UPC Munich Local Division rejects GXD-Bio’s claim against Myriad; EP 3 346 403 revoked

UPC Case Law | 09.02.2026

Summary/Background

The Local Division (LD) Munich dismissed GDX-Bio’s infringement action against Myriad and revoked EP 3 346 403 for added matter.

EP 3 346 403 claimed a method for quantifying expression level of a target gene in FFPE breast cancer tissue comprising normalizing an expression level of the target gene to that of the endogenous reference gene OAZ1.

Myriad was represented by HOFFMANN EITLE (Joachim Renken, Niels Hölder, Michael Pfeifer, and Claudia Unsin).

Added matter

The LD revoked EP 3 346 403 as it contained added matter. The application to amend the patent based on Auxiliary Requests 1, 2, or 3, filed by GXD-Bio in response to Myriad’s counterclaim for revocation, was refused for the same reasons. In its assessment of added matter, the LD applied what is referred to as the “gold standard” for assessing added matter, as laid out by the Court of Appeal [1], and provided its concrete, reasoned findings.

Original claim 25 disclosed a method for quantifying an expression level of a target gene comprising three steps but did not mention that the target gene is in an FFPE tissue sample of human breast cancer. The LD found that this feature was not disclosed as belonging to the invention. The objective of the invention was instead to identify universal reference genes that are expressed in many different human tissues and are not limited to their use on specific tissue samples or in specific studies. For the LD, it was clear the person skilled in the art will set out to read the application with this background in mind [2].

GXD-Bio relied on Experimental Example 7 as further alleged basis for claim 1. While Experimental Example 7 disclosed expression stability (which is an important property for a reference gene) of 12 novel reference genes for a pool of different FFPE tissue samples, including breast cancer FFPE samples, it did not provide any information on expression stability in the specific FFPE tissues (such as breast cancer tissue). With this understanding, the skilled person would merely derive from this disclosure that the novel reference genes could be applied to FFPE samples in general. The LD followed Myriad’s argument that Experimental Example 7 is not an embodiment of original claim 25 [3] and ruled that “it is not sufficient that breast cancer tissue samples are disclosed. Rather, they must also be disclosed as belonging directly and unambiguously to the invention. However, this cannot be established as the application teaches the person skilled in the art that the technical problem is to find universal genes which are not limited to use on specific tissue samples.” [4]

The LD’s consideration of a “disclosure as belonging to the invention” in the context of added matter is somewhat reminiscent of the German case law, e.g. the Federal Court of Justice (BGH) decision X ZR 75/08 – Reifenabdichtmittel. [5] In this decision, the BGH found that the disclosure of a product containing certain components does not equally disclose a product consisting of these components as belonging to the invention in the absence of any indication in the original application text that it is particularly advantageous or otherwise desired that the product exclusively consists of these components. [6] 

The LD further held that “[t]here is no pointer in the application … that the FFPE breast cancer tissue samples are particularly preferred”. [7] The lack of a pointer referred to by the LD is a criterion well-established in the case law of the EPO’s Boards of Appeal: a combination of features taken from separate embodiments of the original application is considered to constitute added matter in the absence of a pointer to that particular combination. [8]

Claim 1 further required that OAZ1 is used as a reference gene in the method for quantifying an expression level of a target gene. The LD found that in Experimental Example 7, OAZ1 was (reliably) detected in all 60 FFPE samples, including the 10 FFPE breast cancer tissues samples. Experimental Example 7 was nevertheless found to lack a disclosure of OAZ1 as the (sole) reference gene for normalizing the expression level of a target gene in a FFPE breast cancer sample: OAZ1 was neither described as the best reference gene for FFPE tissue samples nor did the data reveal the expression stability of OAZ1 in FFPE breast cancer tissue samples. [9] Considering the original application in its entirety, the LD concluded that OAZ1 was not singled out as the reference gene in the claimed method, neither for normalizing the expression level of a target gene, nor with regard to FFPE tissue samples [10].

Whereas the LD refers to the singling out from the original application, in the terminology applied by the EPO’s Boards of Appeal, the expression “singling out” usually refers to a claimed feature. In essence, however, it reveals the same underlying concept, namely that singling out a feature in the claim without direct and unambiguous basis in the original application results in added matter, in line with the case law of the EPO’s Enlarged Board of Appeal. [11] Using EPO terminology, one could say that the combination of OAZ1 with FFPE breast cancer sample tissues was singled out from different lists, a list of reference genes and a list of FFPE tissue samples, thereby creating new subject matter in line with the EPO’s “two-list principle”.

Infringement

GXD-Bio’s infringement action against Myriad et al. was dismissed by the LD. Myriad’s EndoPredict test involves measuring of expression levels of eight target genes and three reference genes, i.e. OAZ1, CALM2, and RPL37A. The relative expression of each of the eight target genes is determined by normalization to an average expression level of OAZ1, CALM2, and RPL37A.

The LD construed claim 1 of EP 3 346 403 as requiring normalization to a single reference gene, i.e., to OAZ1 only. In the contested EndoPredict assay, there is only one normalization calculation to the average value of the expression levels of the three reference genes, which is not identical to normalization to the measured value of the expression level of OAZ1 [12].

The LD found that the actual construction of the contested embodiment was decisive, which did not make use of the claimed method. The LD ruled that “[i]t is not acceptable to assume an infringement because the same result as that achieved by the attacked embodiment can also be hypothetically achieved by a method that complies with the patent, especially if this requires further steps to be carried out after obtaining the normalization value according to the patent (calculation of the average with several normalization values of other reference genes) in order to achieve the same result.” [13]

As direct and indirect infringement were both ruled out by the LD, the infringement action was dismissed.

retour