Introduction and background
The discovery and development of innovative anticancer drugs in the modern era requires exploitation of biochemical and molecular differences between cancer cells and their normal counterparts.1,2 Following on from the earlier era of cytotoxic agents, there have been remarkable advances in discovering and developing targeted agents that exploit the molecular dependencies and vulnerabilities of cancer cells.3 Particularly rapid progress has been made in elucidating the genetic differences between normal and tumour cells, and the identification of drug targets is accelerated by high-throughput genome sequencing and other omics profiling campaigns, along with functional ribonucleic acid interface (RNAi) screens.4
I have described the new molecular targeting approach as ‘drugging the cancer genome’ (or kinome).5–8 Figure 1 illustrates how the identification of cancer genes leads to the validation of new molecular targets for drug discovery, in parallel with the equally important identification of pharmacodynamic biomarkers to demonstrate target engagement together with predictive biomarkers for patient selection. These are brought together in personalized, precision diagnosis and treatment.9–11
To support the biomarker-driven approach to drug discovery and development I conceptualized a framework that I named the ‘Pharmacologic Audit Trail’ – now used widely to enable rational decision-making in preclinical discovery and clinical development. It is based on mandatory real-time measurement of pharmacokinetic exposure and pharmacodynamic biomarker modulation, combined with the application of predictive biomarkers enabling selection of responsive models or patients.11–16 The basic principle is to ensure the necessary level of exposure and to achieve the required degree of target modulation in the correctly selected tumour model or cancer patient. Both the targeted drugs and the companion biomarkers are essential for success in the discovery and development of personalized cancer medicines and the integration of these has been a critically important theme of my research.
Early research on hypoxia-targeted drugs
Early in my career I focused on targeting tumour hypoxia, representing a major physiological difference between solid cancers and normal tissues. This hypoxia results from tumours outgrowing the insufficient vasculature. We made important contributions towards the discovery and development of several hypoxia-targeted drugs (e.g. misonidazole, desmethylmisonidazole, benznidazole, pimonidazole, etanidazole, tirapazamine and EO9) used either as radiosensitizers or as bioreductive antitumour agents.17,18 Key contributions from our work were (1) the rational refinement of pharmacokinetic behaviour to give desired tumour and neural tissue exposure, maximizing the therapeutic window,19 and (2) the elucidation and optimization of molecular mechanisms of action, especially the discovery and exploitation of enzymes involved in bioreductive drug activation.20,21
Early research on signal transduction inhibitors
In the 1990s I initiated research to evaluate oncogenic molecular signal transduction pathways as targets for new cancer drugs, focusing on inhibitors of phospholipid signalling24 and the epidermal growth factor receptor (EGFR) tyrosine kinase.25,26
Following this, at AstraZeneca I was responsible for the biology of the EGFR drug discovery project that resulted in the clinical candidate gefitinib (Iressa®, AstraZeneca, London, UK). This was approved for the treatment of non-small cell lung cancers and is especially useful in patients with activating EGFR mutations.27 Gefitinib also showed clinical activity in squamous cell head and neck cancers, in which overexpression of wild-type EGFR is common and characteristic of aggressive disease. We identified determinants of response to gefitinib in this disease setting, including the expression of EGFR and HER2.28 Later we discovered that, in addition to inhibiting EGFR catalytic activity, the related EGFR inhibitor erlotinib (Tarceva®, Roche Ltd, Basel, Switzerland) (and indeed other protein kinase inhibitors) unexpectedly caused chaperone deprivation of the target kinase, leading to proteosomal degradation.29 Our surprising discovery has important implications for future drug design as well as for the optimization of dose and scheduling of kinase inhibitors, including use in combination with HSP90 inhibitors to maximize chaperone deprivation and therapeutic effect (see ‘Drugging the HSP90 cancer chaperone’). Such close integration of basic studies of cancer disease biology and mechanisms of drug action, alongside highly focused drug discovery and development projects, has been a key feature of my work. Continuing studies of the molecular mechanism of drug action are essential, even after approval, as these can reveal unexpected and important findings.
Drugging the HSP90 cancer chaperone
The HSP90 molecular chaperone regulates folding, activity and stability of many so-called client proteins.30–34 In the late 1990s, we and a small number of others recognized that a major attraction of inhibiting HSP90 was the potential to simultaneously block the effects of multiple oncogenic client proteins, including many cancer-causing kinases and transcription factors.34–36 This idea was initially received sceptically by the research community, and pharmaceutical companies were reluctant to work on HSP90 as a drug target because of concerns about high risk, an unprecedented target class and the potential for toxic side-effects. Early risk-taking work by others and ourselves was therefore essential to move this field forward.
Progress on understanding the biology and therapeutic potential of HSP90 was greatly facilitated by the use of natural product inhibitors such as geldanamycin and radicicol as chemical tools – for which we published ‘fitness factors’.37 In an early paper bridging our work on bioreductive quinone prodrugs to our emerging interest in HSP90, we showed that geldanamycin-induced cytotoxicity towards human colon cancer cells did not involve the quinone reductase NQO1 (DT-diaphorase) or c-Src.38 However, we discovered that NQO1 expression and activity was important for the antitumour activity of what became the first clinical candidate HSP90 inhibitor, 17-allyamino-17-demethoxygeldanamycin, also known as 17-AAG or tanespimycin.39 We demonstrated some of the earliest evidence of the antitumour activity of 17-AAG in human tumour xenograft models and revealed that high expression of NQO1, which metabolizes 17-AAG to the more active hydroquinone HSP90 inhibitor, was associated with strong antitumour activity, whereas reduced expression or the presence of a polymorphic variant with low reductase activity and poor stability was responsible for relative intrinsic resistance to 17-AAG.39 Later we showed that loss of NQO1 expression or the expression of the polymorphic variant was responsible for acquired resistance to 17-AAG.40
We demonstrated that depletion of oncogenic HSP90 client proteins by 17-AAG causes blockade of signal transduction through the RAS-RAF-MAP kinase and PI3K-AKT pathways, leading to inhibition of cell cycle progression, cytostasis and apoptosis.41 We also demonstrated early evidence of antiangiogenic properties, which contribute to therapeutic effectiveness.42
We were among the very first groups to use cDNA gene expression microarray profiling to study – unbiased and at scale – the mechanism of action of anticancer agents; we showed that the expression of HSP90 client proteins at the mRNA level was not affected by 17-AAG in colorectal cancer cells whereas HSP70, HSP90β, keratin 8, keratin 18 and caveolin were transcriptionally deregulated.43 These studies informed on global mechanism of action and provided the initial basis for our mechanistic pharmacodynamic biomarker strategy for HSP90 inhibitors, as did our subsequent parallel global mRNA profiling and proteomic analysis.44 We validated a pharmacodynamic biomarker signature of combined client protein depletion and increased heat shock protein expression in human tumour xenograft models and established a pharmacokinetic–pharmacodynamic relationship, which was then used by us and others to support clinical trials with 17-AAG as well as subsequent clinical HSP90 inhibitors.45 We established the required levels of drug exposure and target engagement, forming the basis for successfully using our ‘Pharmacologic Audit Trail’ concept in the clinical development of HSP90 inhibitors and later in preclinical discovery.
We carried out a detailed investigator-initiated, hypothesis-driven, biomarker-led, first-in-human clinical trial of 17-AAG in cancer patients.46 We used our protein biomarker signature to provide proof-of-mechanism and to demonstrate target inhibition in peripheral white blood cells and tumour tissue at well-tolerated doses. Importantly, we also observed prolonged stable disease in two patients with drug-resistant melanoma that we linked to BRAF and NRAS mutations.47
Following up our clinical observations in melanoma, we discovered that BRAF forms that are activated by upstream NRAS mutation or oncogenic BRAF mutations – including the common V600E BRAF mutant – are hypersensitive to HSP90 inhibition.48 Our findings proved to be an early example of what is now a general principle: that mutated oncoproteins are much more dependent on HSP90 than wild-type proteins, providing an additional important rationale for a selective anticancer effect of HSP90 inhibitors. Another highly sensitive HSP90 client oncoprotein is HER2. Depletion of amplified HER2 explained the promising clinical responses observed by others using rigorous response evaluation criteria in solid tumours (RECIST) in trastuzumab (Herceptin®, Roche Ltd, Basel, Switzerland)-refractory HER2-positive breast cancer. These clinical findings agreed with our demonstration of a high level of sensitivity to 17-AAG in a transgenic mouse model of mammary cancer driven by HER2/NEU, as determined by tumour regression and magnetic resonance spectroscopy.49
Despite the promising clinical results, we recognized that 17-AAG had significant limitations as a drug. We reasoned that the benzoquinone moiety of 17-AAG was a source of heterogeneous responsiveness owing to variable NQO1 activity – which, as mentioned, we discovered also provided a basis for acquired resistance in cancer cells40 – and, in addition, that quinone metabolism was a likely mechanism of liver toxicity that we and others showed was dose-limiting with 17-AAG.46 Other liabilities for 17-AAG were its role as substrate for MDR1/P-glycoprotein39 and its limited solubility, leading to formulation challenges.
Recognizing these liabilities but noting the promising molecular and clinical activity that we and others showed with 17-AAG, including responses by rigorous RECIST criteria in HER2-positive, trastuzumab-refractory breast cancer, we established at The Institute of Cancer Research (ICR, London, UK) a new drug discovery project aimed at identifying high-quality, drug-like, small-molecule inhibitors of HSP90. Initially, we took two innovative screening approaches to find chemical starting points. First, we established a high-throughput screening assay for the HSP90 ATPase activity using the yeast HSP90 equivalent, which is closely related to the human enzyme.50 We also built a cell-based screening assay with a readout of HSP90 inhibition based on our biomarker signature.51 We discovered the pyrazole resorcinol compound CCT018159 in both screening formats and showed that this promising chemical starting point was active in tumour cells by the desired HSP90 mechanism, building confidence in the chemical matter.52 Furthermore, we showed that CCT018159 inhibited the human HSP90 ATPase activity, facilitated by our discovery of the human HSP90-activating protein AHA1.53
Importantly, we rapidly solved the X-ray co-crystal structure of CCT018159 with HSP90, enabling us to take a structure-based design approach during the hit-to-lead and lead optimization phases of our drug discovery project.52,54 Together with the early structure–activity relationships that we defined, our structure-based design approach allowed us to make considerable progress towards prototype drugs in the original drug discovery work at the ICR.54
Next, in collaboration with Vernalis, we carried out extensive structure-based iterative lead optimization.55–58 We demonstrated promising molecular and cellular properties – including mechanism-based depletion of multiple client proteins and induction of the heat shock response – with the advanced, highly potent lead compounds of our resorcinylic pyrazole/isoxazole amide series (Figure 2).59 Both our use of high-throughput crystallographic technology and the application of our biomarker signature were crucial during this phase of the project, as was the development of a highly sensitive fluorescence polarization assay allowing quantitation of the potency of our enhanced HSP90 inhibitors.60 One of our advanced leads, VER-50589, showed accumulation in human tumour xenografts – to levels exceeding the GI50 for inhibition of cancer cell proliferation for > 24 hours – resulting in pharmacodynamic target engagement biomarker modulation and significant tumour growth inhibition, thereby confirming the promising therapeutic potential of our compounds.59
Completing lead optimization required us to carry out innovative multiparameter refinement of in vivo pharmacokinetic, pharmacodynamic, efficacy and toleration properties.57 Here we again employed our ‘Pharmacological Audit Trail’ concept. In addition, we used cassette dosing technology to help optimize pharmacokinetic behaviour.61–63 This cassette dosing, combined innovatively with assessment of tumour tissue compound concentrations together with mechanistic biomarker evaluation, was crucial in the successful discovery of the preclinical development candidate known initially as VER-52296, then as NVP-AUY922 and now as AUY922.57,64
Our detailed biological characterization of AUY922 showed inhibition at the HSP90 N-terminal ATP site with a dissociation constant (Kd) of 1–2 nM and potent inhibition of cancer cell proliferation with GI50 values in the range 2–40 nM.64 We demonstrated therapeutic activity with AUY922 in a range of human tumour xenograft models, particularly a HER2-positive breast cancer, consistent with pharmacodynamic biomarker modulation. We also showed that AUY922 inhibits tumour cell chemotaxis and invasion in vitro and reduces metastatic tumour growth in in vivo animal models.64 Furthermore, we demonstrated that AUY922 potently inhibited proliferation, chemomigration and tubular differentiation of human endothelial cells, and showed that antiangiogenic activity was associated with reduced microvessel density in human tumour xenografts.64 Importantly, we demonstrated that AUY922 exhibits high selectivity for HSP90 isoforms compared with other ATPases and kinases; in addition, we showed that AUY922 was remarkably clean in general pharmacology screening panels and we demonstrated that the drug was well tolerated at therapeutically active doses in mice.64,65
Following our discovery of AUY922, it was licensed to Novartis, which has overseen its clinical development as an intravenous agent. The phase I study was completed, with strong involvement of the ICR and our affiliated partner, The Royal Marsden Hospital (London, UK), and our validated biomarkers were successfully used.66 A total of 101 patients were enrolled and dose-limiting toxicities included diarrhoea, asthenia/fatigue, anorexia, atrial flutter and visual symptoms. At 70 mg/m2, AUY922 plasma concentrations were consistent with those that we showed were required for efficacy in human tumour xenograft models.64 Induction of HSP70 in peripheral blood mononuclear cells and depletion of client proteins in tumour tissue was demonstrated – again as seen in our preclinical studies – along with reduction of 18F-FDG glucose uptake by position emission tomography. All of these findings are consistent with the ‘Pharmacological Audit Trail’ we established preclinically, demonstrating the value of this approach. The recommended phase II dose was 70 mg/m2 and the conclusion was that AUY922 shows acceptable tolerability, with single agent and combination studies being initiated in HER2-positive breast cancer, gastric cancer and non-small cell lung cancer.66
The AUY922 candidate has shown striking clinical activity in phase II clinical trials in patients with HER2-positive breast cancer who have become refractory to the standard-of-care agent trastuzumab, and this is consistent with our experience in preclinical models.64 The partial response rate was 23% and the combined partial response plus stable disease rate was 74% [presented at the American Society for Clinical Oncology (ASCO), 2012]. AUY922 has also shown promising activity as a single agent in non-small cell lung cancer with a response rate of ≈30%, including in patients with EGFR mutant cancers exhibiting acquired resistance to EGFR tyrosine kinase inhibitors, ALK-rearranged tumours and EGFR/KRAS/ALK triple wild-type cases (presented at ASCO, 2012). Novartis has announced that its requirements for proof of concept were met for AUY922.67 The broader therapeutic potential of AUY922 is shown by the fact that, in addition to four clinical trials that have been completed, there are 18 clinical trials ongoing with an estimated enrolment of 845 patients.68
As well as AUY922, our collaborative work at the ICR with Vernalis and Novartis led to the identification of the 2-aminothieno[2,3-d]pyrimidine VER-82576/NVP-BEP800 as an orally active development candidate.69 This work showed the power of combining different hit identification strategies, including fragment-based and in silico approaches, coupled to the expertise and assays that we developed and used successfully to discover AUY922.
Our HSP90 drug discovery research has encouraged many pharmaceutical companies to develop their own agents. Thus, more than 20 novel HSP90 inhibitors have now entered clinical trials70 and more than 15 international pharmaceutical or biotechnology companies have cited our research [analysis by the authors of publication data on HSP90 inhibitors obtained from Web of Science (Thomson Reuters, New York City, NY, USA)].
Beyond discovery and optimization of the chemical series described above, we have also carried out a range of informative chemistry and biology studies internally at the ICR and with collaborators – exploring diverse chemical series and informing on binding mechanism and active chemotypes for drug discovery and chemical tools.71–79
Regarding prediction of tumour response to HSP90 inhibition, preclinical and clinical sensitivity is consistent with therapeutic benefit in cancers for which the driver is an oncogenic protein kinase that is amplified, point-mutated or translocated and is also highly dependent on HSP90, e.g. HER2, BRAF, EGFR and ALK.30 However, additional research is needed to further refine predictive biomarkers,80 and we continue to work on this.81
Very importantly, AUY922 and other HSP90 drugs can overcome resistance to molecularly targeted drugs, because a mutated resistant kinase allele remains sensitive to depletion or because the new kinase driver is HSP90 dependent, for example. Interestingly, with the same experimental procedure used to obtain 17-AAG resistance through loss of NQO1 activity (see Drugging the cancer genome),39,40 we demonstrated that there was no acquisition of resistance to AUY922 in human glioblastoma or melanoma cell lines over a prolonged drug exposure period of up to 12 months.40 This suggests that cancer cells may find it difficult to develop resistance to the combinatorial antioncogenic effects of HSP90 inhibitors such as AUY922. Thus, the ability to overcome or prevent drug resistance may be a particularly important feature of the strategy of targeting the molecular chaperone HSP90.
HSP90 biology and new chaperone pathway targets
A strength of our drug discovery work on HSP90 is the closely integrated research on the biology of this molecular chaperone and its role in cancer cells, both in-house and in collaboration with the structural biology group of Laurence Pearl.82 We discovered and elucidated the important role of the kinase co-chaperone CDC37 in the biology of HSP90 and demonstrated by RNAi the therapeutic potential of inhibiting its function in tumour cells.83,84 We have now challenged the dogma that a direct interaction between CDC37 and HSP90 is required for kinase chaperoning and shown that blocking the interaction between CDC37 and HSP90 would not be a therapeutically effective strategy; thus, alternatives to drug CDC37 are needed.85
Following our discovery of the human HSP90-activating protein AHA1,53 we showed that silencing AHA1 decreases client protein activation and increases the sensitivity of cancer cells to HSP90 inhibition.86
Our global mRNA and proteomic profiling study, mentioned earlier, demonstrated that induction of HSP72 and HSC70 expression following HSP90 inhibition is common.44 As these HSP70 isoforms are likely to be involved in HSP90 function and also exert antiapoptotic roles, we hypothesized that dual knockout of HSP72 and HSC70 could provide a beneficial antitumour effect. Importantly, we then showed that selective combinatorial siRNA knockdown of these two HSP70 isoforms phenocopied HSP90 inhibition in causing client protein depletion but without activating the undesired heat shock response seen with HSP90 inhibitors.87 Furthermore, we demonstrated that dual silencing of HSP72 and HSC70 caused massive, tumour cell-specific apoptosis.87 This discovery has provided a rational basis for ourselves and others to search for HSP72/HSC70 inhibitors, thus opening up an innovative area of research with considerable therapeutic potential; it is recognized that HSP70 isoforms are much less readily druggable than HSP90, but the therapeutic opportunity is exciting.88
We remain very interested in additional ways to drug HSP90 and related stress pathways. One such approach is to block the transcription factor HSF1 – not only important in activating the ‘heat shock’ gene expression programme following cellular stress, but also crucial in supporting the malignant state.33,89–91 This will be technically challenging but the potential therapeutic value is, again, considerable.
Given the attractiveness of the HSPs and HSF1 pathway targets we are currently complementing our hypothesis-driven approach by taking various global RNAi-based screening strategies to identify new therapeutic approaches in this area.
Drug combination studies with HSP90 inhibitors
Although the original rationale for HSP90 inhibitors was that they should exhibit a powerful effect on multiple cancer targets, pathways and hallmark traits30,34–36 – which proved to be correct – it is nevertheless likely that HSP90 inhibitors would be even more effective in combination with other drugs. These could be either cytotoxic agents or molecularly targeted therapeutics.
We showed that HSP90 inhibition can be additive with carboplatin in human ovarian cancer models92 and that HSP90 inhibitors potentiate the effects of paclitaxel in ovarian cancer cell lines with activated AKT.93 We demonstrated that AUY922 causes radiosensitization by abrogation of homologous recombination, resulting in mitotic entry with unresolved DNA damage.94
We demonstrated that the mode of cell death induced following HSP90 inhibition is dependent on the expression of the pro-apoptotic protein BAX,95 and showed that combining HSP90 inhibitors with TNF-related apoptosis-inducing ligand [TRAIL (which binds to the death receptors DR4 and DR5)] results in enhancement of apoptosis in human colorectal cancer cells and tumour xenograft models via multiple effects that cause suppression of survival signalling.96
Supporting our findings on human colorectal cancer cells,87 we have shown that targeting HSP70 isoforms can enhance apoptosis induced by HSP90 inhibition in myeloma cells97 and demonstrated that HSP70 inhibition induces myeloma cell death via the intracellular accumulation of immunoglobulin and the generation of proteotoxic stress.98
Combination of HSP90 inhibitors with ATP-competitive inhibitors of oncogenic protein kinase clients of HSP90 is attractive by delivering a double-hit on the target kinase through attacking both its stability and also its catalytic activity at the same time. Furthermore, such combinations can exploit our surprising discovery that ATP-competitive kinase drugs block recruitment to the HSP90–CDC37 chaperone complex29 to maximize chaperone deprivation.
Drugging the PI3 kinome
In contrast to the targeting of HSP90, which is an example of exploiting non-oncogene addiction, drugging PI3 kinase exemplifies therapeutic attack on oncogene addiction. Dependence on the mutated oncogene PIK3CA is very common in human cancer and the PI3 kinase pathway is also frequently activated by loss of the PTEN tumour suppressor gene, and other genetic and epigenetic events.
We carried out high-risk, early-stage research on the discovery of PI3 kinase inhibitors for use as chemical tools and clinical drugs. As with HSP90, we entered this area when most pharmaceutical researchers considered it too risky, especially because lipid kinases had not been drugged before and because there were technical screening challenges and also concern about potential toxicity, particularly through inhibition of insulin signalling. We recognized the pathogenic role of class I PI3 kinases and the therapeutic potential of class I PI3 kinase inhibitors in cancer.5–7,99–104 Initially collaborating with Mike Waterfield (ICR, University College London, London, UK) in 1999, Peter Parker (Cancer Research UK, London Research Institute, London, UK) in 2007 and the Yamanouchi Pharmaceutical Company Ltd (now Astellas Pharma Inc., Tokyo, Japan) we discovered a range of advanced drug-like chemical leads acting on class I PI3 kinases, with or without mammalian target of rapamycin (mTOR), which exhibited promising anticancer properties.105–108 In this collaborative research phase we also discovered a selective inhibitor of the mammalian class III phosphatidylinositol phosphate kinase PIKfyve that blocks PtdIns(3,5)P(2) production and disrupts endomembrane transport and retroviral budding, revealing that the phosphatidylinositol 3,5-bisphosphate pathway is integral to endosome formation, cargo flux and determination of cell morphology.109
Subsequently, we mechanistically elucidated and optimized a range of potent and selective PI3 kinase inhibitors.110–114 Our discovery and detailed characterization of the class I PI3 kinase/mTOR dual inhibitor PI-103 allowed us to pioneer this chemical probe and show proof of concept for pathway modulation and therapeutic activity in human tumour models with deregulated PI3 kinase signalling.112,115 PI-103 subsequently received widespread adoption as a chemical probe. Further structure-based optimization of PI-103 by our ICR team in collaboration with Piramed Pharma – which I founded with Mike Waterfield and Peter Parker and which was acquired by Roche in 2008 – led to our discovery of the pan class I-selective thienopyrimidine PI3 kinase inhibitor initially known as PI-728 and subsequently as GDC-0941,103,110,113,114 the first of such agents to enter the clinic.
Our paper on the biological and pharmacological properties of GDC-0941113 was selected by the American Association of Cancer Research’s journal Molecular Cancer Therapeutics.113,114 as one of 16 ‘outstanding’ studies during the decade 2001–11 judged to ‘have a lasting impact on research and patient care’116
In parallel with our drug discovery and development work on PI3 kinase inhibitors, we identified important pharmacodynamic and predictive biomarkers enabling the ‘Pharmacological Audit Trail’. For pathway modulation we measured and quantified phosphorylation of downstream protein substrates, defining the pharmacokinetic exposures and pharmacodynamic responses required for therapeutic activity.112–114 Figure 3 shows the excellent biomarker modulation and antitumour activity in a human tumour xenograft model with PI3 kinase pathway addiction. Our early research with PI-103 and GDC-0941 identified tumours with KRAS mutations as resistant whereas those exhibiting PIK3CA mutations and PTEN loss were sensitive.112,113 These predictive biomarkers have been fully confirmed in subsequent studies by others and have important implications for patient selection. Work is ongoing to refine even more precise predictive markers of sensitivity.80,99,117
GDC-0941 and related compounds were subsequently licensed to Genentech and GDC-0941 has completed promising phase I and II clinical trials68 demonstrating biomarker modulation and clinical activity (presentations at ASCO, 2011 and 2012). We are playing an important role in these trials at the ICR and The Royal Marsden Hospital.
Subsequently, GDC-0980 was discovered by Piramed and Genentech based on our previous research at the ICR with Piramed, and this drug has also progressed to phase I/II clinical trials.68 GDC-0980 is quite closely related chemically to GDC-0941 and PI-103. Whereas GDC-0941 is selective for class I PI3 kinases and does not significantly inhibit mTOR, GDC-0980 is more similar to PI-103 in potently inhibiting both class I PI3 kinases and mTOR.113 Each profile is biologically and therapeutically attractive and may offer different therapeutic utility.117
Both GDC-0941 and GDC-0980 continue to progress through phase I/II clinical trials.68 GDC-0941 is well tolerated at doses leading to PI3K target modulation, and clinical responses have been observed in breast and ovarian cancer, gastrointestinal stromal tumours, multiple myeloma and melanoma (presented at ASCO, 2010, and ECCO, 2011). There is evidence that GDC-0941 and the MEK inhibitor GDC-0973 given in combination show antitumour activity in melanoma and prostate, pancreatic and lung cancer. GDC-0980 is showing promising clinical activity in phase II trials.68 At the time of writing, one clinical trial with more than 100 patients has almost been completed with activity seen in mesothelioma, a disease of high unmet medical need.68 Eight clinical trials are ongoing worldwide with an estimated enrolment of over 900 patients.
Clinical trials with GDC-0941 and GDC-0980 include both single-agent and combination studies involving cytotoxic and molecularly targeted agents. We found that PI3 kinase inhibitors predominantly exhibit a cytostatic rather than apoptotic effect, with G1 arrest being a common phenotype,112 thus supporting combination strategies. We showed that combining TRAIL with PI3 kinase increases apoptosis in colorectal cancer cells by suppressing survival signals.96 In patients with advanced ovarian cancer, we identified an association between levels of p70S6K and response to subsequent chemotherapy, suggesting potential for PI3 kinase inhibitor–cytotoxic combinations in patients with high levels of this biomarker.118
In addition to progression of clinical candidates arising from our own research, we have had a worldwide impact by stimulating many other PI3 kinase drug discovery programmes in the pharmaceutical industry.117 Thus, over 40 different commercial companies have cited our work. PI3 kinase inhibitors derived from our research are also being developed by Roche for immune inflammation disorders such as rheumatoid arthritis.119
Concluding remarks and future prospects
Over the last two decades we have seen considerable progress in understanding the genetics and biology of cancer and in drugging the cancer genome.4,8,10 Our group has contributed to this progress by identifying numerous innovative drug candidates, of which HSP90 molecular chaperone inhibitors and PI3 kinase inhibitors are leading examples.
On the other hand, although many breakthrough agents have been discovered, therapeutic discovery lags behind genomic advances in drugging the cancer genome.4,8,10,120 Our integrated, multidisciplinary, computational approach has shown that only 5% of cancer genes are currently targeted, meaning that most cancer genes do not yet have small-molecule or antibody drugs acting on them.4,121 But where should we focus our drug discovery research among the increasingly large number of potential targets arising from large-scale genomics and other screening approaches?4
To accelerate drugging the cancer genome, we developed and made freely available [> 60 000 users by Google Analytics (Google, Mountain View, CA, USA)] a sophisticated computational target assessment tool based on our multidimensional canSAR knowledge base.121 This allows us and other users to prioritize the large number of potential targets for detailed biological validation. Target prioritization is built not only on compelling biological data and clinical information on the pathogenic involvement of targets, but also on technical feasibility and, in particular, druggability assessment. Using this integrated multidisciplinary approach we have identified 46 cancer targets that are biologically validated and readily druggable but currently neglected – presenting immediate opportunities for innovative drug discovery.121 Completing the druggable cancer genome requires expansion of the envelope of technical druggability, and we are now making progress on this, for example in targeting protein–protein interactions.
There is no doubt that the biggest clinical challenge that we now face in oncology is drug resistance due to (1) biochemical feedback loops, (2) cellular plasticity and (3) genetic heterogeneity and clonal evolution, especially under the Darwinian selective pressure of therapy.9,10,122 The current approach to overcome drug resistance is using drug combinations, ideally adapted to the changing molecular and genetic characteristics of the cancer.10 Construction of the full repertoire of such cancer genome-based cocktails requires the drugged cancer genome to be greatly extended and eventually completed, including such currently very difficult targets such as RAS and MYC.
Looking further ahead, rational combination strategies will ideally be devised that can drive tumour subclones down evolutionary blind alleys from which they cannot escape. As we now recognize that tumour genes conspire together in complex and malign networks rather than simple linear textbook pathways, a key objective is defining and drugging the optimal oncogenic network targets.121,122 Such network-based targets may not necessarily themselves be oncogenes. HSP90, as discussed here, may represent the first of these as it acts as a hub for so many oncogenic factors and because HSP90 inhibitors are able to overcome and potentially prevent, or at least slow down, the development of resistance. Clearly, the computational sciences are playing an increasing role in understanding and drugging cancer genomes and cancer gene networks.121,122
Finally, it is important to recognize that in dealing with tumour heterogeneity, clonal evolution and drug resistance we will probably need to do more than use targeted drugs alongside traditional chemotherapy, surgery and radiation therapy – we will almost certainly also need to administer them in combination with the very promising new wave of immunological therapies.4 Such multidisciplinary approaches will be essential to ensure that future personalized, precision therapies that exploit the cancer genome can be maximally effective in defeating cancer.
Conflict of interest
Paul Workman is an employee of The Institute of Cancer Research, which has a commercial interest in the discovery and development of anticancer drugs, including HSP90 and kinase inhibitors and operates a ‘Rewards to Discovers’ scheme. Paul Workman is a former employee of AstraZeneca and declares commercial interactions with Yamanouchi (now Astellas), Piramed Pharma (acquired by Roche), Genentech, Vernalis, Novartis, Chroma Therapeutics, Astex Pharmaceuticals, AstraZeneca, Cyclacel, Onyx Pharmaceuticals, Merck Serono, Sareum, Janssen, Wilex, and Nextech Ventures.