Medimmune Ltd v Novartis Pharmaceuticals UK Ltd & Ors

These appeals concern two judgments of Arnold J given in related proceedings. In action HC09C04770 MedImmune Limited (“MedImmune”), formerly known as Cambridge Antibody Technology Limited (“CAT”), alleged Novartis Pharmaceuticals UK Limited (“Novartis”) had infringed European Patents (UK) Nos. 0 774 511 and 2 055 777 (“the 777 patent” or “the patent”) by selling a product called ranibizumab which is used for the treatment of wet age-related macular degeneration of the eye. Novartis disputed infringement and counterclaimed for revocation of the patents.

The judge held that both patents were invalid on the grounds of obviousness and because the claims were not entitled to priority from the relevant priority document, it being conceded that if they were not entitled to priority then the patents were invalid. He also held the claims did not cover the process by which ranibizumab was made. MedImmune was granted permission to appeal against the judge’s findings in respect of both patents but has only pursued the appeal in respect of the 777 patent. It contends the judge erred in making each of these findings. By way of respondent’s notice, Novartis contends the judge ought additionally to have found the 777 patent invalid for insufficiency, and that it was not infringed for the further reason that ranibizumab is not “obtained directly” by means of the claimed process.

In the second action HC11C01304, Novartis sought a declaration that a Supplementary Protection Certificate (“the SPC”) granted in respect of the 777 patent was invalid. It necessarily followed from the judge’s findings in the first action that the SPC was invalid but the judge held that even on the hypothesis that his earlier findings were wrong, the SPC was nonetheless invalid because it was granted in respect of a product which was not identified in the wording of the relevant claim as a product deriving from the process in question. MedImmune appeals against that further finding.

Upon these appeals we have had the benefit of a scientific advisor, Professor Brian Henderson of University College London. Professor Henderson is Professor of Biochemistry at the UCL-Eastman Dental Institute. Prior to the hearing he gave us three “teach-ins” about the background technology. During the hearing he sat with us and provided us with further assistance in understanding the technical issues. He has also read through this judgment in draft for technical accuracy. At no time has Professor Henderson expressed any opinion to us about any of the matters in dispute. He has been immensely helpful and we are very grateful to him.

At the outset of these appeals we decided to hear full argument on the obviousness and priority issues. Having heard those arguments, we reached the clear conclusion that the judge’s decision on obviousness must be upheld and so informed the parties. We also indicated that it appeared to us that conclusion was dispositive of both appeals and that it would not be in the interests of justice to hear further argument on all the other issues. The parties agreed. We also indicated we would provide our reasons in writing in the usual way. That I now do.

The 777 patent claims priority from a number of applications, but the only one of relevance to these proceedings was filed on 12 November 1990. The patent is concerned with a technique called antibody phage display. The invention described in the patent was developed by CAT (as MedImmune was then known) in close co-operation with scientists at the Medical Research Council (“the MRC”) and the patent is therefore jointly owned by MedImmune and the MRC. The MRC was joined to the actions so as to be bound by their result, but has played no part in the proceedings.

Antibody phage display provides a particularly advantageous way of selecting desirable antibody fragments based upon their ability to bind to particular antigens. The method makes use of particular filamentous bacteriophages. Bacteriophages, often referred to simply as phages, are viruses which infect bacteria. They consist of a protein coat or capsid encapsulating nucleic acid. Filamentous phages are non-lytic, meaning they are extruded from an infected bacterial cell which continues to live and produce more phages. Filamentous phages may be distinguished from lytic phages, such as lambda phages, which are only released when the host cell bursts.

A filamentous phage consists of a single-stranded DNA genome packed in a long tube composed of a protein coat. One end comprises two proteins, one of which is called pIII. This particular protein plays a vital role in the process of infection. It has three domains; two N terminal domains extend from the phage surface, and the third, the C terminal domain, is buried in the phage particle. The phage infects a bacterial cell through long slender proteinaceous appendages called pili on the cell surface. The pilus is first bound by the N2 domain of pIII and then the N1 domain of pIII binds to a bacterial surface protein. The phage DNA is then inserted into the bacterium’s cytoplasm where it is used to make more DNA and for protein expression. When enough DNA and protein have been generated, production of new phage particles begins. Assembly of the new phages occurs in the cytoplasmic membrane of the bacterium and various non-coat proteins create a channel for the passage of each nascent phage out of the cell. The DNA is extruded into the phage body, the capsid completed and the phage released from the bacterial membrane.

The 777 patent describes the generation of filamentous phage particles that have an antibody fragment displayed on their surface by fusion to one of the viral coat proteins, particularly the pIII protein, while the DNA sequence encoding that antibody fragment is included within the genome inside the phage particle. The displayed antibody fragment is correctly folded and functional when fused to the viral coat protein, so that it can bind to its antigen or, more precisely, its target epitope. The display of the antibody fragment on the phage surface means that the ability of the displayed antibody fragment to bind to an antigen can be tested in vitro. Typically antigen is immobilised on a solid substrate and then presented with a library of potential binders, the phage antibody library. Those antibody fragments that bind to the antigen will attach to the immobilised antigen while those that do not bind can be washed away. Once the non-binders have been washed away, the selected phage antibody can be released from the antigen by a process known as elution. The essential elements of the process are depicted diagrammatically below:

One of the key features of phage display is that each antibody fragment is physically associated with the nucleic acid encoding its sequence which is contained within the phage upon which the antibody fragment is displayed. Having selected the relevant phage antibody, the sequence of the antibody fragment polypeptide can readily be elucidated by sequencing the DNA of the phage which displayed it. Further, the technique allows scientists to screen very large libraries of different antibody fragments.

The invention required the bringing together of developments in two technologies, antibody engineering and phage display.

Antibodies are protein molecules which are generated by an animal’s immune system to assist in neutralising or destroying foreign matter, for example bacteria and viruses. The surface of bacteria and viruses contains proteins. When such foreign proteins enter the bloodstream, the immune system of the host recognises them as being foreign and sets about trying to destroy or neutralise them by producing antibodies. The individual foreign molecules are known as antigens and the sites in these molecules recognised by the antibodies are called epitopes. Each antibody locks on to a single epitope but, since each epitope is different, different antibodies have to be made to lock on to each of them.

The structure of antibodies, the way they are generated and their functions are described in the technical primer agreed between the parties and at [39] – [62] of the judgment. For the purpose of this appeal, I would emphasise the following.

Antibodies are generated by specialist cells called B lymphocytes. Each cell can produce only a single design of antibody. Nevertheless, humans have the ability to generate a vast array of different antibodies, in part by the rearrangement of multiple small genes encoding variable parts of the antibody protein during B cell development and also by a process called affinity maturation in which the genes of B cells producing antibodies which have detected a particular antigen are hyper-mutated to produce a further range of antibodies, with those clones producing higher affinity antibodies being favoured.

There are five classes of antibodies, the most prevalent being the class known as immunoglobulin-G (“IgG”). An IgG molecule comprises four chains of amino acids held together by chemical bonds, specifically disulphide bonds. There are two identical long chains, referred to as the heavy (H) chains, and two identical short chains, referred to as the light (L) chains, and they are all held together to create a symmetrical Y-shaped molecule. Each of the chains contains a variable domain located at the ends of the arms of the Y. These are referred to as the V_H (heavy chain variable) and V_L (light chain variable) domains. The remaining domains are said to be constant. The domains are referred to as variable and constant to reflect the extent to which the amino acid sequences in them vary from one antibody molecule to another. In each of the heavy and light chain variable domains there are three hypervariable regions, referred to as complementarity-determining regions (“CDRs”), which define the antigen binding sites of the molecule. They sit on a scaffold of the remaining parts of the variable domains called the framework regions.

This account, whilst technically accurate, does not adequately describe the three-dimensional conformation which the antibody molecule adopts in its natural state. The ability of an antibody to bind to an antigen is critically dependent upon this three-dimensional conformation and, specifically, the antigen-binding site of the antibody (called the paratope) is determined by the three-dimensional conformation of the CDRs and the adjacent framework residues.

As Dr William Huse, one of the experts who gave evidence on behalf of Novartis explained, in seeking sources of antibodies for research, scientists have for decades been able to enrich preparations of antibodies by inoculating a laboratory animal with an antigen of interest and purifying the IgG from its serum. Such antibodies are said to be “polyclonal” because they comprise mixtures of different antibodies, each of which binds to a different epitope on the same antigen.

In 1975 George Köhler and César Milstein, working at the Laboratory of Molecular Biology in Cambridge, developed a technique for making monoclonal antibodies, that is to say antibodies which are homogeneous in structure and so have the same binding properties. It involves the immunisation of a host, usually a rodent, harvesting its B cells and fusing them with myeloma cells to produce large numbers of immortal cells known as hybridomas. Dilution of these cells allows individual hybridomas to be isolated and cultured, a process termed cloning. These individual hybridomas (each of which produces a single antibody) are then screened for the clone producing an antibody with the desired properties. This is a laborious process and a positive result is not guaranteed but, if successful, hybridoma cells derived from the selected clone can be used to make monoclonal antibodies in substantial quantities.

Important though it undoubtedly was and remains, the hybridoma technique has a number of significant limitations. First, it is limited to the antibody repertoire of the host. Second, monoclonal antibodies derived from rodent hybridoma cells are strongly immunogenic in humans and induce an immune response known as the human anti-mouse antibody or HAMA response which can neutralise their biological and clinical effects.

Attempts were therefore made to adapt the hybridoma technique to produce antibodies from human B cells. Unfortunately this did not prove possible for a number of reasons. First, human cell derived hybridomas are not as productive as their rodent equivalents and are unstable. Second, it is difficult to select human antibodies against particular antigens because of the ethical issues involved in immunising human volunteers with antigens.

During the 1980s a good deal of research was therefore carried out into methods for obtaining mouse monoclonal antibodies and then making them compatible with the human immune system while maintaining antigen specificity and affinity. One approach was to fuse the antigen-binding variable region of a mouse antibody to a human constant region to create a “chimaeric antibody”. This preserved the antigen binding ability of the mouse antibody while reducing, but not eliminating, the HAMA immune response. A further development called CDR grafting, or simply humanisation, involved taking the CDRs of a mouse antibody and inserting them into a homologous human antibody framework, or changing the human antibody CDRs to the desired mouse sequences. Unfortunately, it was found that putting mouse CDRs into a human framework resulted in a significant loss of affinity.

The early 1980s saw another significant development, namely the polymerase chain reaction (PCR). This is a powerful and versatile method for copying DNA and allows scientists to amplify and isolate particular genes of interest. In the context of hybridoma technology, PCR allowed scientists to amplify the antibody genes in the hybridoma cells and then humanised or chimaeric antibodies could be made by manipulation of those genes. It also allowed scientists easily to make parts of IgG molecules using recombinant methods. For example, it was possible to make each of the two arms of the molecule, called Fab (Fragment antigen-binding) fragments, or the variable domains, called Fv (Fragment variable) fragments.

These techniques also required an expression system, however. It was known that antibodies and antibody fragments could be expressed recombinantly in eukaryotic cells such as Chinese Hamster Ovary (CHO) cells or mouse myeloma cells. But these were rather cumbersome to manipulate. Bacterial expression systems were recognised as an attractive alternative because of their ability to replicate quickly and produce proteins in large quantities. Scientists had also accumulated a considerable body of knowledge and experience of bacterial genetics. But expression of antibodies in bacteria had proved very difficult, the principal problem being that the antibodies did not fold properly and combined together to form insoluble complexes of proteins called “inclusion bodies” which were generally toxic to the host bacteria.

A breakthrough came in May 1988 with the publication in the same edition of Science of the work of two groups working in parallel. Mark Better and others at International Genetic Engineering Inc (Ingene) achieved the expression in E. coli of a chimaeric mouse-human Fab protein (Science, 240, 1041-1043 (1988))(“Better”); and Andreas Plückthun and Arne Skerra from the Max-Planck-Institut für Biochemie described the expression in E. coli of a functional recombinant Fv fragment (Science, 240, 1038-1041(1988)) (“Plückthun”).

As Dr Huse explained, both groups realised that a key step in the folding of antibodies was the formation of disulphide bonds between the heavy and light chains. This required an oxidising environment, which is normally provided by the endoplasmic reticulum of the B cell. Fortunately, the periplasmic space of Gram-negative bacteria (the region of bacteria between the inner and outer membrane) also provides an oxidising environment. The solution arrived at by both groups was to direct the antibody fragment for secretion into the periplasmic space using a leader peptide from a particular enzyme called PelB. The antibody fragments expressed were soluble and fully functional and their affinity was essentially identical to that of the native antibody.

Shortly afterwards, in October 1988, Robert Bird and co-workers at Genex Corporation described the expression in E. coli of recombinant polypeptides comprising a V_L sequence tethered to a V_H sequence to form a single-chain Fv fragment or scFv (sometimes referred to as single-chain antibody or SCA) (Science, 242, 423-426 (1988)) (“Bird”).

The following year saw a series of further important publications demonstrating the use of PCR techniques to make variable chain cDNA libraries. In August 1989 Professor Richard Lerner’s group at the Scripps Institute with co-workers at Pennsylvania State University and Dr Huse at Stratagene published a paper (Proc. Natl. Acad. Sci. USA, 86, 5728-5732 (1989)) (“Lerner”) describing the cloning of a highly diverse V_H library from mouse spleen cells.

In October 1989 Sally Ward and Greg Winter and co-workers at the MRC Laboratory of Molecular Biology in Cambridge published a paper (Nature, 341, 544-546 (1989)) (“Ward”) describing the use of PCR to generate a diverse library of V_H genes from spleen genomic DNA and then the expression in and secretion from E. coli of those V_H domains.

Finally there is the important paper by Dr Huse, Professor Lerner and co-workers in December 1989 (Science, 246, 1275-1281 (1989)) (“Huse”). This describes the use of lambda phage to express in E. coli a combinatorial library of Fab fragments of the mouse antibody repertoire. It explains that, subject to certain reservations, a lambda phage library of the size of the mammalian antibody repertoire may readily be constructed and that very large numbers of clones may be screened each day.

In 1990 the skilled team would have known of various methods for screening a library of the kind described in Huse, as the judge explained at [71]:

“It will be appreciated from what has been said above that researchers often want to screen large libraries of antibodies or antigens for an antibody or antigen of interest. In 1990 an established technique for doing this was a method called “replica plate-lift”. In this method, the vector containing DNA for each member of the library is inserted into bacteria, which are spread onto a plate containing an appropriate growth medium. The bacteria grow into colonies, each colony expressing that particular member of the library in large quantities. Once the colonies have grown, a nitrocellulose filter is overlaid and the proteins of interest in each colony stick to the nitrocellulose and then can be probed with antibody or antigen probes (depending on whether it was an antigen or antibody library). Bound molecules can be detected by autoradiography. Clones showing a positive signal can simply be cut out and re-plated to grow more colonies and amplify the clone (and also separate out the desired clone from any others that may have also been accidentally re-plated). A variant of this technique involves plating out the bacteria at sufficiently high dilution for a uniform lawn of bacteria to grow, which is then [pockmarked] with holes (“plaques”) where the bacteria have been killed by bacteriophage (as to which, see below). This variant was known as “plaque lift”.”

Around 50,000 plaques can be screened on each plate producing a practical upper limit (in terms of the library size that can be screened) of around 10⁶ different clones.

In the meantime, Professor George Smith at the University of Missouri was pursuing a rather different line of research using filamentous phage vectors. He appreciated that the gene III protein of filamentous phage has at least two functional component parts, an N-terminal domain that is exposed on the surface and is necessary for infection; and a C-terminal domain that is buried in the phage; and he had the idea that foreign polypeptides might be displayed on the surface of the phage without impairing their infectivity by fusion at appropriate positions within the N-terminal domain or between that domain and the C-terminal domain.

In the course of a sabbatical from the summer of 1983 to the summer of 1984, Professor Smith cloned fragments of the gene for the EcoRI endonuclease gene into phage and found he was able to display a 57 amino acid fragment of this protein on the surface of the phage, and he demonstrated it bound to an anti-EcoRI antibody. Professor Smith appreciated the power of this technique and believed that a thousand-fold enrichment could be expected after a single round of purification and hoped that antibodies might be used to isolate desired clones from a library of random inserts in a fusion-phage vector. This seminal work was published in Science in 1985 (“Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface”, Science, 288, 1315-1317 (1985)) (“Smith”).

Over the course of the next few years, Professor Smith and his graduate student, Steve Parmley, developed an improved phage-display vector based upon fd-tet, a version of fd filamentous phage modified by the insertion of a gene conferring resistance to the antibiotic tetracycline, which allows bacteria to be selected for successful incorporation of the vector. A further significant improvement was to move the cloning site from between the N-terminal and C-terminal domains to a position two to three amino acids from the N-terminal end, leaving the majority of the gene III sequence uninterrupted. This work was published in Gene in September 1988 (“Antibody-selectable filamentous fd phage vectors: affinity purification of target genes”, Gene 73, 305-318; 1988”) (“Parmley & Smith”).

This paper was the basis of one of the obviousness attacks on the patent at the trial and was the focus of a good deal of attention on the appeal. The authors describe the production of five fusion phage and a series of experiments to assess the particle yield of the clones, their infectivity and the ability of the peptides encoded by the inserted sequences to be bound by known antibodies. It was found that none of the fusion proteins had a significant effect on the yield of phage particles produced following propagation in E. coli. But a number of the inserts did affect the infectivity of the phage particles. Nevertheless, the authors describe affinity purification – which they call “biopanning” – of phage carrying a target insert from a library containing a 10⁸ fold excess of phage without the insert, using only very small amounts of antibody.

The discussion section of the paper is of some importance and contains the following description of the limitations of the technique:

“(a)

fUSE vectors display foreign antigenic determinants with little loss of phage function

The new fusion phage vectors, fUSE1 and fUSE2, accept inserts in gene III with little or no loss of phage function; inserts are stable. The foreign aa [amino acids] encoded in the inserts are expressed on the surface of the phage; two clones carrying fragments of a target gene were shown to express determinants recognized by antibody to the gene product. These results demonstrate the ability of fUSE vectors to accept inserts up to 335 bp (perhaps more) and express the foreign aa encoded in the inserts on the surface of the virion.

Some inserts by their very nature will affect pIII function. Inserts that contain anchor domains or other hydrophobic segments may stop transfer of pIII into the host membrane (Davis and Model, 1995) and presumably would not be tolerated. Inserts that exceed 335 bp may lead to excessive breakdown of the fusion protein or otherwise impair pIII function, so for the time being we recommended [sic] using fragments of 100-300 bp.”

Nevertheless, the authors explained the benefits of phage display as compared to other more conventional methods of screening epitopes in these terms:

“In fUSE vectors, in contrast, the amino acids encoded by the foreign inserts are displayed on the virion itself. This allows recombinant phage to be purified in infectious form by affinity to antibody; thus antibody is used directly to select for the desired clones.”

The authors were particularly interested in the investigation of epitopes, including their mapping with a view to designing, for example, vaccines, as appears from this further passage:

“(e)

Prospects for an ‘epitope library’

An ‘epitope library’ would contain, say, 10⁸ clones expressing a short, synthetic random coding sequence. Such a library might be expected to contain clones reactive with almost any anti-protein antibody, since protein epitopes are typically about 6 aa long and virtually all 64 million 6 aa epitopes would be represented multiple times in different contexts. Biopanning the epitope library with an antibody of interest and sequencing the inserts in a number of positive clones might provide information about the epitope(s) recognized by the antibody, information that could be used, for example, to design vaccines, identify genes, or map epitopes without the need to clone the relevant natural gene fragments.”

Finally, the authors recognised that the true power of the technique could not be exploited fully until they had the ability to generate much larger libraries, noting that libraries with more than 10⁶ clones were difficult to achieve with any vector that must be introduced into host cells by transfection because of the limited capacity of competent cells.

From 1989 to 1990, Dr Jamie Scott, a post doctoral researcher in Professor Smith’s laboratory, therefore set about constructing a large phage peptide library. She and Professor Smith called this library an “epitope library” and demonstrated that antibodies could be used to affinity-select rare clones displaying antibody-binding peptides. This work was published in 1990 in Science (“Searching for Peptide Ligands with an Epitope Library”, Science 249, 386-390) (“Scott & Smith”). The authors describe the construction and characterisation of a mixture of fusion phage theoretically displaying approximately 4 x 10⁷ different hexapeptide epitopes and then the use of biotinylated monoclonal antibodies to select clones by successive rounds of affinity purification.

In the meantime, in 1988, Professor Smith had read Bird - the paper by Robert Bird and his co-workers at Genex Corporation. He was also aware of the work carried out by Dr Winter’s group in Cambridge and Professor Lerner’s group at the Scripps Institute to express recombinant antibody genes in bacteria. He recognised that one of the aims of these workers was to generate and screen libraries of recombinant antibodies for the ability to bind to a particular antigen but appreciated that many of the recombinant antibody constructs were relatively large, comprising about 500 amino acids and two polypeptide chains. The single-chain antibodies described by Robert Bird were, by contrast, relatively short, comprising only about 250 amino acids and one polypeptide chain. It was apparent to Professor Smith that these scFvs might be displayed on his phage display vectors and then screened using the same affinity-selection approaches that had worked for phage peptide libraries.

Professor Smith therefore amended a grant application that he had submitted to the United States National Institute of Health to include the generation of a library of fusion phage displaying, not foreign antigens, but rather antibodies with a great diversity of antigen-binding specificities. These cloned antibodies would be scFvs and would be generated in vitro using degenerate synthetic oligonucleotides. He also made contact with Robert Bird to discuss the possibility of expressing scFvs as part of a phage coat protein and acquiring some of his antibody clones.

On 26 April 1990, Professor Smith gave a presentation at a conference entitled “Vectors for the Cloning Immune Response” which was held at the Banbury Center at the Cold Spring Harbor Laboratory, New York. It was organised by Professor Lerner and Dr Winter, although Dr Winter did not actually attend the conference. This is the presentation which the judge found rendered the patent obvious. The judge made a series of important findings as to the substance of that presentation at [356] – [374] of his judgment. The following are particularly relevant.

First, participants received a letter summarising the background to the conference in these terms:

“Recently the separate groups of Winter and Lerner have cloned and expressed the antibody repertoire in E. coli using plasmid or lambda phage technology (Sastry et al. Proc. Natl. Acad. Sci. USA 1989; Ward et al Nature 1989; Orlandi et al Natl. Acad. Sci. 1989; Huse et al, In Press, Science). Extensions of their results open the possibility of circumventing the hybridoma methodology to prepare antibodies and ultimately may lead to a generic antibody library which would obviate the need to immunize animals. As one begins to approach these goals, questions concerning the size and screening of the antibody repertoire emerge. This meeting will address these issues and hopefully speed up progress toward creating antibodies in vitro.”

At the conference, Professor Smith described in general terms the work the subject of Scott & Smith and introduced at an early stage the idea of putting an scFv on a fusion phage as a means of creating a paratope, that is to say antibody, library.

Professor Smith then discussed the epitope library referred to in Scott & Smith and talked about his technique of biopanning by affinity purification on a plate as described in Parmley & Smith. He explained that his group had shown that the presence of a peptide on pIII did not destroy the infectivity of the phage. He also explained that this enabled fusion phage, which were eluted from the plate, to be propagated and sequenced to identify the nucleotide sequence corresponding to the peptide expressed on the fusion phage.

Professor Smith then said that the same approach “might be useful in screening an antibody library with antibody displayed on pIII and antigen on the plate”, reversing the roles of the antibody and antigen. He illustrated this proposal with the following slide:

I would observe that this slide shows three different phage particles, each displaying a different antibody fragment, only one of which binds to the biotinylated antigen.

Professor Smith explained that his group were going to test this approach by attempting to express an anti-fluorescein SCA (that is to say an scFv) on pIII. He also suggested that, if such an experiment showed that fusion phage expressing this SCA could bind to fluorescein antigen in a panning experiment, the approach might work to identify SCAs from a library of SCAs.

The judge then turned to various concerns raised by Professor Smith and possible solutions should problems arise, and made these findings:

“366.

At this point Professor Smith raised the question “Will it fold right?”, recognising the possibility that in such an environment the conformation of the SCA might be compromised. He explained that some of the pIII protein was embedded in the inner membrane, with the bulk in the periplasm, which was then transferred into the growing virion. He went on, however, to say that, if folding were a concern, use could be made of a known approach for producing recombinant proteins in bacterial cells, where a process of denaturation using 6M urea at pH 2.2 and renaturation is employed in order to achieve the required folding. Such an approach was potentially feasible in the context of fusion phage because phage could survive the conditions involved.

367.

Professor Smith then said that, although degradation might be an impediment to the successful display of large peptides, as had been suggested in Parmley & Smith, the contrary view was that it might be of benefit. The “lon” and “deg” systems of bacteria exist to remove malfolded proteins. He therefore suggested that such degradation might result in the removal of malfolded SCAs, which might work in the experimenter’s favour and make the detection of correctly-folded SCAs (if they were not degraded) cleaner.

368.

Professor Smith also suggested that, if the SCA on the surface of the fusion phage interfered with the phage’s ability subsequently to infect bacteria, it could simply be removed with trypsin.”

Professor Smith also introduced the idea that an SCA which had been isolated from a fusion phage could repeatedly be mutated and selected for better binding.

Finally, Professor Smith described two approaches to making a library. One was to construct antibody libraries from the natural repertoire. The other was to make a synthetic library using degenerate oligonucleotides to produce random CDRs with various specificities, an approach similar to the one he had used to create his hexapeptide library.

Despite the various concerns that Professor Smith had, he explained in the course of his cross-examination that he was trying to sell antibody phage display as an idea to try, that he was trying to be upbeat about the technology and that he was interested in getting people to try it for themselves.

Professor Smith eventually obtained an scFv clone from Genex in May 1990 and over the course of that summer a graduate student, Ned Watson, began a phage antibody experiment under his supervision. The scFv was designed to bind to fluorescein and it was compared to a mutant which did not bind to fluorescein. The experiment showed that the fusion phage with the anti-fluorescein scFv was able to bind specifically to fluorescein rather than rhodamine, whereas the mutant did not bind to fluorescein. However, discrimination between specific and non-specific binding was less than 100-fold but it did demonstrate the principle of antibody phage display.

Professor Smith did not pursue the work further because by then the inventors of the patent, John McCafferty and co-workers (including Dr Winter) at the MRC Laboratory and CAT had published the results of their own work in a letter to Nature (McCafferty et al., “Phage antibodies – filamentous phage displaying antibody variable domains”, Nature, 348, 6 December 1990, 552-554). As Professor Smith said in a passage in his witness statement upon which he was not cross-examined:

“I was satisfied that antibody phage display … was being pursued by the CAT group and by others, including the group at Scripps. I also did not regard myself as having any claim to scientific priority in the field of phage antibodies, because I considered phage display of antibodies to be a fairly obvious extension of the newly-invented single-chain antibodies and our published work on phage display in general (especially … Parmley and Smith …) to the ongoing work by Greg Winter’s group, Lerner’s group at Scripps, and others, whose efforts to ‘clone the immune response’ were very prominent in the community of molecular biologists by the end of 1988.

I have been asked whether I would have continued work on phage display of antibodies were it not for the fact that CAT and Scripps were also working on it. I am sure that I would have done so, as I believed that the strategy was sound and the approach had enormous potential.”

I must now outline aspects of the teaching of the patent because they bear on one of the arguments advanced by MedImmune as part of its attack on the finding of obviousness, namely that the judge mischaracterised the nature of the skilled team to whom the patent is addressed. They are also relevant to the priority issue.

The specification begins at paragraph [0001] by explaining that the invention relates to methods for producing members of specific binding pairs and to binding molecules produced by such methods.

Paragraphs [0002] – [0003] describe the important advance of monoclonal antibodies and how these are traditionally made by establishing an immortal mammalian cell line derived from a single immunoglobulin producing cell secreting one form of antibody molecule with a particular specificity.

Paragraphs [0004] – [0006] describe the structure and functional elements of antibodies and how it has been shown that the function of binding antigens can be performed by fragments of a whole antibody including an Fv fragment consisting of the V_L and V_H domains of a single arm of an antibody. It continues that although the two domains of an Fv fragment are coded for by separate genes, it has proved possible to make a synthetic linker enabling them to be made as a single protein chain.

Paragraphs [0007] – [0010] explain that although monoclonal antibodies produced by human immortal cell lines hold great promise for the treatment of a wide range of diseases, such antibodies have a number of limitations. The first is that immortal antibody producing human cell lines are very difficult to establish and give low yields of antibody. By contrast, equivalent rodent cell lines yield high amounts of antibody but, if administered to humans can generate the HAMA response, leading to harmful hypersensitivity reactions.

The second problem mentioned is a practical one. The number of different specificities expressed at any one time by lymphocytes of the murine immune system is thought to be about 10⁷, and in the case of the human immune system the figure is even higher at about 10¹². However, the hybridoma technology only allows the sampling of about 10³ to 10⁴ specificities.

Paragraphs [0011] – [0013] describe how these problems have, at least in part, been addressed by the application of recombinant DNA methods to the isolation and production of antibodies and fragments of antibodies with antigen binding ability in bacteria such as E. coli. The description continues that the use of bacterial cells has allowed the production of large numbers of binding molecules. It has also afforded an opportunity to produce tailor-made antibodies and fragments. But many of the problems presented by systems based upon immortalised cells persist. In particular, it is still impractical to screen the very large numbers of clones necessary to begin to sample the range covered by natural antibody variation.

Against this background, paragraph [0014] says there is a need for a screening system that will allow the sampling of very large numbers of antibodies or fragments, and the rapid transfer of the DNA encoding them to the next stage.

Paragraph [0015] explains that the most attractive candidates for large scale screening are prokaryotic organisms because they grow quickly, are simple to manipulate and create large numbers of clones. However, although the co-expression in a single host cell of variable H and L chains has been disclosed, that expression produced insoluble protein which required extensive processing to generate antibody fragments with binding activity. The paragraph continues that it has been shown that antibody fragments can be secreted through bacterial membranes by using an appropriate signal peptide but this method requires the screening of individual clones for binding activity in the same way as murine monoclonal antibodies.

Paragraphs [0017] – [0022] refer to various papers and patent applications, some published after the date of filing of the relevant priority document, describing the use of bacteriophages. Of these, I should mention an application by Protein Engineering Corporation WO 90/02809 which proposes the insertion of the coding sequence for bovine pancreatic trypsin inhibitor (BPTI) into a coat protein of the filamentous bacteriophage M13. However, the description continues, the proposal was not shown to be operative. In particular, there was no demonstration of the expression of BPTI sequences as fusions with the coat protein and display on the surface of M13.

The problem of how to use bacteriophages is developed in paragraph [0023]. It says that the protein must be inserted into the phage in such a way that the integrity of the phage coat is not undermined, and the protein itself must be functional and retain its ability to bind its antigen. Then, in paragraph [0024], the description continues that, surprisingly, the applicants have been able to construct a bacteriophage that expresses and displays at its surface a large biologically functional binding molecule which remains intact and infectious. Where the binding molecule is an antibody or an antibody derivative or fragment, the applicants call the package a “phage antibody” or “pAb”.

Paragraph [0025] explains that pAbs have a range of applications in selecting antibody genes encoding antigen binding activities. For example, they may be used to screen large combinatorial libraries and that rounds of selection using pAbs may help in rescuing higher affinity antibodies from such libraries. It continues that pAbs may also allow the construction of entirely synthetic antibodies. Alternatively, they may allow the construction of antibodies which have some synthetic sequences and some naturally derived sequences. Moreover, libraries of pAbs could be generated and particular pAbs selected by binding to an antigen, hypermutated in vitro in the antigen-binding loops or V domain framework regions, and subjected to further rounds of selection and mutagenesis.

Paragraph [0081] defines the term “derivative”:

“This is a substance which derived from a polypeptide which is encoded by the DNA within a selected bacteriophage particle. The derivative polypeptide may differ from the encoded polypeptide by the addition, deletion, substitution or insertion of amino acids, or by the linkage of other molecules to the encoded polypeptide. These changes may be made at the nucleotide or protein level. For example, the encoded polypeptide may be a Fab fragment which is then linked to an Fc tail from another source. Alternatively markers such as enzymes, flouresceins etc may be linked to eg Fab, scFv fragments.”

Paragraph [0082] sets out the elements of the method of claim 1 of the patent. Paragraph [0091] explains that in this method the gene sequence encoding the binding molecule of desired specificity is separated from a general population of filamentous phage particles having a range of specificities because it binds to a specific target. Paragraph [0092] continues that the bound particle may be recovered by washing with an eluant and that variation of the washing conditions permits the recovery of particles with different binding affinities for the epitope. The nucleic acid may then be recovered from the particle and used or modified to produce a derivative.

Paragraphs [0107] - [0115] explain that the applicants have recognised that gene III of filamentous phage is an attractive possibility for the insertion of biologically active foreign sequences but that, prior to the present application, no substantially complete domain or folded unit had been displayed on phage. They continue that they chose to insert the gene coding sequence for biologically active antibody fragments into the gene III region after amino acid 1 of the mature protein and that, in order to retain the structure and function of gene III, the majority of the original amino acids were synthesised after the inserted immunoglobulin sequence. Surprisingly, they say, by manipulating gene III, they have constructed a bacteriophage that displays on its surface large biologically functional antibody, enzyme and receptor molecules while remaining intact and infectious. Further, phages bearing antibodies of desired specificity can be selected from a background of phages not showing that specificity.

There follows a series of examples which demonstrate that the claimed method can achieve a thousand fold enrichment in one round of purification and a million fold enrichment in two such rounds. That brings me to claim 1, the only claim in issue, which reads (broken down into integers):

“[1] A method for producing a molecule with binding specificity for a particular target, which method comprises:

[2] producing a population of filamentous bacteriophage particles displaying at their surface a population of binding molecules having a range of binding properties,

[3] wherein the binding molecules comprise antibody antigen binding domains for complementary specific binding pair members,

[4] wherein the binding molecules are displayed at the surface of the filamentous bacteriophage particles by fusion with a gene III protein of the filamentous bacteriophage particles,

[5] and wherein each filamentous bacteriophage particle contains nucleic acid encoding the binding molecule expressed from the nucleic acid and displayed by the particle at its surface;

[6] selecting for a filamentous bacteriophage particle displaying a binding molecule with a desired binding property by contacting the population of filamentous bacteriophage particles with a particular target

[7] so that individual binding molecules displayed on filamentous bacteriophage particles with the desired binding property bind to said target;

[8] separating bound filamentous bacteriophage particles from the target;

[9] recovering separated filamentous bacteriophage particles displaying a binding molecule with the desired binding property;

[10] isolating nucleic acid encoding the binding molecule from separated filamentous bacteriophage particles;

[11] inserting nucleic acid encoding the binding molecule, or a fragment or derivative thereof with binding specificity for the target, in a recombinant system; and

[12] producing in the recombinant system separate from filamentous bacteriophage particles a molecule with binding specificity for the target,

[13] wherein the molecule is said binding molecule or a fragment or derivative thereof with binding specificity for the target.”

It is well established that a patent specification is addressed to those persons likely to have a practical interest in the subject matter of the invention, and such persons will have practical knowledge and experience of the kind of work in which the invention is intended to be used. They will also be equipped with the common general knowledge in the art, a matter to which I return below.

As the judge explained, in this case there was a dispute as to the identity of the team to whom the patent is addressed. MedImmune contended it is addressed to a team consisting of an immunologist and a molecular biologist, perhaps assisted by a chemist. Novartis argued the patent is addressed to a team of scientists with differing backgrounds in areas such as immunology, in particular antibody structural biology, molecular biology and protein chemistry, but with a common interest in antibody engineering. As the judge identified, the essential difference between the two formulations lies in the degree of specialisation of the team in the field of antibody engineering.

The judge preferred Novartis’ submission on the basis that the evidence showed that real research teams in the field were teams of the kind contended for by Novartis. He added that, in his view, the specification of the patent is consistent with this characterisation of the skilled team.

MedImmune contended that the judge fell into error in so finding because the invention has a broad application and is not confined to antibody engineering. It continued that expertise in immunology and molecular biology is sufficient to implement its teaching.

I have no doubt that the judge identified the skilled team correctly. As Jacob LJ explained in Schlumberger Holdings Ltd v Electromagnetic Geoservices AS [2010] EWCA Civ 819, [2010] RPC 33 at [42], the court will have regard to the reality of the position at the time and the combined skills of real research teams in the art. A little later, at [53], he continued that where the invention involves the use of more than one skill, if it is obvious to a person skilled in the art of any one of those skills, then the invention is obvious. Finally, at [65], he explained that in the case of obviousness in view of the state of the art, a key question is generally “what problem was the patentee trying to solve?” That leads one in turn to consider the art in which the problem in fact lay. It is the notional team in that art which is the relevant team making up the person skilled in the art.

The judge found that by 1990 antibody engineering was an established field. The three leading teams were those led by Dr Winter at the MRC Laboratory of Molecular Biology and CAT, by Professor Lerner at the Scripps Institute and by Andreas Plückthun at the Max-Planck-Institut für Biochemie. Other teams were also interested, including the research group led by Professor Stefan Dübel at the Deutsches Krebsforschungszentrum and teams at Genentech, Genex Corporation, Ingene, SmithKline Beecham and Genetics Institute. All of these teams were likely to have a practical interest in the subject matter of the invention, in methods for preparing binding molecules including, specifically, antibodies and fragments of them, and selecting those with specificity for particular antigens. They had a need for a system which would allow them to screen very large numbers of different binding molecules. The invention was therefore plainly of interest to antibody engineers and the fact that it may have a broader application is neither here nor there.

The common general knowledge of the notional skilled addressee is all that knowledge which is generally known and generally regarded as a good basis for further action by the bulk of those engaged in a particular art: Beloit Technologies Inc v Valmet Paper Machinery Inc [1997] RPC 489 at 494-495. It also includes all that material in the field in which the skilled addressee is working which he knows exists, which he would refer to as a matter of course if he cannot remember it and which he understands is generally regarded as sufficiently reliable to use as a foundation for further work: Raychem Corporation’s Patent [1998] RPC 31 at 40; [1999] RPC 497 at 503-504 .

There was very little dispute between the parties as to the scope of the common general knowledge of the skilled addressee, that is to say the skilled team to which I have referred. It included all the matters set out in the technical background other than those relating to phage display. Specifically it was accepted that, in relation to antibody engineering, it included the Better, Plückthun, Bird, Lerner, Ward and Huse papers. The skilled team would therefore have been well aware of the important advances in the late 1980s involving the use of PCR to clone gene sequences encoding antibody fragments, the generation of libraries of vectors coding for these fragments, the use of these vectors to infect E. coli and the expression of functional antibody fragments. But selecting a molecule of interest from a large library of antibody fragments using the available techniques presented, in the words of Dr Jean-Luc Teillaud, one of the experts who gave evidence on behalf of MedImmune, a formidable challenge.

The dispute between the parties, such as it was, related to whether phage display formed part of the common general knowledge. The judge found the basic concept of phage display at a high level was common general knowledge by November 1990. It was, he held, an established technique, although not one which was in routine use.

MedImmune contended the judge fell into error in making this finding because there had been only six published studies using the technique. As a matter of principle, it continued, a technique so little used in practice cannot amount to common general knowledge.

I am unable to accept this submission. As Aldous LJ explained in Beloit at 497, the fact that a concept has not been used at all does not mean that it cannot form part of the common general knowledge, though it makes it unlikely. In the present case, by November 1990, Professor Smith’s group had published the three papers to which I have referred, namely Smith, Parmley & Smith and Scott & Smith, and at least three other groups had published work on phage display. Moreover, the judge found on the evidence that other groups were also using the technique, including in relation to antibodies, but had not yet published their work. In all these circumstances the judge had ample material upon which to find as a fact that the concept of phage display would have been known to the person skilled in the art.

I agree with the comprehensive judgment of Kitchin LJ. However, I wish to add a few words of my own on the question of obviousness. The EPC provides:

“Art 52 Patentable Inventions

(1)

European patents shall be granted for any inventions which are susceptible of industrial application, which are new and which involve an inventive step.

Art 56 Inventive Step

An invention shall be considered as involving an inventive step if, having regard to the state of the art, it is not obvious to a person skilled in the art.”

These articles find their domestic equivalent in sections 1 and 3 of the Patents Act 1977. As Jacob LJ pointed out in Actavis UK Ltd v Novartis AG [2010] EWCA Civ 82 [2010] FSR 18 (§ 17):

“So at bottom the question is simply whether the invention is obvious. Any paraphrase or other test is only an aid to answering the statutory question.”

The same point is made in Johns-Manville Corporation’s Patent [1967] RPC 479, which is the starting point in domestic law of the idea of “obvious to try”. In that case Diplock LJ said:

“I have endeavoured to refrain from coining a definition of “obviousness” which counsel may be tempted to cite in subsequent cases relating to different types of claims. Patent law can too easily be bedevilled by linguistics and the citation of a plethora of cases about other inventions of different kinds. The correctness of a decision upon an issue of obviousness does not depend upon whether or not the decider has paraphrased the words of the Act in some particular verbal formula. I doubt whether there is any verbal formula which is appropriate to all classes of claims.”

In the same case Willmer LJ said:

“I would, however, desire to associate myself particularly with what Diplock, LJ said as to the undesirability of coining phrases for the purpose of paraphrasing the words of the Act.”

These sentiments seem to have been largely ignored by the profession. It cannot be said too often that the statutory question is: was the invention obvious at the priority date? It is not: was it obvious to try? In my judgment too much elaboration of the statutory question has been attached to it. The questions of the degree of expectation of success and the length of time thought to be needed to undertake a trial have taken on lives of their own. I think that this happened in our case. Insistence on the statutory question is not a novel thought. It is also an obvious one: see Conor Medsystems Inc v Angiotech Pharmaceuticals Inc [2007] EWCA Civ 5 [2007] RPC 20 (§§ 44, 45 per Jacob LJ, approved on appeal: [2008] UKHL 49 [2008] RPC 28 § 42 per Lord Hoffmann; § 49 per Lord Walker; § 55 per Lord Neuberger). In Generics (UK) Ltd v H Lundbeck A/S [2007] EWHC 1040 (Pat) [2007] RPC 32 (§72) Kitchin LJ (as he then wasn’t) said:

“The question of obviousness must be considered on the facts of each case. The court must consider the weight to be attached to any particular factor in the light of all the relevant circumstances. These may include such matters as the motive to find a solution to the problem the patent addresses, the number and extent of the possible avenues of research, the effort involved in pursuing them and the expectation of success.”

This statement of principle was also approved by the House of Lords in Conor Medsystems Inc v Angiotech Pharmaceuticals Inc. One of the important points, to my mind, is that all these considerations interact with each other. In short, it all depends. MedImmune’s argument proceeded on the basis that Novartis needed to establish (a) a fair prospect of success (b) within a reasonable time, as if these were two independent conditions that had to be satisfied. They are not successive hurdles to be jumped; they are no more than aspects of the statutory question: was the invention obvious? We should stick to the statutory question, which has to be applied in all sorts of circumstances and in all sorts of different fields of endeavour.

An invention is, at least usually, either a product or a process. So the statutory question is: was it obvious to make the product or to carry out the process? In order to answer the statutory question it is, of course, necessary to decide what the invention is. As Lord Hoffmann pointed out in Biogen Inc v Medeva plc [1997] RPC 1, 34:

“Whenever anything inventive is done for the first time it is the result of the addition of a new idea to the existing stock of knowledge. Sometimes, it is the idea of using established techniques to do something which no one had previously thought of doing. In that case, the inventive idea will be doing the new thing. Sometimes, it is finding a way of doing something which people had wanted to do but could not think how. The inventive idea would be the way of achieving the goal. In yet other cases, many people may have a general idea of how they might achieve a goal but not know how to solve a particular problem which stands in their way. If someone devises a way of solving the problem, his inventive step will be that solution, but not the goal itself or the general method of achieving it.”

In many “obvious to try” cases, it is the idea of trying that constitutes the inventive step. It was no doubt this that led Sir Donald Nicholls V-C to say in Molnlycke AB v Procter & Gamble Ltd [1994] RPC 49 that:

“… obviousness connotes something which would at once occur to a person skilled in the art who was desirous of accomplishing the end.” (Emphasis added)

However, in our case, the patentee was not the first to have the idea of phage display of antibodies or antibody fragments and using it to screen an antibody library. Professor Smith had already had that idea, and made it public, some months earlier at the Banbury conference. It is plain from the slide that he showed at that conference that his proposal entailed an experiment with antibodies having different binding properties, which are then exposed to antigen. Nor was the patentee the first to find a way of doing it. The patentee used the method that had already been described by Parmley and Smith two years before the priority date. Nor did the patentee solve a particular problem that stood in the way, because as things turned out there was no problem. In short the patentee pursued an identified goal with known means. Where, then, was the inventive step?

In some cases invention can lie in finding out that something that was thought not to work can in fact work. In Pozzoli SpA v BDMO SA [2007] EWCA Civ 588 [2007] FSR 37 Jacob LJ put it this way:

“27 Patentability is justified because the prior idea which was thought not to work must, as a piece of prior art, be taken as it would be understood by the person skilled in the art. He will read it with the prejudice of such a person. So that which forms part of the state of the art really consists of two things in combination, the idea and the prejudice that it would not work or be impractical. A patentee who contributes something new by showing that, contrary to the mistaken prejudice, the idea will work or is practical has shown something new. He has shown that an apparent “lion in the path” is merely a paper tiger. Then his contribution is novel and non-obvious and he deserves his patent.

28 Where, however, the patentee merely patents an old idea thought not to work or to be practical and does not explain how or why, contrary to the prejudice, that it does work or is practical, things are different. Then his patent contributes nothing to human knowledge. The lion remains at least apparent (it may even be real) and the patent cannot be justified.” (Emphasis in original)

This, I think, is why the judge decided that the patent in suit was not obvious over Parmley and Smith. That paper thought that expressing large proteins on filamentous phage would probably not work. But at the Banbury conference Professor Smith’s perception had changed. In the meantime other workers in the field had published important papers advancing this area of scientific endeavour as Kitchin LJ has explained; and Professor Smith’s own thinking had developed. Although he may not have gone as far as saying that his idea of phage display of antibodies or antibody fragments and using them to screen an antibody library definitely would work, he thought it was really worth pursuing; and was selling it as an idea to try. On the basis of his previous work and scientific analysis, he was reasonably confident of success, although he recognised that success was not guaranteed. He was not predicting problems; but he said that if there were problems, there were ways to overcome them, which he identified. It cannot be said that his idea was one which “was thought not to work”; or that there was a “prejudice that it would not work or be impractical”. The experts also agreed that the skilled addressee would not have expected to take a long time to put Professor Smith’s idea into effect, to see whether it did in fact work. In fact it did. Mr Wyand QC argued that if the skilled addressee had been given Professor Smith’s idea he would have gone back to Parmley and Smith to find out how to put it into practice. On reading Parmley and Smith he would have noted the reservations expressed in that paper, which would have put him off. In the first place, whether that would have been the case is a pure question of fact, which the judge resolved against MedImmune. Second, as Mr Thorley QC submitted, if the skilled addressee went back to Parmley and Smith it would have been to find out how to implement Professor Smith’s new idea; not to second guess the idea itself.

Professor Smith had, so to speak, invented a weapon. He had fired it at a particular target. At the Banbury conference he suggested that it could be fired at a different target, which it might well hit. The patentee fired it. It hit the target as Professor Smith had suggested.

Against that background can it be said that the judge was wrong to hold that the invention claimed by the patent in suit was obvious? Biogen Inc v Medeva plc cautions an appellate court against interfering with a trial judge’s multi-factorial value judgment on whether a claimed invention is obvious. In this case there is no need for such caution. In my judgment the judge’s conclusion was amply supported by his findings of fact. I too would dismiss the appeal.

I agree that the appeal should be dismissed for the reasons given by Kitchin LJ. I also agree with and endorse the observations of Lewison LJ on the question of obviousness and in particular the proper approach to interpretation of the statutory provision.