Hawkins-Daarud, Andrea

Abstract Intratumoral heterogeneity poses a significant challenge to the diagnosis and treatment of glioblastoma (GBM). This heterogeneity is further exacerbated during GBM recurrence, as treatment-induced reactive changes produce additional intratumoral heterogeneity that is ambiguous to differentiate on clinical imaging. There is an urgent need to develop non-invasive approaches to map the heterogeneous landscape of histopathological alterations throughout the entire lesion for each patient. We propose to predictively fuse Magnetic Resonance Imaging (MRI) with the underlying intratumoral heterogeneity in recurrent GBM using machine learning (ML) by leveraging image-localized biopsies with their associated locoregional MRI features. To this end, we develop BioNet, a biologically-informed neural network model, to predict regional distributions of three tissue-specific gene modules: proliferating tumor, reactive/inflammatory cells, and infiltrated brain tissue. BioNet offers valuable insights into the integration of multiple implicit and qualitative biological domain knowledge, which are challenging to describe in mathematical formulations. BioNet performs significantly better than a range of existing methods on cross-validation and blind test datasets. Voxel-level prediction maps of the gene modules by BioNet help reveal intratumoral heterogeneity, which can improve surgical targeting of confirmatory biopsies and evaluation of neuro-oncological treatment effectiveness. The non-invasive nature of the approach can potentially facilitate regular monitoring of the gene modules over time, and making timely therapeutic adjustment. These results also highlight the emerging role of ML in precision medicine.

Tumor recurrence in glioblastoma multiforme (GBM) is often attributed to acquired resistance to the standard chemotherapeutic agent, temozolomide (TMZ). Promoter methylation of the DNA repair gene MGMT (O6-methylguanine–DNA methyltransferase) has been associated with sensitivity to TMZ, whereas increased expression of MGMT has been associated with TMZ resistance. Clinical studies have observed a downward shift in MGMT methylation percentage from primary to recurrent stage tumors; however, the evolutionary processes that drive this shift and more generally the emergence and growth of TMZ-resistant tumor subpopulations are still poorly understood. Here, we develop a mathematical model, parameterized using clinical and experimental data, to investigate the role of MGMT methylation in TMZ resistance during the standard treatment regimen for GBM—surgery, chemotherapy, and radiation. We first found that the observed downward shift in MGMT promoter methylation status between detection and recurrence cannot be explained solely by evolutionary selection. Next, our model suggests that TMZ has an inhibitory effect on maintenance methylation of MGMT after cell division. Finally, incorporating this inhibitory effect, we study the optimal number of TMZ doses per adjuvant cycle for patients with GBM with high and low levels of MGMT methylation at diagnosis.

Purpose Glioblastomas, lethal primary brain tumors, are known for their heterogeneity and invasiveness. A growing body of literature has been developed demonstrating the clinical relevance of a biomathematical model, the proliferation-invasion model, of glioblastoma growth. Of interest here is the development of a treatment response metric, days gained (DG). This metric is based on individual tumor kinetics estimated through segmented volumes of hyperintense regions on T1-weighted gadolinium-enhanced and T2-weighted magnetic resonance images. This metric was shown to be prognostic of time to progression. Furthermore, it was shown to be more prognostic of outcome than standard response metrics. Although promising, the original article did not account for uncertainty in the calculation of the DG metric, leaving the robustness of this cutoff in question. Methods We harnessed the Bayesian framework to consider the impact of two sources of uncertainty: (1) image acquisition and (2) interobserver error in image segmentation. We first used synthetic data to characterize what nonerror variants are influencing the final uncertainty in the DG metric. We then considered the original patient cohort to investigate clinical patterns of uncertainty and to determine how robust this metric is for predicting time to progression and overall survival. Results Our results indicate that the key clinical variants are the time between pretreatment images and the underlying tumor growth kinetics, matching our observations in the clinical cohort. Finally, we demonstrated that for this cohort, there was a continuous range of cutoffs between 94 and 105 for which the prediction of the time to progression was over 80% reliable. Conclusion Although additional validation must be performed, this work represents a key step in ascertaining the clinical utility of this metric.

PurposeDespite the intra- and intertumoral heterogeneity seen in glioblastoma multiforme (GBM), there is little definitive data on the underlying cause of the differences in patient survivals. Serial imaging assessment of tumor growth allows quantification of tumor growth kinetics (TGK) measured in terms of changes in the velocity of radial expansion seen on imaging. Because a systematic study of this entire TGK phenotype—growth before treatment and during each treatment to recurrence —has never been coordinately studied in GBMs, we sought to identify whether patients cluster into discrete groups on the basis of their TGK.Patients and MethodsFrom our multi-institutional database, we identified 48 patients who underwent maximally safe resection followed by radiotherapy with imaging follow-up through the time of recurrence. The patients were then clustered into two groups through a k-means algorithm taking as input only the TGK before and during treatment.ResultsThere was a significant survival difference between the clusters ( P = .003). Paradoxically, patients among the long-lived cluster had significantly larger tumors at diagnosis ( P = .027) and faster growth before treatment ( P = .003) but demonstrated a better response to adjuvant chemotherapy ( P = .048). A predictive model was built to identify which cluster patients would likely fall into on the basis of information that would be available to clinicians immediately after radiotherapy (accuracy, 90.3%).ConclusionDichotomizing the heterogeneity of GBMs into two populations—one faster growing yet more responsive with increased survival and one slower growing yet less responsive with shorter survival—suggests that many patients who receive standard-of-care treatments may get better benefit from select alternative treatments.

Microglia are a major cellular component of gliomas, and abundant in the centre of the tumour and at the infiltrative margins. While glioma is a notoriously infiltrative disease, the dynamics of microglia and glioma migratory patterns have not been well characterized. To investigate the migratory behaviour of microglia and glioma cells at the infiltrative edge, we performed two-colour time-lapse fluorescence microscopy of brain slices generated from a platelet-derived growth factor-B (PDGFB)-driven rat model of glioma, in which glioma cells and microglia were each labelled with one of two different fluorescent markers. We used mathematical techniques to analyse glioma cells and microglia motility with both single cell tracking and particle image velocimetry (PIV). Our results show microglia motility is strongly correlated with the presence of glioma, while the correlation of the speeds of glioma cells and microglia was variable and weak. Additionally, we showed that microglia and glioma cells exhibit different types of diffusive migratory behaviour. Microglia movement fit a simple random walk, while glioma cell movement fits a super diffusion pattern. These results show that glioma cells stimulate microglia motility at the infiltrative margins, creating a correlation between the spatial distribution of glioma cells and the pattern of microglia motility.