Browsing by Subject "Brain tumor"

Background: Radiomics is a process of converting medical images into high-dimensional quantitative features and the subsequent mining these features for providing decision support. It is conducted as a potential noninvasive, low-cost, and patient-specific routine clinical tool. Building a predictive model which is reliable, efficient, and accurate is a vital part for the success of radiomics. Machine learning method is a powerful tool to achieve this. Feature extraction strongly affects the performance. Delta-feature is one way of feature extraction methods to reflect the spatial variation in tumor phenotype, hence it could provide better treatment-specific assessment.

Purpose: To compare the performance of using pre-treatment features and delta-features for assessing the brain radiosurgery treatment response, and to investigate the performance of different combinations of machine learning methods for feature selection and for feature classification.

Materials and Methods: A cohort of 12 patients with brain treated by radiosurgery was included in this research. The pre-treatment, one-week post-treatment, and two-month post-treatment T1 and T2 FLAIR MR images were acquired. 61 radiomic features were extracted from the gross tumor volume (GTV) of each image. The delta-features from pre-treatment to two post-treatment time points were calculated. With leave-one-out sampling, pre-treatment features and the two sets of delta-features were separately input into a univariate Cox regression model and a machine learning model (L1-regularized logistic regression [L1-LR], random forest [RF] or neural network [NN]) for feature selection. Then a machine learning method (L1-LR, L2-regularized logistic regression [L2-LR], RF, NN, kernel support vector machine [Kernel-SVM], linear-SVM, or naïve bayes [NB]) was used to build a classification model to predict overall survival. The performance of each model combination and feature type was estimated by the area under receiver operating characteristic (ROC) curve (AUC).

Results: The AUC of one-week delta-features was significantly higher than that of pre-treatment features (p-values < 0.0001) and two-month delta-features (p-value= 0.000). The model combinations of L1-LR for feature selection and RF for classification as well as RF for feature selection and NB for classification based on one-week delta-features presented the highest AUC values (both AUC=0.944).

Conclusions: This work potentially implied that the delta-features could be better in predicting treatment response than pre-treatment features, and the time point of computing the delta-features was a vital factor in assessment performance. Analyzing delta-features using a suitable machine learning approach is potentially a powerful tool for assessing treatment response.

Malignant gliomas are the most common primary brain tumor in adults, with an incidence of five cases per 100,000 persons per year. Grade IV glioblastoma is the most aggressive form and prognosis remains poor despite the current gold-standard first-line treatment – maximal safe resection and combination of radiotherapy with temozolomide chemotherapy – resulting in a median survival of approximately 20 months. Tumor recurrence occurs in virtually all glioblastoma patients, and there currently exists no accepted treatment for these patients. Recent advances in novel directed therapeutics are showing efficacy and have entered clinical trials. This work spans the pre-clinical and clinical development of a bispecific antibody – EGFRvIII:CD3 bi-scFv – for the treatment of malignant gliomas.

Chapter 1 reviews current front-line immunotherapy research in the fields of antibodies, including BiTEs and checkpoint inhibitors, and tumor vaccinations, including peptide and dendritic cell vaccinations. Furthermore, challenges specific to high-grade gliomas as well as opportunities for combination therapies are discussed. Chapter 2 introduces the architecture of the novel bispecific antibody EGFRvIII:CD3 bi-scFv and provides an overview of the molecule’s efficacy in various models. EGFRvIII:CD3 bi-scFv is a truncated antibody with dual specificity. One arm targets the epidermal growth factor receptor mutation variant III (EGFRvIII), a tumor-specific antigen found on glioblastoma. The other arm targets the human CD3 receptor on T cells. As an obligate bispecific antibody, simultaneous binding of both receptors by multiple EGFRvIII:CD3 bi-scFv’s results in the crosslinking of CD3 receptor, activation of T cells, and release of perforin/granzyme which lyses the proximal EGFRvIII-expressing tumor cells. EGFRvIII:CD3 bi-scFv effectively treats orthotopic patient-derived malignant glioma and syngeneic glioblastoma.

Chapter 3 outlines the in-house development of a scalable clinical production process using a WAVE (GE) bioreactor and describes the cGMP-compliant clinical production of EGFRvIII:CD3 bi-scFv. The 250-liter cGMP-production run yielded more than four grams of clinical drug material.

Chapter 4 demonstrates that EGFRvIII:CD3 bi-scFv produced using the cGMP development process is efficacious in both in vitro and in vivo models of glioblastoma. The chapter also describes the approach used to calculate the starting dose for the upcoming first-in-human clinical trial. First-in-human clinical trials require careful selection of a safe yet biologically relevant starting dose. Typically, such starting doses are selected based on toxicity studies in a pharmacologically relevant animal model. However, with the advent of target-specific and highly active immunotherapeutics, both the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have provided guidance that recommend determining a safe starting dose based on a minimum anticipated biological effect level (MABEL) approach. In order to establish a first-in-human dose, as advised by the FDA for bispecific antibodies, this work uses a MABEL approach to select a safe starting dose for EGFRvIII:CD3 bi-scFv, based on a combination of in vitro data, in vivo animal studies, and theoretical human receptor occupancy modeling. Using the most conservative approach to the MABEL assessment, a dose of 57.4 ng EGFRvIII:CD3 bi-scFv/kg body weight was selected as a safe starting dose for a first-in-human clinical study.

Chapter 5 describes the pharmacokinetic properties of EGFRvIII:CD3 bi-scFv, a necessary step in the drug development process. Using microflow liquid chromatography coupled to high resolution parallel reaction monitoring mass spectrometry, and data analysis in Skyline, the chapter first describes the development of a bottom-up proteomic assay for quantification of EGFRvIII:CD3 bi-scFv in both plasma and whole blood. Importantly, a protein calibrator, along with stable isotope-labeled EGFRvIII:CD3 bi-scFv protein, was used for absolute quantification. A PK analysis in a CD3 humanized mouse revealed that EGFRvIII:CD3 bi-scFv in plasma and whole blood has an initial half-life of ~8 minutes and a terminal half-life of ~2.5 hours. These results establish a sensitive, high-throughput assay for direct quantification of EGFRvIII:CD3 bi-scFv without the need for immunoaffinity enrichment. Moreover, these pharmacokinetic parameters will guide drug optimization and dosing regimens in future IND-enabling and Phase I studies of EGFRvIII:CD3 bi-scFv.

Finally, Chapter 6 provides an outlook of the future development of cancer therapeutics for treating malignant gliomas.

Glioblastoma (GBM) is the most common and deadly primary malignant brain tumor. Despite an aggressive multimodal standard of care, prognoses and patient quality of life remain exceptionally poor, due in part to the non-specific and toxic nature of conventional treatment options. By contrast, adoptive cell transfer of T cells genetically modified to express tumor-specific chimeric antigen receptors (CARs) has emerged as a promising approach to targeting brain tumors, given that T cells have migratory capacity within the brain parenchyma, a mechanism to discriminate between normal and neoplastic tissue, and can develop immunological memory. This work spans the development of an effective CAR T-cell immunotherapy strategy targeting the tumor-specific driver mutation, EGFRvIII, which is expressed exclusively by GBM and other cancers but not normal tissue.

Chapters 1 and 2 provide an overview of GBM and the current clinical standard of care, the role of the immune system as it relates to the development and eradication of cancer, and an introduction to various immunotherapy platforms under active preclinical and clinical investigation. Chapter 3 details the historical context of adoptive T-cell immunotherapy and its evolution to present day, detailing our early proof-of-principle studies that led to the inception of the original research described herein. Data presented in Chapter 4 summarizes our translational objectives in implementing CAR T-cell immunotherapy clinically for patients with newly-diagnosed GBM. Chapter 5 addresses a perennial limitation to the immunotherapy of solid tumors by demonstrating an ability of modified CARs to circumvent intratumoral immunosuppression mediated by regulatory T cells. In Chapter 6, we present data that demonstrate, for the first time, a novel role for host lymphodepletion in cellular immunotherapy delivered directly into the brain. Lastly, Chapter 7 contains concluding remarks on the current state of CAR technology and important future directions.

In summary, our work here demonstrates that CAR T cell immunotherapy 1) has curative potential against highly established, orthotopic and syngeneic murine GBM, 2) can be strategically implemented within the current clinical treatment paradigm for GBM, and 3) can overcome a major mechanism of immunosuppression, demonstrating the versatility of gene-modified T cells for the treatment of malignant brain tumors. Together, these studies have paved way for the rationale design of two phase I clinical trials in patients with newly-diagnosed and recurrent EGFRvIII-positive GBM at Duke University.

Initiation and maintenance of a productive anti-tumor immune response requires a functional T-cell repertoire. Disruptions to T-cell function contribute to tumor immune escape, and to failure of the anti-tumor immune response in cancer patients. T-cell dysfunction is particularly severe in certain types of cancers such as glioblastoma (GBM), which is the most common primary malignant brain tumor in adults and is extremely lethal. Despite near universal confinement to the intracranial compartment, GBM frequently depletes both the number and function of systemic T-cells. A lack of understanding of the mechanisms underlying T-cell dysfunction poses challenges to the goal of developing appropriate and meaningful therapeutic platforms. Currently available treatments, including immunotherapies, for GBM and other intracranial diseases have proven ineffective in part because of underlying T-cell dysfunction. Thus, there is an unmet need for therapies that effectively address T-cell dysfunction. In this dissertation, we explore bone marrow T-cell sequestration, a novel mode of T-cell dysfunction present in GBM and other intracranial tumors.

Chapter 1 provides a comprehensive review of the epidemiology, clinical manifestation and diagnosis, and current standard of care for GBM. Chapter 2 outlines immunotherapeutic strategies under investigation for GBM. Chapter 3 describes the fading notion of traditional brain immune privilege but provides the current understanding of how the brain remains immunologically distinct. In Chapter 4, we explore bone marrow T-cell sequestration, and how this mechanism is usurped by GBM and other intracranial tumors to prevent anti-tumor efficacy of T-cell based immunotherapeutic modalities. In Chapter 5, we propose β-arrestin 2 (BARR2) depletion as a strategy to overcome bone marrow T-cell sequestration. In summary, this original work provides encouraging insights for the development of strategies to enhance anti-tumor efficacy of T-cell based immunotherapy for GBM, reversal of bone marrow T-cell sequestration.

Despite decades worth of research, glioblastoma remains one of the most lethal cancers. The identification of IDH1 as a major cancer gene in glioblastoma provides an exceptional opportunity for improving our understanding, diagnostics, and treatment of this disease. In addition to mutations in IDH1, recent studies from our laboratory have characterized the genetic landscape of gliomas and have shown the cooperation between IDH1 mutations and other oncogenic alterations such at TP53 mutations. Normally, IDH1 functions in the oxidative decarboxylation of isocitrate to α–ketoglutarate, however the mutant form confers neomorphic enzymatic activity by producing 2–hydroxyglutarate, an oncometabolite responsible for aberrant methylation in IDH1–mutated tumors, among other mutant IDH1–mediated phenotypes. To determine the role of mutant IDH1 in vivo, we generated a conditional knock–in mouse model. This genetically faithful system is both biologically and clinically relevant and will promote the understanding of mutant IDH1–mediated tumorigenesis while offering a route for therapeutic targeting.

We observed that broad expression of mutant IDH1 throughout the brain leads to hydrocephalus in 80% of animals. In assessing the earliest effects of mutant IDH1 on the brain, we determined mutant IDH1 confers a decrease in the proliferative cells of the subventricular zone of the lateral ventricle, the area which houses the neural stem cells in embryonic and adult animals. Additionally, a perturbation to the normal neural stem cell niche was observed in these animals. Combined, this data suggests that mutant IDH1 may be affecting the signaling pathways involved in differentiation in this population of cells. In vivo and in vitro studies will further elucidate mutant IDH1's effects on the differentiation patterns of neural stem cells expressing mutant IDH1.

To express mutant IDH1 in a more restricted manner and harness spatiotemporal control, we crossed mutant animals to a Nestin–CreERT2 strain of mouse that permits expression of floxed alleles upon treatment with tamoxifen. Animals were sacrificed at the onset of symptoms or at 1–year of age. We observed the development of both low– and high–grade gliomas in approximately 15–percent of E18.5 tamoxifen–treated animals. All tumors were found in a TP53–deleted background with mutant IDH1 being detected in only those tumors with the mutant allele. Lastly, to decrease the latency and increase the penetrance of tumor formation, an orthotopic intracranial injection model was generated to allow for visualization of tumor formation and development, as well as investigation of therapeutic modalities. The models generated and the knowledge gained from these studies will offer an understanding of the biological effects of the most common mutations found in the astrocytic subset of gliomas, bringing us strides closer to determining mechanisms and therapeutic targets for IDH1–mutated cancers.

During development, the precise regulation of the processes of proliferation, migration, and differentiation is required to establish proper organ structure and function and to prevent the deregulation that can lead to disease, such as cancer. Improved understanding of the signals that regulate these processes is therefore necessary to both gain insight into the mechanisms by which organ development proceeds and to identify strategies for treating the consequences of deregulation of these processes. In the cerebellum, some of the factors that regulate these processes have been identified but remain incompletely understood. Our studies have focused on the signals that regulate the migration of cerebellar granule neuron progenitors (GNPs) and the contribution of the SDF-1/CXCR4 signaling axis to postnatal cerebellar development. Using conditional knockout mice to delete CXCR4 specifically in GNPs, we show that loss of CXCR4 results in premature migration of a subset of GNPs throughout postnatal development that are capable of proliferation and survival outside of their normal mitogenic niche. Loss of CXCR4 also causes a reduction in the activity of the Sonic hedgehog (SHH) pathway (the primary mitogen for GNPs) but does not affect GNP proliferation, differentiation, or capacity for tumor formation. Our data suggest that while other factors likely contribute, SDF-1/CXCR4 signaling is necessary for proper migration of GNPs throughout cerebellar development.

In addition to understanding the signals that regulate normal development, the identification of vulnerabilities of established tumors is also necessary to improve cancer treatment. One strategy to improve treatment involves targeting the cells that are critical for maintaining tumor growth, known as tumor-propagating cells (TPCs). In the context of the cerebellar tumor medulloblastoma (MB), we have previously identified a population of TPCs in tumors from patched mutant mice that express the cell surface carbohydrate antigen CD15/SSEA-1. Here, we employed multiple approaches in an effort to target these cells, including a biochemical approach to identify molecules that carry the CD15 carbohydrate epitope as well as an immunotoxin approach to specifically target CD15-expressing cells. Unfortunately, these strategies were ultimately unsuccessful, but an alternative approach that recognized a vulnerability of CD15+ cells was identified. We show that CD15+ cells express elevated levels of genes associated with the G2/M phases of the cell cycle, progress more rapidly through the cell cycle than CD15- cells, and contain an increased proportion of cells in G2/M. Exposure of tumor cells to inhibitors of Aurora and Polo-like kinases, key regulators of G2/M, induces cell cycle arrest, apoptosis and enhanced sensitivity to conventional chemotherapy, and treatment of tumor-bearing mice with these agents significantly inhibits tumor progression. Importantly, cells from human patient-derived MB xenografts are also sensitive to Aurora and Polo-like kinase inhibitors. Our findings suggest that targeting G2/M regulators may represent a novel approach for the treatment of human MB.

SMARCAL1 (SWI/SNF Related, Matrix Associated, Actin Dependent Regulator Of Chromatin, Subfamily A-Like 1) is an ATP-dependent DNA-annealing helicase that reverses stalled replication forks. Its loss of function genetic alterations occurs in a subset of glioblastomas (GBMs) and has been found to be associated with alterative lengthening of telomeres (ALT+) in tumor cells. ALT tumors exploit homologous recombination to maintain telomere length and share common characteristics, including compromised telomere shelterin protein and increased replication stress in the telomere region.We established a SMARCAL1-null, ALT+ primary GBM culture. We show that the primary GBM culture displays stable ALT features and maintains mostly consistent karyotypes after their growth in mice. Transcriptomic profiling of the ALT+ primary GBM cells that have been propagated in vitro and those that have undergone propagation in vivo revealed the effects of microenvironments on the gene expression of these tumor cells. By using a doxycycline-inducible expression system, we show that the ALT+ features in the GBM primary culture can be turned off and on by the restoration or withdraw of exogenous wild-type SMARCAL1, but not its enzymatic dead mutant counterpart. Telomere pull down assays demonstrated the expression of SMARCAL1 effectively attenuates the process of ALT in tumor cells. In supporting the critical roles of ALT for tumor progression, induced restoration of wild-type SMARCAL1, but not its enzymatic dead mutant, effectively suppresses tumor progression in vivo. By taking advantage of our well characterized ALT model system, we are investigating the role of intrinsic DNA damage in tumorigenesis. Intrinsic DNA replicative stress occurs constantly in tumor cells. However, the pathogenic ramifications of replicative stress and the strategies cancer cells undertake to adapt remain to be fully defined. Here, we attempt to address these questions, using isogenic sarcoma and glioma cell line models differing in their intrinsic telomeric replicative stress levels, we show that intrinsic replicative stress promotes cancer stemness in human sarcoma and glioma cells. Further, molecular profiling analysis of human gliomas supports that human gliomas with higher levels of intrinsic replicative stress levels have increased stemness. We show this intrinsic replicative stress-stimulated stemness is accompanied by nonrandom segregation of chromosomes in mitotic cells. More notably, this nonrandom chromosome segregation is associated with asymmetric partition of CD133, a canonical marker for cancer cell stemness, in that the newly synthesized set of chromosomes are placed in one progeny cell while the set serving as templates for DNA replication turns to co-segregate with CD133 in another. We further reveal that this asymmetric co-segregation of chromosomes and CD133 depends on the Wnt/β-catenin signaling pathway. Collectively, these findings identify intrinsic DNA replicative stress as a driver of cancer cell stemness, and suggest a coordinated, Wnt/β-catenin signal-driven process of asymmetrically partitioning DNA and proteins in these cells, potentially as a way of maintaining cellular heterogeneity and population fitness in response to DNA damage. They also highlight and provide new insights into the roles of the Wnt/β-catenin pathway in maintaining tumor cells stemness, and suggest strategies for therapeutically targeting DNA damage-driven stemness in gliomas.