SUMMARY Glioblastoma remains a deadly cancer driven by invasion of tumor cells into the brain. Transcriptomic analyses have revealed distinct molecular subtypes, but mechanistic differences that explain clinical differences are not clear. Here, we show that, as predicted by the motor-clutch model for cell migration, mesenchymal glioma cells are more spread, generate larger traction forces, and migrate faster in brain tissue compared to proneural cells. Despite their fast migration and comparable proliferation rate in vitro, mice with mesenchymal tumors live longer than mice with proneural tumors, which was correlated with an immune response in the mesenchymal mice that included T cell-mediated killing of cancer cells, similar to human tumors. Thus, mesenchymal tumors have aggressive migration, but are relatively immunologically ‘hot’ which suppresses net proliferation. These two features counteract each other and may explain the lack of a strong survival difference between subtypes clinically, while also opening up new opportunities for subtype-specific therapies.

Radiotherapy is involved in 50% of all cancer treatments and 40% of cancer cures. Most of these treatments are delivered in fractions of equal doses of radiation (Fractional Equivalent Dosing (FED)) in days to weeks. This treatment paradigm has remained unchanged in the past century and does not account for the development of radioresistance during treatment. Even if under-optimized, deviating from a century of successful therapy delivered in FED can be difficult. One way of exploring the infinite space of fraction size and scheduling to identify optimal fractionation schedules is through mathematical oncology simulations that allow for in silico evaluation. This review article explores the evidence that current fractionation promotes the development of radioresistance, summarizes mathematical solutions to account for radioresistance, both in the curative and non-curative setting, and reviews current clinical data investigating non-FED fractionated radiotherapy.

Purpose We demonstrate the importance of considering intratumoral heterogeneity and the development of resistance during fractionated radiotherapy when the same dose of radiation is delivered for all fractions (Fractional Equivalent Dosing FED). Materials and Methods A mathematical model was developed with the following parameters: a starting population of 1011 non-small cell lung cancer (NSCLC) tumor cells, 48-hour doubling time, and cell death per the linear-quadratic (LQ) model with α and β values derived from RSIα/β, in a previously described gene expression based model that estimates α and β. To incorporate both inter- and intratumor radiation sensitivity, RSIα/β output for each patient sample is assumed to represent an average value in a gamma distribution with the bounds set to -50% and +50% of RSIα/b. Therefore, we assume that within a given tumor there are subpopulations that have varying radiation sensitivity parameters that are distinct from other tumor samples with a different mean RSIα/β. A simulation cohort (SC) comprised of 100 lung cancer patients with available RSIα/β (patient specific α and β values) was used to investigate 60Gy in 30 fractions with fractionally equivalent dosing (FED). A separate validation cohort (VC) of 57 lung cancer patients treated with radiation with available local control (LC), overall survival (OS), and tumor gene expression was used to clinically validate the model. Cox regression was used to test for significance to predict clinical outcomes as a continuous variable in multivariate analysis (MVA). Finally, the VC was used to compare FED schedules with various altered fractionation schema utilizing a Kruskal-Wallis test. This was examined using the end points of end of treatment log cell count (LCC) and by a parameter described as mean log kill efficiency (LKE) defined as: Results Cox regression analysis on LCC for the VC demonstrates that, after incorporation of intratumoral heterogeneity, LCC has a linear correlation with local control (p = 0.002) and overall survival (p =< 0.001). Other suggested treatment schedules labeled as High Intensity Treatment (HIT) with a total 60Gy delivered over 6 weeks have a lower mean LCC and an increased LKE compared to standard of care 60Gy delivered in FED in the VC. Conclusion We find that LCC is a clinically relevant metric that is correlated with local control and overall survival in NSCLC. We conclude that 60Gy delivered over 6 weeks with altered HIT fractionation leads to an enhancement in tumor control compared to FED when intratumoral heterogeneity is considered.

In Glioblastoma (GBM), tumor spreading is driven by tumor cells’ ability to infiltrate healthy brain parenchyma, which prevents complete surgical resection and contributes to tumor recurrence. GBM molecular subtypes, classical, proneural and mesenchymal, were shown to strongly correlate with specific genetic alterations (Mesenchymal: NF1; Classical: EGFRVIII; Proneural: PDGFRA). Here we tested the hypothesis that a key mechanistic difference between GBM molecular subtypes is that proneural cells are slow migrating and mesenchymal cells are fast migrating. Using Sleeping Beauty transposon system, immune-competent murine brain tumors were induced by SV40-LgT antigen in combination with either NRASG12V (NRAS) or PDGFB (PDGF) overexpression. Cross-species transcriptomic analysis revealed NRAS and PDGF-driven tumors correlate with human mesenchymal and proneural GBM, respectively. Similar to human GBM, CD44 expression was higher in NRAS tumors and, consistent with migration simulations of varying CD44 levels, ex vivo brain slice live imaging showed NRAS tumors cells migrate faster than PDGF tumors cells (random motility coefficient = 30µm2/hr vs. 2.5µm2/hr, p < 0.001). Consistent with CD44 function as an adhesion molecule, migration phenotype was independent of the tumor microenvironment. NRAS and human PDX/MES tumor cells were found to migrate faster and have larger cell spread area than PDGF and human PDX/PN tumors cells, respectively, in healthy mouse brain slices. Furthermore, traction force microscopy revealed NRAS tumor cells generate larger traction forces than PDGF tumors cells which further supports our theoretical mechanism driving glioma migration. Despite increased migration, NRAS cohort had better survival than PDGF which was attributed to enhanced antitumoral immune response in NRAS tumors, consistent with increased immune cell infiltration found in human mesenchymal GBM. Overall our work identified a potentially actionable difference in migration mechanics between GBM subtypes and establishes an integrated biophysical modeling and experimental approach to mechanically parameterize and simulate distinct molecular subtypes in preclinical models of cancer.

Carcinoma cells frequently expand and invade from a confined lesion, or multicellular clusters, into and through the stroma on the path to metastasis, often with an efficiency dictated by the architecture and composition of the microenvironment. Specifically, in desmoplastic carcinomas such as those of the breast, aligned collagen tracks provide contact guidance cues for directed cancer cell invasion. Yet, the evolving dynamics of this process of invasion remains poorly understood, in part due to difficulties in continuously capturing both spatial and temporal heterogeneity and progression to invasion in experimental systems. Therefore, to study the local invasion process from cell dense clusters into aligned collagen architectures found in solid tumors, we developed a novel engineered 3D invasion platform that integrates an aligned collagen matrix with a cell dense tumor-like plug. Using multiphoton microscopy and quantitative analysis of cell motility, we track the invasion of cancer cells from cell-dense bulk clusters into the pre-aligned 3D matrix, and define the temporal evolution of the advancing invasion fronts over several days. This enables us to identify and probe cell dynamics in key regions of interest: behind, at, and beyond the edge of the invading lesion at distinct time points. Analysis of single cell migration identifies significant spatial heterogeneity in migration behavior between cells in the highly cell-dense region behind the leading edge of the invasion front and cells at and beyond the leading edge. Moreover, temporal variations in motility and directionality are also observed between cells within the cell-dense tumor-like plug and the leading invasive edge as its boundary extends into the anisotropic collagen over time. Furthermore, experimental results combined with mathematical modeling demonstrate that in addition to contact guidance, physical crowding of cells is a key regulating factor orchestrating variability in single cell migration during invasion into anisotropic ECM. Thus, our novel platform enables us to capture spatio-temporal dynamics of cell behavior behind, at, and beyond the invasive front and reveals heterogeneous, local interactions that lead to the emergence and maintenance of the advancing front.