Research at Greehey Children’s Cancer Research Institute (Greehey CCRI) has been at the forefront of improving relevant preclinical models for developing new therapeutics.
Cancer in children remains the major cause of disease-related death in the U.S. Although childhood cancers are rare, the use of multimodality therapies (surgery, intensive chemotherapy, and radiation) is curative in approximately 70% of patients. However, while overall 5-year Event-Free Survival now exceeds 80%, patients who have advanced, metastatic, or relapsed disease have dismal outcomes. In patients surviving their cancer, life-long/threatening toxicities result from their therapy. The challenge is to maintain or improve survival rates and reduce toxicities due to treatment.
Acute lymphoblastic leukemia represents 40% of all pediatric cancer in the U.S. but can be classified into multiple molecular subtypes having different outcomes, and potentially responding to various ‘targeted’ therapies. Even for the most common cancer type, the number of patients within a molecular subtype is small, presenting challenges to conducting clinical trials in these populations. Similarly, for even more rare diseases such as medulloblastoma (a CNS cancer), or rhabdomyosarcoma (a tumor of soft tissue), molecular studies have subclassified these cancers into molecularly distinct entities (potentially requiring different therapeutic approaches). With such small patient populations, the development of efficient and accurate preclinical models becomes imperative for identifying novel effective agents or combinations of agents for each molecular type.
Research at Greehey CCRI has been at the forefront of establishing and validating relevant preclinical models for developing new therapeutics. Specifically, researchers now at Greehey CCRI were the first to create paediatric Patient-Derived Xenografts (PDX), where a sample of tumor from the child is directly implanted into an immune-deficient mouse (hence xeno-graft). These models accurately reproduce the histologic and molecular characteristics of the original tumor, therefore offer the potential to be valuable models to identify new therapeutics.
Most PDX models tend to overpredict clinical efficacy
However, challenges remain; the failure rate of cancer drugs being developed for the treatment of adult cancers is about 95%. Considering that preclinical development of most of these drugs involved screening in xenograft models, there is reason to consider the causes of such failures. When considering developing the Pediatric Preclinical Testing Program (PPTP), the NCI convened a group to find appropriate preclinical models and review available data to support or refute the value of using PDX models in the program.1 The significant limitations to preclinical models are outlined (see Table 1). Most PDX models tend to overpredict clinical efficacy – false positives drugs that cause tumor shrinkage in mice but fail to have efficacy in patients. In many instances, this discrepancy is because drug exposures achieved in mice were much more significant than tolerated in patients; human cancer growing in a mouse would respond, but not the tumor growing in the patient.2
A relatively simple ‘fix’ is to use doses and schedules of drug administration in mice that accurately recapitulate drug plasma exposure that can be achieved in (adult) patients. Clinical trials in children almost always are conducted after a Phase 2 Recommended Dose (P2RD) is established in adult patients, usually with pharmacokinetics being defined. Thus, factoring in relevant dosing in preclinical models should at least reduce the risk of identifying false-positive leads.
The second limitation noted was that preclinical models do not represent the genetic diversity of pediatric cancers.1 Molecular characterisation of many hundreds of pediatric cancers over the last decade has only emphasised the genetic and epigenetic heterogeneity of cancer types that were once thought to be quite homogeneous. The problem of accurately representing the diversity of any pediatric cancer type becomes even more critical and challenging. The third putative reason for failure identified was that clinical criteria for a patient response (as defined by RECIST), were more stringent than those criteria used to determine antitumor activity in preclinical models.3
Limitations’ two and three are intimately linked; the more stringent the criteria for defining an antitumor activity, the fewer animals per treatment group are required in preclinical studies. If correct, then within the same resource constraints, more models of each cancer type may be incorporated, hence increasing the genetic/epigenetic diversity. A retrospective analysis of PPTP studies undertaken by Greehey CCRI and other researchers, included the use of 83 PDX models, 67 drugs/biologicals, and over 2100 tumor/drug determinations.4 The significant finding was that using a single mouse per treatment group gave essentially the same result as ten mice/group (solid tumors) or eight mice/group (leukemias).
Single mouse experimental designs
Prospective validation of single mouse experiments has now been accomplished in two studies soon to be published. Using 80 PDX models representing subtypes of acute leukemias, the PPTC prospectively tested two drugs previously evaluated in conventional experimental design (8 mice/group), where the original 7 PDX models showed virtually identical responses in the single mouse experiment. However, for subgroups of leukemias, such as mixed lineage leukemia (MLL), models highly sensitive, or highly resistant, were identified. By expanding the diversity of models examined, additional information on subgroup heterogeneity became apparent. Similar to the leukemia results, using a single mouse experimental design (rather than 10/group) each mouse is representing one of 32 different childhood patient solid tumors, the study identified a nano-formulated camptothecin as highly active. Importantly, sensitivity was highly correlated with irinotecan activity (another camptothecin analogue) in PDX models where both agents were evaluated. Biomarker identification is facilitated as virtually all of the PDX models within the PPTC have been molecularly characterised.5
The feasibility and validity of using a single mouse experimental design have been demonstrated, but how will this potentially impact our ability to identify novel agents better and improve the ‘translatability’ of preclinical results to the clinic? Essentially the single mouse design generates the same type of data that is sought in phase 2 clinical trial; the objective response rate (ORR; the proportion of tumors that shrink based on RECIST criteria) and time to tumor progression (TTP) or time to an event. In most clinical trials, an untreated control arm would be considered unethical, and no controls are included in the single mouse design. What we gain is that within the same resource constraints, about 16-20 fold more pediatric cancer models can be used to identify new active agents compared to traditional approaches. This will incorporate far more diversity from any tumor type, one of the perceived limitations of preclinical testing. Second, generating response data from a large number of models, each representing a different ‘patient’ with the same disease, may allow identification of ‘exceptional responders.’ Identification of a biomarker that segregates with tumor sensitivity may allow stratification of the patient’s in a subsequent clinical trial. Increasing the number of models representing a tumor type may also improve the prediction of subsequent clinical activity against that disease. For example, suppose one uses a conventional design that limits drug evaluation to use of six PDX models (e.g., 6 neuroblastomas). The agent causes regression in two of six models – would one advance the drug? Alternatively, if the drug caused regression in only two of 30 neuroblastoma models, one probably would not prioritise that agent unless there was a robust biomarker for identifying patients likely to benefit.
It is estimated that over 1100 medicines and vaccines are currently in development by U.S. biopharmaceutical companies for treatment of cancer, however with only rare exceptions, are entities being specifically developed for the treatment of pediatric cancer.6 In the U.S., the regulatory environment is changing with the objective of accelerating the rate at which agents enter into pediatric cancer clinical trials. The Research to Accelerate Cures and Equity for Children Act (RACE for Children Act), as part of the Food and Drug Administration (FDA) Reauthorization Act (FDARA), requires the FDA to develop a list of molecular targets of new drugs and biologics in development, that are determined to be substantially relevant to the growth and progression of pediatric cancer. Agents directed at targets on the list may trigger the requirement for pediatric investigations. The intent is to engage Pharma in pediatric testing at an early stage in drug development. Under the RACE for Children Act, FDA may now require pediatric assessments when molecular targets under FDA review are substantially relevant to childhood cancer. Appropriate preclinical models must be developed for the evaluation of what is estimated to be 200 new entities per year and to create new design paradigms that make this possible within resource constraints. While we have focused on the value and limitations of pediatric tumor models, the development of humanised mouse host strains for evaluating immune-oncology agents is also a priority. Although many challenges to curing childhood cancer with an acceptable quality of life remain, this is a time of considerable excitement as the options for developing effective and less toxic treatments have never been greater.
References
1 P.J Houghton, P.C Adamson, and S. Blaney et al. 2002. Testing of new agents in childhood cancer preclinical models meeting summary
2 J.K Peterson, and P.J Houghton. 2004. Integrating pharmacology and in vivo cancer models in preclinical and clinical drug development. doi:10.1016/j.ejca.2004.01.003
3 E.A Eisenhauer, P. Therasse, and J. Bogaerts et al. 2009. New response evaluation criteria in solid tumours:
revised RECIST guideline (version 1.1). doi:10.1016/j.ejca.2008.10.026
4 B. Murphy, H. Yin, and J.M Maris et al. Evaluation of Alternative In Vivo Drug Screening Methodology: A Single Mouse Analysis. doi:10.1158/0008-5472.CAN-16-0122
5 J.L Rokita, K.S Rathi, and M.F Cardenas et al. 2019. Genomic profiling of childhood tumor patient-derived xenograft models to enable rational clinical trial design. Cell reports In Press
6 H.V ErkizanKong, Y, Merchant, and M, Schlottmann. 2009. A small molecule blocking oncogenic protein EWS-FLI1 interaction with RNA helicase – A inhibits growth of Ewing’s sarcoma. doi:10.1038/nm.1983
Peter J. Houghton PhD
Professor
Raushan T. Kurmasheva PhD
Assistant Professor
+1 210 562 9200
GCCRI@uthscsa.edu
