Opinion for the court filed by Circuit Judge PROST.
Opinion concurring-in-part, dissenting-in-part filed by Circuit Judge DYK.
PROST, Circuit Judge.
The present patent infringement case arises from a declaratory judgment action brought in the United States District Court for the District of Columbia on April 11, 2006. Plaintiff Intervet Inc. (“Intervet”) denies infringing U.S. Patent No. 6,368,601 (“'601 patent”), owned by Defendants Merial Limited and Merial SAS (collectively “Merial”), directed to DNA constructs encoding a type of porcine circovirus. The district court entered summary judgment of noninfringement based on its construction of six disputed claim terms. Merial appeals the district court’s claim construction for three of the terms, and, in the alternative, appeals the district court’s summary judgment of non-infringement based on the doctrine of equivalents.
Because we agree with Merial that the district court erred in its construction of two disputed claim terms, we reverse the district court’s claim construction, vacate the judgment of noninfringement, and remand for a finding of whether the accused device infringes under the claim construction articulated herein. Additionally, because we agree with Merial that the district court misapplied the law of prosecution history estoppel, we instruct the district court to consider on remand arguments related to literal infringement and to infringement under the doctrine of equivalents, consistent with the analysis herein.
Background
Postweaning Multisystemic Wasting Syndrome (“PMWS”) is a disease affecting livestock pigs. Researchers at Merial learned that PMWS is associated with a [1285]*1285particular type of porcine circovirus.1 The scientific community was aware of porcine circoviruses prior to MeriaPs findings. Known porcine circoviruses, however, were not observed to be pathogenic, meaning they did not appear to cause disease in infected pigs. Merial filed for the '601 patent pertaining to the discovery of what it described as a previously unknown pathogenic type of porcine circovirus that the inventors dubbed “PCV-2.” PCV-2 stands for “porcine circovirus type II.” MeriaPs patent categorizes previously known, non-pathogenic porcine circoviruses as belonging to “type I” or “PCV-1”. The patent identifies a particular known DNA sequence isolated from pig kidney cells called PK/15 as being representative of type I. The '601 patent then identifies five isolated pathogenic porcine circovirus strains as being representative of type II.
The patentee placed the five representative strains on deposit with the United States Patent and Trademark Office (“PTO”) as part of the description of the invention. The written patent disclosure provides the full DNA sequence for four of these strains, as well as the full sequence of PK/15.
The disclosure explains that the deposited PCV-2 strains had been detected in lesions of pigs with PMWS, but not in healthy pigs. The patent description observes that the sequenced strains exhibit 96% nucleotide homology with each other, and only 76% nucleotide homology with PK/15.2 The description concludes from these observations that there are two types of porcine circoviruses, and that non-pathogenic “type I,” as represented by PK/15, is “thus” distinct from pathogenic “type II,” as represented by the five isolated strains. '601 patent col.l 11.48-62. The disclosure then identifies the subject of the present invention as “the group II porcine circovirus, as defined above, isolated or in the form of purified preparation.” Id. at col.1 11.63-65.
The patent disclosure goes on to analyze the sequenced PCV-2 strains in more detail, providing tables comparing the sizes and alignments of the strains. The disclosure then identifies one of the sequenced strains, designated SEQ ID 4, as being further representative of the other strains, and identifies thirteen open reading frames (“ORFs”) for PCV-2 using that sequence. The '601 specification identified nine of the thirteen disclosed ORFs that are unique to PCV-2, and four that are present in both PCV-2 and PCV-1.
“ORF” is a commonly used term in molecular genetics that has a standard textbook meaning. An ORF is a portion of a gene that contains a sequence of nucleotide bases that may be translated into a protein. Each amino acid of a protein is encoded by a DNA codon. A codon consists of three adjacent nucleotide bases. The first codon in an open reading frame is the “start” codon, which encodes a modified form of methionine. Each amino acid in the polypeptide chain is encoded by a subsequent set of three base pairs, until the translation is terminated at a stop codon that does not itself encode an amino acid, but rather signals the end of translation. Thus, a double-stranded length of DNA can have six different reading frames, depending on the starting base-pair of the first codon and the direction in which the strand is read.3 The length of [1286]*1286an ORF is thus defined by the number of codons that lie between a start codon and a stop codon within the same frame.
Identifying the ORFs of a gene sequence differentiates the portions of the sequence that may encode a protein from the portions that do not encode a protein. It allows those skilled in the art to estimate the size and composition of potential amino acid sequences for the proteins encoded by the gene. Identifying ORFs is especially important in the context of viral or prokaryotic DNA, which can contain several overlapping ORFs in the same gene sequence.
There are two groups of claims in the '601 patent relevant to the present case. The first • group can be represented by independent claim 9, which reads:
9. A vector comprising an isolated DNA molecule comprising a sequence selected from the group consisting of ORFs 1 to 13 of porcine circovirus type II.
The second group can be represented by independent claim 32, which reads:
32. An isolated DNA molecule comprising a nucleotide sequence encoding an epitope which is specific to PCV-2 and not specific to PCV-1.
An epitope is an immunodominant region of a protein, meaning it is the part of an antigen peptide that is recognized by antibodies of the immune system. The patent explains that certain epitopes found among strains of PCV-2 are not present in PCV-1. Thus, the regions of DNA encoding these epitopes are unique to PCV-2. Epitopes unique to PCV-2 are relevant to diagnostics or treatments, because antibodies specific to these epitopes will recognize and bind to the pathogenic PCV-2, but will ignore the benign PCV-1.
Merial’s patent claims cover certain vectors and other DNA molecules that contain portions of the PCV-2 gene sequence. These molecules are believed to be useful in diagnosing and vaccinating against PMWS, by stimulating the production and activity of antibodies against PCV-2.
Intervet is an animal healthcare company that manufactures vaccines for livestock. Intervet developed a vaccine called “Porcine Circovirus Vaccine Type 2” that contains a porcine circovirus nucleotide sequence in a vector for treating PMWS in pigs. Merial alleges that Intervet’s PMWS vaccine uses an infringing PCV-2 sequence.
At a Markman hearing, the United States District Court for the District of Columbia construed six disputed claim terms of the '601 patent. Among these constructions, the district court defined the term “porcine circovirus type II” as consisting of the five nucleotide sequences that Merial placed on deposit with the PTO. The district court construed the term “ORFs 1-13” as the DNA sequences of the thirteen ORFs of SEQ ID 4 listed in the table under Example 13 of the patent. Finally, the distinct court construed claim 32 in its entirety to refer (in part) to a construct comprising at least one DNA molecule that is unique to one of the five sequences on deposit with the PTO.
The district court then entered summary judgment of noninfringement based on these claim constructions. It was undisputed that Intervet’s vaccine contained a nucleotide sequence that was 99.7% homologous to one of the deposited sequences. The accused product was therefore held to be outside the literal claim scope of PCV-2, which required strict identity to one of the five deposited sequences. The district court also held that the doctrine of prosecution history estoppel precluded Merial from arguing that the accused sequence infringed under the doctrine of equivalents.
Merial timely appealed to this court, arguing that the district court erred in its [1287]*1287claim construction and erred in applying the doctrine of prosecution history estoppel to Merial’s equivalence arguments for the accused product. For the reasons discussed below, we agree with Merial that the district court erred in its claim construction and application of prosecution history estoppel.4
Discussion
Claim Construction
Claim construction is a question of law that is reviewed de novo. Markman v. Westview Instruments, Inc., 52 F.3d 967, 978 (Fed.Cir.1995) (en banc). To the extent possible, claim terms are given their ordinary and customary meaning, as they would be understood by one of ordinary skill in the art in question at the time of the invention. Phillips v. AWH Corp., 415 F.3d 1303, 1312-13 (Fed.Cir.2005) (en banc). Idiosyncratic language, highly technical terms, or terms coined by the inventor are best understood by reference to the specification. Id. at 1315. Such understanding is informed, as needed, by the prosecution history, if it is in evidence. Id. Construing the claims in light of the specification does not, however, imply that limitations discussed in the specification may be read into the claims. It is therefore important not to confuse exemplars or preferred embodiments in the specification that serve to teach and enable the invention with limitations that define the outer boundaries of claim scope. Id. at 1323.
It is with an eye to these tenets of claim construction that we review the district court’s Markman order and conclude the district court erred. We discuss each term in turn.
“Poreine Circovirus Type II”
The district court found that the claim term “porcine circovirus type II” was limited to the five sequences that were deposited with the PTO as part of the description of the invention. The district court was persuaded by Internet's arguments that the patent specification defined the invention as being these five sequences, and contained no disclosure from which to infer that any other sequences were also part of the invention.
It is clear enough to us, however, that the patent states that the five deposited strains and listed sequences are “representative of’ a “type of porcine circovirus,” and thus do not constitute the entire scope of the invention. '601 patent col.l 11.60-61 (emphases added). Sequences are representative of the scope of broader genus claims if they indicate that the patentee has invented species sufficient to constitute the genera. Enzo Biochem, Inc. v. Gen-Probe, Inc., 323 F.3d 956, 967 (Fed.Cir.2002); In re Smythe, 480 F.2d 1376, 1383 (C.C.P.A.1973). Here, the deposited strains are representative species of the larger “type II” genus, where the genus is identified and claimed as the invention.
Claims properly directed to a genus may be adequately supported by the patent disclosure if a sufficient number of species is disclosed so as to properly identify the scope of the genus. Id. Here, the patentee has disclosed five species of PCV-2, provided the full sequences for four, and identified the potential coding portions of the sequences. The patentee also provided a counterexample, PCV-1, that by definition lies outside the scope of the claimed genus, as well as a representative species of the counterexample, its sequence, and potential coding portions for the representative.
[1288]*1288Neither the claim itself nor the specification provides a homology threshold above or below which a particular PCV strain is properly considered PCV-2 rather than PCV-1. It refers instead to strains of the invention having “significant serological similarity” and stringent, selective cross-hybridization to the deposited strains over PK/15. The only quantitative boundaries disclosed in the patent are the 96% homology among representative PCV-2 sequences, and the 76% homology between those sequences and the representative of PCV-1.
The patent’s stated conclusion that the disclosed PCV-2 sequences “thus” represent a new type of porcine circovirus is based on the pathogenicity of the isolated strains, as well as the observed homology patterns. See, e.g., '601 patent col.5 11.59— 61. This conclusion comports with the way that viruses are typically classified in the relevant art. Cfi Universal Virus Database of the International Committee on Taxonomy of Viruses available at http:// www.ictvdb.rothamsted.ac.uk/. The invention is then defined as being the type II porcine circovirus, which is in turn “as defined above.” '601 patent col.5 ll.64. Thus, the pathogenicity and homology patterns are the defining properties of the new type of virus. The claim construction of “porcine circovirus type II” is therefore properly limited to porcine circoviruses that have these two defining properties.
We therefore construe the claim term “porcine circovirus type II” to be “a pathogenic pig virus having a circular genome that is at about 96% or more homologous with the four sequences disclosed in the present specification, and about 76% or less homologous with the PK/15 sequence.” Strains that fit this definition would be expected to have strong serological similarity and cross-hybridize to the deposited strains under high stringency conditions. As such, limiting the claim scope according to these properties is not inconsistent with the other descriptive language in the specification.
“ORFs 1-13”
The district court’s claim construction of ORFs 1-13 defines the claim scope as consisting of the DNA sequence of the thirteen ORFs enumerated in Example 13 of the patent specification as those ORFs apply to SEQ ID 4. Merial argues that the term should instead read on any translatable length of DNA between a start and stop codon in the PCV-2 gene sequence. Although the district court is correct that the disclosed ORFs define the claim term, the court erred in confining the scope of the term to the precise limits of the representative ORFs listed in Example 13, and the exact DNA sequence of SEQ ID 4.
The ORFs listed in Example 13 are identified as corresponding to one representative PCV-2 sequence, designated in the patent as SEQ ID 4. Although the patent explains that the listed ORFs are identical for some of the deposited strains of PCV-2, it also identifies some variation. The specification explains that the ORFs listed in the table are representative, and one of skill in the art would understand that slight natural variation is to be expected. Indeed, limiting the construction of the term to the exact ORF sequences of SEQ ID 4 would even exclude from the claimed ORFs two of the four sequenced strains of PCV-2, the ORF variations for which sequences are expressly disclosed following the table in Example 13. Thus, we hold that the district court’s construction is improperly narrow in scope.
We reject the dissent’s position that the specification limits “ORFs 1-13” to the ORFs of the four sequenced strains. The discussion in Example 13, which explains that the limits of ORFs 1 to 13 are “identical” for certain sequenced strains (and not [1289]*1289for others), strongly implies that the term “ORFs 1-13” does not refer to a specifically defined list of limits, but instead contemplates the potential for variation in any given strain of PCV-2. Furthermore, the specification describes the analysis set forth in Example 13 as “representative of the other circovirus strains associated with the multi-systemic wasting syndrome.” We have already construed that set of circovirus strains to be broader than just the four sequenced strains, so it would be incongruous to selectively impose the narrower construction here, as the dissent suggests.
We note that because isolates of the same viral type will have essentially the same proteins, they will have the same number of ORFs. The ORFs will be approximately the same size and located in the same relative regions of the genome. By identifying the thirteen ORFs of representative sequence SEQ ID 4, the specification purports to disclose to one of skill in the art the expected ORFs of all PCV-2 isolates. Thus a broader claim construction that allows for some variation in the precise limits of the ORFs and of the underlying DNA sequence is consistent with the expectations of a skilled artisan reading the patent disclosure.
Thus the term ORFs 1-13 is properly construed as “lengths of translatable DNA between pairs of start and stop codons, corresponding to the 13 ORFs identified in the patent specification.” ORFs of some PCV-2 strains may not have limits 100% identical to the thirteen illustrated in the patent, but one of skill in the art would readily recognize those ORFs as corresponding to ORFs identified in the patent. Indeed, ORFs 1-13 could correspond to ORFs in other circoviruses, or even other species, as indicated by the examiner’s initial rejection of the claim. It is the “of porcine circovirus type II” limitation, rather than Example 13, that confines' the claim scope to ORFs of PCV-2.
“Specific to PCV-2 and Not Specific to PCV-1”
The parties below could not agree on what terms of claim 32 were disputed, and the district court decided to construe the claim in its entirety. The district court construed claim 32 to mean “an isolated DNA molecule that includes, but is not necessarily limited to, a DNA sequence which codes for an immunodominant region of a protein, wherein the sequence is from the genome of a PCV-2 circovirus, and not from the genome of a PCV-1 circovirus.” The district court explained that due to the “comprising” transition term, the claim may read on molecules that contain sequences that encode epitopes common to PCV-1 and PCV-2, as long as the molecule contains at least one sequence that encodes an epitope unique to PCV-2. We see no error in this construction, and it appears that at the time of the Markman hearing, Merial did not see any either.
Merial challenges this construction on appeal because in the district court’s subsequent infringement analysis, the court explained that the part of the claim construction specifying that the sequence be “from” the genome of a PCV-2 circovirus, etc., excluded sequences that were physically derived from a non-PCV-2 source. Merial argues that such a manufacturing requirement has no place in a proper analysis of this claim, and is inconsistent with the district court’s otherwise correct claim construction. We agree. For purposes of our review of the district court’s opinion, we focus our analysis on the term “specific to” in claim 32, since it appears that this term is the hook for the requirement that the sequence be unique to and derived from PCV-2.
[1290]*1290As Intervet explains, the term “specific to” is a specialized term of art in immunology that typically refers to one structure’s proclivity for binding to another structure. For example, antibodies will attack a viral antigen if paratopes of those antibodies are “specific to” an epitope in the viral antigen. The specialized definition of this term does not make sense in the context of claim 32, however, because the claimed epitope is not described as binding to porcine circoviruses; it is described as located within a porcine circovirus. The epitope is thus bound by antibodies that are “specific to” PCV-2. In light of the patent description and a general understanding of the relevant art, the claim would be understood by one of skill in the art to be using the term “specific to” in a colloquial or non-technical sense. Cf. Intervet, Inc. v. Merial Ltd., No. 1:06-cv-00658, 2007 WL 5685349 (D.D.C. Nov. 28, 2007) (claim construction order at 21). As construed, a nucleotide sequence encoding an epitope that is specific to PCV-2 and not specific to PCV-1, as that term is used in claim 32, is a nucleotide sequence that encodes part of a polypeptide sequence of PCV-2, but not part of a polypeptide sequence of PCV-1. More specifically, it encodes at least one epitope found on the PCV-2 virus, but not found on the PCV-1 virus.
The district court found that Intervet’s vaccine could not have contained a sequence encoding an epitope specific to PCV-2 because the sequence was derived from a non-PCV-2 source. This analysis may be mooted by our reversal of the district court’s claim construction of “PCV-2”, since it is no longer clear that the source of the sequence in Intervet’s product is not PCV-2. Nevertheless, to the extent that the district court’s application of its claim construction requires that the encoded epitope be unique to PCV-2 among all possible antigens, it is erroneous. If the term “specific to PCV-2” meant that the epitope must be found only on PCV-2 and no other antigen, then the subsequent limitation “and not specific to PCV-1” would be redundant. Thus an infringing epitope may be common to PCV-2 and some other antigen, as long as it is not also common to PCV-1. Whether one isolates the sequence directly from a PCV-2 virus or engineers a sequence obtained from another source such that it encodes a PCV-2 epitope makes no difference to the proper application of the district court’s otherwise correct claim construction.
Accordingly, we reverse the district court’s claim constructions of the terms “porcine cmcovirus type II” and “ORFs 1-13,” clarify the proper interpretation of the construction of the term “specific to PCV-2 and not specific to PCV-1,” and remand the question of infringement for a determination consistent with the claim constructions articulated herein.
Doctrine of Equivalents
The district court found that prosecution history estoppel precluded Merial from invoking the doctrine of equivalents. Merial was thus estopped from arguing that the accused PCV strain was equivalent to the claimed “porcine circovirus type II,” as that term was construed by the district court.5 The district court erred, however, in applying controlling Federal Circuit and Supreme Court law to the prosecution history of the '601 patent. As a result, the scope of the district court’s bar on Merial’s ability to invoke the doctrine of equivalents was overly broad.
Whether prosecution history estoppel applies to a particular argument, [1291]*1291and thus whether the doctrine of equivalents is available for a particular claim limitation, is a question of law. Bai v. L & L Wings, Inc., 160 F.3d 1350, 1354 (Fed.Cir.1998); Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1460 (Fed.Cir.1998) (en banc). Where an amendment narrows the scope of the claims, and that amendment is adopted for a substantial reason related to patentability, the amendment gives rise to a presumption of surrender for all equivalents that reside in “the territory between the original claim and the amended claim.” Festo Corp. v. Shoketsu Kinzoku Kogyo Kabushiki Co., 535 U.S. 722, 740, 122 S.Ct. 1831, 152 L.Ed.2d 944 (2002) {Festo VIII). This presumption can be overcome by showing that “at the time of the amendment one skilled in the art could not reasonably be expected to have drafted a claim that would have literally encompassed the alleged equivalent.” Id. at 741, 122 S.Ct. 1831. One way to make this showing is to demonstrate that “the rationale underlying the narrowing amendment bore no more than a tangential relation to the equivalent in question.” Festo Corp. v. Shoketsu Kinzoku Kogyo Kabushiki Co., 344 F.3d 1359, 1368 (Fed. Cir.2003) (en banc) (Festo IX). Although there is no hard-and-fast test for what is and what is not a tangential relation, it is clear that an amendment made to avoid prior art that contains the equivalent in question is not tangential. See Pioneer Magnetics, Inc. v. Micro Linear Corp., 330 F.3d 1352, 1357 (Fed.Cir.2003).
The applicability of prosecution history estoppel does not completely bar the benefit of the doctrine of equivalents from all litigation related to the amended claim. See Festo VIII, 535 U.S. at 737-38, 122 S.Ct. 1831 (“There is no reason why a narrowing amendment should be deemed to relinquish equivalents ... beyond a fair interpretation of what was surrendered.”) The scope of the estoppel must fit the nature of the narrowing amendment. A district court must look to the specifics of the amendment and the rejection that provoked the amendment to determine whether estoppel precludes the particular doctrine of equivalents argument being made.
Merial’s independent claim 9 as originally drafted read, “9. A vector comprising an isolated DNA molecule comprising a sequence selected from the group consisting of ORFs 1-13.” The examiner rejected this claim, noting that for purposes of the rejection “[t]he ORFs are assumed to be derived from porcine circovirus, but as written, the claims could encompass ORFs from any organism.” The claim was then rejected in view of ORFs from PK/15. The inventors disagreed that these ORFs could be derived from any other organism, and argued that the specification defined ORFs 1-13 based on the limits of the ORFs in the PCV-2 genome. Nevertheless, the claim was amended to add the limitation that the ORFs were “of porcine circovirus type II”. The examiner then allowed the claim.
We agree with the district court that this amendment was a narrowing amendment, despite Merial’s arguments that it was merely clarifying. As noted in the patent specification, four of the thirteen claimed ORFs are present in the “type I” circovirus. The original claim only required that a vector comprise a nucleotide sequence comprising one of the thirteen ORFs. Thus, the claim as originally written read on ORFs of PCV-1, and was properly rejected over PK/15. We therefore also agree with the district court that the amendment was substantially related to patentability.
The amendment thus raises a presumption of surrender for all equivalents that reside in the territory between the identified ORFs of PCV-2 and ORFs of PCV-1, as well as corresponding ORFs, if any, for any non-porcine organism. Merial is thus [1292]*1292estopped from arguing that ORFs of pathogenic circoviruses found in other organisms are equivalent to ORFs of PCV-2. It is also estopped from arguing that ORFs of a pathogenic strain of PCV-1 are equivalent to ORFs of PCV-2, despite the strain having strong homology with PK/15 and weak homology with the representative strains disclosed in the patent. Merial is not, however, estopped from arguing that a pathogenic porcine viral sequence with over 99% nucleotide homology with one of the five representative strains is equivalent to that strain.6 Such a draconian preclusion would be beyond a fair interpretation of what was surrendered. Cf. Festo VIII, 535 U.S. at 737-38, 122 S.Ct. 1831. The rationale for the amendment was to narrow the claimed universe of ORFs down to those of PCV-2, and bore only a tangential relation to the question of which DNA sequences are and are not properly characterized as PCV-2. Cf. Festo IX, 344 F.3d at 1369.
The district court thus erred in finding that prosecution history estoppel precluded Merial from arguing that the accused product is equivalent to one of the exemplary embodiments of the asserted claim. The district court is thus instructed on remand to compare the accused product with the claims as construed herein for a determination of infringement literally or pursuant to the doctrine of equivalents, if applicable.
Conclusion
The district court erred in construing the disputed claims of the patent in suit and in barring the doctrine of equivalents from its infringement analysis. Accordingly, we reverse the district court’s claim construction, vacate the entry of summary judgment of noninfringement, and remand to the district court with instructions to determine, consistent with the analysis in this opinion, whether the accused product infringes the asserted claims of the '601 patent.
RE VERSED-IN-PART, VACATED-IN-PART, AND REMANDED