Publications

A Phase II study to screen for anti-melanoma activity of the heat shock protein 90 (HSP90) inhibitor, 17-AAG (17-allylamino-17-demethoxygeldanamycin) was performed. The primary endpoint was the rate of disease stabilisation in patients with progressive, metastatic melanoma treated with 17-AAG. Secondary endpoints were to determine: the toxicity of 17-AAG, the duration of response(s), median survival and further study the pharmacokinetics and pharmacodynamics of 17-AAG.

This study defined the risk of serious toxicity in phase I trials of molecularly targeted agents (MTA).

This clinical trial investigated the safety, tolerability, pharmacokinetic (PK), and pharmacodynamic (PD) profile of CHR-3996, a selective class I histone deacetylase inhibitor. Patients and Methods: CHR-3996 was administered orally once a day. This phase I trial used a 3+3 dose-escalation design. PK profiles were analyzed by liquid chromatography-tandem mass spectroscopic methods and PD studies were conducted using ELISA studying histone H3 acetylation in peripheral blood mononuclear cells.

Evidence that the phosphoinositide 3-kinase (PI3K) pathway is deregulated in ovarian cancer is largely based on the analysis of surgical specimens sampled at diagnosis and may not reflect the biology of advanced ovarian cancer. We aimed to investigate PI3K signaling in cancer cells isolated from patients with advanced ovarian cancer. Ascites samples were analyzed from 88 patients, of whom 61 received further treatment. Cancer cells were immunomagnetically separated from ascites, and the signaling output of the PI3K pathway was studied by quantifying p-AKT, p-p70S6K, and p-GSK3β by ELISA. Relevant oncogenes, such as PIK3CA and AKT, were sequenced by PCR-amplified mass spectroscopy detection methods. In addition, PIK3CA and AKT2 amplifications and PTEN deletions were analyzed by FISH. p-p70S6K levels were significantly higher in cells from 37 of 61 patients who did not respond to subsequent chemotherapy (0.7184 vs. 0.3496; P = 0.0100), and this difference was greater in patients who had not received previous chemotherapy. PIK3CA and AKT mutations were present in 5% and 0% of samples, respectively. Amplification of PIK3CA and AKT2 and deletion of PTEN was seen in 10%, 10%, and 27% of samples, respectively. Mutations of PIK3CA and amplification of PIK3CA/AKT2 or deletion of PTEN did not correlate with levels of p-AKT, p-p70S6K, and p-GSK3β. In patients with advanced ovarian cancer, there is an association between levels of p-p70S6K and response to subsequent chemotherapy. There is no clear evidence that this is driven specifically by PIK3CA or AKT mutations or by amplifications or deletion of PTEN.

The importance of ERBB2/NEU/HER2 in the response of breast tumours to the heat shock protein 90 (HSP90) inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG; tanespimycin) has been demonstrated in the clinic. ERBB2 is an oncoprotein client that is highly dependent on HSP90. This and other oncogenic client proteins (e.g. B-RAF, C-RAF, ALK and CDK4) are depleted by 17-AAG in both animal tumours and patients. Here we investigate by Magnetic Resonance Spectroscopy (MRS) the metabolic response of 17-AAG in spontaneous, NEU/HER2 driven mammary tumours in transgenic MMTV-NEU-NT mice and in cells isolated and cultured from these tumours.

This phase I study assessed the maximum tolerated dose (MTD), dose-limiting toxicities (DLT), safety, pharmacokinetics, pharmacodynamics, and clinical activity of the first-in-class dual MEK/RAF inhibitor, RO5126766.

Over the past decade, whole genome sequencing and other 'omics' technologies have defined pathogenic driver mutations to which tumor cells are addicted. Such addictions, synthetic lethalities and other tumor vulnerabilities have yielded novel targets for a new generation of cancer drugs to treat discrete, genetically defined patient subgroups. This personalized cancer medicine strategy could eventually replace the conventional one-size-fits-all cytotoxic chemotherapy approach. However, the extraordinary intratumor genetic heterogeneity in cancers revealed by deep sequencing explains why de novo and acquired resistance arise with molecularly targeted drugs and cytotoxic chemotherapy, limiting their utility. One solution to the enduring challenge of polygenic cancer drug resistance is rational combinatorial targeted therapy.

This study aimed to evaluate any correlations between baseline creatinine clearance and the development of grade 3/4 toxicities during treatment within oncology phase I trials of molecularly targeted agents where entry criteria mandate a serum creatinine of ≤ 1.5 × the upper limit of normal.

This study assessed the safety, tolerability, pharmacokinetics and pharmacodynamics of the first-in-class dual mammalian target of rapamycin complex (mTORC)1/mTORC2 inhibitor, AZD8055.

Anti-cancer clinical drug development is currently costly and slow with a high attrition rate. There is thus an urgent and unmet need to integrate pharmacodynamic biomarkers into early phase clinical trials in the framework provided by the "pharmacologic audit trail" in order to overcome this challenge. This review discusses the rationale, advantages and disadvantages, as well as the practical considerations of various tissue-based approaches to perform pharmacodynamic studies in early phase oncology clinical trials using case histories of molecular targeting agents such as PI3K, m-TOR, HSP90, HDAC and PARP inhibitors. These approaches include the use of normal "surrogate" tissues such as peripheral blood mononuclear cells, platelet-rich plasma, plucked hair follicles, skin biopsies, plasma-based endocrine assays, proteomics, metabolomics and circulating endothelial cells. In addition, the review discusses the use of neoplastic tissues including tumor biopsies, circulating tumor DNA and tumor cells and metabolomic approaches. The utilization of these tissues and technology platforms to study biomarkers will help accelerate the development of molecularly targeted agents for the treatment of cancer.

Background:We investigated the association between skin rash and plasma creatine kinase (CK) levels in oncology phase I trials.Methods:We analysed data from 295 patients treated at our institution within 25 phase I trials which included CK measurements in the protocol. Trials involved drugs targeting EGFR/HER2, m-TOR, VEGFR, SRC/ABL, aurora kinase, BRAF/MEK, PARP, CDK, A5B1 integrin, as well as oncolytic viruses and vascular disrupting agents.Results:Creatine kinase measurements were available for 278 patients. The highest levels of plasma CK during the trial were seen among patients with Grade (G) 2/3 rash (median 249 U l(-1)) compared with G1 (median 81 U l(-1)) and no rash (median 55 U l(-1)) (P<0.001). There was a significant reduction in CK after the rash resolved (mean 264.2 vs 100.1; P=0.012) in 25 patients, where serial CK values were available. In vitro exposure of human keratinocytes to EGFR, MEK and a PI3Kinase/m-TOR inhibitor led to the increased expression of CK-brain and not CK-muscle or mitochondrial-CK.Conclusion:Plasma CK elevation is associated with development of skin rash caused by novel anticancer agents. This should be studied further to characterise different isoforms as this will change the way we report adverse events in oncology phase I clinical trials.

Tumor genomic instability and selective treatment pressures result in clonal disease evolution; molecular stratification for molecularly targeted drug administration requires repeated access to tumor DNA. We hypothesized that circulating plasma DNA (cpDNA) in advanced cancer patients is largely derived from tumor, has prognostic utility, and can be utilized for multiplex tumor mutation sequencing when repeat biopsy is not feasible. We utilized the Sequenom MassArray System and OncoCarta panel for somatic mutation profiling. Matched samples, acquired from the same patient but at different time points were evaluated; these comprised formalin-fixed paraffin-embedded (FFPE) archival tumor tissue (primary and/or metastatic) and cpDNA. The feasibility, sensitivity, and specificity of this high-throughput, multiplex mutation detection approach was tested utilizing specimens acquired from 105 patients with solid tumors referred for participation in Phase I trials of molecularly targeted drugs. The median cpDNA concentration was 17 ng/ml (range: 0.5-1600); this was 3-fold higher than in healthy volunteers. Moreover, higher cpDNA concentrations associated with worse overall survival; there was an overall survival (OS) hazard ratio of 2.4 (95% CI 1.4, 4.2) for each 10-fold increase in cpDNA concentration and in multivariate analyses, cpDNA concentration, albumin, and performance status remained independent predictors of OS. These data suggest that plasma DNA in these cancer patients is largely derived from tumor. We also observed high detection concordance for critical 'hot-spot' mutations (KRAS, BRAF, PIK3CA) in matched cpDNA and archival tumor tissue, and important differences between archival tumor and cpDNA. This multiplex sequencing assay can be utilized to detect somatic mutations from plasma in advanced cancer patients, when safe repeat tumor biopsy is not feasible and genomic analysis of archival tumor is deemed insufficient. Overall, circulating nucleic acid biomarker studies have clinically important multi-purpose utility in advanced cancer patients and further studies to pursue their incorporation into the standard of care are warranted.

A phase I study to define toxicity and recommend a phase II dose of the HSP90 inhibitor alvespimycin (17-DMAG; 17-dimethylaminoethylamino-17-demethoxygeldanamycin). Secondary endpoints included evaluation of pharmacokinetic profile, tumor response, and definition of a biologically effective dose (BED).

Brain metastases (BM) are treated with surgical resection when feasible. Unfortunately this occurs only in a small subset of patients. The optimal treatment for patients with intracranial metastases non amenable to surgical resection has not been identified. Radiotherapy improves symptom control and survival but long-term local control has been poor. Conventional chemotherapies have generally produced disappointing results possibly due to their limited ability to penetrate the blood brain barrier. Therefore, newer treatments are required for patients with unresectable BM. Targeted therapies such as bevacizumab, erlotinib, gefitinib, sunitinib and sorafenib, are all licensed and have demonstrated improved survival in patients with metastatic disease. In this review we will present current data on targeted therapies that have been approved for the treatment of malignant tumours and we discuss the evidence of their use in patients with BM.

First-line treatment options utilizing chemotherapy and cytokine-based treatments for patients with metastatic melanoma (MM) are unsatisfactory. We analyzed the clinical outcomes of patients with MM treated in phase I trials of novel agents. We hypothesized that patients included in phase I clinical trials did not have worse outcomes than with the chemotherapy and cytokine-based first-line treatment.

Anticancer drug development remains slow, costly and inefficient. One way of addressing this might be the use of predictive biomarkers to select patients for Phase I/II trials. Such biomarkers, which predict response to molecular-targeted agents, have the potential to enrich these trials with patients more likely to benefit. Doing so could maximize the efficiency of anticancer drug development by facilitating earlier clinical qualification of predictive biomarkers and generating valuable information on cancer biology. In this review, we suggest a new model of early clinical trial design, which incorporates patient selection through predictive molecular biomarkers for selected targeted agents.

In part A, the aim was to define the maximum tolerated dose (MTD) of the hydrogen sulfate (Hyd-Sulfate) oral capsule formulation of the mitogen-activated protein kinase kinase inhibitor AZD6244 (ARRY-142886). In part B, the aim was to compare the pharmacokinetic profile of the new Hyd-Sulfate capsule with the initial AZD6244 free-base suspension and further characterize the pharmacodynamic profile and efficacy of the new formulation.

Unplanned hospital admissions (UHAs) in the context of oncology Phase I trials are important, yet rarely reported.

There is clinical evidence to suggest that tumour necrosis factor-α (TNF-α) may be a therapeutic target in renal cell carcinoma (RCC). Multi-targeted kinase inhibitors, such as sorafenib and sunitinib, have become standard of care in advanced RCC. The anti-TNF-α monoclonal antibody infliximab and sorafenib have differing cellular mechanisms of action. We conducted a phase I/II trial to determine the safety and efficacy of infliximab in combination with sorafenib in patients with advanced RCC.

Heat shock protein 90 (HSP90) is a ubiquitously expressed chaperone that is involved in the posttranslational folding and stability of proteins. Inhibition at the NH(2)-terminal ATP-binding site leads to the degradation of client proteins by the ubiquitin proteasome pathway. Inhibition of HSP90 leads to the degradation of known oncogenes, such as ERB-B2, BRAF, and BCR-ABL, leading to the combinatorial blockade of multiple signal transduction pathways, such as the RAS-RAF-mitogen-activated protein/extracellular signal-regulated kinase kinase-extracellular signal-regulated kinase and phosphatidylinositol 3-kinase pathways. Multiple structurally diverse HSP90 inhibitors are undergoing early clinical evaluation. The clinical focus of these drugs should be solid tumors, such as breast, prostate, and lung cancers, along with malignant melanoma, in addition to hematologic malignancies, such as chronic myeloid leukemia and multiple myeloma. HSP90 inhibitors can be used as single agents or in combination with other targeted treatments or conventional forms of treatment such as chemotherapy and radiotherapy. Clinical trials evaluating efficacy of these agents should include innovative designs to capture cytostasis evidenced by clinical nonprogression and enrichment of patient populations by molecular characterization. The results of clinical trials evaluating the efficacy of drugs targeting this exciting target are awaited.

Compared with conventional chemotherapy, rationally designed molecularly targeted agents may be more likely to have antitumor activity in selected tumor subgroups driven by the oncogenic signals targeted by these compounds and a different side-effect profile. The use of biomarkers in the early clinical trials of these new anticancer agents has the potential to increase study participants' benefit from early clinical trials, accelerate the drug development process, maximize the ability to generate important biological information about human cancer, and decrease the risk of late and costly drug attrition.

Phase-I trials traditionally involve dose-escalation to determine the maximal tolerated dose (MTD). With conventional chemotherapy, efficacy is generally deemed to be dose-dependent, but the same may not be applicable to molecularly targeted agents (MTAs). We analysed consecutive patients included in Phase-I trials at the Royal Marsden Hospital from 5 January 2005 to 6 June 2006. We considered only trials of monotherapy MTAs in which the MTD was defined. Three patient cohorts (A, B, and C) were identified according to the dose received as a percentage of the final trial MTD (0-33%, 34-65%, >66%). Potential efficacy was assessed using the non-progression rate (NPR), that is, complete/partial response or stable disease for at least 3 months by RECIST. A total of 135 patients having progressive disease before enrolment were analysed from 15 eligible trials. Median age was 57 years (20-86); male : female ratio was 1.8 : 1. Cohort A, B, and C included 28 (21%), 22 (16%), and 85 (63%) patients; NPR at 3 and 6 months was 21% and 11% (A), 50% and 27% (B), 31% and 14% (C), respectively, P=0.9. Median duration of non-progression (17 weeks; 95% CI=13-22) was not correlated with the MTD level, P=0.9. Our analysis suggests that the potential for clinical benefit is not confined to patients treated at doses close to the MTD in Phase-I trials of MTAs.

Early clinical trials represent a crucial bridge between preclinical drug discovery and the especially resource-intense randomized phase III trial-the definitive regulatory hurdle for drug approval. High attrition rates and rising costs, when coupled with the extraordinary opportunities opened up by cancer genomics and the promise of personalized medicine call for new approaches in the conduct and design of phase I/II trials. The key challenge lies in increasing the odds for successful and efficient transition of a compound through the drug development pipeline. The incorporation of scientifically and analytically validated biomarkers into rationally designed hypothesis-testing clinical trials offers a promising way forward to achieving this objective. In this article, we provide an overview of biomarkers in early clinical trials, including examples where they have been particularly successful, and the caveats and pitfalls associated with indiscriminate application. We describe the use of pharmacodynamic end points to demonstrate the proof of modulation of target, pathway, and biologic effect, as well as predictive biomarkers for patient selection and trial enrichment. Establishing the pharmacologic audit trail provides a means to assess and manage risk in a drug development program and thus increases the rationality of the decision-making process. Accurate preclinical models are important for pharmacokinetic-pharmacodynamic-efficacy modeling and biomarker validation. The degree of scientific and analytical validation should ensure that biomarkers are fit-for purpose, according to the stage of development and the impact on the trial; specifically they are either exploratory or used to make decisions within the trial. To be maximally useful at an early stage, these must be in place before the commencement of phase I trials. Validation and qualification of biomarkers then continues through clinical development. We highlight the impact of modern technology platforms, such as genomics, proteomics, circulating tumor cells, and minimally invasive functional and molecular imaging, with respect to their potential role in improving the success rate and speed of drug development and in interrogating the consequences of therapeutic intervention and providing a unique insight into human disease biology. With these technologies already having an impact in the clinic today, we predict that further future advances will come from the application of network analysis to clinical trials, leading to individualized systems-based medicine for cancer.

To study the interactions of the heat shock protein 90 (HSP90) inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG) and carboplatin in vitro and in vivo.

Oncogenic BRAF and NRAS mutations are frequent in malignant melanoma. BRAF that is activated by the common V600E and other mutations, as well as by upstream NRAS mutations, has been shown to require the molecular chaperone heat shock protein 90 (HSP90) for stabilization and is depleted by the HSP90 inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG)]. Here, we explore the possible relationship between tumor BRAF and NRAS mutations and clinical response to 17-AAG in six patients with metastatic malignant melanoma who received pharmacologically active doses of 17-AAG as part of a phase I clinical trial. One patient with disease stabilization for 49 months had a (G13D)NRAS mutation and (WT)BRAF. A second patient who had stable disease for 15 months had a (V600E)BRAF mutation and (WT)NRAS. These preliminary results suggest that BRAF and NRAS mutation status should be determined in prospective phase II studies of HSP90 inhibitors in melanoma.

Sixty-three patients received capecitabine at 1000 mg/m2 twice daily every 2 out of 3 weeks as first-line treatment for advanced disease at our institution. Forty-five patients (71%) had previously received adjuvant or neoadjuvant chemotherapy. The median number of capecitabine cycles administered was 5(1-40). Forty-eight patients had measurable disease with response rate (RR) of 29%. The median time to progression (TTP) was 18(2-122) weeks. Seven patients (11%) had TTP of >1 yr, four of whom received more than 10(24-40) cycles of capecitabine. Thirty-seven percent of patients still needed dose reductions. Our retrospective audit is consistent with a previously published study which used a higher starting dose of capecitabine as first-line chemotherapy. For a subgroup of patients, capecitabine can result in a long TTP with minimal toxicity. The benefit of continuing capecitabine beyond a fixed number of cycles should be investigated further. Schedules using even lower doses of capecitabine for longer periods may also be of interest.

The promising antitumor activity of 17-allylamino-17-demethoxygeldanamycin (17AAG) results from inhibition of the molecular chaperone heat shock protein 90 (HSP90) and subsequent degradation of multiple oncogenic client proteins. Gene expression microarray and proteomic analysis were used to profile molecular changes in the A2780 human ovarian cancer cell line treated with 17AAG. Comparison of results with an inactive analogue and an alternative HSP90 inhibitor radicicol indicated that increased expression of HSP72, HSC70, HSP27, HSP47, and HSP90beta at the mRNA level were on-target effects of 17AAG. HSP27 protein levels were increased in tumor biopsies following treatment of patients with 17AAG. A group of MYC-regulated mRNAs was decreased by 17AAG. Of particular interest and novelty were changes in expression of chromatin-associated proteins. Expression of the heterochromatin protein 1 was increased, and expression of the histone acetyltransferase 1 and the histone arginine methyltransferase PRMT5 was decreased by 17AAG. PRMT5 was shown to be a novel HSP90-binding partner and potential client protein. Cellular protein acetylation was reduced by 17AAG, which was shown to have an antagonistic interaction on cell proliferation with the histone deacetylase inhibitor trichostatin A. This mRNA and protein expression analysis has provided new insights into the complex molecular pharmacology of 17AAG and suggested new genes and proteins that may be involved in response to the drug or be potential biomarkers of drug action.

Patients with metastatic breast cancer (MBC) are increasingly offered third line chemotherapy. We have reviewed the response rate (RR), time to progression (TTP) and survival of 149 patients in this setting and have investigated factors that influence their outcome. The RR, TTP and survival were 30%, 4 and 8 months, respectively, and should serve as a benchmark for future studies. Response to previous chemotherapy was the only independent variable predicting RR, TTP and survival, p=0.025, 0.04 and 0.004, respectively. Thirty-two percent of patients did not respond to the first two lines of chemotherapy and had a lower RR and a significantly shorter TTP and survival. In conclusion, third line chemotherapy for MBC is sometimes effective in patients who have responded to previous chemotherapy. Patients who do not respond to the first two lines of chemotherapy should be considered for clinical trials or supportive care.

Chemotherapy induced neutropenia has been shown to be associated with improved treatment outcomes in selected solid tumours. We studied the association of chemotherapy induced neutropenia with treatment related outcomes in small cell lung cancer (SCLC).

Heat shock proteins, including HSP90, contribute to the stabilization of a number of key molecules implicated in proliferative pathways in cancer. Geldanamycin and related compounds are competitive inhibitors of HSP90 at the N-terminal ATP binding site. The geldanamycin derivative 17-allylamino, 17-demethoxygeldanamycin (17-AAG) was the first HSP90 inhibitor to enter clinical development. Here we describe a patient with metastatic malignant melanoma treated with 17-AAG. She enjoyed a period of stable disease for a period of 49 months on treatment. Prolonged disease stabilization may prove to be a clinically meaningful outcome for many molecularly targeted agents and the design of phase II studies of these novel therapies needs to consider the use of progression-free survival as a valid endpoint.

To study the toxicity and pharmacokinetic-pharmacodynamic profile of 17-allylamino, 17- demethoxygeldanamycin (17-AAG) and to recommend a dose for phase II trials.

To establish the pharmacokinetic and pharmacodynamic profile of the heat shock protein 90 (HSP90) inhibitor 17-allylamino, 17-demethoxygeldanamycin (17-AAG) in ovarian cancer xenograft models.

The potential clinical applications of the prototype first-in-class Hsp90 inhibitor 17AAG and other emerging Hsp90 drugs are very exciting. Rigorously planned and executed clinical trials, incorporating measurement of appropriate biomarkers and pharmocodynamic endpoints are critical for selecting the optimal dose and schedule. A detailed understanding of the molecular mode of action of Hsp90 inhibitors alongside the elucidation of the molecular pathology of individual cancers will help us to identify tumour types and individual patients that will benefit most from treatment. Careful in vitro and in vivo experiments are needed to choose the most potentially advantageous combination studies. It is important to construct a pharmacologic audit trail linking molecular biomarkers and pharmacokinetic and pharmacodynamic parameters to tumour response endpoints. Phase I clinical studies with 17AAG have shown that the drug can be given at does that are well tolerated and that also achieve active pharmacokinetic exposures and elicit molecular signatures of gene and protein expression that are indicative of Hsp90 inhibition. Furthermore, examples of disease stabilisation have been documented, consistent with the generally cytostatic responses that are seen in animal models. Selecting tumour types for Phase II clinical trials must involve balancing 1) our knowledge of molecular response determinants, such as the expression of and dependence upon key client proteins and 2) more pragmatic evidence of antitumour activity in the relevant preclinical models. Examples of likely disease targets include chronic myeloid leukaemia, melanoma, breast, ovarian, brain, thyroid, colorectal and prostate cancer.

17-allylamino,17-demethoxygeldanamycin (17AAG) is a novel anticancer drug that inhibits heat shock protein 90 (Hsp90), resulting in proteasomal degradation of several oncogenic proteins. We used phosphorus magnetic resonance spectroscopy (31P-MRS) to determine whether 17AAG treatment leads to alterations in phospholipids that could serve as pharmacodynamic markers for tumor response to 17AAG.

Epirubicin, cisplatin and continuous 5-fluorouracil (5-FU) infusion (ECF) has been reported to result in high clinical response rates in advanced gastro-oesophageal adenocarcinoma and is currently the 'gold standard' chemotherapy regimen for this tumour site. Despite this, its role as preoperative (neoadjuvant) treatment is unproven and therefore remains under investigation. We report our experience using ECF (intravenous epirubicin 50 mg/m2 and cisplatin 60 mg/m2 every 3 weeks, with continuous infusion of 5-FU 200 mg/m2 per day) as preoperative treatment in locally advanced adenocarcinoma of the lower oesophagus, gastro-oesophageal junction and stomach. Of the 23 patients treated (median age 54 years), 19 had potentially resectable disease, four were unresectable and seven had radiological evidence of lymph node involvement. A median of four cycles of ECF was delivered (range 1-6). Ten of 12 patients (83%) with dysphagia reported improvement of symptoms. Clinical disease progression occurred in six patients (26%) during chemotherapy. WHO grade 3 or 4 toxicity occurred in six patients (26%): four haematological, one mucositis, one vomiting. Seventeen patients (74%) proceeded to surgery; 14 (61%) were resected and three were unresectable. There were two (12%) postoperative deaths from respiratory failure. Major pathological response was seen in three patients (13%): one pathological complete response, two microscopic residual disease. Two patients had Stage II (T2N(0-1)) disease and nine were Stage III (T(3-4)N(0-1)). None of the patients with initially unresectable disease was rendered resectable. After a median follow-up interval of 33 months (range 26-53), the overall median survival was 12 months and 2-year survival was 30%. All patients who were initially unresectable or had radiological evidence of lymph node involvement have died. Therefore, despite good symptomatic response rates, ECF chemotherapy given in the preoperative setting did not appear to improve the outcome of patients with unresectable or radiologically lymph node-positive gastro-oesophageal adenocarcinoma. The role of ECF chemotherapy in resectable tumours is unclear and is currently under investigation in the randomized MRC Adjuvant Gastric Infusional Chemotherapy (MAGIC) study.

A number of molecular therapeutic agents, derived from exploiting our knowledge of the oncogenic pathways that are frequently deregulated in cancer, are now entering clinical trials. One of these is the novel agent 17-allylamino-17-demethoxygeldanamycin that acts to inhibit the hsp90 molecular chaperone. Treatment of four human colon cancer cell lines with iso-effective concentrations of this agent resulted in depletion of c-raf-1 and akt and inhibition of signal transduction. We have used gene expression array analysis to identify genes responsive to treatment with this drug. The expression of hsp90 client protein genes was not affected, but hsc hsp70, hsp90beta, keratin 8, keratin 18 and caveolin-1 were deregulated following treatment. These observations were consistent with inhibition of signal transduction and suggested a possible mechanism of resistance or recovery from 17-allylamino-17-demethoxygeldanamycin treatment. The results shed light on the molecular mode of action of the hsp90 inhibitors, and suggest possible molecular markers of drug action for use in hypothesis testing clinical trials. Oncogene (2000) 19, 4125 - 4133