Development and Characterization of a Clinically Relevant Mouse Model of Melanoma Brain Metastasis

Melanoma has a high predilection for neuronal dissemination, with 44% to 64% of patients developing brain metastases (B-Mets) during the course of metastatic progression which represents the leading cause of death in most of these cases (Davies et al., 2011; Johnson et al., 1996; Sampson et al., 1998). Current treatments for melanoma B-Mets aim to slow disease progression and are determined by their number and location (Eichler et al., 2007; Elliott et al., 2011; Rush et al., 2011). Preclinical models of melanoma B-Mets are fundamental for studying the biology behind the dissemination of melanoma cells to the brain and evaluating new therapeutic regimens’ efficacy.

Over the years, several groups have reported the development of either experimental or spontaneous models of melanoma B-Mets with the majority using the highly invasive intra-carotid (IC) surgery as a standard technique for tumor induction (Fujimaki et al., 1996; Huang et al., 2008; Xie et al., 2006; Yano et al., 2000; Zhang et al., 2009, Cruz-Munoz et al., 2008). The scarcity of clinically relevant, well characterized models has hindered our understanding of the mechanism of melanoma brain tropism and response to therapies. The need to effectively examine different aspects of metastatic dissemination in vivo has motivated investigators to develop imaging strategies in which MRI is best positioned for intracranial imaging. MRI was used recently in B-Mets murine studies for monitoring the temporal and spatial development of breast lesions (Perera, Ribot et al. 2012). In melanoma preclinical studies, the use of in vivo MRI was limited to the detection of B-Mets (Izraely et al., 2012). Hence, there is still a need to establish the tumor growth profile and pattern of melanoma B-Mets using a 3D quantitative imaging analysis.

Here, we aimed to establish an experimental melanoma B-Mets mouse model using the B16F10 murine and the 5B1 human cell lines by means of minimally invasive ultrasound-guided intra-cardiac injection (UGICI). Our goals were to characterize and compare the growth pattern of the developing B-Mets in both models using MRI with micrometric resolution termed micro-MRI (μ-MRI). By utilizing large cohorts of mice (n=80), we intended to overcome the following limitations imposed by in vivo MRI: the slow throughput, the prohibitive cost per scan, as well as limited access to scan time. We opted to use an ex vivo setup devised in our laboratory which enabled simultaneous imaging of multiple whole mouse heads placed in a larger MRI probe. Four heads could be scanned simultaneously during one unattended overnight session (Data S1). Our approach capitalized on the paramagnetic nature of melanin present in both tumor cell lines; the B16F10 murine and to a lesser extent the 5B1 human cell line. The signal brightening endogenous effect of melanoma in tracer-free T1-weighted MRI datasets correlates with the percentage of melanin-containing cells. This feature combined with the signal darkening effect of a T2*-weighted sequence used in our protocol is specific for melanotic tumors but not in amelanotic tumors. Conversely, the latter exhibit T1-weighted hypointense or isointense signal and T2-weighted hyperintense or isointense contrast (Isiklar et al.1995) (Data S2).

Our first goal was to establish a safer and physiologically relevant tumor induction route. For that, we used (1×10⁵cells/100μl/mouse) B16F10 cells and C57Bl6 mouse strain (n=18) in each group to compare the safety and efficacy of our proposed UGICI route to the standard IC injection. The UGICI closely simulate the spread of melanoma cells from the blood stream to the brain tissue. Melanoma cells are introduced through the left ventricle allowing them to circulate freely throughout the entire intact vascular system. This provides the cells with an equal opportunity to colonize different organs and thereby objectively reflects inherent tissue tropism. The use of ultrasound to visualize the advancing needle into the left ventricle of the heart in real time reduced the number of puncture attempts and allowed us to monitor and confirm the proper and uniform transmission of cells into each mouse (Data S3). Comparing this novel minimally invasive technique of tumor induction to the traditional IC injection, we noticed a distinct difference in mortality rate ranging from 20% to 25% for the IC to ~1% with the UGICI. Moreover, we recorded a marked decrease in surgery-related morbidity, evident by stable mouse weight reflecting a better ability of mice to mobilize and feed following UGICI (Data S4). Additionally, the time required to induce the animals was significantly shorter using the UGICI (10–12 minutes) compared to the IC (35–60 minutes) allowing for large scale experiments to be conducted, which is essential for achieving meaningful, statistically-sound data.

Taking advantage of the safer and more efficient UGICI route for tumor induction, we compared the growth pattern and distribution of two experimental melanoma models based on the B16F10 murine and the 5B1 human cell lines. While the B16F10 has no reported predilection for a particular site of metastasis, the 5B1, an in vivo selected subpopulation of the WM239A cell line, was previously reported to show enhanced brain tropism (Cruz-Munoz et al., 2008). In this system, B16F10 cells showed relatively low brain tropism with 27.5% brain penetrance and high affinity to extra-cranial sites while the 5B1 cells had a high predilection for brain tissue with a penetrance rate of 70% and less extra-cranial organ involvement (Data S5). In addition, the extra-cranial metastatic lesions in the 5B1 model were less aggressive and often unilateral (Data S6). The ex vivo MRI protocol implemented was able to recapitulate the characteristic in vivo clinical radiological findings of melanoma B-Mets in both models: bright enhancing areas on a T1-weighted sequence and susceptibility-induced signal darkening on T2*-weighted imaging (Data S7). These findings are used radiologically to distinguish melanoma B-Mets from other types of B-Mets for which the systemic injection of a contrast agent is usually required in the clinical setting (Gaviani et al., 2006). Although T2*-based imaging did not demonstrate systematic tumor detection via signal darkening, the blooming effect associated with this sequence proved useful in combination with T1-hyperintensity in presence of poor contrast and for identifying small lesions. Hence, the evaluation of the tumor volumes in this study relied principally on T1-weighted imaging in which T2*-darkening was complementary when needed.

Ex vivo data acquired from the B16F10 model (n=40) revealed exclusive ventricular and leptomeningeal spread (Data S8a-c), while 5B1 injected mice (n=40) showed generally parenchymal lesions, reflecting the phenotype mostly seen in the clinic (Chen et al., 2012) (Data S8d-f). In addition, MRI allowed for the exploration of the multicentric nature of the two melanoma cell lines used, where 67.9% of the 5B1 injected mice with brain tumors (n=28) showed multiple lesions (range 1 - 16, median: 3) compared to 18% in the B16 mouse model (n=11) (range 1 - 3, median: 1) (Data S9b, c). The 5B1 cells were injected in immune deficient athymic/nude (nu/nu) mice which may account for the higher number of B-Mets developing in the 5B1 model. However, the relatively lower number of extra-cranial metastasis compared to the B16F10 immune competent model suggests that an intrinsically higher brain tropism of the 5B1 cell line, rather than a more permissive host immune system, allows more cells to actively seed and adapt to brain tissue microenvironment. Tumor segmentation, 3D reconstruction and quantitative volumetric analysis at different time points, post-tumor induction (Data S10a for 5B1 model and S10b for B16F10 model) revealed differences in growth pattern of tumors in both models; B16F10 B-Mets were detected on day 15 post-tumor induction at a volume of 0.65 mm³ and exhibited a rapid pattern of growth over a short time frame of 8 days with some lesions detected at a volume of 44.7mm³ on day 23 post-tumor induction (range: 0.14 mm³ - 44.7 mm³, median: 2.01 mm³). The 5B1 metastatic lesions had a delayed onset of tumor detection on day 27 post-injection at a starting volume of 0.03mm³. In addition, the 5B1 lesions were less aggressive and showed a slower growth rate, acquiring a maximum detected volume of 34.01mm³ (range 0.02 mm³ - 34.01 mm³, median: 0.37 mm³) on day 56 post-tumor induction (Data S9a).

Using 5B1 melanoma human cell line and UGICI for tumor induction together with our novel ex vivo MRI technique we provide the field with a clinically relevant B-Mets experimental model and a faithful tool to assess changes in tumor behavior in preclinical therapeutic studies.