Melamed Udel 0060D 1

TARGETING HEDGEHOG SIGNALING WITH

POLYCATION-SPHERICAL NUCLEIC ACID NANOPARTICLES

Jilian R. Melamed

A dissertation submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Biomedical Engineering

Fall 2018

TARGETING HEDGEHOG SIGNALING WITH

POLYCATION-SPHERICAL NUCLEIC ACID NANOPARTICLES

FOR GLIOBLASTOMA THERAPY

Jilian R. Melamed

Approved: ______Dawn Elliott, Ph.D. Chair of the Department of Biomedical Engineering

Approved: ______Babatunde Ogunnaike, Ph.D. Dean of the College of Engineering

Approved: ______Douglas J. Doren, Ph.D. Interim Vice Provost for Graduate and Professional Education

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Emily Day, Ph.D. Professor in charge of dissertation

Signed: ______Millicent Sullivan, Ph.D. Member of dissertation committee

Signed: ______Christopher Kloxin, Ph.D. Member of dissertation committee

Signed: ______Jennifer Sims-Mourtada, Ph.D. Member of dissertation committee

ACKNOWLEDGMENTS

First, thank you to my advisor, Dr. Emily Day, for your support and guidance over the years. Thank you for pushing me to grow into the knowledgeable scientist I am today! I am truly grateful for your mentorship, your scientific and non-scientific wisdom, and for all of the time and resources you have invested in me. Thank you to my committee, Dr. Millicent Sullivan, Dr. Jennifer Sims-Mourtada, and Dr. Christopher Kloxin, for your support and helpful insights over the years. Next, thank you to the Day Lab for your support and camaraderie throughout this journey: Rachel Riley, Danielle Valcourt, Jenna Harris, Megan Dang, Benjamin Luo, Ritu Goyal, Chintan Kapadia, and Jianxin Wang. You have all been a joy to work with, and it’s been amazing to watch this lab grow from two first year graduate students to the productive research machine that it is today. Special thanks to the talented cohort of undergraduates I’ve had the privilege of mentoring: Brittany Fay, Nicole Kreuzberger, Stephen Ioele, Ariel Hannum, and Violet Ullman. I’m incredibly proud of all of you, your research accomplishments, and the scientists you have become. Thank you for dedicating your time and your talents to research; I appreciate your contributions tremendously! I would like to acknowledge the Bioimaging Staff at Delaware Biotechnology Institute for training me on various microscopy techniques. In particular, thank you to Michael Moore for sharing your extensive microscopy knowledge with me. Thank you to Jeff Caplan, Sylvain Le Marchand, and Shannon Modla for all of your microscopy assistance and knowledge.

iv Special thanks to Joshua Morgan for your endless wisdom and mentoring. I cannot begin to describe how grateful I am for your insights on experiments and troubleshooting, your scientific and professional advice, your time spent helping me learn various experimental and analysis techniques, and most importantly your friendship. I would not be the scientist I am today without you! Next, I’d like to thank my friends and family for being some of my fiercest supporters over the years. Thank you to all of my UD BME friends for the many laughs, adventures, and conversations. To my Statesmen crew: Kyle Trask, Joshua Yost, Tom Alderson, and Rachael Piorko, thank you for your love and support throughout this adventure. Thank you to my wonderful parents for your unwavering encouragement, and thank you to Tess, Matt, Addie, and Lily for being my best friends. I’m so grateful for all of you and your support throughout my PhD and for years to come. Finally, I’d like to acknowledge financial support from the National Defense Science and Engineering Fellowship Program.

v DEDICATION

To my parents, for their unconditional love and support. And to Christie Rose Shimak, whose fight against brain cancer inspired me to pursue a career in biomedical research.

vi TABLE OF CONTENTS

LIST OF FIGURES ...... xii ABSTRACT ...... xxi

Chapter

1 INTRODUCTION ...... 1

1.1 A Brief Introduction to Cancer Biology and Glioblastoma ...... 3

1.1.1 Tumorigenesis and the Propagation of Cancer ...... 3 1.1.2 Introduction to Glioblastoma ...... 6 1.1.3 Current Treatment Strategies for Glioblastoma ...... 9 1.1.4 Glioblastoma Stem Cells and Hedgehog Signaling as a Therapeutic Target ...... 10

1.2 Nanotechnology in Glioblastoma Therapy ...... 15

1.2.1 Challenges and Opportunities for Nanotechnology in Glioblastoma Therapy ...... 17 1.2.2 Nanoparticles for Drug Delivery to Glioblastoma Tumors ...... 19 1.2.3 Nanoparticles as Vectors for Gene Therapy ...... 21

1.3 Nanotechnology in RNA Interference ...... 23

1.3.1 RNA Interference and Rational Design of Nanoparticles ...... 24 1.3.2 Lipid-Based Nanoparticles for siRNA Delivery ...... 29 1.3.3 Polymer-Based Nanoparticles for siRNA Delivery ...... 30 1.3.4 Spherical Nucleic Acid Nanoparticles for siRNA Delivery ...... 31

1.4 Conclusions: Opportunities to Advance Glioblastoma Therapy with Hedgehog-Targeted Polycation-Spherical Nucleic Acid Nanoparticles . 36

2 METHODS FOR NANOPARTICLE SYNTHESIS AND CHARACTERIZATION ...... 38

2.1 Introduction ...... 38 2.2 Preparation of Gold Nanoparticles ...... 38

vii 2.2.1 Frens Method for Gold Nanoparticle Synthesis ...... 39 2.2.2 RNase Deactivation ...... 40

2.3 Synthesis of Gold-Based Spherical Nucleic Acid Nanoparticles ...... 40

2.3.1 siRNA Duplex Preparation ...... 40 2.3.2 Gold Nanoparticle Functionalization with siRNA and Polyethylene Glycol ...... 40

2.4 Coating Spherical Nucleic Acids with Polyethylenimine ...... 41 2.5 Characterization Methods ...... 42

2.5.1 Electron Microscopy ...... 42 2.5.2 UV-Visible Spectroscopy ...... 43 2.5.3 Dynamic Light Scattering and Zeta Potential Analysis ...... 43 2.5.4 Quantifying siRNA Loading onto Particles ...... 44 2.5.5 Quantifying Polyethylenimine Loading onto Particles ...... 46

2.6 Conclusions ...... 46

3 EVALUATION OF HEDGEHOG SIGNALING AS A THERAPEUTIC TARGET FOR GLIOBLASTOMA ...... 48

3.1 Introduction ...... 48 3.2 Materials and Methods ...... 52

3.2.1 Cell Culture and Transient Transfections ...... 52 3.2.2 Immunofluorescent Staining and Image Analysis ...... 52 3.2.3 Western Blotting ...... 54 3.2.4 Senescence Analysis ...... 55 3.2.5 Dose Response, Viability Assay, and Synergy Assessment ...... 55 3.2.6 Assessment of Drug Efflux Transporter Activity ...... 56 3.2.7 Proliferation Assay ...... 57 3.2.8 Neurosphere Growth and Apoptosis Analysis ...... 57 3.2.9 Quantitative Real-Time Polymerase Chain Reaction (qPCR) ..... 59 3.2.10 Statistical Analysis ...... 59

3.3 Results and Discussion ...... 60

3.3.1 U87 and T98G GBM Cells Exhibit Active Hh Signaling Required for Proliferation ...... 60 3.3.2 Silencing Gli1 Potentiates GBM Cell Response to TMZ by Decreasing Multidrug Efflux Activity ...... 62

viii 3.3.3 Suppressing Hh Signaling Modulates p53 and MGMT Expression in GBM Cells ...... 65 3.3.4 Silencing Gli1 Does Not Induce Apoptosis in GBM Cells without TMZ Co-Treatment ...... 68 3.3.5 Silencing Gli1 Induces Senescence in GBM Cells in a Manner Dependent on the Absence of PTEN ...... 71 3.3.6 Hh Inhibition and TMZ Co-Treatment Promotes Apoptosis in Neurospheres and Impairs Neurosphere Growth ...... 73 3.3.7 Discussion ...... 77

3.4 Conclusions ...... 83

4 COMPARISON OF POLYETHYLENIMINE-SPHERICAL NUCLEIC ACID NANOPARTICLES VERSUS POLYPLEXES FOR GENE REGULATION ...... 84

4.1 Introduction ...... 84 4.2 Materials and Methods ...... 87

4.2.1 SNA and PEI-SNA Synthesis and Characterization ...... 87 4.2.2 PEI-siRNA Polyplex Synthesis ...... 87 4.2.3 Particle Characterization and Evaluation of Serum Stability ...... 87 4.2.4 Cell Culture and Stable Gene Expression ...... 88 4.2.5 Cellular Uptake Analysis ...... 89 4.2.6 Lysosomal Trafficking and Imaging ...... 89 4.2.7 siRNA and PEI Co-Trafficking Analysis ...... 90 4.2.8 Toxicity Assessment ...... 91 4.2.9 Gene Knockdown Assessment ...... 91 4.2.10 Statistical Analysis ...... 92

4.3 Results and Discussion ...... 92

4.3.1 Nanoparticle and Polyplex Synthesis and Characterization ...... 92 4.3.2 Evaluation of PEI-SNA and Polyplex Serum Stability ...... 96 4.3.3 PEI-SNAs Undergo Enhanced Cellular Uptake Relative to PEI-siRNA Polyplexes or SNAs ...... 97 4.3.4 PEI-SNAs and Polyplexes Remain Intact Following Endocytosis ...... 99 4.3.5 PEI-SNAs Exhibit Decreased Lysosomal Accumulation Relative to PEI-siRNA Polyplexes ...... 101 4.3.6 PEI-SNAs Mediate GFP Silencing at Dramatically Lower siRNA Doses than PEI-siRNA Polyplexes ...... 104 4.3.7 PEI-SNAs Improve Polycation Cytocompatibility Relative to Polyplexes ...... 107

ix 4.3.8 Discussion ...... 110

4.4 Conclusions ...... 117

5 EVALUATION OF GLI1-TARGETED POLYETHYLENIMINE- SPHERICAL NUCLEIC ACID NANOPARTICLES AGAINST GLIOBLASTOMA CELLS ...... 119

5.1 Introduction ...... 119 5.2 Materials and Methods ...... 123

5.2.1 Nanoparticle Synthesis and Characterization ...... 123 5.2.2 Cell Culture and Stable Gene Expression ...... 124 5.2.3 Assessment of Endocytosis Pathways ...... 125 5.2.4 Confocal Microscopy and Image Analysis ...... 125 5.2.5 Gene Expression Analysis by qPCR ...... 126 5.2.6 Evaluation of Cell Proliferation, Senescence, and Viability ..... 126 5.2.7 Assessment of Self-Renewal and Neurosphere TMZ Response 127 5.2.8 Statistical Analysis ...... 128

5.3 Results and Discussion ...... 128

5.3.1 Evaluation of PEI-SNA Endocytosis Mechanism ...... 128 5.3.2 Evaluation of PEI-SNA Intracellular Trafficking ...... 130 5.3.3 Gli1 PEI-SNAs Regulate the Expression of Hh Signaling Components and Downstream Target Genes ...... 133 5.3.4 Gli1 PEI-SNAs Slow Proliferation, Induce Senescence, and Reduce Chemoresistance of GBM Cells ...... 134 5.3.5 Gli1 PEI-SNAs Reduce Stemness and Impair Self-Renewal of GBM Cells ...... 138 5.3.6 Gli1 PEI-SNAs Potentiate Neurosphere Response to TMZ Chemotherapy ...... 141

5.4 Conclusions ...... 144

6 CONCLUSIONS AND FUTURE DIRECTIONS ...... 145

6.1 Introduction ...... 145 6.2 Hh/Gli1 Signaling as a Target for GBM Therapy ...... 145 6.3 Polycation-Spherical Nucleic Acid Nanoparticles as RNAi Therapeutics ...... 147 6.4 Future Directions ...... 149

REFERENCES ...... 154

x Appendix

JOURNAL PERMISSIONS ...... 182

Permission from Oncotarget ...... 182 Permission from Elsevier ...... 182 Permission from Molecular Pharmaceutics ...... 183

xi LIST OF FIGURES

Figure 1.1. Current models for tumorigenesis. Modified from Bradshaw et al. (2016) Front Surg [10]...... 4

Figure 1.2. The hallmarks of cancer. Reproduced from Hanahan and Weinberg. (2000) Cell [8]...... 6

Figure 1.3. GBM tumors are characterized by rapid, infiltrative growth, inflammation, and a necrotic core. Reproduced from (left) Brat and Van Meir. (2004) Lab Invest [24], (right) Lombardi and Assem. (2016) Glioblastoma [20]...... 7

Figure 1.4. Cancer stem cells as drivers of tumor progression. Reproduced from Lathia et al. (2015) Genes Dev [47]...... 11

Figure 1.5. Schematic depicting Hedgehog signaling...... 13

Figure 1.6. Nanoparticles can enable multiple functionalities from a single construct, including the controlled release of drugs, presentation of nucleic acids and antibodies for targeting or therapy, and stealth agents to reduce immune clearance and increase circulation time...... 16

Figure 1.7. Tumor vasculature is characterized by chaotic, leaky morphology (top). Reproduced from Jain et al. (2005) Science [78]. Leaky tumor vasculature allows passive accumulation of nanoparticles within a tumor, while active targeting can improve intratumoral nanoparticle retention (bottom). Reproduced from Dong and Mumper (2010) Nanomedicine [77]...... 17

Figure 1.8. Structure of the blood-brain barrier. Adapted from Glaser et al. (2017) Front. Pharmacol [79]...... 19

Figure 1.9. RNA interference allows specific silencing of the targeted gene. Reproduced from Crooke. (2004) Curr Mol Med [101]...... 25

Figure 1.10. Barriers to siRNA delivery. Modified from Whitehead et al. (2009) Nat Rev Drug Discov [105]...... 27

xii Figure 1.11. Endocytosed siRNA nanocarriers must avoid retention within the endolysosomal system to effectively deliver their cargo to RISC in the cytosol. Modified from Wittrup et al. (2015) Nat Biotechnol [116]...... 29

Figure 1.12. SNAs consist of an inner core material densely coated with radially- oriented nucleic acids. Reproduced from Kapadia, Melamed, Day (2018) BioDrugs [135], https://mirkin- group.northwestern.edu/project/spherical-nucleic-acids/...... 33

Figure 1.13. SNAs accumulate within cells without the need for auxiliary transfection agents, unlike bare nucleic acids. Reproduced from Choi et al. (2013) PNAS [146]...... 34

Figure 1.14. SNAs exhibit efficient BBB penetration using an in vitro BBB model. A) Schematic depicting noncontact in vitro BBB model consisting of co-cultured BMECs and astrocytes. SNAs were introduced to the luminal side of BMECs and allowed to measure accumulation within astrocytes, requiring passage through the endothelial/transwell barrier. B) Cy5-SNAs accumulate with BMECs expressing vimentin and occluding, characteristic of the BBB in vivo. C) Cy5-SNAs accumulate within GFAP-expressing astrocytes after successfully traversing the BMEC/transwell barrier. Reproduced from Jensen et al. (2013) Sci Trans Med [143]...... 35

Figure 1.15. Endocytosed SNAs accumulate primarily within late endosomes. Modified from Wu et al. (2014) JACS [147]...... 36

Figure 2.1. The Frens method for gold nanoparticle synthesis occurs as a three-step process: reduction, nucleation, and growth...... 39

Figure 2.2. Schematic representing siRNA conjugation to gold nanoparticle surfaces. Reproduced from Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153]...... 41

Figure 2.3. TEM images showing bare 15 nm AuNPs (left) and SNAs counterstained with uranyl acetate to visualize the siRNA shell (right). 42

Figure 2.4. UV-visible spectroscopy characterizing AuNPs, SNAs, and PEI-SNAs. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 43

Figure 2.5. Schematic depicting an OliGreen assay to quantify siRNA loading onto gold nanoparticle surfaces. Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153]...... 45

xiii Figure 2.6. Representative antisense and sense strand loading into 15 nm AuNPs. Reproduced from Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153]...... 46

Figure 3.1: U87-MG and T98G GBM cells exhibit active Hh signaling via Gli1. rhShh increases Gli1 expression and nuclear translocation in (A) U87-MG but not (B) T98G GBM cells by immunofluorescence. Scale bars = 100 μm. (C) Quantitative image analysis reveals that U87-MG Gli1 intensity is significantly increased by ~30% in the nucleus and by ~40% in the cytoplasm relative to that in control cells. Data are shown as mean ± standard deviation from 3 independent experiments, *p < 0.05 by Student’s t-test relative to control. Changes in T98G Gli1 intensity are insignificant. # no significance by Student’s t-test. (D) By EdU incorporation and flow cytometry analysis, silencing GLI1 decreases the proliferation of U87-MG and T98G cells by ~60% and ~44%, respectively, relative to siScr. Data are shown as mean ± standard deviation, *p = 0.03, **p = 0.002 by paired t-test relative to control. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 61

Figure 3.2: Silencing Gli1 influences GBM cell response to TMZ. By AlamarBlue assay, U87-MG and T98G metabolic activity is significantly decreased with combined Gli1 silencing and low-dose TMZ treatment. Data are shown as mean ± standard deviation, *p < 0.01 by one-way ANOVA with post-hoc Tukey. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 63

Figure 3.3 Assessment of additive or synergistic effects between Gli1 silencing and TMZ treatment. *p<0.05 by one-way ANOVA and post-hoc Tukey. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 64

Figure 3.4: Silencing Gli1 influences reduces multidrug efflux activity of GBM cells. Silencing Gli1 reduces multidrug efflux activity in both U87- MG and T98G cells. Flow cytometry reveals that silencing Gli1 prior to incubating cells with Rhodamine123 increases cellular Rhodamine123 intensity by 2.5-fold and 1.5-fold in U87-MG and T98G cells, respectively (C). Data are shown as mean ± standard deviation, *p = 0.0008, #p = 0.07 by paired t-test relative to siScr. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 65

xiv Figure 3.5: Hh inhibitors modulate p53 and MGMT expression in GBM cells. Nuclear p53 staining intensity increases with GANT61-mediated Hh inhibition in (A,C) U87-MG cells by 43.4%, but decreases in (B,C) T98G cells by 21.5% relative to that in control cells. Data are shown as mean ± standard deviation from 3 independent experiments, *p<0.005 by Student’s t-test. Scale bars = 50 µm. (D) By Western blotting, T98G MGMT expression decreases in a TMZ-dependent manner. TMZ-induced downregulation of MGMT may be potentiated by Gli1 silencing. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 67

Figure 3.6. Silencing Gli1 does not correlate with apoptosis in GBM cells. (A) By Western blotting, proteins associated with apoptosis induction do not increase with Gli1 silencing. (B) Gli1 silencing does not significantly increase AnnexinV-FITC/PI staining, # no significant difference in the fraction of viable, early apoptotic, late apoptotic, or necrotic cells by one-way ANOVA with posthoc Tukey. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 70

Figure 3.7. Silencing Gli1 induces senescence in U87-MG cells in a manner dependent on the absence of PTEN. (A) SAβGal staining (teal) demonstrates that silencing Gli1 induces senescence in U87-MG cells. (B) PTEN expression reverses siGli1-induced senescence, *p<0.05. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 72

Figure 3.8. By qPCR, growing U87-MG cells in sphere culture increases the expression of genes associated with a stem-like phenotype, **p<0.01, *p<0.05 by Student’s t-test. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 73

xv Figure 3.9. Hh inhibitors and TMZ cooperatively promote apoptosis in U87-MG neurospheres and suppress neurosphere growth. (A) Flow cytometric scatterplots displaying AnnexinV-FITC and PI staining intensities in U87-MG cells grown as neurospheres in medium containing GANT61 (0 or 10 µM) and TMZ (0 or 50 µM TMZ) from one representative experiment. (B) The fraction of apoptotic (AnnexinV+) U87-MG cells grown as neurospheres in medium containing GANT61 and TMZ (50 µM TMZ), summarized across 3 independent experiments. Data shown are means ± standard deviation from 3 independent experiments, *p<0.05, **p<0.01 by one-way ANOVA with post-hoc Fisher’s least significant difference test. (C) Representative brightfield images of U87-MG neurospheres grown for one week in medium containing GANT61 and TMZ (10 µM TMZ). Scale bar = 500 µm. (D) (Left) Number of spheres counted by automated image analysis, averaged across 9 independent replicates. Data shown are means ± standard deviation from 9 independent experiments, *p<0.05, ** p<0.01 by one-way ANOVA with post-hoc Tukey test. (Right) Sphere size (projected area) as determined by automated image analysis from one representative experiment. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 76

Figure 3.10. Synergy assessment for neurosphere apoptosis in response to co- treatment with TMZ and GANT61. The fraction of measured viable (AnnexinV-, PI-) cells was significantly decreased relative to the projected viability for neurospheres co-treated with TMZ and 10 or 15 µM GANT61. Data shown are means ± standard deviations, *p<0.05 by one-way ANOVA with post-hoc Fisher’s least significant difference test. Reproduced from Melamed et al. (2018) Oncotarget [171]...... 76

Figure 4.1. SNA and PEI-SNA synthesis and characterization. A) Schematic describing the synthesis of SNAs and PEI-SNAs. B) siRNA and PEI loading characterization on SNAs and PEI-SNAs. Data are means +/- standard deviations, # = no significance. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 93

Figure 4.2. Polyplex characterization. A) Polyplex synthesis scheme. B) Gel retardation assay showing complete encapsulation of siRNA within polyplexes for N/P ratios greater than 2. C) Effect of N/P ratio on polyplex cytocompatibility. The siRNA concentration was held constant at 500 nM. *p<0.05 by one-way ANOVA with post-hoc Tukey, relative to control. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 94

xvi Figure 4.3. Nanoparticle and polyplex characterization. A) Zeta potential and DLS measurements for AuNPs, SNAs, PEI-SNAs, and PEI-siRNA polyplexes. Data are means +/- standard deviations, # = no significance. B) Representative TEM images of AuNPs, SNAs, PEI- SNAs, and polyplexes. Grids were counterstained with uranyl acetate to visualize siRNA as light contrast. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 95

Figure 4.4. Serum stability of SNAs, PEI-SNAs, and polyplexes. A) Zeta potential and B) DLS measurements for SNAs, PEI-SNAs, and polyplexes incubated in FBS. Data are means +/- standard deviations. X = couldn’t record accurate measurement due to particle aggregation. # no significant difference, * p < 0.01 relative to 0% FBS control, ** p < 0.0001. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 97

Figure 4.5. Extent and kinetics of nanocarrier binding/internalization depends on nanocarrier surface chemistry and architecture. A) Representative flow cytometric histograms showing cellular uptake of equivalent Cy5-siRNA payloads via SNAs, polyplexes, and PEI-SNAs with increasing incubation times. B) Summary of flow cytometry data in (A) across three independent experiments. Data are median fluorescence intensities (MFI) +/- standard deviations, *p < 0.01. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 98

Figure 4.6. Relative trafficking of siRNA and PEI from PEI-SNAs vs polyplexes. A) Representative confocal microscopy images showing that Cy5- siRNA remains mostly colocalized with TRITC-PEI from both polyplexes and PEI-SNAs following 24 hours incubation with cells (scale bar = 20 µm). Manders’ colocalization coefficient for fractional overlap of Cy5-siRNA and TRITC-PEI is shown in yellow on the merged brightfield image (MCC 1 = Cy5-siRNA colocalized with TRITC-PEI, MCC 2 = TRITC-PEI colocalized with Cy5-siRNA). Quantitative assessment of colocalization depicting average fractional overlap of B) Cy5-siRNA with TRITC-PEI (MCC 1) and C) TRITC- PEI with Cy5-siRNA (MCC 2) in cells exposed to PEI-SNAs or polyplexes for 24 or 48 hours across three independent replicates +/- standard deviations, no significant differences by Student’s t-test. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 100

xvii Figure 4.7. Intracellular trafficking of siRNA depends on nanocarrier architecture. A) Representative confocal microscopy images showing Cy5-siRNA colocalization with early endosomes tagged via a Rab5-GFP fusion protein (scale bar = 20 µm). Yellow boxes indicate magnified regions of siRNA colocalized with Rab5+ early endosomes. B) Representative confocal microscopy images showing Cy5-siRNA colocalization with lysosomes tagged via a LAMP1-mGFP fusion protein. Areas of colocalization appear yellow in the merged brightfield image (scale bar = 20 µm). Manders’ colocalization coefficient for fractional overlap of Cy5-siRNA and LAMP1-GFP is shown in yellow on the merged brightfield image (MCC 1 = Cy5-siRNA colocalized with LAMP1-GFP, MCC 2 = LAMP1-GFP colocalized with Cy5-siRNA). Quantitative assessment of colocalization depicting average fractional overlap of C) Cy5-siRNA with LAMP1-GFP (MCC 1) and D) LAMP1-GFP with Cy5-siRNA (MCC 2) across three independent replicates +/- standard deviations, *p=0.05 by Student’s t-test. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 103

Figure 4.8. Confocal microscopy showing SNAs colocalize heavily with lysosomes stained with LysoTracker Red. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 104

Figure 4.9. GFP silencing efficacy assessed by flow cytometry using PEI-SNAs or polyplexes. Representative flow cytometric histograms from one representative experiment depict dose-dependent GFP silencing using A) PEI-SNAs or B) polyplexes. Summary flow cytometry GFP silencing results using C) PEI-SNAs or D) polyplexes averaged across three independent replicates +/- standard deviations, *p<0.01 by one- way ANOVA with posthoc Tukey. Median fluorescence values for experimental samples are normalized to that of an untreated control. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 106

Figure 4.10: GFP silencing efficacy of SNAs delivering 400 nM siRNA. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 107

xviii Figure 4.11. Cytocompatibility profiles for PEI-SNAs and polyplexes. The cytocompatible dosing range of both A) PEI-SNAs and B) polyplexes was determined using an MTT assay. Data are means +/- standard deviations, *p<0.01 by one-way ANOVA with posthoc Tukey. Results were confirmed using PI staining and flow cytometric analysis in live cells exposed to C) PEI-SNAs or B) polyplexes. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151]...... 109

Figure 5.1. PEI-SNAs targeting Gli1 were developed to reduce the chemoresistance and stemness of glioblastoma cells...... 122

Figure 5.2. Uptake mechanism for PEI-SNAs. *p<0.05, **p<0.005 relative to cells receiving no inhibitors by one-way ANOVA with posthoc Tukey. scale= 50 µm...... 130

Figure 5.3. Intracellular trafficking of PEI-SNAs. A) Confocal microscopy was used to visualize Cy5-PEI-SNA localization to endocytic compartments after 24 hours incubation with cells. Endocytic compartments were labeled by stably expressing GFP-tagged markers for early endosomes (Rab5+), recycling endosomes (Rab11+), late endosomes (Rab7+), or lysosomes (LAMP1+). Scale bar = 20 µm. B) Results from quantitative colocalization analysis to calculate the fractional overlap of Cy5-PEI-SNAs with GFP-endosomal markers and vice versa. MCC = Manders’ colocalization coefficient, *p<0.05 by one-way ANOVA with posthoc Tukey...... 132

Figure 5.4. Gli1 PEI-SNAs reduce the mRNA expression of Gli1 and downstream target genes by qPCR. Gene expression is normalized to that of GAPDH, and data shown are means +/- SEM, *p<0.05 relative to Scr PEI-SNA control by Student’s t-test...... 134

Figure 5.5. Gli1 PEI-SNAs reduce proliferation and chemoresistance. A) By EdU assay, Gli1 PEI-SNAs reduce U87 proliferation by ~30%. Flow cytometric histograms (left) and quantification of EdU+ cells (right). Data are means +/- STDs, *p=0.02. B) SAβGal staining (teal) demonstrating that Gli1PEI-SNAs induce senescence in U87 cells. Scale = 100 µm. C) By MTT assay, Gli1 PEI-SNAs reduce U87 metabolic activity alone and in combination with TMZ, *p<0.01 relative to Scr PEI-SNA control with equivalent TMZ dose by one- way ANOVA with posthoc Tukey...... 137

xix Figure 5.6. Gli1 PEI-SNAs reduce stemness and impair self-renewal of U87 cells. A) Schematic depicting neurosphere culture model and experimental design; red cells illustrate GSCs. B) qPCR showing expression of genes associated with stemness following exposure to PEI-SNAs. Gene expression is normalized to that of GAPDH. Data are means +/- STDs, *p<0.001 relative to Scr PEI-SNA. C) Representative brightfield images of neurospheres cultured from U87 cells after exposure to PEI-SNAs. Scale = 200 µm. D) Gli1 PEI-SNAs reduce the size and number of neurospheres formed, as measured from 25 tiled brightfield images per treatment group per experiment, *p=0.03 by Student’s t-test...... 140

Figure 5.7. Gli1 PEI-SNAs potentiate neurosphere response to TMZ chemotherapy. A) Confocal microscopy visualizing Gli1 PEI-SNA distribution into small (top) and large (bottom) neurospheres. Scale = 100 µm. B) Flow cytometric histogram of Cy5-PEI-SNA uptake by cells grown as neurosperes. C) Flow cytometric analysis of the mechanism by which Gli1 PEI-SNAs distribute throughout neurospheres. Data are geometric mean fluorescence intensity (GMFI) +/- STD normalized to cells receiving no inhibitors, *p<0.05 by one-way ANOVA with posthoc Tukey. D) Experimental timeline for determining effect of co- treating neurospheres with Gli1 PEI-SNAs and TMZ. E) Flow cytometric density plots of AnnexinV/PI apoptosis analysis of neurospheres co-treated with Gli1 PEI-SNAs and TMZ. F) Summary of AnnexinV/PI apoptosis analysis. Data are means +/- standard deviations from n=2 replicates. *p<0.05 by one-way ANOVA with posthoc Fisher’s least significant difference test...... 144

xx ABSTRACT

Glioblastoma (GBM) is the most common and lethal primary brain tumor in adults, with median patient survival of just 14.6 months from the time of diagnosis. The current standard of care is surgery followed by radiation with concomitant chemotherapy; however, recurrence is inevitable and nearly 100% of patients ultimately succumb to their disease. A growing body of evidence has attributed GBM initiation and recurrence to a rare subpopulation of tumor cells characterized by an aggressive stem cell-like phenotype, called glioblastoma stem cells (GSCs). GSCs are highly refractory to radiation and chemotherapy, and they have been demonstrated to repopulate the tumor following treatment. Therefore, new strategies are desperately needed to eradicate both GSCs and differentiated, bulk GBM cells to improve patient outcomes. One potential way this may be achieved is by using RNA interference (RNAi) to suppress the expression of genes that drive GBM progression. Towards this goal, this thesis examines the role that Gli1, a key mediator of Hedgehog signaling, plays in the progression and chemotherapy resistance of GBM, and it develops a new type of polycation-nanoparticle hybrid to suppress Gli1 expression in GBM cells as a strategy for GBM therapy. This is accomplished through three objectives. First, this thesis investigates the utility of suppressing Hedgehog (Hh)/Gli1 signaling as an adjuvant to temozolomide (TMZ) chemotherapy using in vitro models of GBM. We found that small interfering RNA (siRNA) against Gli1 could reduce GBM cell proliferation, metabolic activity, and drug efflux activity. Surprisingly,

xxi silencing Gli1 without TMZ co-treatment did not induce apoptosis in GBM cells; rather, it induced senescence in a loss of PTEN-depended manner. Excitingly, this work showed that pharmacological Gli inhibition in combination with TMZ could induce apoptosis in stem-like GBM cells grown as neurospheres. These results demonstrate that targeting Hh/Gli1 signaling offers a promising strategy to reduce the chemoresistance of GBM cells, and therefore we concluded that Gli1 is a potent target for therapeutic manipulation by siRNA nanocarriers. While siRNA-mediated gene regulation has emerged a promising strategy to halt tumor growth, clinical translation of RNAi remains challenging because siRNA is rapidly cleared from circulation, easily degraded by nucleases that are present in physiological conditions, and cannot passively cross cell membranes due to its large size and negative charge. Nanoparticle-based siRNA carriers have been made to overcome these issues, but their translation is hindered by limited knowledge regarding the parameters that regulate their interactions with biological systems. To address this, we investigated the influence of polycation-based nanocarrier architecture on intracellular siRNA delivery. The cellular interactions of two polycation-based siRNA carriers with different siRNA orientation were compared: (1) polyethylenimine-coated spherical nucleic acids (PEI-SNAs), in which polyethylenimine is wrapped around a spherical nucleic acid core containing radially- oriented siRNA, and (2) randomly assembled polyethylenimine-siRNA polyplexes that lack controlled architecture. PEI-SNAs outperformed PEI-siRNA polyplexes in terms of net cellular uptake, lysosomal evasion, gene regulation potency, and cytocompatibility. These results highlight the importance of siRNA architecture in

xxii efficient nanocarrier design, and prompted us to explore the utility of PEI-SNAs targeting Gli1 as a therapy for GBM. In the final objective of this thesis, PEI-SNAs targeting Gli1 were developed and evaluated against GBM cells. We found that Gli1 PEI-SNAs successfully silence tumor-promoting Hedgehog pathway genes and downstream target genes that promote the chemoresistant phenotype of GBM. This corresponds with decreases in GBM cell proliferation and metabolic activity and an induction of GBM cell senescence. Most importantly, we observed that Gli1 PEI-SNAs impair the self-renewal capacity of GBM cells and substantially improve neurosphere response to low doses of TMZ. These results demonstrate that Gli1 PEI-SNAs can reduce GBM resistance to therapy, which is a major hurdle to eradication of this disease. Overall, this thesis advances the field of RNAi for GBM by showing that Gli1 is an impactful target for GBM therapy, and by developing novel siRNA nanocarriers to target this gene in GBM cells. With continued in vivo investigation and nanocarrier optimization, Gli1-targeted RNAi therapeutics have great potential to alleviate drug resistance and recurrence for GBM patients.

xxiii Chapter 1

INTRODUCTION

Despite tremendous improvements in diagnosis and treatment strategies, cancer remains a leading cause of death worldwide. In 2012, there were 14.1 million new cancer cases diagnosed and 8.2 million cancer-related deaths worldwide, more than half of which occurred in developing nations. By 2030, it is estimated that the number of new cancer cases will rise to 23.6 million [1]. In the United States, the overall cancer death rate decreased by 25% between 1990 and 2014, producing a projected 30% increase in the number of cancer survivors by 2026 [2]. While this highlights the tremendous progress made in cancer diagnosis and treatment, much work remains. Educational movements have decreased rates of behavioral cancer causes such as smoking, but additional cancer risk factors including the aging population and increasing obesity rates have provided new challenges for the fight against cancer. Cancer itself is not a new problem, nor are the primary treatment strategies still used today: surgery, radiation, and chemotherapy. The first record of cancer dates back to 2,500 BCE, when the Egyptian doctor Imhotep recorded treatments for various diseases but in regard to breast cancer, wrote “there is none” [3]. Among the first documented surgeries is a record from 440 BCE about a procedure performed on Atossa, Queen of Persia, for a lump in her breast [3]. It was not until 1882 after the advent of anesthesia and antiseptics that William Halsted introduced the radical

1 mastectomy, which involved removal of the breast, muscles, and axillary lymph nodes [4]. Revolutionary at the time, this approach to surgical breast cancer removal was still widely used in the 1970s, until sufficient data had accumulated to demonstrate this approach offered no survival benefit and the feminist movement opposed the procedure [3,5]. (Actually, survival data presented by Halsted himself in the early 1900s demonstrated this, but the radical surgery movement had already become wildly popular, with increased resection imbuing surgeons with higher status [3].) This eventually gave rise to the minimally-invasive trend that prevails today. Radiation therapy was first introduced in the early 1900s, after the discovery of X-rays by Wilhelm Conrad Runtgen and the discovery of radium by Maria and Pierre Curie. Chemotherapy was introduced the most recently of the three aforementioned cornerstone cancer treatment strategies. Sidney Farber, a pediatric pathologist at Boston Children’s Hospital referred to as the “doctor of the dead,” first used aminopterin, a potent folate antagonist, to treat children with acute lymphoblastic leukemia (ALL) in 1947 [6]. At a time when ALL was practically a death sentence, 10 of the 16 children Farber treated achieved temporary remission, marking a groundbreaking accomplishment for cancer therapy [3,6]. These three strategies have evolved substantially since their early stages with the introduction of targeted therapies and advanced surgical techniques, and patient prognosis has improved dramatically for many cancers. However, many aggressive forms of cancer remain incurable, and new therapies are desperately needed to improve patient outcomes. Working towards the goal of developing therapeutic strategies that offer high specificity for tumor tissue while leaving healthy tissue intact, gene therapy has emerged as a clinically promising strategy to exploit genetic alterations common to

2 many cancers. In this dissertation, research is presented that investigates a nanoparticle platform to enable gene regulation as a therapeutic strategy for glioblastoma, an aggressive form of brain cancer that remains incurable. This approach offers several advantages over traditional therapies, which are discussed throughout this dissertation. In this chapter, the cancer biology and the challenges and progress towards successful glioblastoma therapy will be discussed. Further, developments in gene therapy and nanomedicine to enable it will be reviewed. This chapter contains sections adapted from: Kapadia CH, Melamed JR, Day ES. Spherical Nucleic Acid Nanoparticles: Therapeutic Potential. BioDrugs (2018); 32(4):297-309.

1.1 A Brief Introduction to Cancer Biology and Glioblastoma To successfully treat cancer, it is crucial to first understand its nature. Here, the mechanisms that contribute to tumor initiation and progression will be discussed.

1.1.1 Tumorigenesis and the Propagation of Cancer Tumorigenesis can be explained by two prevailing models that likely each contribute towards the propagation of cancer. The first model is the clonal evolution model, or the stochastic model, for tumor growth (Figure 1.1A). This model states that tumorigenesis is a multi-step process in which normal human cells sustain a series of mutations over time, which disrupts the homeostatic signaling that normally maintains the balance between growth and anti-growth signals [7]. Commonly, such mutations occur to tumor suppressor genes, which are normally responsible for controlling cell growth, or to proto-oncogenes, which signal the onset of cell division or regulate apoptotic cell death [8]. The result of accumulated mutations to these homeostatic regulatory pathways is a cluster of cells that exhibit uncontrolled proliferation and

3 initiate tumor formation. These malignant cells can divide randomly and sustain additional mutations, which produces subpopulations within a tumor that give rise to heterogeneity throughout the tumor [9]. Therefore, the clonal evolution model states that all or most cells within a tumor are capable of initiating a tumor or driving tumor progression.

Figure 1.1. Current models for tumorigenesis. Modified from Bradshaw et al. (2016) Front Surg [10].

The second model for tumorigenesis is the hierarchical stem cell model (Figure 1.1B). According to this model, tumors arise from a rare population of pluripotent cells called cancer stem cells (CSCs), which are capable of self-renewal and asymmetric division. This can produce additional tumorigenic CSCs, non-tumorigenic progenitor cells, or differentiated cancer cells [10]. CSCs are highly refractory to chemotherapies in part due to their characteristic rapid DNA damage response and high expression of anti-apoptotic proteins [11]. Further, CSCs are enriched in tumors

4 following radiation therapy or chemotherapy, and they are consequently responsible for tumor recurrence [12,13]. Importantly, these models for tumorigenesis are not mutually exclusive, and evidence demonstrates that clonal evolution or cancer cell de-differentiation may contribute to the formation of CSCs [10]. Accordingly, an additional mixed model has been proposed in which mutations acquired by CSCs may give rise to a dominant CSC population with enhanced self-renewal capacity [14]. CSCs will be discussed in greater detail later in this thesis. Hanahan and Weinberg proposed that the sum of these acquired mutations produce several alterations to cell physiology that are essential for malignant tumor growth [8]. These characteristics, termed “hallmarks of cancer” include resistance to apoptosis, sustaining proliferative signaling, evasion of growth suppressors, enabling replicative immortality, activation of invasion and metastasis, and induction of angiogenesis (Figure 1.2). After it was proposed that an inflammatory microenvironment constituted a seventh hallmark [15], these characteristics were later expanded upon to include newly emerging hallmarks: deregulating cellular energetics and avoiding immune destruction, and their enabling characteristics: genome instability and tumor-promoting inflammation [16]. Importantly, the molecular aberrations that give rise to these behaviors are highly variable both among and within tumors. The development of successful next-generation cancer therapies will require greater mechanistic insight into the propagation of specific tumor types and the molecular landscapes that enable tumor progression in individual patients.

Figure 1.2. The hallmarks of cancer. Reproduced from Hanahan and Weinberg. (2000) Cell [8].

1.1.2 Introduction to Glioblastoma Glioblastoma (GBM) is the most common and lethal primary brain tumor in adults. Despite substantial advances in surgery, radiotherapy, and chemotherapy over the last several decades, GBM remains an incurable disease, boasting a dismal median patient survival of 14.6 months. Fewer than 10% of patients survive beyond 5 years from diagnosis, and nearly 100% of patients ultimately succumb to disease [17]. Although it can also effect children, GBM incidence increases with age; GBM risk is highest for patients aged 75-84 years and decreases past 85 years [17]. The standard of care for newly diagnosed GBM is surgery, followed by radiation and chemotherapy [18,19]. Despite this multimodal intervention, median progression-free survival remains low at 6.9 months, and recurrence is inevitable [19]. New treatment modalities are desperately needed to improve patient outcomes. Here, the biology of GBM and current treatment strategies will be discussed.

6 GBM is characterized by its rapid, infiltrative growth pattern and extensive vascularization, which complicate both surgical resection and the detection of cells that have migrated far from the original tumor site [20,21]. GBM tumors consist of an outer region of rapidly dividing cells supported by an inflammatory microenvironment, all surrounding a necrotic tumor core [20,22]. The boundary between the tumor and surrounding normal tissue is called the interface, beyond which is the parenchymal brain zone consisting of healthy tissue, tumor infiltrate, and isolated tumor cells [23]. This is illustrated in Figure 1.3.

GBM can be classified into four primary molecular subtypes with unique gene expression profiles that dictate disease initiation and progression: mesenchymal (26.7%), classical (25%), proneural (16.9%), and neural (14.3%) [20]. Of these, mesenchymal and classical GBM are the most aggressive, typically occur in older patients, and are associated with the worst outcomes [25]. Proneural and neural GBM

7 tumors are associated with younger patients and better overall survival than mesenchymal and classical tumors [25]. To complicate things further, multiple molecular subtypes can exist within a GBM tumor, introducing additional intratumor heterogeneity [26]. The extreme genetic instability and epigenetic abnormalities in GBM produce dysregulation in multiple pathways that govern the ability of GBM cells to proliferate, recruit vasculature, invade surrounding tissue, and resist therapy. The most commonly affected pathways are those involving receptor tyrosine kinases (RTKs), which stimulate proliferation and angiogenesis in response to growth factor signaling. EGFR signaling, which induces proliferation and migration of GBM cells, is often upregulated by receptor overexpression or mutation [27]. The most common mutant variant, EGFRviii, is a tumor-specific constitutively active mutant occurring in 25- 64% of all GBMs that enhances proliferation, cell metabolism, and angiogenesis through downstream modulation of PI3K/Akt/mTOR and NF-κB signaling pathways [28]. Upregulated EGFR signaling can also activate oncogenic Ras signaling to drive cell cycle progression, survival, and migration via PI3K/Akt [29,30]. RTK mutations are also implicated in aberrant PI3K/PTEN/Akt signaling, which promotes proliferation and survival [30]. PTEN is a proto-oncogene that normally negatively regulates this pathway; however, PTEN function is lost in 36% of GBMs either through deletion or constitutively active PI3K mutants [31,32]. Loss of tumor suppressive functions, such as those exerted by the retinoblastoma (Rb) pathway or p53, are also common to GBMs. Rb normally regulates cell cycle progression by binding the E2F transcription factors, which prevents the transcription of genes necessary for mitosis and causes G1/S cell cycle

8 arrest [33]. Loss of Rb function through mutation or epigenetic silencing prevents this negative feedback to drive cell cycle progression [34]. Similarly, p53 normally regulates cell cycle progression, DNA damage response (DDR), and apoptosis in response to stress stimuli [35]. Upon recognition of DNA damage, p53 activates DDR and upregulates p21 to halt cell cycle progression until the damage is repaired. This process is negatively regulated by MDM2, a proto-oncogene [36]. GBMs often harbor missense p53 mutations or overexpression of MDM2 to deregulate this pathway, promoting cell cycle progression and opposing apoptosis [35]. Although mutations to key pathways are reasonably well understood, GBM tumors can harbor any number and combination of these mutations with substantial heterogeneity throughout a tumor, and patient-to-patient variability is high. Consequently, there is no definite number of mutations or series of events sufficient to induce gliomagenesis, nor are there therapeutic strategies sufficient to combat all dysregulated signaling in GBM.

1.1.3 Current Treatment Strategies for Glioblastoma The standard of care for newly diagnosed GBM is surgery, followed by radiation and chemotherapy [18,19,37]. The goal of surgery is to achieve maximal safe total resection while preserving function. Greater total resection is associated with better overall survival, and it has been demonstrated that patients who receive at least 78% resection face more favorable prognosis [38], as determined by MRI 72 hours post-operation. The implantation of carmustine wafers (Gliadel) after surgical resection is FDA-approved for GBM and anaplastic glioma, but the severity of adverse events and questionable survival benefit limit the utility of this strategy [39]. Patients subsequently receive concomitant TMZ chemotherapy and radiotherapy followed by adjuvant TMZ, as determined by the phase III trial EORTC 26981 [18]. The addition

9 of TMZ to the previous regimen of surgery plus radiation successfully increased overall survival from 12.1 months to 14.6 months [19]. The success of this regimen is known to depend on the expression of O6-methylguanine–DNA methyltransferase (MGMT), an enzyme that repairs O6-methylguanine DNA adducts resulting from TMZ treatment [40]. MGMT promoter methylation status serves as a prognostic indicator of TMZ response, though the clinical utility of this is limited by the lack of alternative treatment strategies for patients with active MGMT [37]. Bevacizumab, a recently developed targeted therapy, is a humanized monoclonal antibody that binds VEGF-A to prevent its binding to the VEGF receptor that demonstrated exciting response rates against recurrent GBM [41,42]. It further demonstrated an increase in progression free survival in two randomized, placebo-controlled phase III trials (10.7 vs 7.3 months), but it was also associated with a decline in neurocognitive function and did not improve overall survival. No additional therapeutic strategies have succeeded in improving overall survival in combination with the current standard of care. Overall, patient survival remains poor, tumor recurrence is inevitable, and new therapies are needed to improve survival rates and time while maintaining patient quality of life.

1.1.4 Glioblastoma Stem Cells and Hedgehog Signaling as a Therapeutic Target The inevitable disease recurrence characteristic of GBM has recently been attributed to a rare subpopulation of GBM cells termed glioblastoma stem-like cells, or GSCs. These cells were first identified by Singh et al., who isolated a CD133+ population of cells from human brain tumors and found that only CD133+ cells could successfully initiate tumor formation in vivo [43]. Although much remains unknown regarding the origin of GSCs and their defining molecular characteristics remain

10 controversial [44,45], several functional characteristics have been identified that define this population (Figure 1.4). GSCs promote tumor growth through means comparable to how normal stem cells give rise to functional tissues: they demonstrate sustained self-renewal and produce differentiated progeny [46]. In addition, GSCs are characterized by the ability to initiate brain tumors upon implantation into animal models [46].

Figure 1.4. Cancer stem cells as drivers of tumor progression. Reproduced from Lathia et al. (2015) Genes Dev [47].

In addition to their role in tumor initiation, GSCs are known to contribute to therapy resistance. GSCs are maintained by the hypoxic core of GBM tumors rather than the vascularized environment necessary to support differentiated GBM cells [48]. This necrotic core region is therefore more likely to exhibit resistance to radiation and chemotherapy [49]. Accordingly, GSCs can resist radiotherapy through preferentially enhanced DNA damage response [12]. The ability of GSCs to resist cytotoxic stimuli

11 has been further linked to aberrations in Notch, NF-KB, EZH2, and PARP, demonstrating that multiple mechanisms likely contribute to GSC therapy resistance [46]. Crucially, GSCs have been identified as drivers of tumor recurrence following chemotherapy [13], highlighting the need for new therapeutic strategies to successfully eliminate GSCs. Several signaling pathways have been identified to maintain the aggressive, stem-like phenotype of GSCs. Among these, developmental pathways such as Wnt, Notch, and Hedgehog (Hh) have emerged as key pathways required for GSC maintenance [50]. Aberrant Wnt activation in GSCs promotes tumor growth through enhanced β-catenin transcriptional activity [51], and Wnt/β-catenin signaling regulates the expression of PLAGL2 to maintain the stemness of GSCs [52]. Notch signaling is essential for preventing differentiation of neural stem cells, and aberrant Notch activation induces a highly proliferative stem-like phenotype in astrocytes [53], and Notch inhibitors can potentiate the effects of chemotherapy in GSCs [54,55]. Hh signaling has also been identified as a key pathway for GSC self-renewal and tumorigenicity [56,57]. Further, only Hh signaling has been demonstrated to be indispensable for maintaining the self-renewal and tumorigenic capacity of GSCs, whereas Wnt and Notch were not found to be required for maintaining these phenotypes [58]. Hh signaling is a highly conserved developmental pathway that is minimally active in most differentiated adult tissue, but it also plays an impactful role in the progression and maintenance of several cancers, including GBM. Hh signaling is activated when Hh ligands, most notably sonic hedgehog (Shh), bind the extracellular domain of the Patched (Ptc) transmembrane receptor. Ptc then relieves its suppression

12 of Smoothened (Smo), G-protein coupled receptor (GPCR)-like protein, which in turn initiates an intracellular signaling cascade that translocates members of the Gli family of zinc finger transcription factors to the nucleus to regulate the expression of target genes (Figure 1.5). Within this family of transcription factors, upregulated Gli1 expression and transcriptional activity is associated with poor patient prognosis in several cancers [59]. In addition to its role in maintaining GSC stemness, Gli1 is known to regulate proliferation and cell cycle progression [56,60], cell survival [60,61], migration and invasion [62], and response to cytotoxic stimuli such as chemotherapy [55,56,63].

Figure 1.5. Schematic depicting Hedgehog signaling.

Due to its crucial role in the progression of various cancers and maintaining CSCs, several agents have been developed to target Hh signaling by inhibiting either Smo or Gli1 [64]. Therapies targeting Smo are the furthest along in development; two Smo inhibitors, Vismodegib and Sonidegib, have already been FDA-approved to treat

13 basal cell carcinoma (BCC) and are currently in clinical trials for several cancers [64]. In preclinical studies, Sonidegib was demonstrated to impair epithelial mesenchymal transition and self-renewal of GSCs [65]. However, in a later clinical trial investigating Vismodegib to treat patients with metastatic BCC, it was found that after three months, an acquired Smo mutation prevented successful BCC regression [66]. A similar result was found in patients with Shh-driven medulloblastoma [67]. Additionally, a Phase II clinical trial was recently conducted investigating Vismodegib to treat patients with recurrent GBM; however, there was no observed improvement in overall survival or progression-free survival [68]. Targeting Gli1, which acts downstream of Smo, may be a rational strategy to bypass Smo mutations for successful Hh-targeted therapy. A cellular screen identified two Gli antagonists, GANT58 and GANT61, that demonstrated the ability to reduce tumor cell proliferation in a Gli-dependent manner [69]. GANT61, the more potent and specific of the two, has demonstrated efficacy against rhabdomyosarcoma, osteosarcoma, neuroblastoma, and ovarian cancer in preclinical studies, but no clinical trials are currently investigating this compound [64]. Arsenic trioxide (ATO) is a Gli1 and Gli2 inhibitor that has been FDA-approved to treat acute promyelocytic leukemia [70]. ATO has demonstrated efficacy against glioblastoma and GSCs in preclinical studies [71,72] and is currently being investigated in a phase I/II clinical trial to treat malignant glioma patients in combination with TMZ and radiotherapy [73]. These results warrant the continued development of Hh-targeted therapies to treat GBM, particularly therapies targeting Gli1. Accordingly, Chapter 3 of this thesis evaluates Gli1 as a therapeutic target for GBM, and Chapter 5 characterizes a novel strategy to inhibit Gli1.

14 1.2 Nanotechnology in Glioblastoma Therapy Nanotechnology has the potential to overcome some of the limitations associated with conventional cancer therapies, including nonspecific toxicity, short circulation half-life, and poor bioavailability in the brain. Nanoparticles, materials roughly on the order of 10-200 nm, present the advantage that they are large enough to incorporate multiple functionalities into a single construct, while they are still small enough to navigate through the bloodstream and diffuse through complex biological milieu. Over the past several decades, nanoparticle technology has seen rapid development in the field of drug delivery to tumors [74]. Nanoparticles have been employed as vehicles to encapsulate drugs and achieve extended release at the tumor site, as contrast agents to improve the sensitivity and specificity of medical imaging techniques, and as stimuli-responsive carriers designed to release cargo in the presence of environmental cues. Moreover, due to their large surface area-to-volume ratio, nanoparticles can be functionalized with bioactive molecules such as antibodies, peptides, or aptamers to specifically bind desired cells within a targeted disease site, or they can be coated with bioinert “stealth” molecules such as poly(ethylene) glycol (PEG) or hyaluronic acid to disguise them from the immune system and extend their circulation time (Figure 1.6).

Figure 1.6. Nanoparticles can enable multiple functionalities from a single construct, including the controlled release of drugs, presentation of nucleic acids and antibodies for targeting or therapy, and stealth agents such as PEG to reduce immune clearance and increase circulation time.

As an advantage towards engineering effective cancer therapeutics, nanoparticles can exploit vascular abnormalities within the tumor to improve their intratumoral retention. Vasculature in healthy tissue is highly organized, tightly regulated by homeostatic mechanisms, and maintains effective lymphatic drainage. In contrast, solid tumor vasculature, which emerges rapidly in response to aberrant angiogenic signaling, exhibits chaotic, tortuous morphology that leads to severe functional abnormalities including leakiness and poor drainage from the tumor site, shown in Figure 1.7 [75]. This facilitates enhanced nanoparticle retention within the tumor, and this phenomenon is known as the enhanced permeability and retention (EPR) effect [76]. Nanoparticle targeting to tumors that relies on the EPR effect is known as passive targeting. Functionalizing nanoparticles with bioactive ligands that bind overexpressed receptors on tumor cells can further improve retention at the tumor site (Figure 1.7), known as active targeting [77]. Both strategies are widely used to achieve nanoparticle delivery to solid tumors. In this section, the advantages and

16 challenges for nanoparticles in GBM therapy will be discussed, and nanoparticle systems delivering therapeutics to GBM tumors will be reviewed.

1.2.1 Challenges and Opportunities for Nanotechnology in Glioblastoma Therapy Perhaps the greatest challenge towards achieving nanoparticle delivery to GBM tumors is successfully crossing the blood-brain barrier (BBB) to achieve sufficient concentrations of therapeutic cargo within the tumor. The BBB is a selectively permeable endothelial barrier that prevents neurotoxins from entering the brain and consequently prevents many conventional chemotherapeutics and nanoparticle formulations from entering tumors. While the vasculature often exhibits

17 increased permeability in solid tumors as discussed above, the BBB remains intact at sites of infiltrating gliomas or micrometastases [79]. The BBB, depicted in Figure 1.8, is formed by brain microvascular endothelial cells (BMECs) lacking pinocytosis and fenestrations due to the presence of tight junctions that maintain high electrical resistance across the barrier. These tight junctions consist of mainly occludins and claudins, which restrict the permeability of small molecules < 800 Da [79]. The basement membrane consists of ECM components secreted by BMECs, astrocytes, and pericytes [80]. Due to tight regulation of the BBB, it is estimated that only 2% of small molecule drugs (< 400 Da) and no large molecules are capable of passively diffusing across this barrier [81]. However, multifunctional nanoparticles offer opportunities to exploit BBB biology to penetrate GBM tumors with greater success than many small molecule drugs.

Figure 1.8. Structure of the blood-brain barrier. Adapted from Glaser et al. (2017) Front. Pharmacol [79].

1.2.2 Nanoparticles for Drug Delivery to Glioblastoma Tumors Nanoparticles have been investigated as carriers to improve drug delivery to GBM tumors with the goals of enhancing BBB penetration of small molecule drugs, promoting tumor-specific therapy with molecular targeting, and providing extended drug release at the tumor site. Liposomes and lipid nanoparticles can efficiently encapsulate hydrophobic chemotherapy drugs such as doxorubicin, paclitaxel, and TMZ to improve their aqueous solubility. Liposomal TMZ has previously demonstrated improved circulation time and brain accumulation relative to free TMZ following intravenous injection into mice [82]. Another recent study comparing the efficacy of free TMZ and liposomal TMZ delivered via convection-enhanced delivery found that both TMZ formulations improved the survival of GBM-bearing rats

19 equally, while the liposomal formulation significantly decreased tumor edema relative to free TMZ [83]. The liposomal drug formulation furthest along in clinical development for GBM therapy is liposomal irinotecan (NL CPT-11). Convection- enhanced delivery of NL CPT-11 significantly improved intracranial drug half-life relative to free drug, which translated to a significant increase in median survival (100 days) in an intracranial U87 xenograft model [84]. A phase I trial is currently ongoing to evaluate intravenous delivery of NL CPT-11 in patients with recurrent high-grade gliomas [85]. Using in vitro BBB models, solid lipid nanoparticles (SLNs) have demonstrated enhanced bevacizumab and paclitaxel BBB permeability, warranting the further develop of SLN drug formulations for GBM [86,87]. A separate class of degradable polymer nanoparticles can provide extended release to sustain therapeutic drug concentrations at the tumor site for a longer duration. Among the most widely investigated polymer nanoparticles for drug delivery, poly(lactide co-glycolide) (PLGA) nanoparticles are highly biocompatible, degradable, and offer straightforward synthesis, favoring their continued development. Convection-enhanced delivery of dithiazanine iodide-loaded PLGA nanoparticles significantly increased the survival of rats bearing brain cancer stem cell-derived xenografts [88]. Intravenously administered PLGA nanoparticles loaded with camptothecin (CPT), which failed in clinical trials due to dose-limiting toxicity, significantly improved survival relative to free CPT in immunocompetent mice bearing GBM tumors [89]. Polymeric micelles and dendrimers have also been developed as nanoscale materials capable of delivering drugs to GBM tumors [79].

20 1.2.3 Nanoparticles as Vectors for Gene Therapy Over the last several decades, gene therapy has emerged as a promising new paradigm for cancer therapy. Whereas small molecule chemotherapeutics generally target rapidly dividing cells by inducing DNA damage or stabilizing microtubules to prevent cell division, gene therapy offers the ability to directly target the dysregulated signaling pathways that promote cancer progression. Overall, there are two approaches to correct a dysregulated signaling pathway: by inducing expression of a tumor suppressor gene that has been lost or by silencing an oncogene that has been aberrantly upregulated. Gene expression can be induced by delivering plasmid DNA (pDNA) or messenger RNA (mRNA) encoding the desired gene, while the expression of an oncogene can be silenced by delivering complementary small interfering RNA (siRNA), microRNA (miRNA), or antisense DNA. Protein delivery is also of more recent interest for inducing gene expression but is limited by the large size and poor in vivo stability of proteins [90] and will not be discussed in this thesis. Several challenges hinder the successful delivery of nucleic acids to mediate gene therapy. Nucleic acids are highly susceptible to degradation by nucleases present ubiquitously in physiological conditions. Further, nucleic acids, particularly RNA molecules, can be immunogenic and are subject to rapid clearance by macrophages. Nucleic acids are also much larger than small molecule drugs and carry a negative charge, prohibiting their entry into cells. Finally, nucleic acids that have entered cells by endocytic processes must escape the endosome to reach the intracellular compartments necessary for their activity. DNA must pass through small nuclear pores to enter the nucleus, mRNA must reach the ribosomes, and siRNA or miRNA must reach the cytosol to engage with the RNA-induced silencing complex (RISC). To

21 overcome these barriers to nucleic acid delivery, carriers are needed to protect the nucleic acid cargo and efficiently reach the target site. Efforts towards developing vectors for translatable nucleic acid delivery can be grouped into two broad categories: viral vectors and non-viral vectors. While both viral and non-viral vectors have been investigated in clinical trials for systemic nucleic acid delivery, ~70% of gene therapy trials to date have utilized modified viruses, where lentiviruses are most commonly used for ex vivo gene transfer and adeno- associated viruses are most commonly used for in vivo gene transfer [91,92]. However, viral vectors are limited by broad tropism to many cell and tissue types and risk random gene integration into the host genome (insertional mutagenesis), horizontal gene transmission to off-target cells, vertical gene transmission to germline cells, and strong immunogenicity [92]. Further, viral vectors are limited from a production standpoint by low nucleic acid packaging efficiency and difficulty in scaling-up vector production [91]. Non-viral vectors, such as nanoparticles, can overcome several limitations posed by viral nucleic acid delivery vectors. Nanomaterials engineered to deliver nucleic acids, despite their lower gene transfer efficiency, are often easier to synthesize and offer the potential to deliver larger payloads than viruses [93,94]. Further, nanoparticles offer increased safety and design versatility not achievable with viral vectors. As discussed previously in the context of drug delivery, multifunctional nanoparticles can be engineered to enable selective tumor targeting, cargo encapsulation, and controlled release. These advantages also support their use as nucleic acid delivery vehicles.

22 In fact, a substantial body of work has demonstrated that nanoparticle- mediated gene regulation can dramatically improve the efficacy of chemotherapy and radiotherapy by controlling the expression of genes that influence drug response, either from a single multifunctional nanoparticle platform (in the case of chemosensitization) or delivered as separate entities. For example, miR-21 was one of the first miRNAs identified as an oncomiR that drives proliferation, migration, and survival in several cancers, including GBM. Cationic solid lipid nanoparticles and PLGA nanoparticles delivering antisense miR-21 to silence its expression and halt its activity have been investigated to sensitize GBM cells to TMZ chemotherapy and radiotherapy [95,96]. Similarly, nanoparticles delivering pDNA to induce the expression of pro-apoptotic TRAIL and wild-type p53 have also yielded improved TMZ sensitivity in preclinical studies [97,98]. A cationic liposome carrying p53 cDNA is currently being evaluated in a phase II clinical trial with the goal of suppressing MGMT expression and restoring apoptotic signaling to sensitize tumors to TMZ chemotherapy [99]. These promising results warrant the continued development of nanoparticles to regulate the expression of genes governing drug response in GBM tumors.

1.3 Nanotechnology in RNA Interference While gene regulation with non-viral vectors has demonstrated great potential to combat GBM therapy resistance in preclinical studies, no gene regulation-based therapeutics have received FDA approval for GBM therapy. Much research is needed rationally design vectors to achieve safe and efficient nucleic acid delivery. This section will focus on recent efforts to develop nanomaterials to enable clinically

23 translatable RNA interference as a therapeutic strategy to regulate gene expression in GBM.

1.3.1 RNA Interference and Rational Design of Nanoparticles First reported by Fire and Mello in 1998 [100], RNA interference (RNAi) is a clinically promising strategy to silence the expression of genes that promote disease progression with high potency and specificity. RNAi occurs naturally in cells when long double-stranded RNA undergoes cleavage by the enzyme Dicer, producing short double-stranded RNA fragments 19-23 base pairs in length, termed small interfering RNA (siRNA). As a therapeutic strategy, exogenous siRNA can be delivered to cells to silence the expression of the desired gene target. Upon delivery to cells, siRNA associates with the protein complex RISC in the cytosol. Argonaute 2, a protein within the RISC complex, unwinds the duplex and retains the antisense strand, or the “guide strand,” while the sense strand is degraded. The guide strand aligns RISC with complementary mRNA, which initiates mRNA degradation to silence the expression of the targeted gene (Figure 1.9). However, clinical implementation of RNAi has been limited because systemically delivered siRNA is rapidly cleared from circulation due to phagocytic uptake and renal clearance. Additionally, siRNA molecules are too large to cross the capillary endothelial barrier and too large and negatively charged to efficiently enter cells without a carrier. Further, siRNA is highly susceptible to degradation by nucleases and engages toll-like receptors TLR-7 and TLR-3 to induce nonspecific immune activation. Consequently, much research has sought to identify strategies to effectively deliver siRNA. A key challenge that remains to be fully addressed is the development of strategies to deliver siRNA efficiently to target cells while avoiding nonspecific toxicity due to either the employed siRNA carrier or due to

24 off-target siRNA delivery. In this section, strategies to improve the clinical practicality of siRNA therapeutics will be discussed, including 1) conjugating small molecules to the siRNA, 2) chemically-modifying the siRNA bases directly, and 3) developing nanoparticle carriers for siRNA delivery.

Figure 1.9. RNA interference allows specific silencing of the targeted gene. Reproduced from Crooke. (2004) Curr Mol Med [101].

One strategy investigated to render siRNA more amenable to systemic delivery has been to conjugate small molecules to siRNA to improve its circulation half-life and cellular uptake. For example, previous research has explored siRNA conjugated to PEG and to cell-penetrating peptides to improve its therapeutic utility [102,103]. Further advanced in clinical development are dynamic poly-conjugate (DPC)-siRNAs

25 and N-acetylgalactosamine (GalNAc)-siRNA conjugates, which are proprietary technologies under development by Arrowhead Research Corporation and Alnylam Pharmaceuticals, respectively [91]. A second strategy to enable therapeutic siRNA delivery is chemically modifying the siRNA bases to reduce immunogenicity and susceptibility to nuclease degradation. Modifications to the ribose backbone, such as with 2′-fluorine (2′-F) or 2′-O-methyl groups (2′O-Me), can improve RNA stability against nucleases and reduce immunogenicity. In particular, 2′-O-Me bases have demonstrated sufficient success in preclinical and early clinical studies that they are currently under investigation in three phase II-III clinical trials [104]. Other modifications under investigation include locked nucleic acids (LNA) and unlocked nucleic acids (UNA), and substituting the phosphodiester backbone linkage with phosphorothioate linkages [104,105]. The strategy that has perhaps gained the most traction in research and clinical development is the use of nanoparticle carriers to deliver siRNA to the targeted disease site. To be effective, siRNA nanocarriers must overcome several physiological barriers that limit their delivery. Intravenously injected siRNA nanocarriers must avoid clearance and degradation in the bloodstream, transport across the vascular endothelium, diffuse through extracellular matrix, enter cells, escape the endosome, and deliver siRNA to RISC (Figure 1.10). In particular, achieving endosomal escape represents a major bottleneck for siRNA therapeutics, and an expansive field of research has devoted effort towards discovering novel methods and materials to enable efficient endosomal escape using biocompatible materials.

Figure 1.10. Barriers to siRNA delivery. Modified from Whitehead et al. (2009) Nat Rev Drug Discov [105].

The interactions between siRNA nanocarriers and biological systems are determined by their physicochemical and bioactive properties, such as the size, shape, and surface chemistry of the nanoparticle [106]. For example, nanoparticle size influences both the biodistribution and cellular uptake profile. Previous research has demonstrated that particles within the range of 30-200 nm exhibit the longest systemic circulation and accumulate to the greatest extent within tumors [107], where nanoparticles in the smaller end of this range exhibit the greatest penetration into tumors [108]. Spherical nanoparticles ~50 nm in diameter enter cells most rapidly and to the greatest extent [109], while the optimal size for cellular uptake also depends on nanoparticle shape [106]. Surface chemistry, including surface charge and the presence of bioactive ligands, plays a crucial role in determining the siRNA delivery efficacy of nanomaterials. It is well known that cationic nanomaterials undergo greater

27 cellular uptake than anionic nanomaterials due to their increased electrostatic interactions with negatively-charged cell membranes [110–112]. However, this interaction becomes more complex following in vivo systemic delivery because cationic nanomaterials rapidly adsorb a layer of negatively-charged serum proteins [105]. The resulting protein corona has various consequences on the in vivo fate of nanomaterials, including the potential to either stabilize or destabilize the nanoparticles, induce a nonspecific immune response, or shield the nanoparticles from interacting with their target cells to render them ineffective [106,113]. Further, the presence of bioactive ligands can dramatically influence the interactions of siRNA nanocarriers with biological systems. As discussed previously, molecules such as antibodies, peptides, or aptamers can be used to selectively bind receptors overexpressed by tumor cells to improve nanoparticle retention within the tumor [77]. However, such bioactive molecules also influence the intracellular trafficking profile of endocytosed nanomaterials to determine their siRNA cytosolic delivery efficacy. siRNA nanocarriers are most frequently taken up by receptor- mediated endocytic processes, where the exact pathway experienced by a nanomaterial is determined by both the physicochemical properties of the nanoparticle and the cell type exposed to the nanoparticle [114,115]. Following endocytosis, nanoparticles are typically routed through the classical endolysosomal pathway, where nanomaterials that engage cell surface receptors are taken up into early endosomes that acidify to ultimately fuse with late endosomes and lysosomes (Figure 1.11). Upon accumulating within lysosomes, any remaining entrapped cargo is degraded and recycled out of the cell. Because it is crucially important for siRNA nanocarriers to avoid retention within the endolysosomal pathway to be effective, much research has explored strategies to

28 maximize the endosomal escape efficiency of nanoparticles, including the use of lipid- based nanoparticles, synthetic cationic polymers, and cell penetrating peptides (CPPs).

Figure 1.11. Endocytosed siRNA nanocarriers must avoid retention within the endolysosomal system to effectively deliver their cargo to RISC in the cytosol. Modified from Wittrup et al. (2015) Nat Biotechnol [116].

1.3.2 Lipid-Based Nanoparticles for siRNA Delivery Lipid-based nanocarriers were among the first vehicles under investigation for siRNA and remain the furthest along in clinical development [91,104,105,117]. Lipid nanoparticles (LNPs) generally contain a mixture of cationic lipids to promote association with cell membranes and helper phospholipids that enhance transfection efficiency by fusing with cell and endosomal membranes and PEGylated lipids to improve the circulation time and colloidal stability of LNPs [118]. Since their initial development, the cytocompatibility and biodegradability of LNPs has dramatically improved; however, they are known to induce nonspecific immunotoxicity at high doses [118]. Among sixteen LNP-formulated siRNAs in clinical trials as of 2016

29 [118], Alnylam Pharmaceuticals became the first to announce positive phase III trial results for an siRNA therapeutic investigating patisiran in hereditary transthyretin amyloidosis [119]. The success of LNPs for siRNA delivery to date warrant their further development as therapeutics for aggressive diseases such as GBM.

1.3.3 Polymer-Based Nanoparticles for siRNA Delivery Cationic polymers also offer an attractive approach to siRNA delivery. Cationic polymers can efficiency bind and condense negatively charged siRNAs to form nanoscale polyplexes capable of entering cells through nonspecific mechanisms due to their strong positive surface charge. Further, cationic polymers with high buffering capacities can efficiently induce endosomal escape. One proposed mechanism for this is that buffering induces ion and water accumulation within the endosome until osmotic pressure increases to rupture the endosome, known as the “proton sponge effect.” While much research supports this hypothesis [120–123], it has been demonstrated that this effect is not accompanied by a change in lysosomal pH [124], and the validity of this theory remains controversial. As a potent polycation containing a high density of amine groups, polyethylenimine (PEI) is among the most widely investigated synthetic polymers for nucleic acid delivery [125,126]. Further, PEI is attractive from an engineering standpoint for its wide availability in a range of molecular weights in linear or branched configuration, low cost, and ease of modification for multimodal delivery systems [126]. PEI polyplexes delivering siRNA against VEGF demonstrated the ability to reduce angiogenesis and subsequent tumor growth in a murine model of Ewing’s sarcoma [127]. Additionally, intrathecally administered PEI polyplexes delivering siRNA against a nociceptor subunit effectively reduced formalin-induced

30 pain in rat models, demonstrating its efficacy in the central nervous system [128]. PEI polyplexes have been further evaluated against murine GBM models; PEI-mediated survivin silencing successfully impaired the growth of subcutaneous GBM tumors, improving the efficacy of co-administered doxorubicin and prolonging overall survival [129]. However, PEI and other polycationic materials are plagued by toxicity that precludes their clinical utility. This high toxicity is due to the presence of primary amines that impart a high positive charge. Indeed, PEI destabilizes cellular membranes, induces mitochondrial dysfunction, and may trigger complement activation in vivo [125]. This is particularly true for higher molecular weight PEI molecules, and some research has demonstrated that low molecular weight PEI exhibits decreased toxicity while retaining transfection efficiency [130]. Because of the robust ability of PEI to deliver nucleic acids, numerous efforts are ongoing to modify PEI to mitigate its cytotoxic effects while retaining its potent nucleic acid delivery efficacy [125,126,131]. For example, PEGylated PEI molecules have been employed to deliver siRNA, though this strategy was interestingly not translatable to PEI-mediated delivery of DNA [132]. Additionally, full deacylation of PEI not only enhanced its biocompatibility, but also increased gene delivery efficiency and specificity to mouse lungs [133]. Therefore, while PEI is unlikely to provide clinical utility in its unmodified state, its success as a delivery agent for siRNA warrants continued efforts to reduce its toxicity towards clinical application.

1.3.4 Spherical Nucleic Acid Nanoparticles for siRNA Delivery Of the most recently developed nanoparticle-based nucleic acid delivery platforms, spherical nucleic acids (SNAs) have emerged as a promising tool for

31 nucleic acid delivery, enabling their use in various biomedical applications [134–136]. SNAs consist of a nanoparticle core functionalized with a densely packed shell of radially-oriented nucleic acids. The three-dimensional structure of SNAs confers unique physicochemical and biological properties that favor their use as nucleic acid delivery vehicles. The first SNAs were developed using 13 nm solid gold spheres coated with thiolated single-stranded DNA and were used to guide nanoparticle assembly into larger ordered structures [137]. Gold nanoparticles were chosen as an ideal SNA core due to their ease of synthesis, tailorable and homogeneous size, simple gold-thiol conjugation chemistry, and unique optical properties. Since then, next- generation SNAs have been developed as gene regulatory agents carrying antisense DNA [138,139], siRNA [140–143] and miRNA [144] templated onto various nanoparticle cores including gold, quantum dots, iron oxide, silica, crosslinked polymer cores and micellar polymers, proteins, and chemotherapy drugs [134– 136,145]. These constructs are depicted in Figure 1.12. Mechanistic studies have demonstrated that the unique properties of SNAs are independent of the core and are purely dictated by the unique architecture of the nucleic acid shell [146,147].

Figure 1.12. SNAs consist of an inner core material densely coated with radially- oriented nucleic acids. Reproduced from Kapadia, Melamed, Day (2018) BioDrugs [135], https://mirkin-group.northwestern.edu/project/spherical- nucleic-acids/.

The unique architecture of SNAs imparts them with biological properties distinct from those of bare, linear nucleic acids. Unlike bare nucleic acids, SNAs can be taken up by almost any cell type (>50 cell types to date) without the need for auxiliary transfection agents (Figure 1.13) via a lipid-raft dependent and caveolae- mediated pathway [138,146]. SNAs are nontoxic to cells, illicit minimum innate immune response (25 fold less than lipoplexes carrying the same DNA), and they have no effect on blood chemistry in vivo [143,148]. Together, these unique properties allow SNAs to be utilized as tools for diverse biomedical applications.

Figure 1.13. SNAs accumulate within cells without the need for auxiliary transfection agents, unlike bare nucleic acids. Reproduced from Choi et al. (2013) PNAS [146].

Among the most successful applications of SNAs has been regulating gene expression in preclinical models of GBM tumors. These include SNAs delivering siRNA against Bcl2L12 [143], an anti-apoptotic gene overexpressed in GBM, or SNAs delivering mimics of miR-182 [144], a miRNA downregulated in GBM that suppresses Bcl2L12. Unexpectedly, these constructs demonstrated excellent BBB penetration in both an in vitro BBB model (Figure 1.14) and in in vivo studies. Excitingly, both siRNA- and miRNA-based SNAs could enter GBM tumors in mice following intravenous administration to suppress intratumoral Bcl2L12 expression, and this slowed tumor growth and prolonged animal survival [143,144]. Additionally, these SNAs exhibited no apparent toxicity, as evidenced by histopathology of various tissues and analysis of blood chemistry [143,144]. SNAs have also been developed to suppress the DNA repair protein MGMT, which potentiated the effects of co- administered temozolomide in murine tumor models [149]. Further, their successful delivery of siRNA and miRNA to glioblastoma tumors has prompted the first clinical trial evaluating their use as therapeutics for glioblastoma and gliosarcoma [150].

34 Together, these findings demonstrate the immense potential of SNAs as a new treatment for GBM, either alone or in combination with other modalities.

Figure 1.14. SNAs exhibit efficient BBB penetration using an in vitro BBB model. A) Schematic depicting noncontact in vitro BBB model consisting of co- cultured BMECs and astrocytes. SNAs were introduced to the luminal side of BMECs and allowed to measure accumulation within astrocytes, requiring passage through the endothelial/transwell barrier. B) Cy5-SNAs accumulate with BMECs expressing vimentin and occluding, characteristic of the BBB in vivo. C) Cy5-SNAs accumulate within GFAP-expressing astrocytes after successfully traversing the BMEC/transwell barrier. Reproduced from Jensen et al. (2013) Sci Trans Med [143].

However, one major limitation of SNAs is that they lack a material or bioactive component to directly facilitate their escape from endocytic compartments. Correspondingly, the mechanism by which SNAs escape endosomes to regulate gene expression remains unknown. Studies elucidating the mechanisms by which cells internalize SNAs found that SNAs engage scavenger receptors on cell surfaces to

35 trigger endocytosis and subsequently accumulate within late endosomes, as shown in Figure 1.15 [146,147]. Maximizing endolysosomal escape is a bottleneck that must be overcome to recognize the full clinical potential of SNAs. Future research incorporating materials such as cationic polymers and cell penetrating peptides into SNA design may improve their cytosolic delivery, but in utilizing these materials, delivery efficacy must be balanced with cytotoxicity [105,151]. Chapter 4 and 5 of this thesis explore one potential polycation-SNA hybrid material as a platform for improved siRNA delivery.

Figure 1.15. Endocytosed SNAs accumulate primarily within late endosomes. Modified from Wu et al. (2014) JACS [147].

1.4 Conclusions: Opportunities to Advance Glioblastoma Therapy with Hedgehog-Targeted Polycation-Spherical Nucleic Acid Nanoparticles This thesis seeks to explore novel therapies for GBM targeting the Hedgehog signaling pathway using an RNAi-based strategy. Chapter 2 describes the procedures

36 used to synthesize the nanoparticles employed throughout this thesis. GBM therapy often fails due to intrinsic or acquired chemoresistance, so in Chapter 3, Hedgehog/Gli1 signaling is investigated as a novel therapeutic target to improve the response of GBM cells to TMZ chemotherapy and eliminate GSCs. This initial work was done using commercially available transfection agents to deliver Gli1-targeted siRNA to GBM cells in order to thoroughly evaluate the consequences of Gli1 silencing on GBM cells and dissect the signaling pathways responsible for mediating those effects. Next, a novel nanoparticle platform was developed to enable RNAi targeting Gli1. In Chapter 4, the role of nanoscale architecture in siRNA delivery is investigated to maximally promote gene silencing while maintaining high biocompatibility. Specifically, the biological interactions of PEI-siRNA polyplexes were compared to that of PEI-coated SNAs (PEI-SNAs). This work highlights the importance of nanoscale architecture in improving the cellular uptake, cytosolic delivery, and gene silencing potency of siRNA nanocarriers while simultaneously alleviating the toxicity of polycations. Finally, Chapter 5 describes the capacity of Gli1-targeted PEI-SNAs to reduce the stemness and chemoresistance of GBM cells. Chapter 6 summarizes the importance of these findings and describes future directions to further the advancement of Hedgehog-targeted RNAi therapeutics for GBM management.

37 Chapter 2

METHODS FOR NANOPARTICLE SYNTHESIS AND CHARACTERIZATION

2.1 Introduction Gold nanoparticles (AuNPs) have been exploited for their unique optical, electronic, and chemical properties since ancient Roman times when they were used to stain glass for their vibrant red color. Since then, biomedical researchers have developed AuNPs as drug delivery vehicles, image contrast agents, and photothermal therapeutics, relying on their ease of synthesis, characterization, and functionalization, excellent long-term stability, and biocompatibility [152]. This chapter presents methods for synthesizing 15 nm AuNPs, functionalizing them with siRNA and poly(ethylene) glycol (PEG), subsequently coating them with PEI, and characterizing each step of this process. Sections of this chapter have been adapted from Melamed et al. (2017) Biomedical Nanotechnology. Methods in Molecular Biology [153].

2.2 Preparation of Gold Nanoparticles AuNPs were synthesized using the method developed by Frens [154] based on results obtained by Turkevich [155]. Briefly, this method involves the reduction of gold (III) chloride by sodium citrate to produce a solution of monodisperse, citrate- stabilized AuNPs. This occurs as a three-step process: 1) gold chloride is reduced to ionic gold, which increases in concentration until supersaturation induces 2) nucleation and the formation of AuNPs, and then 3) AuNPs grow until the reaction

38 stops (Figure 2.1). Reducing the amount of sodium citrate in the reaction allows control over the size of resulting AuNPs (reducing citrate increases AuNP size), which can range from 12 – 150 nm [154]. The method presented here will produce 15 nm AuNPs.

Figure 2.1. The Frens method for gold nanoparticle synthesis occurs as a three-step process: reduction, nucleation, and growth.

2.2.1 Frens Method for Gold Nanoparticle Synthesis To begin, a 1 L flask with a port and septum was rinsed with aqua regia, a fuming reaction of 3 parts hydrochloric acid to 1 part nitric acid, and then with DI water. Gold chloride (0.197 g) was added to 500 mL of DI water in the flask and heated to boiling under reflux while stirring vigorously. Sodium citrate (0.57 g) was dissolved in 1-2 mL DI water and injected rapidly into the flask. The reaction was allowed to proceed for 15 minutes under reflux, and the solution was removed from heat to cool to room temperature upon reaction completion, as indicated by color

39 stabilization. The AuNP solution was filtered through a 0.45 µm filter and transferred to a media bottle for storage at room temperature.

2.2.2 RNase Deactivation RNases in the AuNP solution must be deactivated prior to functionalizing with siRNA. Diethylpyrocarbonate (DEPC) was added to the AuNP solution (0.1% v/v) and heated to 37ºC for several hours to deactivate RNases. After several hours, DEPC was deactivated by autoclaving the solution at 121ºC for 40 minutes. The RNase-free AuNP solution was stored at room temperature.

2.3 Synthesis of Gold-Based Spherical Nucleic Acid Nanoparticles SNAs were prepared according to previously reported methods, detailed below [141,153].

2.3.1 siRNA Duplex Preparation siRNA sense and antisense strands were suspended in nuclease-free duplex buffer (30mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and 100mM potassium acetate, pH 7.5). The sense sequences contained a 3’ 36-unit PEG spacer and thiol to allow binding to the gold nanoparticle surface. Sense and antisense strands were mixed in a 1:1 molar ratio, heated to 95ºC, and slowly cooled to 37ºC over 1 hour to form duplexes. Duplexed siRNA was stored on ice prior to immediate usage or at -80ºC for long-term storage.

2.3.2 Gold Nanoparticle Functionalization with siRNA and Polyethylene Glycol RNase-free AuNPs (10 nM) were suspended in 0.2% Tween-20 and 150 mM NaCl. siRNA duplexes were added at a concentration of 1 nmol siRNA per 1 mL 10

40 nM AuNPs, and the solution was sonicated for ~10 seconds and gently rocked at room temperature for 4 hours. The concentration of NaCl was increased to 350 mM, briefly sonicated, and rocked overnight. Particles were passivated by adding 2 kDa methoxy- PEG-thiol to a concentration of 30 µM, then briefly sonicated and rocked for 4 hours. Unbound molecules were removed by centrifuging the particles at 21,000 x g for 30 minutes at 14ºC and resuspending in nuclease-free PBS three times. Purified SNAs were stored at ~100 nM in PBS at 4ºC until use. This procedure is summarized in Figure 2.2.

Figure 2.2. Schematic representing siRNA conjugation to gold nanoparticle surfaces. Reproduced from Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153].

2.4 Coating Spherical Nucleic Acids with Polyethylenimine SNAs were diluted to 10 nM in DI water, then 25 kDa branched PEI was added to a final concentration of 1 mg/mL. The solution was sonicated at room temperature for 15 minutes, then purified once by centrifugation at 21,000 x g for 30 minutes at 14ºC. PEI-coated SNAs (PEI-SNAs) were resuspended in nuclease-free PBS and stored at 4ºC until use.

41 2.5 Characterization Methods Methods used to characterize nanoparticles at each synthesis step include electron microscopy, UV-visible spectroscopy, dynamic light scattering, and zeta potential analysis. An OliGreen assay is used to quantify siRNA loading onto AuNPs. These methods are described below.

2.5.1 Electron Microscopy Transmission electron microscopy (TEM) was used to determine the size of uncoated AuNPs and visualize the structure of SNAs and PEI-SNAs. Samples were prepared for TEM imaging by incubating a drop of sample solution on a poly-L- lysine-coated 400 mesh formvar copper grid (Electron Microscopy Sciences, Hatfield, PA). To visualize the siRNA shell on SNAs and PEI-SNAs, grids were with 2% uranyl acetate. Imaging was performed using a Zeiss LIBRA 120 TEM. Representative TEM images of bare 15 nm AuNPs and SNAs are shown in Figure 2.3.

Figure 2.3. TEM images showing bare 15 nm AuNPs (left) and SNAs counterstained with uranyl acetate to visualize the siRNA shell (right).

42 2.5.2 UV-Visible Spectroscopy UV-visible spectroscopy was used to measure the extinction of AuNP solutions, which is useful for determining the concentration and aggregation state of AuNPs. AuNP solutions were diluted in DI water and analyzed using a Cary 60 spectrophotometer (Agilent Technologies, Santa Clara, CA). Functionalized AuNPs exhibit a red-shift in their extinction spectra relative to uncoated AuNPs, which is shown for AuNPs, SNAs, and PEI-SNAs in Figure 2.4.

Figure 2.4. UV-visible spectroscopy characterizing AuNPs, SNAs, and PEI-SNAs. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

2.5.3 Dynamic Light Scattering and Zeta Potential Analysis Dynamic light scattering (DLS) and zeta potential measurements were made using a Litesizer500 (Anton Paar, Graz, Austria). DLS was used to measure the hydrodynamic diameter of nanoparticles. Nanoparticles suspended in solution are constantly undergoing random Brownian motion, and the speed at which Brownian motion occurs is dependent on the size of the nanoparticles, including the core material and surface-bound molecules. Larger particles exhibit slower Brownian

43 motion, while smaller particles exhibit faster Brownian motion. DLS records the light scattered by particles over time to measure the hydrodynamic diameter of particles, which is larger than the true diameter measured by TEM. Zeta potential measurements are used to measure the charge of nanoparticles. Charged nanoparticles in solution adsorb an electrical double layer containing tightly bound, oppositely-charged ions close to the surface and more loosely bound ions further away, and the edge of this layer is referred to as the slipping plane. These ions move with the nanoparticle as moves in solution. Zeta potential is the difference in potential between the slipping layer and the surrounding solvent. Greater zeta potential magnitude indicates increased sample stability, and zeta potential measurements were used to confirm that AuNPs were successfully coated with negatively charged siRNA or positively charged PEI.

2.5.4 Quantifying siRNA Loading onto Particles The number of RNA duplexes per SNA was determined using a Quant-IT OliGreen assay (ThermoFisher Scientific, Waltham, MA) as previously described [153]. To measure antisense loading, SNAs were heated to 45ºC in 8 M urea to dehybridize the siRNA. SNAs were diluted in 0.05% Tween-20 and pelleted by centrifugation, and the antisense-containing supernatant was collected. Supernatant samples were incubated with components of the OliGreen kit and the sample fluorescence intensity was compared to that of an antisense standard curve to determine the number of antisense strands per SNA. This process is depicted in Figure 2.5. To measure sense loading, SNA pellets were collected following siRNA dehybridization and antisense removal, then washed by centrifugation (21,000 x g, 30 minutes, 14ºC) to remove any remaining antisense strands. Sense-containing pellets

44 were resuspended in 2 M β-mercaptoethanol, wrapped tightly in foil, and gently rocked at room temperature overnight to displace the thiol from the AuNP surface. The solution was diluted in 0.05% Tween-20 and centrifuged (21,000 x g, 30 minutes, 14ºC) to pellet AuNPs, and the sense-containing supernatant was collected. Supernatant samples were incubated with components of the OliGreen kit and the sample fluorescence intensity was compared to that of a sense standard curve to determine the number of sense strands per SNA. Representative antisense and sense loading is shown in Figure 2.6.

Figure 2.5. Schematic depicting an OliGreen assay to quantify siRNA loading onto gold nanoparticle surfaces. Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153].

Figure 2.6. Representative antisense and sense strand loading into 15 nm AuNPs. Reproduced from Melamed et al. (2017) Biomedical Nanotechnology. Methods Mol Bio [153].

2.5.5 Quantifying Polyethylenimine Loading onto Particles The amount of PEI per SNA was determined using a 2,4,6-trinitrobenzene sulfonic acid TNBS assay for colorimetric amine detection as described previously [156]. After coating SNAs with PEI, PEI-containing supernatant was collected and diluted in MES buffer (0.1 M MES, 0.15 M NaCl, pH 4.7). Samples were incubated with 0.025 w/v% TNBS (30 min at 37ºC, pH 9.5), and the reaction was stopped with sodium dodecyl sulfate (SDS) and HCl. The sample absorbance at 344 nm was measured and compared to that of a PEI standard curve to determine the amount of PEI remaining in the supernatant following PEI-SNA collection.

2.6 Conclusions The methods in this chapter describe the synthesis and characterization of AuNPs and functionalization with siRNA and PEI to form SNAs and PEI-SNAs, respectively. The synthesis and characterization methods described in this chapter

46 were used throughout this dissertation to prepare nanoparticles and to determine appropriate dosing of AuNPs, siRNA, and PEI for all studies.

47 Chapter 3

EVALUATION OF HEDGEHOG SIGNALING AS A THERAPEUTIC TARGET FOR GLIOBLASTOMA

3.1 Introduction Glioblastoma (GBM) is the deadliest form of brain cancer and represents the most common central nervous system tumor in adults, accounting for 45.6% of all malignant primary brain tumors [17]. Standard treatment for newly diagnosed primary GBM includes surgical resection, radiation, and chemotherapy [157,158]. Temozolomide (TMZ) is the current frontline chemotherapeutic for GBM and acts as an alkylating agent to induce DNA damage and trigger cell death. While the addition of TMZ chemotherapy to the standard radiotherapy has increased median patient survival time from 12.1 months to 14.6 months, tumor recurrence remains virtually inevitable, and nearly 100% of patients ultimately succumb to disease [19]. GBM recurrence is driven in large part by intrinsic or acquired chemoresistance [18], and tumors that have recurred often demonstrate enhanced resistance to chemotherapy and are more invasive than the primary lesion [157]. GBM drug resistance is mediated both by signaling mechanisms throughout the tumor and by an aggressive subpopulation of tumor cells known as glioma stem- like cells, or GSCs. In both GSCs and non-GSCs, deregulated DNA damage repair mechanisms oppose TMZ-mediated cytotoxicity by repairing DNA damage caused by TMZ to prevent apoptosis [159,160]. In particular, upregulated MGMT, an enzyme that removes alkyl groups transferred to DNA by TMZ, is a known prognostic

48 indicator of poor response to TMZ [40]. Additionally, mutations to the p53/Mdm2/PTEN tumor suppressor axis promote cell cycle progression to suppress TMZ-induced apoptosis [36,161]. Further, upregulation of EGFR signaling through wild-type EGFR and the commonly expressed EGFRvIII mutant stimulate survival signaling through the Ras/Raf/MAPK and PI3k/Akt/mTOR pathways to promote TMZ resistance [28,162]. In parallel, TMZ resistance is provided by a subpopulation of multipotent, slow-cycling GSCs that evade chemotherapeutic apoptosis and repopulate the tumor with resistant cells following treatment [51,163,164]. While much remains unknown regarding their biology, these cells may be identified by various markers (CD133, CD44, nestin, ALDH1, Oct4, Sox2), and differential expression of these markers suggest a stemness hierarchy that may correlate with aggressiveness [10]. Elimination of GSCs remains a barrier to complete tumor eradication. A growing body of evidence attributes tumor resistance phenotypes and the maintenance of GSCs to upregulated developmental pathways, such as Notch, Wnt, and Hedgehog (Hh) signaling [50,55,58,165,166]. In particular, research has demonstrated that while several developmental pathways contribute to GBM progression, Hh signaling is indispensable for GSC proliferation and tumorigenesis [58] and may be an impactful target for new GBM treatment strategies. Hh signaling is a highly conserved developmental pathway that is normally minimally active in differentiated adult tissue, but also plays an impactful role in the progression and maintenance of several cancers, including GBM. Hh signaling is activated when Hh ligands, most notably sonic hedgehog (Shh), bind the extracellular domain of the Patched (Ptc) transmembrane receptor. Ptc then relieves its suppression of

49 Smoothened (Smo), a second transmembrane protein, which in turn initiates an intracellular signaling cascade that translocates members of the Gli family of zinc finger transcription factors to the nucleus to regulate the expression of target genes. Within this family of transcription factors, upregulated Gli1 expression and transcriptional activity is associated with poor patient prognosis in several cancers [59]. Further, Gli1 is known to regulate proliferation and cell cycle progression [56,60], cell survival [60,61], migration and invasion [62], cancer cell stemness and self-renewal [56], and response to chemotherapy [55,56,63]. In this study, we investigated the role of Hh/Gli1 signaling in GBM resistance to TMZ. Specifically, we evaluated the molecular and phenotypic consequences of Gli1 inhibition as an adjuvant to TMZ treatment to assess the therapeutic potential of this co-treatment strategy. In this work, we used U87-MG and T98G cells as established in vitro models of GBM. These models were chosen because they both exhibit active Hh signaling as indicated by Gli1 expression and nuclear localization, but they differ in the expression of known molecular contributors to TMZ resistance. For example, U87-MG cells express wild-type p53, while T98G cells express a mutant p53 variant [161]. Although the role of p53 variants in GBM are not fully understood, evidence suggests that wild-type p53 generally retains tumor suppressive functions, while mutant p53 may promote tumor progression [35,161,167] Additionally, T98G cells, but not U87-MG cells, express high levels of MGMT, which is a primary mechanism by which GBM cells resist alkylating chemotherapies [161,168,169]. Because MGMT contains a Gli1 binding domain and consequently may be regulated by Hh signaling [170], MGMT expression may influence GBM cell response to co- treatment with Hh/Gli1 inhibitors and TMZ. Thus, we aimed to capture these key

50 phenotypic differences characteristic of GBM resistance mechanisms with our choice of established cell models. In this chapter, we show that silencing Gli1 prior to treating cells with TMZ increases the cytotoxicity of TMZ against GBM cells. We provide additional evidence that silencing Gli1 expression reduces the proliferation of U87-MG and T98G cells to abrogate disease progression. We also demonstrate that silencing Gli1 promotes sensitivity to TMZ by broadly reducing efflux behavior attributed to multidrug transporters. Further, we show that Hh pathway inhibition induces the expression of wild-type, but not mutant p53, suggesting that silencing Gli1 may induce tumor suppression via a p53-dependent mechanism. We initially hypothesized that Gli1 silencing without TMZ co-treatment would induce apoptosis via p53, however, we observed activation of separate tumor suppressive pathway. Specifically, we found that silencing Gli1 induces senescence rather than apoptosis, and this occurs via a mechanism that depends on the absence of PTEN. Finally, we demonstrate that combined Hh/Gli1 inhibition and TMZ treatment induces apoptosis and suppresses the growth of U87-MG cells cultured as neurospheres, suggesting an abrogation of glioma stem cell-like behavior. In aggregate, this data warrants the continued investigation of Hh-targeted therapies as adjuvants for GBM management. This chapter contains sections adapted from Melamed et al. (2018) Oncotarget [171].

51 3.2 Materials and Methods

3.2.1 Cell Culture and Transient Transfections U87-MG and T98G cells were purchased from American Type Culture Collection (ATCC, Manassas, VA) and cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Cells were maintained in a humidified incubator at 37°C, 5% CO2. RNA interference was performed using Dharmafect 1 (Dharmacon, Lafayette, CO) according to the manufacturer’s protocol to deliver 100 nM siRNA (Integrated DNA Technologies,

Coralville, IA) to cells at approximately 50-60% confluence. Four hours post- transfection, media containing transfection reagents was removed and replaced with fresh media, and cells were incubated at 37°C, 5% CO2 for 72 hours prior to continued experimentation. siRNA sequences used in this study are as follows (written 5' to 3'): siGli1 sense: CCA GGA AUU UGA CUC CCA ATT, siGli1 antisense: UUG GGA GUC AAA UUC CUG GCT, siScramble (siScr) sense: GUG CAC CAA CGA CUU AUC ATT, siScr antisense: UGA UAA GUC GUU GGU GCA CT. Plasmid DNA transfections were performed using Mirus TransIT X2 (Mirus Bio LLC, Madison, WI) to deliver 25 ng/mL pcDNA3-FLAG PTEN (Addgene #78777) or pcDNA3-FLAG HA (Addgene #10792) to cells at approximately 80% confluence. Twenty-four hours post-transfection, media containing transfection reagents was removed and replaced with fresh media, and cells were incubated at 37°C, 5% CO2 for 48 hours prior to analysis.

3.2.2 Immunofluorescent Staining and Image Analysis Cells were seeded on coverslips in 24-well plates and grown overnight prior to treatment with either Hh activating agents or Hh inhibitors. For Hh activation

52 experiments, cells were treated with recombinant human sonic hedgehog (rhShh) (R&D Systems, Minneapolis, MN) at 25 ng/ml to induce Hedgehog signaling and were incubated 48 hours at 37°C, 5% CO2. For chemical Hh inhibition experiments, cells were treated with 10 µM GANT61 (Cayman Chemical Company, Ann Arbor,

MI) and were incubated 48 hours at 37°C, 5% CO2. Subsequently, coverslips were washed in phosphate buffered saline (PBS), fixed in ice-cold acetone (for Gli1 staining) or 4% formaldehyde (for p53 staining), and blocked in PBS containing 1% bovine serum albumin (BSA), 0.2% cold-fish gelatin, and 0.1% Tween-20 for 1 hour at room temperature. Coverslips were then probed with antibodies under the following conditions: 1:500 rabbit anti-Gli1 (Proteintech, Rosemont, IL) or 1:500 mouse anti- p53 (Santa Cruz Biotechnology, Santa Cruz, CA) diluted in blocking buffer at 4°C overnight, followed by incubation with AlexaFluor 488-conjugated secondary antibodies diluted in blocking buffer (1:1000, ThermoFisher Scientific, Waltham, MA) for 1 hour at room temperature. Cells were rinsed in PBS supplemented with 0.1% Tween-20, counterstained with 4',6-diamidino-2-phenylindole (DAPI) to identify nuclei and phalloidin to identify actin cytoskeleton (Cell Signaling Technology, Danvers, MA), mounted on slides using gelvatol, and imaged on a Zeiss AxioObserver.Z1 microscope (Zeiss, Thornwood, NY). For each condition, 25 images were acquired from the central region of the coverslip using automated X, Y, and focus positioning. Each image set was checked by a masked observer for out of focus images, which were discarded from analysis, resulting in sets of 22-25 images encompassing a minimum of 1000 cells per set. Each image was analyzed using a custom image analysis script in MATLAB (2016a; Mathworks, Natick, MA). Briefly, the image was

53 segmented into nuclear and cytoplasmic regions and mean staining intensity was determined for each region. To identify the nuclear compartment, the DAPI stain was normalized and thresholded. To identify the cytoplasmic compartment, the phalloidin stain was normalized, morphologically closed with a 6 µm disk element, and thresholded. The thresholded stains were used as masks to determine the mean intensity of Gli1 and p53 staining in the two components. Data was analyzed frame by frame and averaged within each set.

3.2.3 Western Blotting Protein samples were prepared by lysing cells in radioimmunoprecipitation assay (RIPA) buffer (Amresco, Solon, OH) supplemented with 2X Halt Protease and Phosphatase Inhibitor Cocktail (ThermoFisher Scientific, Waltham, MA) on ice. Sample protein concentration was measured using a detergent-compatible modified Lowry assay (Bio-Rad, Hercules, CA) relative to a BSA standard. Lysate was denatured in Laemmli buffer (Amresco, Solon, OH) at 99°C for 20 minutes, and 30 µg protein was loaded per well in a Bolt® 4-12% Bis-Tris gel (ThermoFisher Scientific, Waltham, MA) and separated by electrophoresis at 135V for 1 hour. Protein was transferred to a 0.45 µm nitrocellulose membrane, which was subsequently blocked in 5% nonfat milk in tris-buffered saline containing 0.1% Tween-20 (TBST). Membranes were probed with primary antibodies at 4°C overnight, followed by incubation with anti-rabbit (1:25000) or anti-mouse (1:25000) horseradish peroxidase-conjugated secondary antibodies (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD) for 1 hour at room temperature. Primary antibodies were obtained from Cell Signaling Technology (Danvers, MA) (MGMT, α-tubulin, β-actin) or Santa Cruz Biotechnology (Rb, p21, PTEN, PUMA, Bid, MDM2). Protein bands were detected by

54 chemiluminescence using VisiGloTM Select Chemiluminescent Substrate (Amresco, Solon, OH) and imaged on a ChemiDoc-It2 Imager (UVP, Upland, CA).

3.2.4 Senescence Analysis Cells were seeded in 24-well plates at a density of 25,000 cells/well and cultured overnight prior to transfection with siGli1 or pPTEN (or relevant controls). Cellular senescence was analyzed using a Senescence Associated β-Galactosidase (SAβGal) kit (Cell Signaling Technology, Danvers, MA) according to the manufacturer’s protocol. Stained cells were imaged using a Zeiss Axioobserver.Z1 microscope equipped with a color camera. For each condition, 25 images were acquired from the central region of the well plate using automated X, Y, and focus positioning. SAβGal staining was quantified using a custom MATLAB script. Briefly, cells were segmented from brightfield images using the Sobel method. RGB images were converted to HSV colorspace, and positive SAβGal staining regions within cells were thresholded using hue values from 0.265 – 0.58. The total area of SAβGal- positive regions was divided by the total area occupied by cells to obtain the fraction of SAβGal-positive area for each condition.

3.2.5 Dose Response, Viability Assay, and Synergy Assessment GBM cytoxicity induced by temozolomide (TMZ, reconstituted in DMSO; Sigma-Aldrich, St. Louis, MO) chemotherapy was investigated following Hedgehog inhibition by silencing Gli1. Cells were seeded in 96-well plates at 5000 cells/well and transfected with siGli1 (or siScr) as described above. Subsequently, cells were treated with increasing TMZ dosages (0-1500 µM) for 48 hours, and viability was assessed using an AlamarBlue assay (ThermoFisher Scientific, Waltham, MA) according to the

55 manufacturer’s protocol with a 2-hour AlamarBlue incubation. Fluorescence intensity (Ex 550 nm/Em 585 nm) was recorded using a Synergy H1 Plate Reader (BioTek, Winooski, VT). Cells transfected with siScr and receiving no TMZ were taken to exhibit 100% metabolic activity, and cells treated with DMSO equivolume to the highest TMZ dosage diluted in media were used to verify that toxicity was due to TMZ treatment rather than DMSO-induced membrane permeabilization, since TMZ was reconstituted in DMSO prior to dilution in media. Therapeutic synergy between Gli1 silencing and TMZ was assessed using previously reported methods [172]. For each group co-treated with siGli1 and TMZ, a projected additive effect was calculated by multiplying the fraction of metabolic activity reduced by Gli1 silencing alone by the fraction of metabolic activity reduced by each TMZ dose. Then, the observed metabolic activities with co-treatment were compared to the projected values by one-way ANOVA with post-hoc Tukey. Co- treatments producing metabolic activities statistically greater than the projected value were taken to be antagonistic, statistically insignificant from the projected value were taken to be additive, and statistically lower than the projected value were taken to be synergistic.

3.2.6 Assessment of Drug Efflux Transporter Activity Rhodamine123 (Sigma-Aldrich, St. Louis, MO) efflux was measured to evaluate the impact of Hh/Gli1 inhibition on multidrug resistance (MDR) efflux transporter activity because it is a known substrate for membrane efflux transporters and its fluorescence is easily detectable [172,173]. Cells were seeded in 24-well plates and grown overnight, then transfected with siGli1 or siScr as described previously. Transfected cells were subsequently incubated with 5 µM Rhodamine123 for 20

56 minutes at 37°C, 5% CO2, then the Rhodamine123-containing medium was removed, cells were replenished with fresh media and incubated for another 20 minutes at 37°C,

5% CO2 to allow for efflux. Treated cells were harvested by trypsinization, fixed in 4% formaldehyde, resuspended in PBS and analyzed for Rhodamine123 fluorescence (Ex 488 nm/Em 533/30 nm) by flow cytometry using a BD AccuriTM C6 cytometer (Becton Dickinson Biosciences, Franklin Lakes, NJ).

3.2.7 Proliferation Assay Cellular proliferation following Hedgehog inhibition by Gli1 silencing was evaluated using a Click-iT® EdU Proliferation Assay (ThermoFisher Scientific, Waltham, MA). Cells seeded in 24-wells were transfected with siRNA as described previously and then incubated with 10 µM EdU for 16 hours at 37°C, 5% CO2. Cells were then harvested by trypsinization, washed in 1% BSA in PBS, fixed in 4% formaldehyde, permeabilized with 0.05% saponin, and stained according to the manufacturer’s protocol. EdU incorporation was measured by flow cytometry (Ex 488 nm/Em 533/30 nm) using a BD AccuriTM C6 cytometer (Becton Dickinson Biosciences, Franklin Lakes, NJ).

3.2.8 Neurosphere Growth and Apoptosis Analysis Neurospheres were grown from a single-cell suspension of U87-MG’s in NeuroCult NS-A (STEMCELL Technologies, Vancouver, BC, Canada) medium supplemented with recombinant human epidermal growth factor (EGF, 20 ng/mL), recombinant human basic fibroblast growth factor (bFGF, 10 ng/mL), and heparin sulfate (2 µg/mL). Cells were plated in ultra-low adhesion 24-well plates at a density of 10,000 cells/mL for growth analysis in medium containing 10 µM TMZ and 10, 15,

57 or 20 µM GANT61 for one week at 37°C, 5% CO2. Controls contained DMSO and/or ethanol (which are used to reconstitute TMZ and GANT61, respectively) at volumes equivalent to the highest TMZ and/or GANT61 doses used in the experiment. After one week, neurospheres were imaged for size and sphere forming efficiency analysis. Imaging was conducted on a Zeiss AxioObserver.Z1 microscope using automated stage control to image each well in its entirety. Following acquisition, images were stitched using Zeiss Efficient Navigation software (ZEN 2.0; Zeiss) and exported as single field images on the OME-TIFF standard and analyzed using a custom MATLAB script. Briefly, the well exterior was discarded by mapping an active contour to the high gradient region at the well edge [174]. After discarding the well exterior, varying background illumination was corrected using morphological reconstruction-based top-hat and bottom-hat filters [175] followed by homomorphic filtering. Image contrast was globally enhanced using homomorphic filtering and phase-preserving dynamic range compression and locally enhanced using the CLAHE algorithm [176,177]. Following enhancement, spheres were identified as increases in local image entropy and touching spheres were separated using watershed-based segmentation. Automated sphere identification was manually confirmed for each image analyzed. Sphere growth data were reported as the average number of spheres per treatment group across 9 independent replicates, where spheres were defined as objects with diameter greater than 50 µm and eccentricity less than 0.8, and as size (sphere projected area) distributions for each treatment group. To evaluate apoptosis in U87-MG neurospheres treated with TMZ ± GANT61, neurospheres grown from a single-cell suspension at a density of 25,000 cells/mL were harvested after 6 days growth in NeuroCult NS-A medium ± 50 µM TMZ ± 0- 15 µM GANT61, trypsinized

58 to a single-cell suspension, and analyzed using an AnnexinV-FITC/propidium iodide (PI) staining kit (Cayman Chemical Company, Ann Arbor, MI) according to the manufacturer’s protocol. Fluorescence intensity was measured by flow cytometry analysis using a BD AccuriTM C6 cytometer, with Annexin V-FITC recorded using Ex 488 nm/Em 533/30 nm (FL-1) and PI recorded using Ex 488nm/Em 670LP (FL-3). Measurements were appropriately corrected for spillover using standard color correction procedures.

3.2.9 Quantitative Real-Time Polymerase Chain Reaction (qPCR) qPCR was used to confirm that U87-MG neurospheres exhibit a stem-like phenotype relative to adherent cultures. Cells were grown as neurospheres as described previously for one week, then mRNA was isolated using an Isolate II RNA Mini Kit (Bioline, Taunton, MA). The relative expression of CD133, Nanog, and Sox2 were measured using SensiFASTTM SYBR® One-Step master mix and normalized to that of GAPDH. qPCR was performed on a LightCycler® 96 (Roche Diagnostics Corporation, Indianapolis, IN). Primer sequences were as follows: CD133 F - GGACCCATTGGCATTCTC, CD133 R - CAGGACACAGCATAGAATAATC, Nanog F – AAATTGGTGATGAAGATGTATTCG, Nanog R – GCAAAACAGAGCCAAAAACG, Sox2 F – AACATGATGGAGACGGTGCTGAA, Sox2 R – CAGCCGTTCATGTGCGCGTA. Gene expression in cells grown as neurospheres was compared to that of cells grown in adherent conditions.

3.2.10 Statistical Analysis Differences between two groups were assessed for statistical significance by Student’s t-test for each cell line independently. TMZ dose response data and sphere

59 growth data were analyzed using a one-way analysis of variance (ANOVA) with posthoc Tukey test using MATLAB software (MathWorks, Natick, MA). Neurosphere apoptosis data were analyzed using one-way ANOVA with post-hoc Fisher’s least significant difference test. Flow cytometry data was analyzed using either FlowJo software (Tree Star Inc, Ashland, OR) or CFlow software (Becton Dickinson Biosciences, Franklin Lakes, NJ). Differences were considered statistically significant at p < 0.05. Data shown represents the mean ± standard deviation of three independent experimental replicates.

3.3 Results and Discussion

3.3.1 U87 and T98G GBM Cells Exhibit Active Hh Signaling Required for Proliferation First, we validated that both U87-MG and T98G cells exhibit active Hh signaling. Nuclear localization of Gli1 was taken to indicate Hh pathway activation, as active Hh signaling produces Gli1 transcriptional activity and cytoplasmic Gli1 undergoes proteasomal degradation [178]. Cells were treated with recombinant human Shh (rhShh) for 48 hours and assessed for Gli1 expression using immunofluorescence. Images obtained using fluorescence microscopy reveal that Gli1 is present in both the nucleus and cytoplasm of untreated U87-MG and T98G cells, suggesting that Hh signaling is active in both cell lines. Further, stimulation with rhShh increases U87- MG Gli1 staining intensity by ~30% in the nucleus and ~40% in the cytoplasm (Figure 3.1AC). In contrast, Gli1 staining intensity is conserved with rhShh treatment in T98G cells (Figure 1BC), indicating that pathway activity is already maximal in untreated cells, that Gli1 is primarily regulated by Indian or Desert Hh ligands, or by noncanonical signaling mechanisms.

61 Gli1 is well understood to positively regulate cell proliferation through direct transcriptional upregulation of cyclinD/E, Rb tumor suppressor inhibitors, c-Myc, and other pro-proliferative genes [179–181]. Therefore, we hypothesized that silencing Gli1 should decrease the proliferation of both U87-MG and T98G cells. To test this, U87-MG and T98G cells were transiently transfected with siRNA against Gli1 (siGli1) or a scrambled control siRNA (siScr) and subsequently analyzed by an EdU incorporation assay. Flow cytometric analysis revealed that silencing Gli1 decreased the fraction of proliferating, EdU+ U87-MG cells by ~60% and the fraction of EdU+ T98G cells by ~44% (Figure 3.1D). These results further support the presence of Hh pathway activity in U87-MG and T98G cells even without stimulation with exogenous rhShh.

3.3.2 Silencing Gli1 Potentiates GBM Cell Response to TMZ by Decreasing Multidrug Efflux Activity Having established that Hh signaling is active in both of our GBM cell models and contributes to cell proliferation, we asked whether Hh/Gli1 activity contributes to cellular resistance to TMZ. To test this, U87-MG and T98G cells were transiently transfected with siRNA targeting Gli1 (siGli1) or a scrambled control (siScr), then treated with a range of TMZ doses (0 – 1500 µM). Treatment efficacy was evaluated using an AlamarBlue assay to measure cellular metabolic activity. In U87-MG cells, we found that silencing Gli1 prior to TMZ treatment decreases metabolic activity up to 30% for TMZ doses through 1000 µM, past which we no longer observed increased efficacy with co-treatment (Figure 3.2).

We further conducted simple analysis to determine whether the treatments employed in tandem were synergistic, additive, or antagonistic. Briefly, the projected additive effect was calculated by multiplying the metabolic activity fraction for each treatment (siGli1 or TMZ) individually. If the measured effect was greater than the projected additive, the co-treatment was considered synergistic. Conversely, if the measured effect was less than the projected additive, the co-treatment was considered antagonistic. Using this analysis method, we found that co-treating U87-MG cells with siGli1 and TMZ produces an additive decrease in metabolic activity for all TMZ doses tested (Figure 3.3). In T98G cells, while silencing Gli1 alone significantly reduced metabolic activity, co-treatment with TMZ induced an additive therapeutic effect at 250 µM TMZ, and an antagonistic therapeutic effect was exerted at higher TMZ doses (Figure 3.2, 3.3).

Figure 3.3 Assessment of additive or synergistic effects between Gli1 silencing and TMZ treatment. *p<0.05 by one-way ANOVA and post-hoc Tukey. Reproduced from Melamed et al. (2018) Oncotarget [171].

These results suggest that additional signaling mechanisms are important for regulating GBM cell response to TMZ, especially in T98G cells, but support that Hh inhibitors may improve the anticancer efficacy of low TMZ doses. We further sought to delineate mechanisms by which silencing Gli1 may improve GBM cell TMZ response. Previous research has demonstrated that Gli1 transcriptionally regulates several transmembrane efflux transporters, such as MDR1/ABCB1, MRP1, LRP, and BCRP/ABCG2, which mediate chemotherapy resistance in several cancers [63,182,183]. Therefore, we asked whether silencing Gli1 could translate to a net reduction in GBM cell multidrug efflux activity. To test this, we used a dye-based efflux activity assay in which intracellular rhodamine intensity is taken to indicate small molecule retention. Rhodamine123 is a suitable dye choice for these experiments because it is a known substrate for multidrug efflux transporters [173]. In these experiments, U87-MG and T98G cells were transiently transfected with siGli1 or siScr, then incubated with Rhodamine123 for 20 minutes. The dye was removed and cells were further incubated in fresh media for an additional 20 minutes.

64 Cells were then trypsinized and analyzed for Rhodamine123 intensity by flow cytometry. Flow cytometry analysis reveals that silencing Gli1 can increase the intracellular Rhodamine123 intensity by 2.5-fold in U87-MG cells and 1.5-fold in T98G cells (Figure 3.4). Therefore, we conclude that silencing Gli1 can reduce multidrug efflux activity to improve the retention of small molecule chemotherapeutics.

Figure 3.4: Silencing Gli1 influences reduces multidrug efflux activity of GBM cells. Silencing Gli1 reduces multidrug efflux activity in both U87-MG and T98G cells. Flow cytometry reveals that silencing Gli1 prior to incubating cells with Rhodamine123 increases cellular Rhodamine123 intensity by 2.5-fold and 1.5-fold in U87-MG and T98G cells, respectively (C). Data are shown as mean ± standard deviation, *p = 0.0008, #p = 0.07 by paired t-test relative to siScr. Reproduced from Melamed et al. (2018) Oncotarget [171].

3.3.3 Suppressing Hh Signaling Modulates p53 and MGMT Expression in GBM Cells Next, we asked whether Hh pathway inhibition could restore tumor suppressive activity in GBM cells. To evaluate tumor suppressive activity in U87-MG and T98G cells, we used immunofluorescent staining to assess the relative expression

65 and localization of p53. Wild-type p53 transcriptionally regulates multiple genes involved in apoptosis, cell cycle arrest, DNA repair, and senescence in response to DNA damage such as that induced by hypoxia, radiation, or chemotherapy [35]. Thus, upregulation of wild-type p53 should induce tumor suppressive functions to mediate GBM cell death in response to chemotherapy. The role of mutant p53 in GBM is more poorly understood, and mutant p53 may promote both tumor suppressive and tumorigenic cell signaling. Both wild-type and mutant p53 are represented in this study, where U87-MG cells express wild-type p53 and T98G cells express a mutant p53 variant [161]. Cells seeded on coverslips were treated with the pharmacological Gli inhibitor GANT61 [69] for 48 hours, then stained for p53 and analyzed by quantitative fluorescence microscopy. We demonstrate that GANT61 significantly increases nuclear p53 by 43.4% in U87-MG cells (Figure 3.5AC) and decreases nuclear p53 by 21.5% in T98G cells (Figure 3.5BC).

Figure 3.5: Hh inhibitors modulate p53 and MGMT expression in GBM cells. Nuclear p53 staining intensity increases with GANT61-mediated Hh inhibition in (A,C) U87-MG cells by 43.4%, but decreases in (B,C) T98G cells by 21.5% relative to that in control cells. Data are shown as mean ± standard deviation from 3 independent experiments, *p<0.005 by Student’s t-test. Scale bars = 50 µm. (D) By Western blotting, T98G MGMT expression decreases in a TMZ-dependent manner. TMZ-induced downregulation of MGMT may be potentiated by Gli1 silencing. Reproduced from Melamed et al. (2018) Oncotarget [171].

67 We further investigated the potential role of MGMT in T98G resistance to TMZ, since MGMT is expressed highly in T98G cells. One previous study investigating other GBM cell lines reported that MGMT can be up-regulated in response to TMZ [160], so we asked whether TMZ-mediated upregulation of MGMT could be driving TMZ resistance in T98G cells. T98G cells were transfected with siGli1 or siScr and subsequently treated with 250 – 500 µM TMZ for 48 hours, then lysed and assessed for MGMT expression by Western blotting. Our results show that MGMT expression is reduced in T98G cells treated with TMZ in a dose-dependent manner, and this MGMT downregulation may be enhanced when TMZ is combined with siGli1 (Figure 3.5D). Accordingly, other mechanisms may be predominantly responsible for T98G TMZ resistance under Hh pathway inhibition.

3.3.4 Silencing Gli1 Does Not Induce Apoptosis in GBM Cells without TMZ Co- Treatment Because Hh pathway suppression increased wt-p53 in U87-MG cells and decreased mut-p53 in T98G cells, we hypothesized that silencing Gli1 might induce apoptosis in GBM cells even in the absence of TMZ co-treatment. To test this, we first evaluated the expression of proteins downstream of p53 that might be upregulated in p53-mediated apoptosis. Contrary to our expectation, in U87-MG cells treated with siGli1, we observed decreases in Rb, p21, Bid, and MDM2 expression (Figure 3.6A). In T98G cells, we observed slight increases in PTEN and PUMA expression and decreases in Rb, Fas, Bid, and MDM2 expression by Western blotting (Figure 3.6A). To validate these surprising results and ensure that p53-mediated apoptosis was not occurring by another signaling mechanism, we used AnnexinV-FITC/PI staining to directly assess apoptosis in cells treated under identical conditions. Surprisingly, we

68 observed a slight but insignificant increase in the fraction of apoptotic cells (AnnexinV+, PI±) following Gli1 silencing, with 14 ± 3.6 % of cells treated with siGli1 staining AnnexinV+ versus 9.9 ± 0.9% of cells treated with siScr (Figure 3.6B). Therefore, we concluded that direct induction of apoptosis by a p53-dependent mechanism is unlikely to cause the previously observed decreases in proliferation and metabolic activity with Gli1 silencing.

Figure 3.6. Silencing Gli1 does not correlate with apoptosis in GBM cells. (A) By Western blotting, proteins associated with apoptosis induction do not increase with Gli1 silencing. (B) Gli1 silencing does not significantly increase AnnexinV-FITC/PI staining, # no significant difference in the fraction of viable, early apoptotic, late apoptotic, or necrotic cells by one- way ANOVA with posthoc Tukey. Reproduced from Melamed et al. (2018) Oncotarget [171].

70 3.3.5 Silencing Gli1 Induces Senescence in GBM Cells in a Manner Dependent on the Absence of PTEN Our studies produced the surprising result that although silencing Gli1 reduces GBM cellular metabolic activity alone and in combination with TMZ and modulates p53 expression, this does not appear to initiate apoptosis in the absence of TMZ co- treatment. Therefore, we asked whether we could be observing senescence rather than apoptosis in response to siGli1 alone. To test this, we used senescence associated β- galactosidase (SAβGal) staining to identify senescent cells following Gli1 silencing. We observed a marked increase in SAβGal staining (Figure 3.7A) in U87-MG cells treated with siGli1, but not in T98G cells. Interestingly, previous research has identified that loss of PTEN can induce premature senescence, which may have a compensatory role for apoptosis in this context [184]. Because U87-MG cells exhibit loss of PTEN while T98G cells do not (Figure 3.6A), we hypothesized that loss of PTEN might explain the observed senescent phenotype in response to siGli1. Indeed, we observed that inducing PTEN expression reversed this behavior; co-transfecting U87-MG cells with siGli1 and a control plasmid (pHA) significantly increased SAβGal staining by 2.3-fold relative to cells co-transfected with siScr and pHA, while co-transfection with siGli1 and pPTEN reversed this phenotype (Figure 3.7BC). Therefore, we concluded that silencing Gli1 induces senescence in a manner dependent on the absence of PTEN.

3.3.6 Hh Inhibition and TMZ Co-Treatment Promotes Apoptosis in Neurospheres and Impairs Neurosphere Growth Having established that Hh inhibition induces senescence in a loss of PTEN- dependent manner as a standalone therapy, we were interested in the fate of GSCs co- treated with Hh inhibitors and TMZ. To model this, we used an in vitro model of neurosphere formation in which neurospheres were grown from a single cell suspension in serum-free, chemically-defined medium. Importantly, GBM cells grown as neurospheres are more sensitive to clinically-relevant TMZ doses than GBM cells grown in adherent culture [185], so we anticipated that neurospheres would be a more appropriate model to elucidate the effects of Hh inhibition and TMZ co-treatment. We validated that U87-MG cells grown as neurospheres are more stem-like than adherent cells using qPCR to measure changes in the expression of genes associated with this phenotype (Figure 3.8).

73 We found that U87-MG neurospheres exhibit a 10-fold increase in CD133 expression and a 5-fold increase in Nanog expression (Figure 3.8). Next, we investigated the extent of apoptosis induced in U87-MG GBM neurospheres by co- treatment with GANT61 and TMZ using AnnexinV/PI staining. We found that, as a monotherapy, only the highest tested GANT61 dose (15 µM) is sufficient to induce a significant increase in apoptotic (AnnexinV+, PI±) cells relative to untreated controls (Figure 3.9), which increased the fraction of apoptotic cells to 41%. While TMZ alone induces apoptosis in 54% of cells, the addition of 10 or 15 µM GANT61 significantly increases the fraction of apoptotic (AnnexinV+, PI±) cells to 83% or 93%, respectively (Figure 3.9). Further, synergy analysis using the previously described method revealed that co-treating U87-MG neurospheres with TMZ and 10 or 15 µM GANT61 produced a statistically significant synergistic decrease in the fraction of viable (AnnexinV-, PI-) cells (Figure 3.10). This data supports the hypothesis that co- treatment with TMZ and GANT61 may cooperatively promote apoptosis in GBM neurospheres. We next investigated the consequences of GANT61/TMZ-mediated apoptosis on neurosphere growth. Using brightfield microscopy, we observed a striking decrease in neurosphere size induced by GANT61/TMZ co-treatment and correlating with increased doses of GANT61 (Figure 3.9CD, 3.10) Quantitative image analysis also revealed that GANT61 significantly reduced the number of neurospheres formed with or without co-treatment with TMZ (Figure 3.9CD).

75 Figure 3.9. Hh inhibitors and TMZ cooperatively promote apoptosis in U87-MG neurospheres and suppress neurosphere growth. (A) Flow cytometric scatterplots displaying AnnexinV-FITC and PI staining intensities in U87-MG cells grown as neurospheres in medium containing GANT61 (0 or 10 µM) and TMZ (0 or 50 µM TMZ) from one representative experiment. (B) The fraction of apoptotic (AnnexinV+) U87-MG cells grown as neurospheres in medium containing GANT61 and TMZ (50 µM TMZ), summarized across 3 independent experiments. Data shown are means ± standard deviation from 3 independent experiments, *p<0.05, **p<0.01 by one-way ANOVA with post-hoc Fisher’s least significant difference test. (C) Representative brightfield images of U87- MG neurospheres grown for one week in medium containing GANT61 and TMZ (10 µM TMZ). Scale bar = 500 µm. (D) (Left) Number of spheres counted by automated image analysis, averaged across 9 independent replicates. Data shown are means ± standard deviation from 9 independent experiments, *p<0.05, ** p<0.01 by one-way ANOVA with post-hoc Tukey test. (Right) Sphere size (projected area) as determined by automated image analysis from one representative experiment. Reproduced from Melamed et al. (2018) Oncotarget [171].

76 3.3.7 Discussion In this study, we evaluated the therapeutic potential of Hh pathway inhibition as an adjuvant to TMZ chemotherapy using in vitro models of GBM. Through this work, we sought to identify key molecular and phenotypic changes that occur in GBM cells because of co-treatment with Hh pathway inhibitors and TMZ. We found that silencing Gli1 before administering TMZ to GBM cells enhanced the cytotoxic effects of chemotherapy, and this effect is likely dependent on the status of additional GBM- promoting genes. Further, this enhanced TMZ cytotoxicity correlated with a decrease in multidrug efflux activity and an increase in wild-type p53, a key mediator of apoptosis, cell cycle arrest, and senescence [35]. Finally, we found that co-treating GBM cells in neurosphere culture with Hh pathway inhibitors and TMZ promotes apoptosis and reduces neurosphere formation, suggesting an inhibition of glioma stem cell-like behavior. Overall, this work provides evidence that Hh pathway inhibition may overcome cellular mechanisms that promote TMZ resistance, and the data warrant continued investigation of this combination treatment. Previous research has demonstrated that Hh inhibition through gene regulatory or pharmacological means can sensitize GBM cells to TMZ [55,56,186] and has suggested a number of possible mechanisms to support these findings. Although many researchers either do not observe MGMT expression in U87-MG cells [161,168,187] or observe very little MGMT expression [188,189], one study reported that GANT61- mediated Hh pathway suppression improved TMZ sensitivity by repressing MGMT and Notch proteins in U87 and U251 cells [186]. Consistently, another investigation reported that targeting both Hh and Notch signaling could further improve TMZ sensitivity in GBM cells [55]. In the present study, we demonstrate that silencing Gli1 prior to TMZ treatment can additively improve chemotherapeutic efficacy in U87-MG

77 cells cultured in adherent conditions (Figure 3.2). Notably, U87-MG cells are highly refractory to TMZ treatment in adherent culture. This is evident in that minimal U87- MG toxicity is observed for TMZ doses up to 500 µM, approximately 10-fold higher than the maximum clinically feasible TMZ dose [190]. Accordingly, it is encouraging that the therapeutic benefit afforded by co-treatment was most prominent for lower TMZ doses, suggesting that the combination of Hh inhibitors and low-dose TMZ may improve treatment outcomes while mitigating off-target toxicity. We can extend these findings to U87-MG cells grown in neurosphere culture, thought to be more representative of a GSC-rich population. Interestingly, we observed a synergistic, or greater than additive, therapeutic effect induced by co-treatment with GANT61 and TMZ in these models (Figure 3.9,3.10). We suspect this is because Hh signaling is upregulated in GSCs relative to non-GSC GBM cells [55] and neurospheres may consequently be more sensitive to its suppression. Additionally, some GBM cell lines grown as neurospheres exhibit greater sensitivity to chemotherapeutic agents including carmustine and TMZ than GBM cell lines grown in adherent culture [185]. Surprisingly, and in contrast to our observations in U87-MG cells, our studies show that silencing Gli1 prior to treating T98G cells with TMZ chemotherapy at doses greater than 250 µM produced an antagonistic therapeutic effect (Figure 3.2,3.3). Though we initially hypothesized that this could be due to MGMT upregulation, we found that siGli1/TMZ co-treatment decreases MGMT expression at TMZ doses greater than 250 µM (Figure 3.5), and we concluded that other mechanisms are responsible for this antagonistic therapeutic effect. We speculate that this may occur because silencing Gli1 may relieve Gli1-mediated suppression of mutant p53, which

78 can oppose TMZ cytotoxicity [35,36]. These studies warrant future investigation to delineate the cause of this differential drug response. Further, Hh signaling has been linked to the upregulation of drug efflux transporters that reduce the intracellular concentration of small molecule chemotherapeutics, rendering them ineffective and imparting drug resistance [63,182,183,191]. While several investigations have shown that Hh inhibition decreases the expression of one or more such transporters at the protein level, we demonstrate here that silencing Gli1 does produce a net decrease in multidrug efflux activity. In these studies, we used Rhodamine123 as a tracer dye to indicate efflux activity, as Rhodamine123 is a known substrate for such transporters. Our data show that silencing Gli1 prior to incubating U87-MG cells with Rhodamine123 can significantly improve intracellular dye intensity, and therefore retention, by ~2.5-fold. T98G cells exhibit a slight but statistically insignificant increase in Rhodamine123 retention (Figure 3.4). These results are consistent with our chemosensitization studies, which revealed that silencing Gli1 increases TMZ response in U87-MG cells to a much greater degree than in T98G cells. It is also of great importance to consider the impact Hh inhibition might have on signaling pathways with well-established significance in TMZ resistance. Both wild-type and mutant p53 have been implicated in the progression and therapeutic response of GBM tumors. Wild-type p53 is well known to mediate apoptosis, cell cycle arrest, and senescence in response to genotoxic stress through a complex tumor suppressive signaling network. Despite the presence of altered p53 in 25-30% of primary GBM cases [35], much remains unknown regarding the phenotypic consequences of p53 mutants on GBM progression. Previous research has identified a

79 Gli1-p53 negative feedback loop present in neural stem cells and brain tumor cells, where Gli1 knockdown was found to increase both p53 and active phospho-serine15 p53 in U87-MG cells [192]. However, additional signaling mechanisms also regulate the inverse relationship between Gli1 and p53. Upstream of Gli1, constitutively activated Smoothened mutants upregulate MDM2, which then represses p53 tumor suppression activity [193]. A Hh-p53 negative regulatory loop is further maintained due to competitive Gli1 and p53 binding to the coactivator TATA Binding Protein Associated Factor 9 (TAF9) [194] and Nanog-mediated upregulation of Gli1 and downregulation of p53 [195]. Consistent with these prior findings, we have demonstrated that pharmacological Gli inhibition can increase p53 expression U87- MG cells. However, we also report the surprising result that this relationship is not conserved in T98G cells (Figure 3.5A-C). Because Hh pathway suppression increased wt-p53 in U87-MG cells and decreased mut-p53 in T98G cells, we hypothesized that silencing Gli1 might induce apoptosis in GBM cells, even in the absence of TMZ co-treatment. To test this, we assessed the expression of proteins downstream of p53 known to mediate apoptosis. Surprisingly, we did not find increased expression of proteins involved in p53- mediated apoptosis (Figure 3.6A), nor did we observe significant increases in AnnexinV-FITC/PI staining with Gli1 silencing (Figure 3.6B). This unexpected result led us to consider senescence as an alternative explanation for the observed decreases in proliferation and metabolic activity with siGli1 alone. Interestingly, we did find that silencing Gli1 induced a senescent phenotype in U87-MG cells but not in T98G (Figure 3.7A). While silencing Gli1 and Gli2 has recently been demonstrated to induce senescence to restore chemosensitivity in melanoma cells [196], Hh signaling

80 has not yet been linked to senescence in GBM. Thus, our results provide novel insight into Hh signaling as a means of evading senescence in GBM cells. Further, previous research has identified that loss of PTEN can induce premature senescence, which may have a compensatory role for apoptosis in this context [184]. Here, we demonstrate that while siGli1 induces senescence, expressing PTEN simultaneously with Gli1 silencing can prevent this phenotype (Figure 3.7B). While p53 and loss of PTEN have both been implicated in glioma cell senescence, we report the novel finding that senescence can also be induced by Hh pathway suppression in the absence of PTEN. While much remains unknown about the role of senescence in tumor progression, senescence is thought to suppress tumor growth in that senescent cells no longer exhibit limitless replicative potential [197]. However, senescent cells that have incurred DNA damage also acquire a senescence-associated secretory phenotype (SASP), in which senescent cells secrete pro-inflammatory factors that can promote tumor progression [198]; this warrants future research on the effects of siGli1-induced SASP on GBM progression. Also heavily implicated in TMZ resistance, MGMT is a DNA repair enzyme that removes alkyl groups transferred to the O6 position of guanine by TMZ to oppose drug toxicity. Though MGMT expression is primarily regulated epigenetically [40], a Gli1-binding domain has been identified within the MGMT promoter in medulloblastoma [170], and Hh inhibition has been linked to MGMT downregulation in GBM cell lines [63,186]. Further, additional reports have sought to relate MGMT expression to TMZ treatment [160,188,199], though no consensus has been reached on whether TMZ modulates MGMT expression. For example, one study reported that exposing GBM cell lines to TMZ in a cyclic manner (3 days of treatment followed by

81 3 days without drug, repeated twice) resulted in MGMT protein upregulation in SF268 GBM and SK-N-SH neuroblastoma cell lines [160]. Another recent report demonstrated that MGMT protein levels are not altered following 72 hours of 10 µM TMZ exposure in T98G GBM cells [188]. In the present study, we observed a dose- dependent decrease in MGMT expression by T98G cells following treatment with 250 – 500 µM TMZ for 48 hours (Figure 3.5D). There are numerous possible explanations for this discrepancy, as MGMT is regulated by several factors, including mutant p53 [161], MEK-ERK [200], mTOR [188], hypoxia [51,201], and microRNAs [202,203]. Further, the TMZ dose-dependent MGMT decrease we observed is enhanced with Gli1 silencing (Figure 3D). Our data indicate that TMZ treatment and Gli1 silencing cooperatively deplete MGMT expression, which is consistent with previous reports [63,186]. Our studies, along with previous research, highlight the importance of investigating combination therapies in multiple GBM models to better understand differential response to treatment and predict efficacious therapy combinations. Having established that Hh inhibition induces senescence in a loss of PTEN- dependent manner as a standalone therapy, we were interested in the fate of GBM neurospheres co-treated with Hh inhibitors and TMZ. Our studies revealed that pharmacological Hh inhibition with GANT61 in combination with TMZ can induce apoptosis in U87-MG neurospheres, which consequently reduces both neurosphere size and the number of neurospheres formed (Figure 3.9,3.10). Interestingly, our studies reveal that while TMZ as a monotherapy reduces the number of large spheres (imaged area > 1 × 105 µm2) with no significant change in the total number of spheres, the addition of GANT61 does significantly reduce the sphere count. This is supported by previous work, which shows that pharmacological or genetic Hh pathway

82 suppression, but not TMZ alone, can prevent GSC tumorigenicity in mice [56], and Hh and Notch pathway inhibition enhances TMZ sensitivity in CD133+ stem-like cells [55]. Our results demonstrate that GANT61/TMZ co-treatment suppresses the growth of anchorage-independent U87-MG cells thought to be more representative of a GSC-rich subpopulation. Taken together with our previous results, this suggests that senescence mediated by siGli1 monotherapy is indeed tumor suppressive and potentiates the effects of TMZ chemotherapy.

3.4 Conclusions Our studies indicate that Hh pathway suppression may enhance the cytotoxic effects of TMZ against GBM, particularly those expressing low levels of MGMT and wild-type p53. We demonstrated that Hh inhibition reduces multidrug efflux activity, modulates p53 and MGMT expression, induces senescence dependent on the absence of PTEN, and reduces the growth of glioma cells exhibiting a stem-like phenotype. Overall, our data highlight the importance of investigating how Hh-targeted therapies interact with pathways that promote cellular resistance mechanisms to ultimately achieve maximal tumor reduction and prevent recurrence.

83 Chapter 4

COMPARISON OF POLYETHYLENIMINE-SPHERICAL NUCLEIC ACID NANOPARTICLES VERSUS POLYPLEXES FOR GENE REGULATION

4.1 Introduction Despite the exciting therapeutic potential of small interfering RNA (siRNA)- mediated gene silencing to treat diseases with a genetic basis, physiological barriers to siRNA delivery have hindered its clinical translation. siRNA is highly susceptible to nuclease degradation, is rapidly cleared from circulation, and cannot passively enter cells due to its large size and negative charge. Consequently, novel carriers are needed to enable the clinical translation of siRNA. To successfully deliver siRNA, intravenously injected carriers must evade immune clearance from the bloodstream, transport across the vascular endothelium to reach the target tissue, diffuse through the extracellular matrix, enter the diseased cells, escape from endocytic vesicles, and deliver siRNA to RNA-induced silencing complex (RISC) [105]. These unresolved challenges have prompted widespread investigation of nanoscale materials to effectively protect and deliver siRNA to diseased tissues. Polycationic nanoscale materials have demonstrated broad utility for nucleic acid delivery due to their abilities to condense and protect nucleic acids from nucleases and rapidly enter cells [91,204]. Such materials include synthetic polymers and polymers derived from amino acids, cell penetrating peptides inspired by viruses, and lipid-based carriers. As a potent polycation containing a high density of amine groups, polyethylenimine (PEI) is among the most widely investigated synthetic

84 polymers for nucleic acid delivery [125,126]. This is in large part because PEI can enable nucleic acid escape from intracellular endocytic vesicles due to its high buffering capacity [124,205], overcoming a significant cellular barrier to nucleic acid delivery. Further, PEI is attractive from an engineering standpoint for its wide availability in a range of molecular weights and degrees of branching, low cost, and ease of modification for multimodal delivery systems [126]. However, PEI and other polycationic materials are plagued by toxicity that precludes their clinical utility. This high toxicity is due to the presence of primary amines that impart a high positive charge. Indeed, PEI destabilizes cellular membranes, induces mitochondrial dysfunction, and may trigger complement activation in vivo [125]. Consequently, numerous efforts are ongoing to modify PEI to mitigate its cytotoxic effects while retaining its potent nucleic acid delivery efficacy [125,126,131]. Towards this goal, we hypothesized that changing the architecture and presentation of PEI to target cells may enhance siRNA delivery while minimizing cytotoxicity. Here, we demonstrate that hybrid polycationic-nanoparticle carriers can outperform standalone polycations, supporting the growing body of literature that suggests nanoscale architecture and surface chemistry play key roles in gene silencing. In this chapter, we exploit the unique surface structure of spherical nucleic acid nanoparticles (SNAs) as agents to modify PEI presentation to cells in a hybrid delivery system. SNAs have recently emerged as a new class of gene regulatory agents with attractive properties that may potentiate the gene silencing capacity of PEI-based siRNA carriers. SNAs consist of a spherical nanoparticle core coated with densely- packed nucleic acids that are oriented radially from the core surface [141,206]. This architecture imparts distinct properties that favor nucleic acid delivery in biological

85 systems. Most notably, SNAs undergo cellular entry without the need for ancillary transfection agents, exhibit enhanced stability against nucleases due to steric and electrostatic hindrances provided by their architecture, and demonstrate excellent biocompatibility in animal models [207–209]. Further, this architecture has been previously demonstrated to facilitate siRNA complexation with poly(β-amino ester)s, which fail to form complete nanoparticles with unbound siRNA likely due to the rigidity of siRNA molecules [210]. SNAs have been previously demonstrated to effectively deliver siRNA and microRNA to glioblastoma tumors following systemic injection [143,144], and their success in animal models has recently prompted the first clinical trial investigating these constructs as therapeutics for glioblastoma multiforme and gliosarcoma [150]. To test our overarching hypothesis that nanoscale architecture can improve polycation-mediated siRNA delivery, we investigate the cellular uptake, intracellular trafficking, cytocompatibility, and gene regulation efficacy of PEI-coated SNAs (PEI- SNAs) versus PEI-siRNA polyplexes. We find that PEI-SNAs undergo enhanced cellular uptake relative to both SNAs and polyplexes and exhibit reduced accumulation within lysosomes relative to polyplexes. The improved uptake and trafficking profile achieved with PEI-SNAs translates to dramatically enhanced gene silencing capabilities using lower siRNA doses. Importantly, potent gene silencing with PEI-SNAs is achieved within a cytocompatible dosing range. Taken together, our data demonstrate the potential for the unique architecture of SNAs to improve the gene regulation efficiency of polycationic materials, and vice versa, for cationic materials to improve the gene regulation efficiency of SNAs. These observations confirm that nanoscale architecture and surface chemistry play a critical role in gene regulation and

86 warrant further investigation of the intracellular trafficking mechanisms responsible for determining the fate of hybrid cationic-nanoparticle siRNA carriers. This chapter contains sections adapted from Melamed et al. (2018) Molecular Therapy: Nucleic Acids [151].

4.2 Materials and Methods

4.2.1 SNA and PEI-SNA Synthesis and Characterization SNAs and PEI-SNAs were prepared and characterized as described in Chapter 2. siRNA sequences used are as follows: GFP – GGC UAC GUC CAG GAG CGC ACC dTdT, Scr – ACG CGA CCG UGC CGA UCG GCA dTdT, Negative Control siRNA #1 (ThermoFisher Scientific, Waltham, MA).

4.2.2 PEI-siRNA Polyplex Synthesis PEI-siRNA polyplexes were prepared using standard methods [211]. siRNA and PEI were diluted separately in 10 mM 4-(2-hydroxyethyl)-1- piperazineethanesulfonic acid) (HEPES) buffer at pH 7.2, then combined at an N/P ratio of 6/1, vortexed for 5 seconds, and incubated at room temperature for 20 minutes. Polyplexes were further diluted in HEPES buffer for characterization or cell culture medium for experiments without further purification.

4.2.3 Particle Characterization and Evaluation of Serum Stability Samples were prepared for TEM imaging by incubating a drop of sample solution on a poly-L-lysine-coated 400 mesh formvar copper grid (Electron Microscopy Sciences, Hatfield, PA). Grids were subsequently dried and counterstained with 2% uranyl acetate prior to imaging on a Zeiss LIBRA 120 TEM.

87 AuNPs, SNAs, and PEI-SNAs were characterized by UV-visible spectroscopy using a Cary 60 spectrophotometer (Agilent Technologies, Santa Clara, CA). Dynamic light scattering (DLS) and zeta potential measurements were made using a Litesizer500 (Anton Paar, Graz, Austria). SNAs and PEI-SNAs were diluted in DI water to 0.67 nM for DLS or to 1.6 nM for zeta potential measurements. Polyplexes were diluted in HEPES buffer to a final RNA concentration of 200 nM for DLS or to 333 nM for zeta potential measurements. Zeta potential measurements were replicated at least 8 times per sample or until quality criteria was met. Serum stability was evaluated by incubating particles in 0, 1, 5, or 10% FBS diluted in PBS for two hours at 37ºC with gentle shaking. Samples were further diluted in PBS to the concentrations described above for DLS and zeta potential analysis.

4.2.4 Cell Culture and Stable Gene Expression U87-MG and U373 glioma cells (ATCC, Manassas, VA) and 293TN cells (System Biosciences, Palo Alto, CA) were cultured in DMEM (VWR, Radnor, PA) supplemented with 10% FBS (Gemini Bio-Products, West Sacramento, CA) and 1% penicillin/streptomycin (ThermoFisher Scientific, Waltham, MA). Cells were maintained in a humidified environment at 37ºC, 5% CO2. U373 cells were stably transfected with eGFP as described previously [212–214]. Briefly, cells were transfected with a murine stem cell virus (MSCV) retroviral vector encoding an eGFP- Firefly luciferase fusion protein and sorted for eGFP positivity. U87-MG cells were stably transfected with Rab5-GFP or LAMP1-GFP using standard procedures. Briefly, Rab5-GFP (Addgene # 56530) or LAMP1-GFP (Addgene # 34831) was cloned into a lentiviral transfer vector (System Biosciences, Palo Alto, CA) by restriction enzyme digest. Lentiviral particles were produced by triple-transfecting (TransIT-Lenti

88 transfection reagent; Mirus Bio LLC, Madison, WI) 293TN cells with either transfer vector and lentiviral packaging and envelope plasmids (Addgene #12260,12259). Lentivirus was harvested, filtered, and diluted in cell culture medium to transduce U87-MG cells. Cells stably expressing the desired fusion protein were selected with 1 µg/mL puromycin (VWR, Radnor, PA).

4.2.5 Cellular Uptake Analysis U87-MG cells were seeded in 24-well culture plates at a density of 50,000 cells/well and cultured overnight. Cells were incubated with SNAs, PEI-SNAs, or PEI-siRNA polyplexes labeled with Cy5-siRNA (Integrated DNA Technologies, Coralville, IA) diluted in complete culture medium at concentrations normalized to siRNA payload (20 nM siRNA) for various times ranging from 15 minutes to 24 hours at 37ºC, 5% CO2. Cells were washed with PBS, trypsinized, and resuspended in PBS for Cy5 intensity analysis by flow cytometry using a NovoCyte flow cytometer (ACEA Biosciences, Inc., San Diego, CA).

4.2.6 Lysosomal Trafficking and Imaging U87 cells expressing Rab5-GFP or LAMP1-GFP were seeded in 35 mm glass bottom dishes at a density of 150,000 cells/dish and cultured overnight in a humidified incubator. Cells were incubated with PEI-SNAs (10 nM siRNA) or polyplexes (200 nM siRNA) prepared with Cy5-siRNA for 0-24 hours at 37ºC, 5% CO2, washed three times with PBS to remove uninternalized constructs, counterstained with CellMask Orange (ThermoFisher Scientific, Waltham, MA) and fed with phenol red-free complete medium. Live cell imaging was performed using a Zeiss LSM880 confocal microscope equipped with an incubated stage to maintain cells during imaging. Z-

89 stacks were acquired to analyze PEI-SNA/polyplex colocalization throughout the entire volume of the cell. Quantitative colocalization analysis was used to determine the fraction of siRNA conjugates present within lysosomes and the fraction of lysosomes containing siRNA conjugates. Image analysis was performed using three independent data sets, each consisting of 5-7 images (corresponding to 10-20 cells) per treatment group. Regions of interest (ROIs) were defined by manually tracing the outlines of individual cells (and excluding nuclei) using ImageJ [215] (NIH, Bethesda, MD) and imported into MATLAB (MathWorks, Natick, MA) for quantitative colocalization analysis using a custom script. Both the Cy5-PEI-SNA/polyplex and LAMP1-GFP channels were median filtered with a 3-by-3-by-3 neighborhood and top-hat filtered by reconstruction using a 2 µm disk element. Manders’ colocalization coefficients [216] were calculated for each ROI within each image stack, and all ROIs for each treatment group were averaged within an experimental replicate. Statistics were performed across averages from three independent replicates.

4.2.7 siRNA and PEI Co-Trafficking Analysis Intracellular construct stability was evaluating by observing the relative intracellular fates of siRNA and PEI by confocal microscopy. PEI was labeled with tetramethylrhodamine isothiocyanate (TRITC) using previously reported methods [217]. Briefly, PEI (1 mL at 10 mg/mL in DI water) was reacted with TRITC (11 uL at 10 mg/ml in DMF) overnight to label 1% of the primary amines. The product was lyophilized and stored dried at -20ºC until resuspension in DI water for experiments. PEI-SNAs and polyplexes were synthesized using TRITC-PEI and Cy5-siRNA as previously described. For siRNA and PEI co-trafficking analysis, U87-MG cells were seeded in 35 mm glass bottom dishes at a density of 150,000 cells/dish and cultured

90 overnight. Cells were incubated with PEI-SNAs (10 nM siRNA) or polyplexes (200 nM siRNA) prepared with Cy5-siRNA for 24 hours at 37ºC, 5% CO2, washed three times with PBS to remove uninternalized constructs, and fed with fresh medium. Cells were either imaged immediately or incubated and additional 24 hours prior to imaging. Confocal microscopy was performed and quantitative colocalization analysis was conducted as described above.

4.2.8 Toxicity Assessment U87 cells were seeded in 96-well culture plates at a density of 2500 cells/well and cultured overnight. Cells were incubated with PEI-SNAs or PEI-siRNA polyplexes in complete culture medium for 24 hours at 37ºC, 5% CO2. Construct- containing medium was removed and cells replenished with fresh medium, then incubated a further 24 hours at 37ºC, 5% CO2. Cellular metabolic activity was assessed by MTT assay (ThermoFisher Scientific, Waltham, MA) according to the manufacturer’s protocol. These results were corroborated by live cell PI staining to detect cells that have undergone membrane permeabilization. U373.eGFP cells were exposed to PEI-SNAs or polyplexes under the previously stated conditions, then trypsinized, stained with PI, and analyzed by flow cytometry using a NovoCyte flow cytometer. IC50 values for PEI-SNAs and polyplexes were calculated using MATLAB software.

4.2.9 Gene Knockdown Assessment U373.eGFP cells were seeded in 12-well culture plates at a density of 25,000 cells/well and cultured overnight. Cells were incubated with PEI-SNAs or PEI-siRNA polyplexes diluted in complete culture medium for 24 hours at 37ºC, 5% CO2. PEI-

91 SNA- or polyplex-containing medium was removed and cells were replenished with fresh complete culture medium and incubated a further 24 hours at 37ºC, 5% CO2. Cells were trypsinized and resuspended in PBS for GFP intensity analysis by flow cytometry using a NovoCyte flow cytometer.

4.2.10 Statistical Analysis All experiments were repeated in triplicate and data are reported as means ± standard deviations across three independent replicates unless otherwise stated. Groups with significant differences were determined using Student’s t-test (when two groups were compared) and one-way analysis of variance (ANOVA) with post-hoc Tukey test (when three or more groups were compared) completed in MATLAB software (MathWorks, Natick, MA). Differences were considered significant at p < 0.05. Flow cytometry data was analyzed using NovoExpress software (ACEA Biosciences, Inc., San Diego, CA). DLS and zeta potential data were analyzed using Kalliope software (Anton Paar, Graz, Austria).

4.3 Results and Discussion

4.3.1 Nanoparticle and Polyplex Synthesis and Characterization Citrate-capped 15 nm gold nanoparticles (AuNPs) were prepared in-house using the Frens method [154] and subsequently functionalized with siRNA and polyethylene glycol (PEG) via thiolated ligands according to established procedures to produce SNAs [143,153]. To prepare PEI-SNAs, purified SNAs were coated with PEI via electrostatic adsorption, and purified again to remove unbound PEI (Figure 4.1A).

An OliGreen assay [153] revealed that SNAs contained ~38 siRNA duplexes per particle, and a TNBS assay [156] determined that ~923 PEI molecules are associated with each PEI-SNA (Figure 4.1B). At each step, the resulting AuNP conjugates were characterized using UV-visible spectroscopy, dynamic light scattering (DLS), zeta potential, and transmission electron microscopy (TEM). PEI-siRNA polyplexes were produced using standard methods with an N/P ratio of 6 (Figure 4.2) [211]. This N/P ratio was chosen to completely complex siRNA with PEI, as demonstrated by a gel retardation assay (Figure 4.2B) and to optimally balance transfection efficiency promoted by higher N/P polyplexes [218] with cytotoxicity (Figure 4.2C). Polyplexes were further characterized by DLS, TEM, and zeta potential measurements.

Functionalizing AuNPs with siRNA and PEG increased the hydrodynamic diameter from 22 nm to 46 nm, and coating with PEI further increased the

94 hydrodynamic diameter to ~175 nm, similar in size to the ~140 nm polyplexes (Figure 4.3A). Because UV-visible spectroscopy (Figure 2.4) did not reveal any signs of aggregation (i.e., broadening or red-shifting of the characteristic ~520 nm AuNP absorbance maximum) and monodisperse PEI-SNAs were visible by TEM (Figure 4.3B), we attributed this large increase to the size and number of PEI molecules associated with each PEI-SNA rather than to particle aggregation. Zeta potential measurements further confirmed successful functionalization, with AuNPs and SNAs exhibiting net negative surface charges (-30 mV and -14 mV, respectively), while PEI- SNAs and polyplexes exhibited net positive surface charges (both ~27 mV) (Figure 4.3A).

Figure 4.3. Nanoparticle and polyplex characterization. A) Zeta potential and DLS measurements for AuNPs, SNAs, PEI-SNAs, and PEI-siRNA polyplexes. Data are means +/- standard deviations, # = no significance. B) Representative TEM images of AuNPs, SNAs, PEI-SNAs, and polyplexes. Grids were counterstained with uranyl acetate to visualize siRNA as light contrast. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

95 4.3.2 Evaluation of PEI-SNA and Polyplex Serum Stability To assess the serum stability of PEI-SNAs and polyplexes, constructs were incubated in 1%, 5%, or 10% FBS for two hours and subsequently evaluated by zeta potential measurements and DLS to determine whether or not exposure to serum affects their surface chemistry or colloidal stability. PEI-SNAs suspended in 1% FBS aggregated quickly and visibly crashed out of solution, while PEI-SNAs remained stabled in 5-10% FBS. This indicates that serum proteins are crucial for maintaining the colloidal stability of PEI-SNAs. Following two hours incubation with FBS, both PEI-SNAs and polyplexes undergo charge reversal and maintain zeta potential values of ~-8 mV for all tested FBS concentrations (Figure 4.4A), indicating that negatively charged serum proteins associate with both particle types. For comparison, SNAs do not undergo significant changes in zeta potential under the same conditions. We further investigated the effects of serum on hydrodynamic diameter (Figure 4.4B). Surprisingly, both PEI-SNAs and polyplexes exhibited significant decreases in hydrodynamic diameter after incubation with 5-10% FBS. Taken together, these results lead us to speculate that serum proteins displace PEI that is loosely bound to the periphery of both particle types and that this interaction induces a stabilizing effect.

4.3.3 PEI-SNAs Undergo Enhanced Cellular Uptake Relative to PEI-siRNA Polyplexes or SNAs Flow cytometry was used to quantitatively examine the extent to and rate at which U87-MG glioma cells bind and internalize SNAs, PEI-SNAs, or polyplexes prepared with Cy5-labeled siRNA. SNAs exhibited the lowest degree of binding and uptake after 24 hours and achieved maximum cellular siRNA fluorescence intensity after 4 hours of continuous incubation with U87-MG cells (Figure 4.5A,B). In contrast, polyplexes and PEI-SNAs undergo increasing binding and uptake through 24 hours of continuous incubation with U87-MG cells and achieve 5- or 81-fold increases in cellular siRNA fluorescence intensity relative to SNAs, respectively (Figure 4.5A,B), after 24 hours. Further, binding and uptake of PEI-SNAs occurs far more rapidly than that of polyplexes, with cellular accumulation of Cy5-siRNA appearing within 15 minutes when delivered via PEI-SNAs versus 4 hours via polyplexes. PEI- SNAs achieve a 28-fold higher Cy5-siRNA intensity relative to polyplexes over 4 hours and a 16.5-fold increase over 24 hours (Figure 4.5A,B). The increased binding

97 and uptake of polyplexes and PEI-SNAs relative to SNAs is likely due to their equivalent net positive surface charge, but we attribute the dramatically enhanced uptake of PEI-SNAs relative to polyplexes to differences in surface structure resulting from the spherical architecture of SNAs.

98 4.3.4 PEI-SNAs and Polyplexes Remain Intact Following Endocytosis To determine whether siRNA dissociates from either PEI-SNAs or polyplexes following endocytosis, confocal microscopy was used to observe the relative intracellular trafficking of Cy5-siRNA and TRITC-PEI from each carrier. Cells were incubated with PEI-SNAs or polyplexes for 24 hours, then imaged or incubated a further 24 hours prior to imaging. Quantitative colocalization analysis was employed to calculate the fractional overlap of Cy5-siRNA and TRITC-PEI using Mander’s Colocalization Coefficients (MCC). Our studies demonstrate that siRNA and PEI from both carriers remain heavily colocalized (all MCC ~0.95) up to 24 hours after endocytosis (Figure 4.6). This data demonstrates that either complete siRNA dissociation from the carrier is not necessary to enable gene silencing or that the fraction of dissociated siRNA is too small to detect using these methods. However, we are encouraged that PEI-SNAs exhibit stability in physiological conditions at least equal to that of polyplexes, which have been widely investigated through in vivo gene regulation studies [125]. Thus, this data supports the use of polycation-wrapped SNAs as potentially translatable siRNA carriers.

Figure 4.6. Relative trafficking of siRNA and PEI from PEI-SNAs vs polyplexes. A) Representative confocal microscopy images showing that Cy5-siRNA remains mostly colocalized with TRITC-PEI from both polyplexes and PEI-SNAs following 24 hours incubation with cells (scale bar = 20 µm). Manders’ colocalization coefficient for fractional overlap of Cy5-siRNA and TRITC-PEI is shown in yellow on the merged brightfield image (MCC 1 = Cy5-siRNA colocalized with TRITC-PEI, MCC 2 = TRITC- PEI colocalized with Cy5-siRNA). Quantitative assessment of colocalization depicting average fractional overlap of B) Cy5-siRNA with TRITC-PEI (MCC 1) and C) TRITC-PEI with Cy5-siRNA (MCC 2) in cells exposed to PEI-SNAs or polyplexes for 24 or 48 hours across three independent replicates +/- standard deviations, no significant differences by Student’s t-test. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

100

4.3.5 PEI-SNAs Exhibit Decreased Lysosomal Accumulation Relative to PEI- siRNA Polyplexes An unresolved barrier to nanoparticle-mediated siRNA delivery is maximizing the release of internalized cargo from endocytic vesicles into the cytosol to interact with RNAi machinery and achieve gene silencing. This is evident in recent research demonstrating that only ~1-2% of internalized siRNA reaches the cytosol to regulate gene expression [219,220]. In the classical endocytic pathway, siRNA nanocarriers interact with cellular surface receptors to initiate endocytosis, are taken up into early endosomes, then are trafficked through late endosomes until finally reaching lysosomal compartments, where they are enzymatically degraded. First, we demonstrated that both PEI-SNAs and polyplexes undergo endocytosis by visualizing their uptake into Rab5+ early endosomes. Rab5+ early endosomes were labeled by stably expressing a Rab5-GFP fusion protein in U87-MG cells, and PEI-SNAs or polyplexes were labeled using Cy5-siRNA. Cells were incubated with PEI-SNAs or polyplexes, then imaged immediately. Using confocal microscopy, we found that PEI- SNAs could be first detected within Rab5+ early endosomes within 60-90 minutes (Figure 4.7A). Consistent with our previous flow cytometry data, polyplex internalization appeared slightly slower, with early endosomal accumulation detectable within 120 minutes (Figure 4.7A). Because successful siRNA nanocarriers should reduce lysosomal accumulation to prevent the degradation of internalized siRNA, we investigated the degree of nanocarrier colocalization with lysosomes using cells stably expressing LAMP1-GFP. Cells were incubated with PEI-SNAs or polyplexes for 24 hours, washed to remove unbound particles, then imaged live using

101 confocal microscopy (Figure 4.7B). Quantitative colocalization analysis was conducted to calculate the fractional overlap of each probe. Excitingly, PEI-SNAs demonstrated a significant decrease in colocalization with lysosomes relative to polyplexes (Figure 4.7C, MCC=0.83 for polyplexes and MCC=0.75 for PEI-SNAs). For comparison, SNAs lacking a PEI coating exhibited high accumulation within lysosomes, similar to polyplexes (MCC = 0.85, Figure 4.8). Because endosomal escape is well understood to be a highly inefficient process [220], we believe the observed 8% decrease in lysosomal accumulation for PEI-SNAs relative to polyplexes is biologically impactful. There was no significant difference in the fractional overlap of LAMP1-GFP with Cy5-siRNA for PEI-SNAs and polyplexes, indicating that the fraction of total lysosomes containing siRNA is the same in each case (Figure 4.7D).

102

Figure 4.7. Intracellular trafficking of siRNA depends on nanocarrier architecture. A) Representative confocal microscopy images showing Cy5-siRNA colocalization with early endosomes tagged via a Rab5-GFP fusion protein (scale bar = 20 µm). Yellow boxes indicate magnified regions of siRNA colocalized with Rab5+ early endosomes. B) Representative confocal microscopy images showing Cy5-siRNA colocalization with lysosomes tagged via a LAMP1-mGFP fusion protein. Areas of colocalization appear yellow in the merged brightfield image (scale bar = 20 µm). Manders’ colocalization coefficient for fractional overlap of Cy5-siRNA and LAMP1-GFP is shown in yellow on the merged brightfield image (MCC 1 = Cy5-siRNA colocalized with LAMP1-GFP, MCC 2 = LAMP1-GFP colocalized with Cy5-siRNA). Quantitative assessment of colocalization depicting average fractional overlap of C) Cy5-siRNA with LAMP1-GFP (MCC 1) and D) LAMP1-GFP with Cy5- siRNA (MCC 2) across three independent replicates +/- standard deviations, *p=0.05 by Student’s t-test. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

103

Figure 4.8. Confocal microscopy showing SNAs colocalize heavily with lysosomes stained with LysoTracker Red. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

4.3.6 PEI-SNAs Mediate GFP Silencing at Dramatically Lower siRNA Doses than PEI-siRNA Polyplexes To compare the gene silencing potency of polyplexes and PEI-SNAs, we targeted GFP as a model gene in U373 glioblastoma cells stably expressing eGFP. Cells were treated with polyplexes or PEI-SNAs at various siRNA payloads for 24 hours in complete culture medium, replenished with fresh medium and incubated a further 24 hours, then harvested and analyzed for GFP expression by flow cytometry. PEI-SNAs induced a significant decrease in GFP fluorescence at a low dose of 32 nM siRNA and yielded an impressive 70% reduction in GFP relative to untreated cells when dosed at 60 nM siRNA (Figure 4.9A,C). Interestingly, polyplex-mediated RNAi required nearly a 10-fold increase in total siRNA payload to reduce GFP expression. At siRNA doses up to 200 nM, no silencing was observed. At siRNA doses of 500 and 1000 nM, significant GFP silencing was observed, and GFP intensity was reduced by 81% relative to untreated cells (Figure 4.9B,D). From this data, we calculated the siRNA dosage required to achieve 50% gene silencing from either carrier (ED50) and

104 found that 38 nM siRNA is required to achieve 50% GFP silencing via PEI-SNAs, while 403 nM siRNA is required via polyplexes (Figure 4.9B,D). For comparison, 400 nM siRNA delivered via SNAs lacking a PEI coating was insufficient to silence GFP expression under the conditions tested (Figure 4.10). Our studies demonstrate that PEI-SNAs enhance polycation-mediated GFP silencing by 10-fold. We attribute the dramatically enhanced silencing potency of PEI-SNAs relative to polyplexes to the unique three-dimensional presentation of siRNAs and polycations in this system.

105

Figure 4.9. GFP silencing efficacy assessed by flow cytometry using PEI-SNAs or polyplexes. Representative flow cytometric histograms from one representative experiment depict dose-dependent GFP silencing using A) PEI-SNAs or B) polyplexes. Summary flow cytometry GFP silencing results using C) PEI-SNAs or D) polyplexes averaged across three independent replicates +/- standard deviations, *p<0.01 by one-way ANOVA with posthoc Tukey. Median fluorescence values for experimental samples are normalized to that of an untreated control. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

106

Figure 4.10: GFP silencing efficacy of SNAs delivering 400 nM siRNA. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

4.3.7 PEI-SNAs Improve Polycation Cytocompatibility Relative to Polyplexes The cytocompatibility profiles of PEI-SNAs and polyplexes were evaluated by 3-(4,5)-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay and live cell propidium iodide (PI) staining with flow cytometry analysis. PEI-SNAs are cytocompatible as indicated by insignificant changes in metabolic activity at siRNA doses up to 60 nM (corresponding to 34 µg/mL PEI) (Figure 4.11A). Likewise, polyplexes are cytocompatible at siRNA doses up to 750 nM (corresponding to 8 µg/mL PEI) (Figure 4.11B). MTT assay results were further used to calculate half- maximal inhibitory concentration (IC50) values for both PEI-SNAs and polyplexes.

PEI-SNAs exhibit an IC50 of 80 nM siRNA, corresponding to 45 µg/mL PEI (Figure

4.11A). Polyplexes exhibit an IC50 of 920 nM siRNA, corresponding to 9.3 µg/mL PEI (Figure 4.11B). The observed decreases in relative metabolic activity correlate with an increase in the fraction of PI+ cells, supporting the results obtained by MTT assay (Figure 4.11C,D). From these studies, we conclude that both PEI-SNAs and polyplexes are cytocompatible within a dosing range sufficient for gene silencing. Further, despite the known cytotoxicity of PEI at concentrations used for nucleic acid

107 delivery, we observed that higher PEI concentrations delivered via PEI-SNAs were more cytocompatible than lower PEI concentrations delivered via polyplexes. This may be due to enhanced electrostatic interaction with the radially-oriented siRNA present on PEI-SNAs, presence of the negatively charged gold nanoparticle core, or altered presentation of the PEI molecules to cells. This intriguing result warrants future investigation of the relationship between polycation presentation from nanomaterials and cytotoxicity.

108

Figure 4.11. Cytocompatibility profiles for PEI-SNAs and polyplexes. The cytocompatible dosing range of both A) PEI-SNAs and B) polyplexes was determined using an MTT assay. Data are means +/- standard deviations, *p<0.01 by one-way ANOVA with posthoc Tukey. Results were confirmed using PI staining and flow cytometric analysis in live cells exposed to C) PEI-SNAs or B) polyplexes. Reproduced from Melamed et al. (2018) Mol Ther Nucleic Acids [151].

109 4.3.8 Discussion Our findings demonstrate that hybrid PEI-SNA constructs with defined siRNA architecture mediate RNAi more efficiently than randomly assembled PEI-siRNA polyplexes. We observed that PEI-SNAs enable greater cellular uptake of siRNA, undergo decreased lysosomal accumulation, and efficiently silence GFP expression at lower siRNA doses relative to polyplexes. Additionally, we found that both PEI-SNAs and polyplexes were cytocompatible within a dosing range sufficient to promote gene silencing. From these results, we conclude that the architecture of siRNA nanocarriers plays an important role in their cellular interactions and ultimate gene regulation efficacy. To understand how exposure to biological milieu will affect PEI-SNAs and polyplexes, we performed serum stability studies in which constructs were incubated in 1%, 5%, or 10% FBS for two hours and subsequently evaluated by zeta potential measurements and DLS. We found that serum proteins are crucial for maintaining the colloidal stability of PEI-SNAs and possibly polyplexes to a lesser extent. Following two hours incubation with FBS, both PEI-SNAs and polyplexes undergo charge reversal for all tested FBS concentrations (Figure 4.4A), indicating that negatively charged serum proteins associate with both particle types. Further, both PEI-SNAs and polyplexes exhibited significant decreases in hydrodynamic diameter after incubation with 5-10% FBS (Figure 4.4B). Based on the observed charge reversal and partial hydrodynamic diameter decrease, we speculate that serum proteins displace PEI that is loosely bound to the periphery of both particle types and that this interaction induces a stabilizing effect. Because we used FBS to model exposure to physiological environments, we would expect that the observed protein corona is largely composed of bovine serum albumin (BSA). While BSA is negatively charged and likely

110 decreases the cellular adhesion of our cationic nanoparticles, it may improve their cytocompatibility by screening strong positive charges [113]. Further, incorporation of albumin around PEI-DNA and PEI-siRNA polyplexes was previously demonstrated to increase cellular uptake of polyplexes and improve transfection efficiency [221,222]. However, a limitation of our study is that FBS lacks complement proteins found in plasma, so future work should seek to understand how these proteins will interact with PEI-SNAs and polyplexes as well. We found that cells bind and take up PEI-SNAs most rapidly and to the greatest extent of the siRNA carriers we investigated (Figure 4.5). We attribute this to several physicochemical characteristics of PEI-SNAs. Notably, cationic nanomaterials undergo greater cellular uptake than anionic nanomaterials because positively charged constructs electrostatically associate with negatively charged cell membranes to promote binding to the cell surface [106,111,223]. It is therefore unsurprising that cationic PEI-SNAs and polyplexes undergo enhanced cellular uptake relative to anionic SNAs (Figure 4.5). However, initial SNA binding and uptake was more rapid than that of polyplexes (Figure 4.5), demonstrating that surface charge is not the only property responsible for dictating cellular uptake. We attribute the more rapid binding and uptake of SNAs relative to polyplexes and the maximal uptake of PEI-SNAs over all carriers to the surface structure of SNAs and PEI-SNAs. SNAs contain a high density of radially-oriented siRNA at their surface, and this architecture allows multiple siRNA strands from a single particle to simultaneously interact with cell surface receptors. As a result, SNAs engage in multivalent binding [224], a property known to enhance the cellular uptake of nanomaterials [225–227]. Further, SNA architecture enhances nucleic acid secondary structure, such as G-quadruplex

111 formation in G-rich SNAs [228]. This further engages scavenger receptors to increase cellular uptake and may explain why SNAs undergo more rapid binding and uptake than polyplexes at early timepoints despite their negative charge. This may also explain the enhanced binding and uptake of PEI-SNAs over polyplexes (Figure 4.5). Wang et al. made the similar observation that ligand organization determines the uptake of peptide-targeted nanoparticles. They found that nanoparticle-bound peptide- lipids are more monomeric and deaggregated than peptide-lipids in aqueous solution, which correlated with improved cellular uptake and emphasizes the importance of ligand orientation and presentation to cells [227]. Our studies suggest that, in the siRNA carriers we investigated, both surface charge and structure play a greater role in regulating cellular interactions than nanoparticle size, another physicochemical property known to influence biological interactions of nanomaterials [106]. This is evident in that SNAs, which are optimally sized to maximize cellular uptake (50 nm; Figure 1) [109], undergo decreased binding and uptake relative to larger PEI-SNAs and polyplexes (~150 nm; Figure 4.3). Finally, the differential uptake between PEI-SNAs and polyplexes could be explained by differences in the composition of each construct. The AuNP core, which serves as a template around which siRNA is spherically arranged, is present in PEI- SNAs but not in polyplexes. However, previous studies investigating the role of core materials in SNA cellular interactions suggest that the presence of the AuNP core has negligible impact on cellular binding and uptake. For example, both hollow SNAs lacking a core material and SNAs containing a quantum dot core undergo cellular uptake and intracellular trafficking similar to that of AuNP-based SNAs [139,147]. Another important composition difference between PEI-SNAs and polyplexes is the

112 ratio of polycation to siRNA. The goal of this study was to compare optimal conditions for siRNA delivery between PEI-SNAs and polyplexes, so we synthesized polyplexes using N/P = 6/1 to achieve this goal. We chose this ratio to maximize polycation content for efficient transfection while minimizing toxicity (Figure 4.2) and to generate carriers of similar size and surface charge to PEI-SNAs (Figure 4.3A). This is in good agreement with existing literature on optimal siRNA polyplex N/P ratios [211,229]. For polyplexes generated with 25 kDa bPEI and siRNA, N/P = 6/1 corresponds to 0.0101 µg/mL PEI per nM siRNA. PEI-SNAs were synthesized by incubating SNAs in excess PEI to prevent aggregation due to interparticle bridging [211], and then unbound PEI was removed by centrifugation to yield purified constructs. Using a TNBS assay for amine quantification, we determined that purified PEI-SNAs contain 0.55 µg/mL PEI per nM siRNA. However, despite containing higher PEI content per siRNA, PEI-SNAs appear less cytotoxic than polyplexes per

PEI payload. Cytocompatibility analysis found that PEI-SNAs yield an IC50 value of

80 nM siRNA, corresponding to 45 µg/mL PEI, while polyplexes yield an IC50 value of 920 nM siRNA, corresponding to 9.3 µg/mL PEI (Figure 4.11A,B). This result was surprising, given the known cytotoxicity of PEI. In agreement with our findings, alternating electrostatic assembly of siRNA and PEI or other materials designed to promote siRNA release around AuNPs has previously demonstrated reduced toxicity relative to PEI-siRNA complexes alone [230,231]. Further, previous research has demonstrated that complexing PEI with additional anionic polymers and nucleic acids also reduces PEI toxicity relative to PEI-nucleic acid complexes alone [232,233]. Based on our results and the results of others, we conclude that the presentation of PEI to cells is important in determining its cytotoxicity and that the architecture afforded

113 by PEI-SNAs offers greater cytocompatibility than PEI assembled randomly in polyplexes. To understand how nanocarrier architecture impacts intracellular stability, we evaluated the relative intracellular fates of siRNA and PEI from either carrier by confocal microscopy using dual-labeled PEI-SNAs and polyplexes synthesized with Cy5-siRNA and TRITC-PEI. Colocalization analysis determined that siRNA and PEI remain heavily colocalized from both carriers through 24 hours incubation with cells, as evidenced by all MCCs remaining ~0.94-0.96 (Figure 4.6A,B). This suggests that the electrostatic interactions between siRNA and PEI are strong enough to prevent significant dissociation of either carrier over the timepoints we investigated. While the spatial resolution of confocal microscopy may limit our ability to detect dissociation of small amounts of siRNA or PEI, our findings agree with a growing body of work that suggests strong polycation-siRNA interactions may prevent substantial intracellular dissociation. One such investigation comparing the self-assembly mechanisms of PEI-siRNA and PEI-DNA found that PEI-siRNA complexation was more thermodynamically favorable and produced more stable complexes than PEI- DNA [229], which may explain why we do not observe significant dissociation of PEI-SNAs or polyplexes. Moreover, the high stability of PEI-SNAs and polyplexes may be especially advantageous given that our data shows that both complexes reside largely within lysosomes (Figure 4.7). An elegant Förster Resonance Energy Transfer-based study demonstrated that PEG/PEI-DNA polyplex dissociation within the endo-lysosomal system negatively correlated with transfection efficiency [218]. This finding is corroborated by another study that found that siRNA polyplexes made with polymers

114 containing double cationic charges per monomer bind more strongly to siRNA and improve transfection efficiency [234]. Therefore, complexes that are robust against dissociation, particularly within endo-lysosomal compartments, may prove more efficient as nucleic acid carriers. Similarly, an unresolved challenge to siRNA delivery is reducing the net accumulation of siRNA within lysosomes, which ultimately leads to degradation of internalized siRNA. Most siRNA carriers are internalized by receptor-mediated endocytosis and are subsequently trafficked through early and late endosomes, which acidify and ultimately fuse with lysosomes. Once trapped within lysosomes, internalized cargo is degraded and rendered inactive. We found that PEI-SNAs associate with Rab5+ early endosomes within 60-90 minutes, while polyplexes are first detectable in early endosomes within 120 minutes (Figure 4.7A). We further demonstrate that PEI-SNAs undergo significantly decreased lysosomal accumulation relative to polyplexes (Figure 4.7B-D), suggesting that PEI-SNAs may act more efficiently as siRNA carriers. Our results are consistent with previous research investigating the use of AuNP/polycation hybrid materials for siRNA delivery. For example, one such study demonstrated that AuNPs functionalized with alternating layers of poly(allylamine hydrochloride) and siRNA or PEI and siRNA exhibited decreased colocalization with lysosomes relative to Lipofectamine 2000 or AuNPs functionalized with alternating layers of poly(diallyl dimethyl ammonium chloride) and siRNA [235]. Additionally, incorporating polymers designed to degrade at endosomal pH into AuNP-based siRNA delivery systems can further improve endosomal escape [231,236]. Because endosomal escape is well understood to be a highly inefficient process, we believe that our observed 8% decrease in PEI-SNA

115 colocalization with lysosomes relative to polyplexes may be biologically impactful. Notably, a study investigating the intracellular trafficking of lipid nanoparticles, among the most successful and widely investigated siRNA carriers [220], determined that only 1-2% of internalized siRNA escapes endosomes to reach the cytosol [219]. Another study further elucidated the inefficient nature of endosomal escape through a series of experiments directly visualizing rupturing endosomes. This study determined that lipoplex-mediated siRNA delivery induces between 1 and 5 endosomal escape events per cell over a period of several hours and that endosomes that do rupture only release about half of their internalized siRNA [237]. Therefore, given the documented inefficiency of endosomal escape, we hypothesize that the observed 8% decrease in lysosomal accumulation afforded by PEI-SNAs is biologically significant and likely contributes to their enhanced gene silencing potency. Finally, gene silencing studies demonstrated that PEI-SNAs enhance siRNA- mediated GFP silencing by 10-fold. Our data shows that 38 nM siRNA is required to achieve 50% GFP silencing via PEI-SNAs, while 403 nM siRNA is required via polyplexes (Figure 4.9B,D). The high siRNA dose required by polyplex-mediated gene silencing is consistent with what others have observed for PEI-based polyplexes; for example, one study reported that at least 200 nM siRNA was required to silence luciferase gene expression using 25 kDa bPEI polyplexes [211]. Interestingly, another study demonstrated that 25 kDa bPEI was insufficient to silence luciferase expression, while robust gene silencing was observed using PEI modified by succinylation at 10% of amines [131]. This improved silencing efficacy was attributed to the decrease in toxicity afforded by succinylation, enabling the use of more polycation. Our results and the results of others further support this conclusion, suggesting that improving the

116 cytocompatibility of PEI by chemical or structural modification is a promising strategy to enhance siRNA delivery [125]. Our studies reveal that improving the cytocompatibility of PEI by altering its presentation to cells can enhance the net cellular uptake of siRNA and reduce lysosomal accumulation to dramatically enhance gene silencing potency. While our investigation sought to evaluate the role of siRNA architecture in polycation-mediated gene silencing at the cellular level, future research should employ animal models to determine the role of nanocarrier architecture on biodistribution and gene regulation efficacy at a disease site. One recent investigation of ligand presentation for RNA targeting to tumors demonstrated the importance of RNA orientation in maximizing ligand exposure to biological systems [238]. Similarly, another study determined that AuNPs functionalized with siRNA/PEG-poly(L-lysine)- thiol complexes exhibited significantly greater intratumoral accumulation following intravenous injection than siRNA/PEG-poly(L-lysine) polyplexes [239]. Because these constructs exhibit architecture similar to PEI-SNAs, we would expect to see similar results using our system.

4.4 Conclusions In conclusion, nanocarrier architecture and the resulting orientation of therapeutic cargo can be engineered to promote siRNA delivery. To demonstrate this, we compared the cellular interactions of two PEI-based siRNA carriers with similar size and surface charge, but different architecture: PEI-SNAs, in which PEI is wrapped around a highly oriented SNA nanoparticle core, and randomly assembled PEI-siRNA polyplexes lacking controlled architecture. We found that PEI-SNAs undergo enhanced and more rapid cellular uptake than polyplexes. Cellular uptake of

117 nanoparticles is broadly understood to rely on size and surface charge, but our evidence suggests a prominent role for architecture as well. We further demonstrate using confocal microscopy studies that while PEI-SNAs and polyplexes exhibit similar intracellular stability as evidenced by colocalization analysis of fluorescent siRNA and PEI, PEI-SNAs undergo decreased accumulation within lysosomes, identifying another advantage conferred by their architecture. Indeed, these advantageous cellular interactions enhanced the siRNA delivery potency of PEI-SNAs by 10-fold relative to polyplexes. Finally, cytocompatibility studies suggest that the association of PEI with a complex SNA core decreases its cytotoxicity relative to PEI- based polyplexes, allowing the use of more polycation. Our studies provide critical insight into a novel design parameter for engineering siRNA carriers based on polycations and warrant future investigation of nanocarrier architecture on cellular, organ, and organism-level interactions.

118 Chapter 5

EVALUATION OF GLI1-TARGETED POLYETHYLENIMINE-SPHERICAL NUCLEIC ACID NANOPARTICLES AGAINST GLIOBLASTOMA CELLS

5.1 Introduction RNA interference (RNAi) therapeutics have received tremendous attention recently for their potential to revolutionize the management of diseases with a known genetic basis. In a clinical setting, RNAi offers the ability to silence the expression of genes that promote disease progression with greater potency and specificity than small molecule drugs. However, one unresolved challenge towards recognizing the clinical potential of RNAi is maximally delivering siRNA to the targeted disease site [74,105]. This is because siRNA is highly susceptible to degradation by nucleases present ubiquitously in physiological conditions, is cleared rapidly from circulation, and cannot cross cellular membranes due its large size and negative charge. As a result, there is a need to develop carriers to protect siRNA and efficiently deliver it to sites of disease. One of the greatest challenges in developing clinically translatable siRNA carriers is achieving a balance between efficacy and toxicity; many strongly cationic carriers that are highly effective for intracellular siRNA delivery are also toxic due to their tendencies to destabilize cellular membranes and trigger immune responses [125,126,240]. The reverse is also true: many materials with more favorable biocompatibility profiles are less effective as transfection agents. Towards the goal of identifying carriers to maximize siRNA delivery efficacy and minimize toxicity, we have recently developed a hybrid delivery vehicle

119 consisting of a spherical nucleic acid (SNA) core and a polyethylenimine (PEI) shell that provides greater cellular uptake and endosomal escape than highly biocompatible SNAs and greater cytocompatibility and transfection efficiency than PEI-siRNA polyplexes [151]. SNAs, which consist of radially-oriented, densely-arranged siRNA stabilized on a gold nanoparticle core, exhibit controlled siRNA architecture that imparts unique properties favoring siRNA delivery to biological systems. Most notably, SNAs are rapidly taken up by >50 cell types, provide steric and electrostatic hindrances against endonucleases, and do not induce an immune response in animal models [143,206,241]. Further, their successful delivery of siRNA and miRNA to glioblastoma tumors has prompted the first clinical trial evaluating their use as therapeutics for glioblastoma and gliosarcoma [143,144,150]. However, their tendency to accumulate within late endosomes limits their transfection efficiency [147]. Simultaneously, polycationic materials such as PEI have been widely investigated as siRNA carriers for their ability to encapsulate and protect nucleic acids and rapidly enter cells. In particular, PEI is strongly cationic due to its high amine content, which enables PEI-siRNA polyplexes to overcome a major bottleneck to their intracellular delivery: achieving endosomal escape [125,126]. However, toxicity common to polycationic materials has limited its clinical translation [91]. Interestingly, we found that PEI-wrapped SNAs afford enhanced cellular uptake and endosomal escape to improve gene silencing efficacy while dramatically reducing the cytotoxicity of PEI [151], warranting continued investigation of these constructs as therapeutic gene regulatory agents for diseases with a genetic basis. One devastating disease that might benefit from the continued development and application of RNAi therapeutics is glioblastoma multiforme (GBM). GBM is the

120 most common and lethal neurological tumor in adults, representing nearly 50% of all malignant primary brain tumors [17]. The three main treatment strategies, surgery, radiation, and chemotherapy, often fail to completely eradicate the disease, in part due to intrinsic or acquired resistance to therapy characteristic of GBM tumors. Consequently, tumor recurrence is inevitable and nearly 100% of patients eventually succumb to disease [157,158]. Recently, a growing body of work has identified a role of developmental pathways in GBM progression. One such pathway is the Hedgehog (Hh) signaling pathway [56,63,171]. During development, Hh signaling plays a key regulatory role in tissue patterning and stem cell maintenance. It is subsequently inactivated in most differentiated adult tissues, during which the Ptch1 transmembrane receptor represses Smo, a G-protein coupled receptor (GPCR)-like protein. However, aberrant pathway activation is implicated in many cancers, including GBM. In GBM, this aberrant activation is most commonly initiated when extracellular Sonic Hh ligand binds to Ptch1, which relieves its suppression of Smo to drive an intracellular signaling cascade that ultimately results in the translocation of the Gli1 transcription factor to the nucleus, where it transcriptionally regulates the expression of genes that promote GBM progression. When activated in GBM, Hh/Gli1 signaling induces proliferation and survival signaling, and also maintains an aggressive subpopulation of GBM cells called glioblastoma stem cells (GSCs) [56,57,171]. GSCs are pluripotent, highly tumorigenic cells that resist therapy and generate the bulk of GBM tumors. This is enabled by their ability to divide asymmetrically; they can divide to produce additional GSCs by self-renewal, or they can divide and differentiate into non- tumorigenic progenitor and differentiated GBM cells [10,46]. Further, GSCs cycle slowly, rendering them highly refractory to radiation and chemotherapies that target

121 rapidly dividing cells. As a result, traditional therapies fail to eliminate GSCs, which in turn drive tumor recurrence. Due to their highly aggressive characteristics, GSCs must be eradicated to achieve complete tumor regression. GSCs may be eliminated by targeting the developmental pathways that maintain them, such as Hh signaling. Therefore, we hypothesized that GBM progression could be halted by targeting Gli1 with RNAi therapeutics. In this work, we developed Gli1-targeted PEI-SNAs, characterized their cellular uptake and intracellular trafficking mechanisms, and demonstrated that they can reduce the chemoresistance and stemness of GBM cells (Figure 5.1).

Figure 5.1. PEI-SNAs targeting Gli1 were developed to reduce the chemoresistance and stemness of glioblastoma cells.

122 Our results demonstrate that PEI-SNAs bind cells via scavenger receptors and undergo both dynamin-dependent, caveolae-mediated endocytosis and macropinocytosis. Following endocytosis, the majority of PEI-SNAs are routed to late endosomes and lysosomes. Despite this, Gli1 PEI-SNAs can successfully reduce the expression of the Hh signaling components Gli1 and Smo by ~30%, and this corresponds to a ~30% reduction in Hh transcriptional target genes that promote GBM progression including CyclinD1, c-Myc, Bcl2, and ABCG2. Gene regulation by Gli1 PEI-SNAs also mediates a 30% reduction in GBM cell proliferation and distinguishable onset of cellular senescence. Further, Gli1 PEI-SNAs reduce the metabolic activity of GBM cells by ~60% alone or in combination with TMZ. Importantly, Gli1 PEI-SNAs impair the self-renewal capacity of GBM cells as indicated by a 30-40% reduction in the expression of stemness genes and by impairing the formation of neurospheres. This translates to a substantial improvement in neurosphere chemosensitivity as demonstrated by a two-fold increase in the fraction of cells undergoing apoptosis in response to low doses of TMZ. These findings underscore the potential for siRNA therapeutics targeting Gli1 to reduce GBM resistance to therapy and warrant the continued development of polycation-SNA hybrid siRNA carriers for potent gene regulation.

5.2 Materials and Methods

5.2.1 Nanoparticle Synthesis and Characterization SNAs and PEI-SNAs were prepared and characterized as described in Chapter 2. siRNA sequences used are as follows: Scr, 5’-UGA UAA GUC GUU GGU GCA CdT-3’; Gli1, 5’-UUG GGA GUC AAA UUC CUG GCdT-3’. Using an OliGreen

123 assay to measure siRNA loading,[153] Scr-SNAs contained 53.3 +/- 6.5 duplexes and Gli1-SNAs contained 58.7 +/- 11.2 duplexes. All loading was measured prior to coating SNAs with PEI.

5.2.2 Cell Culture and Stable Gene Expression U87-MG cells were purchased from American Type Culture Collection (ATCC, Manassas, VA), cultured in Dulbecco’s Modified Eagle Medium (DMEM; VWR, Radnor, PA) supplemented with 10% fetal bovine serum (FBS; Gemini Bio- Products, West Sacramento, CA), and maintained in a humidified incubator at 37ºC,

5% CO2. For neurosphere experiments, U87-MG cells were seeded as a single-cell suspension in low adhesion plates cultured in NeuroCult NS-A (STEMCELL Technologies, Vancouver, BC, Canada) medium supplemented with recombinant human epidermal growth factor (EGF, 20 ng/mL), recombinant human basic fibroblast growth factor (bFGF, 10 ng/mL), and heparin (2 µg/mL). To study the intracellular trafficking of PEI-SNAs, U87-MG cells were stably transformed with GFP-tagged endosomal markers using standard lentiviral transduction procedures. Briefly, Rab5- GFP (Addgene # 56530), Rab7-GFP (Addgene # 12605), Rab11-GFP (Addgene # 12674), or LAMP1-GFP (Addgene # 34831) were cloned into a lentiviral transfer vector (System Biosciences, Palo Alto, CA) by restriction cloning. Lentiviral particles were produced by triple-transfecting (TransIT-Lenti transfection reagent; Mirus Bio, Madison, WI) 293TN cells (System Biosciences, Palo Alto, CA) with either transfer vector and lentiviral packaging and envelope plasmids (Addgene #12260,12259). Lentivirus was harvested, filtered, and diluted in cell culture medium to transform U87-MG cells. Cells stably expressing the desired protein were selected with 1 mg/mL puromycin (VWR, Radnor, PA).

124 5.2.3 Assessment of Endocytosis Pathways Endocytosis pathways responsible for the cellular uptake of PEI-SNAs were investigated using flow cytometry and fluorescence microscopy. U87-MG cells were seeded in 24-well culture plates to 60-70% confluence and treated with inhibitors to various components of endocytic pathways according to the conditions detailed in Table S1. Treated cells were incubated with Cy5-labeled PEI-SNAs (40 nM siRNA) for four hours and either trypsinized for flow cytometry analysis using a NovoCyte flow cytometer (ACEA Biosciences, San Diego, CA) or fixed in 4% formaldehyde, counterstained with DAPI and DyLightTM488-Phalloidin (Cell Signaling Technology, Danvers, MA), and mounted on slides using gelvatol mounting medium for fluorescence microscopy using a Zeiss AxioObserver.Z1 microscope (Zeiss, Thornwood, NY). PEI-SNA distribution into neurospheres was conducted similarly by culturing neurospheres for 1 week prior to treatment with endocytosis inhibitors and PEI-SNAs and subsequently dissociating them for flow cytometry analysis or imaging them intact using a Zeiss LSM880 confocal microscope to collect z-stacks of whole spheres.

5.2.4 Confocal Microscopy and Image Analysis U87-MG cells stably expressing Rab5-GFP, Rab7-GFP, Rab11-GFP, or LAMP1-GFP were seeded in 35 mm glass bottom dishes to 60-70% confluence. Cells were incubated with Cy5-labeled PEI-SNAs for 24 hours, counterstained with CellMask Orange (Thermo Fisher Scientific, Waltham, MA), and fed with FluoroBrite DMEM Media (Thermo Fisher Scientific, Waltham, MA). Cells were imaged live using a Zeiss LSM880 confocal microscope equipped with an incubated stage. Z- stacks were acquired to analyze PEI-SNA/endosomal colocalization throughout the

125 entire volume of cells. Quantitative colocalization analysis was performed as previously described in Chapter 4 to calculate Mander’s colocalization coefficients [216] for each endosomal marker.

5.2.5 Gene Expression Analysis by qPCR The gene regulation potency of Gli1 PEI-SNAs was evaluated using qPCR. U87-MG cells were seeded at 100,000 cells/well in a 12-well culture plate and grown overnight. Cells were incubated with PEI-SNAs (50 nM siRNA) for 24 hours in complete medium, fed with fresh medium, and incubated a further 48 hours at 37ºC,

5% CO2. RNA was isolated using an Isolate II RNA Mini Kit (Bioline, Taunton, MA), and qPCR was performed using SensiFASTTM SYBR® One-Step master mix on a LightCycler® 96 (Roche Diagnostics Corporation, Indianapolis, IN). Gene expression was normalized to that of GAPDH. Primer sequences are listed in Table S2.

5.2.6 Evaluation of Cell Proliferation, Senescence, and Viability Gli1 PEI-SNAs were evaluated for their ability to reduce GBM cell proliferation, senescence, and viability in combination with TMZ. To measure proliferation, U87-MG cells were treated with Gli1 PEI-SNAs as described for qPCR analysis, then assessed using an EdU assay (Thermo Fisher Scientific, Waltham, MA). Briefly, treated U87-MG cells were incubated with 10 µM EdU for 16 hours, then trypsinized, fixed in 4% formaldehyde and permeabilized with 0.05% saponin prior to staining according to the manufacturer’s protocol. EdU incorporation was detected by an AlexaFluor488-azide and measured by flow cytometry (Ex 488nm/Em 530/30 nm) using a NovoCyte flow cytometer. Senescence was evaluated in cells treated with PEI- SNAs using a Senescence Associated β-Galactosidase (SAβGal) kit (Cell Signaling

126 Technology, Danvers, MA) according to the manufacturer’s protocol. Stained cells were imaged using a Zeiss Axioobserver.Z1 microscope equipped with a color camera. U87-MG cell metabolic activity (taken to correlate with viability) was assessed following co-treatment with Gli1 PEI-SNAs and TMZ. Cells were seeded in 96-well plates at a density of 2500 cells/well and treated with Scr or Gli1 PEI-SNAs as described above. Treated cells were then exposed to TMZ concentrations ranging from 0 – 1000 µM for an additional 72 hours, then evaluated using an MTT assay (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s protocol. Each treatment group was performed in triplicate, and separate cells treated with equivalent volumes of DMSO (used to reconstitute and store TMZ) were used to ensure that toxicity was due to TMZ treatment rather than the DMSO vehicle.

5.2.7 Assessment of Self-Renewal and Neurosphere TMZ Response A neurosphere culture model was used to assess the impact of Gli1 PEI-SNAs on the self-renewal capacity of U87-MG cells. First, U87-MG cells were seeded in standard adherent culture at a density of 25,000 cells/mL in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Cells were treated with Gli1 PEI-SNAs or Scr PEI-SNAs for 24 hours at 37°C, 5% CO2, fed with fresh medium, and incubated a further 48 hours at 37°C, 5% CO2. Cells were then trypsinized and seeded in suspension at a density of 10,000 cells/mL to grow as neurospheres as described above for one week at 37°C, 5% CO2. After 7 days, the entirety of each well was imaged using a Zeiss AxioObserver.Z1 microscope equipped with automated stage control. Images were stitched using Zeiss Efficient Navigation software (ZEN 2.0; Zeiss), exported for analysis, and then spheres were counted and

127 measured in ImageJ. The sphere size reported is the diameter of the projected area imaged by brightfield microscopy. To assess the effect of Gli1PEI-SNAs on neurosphere response to TMZ, we used a similar neurosphere model. Cells were pre-treated with PEI-SNAs in adherent culture and then seeded as neurospheres as described above. At 72 hours post-seeding of the neurospheres, cells were exposed to PEI-SNAs for an additional 24 hours, and then resuspended in fresh medium containing TMZ. Spheres were incubated in TMZ for 72 hours and subsequently assessed for apoptosis using an AnnexinV-FITC/PI assay (Cayman Chemical Company, Ann Arbor, MI). AnnexinV-FITC/PI staining was analyzed on a NovoCyte flow cytometer.

5.2.8 Statistical Analysis All experiments were performed in triplicate, and data represent means +/- standard deviations from three independent replicates unless otherwise indicated. Groups with significant differences were identified using one-way ANOVA with posthoc Tukey test (or Student’s t-test when only two groups were compared), and differences were considered significant at p < 0.05. Statistical tests were performed in MATLAB software (MathWorks, Natick, MA), and flow cytometry data was analyzed using NovoExpress software (ACEA Biosciences, San Diego, CA).

5.3 Results and Discussion

5.3.1 Evaluation of PEI-SNA Endocytosis Mechanism To begin our evaluation of Gli1 PEI-SNAs, we were interested in understanding the mechanism by which PEI-SNAs are taken up by cells. Importantly, the mechanism of endocytosis can determine the intracellular fate of the siRNA cargo,

128 which must reach the cytosol to facilitate gene silencing. Based on previous studies that have separately demonstrated both PEI-based polyplexes [242] and SNAs [146] undergo clathrin-independent, caveolae-mediated endocytosis, we expected to observe similar results. We further anticipated that PEI-SNA would bind to cells via class A scavenger receptors, which has been previously reported for SNAs [224]. For our studies, we used chemical inhibitors to different endocytosis mechanisms to determine which pathways were necessary for cellular uptake of PEI-SNAs fluorescently labeled with Cy5-siRNA. Cells were subsequently analyzed by flow cytometry and fluorescence microscopy to identify changes in both net cellular association and in cellular localization of PEI-SNAs. As expected, we found that fucoidan (FCD), a potent scavenger receptor inhibitor, significantly decreases the net cellular uptake of PEI-SNAs by up to 75%, and little association of PEI-SNAs with cells was detected by fluorescence microscopy (Figure 5.2). Similarly, we found that methyl-β- cyclodextrin (MβCD), a caveolae/lipid raft inhibitor, but not chlorpromazine (CPZ), a clathrin inhibitor, significantly reduces the uptake of PEI-SNAs by up to 50% (Figure 5.2). While cargo internalized by clathrin or caveolae-dependent mechanisms can ultimately be routed to lysosomes, caveolae can also fuse with intermediate vesicles called caveosomes, which do not acidify and can avoid lysosomal trafficking in some cases [114,243]. Interestingly, one study found that polyplexes taken up by clathrin- mediated endocytosis are routed to lysosomes for degradation, but polyplexes taken up by caveolae-dependent mechanisms were more likely to evade lysosomes and induce efficient transfection [242]. We additionally found that the cellular uptake of PEI- SNAs occurs in a dynamin-dependent manner, indicated by a significant 50% reduction in PEI-SNA uptake in cells treated with Dynasore (Figure 5.2). Cytochalasin

129 D, which disrupts actin polymerization to inhibit phagocytosis, which typically only occurs in specialized cells, does not significantly reduce PEI-SNA uptake (Figure 5.2).

Figure 5.2. Uptake mechanism for PEI-SNAs. *p<0.05, **p<0.005 relative to cells receiving no inhibitors by one-way ANOVA with posthoc Tukey. scale= 50 µm.

5.3.2 Evaluation of PEI-SNA Intracellular Trafficking Having demonstrated that PEI-SNAs undergo both dynamin-dependent, caveolae-mediated endocytosis and macropinocytosis, we next sought to understand the consequences of this uptake mechanism on the intracellular trafficking of PEI- SNAs. While our previous research demonstrates that PEI-SNAs are visible in early endosomes within 1 hour and undergo reduced lysosomal accumulation relative to SNAs and PEI-siRNA polyplexes [151], we are further interested in identifying other subcellular compartments to which PEI-SNAs traffic. In these studies, we used U87- MG cells engineered to stably express fluorescently-tagged endocytic compartments, including Rab5-GFP, Rab7-GFP, Rab11-GFP, and LAMP1-GFP to label early endosomes, late endosomes, recycling endosomes, and lysosomes, respectively. Cells

130 were incubated with Cy5-PEI-SNAs for 24 hours and imaged by confocal microscopy to capture z-stacks containing the entire volume of the cells. Images were subsequently analyzed to determine Manders’ colocalization coefficients (MCCs) as a quantitative measure of the fractional overlap of the Cy5 and GFP fluorescent signals [216]. MCCs range from 0 – 1, where MCC = 0 indicates no colocalization, while MCC = 1 indicates perfect colocalization. Cy5-PEI-SNAs colocalize to the greatest extent with LAMP1-GFP (MCC = 0.87 +/- 0.04) and Rab7-GFP (MCC = 0.84 +/- 0.04), which was significantly greater than colocalization with Rab11-GFP (MCC = 0.73 +/- 0.06) or Rab5-GFP (MCC = 0.65 +/-0.04, Figure 5.3 A,B). Additionally, LAMP1-GFP colocalizes with Cy5-PEI-SNAs (MCC = 0.61 +/- 0.002) to a significantly greater extent than other endosomal markers (MCCs all ~0.45, Figure 5.3A,B). Therefore, we conclude that PEI-SNAs accumulate to the greatest extent within Rab7+ late endosomes and LAMP1+ lysosomes. This is consistent with previous studies, which demonstrate that avoiding retention within the endolysosomal network remains a challenge for RNAi therapeutics [116,151,219]. However, PEI- SNAs appear to be continually endocytosed through 24 hours incubation, as indicated by relatively strong colocalization with Rab5+ early endosomes (MCC > 0.5). Unexpectedly, we also observed relatively high colocalization with Rab11+ recycling endosomes, suggesting a mechanism by which PEI-SNAs may be exocytosed. While it is unusual for nanoparticles to exhibit colocalization with Rab11+ vesicles [244], previous studies investigating the intracellular fate of SNAs have demonstrated that while the gold core is retained within lysosomes, the oligonucleotide shell is gradually cleared from cells over a period of 24 hours, likely due to nuclease degradation and

131 oligonucleotide exocytosis [147]. A similar process may explain our results, which show that Cy5-siRNA can colocalize with Rab11+ recycling endosomes.

Figure 5.3. Intracellular trafficking of PEI-SNAs. A) Confocal microscopy was used to visualize Cy5-PEI-SNA localization to endocytic compartments after 24 hours incubation with cells. Endocytic compartments were labeled by stably expressing GFP-tagged markers for early endosomes (Rab5+), recycling endosomes (Rab11+), late endosomes (Rab7+), or lysosomes (LAMP1+). Scale bar = 20 µm. B) Results from quantitative colocalization analysis to calculate the fractional overlap of Cy5-PEI- SNAs with GFP-endosomal markers and vice versa. MCC = Manders’ colocalization coefficient, *p<0.05 by one-way ANOVA with posthoc Tukey.

132 5.3.3 Gli1 PEI-SNAs Regulate the Expression of Hh Signaling Components and Downstream Target Genes Next, we evaluated the gene regulation potency of Gli1 PEI-SNAs using qPCR to measure mRNA expression of Hh signaling components and downstream target genes known to promote GBM progression. Gli1 PEI-SNAs significantly reduce Gli1 expression by 30% relative to PEI-SNAs carrying a scrambled control siRNA sequence (Scr PEI-SNAs; Figure 5.4). We also observed a significant 30% decrease in the expression of the GPCR-like protein Smo (Figure 4), which is also considered an oncogene that can induce aberrant Hh signaling to further drive cancer progression [245]. Mutations to Smo are common to many cancers, and therapeutics targeting Smo often fail in the clinic because acquired Smo mutations render tumor cells refractory to Smo-targeted therapy [246]. Our results demonstrate that Gli1-targeted RNAi can reduce the expression of both oncogenes. Interestingly, we did not observe significant differences in the expression of the transmembrane receptor Ptch1. Ptch1 normally inhibits Smo to suppress Hh activity, and is thus considered the tumor suppressor of the Hh pathway, so we were encouraged that Gli1 PEI-SNAs did not reduce the expression of this gene. We also observed ~30% decreases in the pro-GBM Gli1 transcriptional target genes cyclin D1, c-Myc, Bcl-2, and ABCG2 (Figure 5.4). Cyclin D1 promotes cell cycle progression through the G1/S transition, and its overexpression in GBM correlates with poor prognosis. Further, silencing cyclin D1 inhibits proliferation, induces apoptosis, and reduces the invasive capacity of GBM cells [247]. While c-Myc also promotes proliferation in numerous cancers, it has been further recognized as a key regulator of glioma cancer stem cells [10,248]. Bcl-2 opposes apoptosis and contributes substantially to GBM therapy resistance, and inhibiting Bcl-2 can enhance the response of glioma cells to TMZ [249]. ABCG2 is an

133 ABC transporter known for its drug efflux activity to promote cell survival and also regulates self-renewal and the expression of stemness genes in glioma cells [250]. Taken together, these results demonstrate that Gli1 PEI-SNAs can downregulate multiple genes that contribute to GBM progression, chemoresistance, and stemness, and we expect the observed gene regulation to exert tumor suppressive effects.

5.3.4 Gli1 PEI-SNAs Slow Proliferation, Induce Senescence, and Reduce Chemoresistance of GBM Cells To test whether the gene regulation capacity of Gli1 PEI-SNAs is sufficient to induce a tumor suppressive response against GBM cells, we investigated the impact of Gli1 PEI-SNAs on proliferation, senescence, and response to TMZ. Using a 5-ethynyl- 2ʹ-deoxyuridine (EdU) assay to measure proliferation, we found that Gli1 PEI-SNAs significantly reduce the fraction of proliferative (EdU+) U87-MG cells by ~30%

134 relative to cells treated with Scr PEI-SNAs (Figure 5A). This is consistent with our measured decreases in cyclin D1 and c-Myc expression (Figure 5.4) and with previous reports demonstrating that suppressing Gli1 reduces GBM proliferation [56,171]. In parallel with our observed decrease in proliferation, we recently reported that silencing Gli1 can induce senescence in PTEN-deficient U87-MG cells [171]. To determine whether Gli1 PEI-SNAs could also elicit this effect, we employed a senescence associated β-galactosidase (SAβGal) assay to visually identify cells undergoing senescence. SAβGal staining demonstrated that U87-MG cells treated with Gli1 PEI- SNAs broadly undergo senescence, as indicated by teal SAβGal staining, while cells treated with Scr PEI-SNAs do not (Figure 5.5B). During senescence, cells enter a stable state of cell cycle arrest but remain metabolically active [251]. While much remains unknown regarding the anticancer implications of senescence, a growing body of research has demonstrated that senescence can induce tumor suppressive effects and even compensate for apoptosis in some contexts, such as when expression of the tumor suppressor PTEN is lost [184]. Interestingly, senescent cells acquire a senescence-associated secretory phenotype (SASP), which can trigger immune responses to either promote tumor progression or promote immune clearance of tumor cells [198,252]. In aggregate, our data suggests that Gli1 PEI-SNA-mediated senescence exerts tumor suppressive effects, though future research should continue to investigate the effects of the SASP induced by Gli1 silencing on GBM progression. Next, we were interested in investigating the effects of Gli1 PEI-SNAs on GBM cell sensitivity to the frontline chemotherapy, TMZ. One way we could envision using such a Gli1-targeted therapy is to first administer our Gli1 RNAi therapeutic to downregulate cellular resistance mechanisms and subsequently treat with TMZ. To

135 model this schedule, we first delivered treated U87-MG cells with Gli1 PEI-SNAs, incubated the cells for 72 hours, then treated cells with TMZ doses ranging from 0 – 1000 µM for an additional 72 hours and measured cellular metabolic activity using a 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide (MTT) assay. Our results demonstrate that Gli1 PEI-SNAs significantly reduce U87-MG metabolic activity by ~60% alone and in combination with TMZ (Figure 5.5C). Interestingly, TMZ alone was ineffective against U87-MG cells except at the highest dose tested (1000 µM), which is 20-fold higher than the maximum clinically feasible TMZ dose at the tumor site [190]. Our own previous studies and studies conducted by others agree that U87-MG cells grown in adherent culture are indeed highly refractory to low TMZ doses [171,185]. However, when TMZ was applied following treatment with Gli1 PEI-SNAs, we observed a dramatic reduction in U87-MG metabolic activity relative to cells treated with TMZ and Scr PEI-SNAs, particularly at low TMZ doses. This suggests that Gli1 PEI-SNAs might reduce the required dose of TMZ to achieve a therapeutic effect.

136

Figure 5.5. Gli1 PEI-SNAs reduce proliferation and chemoresistance. A) By EdU assay, Gli1 PEI-SNAs reduce U87 proliferation by ~30%. Flow cytometric histograms (left) and quantification of EdU+ cells (right). Data are means +/- STDs, *p=0.02. B) SAβGal staining (teal) demonstrating that Gli1PEI-SNAs induce senescence in U87 cells. Scale = 100 µm. C) By MTT assay, Gli1 PEI-SNAs reduce U87 metabolic activity alone and in combination with TMZ, *p<0.01 relative to Scr PEI- SNA control with equivalent TMZ dose by one-way ANOVA with posthoc Tukey.

137 5.3.5 Gli1 PEI-SNAs Reduce Stemness and Impair Self-Renewal of GBM Cells We were next interested in whether Gli1 PEI-SNAs could reverse the stemness of GBM cells to impair the self-renewal capacity of GSCs. First, we determined the extent to which Gli1 PEI-SNAs could reduce the expression of stemness genes in adherent-cultured cells. By qPCR, we found that Gli1 PEI-SNAs significantly reduce the expression of CD133 and Nanog by 38% and 25%, respectively (Figure 5.6B). We also observed a slight but insignificant decrease in Sox2 expression (Figure 5.6B). To determine whether these changes are sufficient to impair self-renewal, we used a sphere-formation assay in which U87-MG cells are seeded in suspension to form multicellular neurospheres that exhibit increased stemness relative to adherent- cultured cells [171]. In these experiments, U87-MG cells were pre-treated with PEI- SNAs for 72 hours, then dissociated and suspended in serum-free medium supplemented with growth factors to form neurospheres over a period of 7 days (Figure 5.6A). Brightfield images show that neurospheres grown from cells pre-treated with Gli1 PEI-SNAs are significantly smaller than those grown from cells pre-treated with Scr PEI-SNAs, exhibiting a ~40 µm decrease in average projected diameter (Figure 5.6C,D). Further, we found that treatment with Gli1 PEI-SNAs reduced the total number of neurospheres formed by 26% (Figure 5.6D), though this result was not statistically significant. Notably, we detected this reduction in size and number of neurospheres 10 days after Gli1 PEI-SNA treatment, whereas previous work demonstrates that a single transfection with SNAs can silence gene expression up to 96 hours [143]. Here, we demonstrate that PEI-SNAs are a useful vehicle for achieving sustained effects of Gli1-targeted therapy. Consistent with our other results, this suggests that Gli1 PEI-SNAs can reduce the number of stem-like cells capable of forming neurospheres through self-renewal. Our findings are corroborated by prior

138 studies, which reported that either transient transfection of siGli1 or treatment with a Smo inhibitor into GBM neurospheres significantly hinder neurosphere growth.[56,57] Similar results have been obtained using the pharmacological Gli inhibitor, GANT61, encapsulated within PLGA nanoparticles. PLGA-GANT61 reduced tumorsphere formation in colon and breast cancer cell lines.[253] The ability of Hh inhibitors to decrease the number of cancer stem-cells (CSCs) has been demonstrated in vivo as well; one study reported that PLGA-PEG nanoparticles delivering another pharmacological Gli inhibitor, HPI-1, reduced the number of ALDH+ CSCs in a murine orthotopic pancreatic cancer xenograft [254]. In totality, these results suggest that Gli1 PEI-SNAs can impair the self-renewal capacity of U87- MG cells and may be useful to eliminate the aggressive GSC subpopulation.

139

Figure 5.6. Gli1 PEI-SNAs reduce stemness and impair self-renewal of U87 cells. A) Schematic depicting neurosphere culture model and experimental design; red cells illustrate GSCs. B) qPCR showing expression of genes associated with stemness following exposure to PEI-SNAs. Gene expression is normalized to that of GAPDH. Data are means +/- STDs, *p<0.001 relative to Scr PEI-SNA. C) Representative brightfield images of neurospheres cultured from U87 cells after exposure to PEI-SNAs. Scale = 200 µm. D) Gli1 PEI-SNAs reduce the size and number of neurospheres formed, as measured from 25 tiled brightfield images per treatment group per experiment, *p=0.03 by Student’s t-test.

140 5.3.6 Gli1 PEI-SNAs Potentiate Neurosphere Response to TMZ Chemotherapy Finally, we were interested in whether treating neurospheres with Gli1 PEI- SNAs could improve neurosphere response to TMZ. We first confirmed that PEI- SNAs could enter neurospheres by confocal microscopy and flow cytometry. Confocal microscopy demonstrates that Cy5-labeled PEI-SNAs can easily penetrate small neurospheres (<100 µm), but accumulate to the greatest extent within the periphery of larger neurospheres (Figure 5.7A) within 24 hours. Flow cytometry analysis of neurospheres dissociated after treatment demonstrates that Cy5-labeled PEI-SNAs can enter 91.3% of cells grown as neurospheres (Figure 5.7B). Taken together with our microscopy data, this suggests that PEI-SNAs can penetrate neurospheres greater than 100 µm in diameter but show greater accumulation near the periphery than in the center. We were further interested in examining the mechanism by which PEI-SNAs penetrate neurospheres. We found that the scavenger receptor inhibitor FCD reduces the uptake of PEI-SNAs by neurosphere-cultured cells by 43%, and the caveolae/lipid raft inhibitor MβCD reduces uptake by 64% (Figure 5.7C), demonstrating that these are mechanisms are required for PEI-SNA distribution through neurospheres. To investigate the capacity for Gli1 PEI-SNAs to potentiate neurosphere response to TMZ, we treated cells according to the timeline in Figure 5.7D. Adherent- cultured cells were pre-treated with PEI-SNAs for 72 hours, seeded as neurospheres, and incubated a further 72 hours. Spheres were treated with a second PEI-SNA pulse for 24 hours, resuspended in TMZ-containing medium, and incubated 72 hours. Spheres were dissociated and analyzed for apoptosis using an AnnexinV-FITC/PI assay. We found that spheres primed with Gli1 PEI-SNAs prior to low-dose TMZ treatment (50 µM) exhibited a nearly two-fold increase in the fraction of apoptotic cells relative to cells primed with Scr PEI-SNAs (Figure 5.7E,F). Gli1 PEI-SNAs also

141 increased the fraction of apoptotic cells in response to 100 and 200 µM TMZ by 62% and 50%, respectively. This is consistent with previous research, which demonstrated that the pharmacological Hedgehog pathway inhibitor cyclopamine can increase the fraction of Caspase3+ apoptotic GBM stem cell cultures in combination with TMZ [56]. An additional report found that cyclopamine could potentiate the cytotoxic effects of TMZ in CD133+ glioma stem cells [55]. However, the clinical translation of cyclopamine has been deterred by severe toxicity and rapid clearance from circulation [64]. Further, cyclopamine targets Smo, upstream of Gli1 in the Hedgehog signaling pathway, and a growing body of work has demonstrated that cancer cells frequently acquire resistance to Smo antagonists, potentially limiting their utility [255]. Among Gli1 inhibitors furthest along in development is GANT61, which, despite its potency in preclinical models, is poorly stable at physiological conditions [255]. Further, the ability of GANT61 to cross the blood-brain barrier remains poorly understood. Because SNAs have previously demonstrated sufficient stability in physiological conditions and can accumulate within glioma xenografts and improve tumor response to TMZ [143,144,149], we are hopeful that a Gli1-targeted construct might behave similarly. Based on our results and the results of others, we hypothesize that Gli1 PEI- SNAs may also accumulate within GBM tumors in vivo to exert antitumor effects that also extend to GSCs, though this remains to be evaluated in future work.

142

143 Figure 5.7. Gli1 PEI-SNAs potentiate neurosphere response to TMZ chemotherapy. A) Confocal microscopy visualizing Gli1 PEI-SNA distribution into small (top) and large (bottom) neurospheres. Scale = 100 µm. B) Flow cytometric histogram of Cy5-PEI-SNA uptake by cells grown as neurosperes. C) Flow cytometric analysis of the mechanism by which Gli1 PEI-SNAs distribute throughout neurospheres. Data are geometric mean fluorescence intensity (GMFI) +/- STD normalized to cells receiving no inhibitors, *p<0.05 by one-way ANOVA with posthoc Tukey. D) Experimental timeline for determining effect of co-treating neurospheres with Gli1 PEI-SNAs and TMZ. E) Flow cytometric density plots of AnnexinV/PI apoptosis analysis of neurospheres co-treated with Gli1 PEI-SNAs and TMZ. F) Summary of AnnexinV/PI apoptosis analysis. Data are means +/- standard deviations from n=2 replicates. *p<0.05 by one-way ANOVA with posthoc Fisher’s least significant difference test.

5.4 Conclusions In this work, we have demonstrated that Gli1 PEI-SNAs can improve the response of GBM cells and aggressive GBM neurospheres to TMZ chemotherapy. We found that Gli1 PEI-SNAs bind scavenger receptors on GBM cells to undergo endocytosis in a caveolae/lipid raft/dynamin-dependent manner and that this leads to trafficking through the classical endolysosomal pathway. Despite many of the nanoparticles accumulating within late endosomes and lysosomes, Gli1 PEI-SNAs could silence the expression of genes associated with the Hedgehog signaling pathway that promote the aggressive, chemoresistant phenotype of GBM. Further, this leads to a decrease in proliferation that correlates with an onset of GBM cell senescence, as well as a decrease in metabolic activity with or without TMZ co-treatment. Importantly, Gli1 PEI-SNAs reduced the growth of stem-like neurospheres and sensitized neurospheres to low doses of TMZ chemotherapy. These results warrant further development of PEI-SNAs and Gli1-targeted therapies to alleviate drug resistance and recurrence for GBM patients.

144 Chapter 6

CONCLUSIONS AND FUTURE DIRECTIONS

6.1 Introduction This thesis has sought to advance the field of nanoparticle-mediated RNAi for GBM therapy by evaluating Hh/Gli1 signaling as an impactful therapeutic target to eliminate GSCs and differentiated GBM cells. Further, it has developed a novel polycation-spherical nucleic acid composite nanomaterial to enable efficient gene regulation while mitigating the toxicity associated with cationic polymers. Finally, Gli1-targeted PEI-SNAs were demonstrated to reduce the chemoresistance and stemness of GBM cells. In this chapter, the significance of this work will be discussed and future directions to further advance Gli1-targeted RNAi therapeutics will be considered.

6.2 Hh/Gli1 Signaling as a Target for GBM Therapy Resistance to TMZ chemotherapy substantially hinders successful GBM treatment, contributing to an almost 100% mortality rate. Towards the goal of alleviating GBM chemoresistance, suppressing Hh signaling was investigated as an adjuvant to TMZ chemotherapy in GBM cell lines in Chapter 3 [171]. RNAi targeting Gli1 was found to reduce cell metabolic activity in combination with TMZ and reduce the multidrug efflux activity of GBM cells. Additionally, pharmacological Gli inhibition was found to modulate nuclear p53 expression and decrease MGMT expression in combination with TMZ. Importantly, it was demonstrated that Gli

145 inhibition increases apoptosis in glioma stem-like cells in combination with TMZ, and this reduces the size and number of neurospheres grown from glioma stem-like cells. These findings were consistent with previous studies demonstrating that Hh signaling is essential for maintaining GSCs and that pharmacological suppression of either Smo or Gli proteins can potentiate the effects of chemotherapy [55–57,63]. Chapter 3 describes the novel finding that while Gli1 silencing does not induce apoptosis in the absence of TMZ co-treatment, Gli1 silencing without TMZ co- treatment induces a robust senescent response in PTEN-deficient cells but not in PTEN-proficient cells. Senescence induced by Gli1 silencing was further confirmed to depend on the absence of PTEN. While previous research has defined a clear role of Hh/Gli1 signaling in cell cycle progression such as through the regulation of Cyclin D and Cyclin E in both cancer and normal development [60,256,257], no work has previously demonstrated a link between Hh signaling and the onset of senescence. Though Gli1 inhibition has previously been linked to cell cycle arrest in GBM cells [60], senescence represents a distinct state of irreversible cell cycle arrest that could have profound implications for the utility of Hh-targeted therapies. While senescence can induce tumor suppressive consequences in that cells that undergo senescence no longer maintain the ability to proliferate indefinitely, the role of senescence in tumor progression remains poorly understood [197]. Most notably, senescent cells acquire a senescence-associated secretory phenotype (SASP) in which pro-inflammatory cytokines are released that have the potential to drive tumor progression [197]. Therefore, the findings presented in this thesis suggest that the therapeutic benefits of Hh-targeted therapies are likely to be highly dependent on the status of additional signaling components, such as PTEN expression. Interestingly, while silencing Gli1

146 without TMZ co-treatment induced senescence in adherent-cultured PTEN-deficient cells, pharmacological Gli1 suppression with TMZ co-treatment induced apoptosis in neurosphere-cultured PTEN-deficient cells, suggesting that stemness genes likely contribute to mediating the cell fate decision between apoptosis and senescence. Future work should seek to further dissect the molecular pathways responsible for this response using more clinically-relevant models of PTEN(+/-) GBM tumors, such as molecularly-classified patient-derived murine xenografts or genetic murine models of GBM, such as a cre-lox system with PTEN deletion [258].

6.3 Polycation-Spherical Nucleic Acid Nanoparticles as RNAi Therapeutics Clinical translation of siRNA nanocarriers is hindered by limited knowledge regarding the parameters that regulate interactions between nanocarriers and biological systems. To address this, Chapter 4 investigated the influence of polycation-based nanocarrier architecture on intracellular siRNA delivery [151]. The cellular interactions of two polycation-based siRNA carriers that have similar size and surface charge but different siRNA orientation were compared: (1) PEI-SNAs, in which polyethylenimine is wrapped around a spherical nucleic acid core containing radially- oriented siRNA, and (2) randomly assembled PEI-siRNA polyplexes that lack controlled architecture. It was found that PEI-SNAs undergo enhanced cellular uptake and lysosomal evasion, exhibit dramatically enhanced gene silencing potency, and reduce the cytotoxicity of PEI relative to PEI-siRNA polyplexes. These studies provide critical insight into design considerations for engineering siRNA carriers and warrant future investigation of how nanocarrier architecture influences cellular, organ, and organism-level interactions. While previous research essentially creating poly(β-amino ester) (PβAE)-wrapped SNAs demonstrated that

147 these constructs exhibit excellent gene silencing, a direct comparison to PβAE-siRNA polyplexes could not be made because these materials alone failed to form complete polyplexes [210]. Taken together, these results suggest that another advantage conferred by a polycation-SNA hybrid system is increased colloidal stability relative to a polycation-siRNA polyplex system. Chapter 4 demonstrates that SNA core architecture can dramatically improve the biological interactions of polycation-based siRNA delivery systems. These findings have important implications towards the field of polycation-based siRNA delivery because the clinical translation of polycations is notoriously limited by toxicity. This work presents a novel strategy to reduce the toxicity of polycations while simultaneously improving the gene silencing potency of standalone SNAs. Future work to further the development of polycation-SNAs will be discussed in greater detail in the next section. In Chapter 5, PEI-SNAs targeting Gli1 were developed and evaluated against GBM cells. Gli1 PEI-SNAs were found to silence tumor-promoting Hedgehog pathway genes and downstream target genes that promote the chemoresistant phenotype of GBM. This was found to slow proliferation, induce senescence, and decrease cellular metabolic activity in combination with TMZ. Most importantly, Gli1 PEI-SNAs impaired the self-renewal capacity of GBM cells and substantially improved neurosphere response to low doses of TMZ. These results underscore the potential for siRNA therapeutics targeting Gli1 to reduce GBM resistance to therapy and warrant further development of Gli1 PEI-SNAs to alleviate drug resistance and recurrence for GBM patients.

148 6.4 Future Directions This thesis aimed to advance GBM therapy by evaluating a novel therapeutic target and developing a novel nanoparticle system for RNAi against this target. As the results of this work demonstrate, much further research is needed to best utilize Hh- targeted therapies for GBM patients and to maximally induce gene silencing while minimizing toxicity and off-target effects. The results presented in Chapter 3 underscore the importance of effectively predicting which GBM patients are going to be most susceptible to Hh-targeted therapy. In this work, PTEN status was identified as a molecular indicator of whether GBM cells treated with Gli1 siRNA will undergo senescence, which can either have tumor suppressive or tumor-promoting implications. It is likely that other major signaling components also influence cell fate decisions in response to Gli1 silencing, such as p53, Rb, IDH1, NF1, EGFR, PDGFR, or other genes commonly mutated or lost in GBM, and future research should seek to further dissect the signaling pathways that mediate cellular response to Hh-targeted therapy. Further, it is probable that patient response to Gli1-targeted therapy is subtype-specific. Of the four GBM subtypes (mesenchymal, classical, proneural, and neural), mesenchymal and proneural/neural GBMs may be most likely to respond to Hh-targeted therapy, as Gli1 expression is associated both with a mesenchymal-like phenotype [259,260] and with stem cell-like phenotypes common to proneural/neural GBMs [25]. Therefore, future work should investigate the efficacy of Hh-targeted therapies against individual GBM subtypes to choose the best strategy for individual patients. Additionally, the observation that Gli1 silencing induces senescence in a loss- of-PTEN-dependent manner is interesting, but the clinical implications of this response remain unclear. Future research should seek to dissect the SASP acquired by

149 PTEN-deficient cells to understand whether the consequences of siGli1-induced senescence will exert primarily tumor suppressive or tumor-promoting effects. Further, due to the innate heterogeneity throughout GBM tumors, it is likely that the response to Gli1-targeted therapy will not be uniform throughout a tumor, introducing the possibility that some tumor cells will undergo senescence while others will not. To best understand the impact of Gli1-targeted therapy, future research should elucidate the effect of a senescent subpopulation of GBM cells on the rest of the tumor. Previous research has demonstrated that senescent cells can secrete factors that induce extracellular matrix remodeling, proliferation, and angiogenesis. However, the link between Hh pathway suppression and the onset of senescence has never been investigated in GBM and therefore warrants future research to study these behaviors. Chapter 4 demonstrates that PEI-SNAs can outperform PEI-siRNA polyplexes by increasing net cellular uptake, decreasing lysosomal accumulation, improving gene silencing potency, and reducing cytotoxicity. However, the mechanism by which this occurs remains largely unknown. It was hypothesized that the architecture provided by PEI-SNAs increases the strength of the electrostatic interaction between siRNA and PEI to enhance the stability of the construct and prevent dissociation of free PEI, which induces toxicity. This work initially sought to investigate this by confocal microscopy, in which PEI-SNAs and polyplexes were dual-labeled with TRITC-PEI and Cy5-siRNA. If PEI-SNAs exhibited increased stability relative to polyplexes, TRITC and Cy5 signals would have been highly colocalized, whereas polyplex dissociation may have been indicated by decreased signal colocalization. However, TRITC and Cy5 signals remained highly colocalized from each construct, suggesting that traditional confocal microscopy lacks the spatial resolution necessary to detect

150 potentially very small amounts of TRITC-PEI or Cy5-siRNA dissociating from the complexes. The use of super-resolution microscopy and a Förster Resonance Energy Transfer-based system to indicate loss of colocalization may provide higher quality data to investigate this hypothesis. This hypothesis could additionally be tested using heparin displacement and gel electrophoresis to evaluate binding strength of the complexes. Further, the studies reported in Chapter 4 were conducted entirely using in vitro models, and future work should seek to evaluate whether the observed advantage of PEI-SNAs over polyplexes translates to in vivo models. In particular, future research should compare the biodistribution profiles of each complex, circulation half- life, ability to accumulate within tumors, toxicity to major organs, immunogenicity, and gene silencing potency of each complex. Chapter 5 investigated Gli1-targeted PEI-SNAs to reduce the chemoresistance and stemness of GBM cells. This work was also conducted purely using in vitro models, and future work should next evaluate the therapeutic efficacy of Gli1 PEI- SNAs in combination with TMZ chemotherapy using a subcutaneous or orthotopic GBM xenograft, either derived from established cell lines (U87, etc.) or from patient tumor tissue. Outcomes that would provide crucial information towards the development of PEI-SNAs include tumor growth, animal survival, GSC content within treated tumors, and the capacity for treated and resected tumors to propagate additional tumors in mice. While the evaluation of Gli1 PEI-SNAs in this thesis demonstrated the excellent potential for a composite polycation-SNA system to offer efficient gene regulation, the high toxicity associated with 25 kDa branched PEI [125,261–263]

151 likely hinders the clinical utility of these constructs. Much research is currently ongoing to modify PEI in order to reduce its toxicity while retaining its potency, and these approaches might improve the performance of PEI-SNAs as well [125,126,131]. For example, PEGylated PEI molecules have been employed to deliver siRNA [132], which would offer the advantage of neutralizing some of the strong positive charge of PEI and also improve the circulation time of PEI-SNAs. Along these lines, primary amines that render PEI too toxic for clinical use could be used to functionalize PEI- SNAs with additional targeting or stealth agents. Hyaluronic acid (HA) is well-known to bind the CD44 receptor overexpressed on GBM cells and to also act as a stealth agent for intravenously injected nanomaterials [264,265]. Incorporating HA into PEI- SNAs could therefore serve the dual purpose of providing an active targeting scheme to improve intratumoral retention and mitigate the toxicity of PEI. Further, while SNAs have been previously demonstrated to cross the BBB [143], the increased size of PEI-SNAs may prohibit BBB penetration [151]. However, exploiting primary amines within the PEI backbone to functionalize PEI-SNAs with ligands targeting transcytosis pathways, such as angiopep-2 or transferrin molecules may overcome this potential barrier [266–268]. Finally, polycations with enhanced biocompatibility or biodegradability could be employed to replace PEI in this system. For example, poly- L-lysine and poly-L-arginine have been widely employed as siRNA delivery vehicles and may improve the biocompatibility of a polycation-SNA system [269–272]. Finally, perhaps the greatest advantage of a nanotechnology-based delivery system is the ability to impart a single construct with multiple functionalities. Here, gold nanoparticles were used as the SNA core for their synthesis and functionalization simplicity. However, a solid gold nanoparticle core only acts as a scaffold, and SNAs

152 have been previously synthesized with numerous functional core materials, including quantum dots, liposomes, iron oxide nanoparticles, gold-silica nanoshells, and others [135]. PEI- or polycation-SNAs could be synthesized with quantum dots or iron oxide nanoparticles as the core material to incorporate additional imaging modalities, with liposomes as the core to incorporate chemotherapy drugs, or with gold-silica nanoshells as the core to enable photothermal therapy of GBM tumors. Though gene regulation with nanomaterials remains in its infancy, RNAi enabled by nanomedicine has the potential to transform the treatment of cancer and other diseases with a known genetic basis. Much work remains in order to 1) identify effective therapeutic targets and predict which patients will benefit from molecularly- targeted therapies, 2) fully elucidate the structural parameters that govern the interactions of nanomaterials and biological systems, and 3) rationally-design RNAi therapeutics to maximize therapeutic efficacy while minimizing off-target effects. Ultimately, nanoparticle-mediated RNAi could offer a minimally-invasive and potent therapeutic strategy for GBM and other genetic diseases that could dramatically improve patient outcomes.

153 REFERENCES

1. National Cancer Institute. Cancer Statistics [Internet]. 2018. Available from https://www.cancer.gov/about-cancer/understanding/statistics

2. Noone A, Howlader N, Krapcho M, Miller D, Brest A, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis D, Chen H, Feuer E, Cronin K. SEER Cancer Statistics Review, 1975-2015, National Cancer Institute. Bethesda, MD [Internet]. 2018. Available from https://seer.cancer.gov/csr/1975_2015/

3. Mukherjee S. The Emperor of All Maladies. Scribner; 2010.

4. Plesca M, Bordea C, El Houcheimi B, Ichim E, Bludaru A. Evolution of radical mastectomy for breast cancer. J Med Life. 2016; 9: 183–6.

5. Maddox W, Capenter JJ, Laws H, Soong S, Cloud G, Urist M, Balch C. A randomized prospective trial of radical (Halsted) mastectomy versus modified radical mastectomy in 311 breast cancer patients. Ann Surg. 1983; 198: 207–12.

6. Farber S. Some observations on the effect of folic acid antagonists on acute leukemia and other forms of incurable cancer. Blood. 1949; 4: 160–7.

7. Nowell P. The clonal evolution of tumor cell populations. Science (80- ). 1976; 194: 23–8.

8. Hanahan D, Weinberg RA. The Hallmarks of Cancer. Cell. 2000; 100: 57–70.

9. Campbell LL, Polyak K. Breast tumor heterogeneity: Cancer stem cells or clonal evolution? Cell Cycle. 2007; 6: 2332–8. doi: 10.4161/cc.6.19.4914.

10. Bradshaw A, Wickremsekera A, Tan ST, Peng L, Davis PF, Itinteang T. Cancer Stem Cell Hierarchy in Glioblastoma Multiforme. Front Surg [Internet]. 2016; 3: 1–15. doi: 10.3389/fsurg.2016.00021.

11. Beck B, Blanpain C. Unravelling cancer stem cell potential. Nat Rev Cancer [Internet]. Nature Publishing Group; 2013; 13: 727–38. doi: 10.1038/nrc3597.

12. Bao S, Wu Q, McLendon RE, Hao Y, Shi Q, Hjelmeland AB, Dewhirst MW, Bigner DD, Rich JN. Glioma stem cells promote radioresistance by preferential

154 activation of the DNA damage response. Nature. 2006; 444: 756–60. doi: 10.1038/nature05236.

13. Chen J, Li Y, Yu T-S, McKay RM, Burns DK, Kernie SG, Parada LF. A restricted cell population propagates glioblastoma growth after chemotherapy. Nature. 2012; 488: 522–6. doi: 10.1038/nature11287.

14. Adams JM, Strasser A. Is tumor growth sustained by rare cancer stem cells or dominant clones? Cancer Res. 2008; 68: 4018–21. doi: 10.1158/0008- 5472.CAN-07-6334.

15. Colotta F, Allavena P, Sica A, Garlanda C, Mantovani A. Cancer-related inflammation, the seventh hallmark of cancer: links to genetic instability. Carcinogenesis. 2009; 30: 1073–81.

16. Hanahan D, Weinberg RA. Hallmarks of cancer: The next generation. Cell [Internet]. Elsevier Inc.; 2011 [cited 2014 Jul 9]; 144: 646–74. doi: 10.1016/j.cell.2011.02.013.

17. Ostrom QT, Gittleman H, Liao P, Rouse C, Chen Y, Dowling J, Wolinsky Y, Kruchko C, Barnholtz-Sloan J. CBTRUS Statistical Report: Primary Brain and Central Nervous System Tumors Diagnosed in the United States in 2007-2011. Neuro Oncol. 2014; 16: iv1-iv63. doi: 10.1093/neuonc/nou223.

18. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, Belanger K, Brandes AA, Marosi C, Bogdahn U, Curschmann J, Janzer RC, Ludwin SK, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005; 352: 987–996.

19. Stupp R, Hegi ME, Mason WP, van den Bent MJ, Taphoorn MJ, Janzer RC, Ludwin SK, Allgeier A, Fisher B, Belanger K, Hau P, Brandes AA, Gijtenbeek J, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol [Internet]. Elsevier Ltd; 2009; 10: 459–66. doi: 10.1016/S1470-2045(09)70025-7.

20. Lombardi M, Assem M. Glioblastoma Genomics: A Very Complicated Story. In: De Vleeschouwer S, editor. Glioblastoma. 2016. p. 3–25.

21. Armento A, Ehlers J, Schotterl S, Naumann U. Molecular Mechanisms of Glioma Cell Motility. Glioblastoma. 2016. p. 73–94.

22. Lombard A, Goffart N, Rogister B. Glioblastoma circulating cells: Reality, trap or illusion? Stem Cell Int. 2015; 2015.

155 23. Aubry M, de Tayrac M, Etcheverry A. “From the core to beyond the margin”: A genomic picture of glioblastoma intratumor heterogeneity. Oncotarget. 2015; 6: 12094–109.

24. Brat DJ, Van Meir EG. Vaso-occlusive and prothrombotic mechanisms associated with tumor hypoxia, necrosis, and accelerated growth in glioblastoma. Lab Investig. 2004; 84: 397–405. doi: 10.1038/labinvest.3700070.

25. Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD, Misra A, Nigro JM, Colman H, Soroceanu L, Williams PM, Modrusan Z, Feuerstein BG, et al. Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell. 2006; 9: 157–73. doi: 10.1016/j.ccr.2006.02.019.

26. Sottoriva A, Spiteri I, Piccirillo SGM, Touloumis A, Collins VP, Marioni JC, Curtis C, Watts C, Tavare S. Intratumor heterogeneity in human glioblastoma reflects cancer evolutionary dynamics. Proc Natl Acad Sci [Internet]. 2013; 110: 4009–14. doi: 10.1073/pnas.1219747110.

27. Huang P, Xu A, White F. Oncogenic EGFR signaling networks in glioma. Sci Signal. 2009; 2: http://dx.doi.org/10.1126/scisignal.287re6.

28. Gan HK, Cvrljevic AN, Johns TG. The epidermal growth factor receptor variant III (EGFRvIII): Where wild things are altered. FEBS J. 2013; 280: 5350–70. doi: 10.1111/febs.12393.

29. Rajasekhar V, Wiale A, Socci N, Wiedmann M, Hu X, Holland E. Oncogenic Ras and Akt signaling contribute to glioblastoma formation by differential recruitment of existing mRNAs to polysomes. Mol Cell. 2003; 12: 889–901.

30. Zhu Y, Parada LF. The molecular and genetic basis of neurological tumours. Nat Rev Cancer. 2002; 2: 616–26. doi: 10.1038/nrc866.

31. Cerami E, Demir E, Schultz N, Taylor BS, Sander C. Automated network analysis identifies core pathways in glioblastoma. PLoS One. 2010; 5: e8919.

32. Filbin MG, Dabral SK, Pazyra-Murphy MF, Ramkissoon S, Kung AL, Pak E, Chung J, Theisen MA, Sun Y, Franchetti Y, Sun Y, Shulman DS, Redjal N, et al. Coordinate activation of Shh and PI3K signaling in PTEN-deficient glioblastoma: new therapeutic opportunities. Nat Med [Internet]. Nature Publishing Group; 2013; 19: 1518–23. doi: 10.1038/nm.3328.

33. Massagué J. G1 cell-cycle control and cancer. Nature. 2004; 432: 298–306.

156 34. Knudsen E, Wang J. Targeting the RB-pathway in cancer therapy. Clin Cancer Res. 2010; 16: 1094–9.

35. England B, Huang T, Karsy M. Current understanding of the role and targeting of tumor suppressor p53 in glioblastoma multiforme. Tumor Biol. 2013; 34: 2063–74. doi: 10.1007/s13277-013-0871-3.

36. Hientz K, Mohr A, Bhakta-Guha D, Efferth T. The role of p53 in cancer drug resistance and targeted chemotherapy. Oncotarget [Internet]. 2016; 8: 8921–46. doi: 10.18632/oncotarget.13475.

37. Fernandes C, Costa A, Osorio L, Costa Lago R, Linhares P, Carvalho B, Caeiro C. Current Standards of Care in Glioblastoma Therapy. Glioblastoma. 2016. p. 197–241.

38. Sanai N, Berger M. Intraoperative stimulation techniques for functional pathway preservation and glioma resection. Neurosurg Focus. 2010; 28: E1.

39. Bregy A, Shah AH, Diaz M V, Pierce HE, Ames PL, Diaz D, Komotar RJ. The role of Gliadel wafers in the treatment of high-grade gliomas. Expert Rev Anticancer Ther. 2013; 13: 1453–61.

40. Hegi., Marie-france, Tribolet N De, Weller M, Kros JM, Hainfellner JA, Mason W, Mariani L, Bromberg JEC, Hau P, Stupp R. MGMT Gene Silencing and Benefit from Temozolomide in Glioblastoma. N Engl J Med. 2005; : 997–1003.

41. Kreisl TN, Kim L, Moore K, Duic P, Royce C, Stroud I, Garren N, Mackey M, Butman JA, Camphausen K, Park J, Albert PS, Fine HA. Phase II trial of single-agent bevacizumab followed by bevacizumab plus irinotecan at tumor progression in recurrent glioblastoma. J Clin Oncol. 2009; 27: 740–5. doi: 10.1200/JCO.2008.16.3055.

42. Friedman HS, Prados MD, Wen PY, Mikkelsen T, Schiff D, Abrey LE, Yung WKA, Paleologos N, Nicholas MK, Jensen R, Vredenburgh J, Huang J, Zheng M, et al. Bevacizumab alone and in combination with irinotecan in recurrent glioblastoma. J Clin Oncol. 2009; 27: 4733–40. doi: 10.1200/JCO.2008.19.8721.

43. Singh SK, Hawkins C, Clarke ID, Squire JA, Bayani J, Hide T, Henkelman RM, Cusimano MD, Dirks PB. Identification of human brain tumour initiating cells. Nature [Internet]. 2004; 432: 396–401. doi: 10.1038/nature03031.1.

44. Beier D, Hau P, Proescholdt M, Lohmeier A, Wischhusen J, Oefner PJ, Aigner L, Brawanski A, Bogdahn U, Beier CP. CD133+ and CD133- glioblastoma-

157 derived cancer stem cells show differential growth characteristics and molecular profiles. Cancer Res. 2007; 67: 4010–5. doi: 10.1158/0008-5472.CAN-06- 4180.

45. Wang J, Sakariassen P, Tsinkalovsky O, Immervoll H, Bøe SO, Svendsen A, Prestegarden L, Røsland G, Thorsen F, Stuhr L, Molven A, Bjerkvig R, Enger P. CD133 negative glioma cells form tumors in nude rats and give rise to CD133 positive cells. Int J Cancer. 2008; 122: 761–8. doi: 10.1002/ijc.23130.

46. Lathia J, Mack S. Cancer stem cells in glioblastoma. Genes Dev [Internet]. 2015; 29: 1203–17. doi: 10.1101/gad.261982.115.tumors.

47. Lathia J, Mack S. Cancer stem cells in glioblastoma. Genes … [Internet]. 2015; : 1203–17. doi: 10.1101/gad.261982.115.tumors.

48. Kaur B, Khwaja F, Severson E, Matheny S, DJ B, Van Meir E. Hypoxia and the hypoxia inducible- factor pathway in glioma growth and angiogenesis. Neuro Oncol. 2005; 7: 134–53.

49. Cheray M, Begaud G, Deluche E, Nivet A, Battu S, Lalloue F, Verdier M, Bessette B. Cancer Stem-Like Cells in Glioblastoma. Glioblastoma. 2016. p. 59–71.

50. Takebe N, Miele L, Harris PJ, Jeong W, Bando H, Kahn M, Yang SX, Ivy SP. Targeting Notch, Hedgehog, and Wnt pathways in cancer stem cells: clinical update. Nat Rev Clin Oncol [Internet]. Nature Publishing Group; 2015; . doi: 10.1038/nrclinonc.2015.61.

51. Liebelt BD, Shingu T, Zhou X, Ren J, Shin SA, Hu J. Glioma Stem Cells: Signaling, Microenvironment, and Therapy. Stem Cells Int. 2016; 2016.

52. Zheng H, Ying H, Wiedemeyer R, Yan H, Quayle SN, Ivanova E V., Paik JH, Zhang H, Xiao Y, Perry SR, Hu J, Vinjamoori A, Gan B, et al. PLAGL2 Regulates Wnt Signaling to Impede Differentiation in Neural Stem Cells and Gliomas. Cancer Cell [Internet]. Elsevier Ltd; 2010; 17: 497–509. doi: 10.1016/j.ccr.2010.03.020.

53. Jeon HM, Jin X, Lee JS, Oh SY, Sohn YW, Park HJ, Kyeung MJ, Park WY, Nam DH, DePinho RA, Chin L, Kim H. Inhibitor of differentiation 4 drives brain tumor-initiating cell genesis through cyclin E and notch signaling. Genes Dev. 2008; 22: 2028–33. doi: 10.1101/gad.1668708.

54. Gilbert CA, Daou MC, Moser RP, Ross AH. Gamma-Secretase Inhibitors Enhance Temozolomide Treatment of Human Gliomas By Inhibiting

158 Neurosphere Repopulation and Xenograft Recurrence. Cancer Res. 2010; 70: 6870–9. doi: 10.1158/0008-5472.CAN-10-1378.

55. Ulasov I V, Nandi S, Dey M, Sonabend AM, Lesniak MS. Inhibition of Sonic hedgehog and Notch pathways enhances sensitivity of CD133(+) glioma stem cells to temozolomide therapy. Mol Med [Internet]. 2011 [cited 2014 Jul 15]; 17: 103–12. doi: 10.2119/molmed.2010.00062.

56. Clement V, Sanchez P, de Tribolet N, Radovanovic I, Ruiz i Altaba A. HEDGEHOG-GLI1 Signaling Regulates Human Glioma Growth, Cancer Stem Cell Self-Renewal, and Tumorigenicity. Curr Biol. 2007; 17: 165–72. doi: 10.1016/j.cub.2006.11.033.

57. Bar EE, Chaudhry A, Lin A, Fan X, Schreck K, Matsui W, Piccirillo S, Vescovi AL, DiMeco F, Olivi A, Eberhart CG. Cyclopamine-mediated hedgehog pathway inhibition depletes stem-like cancer cells in glioblastoma. Stem Cells. 2007; 25: 2524–33. doi: 10.1634/stemcells.2007-0166.

58. Takezaki T, Hide T, Takanaga H, Nakamura H, Kuratsu JI, Kondo T. Essential role of the Hedgehog signaling pathway in human glioma-initiating cells. Cancer Sci. 2011; 102: 1306–12. doi: 10.1111/j.1349-7006.2011.01943.x.

59. Cheng J, Gao J, Tao K, Yu P. Prognostic role of Gli1 expression in solid malignancies: a meta-analysis. Sci Rep [Internet]. Nature Publishing Group; 2016; 6: 22184. doi: 10.1038/srep22184.

60. Wang K, Pan L, Che X, Cui D, Li C. Gli1 inhibition induces cell-cycle arrest and enhanced apoptosis in brain glioma cell lines. J Neurooncol [Internet]. 2010 [cited 2014 Jul 15]; 98: 319–27. doi: 10.1007/s11060-009-0082-3.

61. Sarangi A, Valadez JG, Rush S, Abel TW, Thompson RC, Cooper MK. Targeted inhibition of the Hedgehog pathway in established malignant glioma xenografts enhances survival. Oncogene [Internet]. Nature Publishing Group; 2009; 28: 3468–76. doi: 10.1038/onc.2009.208.

62. Wang K, Pan L, Che X, Cui D, Li C. Sonic Hedgehog/GLI signaling pathway inhibition restricts cell migration and invasion in human gliomas. Neurol Res [Internet]. 2010; 32: 975–80. doi: 20444323. ₁

63. Cui D, Xu Q, Wang K, Che X. Gli1 is a potential target for alleviating multidrug resistance of gliomas. J Neurol Sci [Internet]. Elsevier B.V.; 2010 [cited 2014 Aug 7]; 288: 156–66. doi: 10.1016/j.jns.2009.09.006.

64. Rimkus T, Carpenter R, Qasem S, Chan M, Lo H-W. Targeting the Sonic

159 Hedgehog Signaling Pathway: Review of Smoothened and GLI Inhibitors. Cancers (Basel) [Internet]. 2016; 8: 22. doi: 10.3390/cancers8020022.

65. Fu J, Rodova M, Nanta R, Meeker D, Van Veldhuizen PJ, Srivastava RK, Shankar S. NPV-LDE-225 (Erismodegib) inhibits epithelial mesenchymal transition and self-renewal of glioblastoma initiating cells by regulating miR- 21, miR-128, and miR-200. Neuro Oncol. 2013; 15: 691–706. doi: 10.1093/neuonc/not011.

66. Sekulic A, Migden MR, Oro AE, Dirix L, Lewis KD, Hainsworth JD, Solomon J a., Yoo S, Arron ST, Friedlander P a., Marmur E, Rudin CM, Chang ALS, et al. Efficacy and Safety of Vismodegib in Advanced Basal-Cell Carcinoma. N Engl J Med. 2012; 366: 2171–9. doi: 10.1056/NEJMoa1113713.

67. Yauch RL, Dijkgraaf GJP, Alicke B, Januario T, Ah CP, Holcomb T, Pujara K, Stinson J, Callahan CA, Tang T, Bazan JF, Kan Z, Seshagiri S, et al. Smoothened Mutation Confers Resistance to a Hedgehog Pathway Inhibitor in Medulloblastoma. Science (80- ). 2009; 337: 572–5.

68. National Cancer Institute. GDC-0449 in Treating Patients With Recurrent Glioblastoma Multiforme That Can Be Removed by Surgery [Internet]. 2017. Available from https://clinicaltrials.gov/ct2/show/results/NCT00980343?term=vismodegib&co nd=glioblastoma&rank=1

69. Lauth M, Bergström A, Shimokawa T, Toftgård R. Inhibition of GLI-mediated transcription and tumor cell growth by small-molecule antagonists. Proc Natl Acad Sci U S A [Internet]. 2007; 104: 8455–60. doi: 10.1073/pnas.0609699104.

70. List A, Beran M, DiPersio J, Slack J, Vey N, Rosenfeld CS, Greenberg P. Opportunities for Trisenox® (arsenic trioxide) in the treatment of myelodysplastic syndromes. Leukemia. 2003; 17: 1499–507. doi: 10.1038/sj.leu.2403021.

71. Yoshimura Y, Shiino A, Muraki K, Fukami T, Yamada S, Satow T, Fukuda M, Saiki M, Hojo M, Miyamoto S, Onishi N, Saya H, Inubushi T, et al. Arsenic trioxide sensitizes glioblastoma to a Myc inhibitor. PLoS One. 2015; 10: 1–16. doi: 10.1371/journal.pone.0128288.

72. Ding D, Lim KS, Eberhart CG. Arsenic trioxide inhibits Hedgehog, Notch and stem cell properties in glioblastoma neurospheres. Acta Neuropathol Commun [Internet]. 2014; 2: 31. doi: 10.1186/2051-5960-2-31.

160 73. Northwestern University. Arsenic Trioxide, Temozolomide, and Radiation Therapy in Treating Patients With Malignant Glioma That Has Been Removed By Surgery. 2018; . Available from https://clinicaltrials.gov/ct2/show/NCT00275067?term=NCT00275067&rank=1

74. Wilhelm S, Tavares AJ, Dai Q, Ohta S, Audet J, Dvorak HF, Chan WCW. Analysis of nanoparticle delivery to tumours. Nat Rev Mater [Internet]. 2016; 1: 16014. doi: 10.1038/natrevmats.2016.14.

75. Jain RK. Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science [Internet]. 2005 [cited 2014 Jul 9]; 307: 58–62. doi: 10.1126/science.1104819.

76. Maeda H, Wu J, Sawa T, Matsumura Y, Hori K. Tumor vascular permeability and the EPR effect in macromolecular therapeutics: a review. J Control Release [Internet]. 2000; 65: 271–84. Available from http://www.ncbi.nlm.nih.gov/pubmed/10699287

77. Dong X, Mumper RJ. Nanomedicinal strategies to treat multidrug-resistant tumors: current progress. Nanomedicine (Lond) [Internet]. 2010; 5: 597–615. doi: 10.2217/nnm.10.35.

78. Jain RK. Normalization of tumor vasculature: An emerging concept in antiangiogenic therapy. Science (80- ). 2005; 307: 58–62. doi: 10.1126/science.1104819.

79. Glaser T, Han I, Wu L, Zeng X. Targeted nanotechnology in glioblastoma multiforme. Front Pharmacol. 2017; 8: 1–14. doi: 10.3389/fphar.2017.00166.

80. Obermeier B, Daneman R, Ransohoff RM. Development, maintenance and disruption of the blood-brain barrier. Nat Med [Internet]. Nature Publishing Group; 2013; 19: 1584–96. doi: 10.1038/nm.3407.

81. Pardridge W. The Blood-Brain Barrier and Neurotherapeutics. NeuroRx. 2005; 2: 1–2. doi: 10.1111/j.1528-1167.2006.00817.x.

82. Gao J, Wang Z, Liu H, Wang L, Huang G. Liposome encapsulated of temozolomide for the treatment of glioma tumor: preparation, characterization and evaluation. Drug Discov Ther. 2015; 9: 205–12. doi: 10.5582/ddt.2015.01016.

83. Nordling-David MM, Yaffe R, Guez D, Meirow H, Last D, Grad E, Salomon S, Sharabi S, Levi-Kalisman Y, Golomb G, Mardor Y. Liposomal temozolomide drug delivery using convection enhanced delivery. J Control Release. 2017;

161 261: 138–46. doi: 10.1016/j.jconrel.2017.06.028.

84. Noble CO, Krauze MT, Drummond DC, Yamashita Y, Saito R, Berger MS, Kirpotin DB, Bankiewicz KS, Park JW. Novel nanoliposomal CPT-11 infused by convection-enhanced delivery in intracranial tumors: Pharmacology and efficacy. Cancer Res. 2006; 66: 2801–6. doi: 10.1158/0008-5472.CAN-05- 3535.

85. University of California San Francisco. A Phase I Trial of Nanoliposomal CPT- 11 (NL CPT-11) in Patients With Recurrent High-Grade Gliomas [Internet]. clinicaltrials.gov. 2015. p. NCT00734682. Available from https://clinicaltrials.gov/ct2/show/NCT00734682?term=nanoparticle&cond=gli oblastoma&rank=4

86. Battaglia L, Gallarate M, Peira E, Chirio D, Solazzi I, Giordano SMA, Gigliotti CL, Riganti C, Dianzani C. Bevacizumab loaded solid lipid nanoparticles prepared by the coacervation technique: Preliminary in vitro studies. Nanotechnology. IOP Publishing; 2015; 26. doi: 10.1088/0957- 4484/26/25/255102.

87. Chirio D, Gallarate M, Peira E, Battaglia L, Muntoni E, Riganti C, Biasibetti E, Capucchio MT, Valazza A, Panciani P, Lanotte M, Annovazzi L, Caldera V, et al. Positive-charged solid lipid nanoparticles as paclitaxel drug delivery system in glioblastoma treatment. Eur J Pharm Biopharm [Internet]. Elsevier B.V.; 2014; 88: 746–58. doi: 10.1016/j.ejpb.2014.10.017.

88. Zhou J, Patel TR, Sirianni RW, Strohbehn G, Zheng M-Q, Duong N, Schafbauer T, Huttner AJ, Huang Y, Carson RE, Zhang Y, Sullivan DJ, Piepmeier JM, et al. Highly penetrative, drug-loaded nanocarriers improve treatment of glioblastoma. Proc Natl Acad Sci [Internet]. 2013; 110: 11751–6. doi: 10.1073/pnas.1304504110.

89. Householder KT, Diperna DM, Chung EP, Wohlleb GM, Dhruv HD, Berens ME, Sirianni RW. Intravenous delivery of camptothecin-loaded PLGA nanoparticles for the treatment of intracranial glioma. Int J Pharm [Internet]. Elsevier B.V.; 2015; 479: 374–80. doi: 10.1016/j.ijpharm.2015.01.002.

90. Vaishya R, Khurana V, Patel S, Mitra AK. Long-term delivery of protein therapeutics. Expert Opin Drug Deliv. 2015; 12: 415–40.

91. Yin H, Kanasty RL, Eltoukhy AA, Vegas AJ, Dorkin JR, Anderson DG. Non- viral vectors for gene-based therapy. Nat Rev Genet [Internet]. Nature Publishing Group; 2014; 15: 541–55. doi: 10.1038/nrg3763.

162 92. Mingozzi F, High KA. Therapeutic in vivo gene transfer for genetic disease using AAV: Progress and challenges. Nat Rev Genet. 2011; 12: 341–55. doi: 10.1038/nrg2988.

93. Pack DW, Hoffman AS, Pun S, Stayton PS. Design and development of polymers for gene delivery. Nat Rev Drug Discov. 2005; 4: 581–93. doi: 10.1038/nrd1775.

94. Putnam D. Polymers for gene delivery across length scales. Nat Mater. 2006; 5: 439–51. doi: 10.1038/nmat1645.

95. Griveau A, Bejaud J, Anthiya S, Avril S, Autret D, Garcion E. Silencing of miR-21 by locked nucleic acid-lipid nanocapsule complexes sensitize human glioblastoma cells to radiation-induced cell death. Int J Pharm [Internet]. Elsevier B.V.; 2013; 454: 765–74. doi: 10.1016/j.ijpharm.2013.05.049.

96. Ananta JS, Paulmurugan R, Massoud TF. Nanoparticle-Delivered Antisense MicroRNA-21 Enhances the Effects of Temozolomide on Glioblastoma Cells. Mol Pharm. 2015; 12: 4509–17. doi: 10.1021/acs.molpharmaceut.5b00694.

97. Mangraviti A, Tzeng SY, Gullotti D, Kozielski KL, Kim JE, Seng M, Abbadi S, Schiapparelli P, Sarabia-Estrada R, Vescovi A, Brem H, Olivi A, Tyler B, et al. Non-virally engineered human adipose mesenchymal stem cells produce BMP4, target brain tumors, and extend survival. Biomaterials [Internet]. Elsevier Ltd; 2016; 100: 53–66. doi: 10.1016/j.biomaterials.2016.05.025.

98. Kim SS, Rait A, Kim E, Pirollo KF, Nishida M, Farkas N, Dagata JA, Chang EH. A nanoparticle carrying the p53 gene targets tumors including cancer stem cells, sensitizes glioblastoma to chemotherapy and improves survival. ACS Nano. 2014; 8: 5494–514. doi: 10.1021/nn5014484.

99. SynerGene Therapeutics I. Phase II Study of Combined Temozolomide and SGT-53 for Treatment of Recurrent Glioblastoma. clinicaltrials.gov [Internet]. 2017; . Available from https://clinicaltrials.gov/ct2/show/NCT02340156?term=nanoparticle&cond=Gli oblastoma&rank=3

100. Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE, Mello CC. Potent and specific genetic interference by double-stranded RNA in caenorhabditis elegans. Nature. 1998; 391: 806–11. doi: 10.1038/35888.

101. Crooke S. Antisense strategies. Curr Mol Med. 2004; 4: 465–87.

102. Oishi M, Nagasaki Y, Itaka K, Nishiyama N, Kataoka K. Lactosylated

163 poly(ethylene glycol)-siRNA conjugate through acid-labile ??-thiopropionate linkage to construct pH-sensitive polyion complex micelles achieving enhanced gene silencing in hepatoma cells. J Am Chem Soc. 2005; 127: 1624–5. doi: 10.1021/ja044941d.

103. Muratovska A, Eccles MR. Conjugate for efficient delivery of short interfering RNA (siRNA) into mammalian cells. FEBS Lett. 2004; 558: 63–8. doi: 10.1016/S0014-5793(03)01505-9.

104. Titze-de-Almeida R, David C, Titze-de-Almeida SS. The Race of 10 Synthetic RNAi-Based Drugs to the Pharmaceutical Market. Pharm Res [Internet]. Pharmaceutical Research; 2017; . doi: 10.1007/s11095-017-2134-2.

105. Whitehead KA, Langer R, Anderson DG. Knocking down barriers: advances in siRNA delivery. Nat Rev Drug Discov. 2009; 8: 129–38. doi: 10.1038/nrd3182.

106. Albanese A, Tang PS, Chan WCW. The effect of nanoparticle size, shape, and surface chemistry on biological systems. Annu Rev Biomed Eng [Internet]. 2012 [cited 2014 Jul 9]; 14: 1–16. doi: 10.1146/annurev-bioeng-071811- 150124.

107. Jain RK, Stylianopoulos T. Delivering nanomedicine to solid tumors. Nat Rev Clin Oncol. 2010; 7: 653–64.

108. Perrault SD, Walkey C, Jennings T, Fischer HC, Chan WCW. Mediating tumor targeting efficiency of nanoparticles through design. Nano Lett [Internet]. 2009; 9: 1909–15. doi: 10.1021/nl900031y.

109. Chithrani BD, Chan WCW. Elucidating the mechanism of cellular uptake and removal of protein-coated gold nanoparticles of different sizes and shapes. Nano Lett. 2007; 7: 1542–50. doi: 10.1021/nl070363y.

110. Slowing I, Trewyn BG, Lin VS-Y. Effect of Surface Functionalization of MCM-41-Type Mesoporous Silica Nanoparticles on the Endocytosis by Human Cancer Cells. J Am Chem Soc [Internet]. 2006; 128: 14792–3. doi: 10.1021/ja0645943.

111. Arvizo RR, Miranda OR, Thompson MA, Pabelick CM, Bhattacharya R, David Robertson J, Rotello VM, Prakash YS, Mukherjee P. Effect of nanoparticle surface charge at the plasma membrane and beyond. Nano Lett. 2010; 10: 2543–8. doi: 10.1021/nl101140t.

112. Thorek DLJ, Tsourkas A. Size, charge and concentration dependent uptake of iron oxide particles by non-phagocytic cells. Biomaterials. 2008; 29: 3583–90.

164 doi: 10.1016/j.biomaterials.2008.05.015.

113. Corbo C, Molinaro R, Parodi A, Toledano Furman NE, Salvatore F, Tasciotti E. The impact of nanoparticle protein corona on cytotoxicity, immunotoxicity and target drug delivery. Nanomedicine [Internet]. 2016; 11: 81–100. doi: 10.2217/nnm.15.188.

114. Munsell E V., Ross NL, Sullivan MO. Journey to the center of the cell: current nanocarrier design strategies targeting biopharmaceuticals to the cytoplasm and nucleus. Curr Pharm Des. 2016; 22: 1227–44. doi: 10.1016/j.cogdev.2010.08.003.Personal.

115. Sahay G, Alakhova DY, Kabanov A V. Endocytosis of Nanomedicine. J Control Release. 2010; 145: 182–95. doi: 10.1016/j.jconrel.2010.01.036.Endocytosis.

116. Wittrup A, Ai A, Liu X, Hamar P, Trifonova R, Charisse K, Manoharan M, Kirchhausen T, Lieberman J. Visualizing lipid-formulated siRNA release from endosomes and target gene knockdown. Nat Biotechnol [Internet]. Nature Publishing Group; 2015; 33: 1–9. doi: 10.1038/nbt.3298.

117. Granot Y, Peer D. Delivering the right message: Challenges and opportunities in lipid nanoparticles-mediated modified mRNA therapeutics—An innate immune system standpoint. Semin Immunol [Internet]. Elsevier; 2017; 34: 68– 77. doi: 10.1016/j.smim.2017.08.015.

118. Zatsepin TS, Koteliansky V. Lipid nanoparticles for targeted siRNA delivery – going from bench to bedside. Int J Nanomedicine. 2016; 11: 3077–86. doi: 10.2147/IJN.S106625.

119. Sheridan C. With Alnylam’s amyloidosis success, RNAi approval hopes soar. Nat Biotechnol [Internet]. 2017; 35: 995–7. doi: 10.1038/nbt1117-995.

120. Vermeulen LMP, Brans T, Samal SK, Dubruel P, Smedt SC De, Remaut K, Braeckmans K. Endosomal Size and Membrane Leakiness Influence Proton Sponge-Based Rupture of Endosomal Vesicles. 2018; . doi: 10.1021/acsnano.7b07583.

121. Creusat G, Rinaldi AS, Weiss E, Elbaghdadi R, Remy JS, Mulherkar R, Zuber G. Proton sponge trick for ph-sensitive disassembly of polyethylenimine-based sirna delivery systems. Bioconjug Chem. 2010; 21: 994–1002. doi: 10.1021/bc100010k.

122. Akinc A, Thomas M, Klibanov AM, Langer R. Exploring polyethylenimine-

165 mediated DNA transfection and the proton sponge hypothesis. J Gene Med. 2005; 7: 657–63. doi: 10.1002/jgm.696.

123. Rehman ZU, Hoekstra D, Zuhorn IS. Mechanism of polyplex- and lipoplex- mediated delivery of nucleic acids: Real-time visualization of transient membrane destabilization without endosomal lysis. ACS Nano. 2013; 7: 3767– 77. doi: 10.1021/nn3049494.

124. Benjaminsen R V, Mattebjerg MA, Henriksen JR, Moghimi SM, Andresen TL. The possible "proton sponge " effect of polyethylenimine (PEI) does not include change in lysosomal pH. Mol Ther [Internet]. 2013; 21: 149–57. doi: 10.1038/mt.2012.185.

125. Hall A, Lächelt U, Bartek J, Wagner E, Moghimi SM. Polyplex Evolution: Understanding Biology, Optimizing Performance. Mol Ther. Elsevier Ltd.; 2017; 25: 1476–90. doi: 10.1016/j.ymthe.2017.01.024.

126. Neuberg P, Kichler A. Recent developments in nucleic acid delivery with polyethylenimines [Internet]. Advances in Genetics. Elsevier; 2014. 263-288 p. doi: 10.1016/B978-0-12-800148-6.00009-2.

127. Guan H, Zhou Z, Wang H, Jia S-F, Liu W, Kleinerman ES. A small interfering RNA targeting vascular endothelial growth factor inhibits Ewing’s sarcoma growth in a xenograft mouse model. Clin Cancer Res. 2005; 11: 2662–9. doi: 10.1158/1078-0432.CCR-04-1206.

128. Tan PH, Yang LC, Shih HC, Lan KC, Cheng JT. Gene knockdown with intrathecal siRNA of NMDA receptor NR2B subunit reduces formalin-induced nociception in the rat. Gene Ther. 2005; 12: 59–66. doi: 10.1038/sj.gt.3302376.

129. Hendruschk S, Wiedemuth R, Aigner A, Topfer K, Cartellieri M, Martin D, Kirsch M, Ikonomidou C, Schackert G, Temme A. RNA interference targeting survivin exerts antitumoral effects in vitro and in established glioma xenografts in vivo. Neuro Oncol [Internet]. 2013; 12: 28–36. doi: 10.1093/neuonc/nou223.

130. Werth S, Urban-Klein B, Dai L, Höbel S, Grzelinski M, Bakowsky U, Czubayko F, Aigner A. A low molecular weight fraction of polyethylenimine (PEI) displays increased transfection efficiency of DNA and siRNA in fresh or lyophilized complexes. J Control Release. 2006; 112: 257–70. doi: 10.1016/j.jconrel.2006.02.009.

131. Zintchenko A, Philipp A, Dehshahri A, Wagner E. Simple modifications of branched PEI lead to highly efficient siRNA carriers with low toxicity. Bioconjug Chem. 2008; 19: 1448–55. doi: 10.1021/bc800065f.

166 132. Malek A, Czubayko F, Aigner A. PEG grafting of polyethylenimine (PEI) exerts different effects on DNA transfection and siRNA-induced gene targeting efficacy. J Drug Target [Internet]. 2008; 16: 124–39. doi: 10.1080/10611860701849058.

133. Thomas M, Lu JJ, Ge Q, Zhang C, Chen J, Klibanov a. M. Full deacylation of polyethylenimine dramatically boosts its gene delivery efficiency and specificity to mouse lung. Proc Natl Acad Sci [Internet]. 2005; 102: 5679–84. doi: 10.1073/pnas.0502067102.

134. Cutler JI, Auyeung E, Mirkin C a. Spherical nucleic acids. J Am Chem Soc [Internet]. 2012; 134: 1376–91. doi: 10.1021/ja209351u.

135. Kapadia CH, Melamed JR, Day ES. Spherical Nucleic Acid Nanoparticles: Therapeutic Potential. BioDrugs [Internet]. Springer International Publishing; 2018; : 1–13. doi: 10.1007/s40259-018-0290-5.

136. Barnaby SN, Sita TL, Petrosko SH, Stegh AH, Mirkin CA. Therapeutic Applications of Spherical Nucleic Acids. Cancer Treatment and Research [Internet]. 2015. p. 23–51. doi: 10.1007/978-3-319-16555-4.

137. Mirkin CA, Letsinger RL, Mucic RC, Storhoff JJ. A DNA-based method for rationally assembling nanoparticles into macroscopic materials. Nature. 1996; 382: 607–9.

138. Rosi NL, Giljohann D a, Thaxton CS, Lytton-Jean AKR, Han MS, Mirkin C a. Oligonucleotide-modified gold nanoparticles for intracellular gene regulation. Science [Internet]. 2006 [cited 2014 Jul 15]; 312: 1027–30. doi: 10.1126/science.1125559.

139. Young KL, Scott AW, Hao L, Mirkin SE, Liu G, Mirkin C a. Hollow spherical nucleic acids for intracellular gene regulation based upon biocompatible silica shells. Nano Lett [Internet]. 2012; 12: 3867–71. doi: 10.1021/nl3020846.

140. Zheng D, Giljohann D a, Chen DL, Massich MD, Wang X-Q, Iordanov H, Mirkin C a, Paller AS. Topical delivery of siRNA-based spherical nucleic acid nanoparticle conjugates for gene regulation. Proc Natl Acad Sci U S A [Internet]. 2012 [cited 2014 Jul 16]; 109: 11975–80. doi: 10.1073/pnas.1118425109.

141. Giljohann DA, Seferos DS, Prigodich AE, Patel PC, Mirkin CA. Gene regulation with polyvalent siRNA-nanoparticle conjugates. J Am Chem Soc [Internet]. 2009; 131: 2072–3. doi: 10.1021/ja808719p.

167 142. Nemati H, Ghahramani MH, Faridi-Majidi R, Izadi B, Bahrami G, Madani SH, Tavoosidana G. Using siRNA-based spherical nucleic acid nanoparticle conjugates for gene regulation in psoriasis. J Control Release [Internet]. Elsevier; 2017; 268: 259–68. doi: 10.1016/j.jconrel.2017.10.034.

143. Jensen SA, Day ES, Ko CH, Hurley LA, Luciano JP, Kouri FM, Merkel TJ, Luthi AJ, Patel PC, Cutler JI, Daniel WL, Scott AW, Rotz MW, et al. Spherical nucleic acid nanoparticle conjugates as an RNAi-based therapy for glioblastoma. Sci Transl Med [Internet]. 2013 [cited 2014 Jul 12]; 5: 209ra152. doi: 10.1126/scitranslmed.3006839.

144. Kouri FM, Hurley LA, Day ES, Hua Y, Merkel TJ, Queisser A, Peng C, Ritner C, Hao L, Daniel WL, Sznajder JI, Chin L, Giljohann DA, et al. miR-182 integrates apoptosis , growth and differentiation programs in Glioblastoma. Genes Dev. 2015; 2: 732–45. doi: 10.1101/gad.257394.114.732.

145. Tan X, Lu X, Jia F, Liu X, Sun Y, Logan JK, Zhang K. Blurring the Role of Oligonucleotides: Spherical Nucleic Acids as a Drug Delivery Vehicle. J Am Chem Soc [Internet]. 2016; : jacs.6b07554. doi: 10.1021/jacs.6b07554.

146. Choi CHJ, Hao L, Narayan SP, Auyeung E, Mirkin C a. Mechanism for the endocytosis of spherical nucleic acid nanoparticle conjugates. Proc Natl Acad Sci U S A [Internet]. 2013 [cited 2014 Jul 14]; 110: 7625–30. doi: 10.1073/pnas.1305804110.

147. Wu XA, Choi CHJ, Zhang C, Hao L, Mirkin CA. Intracellular fate of spherical nucleic acid nanoparticle conjugates. J Am Chem Soc. 2014; 136: 7726–33. doi: 10.1021/ja503010a.

148. Massich MD, Giljohann DA, Seferos DS, Ludlow LE, Horvath CM, Mirkin CA. Regulating Immune Response Using Polyvalent Nucleic Acid - Gold Nanoparticle Conjugates. Mol Pharm. 2009; 12: 662–8.

149. Sita TL, Kouri FM, Hurley LA, Merkel TJ, Chalastanis A, May JL, Ghelfi ST, Cole LE, Cayton TC, Barnaby SN, Sprangers AJ, Savalia N, James CD, et al. Dual bioluminescence and near-infrared fluorescence monitoring to evaluate spherical nucleic acid nanoconjugate activity in vivo. Proc Natl Acad Sci [Internet]. 2017; : 201702736. doi: 10.1073/pnas.1702736114.

150. Northwestern University. NU-0129 in Treating Patients with Recurrent Glioblastoma or Gliosarcoma Undergoing Surgery. [Internet]. clinicaltrials.gov. 2017. Available from https://clinicaltrials.gov/ct2/show/NCT03020017

151. Melamed JR, Kreuzberger NL, Goyal R, Day ES. Spherical Nucleic Acid

168 Architecture Can Improve the Efficacy of Polycation-Mediated siRNA Delivery. Mol Ther Nucleic Acids [Internet]. Elsevier Ltd.; 2018; 12: 207–19. doi: 10.1016/j.omtn.2018.05.008.

152. Giljohann DA, Seferos DS, Daniel WL, Massich MD, Patel PC, Mirkin C a. Gold nanoparticles for biology and medicine. Angew Chemie - Int Ed [Internet]. 2010 [cited 2014 Jul 10]; 49: 3280–94. doi: 10.1002/anie.200904359.

153. Melamed J, Riley R, Valcourt D, Billingsley M, Kreuzberger N, Day E. Quantification of siRNA Duplexes Bound to Gold Nanoparticle Surfaces. Methods in molecular biology. 2017. p. 1–15. doi: 10.1007/978-1-61779-052- 2_1.

154. Frens G. Controlled Nucleation for the Regulation of the Particle Size in Monodisperse Gold Suspensions. Nat Phys Sci [Internet]. 1973; 241: 20–2. doi: 10.1038/physci241020a0.

155. Turkevich J, Stevenson P, Hillier J. The formation of colloidal gold. J Phys Chem [Internet]. 1953; 57: 670. doi: 10.1021/j50508a015.

156. Grotzky A, Manaka Y, Fornera S, Willeke M, Walde P. Quantification of α- polylysine: a comparison of four UV/Vis spectrophotometric methods. Anal Methods [Internet]. 2010; 2: 1448. doi: 10.1039/c0ay00116c.

157. Weller M, Cloughesy T, Perry JR, Wick W. Standards of care for treatment of recurrent glioblastoma-are we there yet? Neuro Oncol. 2013; 15: 4–27.

158. Seystahl K, Wick W, Weller M. Therapeutic options in recurrent glioblastoma – an update. Crit Rev Oncol Hematol [Internet]. Elsevier Ireland Ltd; 2016; 99: 389–408. doi: 10.1016/j.critrevonc.2016.01.018.

159. Messaoudi K, Clavreul A, Lagarce F. Toward an effective strategy in glioblastoma treatment. Part I: resistance mechanisms and strategies to overcome resistance of glioblastoma to temozolomide. Drug Discov Today. 2015; 20: 899–905. doi: 10.1016/j.drudis.2015.02.011.

160. Perazzoli G, Prados J, Ortiz R, Caba O, Cabeza L, Berdasco M, Gónzalez B, Melguizo C. Temozolomide Resistance in Glioblastoma Cell Lines: Implication of MGMT, MMR, P-Glycoprotein and CD133 Expression. PLoS One [Internet]. 2015; 10: e0140131. doi: 10.1371/journal.pone.0140131.

161. Wang X, Chen JX, Liu YH, You C, Mao Q. Mutant TP53 enhances the resistance of glioblastoma cells to temozolomide by up-regulating O6- methylguanine DNA-methyltransferase. Neurol Sci. 2013; 34: 1421–8. doi:

169 10.1007/s10072-012-1257-9.

162. Thorne AH, Zanca C, Furnari F. Epidermal growth factor receptor in glioblastoma. Neuro Oncol [Internet]. 2016; 0: 1–5. doi: 10.1093/neuonc/nov319.

163. Codrici E, Enciu AM, Popescu ID, Mihai S, Tanase C. Glioma Stem Cells and Their Microenvironments: Providers of Challenging Therapeutic Targets. Stem Cells Int. 2016; 2016. doi: 10.1155/2016/5728438.

164. Sampetrean O, Saya H. Characteristics of glioma stem cells. Brain Tumor Pathol. 2013; 30: 209–14. doi: 10.1007/s10014-013-0141-5.

165. Yi Y, Hsieh I-Y, Huang X, Li J, Zhao W. Glioblastoma Stem-Like Cells: Characteristics, Microenvironment, and Therapy. Front Pharmacol [Internet]. 2016; 7: 1–14. doi: 10.3389/fphar.2016.00477.

166. Lee Y, Lee J-K, Ahn SH, Lee J, Nam D-H. WNT signaling in glioblastoma and therapeutic opportunities. Lab Investig [Internet]. Nature Publishing Group; 2015; 00: 1–14. doi: 10.1038/labinvest.2015.140.

167. Miao W, Liu X, Wang H, Fan Y, Lian S, Yang X, Wang X, Guo G, Li Q, Wang S. P53 upregulated modulator of apoptosis sensitizes drug-resistant U251 glioblastoma stem cells to temozolomide through enhanced apoptosis. Mol Med Rep. 2015; 11: 4165–73. doi: 10.3892/mmr.2015.3255.

168. Bobustuc GC, Baker CH, Limaye A, Jenkins WD, Pearl G, Avgeropoulos NG, Konduri SD. Levetiracetam enhances p53-mediated MGMT inhibition and sensitizes glioblastoma cells to temozolomide. Neuro Oncol. 2010; 12: 917–27. doi: 10.1093/neuonc/noq044.

169. Munoz JL, Rodriguez-Cruz V, Greco SJ, Ramkissoon SH, Ligon KL, Rameshwar P. Temozolomide resistance in glioblastoma cells occurs partly through epidermal growth factor receptor-mediated induction of connexin 43. Cell Death Dis [Internet]. Nature Publishing Group; 2014; 5: e1145. doi: 10.1038/cddis.2014.111.

170. Yoon JW, Gilbertson R, Iannaccone S, Iannaccone P, Walterhouse D. Defining a role for Sonic hedgehog pathway activation in desmoplastic medulloblastoma by identifying GLI1 target genes. Int J Cancer. 2009; 124: 109–19. doi: 10.1002/ijc.23929.

171. Melamed JR, Morgan JT, Ioele SA, Gleghorn JP, Sims-mourtada J, Day ES. Investigating the role of Hedgehog/GLI1 signaling in glioblastoma cell

170 response to temozolomide. Oncotarget. 2018; 9: 27000–15.

172. Fay BL, Melamed JR, Day ES. Nanoshell-mediated photothermal therapy can enhance chemotherapy in inflammatory breast cancer cells. Int J Nanomedicine. 2015; 10: 6931–41.

173. Forster S, Thumser AE, Hood SR, Plant N. Characterization of rhodamine-123 as a tracer dye for use in in vitro drug transport assays. PLoS One. 2012; 7. doi: 10.1371/journal.pone.0033253.

174. Kass M, Witkin A, Terzopoulos D. Snakes: active contour models. Int J Comput Vis. 1988; 1: 321–31.

175. Vincent L. Morphological Grayscale Reconstruction in Image Analysis: Applications and Efficient Algorithms. IEEE Trans Image Process. 1993; 2: 176–201.

176. Kovesi P. Phase Preserving Tone Mapping of Non-Photographic High Dynamic Range Images. Proc Digit Image Comput Tech Appl. 2012; 2012.

177. Zuiderveld K. Contrast Limited Adaptive Histograph Equalization. Graph Gems IV San Diego Acad Press Prof. 1994; : 474–85.

178. Huntzicker EG, Estay IS, Zhen H, Lokteva LA, Jackson PK, Oro AE. Dual degradation signals control Gli protein stability and tumor formation. Genes Dev. 2006; 20: 276–81. doi: 10.1101/gad.1380906.GENES.

179. Pasca di Magliano M, Hebrok M. Hedgehog signalling in cancer formation and maintenance. Nat Rev Cancer [Internet]. 2003 [cited 2014 Sep 17]; 3: 903–11. doi: 10.1038/nrc1229.

180. Amakye D, Jagani Z, Dorsch M. Unraveling the therapeutic potential of the Hedgehog pathway in cancer. Nat Med [Internet]. 2013; 19: 1410–22. doi: 10.1038/nm.3389.

181. Sims-Mourtada J, Izzo JG, Apisarnthanarax S, Wu T-TT, Malhotra U, Luthra R, Liao Z, Komaki R, van der Kogel A, Ajani J, Chao KSC. Hedgehog: an attribute to tumor regrowth after chemoradiotherapy and a target to improve radiation response. Clin cancer Res [Internet]. 2006 [cited 2014 Jul 15]; 12: 6565–72. doi: 10.1158/1078-0432.CCR-06-0176.

182. Munoz JL, Rodriguez-Cruz V, Ramkissoon SH, Ligon KL, Greco SJ, Rameshwar P. Temozolomide resistance in glioblastoma occurs by miRNA-9- targeted PTCH1, independent of sonic hedgehog level. Oncotarget [Internet].

171 2015; 6: 1190–201. Available from http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4359226&tool=pmc entrez&rendertype=abstract

183. Sims-Mourtada J, Izzo JG, Ajani J, Chao KSC. Sonic Hedgehog promotes multiple drug resistance by regulation of drug transport. Oncogene [Internet]. 2007 [cited 2014 Jul 15]; 26: 5674–9. doi: 10.1038/sj.onc.1210356.

184. Lee J-J, Kim BC, Park M-J, Lee Y-S, Kim Y-N, Lee BL, Lee J-S. PTEN status switches cell fate between premature senescence and apoptosis in glioma exposed to ionizing radiation. Cell Death Differ [Internet]. Nature Publishing Group; 2011; 18: 666–77. doi: 10.1038/cdd.2010.139.

185. Mihaliak AM, Gilbert CA, Li L, Daou MC, Moser RP, Reeves A, Cochran BH, Ross AH. Clinically relevant doses of chemotherapy agents reversibly block formation of glioblastoma neurospheres. Cancer Lett. 2010; 296: 168–77. doi: 10.1016/j.canlet.2010.04.005.

186. Li J, Cai J, Zhao S, Yao K, Sun Y, Li Y, Chen L, Li R, Zhai X, Zhang J, Jiang C. GANT61, a GLI inhibitor, sensitizes glioma cells to the temozolomide treatment. J Exp Clin Cancer Res [Internet]. Journal of Experimental & Clinical Cancer Research; 2016; 35: 184. doi: 10.1186/s13046-016-0463-3.

187. Munoz JL, Rodriguez-Cruz V, Greco SJ, Nagula V, Scotto KW, Rameshwar P. Temozolomide Induces the Production of Epidermal Growth Factor to Regulate MDR1 Expression in Glioblastoma Cells. Mol Cancer Ther [Internet]. 2014; 13: 2399–411. doi: 10.1158/1535-7163.MCT-14-0011.

188. Smalley S, Chalmers AJ, Morley SJ. mTOR inhibition and levels of the DNA repair protein MGMT in T98G glioblastoma cells. Mol Cancer [Internet]. 2014; 13: 144. doi: 10.1186/1476-4598-13-144.

189. Kohsaka S, Wang L, Yachi K, Mahabir R, Narita T, Itoh T, Tanino M, Kimura T, Nishihara H, Tanaka S. STAT3 Inhibition Overcomes Temozolomide Resistance in Glioblastoma by Downregulating MGMT Expression. Mol Cancer Ther [Internet]. 2012; 11: 1289–99. doi: 10.1158/1535-7163.MCT-11- 0801.

190. Biswas NK, Chandra V, Sarkar-Roy N, Das T, Bhattacharya RN, Tripathy LN, Basu SK, Kumar S, Das S, Chatterjee A, Mukherjee A, Basu P, Maitra A, et al. Variant allele frequency enrichment analysis in vitro reveals sonic hedgehog pathway to impede sustained temozolomide response in GBM. Sci Rep [Internet]. 2015; 5: 7915. doi: 10.1038/srep07915.

172 191. Amable L, Fain J, Gavin E, Reed E. Gli1 contributes to cellular resistance to cisplatin through altered cellular accumulation of the drug. Oncol Rep. 2014; 32: 469–74. doi: 10.3892/or.2014.3257.

192. Stecca B, Ruiz i Altaba A. A GLI1-p53 inhibitory loop controls neural stem cell and tumour cell numbers. EMBO J. 2009; 28: 663–76. doi: 10.1038/emboj.2009.16.

193. Abe Y, Oda-Sato E, Tobiume K, Kawauchi K, Taya Y, Okamoto K, Oren M, Tanaka N. Hedgehog signaling overrides p53-mediated tumor suppression by activating Mdm2. Proc Natl Acad Sci U S A. 2008; 105: 4838–43. doi: 10.1073/pnas.0712216105.

194. Yoon JW, Lamm M, Iannaccone S, Higashiyama N, Leong KF, Iannaccone P, Walterhouse D. P53 modulates the activity of the GLI1 oncogene through interactions with the shared coactivator TAF9. DNA Repair (Amst) [Internet]. Elsevier B.V.; 2015; 34: 9–17. doi: 10.1016/j.dnarep.2015.06.006.

195. Zbinden M, Duquet A, Lorente-Trigos A, Ngwabyt S-N, Borges I, Ruiz i Altaba A. NANOG regulates glioma stem cells and is essential in vivo acting in a cross-functional network with GLI1 and p53. EMBO J [Internet]. Nature Publishing Group; 2010; 29: 2659–74. doi: 10.1038/emboj.2010.137.

196. Faião-Flores F, Alves-Fernandes DK, Pennacchi PC, Sandri S, Vicente ALSA, Scapulatempo-Neto C, Vazquez VL, Reis RM, Chauhan J, Goding CR, Smalley KS, Maria-Engler SS. Targeting the hedgehog transcription factors GLI1 and GLI2 restores sensitivity to vemurafenib-resistant human melanoma cells. Oncogene [Internet]. Nature Publishing Group; 2017; 36: 1849–61. doi: 10.1038/onc.2016.348.

197. Campisi J, d’Adda di Fagagna F. Cellular senescence: when bad things happen to good cells. Nat Rev Mol Cell Biol. 2007; 8: 729–40. doi: 10.1038/nrm2233.

198. Coppé J-P, Desprez P-Y, Krtolica A, Campisi J. The Senescence-Associated Secretory Phenotype: The Dark Side of Tumor Suppression. Annu Rev Pathol Mech Dis [Internet]. 2010; 5: 99–118. doi: 10.1146/annurev-pathol-121808- 102144.

199. Stepanenko AA, Andreieva S V., Korets K V., Mykytenko DO, Baklaushev VP, Huleyuk NL, Kovalova OA, Kotsarenko K V., Chekhonin VP, Vassetzky YS, Avdieiev SS, Dmitrenko V V. Temozolomide promotes genomic and phenotypic changes in glioblastoma cells. Cancer Cell Int [Internet]. BioMed Central; 2016; 16: 36. doi: 10.1186/s12935-016-0311-8.

173 200. Sato A, Sunayama J, Matsuda KI, Seino S, Suzuki K, Watanabe E, Tachibana K, Tomiyama A, Kayama T, Kitanaka C. MEK-ERK signaling dictates DNA- repair gene MGMT expression and temozolomide resistance of stem-like glioblastoma cells via the MDM2-p53 axis. Stem Cells. 2011; 29: 1942–51. doi: 10.1002/stem.753.

201. Fidoamore A, Cristiano L, Antonosante A, Angelo M, Giacomo E Di, Astarita C, Giordano A, Ippoliti R, Benedetti E, Cimini A. Glioblastoma Stem Cells Microenvironment : The Paracrine Roles of the Niche in Drug and Radioresistance. Stem Cells Int. 2015; 2016. doi: 10.1155/2016/6809105.

202. Khalil S, Fabbri E, Santangelo A, Bezzerri V, Cantù C, Di Gennaro G, Finotti A, Ghimenton C, Eccher A, Dechecchi M, Scarpa A, Hirshman B, Chen C, et al. miRNA array screening reveals cooperative MGMT-regulation between miR-181d-5p and miR-409-3p in glioblastoma. Oncotarget [Internet]. 2016; 7. doi: 10.18632/oncotarget.8618.

203. Gao Y, Chen X, Liu H. Up-regulation of miR-370-3p restores glioblastoma multiforme sensitivity to temozolomide by influencing MGMT expression. Sci Rep [Internet]. Nature Publishing Group; 2016; 6: 32972. doi: 10.1038/srep32972.

204. Wittrup A, Lieberman J. Knocking down disease: A progress report on siRNA therapeutics. Nat Rev Genet. Nature Publishing Group; 2015; 16: 543–52. doi: 10.1038/nrg3978.

205. Singh B, Maharjan S, Park TE, Jiang T, Kang SK, Choi YJ, Cho CS. Tuning the buffering capacity of polyethylenimine with glycerol molecules for efficient gene delivery: Staying in or out of the endosomes. Macromol Biosci. 2015; 15: 622–35. doi: 10.1002/mabi.201400463.

206. Cutler JI, Zhang K, Zheng D, Auyeung E, Prigodich AE, Mirkin C a. Polyvalent nucleic acid nanostructures. J Am Chem Soc. 2011; 133: 9254–7. doi: 10.1021/ja203375n.

207. Seferos DS, Prigodich AE, Giljohann DA, Patel PC, Mirkin CA. Polyvalent DNA nanoparticle conjugates stabilize nucleic acids. Nano Lett. 2009; 9: 308– 11. doi: 10.1021/nl802958f.

208. Barnaby SN, Lee A, Mirkin C a. Probing the inherent stability of siRNA immobilized on nanoparticle constructs. Proc Natl Acad Sci U S A [Internet]. 2014; 111: 9739–44. doi: 10.1073/pnas.1409431111.

209. Barnaby SN, Perelman GA, Kohlstedt KL, Chinen AB, Schatz GC, Mirkin CA.

174 Design Considerations for RNA Spherical Nucleic Acids (SNAs). Bioconjug Chem [Internet]. 2016; . doi: 10.1021/acs.bioconjchem.6b00350.

210. Lee JS, Green JJ, Love KT, Sunshine J, Langer R, Anderson DG. Gold, poly(beta-amino ester) nanoparticles for small interfering RNA delivery. Nano Lett [Internet]. 2009; 9: 2402–6. doi: 10.1021/nl9009793.

211. Grayson ACR, Doody AM, Putnam D. Biophysical and Structural Characterization of Polyethylenimine-Mediated siRNA Delivery in Vitro. 2006; 23: 1868–76. doi: 10.1007/s11095-006-9009-2.

212. Day ES, Zhang L, Thompson PA, Zawaski JA, Caffes CC, Gaber MW, Blaney SM, West JL. Vascular-targeted photothermal therapy of an orthotopic murine glioma model. Nanomedicine. 2011; . doi: 10.2217/NNM.11.189.

213. Day ES, Thompson P a, Zhang L, Lewinski NA, Ahmed N, Drezek R a, Blaney SM, West JL. Nanoshell-mediated photothermal therapy improves survival in a murine glioma model. J Neurooncol [Internet]. 2011 [cited 2014 Jul 16]; 104: 55–63. doi: 10.1007/s11060-010-0470-8.

214. Ahmed N, Ratnayake M, B S, Perlaky L, Dotti G, Wels W, Chattacharjee M, Gilbertson R, Shine H, Weiss H, Rooney C, Heslop H, Gottschalk S. Regression of experimental medulloblastoma following transfer of her2-specific t cells. Cancer Res. 2007; 67: 5957–64.

215. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez J-Y, White DJ, Hartenstein V, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods. United States; 2012; 9: 676–82. doi: 10.1038/nmeth.2019.

216. Manders E, Verbeek F, Aten J. Measurement of co-localization of objects in dual-colour confocal images. J Microsc. 1993; 169: 375–82.

217. Goyal R, Tripathi SK, Vazquez E, Kumar P, Gupta KC. Biodegradable poly(vinyl alcohol)-polyethylenimine nanocomposites for enhanced gene expression in vitro and in vivo. Biomacromolecules [Internet]. 2012; 13: 73–83. doi: 10.1021/bm201157f.

218. Ulasov A V, Khramtsov Y V, Trusov GA, Rosenkranz AA, Sverdlov ED, Sobolev AS. Properties of PEI-based Polyplex Nanoparticles That Correlate With Their Transfection Efficacy. Mol Ther [Internet]. The American Society of Gene & Cell Therapy; 2011; 19: 103–12. doi: 10.1038/mt.2010.233.

219. Gilleron J, Querbes W, Zeigerer A, Borodovsky A, Marsico G, Schubert U,

175 Manygoats K, Seifert S, Andree C, Stöter M, Epstein-Barash H, Zhang L, Koteliansky V, et al. Image-based analysis of lipid nanoparticle–mediated siRNA delivery, intracellular trafficking and endosomal escape. Nat Biotechnol [Internet]. 2013; 31: 638–46. doi: 10.1038/nbt.2612.

220. Shi J, Kantoff PW, Wooster R, Farokhzad OC. Cancer nanomedicine: progress, challenges and opportunities. Nat Rev Cancer [Internet]. Nature Publishing Group; 2017; 17: 20–37. doi: 10.1038/nrc.2016.108.

221. Syga MI, Nicolì E, Kohler E, Shastri VP. Albumin incorporation in polyethylenimine-DNA polyplexes influences transfection efficiency. Biomacromolecules. 2016; 17: 200–7. doi: 10.1021/acs.biomac.5b01308.

222. Nicolì E, Syga MI, Bosetti M, Shastri VP. Enhanced Gene Silencing through Human Serum Albumin-Mediated Delivery of Polyethylenimine-siRNA Polyplexes. 2015; : 1–16. doi: 10.1371/journal.pone.0122581.

223. Gratton SEA, Ropp PA, Pohlhaus PD, Luft JC, Madden VJ, Napier ME, DeSimone JM. The effect of particle design on cellular internalization pathways. Proc Natl Acad Sci [Internet]. 2008; 105: 11613–8. doi: 10.1073/pnas.0801763105.

224. Patel PC, Giljohann D a, Daniel WL, Zheng D, Prigodich AE, Mirkin C a. Scavenger receptors mediate cellular uptake of polyvalent oligonucleotide- functionalized gold nanoparticles. Bioconjug Chem [Internet]. 2010; 21: 2250– 6. doi: 10.1021/bc1002423.

225. Chakrabarti A, Witsenburg JJ, Sinzinger MD, Richter M, Wallbrecher R, Cluitmans JC, Verdurmen WPR, Tanis S, Adjobo-Hermans MJW, Rademann J, Brock R. Multivalent presentation of the cell-penetrating peptide nona-arginine on a linear scaffold strongly increases its membrane-perturbing capacity. Biochim Biophys Acta - Biomembr. Elsevier B.V.; 2014; 1838: 3097–106. doi: 10.1016/j.bbamem.2014.08.001.

226. Giljohann D a., Seferos DS, Patel PC, Millstone JE, Rosi NL, Mirkin C a. Oligonucleotide loading determines cellular uptake of DNA-modified gold nanoparticles. Nano Lett. 2007; 7: 3818–21. doi: 10.1021/nl072471q.

227. Wang M, Miller AD, Thanou M. Effect of surface charge and ligand organization on the specific cell-uptake of uPAR-targeted nanoparticles. J Drug Target [Internet]. 2013; 21: 684–92. doi: 10.3109/1061186X.2013.805336.

228. Chinen AB, Guan CM, Mirkin C a. Spherical Nucleic Acid Nanoparticle Conjugates Enhance G-Quadruplex Formation and Increase Serum Protein

176 Interactions. Angew Chemie [Internet]. 2014 [cited 2014 Nov 13]; : n/a-n/a. doi: 10.1002/ange.201409211.

229. Zheng M, Pavan GM, Neeb M, Schaper AK, Danani A, Klebe G, Merkel OM, Kissel T. Targeting the blind spot of polycationic nanocarrier-based siRNA delivery. ACS Nano. 2012; 6: 9447–54. doi: 10.1021/nn301966r.

230. Elbakry A, Zaky A, Liebl R, Rachel R, Goepferich A, Breunig M. Layer-by- layer assembled gold nanoparticles for sirna delivery. Nano Lett. 2009; 9: 2059–64. doi: 10.1021/nl9003865.

231. Chen Z, Zhang L, He Y, Shen Y, Li Y. Enhanced shRNA delivery and ABCG2 silencing by charge-reversible layered nanocarriers. Small. 2015; 11: 952–62. doi: 10.1002/smll.201401397.

232. Patnaik S, Arif M, Pathak A, Kurupati R, Singh Y, Gupta KC. Cross-linked polyethylenimine-hexametaphosphate nanoparticles to deliver nucleic acids therapeutics. Nanomedicine Nanotechnology, Biol Med [Internet]. Elsevier Inc.; 2010; 6: 344–54. doi: 10.1016/j.nano.2009.07.007.

233. Tripathi SK, Goyal R, Ansari KM, Ravi Ram K, Shukla Y, Chowdhuri DK, Gupta KC. Polyglutamic acid-based nanocomposites as efficient non-viral gene carriers in vitro and in vivo. Eur J Pharm Biopharm [Internet]. Elsevier B.V.; 2011; 79: 473–84. doi: 10.1016/j.ejpb.2011.07.008.

234. Varkouhi AK, Mountrichas G, Schiffelers RM, Lammers T, Storm G, Pispas S, Hennink WE. Polyplexes based on cationic polymers with strong nucleic acid binding properties. Eur J Pharm Sci [Internet]. Elsevier B.V.; 2012; 45: 459–66. doi: 10.1016/j.ejps.2011.09.002.

235. Zhao E, Zhao Z, Wang J, Yang C, Chen C, Gao L, Feng Q, Hou W, Gao M, Zhang Q. Surface engineering of gold nanoparticles for in vitro siRNA delivery. Nanoscale [Internet]. 2012; 4: 5102. doi: 10.1039/c2nr31290e.

236. Han L, Zhao J, Zhang X, Cao W, Hu X, Zou G, Duan X, Liang XJ. Enhanced siRNA delivery and silencing gold-chitosan nanosystem with surface charge- reversal polymer assembly and good biocompatibility. ACS Nano. 2012; 6: 7340–51. doi: 10.1021/nn3024688.

237. Wittrup A, Ai A, Liu X, Hamar P, Trifonova R, Charisse K, Manoharan M, Kirchhausen T, Lieberman J. Visualizing lipid-formulated siRNA release from endosomes and target gene knockdown. Nat Biotechnol [Internet]. Nature Publishing Group; 2015; 33: 870–6. doi: 10.1038/nbt.3298.

177 238. Pi F, Binzel DW, Lee TJ, Li Z, Sun M, Rychahou P, Li H, Haque F, Wang S, Croce CM, Guo B, Evers BM, Guo P. Nanoparticle orientation to control RNA loading and ligand display on extracellular vesicles for cancer regression. Nat Nanotechnol. Springer US; 2018; 13: 82–9. doi: 10.1038/s41565-017-0012-z.

239. Kim HJ, Takemoto H, Yi Y, Zheng M, Maeda Y, Chaya H, Hayashi K, Mi P, Pittella F, Christie RJ, Toh K, Matsumoto Y, Nishiyama N, et al. Precise engineering of siRNA delivery vehicles to tumors using polyion complexes and gold nanoparticles. ACS Nano. 2014; 8: 8979–91. doi: 10.1021/nn502125h.

240. Mitra M, Kandalam M, Rangasamy J, Shankar B, Maheswari UK, Swaminathan S, Krishnakumar S. Novel epithelial cell adhesion molecule antibody conjugated polyethyleneimine-capped gold nanoparticles for enhanced and targeted small interfering RNA delivery to retinoblastoma cells. Mol Vis [Internet]. 2013; 19: 1029–38. Available from http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3654846&tool=pmc entrez&rendertype=abstract

241. Massich MD, Giljohann DA, Schmucker AL, Patel PC, Mirkin C a. Cellular response of polyvalent oligonucleotide-gold nanoparticle conjugates. ACS Nano. 2010; 4: 5641–6. doi: 10.1021/nn102228s.

242. Rejman J, Bragonzi A, Conese M. Role of clathrin-and caveolae-mediated endocytosis in gene transfer mediated by lipo-and polyplexes. Mol Ther [Internet]. The American Society of Gene Therapy; 2005; 12: 468–74. doi: 10.1016/j.ymthe.2005.03.038.

243. Pelkmans L, Kartenbeck J, Helenius A. Caveolar endocytosis of simian virus 40 reveals a new two-step vesicular-transport pathway to the ER. Nat Cell Biol. 2001; 3: 473–83. doi: 10.1038/35074539.

244. Sandin P, Fitzpatrick LW, Simpson JC, Dawson KA. High-speed imaging of rab family small GTpases reveals rare events in nanoparticle trafficking in living cells. ACS Nano. 2012; 6: 1513–21. doi: 10.1021/nn204448x.

245. Ng JMY, Curran T. The Hedgehog’s tale: developing strategies for targeting cancer. Nat Rev Cancer [Internet]. 2011; 11: 493–501. doi: 10.1038/nrc3079.

246. Metcalfe C, De Sauvage FJ. Hedgehog fights back: Mechanisms of acquired resistance against smoothened antagonists. Cancer Res. 2011; 71: 5057–61. doi: 10.1158/0008-5472.CAN-11-0923.

247. Wang J, Wang Q, Cui Y, Liu ZY, Zhao W, Wang CL, Dong Y, Hou L, Hu G, Luo C, Chen J, Lu Y. Knockdown of cyclin D1 inhibits proliferation, induces

178 apoptosis, and attenuates the invasive capacity of human glioblastoma cells. J Neurooncol. 2012; 106: 473–84. doi: 10.1007/s11060-011-0692-4.

248. Wang J, Wang H, Li Z, Wu Q, Lathia JD, McLendon RE, Hjelmeland AB, Rich JN. c-Myc is required for maintenance of glioma cancer stem cells. PLoS One. 2008; 3. doi: 10.1371/journal.pone.0003769.

249. Voss V, Senft C, Lang V, Ronellenfitsch MW, Steinbach JP, Seifert V, Kogel D. The Pan-Bcl-2 Inhibitor (-)-Gossypol Triggers Autophagic Cell Death in Malignant Glioma. Mol Cancer Res [Internet]. 2010; 8: 1002–16. doi: 10.1158/1541-7786.MCR-09-0562.

250. Wee B, Pietras A, Ozawa T, Bazzoli E, Podlaha O, Antczak C, Westermark B, Nelander S, Uhrbom L, Forsberg-Nilsson K, Djaballah H, Michor F, Holland EC. ABCG2 regulates self-renewal and stem cell marker expression but not tumorigenicity or radiation resistance of glioma cells. Sci Rep [Internet]. Nature Publishing Group; 2016; 6: 1–9. doi: 10.1038/srep25956.

251. Campisi J, D’Adda Di Fagagna F. Cellular senescence: When bad things happen to good cells. Nat Rev Mol Cell Biol. 2007; 8: 729–40. doi: 10.1038/nrm2233.

252. Childs BG, Baker DJ, Kirkland JL, Campisi J, van Deursen JM. Senescence and apoptosis: dueling or complementary cell fates? EMBO Rep [Internet]. 2014; 15: 1139–53. doi: 10.15252/embr.201439245.

253. Borah A, Palaninathan V, Girija AR, Balasubramanian S, Rochani AK. Poly- lactic-co-glycolic acid Nanoformulation of Small Molecule Antagonist GANT61 for Cancer Annihilation. 2017; : 1–10.

254. Chenna V, Hu C, Pramanik D, Aftab BT, Karikari C, Campbell NR, Hong S-M, Zhao M, Rudek M a., Khan SR, Rudin CM, Maitra a. A Polymeric Nanoparticle Encapsulated Small-Molecule Inhibitor of Hedgehog Signaling (NanoHHI) Bypasses Secondary Mutational Resistance to Smoothened Antagonists. Mol Cancer Ther. 2012; 11: 165–73. doi: 10.1158/1535- 7163.MCT-11-0341.

255. Infante P, Alfonsi R, Botta B, Mori M, Di Marcotullio L. Targeting GLI factors to inhibit the Hedgehog pathway. Trends Pharmacol Sci [Internet]. Elsevier Ltd; 2015; 36: 547–58. doi: 10.1016/j.tips.2015.05.006.

256. Sun Y, Guo W, Ren T, Liang W, Zhou W, Lu Q, Jiao G, Yan T. Gli1 inhibition suppressed cell growth and cell cycle progression and induced apoptosis as well as autophagy depending on ERK1/2 activity in human chondrosarcoma cells.

179 Cell Death Dis [Internet]. 2014 [cited 2014 Jul 15]; 5: e979. doi: 10.1038/cddis.2013.497.

257. Bermudez O, Hennen E, Koch I, Lindner M, Eickelberg O. Gli1 mediates lung cancer cell proliferation and Sonic Hedgehog-dependent mesenchymal cell activation. PLoS One [Internet]. 2013 [cited 2014 Jul 15]; 8: e63226. doi: 10.1371/journal.pone.0063226.

258. Stylli SS, Luwor RB, Ware TMB, Tan F, Kaye AH. Mouse models of glioma. J Clin Neurosci [Internet]. Elsevier Ltd; 2015; 22: 619–26. doi: 10.1016/j.jocn.2014.10.013.

259. Joost S, Almada LL, Rohnalter V, Holz PS, Vrabel AM, Fernandez-Barrena MG, McWilliams RR, Krause M, Fernandez-Zapico ME, Lauth M. GLI1 inhibition promotes epithelial-to-mesenchymal transition in pancreatic cancer cells. Cancer Res. 2012; 72: 88–99. doi: 10.1158/0008-5472.CAN-10-4621.

260. Lei J, Ma J, Ma Q, Li X, Liu H, Xu Q, Duan W, Sun Q, Xu J, Wu Z, Wu E. Hedgehog signaling regulates hypoxia induced epithelial to mesenchymal transition and invasion in pancreatic cancer cells via a ligand-independent manner. Mol Cancer [Internet]. Molecular Cancer; 2013 [cited 2014 Jul 15]; 12: 66. doi: 10.1186/1476-4598-12-66.

261. Moghimi SM, Symonds P, Murray JC, Hunter AC, Debska G, Szewczyk A. A two-stage poly(ethylenimine)-mediated cytotoxicity: Implications for gene transfer/therapy. Mol Ther. 2005; 11: 990–5. doi: 10.1016/j.ymthe.2005.02.010.

262. Huh SH, Do HJ, Lim HY, Kim DK, Choi SJ, Song H, Kim NH, Park JK, Chang WK, Chung HM, Kim JH. Optimization of 25 kDa linear polyethylenimine for efficient gene delivery. Biologicals. 2007; 35: 165–71. doi: 10.1016/j.biologicals.2006.08.004.

263. Khansarizadeh M, Mokhtarzadeh A, Rashedinia M, Taghdisi S, Lari P, Abnous K, Ramezani M. Identification of possible cytotoxicity mechanism of polyethylenimine by proteomics analysis. Hum Exp Toxicol [Internet]. 2016; 35: 377–87. doi: 10.1177/0960327115591371.

264. Asher R, Bignami a. Hyaluronate binding and CD44 expression in human glioblastoma cells and astrocytes. Exp Cell Res. 1992; 203: 80–90. doi: 0014- 4827(92)90042-7 [pii].

265. Hayward SL, Wilson CL, Kidambi S. Hyaluronic acid-conjugated liposome nanoparticles for targeted delivery to CD44 overexpressing glioblastoma cells. Oncotarget. 2016; 7. doi: 10.18632/oncotarget.8926.

180 266. Gao X, Qian J, Zheng S, Xiong Y, Man J, Cao B, Wang L, Ju S, Li C. Up- regulating blood brain barrier permeability of nanoparticles via multivalent effect. Pharm Res. 2013; 30: 2538–48. doi: 10.1007/s11095-013-1004-9.

267. Chang J, Paillard A, Passirani C, Morille M, Benoit J, Betbeder D, Garcion E. Transferrin adsorption onto PLGA nanoparticles governs their interaction with biological systems from blood circulation to brain cancer cells. Pharm Res. 2012; 29: 1495–505.

268. Yan F, Wang Y, He S, Ku S, Gu W, Ye L. Transferrin-conjugated, fluorescein- loaded magnetic nanoparticles for targeted delivery across the blood-brain barrier. J Mater Sci Mater Med. 2013; 24: 2371–9.

269. Goyal R, Kapadia CH, R MJ, Riley RS, Day ES. Layer-by-Layer Assembled Gold Nanoshells for the Intracellular Delivery of miR-34a. Cell Mol Bioeng. 2018; : 1–14. doi: 10.1007/s12195-018-0535-x.

270. Choi J-H, Kim S-O, Linardy E, Dreaden EC, Zhdanov VP, Hammond PT, Cho N-J. Influence of pH and Surface Chemistry on Poly( L -lysine) Adsorption onto Solid Supports Investigated by Quartz Crystal Microbalance with Dissipation Monitoring. J Phys Chem B. 2015; 119: 10554–65. doi: 10.1021/acs.jpcb.5b01553.

271. Deng ZJ, Morton SW, Ben-Akiva E, Dreaden EC, Shopsowitz KE, Hammond PT, Kevin E, Shopsowitz KE, Hammond PT. Layer-by-layer nanoparticles for systemic codelivery of an anticancer drug and siRNA for potential triple- negative breast cancer treatment. ACS Nano [Internet]. 2013; 7: 9571–84. doi: 10.1021/nn4047925.

272. Poon Z, Lee JB, Morton SW, Hammond PT. Controlling in vivo stability and biodistribution in electrostatically assembled nanoparticles for systemic delivery. Nano Lett. 2011; 11: 2096–103. doi: 10.1021/nl200636r.

181 Appendix

JOURNAL PERMISSIONS

A.1 Permission from Oncotarget This permission covers the research presented in Chapter 3. Open-Access License No Permission Required

Oncotarget applies the Creative Commons Attribution 3.0 License (CC BY 3.0) to all works we publish (read the human-readable summary or the full license legal code). Under the CC BY, authors retain ownership of the copyright for their article, but authors allow anyone to download, reuse, reprint, modify, distribute, and/or copy articles in Oncotarget, so long as the original authors and source are cited. No permission is required from the authors or the publishers. In most cases, appropriate attribution can be provided by simply citing the original article. If the item you plan to reuse is not part of a published article (e.g., a featured issue image), then please indicate the originator of the work, and the volume, issue, and date of the journal in which the item appeared. For any reuse or redistribution of a work, you must also make clear the license terms under which the work was published. This broad license was developed to facilitate open access to, and

182 free use of, original works of all types. Applying this standard license to your own work will ensure your right to make your work freely and openly available.

A.2 Permission from Elsevier (Molecular Therapy-Nucleic Acids) This permission covers the research presented in Chapter 4. Written permission may not need to be obtained in certain circumstances, such as the following:

Public domain works are not protected by copyright and may be reproduced without permission, subject to proper acknowledgement. This includes works for which copyright has expired (for example, any US work published prior to 1923), works that are not copyrightable by law (for example, works prepared by US government employees as part of their official duties), and works expressly released into the public domain by their creators. (Permission would however be required to reuse the final formatted, edited, published version of a public domain journal article, for example, as this version is owned by the publisher.)

Open access content published under a CC-BY user license, as well as open access content published under other types of user licenses depending on the nature of your proposed re-use (for example, commercial vs. nonprofit use), may not require written permission, subject to proper acknowledgement. Permissions vary depending on the license type, and we recommend that readers check the license details carefully before re-using the material.

Creating an original figure or table from data or factual information that was not previously in figure or table format typically does not require permission, subject to proper acknowledgement of the source(s) of the data.

183 A.3 Permission from Molecular Pharmaceutics This permission covers the research described in Chapter 5. Reuse/Republication of the Entire Work in Theses or Collections: Authors may reuse all or part of the Submitted, Accepted or Published Work in a thesis or dissertation that the Author writes and is required to submit to satisfy the criteria of degree-granting institutions. Such reuse is permitted subject to the ACS’ “Ethical Guidelines to Publication of Chemical Research.”

184