“Of mice and men”

Abstract

Human tumour xenografts implanted subcutaneously (s.c.) into immunosuppressed mice have played a significant role in preclinical anticancer drug development for the past 25 years. Their use as a predictive indicator of probable clinical activity has been validated for cytotoxics. A retrospective analysis for 39 compounds where both extensive xenograft testing and Phase II clinical data were available, performed by the National Cancer Institute (NCI), has shown that 15/33 agents (45%) with activity in more than one-third of xenografts showed clinical activity (P=0.04). However, with the exception of non-small cell lung cancer, activity within a particular histological type of the xenograft generally did not predict for clinical activity in the same tumour. Today, the question (largely unanswered) is how useful is the xenograft model (particularly the traditional s.c. model) in contemporary cancer drug discovery? There are many variables when conducting xenograft experiments which impact on outcome; viz, site of implantation, growth properties of the xenograft and size when treatment is initiated, agent formulation, scheduling, route of administration and dose and the selected endpoint for assessing activity. The xenograft model remains of value in current preclinical cancer drug development, especially when such studies give due consideration to the above variables and are based on sound mechanistic (e.g. status of the selected target in the chosen model) and pharmacological (e.g. use of formulated agent) principles. Dependent upon the drug target, a slowing of xenograft tumour growth (cytostatic effect) rather than tumour shrinkage might be the major observed effect. Human tumour xenografts are also particularly useful in determining pharmacodynamic markers of response for subsequent clinical application. Nevertheless, it needs to be kept in mind that the use of xenografts is relatively time-consuming and expensive, raises animal ethical issues and there are instances where the model is inappropriate as a likely predictor of clinical outcome (e.g. inhibitors of the metastatic process and anti-angiogenic strategies as the vasculature is of murine origin).

Introduction

Contemporary anticancer drug development is a multi-million dollar and time-consuming business. Typically, from concept to the completion of Phase III clinical trials and gaining regulatory approval requires in excess of 10 years and as much as 500 million dollars. For the first quarter of a century of modern cancer drug development (circa 1945–1969), thousands of generally randomly produced molecules were tested in mice bearing rapidly growing murine leukaemias (e.g. P388 and L1210) [1]. In 1969, came the first report of the growth of a human tumour in an immunodeficient “nude” (athymic) mouse [2]. Since then, human tumour xenografts grown in nude [3] or in mice with severe combined immunodeficiency (SCID) [4] have covered all of the major tumour types and represented the mainstay of preclinical anticancer drug development testing in vivo.

The modern paradigm for anticancer drug discovery, as widely used by drug companies and within some academic groups, comprises a series of carefully constructed steps that are designed to rapidly and efficiently allow “proof of principle”, pharmaceutically-tractable, molecules to be tested in Phase I and II clinical trials. Such a cascade (Fig. 1) may be envisaged as a large number of molecules feeding into a series of iterative stop/go tests of increasing biological complexity. The concept of “therapeutic index”, that is the demonstration of antitumour efficacy at doses well below those causing severe toxicities, is also a long-established paradigm in preclinical drug discovery [5]. A key part of the preclinical stage of the process, and often representing a significant bottleneck, is the demonstration of antitumour efficacy in a “relevant” tumour model in vivo.

However, what is the place for the nude mouse xenograft model in cancer drug development today in this post-genomic era? What are its values? What are its limitations? Is testing in vivo even required—could predictive answers be obtained solely using human tumour cells in culture or even using in silico methodology, thus avoiding the increasing ethical issues raised by animal experimentation? As opposed to the previous era of “cytotoxic” cancer drugs, contemporary cancer drug development encompasses a wide variety of approaches generally based on attacking specific molecular targets where often, cytostatic rather than cytotoxic effects may be predicted. Thus, a slowing of tumour growth rather than shrinkage may occur. This may require a re-evaluation of the in vivo models developed and validated using cytotoxic drugs when testing such agents. Even within mouse models, is it the case that, in some instances, murine syngeneic models (often dismissed since the advent of human tumour models), transgenic models or orthotopic models may be more appropriate to use than human subcutaneously (s.c.) implanted xenografts in immune-suppressed animals? In some cases, maybe there is no appropriate preclinical model? In an industry where time is paramount, should we dispense with relatively slow and laborious xenograft efficacy determinations in favour of some more rapid, higher-throughput alternative (e.g. the hollow fibre assay [6])?

Hence, this article aims to provide a critical review of the role of human tumour xenografts transplanted in athymic (or SCID) mice in contemporary cancer drug development. Both values and liabilities will be discussed specifically in the context of experience with firstly cytotoxic platinum-based molecules and secondly with small molecules targeted against contemporary cancer targets (e.g. farnesyltransferase, heat shock protein 90, telomerase and tumour vasculature).