The question in this patent infringement action is whether a protein, formed through recombinant DNA technology, infringes, under the doctrine of equivalents, any of three patents: a patent directed to a natural protein extracted from certain human cancer cells; a patent directed to the materials needed to produce the natural protein through recombinant DNA technology, i.e., the DNA sequence encoding the protein, the expression vector containing the sequence, and the microorganism or cell culture capable of expressing the protein; or a patent directed to the process of producing the natural protein through recombinant DNA technology. The United States District Court for the District of Delaware found in favor of the patent owners/licensees (and their agent), plaintiffs Genentech, Inc. (Genentech), Innovi N.V. (Innovi), and Leuven Research & Development VZW (Leuven), holding that the Genetics defendants, to wit, Genetics Institute, Inc. (Institute) and Genetics Manufacturing, Inc. (GI Manufacturing), infringed under the doctrine of equivalents U.S. Patent Nos. 4,752,603 (the '603 patent), 4,766,075 (the '075 patent), and 4,853,330 (the '330 patent). That judgment was entered by the court on April 6, 1990 in consolidated Civil Action Nos. 88-330 and 89-407 following a jury trial, and became final on July 15, 1992 when the court denied defendants' motions for judgment as a matter of law (JMOL)1 or, in the alternative, a new trial. Genentech Inc. v. Wellcome Foundation Ltd., 798 F. Supp. 213, 24 USPQ2d 1782 (D.Del.1992). The Genetics defendants appeal. We find that the judgment of the trial court is not sustainable under the law, and reverse.

The protein tissue plasminogen activator (t-PA) plays an important role in the dissolution of fibrin clots in the human body. The body forms such clots typically to breach a rupture in a blood vessel. When they are no longer needed, they are dissolved through the action of plasmin, an enzyme which binds to the fibrin and severs the bonds between the fibrin molecules. Since plasmin circulates through the blood in an inactive form called plasminogen, a mechanism must be provided to activate the plasminogen and convert it to plasmin when a clot is targeted for dissolution by the body. The protein t-PA serves as that mechanism.

Unfortunately, a pathological clot known as a 'thrombus' sometimes forms in intact vessels and causes life-threatening conditions. When a thrombus occurs, the normal amount of t-PA circulating in the body may not be effective to produce plasmin fast enough to dissolve the clot, and avoid the risk of heart muscle damage or death. An additional dosage of a material which activates the plasminogen is often necessary to dissolve the clot rapidly. Several materials, such as natural t-PA extracted from human cells, streptokinase, or urokinase, were known to perform this function, although imperfectly, either because, in the case of streptokinase and urokinase, of undesirable side effects and low affinity to fibrin, and in the case of natural t-PA, the inability to derive clinically effective volumes from known sources.

Plaintiff Leuven then set to work to find a way to produce natural t-PA in a commercially useful way, i.e., in sufficient quantities and at a sufficient level of purity and effectiveness to meet commercial demands. This task was assigned to three of Leuven's scientists--Drs. Collen, Rijken, and Matsuo. They discovered that a commercially useful quantity and purity of natural t-PA could be produced from human melanoma cell cultures.

This discovery is the subject of the '603 patent, the sole independent claim of which reads:

1. Human plasminogen activator, having thrombolytic properties, immunologically distinct from urokinase and having a specific activity of about 500,000 IU/mg. using the WHO First International Reference Preparation of t-PA (tissue plasminogen activator) as assay standard or a specific activity of about 90,000 IU/mg. using the WHO First International Reference Preparation of urokinase as assay standard.

Meanwhile, plaintiff Genentech set about pursuing the same objective, a commercially useful process for producing natural t-PA, but by a different route--recombinant DNA technology.2 That task was assigned to four Genentech scientists--Drs. Goeddel, Kohr, Vehar, and Pennica. They ultimately discovered such a process as well as the intermediate products used in the process, i.e., the DNA sequence encoding human t-PA, the expression vector containing that sequence, and the microorganism or cell culture capable of expressing human t-PA using that vector.

The intermediate products are the subject of the claims of the '075 patent, of which claims 1, 3, and 8 are representative:

1. A DNA isolate consisting essentially of a DNA sequence encoding human tissue plasminogen activator.

* * * * * *

3. A recombinant expression vector containing a DNA sequence encoding human tissue plasminogen activator, wherein the vector is capable of expressing human tissue plasminogen activator in a transformed microorganism or cell culture.

* * * * * *

8. A cell culture capable of expressing human tissue plasminogen activator, obtained by transforming a mammalian cell line with a vector according to claim 3.

The process itself is the subject of the claims of the '330 patent, of which claims 1, 8, and 12 are representative:

1. A process which comprises expressing a DNA sequence encoding human tissue plasminogen activator in a recombinant host cell, said recombinant host cell being a microorganism or cell culture transformed with an expression vector containing said DNA sequence.

* * * * * *

8. A process for producing recombinant human tissue plasminogen activator comprising:

(a) growing recombinant cells in a growth medium, said cells being a microorganism or cell culture transformed with an expression vector containing DNA encoding human tissue plasminogen activator; and

(b) simultaneously expressing said DNA, thereby producing recombinant human tissue plasminogen activator.

* * * * * *

12. A process for producing recombinant human tissue plasminogen activator comprising:

(a) transforming a microorganism or cell culture with a replicable vector containing DNA encoding human tissue plasminogen activator; and(b) expressing said DNA in said transformed microorganism or cell culture.

The plaintiffs in this action consist of Leuven, the owner of the '603 patent; Genentech, the owner of the '075 and '330 patents, and the exclusive licensee of the '603 patent; and Innovi, Leuven's agent to assist it in licensing its technology rights. They initiated this action on June 21, 1988, the day the '603 patent issued, alleging infringement of the '603 patent, and subsequently amended their complaint after the '075 patent issued on August 23, 1988 to allege infringement of that patent. When the '330 patent issued on August 1, 1989, plaintiffs initiated a separate action against defendants alleging infringement of that patent; that action was subsequently consolidated with the other.

The original defendants consisted of the Genetics defendants--Institute and GI Manufacturing--as well as the Wellcome defendants--The Wellcome Foundation Ltd. (Foundation); Wellcome Biotechnology Ltd. (Biotechnology); Burroughs Wellcome Co. (Burroughs); BW Manufacturing, Inc. (BW Manufacturing); and WelGen Manufacturing, Inc. (WelGen).3 GI Manufacturing is a wholly-owned subsidiary of Institute. BW Manufacturing is a wholly-owned subsidiary of Burroughs, which in turn is a wholly-owned subsidiary of Foundation. GI Manufacturing and BW Manufacturing jointly own WelGen. That entity was, at the time this action was initiated, admittedly constructing a facility in the United States for the commercial production of t-PA.

According to plaintiffs' allegations, the Genetics and Wellcome defendants acted in concert to make, use, and import into the United States natural t-PA, or variants of natural t-PA, produced through recombinant DNA technology that infringe the patents-in-suit. There are two accused products--met-t-PA and FE1X--both of which are structurally distinct from natural t-PA.4 The Genetics defendants are alleged to have manufactured in the United States the FE1X product for commercial purposes. The Wellcome defendants are alleged to have manufactured in the United Kingdom the met-t-PA product, and imported that product into the United States for commercial purposes.

In response to plaintiffs' allegations, defendants denied infringement and asserted the affirmative defenses that the three patents-in-suit were invalid, and unenforceable due to inequitable and fraudulent conduct during prosecution. Defendants also asserted counterclaims alleging that plaintiffs' procurement and enforcement of the three patents-in-suit against them constituted an antitrust violation and unfair competition.

The parties then filed cross-motions for summary judgment on the infringement question in relation to the '603 and '075 patents. The '330 patent was not the subject of those motions. Shortly before trial was to commence, on March 8, 1990, the court granted defendants' motions in part. Specifically, it found that the accused products did not literally infringe the '603 and '075 patents, but it reserved the doctrine of equivalents issue for trial. Genentech Inc. v. The Wellcome Foundation Ltd., 14 USPQ2d 1363, 1990 WL 69187 (D.Del.1990). In concluding that defendants' products did not literally infringe the '603 and '075 patents, the court focused on the "human plasminogen activator" limitation recited in the '603 claims and the "human tissue plasminogen activator" limitation appearing in the '075 claims. It interpreted these phrases to mean the full length amino acid sequence of human t-PA plus any "naturally-occurring allelic variant" thereof. Id. at 1369. Since neither of the accused products contains the full length sequence of natural t-PA or any naturally-occurring variant thereof, the court concluded they did not literally meet that limitation. Id. at 1369-70.

The court also focussed on the "specific activity of about 500,000 IU/mg." limitation recited in the '603 claims. Although that limitation is not expressly recited in the '075 claims, the court found it was implicit in those claims. Id. at 1368. Based on representations made to the United States Patent and Trademark Office (PTO) during the '603 prosecution, the court interpreted that phrase to mean something "significantly above" a specific activity of about 266,000 IU/mg., to distinguish the '603 claims from prior work of one of the '603 inventors, Dr. Rijken, in which natural t-PA with a specific activity of 266,000 IU/mg. had been isolated. Id. at 1368. It found that that limitation was likewise not literally met by either of the accused products, FE1X, because of plaintiff's failure to prove the specific activity level of that product, and met-t-PA, because the proven specific activity level was not "about 500,000 IU/mg.". Id. at 1369-70.

Subsequently, on March 15, 1990, the court commenced a jury trial on the doctrine of equivalents issue. After 15 days of testimony, the trial court instructed the jury. Although the court issued general instructions on the issues of claim construction, the doctrine of equivalents, and prosecution history estoppel, the court refused defendants' request that the court instruct the jury on the construction and interpretation of the claims it had previously utilized in resolving the literal infringement issue.

After deliberating for two hours and forty-eight minutes, the jury returned special verdicts finding that (1) the accused product manufactured by the Wellcome defendants--met-t-PA--infringed under the doctrine of equivalents the '603 and '330 patents; (2) the accused product manufactured by the Genetics defendants--FE1X--infringed under the doctrine of equivalents the '603, '075, and '330 patents;5 and (3) all three patents were not proved invalid or unenforceable. In addition, the jury determined that defendants had not shown plaintiffs committed antitrust violations or unfair competition. On April 6, 1990, the court entered judgment in accordance with the jury's special verdicts.

On April 20, 1990, defendants filed their motions for JMOL or, in the alternative, for a new trial. Some two years later, in the decision that gave rise to this appeal, the trial court denied those motions. The Genetics defendants then filed this appeal. One of the Wellcome defendants--WelGen, the joint venture--also appealed; however, on October 22, 1992, it dismissed its appeal with prejudice. None of the other Wellcome defendants appealed. While this appeal was pending, the Wellcome defendants announced their intention to discontinue development of a t-PA product. Thus the issue of infringement by met-t-PA is not involved in this appeal. Nor did plaintiffs/appellees cross-appeal the partial summary judgment against them on the literal infringement question. The only issues now on appeal then are whether FE1X infringes one or more of the '603, '075, or '330 patents under the doctrine of equivalents.

The Genetics defendants state three separate grounds upon which they assert that the judgment of the trial court should be reversed. First, there was a lack of substantial evidence showing that FE1X met under the doctrine of equivalents the specific activity limitation expressly recited in the '603 claims and implicit in the claims of the other two patents-in-suit. Second, there was a failure to provide particularized testimony sufficient to support the equivalence finding in relation to the "human tissue plasminogen activator" limitation recited in the '075 and '330 claims as required by Malta v. Schulmerich Carillons, Inc., 952 F.2d 1320, 21 USPQ2d 1161 (Fed.Cir.1991), cert. denied, --- U.S. ----, 112 S. Ct. 2942, 119 L. Ed. 2d 566 (1992). Third, there was a complete absence of proof of any involvement by GI Manufacturing in infringing activity.

In the alternative, the Genetics defendants state four separate grounds upon which they assert they are entitled to a new trial. First, the jury was not provided sufficient guidance--including an instruction informing it of the construction and interpretation of the '603 and '075 claims the court had adopted in resolving on summary judgment the literal infringement question in relation to those patents--with which to fairly resolve the doctrine of equivalents issue. Second, the jury was not advised of the adverse consequences to defendants of a finding of infringement. Third, relevant portions of the prosecution histories of the patents-in-suit--relating to the treatment received in the British courts by the British counterparts of the patents-in-suit--were improperly excluded. Fourth, the verdict is against the great weight of the evidence and would stifle research and development in the field of biotechnology.

Before addressing these points, it is necessary to resolve three threshold issues of claim construction or interpretation.6 First, whether, as the Genetics defendants assert, the specific activity limitation appearing in the '603 claims is implicit in the '075 and '330 claims. Second, whether, as the Genetics defendants further assert, the 500,000 figure appearing as part of that limitation means 500,000 IU/mg. as measured using the bovine fibrin plate assay. Third, what is the literal meaning of the phrase "human tissue plasminogen activator" appearing in the '075 and '330 claims. These issues are issues of law--they are classic issues of claim construction, that is, what do the claims mean. See North American Vaccine v. American Cyanamid Co., 7 F.3d 1571, 1575, 28 USPQ2d 1333, 1336 (Fed.Cir.1993), cert. denied, --- U.S. ----, 114 S. Ct. 1645, 128 L. Ed. 2d 365 (1994); Johnston v. IVAC Corp., 885 F.2d 1574, 1579-80, 12 USPQ2d 1382, 1385-86 (Fed.Cir.1989). They are for the court to decide and explicate on the record. See Read Corp. v. Portec Inc., 970 F.2d 816, 822, 23 USPQ2d 1426, 1432 (Fed.Cir.1992). Since the trial court did not instruct the jury on these issues, and did not appear to have utilized its prior claim interpretation in ruling on the motions for JMOL, it is necessary for us to resolve these issues now in ruling on the court's denial of Genetics' motions for JMOL. Id. at 822-23, 23 USPQ2d at 1432. We address each of these threshold points in order.

a. The specific activity limitation

We begin with the question of whether, as the Genetics defendants assert, the specific activity limitation appearing in the '603 claims is implicit in the '075 and '330 claims.

The Genetics defendants first assert that plaintiffs/appellees are precluded from arguing that the limitation is not implicit in those claims due to plaintiffs' failure to appeal the district court's grant of summary judgment (by filing a cross-appeal to the present appeal). As noted previously, in that decision, the trial court determined that the limitation was implicit in the '075 claims. Thus, according to the Genetics defendants, that determination is "law of the case."7

We disagree. "The general rule is that, without taking a cross-appeal, the prevailing party may present any argument that supports the judgment in its favor." Radio Steel & Mfg. Co. v. MTD Products, Inc., 731 F.2d 840, 843, 221 U.S.P.Q. (BNA) 657, 660 (Fed.Cir.), cert. denied, 469 U.S. 831, 105 S. Ct. 119, 83 L. Ed. 2d 62 (1984); see 9 James W. Moore, Moore's Federal Practice, p 204.11, at 4-47 (2d ed. 1993). Here, were we to adopt a claim construction more favorable to plaintiffs' case than the construction adopted by the district court on summary judgment, that would have no different result than affirmance of the trial court's April 6, 1990 judgment. Accordingly, plaintiffs are not precluded from arguing a construction in support of the judgment that is different from that announced by the trial court.

In any event, the Genetics defendants argue, the specific activity limitation appearing in the '603 claims is inherent in the phrase "human tissue plasminogen activator" appearing in the '075 and '330 claims. Again, we disagree. Assuming for the sake of argument that the phrase "human tissue plasminogen activator" is ambiguous, there is no basis in the '330 and '075 specifications and prosecution histories for reading a specific activity limitation into the phrase. The only evidence identified by the Genetics defendants as a plausible basis for doing so is the '603 specification and prosecution history. However, that patent is completely unrelated to the '330 and '075 patents. It was prosecuted by a different entity than the others. The underlying research was conducted by different individuals than the research underlying the others. And finally, while the specific activity concept is a definition of purity critical to the patentability of the '603 claims--it is the critical distinction of those claims over the less purified materials constituting the relevant prior art8--the same cannot be said of the '330 and '075 claims, in which other limitations serve to distinguish the claimed subject matter over the prior art. Thus, that documentation cannot serve as the basis for reading the limitation into the phrase. We conclude that the '330 and '075 patents contain no implied specific activity limitation. The avoidance of this limitation by FE1X thus cannot provide the basis for a finding of non-infringement under the doctrine of equivalents in relation to these patents.

b. The meaning of "500,000 IU/mg."

The next question to resolve is the meaning of the numerical figure "500,000 IU/mg." appearing as part of the specific activity limitation of the '603 claims. The Genetics defendants argue that the figure means 500,000 IU/mg. as measured using a bovine fibrin plate assay. The plaintiffs, by contrast, argue that the figure is not limited to any specific assay type.

We agree with the Genetics defendants. According to the prior art of record, the numerical measurement of specific activity of t-PA can vary by more than a factor of three depending on the specific assay used.9 Thus, in order for the 500,000 figure to serve its intended purpose of distinguishing over the prior art value of 266,000 IU/mg., it is necessary to assign that figure a specific assay type. Since the claim is silent on this point10, we must look elsewhere for the answer. See Hormone Research Found. v. Genentech, Inc., 904 F.2d 1558, 1562, 15 USPQ2d 1039, 1043 (Fed.Cir.1990), cert. dismissed, 499 U.S. 955, 111 S. Ct. 1434, 113 L. Ed. 2d 485 (1991).

The '603 prosecution history reveals that the 500,000 figure was added to the '603 claims during prosecution to distinguish over prior work of Dr. Rijken's. That work had resulted in a specific activity of 48,000 IU/mg., when expressed in terms of the prevailing urokinase standard, and 266,000 IU/mg., when expressed in terms of the new and preferred t-PA standard that had evolved after the '603 filing date.11

The specific activity of the claimed product was initially only available in terms of the urokinase standard. When so expressed, the figure was 90,000 IU/mg. Dr. Collen was assigned the task of expressing the figure in terms of the new standard. That work resulted in the 500,000 figure. An article contained in the '603 prosecution history shows that this figure was determined using "bovine fibrin films", i.e., a bovine fibrin plate assay.12 Apparently, that assay was used so that the figure would be compatible with Dr. Rijken's prior work, in which specific activity was measured using the bovine fibrin plate assay13, and with the work which gave rise to the filing of the '603 patent, in which specific activity was likewise measured using the bovine fibrin plate assay. Col. 5, ll. 1-6. Accordingly, we conclude that the 500,000 figure means IU/mg. as measured using the bovine fibrin plate assay.

c. The definition of "human tissue plasminogen activator"

We next address the question of the meaning of the phrase "human tissue plasminogen activator" appearing in the '075 and '330 claims. Since a definition of that phrase cannot be extracted from the claims themselves, we look to the specification14 for guidance. See McGill Inc. v. John Zink Co., 736 F.2d 666, 674, 221 U.S.P.Q. (BNA) 944, 949 (Fed.Cir.1984), cert. denied, 469 U.S. 1037, 105 S. Ct. 514, 83 L. Ed. 2d 404 (1984).

A problem now arises because there are at least four possible definitions of the phrase set forth in the specification.15 First, there is a narrow structural definition: t-PA produced through recombinant DNA technology but having the same structure as natural t-PA.16 Second, there is a broader structural definition: all products containing the "essential"17 Kringle region, and the Serine Protease region.18 Third, there is an even broader structural definition: all products containing just the enzymatically active portion, i.e., the Serine Protease portion.19 Fourth, there is a functional definition: "[I]t is capable of catalyzing the conversion of plasminogen to plasmin, binds to fibrin, and is classified as a t-PA based on immunological properties as set forth hereinabove." Col. 6, ll. 15-19.

These diverse definitions reflect either inartful drafting, a conscious attempt to create ambiguity about the scope of the claims, or a desire to claim a wide variety of materials not described or enabled in the specification. For example, it appears that some of the products contained within the third definition will not be capable of binding to fibrin because they lack the regions which have been identified as playing a role in the fibrin binding process.20 Yet, the fourth definition identifies fibrin binding as an essential functional attribute of human t-PA.

An appropriate method for resolving the issue is to avoid those definitions upon which the PTO could not reasonably have relied when it issued the patent. That is an appropriate method to follow because it avoids the possibility of an applicant obtaining in court a scope of protection which encompasses subject matter that, through the conscious efforts of the applicant, the PTO did not examine.21 An applicant should not be able deliberately to narrow the scope of examination to avoid during prosecution scrutiny by the PTO of subject matter with the objective of more quickly obtaining a patent (or avoiding the risk of an estoppel), and then obtain in court, either literally or under the doctrine of equivalents, a scope of protection which encompasses that subject matter. See North American Vaccine Inc., 7 F.3d at 1577, 28 USPQ2d at 1337.

Under this approach, the first definition is the appropriate one to adopt because of the four, it is the most consistent with the limited form in which the claims are drafted22, and the others are hopelessly overbroad. As Dr. Goeddel testified, an infinite number of permutations of natural t-PA are covered by these other definitions. Many of these permutations would not be capable of binding to fibrin and would thus be inoperative. There is no basis provided in the specification for determining which of these permutations are operative and which are not. The point is supported by Dr. Larsen's testimony to the effect that the properties of these permutations were "totally unpredictable." Genentech acknowledged this point in two international patent applications it filed directed to the structural features which define FE1X in relation to natural t-PA.23 The determination of which permutations are operative would thus require an undue amount of experimentation.24 Thus, we are unwilling to say that the specification satisfies the enablement requirement of 35 U.S.C. Sec. 112 p 1 (1988) with respect to these broader definitions25, or that the PTO could have relied on these definitions in issuing the patent. See In re Fisher, 427 F.2d 833, 838-40, 166 U.S.P.Q. (BNA) 18, 23-24 (CCPA 1970); see also Amgen Inc. v. Chugai Pharmaceutical Co. Ltd., 927 F.2d 1200, 1212-14, 18 USPQ2d 1016, 1026-28 (Fed.Cir.), cert. denied sub nom., Genetics Inst., Inc. v. Amgen, Inc., --- U.S. ----, 112 S. Ct. 169, 116 L. Ed. 2d 132 (1991).

Plaintiffs argue that a protein (25E10) discussed in the specification provides a basis for interpreting the phrase "human tissue plasminogen activator" appearing in the claims broadly, on the theory that claims are to be interpreted in light of the disclosure of examples in the specification. See SmithKline Diagnostics Inc. v. Helena Laboratories Corp., 859 F.2d 878, 8 USPQ2d 1468 (Fed.Cir.1988). The difficulty with that argument is that 25E10 is not natural t-PA, and its disclosure therefore does not aid in interpreting the claims.26

We conclude therefore that the phrase "human tissue plasminogen activator" appearing in the '075 and '330 claims means natural t-PA.27

4. The Jury Findings Regarding the Specific Activity Limitation

Having completed our resolution of the three threshold claim interpretation questions, we consider in order the points raised by the Genetics defendants keeping in mind our standard of review: whether the jury's express or implied findings of fact are supported by substantial evidence. See Read, 970 F.2d at 821, 23 USPQ2d at 1431. We begin with the question of whether the jury's implied conclusion--that the specific activity limitation appearing in the '603 claims was met by FE1X either literally or equivalently--is supported by substantial evidence.28 We conclude that it is not.

As we have said, "substantial evidence is more than a mere scintilla. It means such relevant evidence as a reasonable mind might accept as adequate to support a conclusion." Biodex Corp. v. Loredan Biomedical Inc., 946 F.2d 850, 859, 20 USPQ2d 1252, 1259 (Fed.Cir.1991), cert. denied, --- U.S. ----, 112 S. Ct. 2957, 119 L. Ed. 2d 579 (1992) (quoting Consolidated Edison Co. v. NLRB, 305 U.S. 197, 229, 59 S. Ct. 206, 216, 83 L. Ed. 126 (1938)). Plaintiffs point to the following three pieces of evidence: (1) Dr. Larsen's testimony that the specific activity of FE1X is in the range from 350,000 to 450,000 IU/mg., (2) statements in an internal Institute document to the effect that the specific activity of FE1X is "similar" to that of natural t-PA29, and (3) a publication co-authored by Dr. Collen stating that the specific activity of FE1X is 440,000 IU/mg.30

To begin with, the specific activity measurements reported by Dr. Larsen were made using the chromogenic substrate assay, and there is no evidence that measurements made using this assay are comparable to those made using the bovine fibrin plate assay. In fact, all the evidence of record addressing this subject indicates that the two methods are not comparable. Dr. Larsen's testimony on this point is representative:

A Excuse me, I believe [the chromogenic assay measurements] requires a clarification. You're really talking about two totally different assays.

Q I understand that.

A So it's apples and oranges. You cannot compare the two numbers....

Q I understand. And at your deposition when you were asked about the specific activity for FE1X was measured against the international t-PA, you said it ranged from 350,000 to 450,000 IU's per milligram?

A Using a chromogenic substrate assay, and not the fibrin plate assay which appears to be the standard for measuring plasminogen activator activity with respect to these patent disputes. So it is completely distinct from anything that appears to be relevant to me in this trial.

Plaintiffs assert that a publication co-authored by Dr. Larsen shows that results obtained using the chromogenic assay are comparable to those obtained using the bovine assay.31 We cannot agree. Although that document presents results obtained using the two methods, it hardly establishes that the two methods are comparable. In fact, it implies just the opposite. The two figures reported for FE1X are very different: 334 ± 45 units/ug. for the chromogenic assay, and 440 units/ug. for the bovine fibrin plate assay.

Likewise, the statements in the internal Institute report do not provide the necessary support. First, they are qualitative whereas the necessary comparison is quantitative. Second, they are based on a synthetic assay which, unlike the bovine fibrin plate assay, does not measure the ability to dissolve clots.

Finally, the 440,000 figure reported in the publication co-authored by Dr. Collen was measured from material prepared by Leuven; it is not clear what relationship that material bears to the accused product that is the subject of this lawsuit. Dr. Collen's testimony on this precise point was equivocal. When asked whether this material was the same substance as the accused product, he merely said the "description" was the same. Moreover, other measurements reported in that publication and taken in the same manner fall outside the range of permissible scientific error and thus bring into question all the measurements reported.32

The only evidence in the record which is probative on the question of the specific activity of FE1X is the testimony of plaintiffs' expert Dr. Mann. According to that testimony, the specific activity of FE1X using the bovine fibrin plate assay is 253,800 IU/mg. plus or minus 18%, i.e., 208,116 to 299,484. That is significantly closer to the prior art value of 266,000 than it is to the claimed range33, and may even be less than that value34. FE1X is thus outside the permissible range of equivalents through the application of prosecution history estoppel.35 No reasonable jury could have concluded otherwise. The trial court thus erred in denying the Genetics defendants' motions for JMOL in relation to the '603 patent under the doctrine of equivalents.

5. The Jury Findings Regarding the Human Tissue Plasminogen

We next consider whether there is evidence in the record sufficient to support the jury's implied conclusion that the "human tissue plasminogen activator" limitation appearing in the '075 and '330 claims is met by FE1X.36 FE1X does not literally meet the limitation--it is not natural t-PA. Thus, the question is whether the evidence supports a finding that this limitation is met by an equivalent element of FE1X under the doctrine of equivalents.

To support such a finding, the evidence must be sufficiently particularized to meet the three prong test of equivalency enunciated in Graver Tank & Mfg. Co. v. Linde Air Products Co., 339 U.S. 605, 608, 70 S. Ct. 854, 856, 94 L. Ed. 1097, reh'g denied, 340 U.S. 845, 71 S. Ct. 12, 95 L. Ed. 620 (1950) requiring a showing of substantial identity of function, way, and result. Malta, 952 F.2d at 1327, 21 USPQ2d at 1166; Lear Siegler Inc. v. Sealy Mattress Co., 873 F.2d 1422, 1425-26, 10 USPQ2d 1767, 1770 (Fed.Cir.1989). If any one of the prongs is unsupported, the finding of equivalency cannot stand.

At the outset, we are confronted with a problem: The issue of whether the 'way' or 'result' prongs are met is highly dependent upon how broadly one defines the "function" of human t-PA. If, as the trial court thought, a broad definition is appropriate--stimulating "the dissolution of fibrin clots through the cleavage of plasminogen to plasmin"37--then it is difficult to imagine how FE1X, or any version of t-PA for that matter, would avoid infringement under the doctrine of equivalents because t-PA, or any operative variant, would by definition necessarily perform this function in the same general way with the same general results. However, if the definition of t-PA set forth in the specification is adopted--catalyzing the conversion of plasminogen to plasmin, [and] bind[ing] to fibrin38--then the equivalence question becomes a much closer one.

The operative definition for purposes of equivalency analysis is the intended function as seen in the context of the patent, the prosecution history, and the prior art. Graver Tank, 339 U.S. at 609, 70 S.Ct. at 856; Zenith Lab. Inc. v. Bristol-Myers Squibb Co., 19 F.3d 1418, 1425, 30 USPQ2d 1285, 1291 (Fed.Cir.1994). Based on our review of these materials, we conclude that no reasonable jury could have adopted the broad definition suggested by the trial court. As noted, the specification expressly defines fibrin binding as a critical component of the "function" of human t-PA. Col. 6, ll. 16-19. Other evidence confirms this. According to a British patent application filed by Foundation, it is critical in a therapeutic sense--it reduces the risk of hemorrhaging.39 Moreover, as Drs. Goeddel and Collen testified, the fibrin binding affinity of human t-PA is a critical distinction between this protein and the two prior plasminogen activators, urokinase and streptokinase. Thus, a functional definition of t-PA which ignores this distinction would result in a range of equivalents which impermissably reads on the prior art.40 See Zenith Lab., 19 F.3d at 1425, 30 USPQ2d at 1291 n. 9.

We conclude that the "function" of human t-PA for purposes of the equivalency analysis includes fibrin binding, and no reasonable jury could have concluded otherwise. In light of this definition, we find that the record is devoid of any particularized evidence or linking argument showing that FE1X functions in substantially the same way as human t-PA or achieves substantially the same results.41

Plaintiffs point to the testimony of several witnesses to the effect that the Kringle 2 (K2) region of amino acids is present in both FE1X and human t-PA, and that this region plays a role in the ability of both to bind to fibrin. Dr. Larsen testified:

Q And FE1X binds the fibrin, doesn't it?

A Weakly.

Q But it binds the fibrin?

A Weakly.

Q Weakly. And it binds the fibrin through the second kringle region, doesn't it?

A I don't know.

Q You don't know how it binds the fibrin?

A I mean that is speculation in the literature, but there are conflicting reports to that point.

Q But there are literature reports that suggest that the finger region is involved in the binding to fibrin, as well as the kringle two region; isn't that correct?

A That is what is generally believed, yes.

Q And natural t-PA also binds to fibrin through kringle two, doesn't it?

A Assuming that its kringle two, yes.

Q Does GI's FE1X bind to fibrin?

A GI's bind to fibrin? I think it does, but in a weaker way than the full-length material.

Q Binds in a weaker way?

A Yes.

Q In your view, does it bind to fibrin in a completely different way than the full-length human t-PA?

A No. Clearly there are common elements in the binding of the two t-PA's.

Dr. Larsen's testimony is speculative, and Dr. Wetlaufer's testimony is tentative and conclusory. Even assuming that the K2 region does play a role in the binding function of both, that hardly establishes that the two bind to fibrin in substantially the same way with substantially the same results, particularly in view of the overwhelming and undisputed evidence that the two possess dramatically different properties and structure.

First, there is the undisputed testimony of Drs. Collen and Larsen that the fibrin binding affinity of FE1X is less than half, i.e., about 40%, of that of human t-PA as a consequence of the deletion of the E and F regions in FE1X.

Second, there is undisputed evidence showing that the amino acid substitution at position 117 of the K1 region eliminates a glycosylation site, and thus prevents a carbohydrate side chain from being attached to the protein during post-translational modifications by the host mammalian cell. According to Dr. Larsen, the deletion of the F and E regions standing alone would result in a protein that would be therapeutically ineffective in that it would be incapable of binding to fibrin at all; preventing the attachment of the carbohydrate side chain increases the binding affinity of FE1X sufficiently to make it therapeutically effective. Thus, the mode of binding is hardly substantially the same.

Third, there is the undisputed evidence that FE1X behaves significantly differently than human t-PA in the human body. It has a half-life42 about ten times that of natural t-PA and it has a significantly decreased affinity for binding to endothelial cells43 in relation to human t-PA. Genentech acknowledged the significance of these advantages in the two international patent applications noted earlier in this opinion.44 Thus, the results achieved are hardly substantially the same.

We are mindful that the state of the science in this area of endeavor is very imprecise.45 Thus, it would be inappropriate to interpret Malta as requiring plaintiffs/appellees to prove the specific mechanism by which FE1X binds to fibrin, or to prove that the different properties and structure exhibited by FE1X bear no relation to the binding function. Our only point is that the showing that the K2 region plays a role in the binding function of each is insufficient, particularly in view of the profound differences in the properties and structure possessed by each.

For all the foregoing reasons, we conclude the trial court erred in denying the Genetics defendants' motion for JMOL in relation to the '075 and '330 patents under the doctrine of equivalents.46

The trial judge should have granted JMOL in favor of the Genetics defendants on the doctrine of equivalents findings by the jury because:

1. The specific activity limitation appearing in the '603 claims means specific activity as measured using the bovine fibrin plate assay.

2. Considering the '603 prosecution history, the specific activity of FE1X does not as a matter of law meet the specific activity limitation appearing in the '603 claims under the doctrine of equivalents.

3. The human tissue plasminogen activator limitation appearing in the '075 and '330 claims means natural t-PA. The jury's implied conclusion that FE1X meets the human tissue plasminogen activator limitation appearing in the '075 and '330 claims under the doctrine of equivalents is not supported by substantial evidence.

For all the foregoing reasons, we reverse.47

REVERSED.

LOURIE, Circuit Judge, concurring.

I join the opinion of the majority, but I would adopt a different means to interpret the expression "human tissue plasminogen activator" (t-PA). Moreover, I wish to point out an additional reason why I believe that the accused FE1X cannot infringe a claim to t-PA under the doctrine of equivalents.

The independent claims at issue in the '075 and '330 recombinant patents contain no definition for the DNA isolate other than that it encodes human t-PA. Such a claim, defining a substance only by its function, encompassing all substances that accomplish that result, is akin to a single means claim, which might fail to satisfy the definiteness requirement of 35 U.S.C. Sec. 112. See Fiers v. Sugano, 984 F.2d 1164, 1171, 25 USPQ2d 1601, 1606 (Fed.Cir.1993). Such a claim avoids that problem only when the term "human tissue plasminogen activator" has definitive meaning, when it describes a specific compound. That term is in fact definite, as it is well established that human t-PA is a protein consisting of 527 amino acids. Thus, all the claims which depend upon a definition of human t-PA are limited to that specific compound and any other compounds considered under law as equivalent thereto.* The trial court, of course, held that there was no literal infringement.

Under the doctrine of equivalents, an accused compound can be held to infringe if, inter alia, it represents only an insubstantial change from the claimed compound. See Graver Tank & Mfg. Co. v. Linde Air Prods. Co., 339 U.S. 605, 607, 610, 70 S. Ct. 854, 855, 857, 94 L. Ed. 1097 (1950). The accused compound in this case consists of a protein that contains 446 amino acids, 15% fewer than the t-PA referred to in the claims. This is not an insubstantial change, but a substantial one. Moreover, it has ten times the half-life of natural t-PA. The t-PA recited in the claims was not copied, since the accused FE1X is a very different material, independently invented and developed, requiring an estimated 130 man-years, and costing $20 million. If claims are to have any meaning, as a matter of law such a substance cannot be held to be infringing.

Thus, this case illustrates the problem that results not only when a court fails to construe the claims for the jury, but also when the fact-finder unduly focuses only on the function, way, result analysis of Graver Tank, 339 U.S. at 608, 70 S.Ct. at 856. These limited means of analysis fail to fully elucidate the issue, especially when the patented material is a chemical, as it is here. Is the increased half-life part of the "way" analysis or is it a different "result"? Is the binding to fibrin "function," as stated by the majority, or is it part of the "way" t-PA dissolves clots? These questions illustrate the shortcomings of the function, way, result tests which relate to "how" a substance works, i.e., what it does, rather than what it is, which claims purport to define. The other aspects of Graver Tank, if properly considered by the fact-finder, would have led to a sounder result. The substantiality of the difference between the accused and claimed compounds, the fact of independent development, and the lack of copying, all lead to a conclusion of lack of infringement.