Scinopharm Taiwan Ltd v Eli Lilly & Company

This is an action by Scinopharm Taiwan Limited (“Scinopharm”) to revoke EP (UK) No 0,577,303 (“the Patent”) which is owned by Eli Lilly & Company (“Lilly”). The Patent has an earliest priority date of 22 June 1992 and relates to a new process for making a compound called gemcitabine, which is a drug used for the treatment of cancer and sold by Lilly under the name GEMZAR. It is of considerable commercial significance with worldwide sales in 2007 of about US $ 1.36 billion.

Gemcitabine is a nucleoside analogue. Nucleosides themselves are the building blocks of DNA and RNA and are composed of a sugar component linked to a nucleobase component by a glycosidic linkage. Gemcitabine has a very similar structure to a nucleoside called deoxycytidine. Both are depicted below:

It can be seen the two molecules differ only in that gemcitabine has two fluorine atoms at the C-2 position on the sugar component instead of two hydrogen atoms. It is these fluorine atoms which cause gemcitabine to interfere with the process of DNA replication and so make it an efficacious anti-cancer therapeutic agent.

It has been known for many years how to make nucleosides by a nucleophilic reaction in which a substituent (the nucleofugal or leaving group) attached to the carbon atom at the C-1 position on the sugar (the electrophile) is displaced and a new bond is formed with the nucleobase (the nucleophile). In the case of gemcitabine a particular synthesis disclosed in the prior art reacts a benzoyl protected sugar which has a methane sulphonate (mesylate) leaving group with a silylated cytosine, and then the benzoyl protecting groups are removed. The reaction can be represented like this:

A nucleoside can have two isomers (known as α and β anomers) depending on the orientation of the glycosidic bond. Drawings can be used to illustrate the approximate 3-dimensional shape of these anomers in which, by convention, the observer looks slightly down on the more or less flat five membered sugar ring with the sugar ring oxygen atom behind the plane of the page and with the sugar ring C-2 and C-3 atoms in front of the plane of the page. All naturally occurring anomers are β anomers, and so is gemcitabine. The structure of gemcitabine and its α anomer are shown below:

It can be seen that in the case of the β anomer the nucleobase component lies above the five membered ring of the sugar component whereas in the case of the α anomer the nucleobase component lies below the ring of the sugar component. This difference in orientation is highly significant – it is only the β anomer which has the desired activity.

Gemcitabine has been sold in the UK since 1995 and was protected by a basic patent (EP 0,122,707) which was the subject of an SPC which expired on 5 March 2009.

The Patent is a process patent which describes an improved way of making 2-deoxy-2,2,-difluoro-β-nucleosides, and gemcitabine in particular. The process is used by Lilly for its commercial production of gemcitabine and is valuable because it results in an excess of the β anomer of the nucleoside analogue over the unwanted α anomer and therefore it permits the manufacture of gemcitabine (and other 2-deoxy-2,2,-difluoro β nucleosides) in higher yields.

Lilly recognises that, as from March 2009, it will be unable to prevent competitors such as Scinopharm from making and selling gemcitabine. However it does seek to defend what it describes as its high yielding - and therefore more efficient and valuable - process described in the Patent.

Scinopharm evidently does wish to make use of the process described in the Patent and seeks its revocation. Its primary ground of attack is obviousness but it also contends the Patent is insufficient.

The obviousness allegation relies upon the following eight prior publications:

The attack was founded primarily upon Chou, which explicity refers to Hertel 3. The parties agreed the two documents should therefore be read together. The case based upon Hubbard, Kawakami, Vemishetti and Howell was not that they rendered the Patent obvious of themselves but that they should be considered (individually and collectively) together with Hertel 3 and Chou and that the Patent was obvious in the light of each of these combinations.

A patent specification is addressed to the skilled person who has a practical interest in the subject matter of the invention and practical knowledge and experience of the kind of work in which the invention is intended to be used: Catnic v Hill & Smith [1982] RPC 183 at 242-243. Moreover, the skilled person reads the specification with the benefit of the common general knowledge of the art: Kirin Amgen v Hoechst Marion Roussel [2005] RPC 9 at [33]. That common general knowledge will include the specific knowledge and prejudices of those most closely involved with the actual field with which the patent is concerned: Mayne Pharma v Debiopharm SA [2006] EWHC 1123 (Pat) at [3]-[4].

In the present case the Patent is directed to those interested in the synthesis of 2-deoxyfluoronucleosides and their analogues and addresses the need to provide such compounds, including gemcitabine, in high yields. In my judgment the skilled person is therefore a process chemist with a PhD in nucleoside and glycosylation chemistry or with several years of practical experience in the subject or who has read into the subject sufficiently to educate himself to the level of someone with such experience.

Each party called one expert witness. Scinopharm’s expert was Professor Ian Fleming. He is a distinguished organic chemist and currently Emeritus Professor of Chemistry at the University of Cambridge. He has spent almost his whole career at Cambridge, the only substantial absences being a postdoctoral year at Harvard University working with the Nobel Prize-winner Professor R.B. Woodward, and sabbatical periods as a Visiting Professor at McGill University, Montreal, at the University of Wisconsin, Madison and again at Harvard University. He was awarded the Tilden lectureship at the Royal Society of Chemistry in 1981 and the Prize for Organic Synthesis in 1984. He was elected a member of the Royal Society in 1993.

Most of Professor Fleming’s research has been in the area of organic synthesis, with an emphasis on understanding the mechanisms of organic chemical reactions in order to improve synthetic procedures and to find new synthetic pathways. In the course of his career he has been consulted by several chemical companies and provided advice and made suggestions in relation to problems and issues in synthesis. Not surprisingly, he has had many occasions to study and be aware of stereochemistry and, indeed, has had a life-long interest in and appreciation of the complexities of stereochemistry, and specifically the stereochemistry of chemical reactions. However, he has not himself carried out research in the field of nucleosides.

Lilly makes no criticism of Professor Fleming himself, and rightly so. He is a leader in his field and his evidence was clear and cogent. Nevertheless, Lilly suggests that Professor Fleming’s approach to the case was tainted by hindsight in that although he properly tried to read himself into the art, he only did so having read the cited prior art. In my judgment this criticism is unwarranted. It is true that Professor Fleming carried out a search of the literature after reading the cited prior art. But this did not influence the nature of the search he conducted. In order to determine how quickly and readily published material would have been available to the skilled person aiming to enter the field at the priority date, he simply spent an hour in the library with the indexes of Chemical Abstracts and searched them under words like nucleoside, concentrating on subheadings like deoxy and preparation, and preferably both. He also looked at ribose, deoxyribose, cytidine and deoxycytidine, and the subheadings there, all of which was perfectly reasonable. However, it is right to record that nucleoside synthesis has not been a subject of Professor Fleming’s research and consequently he was not familiar with some of the practical problems that those working in the field had experienced. I recognise this is a matter to which I must have regard in considering the allegation of obviousness.

Lilly’s expert was Professor Boons. He is Franklin Professor of Chemistry at the Complex Carbohydrate Research Center and Department of Chemistry of the University of Georgia in the United States. He received his Bachelor degree in Chemistry at the University of Leiden in 1987 and thereafter worked for six months as a research scientist at the Dutch pharmaceutical company Organon on a project concerning the chemical synthesis of modified nucleosides for the development of antiviral drugs. He then returned to the University of Leiden where he obtained his PhD in 1991. Since that time his research interests have focussed on carbohydrate chemistry and biology, glycosylation chemistry and nucleoside chemistry, and his laboratory is recognised for the development of novel methods for stereoselective glycosylations and their application for the preparation of compounds of biological importance. Over the years he has acted as a consultant to a number of biotechnology and pharmaceutical companies and several of his research programmes have been funded by industry.

Professor Boons proved himself well equipped to give evidence as to the state of the art of glycoside and nucleoside chemistry at the priority date and he gave his evidence in a direct and convincing manner. Scinopharm submits that the value of his evidence was compromised by his inclination hastily to assume a contrary position to propositions put to him in cross examination. It is also said his answers tended to be long and not focussed on the questions being asked of him. I do not accept these criticisms. Certainly the answers given by Professor Boons tended to be full but were not unduly so and, in so far as he gave answers which he subsequently qualified, I am satisfied that this was simply because he did not properly understand the questions being put to him. Overall I found Professor Boons to be an excellent witness and his evidence of great assistance.

The common general knowledge 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 the art to which the invention relates: Beloit Technologies Inc v Valmet Paper Machinery Inc [1997] RPC 489 at pages 494-495. It also includes all the material which the skilled person knows exists and which he would refer to as a matter of course if he cannot remember it and which he generally understands is sufficiently reliable to use as a foundation for further work: Raychem Corporation’s Patents [1988] RPC 31 at 40. It does not, however, include everything which may be turned up on a literature search. As Floyd J said in ratiopharm v Napp [2008] EWHC 3070 at [159]:

“Finally, the common general knowledge does not include knowledge which does not inform the skilled person's approach from the outset. As Kitchin J said in Generics (UK) v Daiichi Pharmaceutical [2008] EWHC 2413 (Pat):

" I can readily accept that, faced with a disclosure which forms part of the state of the art, it may be obvious for the skilled person to seek to acquire further information before he embarks on the problem to which the patent provides a solution. But that does not make all such information part of the common general knowledge. The distinction is a fine one but it may be important. If information is part of the common general knowledge then it forms part of the stock of knowledge which will inform and guide the skilled person's approach to the problem from the outset. It may, for example, affect the steps it will be obvious for him to take, including the nature and extent of any literature search."

Whether knowledge is common and general depends on the considerations explained by Aldous LJ in Beloit. If information does not satisfy that criterion, it does not become common general knowledge by postulating a set of steps that the skilled team might take to find it if they had already embarked on an attempt to solve a particular problem. That is not to say that it is illegitimate, in assessing an obviousness attack, to take account of material which would inevitably be found and treated as reliable in consequence of a step or steps which it is obvious to take. If the material so found is such as would be accepted, then it may assist in showing obviousness of a further step. But what it cannot be used for is in support of an argument that the series of steps being undertaken were obvious from the start.”

These observations have a particular relevance in the context of the present case in the light of Scinopharm’s contention that it was obvious to read various items of prior art together.

Much of the general background necessary for an understanding of the Patent was not in dispute and I have taken the non contentious aspects of the exposition which follows largely from the reports of Professor Fleming and Professor Boons, for which I am most grateful.

The essential principles of chirality were accepted to be common general knowledge. Isomers are different compounds which have the same molecular formula. If isomers differ from each other only in the way the atoms are oriented in space then they are called stereoisomers. A molecule is chiral if its mirror image is not super-imposable on itself. In the case of a tetrahedral carbon, a molecule is chiral when it has four different groups. A chiral molecule and its mirror image are called enantiomers.

Enantiomers usually have identical chemical and physical properties except for the rotation of polarised light, which is by equal amounts but in opposite directions. A mixture of equal quantities of enantiomers is termed a racemic mixture and overall will not rotate polarised light. In general, each member of a pair of enantiomers has different biological properties. However, the extent of any differences between enantiomers will vary from one pair of enantiomers to another.

If a molecule has more than one chiral center it has, in general, 2ⁿ stereoisomers (where n = number of chiral centers), which cannot all be enantiomers. In this respect, stereoisomers that are not mirror images are termed diastereoisomers. Diastereoisomers often have different physical, chemical and biological properties.

The basic mechanisms of nucleophilic substitution reactions were also agreed to be common general knowledge and have for many years been taught to undergraduates. They are a fundamental class of organic transformations in which a nucleophile (represented below as ‘Nuc’) attacks a positive or partially positive charge of an atom attached to a group or atom called the nucleofugal or leaving group (represented below as ‘LG’). The positive or partially positive atom is referred to as an electrophile (represented below as ‘EL’). A nucleophile is a molecule or ion that can provide a pair of electrons (represented below as ‘:’) to be shared with another atom in the formation of a new covalent bond.

There are two main mechanisms by which such reactions can occur, which are called S_N1 and S_N2 reactions.

The S_N1 reaction involves two separate chemical steps. The first is a heterocyclic cleavage of the leaving group of the electrophile to give a trigonal positively charged carbon (also referred to as a carbocation) and the leaving group, which can be depicted thus:

This reaction is then followed by the combination of the carbocation with the nucleophile to give a product:

As the figure explains, the carbocation is planar and hence the nucleophile can attack it from either side to give one or other of two products. In the simple reaction depicted, attack on each surface is equally probable and the product mixture is racemic. If, on the other hand, the faces are not identical then a mixture which is not 1:1 is likely to result, a matter of some relevance in the context of the issues which arise in this case.

Generally, the rate of the reaction is determined by the first step. Since this involves only the substrate and not the nucleophile, neither the nature nor the concentration of the nucleophile makes any difference to the rate of the reaction.

In an S_N2 reaction, the nucleofugal group does not leave until it is pushed out by the attack of the nucleophile. This reaction always takes place in a single step and the nucleophile always attacks on the opposite side of the carbon atom from the nucleofugal group. Hence any stereochemistry of the starting product is preserved, save that it is inverted:

A number of factors which tend to influence S_N1 and S_N2 reactions were not in dispute. As mentioned, in a pure S_N1 reaction, the rate of the reaction does not depend upon the concentration of the nucleophile. But it does depend on the concentration of the electrophile. By contrast, in a pure S_N2 reaction, the rate of the reaction will depend upon the concentration of both the electrophile and the nucleophile. Some reactions are said to be borderline in that they exhibit characteristics of both S_N1 and S_N2 reactions. In such cases, increasing the concentration of both the nucleophile and electrophile will increase the overall rate of both the S_N1 and S_N2 reactions, but will increase the rate of the S_N2 reaction more. At an extreme, these reactions can sometimes be conducted under fusion conditions, that is to say in the absence of any solvent at all.

A number of other factors were known to have an influence on the propensity of a reaction to take place by the S_N1 or S_N2 pathways. One is charge stabilisation. In particular, factors which tend to lead to a stabilisation of the carbocation will favour an S_N1 reaction. Another is the structure of the nucleophile. A highly reactive nucleophile will speed up an S_N2 reaction because the nucleophile is involved in the rate determining step of the reaction. The nature of the leaving group is also of considerable importance. The S_N1 pathway is particularly dependent on leaving group ability because it requires cleavage of the bond between the carbon atom and the leaving group without the assistance of the nucleophile. As Professor Boons put it, increasing the leaving group ability will also increase the rate of an S_N2 reaction; just not as much as for the S_N1 case. The experts were also agreed that the nature of the solvent can also have a dramatic effect on the speed of an S_N1 reaction with good ionising solvents (such as water, methanol and acetonitrile) increasing the rate of the reaction by stabilising the carbocation and poorly ionising solvents (such as dichloromethane and chloroform) slowing the reaction down. S_N2 reactions tend to be less affected by changes in solvent, save that highly polar aprotic solvents (such as DMSO and DMF) can speed them up.

Sugars having five carbon atoms are called pentoses. This action is concerned with the common form of pentose, namely furanose. The carbon atoms are numbered 1-5, as shown in the drawing below:

Furanoses sometimes lack one or more of the OH groups. The one with the C-2 OH group missing and the C-3 OH group on the opposite side from the C-5 side chain is a particular furanose called D-2-deoxyribose and was drawn by Professor Fleming in his first report as structure 7:

I should also explain that D-2-deoxyribose has two diastereomers. The carbon atom at C-1 is called the anomeric carbon atom and these particular diastereomers are therefore called anomers. In the α form the OH attached to C-1 lies behind the page and in the β form it lies in front of the page, as shown in Professor Fleming’s structures 8 and 9:

Glycosylation is the general term for a reaction in which a fragment is attached to C-1 of a sugar. Hence a glycosylation reaction is used to produce nucleosides which consist of a sugar component attached to a heterocyclic aromatic base.

For many years nucleoside synthesis has been performed by a particular glycosylation reaction known as the Hilbert-Johnson reaction in which a protected 2-deoxyribosyl halide as the electrophile reacts with a nucleoside base as the nucleophile. In this reaction the halide leaves with the electrons to form a negatively charged ion and is replaced by the nucleobase, and the problem of regio-selectivity is solved by the use of protecting groups. The general reaction is illustrated below, with a chlorine atom as the halide and the protecting groups shown as X, Z and W:

Over the years various improvements to this basic reaction have been developed. In the 1960s it was found that silylation of the nucleobase (Z=W= Me₃Si) made it more nucleophilic and also facilitated its purification. It was also found that the reaction could be promoted by the use of metal salts which increased the ability of the halogen to leave the electrophile. Sometime later Professor Helmut Vorbrüggen found that anomeric acetates could be condensed with silylated nucleobases to give the desired nucleosides and this reaction could be activated or promoted by a variety of reagents such as trimethylsilyl triflate.

It will be apparent from the foregoing, and as Professor Fleming explained, there are two possible starting materials and two possible products differing in their stereochemistry. If the starting material is either the b or the a chloride, and the pathway is S_N1, by way of a cationic intermediate (27), there will be two products, the a and the b nucleosides, typically in a 1:1 ratio, regardless of which chloride anomer was used as the substrate, as shown in Professor Fleming’s Scheme 2:

By contrast, if the reaction pathway is pure S_N2, then it will produce complete inversion. If the starting material is a 1:1 mixture of the b and a chlorides, then the product will be a 1:1 mixture of the a and the b nucleosides. If, on the other hand, the starting material is the b chloride then the product will be the a nucleoside.

The foregoing description was accepted by the parties to be common general knowledge. However, they were not in agreement as to the extent to which the skilled person would be familiar with S_N2 glycosylation reactions, with Eli Lilly contending and Scinopharm disputing that they were rare.

Professor Boons was clear and consistent throughout his cross examination that in carbohydrate chemistry S_N2 reactions with full inversion of configuration were extremely rare. As he explained, the skilled person would be thinking about S_N1 reactions, oxocarbenium ions and how to influence the anomeric ratio of the a and the b nucleosides by other means.

Professor Fleming said in his first report that it has been known since the 1930s that those substrates that have substituents that give rise to a lower energy cation encourage the reaction to take the S_N1 pathway; and that riboses with a nucleofuge attached to C-1 inherently have a propensity to take the S_N1 pathway because the oxygen atom in the ring is just such a substituent. To persuade a substrate to follow the stereochemically more useful S_N2 pathway requires that other factors shift the conditions towards those favouring an S_N2 reaction. In cross examination he accepted that the Hilbert-Johnson reaction normally has the characteristics of an S_N1 reaction and depends on the capture of a carbocation at C-1 of the sugar moiety by the most electronegative nitrogen on the base.

In the end, therefore, I do not believe there was much between the experts. I am satisfied that the skilled person would have considered it likely (at the least) that glycosylation reactions to produce nucleosides would be proceeding by the S_N1 pathway.

Despite the inherent nature of S_N1 reactions, various techniques are and were available to try and secure some stereochemical control over their outcome. One of the most promising is called neighbouring group participation. This involves the use of a group such as an ester at the C-2 position to form an intermediate compound which only allows the nucleophile to attack from one face of the molecule. A Vorbrüggen reaction using this technique, an acetate leaving group and trimethylsilyl triflate was drawn by Professor Boons as figure 2 of his first report:

It was also known that in specific cases a measure of stereochemical control could be achieved in S_N1 reactions by judicious use of reaction conditions and protecting groups.

Nevertheless, it was something of a hit and miss affair and I am quite satisfied that it was recognised that stereochemical control of glycosidic reactions to produce nucleosides was difficult to achieve. Thus in Basic Principles in Nucleic Acid Chemistry, edited by Paul Ts’o, Academic Press, 1974, Leon Goodman wrote, in what was accepted by Professor Fleming to be a fair statement:

“A great deal of data exists on the anomeric nature of the products formed by the condensation of a sugar derivative and a heterocyclic base. The stereo-specificity of the coupling reaction depends on the method of condensation, and, within a given method, can vary on the choice of solvent, temperature and catalysts, and be very dependent on the structures of the reactants. In spite of the huge amount of effort devoted to the synthesis of nucleosides, there have been essentially no studies specifically directed toward elucidating the detailed mechanism of any of the condensation methods. The stereochemistry of the product provides the best basis for speculation about these mechanisms. ”

In relation to reactions of pentose sugar moieties without neighbouring group participation, Goodman observed:

“The use of acetylated glycosyl halides of 2-deoxy-D-erythro-pentose, which lack a participating group at C-2 give mixtures, as predicted, of α and β anomers in reactions with heavy metal salts of both purines and pyrimidines”

In other words, in the absence of neighbouring group participation one would predict the production of a mixture of the a and the b nucleosides.

Similarly, in the textbook Nucleic Acids in Chemistry and Biology by Michael Blackburn and Michael Gait, IRL Press, 1990, the authors explained that the silyl Hilbert-Johnson method worked very well for a large number of nucleoside analogues with modified bases but that it suffered from a lack of precise control of regio- and stereoselectivity because it normally had the characteristics of an S_N1 reaction and depended upon the capture of a carbocation at C-1 of the sugar component by the most electronegative nitrogen on the base. They also emphasised that a measure of anomeric control can be achieved by the use of neighbouring group participation but that this is not possible in the case of 2-deoxy and 2-deoxy-2-fluoro- sugars.

The same picture emerges from a number of publications sometime after the priority date. Professor Boons wrote in his textbook Organic Synthesis with Carbohydrates, Sheffield Academic Press, 2000, that the stereochemical formation of a glycoside linkage was one of the most challenging aspects of oligosaccharide synthesis, that 2-deoxy derivatives proceed through an intermediate oxocarbenium ion and that the stereoselective synthesis of 2-deoxy-β-glycosides was still a major challenge. As to this latter observation, Professor Fleming accepted that it was indeed “a challenge” but suggested it was one which had been met many times.

Professor Vorbrüggen himself wrote in these terms in his textbook Handbook of Nucleoside Synthesis, John Wiley, 2001:

“Although the syntheses of 2’-deoxynucleosides from suitable derivatives of 2-deoxysugars and salts of acidic heterocyclic bases or silylated heterocyclic bases have already been discussed, the different methods for their preparation are summarized here. These syntheses are among the most difficult in nucleoside chemistry...”

Professor Fleming accepted this statement as “probably” fair as of 1992. Professor Vorbrüggen continued by identifying three main problems, the third of which was that:

“the ratios of the desired natural β anomer to the undesired α anomer are often difficult to control and to reproduce”.

As to this, Professor Fleming accepted that it was a statement of fact which Professor Vorbrüggen knew far more about than he did and so it was probably correct.

Benjamin Davis and Anthony Fairbanks of Oxford University made a number of observations to similar effect in Carbohydrate Chemistry, OUP, 2002. In respect of the mechanism of nucleophilic substitution at the anomeric centre, they observed:

“In fact control of the stereochemical outcome of nucleophilic substitution at this position is one of the most difficult tasks faced by a synthetic organic chemist”.

Professor Fleming accepted this as fair, as he did the following further statements by the same authors in respect of glycosylation reactions:

“Although conceptually extremely simple, this operation has in fact been frustrating chemists for nearly a hundred years ….”.

And:

“Although a completely general approach to total stereocontrol of the formation of any glycosidic bond remains as yet unrealised, we shall see that in many specific cases high levels of stereocontrol may be achieved by judicious choice of reaction conditions and protecting groups”.

In the light of all these publications and the evidence I am satisfied the skilled person would have considered the task of securing stereochemical control in glycosylation reactions involving 2-deoxy sugars as being both challenging and difficult. But that is not to say the skilled person had nothing at all in his armoury. Both experts were agreed that, at least in the case of five membered sugar moieties, the skilled person could attempt to secure some stereochemical control over the S_N1 reaction by adjusting the reaction conditions and the protecting groups so as to favour attack of one face of the oxocarbenium ion intermediate over the other. As Professor Boons elaborated, the five membered sugar moiety is very flexible and this allows the protecting groups to have a steric effect. In addition, the protecting groups may themselves have a dipole. Both factors may result in one particular side of the oxocarbenium ion being more susceptible to attack by the nucleophile.

The Patent is entitled “Stereoselective glycosylation process” and, as the description explains at the outset, relates to a stereoselective glycosylation process for the preparation of 2’-deoxyfluoronucleosides. It explains at lines 5-14 first, why these compounds are of interest and the need to provide them in high yield:

“The continued interest in the synthesis of 2'-deoxyfluoronucleosides and their analogues is predicated on their successful use as therapeutic agents for treating viral and cancerous diseases. A compound of particular interest is gemcitabine; see European Patent Specification No. 211354 and U.S. Patent No. 4,526,988. Since these compounds are beta nucleosides, there is a need to provide such compounds in high yield. ”

And second, the problem with existing processes:

“A critical step in the synthesis of 2'-deoxyfluoronucleosides is the condensation or glycosylation of the nucleobase and carbohydrate to form a N-glycoside bond. However, processes for synthesis of 2'-deoxynucleosides are typically non-stereoselective forming mixtures of alpha and beta nucleosides. For instance, U.S. Patent 4,526,988 [Hertel 1] did not stereoselectively produce 2-deoxy-2,2-difluoro- beta -nucleosides but instead produced a 4:1 alpha to beta anomer ratio of 2-deoxy-2,2-difluoronucleoside. Even optimizing the protecting groups could not increase the alpha to beta ratio beyond 1:1; see U.S. Patent No. 4,965,374 [a patent in the name of Chou, but which is not cited as prior art in this case] which utilized benzoyl blocking groups on the carbohydrate. ”

The Patent then sets out the consistory clause of claim 1. It has helpfully been broken down by Scinopharm into the following key integers:

deblocking to form the compound of formula (I).

Scinopharm fairly highlight a number of features of the claim. The first is that the claimed process is for preparing a β anomer enriched nucleoside. Enrichment is defined by the Patent at page 5, line 44 in this way:

“The phrase “anomer enriched” alone or in combination refers to an anomeric mixture wherein the ratio of a specified anomer is greater than 1:1 and includes substantially pure anomer.”

This is a broad definition, indeed as broad as it could possibly be. Anything in which the ratio of β anomer to α anomer is greater than 1:1 satisfies the claim.

The second is that there must be an S_N2 displacement. But the claim does not require that the reaction takes place exclusively by the S_N2 pathway. It also embraces what I have described as borderline reactions which proceed partly by the S_N2 pathway and partly by the S_N1 pathway, provided that they result in β anomer enrichment.

The third is that there is nothing in the claim which limits it to reactions which produce a high yield.

This brings me to the description of the discovery which underpins the invention. On page 6 lines 47-53, it is stated:

“A key feature of the present invention is the discovery that novel α anomer enriched carbohydrate intermediate of formula (III) … can be reacted under nucleophilic displacement conditions which favor inversion (i.e. S_N2) to provide the β anomer enriched nucleosides of formula (I).

To obtain an efficient reaction between the nucleobase and the α anomer enriched carbohydrate of formula (II), an appropriate leaving group (Y) must be stereoselectively attached to the lactol to activate the lactol and generate the α anomer enriched carbohydrate of formula (II). However, the leaving group selected depends on the nucleobase chosen and the glycosylation conditions selected.”

In short, Lilly says it has discovered the reaction can be persuaded to proceed by the S_N2 pathway.

The specification then explains from page 6, line 54 how the α anomer enriched carbohydrate of formula (II) can be prepared. This is of some importance. It is to be remembered that the S_N2 pathway preserves but inverts the chirality of the starting material. Hence performing an S_N2 reaction on a 1:1 mixture of the α and β anomers of the starting material will produce a 1:1 mixture of the α and β anomers of the nucleoside and consequently be no better than an S_N1 reaction. Specifically the specification refers to two methods described in different patents, namely EP-A-0 576 229 and EP-A-0 577 302. These are not prior art, but rather share the same priority date as the Patent, a matter which has some bearing on the allegation of obviousness, as I shall explain.

The specification then elaborates how the reaction of the invention is to be performed. At page 8, line 35 it says that glycosylation reactions typically require protecting the hydroxyl groups of the lactol to prevent them from reacting with the nucleobase. It continues that suitable protecting groups are known in the art, with the most preferred being benzoyl.

At page 8, line 54 it is said that at least an equimolar amount of nucleobase must be employed relative to the amount of carbohydrate. However, it is preferable to use an excess of nucleobase, and most preferably 15 to 20 molar equivalents.

The specification then turns to the nucleobases and from page 8, line 57 to page 9, line 28 describes how they too must have protecting groups and those which are preferred.

Solvents and suitable conditions of temperature are described from page 9, line 35 to page 11, line 34. The solvent must be inert to the glycosylation reaction and a high boiling inert solvent and a solution having a carbohydrate concentration of at least 20% are preferred. One of the more preferred solvents is anisole. Particular leaving groups for use with high boiling point solvents are described on page 9 at lines 55-58. On page 10, from line 45 it is said that if the carbohydrate contains a fluoro sulfonyloxy group it is unstable above room temperature and that the reaction of the invention must therefore be carried out at lower temperatures, and suitable solvents and nucleobases for use in such conditions are described.

From page 11, line 35 to page 12, line 57 the specification explains that the nucleobase may be converted to a metal cation salt to increase its nucleophilic reactivity.

At page 13, line 9 it is said that the reaction may be run in the absence of a solvent, that is to say under fusion conditions. In this case the temperature must be sufficient to convert the α anomer carbohydrate to a molten phase. Suitable leaving groups and nucleobases are then described.

Finally by way of general teaching, the specification discloses the use of suitable catalysts and, from page 15, line 2, the removal of the protecting groups.

The specification then proceeds to describe 58 examples which were summarised and grouped by Professor Fleming as follows:

“80.

These are the main groups of recipes in the Examples in the Patent:

(i)

The α-enriched mesylate with the silylated cytosine typically (Example 5 on p. 16) at 115-120 °C for 7 hours in anisole giving a β: α ratio of 7.3:1 and a yield of 79.5%.

(ii)

The same reagents but neat (without solvent) typically (Example 48 on p. 31) at 130 °C for 1 hour giving a β:α ratio of 3.9:1 and a yield of 43%.

(iii)

The same reagents but in acetonitrile with added salts typically (Example 26 on p. 24) potassium trifluoromethansulfonate at 75 °C for 45 hours giving a β:α ratio of 7.2:1 and a yield of 62.4%. A direct comparison of a salt-catalysed and an uncatalysed reaction is reported in Examples 27 (6.7:1, 59.3%) and 28 (3.4:1, 77%), respectively.

(iv)

The α-enriched triflate, carefully made and kept at or below –65°C, with the silylated cytosine in dichloromethane typically (Example 17 on p. 22) at 23 °C for an unspecified time giving a β: α ratio of 2.5:1 and a yield of 68%.

(v)

The α-enriched p-bromobenzenesulfonate with the potassium salt of N-pivaloylcytosine (Example 49 on p. 32) at 65 °C for 5.5 hours in acetonitrile giving a β: α ratio of 3.9:1 and a yield of 20%.

81.

In addition, other Examples record the use of the potassium salts of variously protected purines (Examples 30-42, 54-55), silylated uracil (Example 44), the sodium salt of a purine (Example 53) and of benzylcytosine (Example 56), triazole salts (Examples 51, 52 and 57) and chloropurines (Example 58), and as electrophiles the α-iodide (Examples 50 and 56) and bromide (Example 51).

82.

Most of the Examples, i.e. 1-6, 8-11, 13-29, 43, 45 and 48 give blocked gemcitabine straightforwardly, and 12, 46, 47, 49, 50 and 56 give differently blocked versions which could easily be converted to gemcitabine.”

His overall conclusion that a careful reading reveals some trends in the work and supports the contention that a large excess of nucleobase, a high concentration of reagents, and a non-polar solvent all encourage the S_N2 reaction, as any chemist would expect was, however, hotly contested on the facts of this case, as I shall explain in addressing the cited prior art.

Professor Boons also drew some helpful general conclusions in paragraph 159 of his first report:

“The numerous examples show that β-selectivity can be accomplished by employing the following reaction conditions:

(a)

an excess of nucleobase favors the production of the β-anomer (see example 7)

(b)

using mesylate or tosylate as a leaving group requires the use of very apolar solvents (such as xylenes) and requires the reaction to be carried out at high temperatures (105-135^oC)

(c)

using triflate as a leaving group requires the use of a solvent such a dichloromethane and a much lower reaction temperature (many of the examples are between -60 and 20 ^oC )

(d)

the addition of non-nucleophilic salts appears to improve the β- selectivity

(e)

fusion conditions can be employed to prepare β-enriched nucleosides.”

One important point arises from this disclosure which it is convenient to mention at this stage but which is reinforced by the cited prior art. The teaching of the Patent is not that any reaction can be made to proceed by the S_N2 pathway by adjusting any single one of the described parameters. Particular combinations of, for example, nucleophile concentration and solvent will not necessarily induce an S_N2 reaction. This much appears to be common ground. As Scinopharm itself contends in its closing submissions: “It is not sufficient for the claim to require at least a molar equivalent of nucleobase. It has not been shown that as a general matter this particular ratio of itself will result in β enrichment.” The Patent does, however, disclose that the claimed reaction can be made to proceed by the S_N2 pathway by appropriate adjustment of all the described parameters and it provides numerous examples and general teaching as to how this can be achieved.

The general legal principles are well established and I did not understand them to be in dispute. It is convenient to address the question of obviousness by using the structured approach explained by the Court of Appeal in Pozzoli v BDMO [2007] EWCA Civ 588; [2007] FSR 37. This involves the following steps:

The last question requires the court to take into account all the relevant circumstances. It is also important to have in mind the nature of the obviousness case which is being advanced. As I shall elaborate, Scinopharm says it was obvious to investigate the prior art and, moreover, it was obvious to investigate the conditions that might induce an S_N2 reaction. It is therefore, I think, another ‘obvious to try’ case, and one of the kind considered by the House of Lords in Conor v Angiotech [2008] UKHL 49. There Lord Hoffmann said at [42]:

“In the Court of Appeal, Jacob LJ dealt comprehensively with the question of when an invention could be considered obvious on the ground that it was obvious to try. He correctly summarised the authorities, starting with the judgment of Diplock LJ in Johns-Manville Corporation's Patent [1967] RPC 479, by saying that the notion of something being obvious to try was useful only in a case in which there was a fair expectation of success. How much of an expectation would be needed depended upon the particular facts of the case. As Kitchin J said in Generics (UK) Ltd v H Lundbeck A/S [2007] RPC 32 , para 72:

“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.””

I think it is also right to note that although motivation may be one of the factors which it is proper to consider, its relevance will depend upon the facts of the case. It is certainly not an essential requirement of a conclusion of obviousness, as the Court of Appeal made clear in Asahi Medical v Macopharma [2002] EWCA Civ 466 at [23] – [24]:

“23.

Mr Thorley also submitted that the judge had wrongly rejected his submissions that were recorded by the judge as paragraph 46 of his judgment. He submitted that an invention would not be obvious unless there was some motivation to implement the disclosure in the prior art and to take the steps required to arrive at the invention. In certain cases that can be right. Such cases are usually those where the invention lies in the idea of taking a step. However, motivation may not be a requirement. The fact that nobody would dream of making a plate one inch bigger than the standard size does not mean that there would be invention in making one. In Pharmacia Corporation v Merck & Co Inc [2001] EWCA Civ 1610, I cited this passage from the speech of Laddie J in Hoechst v Celanese Corp v BP Chemicals Ltd [1997] FSR 547at 573:

“Before a step from the prior art can be held to be obvious there must be some reason why the man skilled in the art would wish to take it. If he has a problem and the step would occur to him as a solution to it, then he has a reason. But there is no requirement that it be demonstrated that the step would have been expected to produce significant commercial advantages. The problem might be very small. The courts will assume that he may just want an alternative way of achieving essentially the same result as in the prior art. Thus were workshop modifications, none of which would be expected to produce significant technical or commercial benefits are still obvious. To adopt an example sometimes given by Jacob J., if it is known to make a 5-inch plate, it is obvious to make a 5¼ -inch plate. Technicians and businessmen frequently want to make trivial variations in established or known products. Similarly if the prior art discloses two wooden parts held together by screws it would be obvious to glue them, even if so doing would not be expected to advance the industry. The notional addressee is likely to want to use materials readily at hand to make essentially the same thing as is disclosed in the prior art. That is sufficient motivation and the use of those materials is, accordingly, obvious. When the defendants argue that Hingorani or any of his readers is entitled to use any “natural extension” or “obvious variant” of his concept, they are correct if by that they mean the type of workshop modification or alternative discussed above. But it was not and could not be suggested by any witness that changing the medium from aqueous to organic and changing the resin was a mere workshop variant of what is set out in Hingorani.”

24.

I continued:

“124.

That statement of the law was, I expect, apt on the facts of that case, but should not be followed generally. A step from the prior art, albeit made without reason, can still be obvious. The judge categorises such a step as workshop modifications and, in so doing, introduces a test not in the statute, namely whether the step from the prior art was a workshop modification. The statutory test is obviousness and any modification which is obvious will not be patentable, whereas one which is not obvious will be. The true test, as made clear in Windsurfing, is to ask whether the invention was obvious. Whether or not there is a reason for taking the step from the prior art may well be an important consideration, but that does not mean that it is an essential requirement of a conclusion of obviousness.”

25.

The judge did not in paragraph 46 of his judgment fall into the error of principle that Mr Thorley submitted that he had. What he said in that paragraph has to be read in the light of the conclusion he had reached in paragraph 45. The judge concluded that the step from the prior art was a “workshop variation” and therefore was an obvious step. Mr Thorley had submitted that it was not obvious because the skilled person would not in practice have thought of implementing it at all. If the step from the prior art lacked invention, then it mattered not whether anybody would have thought of implementing it. The public are entitled to make obvious modifications. Whether they would want to do so will depend upon a variety of factors which could include such things as cost and the attitudes of users.”

There is one other matter it is convenient to mention at this stage. Scinopharm’s case depends, in part, upon reading various items of prior art together. It contends it is permissible to do this if they are in the same technical field. I do not agree. In my judgment it is only permissible to read two documents together if it is obvious to do so, as the Court of Appeal made clear in Smithkline Beecham v Apotex Europe [2005] FSR 23 at [96]:

“96.

I think the Judge erred in principle here. The skilled man has his common general knowledge — the mental tools of his trade — but no more. The law of obviousness supposes that he can be given any individual piece of prior art and read it with that knowledge. The piece of prior art forms part of the “state of the art”. What he cannot do is to just link one piece of prior art with another, unless so to do would itself be uninventive. No-one disputes what Lord Reid said in Technograph v Mills & Rockley [1972] RPC 346 at page 355:

“In dealing with obviousness, unlike novelty, it is permissible to make a ‘mosaic’ out of the relevant documents, but it must be a mosaic which can be put together by an unimaginative man with no inventive capacity.””

The question whether it is obvious to read two documents together is one to be considered in the light of the particular circumstances of each case. Relevant factors may include whether one document refers to the other or whether one or both documents would be found on a literature search of the kind the skilled person would routinely carry out before attempting to find a solution to the problem the patent addresses.

In these papers Hertel and his colleagues at Lilly described the results of their program of research to synthesise a new family of nucleosides having two fluorine atoms attached to C-2 of the sugar component and which they hoped would have value as antiviral or anticancer agents.

Specifically Hertel discloses the synthesis of α and β nucleosides by the Vorbrüggen procedure drawn by Professor Fleming as scheme 5 and reproduced below (roman numerals correspond to the numbering in Hertel 2 and bold numerals to the numbering in Hertel 3):

The reaction has a number of steps. In the first, the β-hydroxy ester (IV) is synthesised by combining a sugar-derived aldehyde (V) with a fluorine containing α-bromo ester. This is then converted into the lactone (III). The OH groups of the lactone are protected with protecting groups, typically as tert-butyldimethylsilyl (TBS) ethers 4, and the lactone is reduced to the lactol 5, which will be a mixture of the α and β anomers. The lactol is converted to a mixture of the α and β mesylates 7, equipping the ribose with a sulfonyloxy nucleofugal group. The mesylates are then combined with equimolar amounts of silylated pyrimidines to give, after hydrolysis, mixtures of the α and β nucleosides 8.

Hertel 3 focusses on gemcitabine and explains that condensation in dichloroethane in the presence of the Lewis acid trimethylsilyl triflate gave a 40% yield of the α anomer and a 10% yield of the β anomer after HPLC separation. An alternative method involving fusion gave a yield of 20% α and 5% β after separation of the anomers by HPLC. Thus in both syntheses, the ratio of the isolated products was about 4:1, but the yield of the fusion reaction was much reduced.

The authors continue that the presence of the electron withdrawing difluoro substituents at C-2 should make dissociation of the mesylate ion from the sugar more difficult and consequently condensation of the sugar mesylate with silylated bases should proceed in large measure by the S_N2 mechanism. However, investigation of the NMR spectra of the disilyl sugar (5) and the disilyl mesylated sugar (7) revealed the anomeric ratio at C-1 to be approximately 1:1. These data led them to conclude that the predominant formation of the α nucleoside must therefore involve facial differences of the sugar toward the nucleophile in an S_N1 type mechanism.

Of further note in the context of the issues arising in this case is that the adoption of fusion conditions did not alter the α:β anomeric ratio of the nucleoside reaction product but simply resulted in a reduction in yield.

Chou was published in 1992, some four years after Hertel 3 and reveals the results of further work by another team at Lilly to improve the yield of gemcitabine over that achieved by Hertel and his colleagues. At the outset of the paper the authors refer back to the Hertel synthesis and explain it is not suitable for kilogram scale production. The authors continue that they have used the same synthetic scheme but selected benzoyl over TBS as the protecting group for the hydroxyl groups. With this modification, a crucial selective crystallisation is now possible, that is to say a crystallisation of the desired ribonolactone (2a) from its diastereomeric mixture consisting of 2a and 2b. Also the crystallisation of gemcitabine from a 1:1 anomeric mixture was accomplished, this being a vast improvement over the 4:1 mixture when TBS was used as the protecting group.

The first part of the reaction scheme was drawn by Professor Fleming as scheme 3 and is reproduced below:

The paper describes each stage of the synthesis from 3-8a/b, explains that the mixture of isomers 8a/b is benzoylated and then fractionally crystallised to give the anomerically pure lactone 2a. This lactone is then reduced to a mixture of the anomers 9a/b. The mixture of mesylates 10a/b is then prepared from the mixture of lactols 9a/b. These could be separated by fractional crystallisation and characterised by their melting points, their combustion analyses and their rotational and NMR data.

The authors continue with a description of the use of the Vorbrüggen procedure for nucleoside synthesis in two worked examples. In one, carried out on a kilogram scale, the mesylate mixture was reacted with an excess of silylated cytosine in the presence of trimethylsilyl triflate in dichloroethane overnight at 83ºC. After deblocking, the reaction produced a yield of 49.2% of which 47.3% was the desired gemcitabine β anomer and 52.7% was the wrong α anomer. In the other, conducted on a smaller scale, the mesylate mixture was again reacted with an excess of silylated cytosine in the presence of trimethylsilyl triflate in trichloroethane for 18 hours at 113ºC. In this case, after partial deblocking, it was found the reaction produced a yield of 86.9% of which 43.8% was the desired gemcitabine β anomer and 56.2% was the wrong α anomer.

The authors also carried out experimental work to determine the reaction mechanism. For this purpose they used each of the separated mesylates 10a and 10b with silylated cytosine in non polar solvents and at high temperature. They found that a 1:1 mixture of nucleoside anomers was produced whatever the stereochemistry of the mesylate used as the starting material and deduced that the reaction was taking place by the S_N1 pathway via an oxygen stabilised cationic intermediate. This part of the reaction was drawn by Professor Fleming as his reaction scheme 4:

The stabilised cationic intermediate or carbocation 15 is attacked by the nucleophile 12 on either face so producing a 1:1 mixture of the nucleoside anomers 14a/b irrespective of whether the reaction is carried out using the α mesylate 10a or the β mesylate 10b. Indeed the authors themselves state:

“We believe the reaction proceeds via an S_N1 pathway involving an oxonium ion intermediate 15 giving rise to a 1:1 mixture of nucleoside anomers.”

The authors then describe further experimentation to confirm that both the mesylates 10a/b and the product nucleosides 14a/b are configurationally stable under the reaction conditions, showing that rapid equilibration of the anomers, before and after the coupling, is not responsible for the loss of stereocontrol. In light of this further work they say:

“Therefore, we concluded that the observed 1:1 β/α ratio of nucleosides was not the result of epimerization of either the mesylate starting material or the nucleoside product, but rather a consequence of the predominant reaction mechanism.”

Professor Fleming concluded in his first report that the paper had at this stage taught the skilled person two important matters:

Professor Boons arrived at the same conclusion. In his first report he said that the skilled person would read Chou together with Hertel 3 and would have understood that anomeric mesylates of 2-deoxy-2’-2’difluoro sugar have a tendency to glycosylate by an S_N1 mechanism and, moreover, the skilled person would have understood there would be no gain by performing the reaction by an S_N2 substitution because the mesylates were obtained as anomeric mixtures. In his second report he further elaborated his view that the skilled person would take from Chou that the reaction follows the limited S_N1 pathway and that this is not a borderline case because, regardless of the anomeric configuration of the starting material, a 1:1 mixture of nucleoside anomers is produced.

It is convenient to consider these publications together because that is the way the case of obviousness was advanced and because it is accepted that the skilled person reading Chou would inevitably refer to Hertel 3. No separate case was advanced based upon Hertel 1 or Hertel 2.

There is to my mind only one difference of substance between claim 1 of the patent and the disclosure in Chou and Hertel 3, namely that this prior art does not disclose a process for preparing a relevant β anomer enriched nucleoside by conducting an S_N2 displacement reaction. It is to be noted that Chou does describe a glycosylation reaction using an α anomer enriched carbohydrate of a relevant formula because the authors carried out their investigation into the reaction mechanism using each of the separated mesylates 10a and 10b. However, these reacted by the S_N1 pathway rather than the S_N2 pathway and so produced a 1:1 mixture of nucleoside anomers, and, more particularly, the α mesylate did not produce a β anomer enriched nucleoside.

The argument advanced by Scinopharm is attractive and straightforward. It points out that the protecting groups and the leaving groups described in Chou are the same as those falling in the claims of the Patent and that Chou teaches the use of non polar solvents. It then contends, supported by the evidence of Professor Fleming in his first report, as follows.

β enrichment by an S_N2 pathway is obviously attractive to a person skilled in the art. Whether the α mesylate starting material is formed by fractional crystallisation (as in Chou) or in some other way, there are benefits in terms of efficiency to work with anomerically pure material in the next stage of the process. It means that less solvent can be used and β enrichment of the nucleoside would make its isolation easier.

In any event, an industrial chemist contemplating the production of gemcitabine would be expected to investigate the conditions that might produce an S_N2 reaction.

A variety of standard factors and conditions could and would be investigated for this purpose. One of the obvious things to do would be to increase the concentration of the nucleophile and try fusion conditions. Another would be to use salts of the nucleobase which are more nucleophilic. A third would be to use a large excess of silylated nucleophile, in the same volume of solvent. Of all of these, Scinopharm focusses particularly on an increase in the concentration of nucleophile.

In summary, the skilled person, interested in the production of gemcitabine and having read Chou, would consider that the reaction conditions could be adjusted to increase the amount of the reaction that proceeds by the S_N2 pathway. As part of his examination of this issue it was obvious, indeed likely, that he would increase the relative concentration of the nucleophile and thus perform the process of claim 1 of the Patent.

In assessing this attack I think it is important to be clear about the starting point. I have no doubt that the skilled person would have read Chou and Hertel 3 with interest with the aim of finding a better way to make gemcitabine. But I do not accept he would have approached the documents with a predisposition to look for ways of persuading the reaction to proceed by the S_N2 pathway; indeed quite the contrary, as I shall explain.

At the outset it must be remembered that the skilled person would have come to a reading of these documents with the benefit of all the common general knowledge to which I have referred. There are several aspects of this which are of particular relevance. First, he would have known that S_N2 glycosylation reactions are rare, if not extremely so. In the course of his cross examination, Professor Fleming was unable to recall any text which taught increasing the concentration of a nucleophile in a glycosylation reaction to promote an S_N2 pathway over an S_N1 pathway. Second, it was recognised that retaining stereochemical control of glycosylation reactions was therefore very difficult. Various techniques were available, of which neighbouring group participation was one of the most promising, but this requires a suitable group at the C-2 position. This apart, retention of stereochemical control had proved a frustration to chemists and was recognised to be a challenge. There were avenues to explore, such as by judicious use of reaction conditions and protecting groups, but the general knowledge did not provide any signposts as to what would be successful. Third, and importantly, the skilled person would have been well aware that increasing the concentration of the nucleophile will only increase the rate of an S_N2 reaction if it is already occurring, albeit slowly. If a particular glycosylation is only proceeding by the S_N1 pathway then increasing the concentration of the nucleophile will have no effect on its rate whatsoever; it can only be affected by the concentration of the electrophile.

The skilled person would therefore have approached Chou and Hertel with little expectation that the glycosylation reaction they describe was proceeding other than by the S_N1 pathway. There was one matter which suggested that this reaction might be an exception, namely the presence of the difluoro substituents at C-2 of the sugar moiety and their capacity to prevent or inhibit dissociation of the mesylate ion, and this was something that was clearly in the minds of Hertel and his colleagues. But a careful reading of the papers eliminates that possibility. Chou describes a specific investigation of the reaction mechanism by separating out the mesylates and subjecting them to individual glycosylation reactions. The result was the same in each case, and consistent only with an S_N1 reaction, as the authors themselves conclude, albeit on one occasion by using the guarded expression “predominant reaction mechanism”. Although Professor Fleming was suspicious that Chou measured some ratios rather accurately to two decimal places and other rather critical ones not at all, he accepted there was no evidence of an S_N2 reaction going on, even in the background, or that it might otherwise be a borderline case proceeding by both pathways.

Hertel 3 is entirely consistent. There are two particular aspects of its teaching that bear on this issue. It describes a reaction in which the mesylates are present in a ratio of 1:1, yet the nucleoside products are present in a ratio of 4:1, with the undesirable α nucleoside predominant. This in itself points to an S_N1 reaction with a measure of facial selectivity. But this is reinforced by the investigation of the impact on the reaction of running it under fusion conditions, that is to say at a very high concentration of nucleophile and electrophile. This would be expected to favour any S_N2 reaction which might be occurring. But the results are exactly the same in terms of the ratio of nucleoside products, which remains at 4:1. It is only the yield which is affected, and adversely so.

It must also be borne in mind that all of the above involve a focus on the teaching of Chou and Hertel as to the nature of the reaction pathway. But this is not the heart of Chou’s teaching. The problem addressed by Chou is the same as that addressed by the Patent, namely to find a better way of making gemcitabine and, specifically, a process suitable for kilogram production. Chou highlights the selection of benzoyl protecting groups which has permitted what the authors describe as the crucial selective crystallisation of the desired ribonolactone. It has also permitted crystallisation of gemcitabine from a 1:1 final anomeric mixture, which they consider a vast improvement over the 4:1 mixture produced by Hertel. So the skilled person would understand that Chou has improved Hertel’s process, not by trying to find a way to make the reaction proceed by the S_N2 pathway but by modification of the protecting groups which has allowed the described selective crystallisation and dramatically improved the adverse facial selectivity experienced by Hertel.

As I have said, it was Professor Fleming’s view that the skilled person would further investigate the effect of increasing the concentration of nucleophile with a view to seeing if it revealed an S_N2 pathway. He maintained this in cross examination, as the following passage from Day 2, at 202-203 shows:

“Q. As I understand it, just simply on that, what you say the skilled person would do is take the experimental [conditions] in Chou and add more nucleophile.

A. Yes. The first thing to do is to get it more concentrated because these reactions are done pretty dilute.

Q. Essentially, what you are saying they would is take the experiments in Chou and add more nucleophile.

A. One of the things they would do is that, yes.

Q. First, second?

A. I would cut down the solvent first; but I do not know, you see. This is where the experimentalist does in fact rule because it may turn out that you cannot get them all into solution in less solvent than is here. I do not know what the situation is because I have not handled these compounds myself.

Q. I see.

A. But those are the directions in which my thinking would go, certainly; or rather, I should say, how I think the skilled person would think.

Q. I understand. Just to be absolutely clear -- I should say I am surprised, but I just want to be clear because I do not want there to be any misunderstanding later -- I had understood your evidence to be that you were saying you would take the Chou conditions and add more nucleophile. As I understand it, you are saying that is one of the things you would do. The other thing you would do is what?

A. Well, you would look at the solvents but decide that there was not very much room for making them less polar than they already are and you would make things concentrated. It is the concentration of the nucleophile that is critical.”

In my judgment there are a number of points to note about this evidence. The first is that Professor Fleming recognised there was little prospect of any improvement resulting from a change in solvent. It was the concentration of the nucleophile he considered to be critical. Second, and as I have explained, Hertel and Chou report no evidence of an S_N2 reaction and the skilled person knows that if no part of the reaction is proceeding by the S_N2 pathway then increasing the concentration will have no effect on the outcome. Professor Fleming’s approach requires the skilled person to believe that there is still a prospect that an S_N2 reaction is taking place, despite his common general knowledge and the teaching of both documents. Indeed, it was part of Professor Fleming’s reasoning that the skilled person would, having started with S_N2 in mind, hardly abandon even the prospect of it (Day 2 at 201). I do not accept that this is a legitimate assumption in the light of their teaching. In any event, it is hardly a ringing endorsement that his approach is an obvious one. Moreover, Professor Fleming also accepted that there is no evidence that any background S_N2 reaction could be speeded up sufficiently to produce a material difference to the anomeric outcome, as shown from the following passage of his cross examination from Day 2 at 206:

“Q. So increasing the concentration of the nucleophile by 10 fold, you ought to increase it 10 times. If it is 10 times of very little, it is still very little, is it not?

A. Yes, but it gives you more beta than alpha if you start with alpha mesylate.

Q. The point that I am trying to put to you is, would you accept this, that there is no evidence in Chou that such a ghostly SN2 reaction is large enough already that speeding it up 10 fold will make enough of a difference that you see a material difference to the anomeric outcome?

A. No, there is no evidence in Chou for that.”

I found Professor Boon’s evidence as to the attitude of the skilled person more persuasive. He put it this way in cross examination of Day 3 at 441-442:

“Q. So can I ask you to imagine if you were the skilled person in 1992 with the common general knowledge and so forth that we have, that you set out in your report, and if I can ask you to imagine that Dr. Chou comes to you and says "I had the alpha mesylate. I had stereocontrol at the C-1 position and then I lost it, because it was an S_N1 predominant S_N1 reaction mechanism", and he says, "What shall I do?"

A. He would give you a simple answer that is also in my report. It is, you can get facial selectivity also in an S_N1 mechanism. So if you start playing with your protecting groups, and that is already what the prior art is teaching, if you have trimethyl S, you get mainly alpha, now I change protecting groups and I go to benzoate still an S_N1 pathway but now, wow, I get lots more [beta]. I would suggest to Dr. Chou as the skilled person, let us keep playing with protecting groups. If we can go from 4-1 to 1-1 maybe we can go to 1-4 in favour of the beta anomer. That would be the most logical, what I think the skilled person would suggest to Dr. Chou.”

So he would pursue the teaching of Chou and look at other protecting groups. When pressed as to the possibility of looking further for an S_N2 reaction, he explained at 442-444:

“Q. And is one answer that you might give, I mean, not necessarily your first answer, to Dr. Chou in that situation, where he said to you "I had the stereocontrol, I had the alpha mesylate but I lost it because it was an S_N1 predominant S_N1 reaction mechanism" and he tells you it was a predominant S_N1 reaction mechanism, against that background, is not one answer that you might give to Dr. Chou in that situation, "Well, have you thought of trying the S_N2 pathway?"?

A. I think that Dr. Chou knew quite a bit of carbohydrate chemistry. I think that Dr. Chou, there is another Dr Chiu who is a famous nucleoside chemist -- would know that S_N2 glycosylations by direct nucleophilic substitution in an S_N2 pathway are very, very rare. So I think if the skilled person would have suggested that, I think Dr. Chou would have said, "well, that is not really the way to proceed". I do not think the skilled person would have suggested that in the hypothetical world where the skilled person would make a suggestion I think Dr. Chou would have said that there is really no or very little ground to make this work. The prior art basically shows that, yes, these two difluorides might have promoted an S_N2, but failed. So why keep banging your hand against a door that does not want to open. A wall, not a door that does not want to open? That sounds strange. I sometimes say strange things. But I think you get the point. It works.

MR. MILLER: All that Dr. Chou has done is to run his glycosylation step with the alpha mesylate and found that he got a predominant S_N1 reaction mechanism and lost stereocontrol of the C-1 position. I mean, that is all he has done.

A.

That is in accord with what we generally find in glycoside chemistry. So the two fluorides might make me and the skilled person speculate otherwise. The experiment has been done in two very different settings and it is S_N1. So why keep -- can I say banging your head against the wall?”

It must also be appreciated that the search for an S_N2 pathway requires the skilled person to carry out the fractional crystallisation of the ribonolactones so as to obtain an enriched source of α and β mesylates. It does not simply involve varying the conditions of the synthetic scheme which Chou has proposed. For all the foregoing reasons I do not accept it was an obvious path to take.

I feel reinforced in my conclusion by a number of further matters. The first is motivation. It must be recalled the skilled person is assumed to have been looking for a better way of making gemcitabine and, on the further assumption he was minded to consider progressing down the path for which Scinopharm contends, he would have appreciated that the result of carrying out an S_N2 reaction on a 1:1 anomeric mixture of mesylates would have been a 1:1 anomeric mixture of the nucleosides. Accordingly, it was only a path worth pursuing if the skilled person had a source of the α mesylate. This would have posed a significant problem from a practical perspective. The only known way of obtaining the α mesylate was to use the selective crystallisation described by Chou. But the skilled person would have appreciated this was not a satisfactory approach for a commercial synthesis because it would have meant that only half of the starting material was available for the putative S_N2 reaction unless a way could be found to convert the β mesylate into the α mesylate. This was something on which the experts disagreed. In re-examination Professor Fleming thought it could be achieved by putting the β mesylate into water to hydrolyse it. Professor Boons thought it would not be easy to hydrolyse such a glycoside and there would be a risk of side reactions. On this point I prefer the evidence of Professor Boons, who seemed to me to have greater experience of this kind of reaction. Moreover, even assuming he was considering proceeding down this route, the skilled person would not have expected to achieve a pure S_N2 reaction. As Professor Fleming accepted, the best he could realistically have expected would have been a mixture of S_N1 and S_N2. So he would still have been faced with the need to carry out a selective crystallisation at the end of his process to separate the gemcitabine from the unwanted α nucleoside. Whilst theoretically possible to carry out a parallel S_N1 glycosylation reaction on the β mesylate and so obtain a 1:1 mixture of the nucleoside anomers, Professor Fleming recognised this was not a realistic approach; indeed he described it as “a crazy way to do it”. In summary, therefore, pursuit of an S_N2 mechanism based upon Chou would have resulted in a lower yield and an additional crystallisation step, making it a far from attractive avenue to explore, unless the skilled person had the further insight to suppose that further and better ways of making the α mesylate might become available, as they duly did and are described, though not claimed, in the Patent.

A further consideration arises from the evidence of Professor Fleming which I have quoted at paragraph [113] above. Lilly does not accept that adding more nucleophile to Chou or taking other steps to increase its concentration will necessarily produce a result which demonstrates S_N2 character. As I have mentioned, it is common ground that having an excess of nucleophile is not necessarily sufficient to induce the reaction to proceed by the S_N2 pathway, and such is not promised by the Patent. If it were needed, further evidence is to be found in Hertel 3 itself. In these circumstances I consider Lilly can fairly take the point that Scinopharm has not established what the result of adding more nucleophile to Chou would be. The Patent does not contain an example which shows what would happen and I consider it a real possibility that a modest increase in nucleophile concentration would not reveal an S_N2 reaction, particularly in the light of Professor Fleming’s evidence to which I have referred.

I also consider it relevant that the skilled person had a number of other avenues to explore. The experts were agreed the skilled person would have known it was possible to influence the anomeric outcome of an S_N1 reaction by the use of different protecting groups and reaction conditions. All of these could have been investigated. He would also have recognised that Hertel and Chou were unusual in using sulfonyl leaving groups and he would have considered using others, particularly halides, as Professor Fleming accepted.

Professor Boons also believed the skilled person would have considered changing the sequence of the synthesis. He outlined two possibilities. First, the skilled person would have known that a Vorbrüggen glycosylation of a silylated nucleobase with a ribosyl donor having a C-2 ester participating group will provide selectively a β-linked nucleoside. After removal of the ester to provide a ketone, he believed that the two fluorides could potentially then be installed by treatment with diethylaminosulfurtrifluoride (DAST) or SF₄. He also thought the skilled person might have considered a synthetic scheme to make gemcitabine starting from cytidine, again using DAST. Professor Fleming said he would not be surprised if someone tried such a reaction and described DAST as a cunning but nasty agent which he had never had to use. Professor Boons, on the other hand, had employed it and, though he agreed it required particular precautions, thought it was widely used. I accept the evidence of Professor Boons on this point. He had actual experience of the product and I am satisfied it is not of such a nature as would have deterred the skilled person from using it.

The skilled person would also have appreciated that there are many steps involved in the synthesis of gemcitabine and these provided lots of scope for optimisation and improvement and that, at looking at the synthesis as a whole, there were many areas worthy of study apart from the glycosylation step.

In the light of all the forgoing I have reached the conclusion it was not obvious to carry out a process for preparing a relevant β anomer enriched nucleoside by conducting an S_N2 displacement reaction in the light of Chou and Hertel 3. This allegation of obviousness therefore fails.

Hubbard was published in 1984 in a well respected journal, Nucleic Acid Research. It explains there was a need for substantial quantities of b-2'-deoxynucleosides for potential therapeutic use. The available synthetic routes were said to be unsatisfactory because they usually led to the production of both a and b anomers. It continues with the statement: “If the condensation can be performed under conditions which prevent sugar anomerisation and encourage S_N2-type reaction, it should be possible to obtain high yields of b-2'-deoxynucleoside.” The authors also observe: “We have shown that by a rational choice of solvent, catalyst and other conditions, one can produce very high yields of either almost exclusively b or a nucleoside.”

The reaction was drawn by Professor Fleming as Scheme 6:

Hubbard explains that a prerequisite for the formation of high yields of the b nucleoside is the prevention of anomerisation of the α sugar. As Professor Fleming explained, Hubbard found that the α chlorides were easily interconverted in the relatively polar solvent, acetonitrile, but only slowly in non polar solvents like chloroform. When the silylated pyrimidine was added to a mixture of the α and β chlorides in chloroform, the β chloride was consumed rapidly, and gave the α-nucleoside, but longer reaction times increased the proportion of β nucleoside as the α chloride was then consumed, suggesting that both reactions were proceeding by the S_N2 pathway. Hubbard then used a known procedure to isolate the pure α chloride by crystallisation, and treated it with a silylated pyrimidine in chloroform, and obtained a mixture rich in the β nucleoside, with the inversion of configuration characteristic of an S_N2 reaction. Hubbard also investigated the effect of using a two fold excess of nucleophile and found this also increased the proportion of desired β nucleoside, and of using less nucleophilic pyrimidines and found these decreased stereospecificity. Various catalysts increased both the rate of anomerisation and nucleoside synthesis.

Overall, the experts agreed with Hubbard’s conclusion that the evidence was consistent with the view that this nucleoside synthesis was proceeding by the S_N2 pathway, that sugar anomerisation and decomposition had to be taken into account and that the β chloride reacted more rapidly to give the α nucleoside than did the α chloride to give the β nucleoside. Finally, but importantly, an excess of nucleophile increased the rate of the S_N2 reaction.

The first issue which arises on Hubbard is whether it is permissible to read it with Chou and Hertel. I am satisfied that it is. Professor Fleming found it in a very short time having set about a search to familiarise himself with the art. He picked it out as one of those publications likely to be helpful. Had he not done so then he was clear it would have been found very soon afterwards because of the number of times it is referenced.

As for the differences between Hubbard and the invention of the Patent, these were explained by Professor Boons. He pointed out:

In his view, the skilled person would have understood that the presence of two strongly electron-withdrawing fluorides at C-2 of the sugar moiety would significantly reduce its electrophilicity. Therefore a stereoselective and high yielding glycosylation with such a sugar would require significant alteration of reaction conditions. He also explained that chloride is a commonly used leaving group but is certainly not the most reactive leaving group amongst the halides. Further, he suggested that a comparison of reaction outcomes described in the Hubbard paper and the Patent indicates that reaction parameters such as solvent affect a 2,2-difluoro sugar very differently compared to a 2-deoxy sugar. In his report he then proceeded to identify a series of changes that the skilled person would consider, on the assumption he had been asked to apply the teaching of Hubbard to the problem of preparing 2-deoxy-2,2-difluoronucleosides. I did not understand this evidence to be challenged.

Instead, the case of obviousness put to Professor Boons was that Hubbard would give the skilled person a measure of confidence that the Chou reaction could be made to proceed by the S_N2 pathway, a proposition with which he did not agree. It was his consistent view that S_N2 glycosylation reactions are very rare.

Scinopharm also contends that the skilled person would have understood from Hubbard that a predominant S_N1 reaction could be pushed to an S_N2 reaction by increasing the concentration of the nucleophile. I accept this was so. Indeed, as I have found, it formed part of the common general knowledge. But it is important to have in mind that it only applies to reactions which are proceeding in part by the S_N2 pathway. Increasing the concentration of the nucleophile will have no effect on a reaction that is proceeding only by the S_N1 pathway.

In my judgment the allegation of obviousness based on Hubbard therefore adds nothing to that based on Chou and Hertel and must be rejected.

I can take all these publications together. Once again, the allegation of obviousness is not founded on any one of these papers as such. Rather, it is said that the invention of the Patent was obvious in the light of these papers when read with Hertel and Chou. However, there was no evidence or attempt to prove that a skilled person would have found any of these papers in the light of his common general knowledge or on a reading of Hertel and Chou. In my judgment it is not permissible to present to an expert different items of prior art which are not said to be common general knowledge and ask him whether it would have been obvious to read them together. It may well be the case that the skilled person, presented with Hertel and Chou, would not have considered it necessary to supplement his common general knowledge and, even if he did, he may never have found these further publications. I therefore reject the submission it is permissible to read them, individually or collectively, together with Hertel or Chou. The allegation of obviousness based on these publications fails at the outset.

As for the substance of the papers, they each describe the production of deoxyribonucleosides by S_N2 reactions. But those reactions are different from those of the Patent. They are deployed in support of the same proposition advanced over Hubbard and, for like reasons, add nothing to the case based upon Hertel and Chou and must be rejected.

Of the allegedly independently valid claims, it is now accepted that claims 1, 11, 12, and 13 stand or fall together. The remaining claims in issue are 2, 3, 8, 9 and 14. In the light of my finding on claim 1, the allegation of obviousness against these claims must also fail but, in case this action goes further, I shall state my findings in relation to them, albeit shortly.

Claim 2 is directed to the process of claim 1 with the nucleobase selected from a narrower group, and the reaction is carried out in a solution having a carbohydrate concentration above 20g/100ml and in which the solvent is a high boiling point inert solvent.

Lilly contends this claim is valid over Hertel and Chou in any event because the skilled person would not seek to promote an S_N2 reaction by increasing the concentration of the electrophile. I do not think there is anything in this point. It was accepted by both experts that reducing the amount of solvent will increase the concentration of both the nucleophile and the electrophile and favour the S_N2 pathway. There is therefore nothing in this claim over and above claim 1.

Claims 3 and 14 can be taken together. Claim 14 requires the leaving group to be triflate. Claim 3 is broader and calls for a fluorinated sulfonyloxy leaving group. I am not satisfied these claims were obvious. As Professor Boons explained, these are better leaving groups than mesylate and so more likely to promote an S_N1 reaction. Professor Fleming accepted this was so but did not think it would stop the skilled person from trying. Overall, I again prefer the evidence of Professor Boons. It would have been counter-intuitive to adopt a leaving group which would have been thought to favour the S_N1 pathway.

Claims 8 and 9 require fusion conditions. Fusion as a concept was undoubtedly part of the common general knowledge but, as Professor Boons explained, it was not commonly used in the case of pyrimidines. These have a much a higher melting point than purines and hence the reaction must be conducted at high temperature. This in turn may impact on the yield, just as it did in the case of Hertel. Accordingly, says Lilly, the claims were not obvious. I reject this contention. I am satisfied that all the matters to which Lilly refers were common general knowledge and that the skilled person would have appreciated the high temperature necessary to carry out the reaction might impact on the yield. But claims 8 and 9 are not limited to processes which produce a high yield. They simply require fusion conditions. They add nothing to claim 1.

Scinopharm’s principal case is that the claims are obvious, that is to say it was obvious to seek to obtain enrichment of the β anomer by increasing the concentration of nucleophile. But, Scinopharm continues, if and in so far as Lilly says that the inventive step lies in particular conditions needed to perform an S_N2 reaction, then Lilly has failed to provide the skilled person with sufficient information to enable the invention to be performed across the breadth of the claims. In particular, one of the problems with the Patent is that it does not identify the conditions which will consistently produce β enrichment.

I have considered the allegation of obviousness based upon the cited art and rejected it. The invention is the process of claim 1. The inventive step was the discovery that the α anomer enriched mesylates of the claim could be reacted under nucleophilic displacement conditions which favour the S_N2 pathway and hence inversion to provide β anomer enriched nucleosides.

The question then is whether the claim is enabled across its width. Here I have some sympathy with Scinopharm in that the claims certainly do not identify all those process steps which are necessary to achieve that result and, as I have indicated, both sides contend that having at least a molar equivalent of nucleobase is not itself a sufficient condition. However, Professor Boons has explained in his unchallenged evidence that the Patent provides a large number of examples and the skilled person would be able to identify the general conditions which would enable him to work the invention and he would certainly have been able to follow the teaching to produce a β enriched product, and gemcitabine in particular. The allegation of insufficiency therefore fails on the evidence.

The allegations of obviousness and insufficiency have not been made out. The Patent is valid and the action must be dismissed.