Two Days after Teva v. Eli Lilly: Sum of All Fears Averted?

"Summarizing Judge Dyk’s dissent more simply: one cannot declare the race easy by moving the starting line past the hardest hill."

I. Introduction

Eli Lilly won its first patent battle with Teva by persuading the Patent Office that humanized anti-CGRP antagonist antibodies were obvious: the antibodies were “well known,” it said, and the methods of making them “extensively described,” and humanization of the antibodies was “routine.” Teva’s antibody claims were invalidated. But Lilly ultimately lost its patent war with Teva by unsuccessfully arguing at trial that the same antibodies could not be made and identified across the claimed genus without undue experimentation to treat headache. In the end, Teva prevailed on the method of treatment claims that survived inter partes review at the Patent Office. For its part, Teva traveled the same road but in the opposite direction. It defended its headache treatment patents at the Patent Office by arguing a skilled artisan would have had no reasonable expectation that the claimed antibodies would treat headache. At trial, it persuaded the jury that its patent specification taught that every one of them would.

Those self-colliding positions before the Patent Office and district court set the stage for everything that followed.

On September 30, 2026, in a short order, the Federal Circuit denied Lilly’s petition for rehearing en banc in Teva Pharmaceuticals International GmbH v. Eli Lilly & Co., No. 24-1094 (Fed. Cir. Sept. 30, 2026) (order). The denial left in place an April 2026 panel decision, Teva Pharms. Int’l GmbH v. Eli Lilly & Co., No. 24-1094 (Fed. Cir. Apr. 16, 2026) (Teva, slip op.), that reversed the district court’s post-trial judgment as a matter of law (“JMOL”) that had invalidated Teva’s method-of-treatment patent claims covering the use of humanized anti-CGRP antagonist antibodies to treat headache. The panel’s decision revived the patent claims and reinstated the jury’s verdict of willful infringement and $176.5 million in damages, now subject to further proceedings on remand. Lilly has until December 29, 2026, to ask the Supreme Court to take up the case.

Judge Dyk was the sole dissenter to the Federal Circuit’s denial of rehearing en banc, and he wrote a provocative dissent. Echoing the numerous industry amicus briefs filed in the matter, he said that the panel’s opinion creates an “end-run around Amgen,” the Supreme Court’s recent guidance on patent enablement. Teva Pharms. Int’l GmbH v. Eli Lilly & Co., No. 24-1094, slip op. at 5 (Fed. Cir. Sept. 30, 2026) (Dyk, J., dissenting from denial of rehearing en banc) (Teva, Dyk Dissent). In his view, the panel’s decision creates an unwarranted situation: a patentee whose functional antibody genus could not itself satisfy the enablement requirement can now escape that problem simply by recasting the same genus inside a method-of-use claim.

While well-reasoned, Judge Dyk’s concerns might not be so far reaching.

On October 2, 2026, two days after the Federal Circuit denied rehearing en banc in Teva, the court decided Pioneer Hi-Bred International, Inc. v. Inari Agriculture, Inc., No. 25-1287 (Fed. Cir. Oct. 2, 2026) (nonprecedential) (Pioneer, slip op.). Judge Prost, who wrote Teva, wrote Pioneer. Judge Cunningham, who joined Teva, also joined Pioneer. The third member of the panel was none other than Judge Dyk. The court affirmed a Patent Trial and Appeal Board (“the Board”) decision holding all 33 claims of Pioneer’s U.S. Patent No. 11,371,055 unpatentable for non-enablement of the claimed “dual herbicide-degrading” enzymes in crops. Eighteen of those claims were method claims, reciting the enzyme genus in methods for controlling weeds. The opinion gives no indication that Pioneer asked the court to treat the method claims separately, and Pioneer did not invoke Teva. The method claims simply fell together with the composition claims, as one would expect under Amgen Inc. v. Sanofi, 598 U.S. 594 (2023).

At first glance, the two decisions seem uncomfortable neighbors. In Teva, a method claim survived even though the panel assumed that identifying and making the full universe of qualifying antibodies would require undue experimentation. In Pioneer, the method claims fell after the court determined the patent’s enormous functionally defined enzyme genus was not enabled. If Teva really allows patentees to avoid Amgen by converting a genus claim into a method of use claim, why would that drafting move not have rescued Pioneer?

The answer lies in Teva’s unique facts and an old line of precedent that treats a well-known genus used as a component of a separate invention differently from a genus that is itself the invention. Teva, slip op. at 9. Teva stood on two pillars. The first was that anti-CGRP antagonist antibodies were already well known, a proposition the jury could find in significant part from Lilly’s own statements to the Patent Office. The second was that every humanized anti-CGRP antagonist antibody would treat headache, a proposition Lilly did not dispute on appeal. Applied to Pioneer’s patent, Teva’s first pillar fails on the patentee’s own words: its specification called the dual-function enzyme a “highly novel discovery.” Once the court found that the specification failed to enable representative claim 1 and its “novel” enzyme genus, no separate review of the weed-control method claims was needed. All claims were held invalid.

The two cases therefore share an unexpected feature. In each, a party’s own statements fixed the starting line of the enablement inquiry. In Teva, the panel used the challenger’s words to place the genus in the background art. In Pioneer, the patentee’s words placed it at the heart of the invention.

Read this way, Teva is narrower than the amici and Judge Dyk feared, but not for the reason one might first suppose. Teva’s second pillar that all antagonist antibodies treat headache, though litigated in the Patent Office years earlier, had become sturdy by the time of trial; the treatment of headache, not the antagonist antibodies themselves, was the invention. Whether the recited genus of antagonist antibodies was either known background technology or novel science was precisely where the district court and the Federal Circuit parted ways, and it is where the battles that follow Teva will be waged. Pioneer shows how that fight ends when the genus itself is the novel invention. Teva shows what happens when it is not, especially where the accused infringer has already argued the genus is well known prior art.

II. Lilly’s First Problem: The Card It Had Already Played

The story begins years before the dispute that produced the Federal Circuit’s Teva decision. CGRP is a protein found in humans. When it binds to receptors on certain cells, the cells expand and increase blood flow through blood vessels, a phenomenon associated with headache. Anti-CGRP antagonist antibodies bind to CGRP in a way that inhibits headache-associated activity. Teva, slip op. at 2. Teva’s headache patents, with a November 2006 priority date, claim the use of humanized versions of those antibodies to treat headache. The specification observed that anti-CGRP antagonist antibodies were “known in the art,” cited a product catalog offering a murine (mouse) antibody for sale, disclosed other murine examples, and disclosed prior-art humanization methods. But it disclosed only one humanized antibody, G1, the active ingredient in Teva’s Ajovy product. Id. at 3.

Between August and October 2018, Lilly filed petitions for inter partes review challenging two groups of Teva’s patents. One group, the headache patents, claimed methods of using humanized anti-CGRP antagonist antibodies to treat headache. The other group claimed the antibodies themselves. Id. at 4. In seeking to invalidate the antibody patents for obviousness, Lilly painted the technological landscape as mature and accessible. Lilly told the Board that by November 2006 anti-CGRP antagonist antibodies “were well known in the art” and that the prior art was “replete with exemplary disclosures” of them. Techniques for making such antibodies were, according to Lilly, “extensively described in the prior art.” Humanization, Lilly maintained, “was a well-established and routine procedure.” Id. (quoting J.A. 21417, 21442, 21407).

The IPR results were mixed.

The Board held the challenged antibody claims unpatentable, and the Federal Circuit affirmed. See Teva Pharms. Int’l GmbH v. Eli Lilly & Co., Nos. 2020-1747, 2020-1748, 2020-1750 (Fed. Cir. Aug. 16, 2021). But the headache treatment patents survived. The Board found that Lilly had not shown a reasonable expectation of success in treating headache with anti-CGRP antibodies, given the uncertainty at the time over whether such large molecules would need to cross the blood-brain barrier to work, and the Federal Circuit affirmed. See Eli Lilly & Co. v. Teva Pharms. Int’l GmbH, Nos. 2020-1876, 2020-1877, 2020-1878 (Fed. Cir. Aug. 16, 2021).

Teva would also be haunted by its own positions during those proceedings. In defending its antibody patents before the Patent Office, Teva had argued, as the district court later recounted in its JMOL ruling, that the obviousness of making humanized anti-CGRP antibodies did not establish that they “would necessarily be safe and effective” for use “in treating various diseases.” Teva Pharms. Int’l GmbH v. Eli Lilly & Co., No. 18-cv-12029-ADB, Dkt. 695, at 27 n.14 (D. Mass. Sept. 26, 2023) (Teva, D. Ct. Decision) (citing PTX-924 at 102–06). At trial, Teva argued that the same antibodies were “a well-known class of compounds that could be routinely made, tested, and humanized.” Id. at 44 (quoting Teva’s opposition brief). And the inventors, Teva asserted, “understood and disclosed that all humanized anti-CGRP antagonist antibodies would work in the claimed method.” Id. at 29 (quoting Teva’s opposition brief). Each party, in other words, had argued both sides of the same science, in different forums, for different purposes.

The district court and the Federal Circuit viewed that history very differently.

The district court refused to let Teva lean on the Board’s IPR antibody findings that “‘anti-CGRP antagonist antibodies were well-known in the art,’ and that a POSA could humanize them without disrupting their essential properties.” Teva, D. Ct. Decision, at 25 n.13 (quoting ECF No. 667 at 14). The critical fact, the district court noted, was that the antibody patents at issue in the IPRs did not “recite an intended therapeutic use,” and Teva had conceded as much. Id. The district court found that, although murine antibodies were disclosed generally in Teva’s patent, only one humanized antibody, G1, was disclosed, and “only humanized antibodies fall within the scope of the Asserted Claims.” Id. at 25–26. In the district court’s view, “Humanized anti-CGRP antagonist[] antibodies were a new genus, and the Patents-in-Suit were the first to disclose a method of treating headache utilizing a new genus of antibodies.” Id. at 37–38 (emphasis added).

In contrast, the Federal Circuit made no mention of Teva’s earlier conflicting positions. Instead, the court made Lilly’s earlier IPR assertions the centerpiece of its prior art discussion. A reasonable jury, the court determined, could have found the antibodies and the methods of making them well known “based on Lilly’s own statements.” It also could have found humanization routine, “again, based on Lilly’s own statements.” Teva, slip op. at 12–13. The panel did not rely exclusively on those statements; it also identified supporting testimony from Teva’s experts and from Lilly’s own expert. Id. at 13 nn.12–13. And the novelty of the humanized genus found by the district court was of little importance because murine antibodies were themselves well known and “just a routine . . . procedure away from being humanized.” Id. at 14–15.

Of course, Lilly was not legally estopped from painting a different picture of the prior art at trial, and the Federal Circuit never suggested otherwise. Inter partes review estoppel under 35 U.S.C. § 315(e)(2) reaches only grounds that were or reasonably could have been raised in the review, and those are limited to anticipation and obviousness based on patents and printed publications. See 35 U.S.C. § 311(b). Section 112 enablement and written description were never in play during the IPRs.

The district court’s JMOL decision setting aside the adverse jury verdict reveals Lilly’s yeoman efforts to recast the prior art within the analytical frameworks of enablement and written description. Before the Federal Circuit, however, Lilly could not outrun its own dogging IPR assertions that the antagonist antibody art was well known and routine. With those statements and other evidence in hand, the panel concluded that the anti-CGRP antagonist antibodies were not themselves Teva’s claimed invention. The invention was a new use of an old technology: using anti-CGRP antagonist antibodies to treat headache.

III. Lilly’s Second Problem: A Self-Fulfilling Function

To fully appreciate what became an undisputed tautology in the case, that all anti-CGRP antagonist antibodies treat headache, we must take a closer look at the patent claim language itself. Representative claim 30 of asserted U.S. Patent No. 8,586,045 recites “[a] method for reducing incidence of or treating headache in a human, comprising administering to the human an effective amount of an anti-CGRP antagonist antibody, wherein said anti-CGRP antagonist antibody is a . . . humanized monoclonal antibody.” Teva, slip op. at 3–4.

The claim thus builds in two functional propositions. First, the antibody must qualify as an antagonist: it must bind CGRP in a manner that inhibits CGRP activity. Second, administration of that antagonist must reduce the incidence of or treat headache. Judge Dyk understandably described the claims as reciting an “antagonist limitation” and a “treatment limitation.” Teva, Dyk Dissent at 2.

From the perspective of a skilled artisan reading these facts today, the second limitation almost reads like a foregone conclusion of the first. Before the patents’ 2006 priority date, it was not. Before that date, the link between CGRP and headache had already been established, and small-molecule drugs targeting the CGRP pathway had shown clinical effect. Important to the issues here, murine anti-CGRP antagonist antibodies were also known in the art. But as Teva itself put it, a skilled artisan “‘did not yet know whether the antibodies could be used to prevent headache.’” Teva, D. Ct. Decision, at 6 (quoting Teva’s opposition brief). That is the uncertainty the Board relied upon when it upheld the headache patents against Lilly’s obviousness challenge.

What closed the gap was the inventors’ own work. Led by Dr. Jorg Zeller, the team of inventors obtained a CGRP-blocking antibody from an academic laboratory, generated their own antibodies, and showed in animal models that anti-CGRP blocking antibodies “could really treat migraine.” Id. at 7. At trial, Teva’s position was that once the inventors showed G1 would treat headache, a skilled artisan “would understand that any humanized anti-CGRP antagonist antibody would.” Id. at 13 (quoting Teva’s opposition brief) (emphasis in original). The jury heard testimony that a skilled artisan could have believed the antibodies would treat headache based on what was known in November 2006 together with “the data from the animal test in the Zeller specification.” Id. at 39 n.23 (quoting Trial Tr. 15-64:6–18).

The inventive contribution therefore was demonstrating that antibody antagonism of CGRP treats headache. The antibodies antagonize the headache-inducing effect of the CGRP protein. Thus, if the antibody is a CGRP antagonist, then it treats headache. In 2006, that proposition was not a definitional truism; it had to be shown. But once shown, the functional relationship proved difficult for Lilly to attack at trial. Strictly speaking, Lilly never conceded that every antagonist treats headache; it stopped contesting that a jury could so find. By the time the case reached the Federal Circuit, the proposition had achieved tautological status.

That is not to say, however, that Lilly did not wage attacks on that tautological proposition at the district court.

There, Lilly concentrated its § 112 attack on the antagonist claim limitation. The number of candidate antibodies was very large, Lilly argued, and the specification did not tell a skilled artisan how to determine in advance which ones would antagonize CGRP. The inventors had left to others the “research assignment” of screening and testing those candidates, at a cost in time and money that Lilly characterized as undue experimentation. Teva, slip op. at 21.

Lilly also took several runs at the claims’ treatment limitation. Its principal attack was both technical and legal: Teva could not disclose a single covered antibody and then rely on testimony that a skilled artisan could identify other antibodies, could humanize them, and would find that they all treat headache. Teva, D. Ct. Decision, at 30. Lilly also argued that the specification failed to address the blood-brain-barrier skepticism the Board had found in the IPRs, an argument the district court noted but did not decide. Id. at 39 n.23. In its reply brief, Lilly pointed to evidence that Teva’s G1 antibody had failed to treat certain types of headache. Id. at 43. And it showed that its own antibody, galcanezumab, the active ingredient in its Emgality product, differed from G1 in epitope, sequence, V-gene family, CDR structure, potency, dosing, and the types of headache each was used to treat. Emgality, for example, was effective against cluster headache, while Teva was not pursuing approval to treat cluster headache with G1. Id. at 14–15, 31.

The district court credited much of Lilly’s argument, but not as proof that some members of the genus fail. It used the differences between G1 and Emgality to hold, under AbbVie Deutschland GmbH & Co. v. Janssen Biotech, Inc., 759 F.3d 1285 (Fed. Cir. 2014), that the specification described no antibody structurally similar to, and thus representative of, the accused product. Teva, D. Ct. Decision, at 33–35. On the tautology itself, the district court repeatedly acknowledged that the jury could have credited testimony “that all humanized anti-CGRP antagonist antibodies would treat headache.” Id. at 25; see also id. at 30. Antibodies binding different regions of CGRP, the jury could find, would “still accomplish the claimed function of treating headache.” Id. at 30.

By the time of the appeal, the point was no longer contested. “[A]s Lilly does not dispute,” the Federal Circuit wrote, a reasonable jury could have found that a skilled artisan would understand from the specification that all humanized anti-CGRP antagonist antibodies treat headache. Teva, slip op. at 13. When distinguishing Idenix Pharmaceuticals LLC v. Gilead Sciences Inc., 941 F.3d 1149 (Fed. Cir. 2019), the panel again called it “undisputed” that a reasonable jury could have so found, “thereby obviating any extensive screening to determine which ones do.” Teva, slip op. at 24. The cluster headache evidence survived only as a footnote: the panel held that a reasonable jury could have found Lilly’s support for it “less than clear and convincing.” Id. at 20 n.17.

Lilly’s attempted distinctions ultimately proved to be the hinge on which the Federal Circuit’s analysis turned. Lilly succeeded in demonstrating antagonist antibody heterogeneity. But it failed to break the link between antagonist antibodies and their treatment of headache.

IV. Moving the Analytical Starting Line

The Federal Circuit reviewed the district court’s JMOL decision de novo under First Circuit law, asking whether the facts and inferences, “‘viewed in the light most favorable to the verdict, point so strongly and overwhelmingly’ in Lilly’s favor ‘that a reasonable jury could not have returned the verdict.’” Teva, slip op. at 6 (quoting Acevedo-Diaz v. Aponte, 1 F.3d 62, 66 (1st Cir. 1993)). Lilly bore the burden of proving the facts underlying patent invalidity by clear and convincing evidence. Id. Enablement is a question of law, but its factual underpinnings are reviewed for substantial evidence. Id. at 21.

Neither the district court nor the Federal Circuit disputed that the universe of candidate antibodies was vast and that identifying antibodies with antagonist activity required substantial screening work. Teva’s expert testified that one could not predict from an antibody’s amino acid sequence whether it would antagonize CGRP; antibodies had to be made and tested. Teva, D. Ct. Decision, at 15–16. One of the inventors agreed that “‘you would have to evaluate each anti-CGRP antibody empirically to determine whether or not it [could] inhibit CGRP.’” Id. at 15 (quoting Trial Tr. 3-35:10–15). Obtaining a humanized antagonist required in vitro screening, in vivo animal testing, procurement of the animal antibodies (which had taken the inventors 111 days), and humanization. Id. at 15–16. Lilly’s expert put humanization alone at “many months’ worth of effort” and “maybe half a million dollars”; Teva’s expert said it might have taken several months in 2006 and could cost up to $70,000 today. Id. at 16. The district court found that the process “would collectively take months and cost at least tens of thousands of dollars per antibody.” Id. at 46–47.

The district court and Federal Circuit, however, parted ways on the question of whether that antibody production and screening work was even relevant to the presented § 112 written description and enablement issues.

The district court concluded that it was. Relying on Amgen, Baxalta Inc. v. Genentech, Inc., 81 F.4th 1362 (Fed. Cir. 2023), Wyeth & Cordis Corp. v. Abbott Laboratories, 720 F.3d 1380 (Fed. Cir. 2013), and Idenix, it held that the specification offered only a “roadmap” for a “trial and error” process. Teva, D. Ct. Decision, at 47. And it confronted the headache treatment tautology head on: “even if the jury concluded that the state of the art was generally predictable in the sense that an anti-CGRP antagonist antibodies would treat headache, that does not change the fact that . . . each potential antibody would have to be synthesized and screened for effectiveness.” Id. at 48. Thus, three years before Judge Dyk’s dissent, the district court had already staked out his position.

Like a resolute gambler, the Federal Circuit not only accepted the district court’s opening bid on the § 112 analysis but also significantly raised the bet. The panel declared that it would “assume two things in [Lilly’s] favor, solely for argument’s sake.” Teva, slip op. at 21. First, making all anti-CGRP antagonist antibodies would require screening and testing a very large number of candidates. Second, the time and expense required to make and humanize all of them would constitute “undue experimentation.” Id. at 21–22 (emphasis added).

With those assumptions in place, it would appear the panel gave Lilly a substantial headstart in proving lack of written description and enablement. But as we will see, no headstart was given at all because the panel effectively moved the starting line of the analysis.

The panel continued by observing that Lilly’s § 112 arguments “might be more persuasive” if the patent claims were to the antibody genus itself. If so, “this case would resemble Amgen,” where the patentee had laid claim to an entire antibody genus “for any and all purposes.” Id. at 22. From the panel’s perspective, Teva’s patents claimed neither the antibodies themselves nor all uses of the antibodies. They claimed “only the use of such antibodies for the different, limited purpose of treating headache.” Id. at 22–23. When the patents are read strictly in that manner, the claims permit anyone to make and use the antibodies in any way they choose, except for the treatment of headache.

And these were not novel antibodies. The trial record established that murine anti-CGRP antagonist antibodies were known in the art before the patents’ 2006 priority date. In view of these “well-known” antibodies and their “routine” humanization, the panel reasoned, the more relevant “research assignment” was “determining which humanized anti-CGRP antagonist antibodies treat headache.” Id. at 23 (emphasis in original). And in response to its own inquiry, the panel gave a passing grade to that research assignment by concluding, “That assignment was completed; the specification disclosed that all such antibodies work for that purpose.” Id. (emphasis in original).

Next, the panel addressed the scope of preliminary work required to make, screen, and humanize anti-CGRP antagonist antibodies in the first place: work that it assumed required “undue experimentation.” That work, the court concluded, was in effect unrelated to and separate from the claimed invention. The court reasoned that undertaking to find or make candidate members of the genus would, “in the context of these claims,” be “more akin to extra credit than a necessary research assignment left to others to complete.” Id. (emphasis added).

The metaphor is enlightening because it unveils precisely how the panel arrived at its decision. It did not conclude that generating candidate members of the genus would be easy. It assumed the opposite. Instead, it decided that the work of generating members of the genus was irrelevant to the claimed invention, which begins only after an anti-CGRP antagonist antibody is firmly in hand. The only research assignment left was determining whether a particular antagonist would succeed or fail in the claimed use, and a reasonable jury could find that all of them would succeed. That assignment was therefore complete. Artisans who undertake the arduous upstream prior art steps of making such antibodies, however much time and effort that requires, will be rewarded with antibodies that treat headache. The steps themselves are “extra credit.”

In reaching these conclusions, the panel did not write on a blank slate. Its analysis was fairly grounded in established Federal Circuit precedent addressing the § 112 written description and enablement requirements, which the panel observed “often rise and fall together.” Id. at 21 (quoting Ariad Pharms., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1352 (Fed. Cir. 2010) (en banc)). The panel drew on prior caselaw involving “a well-known genus that is not, itself, the invention.” Id. at 9.

For example, in Ajinomoto Co. v. International Trade Commission, 932 F.3d 1342 (Fed. Cir. 2019), the patent disclosed that enhancing the activity of a bacterium’s yddG gene increased its production of certain amino acids, and the claims covered producing those amino acids from a bacterium in which a “more potent promoter” had been substituted for the gene’s native promoter. The challenger argued that the specification did not disclose a sufficient number of representative species within the genus of “more potent promoters.” The court disagreed, citing substantial evidence that enhancing promoter activity “was well-known” and that a skilled artisan “would have been able to identify more potent promoters by employing common tools.” Id. at 1359. The invention was the discovery about enhancing the yddG gene, “not the well-known techniques” used to enhance it, especially where “the genus of more potent promoters was already well explored.” Id., quoted in Teva, slip op. at 9–10. In Teva, the panel drew the same lines. The humanized antagonist antibodies played the part of the promoters, a known class of tools defined by what they do, and the discovery that they treat headache played the part of the yddG enhancement insight. Teva, slip op. at 12–13.

The panel also relied on In re Herschler, 591 F.2d 693, 700–01 (C.C.P.A. 1979), involving the discovery that DMSO enhances skin penetration, in which a single disclosed steroid sufficed to describe the genus of “physiologically active steroidal agent[s]” carried along with it. Arguably, Herschler provided the analytical framework that the panel applied here. In Herschler, the court distinguished claims to “classes of new compounds per se or . . . processes using those newcompounds” from “claims drawn to the use of known . . . compounds in a manner auxiliary to the invention.” Teva, slip op. at 10 (quoting Herschler, 591 F.2d at 702) (emphasis added). “Were this application drawn to novel ‘steroidal agents,’” the Herschler court reasoned, “a different question would be posed.” Id. (quoting Herschler, 591 F.2d at 701). To fit within Herschler, the panel had to get past the district court’s finding that humanized anti-CGRP antagonist antibodies were “a new genus.” As discussed above, it did so by treating the humanization of well-known murine antibodies as routine.

The oldest of the cases relied upon by the panel is perhaps the most revealing, because it involved the same problem Lilly raised and resolved it in a way similar to Teva’s analysis. In In re Fuetterer, 319 F.2d 259 (C.C.P.A. 1963), the claim was to a tire-tread composition that included “an inorganic salt that is capable of holding a mixture . . . in colloidal suspension in water,” and the specification disclosed only four such salts. The Patent Office rejected the claim under the first paragraph of § 112, reasoning that “not all inorganic salts are capable of performing” the specified function and that “one skilled in the art would not know offhand which inorganic salts are capable of so functioning.” Id. at 260–61, 265, quoted in Teva, slip op. at 11 & n.10. That was Lilly’s argument in 1963 dress: a functionally defined component whose qualifying members could be identified only by testing.

Judge Rich’s plurality opinion reversed. “Appellant’s invention is the combination claimed and not the discovery that certain inorganic salts have colloid suspending properties. We see nothing in patent law which requires appellant to discover which of all those salts have such properties and which will function properly in his combination.” Id. at 265, quoted in Teva, slip op. at 11–12. The applicant, the court added, “has not invented, and is not claiming, colloid suspending agents.” Id. at 266, quoted in Teva, slip op. at 12. “Extra credit,” it seems, has a 1963 ancestor.

Fuetterer also anticipated the problem presented by a redesigned accused product. “If others in the future discover what inorganic salts additional to those enumerated do have such properties,” Judge Rich wrote, the applicant “will have no control over them per se,” but “his claims should not be so restricted that they can be avoided merely by using some inorganic salt not named by appellant in his disclosure.” Id. at 265, quoted in Teva, slip op. at 12. Substitute Lilly’s Emgality for the unnamed salt, and that sentence is the panel’s answer to Lilly’s argument that Teva disclosed nothing resembling Emgality. The answer works, however, only if every member of the class actually performs in the claimed combination. In Fuetterer, that premise rested on the specification’s own assurance that “any inorganic salt which has such properties is usable in his combination,” accepted on an ex parte appeal with no adversary to test it. Id.

Teva had considerably more: a jury-supported finding, which Lilly did not dispute on appeal, that all humanized anti-CGRP antagonist antibodies treat headache. Teva, slip op. at 13, 24. In that respect, Teva rests on firmer ground than the precedent it invoked. Its doctrinal footing is another matter. The panel acknowledged that the Fuetterer plurality is not binding, but noted that Herschler, which is binding, relied heavily on it. Teva, slip op. at 11 n.9. A sixty-year-old plurality opinion about tire treads now carries a good part of the weight of a method-of-treatment decision issued in the shadow of Amgen.

Stripped of its chemistry, the panel’s approach in Teva is a familiar one to patent lawyers. Software patents routinely recite “a processor” or “a general-purpose computer” without teaching anyone how to build one, because “a patent need not teach, and preferably omits, what is well known in the art.” Hybritech Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1384 (Fed. Cir. 1986). The computer is a tool, and the invention lies in what the patent’s claim does with it. In Teva, Lilly’s IPR arguments about the “obviousness” of anti-CGRP antagonist antibodies helped permit the panel to treat those antibodies the same way. The antibody was the general-purpose hardware; the discovery that it treats headache was the program.

That principle does have limits, however, and the Federal Circuit has stated it plainly. Reliance on the knowledge of the art “is merely a rule of supplementation, not a substitute for a basic enabling disclosure.” Genentech, Inc. v. Novo Nordisk A/S, 108 F.3d 1361, 1366 (Fed. Cir. 1997). “It is the specification, not the knowledge of one skilled in the art, that must supply the novel aspects of an invention.” Auto. Techs. Int’l, Inc. v. BMW of N. Am., Inc., 501 F.3d 1274, 1283 (Fed. Cir. 2007). A patentee may borrow the computer from the art. It may not borrow the invention.

The analogy goes only so far, and it strains precisely where the Teva case was fought. “A computer” is a category of hardware; “an anti-CGRP antagonist antibody” is a category defined by a function. Any general-purpose computer will run the program, and an engineer can tell from its specifications whether it qualifies. An antibody offers no such shortcut. Its function cannot be read from its code: Teva’s own expert conceded that no one could tell from an antibody’s amino acid sequence whether it would antagonize CGRP. Teva, D. Ct. Decision, at 15. The true software analogue, then, is not “a general-purpose computer” but “a processor that performs the claimed function,” where identifying which processors qualify requires building and testing them. In software, that kind of claim must disclose how the function is achieved.

The panel in Teva had an answer: the component in Fuetterer was functionally defined as well, an inorganic salt “capable of holding a mixture . . . in colloidal suspension.” Teva, slip op. at 11. Whether a functionally defined genus can still be borrowed from the art as a tool after Amgen, or whether it is part of the invention the specification itself must supply, is exactly the question Judge Dyk would press.

V. Judge Dyk’s Objection: A Claim Limitation Cannot Be Ignored

Judge Dyk’s dissent from the denial of rehearing en banc turns the panel’s analysis on its head. Claim 30 contains two functional limitations: (1) an antagonist antibody and (2) treatment of headache. The panel had assumed that the experimentation required to satisfy the first limitation would be undue but held it irrelevant because the invention lay in the second. Judge Dyk saw it the other way around. The second limitation, he wrote, “adds nothing of substance to the first limitation,” because, as the panel recognized, any antibody satisfying the first limitation would be effective for treating headache. Teva, Dyk Dissent at 2.

The panel and Judge Dyk thus drew opposite conclusions from the same undisputed facts. For the panel, the proposition that every antagonist treats headache meant the inventors’ research assignment was complete. For Judge Dyk, it meant the treatment limitation was merely a reflection of the antibody limitation. What remains in substance is a claim to the antagonist genus itself, precisely what Amgen forbids without full-scope enablement. Neither reading can be refuted from the words of the claim alone; the standoff turns on which limitation embodies the invention. And the stakes were not in doubt. The undisputed record showed that the first limitation, identifying and humanizing antibodies with antagonist properties, required benchtop and animal testing at substantial cost, measured in months and in tens of thousands of dollars per antibody. Id. at 2–3.

Judge Dyk next challenged the panel’s reliance on the “well-known” antagonist antibody prior art and the asserted narrowness of the treatment claims. As to the first, he answered that it was only well known that the genus exists; it was not well known which antibody species had antagonist qualities. A genus may be well known while its specific embodiments are not enabled by knowledge of the genus. Id. at 3 (citing Wyeth, 720 F.3d at 1383–86). As to the second, he rejected the premise that the method claim was narrower than a claim to the antibodies at all. The record suggested no use for the antibodies other than treating headache, so “the practical scope of the method claim is equal to the scope of the compound.” Id. at 4. And even if the method claim were narrower, “the compound still must be enabled since the compound is indisputably a claim limitation.” Id. His conclusion was categorical: “Claiming a method of use that depends on a recited component cannot excuse the enablement requirement of the component.” Id.

The criticism is difficult to dismiss. Section 112 requires enablement of “‘the full scope of the invention as defined by its claims.’” Id. (quoting Amgen, 598 U.S. at 610). A method claim cannot ordinarily be practiced unless its recited components can be obtained. If the claim covers a vast functionally defined antibody genus, and discovering which candidates satisfy that functional limitation requires undue experimentation, the fact that every successful candidate subsequently performs the claimed treatment does not erase the experimentation required to obtain it.

Summarizing Judge Dyk’s dissent more simply: one cannot declare the race easy by moving the starting line past the hardest hill. That concern animated the industry amici as well. Johnson & Johnson and Nagra pointed to post-decision guidance urging practitioners to use “creative claim drafting to skirt the disclosure requirements of Amgen.” Merck and Ipsen warned that the decision “risks foreclosing entire fields of independent scientific development.” Amgen and Sanofi argued that “differing outcomes” between Teva and Amgen would create uncertainty for decisionmakers, the Patent Office, and industry. Id. at 5.

Judge Dyk observed, pointedly, that these were amici “representing the pharmaceutical industry, not known for urging onerous requirements for patentability.” Id. He then posed the obvious hypothetical. The principal use of the PCSK9-blocking antibodies in Amgen was to lower cholesterol. Had Amgen simply claimed a cholesterol-lowering method using those antibodies, its claim would not have been meaningfully narrowed. Yet under the panel’s reasoning, Amgen could have evaded the enablement requirement by doing exactly that. Id.

The Federal Circuit judges reviewing the petition for en banc rehearing were unmoved. The poll failed, and no other judge joined the dissent or wrote separately. Teva Pharms. Int’l GmbH v. Eli Lilly & Co., No. 24-1094, slip op. at 2 & n.1 (Fed. Cir. Sept. 30, 2026) (order) (noting that Judge Newman did not participate).

What the dissent does not do is engage Herschler and Fuetterer. That omission frames the real doctrinal question. If Judge Dyk is right that a recited component must always be enabled across its full scope, it is hard to see how Fuetterer’s salts survive, since the applicant there was expressly excused from discovering which salts worked. If the panel is right, the Herschler line of cases remains good law after Amgen, and the question in each case becomes whether the recited genus is a known tool or the invention itself. The en banc court declined to choose. Until it does, Herschler and Teva will define the battleground of future disputes. Pioneer supplies an unusually timely illustration.

VI. Two Days Later: Pioneer and its Non-Enabled Method Claims

Pioneer involved a different technology, but the structural resemblance to Teva is hard to overlook. The ’055 patent concerns enzymes with what the parties called a “dual herbicide-degrading function”: they make crops resistant to two different classes of herbicides, phenoxy auxins such as 2,4-D and pyridyloxyacetate auxins such as triclopyr and fluroxypyr. Pioneer, slip op. at 2. Representative claim 1 is directed to a transgenic plant cell containing a recombinant polynucleotide encoding an AAD-12 protein. The claim defines the enzyme genus by function, an activity that “enzymatically degrades a phenoxy auxin herbicide and a pyridyloxy auxin herbicide,” and by structure, “at least 85% sequence identity with SEQ ID NO: 2” and “an AAD-12 motif.” Id.; ’055 patent claim 1. In short, a plant modified to express the enzyme has the means to “detox” itself whenever confronted with these herbicides.

The claim set reaches well beyond plant cells. Claims 4 through 11 and 30 through 31 cover plants and their parts and progeny; claim 28 covers seed. Claims 12 through 27, 29, and 33 are methods of controlling weeds. Claim 12 applies a phenoxy auxin or pyridyloxy auxin herbicide to a field containing the plant of claim 4. Claim 29 applies such an herbicide and plants the claimed seed within 14 days. Claim 33 recites the full 85 percent sequence identity genus and AAD-12 motif and requires applying a pyridyloxy auxin herbicide to a field planted with seed containing the encoding polynucleotide. ’055 patent claims 12, 29, 33.

The claimed genus was breathtakingly large. Inari calculated that it included on the order of 10¹⁰⁶ species, “greater than the number of atoms in the universe,” while the specification disclosed only two: SEQ ID NO: 2 and SEQ ID NO: 4, which shares 99.3 percent sequence identity with it. Pioneer, slip op. at 2. Pioneer did not dispute the size of the genus. Even applying the additional guidance in Figure 2 of the patent, which identifies residues that can more likely be varied without changing function, the genus still included 1.23 × 10⁶⁶ species. Id. at 3.

Numerical breadth alone, however, does not capture why Pioneer lost. Pioneer argued that the 85 percent sequence identity limitation was a “common quality running throughout the claimed genus” and that this feature alone ensured that “90+%” of the genus would exhibit the dual herbicide-degrading function. Id. That was a proposition about genus membership: that satisfying the structural limitations would reliably produce the claimed function. It was, in other words, an answer to the same upstream question Lilly had pressed in Teva, and that Teva’s own expert had conceded could not be answered from sequence alone.

Pioneer’s own evidence defeated it. With its patent owner response, Pioneer submitted experimental data, generated more than a decade after the priority date, testing eight enzymes that met the claimed structural limitations. Only two exhibited the dual herbicide-degrading function, “and weakly at that, with ‘only 1% and 2% . . . triclopyr activity.’” Id. (quoting J.A. 651). Inari turned those results against Pioneer, arguing that they showed the patent’s guidance was not reliably correlated with the claimed function. The Board agreed. Id.

On appeal, Pioneer argued that the Board could not rely on post-priority data at all. The court found the argument likely forfeited and, in any event, wrong. In Amgen Inc. v. Sanofi, 872 F.3d 1367, 1375 (Fed. Cir. 2017), the court had held it error to exclude post-priority evidence relevant to whether claims were enabled as of the priority date. The court distinguished In re Hogan, 559 F.2d 595 (C.C.P.A. 1977), and In re Entresto, 125 F.4th 1090 (Fed. Cir. 2025), because those cases rejected the use of later-developed technology that did not exist at the priority date to show nonenablement. Pioneer’s tested variants, by contrast, could all have been created at the priority date, even though Pioneer tested them later. Pioneer, slip op. at 4–5.

On the merits, the court quoted the Supreme Court’s admonition that “[t]he more one claims, the more one must enable.” Id. at 6 (quoting Amgen, 598 U.S. at 610). The massive disparity between the amount claimed and the amount disclosed supported the Board’s finding. But even setting the numbers aside, the predictions of Pioneer’s expert that the disclosed structures correlated with the claimed function were contradicted by Pioneer’s own data. Two of the tested variants differed by only six amino acids, none at positions the specification identified as presumptively important, yet one exhibited the dual function and the other did not. Substantial evidence supported the Board’s finding that the specification and working examples “fail to provide guideposts that would have illuminated a path toward embodiments at the 85% sequence identity level.” Id. at 6–7 (quoting J.A. 30).

Inari thus challenged the size of the genus haystack and succeeded by showing that Pioneer’s own map through that haystack led mostly to dead ends.

Significantly, the court’s analysis addressed only representative claim 1 and did not discuss the method claims at all. It did not need to. The Board had ruled before Teva was decided, and although the appeal was argued on September 14, 2026, months after Teva issued, id. at 6 n.1, Pioneer did not invoke Teva on behalf of its weed-control claims. Pioneer is nonprecedential, and it should not be read as a rejection of a Teva argument that was never made. It is better understood as demonstrating that, in the typical case, Amgen’s default rule still applies: when a method claim incorporates a nonenabled functional genus, the method claim falls with the genus.

VII. Why Teva Could Not Have Saved Pioneer’s Method Claims

The more interesting question is whether Teva would have made a difference had Pioneer argued that the method claims are valid even while reciting a non-enabled genus. Applying Teva’s framework, it appears the argument would have failed.

First, the starting point of Teva’s reasoning depends on finding a “well-known genus that is not, itself, the invention.” Teva, slip op. at 9. The ’055 patent answers that question itself. Its specification explains that the invention “relates in part to the identification of an enzyme that is not only able to degrade 2,4-D, but also surprisingly possesses novel properties,” that “[n]o α-ketoglutarate-dependent dioxygenase enzyme has previously been reported to have the ability to degrade herbicides of both the phenoxyacetate and pyridyloxyacetates auxin herbicides,” and that “[t]his highly novel discovery is the basis of significant herbicide-tolerant crop (HTC) trait and selectable marker opportunities.” ’055 patent col. 4 ll. 44–58 (emphasis added). The specification adds that “[h]eretofore, there was no expectation or suggestion that a plant with both of these advantageous properties could be produced by the introduction of a single gene.” Id. col. 4 ll. 6–9.

That places Pioneer’s weed-control claims beyond the reach of Herschler as applied in Teva. The weed-control claims are “‘processes using those new compounds,’” not “‘the use of known . . . compounds in a manner auxiliary to the invention.’” Teva, slip op. at 10 (quoting Herschler, 591 F.2d at 702). For such claims, Herschler says, “‘a different question would be posed.’” Id. (quoting Herschler, 591 F.2d at 701). That different question is the one Amgen answers. Put in the terms of Automotive Technologies, the dual-function enzyme was the novel aspect of Pioneer’s invention, and it was the specification, not the knowledge of the art, that had to supply it.

Second, in Teva, the panel could treat the upstream work of finding antagonist antibodies as “extra credit” because the genus was known and making its members was routine. In Pioneer, the upstream work was itself the object of the invention. Determining which of the 10¹⁰⁶ structurally qualifying sequences actually perform the dual herbicide-degrading function was not a detour on the way to controlling weeds. It was the only way to know whether a candidate belonged to the claimed genus at all, and Pioneer’s own data showed that the patent’s guideposts did not lead there. Unlike the antibodies in Teva, making a structurally qualifying candidate does not ensure that it has the dual function, let alone that it will control weeds. That is a research assignment Amgen forbids a patentee to leave to others, and one Tevadid not purport to excuse.

Third, Pioneer’s method claims would have run into a rule that Teva itself acknowledged. A method claim that recites a compound defined only by the very function the method performs is, in substance, a claim to the compound. In Rochester, the court called the patentee’s attempt to distinguish such a method claim from a compound claim a “semantic distinction without a difference.” Univ. of Rochester v. G.D. Searle & Co., 358 F.3d 916, 926 (Fed. Cir. 2004), quoted in Teva, slip op. at 16. The Teva panel accepted the rule but held that Teva’s claims fell outside it. The claims in Rochester and Ariadwere, in effect, “methods of doing X using ‘something that does X.’” Teva, slip op. at 17. Teva’s claims were not. Antagonizing CGRP and treating headache are related, but they are different things, and the record showed that as of the 2006 priority date a skilled artisan could not reasonably have expected the first to produce the second. That gap is why Teva’s method patents survived Lilly’s obviousness challenge, and why the panel could treat the treatment limitation as the invention rather than a restatement of the antibody’s function.

The limitations in Pioneer’s method claims have no such separation. The enzyme is defined by its ability to degrade phenoxy and pyridyloxy auxin herbicides. The method controls weeds by applying those same herbicides to a field whose crops carry the enzyme. The crop survives because the enzyme does what defines it; the weeds die because they lack it. The method is simply the enzyme’s defining function put to work in a field, a method of doing X using something that does X. Judge Dyk’s “practical scope” objection, that a method claim covering a compound’s only use is no narrower than a claim to the compound itself, applies to Pioneer’s method claims with full force.

Teva’s treatment of Rochester points the same way in another respect. The panel distinguished Rochester in part because the compounds used in the claimed methods there were neither adequately disclosed nor known in the art. Teva, slip op. at 16–17. Pioneer disclosed two nearly identical sequences. The rest of its genus was unknown to anyone.

And the only two tested variants that exhibited the dual function showed only 1 percent and 2 percent triclopyr activity. Pioneer, slip op. at 3. Whether enzymes that weak would protect a crop from a pyridyloxy auxin herbicide applied at field rates, as claim 33 requires, is a question no tribunal needed to decide. But it suggests that the clean, undisputed bridge from genus membership to effective claimed use that sustained the panel’s decision in Teva would have been “a bridge too far” in Pioneer.

The emerging reconciliation of Teva with Amgen is that Teva is heavily fact dependent. Where the recited genus is established in the art rather than contributed by the inventor, where methods for obtaining its members are part of ordinary technical knowledge, and where the record supports a finding that qualifying members uniformly perform a newly discovered use, the patentee may not need to identify or make every species before claiming that use. That is considerably narrower than saying a patentee may take any nonenabled functional genus, add “a method of” to the preamble, and leisurely walk around Amgen. When the genus is the invention, a method formulation will not rescue the claim.

The remaining doctrinal tension should not be minimized. Teva never supplies a test for determining when a component genus is sufficiently “known” to be treated as background technology despite the difficulty of identifying its members, and the case itself shows how much turns on that precise answer. The district court looked at the claimed genus of humanized antagonists and found it new. The Federal Circuit looked one step back, at the prior art murine antibodies and the routine humanization that converts them, and found it known. Neither court was wrong about the facts it chose to rely upon. But their choices dictated the starting lines of their enablement inquiries.

Future courts will have to decide which of several questions controls: whether the genus was known in concept or in its members, whether representative species were available in the art, whether methods for generating candidates were routine, whether membership in the genus could be predicted without testing, whether the function defining the genus was itself the discovery, and whether the claimed use meaningfully narrows the scope of the patent monopoly. Teva touches each of these without saying which is decisive. Judge Dyk’s dissent lives in between those gaps: knowing that a genus exists is not the same as knowing which of its members have the required property. But uncertainty about where that particular line falls is far different from establishing the broad claim drafting loophole feared by Judge Dyk and the amici.

VIII. What Remains of the Amici’s Fears

The amici’s concern was understandable. Amgen had recently reaffirmed a demanding full-scope enablement requirement for broad functional antibody claims. Teva then appeared to permit a patentee to avoid that requirement by claiming a particular use of the same type of functional genus. From the perspective of pharmaceutical research, the fear was that early patentees could lock up downstream therapeutic programs without having taught the industry how to obtain the full range of compounds covered by their claims.

But Teva is a unicorn. Its record was unusually favorable to the patentee. The challenger had previously characterized the relevant genus as well known, methods of making the antibodies as extensively described, and humanization as routine. The specification disclosed murine antibodies, a commercially available example, and humanization techniques. The proposition that every qualifying humanized antagonist would treat headache went undisputed on appeal. The trial record was built before the Supreme Court decided Amgen. Teva, D. Ct. Decision, at 50 n.27. And the case arrived on appeal from a judgment that set aside a jury verdict, with the challenger bearing a clear-and-convincing burden. Each of those facts worked toward the panel’s conclusion that screening for undisclosed species was merely “extra credit.”

The comfort Pioneer offers is real but should be stated precisely. Challengers should not expect to successfully defeat Teva-style claims by attacking the functional bridge between genus and use once the specification has sufficiently supplied it. When the patentee’s contribution is the discovery that a known class of compounds achieves a therapeutic effect, the bridge is the invention, and the patent will usually prove it. The contest will instead be fought over the first pillar.

Judge Dyk’s Amgen hypothetical illustrates the point rather than refuting it. Amgen’s contribution was the antibodies themselves, the genus that bound PCSK9’s sweet spot and blocked it; that genus was not a known tool waiting for a new use. Under a strict reading of Teva, a cholesterol-lowering method claim built on that genus would face the same question Pioneer answered. The exposure the amici fear lies in a narrower setting: method patents filed after a target and compounds acting on it had already been explored in the art, as murine anti-CGRP antagonist antibodies had been by 2006, and when one could purchase such an antibody from a product catalog. Teva, slip op. at 3.

None of this requires overturning Teva. The opinion will remain precedential, and its “extra credit” language will be quoted. But a future panel applying that precedent would need to find the same factual conditions Judge Prost emphasized when writing the decision. If it cannot, the court can simply say: this case is not Teva. This is Amgen, Baxalta, or Idenix and, as Pioneer illustrates, the method of use claims rise or fall together with the composition claims.

IX. Conclusion

Teva arrived in legal commentary and on social media as a doctrinal shock: a Federal Circuit panel had apparently invented a way around Amgen and called the missing enablement “extra credit.” Read in isolation, the opinion invites that reaction. Read alongside the full record, it tells a different story. Years before trial, Lilly told the Patent Office that anti-CGRP antagonist antibodies were well known and routinely made, and it won. The Federal Circuit later read those same words back to Lilly, and they turned a genus that biology would ordinarily treat as anything but fungible into something closer to a general-purpose computer: a tool, not an invention.

Patent litigation has a long memory. Words offered to win one proceeding rarely stay put; they wait quietly in the record for the next proceeding. The amici feared thatTeva had handed patentees a key to every functional genus. But the key was cut to the unique record of a single case, and Pioneer shows how few doors it opens. There, the patentee’s own words, a “highly novel discovery,” place the enzyme beyond Teva’s reach. In one case, the challenger’s words made the genus a tool. In the other, the patentee’s words made it the invention. The surest protection against the outcome in Teva may never have been in the case law. It was in the record itself.