Review
`
`Annals of Oncology 6'. 871-881, 1995.
`O 1995 Kluwer Academic Publishers. Primed in the Netherlands.
`
`Folate-based thymidylate synthase inhibitors as anticancer drugs
`
`A. L. Jackman' & A. H. Calvertz
`
`‘The CRC Centre for Cancer Therapeutics at the Institute of Cancer Research, Sutton, Surrey; ’I7re Cancer Research Unit, University of
`Newcastle upon Tyne, Newcastle upon Tyne, UK.
`
`Summary
`
`The enzyme, thymidylate synthase (TS) is considered an
`important
`target for the development of new anticancer
`agents. Moreover, the folate—binding site in TS is believed to
`offer better opportunities for the design of highly specific
`inhibitors than the pyrimidine (dUMP) binding site. This
`belief led to the design of N '°-propargyl-5,8-dideazafolic
`acid (CB3717), a quinazoline-based drug which had anti-
`tumour activity in clinical studies. Occasional, but serious
`nephrotoxicity led to the withdrawal of CB37l7 from further
`clinical
`study. More water-soluble and non-nephrotoxic
`analogues were developed with an interesting diversity in bio-
`chemical profile, particularly with respect to interactions
`with the reduced-folate cell membrane carrier (RFC) and
`folylpolyglutamate synthetase (FPGS). An example of a com-
`pound that uses both of these processes well is the quinazo-
`line, ZD1694 (Tomudex), a drug which is about to complete
`phase II] evaluation for colorectal cancer. High chain length
`polyglutamates are formed that are up to 70-fold more
`potent TS inhibitors than the parent drug (Ki tetraglutamate
`-= 1 nM). Furthermore they are retained in cells/tissues for a
`prolonged period. A number of other novel folate-based TS
`inhibitors are currently in pre-clinical or clinical study. For
`
`example, LY2315l4 is a pyrrolopyrirnidine analogue in
`phase I study and, although less potent as a T8 inhibitor, has
`biochemical properties similar to ZDI694. Another com-
`pound
`in
`phase
`I
`study
`is
`the benzoquinazoline,
`BW]843U89 which has somewhat different properties. It is
`a very potent TS inhibitor (Ki = 0.09 nM) and an excellent
`substrate for the RFC (human) and FPGS, but polyglutama-
`tion proceeds to diglutamate only and is not accompanied by
`increased TS inhibition. Another highly water-soluble com-
`pound in pre-clinical development is ZD9331 which was
`specifically designed to use the RFC but not be a substrate
`for FPGS. Potent TS inhibition (Ki = 0.4 nM) was achieved
`through a rational programme of computerised molecular
`modelling of the active site of “IS and a large database of
`structure-activity relationships. Two lipophilic compounds
`were designed to be devoid of interactions with either the
`RFC or FPGS. High resolution crystal complexes of E. coli
`TS were central to obtaining potent TS inhibitors and both
`AG337 (Ki human recombinant TS - 16 nM) and AG331
`(Ki - 12 nM) are in clinical studies. This portfolio of novel
`compounds therefore comprehensively addresses the poten-
`tial of TS as a target for cancer chemotherapy.
`
`Key words: thymidylate synthase, antifolates
`
`Introduction
`
`Thymidylate synthase (TS) catalyses the methylation of
`dUMP to give TMP, which after metabolism to TTP is
`exclusively incorporated into DNA (Fig. 1). This fact
`alone makes TS an attractive target for the develop-
`ment of an anticancer agent. The co-substrate for the
`reaction is the reduced-folate cofactor, 5,10-methylene
`tetrahydrofolate (CHZFI-I4) which becomes oxidised to
`dihydrofolate (Fl-I2) during this reductive methylation
`reaction. Methotrexate (MTX; Fig. 2a) and 5-t]uorou-
`racil (FU; Fig. 2b) were amongst the earliest anticancer
`drugs developed and both partly act through inhibition
`of TS (Fig. 1). MTX primarily inhibits dihydrofolate
`reductase (Dl-IFR), an enzyme which functions to re-
`generate FH, (produced in the TS reaction) to the fully
`reduced tetrahydrofolate (FH,) form (reviewed in [1]).
`This in turn accepts a 1 carbon unit from a donor such
`as serine allowing it to function once more in 1 carbon
`transfer in various folate-dependent reactions such as
`
`TS and the enzymes involved in de novo purine syn-
`thesis (Fig. 1). FU is metabolised intracellularly to pro-
`duce a number of anabolites. One of these, 5-fluoro-
`deoxyuridine monophosphate (FdUMP) binds tightly
`to the dUMP binding site of 'IS [2]. However other
`anabolites of FU are believed, at least in some tumours
`and in some administration protocols, to have other
`effects, particularly after incorporation of the base into
`RNA [3,4].
`MTX and FU have each found roles in the treatment
`
`of certain tumour types, alone or in combination with
`other drugs. For example, MTX is still an important
`single agent treatment for low risk choriocarcinoma [S],
`and, when combined with other drugs, for the treat-
`ment of childhood acute lymphoblastic leukemia
`(maintenance therapy) and non-Hodgkins lymphoma
`[6]. MTX and/or 5-FU, when used in combination with
`other agents, have a place in the treatment of the com-
`mon solid tumours e.g. CMF (cyclophosphamide/
`MTX/5-FU) therapy for breast cancer [7]. Bolus 5-FU
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`RFC modabd
`:0 lD1694 (gluln
`D1094
`LY231614 ——F l.Y23181((glu)n
`BW1843UI9 *5 BW1843Ul9[¢|u)2
`
`5.10-CHQFH4
`
`dUMP
`
`IE1
`
`RNA
`
`DNA
`
`Fig. I. Thymidylate synthase as a target for antifolates and 5-
`fluorouracil.
`
`as a single agent had, for many years, been the standard
`drug for the treatment of colorectal cancer, albeit with
`very limited response rates (~ 10%) and no marked
`improvement on the survival of treated patients (re-
`viewed in [8, 9]). More recently the use of an adjuvant
`F U + Levarnisole regimen for Dukes C colorectal can-
`cer has been associated with a marked improvement in
`survival [10], suggesting a potentially curative role for
`drugs of this type (data from studies randomised
`against FU alone are not available). It is thus clear that
`the place of these drugs in clinical protocols has estab-
`lished the principle that some solid tumours may be
`antimetabolite-sensitive.
`The biochemical pharmacology of both MTX (re-
`viewed in [l]) and FU (reviewed in [9]) has been very
`widely investigated and understood, and it may be
`argued that the vast amount of knowledge accrued is
`out of proportion to the clinical usefulness of either
`drug. However such fundamental research has led to
`the design of alternative administration protocols, drug
`combinations, modulatory and rescue agents that im-
`proved response rates and overcame resistance in par-
`ticular tumour types [8, 9]. Examples include the use of
`high-dose MTX followed by Leucovorin (LV) rescue
`for the treatment of osteogenic sarcomas (11) and the
`sequential dosing of MTX followed by FU for colo-
`rectal cancer (reviewed in l8,9]). The most notable
`example in recent years relating to the use of FU was
`the discovery that the folate cofactor for the TS reac-
`tion can be low and limiting for the formation of the
`stable ternary complex in which FdUMP binds (re-
`viewed in [9]). The provision of LV to elevate the co-
`factor pool forms the basis for the now commonly used
`
`FU/LV clinical protocols for the treatment of advanced
`colorectal cancer [8, 9] as well as for adjuvant therapy.
`However, it is probably fair to conclude that increased
`response rates, in the order of 25%-30%, are seldom
`accompanied by a significant improvement in patient
`survival [8, 9].
`Other antimetabolites have also found roles in can-
`
`cer therapy (particularly the leukemias) but are not
`generally considered particularly active
`in
`solid
`tumours. It may be pertinent to question whether the
`concept of using antimetabolite therapy to treat the
`common (and usually slowly proliferating) cancers is
`flawed, whether we are approaching the wrong targets
`or whether the targets are right but we are just using
`inadequate drugs. Thymidylate synthase is probably
`one of the best examples of a target where, in a few
`years, these answers should be available. This is for a
`number of reasons. First, there has been the rational
`design of specific, folate-based inhibitors of TS in the
`last 15-20 years. Commitment to the rapid develop-
`ment of TS inhibitors for clinical study initially lay with
`the Institute of Cancer Research and their collabora-
`
`tors, ICI Pharmaceuticals (Zeneca Pharmaceuticals)
`which led to the discovery and clinical development of
`CB3717 (Fig. 2c) in 1979 and 1981, respectively [12-
`17}. Since then at least three other drug companies
`(usually with external academic collaborators) have
`been actively involved in this area. It was soon realised
`that the scope for novel chemical modifications was
`large, and that factors other than potent TS inhibition
`were generally important
`for
`the phannacological
`activity of novel compounds. For one of the companies,
`Agouron Pharmaceuticals, the approach to the design
`of novel chemical entities was somewhat different.
`Their starting point was the X-ray crystal structure of
`bacterial TS co—crystallised with FdUMP and CB3717
`[18]. These different approaches have led to an interest-
`ing diversity in chemical structures and biochemical
`profiles. A second reason why elucidation of the useful-
`ness of TS as a target may be forthcoming, is the wealth
`of knowledge available on the mechanisms involved in
`novel compound activity and potential mechanisms by
`which cells may be, or become, resistant. Much of this
`has been the direct result of the large amount of fun-
`damental research into antifolates and fluorinated pyri-
`midines that has been performed by academic organi-
`sations over several decades. Undoubtedly this has
`aided ‘fast-track’, pre-clinical development of new
`folatc-based TS inhibitors and specific examples will be
`highlighted below.
`
`CB3717: The first clinically evaluated folate-based TS
`inhibitor
`
`CB 3717, the first folate-based TS inhibitor to be devel-
`oped clinically, was developed following the discovery
`that certain 2-amino-4-hydroxy quinazoline analogues
`of folic acid both inhibited TS and had antitumour
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`NH;
`
`H
`
`°
`
`F
`
`H2NJ\\N N’
`23. MTX
`
`O H
`
`'
`
`coo“
`
`°J\'.'
`
`I
`
`0
`
`HN
`
`.,.x~. 0
`
`2h. BW 1B43U89
`
`Fig. 2. Structures of some folate-based thymidylatc synthase inhibitors.
`
`activity [16]. The observations both of clinical activity
`and toxicity of CB3717 (Fig. 2c) are salient to the
`development of subsequent analogues and are outlined
`in the context of knowledge of TS inhibitors. The
`potency of CB3717 as a TS inhibitor (Ki ~3 nM) was
`central to its further development as it represented a
`major advance in the search for a specific, potent,
`folate-based inhibitor of this enzyme [12, 19, 20].
`CB3717 was shown to have slightly better folylpoly-
`glutamate synthetase (FPGS) substrate activity than
`MTX and polyglutamate metabolites were formed
`intracellularly [20,21]. These polyglutamates, particu-
`larly those of higher chain length, were shown to be ex-
`quisitely potent TS inhibitors with estimated Ki values
`of ~30 pM [22,23]. Furthermore,
`these polygluta-
`mates were retained inside cells [21]. Later, the rate of
`this polyglutamation was demonstrated to be relatively
`slow compared with certain other quinazoline ana-
`logues, a property assigned to the poor affinity of
`CB3717 for the reduced-folate cell membrane carrier
`
`(RFC) and its moderate affinity for FPGS [24, 25].
`Although many of the finer points of CB3717 prop-
`erties were unknown at the time of its early develop-
`ment, it represented a major advance in this field. The
`problems of testing inhibitors of TS in rodent anti-
`tumour and toxicity models were beginning to be re-
`alised (see below). Nevertheless CB3717 had antitu-
`mour activity attributable to TS inhibition and it was
`therefore advanced to phase I clinical study. At this
`time the development of CB3717 became a joint ven-
`ture with [Cl Pharmaceuticals and a phase I study com-
`
`menced in 1981 at the Royal Marsden Hospital/lnsti-
`tute of Cancer Research. Antitumour activity was
`demonstrated in this and further phase I/ll studies,
`particularly in breast,
`liver and platinum refractory
`ovarian cancer [13, 17, 24, 26, 27]. No activity was
`demonstrated in colorectal cancer [28]. Transient rises
`in liver transaminases were noted but, more seriously,
`sporadic life-threatening toxicity consisting of nephro-
`toxicity coupled with myelosuppression was seen (un-
`published data). A decision was therefore made that,
`although the concept of TS as an antitumour target was
`a good one, CB3717 itself would not be developed
`further. However, an ongoing biological and chemical
`synthetic programme at the Institute of Cancer Re-
`search had identified the poor aqueous solubility of
`CB3717 at acid pH as the cause of the nephrotoxicity,
`and had designed and synthesised a more water-soluble
`analogue (desamino-CB3717) devoid of this toxicity in
`mice [20, 29]. The discovery of desamino-CB3717 was
`the impetus for a new and very vigorous collaboration
`between the Institute of Cancer Research and ICI Phar-
`
`maceuticals which led to the synthesis and evaluation
`of >3,000 quinazoline analogues, most of which had
`"IS as the intracellular locus of action. The compounds
`fall into four broad classes, each with distinct intra-
`cellular biochemical and phannacological properties.
`These properties include different interactions with the
`RFC and FPGS. ZD1694 (Tomudex; in clinical devel-
`opment), ZD933l
`(in preclinical development) and
`further
`lead compounds with different
`features
`emerged from this portfolio of compounds.
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`874
`
`Other phannaceutical companies (and their aca-
`demic collaborators) which have actively sought folate-
`based TS inhibitors have now reached the stage of pre-
`clinical or clinical evaluation. These will be discussed in
`further detail below.
`
`Second generation analogue design
`
`The first compound to emerge from new analogue syn-
`thesis, which was widely investigated, was the 2-desa-
`mino-2-methyl-analogue of CB3717 (ICI 198583;
`MPPDF) [30]. [Cl 198583 had good water-solubility,
`no detectable nephrotoxicity in mice and improved in
`vivo antitumour potency (L1210:lCR) over CB3717
`and desamino-CB3717 [31]. This compound repre-
`sents a class of highly active compounds displaying
`high affinities for the RFC and FPGS as well as for the
`target enzyme, TS. Results indicated that although
`there was no direct correlation between TS inhibition
`
`and cytotoxic potency there was a good correlation
`between the extent of polyglutamation and cytotoxicity
`[24,25]. Furthermore, the rapid polyglutamation of
`some analogues and the increased potency of the poly-
`glutamate metabolites towards TS result in their cel-
`lular pharrnacokinetics/dynamics being different from
`those of the DHFR inhibitor, MTX. Another feature of
`this class of compound is their activity by bolus admini-
`stration in mice, not necessarily predicted from their
`rapid plasma clearance, but consistent with a high
`degree of polyglutamate formation [20,31]. Although
`ease of administration and cytotoxic potency were con-
`siderations that favoured the development of this class
`of TS inhibitor, a positive decision was made to develop
`a drug that took advantage of the fact that FPGS was
`reported to be expressed relatively highly in some
`tumours and believed to explain some of the selective
`activity of MTX in mice ([32,33] and reviewed in [1]).
`A more recent report by Rumberger et al. endorses this
`further [34].
`A general theme in the synthesis and development of
`TS inhibitors, whatever their structural class (several of
`which are described later) is the elegant manner with
`which they can be evaluated in vitro because of the
`array of biological models and assays developed
`[2 5, 35]. However, it is their in vivo evaluation that pro-
`vides the most challenge to their development as drugs.
`One major problem is the high plasma thymidine
`(dThd) level in rodents relative to man (at least 10-fold
`higher) which, through the activity of the dThd salvage
`pathway, is able largely to circumvent any TS inhibition
`[36]. This accounts for the generally low activity of TS
`inhibitors in many mouse tumours and human tumour
`xenografts. Similarly their toxicity to normal proliferat-
`ing tissues is low, particularly in non-chronic admin-
`istration schedules, and therefore not necessarily pre-
`dictive of that in man. Test systems were introduced
`that at least, in part, address their in vivo activity. In-
`hibition of TS by polyglutamatable species in tumour
`
`cells that have been removed at time intervals after
`
`drug injection can serve as a pharmacodynarnic meas-
`urement [20, 31, 37]. Thymidine kinase deficient tu-
`mours such as the L5178Y TK-/- mouse lymphoma
`generally respond to single bolus therapy (polygluta-
`matable compounds) but normal proliferating tissues
`remain unaffected [38,39]. TK competent tumours,
`such as the L5178Y TK+/- [38,39], the L1210:ICR
`leukemia [20,31,37] or some human tumour xeno-
`grafts ]39,40], may be sensitive to TS inhibitors if
`administered for prolonged periods. This is probably,
`in part, the result of a fall in plasma dThd that occurs
`after administration of a TS inhibitor ([36] and unpub-
`lished data). Prolonged administration also gives drug-
`induced nonnal tissue toxicity. Taking these problems
`into account, it emerged that one compound had a bet-
`ter therapeutic index than the others in a short-list of
`active compounds (manuscript in preparation). This
`compound was ZD1694 (lCl D1694; Tomudex; Fig.
`2d) and was selected for pre-clinical and clinical devel-
`opment. The model systems used in the development
`of the other TS inhibitors in clinical study will be dealt
`with under the appropriate section.
`
`ZD1694 (Tomudex), a highly polyglutamatable TS
`inhibitor
`
`ZD1694 (Fig. 2d) has a Ki for isolated mouse and
`human TS of ~60 nM [37,41]. This 20-fold loss in
`inhibition compared with CB3717 is compensated for
`by a ~500-fold increase in cytotoxic potency [37, 42].
`This latter activity was prevented by co-incubation with
`dThd alone indicating that TS is the target enzyme for
`ZD1694 [37,42]. ZD1694 is internalised into cells via
`the RFC (Ki for the inhibition of MTX influx -2 p.M)
`and has a very low Km for mouse liver and human
`FPGS (1.3 p.M) [25, 37, 43]. The corresponding values
`for CB3717 are -40 pM for both proteins and for
`MTX are ~4 and —-166 p.M for the RFC and FPGS,
`respectively [21,25]. ZD1694 is almost completely
`metabolised to polyglutamate forms (tetra and penta-
`glutamates usually predominate) in a variety of tumour
`cells in culture [37, 44] and in certain normal tissues in
`mice such as liver, kidney and gut epithelium [42]. The
`higher polyglutamate forms (triglu and above) have Ki
`values for TS of ~1 nM and are not readily effluxed
`from the cell [37,41]. lt is concluded that ZD1694 is
`active as the polyglutamate metabolites and indeed
`structural modifications that enhance TS inhibition but
`
`largely prevent polyglutamation (for example C7
`methylation), give compounds that are -100-fold less
`potent as antitumour agents
`[24]. Furthennore a
`mechanism of acquired resistance to ZD 1694 is defec-
`tive polyglutamation (at least partly due to decreased
`FPGS activity) which illustrates the importance of this
`metabolising enzyme for drug activity [45]. The con-
`sequence of the formation of slowly effluxable polyglu-
`tamates is that short incubation periods with ZD1694
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`(e.g. 4 h) will give a high level of cytotoxicity which is
`not seen with analogues that carmot be polyglutamated
`[25]. Indeed TTP pools in tumour cells deplete rapidly
`during incubation with the drug and recover very slow-
`ly after resuspension of cells into drug-free medium
`[46].
`After a single i.p. bolus injection of ZD1694 to mice,
`rapid clearance of the drug from the plasma occurs
`(t./B of ~30 min) [47]. A third, much slower phase of
`elrinination is then measured, giving a persistently low,
`but possibly significant, drug level. Some of the anti-
`tumour and toxicological properties of ZD1694 are to
`be found in the literature and will be the subject of
`further communications so that only a very brief sum-
`mary is given here [39, 42, 47-49]. The problems of
`testing TS inhibitors in rodents (high dThd) means that
`very little useful data can be gained as to their thera-
`peutic index and the dose and schedule applicable to
`humans. Nevertheless pharmacodynamic measure-
`ments, such as inhibition of TS in tumour cells in vivo,
`and the use of a TK deficient tumour (L5178Y TK-)
`demonstrated that a single bolus injection of ~ 10 mg/
`kg was very active, consistent with the formation of
`highly retained polyglutamates [37, 39, 50]. Repeat
`daily dosing in TK competent tumour-bearing mice
`gave some antitumour activity (including human tu-
`mour xenografts) and normal tissue toxicity (mainly
`gastrointestinal) [37, 39, 40, 48]. Co-administration of
`dThd prevented the antitumour activity (L1210:lCR)
`and toxicity of ZD1694 which confirmed that, in vivo,
`T8 was the antitumour locus of action and toxicity was
`mechanism related [37]. Consistent with this was the
`fact
`that
`the non-mechanism related nephrotoxicity
`associated with the original compound, CB3717, was
`not observed [39,48]. Antitumour activity in mice was
`also prevented by co-administration of LV which was
`explained by a series of in vitro experiments demon-
`strating competition for cellular uptake and polygluta-
`mation [24, 37].
`The clinical evaluation of ZD1694 began in Europe
`in February 1991, recruitment being principally at the
`Royal Marsden Hospital/Institute of Cancer Research,
`one of the institutes of the co-discovers of the drug [17,
`49, 51]. The starting dose was 0.1 mg/m2 (15 min i.v.
`infusion) every three weeks, which was predicted to be
`a safe starting dose from studies in dogs (data of Zene-
`ca Pharmaceuticals). Dose escalation continued up to
`3.5 mg/m2 and included 61 patients with a range of
`solid tumours
`[51]. Dose-limiting toxicities were
`malaise, gastrointestinal and haematological (leuco-
`penia or thrombocytopenia). Other toxicities which
`were encountered included reversible rises in liver
`
`transaminases, skin rash and anorexia. Three objective
`partial responses were seen (2.6 and 3 mg/m’) in pre-
`viously treated patients with ovarian and breast cancer
`and adenocarcinoma of unknown origin [51]. Clear-
`ance from the plasma was triphasic and the half lives
`for the beta and gamma phases were ~2 and 75 h,
`respectively [51]. No relationship was observed
`
`875
`
`between various pharmacokinetic parameters and re-
`sponse or toxicity. A dose of 3 mg/m’ was recommend-
`ed as the phase 1] dose. Another phase I was per-
`formed by the NCI in the U.S.A. (using the same proto-
`col but in a mainly pre-treated colorectal patient popu-
`lation (76%)), the results of which led to their recom-
`mendation of a phase II dose of 4 mg/ml [52]. How-
`ever the first phase II studies were performed at 3 mg/
`m2 and the general experience seen in the phase I was
`reproduced in these multicentred trials. Objective re-
`sponses (CR and PR) were seen at interim analysis in
`breast (25%), platinum resistant ovarian (8.5%), non-
`small cell lung (10%), and pancreatic carcinomas (14%)
`[53]. The most exciting response rate was seen in colo-
`rectal cancer where, in a study of 176 patients with ad-
`vanced disease, 26% (95% C1. 19°/o—33°/o) had objec-
`tive responses which included 4 complete and 41 par-
`tial responses [54]. A further 30% of patients had
`minor responses. The median time to progression was
`'18 weeks and median survival was 42 weeks. Grade 1]]
`
`and IV toxicities included asthenia (12%), nausea and
`vomiting (11%), diarrhoea (10%) and leucopenia (6%).
`Overall, toxicity was considered acceptable and man-
`ageable. The colorectal response rate is comparable to
`many of the reported FU/LV studies (23%—30%).
`These facts, taken together with the relative ease of
`ZD1694 administration, encouraged the iriitiation of
`the European phase II] study in 1993 (3 mg/m2, 15
`min infusion every three weeks), randomised against
`FU and LV given as 5 daily iv. bolus injections (425
`mg/m2 FU, 20 mg/m2 LV) repeated weeks 4, 8 and
`then 5 weekly (Mayo Regimen). The first results of this
`study (439 patients) at a median follow-up of 5.3
`months are now published [55]. 20% of all patients
`receiving ZD1694, compared with 13% receiving FU/
`LV, had objective partial or complete responses
`(p = 0.059; odds ratio 1.7). In addition, a further 9.4%
`of patients receiving ZD1694 had a 40%—50% reduc-
`tion in measurable lesion size (3% in FU/LV group).
`There was less grade [I] and IV leucopenia and muco-
`sitis in the ZD1694 arm of the study (p<0.001).
`Asymptomatic and reversible rises in transarninases
`were seen more frequently in patients
`receiving
`ZD1694. Overall the authors concluded that ZD1694
`
`compares favourably with FU/LV, with relatively good
`antitumour activity and an acceptable toxicity profile.
`
`LY231514, a pyn-olopyrimidine folate-based TS
`inhibitor
`
`LY231514, a pyrrolopyrimidine folate-based TS inhibi-
`tor (Fig. 2e) serendipitously arose from a chemical pro-
`gramme at Princeton University synthesising analogues
`of DDAT1-IF (Lometrexol), an inhibitor of GAR trans-
`forrnylase [56]. LY231514, further developed by Eli
`Lilly, is a relatively poor inhibitor of isolated recom-
`binant human TS (Ki - 0.34 p.M) but is highly cyto-
`toxic to cultured CCRF-CEM human or mouse leuke-
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`876
`
`mia cells (20 nM) [56]. This seems to be due to its very
`efficient metabolism to polyglutarnates (mouse liver
`FPGS Km = 0.8 nM) with increased TS inhibitory
`potency (pentaglu estimated to be 3.4 nM assuming
`competitive inhibition) [5 6]. Thymidine reversed most
`of the cytotoxic activity, the degree depending on the
`cell line used, which suggests that "I8 is probably the
`primary locus of action although other some other
`folate target may also be involved [56]. In many re-
`spects this agent has biochemical properties similar to
`those of ZD 1694; although less active than ZD1694 as
`an inhibitor of isolated TS it is only marginally less
`active as an inhibitor of cell growth (~ 3-fold; L1210).
`Antitumour studies in vivo demonstrated complete
`inhibition of the growth of a TK- and HX- murine
`lymphoma model (L5178Y TK-/HX-) when admin-
`istered in an i.p. daily X 8 schedule of 12.5 mg/kg
`[56, 57]. Myelosuppression and gastrointestinal toxicity
`were dose-lirniting. LY231514 was also active against
`two human colon tumour xenografts (VRC5 and
`HXGC3) with > 80% tumour inhibition being observed
`at 25 and 50 mg/kg i.p. daily x 10 [56].
`A phase I clinical study (10 min infusion, weekly for
`4 weeks repeated every 42 days) started at 10 mg/m 1/
`week and escalated to the MTD of 40 mg/m2 with neu-
`tropenia being dose-lirniting [58]. Other toxicities in-
`cluded mild fatigue, anorexia and nausea. No objective
`responses were seen in the 24 patients treated. Another
`schedule investigated was a 10 min infusion daily for 5
`days repeated every 21 days [59]. At the starting dose
`of 0.2 mg/m’ some gastrointestinal toxicity was ob-
`served and dose-escalation continued. Another sched-
`
`ule recently reported was a 10 min infusion every 21
`days. An MTD was determined of 600 mg/m’ [60].
`Neutropenia was the dose-limiting toxicity and minor
`responses were observed. Phase 11 studies in colorectal
`cancer are planned I60].
`
`AG331 and AG337, lipophilic TS inhibitors in clinical
`study
`
`Agouron Pharmaceuticals, in their search for lipophilic
`TS inhibitors that interact with the folate cofactor bind-
`
`ing site, have developed two structurally dissimilar
`compounds for clinical study. The potential advantages
`of such inhibitors are their lack of a requirement for
`active uptake into cells or for polyglutamation. While
`these two processes may contribute to the antitumour
`selectivity of classical antifolates in tumours expressing
`high levels of the relevant proteins, they are also impli-
`cated in mechanisms of resistance (i.e., in tumours ex-
`pressing low levels). As a single protein interaction is
`required for their activity i.e. the target enzyme TS, the
`adopted drug design strategy was to use high resolution
`ternary crystal complexes of E. coli TS (originally with
`FdUMP and CB3717) to suggest compounds that may
`have a high binding affinity for T8. An iterative cycle of
`compound synthesis, modelling, and with certain lead
`
`compounds, new crystal complexes led to potent TS
`inhibitors with in vitro growth inhibitory potency
`similar to CB3717 (-1 nM). AG337 (compound 21 in
`[18]) is a quinazoline structure linked through C5 to a
`4-pyridylthio moiety (Fig. 2f) that was designed to
`access a hydrophobic cavity in the active site of the
`enzyme that nonnally associates with the para-amino-
`benzoyl portion of CB3717 or of the natural substrate,
`Cl-I,FH,. The reported Ki for E. coli TS is 49 nM and
`for recombinant human T8 16 nM. Curative anti-
`
`tumour activity was observed in the mouse i.m. or i.p.
`L5178Y TK- lymphoma (i.p. or oral drug administra-
`tion) and significant growth delay was seen against the
`human GC,M/TK- colon tumour xenograft [61].
`AG331 (Fig. 2g; compound 27 in [62]) is structural-
`ly unrelated to the folate cofactor and interesting
`because more conservative studies into structure-activ-
`
`ity relationships would be unlikely to reveal such a
`compound. The Ki for E. coli TS is 1800 nM and for
`human recombinant TS is 2 nM [62]. No explanation is
`provided as to why these values are so different for the
`'18 of different sources. Tumour cell growth inhibition
`(lC5o) falls in the 0.5-1 uM range. AG331 may have a
`second locus in some hepatoma cell lines which is
`hypothesised to be due to metabolism to another drug
`species [63]. Antitumour activity was observed against
`a mouse L5178Y TK- variant implanted i.p. in mice
`[64].
`Clinical studies are in progress for both drugs. A
`pharrnacokinetic and pharrnacodynamic
`study of
`AG337 (dihydrochloride fonn) given as a 24-h con-
`tinuous i.v. infusion (every three weeks) has been re-
`ported [65]. Rapid plasma clearance (t./I - <2 h) was
`observed once the infusion ceased. A dose of 1.35
`
`g/m’ was reached with no antiproliferative toxicity
`being observed [65]. However, plasma deoxyuridine
`(dUrd) was monitored and found to rise during the
`infusion (result of the elevation in dUMP that occurs
`when TS is inhibited) but normalising once the infusion
`was stopped. The rise was not dose-related above 900
`mg/m’. Antitumour studies in vitro and in vivo sug-
`gested that >24 h exposure was required to induce a
`significant cytotoxic effect [66].
`Five days of chronic dosing (75 mg/m’ 4 hourly)
`gave a median growth delay of 14 days against the Hela
`Bu25Tl(— human cervical cell line implanted in nude
`mice [66], while it was inactive when given over a
`shorter time period. This evidence led to the conduct
`of a phase I study of 5-day continuous i.v. infusion. A
`maximum tolerated dose of 1130 mg/m’/day has been
`established with dose limiting toxicities of myelosup-
`pression and mucositis [67]. The dose recommended
`for phase [1 studies is 1000 mg/mi/day.
`Preclinical studies with AG331 indicated that the
`
`plasma half life was considerably longer in dogs (13 h)
`than in rodents (2-4 h) [68]. This was confirmed as
`-20 h in humans in the phase I study. Eighteen pa-
`tients received AG331 (glucuronate salt) at doses es-
`calated between 12.5 and 225 mg/m2 (10 min i.v. infu-
`
`Lilly Ex. 2073
`Sandoz V. Lilly IPR2016-00318
`
`Lilly Ex. 2073
`Sandoz v. Lilly IPR2016-00318
`
`

`
`sion) with side—effects being observed at >130 mg/m2
`(moderate nausea and vomiting) [68]. At the highest
`dose given some mild flushing was observed which is
`believed to be due to histamine release. A similar effect
`
`was observed prec1inically.A phase I study involving a
`5-day continuous infusion has now been reported (25-
`800 mg/mz/day) [69]. Elevated liver function tests
`were evident by day 3 but were not dose-limiting. Phar-
`macokinetics suggested saturable clearance. Phase II
`studies are planned.
`
`BW1843U89, a benzoquinazoline in clinical study
`
`The Wellcome Research Laboratories have developed
`this benzoquinazoline TS inhibitor for clinical study.
`Their drug development programme originally concen-
`trated on the synthesis of lipophilic benzoquinazolines
`[70] lacking the para-aminobenzoyl glutamate chain of
`the natural
`folates, or certain antifolates such as
`CB3717 or ZD1694. Poor water-solubility and low ac-
`tivity of such analogues led to the re-introduction of
`this side chain, which in the case of BW1843U89
`(Fig. 2h), is an isoindolinone modified glutamate. This
`compound is the most potent TS inhibitor described i