filament, then claimed a patent covering all fibrous and textile materials used as filaments. Thomas Edison had tested hundreds of plant materials before finding that a specific bamboo worked best because of its parallel fiber structure. The Court rejected the broad claim. There was no shared quality among all fibrous materials that made them suitable filaments, and letting Sawyer and Man block Edison’s more specific invention was exactly the harm the enablement rule was meant to prevent.⁷

Justice Gorsuch drew the parallel directly. Morse tried to claim all telegraphic communication. Sawyer and Man tried to claim all fibrous filament materials. Amgen tried to claim an entire kingdom of antibodies. The technology changes across centuries; the legal error does not.¹

The Holding

The Court affirmed the district court and the Federal Circuit and held Amgen’s claims invalid for lack of enablement. Its core reasoning was the mismatch between what Amgen claimed and what it taught. Amgen claimed a monopoly over a potentially limitless class of antibodies defined solely by function. Its specification disclosed 26 examples and two search methods that reduced to systematic guessing.¹

The Court was careful to say it was not announcing a new legal framework. It applied the same enablement standard that has existed since the first Patent Act of 1790, and it emphasized that the principle holds “whether the case involves telegraphs devised in the 19th century, glues invented in the 20th, or antibody treatments developed in the 21st.” No field of technology gets a special pass on the disclosure requirement.¹

What the Ruling Changes for Patents Going Forward

The immediate consequence is that broad, functionally defined claims in unpredictable fields face a much harder path to validity. If small structural changes to an invention produce unpredictable results, a patent specification has to do more than describe what the invention does. It has to show how to make and use every variation the patent claims, or provide enough guidance that a skilled person can get there without open-ended experimentation.

That pushes patent strategy toward claims tied to specific structures rather than broad functions. Instead of claiming “any antibody that binds to protein X and blocks activity Y,” an applicant is better served claiming specific antibodies by their amino acid sequences or structural features, along with closely related variants the specification demonstrates will work. The resulting patents are narrower but far more defensible. For companies in research-intensive industries, protecting a class of drugs is likely to require multiple patents filed over time as new discoveries emerge, rather than a single broad filing early in the research cycle.¹

Courts have applied the reasoning quickly. In Baxalta v. Genentech, decided months later, the Federal Circuit invalidated another broad antibody patent. That patent claimed all antibodies that bind to a blood-clotting protein and increase its activity, disclosed sequences for eleven such antibodies, and offered a screening method structurally similar to Amgen’s roadmap. The Federal Circuit called the facts “materially indistinguishable from Amgen” and affirmed summary judgment of invalidity. It also confirmed that Amgen did not disturb prior enablement case law and saw “no meaningful difference” between the longstanding “undue experimentation” standard and the Court’s “unreasonable experimentation” framing.⁸

The reach is not limited to biotechnology. In In re Starrett, the Federal Circuit rejected patent claims for a system described at a high level of abstraction, finding the specification amounted to an invitation to experiment rather than an enabling disclosure. That confirms what the Supreme Court signaled: the enablement principle is technology-neutral. A patent claiming “any machine learning model that achieves outcome X” without enough disclosure to reliably reproduce the full range of covered models faces the same vulnerability that sank Amgen’s claims.

The ruling does not ban functional claims or class claims outright. A patent can still claim a class of inventions if the specification enables the full scope of that class. In a highly predictable field, where a few examples support reliable extrapolation, a broad claim may survive. The problem arises when breadth and unpredictability combine, and the specification offers only a few data points plus instructions to keep searching. That combination is what Amgen, Morse, and Sawyer and Man had in common, and it is what the Court continues to reject.

1
Supreme Court of the United States. Amgen Inc. v. Sanofi, No. 21-757
2
Justia. Amgen Inc. v. Sanofi, 598 U.S. ___ (2023)
3
Supreme Court of the United States. Amgen Inc. v. Sanofi, No. 21-757
4
Office of the Law Revision Counsel. 35 USC 112 – Specification
5
United States Court of Appeals for the Federal Circuit. Amgen Inc. v. Sanofi, No. 20-1074
6
Library of Congress. O’Reilly v. Morse, 56 U.S. (15 How.) 62 (1854)
7
Library of Congress. The Incandescent Lamp Patent, 159 U.S. 465 (1895)
8
United States Court of Appeals for the Federal Circuit. Baxalta Inc. v. Genentech Inc., No. 22-1461