Juno v. Kite: Written Description and the Genus the Patent Could Not Possess

Juno Therapeutics, Inc. v. Kite Pharma, Inc., No. 20-1758 (Fed. Cir. Aug. 26, 2021), is the written-description counterpart to the enablement decisions in Amgen, Idenix, and Wyeth. Writing for the panel, Chief Judge Moore reversed a jury verdict (and the judgment built on it, which exceeded $1.2 billion), holding that Juno’s foundational CAR-T patent failed the written-description requirement of 35 U.S.C. § 112(a). The patent claimed a chimeric antigen receptor with “a binding element that specifically interacts with a selected target,” a phrase broad enough to encompass an essentially limitless universe of antibody fragments. The specification disclosed only two. The decision is a leading statement of how written description disciplines functional genus claims through a different doctrinal lens than enablement.

At a glance

• Case: Juno Therapeutics, Inc. v. Kite Pharma, Inc., No. 20-1758
• Court and date: U.S. Court of Appeals for the Federal Circuit; decided August 26, 2021 (precedential)
• Author: Chief Judge Moore, joined by Judges Prost and O’Malley
• Patent: U.S. Patent No. 7,446,190 (the ‘190 patent), owned by Sloan Kettering and licensed to Juno
• Holding: The asserted claims are invalid for lack of written description; the specification did not show possession of the full claimed scFv genus
• Status: Final; the Supreme Court denied certiorari in November 2022

The technology and the claim

Chimeric antigen receptor (CAR) T-cell therapy re-engineers a patient’s own T cells to recognize and kill cancer cells. The ‘190 patent claimed a nucleic acid encoding a three-part CAR: an intracellular signaling (zeta) domain, a co-stimulatory domain (such as CD28), and an extracellular binding element that targets the cell to its quarry. The binding element is typically a single-chain antibody variable fragment, or scFv: a small engineered antibody piece tuned to recognize a specific antigen.

The claim language at issue was deliberately broad. It required “a binding element that specifically interacts with a selected target,” without restricting the antigen, the antibody, or the scFv’s sequence or structure. Read literally, it covered CARs directed at any target through any scFv. Kite’s accused product, YESCARTA (axicabtagene ciloleucel), uses a CAR with an scFv that binds CD19, an antigen on certain blood cancers. The jury found willful infringement and awarded a $585 million upfront payment plus a 27.6% running royalty. The district court then updated the award to $778,343,501 to reflect YESCARTA revenues through trial, added prejudgment interest, and enhanced damages by 50%, producing a final judgment of $1,200,322,551.50. Kite’s defense was that the claims, so broad, were never adequately described.

The written-description requirement

Written description, like enablement, lives in § 112(a), but it asks a distinct question. Under Ariad Pharmaceuticals v. Eli Lilly, the specification must reasonably convey to a skilled artisan that the inventor possessed the claimed subject matter as of the filing date. For a genus claim, possession is shown in one of two ways: by disclosing a representative number of species spanning the genus, or by identifying structural features common to the members of the genus that allow a skilled artisan to recognize which compounds belong.

The doctrine is not satisfied by a claim that merely recites a desired function. Describing what a molecule must do (bind a selected target) is not the same as describing the molecules that do it. That gap between function and structure is exactly where written description does its work, and it is why a claim can be enabled (a skilled artisan could make scFvs to many targets) yet still lack written description (the inventor did not show possession of the full claimed class).

Why the genus was not described

The scFv genus embraced by the claim was enormous. On the record, the universe of possible scFv sequences runs to “millions of billions,” a figure Kite advanced and Juno did not dispute. Against that breadth, the ‘190 specification disclosed only two scFv species, identified by nothing more than their alphanumeric designations: J591 (derived from an antibody binding the prostate antigen PSMA) and SJ25C1 (derived from an antibody binding CD19). The patent did not disclose the amino acid sequence of either one. Two examples out of a genus that vast cannot be a “representative number,” especially when the two share no special structural relationship that would stand in for the rest.

Nor did the specification supply common structural features. scFvs that bind different targets do not share a predictable sequence or architecture that determines binding; the part of the antibody that does the recognizing (the complementarity-determining regions) varies enormously and unpredictably from one scFv to the next. The specification offered no rule, no shared motif, and no method for distinguishing scFvs capable of binding a given target from those incapable of it. Under both prongs of Ariad, then, the patent failed: too few species, and no common structure. The court concluded that the inventors had not shown possession of the broad genus they claimed, and reversed the verdict.

Notably, narrowing by target did not rescue the patent. Claims 5 and 11 were limited to scFvs binding CD19, yet they fell too: the realm of possible CD19-specific scFvs was itself vast, at most five were known in the art at the priority date, and the ‘190 patent gave no sequence, shape, or characteristic distinguishing scFvs that bind CD19 from those that do not. All four asserted claims (3, 5, 9, and 11) were held to lack written description.

Function versus structure, again

What makes Juno analytically important is its insistence that functional claiming cannot substitute for structural disclosure. The patentee effectively claimed “whatever scFv binds whatever target,” then pointed to two worked examples as proof of possession of the whole. The Federal Circuit’s answer is that possession of a functional genus requires either enough representative members to map the territory or a structural common denominator that defines membership. A binding result supplies neither.

This is the written-description analogue to the enablement holdings in Amgen, Idenix, and Wyeth. Where those cases asked whether the specification taught the full scope, Juno asks whether it showed possession of the full scope. The two requirements often travel together and frequently sink the same overbroad claim, but they are independent: a claim can fail one while satisfying the other, and a careful challenger pleads both.

Open questions

Juno does not quantify how many scFv species would be “representative,” nor how strong a structural commonality must be to carry an antibody genus. Those questions will be litigated antigen by antigen. It also sharpens, without resolving, the tension between rewarding pioneering inventors and policing claim breadth: Juno’s inventors were genuinely early to three-domain CARs, yet the breadth of the binding-element limitation outran their disclosure. Finally, the decision leaves open how foundational platform patents in fast-moving biologic fields should be drafted to claim meaningful scope without overreaching, a problem Juno poses acutely but does not solve.

Implications

• Function is not possession. Claiming a binding element by what it does, without representative species or common structure, is fatal under written description for a large genus.
• Two examples rarely carry a genus this size. Disclose species across the breadth actually claimed, or claim only the scope the examples support.
• Plead both § 112 theories. Written description and enablement are independent; a genus claim that survives one may fall to the other.
• Pioneering does not excuse overclaiming. Even genuinely early inventions are confined to what the specification shows the inventor possessed as of filing.
• Verdict size is no shield. A billion-dollar award rests on valid claims; a written-description defect erases it on appeal.

Frequently asked questions

What is the difference between written description and enablement here? Enablement asks whether the specification teaches a skilled artisan how to make and use the full claimed scope. Written description asks whether the specification shows the inventor actually possessed that full scope at filing. Juno turned on the second: too few example scFvs and no common structure to define the claimed genus.

Why weren’t two disclosed scFvs enough? Because the claim reached a genus the record put in the range of “millions of billions” of possible scFv sequences binding any target. Two species (binding PSMA and CD19) are not “representative” of that range, and the two shared no structural feature that would let a skilled artisan identify the rest of the genus.

Does Juno mean CAR-T patents are unprotectable? No. It means claims must be tied to the disclosure. Claims defined by the structure of the scFvs actually described, such as a disclosed sequence, are a different case; the problem was claiming binding elements by function on the strength of two examples the patent barely described. Note that narrowing by target alone was not enough here: claims 5 and 11, limited to CD19, failed as well.

Authorities and sources

• Juno Therapeutics, Inc. v. Kite Pharma, Inc., No. 20-1758 (Fed. Cir. Aug. 26, 2021). Panel, claim language, the J591/SJ25C1 disclosures, the “millions of billions” record figure, the damages history, and the written-description holding all taken from the slip opinion: U.S. Court of Appeals for the Federal Circuit.
• U.S. Patent No. 7,446,190, “Nucleic acids encoding chimeric T cell receptors” (issued Nov. 4, 2008), inventors Sadelain, Brentjens, and Maher, assigned on its face to Sloan-Kettering Institute for Cancer Research: USPTO.
• Commentary corroborating the Ariad two-prong analysis: Wolf Greenfield and Haug Partners.
• The Supreme Court’s 2022 denial of certiorari reported by IPWatchdog.