Nat Med. 2015 Nov;21(11):1318-25.

High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response

Hui Gao, Juliet Anne Williams, et al.

Oncology Disease Area, Novartis Institutes for Biomedical Research, Cambridge, Massachusetts, USA

 

Abstract:

Profiling candidate therapeutics with limited cancer models during preclinical development hinders predictions of clinical efficacy and identifying factors that underlie heterogeneous patient responses for patient-selection strategies. We established ~1,000 patient-derived tumor xenograft models (PDXs) with a diverse set of driver mutations. With these PDXs, we performed in vivo compound screens using a 1×1×1 experimental design (PDX clinical trial or PCT) to assess the population responses to 62 treatments across six indications. We demonstrate both the reproducibility and the clinical translatability of this approach by identifying associations between a genotype and drug response, and established mechanisms of resistance. In addition, our results suggest that PCTs may represent a more accurate approach than cell line models for assessing the clinical potential of some therapeutic modalities. We therefore propose that this experimental paradigm could potentially improve preclinical evaluation of treatment modalities and enhance our ability to predict clinical trial responses.

PMID: 26479923

 

Supplementary

Development of effective drugs to treat or cure human cancer has been very challenging, reflected by the fact, in the past several decades, over 90% of novel drug candidates entering oncology clinical trials have failed (1,2). Reasons for this low success rate include a lack of understanding of cancer biology and also the failure to model the complexity of human cancer preclinically. In oncology drug research and development, a drug candidate is often evaluated for their anti-tumor activities in cell line-derived models both in vitro and in vivo. One of the caveats of using cell lines is that they have been propagated in vitro for many generations and therefore compose of a selection of cells that grow robustly in 2D on plastic. Although often these cell lines still carry the driver mutations and maintain some characteristics of original patient tumors, they have deviated from truly representing patient disease, which may well contribute to the low success rate of predicting human clinical trial response. Another major caveat in traditional oncology drug discovery is that drug candidates are often evaluated in only a handful biasly chosen xenograft models before being tested in the clinic. We know though, each ‘type’ of cancer, for example colon cancer, is not just one type of cancer, but has many subsets of disease which may have differential responses to therapy. Therefore the approach of choosing only a small selection of in vivo models to evaluate a therapy fails to recapitulate the response heterogeneity seen among different patients, which also may contribute to the high failure rate of oncology drugs in patients.

 

 

WJ-FIG1Figure 1: The Novartis Institutes for Biomedical Research patient-derived tumor xenograft (PDX) collection (PDXE). (A)Distribution of cancer types in the PDXE by lineage. (B) Genomic landscape analysis of pancreatic cancer: left, PDXs; middle, The Cancer Genome Atlas (TCGA); right, Cell Line Encyclopedia (CCLE).

 

Seeking to reduce the number of preclinical drugs that fail along the road to regulatory approval, we initiated an effort to establish a large panel of patient-derived tumor xenograft (PDX) models and assess their values to predict patient response in the clinic (3). These PDXs are established by directly implanting a piece of human tumor fragment in immunocompromised mouse with minimum manipulation. We have generated an extensive collection of over 1,000 models, representing a spectrum of solid cancers (Figure 1). We performed extensive genomic characterization on these models to understand how closely they represent patient tumors and subsequently compared our PDX tumor bank with patient tumors that were collected and characterized in The Cancer Genome Atlas (TCGA). The analyses indicate that our PDXs closely resemble the genomic landscape when assessed using somatic mutation and copy number alterations. We also found that these two data sets were more comparable than data from cell lines, primarily because of a lack of hypermutators in melanoma cell line models. Additionally we found that all major oncogenic driver alterations were captured in the PDXs with similar mutation frequency seen in the patient tumors, and this close alignment is more pronounced than with the cell line bank (CCLE) (Figure 1). We hypothesized that we could harness these PDXs to model inter-patient response heterogeneity in a setting that recapitulates human clinical trials: Our ultimate goal being, to develop cancer therapeutics with a much higher probability of working patients by overcoming limitations associated with in vitro systems and with using a small number of models.

 

 

WJ-FIG2

Figure 2: Reproducibility of PDX-based clinical trials (PCTs). (A) Use of PDXs to model inter-patient response heterogeneity in PCTs. (B) Responses to PI3K inhibitors BKM120 and CLR457. (C) Responses to everolimus/LEE011 combination.

 

To that end, we designed a non-bias PDX-based clinical trial using PDXs from different lineages/indications (Figure 2). In a given PCT, each mouse receiving the therapy of interest bears a unique tumor xenograft from an individual patient. By treating a population of such mice, the therapy’s efficacy against the cancer type in question can be determined. Furthermore, we can capture the heterogeneity of responses between patients by including a large number of PDXs in the trial. Additionally we can compare the relative effectiveness of a treatment across multiple indications to an extent not currently feasible in the clinic. Unlike the traditional preclinical in vivo studies where a drug candidate is often evaluated in an experimental setting using multiple animals per treatment group to facilitate statistical analysis, here for feasibility, we only use one animal per treatment group. We are not concerned with the response of an individual tumor per se, rather the response of the population. But still, the question is, s this approach reproducible? We addressed this by both comparing responses to treatments that modulate same molecular targets in a single trial as well as by comparing responses to same treatments in two independent trials. To our satisfaction, although we anticipated potential variability between in an individual PDX, population responses were consistent: For example, two structurally distinct phosphoinositide 3-kinase (PI3K) inhibitors BKM120 and CLR457 gave an almost identical population response rate in >200 PDXs across six indications (Figure 2) and the effectiveness of a combination treatment LEE011 (cyclin-dependent kinase 4/6 inhibitor, CDK4/6 inhibitor)/everolimus (inhibitor of mammalian target of rapamycin, mTOR inhibitor) was reproducible in 2 distinct  PCT studies (Figure 2). These data clearly support the idea that when screened across a large number of PDXs, the response to both single agents and combinations can be accurately modeled using one animal per PDX model.

 

 

WJ-FIG3

Figure 3: Translatability of PCTs. (A) Response to BRAF inhibitor encorafenib in melanoma PDXs. (B) Response to combination of inhibitors of BRAF (encorafenib) and MEK (binimetinib) in melanoma PDXs.

 

 

The next questions we asked were: How do the responses we see in our PDX trials compare to efficacy actually achieved in human clinical trials? How translatable are our preclinical trials? Interestingly, we have found that the data from PCTs is highly consistent with what we see in humans. For instance, BRAF-mutant PDXs responded well to BRAF inhibition and even better with the addition of a MEK inhibitor (Figure 3). The response rates to the BRAF inhibitor single agent and combination with MEK inhibitor are similar to what have been reported in melanoma patients carrying BRAF mutation in the clinic. Importantly, we found that therapeutic activity in vitro wasn’t necessarily seen in vivo, and vice versa. Novartis’s investigational IGF1R inhibitor LFW527 appeared to increase the efficacy of the MEK1/2 inhibitor binimetinib (MEK162; Array BioPharma) in pancreatic cancer cell lines in vitro (Figure 4). When we combined binimetinib with an antibody against IGF1R and tested it in pancreatic PCTs, no such synergy was observed—the IGF1R inhibitor failed to potentiate the anti-tumor activities of binimetinib as seen in vitro in cell lines (Figure 4). There are several literature reports showing that combinatorial inhibition of IGF1R and MEK1/2 are efficacious not only in cell proliferation assays in vitro but also, importantly, in cell line–derived xenografts in vivo; however, the outcomes of such combinations in the clinic have been very dismal(4-6). On the other hand, the PCT has discovered effective combination partners in vivo which were not uncovered in cell line screening in vitro. For example LEE011, a Novartis’s CDK4/6 inhibitor, significantly increases the response rate and extends progression-free survival when combined with BRAF inhibitor encorafenib (LGX818; Array BioPharma) in BRAF-mutant melanomas (Figure 4). The response rate with this combination is 100% with 70% of PDXs demonstrating complete or partial regression, (in contrast to encorafenib single agent where the response rate is only 65%). A clinical investigation of this combination is currently under way based on these encouraging PCT results.

Importance of the study: The PDXs and the herein described PDX clinical trial concept is the first to evaluate the reproducibility and translatability of the 1x1x1 concept to drug response using an extensive, well-characterized PDX collection. This approach represents a novel experimental paradigm to address tumor biology including the diversity of cancer patients and interrogate targeted therapies in in vivo models that are more relevant to the clinic than traditional oncology models, potentially improving the ability of preclinical oncology studies to predict patient response in the clinic.

 

 

WJ-FIG4

Figure 4: In vitro and in vivo responses to combination of IGF1R inhibitor (LFW527) and MEK inhibitor (binimetinib) in pancreatic cancer. (A) The number of hits from in vitro combination screening in pancreatic cancer cell lines. (B) Response to binimetinib single agent in pancreatic PDXs. (C) Response to IGF1R inhibitor figitumumab in pancreatic PDXs. (D) Response to figitumumab and binimetinib combination in pancreatic PDXs. (E) Response to combination of BRAF inhibitor encorafenib and CDK4/6 inhibitor LEE011.

 

References:

  1. DiMasi, J. A., Reichert, J. M., Feldman, L., and Malins, A. (2013) Clinical approval success rates for investigational cancer drugs. Clin Pharmacol Ther 94, 329-335
  2. Paul, S. M., Mytelka, D. S., Dunwiddie, C. T., Persinger, C. C., Munos, B. H., Lindborg, S. R., and Schacht, A. L. (2010) How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nat Rev Drug Discov 9, 203-214
  3. Gao, H., Korn, J. M., Ferretti, S., Monahan, J. E., Wang, Y., Singh, M., Zhang, C., Schnell, C., Yang, G., Zhang, Y., Balbin, O. A., Barbe, S., Cai, H., Casey, F., Chatterjee, S., Chiang, D. Y., Chuai, S., Cogan, S. M., Collins, S. D., Dammassa, E., Ebel, N., Embry, M., Green, J., Kauffmann, A., Kowal, C., Leary, R. J., Lehar, J., Liang, Y., Loo, A., Lorenzana, E., Robert McDonald, E., 3rd, McLaughlin, M. E., Merkin, J., Meyer, R., Naylor, T. L., Patawaran, M., Reddy, A., Roelli, C., Ruddy, D. A., Salangsang, F., Santacroce, F., Singh, A. P., Tang, Y., Tinetto, W., Tobler, S., Velazquez, R., Venkatesan, K., Von Arx, F., Wang, H. Q., Wang, Z., Wiesmann, M., Wyss, D., Xu, F., Bitter, H., Atadja, P., Lees, E., Hofmann, F., Li, E., Keen, N., Cozens, R., Jensen, M. R., Pryer, N. K., Williams, J. A., and Sellers, W. R. (2015) High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response. Nat Med 21, 1318-1325
  4. Flanigan, S. A., Pitts, T. M., Newton, T. P., Kulikowski, G. N., Tan, A. C., McManus, M. C., Spreafico, A., Kachaeva, M. I., Selby, H. M., Tentler, J. J., Eckhardt, S. G., and Leong, S. (2013) Overcoming IGF1R/IR resistance through inhibition of MEK signaling in colorectal cancer models. Clin Cancer Res 19, 6219-6229
  5. Ebi, H., Corcoran, R. B., Singh, A., Chen, Z., Song, Y., Lifshits, E., Ryan, D. P., Meyerhardt, J. A., Benes, C., Settleman, J., Wong, K. K., Cantley, L. C., and Engelman, J. A. (2011) Receptor tyrosine kinases exert dominant control over PI3K signaling in human KRAS mutant colorectal cancers. J Clin Invest 121, 4311-4321
  6. Molina-Arcas, M., Hancock, D. C., Sheridan, C., Kumar, M. S., and Downward, J. (2013) Coordinate direct input of both KRAS and IGF1 receptor to activation of PI3 kinase in KRAS-mutant lung cancer. Cancer Discov 3, 548-563

 

Acknowledgements: This research was funded by Novartis, Inc. and both authors were employees thereof at the time the study was performed.

 

Contact:

Juliet Anne Williams, Ph.D.

Director of Pharmacology

Oncology Disease Area, Novartis Institutes for Biomedical Rsearch, Cambridge, MA 02139

Email: juliet.williams@novartis.com

 

 

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann