T 0047/22 (Immunoglobulin library/CRESCENDO BIOLOGICS) 05-07-2023

Reasons for the Decision

2. Admittance of documents D26 to D28

2.1 Violation of the right to be heard and admittance of document D26

2.1.1 Under Article 113(1) EPC, decisions may only be based on grounds or evidence on which the parties concerned have had an opportunity to present their comments.

2.1.2 The appellant argued that it had not been able to respond adequately to the document Chowdhury 1998, on which the opposition division had relied in paragraph 48.4 of its decision but which had never previously been admitted or referred to at any stage of the opposition or examination proceedings. The conclusion drawn by the opposition division in view of Chowdhury 1998, i.e. that the authors of document D10 "were of the opinion that immune libraries with improved members could only be achieved by focusing on variations that would not occur in nature" was completely incorrect.

This position was not derivable from the opposition division's preliminary opinion.

Document D26 had been filed in direct response to the new position taken by the opposition division and the new evidence, i.e. Chowdhury 1998, cited by the opposition division.

2.1.3 Point 48.4 of the decision under appeal merely cites verbatim the passage on page 10, lines 29 to 34 of document D10, which references Chowdhury 1998. The opposition division's argument why the claims of the main request were inventive was based on what is taught in document D10 itself and not the document referenced in the cited passage.

The appellant/opponent itself mentioned in its letter of 13 May 2021, page 16, third-to-last paragraph that "D10 describes a method for generating antibodies with higher affinity to a target antigen than that of a parental antibody (see D10, page 8 lines 30-31). D10 teaches the advantages of a targeted approach (D10, passage bridging pages 10 and 11)". Thus, the opposition division has not referred to any new passage of document D10 (see point 48.4 of the decision under appeal).

In line with the discussion on the teaching in the closest prior-art documents in point 43.1 of the decision under appeal and covering the opponent's above-mentioned argument, the opposition division reasoned that document D10 (as well as documents D2 and D23) did not focus on naturally occurring variations but applied random sequence mutations. Paragraph 48.4 of the decision under appeal discusses the teaching of document D10 but does not rely on the content of the Chowdhury 1998 reference itself, beyond the information provided in the relevant sentence on page 10, lines 29 to 31 of document D10.

The board therefore considers that the opposition division did not violate the right to be heard within the meaning of Article 113(1) EPC and that the argument made in paragraph 48.1 of the decision under appeal cannot be used to support the admittance of document D26.

2.2 Other reasons concerning the admittance of document D26

2.2.1 The board fails to see a clear link between the reference to review article D25 and the change in the opposition division's reasoning when comparing paragraph 34 of its preliminary opinion and paragraphs 41.7 and 43.2 of its final decision.

Document D26 may be of prima facie relevance; however, this is only one criterion to be considered by the board when deciding on the admittance of late-filed submissions. Other criteria to be considered include fairness and procedural economy.

2.2.2 In the board's opinion, the decisions referred to by the appellant, namely T 1817/15, T 238/92 and T 1380/04, were all taken in the context of earlier, outdated versions of the Rules of Procedure of the Boards of Appeal.

The appellant also referred to decision T 1213/19. In this case, the entrusted board applied the RPBA 2020 when deciding on the admittance of a document, as the appellant had filed the document only after filing the statement of grounds of appeal. The board decided to admit the document on the basis of its prima facie relevance and the fact that the respondent had known the content of that document before the appeal was filed and had had sufficient time to assess its contents and react appropriately (see Reasons 5 and 7).

2.2.3 In the case in hand, it seems that the issue of whether the prior art itself provided a motivation to replicate the in vivo somatic hypermutation process and was not limited to identification of the SHHs by in silico means, as well as the general concept of identifying SHHs by analysing sequence data of immunoglobulins generated in vivo, had already been discussed in the notice of opposition (see points 9.21 and 9.22) and the patent proprietor's reply (see e.g. page 30, last three paragraphs et seq.).

As these issues have been extensively discussed since the start of the opposition proceedings, D26 could and should have been submitted during the opposition proceedings.

2.2.4 Consequently, document D26 is not admitted into the proceedings under Article 12(4) and (6) RPBA.

2.3 The request for admittance of documents D27 and D28 was made conditional on the admission of document D26. The respondent argued that document D28 should be admitted into the proceedings in any case since it represented the common general knowledge. However, the content of document D28 was invoked exclusively in the context of the respondent's discussion of document D26 in its reply to the statement of grounds of appeal.

Since document D26 was not admitted into the proceedings, the board considers it unnecessary to decide on the request to admit documents D27 and D28.

Main request

1. Claim construction

1.1 Claim 1, step a)

1.1.1 The appellant argued that claim 1 did not require active immunisation and that to put the claim into practice, it was necessary simply to identify related immunoglobulins.

The method of claim 1 encompassed a sequence comparison between the first lead immunoglobulin and one or more related immunoglobulins, i.e. including the comparison of only two or a few sequences. The scope of the expression "one or more sites" was open-ended and not restricted to only SHHs. Hence the method of claim 1 comprised identifying any mutation at any position. Step g) of claim 1 was optional and so was not limiting.

1.1.2 The board agrees that claim 1, step a) does not require active immunisation; however, the identified immunoglobulins need to have been raised against the same target antigen and derive from the same germline sequence as the first lead immunoglobulin. Claim 1 also requires the immunoglobulins to have been raised by immunisation of a transgenic non-human mammal comprising human immunoglobulin genes. Yet this wording does not necessarily imply that the final immunoglobulin comprises human sequences as suggested by the respondent.

1.2 Claim 1, step c)

Claim 1, step c) reads "identifying, based on the sequence comparison, one or more sites at which there are variant amino acid residues [...] wherein the one or more sites at which there are variant amino acid residues comprise somatic hypermutation hot spots targeted during the immune response that are potential sites for modification of the first lead immunoglobulin;" (underlining added by the board).

This wording requires that the "one or more sites at which there are variant amino acid residues" comprise SHHs (targeted during the immune response in the mammal) as potential sites for modification. Thus, the claimed method defines a step of identifying at least one site which has to include SHHs.

2. Disclosure of the invention - Article 83 EPC - claim 1

2.1 Identification of SHHs

2.1.1 Claim 1, step c) requires knowledge of the presence of at least one SHH at the identified variant site(s) (see point 1.2 above). Therefore, it is essential to know whether the person skilled in the art knew how to determine the presence of SHHs.

2.1.2 The application as filed does not provide any instructions on how to generally identify SHHs, nor is there any prior-art reference in that direction. The only method disclosed is the identification of SHHs by aligning multiple, i.e. 20 or more, sequences with the first lead immunoglobulin (see Example 1 and Figures 1 and 4).

The question arises as to whether the identification of SHHs in general was common general knowledge at the filing date without requiring the alignment of at least 20 or more sequences.

In relation to its decision that the subject-matter of claim 1 was sufficiently disclosed, the opposition division referred to documents D4 to D10 as showing that the skilled person would be able to identify SHHs (see paragraph 32 of the decision under appeal).

The appellant contested this and argued that there was no evidence that alignment of only two or a few sequences allowed SHHs to be identified. None of documents D4 to D10 represented the common general knowledge.

The respondent had not shown that identifying SHHs was part of the common general knowledge or that SHHs could be identified by comparing only two or a few sequences.

2.1.3 The board is thus not convinced that the skilled person would have been able to verify whether variations in the amino acid sequence of immunoglobulins were due to SHH maturation by comparing the amino acid sequences of two or only a few immunoglobulins. The board is thus not convinced that the skilled person would have been able to identify SHHs.

Since it has not been shown that the relevant part of the method of claim 1, i.e. the part relating to the comparison of the lead immunoglobulin with only two or very few immunoglobulins, ensures the presence of SHHs as required in step c), the board concludes that the patent application fails to disclose the claimed invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art.

Auxiliary request 1

3. Disclosure of the invention - Article 83 EPC - claim 1

3.1 For the reasons provided in the context of claim 1 of the main request (see point 2.1), the board is of the view that the claimed invention, involving the comparison of the first lead immunoglobulin with at least 20 related immunoglobulins of common lineage that bind the same target antigen as the first lead immunoglobulin, allows SHHs to be identified and PCR-cloning errors to be excluded. See also Figure 4 of the patent and Annex A of document D15, showing that aligning the lead immunoglobulin sequence with those of more than 20, specifically 35, other members of the same lineage makes it possible to identify amino acid positions that have been mutated during the immune response.

3.2 Does a library produced by the claimed method allow for optimisation?

3.2.1 The board agrees with the respondent that the observed variations in the immunoglobulins raised against the same antigen and deriving from the same germline can be expected to be mainly due to the in vivo affinity maturation process by somatic hypermutation.

Occasional PCR-cloning-induced variations in a given sequence cannot be excluded but can be recognised and eliminated by comparing multiple sequences. The appellant has not provided any serious doubts substantiated by verifiable facts suggesting that the invention cannot be carried out by a skilled person.

3.2.2 Not all library members will have an optimised biological property. However, this is inherent to the nature of immunoglobulin libraries which need to be screened for members having the desired optimised biological property. On the basis of the selection of in vivo-matured immunoglobulins derived from a common lead germline sequence, it is credible that immunoglobulins with an optimised biological property in relation to the lead immunoglobulin will be present as members of the library and thus can be found. There is no evidence to the contrary on file.

Thus, the patent application discloses the invention in a manner sufficiently clear and complete for it to be carried out by a person skilled in the art as per Article 83 EPC.

Inventive step - Article 56 EPC - claim 1

The parties started their inventive step reasoning with document D2/D10 or D23 as the closest prior art. The board has no reason to deviate from this.

D10/D2 as the closest prior art

3.3 Document D10 (document D2 is the corresponding scientific article) discloses a method for improving the binding affinity of an immunoglobulin using a phage display library. The variant sequences differ from the parental/lead immunoglobulin on account of at least one random amino acid substitution, the amino acid being encoded by a codon comprising a nucleotide SHH-motif selected from the tetranucleotide A/G-G-C/T-A/T (Pu-G-Py-A/T) or the serine codons AGC or AGT. The amino acid substitution(s) can occur in any of the complementarity-determining regions (CDRs) of the variable heavy (VH) and/or variable light (VL) chain (see page 4, lines 9 to 16 and page 10, lines 9 to 12). The method allows for the generation of Fvs with increased affinity from a small library of variants (see page 9, last paragraph). The method yielded a 15 to 55-fold affinity improvement from a small library of only about 8 000 independent clones (see page 11, paragraph 2).

Example 1 discloses the construction of phage libraries starting from the mesothelin-specific single-chain variable fragment (scFv) referred to as "SS scFv" (see page 3, last paragraph), which comprises in its VL CDR3 two hot spot motifs of the A/G-G-C/T-A/T type and one AGT serine codon. Lastly, residues 89, 93 and 94 were randomly replaced.

3.4 Document D10 implicitly uses the same germline because all the sequences are derived from a single lead immunoglobulin. The immune library made according to the method disclosed in document D10/D2 is also focused on SHHs, as required in the method according to claim 1. However, the immune library of document D10/D2 uses a random amino acid replacement, i.e. all 20 natural amino acids (including the original one) randomly appear in place of the original hot spot amino acid.

D23 as the closest prior art

3.5 Document D23 relates to methods for screening and identifying immunoglobulins with diverse sequences and high affinity to a target antigen by combining computational prediction and experimental screening of a biased library of antibodies (see paragraph [0003]).

The method makes it possible to efficiently generate and screen protein libraries for optimised proteins with desirable biological functions. The process is carried out computationally in a high-throughput manner by mining databases of protein sequences of all organisms, especially human. The method is used in designing antibodies that are diverse in sequence and yet functionally related to each other. On the basis of the designed antibody sequences, a library of antibodies can be constructed to include diverse sequences in the CDRs and/or humanised framework regions (FRs) of a non-human antibody. This library of antibodies can be screened against a wide variety of target molecules for novel or improved functions (see paragraphs [0023] and [0024]).

In one aspect, the method relates to the in silico selection of antibody sequences on the basis of the amino acid sequence of a region in a lead antibody, which is used to search protein sequence databases. The choice of the database depends on the specific functional requirement of the designed motifs. Databases for immunoglobulin sequences of various species or even unrelated sequences in Genbank, Swiss-Prot or Kabat can be used to design the CDRs. A library of diverse antibody sequences can be constructed and screened experimentally in vitro or in vivo for antibody mutants with improved or desired function(s), such as affinity (see paragraphs [0025], [0229] and [0346]). A conventional BLAST analysis may be employed to search for sequences with high homology to the CDR H3 sequence (see paragraph [0726]).

The method may comprise (pre-)selecting from the plurality of tester protein sequences at least two peptide segments that have at least 10% sequence identity with the lead sequence, the selected peptide segments forming a hit library (see paragraphs [0087], [0118] and [0139]), and determining if a member of the hit variants library is structurally compatible with the lead structure template using a scoring function (see e.g. paragraphs [0119], [0141], [0174], [0186] and [0608]).

Paragraph [0552] states that: "Given the availability of a high affinity complex structure as a template, the hit variant library can be computationally pre-screened to reduce the library size, yet remain functionally highly focused compared to traditional libraries generated through complete randomization of amino acids in each position of the lead antibody. Through prediction and construction of the hit variant library in silico, the whole process of protein evolution can be hastened, effectively mimicking the natural process of antibody affinity maturation in a high throughput manner."

Selecting antibodies from a highly diverse library allows for broad coverage of sequences, thereby maximising the chance of finding the optimal sequence(s). To avoid 3D structural incompatibility between the tester and the lead, it is suggested to use expressed protein sequences and filter out the incompatible sequences (see paragraph [0555], [0556], [0560], [0636] or [0663]).

Difference and objective technical problem

3.6 The claimed method differs from the method in document D10 in that:

- The variant immunoglobulins to be included in the library were raised against the same target molecule in a transgenic non-human mammal comprising human immunoglobulin genes.

- The positions to be modified are selected on the basis of SHHs observed in vivo (i.e. not at motif-predicted sites) by comparing at least 20 immunoglobulins with the first lead immunoglobulin.

- The modification at the identified SHH(s) is limited to the variant amino acids observed in the pool of compared immunoglobulin sequences (i.e. does not use a random amino acid replacement).

3.7 The claimed subject-matter differs from the disclosure in the closest prior-art document D23 in that:

- The method compares at least 20 related immunoglobulins with the first lead immunoglobulin.

- All of the immunoglobulins have been raised against the same target antigen in a transgenic non-human animal.

- The positions to be modified are selected on the basis of SHHs observed in vivo.

- The modification at the identified SHH sites is limited to the variant amino acids observed in the pool of compared immunoglobulin sequences.

3.7.1 The appellant referred to paragraphs [0013] to [0015] of the patent as providing evidence that CDR homology was how related immunoglobulins were to be identified according to the invention, which was the same approach as used in document D23 (see paragraph [0726]).

3.7.2 The board does not agree. Claim 1, step a) relates to "identifying at least 20 related immunoglobulins of common lineage". In other words, two conditions need to be met:

i) the selection of "related" immunoglobulins which, according to paragraphs [0013] and [0015] to [0017], requires at least one CDR or FR to have at least 70% homology when compared with the lead immunoglobulin

ii) the immunoglobulins need to be derived from a common, i.e. the same, germline sequence (see paragraph [0014])

Thus, the germline sequence classification is to be considered independent of the at least 70% CDR/FR homology-based "related" feature.

3.8 The objective technical problem starting from either document D10/D2 or D23 as the closest prior art can be formulated as to provide a method for designing an alternative immunoglobulin library for optimising a biological property of a first lead immunoglobulin.

Obviousness

3.9 In assessing the obviousness of the claimed subject-matter, the relevant question is whether or not, having regard to the state of the art, the skilled person faced with this objective technical problem would have modified the method disclosed in document D10/D2 or D23 and arrived at the claimed method.

3.10 Document D10 in combination with document D3 or D22

3.10.1 In its inventive step arguments, the appellant specifically pointed to part b) of claim 42 of document D10, which it asserted would make the skilled person realise that the method was not limited to the identification of the hot spots by in silico means. It was the replication of the in vivo somatic hypermutation process that was key, not the method by which somatic hypermutations were identified or the identity/provenance of the immunoglobulin itself (reference was made to page 11, lines 16 to 19 and the sentence bridging pages 51 and 52).

3.10.2 However, part b) of claim 42 of document D10 cannot be read in isolation and follows part a), which explicitly refers to "providing a nucleic acid molecule encoding an amino acid sequence of a VH or a VL domain of a parental antibody, the nucleic acid molecule comprising at least one parental hot spot codon comprising at least one nucleotide within a hot spot motif" (underlining added by the board). In other words, the skilled person would also read part b) of this claim, in line with the general teaching of document D10, as requiring motif-based prediction of SHHs as a mandatory part of the method.

The paragraph bridging pages 10 and 11 of document D10 sets out other prior attempts to improve antibody affinity and explicitly mentions either random mutations of the CDR residues or mutating particular amino acids found from crystallographic analysis to affect antigen contact. Subsequently, the inventors propose targeting hot spots in the CDRs by random mutation to overcome some of the disadvantages in these methods, thereby teaching away from using the reported prior-art methods, including immunisation or DNA immunisation, for improving antibody affinity.

The board is of the opinion that on the basis of the teaching of document D10/D2 alone, a skilled person would not have arrived at the subject-matter of claim 1 in an obvious way.

3.10.3 With reference to document D3 and the common general knowledge, the appellant argued that the skilled person would know that the in silico modelling approach of document D10 might not capture everything that happens in vivo, and so the skilled person would actually be motivated to identify the somatic hypermutations for a given antibody on the basis of a sequence analysis of what is actually observed in vivo after immunisation. The appellant also pointed to document D22 as teaching the use of immune libraries obtained from host immune systems sensitised against the target antigen.

3.10.4 Document D3 analyses the diversification of camel heavy-chain antibodies by investigating the germline variability and the specific mechanisms of diversification. The authors of document D3 recommend providing a (naive) library that contains all germlines and analysing the positions where mutations occur (see page 922, left-hand column, paragraph 1). It is also explained that the occurrence of hypermutational hotspots, such as the AGY and TAY (Y = C or T) sequences, corresponded to the highest variability at the CDR1 and CDR2 of cDNA sequences (see page 923, right hand column, "Hypermutational hotspots imprinted in the germline").

Thus, a motif-based prediction of SHHs appears to be a reasonable approach (at least for the CDR1 and CDR2).

However, there is no suggestion to use variants derived from the same germline obtained by immunisation with the same antigen, nor is there any hint towards restricting the positions to be mutated only to amino acids observed in vivo.

Consequently, document D3 cannot lead to the method of claim 1.

3.10.5 The book chapter D22 reviews antibody libraries from immune repertoires and teaches that a relatively small immune library can be generated by using the potential of the immune system to enrich the antigen-binding B cells via clonal expansion and perform the affinity maturation (see paragraph bridging pages 549 and 550). Page 624 highlights the advantages of using immune libraries, such as from xenomice, for in vitro modifications including affinity maturation or humanisation of existing potent murine antibodies.

Document D22 is not concerned with methods for designing antibodies using SHHs, and there is no pointer towards amending the method disclosed in document D10 and using immune libraries.

The paragraph bridging pages 10 and 11 of document D10, which discourages the use of the reported prior-art methods (including immunisation or DNA immunisation) to improve antibody affinity, actually points away from amending the method of document D10 in the direction of immunised libraries.

3.10.6 None of the other documents (e.g. document D16 or D9 relied on in the opposition proceedings) teaches replacing the amino acids at observed SHHs only with the variants observed during in vivo affinity maturation against the same target antigen.

3.10.7 In the board's view, on the basis of the teaching of document D10/D2 alone or in combination with the teaching of document D3 or D22, a skilled person would not have arrived at the claimed subject-matter in an obvious way.

3.11 Document D23 in combination with document D22

3.11.1 The appellant further argued that combining the teachings of document D23 with the skilled person's common general knowledge as represented by document D22 would render the subject-matter of claim 1 obvious.

3.11.2 The book chapter D22 is summarised in point 3.10.5 above.

3.11.3 On the basis of the teaching in document D23, the skilled person would not consider using an immune library from xenomice as suggested in document D22, since document D23 emphasises creating an unfocused library in order to have as many variations as possible for screening towards a specific target antigen (see paragraph [0025], [0346], [0555], [0560], [0636] or [0650]).

3.11.4 There is no teaching in either document D23 or document D22 that the lead and variant immunoglobulins in the library have to be derived from the same germline by somatic hypermutation, or that the positions to be modified are selected on the basis of SHHs observed in vivo, or that the modification at the identified SHHs is limited to the variant amino acids observed in the pool of compared immunoglobulin sequences.

No document on file suggests modifying the methods described in document D23 so as to arrive at the method claimed according to claim 1. None of these documents discloses steps d) and e) of claim 1 of auxiliary request 1 on file.

3.12 The board concludes that the claimed subject-matter is not obvious from one of the two alternative closest prior-art documents D10/D2 and D23, either alone or together with any combination document discussed above. Thus, the claimed subject-matter involves an inventive step within the meaning of Article 56 EPC.

In view of this, it is not necessary to consider inventive step starting from an objective technical problem of providing an improved method.