Radiomics of NSCLC: Quantitative CT Image Feature Characterization and Tumor Shrinkage Prediction

Radiomics is the high-throughput extraction and analysis of quantitative image features. For non-small cell lung cancer (NSCLC) patients, radiomics can be applied to standard of care computed tomography (CT) images to improve tumor diagnosis, staging, and response assessment. The first objective of this work was to show that CT image features extracted from pre-treatment NSCLC tumors could be used to predict tumor shrinkage in response to therapy. This is important since tumor shrinkage is an important cancer treatment endpoint that is correlated with probability of disease progression and overall survival. Accurate prediction of tumor shrinkage could also lead to individually customized treatment plans. To accomplish this objective, 64 stage NSCLC patients with similar treatments were all imaged using the same CT scanner and protocol. Quantitative image features were extracted and principal component regression with simulated annealing subset selection was used to predict shrinkage. Cross validation and permutation tests were used to validate the results. The optimal model gave a strong correlation between the observed and predicted shrinkages with . The second objective of this work was to identify sets of NSCLC CT image features that are reproducible, non-redundant, and informative across multiple machines. Feature sets with these qualities are needed for NSCLC radiomics models to be robust to machine variation and spurious correlation. To accomplish this objective, test-retest CT image pairs were obtained from 56 NSCLC patients imaged on three CT machines from two institutions. For each machine, quantitative image features with concordance correlation coefficient values greater than 0.90 were considered reproducible. Multi-machine reproducible feature sets were created by taking the intersection of individual machine reproducible feature sets. Redundant features were removed through hierarchical clustering. The findings showed that image feature reproducibility and redundancy depended on both the CT machine and the CT image type (average cine 4D-CT imaging vs. end-exhale cine 4D-CT imaging vs. helical inspiratory breath-hold 3D CT). For each image type, a set of cross-machine reproducible, non-redundant, and informative image features was identified. Compared to end-exhale 4D-CT and breath-hold 3D-CT, average 4D-CT derived image features showed superior multi-machine reproducibility and are the best candidates for clinical correlation.

CHARACTERIZATION OF TCL1-Tg:P53-/- MICE THAT RESEMBLE HUMAN CHRONIC LYMPHOCYTIC LEUKEMIA WITH 17P-DELETION

Chronic lymphocytic leukemia (CLL) is the most common adult leukemia in the United Statesand Europe. CLL patients with deletion of chromosome 17p, where the tumor suppressor p53 gene is located, often develop a more aggressive disease with poor clinical outcomes. However, the underlying mechanism remains unclear. In order to understand the underneath mechanism in vivo, I have recently generated mice with Eu-TCL1-Tg:p53-/- genotype and showed that these mice develop aggressive leukemia that resembles human CLL with 17p deletion. The Eu-TCL1-Tg:p53-/- mice developed CLL disease at 3-4 months, significantly earlier than the parental Eu-TCL1-Tg mice that developed CLL disease at 8-12 months. Flow cytometry analysis showed that the CD5+/ IgM+ cell population appeared in the peritoneal cavity, bone marrow, and the spleens of Eu-TCL1-Tg:p53-/- mice significantly earlier than that of the parental Eu-TCL1-Tg mice. Massive infiltration and accumulation of leukemia cells were found in the spleen and peritoneal cavity. In vitro study showed that the leukemia cells isolated from the Eu-TCL1-Tg:p53-/- mice were more resistant to fludarabine treatment than the leukemia cells isolated from spleens of Eu-TCL1-Tg mice. Interestingly, TUNEL assay revealed that there was higher apoptotic cell death found in the Eu-TCL1-Tg spleen tissue compared to the spleens of the Eu-TCL1-Tg:p53-/- mice, suggesting that the loss of p53 compromises the apoptotic process in vivo, and this might in part explain the drug resistant phenotype of CLL cells with 17p-deletion. In the present study, we further demonstrated that the p53 deficiency in the TCL1 transgenic mice resulted in significant down-regulation of microRNAs miR-15a and miR16-1, associated with a substantial up-regulation of Mcl-1, suggesting that the p53-miR15a/16-Mcl-1 axis may play an important role in CLL pathogenesis. Interestingly, we also found that loss of p53 resulted in a significant decrease in expression of the miR-30 family especially miR-30d in leukemia lymphocytes from the Eu-TCL1-Tg:p53-/- mice. Such down-regulation of those microRNAs and up-regulation of Mcl-1 were also found in primary leukemia cells from CLL patients with 17p deletion. To further exam the biological significance of decrease in the miR-30 family in CLL, we investigated the potential involvement of EZH2 (enhancer of zeste homolog 2), a component of the Polycomb repressive complex known to be a downstream target of miR-30d and plays a role in disease progression in several solid cancers. RT-PCR and western blot analyses showed that both EZH2 mRNA transcript and protein levels were significantly increased in the lymphocytes of Eu-TCL1-Tg:p53-/- mice relative to Eu-TCL1-Tg mice. Exposure of leukemia cells isolated from Eu-TCL1-Tg:p53-/- mice to the EZH2 inhibitor 3-deazaneplanocin (DZNep) led to induction of apoptosis, suggesting EZH2 may play a role in promoting CLL cell survival and this may contribute to the aggressive phenotype of CLL with loss of p53. Our study has created a novel CLL mouse model, and suggests that the p53/miR15a/16-Mcl-1 axis & p53/miR30d-EZH2 may contribute to the aggressive phenotype and drug resistance in CLL cells with loss of p53.

Galectin-3 Enhances the Malignant Melanoma Phenotype by Regulating Autotaxin

In melanoma patient specimens and cell lines, the over expression of galectin-3 is associated with disease progression and metastatic potential. Herein, we have sought out to determine whether galectin-3 affects the malignant melanoma phenotype by regulating downstream target genes. To that end, galectin-3 was stably silenced by utilizing the lentivirus-incorporated small hairpin RNA in two metastatic melanoma cell lines, WM2664 and A375SM, and subjected to gene expression microarray analysis. We identified and validated the lysophospholipase D enzyme, autotaxin, a promoter of migration, invasion, and tumorigenesis, to be down regulated after silencing galectin-3. Silencing galectin-3 significantly reduced the promoter activity of autotaxin. Interestingly, we also found the transcription factor NFAT1 to have reduced protein expression after silencing galectin-3. Electrophoretic mobility shift assays from previous reports have shown that NFAT1 binds to the autotaxin promoter in two locations. ChIP analysis was performed, and we observed a complete loss of bound NFAT1 to the autotaxin promoter after silencing galectin-3 in melanoma cells. Mutation of the NFAT1 binding sites at either location reduces autotaxin promoter activity. Silencing NFAT1 reduces autotaxin expression while over expressing NFAT1 in NFAT1 negative SB-2 melanoma cells induces autotaxin expression. These data suggest that galectin-3 silencing reduces autotaxin transcription by reducing the amount of NFAT1 protein expression. Rescue of galectin-3 rescues both NFAT1 and autotaxin. We also show that the re-expression of autotaxin in galectin-3 shRNA melanoma cells rescues the angiogenic phenotype in vivo. Furthermore, we identify NFAT1 as a potent inducer of tumor growth and experimental lung metastasis. Our data elucidate a previously unidentified mechanism by which galectin-3 regulates autotaxin and assign a novel role for NFAT1 during melanoma progression.

Developing the follow-up schedules for women with breast cancer

Background: The follow-up care for women with breast cancer requires an understanding of disease recurrence patterns and the follow-up visit schedule should be determined according to the times when the recurrence are most likely to occur, so that preventive measure can be taken to avoid or minimize the recurrence. Objective: To model breast cancer recurrence through stochastic process with an aim to generate a hazard function for determining a follow-up schedule. Methods: We modeled the process of disease progression as the time transformed Weiner process and the first-hitting-time was used as an approximation of the true failure time. The women's "recurrence-free survival time" or a "not having the recurrence event" is modeled by the time it takes Weiner process to cross a threshold value which represents a woman experiences breast cancer recurrence event. We explored threshold regression model which takes account of covariates that contributed to the prognosis of breast cancer following development of the first-hitting time model. Using real data from SEER-Medicare, we proposed models of follow-up visits schedule on the basis of constant probability of disease recurrence between consecutive visits. Results: We demonstrated that the threshold regression based on first-hitting-time modeling approach can provide useful predictive information about breast cancer recurrence. Our results suggest the surveillance and follow-up schedule can be determined for women based on their prognostic factors such as tumor stage and others. Women with early stage of disease may be seen less frequently for follow-up visits than those women with locally advanced stages. Our results from SEER-Medicare data support the idea of risk-controlled follow-up strategies for groups of women. Conclusion: The methodology we proposed in this study allows one to determine individual follow-up scheduling based on a parametric hazard function that incorporates known prognostic factors.^

PRKCa: Identification of a Novel Downstream Target of WT1

Wilms tumor is a childhood tumor of the kidney arising from the undifferentiated metanephric mesenchyme. Tumorigenesis is attributed to a number of genetic and epigenetic alterations. In 20% of Wilms tumors, Wilms tumor gene 1 (WT1) undergoes inactivating homozygous mutations causing loss of function of the zinc finger transcription factor it encodes. It is hypothesized that mutations in WT1 result in dysregulation of downstream target genes, leading to aberrant kidney development and/or Wilms tumor. These downstream target genes are largely unknown, and identification is important for further understanding Wilms tumor development. Heatmap data of human Wilms tumor protein expression, generated by reverse phase protein assay analysis (RPPA), show significant correlation between WT1 mutation status and low PRKCα expression (p= 0.00013); additionally, p-PRKCα (S657) also shows decreased expression in these samples (p= 0.00373). These data suggest that the WT1 transcription factor regulates PRKCα expression, and that PRKCα plays a potential role in Wilms tumor tumorigenesis. We hypothesize that the WT1 transcription factor directly/indirectly regulates PRKCα and mutations occurring in WT1 lead to decreased expression of PRKCα. Prkcα and Wt1 have been shown to co-localize in E14.5 mesenchymal cells of the developing kidney. siRNA knockdown, in-vivo ablation, and tet-inducible expression of Wt1 each independently confirm regulation of Prkcα expression by Wt1 at both RNA and protein levels, and investigation into possible WT1 binding sites in PRKCα regulatory regions has identified multiple sites to be confirmed by luciferase reporter constructs. With the goal of identifying WT1 and PRKCα downstream targets, RPPA analysis of protein expression in mesenchymal cell culture, following lentiviral delivered shRNA knockdown of Wt1 and shRNA knockdown of Prkcα, will be carried out. Apart from Wilms tumor, WT1 also plays an important role in Acute Myeloid Leukemia (AML). WT1 mutation status has been implicated, controversially, as an independent poor-prognosis factor in leukemia, leading to decreased probability of overall survival, complete remission, and disease free survival. RPPA analysis of AML patient samples showed significant decreases in PRKCα/p-PRKCα protein expression in a subset of patients (Kornblau, personal communication); therefore, the possible role of WT1 and PRKCα in leukemia disease progression is an additional focus of this study. WT1 mutation analysis of diploid leukemia patient samples revealed two patients with mutations predicted to affect WT1 activity; of these two samples, only one corresponded to the low PRKCα expression cohort. Further characterization of the role of WT1 in AML, and further understanding of WT1 regulated PRKCα expression, will be gained following RPPA analysis of protein expression in HL60 leukemia cell lines with lentiviral delivered shRNA knockdown of WT1 and shRNA knockdown of PRKCα.

Clinical, microbiological, and immunological markers of gingival and periodontal disease induction and progression

The potential impact of periodontal disease, a suspected risk factor for systemic diseases, presents challenges for health promotion and disease prevention strategies. This study examined clinical, microbiological, and immunological factors in a disease model to identify potential biomarkers that may be useful in predicting the onset and severity of both inflammatory and destructive periodontal disease. This project used an historical cohort design based on data obtained from 47 adult, female nonhuman primates followed over a 6-year period for 5 unique projects where the ligature-induced model of periodontitis was utilized. Standardization of protocols for sample collection allowed for comparison over time. Bleeding and pocket depth measures were selected as the dependent variables of relevance to humans based upon the literature and historical observations. Exposure variables included supragingival plaque, attachment level, total bacteria, black-pigmented bacteria, Gram-negative and Gram-positive bacteria, total IgG and IgA in crevicular fluid, specific IgG antibody in both crevicular fluid and serum, and IgG antibody to four select pathogenic microorganisms. Three approaches were used to analyze the data from this study. The first approach tested for differences in the means of the response variables within the group and among longitudinal observations within the group at each time point. The second approach examined the relationship among the clinical, microbiological, and immunological variables using correlation coefficients and stratified analyses. Multivariable models using GEE for repeated measures were produced as a predictive description of the induction and progression of gingivitis and periodontal disease. The multivariable models for bleeding (gingivitis) include supragingival plaque, total bacteria and total IgG while the second also contains supragingival plaque, Gram-positive bacteria, and total IgG. Two multivariable models emerged for periodontal disease. One multivariable model contains plaque, total bacteria, total IgG and attachment level. The second model includes black-pigmented bacteria, total bacteria, antibody to Campylobacter rectus, and attachment level. Utilization of the nonhuman primate model to prospectively examine causal hypotheses can provide a focus for human research on the mechanisms of progression from health to gingivitis to periodontitis. Ultimately, causal theories can guide strategies to prevent disease initiation and reduce disease severity. ^

14-3-3 zeta overexpression in early stage breast disease and tumor progression

Recent progress in diagnostic tools allows many breast cancers to be detected at an early pre-invasive stage. Thus, a better understanding of the molecular basis of early breast cancer progression is essential. 14-3-3 is a family of highly conserved and ubiquitously expressed proteins that are expressed in all eukaryotic organisms. In mammals there are seven isoforms, which bind to phosphor-serine/threonine residues regulating essential cellular processes such as signal transduction, cell cycle progression, and apoptosis. Our laboratory has discovered that a particular 14-3-3 family member, Zeta, is overexpressed in over 40% of breast tumor tissues. Furthermore, I examined the stage of breast disease in which 14-3-3ζ overexpression occurs and found that increased expression of 14-3-3ζ begins at the stage of atypical ductal hyperplasia, a very early stage of breast disease that confers increased risk for progress toward breast cancer. To determine whether 14-3-3ζ overexpression is a decisive early event in breast cancer, I overexpressed 14-3-3ζ in MCF10A cells, a non-transformed mammary epithelial cell (MEC) line and examined its impact on acini formation in a three dimensional (3D) culture model which simulates a basic unit of structure in the mammary gland. I discovered that 14-3-3ζ overexpression severely disrupted the acini architecture resulting in the disruption of polarity and luminal filling. Both are critical morphological events in the pre-neoplastic breast disease. This thesis focuses on the molecular mechanism of luminal filling. Proper lumen formation is a result of anoikis, a specific type apoptosis of cells not attached to the basement membrane. I found that 14-3-3ζ overexpression conferred a resistance to anoikis. Additionally, 14-3-3ζ overexpression in MCF10A cells and in MECs from 14-3-3ζ transgenic mice reduced expression of p53, which is known to mediate anoikis. Mechanistically, 14-3-3ζ induced hyperactivation of the PI3K/Akt pathway which led to phosphorylation and translocation of the MDM2 to the nucleus resulting in increased p53 degradation. Ectopic expression of p53 restored luminal apoptosis in 14-3-3ζ overexpressing MCF10A acini in 3D cultures. These data suggest that 14-3-3ζ overexpression is a critical event in early breast disease and down-regulation of p53 is one of the mechanisms by which 14-3-3ζ alters MEC acini structure and may increase the risk of progression to breast cancer. ^