New cast for a new era: preclinical cancer drug development revisited

Authors

Grit S. Herter-Sprie

Andrew L. Kung

Kwok-Kin Wong

Doi

10.1172/JCI68340

PMID: 23999436 · DOI: 10.1172/JCI68340 · Journal: The Journal of Clinical Investigation (2013)

TL;DR

A narrative review by Herter-Sprie, Kung, and Wong arguing that the historic poor translation of mouse-derived oncology efficacy data into human clinical results — only 9% of preclinical candidates ultimately gain FDA approval — can be improved by combining patient-derived xenograft (PDX) and genetically engineered mouse (GEM) models with a co-clinical trial design that runs murine and human studies in parallel. The authors compare the strengths and weaknesses of conventional xenografts, PDX, and GEM platforms; outline criteria for an “ideal” GEM model and an “ideal” preclinical study (sample size, randomization, blinded evaluation, PD biomarker definition, alignment of response criteria with RECIST); and use case studies in KRAS-driven lung adenocarcinoma (selumetinib + docetaxel) and pancreatic ductal adenocarcinoma (gemcitabine + saridegib) to illustrate where co-clinical inference succeeded and failed.

Cohort & data

  • This is a review; no primary cohort. The authors synthesize results from previously published preclinical and clinical studies.
  • Illustrative case 1: co-clinical trial in KRAS-mutant LUAD testing selumetinib + docetaxel combination therapy in GEM lung adenocarcinoma models paralleling a human phase II study (Jänne et al., Lancet Oncol 2013) PMID:23999436.
  • Illustrative case 2: gemcitabine + saridegib (IPI-926, Smoothened/Hedgehog inhibitor) in pancreatic ductal adenocarcinoma — phase II trial NCT01130142 was halted for inferior survival in the experimental arm despite positive preclinical signal PMID:23999436.
  • Illustrative case 3: the APL paradigm (acute promyelocytic leukemia, APLPMLRARA) cited as the original co-clinical trial success PMID:23999436.

Key findings

  • Conventional xenografts use cell lines selected for in-vitro growth, become less differentiated and more homogeneous in culture, are typically engrafted subcutaneously (ectopic site), and lack a functional immune system — together explaining their historically weak predictive value PMID:23999436.
  • PDX models preserve molecular diversity and co-transplant human stroma, but show genetic drift on serial passage, partial replacement of human microvasculature by host vessels by three weeks (Hylander et al.), and infiltration by murine cells (>40% of analyzed pancreatic PDXs harbored less human than mouse DNA, Lin et al.) PMID:23999436.
  • PDX engraftment failure rates can exceed 90% depending on tumor type; orthotopic implantation of resected NSCLC under the renal capsule of NOD-scid mice raised engraftment from 30–40% to 90% and enabled drug testing within 6–8 weeks (Dong et al.). Of 32 resected tumors, 16 yielded sufficient tissue and 11 of those showed concordant response to three chemotherapy regimens between mouse and patient, with particularly strong concordance for resistance to conventional therapy PMID:23999436.
  • Hidalgo et al. pilot: PDXs from 14 advanced refractory cancer patients tested across 63 drugs in 232 regimens; PDX-guided treatments were effective for 11 of 14 patients (“mouse avatars”) PMID:23999436.
  • GEM models support orthotopic, autochthonous tumor initiation with an intact innate immune system but show a relatively benign cytogenetic profile compared to human tumors and usually arise from single dominant driver mutations rather than the multi-hit accumulation seen in human malignancies PMID:23999436.
  • GEM models “notoriously fail to recapitulate the metastatic features of human malignancies”; multifocal disease is established for hematologic GEMs but remains under investigation for solid tumor GEM-derived allografts PMID:23999436.
  • Preclinical-vs-clinical methodology gaps the authors highlight: preclinical trials rarely meet validation criteria (multiple models, hypothesis-driven design, randomization, blinded evaluation); tumor latency in mice (6–12 weeks post sexual maturity) versus humans (>55 years for most diagnoses); treatment-naive murine tumors versus heavily pretreated phase I/II human patients; few PDX/GEM models recapitulate metastatic spread, which is the dominant cause of human cancer death PMID:23999436.
  • “Simple diminution of tumor growth in preclinical studies is regarded efficacious as long as statistical significance is achieved,” whereas RECIST in humans requires complete arrest of growth at minimum (regression is more desired) — a misalignment of success criteria PMID:23999436.
  • For cytostatic targeted agents, functional metabolomic changes (PET, fluorescence/bioluminescence imaging) are preferred over mass-based readouts; fluorescence imaging is limited to superficial tumors by ~mm tissue penetration whereas luciferase bioluminescence supports deeper GEM tumors PMID:23999436.
  • Recent in-vivo studies cited by the authors suggest therapeutic response rates in mouse models closely mirror those in the clinic when models and trial design are aligned, despite the absence of drug-drug interactions and supportive-care drugs in murine experiments PMID:23999436.

Genes & alterations

  • KRAS — the authors center their case study on a co-clinical trial in murine KRAS-driven LUAD treated with selumetinib plus docetaxel. The murine arm delivered predictive response data in nine months and surfaced STK11 (LKB1) loss as a genetic modifier of response — an analysis not prospectively included in the human registration trial, prompting retrospective reanalysis and informing subsequent trial design PMID:23999436.
  • STK11 (referred to in the paper as Lkb1/LKB1) — loss of this tumor suppressor “significantly influences overall response and drug resistance to docetaxel and selumetinib combination therapy in lung cancer–bearing mice”; the authors propose Lkb1 status as a predictive biomarker of treatment outcome in KRAS-driven LUAD PMID:23999436.
  • SMO (Smoothened, Hedgehog pathway) — target of saridegib (IPI-926). Preclinical mouse data (Olive et al., Science 2009) showed gemcitabine + saridegib extended survival in pancreatic-cancer-bearing mice; the parallel human phase II trial (NCT01130142) was halted by Infinity Pharmaceuticals after interim analysis showed inferior survival in the experimental arm. Post-clinical mouse experiments suggested long-term continuous Smoothened targeting induces a more aggressive phenotype in early-stage disease PMID:23999436.
  • TP53 (mouse Trp53) — referenced as a component of GEM cancer models (Olive et al. mutant p53 Li-Fraumeni models; Hingorani Trp53R172H + KrasG12D pancreatic GEM with chromosomal instability and widespread metastasis) used to discuss genome destabilization needed to make murine tumors phenocopy human genomic instability PMID:23999436.

Clinical implications

  • The authors advocate for routine deployment of co-clinical trial design — synchronized murine (GEM and/or PDX) and human phase I/II studies — to compress timelines, enable interim biomarker discovery, and let real-time mouse findings inform adjustments in the human arm. They cite the APL paradigm (APLPMLRARA) and the murine LUAD selumetinib + docetaxel co-clinical study as proofs of concept PMID:23999436.
  • Co-clinical trials are pitched as especially valuable for rare cancers and genetically defined subsets where human enrollment is rate-limiting; GEM cohorts can substitute for unattainable patient numbers and accelerate prioritization of combination regimens PMID:23999436.
  • The saridegib + gemcitabine PAAD failure is framed as a cautionary tale: chronic versus acute dosing schedules may produce divergent effects on the tumor microenvironment, and short preclinical dosing windows may miss late-emerging toxicities or paradoxical resistance phenotypes PMID:23999436.
  • The authors recommend PD biomarker–driven dose selection (rather than maximum tolerated dose) for molecularly targeted agents in phase I, and call for alignment of preclinical efficacy endpoints with clinical RECIST so that “success” means the same thing in mouse and human studies PMID:23999436.
  • Progression-free survival is suggested as a better preclinical outcome surrogate than overall survival/lethality in mice, because metastatic spread (the dominant cause of human cancer death) is poorly modeled and murine “lethality” may instead reflect drug toxicity rather than tumor burden PMID:23999436.

Limitations & open questions

  • Stromal and immune fidelity of PDX: human stroma is partially replaced by murine fibroblasts and vasculature over passages; the lack of a functional immune system limits use for immunotherapy assessment. Humanized-mouse strategies (Kalscheuer; Wege; DeRose with co-injected human MSCs) are emerging but not yet routine PMID:23999436.
  • Genetic complexity of GEM models: murine tumors typically arise from a single dominant driver, lack the multi-step mutation accumulation of human malignancies, and have a relatively benign cytogenetic profile. Telomere dysfunction, DNA-damage checkpoint impairment, and defective DNA repair are being explored as ways to engineer human-like genomic instability PMID:23999436.
  • Whether GEM-derived syngeneic allografts can recapitulate enhanced metastasis the way orthotopic xenografts do is “currently under intense investigation” PMID:23999436.
  • The relevance of absent drug-drug interactions and supportive-care drugs in murine studies remains “enigmatic” PMID:23999436.
  • The time frame required to reveal deleterious side effects in chronic-dosing targeted-therapy preclinical studies “is still unknown,” limiting confidence in chronic-tolerance estimates PMID:23999436.
  • The paper is a narrative review with no systematic comparison: the authors explicitly state that “comprehensive comparative studies” of alternative murine models versus drug response/resistance “are still missing today.” Many of the case-study claims (Olive et al. post-clinical Smoothened reanalysis; Cory Abate-Shen interactome work; Mariano Barbacid whole-body inducible knockouts) are sourced to personal communications rather than published data PMID:23999436.

Citations from this paper used in the wiki

  • “approximately 99% of its genes overlap with those of Homo sapiens” — opening of Introduction, p. 3639.
  • “only 9% of candidates demonstrate robust clinical performance and are eventually approved by the FDA” — Current challenges, p. 3639.
  • “PDX-bearing mice, or xenopatients, were treated with 63 drugs in 232 treatment regimens. Based on the response data, clinical treatments were effective for 11 out of 14 patients” — describing Hidalgo et al. pilot, p. 3640.
  • “implantation of resected non–small cell lung cancers under the renal capsules of NOD-scid mice not only significantly increased the engraftment rate (90% vs. 30%–40%) … but also allowed rapid drug testing (within 6–8 weeks)” — Dong et al., p. 3640.
  • “GEM models notoriously fail to recapitulate the metastatic features of human malignancies” — p. 3642.
  • “loss of the tumor suppressor Lkb1 significantly influences overall response and drug resistance to docetaxel and selumetinib combination therapy in lung cancer–bearing mice” — Current clinical investigation, p. 3643.
  • “Infinity Pharmaceuticals halted the phase II clinical trial (NCT01130142) because interim analyses revealed inferior survival in the experimental study arm” — saridegib + gemcitabine pancreatic example, p. 3643.
  • “long-term, continuous targeting of Smoothened induces a more aggressive phenotype and reduced overall survival” — Kenneth P. Olive, personal communication, p. 3643.
  • “a co-clinical study evaluating treatment response in murine Kras-driven lung adenocarcinoma revealed predictive response data in just nine months” — p. 3643, referencing Chen et al., Nature 2012.

This page was processed by crosslinker on 2026-05-09.