England and Wales High Court (Patents Court) Decisions

IN THE HIGH COURT OF JUSTICE
CHANCERY DIVISION
PATENTS COURT

		Royal Courts of Justice Strand, London, WC2A 2LL
		19 May 2004

Mr Justice Laddie:

Introduction:

1. In this action the claimant, Celltech R &D Limited ("Celltech"), claims royalties from Medimmune Inc ("Medimmune") under a patent licence dated 19 January 1998. The licence relates to what are referred to as "Adair Patent Rights". One of those is the German designation of EP 0 626 390 ("the Patent"). Under the licence, Medimmune is obliged to pay royalties on products made or sold in the territory of the licence (which includes Germany) which, but for the licence, would infringe a valid claim of one of the licensed patents. The licence defines a valid claim as meaning a claim of an issued unexpired patent within the Adair Patent Rights which has not been held invalid or unenforceable in an unappealed decision of a court or competent body having relevant jurisdiction. Exclusive jurisdiction for resolving disputes under the licence is reserved to the English courts. There has already been one piece of litigation between these parties under the licence. That related to activities of Medimmune in the United States of America. It has no bearing on the issues raised in this action.
2. In Germany Medimmune manufactures and sells (or has manufactured and sold on its behalf) a product called SYNAGIS. This is an antibody preparation which is used for the treatment of a virus which causes respiratory illness in babies and young children. Celltech alleges that it infringes at least claim 1 of the Patent. The only question arising in this action is whether or not that allegation of infringement is made out. To determine that it is agreed that I must apply the German law on patent infringement to the facts of this case.

Technical Background

3. All living things are made up of cells. All cells contain a number of different chemicals including a group of large molecules called proteins. Different proteins have different functions within the cell. For example some are enzymes which mediate chemical reactions, some are structural, such as those which form part of muscle cells and some are chemical messengers, for example those molecules called hormones. Proteins are made up of strings of building blocks, called amino acids. A protein may be made up of hundreds or thousands of amino acids. There are 20 amino acids. Each is given a convenient name so it is not necessary to refer to it by its full chemical formula. Furthermore, as shorthand, each one can be referred to by an abbreviation. Thus the amino acid Arginine is frequently referred to as Arg, Serine as Ser, Threonine as Thr, Tyrosine as Tyr and so on. Each amino acid has a different chemical structure and exhibits different chemical and physical characteristics from all the others. The structural formulae of a small amino acid, glycine, and a large amino acid, tyrosine, are set out below.

Figure 1:

[Diagram or picture not supplied]
4. The function of each protein is, in large part, determined by the precise sequence of the amino acids of which it is made and the shape into which the protein molecule is folded.
5. Antibodies play a large part in an animal's immune system. When a human is infected with, say, a virus, the latter may multiply and cause illness. For example it may invade and destroy host cells. An effective immune system is one which recognises the virus as "foreign" and then produces countermeasures designed to neutralise it. The creation of antibodies is part of such a system. Although below I will concentrate on virus infections, the same general principles apply to meeting infections with, for example, bacteria. It also applies to the body's reaction to the presence of other types of foreign matter.
6. A virus contains proteins. Some of those will be on its surface. Like every other protein, every surface protein possesses a number of discrete areas which can be recognised as foreign by the immune system of a host which the virus has infected. Each of these areas is known as an antigen. When so recognised, the host system produces antibodies in specialised cells called b -Lymphocytes. A b -Lymphocytes cell can only learn to make one antibody and each antibody is a bespoke chemical which is designed to lock onto a single specific antigen. Thus a foreign protein which contained, say, 8 antigens could, in a perfect immune system, induce the production in the host of 8 different antibodies, each one of which would lock onto a specific antigen. Once an antibody has locked onto, or bound to, its antigen, this may well interfere with the foreign protein's function and the antigen/antibody combination will be recognised by another part of the host's immune system which leads to the combination, and thereby the foreign protein, being destroyed. Once the host system has learnt to produce a particular antibody it will usually remember how to do it. As a result, if there is another later attack by a virus with the same surface protein, the host system can more rapidly swing into action to produce the relevant antibodies. The period during which the host remembers how to make a particular antibody varies. If it were possible to make bespoke antibodies to order on an industrial scale, they could be used to speed up or supplement the host's immune system. Instead of waiting for the immune system to recognise the foreign antigens and to learn how to make suitable antibodies, or to switch into production of an antibody which it had learnt to make in the past, large quantities of ready-made antibodies could be injected directly into the host.
7. Although I have described, in a much simplified way, the way the immune system in a human works, essentially the same system operates in all other animals including, for example, rats, mice and horses. In theory, if one could make a rat produce antibodies to Virus X and those antibodies could be gathered, it would be possible to use those antibodies to tackle an infection of the same virus in humans. Unfortunately it is not that simple.
8. All animals produce very large numbers of antibodies to different antigens on different foreign proteins. The number of such antibodies will vary from host to host and will depend, among other things, on the number of foreign proteins that the host has had to learn to deal with during its life. A laboratory mouse may have hundreds or thousands of antibodies to different antigens in its blood stream. It would be particularly useful to isolate a single antibody which is capable of binding to a single antigen on a particular foreign protein and to make that single antibody in very large quantities. Best of all, as I will explain below, would be if single human antibodies could be made in large quantities. Unfortunately there are ethical and practical problems in doing the latter. But it is possible to produce large quantities of, for example, single antibodies produced by mouse immune systems. The precise way this is done is irrelevant to the issues in this case, but, in essence, it is as follows. A mouse is challenged with the foreign agent, for example Virus X. It produces antibodies to some or all of the antigens on the various virus proteins, particularly those on the virus's surface. As mentioned above, each mouse b -Lymphocyte cell can produce only one antibody. Techniques have been developed which allow one to pick out individual b -Lymphocyte cells. These are then hybridized with a strain of mouse cancer cells. Because they are developed from cancer cells, the hybrid cells are immortal. They continue to multiply and grow more or less indefinitely. As they do so they produce the single antibody which the b -Lymphocyte had learnt to make. The single antibody so produced is called a monoclonal antibody because it comes from a line of cloned cells – that is to say the hybrid ones. Such antibodies, since they are produced from mice, are referred to as murine.
9. If a murine antibody is used in a human they may well lock onto the relevant protein on Virus X and thereby provide or boost the human defence to infection. However there is a significant problem. Antibodies are themselves proteins. A murine antibody includes antigenic sites which would be recognised by the human immune system as foreign. It will learn to make antibodies to the murine antibody. Eventually it may make so much antibody to the murine antibody that, as soon as the murine antibody is injected into the human, all of it is effectively destroyed by the anti-antibody antibodies. This problem is known as the HAMA response. HAMA stands for human antimouse antibody response. The more a human is exposed to murine antibodies, the greater the HAMA response is likely to be. One way of reducing or avoiding this problem would be to make human antibodies and to graft into them, by modern genetic engineering techniques, only those parts of the murine monoclonal antibodies which are responsible for locking onto the Virus X antigen. In this way, antibodies could be produced which are almost completely human – and therefore are much less likely to generate the HAMA response – yet which have the essential Virus X bonding characteristics of the relevant monoclonal murine antibody. Once again, this is not as straightforward as it may sound. To understand why, it is necessary to know something about the structure and function of a typical antibody protein molecule. Both Celltech's expert, Dr Martin, and Medimmune's expert, Professor Rees, give descriptions of this in their respective first reports. What is set out below is based on both descriptions.
10. There are several different types of antibodies. The most common type is called IgG (immunoglobulin G). IgG antibodies all have the same structural pattern: two identical light chains and two identical heavy chains are joined to make one antibody molecule in the shape of a Y. Each light chain contains about 220 amino acids. Each heavy chain contains about 440 amino acids. Thus the whole IgG antibody is a large molecule containing about 1300 amino acids. A light chain is composed of a so-called "variable region" of approximately 110 amino acids in which the amino acid sequence varies depending on the specific antibody and a "constant region" of about 110 amino acids in which the amino acid sequence is largely the same in different antibodies within the same organism. Heavy chains are composed of a variable region of about 110 amino acids and a constant region of about 330 amino acids. Again, the constant regions in such heavy chains are largely the same in different antibodies. An illustration of this Y-shaped configuration is set out in Annex I to this judgment.
11. In this molecule, the variable region of one light chain associates with the variable region of one heavy chain to form an antigen-binding site. The variable regions distinguish one kind of antibody from another and give the antibody its ability to bind to a particular antigen. The antigen-binding sites are at the tips of the Y shaped molecules. The atomic structure of these sites are capable of taking up a shape which fits the structure of the corresponding antigen in a sort of lock and key arrangement. Each antigen binding site is formed by the juxtaposition of six "segments" of amino acids - the so-called Complementarity-Determining Regions, or CDRs - which differ in sequence between different antibodies, and in addition, certain of which vary in length from one antibody to another. Each CDR is made up of a small number of amino acids. There are three CDRs in each of the heavy and light chain variable domains. Additionally, the intervening amino acids in the variable regions of both the heavy chains and light chains that are outside the CDRs form the remainder of the variable region and are known as the "framework" regions. The precise three-dimensional orientation of the CDRs to each other is critical to the shape and structure of the binding site and thus for the proper binding of the antigen. In the drawing in Annex I, each of the three CDRs on a heavy or light chain is represented by a loop.
12. The precise shape (or "conformation") of the CDRs is not determined solely by the sequence of residues from which they are made. They are also affected by the structure and composition of the framework region on which they are located. In an antibody, some of the residues in the framework are on the surface of the antibody when the protein is folded. Others are located within the framework regions. The latter are referred to as packing residues. As a broad general proposition, changing a packing residue within a framework region is more likely to have a major impact on the shape of the framework region than changing a surface residue. However the precise manner in which a change in either a packing or surface residue will affect the conformation of an adjacent CDR and the effect of any such change in conformation will depend on many factors and, even with current knowledge, is not capable of accurate prediction.
13. Two numbering systems have been developed which make it possible to identify in a consistent manner the amino acids in, inter alia, the CDRs. The first is called the Kabat numbering system. It is used to identify amino acids in the heavy and light chains of antibodies. Kabat also defines the segments of the heavy chain and light chain which it refers to as CDRs. A second numbering system was developed by Chothia. It more precisely identifies the boundaries of the CDRs in terms of their three-dimensional relationship to the framework segments. Except for the first CDR of the heavy chain, the "Chothia" loops are entirely contained within the "Kabat" CDRs. Sometimes the number of a residue is preceded by a letter to indicate on which chain it is located. For example, residue 23 on a heavy chain is frequently referred to as H23.
14. Before considering how satisfactory antibodies can be manufactured, there are two other general topics which need to be addressed. The first is the concept of homology. Except for identical siblings, no two humans are identical. All of our genes are extremely similar but there are differences. The genes contain the building codes for all of our numerous proteins. Thus individual proteins within two humans may be the same or very slightly different to each other. As is well known, it is because proteins are not identical from one human to another that human to human transplants sometimes are unsuccessful. The recipient patient's body recognises the grafted organ as foreign. Although it comes from another human, the graft is still sufficiently different to raise an immune response. As a result, to achieve acceptable grafts, first it is necessary to find a human donor who has proteins which are as similar as possible to those in the recipient (so called "tissue matching") and, after the graft has been inserted, the recipient may need a long term course of immuno-suppressive drugs to reduce the risk of an immune response leading to rejection of the graft.
15. In evolutionary terms, we are very closely related to chimpanzees. Their genes and their proteins are very similar to our genes and proteins, but less similar than we are to each other. We are further away genetically from other mammals such as mice and even further from other living creatures like fish and insects. It is possible to compare the amino acid sequence in equivalent proteins from different living creatures. If a protein from one human were to be compared with the equivalent protein in another human their amino acid sequences would be very similar indeed and they will be considered to have a high degree of homology. (The difference between equivalent proteins in humans is what gives rise to the graft rejection referred to above.) If you did the same comparison but between equivalent proteins from, say, a human and a chimpanzee there would still be a high degree of homology but less than that between two humans. Furthermore, since the proteins in each species vary amongst themselves, the degree of difference between the proteins of say a human and a chimpanzee will itself differ depending on which particular human is compared with which particular chimpanzee. Thus a particular protein from human A may be closer, or more homologous, to the equivalent from Chimpanzee X than from Chimpanzee Y and the protein from human B may be less homologous to the chimpanzees' protein than that from human A. The same applies, of course, if one compares human proteins to mouse proteins. There are libraries which contain the DNA sequences (i.e. the parts of the genes) which code for various human proteins. There are frequently a number of such DNA sequences from different individuals. They will produce proteins containing slightly different amino acid sequences. There are also known DNA sequences for proteins in, for example, mice. Once again, there are different sequences derived from different mice. They will be very similar, but not identical to each other. If one is designing a new protein to be made by connecting human parts with mouse parts, it may be desirable to use human and mouse sources which show the least differences. A mouse sequence which is less dissimilar to a human sequence it is replacing in the hybrid protein is likely to be viewed as less "foreign" by a human immune system.
16. The second general concept relates to "substitution" of one amino acid for another. The genetic material in living things is always subject to a process of random mutation. Whether spontaneously or, for example, because of low levels of radiation, the DNA in genes change from time to time. The total genetic sequence in a human's DNA consists of millions of codes. Each amino acid is represented by a sequence of three units of DNA. If one of the units in the DNA is changed by mutation, the amino acid which will be produced will also change, because the new triplet code will relate to a different amino acid. The result will be that the protein which is produced within the cell by "reading" that part of the gene, will contain a different amino acid at one location or, if there is more than one mutation, at a number of locations. Over millions of years there have been enormous numbers of such mutations – it is what has caused humans to evolve from pre-human mammals. Many of such mutations don't work. That is to say, the protein with the altered amino acid is not as good as the original one of which it is a mutation. If, for example, the protein is an enzyme, the mutation may not work at all or as well as the original. If that is the case then, by natural selection, the organism containing the mutation will fail and will disappear. Since mutations are being created spontaneously and randomly all the time, it is to be assumed that similar numbers of mutations will be created for all amino acids. For example, if it is only necessary to change one unit in the triplet code of DNA to turn amino acid A into amino acid B and similarly only one change is necessary to change amino acid A into amino acid C then, over time, there should be similar numbers of proteins produced in which amino acid A has been replaced by B as those where it is replaced by C. If one compared amino acid sequences in proteins, there should therefore be as many B mutants as C mutants. However that is not what happens. This is because some amino acids are much closer in chemical and physical characteristics than others. If amino acid A is very similar to amino acid B but very different to amino acid C, it is much more likely that mutant B rather than C will survive. The more similar A and B are, the more likely it is that the original protein and the B mutant will behave in the same way, with the result that B will not be excluded by natural selection. On the other hand if amino acid C is very different to amino acid A, it is much more likely that the mutant will have a significantly different behaviour to the original protein and it is more likely to fail.
17. It is possible to compare the incidence of amino acids in the same location in equivalent proteins in a number of living creatures. This will give you a picture of which mutations survive best and which survive worst. Where one amino acid is similar to another in the above sense it is said to be a conservative substitution for it. In other words, the substitution of one amino acid by the other is likely to conserve the function of the protein in which the amino acid is located. Conservative substitution is a relative concept. In other words some amino acids are more similar than others. There may therefore be different degrees of conservative substitution. Furthermore, a substitution of an amino acid in a part of a protein which has very little function is less likely to lead to the failure of the mutant protein (and therefore the life form containing it). On the other hand, the substitution of one amino acid by another which is extremely similar in physical and chemical function – i.e. very conservative substitution – may result in a mutant protein that fails because the amino acid is located in a position which is critical to the function of the protein and even a small change there may have a dramatic adverse effect on the protein's performance.
18. In this case, much turns on the interchangeability of two amino acids, serine (Ser) and threonine (Thr). Analysis of natural mutants shows that serine can be conservatively substituted by a number of amino acids. As mentioned above, the degree of interchangeability varies from one amino acid to another. One of the documents referred to in this case ((Matrices for Detecting Distant Relationships by Schwartz) shows that serine can be conservatively substituted by alanine (Ala), asparagine (Asn), aspartic acid (Asp), glycine (Gly), proline (Pro) and threonine (Thr). The concept of conservative substitution was well know at the priority date of the Patent. It was also well known that threonine was similar in many respects to serine and that it was a conservative substitute for it, in the sense explained above.

The Patent

19. The Patent is concerned with the creation of a set of "rules" of general application which will guide workers in the field in the choice of amino acid sequences for hybrid antibodies which will exhibit good antigen binding properties without, at the same time, exhibiting an unacceptable HAMA response. There is no dispute between the parties that those involved in this work were members of a small number of highly regarded teams working at the cutting edge of this technology at the priority date. The manufacture of bespoke antibodies for human clinical use was at its infancy at that time. Indeed, the Patent may have been only the third publication in its specialised field.
20. The Patent starts with a description of the relevant prior art. Because the technology was in its youth, that prior art is limited. It starts by referring to the early experiments which consisted of making "chimeric antibodies" in which the complete variable domain of one antibody is linked to the constant domains derived from another. In particular it refers to work done on taking murine variable domains and grafting them on to human constant domains. As the Patent records:

"Such humanised chimeric antibodies, however, still contain a significant proportion of nonhuman amino acid sequence, i.e. the complete non-human variable domains, and thus may still elicit some HAMA response, particularly if administered over a prolonged period [Begent et al (4)]." (paragraph 0009)

21. In other words, according to the patentee, these humanised chimeric antibodies could work. They bound to the antigen. But the risk of HAMA response was significant, particularly if the chimera was used repeatedly in a human.
22. The next stage was the work done by Dr Winter and his team. This involved taking only the CDR regions from mouse monoclonal antibody and grafting them onto the framework regions of the variable domains of human antibody molecules. In other words, in the illustration in Annex I, the whole of the resulting antibody was human save for the very small CDR loops at the end of the two arms of the molecule, which were mouse sequences. These are referred to as CDR-grafted humanised antibody molecules. The Patent states that its invention relates to humanised antibody molecules made by this, the Winter, technique. This may be significant. Grafting only mouse CDRs onto human antibodies is the Winter design. As will be explained below, the Patent takes this as its starting point. What it teaches is which mouse (or other donor) amino acids, in addition to those in the CDRs, are to be grafted into the hybrid antibody.
23. The Patent goes on to refer to work carried out by Riechmann and his team:

"In Riechmann et al/Medical Research Council it was found that transfer of the CDR regions alone [as defined by Kabat (7) and (8)] was not sufficient to provide satisfactory antigen binding activity in the CDR-grafted product. Riechmann et al found that it was necessary to convert a serine residue at position 27 of the human sequence to the corresponding rat phenylalanine residue to obtain a CDR-grafted product having improved antigen binding activity. This residue at position 27 of the heavy chain is within the structural loop adjacent to CDR1. A further construct which additionally contained a human serine to rat threonine change at position 30 of the heavy chain did not have a significantly altered binding activity over the humanised antibody with the serine to phenylalanine change at position 27 alone."

24. Thus Riechmann had found that just transferring the murine CDRs was not enough. In their new location on the hybrid molecule they would not bind to the antigen. Something had to be done to the supporting framework onto which the CDRs were grafted. Some at least of the amino acids in the framework had to be changed from those present in the human molecule to those present in the donor molecules (in Riechmann's case, the donor was rat). In other words, more than the CDRs had to be donor-like for the antibody to bind properly. As the Patent explains:

"These results indicate that changes to residues of the human sequence outside the CDR regions, in particular in the structural loop adjacent to CDR1, may be necessary to obtain effective antigen binding activity for CDR-grafted antibodies which recognise more complex antigens. Even so, the binding affinity of the best CDR-grafted antibodies obtained was still significantly less than that of the original MAb (i.e. Murine Antibody)"

25. It should be noticed that the Patent does not say that Riechmann provides a detailed analysis of why particular changes of residues from human to rat work. All that is suggested is that changes close to the CDR regions may be necessary. It should be noted that, for reasons which will be explained below, Celltech places some weight on the disclosure that Riechmann's substitution of threonine for serine at position 30 did not appear to alter binding affinity. However it is not suggested that Riechmann offers any explanation as to why the change at residue 30 had no significant effect nor is it suggested that the Patent offers any explanation for this.
26. The Patent goes on to describe work done by Queen and his team. It is not proved that this was published prior art at the priority date. Mr Kitchin QC for the claimant and Mr Watson QC for the defendant agree that, for the purpose of this action, I should treat Queen as if it were not prior art. On the other hand, the description of Queen is an integral part of the teaching of the Patent and helps to elucidate the contribution which the patentee is claiming to have made and wants to monopolise.
27. At paragraph 13 the Patent says:

"Very recently Queen et al (9) have described the preparation of a humanised antibody that binds to the interleukin 2 receptor, by combining the CDRs of a murine MAb (anti-Tac) with human immunoglobulin framework and constant regions. The human framework regions were chosen to maximise homology with the anti-Tac MAb sequence. In addition computer modelling was used to identify framework amino acid residues which were likely to interact with the CDRs or antigen, and mouse amino acids were used at these positions in the humanised antibody.

28. It should be noted that of the available human proteins which could have been used for making the construct, Queen chose the one which was most like the mouse antibody which was the donor of the CDR region, hence the reference to maximizing homology. Queen went on to set out a set of rules which, if adhered to, would be expected to result in the production of antibodies which had the dual benefits of high binding affinity and low risk of HAMA response. They are described in the Patent as follows:

"In WO 90/07861 Queen et al propose four criteria for designing humanised immunoglobulins. The first criterion is to use as the human acceptor the framework from a particular human immunoglobulin that is unusually homologous to the non-human donor immunoglobulin to be humanised, or to use a consensus framework from many human antibodies. The second criterion is to use the donor amino acid rather than the acceptor if the human acceptor residue is unusual and the donor residue is typical for human sequences at a specific residue of the framework. The third criterion is to use the donor framework amino acid residue rather than the acceptor at positions immediately adjacent to the CDRs. The fourth criterion is to use the donor amino acid residue at framework positions at which the amino acid is predicted to have a side chain atom within about 3 A of the CDRs in a three-dimensional immunoglobulin model and to be capable of interacting with the antigen or with the CDRs of the humanised immunoglobulin. It is proposed that criterion two, three or four may be applied in addition or alternatively to criterion one, and may be applied singly or in any combination."

29. The patent goes on to describe an artificial antibody created by Queen using these rules or criteria. It says:

"In the resultant humanised antibody the donor CDRs were as defined by Kabat et al (7 and 8) and in addition mouse donor residues were used in place of the human acceptor residues at positions 27, 30, 48, 66, 67, 89, 91, 94, 103, 104, 105 and 107 in the heavy chain and at positions 48, 60 and 63 in the light chain of the variable region frameworks. The humanised anti-Tac antibody obtained is reported to have an affinity for p55 of 3 x 10⁹ M^-1, about one-third of that of the murine MAb."

30. What is important about this passage is that it identifies which amino acid residues in the human (acceptor) molecule had to be changed into the equivalent murine (donor) amino acids in order to achieve adequate binding. For present purposes it is worth noting that Queen's rules do not require a change to donor residues at positions 23 or 24.
31. Based on the above analysis of what others had done, the Patent indicates the nature of the invention it claims to cover:

"We have further investigated the preparation of CDR-grafted humanised antibody molecules and have identified a hierarchy of positions within the framework of the variable regions (i.e. outside both the Kabat CDRs and the structural loops of the variable regions) at which the amino acid identities of the residues are important for obtaining CDR-grafted products with satisfactory binding affinity. This has enabled us to establish a protocol for obtaining satisfactory CDR-grafted products which may be applied very widely irrespective of the level of homology between the donor immunoglobulin and acceptor framework. The set of residues which we have identified as being of critical importance does not coincide with the residues identified by Queen et al (9)."

32. It should be noted that the patentee does not suggest that all constructs falling outside the protocol will necessarily fail to have useable characteristics of low HAMA response and sufficient affinity. All he is suggesting is that following his protocol will produced good antibodies over a wide field. Furthermore the patentee is promising not just to identify a protocol of general application, but also which residues are of "critical importance". In addition he makes clear that the Patent's list of critical residues does not accord with the residues identified by Queen. As pointed out above, Queen does not suggest that residues H23 and H24 are to be donor. As shall be more fully explained below, the Patent does. They are amongst those "identified as being of critical importance".

The Patent Protocol

33. The Patent contains not only the patentee's protocol but also the "rationale" by which it had been derived. The starting point is an acceptor sequence for the heavy and light chains. That is to say, one must start from the position that all the residues will be those in a human immunoglobulin. This will tend to reduce the likelihood of a HAMA response. The Patent stipulates that the CDRs in the heavy chain (i.e. residues 26-35, 50-65 and 95-102) and the CDRs in the light chain (i.e. residues 24-34, 50-56 and 89-97) must be donor (e.g. from mouse monoclonal antibodies). Thus far, the protocol will result in the production of Winter-type CDR-grafted humanised antibody molecules (see paragraph 22 above). The Patent then proceeds to describe the other substitutions called for under its protocol. This passage is important and is reflected, at least in part, in the claims:

"The positions at which donor residues are to be substituted for acceptor in the framework are then chosen as follows, first of all with respect to the heavy chain and subsequently with respect to the light chain.

2. Heavy Chain

2.1 Choose donor residues at all of positions 23, 24, 49, 71, 73 and 78 of the heavy chain or all of positions 23, 24 and 49 (71, 73 and 78 are always either all donor or all acceptor).

2.2 Check that the following have the same amino acid in donor and acceptor sequences, and if not preferably choose the donor: 2, 4, 6, 25, 36, 37, 39, 47, 48, 93, 94, 103, 104, 106 and 107.

2.3 To further optimise affinity consider choosing donor residues at one, some or any of:

i.. 1,3

ii. 72, 76

iii. If 48 is different between donor and acceptor sequences, consider 69

iv. If at 48 the donor residue is chosen, consider 38 and 46

v. If at 69 the donor residue is chosen, consider 80 and then 20

vi. 67

vii. If at 67 the donor residue is chosen, consider 82 and then 18

viii. 91

ix. 88

x. 9, 11, 41, 87, 108, 110, 112"

34. It is not necessary to consider the required substitutions in the light chain. It follows a similar tiered approach. As Mr Watson puts it, what this is setting out is a hierarchy of substitution at different levels of precision or flexibility. The first level consists of donor residues in all the CDRs (i.e. so that they are of the Winter type) and in all, not some, of positions 23, 24 and 49 of the heavy chain. Residues 71, 73 and 78 are only slightly less rigidly defined than this. They must be all donor or all acceptor. The next level consists of substitutions which, if not mandatory, are preferable (see paragraph 2.3 in the last cited passage from the Patent). The third level consists of fine tuning where the addressee is encouraged to consider substitutions of acceptor by donor residues. From the above, we can see that the mandatory donor residues in the heavy chain (taking the CDRs and the above protocol together) are: 23, 24, 26-35, 49, 50-65 and 95-102. It should be remembered that these are only a very few of the residues in the large heavy chain.
35. The rationale for this hierarchy is set out at pages 8 onwards of the Patent. It deals first with the need for the CDRs to be those of the donor, i.e. to be of the Winter type. Next it considers what it describes as "non-CDR residues which contribute to antigen binding". It explains that the inventor has come to his conclusions by examination of X-ray structures and by carrying out experiments. Some at least of those experiments are described in the specification. To the extent necessary, I shall refer to those below. The Patent then states that the non-CDR residues which contribute to antigen binding can be divided into different groups. In particular it distinguishes between "surface residues near CDR" and "packing residues near the CDRs". As far as the former group is concerned, the Patent states, in relation to the heavy chain:

"Key residues are 23, 71 and 73. Other residues which may contribute to a lesser extent are 1, 3 and 76. Finally 25 is usually conserved but the murine residue should be used if there is a difference."

36. As far as the packing residues are concerned, the Patent states, insofar as material:

"Heavy Chain - Key residues are 24, 49 and 78. Other key residues would be 36 if not a tryptophan, 94 if not an arginine, 104 and 106 if not glycines and 107 if not a threonine. Residues which may make a further contribution to stable packing of the heavy chain and hence improved affinity are 2, 4, 6, 38, 46, 67 and 69. 67 packs against the CDR residue 63 and this pair could be either both mouse or both human. Finally, residues which contribute to packing in this region but from a longer range are 18, 20, 80, 82 and 86. 82 packs against 67 and in turn 18 packs against 82. 80 packs against 69 and in turn 20 packs against 80. 86 forms an H bond network with 38 and 46. Many of the mouse-human differences appear minor e. g. Leu-Ile, but could have an minor impact on correct packing which could translate into altered positioning of the CDRs."

37. Once again, the Patent is drawing a distinction between residues which should be donor, the so called "key" ones, and those which can be either donor or acceptor. It should be noticed from the latter passage that the patentee has contemplated that some specific amino acids in some locations can be replaced by other, specified, residues. These are identified. Even minor changes are stated to be capable of altering the position of the CDRs, that is to say the conformation – and therefore binding ability – of the hybrid antibody. Consistent with the protocol, there is no suggestion in these passages, or anywhere else, that it is permissible, when following the Patent's hierarchy, to substitute acceptor residues for donor residues at locations which it has identified as key or critical.
38. The specification goes on to describe experimental work, including references to X-ray analysis. The overall aim of the Patent is stated as follows:

"In the design of the fully humanised antibody the aim was to transfer the minimum number of mouse amino acids that would confer antigen binding onto a human antibody framework."

39. This is consistent with the rest of the specification. The Patent's objective is to produce fully humanised antibodies, that is to say antibodies which, as far as possible, are made up of acceptor (human) residues. To that end, the minimum number of donor residues were to be added. The rules set out in the Patent help to identify those minimum necessary donor residues.
40. The specification proceeds to describe some of the experiments carried out to determine which residues need to be from the donor. Dr Martin summarised this part of the Patent in his first report as follows:

39. The patent gives some details of experimental work carried out in relation to H23. In an initial set of experiments discussed on pages 17 and 18, the inventors show that CDR-grafted light chain gL221A, in association with CDR-grafted heavy chain gH341A (also referred to as JA185) bound well to antigen (page 18, lines 40-41). However, as discussed in their interim conclusions (page 18, lines 43-55), the same light chain, in association with gH341B (also referred to as JA183) which lacked donor residues at H6, H23 and H24, generated 'only weak binding activity'. From this, they concluded that H6, H23 and H24 were important to maintain good binding. However, from these experiments, no conclusion can be drawn as to the individual importance of these three residues."

41. The last sentence in this passage is Dr Martin's comment, not a statement to be found in the specification. In fact the patentee's interim conclusions are expressed somewhat more strongly:

"INTERIM CONCLUSIONS

It has been demonstrated, therefore, for OKT3 that to transfer antigen binding ability to the humanised antibody, mouse residues outside the CDR regions defined by the Kabat hypervariability or structural loop choices are required for both the light and heavy chains. Fewer extra residues are needed for the light chain, possibly due to the higher initial homology between the mouse and human kappa variable regions. Of the changes seven (1 and 3 from the light chain and 6, 23, 71, 73 and 76 from the heavy chain) are predicted from a knowledge of other antibody structures to be either partly exposed or on the antibody surface. It has been shown here that residues 1 and 3 in the light chain are not absolutely required to be the mouse sequence; and for the heavy chain the gH341B heavy chain in combination with the 221A light chain generated only weak binding activity. Therefore the presence of the 6, 23 and 24 changes are important to maintain a binding affinity similar to that of the murine antibody. It was important, therefore, to further study the individual contribution of the other 8 mouse residues of the kgH341A gene compared to kgH341."

42. This emphasises that certain mouse residues are "required" in the humanised antibodies. It points to seven changes being made – namely 1 and 3 for the light chain and 6, 23, 71, 73 and 76 for the heavy chain – but the first two are not "absolutely essential". The specification then goes on to describe experiments which the patentee apparently believed would demonstrate which of the remaining residues were essential. Some of these experiments are set out in pages 19 et seq. of the specification. The latter consist of making a number of humanised antibodies and testing their binding affinity. For example antibody JA185 is identical to antibody JA203, save that in the former there is a donor residue at 23, whereas in the latter there is an acceptor residue at that location. Similarly JA207 is identical to JA208 save that in the former there is a donor residue at 23, whereas in the latter there is an acceptor residue at that location. Likewise antibody JA205 is identical to antibody JA183, save that the former has a donor residue at location 24 whereas the latter has an acceptor residue there. The performance of these pairs are then compared. In each case the antibody with the donor residue is better than the one with the acceptor residue at the same location. Having analysed all these data, the specification states:

"These and other results lead us to the conclusion that of the 11 mouse framework residues used in the gH341A (JA185) construct, it is important to retain mouse residues at all of positions 6, 23, 24, 48 and 49, and possibly for maximum binding affinity at 71, 73 and 78."

43. It seems to me that the patentee's message is as follows. He has tried to find a uniform construction code which will guarantee, across the antibody field, the production of molecules which will both show high affinity, thereby being effective antibodies, while still exhibiting low HAMA response. In working out his code he has not only considered from a theoretical standpoint the location of various residues (e.g. whether they are packing, CDR or surface residues), but he has also considered structural analysis (X-ray data) and experiments carried out specifically designed to determine how important it is to retain donor residues in particular positions. Some of the experiments are referred to in the specification, but the last quoted passage suggests that not all the experiments have been set out. Among the conclusions he has drawn is that it is necessary to have donor residues at 23 and 24. These are among the very few key or critical residues. This is consistent with the claims.

The Claims:

44. It is only necessary to consider claims 1 and 2. The former, ignoring a proviso which is, for present purposes, irrelevant, reads as follows:

"An antibody molecule having affinity for a predetermined antigen and comprising: a CDR-grafted heavy chain wherein, according to the Kabat numbering system, residues 31 to 35, 50 to 65 and 95 to 102 are donor residues; and a complementary light chain, said CDR-grafted heavy chain having a variable domain comprising predominantly acceptor antibody heavy chain framework residues and donor antibody heavy chain antigen-binding residues, said donor antibody having affinity for said predetermined antigen, wherein, according to the Kabat numbering system, in said CDR-grafted heavy chain, amino acid residues 23, 24, 26 to 30 and 49 at least are additionally donor residues, provided that …" (emphasis added)

45. Claim 2 reads as follows:

"The antibody molecule of claim 1, wherein amino acid residues 71, 73 and 78 in said CDR-grafted heavy chain are additionally donor residues."

46. If one stands back and looks at these claims, it an be seen that claim 1 is for Winter type hybrid antibodies characterised by a requirement that a very small number of antibodies outside the CDR region are to be donor (e.g. mouse). Claims 1 and 2 together emphasise that residues H71, H73 and H78 may or may not be donor. The ones which, at least, are to be present are H23, H24, H26 to H30 and H49. The Patent specification discloses that H23 is the only surface residue which is picked out as needing to be retained as donor (see paragraph 35 above).

The defendant's product

47. There is no dispute that the defendant's product, SYNAGIS, does not fall within the acontextual meaning of the claims. It does not have a donor residue at position 23 in the heavy chain. At that position it retains the residue found in the human antibody. If it had been modified to fall within the wording of the claim, a donor residue, serine, would have been substituted at that location. Instead it has threonine there. Nevertheless, Mr Kitchin argues that there is infringement in accordance with German patent law.

German patent law

48. Both parties provided expert evidence on the relevant German law, that evidence being given by Mr Peter Von Rospatt for the claimant and by Mr Matthias Brandi-Dohrn for the defendant. Although those witnesses disagreed on a few important issues, the parties agreed to forego the right to cross-examine. Their position is that the relevant German case law has been identified and they are content that I should read that for the purpose of resolving relevant areas of dispute.
49. The current German law, like ours, seeks to implement the provisions of Article 69 EPC and the Protocol on Interpretation. Both Mr Kitchin and Mr Watson are agreed that it would not be safe to assume that the outcome of applying the German approach would, in all cases, be the same as the outcome under our domestic law. For that reason it is not possible to take a short cut and apply our law.
50. Under previous German patent law, a liberal approach to the scope of protection applied. There was infringement if the defendant used the "general inventive idea" in the patent. Since the change in domestic legislation in 1981, the German courts have recognised that the EPC has imposed a somewhat more restrictive regime. Now the essential basis for determining the scope of protection is the claims. This was confirmed by the decision of the German Supreme Court, the Bundesgerichtshof ("BGH"), in Formstein (Moulded Kerbstone) BGHZ 98, 12 = GRUR 1986, 803, 6 IIC (1987) p795 (Annex 2):

"In contrast to the legal situation until 1978, the claims are not now merely the starting point but rather the essential basis for determining the extent of protection. Under Sect. 14, 2nd sentence of the Patent Law 1981, the terms of the claim have to be determined by interpretation, taking the description and drawings into consideration. As the Protocol on the Interpretation of Art. 69 (1) EPC (corresponding to sect. 14 of the Patent Law 1981) shows, the interpretation does not only serve the purpose of resolving an ambiguity found in the claims but also of clarifying the technical terms used in the claims as well as the limits and bounds of the invention described therein. . .

The extent of protection of a patent filed after January 1, 1978, is determined as regards the equivalent use of the invention, by the terms of the claims to be ascertained by interpretation. What must be considered is the scope of the invention as it may be recognized by a person skilled in the art. It has to be examined whether a person skilled in the art, based on the invention as claimed, is able to solve the problem solved by the invention as claimed with equivalent means, i.e. to achieve the desired result with different means also leading to that result. Means which the average person skilled in the art, due to his knowledge and skill and based on considerations oriented on the invention as claimed, can identify as being equivalent are generally covered by the extent of the protection conferred by the patent. This is required by the goal of fair remuneration for the inventor under consideration of the aspect of legal certainty."

51. Formstein is not only the foundation for the current approach to the law of infringement in Germany, but it is also the source of an additional defence, called the Formstein objection, which is to the effect that the patent cannot be extended to cover equivalent embodiments if such embodiments were either known or were obvious in the light of the state of the art. This is similar to our Gillette defence. It has been raised by the defendant in this case and will be considered at the end of this judgment.
52. As both experts agree, the German law of infringement has been refined since Formstein, in particular in five decisions of the BGH all given on 12 March 2002. These cases, referred to as "the quintet", are Schneidmesser I and II GRUR 2002, 515, 519 Custodial I and II GRUR 2002, 523, 527 and Plastic Pipe GRUR 2002, 511. I should mention that law in most of the quintet of cases is expressed in identical terms.
53. Before considering that case law, one general point should be made. Under German law, as under English law, it is for the court to construe the patent and to decide the issue of infringement. In doing that it must view the patent through the eyes of the notional skilled, but uninventive, addressee. In this case, both German law experts have expressed views, predictably different, as to whether the defendant's product infringes in accordance with German law. Those views are interesting but do not avoid the necessity for me to make my own mind up on this issue. Furthermore, both Professor Rees and Dr Martin have, to differing degrees, given evidence of how they construe the German patent and the claims. Once again, those views, although interesting and conforming to the arguments advanced by the parties' respective counsel, do not avoid the necessity for me to assess the issue for myself. In addition to that, it is apparent that neither expert could be regarded as equivalent to the notional addressee of the Patent. Both were at the top of the tree in their fields and far more knowledgeable and perceptive than the notional addressee. In particular, Dr Martin analysed the Patent with immense forensic and scientific skill. He detected submerged or implied meanings which, in my view, would not have been apparent to the notional addressee. Indeed one of the features of this trial was the sophistication of the analysis of the Patent and claims in the experts' reports compared with the comparative technical simplicity of the document itself. It needs to be kept in mind that there is no dispute as to the meaning of any of the terminology in the Patent nor as to the meaning of what it describes.
54. According to German law, the first step in determining whether there is infringement is to construe the "semantic" meaning of the claims. If the defendant's product or process falls within that, there is infringement. As indicated above, that is not suggested here. If that type of infringement is not found, the next step is to determine how far away from the semantic meaning, if at all, the scope of protection extends. For this purpose the German courts adopt a three stage test which is similar, but not identical, to the Improver questions.
55. The German approach is most conveniently illustrated by reference to one of the quintet of cases. Like the parties, I will concentrate on Custodial II. It is necessary to start with the German court's approach to the semantic meaning of a claim since, as will be explained below, this forms the foundation for its determination of the non-semantic scope of protection. In particular, it is necessary to appreciate how German courts approach the construction of numbers or figures in a claim. As the BGH explained in Custodial II:

"b) The principles for determining the extent of protection are also to be applied if the patent claim contains specifications of figures or dimensions. Such details participate in the binding nature of the patent claim as the decisive basis for determining the extent of protection. The inclusion of figures or dimensions in the claim shows that they are intended to contribute to determining and hence to delimiting the subject matter of the patent. Consequently, such details must not be regarded as less binding, merely exemplary determinations of the protected technical teaching, as was considered possible in the case law of the legal situation in Germany before entry into effect of Art. 69 of the EPC and the corresponding amendments of national legislation.

c) Like any element of the patent claim, specifications of figures and dimensions are as a matter of principle subject to interpretation. As in other aspects, the decisive factor is the way in which the person skilled in the art understands such details in the overall context of the patent claim, with the description and drawings again being used to illuminate this context. Account must be taken of the fact that specifications of figures and dimensions, by virtue of their objective content, which will also dominate the interpretation by the person skilled in the art, are not uniform but may in different forms refer to factual constellations with very different contents."

56. Thus the German courts take into account the fact that figures and dimensions are, or are capable of being, used to delimit the scope of protection. On the other hand, the way they are interpreted depends upon the context in which they are used. Furthermore, the BGH went on to explain that, because of their ability to be precise, figures and dimensions are different to mere verbal descriptions:

"d) These factors alone prevent the person skilled in the art always ascribing the same fixed meaning to specifications of figures, dimensions or ranges. However, as a rule, he will attach a higher degree of certainty and clarity to such details than to verbal descriptions of the elements of the teaching according to the invention [citation omitted]. Figures as such are unambiguous, while general concepts formulated in linguistic terms constitute a certain degree of abstraction from the object to which they refer. In addition, such concepts, if used in a patent specification, need not necessarily be used in the meaning attached to them by general technical linguistic usage; in this sense the patent specification can constitute its "own lexicon" (c.f. [30 IIC 932 (1999)] - Spannschraube). From the vantage point of the reader skilled in the art, the features given concrete form by means of the specification of figures and dimensions may be interpreted such that the objective success to be achieved according to the invention is determined more precisely and, where appropriate, more narrowly than would be the case for a purely verbal description. Since it is the applicant's responsibility to ensure that everything is contained in the patent claim for which he requests protection, the reader of the patent specification is entitled to assume that this requirement has also been satisfied through the inclusion of figures in the wording of the patent claims. This applies all the more in that the applicant who specifies figures has particular occasion to be fully aware of the consequences of the wording of the claims for the limits of the patent protection requested.

For this reason, a considerably stricter assessment is appropriate than was the practice under the law in Germany before 1978 [citation omitted]. As a matter of principle, an unambiguous figure determines and delimits the protected object exhaustively in this respect; figures above or below are therefore as a rule no longer to be included within the subject matter of the patent claim."

57. It will be seen that, as under English law, the German courts have to take into consideration the fact that the language of the claim is the patentee's choice and that the reader is entitled to assume that he knew what he was doing when seeking to define the monopoly he was seeking.
58. The BGH went on to explain that, even if the semantic meaning contained numerical limits, this did not necessarily avoid the possibility that such limits were subject to usual tolerances. In this as in other respects, whether tolerances are permissible is determined through the eyes of the notional addressee:

"… the decisive factor is the semantic content of the patent claim to be determined with the assistance of the description and drawings. In another context, [a specified angle, e.g. 90^o] may therefore be regarded by the person skilled in the art as a magnitude to be complied with exactly. This also applies as a matter of principle to ranges of figures with limit values [citation omitted]. An interpretation that a value must be complied with exactly will above all correspond to the interpretation of the person skilled in the art following the realization that this is a "critical" value. Accordingly, the way in which a specific figure or dimension in the patent claim is to be understood is a question of the interpretation by the person skilled in the art in the individual case, which is a matter to be determined by the trial judge."

59. A German court must also determine whether the patent monopoly extends beyond the semantic meaning of the claims. As indicated above, this involves a three stage test which is similar to the Improver questions. The parties' experts did not agree on the precise way in which the test should be formulated. The nature of the test was indicated as follows in Custodial II:

"Accordingly, for an embodiment departing from the literal wording of the patent claim to be within the extent of protection, it is not sufficient that (1.) it solves the problem underlying the invention with modified but objectively equivalent means and (2.) specialist knowledge enables the person skilled in the art to recognize the modified means as being equivalent. In the same way that the same effect cannot be determined without focusing on the patent claim, in addition (3.) the considerations that the person skilled in the art must apply must focus on the semantic content of the technical teaching protected in the patent claim in such a way that the person skilled in the art would consider the different embodiment with its modified means as being the specific equivalent solution."

60. It appears to me that the first two of these requirements may be summarised as (i) does the accused device or process work in the same way as, and by equivalent means to, the patent and, if so, (ii) would it be obvious to the notional addressee that the variant is equivalent? The third requirement concentrates on the semantic content of the claim again. It appears that what is intended is, to all intents and purposes, the same as the third Improver question. This is consistent with the following passages in Custodial II:

"As with other elements of the patent claim, the effect according to the claim must not be determined without taking account of the figures and dimensions contained in the claim. Consequently, as a matter of principle it is not sufficient for the inclusion of different embodiments within the extent of protection that in the interpretation of the person skilled in the art the effect of the invention otherwise occurs independently of compliance with the figures. If no other figure than the value according to the claim appears to be equivalent to the person skilled in the art, the extent of protection does not go beyond the semantic content of the patent claim. In the interpretation of the person skilled in the art the effect of the feature determined by figures according to the claim is in such event determined by (exact) compliance with a figure and can therefore necessarily not be obtained by a different figure. In such a case, it is not sufficient for the person skilled in the art to realize that a teaching abstracted from the figures is technically reasonable.

The applicant will not always recognize and exhaust the entire technical contents of the invention; irrespective of the question whether this is legally possible, he is not obliged by law to do so. If, when observed objectively, the patent is restricted to a narrower wording of the claim than would be appropriate according to the technical content of the invention, and thus compared with the state of the art, the specialist in the field is entitled to rely on the fact that protection is correspondingly restricted. The patent holder is then prevented from subsequently claiming protection for something he has not placed under protection. The same applies even if the person skilled in the art realizes that the effect of the invention as such (in the narrower case discussed above) could be achieved beyond the range protected in the patent claim."

61. This appears to be saying that, although it may be obvious to the notional reader that other variants will work as well, they are not covered by the patent if, the reader would conclude, from the teaching of the patent, that the inventor intended not to cover such variants. Similarly the BGH said:

"This Court has repeatedly held that an embodiment cannot fall within the extent of protection of the patent if it fails to make use of a feature of the claim that is essential and decisive for the teaching protected. Accordingly, this is in any event the case if this assessment is the result of essential differences in the effect. However, the case law of this Court has hitherto not included in this assessment cases in which the person skilled in the art's expectation is based not on the technical contents of the feature ("essential and decisive") but instead on the wording of the patent specification as such, i.e., such cases in which the wording in the patent specifications - irrespective of the recognizable technical significance of the feature - communicates to the specialist that the decisive factor for the implementation of the teaching protected by the patent is that the feature is used according to its literal meaning or at least not in the entire range of (assumed) objectively equivalent solutions.

The aforementioned responsibility of the patent holder to ensure that what he requires protection for is set out in the features of the patent claim therefore restricts the protection to what is to be related to the semantic content of its patent claims even in such cases in which the holder - for whatever reasons - has missed this opportunity and the patent, if considered objectively, remains less than a more extensive technical content of the invention."

62. In its Plastic Pipe judgment, the BGH confirmed that this is essentially the same as the third Improver question:

"Ultimately in agreement with the aforesaid, the courts in the United Kingdom, in order to determine an infringement, examine whether the specialist public is entitled to expect and proceed on the basis that according to the patent the decisive factor will be precise compliance with the wording of the patent claim (cf the so called Catnic question: for harmonized law see inter alia Patents Court 1989 F.S.R. 181…..- Improver Corporation v Remington Consumer Products Ltd (Epilady case); Court of Appeal, 1995 R.P.C. 585…-Kastner v Rizla). Related to an individual feature of the patent claim, the issue is whether the feature in question appears to the person skilled in the art as one that can be used exclusively in accordance with the meaning of the words, if the claimed teaching on the technical action to be complied with (cf Court of Appeal ….- Kastner v Rizla Ltd). Such an interpretation is possible particularly in the case of figures and measurements (cf Patents Court ….- Auchinloss v Agricultural and Veterinary Supplies Ltd)".

63. As I understand it, Mr Kitchin, after consultation with Mr Von Rospatt, accepts that under German law an accused device which, to a notional reader, would obviously work in the same way and by equivalent means is not an infringement if the same reader would conclude, from the teaching in the patent, that the patentee intended to exclude variants in general or that variant in particular. The notional reader may think that the patentee has made an obvious mistake, but, as the BGH has pointed out, the latter is not obliged to seek protection for everything which works in accordance with his invention and if he chooses to teach that certain features are essential and decisive, he is free to do so and readers of the patent are entitled to take him at his word.

Does SYNAGIS fall within the Patent?

64. In my view the third question under German patent infringement law is determinative in this case. Both in the claims and the specification the patentee has made clear the extent and the limitations on the freedom to substitute residues in the hybrid antibody. I have set out both the relevant claims and significant extracts from the specification above. The message from all of them is consistent. Some residues have to be donor, some are preferably donor and some are optionally donor. The first group is very small, containing only eight residues other than those in the CDRs. Residue 23 is unambiguously put in that group. The specification explains why that is so. For better or for worse the patentee believes that he has found a recipe which will guarantee effective hybrid antibodies across the whole antibody field. That recipe requires that some residues must be donor. This is not arrived at solely on theoretical grounds. The patentee has gone further and looked at the location of the residues in relation to the CDRs and has analysed their position by means of X-ray analysis. He has also taken the step of testing his theory by experiments, some of which he has set out in the specification. Those experiments show – or the patentee states that they show – that individual acceptor for donor substitutions for the eight residues result in less acceptable antibodies. He tested substitutions of all eight of the critical residues. In my view, on its true construction, the Patent imposes a strict, inviolable requirement in relation to these eight residues that they must be donor. When the claim says that residues 23, 24, 26 to 30 and 49 "at least" are donor residues, it means just that. Nothing less than having all of them as donor residues is acceptable.
65. Notwithstanding the fact that the only issue on infringement turns on the substitution in SYNAGIS of the acceptor for the donor residue at position 23 of the heavy chain, in the references to the claims and the specification set out above, I have referred also to the teaching relating to position 24. The reason for this is as follows. The claimant's case is that conservative substitution of the donor residue at position 23 would be understood to a man skilled in the art to fall within the scope of protection of the Patent. This argument is supported by the teaching relating to position 24 and the evidence of Dr Martin.
66. Dr Martin's view was as follows. First, residues 23 and 24 are close to the CDRs on the heavy chain, the former is a surface residue whereas the latter is a packing residue. This means that change of the latter is more likely to alter the conformation of the adjacent CDR than would change to the former. Furthermore, it is possible to visualise the structure of a typical immunoglobulin antibody and it would be obvious to do so. If this is done it will be seen that between residue 23 and the CDR there are additional residues referred to as the H0 loop. This was illustrated in Dr Martin's first expert report. That illustration is reproduced in Annex II to this judgement.
67. The drawing illustrates only the structure of the very end of the heavy chain. The lime green at the top of the drawing represents the antigen to which the antibody will bind. Three CDR loops are illustrated. They are light blue. Three of the residues which claim 1 says must be donor are identified (H23, H24 and H49). On the left of the drawing between H23 and H24 there is a turquoise projection. This is the H0 loop. Dr Martin says that it would be apparent that this "shields" the adjacent CDR. Therefore changes to the shape and chemistry of residue 23 are less likely to have an impact on CDR-H2. Therefore conservative substitution of residue 23 is unlikely to affect the conformation, and hence binding affinity, of the antibody and furthermore, since it will result in another acceptor residue being present, it almost certainly means that the HAMA response will be no worse. By contrast, changes in residue 24 are much more likely to adversely affect the conformation of the CDRs. The patentee has demonstrated that by making a conservative substitution of this residue (see paragraph 42 above). The result was unacceptable. Dr Martin says that implicit in the specification was the teaching that conservative substitution of residue 23 was likely to work but that similar substitution of residue 24 was shown not to.
68. In addition to this, reliance is placed on the brief reference in the Patent to Riechmann's work referred to at paragraph 23 et seq. above. That shows that a substitution of serine to threonine at position H30 did not significantly affect the binding ability of Riechmann's hybrid antibody. It is said by Celltech that the uninventive addressee would realise that this was a conservative substitution. He would note that it had no impact on antibody activity. He would realise that residue H30 was much closer to the CDR than residue H23 and therefore more likely to alter conformation of the CDR. Therefore he would realise that a similar conservative substitution at H23 would be even less likely to affect CDR conformation and binding ability.
69. There is nothing in the Patent which suggests that the patentee had analysed Riechmann's work in this way.
70. I have no doubt that Dr Martin feels that conservative substitution of residue 23 was worth a try and very likely to work. He reads the Patent with that in mind. I also have no doubt that his views on this are so strong that he reads the absence of any experiment showing the effect of conservative substitution as implicit acknowledgement by the patentee that such substitution would work. As I have said already, Dr Martin is clearly far in advance of the ordinary skilled man in the art. He did not profess to be otherwise. I do not accept that his view represents how the notional addressee would read the Patent. If the patentee had contemplated that any retention of an acceptor residue at 23 (or any of the other seven locations) was possible, he would have said so. He works carefully through the antibody molecule explaining which residues can be substituted, in some cases only allowing substitution when particular amino acids are present. He draws no distinction between 23 and 24, or any of the other six residues, as far as substitutability is concerned. In my view the notional reader is likely to view the teaching that there can not even be conservative substitution of residue 24 as illustrative of how critical retention of the donor residues is in the eight locations. He says nothing to suggest that different considerations apply to residue 23. The addressee would take the claims and the specification to be an instruction from the patentee that there can be no acceptor for donor substitution at residue 23. Furthermore, I do not accept that the uninventive man skilled in the art would read Riechmann as encouraging or supporting conservative substitution at H23. Riechmann does not describe the substitution at 30 as conservative and he ventures no opinion as to why it has no effect on binding, nor does the Patent. Furthermore, it should be remembered that the Patent teaches that conservative substitution of H23's immediate neighbour, H24, does adversely affect binding affinity.
71. I do not accept the claimant's argument that Dr Martin's approach to the teaching of the Patent accords with that of the notional addressee. As Mr Watson argues, at the priority date there was no body of experience to form any judgment as to the effect of residues, or substitution of residues, close to CDRs. Dr Martin accepted under cross-examination that, at that time, no one appeared to have had turned his mind to the concept of conservative substitution in the field of humanised antibodies (see Transcript Day 3 page 156).
72. It appears that Dr Martin's views on using an acceptor residue at location 23 instead of a donor residue is arrived at in spite of the teaching of the Patent. His starting point, as emphasised by Mr Kitchin, is that it was "common general knowledge that serine and threonine, unless they are intimately involved in function, can generally substitute for one another" (see Transcript Day 3 p 165). But the Patent asserts that the residues at location 23, like those at 24 and 49 are intimately involved in function. In the end it appears that Dr Martin's approach, adopted by the claimant, is that the reader would think that the patentee had erred in treating residue 23 as important. This appears from two extracts from his cross-examination:

"Q. Help my Lord with this. The Adair team [i.e. the patentee], we know, looked at crystal structures.

A. Yes.

Q. We know that they looked at KOL.

A. Yes.

Q. Why did they say that 23 was key and, in the first tier in the hierarchy, if they had looked at a picture like illustration 3, they would have seen unambiguously that it was a low priority.

A. I cannot answer that. I do not know why they made the choices that they did, other than the binding data certainly shows you that if you make a serine to lycine change, a non-conservative substitution, then you have a difference."

73. A little later he said:

"MR. WATSON: You are saying that Adair made a wrong analysis on the basis of the crystal structures?

A. I think that that is possible, yes. I think maybe H23 slipped in somehow. They found experimentally that it made a difference when serine lycine change was made, and then grouped it in as surface near the CDRs. It is certainly on the surface. Whether you call it near is a somewhat debatable point. You could argue that it was near if you measure from backbone rather than side chain. My view is that the description of it as being near is somewhat misleading.

Q. So you are the skilled man who goes off and carries out the same work as Adair. He looks at crystal structures.

A. I think this is rather simpler than what was done by Adair. I think they did quite detailed comparative analysis, looking at different structures, checking that the interactions were similar in all of them, rather than just getting a general view of where this is, which is all that I am suggesting as a skilled man one would need to do.

Q. So you eyeball it, conscious that Adair has done much more sophisticated modelling, and you say, "I am going to ignore Adair." Is that what you are saying?

A. Adair has not done sophisticated modelling; he has done analysis of some of the structures. It is the same sort of thing. It is not doing modelling, which was done by Queen. It is just looking at the structures.

Q. Ultimately you are saying that Adair got it wrong and the skilled man would spot that.

A. I think that is a possibility, but he did not get it wrong in that we know that the change at H23 does make a difference. I think that throwing it in with surface residues near to the CDRs is maybe misleading rather than wrong. If it was misleading, it depends what one calls near."

74. The fact that Dr Martin feels that the patentee allowed residue 23 to "slip in" and be treated as more critical than it deserved to be does not help the claimant. Even if that view were shared by the notional addressee – and in my view it would not – one is left with the conclusion that the patentee has imposed a non-acceptor rule for residue 23 when that was not necessary or justified by the science. The German law of infringement does not allow one to correct mistakes made by the patentee. If the patentee has imposed a clear limit on his monopoly which addressees recognise as being unnecessary or based on mistaken science, it is still effective. That is the point of the passages in Custodial II set out in paragraphs 57 and 58 above.
75. For the above reasons I find that SYNAGIS does not fall within the scope of the claims of the Patent. In the light of this, it is not strictly necessary to consider the other arguments advanced by the defendant. Nevertheless, in view of the fact that this case may go further, I will consider these briefly.
76. First I shall consider the dispute between the parties as to whether or not SYNAGIS works in the same way, and by equivalent means to, the antibodies covered by the patent. This question seems to me to raise the same problem as addressed by Hoffmann J in the Improver decision, namely whether something is an equivalent depends upon the level of generality at which one describes the way the invention works. In my view there is nothing in the Patent which suggests that the patentee was claiming that only hybrid antibodies which met his criteria or rules would work. It seems to me that all he was saying was that antibodies made to his rules would work. What happened outside was not his concern. How this impacts on the first part of the German infringement test is not clear to me. For present purposes I will adopt a restrictive view namely that the variant must work in substantially the same way and substantially as well as the version falling within the semantic meaning of the claim.
77. There is no dispute between the parties that the onus is on the claimant to prove this level of equivalence. To meet that reliance was placed on the evidence of Dr Martin, the defendant's own published data concerning its product and on two groups of experiments. I will consider the last of these first.
78. The two groups of experiments run by the claimant consisted of a neutralisation assay and a test, called a Biacore test, for binding affinity. By the time the trial commenced, Mr Kitchin was content to rely primarily on the Biacore test, but without formally abandoning the other experiments. As Mr Kitchin explained in a succinct supplementary skeleton argument, the Biacore test uses a piece of equipment which is known as a Biacore machine the function of which is to measure and compare how much antibody binds to and unbinds from a special chip. After calibration, the experiment starts by binding the antigen of interest to the surface of a chip. The machine then passes the antibody of interest over the chip. It detects the rate of binding and calculates a rate of association or binding (Ka). The antibody is then washed off under controlled conditions. This enables the machine to calculate the rate of dissociation or unbinding (Kd). Kd is then divided by Ka to give a figure (KD) which is the binding affinity of the antibody for that antigen.
79. The experiment was conducted for both the antibody contained within SYNAGIS and for the same antibody but with a donor residue at location 23 on the heavy chain. There is no dispute that the values of Ka, Kd and KD are virtually identical. The claimant says that this proves equivalence.
80. Mr Meade for the defendant argues that the experiments are unreliable. Two major objections were taken. The more significant one was that the two antibodies (synagis-ser and synagis-thr) were applied to the chip in concentrated culture medium. They had not been purified. Mr Meade argues that they should have been applied to the chip in purified form. This was a point put to Professor Easton, one of the claimant's experts, under cross examination. The Professor accepted that the Biacore process calls for the use of test samples of high purity. He accepted that there were impurities present in the experiment. It was put to him that the antibody present could have been a minor part of each sample. He disputed that. Based on this evidence, Mr Meade argues that the results can be ignored. I do not accept that. What is as significant as the questions put to the Professor is what was not put and what was not done. First, although in his second report he said that these experiments demonstrate that the binding affinities of synagis-ser, synagis-thr and commercial SYNAGIS are the same, this was not challenged either directly or indirectly. What the Professor was not asked was whether the impurities meant that the results were unreliable, or the extent of any unreliability. That is a conclusion which Mr Meade invites me to draw without the benefit of help from those who understand this technology. Furthermore Mr Meade argues that the level of impurity is unknown, yet as far as I can see, the defendant neither asked to be told what the level of impurities were or asked for a sample so as to work out the level of impurities for itself. As against this, the only evidence on this issue is contained in a brief passage in the third expert report of Professor Rees. He notes that the operation instructions for the Biacore apparatus requires high purity. He comments that the presence of impurities in the claimant's experiments "allows for the possibility" of interference in the result and that this means that it is "impossible to be certain about the accuracy of the antibody concentrations measured in the various preparations".
81. In my view an attack on the experiments deserves more precision than this.
82. The second point taken by Mr Meade was that the synagis-ser and synagis-thr used for the experiments were produced from a different line of mammalian host cells than those used by his client for the production of its commercial product. For this reason the results relating to the comparison of synagis-ser and synagis-thr does not prove that conservative substitution of residue 23 results in an antibody of substantially the same characteristics. Mr Meade says that the proteins produced by different mammalian cell lines will have different levels of glycosylation. Therefore it may be that in some cases conservative substitution produces substantially identical antibodies and sometimes it does not. The conclusion would be that the claimant would have to carry out parallel experiments in a large number of host cell lines to see whether variants made in each are equivalent. In my view there is nothing in this point. There is no material before me which suggests that glycosylation would be likely to differentially affect the binding affinity of synagis-ser and synagis-thr produced from the same host cell line. It may be, although it is not demonstrated, that an antibody produced in one host cell line will differ from the same antibody produced in another host cell line but that appears to me to be irrelevant to any issue I have to decide.
83. The two other matters relied upon by Mr Kitchin in support of his argument that synagis-ser and synagis-thr are sufficiently similar to meet the requirement of the first issue in Custodial II is the evidence of Dr Martin and the published data concerning the defendant's commercial product. As to the former, Dr Martin's evidence was that one would expect these two antibodies to be substantially the same. He was very confident that that was so although he could not be certain. Although there was considerable cross examination of the witness on the issue of whether or not his expectation or prescience would have been shared by a man in the art at the priority date, there was no substantial attack on the accuracy of his prescience. In other words there was no substantial attack on the assertion that the antibodies were in fact likely to be substantially the same.
84. As far as the characteristics of the defendant's commercial material is concerned, there was no dispute that the published data indicates that it has binding affinity well within the range stated in the Patent to be exhibited by antibodies made in accordance with the claimed protocol.
85. The defendant's approach in relation to this issue has been negative throughout. It has not sought to prove that conservative substitution at residue 23 has a material effect on the binding and HAMA characteristics of the antibody. It has simply challenged the claimant's case that it has no significant effect. In my view it has failed in that challenge. Bearing in mind all three categories of material relied upon by Mr Kitchin, I find that, on a balance of probabilities, conservative substitution of residue 23 has no material affect on the characteristics of the antibody. The first Custodial II question is therefore answered in the claimant's favour.
86. The second Custodial II question is whether it would be obvious to the notional addressee that the antibody with conservative substitution at residue 23 would be equivalent to the one with a donor residue at that location.
87. In answering this, the usual arguments on obviousness are reversed. So as to support its argument of infringement, it is for the claimant to prove that the variation is obvious. Nevertheless it is not suggested that the court's approach to obviousness should be any different for this reason nor that there is any material difference between the German and English approach to obviousness. Mr Kitchin set out in his closing skeleton argument the line of thinking which supports his client's case on this issue. It is as follows:

(i) The patent sets out a protocol which is applicable to a given situation of humanising a mouse monoclonal.

(ii) The skilled person would pick an acceptor sequence of close homology if he could. Assume a variant in which he had ensured that there were donor residues at all the positions called for by the claims except at position 23.

(iii) He or she sees at position 23 the acceptor (threonine) and the donor (serine).

(iv) He or she knows from the common general knowledge that serine and threonine are conservative substitutions. They are of similar size, not charged, nearly the same hydrophilicity/hydrophobicity, and the same polarity (they can make the same H-bond through the presence of the same OH group). The only difference between them is the presence of an extra methyl group in threonine.

(v) He knows from the Patent that it is undesirable to make a change, unless necessary because of the potential HAMA response.

(vi) He knows from the Patent and from visualisation (i.e. by using a computer to generate a virtual model of the antibody molecule) that H23 is a surface residue outside the Kabat and Chothia CDRs. No special warning has been given as to the sensitivity of that position to change.

(vii) He knows from the Patent and the common general knowledge that a serine-threonine change at position 30 in the Chothia loop in Riechmann made no difference.

88. It seems to me that one must have firmly in mind that the obviousness argument starts from the Patent. The question is what would be obvious to the notional address from a reading of that document, taking into account common general knowledge. There may be cases in which he will read the patent in suit and realise that it contains errors or exaggerations. As I have explained above, Dr Martin had such a clear perception of how the various residues were likely to interact that he came to the conclusion that the Patent was misleading or that the patentee had allowed residue 23 to slip into the critical group. Dr Martin's view was that the addressee should have a "remarkably high" level of skill (Transcript Day 3 p 194). I do not accept that this Patent should be treated differently to others and, as indicated above, I do not accept that the uninventive skilled addressee would read the Patent in the way that Dr Martin does. On the contrary he would read the Patent as a major and formidable piece of work. Dr Martin's evidence appears to suggest as much:

"Q. Then when we look at the experimental data, we see that they have shown that changing 23 from human to mouse does improve affinity.

A. Yes.

Q. What they said is from structural analysis they predicted 23 is a key residue and by experiment they have shown that it is.

A. Yes.

Q. You would agree that the patent contains a considerable volume of work?

A. Indeed, yes.

Q. Many man years.

A. Probably, yes. 1989, yes.

Q. There are a lot of constructs they have made.

A. Yes.

Q. Just making one construct is not trivial, is it?

A. Certainly not in 1989.

Q. Leaving aside problems of whether you had a right to work somebody's patent, this is a very nice body of research to be able to carry forward your learning.

A. Yes." (Transcript Day 3 p 161

89. The skilled addressee would have no reason to doubt the accuracy of what the patentee says and he would realise that the patentee had worked out the protocol not only from theoretical considerations but also by determining the structure of the molecules (by X-ray analysis) and by experiments. He had made numerous constructs to test and refine the hierarchy. (See on these points, Dr Martin's cross-examination at Day 3 pages 160 et seq.). I do not think there is anything which would make him suspect that the patentee's apparent belief that residue 23 must be donor was too strongly stated. Everything in the Patent indicates that this was a considered conclusion based on much work. On a balance of probabilities, I do not think he would consider that the rigid donor-only rule for residue 23 should be modified whether by conservative substitution or otherwise. It follows that I reject the argument that this substitution was obvious.
90. Even were each of the seven steps in Mr Kitchin's argument correct, I do not accept that this would lead to a finding of obviousness. His is just the sort of step by step ex post facto argument which Moulton LJ said was unfair in British Westinghouse v Braulik (1910) 27 RPC 209 at 230.
91. Furthermore I do not accept that all seven steps would have been taken by the man skilled in the art. Although as suggested in (iv) the notional addressee would be aware from the common general knowledge that serine and threonine are conservative substitutions, he would know that there were other possible conservative substitutions and he would assume that this was also known to the patentee when he decided on his protocol. Similarly, although as in (v), he would know from the Patent that it is undesirable to make a change, unless necessary because of the potential HAMA response, this would be balanced by the realisation that the patentee's decision to stipulate residue 23 as donor only was made in the face of his own acknowledgement that he was reducing the number of donor residues to a minimum. In other words the patentee's declared desire to reduce the number of donor residues to a minimum emphasises how important it must have been for him to mandate such a residue at position 23. Furthermore I do not accept much of Mr Kitchin's step (vi). Although the evidence proved that visualising the shape of the antibody molecule was within the reach of the skilled worker at the priority date, it was not proved that he would go down that route. This is for two reasons. First, the patentee explains that he has already considered the structure of the molecule. For this purpose he carried out X-ray analysis. There is nothing to suggest that the patentee's conclusions drawn from that analysis are wrong and there would have been no reason for a man skilled in the art to believe that he could have undermined the patentee's conclusions by carrying out molecular visualisations on a computer. Second, even if he did use one of the available computer programmes to generate a picture of the molecule, I accept Professor Rees' evidence to the effect that the addressee could not reliably predict from that that changing residue 23 would be unlikely to alter the conformation of the CDRs. Furthermore I do not accept Mr Kitchin's argument that no special warning was given as to the sensitivity of residue 23 to change. On the contrary, it was picked out in the Patent as one of the residues which was key, or critical, contributed to the antigen binding and was, significantly, near the CDR. Finally I do not accept that he would draw the conclusions suggested in item (vii). I do not accept that he would look at the conservative substitution of residue 30 in Riechmann's paper and conclude that, because that substitution had little effect, a conservative substitution of residue 23 would also have little effect. This is particularly so since the patentee had demonstrated that conservative substitution of the adjacent residue 24 did have an effect.
92. It follows that the claimant fails on Custodial II question 2.
93. Since, for the above reasons, I have held that SYNAGIS does not fall within any claim of the Patent, there is no room for the Formstein objection.