The discovery of gene products with recurrent oncogenic mutations in a variety of cancers has revolutionized the approach to cancer therapy. Targeted therapies designed to inhibit these mutated gene products have resulted in dramatic clinical responses. In melanoma, the discovery of BRAF gain-of-function mutations in approximately 50% of melanomas led to the rapid development of targeted BRAF inhibitors. The BRAF inhibitor vemurafenib has produced unprecedented response rates in patients with BRAF-mutant tumors (Chapman et al., 2011). Despite the exquisite sensitivity of most BRAF-mutant melanomas to BRAF inhibition, a subset of tumors show little or no detectable response to therapy. This observation suggests that in addition to BRAF mutation, other genetic/epigenetic changes within a given tumor may influence sensitivity to vemurafenib.

Along these lines, Chen and colleagues recently reported a ‘co-clinical trial’ in which genetically engineered mouse models (GEMMs) of mutant Kras-driven lung cancer were used in parallel to an ongoing clinical trial in KRAS-mutant non-small-cell lung cancer (NSCLC) (Chen et al. 2012). KRAS is mutated in 15–30% of NSCLC and is associated with poor response to conventional therapy. This study utilized three different Kras-driven GEMMs based on conditional genetic recombination and subsequent spontaneous tumor formation in the lungs of mice. These models included tumors driven by mutant Kras only (Kras model), mutant Kras in combination with p53 inactivation (Kras/p53 model), and mutant Kras in combination with Lkb1 inactivation (Kras/Lkb1 model). p53 and Lkb1 are tumor suppressors that are commonly inactivated in NSCLC (reviewed by Wen et al., 2011). Tumor-bearing mice of these genotypes were randomized into one of three treatment groups: docetaxel cytotoxic therapy, targeted therapy with selumetinib, or combination therapy with both docetaxel and selumetinib. Docetaxel is a standard of care chemotherapy in NSCLC. Selumetinib is a MEK inhibitor; MEK is a downstream mediator of activated KRAS signaling in the MAPK/ERK pathway. Tumor response was determined by serial MRI. This murine co-clinical trial was designed to complement an analogous study in humans, which is testing the efficacy of adding selumetinib to docetaxel therapy in patients with KRAS-mutant NSCLC.

Murine Kras tumors (with wild-type p53 and Lkb1) were the most responsive to docetaxel, with approximately 30% of mice exhibiting a partial response (tumor shrinkage >30%) with this single-agent therapy. Kras/p53 and Kras/Lkb1 tumors were significantly less sensitive to docetaxel therapy, suggesting that inactivation of either of these tumor suppressors affects response. Monotherapy with selumetinib, while slowing tumor growth and decreasing MAPK/Erk pathway activation, did not lead to tumor shrinkage in any of the models. Combination therapy with docetaxel and selumetinib increased response rates to nearly 100% of mice in the Kras model. Similarly, in the Kras/p53 model, response rates were enhanced, with approximately 60% of tumors exhibiting significant tumor shrinkage with combination therapy. Both models showed improved progression-free survival, a metric commonly used to assess treatment responses in human clinical trials. Notably, there was no statistically significant effect on tumor size with combination therapy in Kras/Lkb1 tumors. At the immunohistochemical level, Kras and Kras/p53 tumors showed increased apoptosis and decreased proliferation after combination treatment, while Kras/Lkb1 tumors did not show statistically significant changes in these parameters. Together, these data demonstrated an inherent resistance to combination therapy with docetaxel and selumetinib that was mediated by Lkb1 loss.

FDG-PET is a clinical imaging modality that utilizes a radiolabeled glucose analog (18F-fluoro-2-deoxyglucose, FDG) to visualize tissues with high glucose uptake, such as tumors. Chen and colleagues considered that FDG-PET might provide a human-translatable means by which to assess initial response to therapy. The baseline (pretreatment) FDG-PET signal was established in the three different GEMMs and was found to be significantly different among models. Specifically, FDG uptake was significantly higher in the Kras/Lkb1 tumors compared to the Kras and Kras/p53 tumors. The authors reasoned that LKB1-deficient KRAS-mutant NSCLCs from patients may also show enhanced FDG uptake at baseline, a hypothesis that was supported by comparing PET signal intensity in tumors with and without LKB1 staining by immunohistochemistry. Next, the effect of treatment on FDG-PET signal was tested in the mouse models. While docetaxel treatment alone did not significantly reduce PET positivity in Kras and Kras/p53 tumors, combined therapy with both docetaxel and selumetinib resulted in a significant reduction in FDG uptake. Kras/Lkb1 tumors showed no significant changes in FDG uptake in response to either single-agent or combination therapy. On the basis of these findings, the authors suggest that FDG-PET studies (both pre- and post-treatment) may be useful in assessing initial responses to therapy in patients with NSCLC.

To clarify potential mechanisms by which Lkb1 loss might mediate resistance to docetaxel and selumetinib treatment, Chen and colleagues completed extensive Western blot signaling analyses using protein lysates prepared from the various murine tumors. As would be predicted, Kras tumors showed evidence of MAPK/Erk pathway activation (phosphorylated Erk). However, Kras/p53 tumors showed relatively increased levels of activated MAPK/Erk, while Kras/Lkb1 tumors showed relatively reduced levels of activated MAPK/Erk compared to the Kras tumors. These patterns were also evident in human KRAS-mutant NSCLC with analogous changes to p53 and LKB1. Murine Kras/Lkb1 tumors showed increased activation of other oncogenic pathways including phosphorylated Akt, suggesting that PI3K/Akt signaling may have a more critical role in tumorigenesis than MAPK/Erk signaling in the Kras/Lkb1 tumors. This hypothesis is consistent with resistance to MEK inhibition (selumetinib) and enhanced FDG-PET intensity at baseline. Glucose uptake in cancer is thought to be regulated in a positive fashion by PI3K/AKT signaling (reviewed by Cairns et al., 2011).

A discouraging trend in some newer cancer therapies is that despite often promising efficacy at the outset of treatment, resistance nearly inevitably develops; a finding paralleled in this mouse co-clinical trial. At a mechanistic level, the authors sought to address potential resistance mechanisms to docetaxel/selumetinib treatment in the Kras and Kras/p53 models. These analyses suggested that mechanisms of resistance were likely heterogenous, but seemed to boil down to reactivation of the MAPK/Erk pathway. The possibility of stochastic Lkb1 inactivation as a mechanism of therapy resistance in Kras and Kras/p53 tumors was not explored.

Chen and colleagues lead by example through this study and describe a practical approach through which GEMMs can be used to inform and even direct clinical trials in humans. The authors show that not all KRAS-mutant NSCLCs are created equally and likely also derive characteristics from concurrent functional alterations to proteins such as p53 or LKB1. These findings have broad implications across all cancers, including melanoma. For example, BRAF-mutant melanomas exhibiting differential sensitivity to BRAF inhibition likely owe these differences (at least in part) to the modifying effects of additional, functionally relevant changes to other pathways. Similar co-clinical approaches are very feasible in melanoma given the availability of both mutant Braf-driven melanoma GEMMs and a variety of additional conditional alleles (reviewed by Damsky and Bosenberg, 2010). In particular, it will be of interest to determine whether other specific genetic changes within murine Braf-driven melanomas modulate the degree of response to Braf inhibition. More generally, melanoma GEMMs will not only help to predict pathways that modulate treatment response to BRAF inhibition, but will also enable formal study of the intersection of targeted inhibitors, cytotoxic drugs, and immunomodulatory agents in a genotype-specific fashion.

Genetically engineered mouse models provide several convenient experimental advantages including: (i) the ability to practically optimize dosing, scheduling, and ordering of multiple-agent regimens, (ii) the ability to biopsy tissue at any point before, during, or after treatment/subsequent resistance, (iii) the ability to objectively measure the responses to treatment in an experimentally tractable fashion, (iv) the availability of fluorescent reporter strains that can assist in tumor/metastasis and/or immune-infiltrating cell visualization and purification, and (v) the ability to utilize imaging agents/substrates that allow in vivo, real-time study of the tumor microenvironment, as well as processes such as glucose uptake and apoptosis (intravital imaging reviewed by Pittet and Weissleder, 2011).

At a more fundamental level, GEMMs have the ability to help elucidate the interactions among different genetically altered pathways within a given tumor. Differences in response to targeted therapies are likely a reflection of fundamental differences in the nature of alterations to pathways driving tumorigenesis. This concept is illustrated nicely by Chen et al. when comparing MAPK/Erk and PI3K/Akt pathway activation between Kras/Lkb1 and Kras/p53 tumors. In melanoma, modulation of β-catenin signaling was recently shown to dramatically modulate metastasis in a context-specific manner that depended upon other co-occurring alterations to Braf and the Pten tumor suppressors (Damsky et al., 2011). Notably, β-catenin alteration also affected signaling through MAPK/Erk and PI3K/Akt pathways in these models. More examples of context-specific function and modulating/interacting genetic alterations in GEMMs and human cancers are likely to be described in the coming years.

Genetically engineered mouse models and the co-clinical model will help to delineate the complex relationships between altered signaling pathways in cancers and will also expedite clinical development of novel cancer therapeutics. The recent efforts by Chen et al. emphasize that the presence of a given driving mutation does not guarantee clinical response to a targeted therapy. The co-clinical model will facilitate sub-classification of cancers with the same driving mutations (e.g., KRAS) into biologically meaningful groups (e.g., LKB1 presence or absence) that more accurately correlate with clinical responses. This approach is likely to have important implications in our understanding of basic processes mediating cancer formation, progression, and response to treatment and will ultimately enhance the effectiveness of care for patients with cancer.


  1. Top of page
  2. References