RICH, Circuit Judge.

This appeal is from the September 28, 1988, judgment of the United States District Court for the Northern District of Georgia, Gainesville Division. Intervet America, Inc. (Intervet) sued Kee-Vet Laboratories, Inc., Robert D. Keenum, and Keenum, Inc. (Kee-Vet) for infringement of its patent No. 4,530,831 ('831 patent), entitled "Infectious Bursal Disease Vaccine." The district court held that Intervet failed to prove infringement and did not reach the question of validity of the patent. We reverse and remand.

I [2] A. Background

Intervet's patent claims a vaccine for a poultry disease called infectious bursal disease (IBD) and a method of making the vaccine. IBD particularly attacks chickens and its name derives from an organ in chickens called the "bursa" which is important in maintaining the immune system. If the bursa is damaged, as it may be by IBD, the chicken's ability to resist a host of diseases is diminished. Undiminished, IBD can cause great economic loss.

As is usual with inventions of this type, the inventors deposited vaccine virus samples in the American Type Culture Collection (ATCC) in Rockville, Maryland, to satisfy the requirements of the law. Before any patent issued, a predecessor of Kee-Vet Labs., Inc., Agri-Bio Corporation, obtained some of the samples and used them to grow its own vaccine, which it called Bio-Burs I (pronounced "eye"). Pursuant to 9 CFR 114.9(d), Agri-Bio's United States Department of Agriculture production outline recited the history of the development of Bio-Burs I vaccine as follows:

Twenty vials of bursal disease vaccine virus was [sic] obtained from ICB-clonevac D-78, serial lot No. 11-103 on November 3, 1982. Four sequential passages were made in C[hicken] E[mbryo] F[ibroblasts] by Agri-Bio Corp.

Kee-Vet Labs., Inc., subsequently purchased Agri-Bio's Bio-Burs I operation, and Agri-Bio's production license for Bio-Burs I, was transferred to Kee-Vet which continued to make and sell Bio-Burs I, prompting Intervet's complaint for patent infringement.

A bench trial was held. The district judge limited his decision to the issue of infringement. He construed the claims, relying heavily on part of the patent prosecution history, and held that "Plaintiff's claims limit their vaccine to unattenuated viruses" and that "plaintiff's patent is limited to a single administration scheme." The judge made findings that Bio-Burs I is an attenuated virus and that it is not effective upon a single administration. For those reasons he held that Intervet had failed to prove infringement.

B. Claims

Intervet alleges infringement of claims 1, 4, 5, and 7. They read:

1. A live vaccine effective against Infectious Bursal Disease in poultry upon a single administration to birds at the usual age of vaccination comprising a live Infectious Bursal Disease virus belonging to the strain deposited at the ATCC under No. VR-2041.

4. A virus suspension containing the Infectious Bursal Disease virus deposited at the ATCC under no. VR-2041.

5. A lyophilized virus composition containing the Infectious Bursal Disease virus deposited at ATCC under no. VR-2041.

7. In a method for the preparation of a live vaccine that protects poultry against Infectious Bursal Disease virus which comprises:

a. growing an Infectious Bursal Disease virus on a culture medium selected from the group consisting of embryonated eggs, chicken embryo cells, a culture of bursal cells and newborn mice,

b. subsequently arresting the cultivated virus material, and

c. subjecting the material obtained from step b. to at least one of the following treatments:

i. clarifying by centrifugation and/or filtration;

ii. adding a stabilizing agent;

iii. putting the material in a vessel;

iv. freeze-drying,

the improvement comprising that the Infectious Bursal Disease virus grown in step a. is the virus of the strain deposited in the ATCC under No. VR-2041.

C. Contentions of the parties

Appellant Intervet alleges some ten issues on appeal which, upon analysis, all resolve into the legal question of claim interpretation by the trial court involving essentially only the two issues stated above: (1) are the claims asserted "limited to a single administration scheme" and (2) are the claims limited to vaccines made from unattenuated viruses?

Responding to the question whether the trial court erred in holding that the claims are limited in the two respects just stated, Kee-Vet argues, as it presumably did in the trial court considering the results it achieved there, that without reading these two limitations into the claims from the prosecution history the claims are "vague and ambiguous, or completely trivial." It makes the argument that the form of the claims is comparable to a hypothetical claim reading "A seasoning composition for food comprising a material deposited at the public archives," which "needs interpretating because there is no recital of structure, functions, or properties of the material deposited." On this basis it attempts to justify reading limitations into claims.

D. Issues

Did the trial court err in construing all claims in suit to be limited to (a) a single administration scheme and (b) vaccines made from unattenuated viruses?

II [14] OPINION [15] A. Applicable Law

This appeal is from a judgment below holding the asserted claims 1, 4, 5, and 7 not infringed, given the trial judge's construction of those claims. We agree with Kee-Vet's statement of the law that interpreting the scope of claims is a question of law and that determining the issue of claim infringement is a question of fact, citing McGill v. John Zink Co., 736 F.2d 666, 671-72, 221 USPQ 944, 948 (Fed. Cir.), cert. denied, 469 U.S. 1037, 105 S.Ct. 514, 83 L.Ed.2d 404 (1984). As an issue of law, our review of claim construction is plenary. Corning Glass Works v. Sumitomo Electric U.S.A., Inc., 868 F.2d 1251, 1256, 9 USPQ2d 1962, 1965 (Fed. Cir. 1989).

On the issue of claim construction, which takes into account the specification of the patent, the prosecution history, and the prior art, we find it necessary to cite only the section of this court's recent opinion in E.I. Du Pont De Nemours Co. v. Phillips Petroleum Co., 849 F.2d 1430, 1433, 7 USPQ2d 1129, 1131 (Fed. Cir. 1988), designated "I.A. Claim Interpretation." Kee-Vet also cites this case but has apparently not taken adequate notice of that section's several times repeated statement to the effect that this court has consistently adhered to the proposition that courts cannot alter what the patentee has chosen to claim as his invention, that limitations appearing in the specification will not be read into claims, and that interpreting what is meant by a word in a claim "is not to be confused with adding an extraneous limitation appearing in the specification, which is improper." The court quoted with approval from Autogiro Co. of America v. United States, 384 F.2d 391, 395-96, 155 USPQ 697, 701 (Ct.Cl. 1967), the statement that "No matter how great the temptations of fairness or policy making, courts do not rework claims. They only interpret them." The panel found it necessary in Du Pont to reverse the district court's interpretation of claims which read into them properties of a polymer which were not recited in the claims. We have to do the same here. We refer the lower court to this section of the Du Pont opinion for a complete and up-to-date exposition of this court's "consistent approach in interpreting claims." What it says is, a fortiori, applicable to the impropriety of injecting into claims limitations from the prosecution history. Ambiguity, undue breadth, vagueness, and triviality are matters which go to claim validity for failure to comply with 35 U.S.C. § 112-¶ 2, not to interpretation or construction. We fully subscribe to the proposition that claims must be given the same construction when considering infringement as when considering validity.

B. The supposed single administration scheme limitation

The facts are simple, confusing, and presented the trial judge with a dilemma. He got out of it by making a seemingly logical choice. However, as a matter of well-established patent law, we must hold that the choice was legally wrong.

The facts are these. Responding to the second Patent Office action on August 6, 1984, applicants' attorney attempted to distinguish from the principal prior art reference, an article by Cursiefen et al., on the ground that their vaccination scheme involved two vaccinations with their strain of virus to get protection whereas applicants' strain "takes with a single vaccination at the age of 14 days in the face of MDA [maternally derived antibodies]." In the third office action the examiner said, "The instant claims are not limited to a single vaccination scheme which would patentably distinguish them from the reference." In response, the attorney, taking the examiner at his word, amended three of the claims, including patent claim 1, to refer to "single administration" effectiveness but did not so amend other claims. In the "Remarks" accompanying the amendment, however, the attorney made the unqualified and now admittedly untrue statement that "the claims are restricted to a single vaccination scheme."

Of the asserted patent claims, only claim 1 will be seen to contain such a limitation. Claims 4, 5, and 7 do not have the limitation. Nevertheless, the trial judge took the untrue remark to so limit all claims in the patent. He read into those three claims a limitation which they do not and never did contain simply on the basis of the attorney's remark. This was error.

When it comes to the question of which should control, an erroneous remark by an attorney in the course of prosecution of an application or the claims of the patent as finally worded and issued by the Patent and Trademark Office as an official grant, we think the law allows for no choice. The claims themselves control.

There are, of course, situations in which what an attorney says or does during prosecution may be held against a patentee on the theory of estoppel. For example, when a patentee attempts to expand the literal meaning of a claim under the patent law doctrine of equivalents and the prosecution history shows that the expanded scope would be inclusive of subject matter the attorney had represented to the examiner was not intended to be included in order to get the claim allowed, the patentee may be estopped to contend otherwise. Graver Tank Mfg. Co. v. Linde Air Prods. Co., 339 U.S. 605, 70 S.Ct. 854, 94 L.Ed. 1097 (1950). But that is not the situation here where the patentee was granted broad claims in spite of the statement by the attorney that he was amending them, though he never did so. The examiner was not misled or deceived. The erroneous remark was not the end of the prosecution. The examiner was fully aware of what claims he was allowing. Subsequent to the erroneous remark, the examiner had two telephone interviews with the attorney and made two "examiner's amendments" to the claims himself before formally allowing application claims 1, 4, 7-9, and 15-17, now the claims of the patent. The record does not fully show what was said in the interviews so we do not know what transpired between the examiner and the attorney. The absence of the "single administration" effectiveness limitation from several of the claims may very well have been discussed. The examiner certainly was aware of it. The presumption of validity under 35 U.S.C. § 282 carries with it a presumption the examiner did his duty and knew what claims he was allowing. In any event, the claims as allowed are what we have to deal with and it is not for the courts to say that they contain limitations which are not in them. Du Pont, supra. See also Loctite Corp. v. Ultraseal Ltd., 781 F.2d 861, 228 USPQ 90 (Fed. Cir. 1985).

Claim 1 raises an additional question which has not been decided. Of all the asserted claims, it alone was amended to state that the live vaccine being claimed is "effective against Infectious Bursal Disease upon a single administration to birds at the usual age of vaccination," a statement of an alleged inherent property. This clause is followed by the word "comprising," following which is the statement defining what the vaccine is, namely, "a live Infectious Bursal Disease virus belonging to the strain deposited at the ATCC under No. VR-2041."

The question raised is, first, whether the single administration recitation is part of a preamble of the claim which gives rise to the possibility that it may not be a limitation of the claim. Kropa v. Robie, 187 F.2d 150, 88 USPQ 478 (CCPA 1951). Intervet's brief contends it is part of a preamble and is not a limitation of the claim but "merely a description of an inherent property of the vaccine." Although Intervet seems to be blowing both hot and cold in saying it is a property of the vaccine to be effective on a single administration and in also saying it is not a claim limitation from the infringement standpoint, that is the basis on which it argues that the claim can be infringed by a vaccine which is not effective on a single administration. Kee-Vet's brief, citing Loctite, grants that the "single administration" clause is in a preamble but insists that it is a claim limitation. It does so, however, because of the attorney's remark that all the claims are limited to a virus effective as a vaccine upon a single administration, as the trial court held. Because of that approach, the trial court never dealt with the precise question of claim construction raised by Intervet, namely, whether the single administration clause in claim 1 is or is not a claim limitation. This issue remains unsettled and may or may not have to be dealt with on remand. No other asserted claim contains the clause. A patent is infringed if a single claim is infringed. There are other claims which, contrary to the trial court's view, are broader, at least with respect to how many vaccine doses may be required.

With respect to the significance of preamble statements in claims, in addition to the extended discussion in Kropa v. Robie and what is said in Loctite see this court's more recent comments in In re Stencel, 828 F.2d 751, 754-55, 4 USPQ2d 1071, 1073 (Fed. Cir. 1987). In Loctite, it is to be noted that part of what was in a claim preamble was held to be a limitation and another part of the preamble was not. Each case must be judged on its own facts.

C. The supposed unattenuated virus limitation

The trial court correctly held that it must, in determining the fact of infringement, compare the defendant's product with the claims of the patent, not with the patentee's product. The court also correctly held that the burden is on the patentee to establish infringement. Having found as a fact that Kee-Vet's product, Bio-Burs I, is "attenuated," which "when applied to a virus means a virus of reduced virulence," the court held there was no infringement because "Plaintiff's claims limit their vaccine to unattenuated viruses."

Here again, even more than with the "single administration" supposed limitation, the court was reading limitations into the claims which cannot be found there; and, again, it was reading them in from remarks made by the attorney in the course of arguing distinctions from a cited reference. We have set forth the asserted claims in full above and it is clear that they make no reference whatever to attenuation.

For the reasons already discussed in parts II A and B of this opinion, the trial court erred in reading into the claims a limitation to unattenuated viruses. The claims must be dealt with on the basis of what they say, not remarks made by the attorney in discussion of a reference, and the issue of infringement must be determined accordingly.

On the subject of attenuation, it appears to us that any live IBD virus vaccine is, by definition, an attenuated form of the more virulent virus which causes the disease to be vaccinated against, the so-called "field strain." Attenuation is a relative term and means nothing without knowing the reference point from which attenuation is being measured. The patent itself makes clear that the clone D78 deposited at the ATCC under No. 2041 was a strain the patentee developed by attenuation of a virus taken from the bursa of a chicken infected with the IBD virus. It seems to be undisputed that Bio-Burs I was made from the clone or virus, samples of which were deposited as No. VR-2041. Whether or not it was "attenuated" in the process of being made seems to be a matter of serious dispute together with the related questions whether Bio-Burs I and the No. 2041 viruses are the same or of the same strain. We express no opinion thereon because we are a court of review only and those are unanswered questions of fact for decision by the trial court. The trial court's order of June 30, 1988, states (our emphasis):

Plaintiff argues that it met its burden [of proving infringement] by showing that Bio-Burs I was derived from the strain plaintiff deposited at the ATCC. The court does not find it necessary to decide whether the two viruses are of the same strain. Literal infringement requires the plaintiff to show that the accused device [sic] embodied every element of the claim as properly interpreted. A showing that the viruses derive from the same strain is insufficient to meet this burden.

The conclusion of the last sentence was predicated on the court's determination that two of the presumed claim elements were those which we have held not to be elements of the claims, with the possible exception of one of the elements of claim 1. Infringement of claims 4, 5, and 7 is still a wide open question.

D. Summary

For the foregoing reasons, the decision of the trial court that Bio-Burs I does not infringe the plaintiff's '831 patent on the grounds stated by the trial court is reversed and the case is remanded for a reconsideration of that issue in accordance with this opinion, as well as a consideration of any other issues, such as the validity or enforceability of the patent or any of its claims, properly before the court.

COSTS

Each party shall bear its own costs.

REVERSED AND REMANDED.