The Pennsylvania State University
The Graduate School
Department of Pharmacology
ALTERATIONS IN GENE EXPRESSION AS A RESPONSE
OF TUMOR CELLS TO STRESSES
A Dissertation in
Pharmacology
by
Kathryn Joyce Huber-Keener
2012 Kathryn Joyce Huber-Keener
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
May 2013
The dissertation of Kathryn Joyce Huber-Keener was reviewed and approved* by the following:
Jin-Ming Yang Professor of Pharmacology Dissertation Advisor Chair of Committee
Willard M. Freeman Associate Professor of Pharmacology
Rongling Wu Professor of Statistics
Robert G. Levenson Professor of Pharmacology
Andrea Manni Professor of Medicine
Jong K. Yun Director of the Pharmacology Graduate Program
*Signatures are on file in the Graduate School
ii
ABSTRACT
Solid cancers are the 2nd leading cause of death in adults in the United States.
Understanding the mechanisms by which cancer cells survive under stress is pivotal to decreasing this statistic. Cancer cells are constantly under stress, a factor which needs to be taken into consideration during the treatment of solid cancers. Both intrinsic and extrinsic stresses impact the development and progression of neoplasms, down to the level of individual proteins and genes. Stresses like nutrient deficiency, hypoxia, acidity, and the immune response are present during normal tumor growth and throughout treatment. Tumor cells that survive these stresses are more adept at surviving hostile conditions and are more resistant to current therapies. Stress, therefore, shapes the tumor cell population. Gene expression alterations are the driving feature behind the adaptive ability of cancer cells to these stresses, and therefore, careful examination of these gene expression changes must be undertaken in order to develop effective therapies for solid cancers.
In the present investigations, we first explore the global alterations of gene expression in tamoxifen resistance. Resistance to tamoxifen (Tam), a widely used antagonist of the estrogen receptor (ER), is a common obstacle to successful breast cancer treatment. While adjuvant therapy with Tam has been shown to significantly decrease the rate of disease recurrence and mortality, recurrent disease occurs in one third of patients treated with Tam within 5 years of therapy. A better understanding of gene
iii
expression alterations associated with Tam resistance will facilitate circumventing this problem. Using a next generation sequencing approach and a new bioinformatics model, we compared the transcriptomes of Tam-sensitive and Tam-resistant breast cancer cells for identification of genes involved in the development of Tam resistance. We identified differential expression of 1215 mRNA and 513 small RNA transcripts clustered into ERα functions, cell cycle regulation, transcription/translation, and mitochondrial dysfunction.
The extent of alterations found at multiple levels of gene regulation highlights the ability of the Tam-resistant cells to modulate global gene expression. Alterations of small nucleolar RNA, oxidative phosphorylation, and proliferation processes in Tam-resistant cells present areas for diagnostic and therapeutic tool development for combating resistance to this anti-estrogen agent.
After such a global exploration of cancer cell responses to stress, we next investigated a mechanism of cancer cell survival by exploring the alterations in protein synthesis inhibitor eEF-2K. Studies show that EF-2K plays a role in cell survival through this inhibition of protein synthesis and that its protein levels are increased in cancer.
Post-translational modification of translation machinery is important for its regulation and could be critical for survival of cancer cells encountering stress. Thus, the purpose of our study is to examine the regulation of EF-2K during stress with a focus on the phosphorylation status and stability of EF-2K protein in cancer cells. Using two human glioma cell lines (T98G and LN229), we have found a 2-5 fold increase in EF-2K expression and activity under stress conditions of nutrient deprivation and hypoxia. mRNA levels are only transiently increased and shortly return to normal, while EF-2K
iv
protein levels continue to increase after further exposure to stress. This result could be explained by decreased turnover of EF-2K protein, which has a normal half-life of ~ 6-8 hours in glioma cells, so cycloheximide experiments were used to examine the effect of stress on EF-2K protein stability. A seemingly paradoxical decrease in EF-2K stability
(t1/2 = 2-4h) was found when glioma cells were subjected to stress despite increased protein expression. Phosphorylation may play a role in this altered protein stability as
EF-2K has multiple phosphorylation sites that are phosphorylated by the mTOR/S6 kinase (Ser78 and Ser366) and AMP kinase (Ser398), pathways which would be affected by stress. Therefore, phosphorylation-defective mutants of EF-2K were made to examine the effect of phosphorylation at these sites on EF-2K protein stability. We discovered that the AMP kinase site was pivotal to protein stability as the S398A mutant half-life increased to greater than 24 hours under both normal and stress conditions. Mutating the mTOR pathway sites made EF-2K protein more stable under normal conditions (t1/2 >
24h) but decreased to normal levels under stress conditions (t1/2 = 8h). Inhibiting the mTOR pathway with rapamycin treatment increased protein expression ~ 5 fold and increased EF-2K stability in these mutants under all culture conditions. These data indicate that EF-2K is regulated at multiple levels with phosphorylation playing an important role in protein turnover. The unexpected decrease in EF-2K protein stability during stress may be a compensatory mechanism for an additional level of regulation at the post-transcriptional level that increases EF-2K translation. Due to the importance of translation regulation during stress, it is reasonable to have increased translation of a regulator of protein synthesis while decreasing the same protein’s stability in order to
v
quickly adapt to changing nutrient levels. Further studies will examine the post- transcriptional regulation of EF-2K during stress as these data demonstrate its complex and tight regulation. Understanding the regulation of EF-2K could lead to therapeutics targeting EF-2K that could potentially render cancer cells intolerant to stress and susceptible to current treatments.
Together, our results represent the global impact that stress can have on cancer cells. Our NGS study exemplifies the complexity of alterations caused by exogenous stresses on cancer cells. The wide variety of gene expression changes from genes involved in proliferation and survival to those that regulate energy metabolism and post- transcriptional indicate that stress selection can alter the overall functioning of the cancer cell. The focused eEF-2K study indicates how tightly and complexly the protein synthesis regulator is controlled, which has broad implications for global and specific translation of the numerous gene transcript alterations caused by stress. Understanding how gene expression and proteins are altered in cancer cells by different intrinsic and extrinsic stresses will help current and future researchers to develop novel and more effective methods of overcoming cancer cell survival and resistance to current treatments.
vi
TABLE OF CONTENTS
ABSTRACT ...... iii LIST OF FIGURES ...... xi LIST OF TABLES ...... xii LIST OF ABBREVIATIONS ...... xiii ACKNOWLEDGEMENTS ...... xvi EPIGRAPH ...... xx Chapter 1 ...... 1 Cellular stress, gene expression, and cancer development and progression ...... 1
1.1. Overview of Solid Cancers ...... 1 1.1.1. Breast Cancer ...... 3 1.1.1.1. Breast cancer statistics 3 1.1.1.2. Breast cancer progression 4 1.1.1.3. Breast cancer etiology 6 1.1.2. Glioma ...... 6 1.1.2.1. Glioma statistics 9 1.1.2.2. Glioma progression 9 1.1.2.3. Glioma etiology 10 1.2. Cellular stress ...... 11 1.2.1. Intrinsic stresses during tumor development and progression ...... 13 1.2.2. Extrinsic stress caused by cancer treatments ...... 18 1.2.2.1. Surgery 19 1.2.2.2. Radiation 19 1.2.2.1. Chemotherapy 20 1.2.2.3. Targeted therapies 22 1.2.3. Molecular markers of stress ...... 23 1.3. Common mechanisms of cancer therapy resistance ...... 28 1.3.1. Multi-drug resistance ...... 29 1.3.2. Cancer stem cells ...... 29 1.3.3. Autophagy ...... 30 1.3.4. Modifications in signaling pathways and gene expression ...... 31 1.3.4. Alterations in gene expression ...... 33 1.4. Regulation of gene expression ...... 34 1.4.1. Transcriptional control ...... 34 1.4.2. Regulation of translation ...... 38 1.4.3. Post-transcriptional and post-translational modifications ...... 41 1.4.3.1. Covalent additions 41 1.4.3.2. Cleavage reactions 42 1.5. Use of next generation sequencing in gene expression studies ...... 43 1.5.1. Sequencing Process ...... 44 1.5.2. Platforms ...... 45 1.5.2.1. Illumina HiSeq and Genome Analyzer 46 1.5.2.2. Roche 454 Pyrosequencing 47 1.5.2.3. Helicos Heliscope: 47 1.5.2.4. Applied Biosystems SOLiD Sequencing: 47 1.5.2.5. Life Technologies Ion Torrent Sequencing 48 vii
1.5.3. Advantages and disadvantages of NGS ...... 48 1.5.4. Uses in gene expression studies ...... 51 1.5.4.1. Chromosomal rearrangements 51 1.5.4.2. Transcriptomes, exomes, and gene signatures 52 1.5.4.3. Epigenome 53 1.5.4.3. Binding sites for transcription factors 54 1.6. Significance of my research project ...... 54
Chapter 2 Differential gene expression in tamoxifen-resistant breast cancer cells revealed by a new analytical model of RNA-Seq data ...... 56
2.1. Abstract ...... 56 2.2. Introduction ...... 57 2.2.1. Estrogen and Estrogen Receptor ...... 57 2.2.1.1. ER structure 58 2.2.1.2. Regulation and Activity of ER 60 2.2.2. Estrogen receptor in breast cancer ...... 61 2.2.3. Endocrine therapy in breast cancer ...... 62 2.2.3.1. Aromatase Inhibitors (AIs) 62 2.2.3.2. Selective estrogen receptor down-regulators (SERDs) 62 2.2.3.3. Selective estrogen receptor modulators (SERMs) 63 2.2.3. Tam resistance in breast cancer ...... 64 2.2.5. Technology used for our study of Tam resistance...... 67 2.3. Rationale ...... 69 2.4. Experimental design ...... 70 2.4.1. Cell lines and reagents ...... 70 2.4.2. RNA preparation ...... 71 2.4.3. Library preparation for SOLiD™ NGS sequencing ...... 71 2.4.4. Library preparation for small RNA sequencing ...... 72 2.4.5. Sequencing ...... 73 2.4.6. NGS mapping and expression ...... 73 2.4.7. qRT-PCR validation ...... 74 2.4.8. Statistical models ...... 74 2.4.9. Expression Analysis ...... 76 2.4.10. SIRT3 expression and cell growth assays ...... 77 2.5. Results and Discussion ...... 77 2.5.1. Clustering of gene expression data ...... 77 2.5.1.1. Rationale behind mathematical model 78 2.5.1.2. Validation and comparison of gene expression levels between Tam- sensitive and Tam-resistant breast cancer cells. 79 2.5.1.3. Phenotypic plasticity clustering analysis. 85 2.5.1.4. Effects of Tam resistance on smRNA expression and clustering. 90 2.5.1.5. Gene ontology and clustering analysis of mRNA expression. 92 2.5.2. Comparison to traditional analysis methods and previous studies ...... 97 2.5.3. Effects of SIRT3 expression on Tam resistance ...... 102 2.6. Conclusions ...... 104
Chapter 3 ...... 107 viii
Phosphorylation of elongation factor-2 kinase and the stability of the enzyme under various stress conditions ...... 107
3.1. Abstract ...... 107 3.2. Introduction ...... 109 3.2.1. Eukaryotic elongation factor-2 kinase (eEF-2K) ...... 109 3.2.1.1. eEF-2K structure 111 3.2.1.2. Regulation of protein synthesis by eEF-2K 112 3.2.1.3 Regulation of eEF-2K 112 3.2.1.3.1. Calcium/Calmodulin and autophosphorylation 112 3.2.1.3.2. cAMP-dependent protein kinase (PKA) regulation of eEF-2K 114 3.2.1.3.3. mammalian target of rapamycin (mTOR) regulation of eEF-2K 115 3.2.1 3.4. Adenosine monophosphate-activated protein kinase (AMPK) pathway regulation of eEF-2K 117 3.2.1.3.5. Multiple stress response pathways regulation of eEF-2K 118 3.2.2. Stability of eEF-2K protein ...... 121 3.2.3. eEF-2K expression ...... 123 3.2.3.1. eEF-2K expression in cancer 123 3.2.3.2. Correlation with stress and cellular energy 124 3.3. Rationale ...... 126 3.4. Experimental design ...... 127 3.4.1. Cell lines and culture...... 127 3.4.2. Reagents and antibodies ...... 128 3.4.3. Stress Conditions ...... 128 3.4.4. Real time RT-PCR ...... 128 3.4.5. EF-2K phosphorylation-defective mutants ...... 129 3.4.6. Preparation of cellular extracts and Western blot analysis ...... 129 3.4.7. Mining of mRNA functional elements ...... 130 3.5. Results ...... 130 3.5.1. eEF-2K levels are increased by metabolic stress in glioma cells ...... 130 3.5.2. eEF-2K protein turnover is increased by metabolic stress ...... 132 3.5.4. Phosphorylation sites differentially regulate eEF-2K turnover ...... 134 3.5.5. Effects of inhibition of upstream signaling cascades on eEF-2K stability ..... 135 3.5.6. Determination of RNA elements important for translation of eEF-2K ...... 138 3.6. Discussion ...... 139
Chapter 4: Discussion of the current Studies; ...... 146 Altered signaling and gene expression in cancer cells under stress ...... 146
4.1. Preamble...... 146 4.2. Tam resistance in breast cancer ...... 147 4.2.1. Clinical Studies ...... 147 4.2.2. Potential biomarkers of Tam resistance ...... 149 4.2.3. Current findings: major gene expression changes found by next generation sequencing ...... 150 4.2.3.1. The importance of analytical methods in determining differential gene expression. 151 4.2.3.2. Important ontological groups altered by Tam resistance 152 ix
4.2.3.3. Importance of small RNA in Tam resistance 153 4.2.5. Directions for future studies ...... 155 4.2.5.1. Moving from preclinical to clinical studies: the use of NGS in Tam resistant breast cancer tumor samples 157 4.2.5.2. SIRT3 findings and future experiments 159 4.2.5.3. Mitochondria and energy metabolism 161 4.3. Regulation of eEF-2K in response to metabolic stress ...... 162 4.3.1. Protein synthesis in cancer cells ...... 163 4.3.2. Metabolic stress effects on protein phosphorylation ...... 163 4.3.3. Current findings: Phosphorylation at specific sites of eEF-2K differentially modulates its turnover ...... 164 4.3.3.1. Cellular stress alters eEF-2K protein stability 165 4.3.3.2. Identification of upstream signaling pathways and phosphorylation sites affecting eEF-2K protein turnover 165 4.3.4. Directions for further studies ...... 167 4.3.4.1. Identification of additional upstream pathways that phosphorylate eEF-2K under stressful conditions 168 4.3.4.2. Examination of discordant mRNA and protein levels 169 4.3.4.3. Effect of eEF-2K phosphorylation on specific versus global protein synthesis 170 4.4. Epilogue: The importance of stress response in cancer research ...... 171 REFERENCES...... 174
x
LIST OF FIGURES
Figure 1.1 Breast cancer progression …………………………………………………...5 Figure 1.2 Glioma progression ………………………………………………………….8 Figure 1.3 Typical stresses encountered by cancer cells during tumor growth and progression ……………………………………………………………………………...12 Figure 2.1 Estrogen receptor signaling ……………………………………………...….59 Figure 2.2 NGS identification and comparison of differentially-expressed genes in TamR cells by the Fisher’s exact test …………………………………………………………..79 Figure 2.3 Comparison and validation of differentially-regulated genes by the two significance methods ……………………………………………………………………82 Figure 2.4 Clustering patterns of genes by absolute difference and ratio of expression ……………………………………………………………………………….86 Figure 2.5 Heatmap comparison of differentially-expressed genes by clustering analysis …………………………………………………………………………….…..100 Figure 2.6 Dysregulation of pathways and processes involved in Tam resistance as revealed by NGS …………………………………….. Figure 2.7 Effect of SIRT3 expression on Tam resistance ……………………….…...102 Figure 3.1 Proposed structure of eEF-2K ……………………………………………..109 Figure 3.2 Cellular stresses increase EF-2K expression which is not fully accounted for in mRNA levels ………………………………………………………………………...130 Figure 3.3 Cellular stress decreases eEF-2K stability …………………………………132 Figure 3.4 Phosphorylation sites on EF-2K differentially affect its turnover …………134 Figure 3.5 Inhibition of the mTOR pathway with rapamycin increases EF-2K stability regardless of stress ……………………………………………………………………..135 Figure 3.6 Inhibition of AMP kinase pathway with compound C decreases EF-2K stability under both stressful and non-stressful conditions …………………………….136 Figure 3.7 eEF-2K phosphorylation sites ……………………………………………..139
xi
LIST OF TABLES
Table 1.1 Stresses caused by glioma and breast cancer therapies ……………………18 Table 1.2 The minimal stress proteome as adapted from source ……………………24 Table 1.3 Drug resistance mechanisms ……………………………………………….28 Table 1.4 Comparison of NGS platforms …………………………………………….46 Table 2.1 Significant genes found by both methods …………………………………81 Table 2.2 mRNA exon gene clusters ………………………………………………….92
xii
LIST OF ABBREVIATIONS
ABC ATP binding cassette ACS American Cancer Society AF-1/2 Activation function 1 or 2 AI Aromatase inhibitor AIC Akaike information criterion AMPK AMP kinase APL Acute promyelocytic leukemia AMP/ATP Adenosine mono- or tri-phosphate AUBP AU binding protein bp Base pairs BCNU Carmustine BIC Bayesian information criterion CaM Calmodulin CaMK Calmodulin kinase CBP CREB binding protein CDK Cyclin dependent kinase ChIP-Seq Chromatin immunoprecipitation next generation sequencing CHX Cycloheximide CML Chronic Myeloid Leukemia CNS Central nervous system CREB cAMP response element-binding protein 1 CSC Cancer stem cell CSF Cebrospinal fluid CSR Conserved stress response CSS Charcoal stripped serum DBD DNA-binding domain DCIS Ductal carcinoma in situe ddNTP Dideoxynucletodie DNA Deoxyribonucleic acid DPE Downstream core promoter element E1/2/3/4 Estrone, estradiol, estriol, and estetrol ECM Extracellulr matrix eEF-2 Eukaryotic elongation factor-2 eEF-2K Eukaryotic elongation factor-2 kinase eIF Eukaryotic initiation factor EGF Epidermal growth factor EGFR Epidermal growth factor receptor EM Expectation maximization ER Estrogen receptor or Endoplasmic reticulum
xiii
ERE Estrogen response element FAK Focal adhesion kinase FDA Food and Drug Administration FET Fisher’s exact test FISH Fluourescent in situ hybridization FPKM Fragments per kilobase of exon per million fragments mapped GBM Glioblastoma multiforme GFAP Glial fibrillary acidic protein GRP Glucose related protein HER-2/neu/ErbB2 Human epidermal growth factor 2/Neu HERT Hormone estrogen replacement therapy HIF-1 Hypoxia-inducible factor-1 HSP Heat shock protein IGF-1 Iinsulin growth factor 1 IGF-1R Insulin growth factor 1 receptor IPA Ingenuity pathway analysis IR Ionizing radiation IRES Internal ribosome entry site LCIS Lobular carcinoma in situ MAPK Mitogen-activated protein kinase MDR Multidrug resistance MHCK Myosin heavy chain kinases miRNA MicroRNA mRNA Messenger RNA mTOR Mammalian target of rapamycin NAD/NADH Nicotinamide adenine dinucleotide NCI National Cancer Institute ncRNA Non-coding RNA NF-κB Nuclear factor-KappaB NGS Next generation sequencing NRSF Neuron -restrictive silencer factor ORF Open reading frame PCR Polymerase chain reaction PGC-1 PPAR gamma coactivator 1 PDGFR Platelet-derived growth factor receptor PI3K Phosphoinositide 3-kinase PKA Protein kinase A PNS Peripheral nervous system PR Progesterone receptor PTEN Phosphatase and tensin homolog RB Retionblastoma protein RNA Ribonucleic acid RNA-Seq RNA sequencing by next generation sequencing ROS Reactive oxygen species xiv
rRNA Ribosomal RNA S398A Serine to Alanine phosphorylation-defective mutant at Ser398 on eEF-2K S78/366A Serine to Alanine phosphorylation-defective mutant at Ser78 and Ser398 on eEF-2K SAPK/JNK Stress-activated protein kinase/c-Jun NH2-terminal kinase SEER Surveillance, Epidemiology and End Results Ser Serine SERD Selective estrogen receptor degrader SERM Selective estrogen receptor modulator shRNA Short hairpin RNA siRNA Small intefering RNA SIRT3 Sirtuin 3 SLR SELI-1 like repeats smRNA Small RNA snoRNA Small nucleolar RNA snoRNP Small nucleolar RNA protein SNP Single nucleotide polymorphism snRNA Small nuclear RNA SRE Serum response element Tam Tamoxifen TamR Tamoxifen resistant cell line MTR-3 TamS Tamoxifen sensitive cell line MCF-7 TBP TATA-binding protein TMZ Temezolomide TOP 5’ terminal oligopryimidine tract TRAIL TNF-related apoptosis-inducing ligand tRNA Transfer RNA uORF Upstream open reading frame UPR Unfolded protein response UTR Untranslated region VEGF Vascular endothelial growth factor WT Whole transcriptome
xv
ACKNOWLEDGEMENTS
Lab: I would like to thank the people directly responsible for helping and supporting me through the research presented in this thesis. My lab has provided a wonderful and collaborative atmosphere, which is the result of the exemplary mentorship we receive from my advisor, Dr. Jin-
Ming Yang. Dr. Yang, you are truly worthy of the title of mentor. You live by the motto that our success is your success. You have helped me to grow as an independent scientist, allowing me freedom to explore my hypotheses while guiding me when I had the inevitable troubles that come along with biomedical research. You have provided me with opportunities that I might not have experienced elsewhere, which have been vital to my career development. I only hope that one day I can use the tools that I learned from your example and guidance to mentor my own students half as well as you mentored me. I am eternally grateful to you. Dr. Xingcong Ren, you helped me from day one to get my experiments running, and you have been a constant support to me throughout my research, both personally and experimentally. Dr. Yan Cheng, Dr. Yi Zhang, and
Yu Shan, I have enjoyed getting to know you all, and you have made my time so much more enjoyable in the lab. I am so glad that we were here together to bounce ideas off one another and to support each other as we muddled through the hardships of research and triumphed in one another’s success.
Collaborators: To all those who have ever helped me in my endeavors for this thesis, I thank you. Dr. Rongling Wu and Dr. Zhong Wang, I cannot thank you and your team enough for helping to get our NGS study off the ground. My NGS paper would have gotten nowhere without the development of the NGS pipeline and clustering method. Our collaboration taught me so much about bioinformatics and has been a lesson in teamwork. Dr. Bill Freeman, thank you so much for taking so much time to mentor me in genomics and proteomics. The significance of our xvi
NGS results would have been rubbish without someone to help me learn the tools to analyze the biological relevance of them, so I truly appreciated and enjoyed the opportunity to work with you.
Drs. Maricarmen Silva-Planas, Song Wu, Kent Vrana, Xiuping Liu, and Chang-Gong Liu, thank you for your collaborative efforts on the NGS manuscripts. Our work together has been a prime example of what scientists can accomplish when they work together.
Penn State and the MD/PhD and Pharmacology Programs: I have great appreciation for these two programs for supporting me throughout my years at Penn State. Special thanks to Dr.
Bob Levenson, Dr. Diane Thiboutot, and Barb Koch in the MD/PhD program for all that they have done for me. I wouldn’t have made it without your support and encouragement. You all have been integral to my success. Special thanks also to Dr. Kent Vrana, Dr. Jong Yun, and
Elaine Neidigh for assisting and supporting me through the various successes and challenges of graduate school. Added thanks go to Kathy Simon in the Graduate Office for dealing with all the paperwork problems that occur as a MD/PhD. To Drs. Michael Katzman, George Blackall, and
Kevin Grigsby and to Penn State in general, thank you for all that you have done and all the support.
Committee: To all my committee members, including Dr. Andrea Manni, I thank you for taking the time out of your schedules to be on my committee and to take a vested interest in my education. Your great questions both helped to expand my background knowledge and to focus my expertise.
Friends: I have had the pleasure and honor of being surrounded by great friends during my graduate career at Penn State. Nicole, you are both family and friend, and I consider it serendipitous that we were housed together starting my first year. I cannot thank you enough for all the support you’ve given me both as a friend and as a graduate student ahead of me in school.
Arati and Melissa, my interactions with you as a young graduate student helped me to develop as
xvii
the scientist I am today. Thank you both for your guidance and friendship throughout these years.
Melanie, you are my MD/PhD homie (Dan can’t have you), and I consider myself so lucky to you in my life. Carolina, Vance, Dan, Theresa, and James, I am eternally grateful we started this journey together even if we’ve ended it at different times. Amanda, you have been my med school mentor and friend, and I am so thankful for your support. Brian, Sarah, Shannon, Darren,
Shorena, Mark, Sarah A.C., Brad, Lindsay, Terri, Alex, Kat, Pat, and Sue, you have made life in grad school so much more fun with game nights, book club, and memorable Halloween parties.
Kristin, Su-Fern, James, Shang-Min, Becky, Lauren, James, Meghan, Raghu, SubbaRao, Sung-
Jin, Keen, Tom, Chris, Jamal, Brian, Tony, and Marie, thank you for making coming to work so much more enjoyable every day. To all those not mentioned specifically, I thank you for your friendship, especially those that knew me before this journey ever started.
Family: Last, but certainly not least, I would like to thank my family. There aren’t thanks enough to show my parents, Jim and Marion, how grateful I am for their role in my life both before and during graduate school. You have both been examples of people I wish to be, and you have my eternal thanks for giving me the freedom to make my own choices in life. Your encouragement of my curiosity has helped me to develop into the scientist I am today, and your constant love and support has strengthened me. To my sister Jeni and her husband Doug, thank you for being my East Coast refuge, and I am so glad I made the decision to move closer to you.
A huge thank you goes out to my in-laws, Mel and Betty, for their constant support; I am so happy to call you family. To all my extended family, thank you so much for your constant prayers and support. Your belief in me helped me to get through it. And my biggest thanks go to the family I chose for myself, my husband Jay; I know without a shadow of a doubt that I made the right choice to make you family. You have been my pillar of strength, my fun escape, my
xviii
savior in Word formatting, and the love of my life. Thank you so much for your unwavering love and encouragement.
xix
EPIGRAPH
"I've missed more than 9,000 shots in my career. I've lost almost 300 games. 26 times, I've been trusted to take the game winning shot and missed. I've failed over and over and over again in my
life. And that is why I succeed."
- spoken by basketball legend, Michael Jordan
xx
Chapter 1 Cellular stress, gene expression, and cancer development and progression
Tumor cells, both within solid tumors and during invasion, metastasis, or circulation, exist in surroundings that are subject to a variety of stresses, including metabolic and environmental stresses. Nonetheless, cancer cells survive and can even thrive under hostile conditions, such as hypoxia, nutrient deprivation, and therapeutic regimens. In order to survive, these tumor cells have to find a way to adapt to such an environment by activating certain growth factor and survival pathways while down- regulating cell death mechanisms. In fact, cancer cells adapt so well that they not only survive but proliferate by creating a more hospitable environment through new blood vessel formation and dissemination, even as they endure additional stresses along the way. Alteration of gene expression is a way that allows cancer cells to activate the necessary survival pathways and adapt to adverse condition.
1.1. Overview of Solid Cancers
As the second most common cause of death in the United States behind heart disease, cancer is an important public health concern. It is estimated that approximately
1,638,910 Americans will be diagnosed with cancer in 2012. Shockingly, this large
1
number does not include noninvasive (in situ) cancers or non-melanoma skin cancers. Of those diagnosed, 577,190 will die from their cancers in 2012 [1]. Although there are numerous specific types of malignancies, cancer can be divided into two main categories: solid and non-solid cancers. Non-solid cancers are often referred to as hematological cancers which include lymphoid, hematopoietic, and related tissue-cancers. Although there is both a high public awareness and a plethora of research dollars for hematologlical cancers, these cancers only account for 6.2% of cancer deaths [2]. Thus, approximately
530,000 deaths in the U.S. in 2012 will be due to solid cancers – a number greater than the sum from the next five most common causes of death [3]. The impact of solid cancers on Americans’ health cannot be overstated.
The most commonly diagnosed cancers in males are prostate (29%), lung (14%), and colorectal (9%) cancer, while in female they are breast (29%), lung (14%), and colorectal (9%) cancer. In cancer death statistics, lung cancer switches places with prostate and breast cancer to become the most common cause of cancer death in both males and female (accounting for 29% and 26% of deaths, respectively) [1].
Cancer can be considered a genetic disease in that it is the result of genetic damage. While faulty genes are sometimes inherited, accounting for about 5% of cancers, most often the genetic alterations are caused by other factors, both internal and external [1]. Changes in cellular metabolism and exposure to hormones can change gene expression. Even just aging increases cancer risk as genetic insults accumulate. The effect of aging is well illustrated by the fact 77% of all cancers are diagnosed in patients that are 55 or older. External or environmental factors also carry known risks for cancer,
2
such as exposure to toxins and UV rays [1]. All these risk factors result in the same event: alterations in gene expression that eventually lead to uncontrolled growth of cancer cells.
1.1.1. Breast Cancer
Breast cancer is the result of uncontrolled growth of one or more of its tissues.
The breast is made up of fatty and connective tissue; lymph tissue; and glands. The glands that produce milk are called lobules, and these lobules are connected to the nipple through ducts. In situ breast cancers are grouped into two categories: ductal carcinoma in situ (DCIS), which accounts for 83% of all in situ breast cancers, and lobular carcinoma in situ (LCIS), which describes roughly 11% of all in situ breast cancers. The rest of in situ breast cancers are considered hybrids of DCIS and LCIS or have origins that are unspecified [4]. Invasive breast cancers start in the ducts and/or lobules but eventually break through membrane barriers to infiltrate other tissues. Rarely, breast cancers develop from the connective tissue, which are then called sarcomas. These breast cancers include angiosarcomas of the blood vessels and phyllodes tumors of the periductal stromal cells. Other rare forms of breast cancer include Paget’s disease of the nipple and inflammatory breast cancer, both of which present with inflammation and itching.
1.1.1.1. Breast cancer statistics
Excluding non-melanoma skin cancers, breast cancer accounts for approximately one-third of the cancer diagnoses in women, making it the most common female
3
malignancy and the second leading cause of cancer death in women. According to the
Surveillance, Epidemiology, and End Results (SEER) report from the National Cancer
Institute (NCI), in 2011, approximately 230,480 U.S. women are estimated to have been diagnosed with breast cancer, and 39,520 women were expected to die of the disease [5].
Although it is more common in women, approximately 2,140 men were diagnosed with breast cancer, and 450 men died from it in 2011. The American Cancer Society (ACS) states that the lifetime chance of a woman in the US developing breast cancer in her lifetime is about 1 in 8, while men have a 1 in a 1,000 chance. Of those cancers diagnosed, 57,650 were considered in situ or non-invasive breast cancer, leaving over
170,000 cases of invasive breast cancer [4].
1.1.1.2. Breast cancer progression
The Figure 1.1 shows the general progression of breast cancers. As stated previously, breast cancer development starts with high grade dysplasia which is referred to as in situ (Greek for “in its place”). This abnormal growth of breast tissue cells can proliferate within the tissue but has yet to invade through the basement membrane. If left unattended, in situ breast cancers could develop into an invasive phenotype. As breast cancers progresses, additional genetic insults accumulate and the breast cancer cells are able to invade through basement membranes. Finally, activated stromal cells produce factors that aid cancer cells in metastasis.
4
5
1.1.1.3. Breast cancer etiology
Most breast cancers are not hereditary in nature. However, there are mutated genes that increase the chance of breast cancer, such as DNA repair genes Breast Cancer
1 and 2 (BRCA1 and BRCA2, respectively). If mutations are found in these genes or if a first-degree relative carries the mutation, a patient has a relative risk of being diagnosed with breast cancer greater than 4.0. As with other cancers, one of the major risk factors for breast cancer is aging. 95% of new breast cancer cases and deaths are in women over
40 years of age. Women have a relative risk greater than 4.0 if they are over 65. The other major risk factor for breast cancer is race. In the US, non-Hispanic white women have higher incidence rates of breast cancer than African American women, although African American women are more likely to die from the disease and be diagnosed before age 40 [4].
Risk factors that carry a moderate relative risk (>1 to 2) for developing breast cancer include situations that increase exposure to estrogen: early menarche (<12), late first pregnancy (>30) or no pregnancy, not breastfeeding a child, and late menopause (>55), and oral contraceptives and HERT therapy. Other moderate risks include alcohol use, obesity, first degree relative with breast cancer, and a history of other cancers [4].
1.1.2. Glioma
Gliomas arise from the glia (Greek for “glue”) or supportive tissue of the brain, which account for half the cells in the brain [6]. These non-neural cells form myelin and
6
support the neurons. Part of their supporting role is to supply nutrients and oxygen to oxygen neurons while destroying pathogens and removing dead neurons; they may also be involved in uncharacterized methods of modulating neurotransmission [7]. Several types of glia exist that fall under two categories: macroglia, which are derived from ectodermal tissue, and microglia, which are unique in that they are derived from hemopoietic stem cells. Microglia cells are specialized macrophages that make up approximately 15% of total brain cells. Macroglia cells include the abundant astrocytes which supply neurons with their blood supply, oligodendrocytes which form myelin, ependymal cells that produce CSF, radial glia that provide scaffolding, Schwann cells which myelinate PNS cells, satellite cells which connect ganglia in the PNS, and finally enteric glia which are found in the digestive system. Glial cells are different from neurons in that they retain the ability to undergo mitosis, thus there is the also the possibility for unrestrained cell division leading to gliomas.
Most gliomas are astrocytomas (from astrocytes), although ependymomas, oligodendrogliomas, and mixed types do exist. Astrocytomas are the most common.
They are classified on a low-grade (I or II) or high-grade system (III or IV), depending on whether they are slow-growing. If they develop into a high-grade glioma, these tumors are called glioblastoma multiforme (GBM), which is the most common and lethal form of glioma [8]. GBMs can be either primary, in that they arise de novo, or secondary, where they arise from transformation of lower grade astrocytoma
7
8
1.1.2.1. Glioma statistics
Primary brain cancers account for 2% of all primary tumors with approximately
22,910 new cases of primary brain and CNS tumors being diagnosed in 2011 in the U.S
[1]. Although the incidence of primary brain cancers is low, the mortality rate of glioma is high. Adult patients with GBM who undergo treatment still only have a median survival of only 12-18 months [9]; the 5-year survival rate is only 1% [10]. Brain cancers are the leading cause of solid cancer tumor death in children and make up 27% of child cancer diagnoses (2nd most common). Brain and nervous system cancers account for 2% of all female cancer deaths making it the 10th most common cause of cancer deaths in females [1].
1.1.2.2. Glioma progression
High grade gliomas can progress via two pathways: primary GBM and secondary
GBM (Fig. 1.2). In primary GBM, the original lesion is high-grade at its diagnosis and is considered to have arisen de novo. Primary GMB has a short clinical course with little time for physicians to intervene. Secondary GBM, however, can arise over the course of years. These tumors start as low-grade gliomas and, over time, progress into GBM [11].
A series of genetic insults is thought to play a role in the progression of low-grade astrocytomas into GBM. Large chromosome alterations occur, with loss of heterozygosity of chromosome 10 as the most frequent genetic event in GBM [12].
Tumor suppressor p53 is mutated in approximately 65% of secondary GBMs [13]. Some genetic modifications are found more often in primary GBM, like loss of PTEN [11] and 9
amplification of EGFR [13]. Both primary and secondary GBMs show alterations in cell cycle regulator p16 expression [14]. Histopathological progression from low-grade to high-grade glioma begins with differentiated cells that are diffusely infiltrating an area of the brain. As nuclear atypia and proliferation increase, so does the classification grade.
Final progression to GBM results in undifferentiated cells, areas of leaky blood vessels, necrosis, and continued proliferation of endothelial cells [15]. Diagnostic features of high-grade GBM include glial fibrillary acid protein (GFAP), pseudopalisading necrosis, and microvascular proliferation [16]. Because GBMs rarely metastasize outside of the brain due to the blood brain barrier, recurrences are local. Recurrences are common; up to 80% of patients, regardless of treatment type, will have a local recurrence within 3 cm of the original site [17].
1.1.2.3. Glioma etiology
Glioma can have a hereditary component as glioma has been linked to genetic syndromes like neurofibromatosis 1 and 2, tuberous sclerosis, retinoblastoma, Li-
Fraumeni syndrome, and Turcot’s syndrome. All of these genetic syndromes have genetic alterations that are common in glioma [18]. Additionally, the relative risk of developing GBM associated with having a primary family member diagnosed with GBM is similar to the familial association for relative risk of developing breast cancer [19]. On known environmental cause has been proved to be linked to glioma, and that is high-dose radiation [20]. This could explain the increase in incidence for glioma in aging population as this group of adults will have accumulated radiation exposure over a
10
lifetime. GBM arises most commonly in adults between the ages of 50-60 years old, although 20% of gliomas in children are high-grade as well [21].
1.2. Cellular stress
Internal stresses such as hypoxia, acidity, oxidative stress, and nutrient deprivation already exist within the cellular environment of tumors while external stressors like radiation treatment and genotoxic chemotherapy only worsen the internal factors. Encountering these stresses affects the process of carcinogenesis. Common markers of stress will be discussed along with their roles in induction of energy conservation and cell survival in cancer. Redistribution of energy resources towards survival pathways and away from energy-consuming processes is common.
Cellular stress can cause damage and mutations to numerous proteins, nucleic acid strands, and other macromolecules. The body has an innate reaction called the cellular stress response (CSR) to such damage. In the case of cancer, the tumor is able to highjack the body’s own machinery in order to help the cancerous cells survive usually by taking advantage of intrinsic or stress-related mutations. Thus, at times, stress may only further the growth and survival of tumor cells.
While the type of stress may vary, a common feature of many stresses is something referred to as the oxidative burst characterized by generation of oxidative stress and redox potential changes. Reactive oxygen species (ROSs) can be produced from activation of NADPH oxidase or other cellular oxidases in various membranes of 11
cells. Stressor-specific responses, on the other hand, may be induced differentially depending on the type, severity, and duration of the stress. The following sections will cover these stresses and the resulting responses by normal cells and cancerous cells.
Although, medical professionals, along with patients, are able to control to a certain extent the extrinsic stresses put on patients’ bodies and tumors such as treatments and environmental stressors (e.g. smoking), intrinsic stresses still naturally affect the tumor as it progresses.
Hypoxia Extra- Nutrient cellular Deprivation Signals
Growth pH Factor Imbalance Inhibitors Glioma Cell
Immune Radiation System
Chemo- ROS therapy
Figure 1.3: Typical stresses encountered by cancer cells during tumor growth and progression.
12
1.2.1. Intrinsic stresses during tumor development and progression
While research on cancer cells in laboratories is conducted primarily under nutrient-rich conditions, the micro-environments for cancer cells are actually quite hazardous. Rarely do tumor cells find themselves in conditions of perfect nutrient balance with necessary blood flow and comfortable living spaces. More often, tumor cells are constantly inundated with a barrage of stresses (Fig. 1.3). The internal stresses faced by different types of cancer are not usually unique to the type, but are actually shared by the majority of solid tumors types.
Environmental and metabolic stress occurs during tumor growth and progression.
As cancer cells divide, they take up more space. Normal cells would stop growing through contact inhibition, but malignant cells overcome the signals to inhibit growth and continue to divide. As the tumor expands, it continually outgrows its blood supply.
Tumors larger than 1 mm in diameter can no longer subsist on passive diffusion of nutrients [22]. The cancer cells, thus, go through periods of severe nutrient deprivation and hypoxia until enough tumor cells are able to signal new blood vessel formation or neoangiogenesis. Even after angiogenesis, cells are still subject to stress as the new vessels are prone to collapse due to their abnormal state and harsh surrounding conditions
[23]. During this time, cancer cells must adapt to survive in conditions and intermittent periods of limited amino acids, salts, and oxygen. Some researches indicate that this is when cancer cells start to rely on glycolysis, which continues even when oxygen is available, a phenomenon called the Warburg effect. This contributes to the high metabolic demand of proliferating tumor cells and is a relatively inefficient way of 13
producing energy [24]. Therefore, the cells are put under enormous stress just to keep up with energy production needs and are subjected to further metabolic stress when nutrients become unavailable.
The metabolic demands of the cancer cells are partially responsible for the increased acidity or pH imbalance found in many tumors. For instance, human brain tumors measured with electrodes had a mean pH of 6.8, with measurements as low as 5.9; the normal pH for the human brain is ~7.1 [25]. Such pH imbalance is even found in well vascularized areas of gliomas, thus indicating that tumor cells reside within a highly acidic environment even when oxygen is present. It was originally hypothesized that hypoxia caused the acid buildup, but these new findings mean that hypoxia and acidity are not always linked. The increased energy metabolism of the cancer cells produce hydrogen ions and metabolites like lactic acid and carbonic acid. All these products are actively pumped out of the cell through proton exchangers and other transporters [26]. In cases with decreased perfusion, poor circulation contributes to the buildup of an acidic extracellular environment.
As the tumor grows, the extracellular environment strives to slow down the progress of the cancer. Growth inhibition signals are sent that can either activate or deactivate cellular receptors depending on the need. Cancer cells survive by undergoing mutations in receptors like EGFR and PDGFR, changes that either stop the signaling cascades or rewire the signaling pathways to actually promote cancer cell growth. In this way, cellular proliferation is dissociated from nutrient availability by stress selection of surviving cells.
14
Hypoxia can play a major role in tumor development, with oxygen deprivation actually being necessary for tumor progression through alteration of gene expression, genomic instability, apoptotic dysregulation, and neoangiogenesis. In glioma, hypoxia is believed to be a key player due to the evidence of tumor necrosis in highly malignant forms like GBM [27]. Brain tumors smaller than the previously stated 1 mm cutoff are found to be highly hypoxic and ill-perfused [28]. In fact, the majority of solid tumors are hypoxic, with 40% of breast cancer tumors having oxygen readings that are below the minimal cut-off for radiation treatment [29]. The oxygen deprivation is actually responsible for the growth of elaborate microvascular networks that indicate tumor progression in solid tumors. Even though larger solid tumors, especially GBM, are more vascularized, the blood vessels present are inefficient, and parts of the tumor environment remain hypoxic [23]. Further transformation of the tumor cells occurs as reactive oxygen species (ROS) increase during this time due to production by the mitochondria [30].
Thus, hypoxia not only deprives cells of oxygen but leads to oxidative stress as well.
One of the hallmarks of a cancerous cell or tumor is its ability to invade through the basement membrane of one tissue into another type of tissue. The body has many stop guards in place to prevent this from happening, but somehow glioma cells overcome the challenge. Again, during this time, extracellular signals are sent to the cancer cells informing them to stop growing or to go through programmed cell death. Cell growth pathways are down-regulated by these signals causing severe stress to the cells. Without the normal nutrients or pathway activations, uncancerous cells would die, but tumor cells find a way to overcome the death signals. Cellular stress pathways that involve tumor
15
suppressor p53 and metabolic stress pathways which activate the apoptotic protein Bim are often deregulated in cancers [31]. Therefore, the neoplastic cells are able to overcome invasion preventions.
The immune system is an added stress to cancer cells during all times of tumor progression, but is especially active during invasion and dissemination. While the immune system may ignore some cancer cells that stay in their own tissue, cells from different tissues are recognized by the markers or antigens they display. The innate immune system encounters the cancer cells first, with first-response cells like macrophages, granulocytes, and mast cells attacking foreign cells displaying unknown or altered markers. Even cancer cells that have managed to down-regulate these markers are subjected to hazardous surrounding environments due to the release of ROSs, metalloproteinases, chemokines, and cytokines created by the attack and death of neighboring cancer cells [32]. Dendritic cells transport the antigens from the neoplastic cells to the lymphoid organs in order to mount an adaptive response against the tumor.
Yet somehow, in cases of cancer progression, tumor cells are able to survive these stresses and move to alternate locations. This is partially due to activated innate immune cells and paracrine signals from surrounding cells releasing soluble pro-survival molecules that initiated tumor cells can use to alter their levels of gene transcription, continuing the cell cycle and surviving [33]. Even though the hazardous environment may kill some neoplastic cells, others may develop and thrive due to increased genomic instability from free radicals, creating additional, resistant cancer cells [34]. In fact, chronic inflammation has actually been linked to tumor development. Inflammatory cells
16
can actually help in the angiogenesis and migration of glioma cells by promoting vasculature development and releases extracellular proteases that rebuild and mold the tumor environment [35]. The adaptive immune response eventually builds such that it can clear some of the neoplastic cells, but many of the cancer cells have further transformed so that they are not recognized by the cytotoxic T-cells. Even though the adaptive immune response may initially be helpful, as it continues it further promotes chronic inflammation and stress in the area, thereby contributing to cancer progression.
Those cells able to overcome the response of the immune system have a better chance of surviving invasion and migration into new tissues. After breaking through the basement membrane, invasion may include entrance into nearby blood vessels and lymphatic ducts. Neoplastic cells are able to travel through the circulatory system taking up residence in new areas of “ideal” conditions, again through the process of invasion.
The cells that survive this process are again subjected to the stresses of the immune system, new ECM signals, and lack of designated tumor blood vessels. While it was originally thought that stress would have a negative impact on metastasis, stress signaling pathway actually induces the transcription of metastasis-promoting factors [36].
With the above intrinsic stresses attacking and even helping neoplasms throughout their progression, it is a testament to cancer cell adaptability that any cancerous cells survive. By adapting to these stresses, cancers have selected for the most stress resistant cells, making cancer therapy ineffective. However, while many therapies do produce the same types of stresses already present in the body, the treatments cause harsher and sustained stress, especially when combined.
17
1.2.2. Extrinsic stress caused by cancer treatments
Tumor cells have ways to survive the various internal stresses covered in the previous section, so it is of no surprise that glioma cells are often able to find ways around common therapies due to the similarity of the mechanisms of action of the interventions with the mechanism of the body’s natural defenses. Table 1.1 lists some of the common and experimental therapies used to treat solid cancers and the type of stress they cause. While the mechanisms of action are diverse, the treatments cause the same stresses already encountered by the neoplastic cells during the body’s intrinsic response to aberrant cell growth.
Therapy Mechanism of Action Form of Stress Surgery Evacuation of tumor site De-vascularization with subsequent hypoxia and nutrient deprivation, immune response Radiation Ionizing radiation Oxidative stress, DNA damage External beam radiation Brachytherapy Genotoxic Chemotherapy DNA and organelle damage due to Temozolomide, Procarbazine Nonclassical alkylating agents interruption of replication, induces Carmustine (BCNU), Lomustine Nitrosourea alkylating agents metabolic stress and ROSs Cis-platinum, Carboplatin Platinum DNA crosslinkers Vincristine Mitotic inhibitor Etoposide, Irinotecan Inhibits topoisomerase I or II Monoclonal Antibodies Bevacizumab Anti-angiogenic Hypoxia EGFR – Cetuximab, Tyrosine kinase inhibitors Growth factor signaling inhibition Nimotuzumab leading to oxidative stress Immunotherapies/Vaccines* Immune system response Oxidative stress, DNA damage, metabolic stress Small Molecule Targeted Therapy* Tyrosine kinase inhibitors Growth factor signaling inhibition EGFR – Gefinitib, Erlotinib leading to oxidative stress PDGFR – Imatinib mTOR - Everolimus Hormone Therapy Anti-estrogen activity Decreased hormone proliferation and SERMs - Tamoxifen cell survival signaling leading to SERDs - Fulvestrant oxidative stress AIs - Anastrozole Table 1.1: Stresses caused by glioma and breast cancer therapies 18
1.2.2.1. Surgery
Surgery is almost always used on patients who are surgical candidates.
Debulking of the tumor not only allows for better tissue function but also allows chemotherapies to be more effective by working on a smaller population. Therefore, while surgery should be undertaken in situations where critical structures will not be disrupted, the act is extremely stressful on the brain. Small areas will be cut off from the blood supply creating a hypoxic and nutrient deprived environment leading to metabolic stress for unremoved cancer cells. The death of neighboring cells along with the immune response will cause an increase in oxidative stress and ROS production.
1.2.2.2. Radiation
Radiation is another first-line therapy for many tumors, including glioma and breast cancer. While there are many variations of radiotherapy, ionizing radiation (IR) tends to work through two basic mechanisms that ultimately damage DNA by either charged particles or photons. In the case of photon radiation, like in intensity modulated radiation therapy (IMRT), this technique causes indirect damage that occurs after water is ionized producing free radicals. Double-stranded DNA breaks (DSBs) are the most significant cause of cell death. Photon radiotherapy requires well-oxygenated tumors to create the damaging free radicals, which requires adequate blood supply to all areas of the tumor. Because some tumors are hypoxic, this technique is often relatively unsuccessful long-term [37]. Particle therapy, on the other hand, works by directly damaging the DNA by charged particles. Direct damage can occur through transfer of 19
energy from charged particles like proton, carbon or boron ions that do not require oxygen. These particles can cause DSBs themselves. In either case, there are free radicals and ROS produced by radiation; the ROS are necessary for the efficacy of the treatment, further injuring cells [38]. However, these reactive molecules are released during cell death causing increased stress to surviving cells. The body mounts an immune response to repair and clear damaged cells. Although many cancer patients take steroids to reduce the swelling and inflammation produced by radiotherapy, remaining neoplastic cells are still subjected to large amounts of stress, killing many while further transforming others into radio-resistant tumor cells.
1.2.2.1. Chemotherapy
Chemotherapy is often used in conjunction with surgery and radiotherapy. Most of the common genotoxic chemotherapies for glioma produce their effects by disrupting the DNA strands. Alkylating agents like temozolomide (TMZ) and carmustine (BCNU) primarily work by alkylating the guanine base of DNA leading to cross-linking of the
DNA strands which causes the strands to be unable to uncoil and separate. The platinum drugs, such as carboplatin and cis-platinum, work similarly by using the platinum ion to cross-link the guanine base pairs on the DNA strand. These therapies are more toxic to cells that replicate and proliferate faster, thus making cancer cells more sensitive than normal cells to genotoxic therapy. The stresses to the cell caused by chemotherapy are mostly due to interference of mitosis and induction of DNA repair mechanisms. When the cell is unable to unwind and repair its DNA, it causes apoptosis and metabolic stress.
20
As apoptosis continues, ROS are released into the ECM affecting nearby cells.
Genotoxic stress through ROS is dependent on activation of SAPK/JNK pathway [39].
Thus, chemotherapy stresses cancer cells through different mechanisms.
A general chemotherapy-induced stress response is seen in many types of cancer cells. This is a response to anti-neoplastic agents that can destroy many cancer cells but induce survival and resistance mechanisms in others. In yeast, stress changed the cell cycle and lead to increases in de novo protein synthesis, proliferation, HSP90 expression, and proton pump levels. The first line of defense in severe shock is de novo synthesis of protective proteins [40]. Increasing key membrane component proteins can up- or down- regulate their efficacy to restore ionic balance. Changes in the heat shock protein (HSP) population of the cell due to chemotherapeutic stress also increase HSP27 and HSP70 in resistant cells; these cells are translocated to the nucleus in response to stress, increasing protein synthesis necessary for resistance [41]. Whole body response is also important as hormones production levels can change during stress responses; such hormones can affect the cell cycle or gene transcription. On a smaller scale, cell-to-cell interactions occur between transformed and non-transformed cells involving the transfer of survival signals, thus indicating that the extracellular environment is important. The stress response to chemotherapy-induced hyperthermia can even lead to induction of drug resistance through a general increase in the production of MDR1 gene product, P- glycoprotein [39].
Anti-angiogenic therapies are becoming more common in the management and treatment of cancer and are used to combat the tumor vasculature. Most of the inhibitors,
21
like bevacizumab, are monoclonal antibodies that work by antagonistically binding vascular endothelial growth factors (VEGFs), the factors responsible for signaling growth of blood vessels. Contrary to other cancers, it is thought that anti-angiogenic drugs in glioma could work by transiently normalizing the tumor vasculature [42]. As discussed previously, tumor blood vessels are abnormal and unstable due to the mixture of pro- and anti-angiogenic factors. An angiogenesis inhibitor would override many of these signals.
Although, this might decrease blood vessel formation, it might also stabilize the existing vasculature. By normalizing the vasculature, there could be improved delivery of chemotherapeutic agents. Either way, the neoplastic cells would be subject to stress caused either by hypoxia and nutrient deprivation or increased concentrations of anticancer drugs.
1.2.2.3. Targeted therapies
Progress in targeted therapies for cancer has been made in recent years.
Numerous small molecule inhibitors are being tested in clinical trials to antagonize the commonly mutated or over-expressed growth factor pathways. These inhibitors, like erlotinib which works on EGFR and imatinib for PGDFR, work by intracellularly binding the tyrosine kinase receptors, interrupting the downstream PI3K and MAPK signaling cascades. Monoclonal antibodies like cetuximab and nimotuzumab (EGFR inhibitors) work similarly, except they bind extracellulary to the growth factor receptors. Most of the stress caused by these antagonists is through decreased growth factor signaling and the resulting metabolic stress.
22
All these therapies cause stresses already encountered by tumor formation, but the duration and severity of the stresses during therapy is more extreme. Prolonged exposure to these stresses can induce cell death programming more effectively than short, intermittent periods. However, in most cases, some cancer cells do survive. They evade the immune system and death signals, selected for by their unique mutations leading to therapy resistance. It is therefore important to determine accurate markers to identify these cells and to classify the mechanisms through which they survive.
1.2.3. Molecular markers of stress
While cancer cells may be adept at surviving cellular stress, they do show indicators of the stresses they endure. These indicators or markers may eventually be exploited to determine what stresses the cancer cells are under, and thus what types of stress they may be more susceptible to if subjected further.
When cells encounter stress, certain elements of the stress response are universal.
There is a highly conserved minimal stress proteome that is shared among species. In a paper by Dieter Kultz, a list of the 41 proteins needed for the minimal stress proteome was compiled (Table 1.2) [43] .
While this table is not exhaustive for all proteins involved in the stress response, nor does it list the most reliable markers, it does indicate that cells all have a fundamental basic response to stress. The response is referred to as the conserved stress response
(CSR). Various stresses may induce different proteins and markers, but certain responses are unchanged between stresses, even amongst species. 23
Minimal Stress Proteome
Redox regulation DNA damage sensing/repair Fatty acid/lipid metabolism Aldehyde reductase MutS/MSH Long-chain fatty acid ABC transporter Glutathione reductase MutL/MLH Multifunctional beta oxidation protein Thioredoxin Topoisomerase I/III Long-chain fatty acid CoA ligase Peroxiredoxin RecA/Rad51 Superoxide dismutase MsrA/PMSR Molecular chaperones Energy metabolism SelB Petidyl-prolyl isomerase Citrate synthase (Krebs cycle) Proline oxidase DnaJ/HSP40 Ca2+/Mg2+-transporting ATPase Hydroxyacylglutathione GrpE (HSP70 cofactor) Ribosomal RNA hydrolase 6 methyltransferase NADP-dependent HSP60 chaperonin Enolase (glycolysis) oxidoreductase YMN1 Putative oxidoreductase YIM4 DnaK/HSP70 Phosphoglucomutase Aldehyde dehydrogenase Isocitrate dehydrogenase Protein degradation Other functions Succinate semialdehyde FtsH/proteasome-regulatory subunit Inositol monophosphatase dehydrogenase Quinone oxidoreductase Lon protease/protease La Nucleoside diphosphate kinase Glycerol-3-phosphate Serine protease Hypothetical protein YKP1 dehydrogenase 2-hydroxyacid dehydrogenase Protease II/prolyl endopetidase phosphogluconate Aromatic amino acid dehydrogenase aminotransferase Aminobutyrate aminotransferase Table 1.2: The minimal stress proteome as adapted from source: Kultz, 2005 [43].
The general response of cells to stress originally focused on three types of proteins: heat shock proteins (HSPs), glucose-regulated proteins (GRPs) and ubiquitin- associated proteins, all of which are inter-related [44]. Of these three types of proteins,
HSPs have been studied the most thoroughly. HSPs are induced during stress as a
24
protective mechanism. While HSPs ordinarily play a more mundane role in the cell, folding proteins into their appropriate tertiary structures and facilitating steroid hormone binding, the subjection of cells to stress activates heat shock transcription factors (HSFs), allowing the transcription of stress-related HSPs like HSP27, HSP70 and HSP90 [45]. In neoplasms, binding of HSP90 to p53 mutants in the cytoplasm can further the damage caused by stress [46]. This is because p53 functions in the nucleus, and it leads to enhanced HSP70 transcription which allows for cancer cell growth [47]. These HSPs, along with others, have been linked to cancer therapy resistance.
The unfolded protein response (UPR) has gained increasing attention as a fundamental stress reaction caused by changes in the cellular redox potential, energy status, or Ca2+ levels, leading to unfolded or misfolded proteins within the lumen of the endoplasmic reticulum (ER). This is also known as ER stress response [48]. ER stress is closely linked to hypoxia and glucose depletion. Misfolded proteins can be a problem due to their propensity to aggregate together and cause harmful accumulations. The role of UPR is to stop protein translation, arrest the cell cycle, and to signal pathways that increase activation of protein folding chaperones, some of which are HSPs. Ultimately,
UPR leads to cell death through apoptosis if translation is halted for a prolonged period.
GRPs are related to UPR and are actually just specialized HSPs that are found in the ER of the cell. In fact, Grp78 is the protein responsible for chaperoning the misfolded proteins and signaling downstream activators of the UPR. Another GRP, grp94 or
HSP90B1, is actually essential for immune responses as it is a chaperone that regulates both innate and adaptive immunity through secretory pathways [49]. Upregulation of
25
these proteins is often seen during stress, and thus could represent markers for stress induction.
Many stresses signal through the stress-activated protein kinase/c-Jun NH2- terminal kinase (SAPK/JNK) pathway, which is activated by a variety of extracellular signals and stresses. These kinases are part of the larger superfamily known as mitogen- activated protein kinases (MAPKs), which control many intracellular events. SAPK is activated by SEK1 or MKK4. The SAPK/JNK pathway is activated by stresses such as hypoxia, radiation, drug therapy, ROS, and inflammatory molecules [39]. They signal through a variety of receptors, including G-protein coupled receptors (GPCRs), cytokine receptors (TNFα), death receptors (Fas), and antigen receptors. SAPK/JNK pathways can control proliferation, apoptosis, transformation and differentiation along with migration. In response to many types of stress such as radiation and hypoxia, this pathway signals for mitochondrial-dependent apoptosis [50]. Thus, the SAPK/JNK signaling cascade is a protective mechanism for cells. However, any mutations or aberrant signaling could also lead to further glioma progression. Up-regulation of proteins involved in these pathways is a good indicator of cellular stress.
Additionally, other markers of general stress have also been found. The MDR1
(multi-drug resistance 1) gene, which encodes the P-glycoprotein responsible for reducing drug accumulation in cancer cells, is actually induced by stresses like acidity, drug treatment, and radiation [51]. Thus, cancer cells under stress have developed multiple mechanisms to evade cell death. The original intent for non-transformed, normal cells was for them to be able to pump out toxins encountered in their environment
26
for survival purposes. Transformed cancer cells have adapted those responses to their own needs.
Some indicators of specific stresses have also been revealed. An example of a marker for specific stress can be found in hypoxia. The transcription factor HIF-1
(hypoxia-inducing factor-1) is a major regulator of the cellular hypoxia response, which binds to hypoxia-responsive elements (HREs) leading to the transcription of genes involved in cell survival, metabolism, angiogenesis and invasion. It can increase expression of glycolysis genes and VEGF protein [52]. The expression of HIF-1 is increased in cancers usually through induction by EGFR or other growth factor signaling of the PI3 kinase pathway and loss of the tumor suppressors p53 and PTEN. HIF-1 expression, and thus hypoxic stress, in tumors can be determined by immunohistochemical staining. Another indicator of stress linked to HIF-1 is NF-κB induction. NF-κB activation leads to the rapid transcription of important genes involved in the stress response. Because it is a transcription factor, it is often thought of as a first line of defense, especially against activators of the immune system [53]. NF-κB is able to regulate many proteins involved in proliferation and survival, including HIF-1.
Another isoform, HIF-2α, appears to be a specific marker for pH imbalance as it is increased with exposure to acidic stress [54].
Overall, there are numerous molecular markers of stress in cancer cells. Many are the result of a general response to stress, but as research continues, better markers for specific stress, like HIF-1 in the case of hypoxia, will be developed as our understanding continues to grow. These markers may eventually help clinicians to positively identify
27
the stresses to which the tumor is subjected, which will inevitably lead to more effective cancer treatment.
1.3. Common mechanisms of cancer therapy resistance
Category of Mechanism Important molecules and Resistance markers Multi-drug Resistance Drug Efflux P-glycoprotein (MDR) ABCs Membrane channels Drug binding Albumin α, β‚ γ globulins Drug Metabolism Cytochrome P450s Phase II Conjugating enzymes Resistance to Bcl-2 proteins Apoptosis Survivin XIAP Cancer Stem Cells Genetic alterations in CD133+ and CD44+/CD24- (CSC) stem cell niches ALDH1+ Notch Cell dedifferentiation CK19 ß-catenin PTC Autophagy Cell self-digestion Beclin-1 eEF-2K LC3-II Gene expression Growth factor and EGFR, ErbB2, PDGFR, IGFR alterations survival pathway up- B-raf, KRAS regulation Rb, p53 Table 1.3: Common drug resistance mechanisms. 28
1.3.1. Multi-drug resistance
Multi-drug resistance or MDR is a common occurrence where cells are resistant to multiple drugs that have different mechanisms of action or structures (Table 1.3), a term that has been adopted in the field of cancer research [55]. Two types of MDR exist in cancer cells: acquired resistance and intrinsic resistance. Several mechanisms of MDR have been discovered. The classical MDR mechanism is the over-expression of ATP- binding cassette (ABC) family members like the membrane transporter, P-glycoprotein
(P-gp) or other membrane carriers and channels [56], which efflux unrelated drugs from the cell [57]. Sometimes, drug concentration and therefore drug influx is affected by binding proteins like albumin, which can dramatically lower the free drug concentrations
[58]. Drug structure and therefore efficacy can be altered by several means. Phase I
(cytochrome P450s) and Phase II conjugating-enzymes that are meant to metabolize drugs and protect the body can lead to chemotherapy resistance [59,60]. Resistance to apoptosis is an intrinsic characteristic of many cancer cells, which renders many drugs that successfully make it into the cell less effective [61].
1.3.2. Cancer stem cells
Cancer stem cells (CSCs) are hypothesized to be the tumor-initiating cells that are responsible for much of treatment resistance [62]. CSCs have the stem cell properties of self-renewal, differentiation, and maintenance of homeostasis [63]. However, they do differ from normal stem cells in that they can initiate tumorigenesis while also being able to invade tissues [64]. This could be due to the minor genetic alterations being sufficient 29
to transform cells that are already self-renewing. Also, some researchers hypothesize that
CSCs may be derived from differentiated cells that have reverted to a stem cell phenotype instead of normal adult stem cells [65]. Regardless of their origin, CSCs have been implicated in MDR and cancer cell survival because of their innate ability to survive assaults on the cell. CSCs could be the reason for the poor prognosis associated with glioma heterogeneity [66] and the intrinsic heterogeneity of breast cancer [67]. CSCs have a variety of markers that are similar to normal stem cells. For instance, in glioma,
CSCs are recognized by their CD133+ status [68]. Breast cancer stem cells often have
CD44+/CD24- and ALDH1+ markers [69]. Some of these stem cell markers are actually increased after treatments like radiation [70] and alkylating agents [71]; this could mean that some cancer therapies actually select for the more resistant CSC populations.
1.3.3. Autophagy
As previously stated, cancer cells are able to employ several ways to overcome stress and the resulting energy depletion. Autophagy, the catabolic recycling of the cell’s own components, takes advantage of this idea as a survival mechanism, especially during metabolic stress [72]. The autophagic process allows the tumor cells to reallocate amino acids, fatty acids and other macromolecules for energy and go into a hibernation-like state where they have decreased energy needs as a strategy of survival until the surrounding environment is more favorable [73]. Thus, autophagy can also be seen as a temporary protective response of cells. Since autophagy can serve as a cell survival mechanism, it is unsurprising that cancer cells would adapt to use it to their advantage. 30
Disruption of the PI3K/Akt pathway has been associated with autophagy induction as well as stimulation of the AMP kinase pathway. mTOR inhibits autophagy through activating one of its downstream targets, S6 kinase [74]. Autophagy has been shown to be induced in a wide range of therapies. Radiation was one of the first therapies shown to cause cancer cells to undergo autophagy [75],. Many cytotoxic chemotherapies, such as temozolomide (TMZ), camptothecin, and cisplatin, cause autophagy in cancer as a protective mechanism [76,77,78]. Even targeted therapies such as growth factor inhibitors induce autophagy. Inhibition of PDGFR and EGFR with imatinib and erlotinib, respectively, induce autophagy but not apoptosis [79]. Studies do indicate, however, that inhibiting autophagy may re-sensitize cancer cells to currently used treatments, providing a way around tumor resistance [80].
1.3.4. Modifications in signaling pathways and gene expression
A plethora of alterations in signaling pathways have been found in various cancer cells. Mitogenic pathways leading to tumor proliferation are often altered in neoplastic cells. Mutations or modifications in growth factor signaling have become an important mechanism of resistance to therapies. A variety of tyrosine kinase and growth factor signaling have been shown to be up-regulated in tumors [81]. Many cancers have altered signaling in EGFR, IGF-IR, and other ErbB receptors that lead to their constitutive activation [82]. This causes downstream proliferation pathways to be signaled. Even without upstream signaling, these pathways can be altered themselves. Resistance through growth factor pathways like HER-2 in breast cancer often involve mutations in 31
downstream molecules such as PI3K [83]. In the case of other EGFR targeted therapies, constitutively active KRAS mutations are common in cases of resistance to growth factor inhibition [84]. The MAPK pathway is often affected by B-Raf mutations [85] and loss of PTEN [86,87]. Targeting such altered proteins however is not always the solution.
Isoforms of the upstream Ras kinases can cooperate with growth signals to overcome
BRAF inhibitors in cancers like melanoma and lung, where it bypasses BRAF and instead activates CRAF and thus MAPK signaling [88]. Resistance to the inhibition of the mitogenic mTOR pathway can lead to increased growth factor and Akt signaling [89].
Evasion of growth suppression and DNA damage by treatments is often accomplished in cancer cells by changes in regulation of the DNA damage controls.
Resistance to apoptosis is common in high-grade tumors, which decreases the efficacy of drugs that cause cell death through apoptotic pathways [90]. Tumor suppressor gene p53 plays an important role in cell cycle regulation in response to intracellular stresses and
DNA damage. Under harmful conditions, it stops the cell cycle and can signal apoptosis of the cell if the damage is too great. p53t is often dysregulated or mutated in cancers; in fact, it is the most frequently mutated gene in human cancers [91]. Another tumor suppressor RB (retinoblastoma-associated) regulates entrance into the growth-division part of the cell cycle and is considered a gate-keeper for cell cycle progression [92].
Mutations or loss of RB allows proliferation of cells that have often signaled for division to stop. Many cancer therapies cause damage to the cell that trigger a halt in the cell cycle, but mutations in these controls can lead to unregulated growth of damaged and
32
cancerous cells. These cells that have altered cell cycle regulation are then able to evade senescence caused by some chemotherapies [93].
The tumor microenvironment is also pivotal for the development of resistance to cancer therapy. The nearby stroma of solid tumors can play a role in the pro-invasion adaptive response [94], and many anti-angiogenic therapies fail because the stroma gives off pro-angiogenic signals [95]. Both fibroblasts [96] and tumor-associated endothelial cells [97] can promote drug resistant phenotypes through a dysfunctional extracellular matrix (ECM).
1.3.4. Alterations in gene expression
Although many of the above mechanisms of therapy resistance could be considered to alter gene expression, there are a few traditional methods of gene expression modifications that have not been mentioned. The idea of genetic instability and the origin of drug-resistance has been around since the 1950’s when it was first implicated in MDR of leukemia cells [98]. Large alterations in gene expression occur with the chromosomal instability and rearrangements that occur within cancer cells, and such large-scale alterations in gene expression can lead to therapy resistance [99].
Epigenetics can also play a role in large-scale changes gene expression. Many tumors have found to have hyper-methylated regions of DNA that repress transcription of certain genes while other areas have hypo-methylation of regions like centromeres [100].
Alterations in both methylation and histone acetylation have been implicated in cancer therapy resistance [101]. Mutations to DNA repair enzymes like BRCA1 and BRCA2 in 33
breast cancer can cause additional mutations and alterations in the expression of a range of unrelated genes which lead to resistance of microtubule therapies [102]
Smaller scale or specific modifications in gene expression are also common.
Amplifications of different genes like the previously mentioned EGFR [82] or dihydrofolate reductase [103] result in therapy resistant phenotypes. Micro-RNAs
(miRNA), small RNAs that affect RNA stability and translation, have been implicated in numerous resistant cancers such as mir-221/222 in tamoxifen resistance in breast cancer
[104] and miR-30b/c in gefitinib resistance in lung cancer [105]. Mutations of specific genes have been correlated with resistance in numerous cancer types [106]
1.4. Regulation of gene expression
1.4.1. Transcriptional control
Genes are found encoded in the double-stranded DNA of our cells. One strand is considered the protein coding strand (5’ to 3’) that contains the actual gene while the complementary anti- parallel strand is the template strand (3’ to 5’). The template strand is copied into an mRNA which is identical to the coding strand, except for the fact that the RNA strand’s thymines (T) are replaced with uracil (U) nucleotides. The process of copying DNA into RNA is known as transcription.
Transcription is controlled by a system of DNA regulatory sequences and DNA- binding proteins called transcription factors. RNA polymerases actually transcribe the
34
DNA. The most important polymerase is RNA polymerase II which is the main polymerase responsible for transcribing all protein-coding genes in addition to non- coding RNAs (ncRNAs) such as microRNA (miRNA), small nucleolar RNA (snoRNA), small nuclear RNA (snRNA) and long ncRNA. Transcription can be regulated at the level of RNA polymerase II by its phosphorylation status which has been shown to affect the rate of transcription and splicing [107]
Regulation of transcription can occur through genetic promoters and inhibitors, modulation of transcription machinery and epigenetic changes. The most direct method are the regulatory DNA binding sites and factors. The sites include promoters, enhancers, insulators, repressors, and silencers. Transcription factors translate intracellular signals into action on DNA binding sites and regulatory complexes. RNA
Polymerase II is able to bind to initiation elements after its co-regulatory complex proteins bind to nearby promoters; the most famous of these promoters is the TATA box which binds the TATA-binding proteins (TBP) part of the transcription factor complex at approximately 25 bp upstream of the start of transcription. Other eukaryotic promoters include DPE elements (AGAC) and the CAAT box. There is a basal rate of transcription but activators and repressors are responsible for additional alterations in transcription rate.
Transcription produces primary transcripts of RNA called pre-mRNAs. Before protein synthesis can occur, the pre-mRNA must undergo post-transcriptional modifications to become mature messenger RNA (mRNA), which is then translated into proteins. Alterations in pre-mRNA processing lead to changes in gene expression. The
35
pre-mRNA is capped with a 7-methylguanosine on its 5’ end by the cap MTase to protect the RNA from degradation of the 3’5’ phosphodiester bonds [108]. Polyadenylation is done on the 3’ end after it is cleaved at a 5’-CA-3’ sequence in the mRNA. The
Polyadenylate Polymerase (PAP) in complex with other proteins adds the poly(A) tail consisting of approximately 200 adenines (A) which is added to the 3’ end to ready the mRNA for export and initiation of translation [109]. The poly(A) tail is able to bind poly(A) binding proteins which aid in protection from ribonuclease digestion. This only occurs in pre-mRNAs which include a polyadenylation signal (5’-AAUAAA-3’) and also serves to protect from degradation [110]. This sequence is often surrounded by cis- elements, which is responsible for alternative polyadenylation that functions to express different isoforms of a protein [111]. Polyadenylation is more frequent in oncogenes of cancer cells which is hypothesized to occur because it shortens the 3’-UTR, deleting important miRNA regulation sites [112]. Another element in the 3’-UTR responsible for mRNA stability are AU-rich elements (AREs) that regulate turnover of mRNAs [113].
AREs most often function to limit over-expression which could be detrimental to cells by binding AU-binding proteins (AUBPs) which lead to their degradation, although AUBP binding can also stabilize some mRNA [114]. At either of these two steps, alterations in capping or interactions of the poly(A) tail will have repercussions in gene expression.
A process especially relevant to genome and transcriptome research is RNA splicing. Pre-mRNAs contain both exons and introns, and the splicing process removes some introns while splicing together exons and any leftover introns. Alternative splicing of genes can lead to different proteins having the same parent gene; it is often signaled by
36
the presence of a splicing regulatory element (SRE) in the pre-mRNA which can enhance, silence, or even frameshift the splicing process [115]. SRE binding of ribonucleoproteins (RBPs) particles can regulate the splicing of mRNA within different tissues and environmental conditions [116]. RBPs are especially important in this process as they protect mRNA from degradation. They also help, along with molecular motors and the cytoskeleton, to localize the mRNA to the area of the cell that is signaled by its zipcode or localization element in its ‘3-UTR [117]. mRNA can contain multiple zipcodes which can take the form of sequence elements or complex secondary structures, and these can be enhanced or silenced by the binding of RBPs [118].
Non-coding RNAs have their own processing sequence. rRNAs start as pre- rRNA which are cleaved and modified with 2’-O-methylation and pseudouridines by snoRNAs. The snoRNAs are able to base-pair with the rRNA while the proteins catalyze the reactions. These snoRNAs often form complexes with other proteins to become snoRNPs. The snoRNA’s themselves are methylated or psuedouridinilated by small
Cajal body specific RNAs (scRNAs), which are similar to snoRNAs in structure. tRNA’s have their 5’ and 3’ ends removed by the enzymes RNase P and tRNAase Z, respectively, after which a 3’ CCA tail is added. miRNA’s have additional processing. They are capped and tailed like mRNA, and then processed to pre-miRNA 70-nucleotide loops by
Drosha and Pasha in the nucleus. These are then exported into the cytoplasm where the endonuclease Dicer processes it into a mature miRNA.
After the pre-RNAs are processed into mature RNAs, most are then exported from the nucleus into the cytoplasm by nuclear pores. ncRNA are then ready to function
37
as this is their final gene product. mRNA, however, must still be translated into proteins into order to fulfill their role in gene expression. These mature mRNA carry a protein sequence that can be translated into proteins by the ribosome.
1.4.2. Regulation of translation
mRNAs are made of three parts: both a 3’ and 5’ untranslated region (UTR) and a protein coding region or open reading frame (ORF). The untranslated regions are involved in the regulation of translation while the protein coding region is made up of triplets of nucleotides called codons, which correspond to one amino acid in the protein.
Codons are recognized with complementary anti-codons found on the ends of tRNAs which carry the amino acids. The ribosome complex contains enzymes that catalyze reactions that link the amino acids together into a protein chain. One mRNA can be the template for many identical proteins, often in a chain of ribosomes.
Once the chain of amino acids is released by the ribosome, the unstructured polypeptide is then folded by chaperone enzymes into their three-dimensional structure.
Finally, proteins are transported to the parts of the cell where they function or they are translocated into the endoplasmic reticulum (ER) and golgi apparatus for final modifications before export from the cell.
While only approximately 2% of the genome is protein coding [119], almost all of the genome is transcribed. These non-coding region transcripts are not translated, however, and instead play other roles in the cell. Despite the presence of numerous
38
ribonucleases and other cutting enzymes, the average half-life of an mRNA is over 6 hours [120].
Traditional translation initiation occurs when the eIF4F protein complex binds the
5’cap. This complex is made of the RNA helicase (eIF4A), its activator (eIF4B), the cap-binding protein (eIF4E), and an additional protein which interacts with the PABPs and other regulatory molecules (eIF4G) [121]. Meanwhile, the 40S ribosome small sub- unit starts complexing with elongation factor proteins and initiator tRNA to bind mRNA and scan from the 5’ end towards 3’ end to find the start codon (AUG). The large 60S ribosome unit joins with the small unit to form the 80S ribosome after several initiation factors dissociate [122].
Different sequence motifs exist within mRNAs that determine how they are translated, like internal ribosome entry sites (IRESs), terminal oligopyrimidines (TOPs), and upstream ORFs (uORFs) [123]. The binding of the mRNA to the ribosome is commonly guided by its 5’cap. However, not all mRNAs are 5’-capped, and these mRNAs often have a structural motif called an internal ribosome entry site (IRES) located in the 5’-UTR. IRESs can direct cap-independent initiation to occur by binding eIF4G or the small ribosome unit directly without the help of the eIF4F complex.
Alterations in the expression and activity of any of these initiation factors and associated proteins can regulate protein synthesis. The presence of additional 5’-UTR functional
RNA elements such as TOPS and uORFs can also alter gene expression. TOPs are normally found in ribosome and elongation factor mRNAs whose transcription is dependent upon nutrients, and their presence or absence would affect global translation
39
[124]. uORFs can regulate the speed at which elongation proceeds while providing a protected environment for mRNAs in the ribosome [125]. Together, these 5’-UTRs present in different mRNAs can affect both global and specific gene expression.
Once mRNAs are allowed to undergo translational initiation and elongation, they go through multiple rounds of translation. When the cell no longer needs the mRNA or a disease process is underway, the mRNA is degraded. This is signaled through additional cis-elements on the mRNA and enzymes. The most common pathway is for the poly(A) tail to be shortened by poly(A) nuclease, decapped, and cleaved [126]. Polyadenylation- independent pathways include other enzymatic cleavage enzymes, oligo(U) additions, and miRNA interactions, in addition to the previously mentioned AREs. Iron rich elements (IREs) are another stabilization motif in mRNAs that are involved in the homeostasis of iron within the cell [127]. Iron regulatory proteins can bind to 5’-UTR
IREs near the cap to promote ribosomal dissociation and degradation of the mRNA or bind to 3-UTR to stabilize mRNAs and promote translation [128]. miRNA are a large regulator as well. Riboswitches are common in bacteria, but are also found in eukaryotes
[129]. These mRNA catalytic domains can change conformation when bound to molecules or metal ions and affect anything from transcription (early termination) to translation (inhibition of initiation)
Epigenetics control the higher order structure of DNA which can be modified with non-sequence specific DNA binding proteins and chemical alterations. The function of epigenetics is to make the DNA more or less accessible for transcription. Chromatin is controlled by binding of histones which allow areas of the genome to be expressed, while
40
methylation of DNA usually silences transcription. Methylation is an inheritable feature of transcription regulation [100]. Histone acetylation, however, is not. The effect of acetylation of histone residues is not fully understood as both hyper- and hypo- acetylation can be associated with transcription [130]. Epigenetics has been increasingly recognized as a vital part of gene expression.
1.4.3. Post-transcriptional and post-translational modifications
As discussed previously, post-transcriptional modifications are important for gene expression. Gene expression can be regulated even at the level of nuclear export of mRNAs. Degradation of mRNAs also plays an important role in gene expression, and this can be a tricky proposition as an mRNA has to travel across the cell. Small interfering RNA like miRNA can target mRNA for destruction.
Two main groups of post-translation modification to protein structure can occur: covalent additions of chemical groups to a side chain group or cleavage of the actual peptide backbone.
1.4.3.1. Covalent additions
One of the most well known covalent modifications is phosphorylation – the reversible catalytic addition of a phosphate group to a protein side chain. A hydroxide side chain is exchanged for a dianionic phosphate group catalyzed by a protein kinase.
The main sites of mammalian phosphorylation take place on serine (S), threonine (T), and tyrosine (Y) residues, although phosphorylation of histidine (H) and aspartate (D) 41
can occur [131]. Phosphorylation induces conformational changes in proteins which restructures the protein to an active or inactive state and often initiates signaling in a pathway [132].
Other common covalent additions include acylation, alkylation, glycosylation, and oxidation. In acylation, carbon chains with hydroxyl groups are added to amino acids, which includes modifications like acetylations (such as with histone regulation of gene expression). Acylations are involved in the linker for post-translational modifications that add proteins or peptides like ubiquitination (to target proteins for degradation and SUMOylation (localization signals) [133]. Alkylations involve additions of single-bonded carbon groups; methylation of histones on lysines or arginines are a prime example [134]. Glycosylation or the addition of glycan (sugar residues), which performed in the endoplasmic reticulum, is often seen in lipids or other proteins that are to be secreted from the cell [135]. Finally, oxidation reactions which cross-link proteins are also common; the prototypical reaction for this category are the disulfide bonds formed from oxidation of cysteine residues [136].
1.4.3.2. Cleavage reactions
The other main group of post-translational protein modifications, cleavage reactions, serves important functions in the lifespan and activity of different proteins.
While typically thought of in protein degradation, specific cleavages of proteins are necessary for their function within the body. Secretory proteins must undergo cleavage
42
of their N-terminal 25-30 amino acids to be secreted [137], and many protein precursors, such as the proteasome, must be autocleaved in order to be activated [138].
1.5. Use of next generation sequencing in gene expression studies
The human genome has approximately 3.2 billion bps made up of four types of nucleotides whose sequence can be ordered in approximately 4^3,200,000,000 permutations of the human genome [119]. The number of possible genomes is mind- boggling, and having that many possible combinations of sequence makes the importance of appropriately determining sequence nucleotides even more critical. Undetermined nucleotides (N) add up and decrease the accuracy of our reference genomes, a fact that is important when determining disease mutations. Thus, better sequencing technology has been developed to try to overcome previous limitations and lack of precision plagued by previous systems.
Traditional Sanger sequencing used to sequence the original human genome took a “shotgun de novo” approach by fragmenting genomic DNA and then cloning it into a plasmid vector that transformed E. coli to amplify the DNA. This DNA was then sequenced by incorporating fluorescently labeled dideoxynucleotides (ddNTPs) which would stop the extension of the DNA. The sequence is then determined with a capillary based gel and electrophoresis. Overlap in the random fragments would be used to assemble longer sequences [139]. Since about 2005, a new version of sequencing has 43
been available called next generation sequencing (NGS) (also known as massively parallel, ultra-high throughput, or deep sequencing).
NGS was developed because there was a high demand for lower-cost and faster sequencing of large sets of genes and genomes. The first human genome was a collaborative effort from multiple laboratories across multiple continents that was only published after 12 years of sequencing. The total cost for the first genome is estimated to have been greater than $3 billion [140]. Today, makers of one of the next generation sequencing technologies claim that they can sequence a preliminary human genome for
$1,000 in less than 2 hours; however, even higher accuracy next generation platforms can still do it for $5,000 in a week’s time, a vast improvement over previous versions [141].
1.5.1. Sequencing Process
NGS is still a “sequencing by synthesis” method. The basic premise of NGS is that sequencing of millions of fragments or reads are done in parallel, thus decreasing run times dramatically. The process involves creating a cDNA library which is fragmented and then ligated to different types of universal adaptors. These fragments are then amplified in miniature individual PCR reactions that can be sequenced in parallel (hence, massively parallel) using arrays or flow cells that allow for millions of reads. Some sequence from one end (which would have its own biases) while newer technology allows for paired-end reads.
The depth or redundancy of the sequencing is important in deciphering the accuracy of the sequence. This involves knowing the mean coverage of the genome and 44
the percent of bases covered at least N number of times. Usually, 100x coverage and
90% (of N=20x) is needed for comparison studies of genomes [142]. Quality scores of the individual nucleotides are important as well, and this is determined by platform software. Reads of bad quality are removed, which can mean that too many Ns were found in a read or that the read could not be mapped. Expression scoring after mapping is where NGS can be incredibly useful in that it gives quantitative data for transcript reads. For exomes, it is easier since many genes are annotated. Scoring can be accomplished by summing the number of reads or scoring each nucleotide position on an exon and normalizing to the gene length. Transcriptome analyses can give vital new information on alternative splicing or isoforms hinted at by exon-exon spans, while statistical analyses can be used to examine non-random associations between the number of reads and the sample for comparison.
1.5.2. Platforms
Several commercial platforms are available. Some platforms are more suited for certain uses than others. A comparison of these sequencers is listed in Table 1.4 [141].
45
Platform Method Read Uses Advantage Disadvantages length Illumina reversible 4-color 75-100 SNP analysis, whole Most used Low multiplex Hi-Seq dye termination bp genome and exome method with sequencing many analysis tools 454 Bead-based >500 bp Confirmatory high accuracy High error rate Pyrosquencing pyrosequencing sequencing, SNP with long reads for homo- detection performed polymer quickly SOLiD Dinucleotide 50-75 bp SNP analysis, Less error with Long run times ligation with 4- whole genome and 2-base system color exome sequencing
Helicos PolyA-tail with 4- 35 bp Whole genome and Unbiased reads High error color exome, single molecule sequencing,
Ion Torrent Seminconductor 50 bp Targeted sequencing Fastest and High error, low non-optical, cheapest parallelism standard sequencing Table 1.4: Comparison of NGS platforms
1.5.2.1. Illumina HiSeq and Genome Analyzer
Solexa, which was bought by Illumina, created the first NGS platform, and their
NGS products are still the most widely used. The Illumina platform has been through numerous improvements, but still relies on the same basic technology. It uses a bead- based system which ligates fragments to adapter sequences, performs of bridge amplification of the fragments, and then reads fluorescently labeled nucleotides that incorporated one at a time using reversible dye terminators [143]. It is a short-read system that can do eight lanes in parallel to increase sequencing depth. Disadvantages of this platform include its low multiplexing capability and its high rate of aberrant nucleotide incorporation.
46
1.5.2.2. Roche 454 Pyrosequencing
The 454 platforms performs thousands of simultaneous emulsion PCR reactions on primer-coated beads that each hold a single DNA template. These PCR reactions are then loaded onto an array plate and incorporation of nucleotides that release pyrophosphate which is then read by an optical camera. The platform has quick run times but does not handle homopolymer repeats well. [144]. For this reason, it seems to be mostly used for confirmatory sequencing or smaller genomes
1.5.2.3. Helicos Heliscope:
On the Heliscope platform, fragments of DNA are polyadenylated and then hybridized to poly(dT) oligonucleotides on flow cell. It also uses an optical camera to record nucleotide incorporation. This platform is of note since it was the first to be able to perform single molecule sequencing [145].
1.5.2.4. Applied Biosystems SOLiD Sequencing:
The SOLiD acronym stands for Supported Oligonucleotide Ligation and
Detection). It is aptly named since it uses a process that uses DNA ligase to ligate oligonucleotides after emulsion PCR in process that is sometimes referred to as polony sequencing. Di-nucleotide bases are then incorporated, detected with an optical camera, and then sequences are decoded color-space coding algorithms. The SOLiD platform has been especially useful for whole genome and exome sequencing, in addition to SNP discovery. It has a lower error rate due to 2-base encoding, but its disadvantages are that 47
run times are longer and the color-space coding leads to a more complex data analysis process [144].
1.5.2.5. Life Technologies Ion Torrent Sequencing
This new commercial platform uses a new method of reading fragment sequences.
While it still uses standard DNA polymerase, the process of sequencing is referred to as
Ion semiconductor sequencing which detects hydrogen ions released during the creation of DNA polymers. The technology is still relatively new, but this platform and process has the potential for large commercial success.
1.5.3. Advantages and disadvantages of NGS
The obvious advantage of the massively parallel sequencing performed by NGS is that the process is much faster than traditional methods. Due to the high-throughput and tiny sequencing reactions, the cost is definitely cheaper than Sanger sequencing and will only continue to decrease. Using NGS for exome sequencing would be cost efficient for patients since it represents only 1-2% of the genome However, at this point, NGS might be more appropriate for large gene studies rather than experiments focused on only a few genes. Sequencing of a single well-annotated gene might be more accurate and less expensive with normal Sanger sequencing.
While NGS might not be appropriate for a small gene study, it is probably the best platform for working with small amounts of samples and genetic material such as with
48
aspiration biopsies and circulating tumor cell profiling. NGS technology is able to sequence a single cell, so small sample size is not an obstacle.
The accuracy of NGS also allows for the discovery of single nucleotide polymorphisms (SNPs) and mutations, which in turn will allow scientist to form a more appropriate average human reference genome. NGS has been used to compare human genomes and unrelated individuals have been shown to have more than two million bp differences in their DNA [146]. This fact is important if databases comparing normal and diseased tissues are going to be able to decipher the difference between normal versus disease polymorphisms. NGS SNP studies are necessary to find out what variations are within normal ranges. At a larger level, NGS is a great tool for looking for homozygous and heterozygous deletions and amplifications in addition to translocations.
A more accurate human reference genome will also aid in the discovery of new genes and smRNA, as transcripts will be able to be more accurately mapped to the genome. NGS will play a key role in the discovery of the role that uncharacterized transcripts play within the cell and human body. The hope is that NGS will give an unbiased approach to the discovery of novel transcribed regions. Theoretically, a library could be resequenced over and over again to reach a level where every single transcript is recorded which would allow scientists to be able to examine the true diversity of the transcriptome
NGS is especially valuable for gene expression studies as compared to microarrays since NGS data represents absolute number of reads, which are able to be counted. Range of expression is only limited by depth of sequencing, not the range of the
49
scanner as in microarray studies. Also, NGS produces little to no noise, which allows for the detection of lowly-expressed transcripts [147].
The major disadvantage of NGS is the bioinformatics challenge that it poses.
Even storage of sequence read information is difficult due to the size of the data files.
The size of the files also creates the need for networked computer clusters to analyze the data. No personal computer has the computing power to run such large amounts of data, which limits some labs ability to do NGS projects. Another limiting factor for labs is that no “box-standard” software packages exist for data analysis. Thus, a bioinformatician is usually necessary to even start to decipher the data derived from NGS runs. Mapping can be an issue with the small reads created by NGS. Repetitive sequences can be difficult to map, and they are important as these account for nearly 50% of the genome [119]; many times they can align to more than one area on the genome. Paired-end or mate-pair refers to sequencing both ends of a fragment which creates a larger read with known ends (~120 bp). This can allow for unambiguous mapping to the genome despite repeats. It increases read mapping from 85% to 93%, but also increases the run time and cost [148].
A number of unintended consequences or issues can occur with NGS as it starts to be used in a clinical setting. Some were mentioned previously like the unknown significance of DNA variants or their penetrance and expressivity within different individuals. Large ethical dilemmas come into play as doctors will have to determine how incidental findings and impact on patients and families, consider information privacy, encounter paternity issues, and train medical staff in genetics [149]. Some safeguards have already been put into place in the US; the Genetic Information
50
Nondiscrimination Act of 2008 legislation was passed in the US, which prohibits insurance company from using genetic information against patients.
1.5.4. Uses in gene expression studies
Within the last few years, there has been an explosion of NGS sequencing studies examining the genomes of many different types of species from wholly mammoths to bacteria to Neanderthals [150,151,152]. Cancer genomes are starting to be examined as in the case of a study that sequenced estrogen receptor-positive metastatic lobular breast cancer [153]. Even single cell genomes have been sequenced (single nucleus sequencing). This could be used to look at genomic heterogeneity within tumors [154].
All these works will be pivotal for creating accurate reference genomes, which will be necessary as NGS is increasingly being used gene expression studies. All the uses for
NGS in cancer discovery are probably not even known yet, but there is a subset of uses for which NGS has been single out as an important tool.
1.5.4.1. Chromosomal rearrangements
One of the many uses of NGS is that it has the ability to look at breakpoints for chromosome translocations and inversion discovery [155]. Thus, it is used frequently to study chromosomal rearrangements in cancer. Some studies have examined cancer cell lines [156], while others have looked at patient samples. The first whole genome sequencing on prostate cancer versus normal tissue was done using NGS in conjunction with FISH in order to look at chromosomal rearrangements [157]. Lung cancer [158] and 51
numerous leukemias, such as CML [159] and APL [160], have had their chromosome rearrangements examined with NGS. Breast cancer in particular has had several important studies on chromosome alterations. Both structural rearrangements and mutations were examined in basal-like breast cancer from blood, tumor, metastasis, and xenograft [161], while a paired-end NGS study found a variety of somatic rearrangements in the human breast cancer genome indicating the diversity of rearrangements in breast cancer [162].
1.5.4.2. Transcriptomes, exomes, and gene signatures
Transcriptome sequencing has become a popular use for NGS – this sequencing of RNA is referred to as RNA-Seq. Studies on focused on mRNA expression are referred to as exome projects. These different types of gene expression studies will help us to better understand how cancer cells are able to survive despite the hazardous environments they encounter. It has lead to the discovery of numerous new small RNAs [163,164,165] in addition to its uses in protein-coding gene expression. Exomes of AML [166], melanoma [167], breast, and colon cancer [168] were a few of the first studies to compare the full range of variations between normal and cancerous tissue with NGS.
In addition to examining entire transcriptomes, NGS has been used for very specific examples of gene expression regulation. NGS can examine resistance mechanisms of cancer, such as the study that determined the gene expression changes in response to targeted kinase inhibitors in adenocarcinoma of the tongue [169].
Transcriptome analysis has been used for new biomarker applications; for instance, one
52
research team studied DNA circulating in the serum of breast cancer patients compared to controls [170]. Another value of NGS transcriptome or exome sequencing is that allows scientist to see the differences in splicing variation of genes that occur during different parts of the cell cycle and states of proliferation [171]. All these differences in gene expression found by NGS will be pivotal to overcoming cancer survival.
1.5.4.3. Epigenome
While technically a version of gene expression studies, epigenomic experiments highlight a higher level of regulation of gene expression and are often put in their own category. Although NGS is just starting to be used for epigenome studies, the field holds great promise. Currently, NGS have been used to map chromatin and DNA methylation in stem cells, normal, and cancer cells [172].
A popular use for NGS in epigenetics is the examination of promoter CpG island
DNA hypermethylation – which stabilizes transcriptional repression and leads to loss of gene function. Most studied epigenetic change in cancer [173]. 5-10% of the promoter
CpG islands in a cancer patient have abnormal gains in methylation [174]. Methylation of genes and sequences that produce miRNA can lead to up-regulation of oncogenic targets [175]
DNA methylation can also occur up- and down-stream of promoter islands, and called shores. Losses in methylation have also been shown to be important and more widespread [173]. These losses are actually in CpG-poor regions and span across multiple chromosomes, with discreet increases in methylation in certain promoters [176]
53
1.5.4.3. Binding sites for transcription factors
ChIP-Seq combines traditional chromatin immunoprecipitation (ChIP) with NGS to sequence the DNA that binds to the proteins of interest while eliminating the need for hybridization. This technique can easily perform a whole genome coverage and has far better resolution. ChIP-Seq was used for binding studies of transcription factors NRSF
(neuron-restrictive silencer factor) [177] and STAT1(signal transducer and activator of transcription 1) [178]; these studies were able to confirm previous microarray studies while adding additional sites. Other non-transcription factor DNA or RNA-binding proteins have been studied with NGS; histone-methylation marks, histone variants and
RNA polymerase II binding were all studied in a genome-wide analysis [179]. Even
DNase-hypersensitive sites, sites associated with regulatory elements, can be studied without chromatin proteins. Studies have found that these DNase-hypersensitive sites are not even associated with promoter regions, giving new insight into open DNA regions
[163].
1.6. Significance of my research project
Ultimately, the goal of this research is to advance the understanding of how gene expression is altered in cancer cells that are under stress. The objective of our global study of Tam resistant is to overcome the limitations of previous studies by developing a comprehensive analysis of the transcriptome changes involved in Tam resistance in
54
breast cancer using the NGS method. NGS allows for unbiased analysis and exploration of all possible cellular molecules and pathways. Tam resistance is a complex problem, and the field would benefit tremendously from studies examining global changes with
NGS, which have not been previously explored. The objective of our focused, mechanistic study on the stability of protein synthesis regulator eEF-2K is to better understand how translation is regulated in cancer cells under stress and to determine the role that phosphorylation plays in eEF-2K turnover. The differential regulation of eEF-
2K stability and phosphorylation in response to upstream signaling could have function consequences for targeting eEF-2K or protein synthesis in future cancer therapeutics.
Together, the results from these studies will support the idea that stress has global impact on cancer cell gene expression, both through general and specific mechanisms, which has broad implications for future cancer treatment strategies.
55
Chapter 2
Differential gene expression in tamoxifen-resistant breast cancer cells revealed by a new analytical model of RNA-Seq data
2.1. Abstract
Resistance to tamoxifen (Tam), a widely used antagonist of the estrogen receptor
(ER), is a common obstacle to successful breast cancer treatment. While adjuvant therapy with Tam has been shown to significantly decrease the rate of disease recurrence and mortality, recurrent disease occurs in one third of patients treated with Tam within 5 years of therapy. A better understanding of gene expression alterations associated with
Tam resistance will facilitate circumventing this problem. Using a next generation sequencing approach and a new bioinformatics model, we compared the transcriptomes of Tam-sensitive and Tam-resistant breast cancer cells for identification of genes involved in the development of Tam resistance. We identified differential expression of
1215 mRNA and 513 small RNA transcripts clustered into ERα functions, cell cycle regulation, transcription/translation, and mitochondrial dysfunction. The extent of alterations found at multiple levels of gene regulation highlights the ability of the Tam- resistant cells to modulate global gene expression. Alterations of small nucleolar RNA, 56
oxidative phosphorylation, and proliferation processes in Tam-resistant cells present areas for diagnostic and therapeutic tool development for combating resistance to this anti-estrogen agent.
2.2. Introduction
2.2.1. Estrogen and Estrogen Receptor
The primary female sex hormone estrogen comes in three forms: estrone (E1), estradiol (E2 or 17β-estradiol), and estriol (E3). Of these estrogens, estradiol has the strongest potency despite being less plentiful than estriol. Estradiol plays a major role in non-pregnant females while estriol and a fourth estrogen – estetrol (E4) – become the principal players during pregnancy. Estrogens are primarily produced in developing follicles in the ovaries, although secondary sources include the liver, adrenal glands, breasts, and adipose tissue, which explains their presence and activity within males.
Steroid synthesis occurs from the conversion of male sex hormones, androgens, into estrogen by the important enzyme aromatase. Estrogen hormones are then circulated around the body where they easily diffuse across cell membranes.
Once inside the cell, estrogen has many functions. In females, estrogen promotes the development of female secondary characteristics; regulation of the menstrual cycle; and coagulation, while in males, estrogen can regulate the maturation of sperm. In both sexes, estrogen regulates cholesterol, is beneficial for heart health, and reduces bone
57
resorption and can be particularly protective in pre-menopausal women. Many physiological processes are affected by estrogens including proliferation and cell survival
Estrogen acts through its binding to the estrogen receptor (ER). ER is a member of nuclear receptor family, which are ligand-inducible transcription factors. ER has a hydrophobic pocket for estrogen, which is left exposed for co-regulators. There are two types of estrogen receptor, ERα and ERβ. Both can bind estrogen and other small lipophilic molecules, but are found in different concentrations and locations throughout the body. The liver has primarily ERα, while the gastrointestinal tract has mostly ERβ.
The breast, central nervous system, cardiovascular system, bone, and urogenital tract have a combination of both ERα and ERβ [180].
2.2.1.1. ER structure
Both types of ER contain six domains. They have hydrophobic ligand binding domain known as AF-2, which consists of 12α-helices [181] and a growth factor binding domain called AF-1, separated on opposite ends of the protein by a hinge domain. These domains also surround a central DNA-binding domain which binds the small palindromic
DNA sequences known as estrogen response elements (ERE) on the chromosome. When a ligand binds AF-2, this mediates the interaction with co-activators the increase ER transcriptional activity. ERα and ERβ share 96% homology of their DNA binding domains (DBD) [182] and differ mostly in their AF-1 domains. ER-β transcriptional activity is negligible compared to ERα activity [183].
58
59
2.2.1.2. Regulation and Activity of ER
A series of co-activators and co-repressors add levels of regulation to estrogen responses. Co-activators like SRC-1 that are able to modify histone acetylation are necessary to decompact chromatin. Other co-activators such as AIB1 and p300/CBP help to form a functional ER complex. Co-repressors like SMRT act as silencers and are often histone deactyltransferases [184].
Through the estrogen receptor, estrogen can control gene expression through a variety of classical genomic and nongenomic pathways (Fig. 2.1). In the classical mechanism ER can form homo- or heterodimers of ERα and ERβ. Estrogen binds to ER which then dimerizes, disassociates from heat shock proteins, and enters the nucleus to bind to EREs through its DBD [185]. Other genomic pathways exist where ER does not bind DNA directly. Instead, it modulates transcription through tethering with other transcription factors like activator protein 1 (AP-1), specificity protein 1 (Sp1), and nuclear factor-KB (NF-KB) [184]. Thus, non-ERE genes can be transcribed through estrogen activation.
ER can also be activated through several of its phosphorylation sites which are downstream of kinases such as MAPK and PI3K/Akt. Serines 104, 106, 118, and 167 are of particular importance since they lie within the AF-1 region. MAPK phosphorylation of Ser118 has been shown to up-regulate ERE activity of ER [186], while PI3K/Akt stimulates ER activity through Ser167 [187]. ERα can also be methylated at R260 by protein arginine N-methyltransferase 1 (PRMT1). This leads to its association with
60
PI3K, SRC and focal adhesion kinase (FAK) which in turn activates Akt [188] (Fig.
2.1C).
Estrogen can also rapidly work through nongenomic pathways. Other nuclear steroid receptors have previously been shown to interact with the inner plasma membrane through proteins like caveolin-1[189], rapidly relaying extracellular signals through nongenomic signaling pathways. ER can be found at the cell-membrane and has a binding domain for many growth factors including IGF-1R, HER2, EGFR, and GPR30.
Cytoplasmic sequestration of ER can also occur through interaction with HER2, which then signals through ERK1/2 [190].
2.2.2. Estrogen receptor in breast cancer
ER-positive tumors comprise approximately 70% of breast cancers [191]. ERα is expressed in 15-30% of luminal epithelial cells in normal breast tissue, which usually results in transcription of genes [192]. ERβ is expressed throughout all breast tissue
[193]. In addition to being co-expressed with ERα, it has functions within the stroma and myoepithelial cells. Its role in proliferation and neoplasms is still unknown; however,
ERβ is necessary for normal mammary development [194]. New studies suggest that
ERβ may be involved in pro-apoptotic and anti-proliferative actions, so it is unsurprising that it is often down-regulated in breast cancer. Many of these genes are involved in cell cycle and proliferation because during a normal menstrual cycle, proliferation of breast lobules increases before menstruation [195]. Estrogen stimulation of breast cells can occur of through paracrine signaling, such that ER negative cells can have their 61
proliferative machinery induced. ERα expression increases in breast cancer, and these cells now can shift to autocrine growth [191].
2.2.3. Endocrine therapy in breast cancer
The goal of the treatment for hormone-dependent breast cancer is to inhibit the activity of estrogen on tumor cells. This can be done by 1) preventing estrogen synthesis,
2) down-regulating ER protein levels, and 3) inhibiting estrogen binding.
2.2.3.1. Aromatase Inhibitors (AIs)
AIs work by inhibiting aromatase and thus the formation of estrogens from androgens. These are usually the most clinically useful in post-menopausal women who have failed tamoxifen therapy. However in recent years, AIs have had success as first- line adjuvant therapy for ERα positive breast cancers [196].
2.2.3.2. Selective estrogen receptor down-regulators (SERDs)
One commercially available therapy belongs to this group, fulvestrant. It is referred to as a SERD because it is a pure ER antagonist that binds with a 100-fold greater affinity for ER than does tamoxifen, thereby inhibiting ER dimerization [197].
This decreases overall ER protein levels. Interestingly, tumors resistant to tamoxifen are still many times responsive to fulvestrant treatment [198].
62
2.2.3.3. Selective estrogen receptor modulators (SERMs)
SERMS are so named because of their ability to act as ER antagonists in some tissues (breast) while functioning as agonists (bone and heart) in others. They work in a variety of ways including causing conformational changes in the estrogen receptor that prevent ligand and co-regulatory proteins from binding and affect ER expression and gene activation through non-ERE pathways. Tam is the most commonly used SERM in breast cancer treatment. Within the body, Tam is converted to its predominant active forms which are 4 hydroxytamoxifen (4-OHT) and endoxifen. This SERM then blocks the effects of estrogen in breast cancer cells by competitively interacting with the ER through its LBD and altering the conformation of ER which inhibits co-activator associations. This prevents the ER-dimerization-mediated transcription through EREs of various genes. Interestingly, in recent years Tam has also been used to successfully treat some ER-negative breast tumors [199]. One important risk of tamoxifen treatment is an increase in endometrial cancer, as it acts as an agonist in uterine tissue [200].
Second line SERM therapies include raloxifine and toremifine. Raloxifine is approved for both breast cancer treatment and the treatment and prevention of osteoporosis in postmenopausal women. It differs from Tam in that it is an antagonist of both breast and uterine tissue; thus it does not carry the risk of endometrial cancer [201].
Toremifine is a triphenylethylene derivative is similar to Tam in its mechanism of action and side effect profile; however, it does also not increase the risk of endometrial cancer
[202]. It has been approved by the FDA for restricted use in post-menopausal women with metastatic breast cancer. Information on these drugs’ long-term effects and cancer 63
benefits are limited. Thus, they are still not considered first-line therapy, and the search for the perfect SERM continues.
2.2.3. Tam resistance in breast cancer
The benefits of hormonal therapy have often been limited by resistance to this drug. Approximately one-third of early-stage breast cancer patients will become resistant to Tam over the 5-year treatment period [203], making resistance to Tam treatment one of the major obstacles to the successful treatment of breast cancer. Resistance to tamoxifen can fall into two categories: intrinsic and acquired. A main mechanism of intrinsic resistance to Tam is a lack of ER expression in the tumor. Other mechanisms include inactivating mutations of the cytochrome P450 2D6 allele, which is responsible for the conversion of Tam to its active form [204]. Seeing as only 8% of Caucasian women posses the CYP2D6 mutation and over two-third of breast cancers express ERα, acquired mechanism of resistance have been hypothesized to be responsible for the majority of Tam resistance.
Large microarray studies have already revealed several mechanisms of acquired
Tam resistance, including increased metabolism of Tam [205], loss or alterations of ERα and ERβ expression [206,207,208], estrogen hypersensitivity [209], altered expression of co-regulators [184], and miRNA interference [104].
Tyrosine kinase signaling plays a significant role in tamoxifen resistance due to the ability of signaling cascades to phosphorylate and affect ER actions (Fig. 2.1).
Crosstalk exists between growth factors like IGF, EGFR and ERBB2 (HER2). Pure anti- 64
estrogens abrogate these effects but Tam does not [210]. Increases in all of these pathways confers resistance to Tam, many times by activating Erk or the PI3K signaling cascades [184]. Over-expression and amplification of ERBB2 is a major mechanism of tamoxifen resistance, as is the loss of transcriptional repressors [211]. Src tyrosine kinases, such as SRC, and downstream molecules are also over-expressed in Tam resistance. Focal adhesion molecules like BCAR1 and BCAR3 have been implicated as well [184].
Cell cycle regulators predictably affect Tam resistance due to the cytostatic nature of the drug. Tam induces a G1 phase-specific cell cycle arrest [212]. Molecules involved in this part of the cell cycle are often dysregulated in tamoxifen resistance.
Decreases in CDK inhibitors such as p21 and p27 lead to resistance as does overexpression of cyclins like cyclin D1 and cyclin E1 along with MYC, which can lead to inactivation of the tumor suppressor RB [182].
In areas of breast tumors where Tam concentrations are high, Tam can act as a cytotoxic drug. Apoptosis can be induced by the cellular stress response [213].
Resistance to apoptosis in Tam-treated breast cancer cells and tumors has been shown with increases in anti-apoptotic molecules like BCL-2 and BCL-X along with decreases in expression of pro-apoptotic molecues such as BAK, BIK, and caspase 9 [214]. These could be a result of changes in the tyrosine kinase and growth factor signaling pathways, although NF-κB and unfolded protein response modulator X-box-binding protein 1
(XBP1) appear to play a role in cell survival as well in tamoxifen resistant tumors [215].
65
Truncated versions of ERα exist (ERα36) that have been found to reduce response to endocrine therapy when it is co-expressed with full-length ERα[207] Alternative forms of ERβ, such as ERβ/cx and ERβ5, have proven to be involved in endocrine therapy response as well [208,216]. Estrogen-related receptor EERγ is also over- expressed in Tam-resistant breast cancer [217].
Increased API and NF-KB activity have also been associated with endocrine resistance, which allows for ER to have transcriptional activity without binding to EREs
[218]. Other co-activators like nuclear receptor co-activator 3 (NCOA3, also known as
AIB1 or SRC3) can be over-expressed or phosphorylated to lead to ERα transcription that is associated with resistance to Tam in patients [219].
Taking into account these various Tam resistance mechanisms, medical centers have clinical trials underway with the primary objective of overcoming resistance to
Tam. According to clinicaltrails.gov, several studies are actively recruiting patients or underway in Phase II and III trials to study tamoxifen resistance mechanisms and treatment options. One observational study is examining the serum of patients who are taking tamoxifen for expression of 15 miRNAs both before and after recurrences of breast cancer for comparison with patients who remain sensitive to breast cancer (study
#NCT01612871), while another is examining both protein and genetic biomarkers of the same process (study #NCT00899197). As for therapeutic options, blocking the previously mentioned signaling cascades that are shown to be up-regulated or important in Tam resistance could lead to restoration of sensitivity to Tam. Two studies have been completed which examined combination treatments of Tam and gefitinib – an EGFR
66
tyrosine kinase inhibitor – in patients who have been shown to be Tam resistant
[220,221]. A large Phase II trial showed significant benefit to patients who were still currently ER+/PR+ with a clinical benefit rate (CBR) of 33.3% [220]. New interventional treatment combination studies for Tam resistance include the drugs MK-
0752 (a gamma secretase inhibitor) and Lapatinib (HER2 inhibitor), which are both being used in combination with Tam and other anti-estrogens to determine whether these unique drug combinations can restore sensitivity to Tamoxifen . A mixture of both biomarker prediction studies and treatment options will be necessary to conquer the devastating outcomes of resistance to Tam in breast cancer patients.
2.2.5. Technology used for our study of Tam resistance
While global microarray studies have been performed, some were limited to a chosen set of genes, while others were genome-wide studies that still did not include small RNA analysis and focused instead on the protein-coding genome. In order to improve the effectiveness of Tam therapy, a more comprehensive understanding of the molecular mechanisms and pathways determining Tam sensitivity would help overcome this clinical problem.
In the current study, next generation sequencing (NGS) technology was used to identify the genes and pathways potentially involved in Tam resistance through a global analysis of the transcriptomes in Tam-sensitive (TamS) and Tam-resistant (TamR) breast cancer cells. NGS, or deep sequencing, offers a powerful platform for characterization of altered gene expression, as it allows for a more unbiased exploration of all areas of the 67
genome and transcriptome. RNA-Seq can overcome microarray-associated problems with cross hybridization of similar sequences and allows for single nucleotide resolution, as well as reducing under-representation or the omission of low abundance sequences
[222]. Although one study has come out recently using NGS to explore tamoxifen resistance [223], this investigation used deep sequencing after an initial shRNA screen that contained no small RNA targets and was done to choose the genes to be explored with NGS, thus limiting the power of this state-of-the-art technology to give a less biased view of the transcriptome. While it is recognized that previous biological knowledge can be important in developing some biologically relevant clustering models, new relationships between molecules can be missed by using such a technique. Thus, we present an alternative analytical method.
As the RNA-Seq field is relatively new, analysis models must be tested and compared for their ability to accurately analyze genomic data. Traditional approaches for pattern identification, such as hierarchical clustering or other partitioning methods, are based on cluster analysis for differential gene expression under one specific condition or treatment [224], without considering the mechanisms behind differential expression across environments. These approaches can cluster genes into different groups according to their known functions, but are not able to catalogue genes based on the patterns of how different genes respond to different environmental signals. The difference in expression of the same gene between environments, called phenotypic plasticity, plays an important role in the adaptation of organisms or cells to environmental changes [225,226].
Therefore, we developed an algorithmic model for clustering genes based on
68
environment-dependent differences and ratios by incorporating these measures into a mixture model framework, in which an optimal number of gene clusters can be estimated and the patterns of gene expression plasticity tested. Because of the integration of intrinsic environment-dependent plasticity, results from our model are biologically more relevant than those from traditional clustering approaches using a single environment, which rely on known functional similarities or a predetermined number of gene clusters.
Using this new method, we found that large global changes occur in TamR cells, with differential expression of many genes involved in transcriptional/translational control as well as cell cycle and mitochondrial dysfunction. Through clustering, we identified patterns of differential expression in response to differences between TamS and
TamR cells, with similar functions often clustered together in expression. Through our approach, 1215 mRNA and 513 small RNA (smRNA) transcripts were identified as significantly differentially-expressed, indicating that resistance to Tam is multi-faceted, derived from global changes in gene expression, and involves multiple pathways.
2.3. Rationale
Tamoxifen (Tam) is a commonly used adjuvant therapy for 70% of breast cancer patients that express estrogen receptor [191]. Estrogen receptor signaling can stimulate cancer cells to grow, and blocking this receptor with Tam has significantly decreased breast cancer mortality. However, only about half of patients prescribed the drug respond to tamoxifen [203]. Thus, resistance to tamoxifen therapy is a large clinical problem. 69
New sequencing technology has been developed during the last 10 years that allows for a global look at all of the genetic expression or transcriptome changes that occur with the acquisition of tamoxifen resistance. Examining the transcriptome of tamoxifen-resistant breast cancer with RNA-Seq has identified new potential mechanisms and a number of new genetic alterations in tamoxifen resistance. We developed a new analytical model to cluster genes into groups, and categorizing these genes and clusters has lead us to determine that gene expression is altered on multiple levels in tamoxifen resistance, along with numerous changes in the energy-producing machinery and proliferation signals that determine survival of cancer cells. Our study indicates that targeting singular pathways may not be an effective treatment for overcoming tamoxifen resistance while suggesting new areas to investigate for therapeutic targets and a personalized medicine approach to the treatment of breast cancer.
2.4. Experimental design
2.4.1. Cell lines and reagents
Parental MCF-7 cells were grown in DMEM (Hyclone) supplemented with 5% fetal bovine serum (FBS) (Hyclone), 100 units/mL penicillin, and 100 µg/mL streptomycin. The MTR-3 line (MCF-7 Tamoxifen-Resistant-3) was derived from the parental MCF-7 cells by continuously culturing the cells in the presence of 1 µM Tam
(Sigma Aldrich) in phenol red-free DMEM (Hyclone) supplemented with 5%
70
charcoal/dextran-stripped fetal bovine serum (CSS) (Hyclone) and antibiotics. Estrogen independent cells (MCF-7-E2) were derived from parental MCF-7 cells grown in phenol red-free DMEM (Hyclone) supplemented with 5% CSS (Hyclone) and antibiotics. Cells were maintained at 37ºC in a humidified atmosphere containing 5% CO2/95% air.
2.4.2. RNA preparation
Total RNA was prepared with TRIZOL Reagent (Invitrogen) from MCF-7 and
MTR-3 cells grown under preferred culture conditions as described above. RNA was extracted and isolated as recommended by the manufacturer. Sample integrity was verified with Nanodrop 1000 and Agilent Bioanalyzer 2100.
2.4.3. Library preparation for SOLiD™ NGS sequencing
Library preparation for both whole-transcriptome sequencing and small RNA sequencing was performed using Applied Biosystems Inc’s (ABI) small RNA expression
Kit (SREK) based on SOLiD WT and small RNA sequencing protocols provided by ABI. rRNAs were depleted from total RNA using the Invitrogen Ribominus Eukaryotic Kit
(Life Technologies Corp). 05~1 ug of rRNA-depleted total RNA were fragmented by
RNase III. The fragmented rRNA-depleted total RNA were hybridized and ligated with adaptor mix A from the SOLiD small RNA Expression kit (SREK kit). Next, reverse transcription was performed to generate cDNA templates. The cDNA were size-selected in 100~200 nts from Novex 6% TBE-urea gel (Invitrogen). The excised gel piece containing 100-200 nt DNA was split vertically into 4 pieces using a clean razor blade. 71
The cDNA selected were further amplified using the supplied primer set and ~12–15 cycles of PCR. The purified PCR products as a library, in size ranges of 150~250 bp, and containing 50-150 bp cDNA inserts, quantitated and qualitated by Agilent Bioanalyzer
2100, were prepared for the next steps of emulsion PCR for preparation of template beads.
2.4.4. Library preparation for small RNA sequencing
The sample containing small RNA was hybridized with adaptor mix A provided in the small RNA expression Kit (SREK). The adaptor mixes are sets of RNA/DNA oligonucleotides with a single-stranded degenerate sequence at one end and a defined sequence required for SOLiD sequencing at the other end. Hybridizing and ligating the sample with adaptor mix A sequentially yields the template for SOLiD sequencing from the 5’ ends of the small RNA. The small RNA population with ligated adaptor was reverse transcribed to generate cDNA. To meet the sample quantity requirement for
SOLiD sequencing, and to append the required terminal sequences to each molecule, the cDNA library was amplified using one of the supplied primer set and ~12–15 cycles of
PCR. The individual library PCR products were purified and size-selected in the range of
108–135 bp PCR products containing small RNA of 18–40 nt inserts by electrophoresis on 6% polyacrylamide gel in preparation for the subsequent step of emulsion PCR, in which the molecules were attached to beads.
72
2.4.5. Sequencing
The individual prepared libraries were quantified as templates for emulsion PCR; the template molecules were attached to beads, enriched for adaptor P2, and immobilized to the slide according to the ABI SOLiD emulsion. The library template beads were titrated by Work Flow Analysis (WFA) run to determine the percentage of P2-positive beads in the total template of beads before they were deposited onto slides for sequencing. The sequencing runs were performed on a SOLiD v 3.5 for both WT-seq and small RNA-Seq. The number of P2 positive template beads (equal to the number of transcripts) deposited on the sample slide were 71,250,509 and 69,005,180 of 50 nt length for WT sequencing, and P2 positive beads for small RNA sequencing were
35,686,597 and 35,176,389 of 35 nt length for smRNA sequencing, of TamS and TamR cell lines, respectively
2.4.6. NGS mapping and expression
Fifty bp and thirty-five bp reads (for WT and smRNA, respectively) were assessed for quality and mapped to the reference human genome (hg18) by the software
Maq: Mapping and Assembly of Qualities. Whole transcriptomes for the two cell lines were constructed and compared for their gene expression. One hundred and forty million total reads were produced by sequencing, and ~ 50% of them mapped to the genome after initial quality control measures. Applied Biosystems WT and smRNA Analysis Pipelines were used to confirm results and score expression. Histogram analysis of the log2(# of case reads/# of control reads) provided gene candidates that were differently expressed 73
between the tamoxifen sensitive and resistant cells with a 1.7 fold criterion for the traditional model.
2.4.7. qRT-PCR validation
Potential gene candidates were validated using TaqMan Gene Expression assays. cDNA was made from previously harvested total RNA of MCF-7, MCF-7 estrogen- independent cells, and MTR-3 cells (Roche). The products were tested for purity using spectrophotometry (Aligent Nanodrop). RT-PCR was performed using TaqMan Gene
Expression Assays (Applied Biosystems) on a Statagene Mx3005P (Aligent
Technologies). GAPDH was used to normalize samples for comparison.
2.4.8. Statistical models
We implemented a novel statistical model for identifying the patterns and differences in smRNA and mRNA expression in TamS and TamR cells. Consider m genes are detected in both TamS and TamR cells. Because of their functional similarities and differences, these genes can be clustered into different groups. Let (y1i,y2i) denote the expression data for gene i from these two cell lines, respectively. We can describe the differential expression of gene i using the absolute difference (i.e., yi = y1i – y2i) or ratio
(zi = y2i/y1i) of the gene’s expression between the two cell types. Genes will be clustered into different groups based on the differences and ratios between individual genes in the
TamR and Tams S cell lines using a mixture-based likelihood model:
74
n (1) L( | y) 1 p1(yi ) ... J pJ (yi ) for the difference model i1
n (2) L( | z) 1h1(zi ) ...LhL (zi ) for the ratio model i1
where (1,..., J ) and (1,...,L ) are a set of proportions that each correspond to a different gene group under the difference and ratio model, respectively; pj(yi) and hl(zi) are the discrete probability distributions of differential expression for group j (j = 1, …, J) for the difference model and group l (l = 1, …, L) for the ratio model. The expression reads of genes in each cell type are thought to obey a Poisson distribution (16), thus the distribution of the read differences and ratios between the two cell types is modeled by specific functions.
The EM (expectation maximization) algorithm can be conveniently used to estimate the means of gene expression in TamS (Sj) and TamR cells (Rj) for group j under the difference model. Similarly, the means of gene expression in TamS (Sl) and
TamR cells (Rl) for group l can also be estimated. The optimal number of clusters is determined by a model selection criterion, such as commonly used Akaike information criterion (AIC) [227] or Bayesian information criterion (BIC) [228]. In this article, both the AIC and BIC values under different numbers of clusters were calculated to be the same; thus, only the AIC values were reported. The optimal number of clusters corresponds to the minimum AIC value. After this is determined, our model allows the following biologically meaningful tests.
75
Test 1: For a given cluster group, do TamS and TamR cells differ? This can be done by testing:
H0: µSj = Rj (j = 1, …, J) for the difference model
H0: µSl = Rl (l = 1, …, L) for the ratio model
If the H0 is rejected, this group of genes is expressed differently between TamS and TamR cells, indicating that they may be involved in drug resistance and can be viewed as a biomarker of drug response;
Test 2: For a pair of genes, do they interact with each other to determine drug resistance? This can be done by testing:
H0: (j = 1, …, J) for the difference model Sj1 Sj2 Rj1 Rj 2
H0: (l = 1, …, L) for the ratio model Sl1 Sl2 Rl1 Rl 2
A rejection means that these two groups of genes have significant interaction effects on drug response.
2.4.9. Expression Analysis
Gene network and pathway analyses were conducted using the Ingenuity Pathway
Analysis (IPA, Ingenuity® Systems) and GeneGO (Thomson Reuters) software.
Functional analysis of the resistant cell lines was performed using IPA with a 1.7-fold change criteria and a P value of <0.01.
76
2.4.10. SIRT3 expression and cell growth assays
SIRT3 protein levels were determined by Western blot with SIRT3 antibodies from Cell Signaling. MTR-3 cells were transiently transfected with SIRT3 siRNA
(Sigma) using Invitrogen Lipofectamine RNAiMax Transfection reagent according to the manufacturer’s instructions. MCF-7 cells were transiently transfected with SIRT3 over- expression vector (pcDNA3.1+) generously donated by Dr. Chu-Xia Deng at NIH with
FuGENE 6 transfection reagent (Roche) according to the manufacturer’s instructions.
Cells were then exposed to escalating doses of Tam for a period of 5 days. Cell growth was determined by Sulforhodamine B (Sigma In Vitro Toxicology assay kit) according to previous study methods [229].
2.5. Results and Discussion
2.5.1. Clustering of gene expression data
A new algorithmic model was created to cluster genes based on environment- dependent differences and ratios. This incorporation of intrinsic environment-dependent plasticity fully takes into account the differences in environment or treatment conditions to produce more biologically relevant clustering results.
77
2.5.1.1. Rationale behind mathematical model
When microarrays were first developed, a z-test or Poisson distribution was commonly used to determine differential gene expression, where an expected distribution of gene expression levels is compared to the actual data. This results in the standard method of genes with expression levels that are at least 2 standard deviations away from the average being deemed as differentially expressed and not just the result of chance.
This method uses a log2 ratio data of normalized read counts. However, the z-test is easily influenced by sample size. Too large of samples can artificially inflate the number of differentially expressed gens while too small a sample size may over-estimate what the standard deviation actually is [230].
By comparison, the Fisher’s Exact Test (FET) is often used for small sample sizes as it is usually considered the most powerful and unbiased test for small sample sizes; it has no minimal sample size. It takes into account the exact probability value rather than the just the approximate value as does the Z-test. Thus, it does not rely on normality; it uses an exact distribution of the data rather than an “assumed” normality. Since sequencing data for a transcript obtained with NGS read as counts, Fisher’s exact test
(FET) was typically used to detect the difference of gene expression across replicates
[224]. We compared this FET method to the Poisson determination to examine their influence on the selection of differentially expressed genes [230].
Traditional approaches for pattern identification are based on cluster analysis for gene expression in one replicate [224], without considering the mechanisms behind differential expression. The differential expression of a gene across replicates can be 78
described as the difference or ratio of expression values between the replicates. We develop a mixture model for clustering genes into distinct groups based on the differences and ratios calculated.
2.5.1.2. Validation and comparison of gene expression levels between Tam-sensitive and Tam-resistant breast cancer cells.
In order to reveal the potential genes and mechanisms involved in resistance to
Tam, we used a NGS approach with ABI SOLiD3 technology as a means of examining and comparing the transcriptomes of TamS and TamR breast cancer cell lines. These cell lines were previously characterized for tamoxifen resistance [229,231], which was confirmed before sequencing. Experimental procedures are summarized in Figure 2.2A.
A total of 71,250,509 and 69,005,180 reads, for TamS and TamR cells respectively, were sequenced. Gene expression of parental MCF-7 (TamS) cells was used as a baseline for up- or down-regulation of expression in TamR cells. Gene expression data by RNA-Seq are generally thought to follow a Poisson distribution [232]. To check whether our data are Poisson-distributed, we calculated chi-square goodness of fit test statistics for read counts observed in TamS and TamR cell lines, respectively. The calculated test statistics by assuming the Poisson distribution are smaller than critical thresholds, suggesting that these RNA reads obey a Poisson distribution (P > 0.90). Based on this two-standard deviation criterion of mRNA expression which indeed followed a Poisson distribution
(Fig. 2.2B), we found that 667 mRNAs were significantly differentially-expressed between the TamS and TamR cell lines.
79
Figure 2.2. NGS identification and comparison of differentially-expressed genes in TamR cells by Fisher’s exact test. (A) Total RNA from human breast cancer cell lines MCF-7 (TamS) and MTR-3 (TamR) were collected and subjected to the next generation sequencing process. (B) Gene expression followed a Poisson distribution with significantly differentially-expressed genes two standard deviations from the mean in the traditional method. Fisher’s exact test was used to test the significance of differential 80
expression of genes between two treatments. The change of the normalized smRNA exon reads (C → B) and intron reads (D → C), and exon reads for mRNA genes (E → D) from TamS to TamR cells is plotted against the mean expression between these two types of cells for the new method. Purple dots represent significantly expressed genes as determined by FET; gray dots represent genes with similar expression. The red horizontal line at zero provides visualization for the signs of differential expression.
To better analyze and categorize the transcriptome differences associated with
Tam resistance, including analysis of smRNA, we used the Fisher’s Exact Test (FET), in which significance was assessed with the normalized data by FPKM (fragments per kilobase of exon per million fragments mapped). This allows for analysis of smRNA
(which may map to unidentified genome regions with no recognized gene lengths) in addition to mRNA and more accurately deals with variation between different treatments or cell lines [233]. FET was therefore also used to analyze the significance of differential expression between the TamS and TamR cells for each gene, a method which has recently gained favor in microarray analysis [234]. Among a total of 7713 small RNAs,
513 display significant differences in exon reads (Fig. 2.2C) between the two cell types.
For intron reads, 55 smRNAs were differentially-expressed (Fig. 2.2D). From a total of
23,561 mRNA genes, 1215 were differentially-expressed (870 up-regulated and 335 down-regulated) between the TamR and TamS cells (Fig. 2.2E). Interestingly, upon comparison of the mRNA expression, only 150 genes were found by both the “two- standard deviation” method and FET (Fig. 2.3A). Table 2.1 lists the most differentially- expressed genes found by both tests.
81
Table 2.1: Significant genes found by both methods
Gene Function Fold-change P-value
Up-regulated Genes in TamR cells ANKRD32 Ankyrin repeat domain: cell-cell adhesion and cell 12.25 0.015 structure ABHD10 Alpha-beta hydrolase 11.81 0.002
INTS12 Integrator complex subunit: associates with RNA 9.19 2.00E-04 polymerase II SIRT3 Sirtuin 3: deacetylase 9.19 .015
TATDN1 Putative deoxyribonuclease: alternative splicing 8.75 7.65E-05
UBC Ubiquitin C: ubiquitination 8.75 0.008
CAV2 Caveolin-2: formation of caveolae 8.31 0.050
ATP5E ATP synthase: oxidative phosphorylation 7.61 2.48E-81
HIST1H2BM Histone 1: gene expression 7.44 4.31E-09
RAB27B Ras oncogene: vesicular fusion and trafficking 7.00 0.003
Down-regulated Genes in TamR cells RPLP1 60S ribosomal protein: translation -22.86 1.42E-12
SLC12A9 Solute carrier: membrane transport -18.29 0.034
REEP6 Receptor accessory: cell surface receptor expression -11.43 0.001
IFITM2 Interferon induced transmembrane protein: cell cycle -11.43 4.5E-05 arrest and apoptosis NDUFS6 NADH dehydrogenase: oxidative phosphorylation -9.14 .001
TSSC4 Tumor suppressing subtransferable: -8.00 0.016
TMSB15B Thymosin β: actin binding -6.86 0.027
HIST1H3E Histone 1: gene expression -6.86 0.003
CSNK2A2 Casein kinase: PI3K and Wnt signaling -6.10 0.016
ATP6V0E2 ATP synthase: oxidative phosphorylation -5.94 0.007
82
Figure 2.3: Comparison and validation of differentially-regulated genes by the two significance methods. (A) Venn diagram of overlap of significant genes found by simple calculation of two standard deviations or the Fisher’s exact test. (B) Validation of mRNA levels of selected genes found by NGS was performed on MCF-7 (TamS), estrogen independent TamS cells (MCF-7-E2), and MTR-3 (TamR) cells by qRT-PCR. The log ratio of MCF-7-E2 or TamR to TamS gene expression is shown to indicate up- or down- regulation. GAPDH was used as a control. Each point represents mean ± S.D. of triplicate 83
determinations; results shown are the representative of three identical experiments. *p<0.05; t-test. (C) GeneGO (Thomson Reuters) network analysis of most significant networks dysregulated in TamR cells. Red circles with a red dot in the middle next to the proteins indicate up-regulation in TamR cells. The different shapes indicate different classes of proteins. Green lines indicate activation while red lines indicate inhibition; gray lines are unspecified interactions.
For preliminary verification of differential expression between the TamS and
TamR cell lines, we chose ten genes found by both statistical tests (five of which were up-regulated and five down-regulated in TamR cells) and compared their mRNA levels using quantitative RT-PCR. An additional treatment group of TamS cells grown in phenol red-free media, which acts as an estrogen mimic [235], was added to explore the effects of estrogen independence on the gene expression changes. Three replicates from cell culture experiments were prepared on three separate days that were distinct from those used for NGS. Figure 2.3B shows the mRNA levels of the selected genes as determined by qRT-PCR. The qRT-PCR confirmed the general up- or down-regulation of the genes. Quantitatively, the fold-changes observed were usually smaller in the qPCR analyses than the NGS by approximately 2-fold. The down-regulated genes in
TamR cells, GTSE1,IFITM2, and mir-1974 showed a 6-fold difference by NGS but only a 2 to 4-fold difference by qRT-PCR, while genes CCDN1 and U2AF1 showed a more moderate decrease of a 2-fold difference which was similar to their 1.7- difference found by NGS. Although all the down-regulated genes were more down-regulated in TamR cells than in TamS cells grown without estrogen, it was interesting that JUNB and mir-
1974 trended towards an up-regulation under estrogen independent conditions, suggesting a distinct mechanism for the emergence of tamoxifen resistance. TamR upregulated genes, ATP5E, CCDN1, SIRT3, UBC, and mir-21,a, showed a 7-9 fold difference by
84
NGS but only a 2-4 fold difference by qRT-PCR. While ATP5E, SIRT3, and mir-21 all had increased expression levels under estrogen independent conditions, all the up- regulated genes were increased further when the cells were tamoxifen resistant. Thus, while some of the validated genes have altered expression as they become estrogen independent, further alterations in expression appear to be necessary for the development of resistance to tamoxifen. An initial ontological exploration of both methods’ sets of statistically significant genes indicated that genes related to ESR1 (estrogen receptor alpha) comprised the most enriched pathways (Fig. 2.3C). This validates the significance of our data set in comparison to previous studies [182,184,236].
2.5.1.3. Phenotypic plasticity clustering analysis.
Due to the large number of differentially-expressed genes, we next sought a method to categorize the genes based on their levels of differential expression to determine if any new patterns emerged. Because many traditional clustering methods create clusters based on known gene function similarities, they fail to recognize novel patterns of gene expression. Other methods that do not rely on gene function are usually limited because they force genes to fit into one of a predetermined number of gene clusters that can create false relationships between genes. To overcome these limitations, we developed difference and ratio models (see the Methods) that take into account phenotypic plasticity of gene expression and cluster the FET significant genes into different groups based on the pattern of differential expression between TamS and TamR cells (Figs. 2.4-2.5). Phenotypic plasticity of gene expression can be measured as
85
absolute differences or ratios of expression levels between environments or treatments.
The difference model determines the assignment of genes to particular clusters based on absolute differences in gene expression levels from one environment (TamS) to the next
(TamR), whereas the ratio model identifies expression patterns according to relative difference of gene expression. The optimal number of clusters is determined by a model selection criterion, such as the commonly used Akaike information criterion (AIC)[227] or Bayesian information criterion (BIC) [228]. As is common, BIC here provides a consistent result with AIC. Therefore, we only report the results from AIC to avoid confusion and redundancy. In this study, the AIC values under different numbers of clusters are calculated, with an optimal number of clusters corresponding to the minimum
AIC value.. The two models for absolute difference and ratio of expression may produce similar results, but meanwhile, they are complementary in identifying particular clusters.
Detailed method and validation is unpublished as of yet.
The AIC criterion calculated from the difference model favors the choice of five clusters for the 513 exon genes and four clusters for 55 intron genes of the differentially- expressed small RNA genes. Figures 2.4A and 2.4B plot the patterns of absolute difference in smRNA gene expression in the TamS and TamR cells, showing marked differences in the pattern of differential expression. The majority of exon genes fall into
Cluster 3 which represents low expression genes (Fig. 2.4A). It should be pointed out that for these weakly expressed clusters in both cell types, some sub-clusters
86
Figure 2.4: Clustering patterns of genes by absolute difference and ratio of expression. Clustering as determined by the difference model for smRNA exon reads (A) and intron reads (B), as well as mRNA genes for exon reads (C) in TamS (S) and TamR (R) cells. Clustering as determined by the ratio model for smRNA exon reads (D) and intron reads (E), as well as mRNA exon reads (F) in TamS (S) and TamR (R) cells. The number in parentheses corresponds to the number of genes in each cluster. 87
Figure 2.5: Heatmap comparison of differentially-expressed genes by clustering analysis. Heatmaps showing results of the clustering of small RNA exons (A) and introns (B), as well as mRNA exons (C) absolute difference gene expression (R-S) between TamS (S) and TamR (R). Heatmaps showing results of the clustering of small RNA exons (D) and introns (E), as well as mRNA exons (F) ratio gene expression (R/S) between TamS (S) and TamR (R). Gene expression levels are displayed for R and S on a log(absolute values) scale. (R-S) are absolute values while (R/S) values display a fold-change from R to S cells. Clustering groups are represented by different colors above the heatmaps. P-values were calculated using a χ-squared test. 88
may exist in terms of the relative difference which would be found with the ratio model.
In general, the counts of intron reads are strikingly low compared with exon reads (Fig.
2.4B). Distinct patterns of absolute difference in gene expression can also be detected for total exon reads for mRNA. Introns were not included for mRNA analysis as they do not accurately portray the genes being expressed. Among 1215 significant mRNA genes, we detected three clusters based on the AIC criterion (Fig. 2.4C), with the majority of genes falling into the low expression Cluster 2 with little absolute difference between the cell types. Overall, these results suggest that the difference model is effective for large differences in gene expression, but genes that have low expression could be inaccurately categorized as having no change between treatments.
The ratio model was better able to cluster genes together that had lower absolute expression but a high degree of difference in expression between TamS and TamR cells
(Fig. 2.4D-F); up- and down-regulation is more evident in this format. In this model, fewer genes were clustered due to some genes only being expressed in one cell line. For smRNA gene expression, the model found four clusters for the exon-significant genes
(Fig. 2.4D), while the absolute difference method found five (Fig. 2.4A). Intron-gene expression was clustered into three groups (Fig. 2.4E). For the differentially-expressed mRNA genes, the genes clustered into three groups again (Fig. 2.4F). As expected, the ratio model was better able to capture the nuances of the fold-changes than the absolute difference method, but the absolute difference method was superior at clustering genes simply by their absolute low or high levels of expression.
89
Taken together, this model was able to cluster differentially-expressed genes into groups with similar degrees of expression differences. With the model accurately taking into account the statistical ramifications of comparing across different environmental groups rather than just across multiple samples of the same treatment, our next goal was to determine if the genes within clusters have any significant known relationships to one another.
2.5.1.4. Effects of Tam resistance on smRNA expression and clustering.
In order to better understand how different types of smRNA were affected in
TamR cells, we next examined the smRNA clusters. Based on clustering by absolute difference, almost all genes were designated to a single cluster (Cluster 3) in both the exon and intron analyses. These clusters were low expression genes that showed little difference between TamS and TamR cells when measured on an absolute expression scale (Fig. 2.4A-B). The majority of these differentially-expressed small RNA aligned to known small nucleolar RNA (snoRNA) genes as well as other non-coding RNA (ncRNA) regions. snoRNAs were both up- and down-regulated in TamR cells. This relatively new category of non-coding RNA was originally thought to be unimportant or to only have effects on the chemical modifications of other RNA molecules [237]. However, there is recent evidence showing that snoRNAs can act much in the same way as micro-RNA
(miRNA), regulating gene expression [238]. Other ncRNA categories included those related to histone modification, small cajal nucleolar bodies, and vault RNA, with one notable exception: miRNA mir-16-2 was found in Cluster 3. This miRNA normally stops
90
E2F control of proliferation [239] and its down-regulation would allow proliferation to continue. smRNA exon Clusters 1, 2, and 5 contained only a few transcripts that were differentially-expressed, and all aligned sequences were mapped to snoRNA genes. The remaining group, Cluster 4, which contains moderately-expressed genes with little absolute difference but significant fold-change between TamS and TamR, did include one interesting transcript – RMRP (RNA component of mitochondrial processing endonuclease), a ncRNA that binds several proteins to create the endonuclease complex controlling mitochondrial transcription.
Comparison of the ratios of smRNA expression revealed much of the same alteration of snoRNA as well as other significant ncRNA. Most differentially-expressed smRNA that were not labeled as snoRNA were mapped to regions that were generally categorized as nonspecific ncRNA, open reading frames, and transcription regulation.
However, there were some significant changes in miRNA. In Cluster 4, we found that known oncomir mir-21 expression was increased in TamR cells by ~ 5-fold, as was uncharacterized mir-1259. Up-regulation of mir-93 and mir-125A, which are involved in invasion, migration and metastasis [240,241], was observed in Cluster 4. Cluster 2 contained newly discovered mir-1974, a mitochondrialy-targeted miRNA [242] found to be decreased in adrenocortical carcinoma [243]. In addition to these specific miRNAs, other areas of smRNA dysregulation include transcripts that lead to alteration of transcription by modification of histone acetylation and methylation proteins. smRNA from several histone-associated proteins like Histone 1 complexes A-D as well as histone acetyltransferases (MYST4) and methylators (MBD1) were only found in TamR cells, and
91
thus were not included in the clustering analysis in the ratio setting. However, such binary “on/off” expression suggests a strong role in mediating drug resistance.
In general, smRNA analysis of TamR and TamS breast cancer cells illuminated large alterations of snoRNA levels. This study provides support for the exploration of snoRNA in the cancer genome and drug resistance phenotype. Clustering analysis did not appear to cluster genes based on function, but analysis is restricted by the limited characterization of snoRNA and other ncRNA. As the field progresses, these snoRNA may be better categorized and the significance of the clusters may become apparent.
Additional limitations of the smRNA analysis lie in the fact that some transcripts aligned to protein-coding exons of genes. While many of these genes may be subject to alternative splicing leading to the creation of smRNAs, the actual function of these smRNAs could be unrelated to the function of the gene. For this reason, we did not include analyses with these alignments. Overall, the existence of so many snoRNAs, miRNAs, and smRNA transcripts related to gene expression (histone modification, mitochondrial transcription, etc.) implicate the intricate regulation of a large set of gene expression changes in the development of Tam resistance.
2.5.1.5. Gene ontology and clustering analysis of mRNA expression.
To better understand the wide-range of altered mRNA transcripts and proteins in
TamR breast cancer cells, we performed a gene ontology and pathway analysis of the differentially-expressed genes and clusters. Analysis was relegated to exon-significant mRNA transcripts since these can be verified as protein coding regions.
92
Table 2.2: mRNA exon gene clusters Category Function Molecules Cluster 1 (down-regulated) Mitochondria ATP synthases ATP: 5J2, 6V0E2 Gene expression Ribome 60S RPL: 17, 27, 28, 35, 39, 41 , P0, P1 Splicesome U2AF: 1, 2 Transcription factor JUNB Histone-associated HIST1H: 1C, 2AE, 2BD, 2BO, 3E, 4A, 4D Cluster 2 (moderately up-regulated) Mitochondria ATP synthases ATP: 1F1, 5A1, 5B, 5I, 5O, 6VOE1,6VOE1, 8B1, Gene expression Histone-associated HIST1H: 2AC, 2AM, 3F, 3J, 4H, HIST2H: 2AB, 2AC Histone-binding HINT1 Ribosome40S RPS: 4X, 5, 6, 8, 21, 23, 24, 25,27 Ribosome 60S RPL: 3, 5,10A, 11, 13A, 23, 30, 36, 37, 38, P2 Initiation factors eIF: 2A, 3E, 3H, 3M, 4A1, 4G2, 5, 6 Elongation factors eEF1E1 Proteosome PSM: A1, A2, A4, A5, A7, B1, B2, C2, D6, D7, D10, D12, G3 Mitochondria NADH NDUF: A1, A4, A6, B2, C1, S3, S4 dehydrogenases Cytochrome c COX: 6C, 7A2L, 7B, 7C,16 Mitochondrial MRPL: 16, 27, 32, 39, 47, 50, 53 MRPS: ribosome proteins 7, 17, 21, 22, 23 Cell cycle Cyclins CDK1, CDKN3, CCNB1, CCNC Retinoblastoma RB1 Cluster 3 (highly up-regulated) Mitochondria ATP synthases ATP: 5E, 6V1D, 6V1H Estrogen receptor ESR1 pathway CCNC, HRAS, MAPK1, NCOA, NRAS, NRIP1, PHB2, SRA1, TAF7 Gene expression HNF4a targets ABD10, DPH5, NOP6, E2F Multi-drug Caveolins CAV2 resistance 93
Using the absolute difference method, the majority of genes fell into one cluster,
Cluster 2. Figure 2.4C shows that this cluster contains genes with low levels of expression and little absolute difference in gene expression. This is to be expected, as most transcripts are not highly expressed. Clusters 1 and 3 contain transcripts mostly from snoRNA regions, with varying levels of up- and down-regulation in TamR cells.
While Cluster 2 contains snoRNA transcripts as well, it also includes miRNAs mir1248, mir1291, and mir1978, which are slightly up-regulated with moderate absolute expression levels in TamR cells. So far, these miRNAs have not been associated with any disease state. Transcripts for the non-protein coding RMRP, that was also found by smRNA analysis, clustered in this group as well. The assignment of RMRP and snoRNA to both smRNA and mRNA is unsurprising due to their intermediary sizes before processing ranging from 60 - 350 bp. Transcripts from the mRNA analysis designated as miRNA are probably the result of unprocessed transcripts or previously named miRNAs being assigned to areas of alternative splicing of unknown genes.
The clustering analysis gave a more substantial set of results for the mRNA transcripts using the ratio method that compares the relative difference of individual genes expression from TamS to TamR cells (Table 2.2). Of the three clustering groups,
Cluster 1 contains all of the down-regulated genes in TamR cells. Analysis of the biological functions and pathways contained in Cluster 1 genes indicates a high level of modification of mitochondrial oxidative phosphorylation and gene expression regulation.
Some of the ATP synthase genes are down-regulated by 2-fold, as are ribosomal proteins
60S proteins. Transcripts from splicesome genes U2AF1 and U2AF2 are decreased by 5-
94
and 2-fold, respectively, in TamR cells. Expression of the transcription factor JUNB, which binds and represses AP-1, was decreased by 5.5-fold, which has previously been shown to be linked to increased cell cycle progression and lack of response to Tam [244].
In general, the down-regulation of Cluster 1 may be required for specific gene expression changes that allow Tam resistance to occur. Changes (up or down) in energy metabolism molecules have been observed previously with Tam treatment [245] and might be necessary for altered global gene expression.
Cluster 2 exhibits a more moderate increase in gene expression of TamR cells, many of which are related to expression of transcripts and proteins. Pathway and function analysis shows this cluster to have the most diverse set of gene functions with alterations in mitochondria, transcription, translation, cell cycle, and ubiquitination.
Transcription regulation is altered with a number of histone-associated genes that are up- regulated 2- to 3-fold as well as histone binding protein HINT1. Protein synthesis is affected on several levels. Ribosomal transcripts that code for proteins rather than rRNA were also increased such as 40S (RPSs: ribosomal protein S) small subunit and 60S
(RPLs: ribosomal protein L) large subunit ribosomal proteins; interestingly, mitochondrial ribosomal proteins (MRPLs and MPRSs) were also increased. Another level of protein regulation was found with increases in translational machinery including up-regulation of initiation factors (eIFs), as well as increased elongation factor eEF1E1.
Finally, post-translation modification is also up-regulated with an increase in proteosomal
PSMs (proteosome/macropain) subunits.
95
The TamR breast cancer cells in Cluster 2 are also characterized by expression of cell cycle and mitochondrial energy metabolism genes. Molecules involved in the progression of cell cycle are moderately up-regulated in TamR cells. Cyclin D1 kinase
(CDK1) and Cyclin D3 kinase inhibitor (CDKN3) are increased as are Cyclin B (CCNB1) and Cyclin C (CCNC). Master regulator RB1 (retinoblastoma 1) is increased as well.
We also found a 2-fold increase in various E2F5 transcripts in TamR cells, as well as a decrease in mir-16-2, an E2F negative regulator miRNA which stops E2F1 control of proliferation [239]. The increase in E2F transcripts is probably partially due to the activation of HNF4a as they are known targets of the transcription factor. Multiple components of mitochondria are altered as well. In addition to the increase in mitochondrial specific ribosome proteins, NADH dehydrogenase subunits (NDUFs) are increased as are cytochrome c oxidases (COXs). Drug resistance has been previously linked to changes in cell cycle [246] and oxidative phosphorylation with a decreased use of glycolysis [246,247].
Cluster 3 genes, which were highly up-regulated in TamR cells, also contained transcripts related to dysfunctional mitochondria and oxidative phosphorylation, in addition to those related to proliferation and drug resistance. Genes from the ESR1 pathway were also increased including downstream proliferation activators and nuclear receptor regulators. These genes include HRAS, MAPK1, NCOA, NRAS, NRIP1, SRA1, and TAF7. Different ATP synthases were affected, increasing 4- to 7-fold in TamR cells.
Interestingly, activation of transcription factor HNF4a (hepatocyte nuclear factor 4, alpha) genes was found with increases in targets without an increase in HNF4a mRNA.
96
HNF4a has not previously been associated with breast cancer, but has been linked to types of ovarian and liver cancer [248] in addition to changing the expression of drug metabolism enzymes [249]. Caveolin-2 (CAV2), which is involved in creating caveolae or invaginations of the cell membrane, was also increased ~ 8-fold in Tam resistant cells.
CAV2 expression is associated with poor prognosis in breast cancer patients [250] and with multi-drug resistance in multiple cancer types [251]. Overall, Cluster 3 contains many of the traditional molecules associated with Tam resistance, including those related to ESR1 along with multi-drug resistance molecules.
Taken as a whole, while all three clusters contain many of the same type of genes
– mitochondrial and those related to gene expression – Cluster 2 stands apart with its inclusion of additional categories of modifications. Specific regulation of gene expression with histone modification, translation factors, and proteasome components is found exclusively in this cluster as are oxidative phosphorylation members relating to
NADH dehydrogenases and cytochrome c. The designation of a variety of gene types to
Cluster 2 is unsurprising since this cluster represents moderately altered genes, and most genes would be expected to only have moderate expression changes rather than dramatic ones.
2.5.2. Comparison to traditional analysis methods and previous studies
A comparison of the ontology of genes found to be differentially-expressed between TamS and TamR cells validated the results of the NGS study. Numerous previous investigations have explored the mechanism of Tam resistance, from large 97
microarray [252] and shRNA [223] screens to focused mechanism studies. These studies have linked pathways related to estrogen receptor to be of great importance to Tam resistance, which was confirmed by our study. Both the FET and two-standard deviation significance test showed that molecules related to ER-α (ESR1) pathways were found to be the most enriched genes as analyzed by GeneGo network and pathway analysis (Fig.
2.3C), such as DDX5, MAPK, and NRIP genes.
The other pathways implicated by the traditional two-standard deviation method are also in agreement with previous studies, including down-regulation of cell death/apoptosis molecules [253] and up-regulation of cell cycle regulators [254] and metabolic genes [255]. Tam resistant cell lines and tumors are known to have dysregulated cell cycle and apoptotic pathways in an attempt to survive long-term treatment and to overcome the cytostatic effects of Tam. One of our qRT-PCR genes confirmed to be up-regulated in TamR cells, the gene for Cyclin D1, has been shown to be increased in the plasma of breast cancer patients that have poor outcomes and are non- responsive to tamoxifen [256]. Although the FET analysis found many of the same metabolic and cell cycle regulators, apoptotic and cell death regulators were not among the most prominent molecules found by our new method, which may indicate that a variety of analytical methods should be used when exploring RNA-Seq data. While in agreement with the traditional ontology analysis, the expanded study, which uses the FET significance test and our model, gave a broader picture of the vast changes between these two cell lines. The ability of the TamR cells to change global expression of so many genes is highlighted by the amount of alteration at all levels of transcriptional and
98
translational control including changes in epigenetic regulation, transcription factors, and post-translational regulation. smRNA is altered in both miRNA and snoRNA forms, emphasizing the complexity and dysregulation of smRNA in Tam resistant breast cancer.
Previously known miRNAs were also implicated, with increases in mir-21 expression which targets tumor suppressors [257], as well as up-regulation of mir-93 and mir-125A
[240,241], which are both involved in the development of more aggressive phenotypes capable of migration and metastasis. These results together represent a global change in the way that genes are transcribed and expressed.
Additionally, along with traditional mechanisms of Tam resistance such as up- regulation of ESR1 and other proliferative pathways, alterations in mitochondrial function occurred in TamR cells. This is not unexpected as new modes of energy metabolism would be necessary to have a global control of gene expression, and glycolysis may no longer provide adequate energy supplies. Altered oxidative phosphorylation has been linked previously to increased drug resistance [247].
Several large microarray studies on patient breast cancer tumors were previously reported which presented gene signatures associated with tamoxifen resistance and breast cancer recurrence. While these studies were limited by the fact that they did not explore smRNA expression, they did look at the prognostic value of gene expression signatures of tamoxifen resistant tumors. When the microarray gene signatures from Loi et al.[258], Jansen et al [259], and Ma et al [260] were compared to our differentially expressed genes, relatively few genes overlapped. Out of the 181 transcripts reported as differentially expressed by Jansen et al., 173 could be matched confidently to our data.
99
Of the 173 matched we observed 17 changed in expression. A similar 10% overlap in genes was found in our comparison with Loi et al where we confidently matched 173 out of 181 genes, while 17 out of these 173 changed expression in our dataset. We had no overlap with Ma et al. It should be noted that the gene signatures of these three previous microarray studies do not have a single gene that overlaps with one another, despite the studies using some of the same patient tumor datasets. In our case, this could be due to the difference in using a different technology. Microarrays measure one part of the gene, which are usually 3' biased. With the sequencing approach, reads are measured across the gene. The differences could be attributed to the fact that these different methods are not analyzing the exact same thing. In general, there have been many challenges with the reproducibility of replication studies [261].
The diversity of molecules involved in Tam resistance was also established in a recent meta-analysis of three separate microarray studies [262]. The systematic study by
Huang et al. examined the 275, 130, and 252 genes found in three public microarray data sets (GSE6532, GSE9195 and GSE9893), respectively, comparing Tam-sensitive and
Tam-resistant breast cancer samples. While the authors found little overlap in the actual genes between datasets, they did find a general theme of cell cycle and proliferation transcription factor over-expression including an increase in activation of various E2F’s in tamoxifen resistant cells in all three studies. In fact, E2F gene expression was the only common molecules between all three studies. They concluded that Tam-resistant cells were highly proliferative compared to their sensitive counterpart, a finding that is corroborated by our current study. Specifically, we also found an increase in E2F5
100
Figure 2.6: Dysregulation of pathways and processes involved in Tam resistance as revealed by NGS. Changes in E2F control of proliferation are in agreement with previous clinical sample studies with increases in E2F5 (A) and mir16-2 (B) expression in TamR cells. (C) Pathway analysis of clusters revealed several important areas of dysregulation in Tam resistance: traditional Tam resistant ESR1 (1) and proliferation (2) pathways are up-regulated in TamR cells, as are molecules involved in cell cycle progression (3). Oxidative phosphorylation is altered (4). Transcription was affected with modification of histone and transcription factor expression (5). Expression of transcripts was altered by the large number of smRNA molecules that were dysregulated, particularly in snoRNA (6) and miRNA (7) expression. Translation of proteins is affected in Tam resistance as well with up-regulation of ribosomal and translational machinery (8). Protein expression was also affected by an up-regulation of proteosomal proteins in TamR cells (9). 101
transcripts in TamR cells (Fig. 2.6A), as well as a decrease in general E2F negative regulator mir-16-2 (Fig. 2.6B). Thus, while our study was performed in breast cancer cell lines, these findings support the validity of our method and its significance for clinical cases. Targeting of E2Fs may be a promising area for the development of adjuvant therapies that may sensitize breast cancer cells to Tam treatment.
2.5.3. Effects of SIRT3 expression on Tam resistance
As our qRT-PCR results indicated, SIRT3 mRNA levels were increased in TamR cells. This mitochondrial deacetylase has been shown to be up-regulated in node-positive breast cancer [263]. Further investigation of SIRT3 protein levels revealed that SIRT3 protein levels were also increased in TamR cells by approximately 3-fold (Fig 2.7A).
This led us to examine if modulation of SIRT3 levels would have an effect on Tam resistance.
TamR cells transfected with SIRT3 siRNA and then treated with Tam show an increased sensitivity to Tam as indicated by a Tam IC50 decrease of ~ 50% in TamR cells
(Fig. 2.7B). The reciprocal experiment was performed on TamS cells that were transfected with a SIRT3 over-expression vector and treated with Tam. While TamS cells were still sensitive to Tam treatment, the IC50 concentration rose by 5-fold (Fig.
2.7B). Protein levels of SIRT3 in TamS cells increased as cells were exposed to 1 µM of
Tam over a 5 day period (Fig. 2.7C). Together, these results indicate that SIRT3 can modulate resistance to Tam and increases in SIRT3 protein levels are an early indicator
102
103
of exposure to Tam. Further experiments are needed to explore the mechanism by which
SIRT3 affects Tam resistance.
2.6. Conclusions
Our study highlights the ability of NGS to profile and characterize transcriptome changes in Tam-resistant breast cancer. RNA-Seq analysis of gene expression of Tam- sensitive and Tam-resistant breast cancer cells led to the identification of 1215 mRNA and 513 smRNA transcripts that were differentially-expressed. Validation of the ability of NGS to discover experimentally relevant findings was exemplified by the results of
SIRT3 modulation of Tam resistance. The sheer number of differentially-expressed genes demonstrates – quite effectively – that resistance to Tam is not only through changes in an individual molecule or pathway, but is the result of global changes in gene expression (Figs. 2.5-2.6).
RNA-Seq and NGS allow for an unbiased search for these pivotal transcriptome modifications. Regardless of these advantages, use of NGS has been limited by the lack of suitable analysis tools for the large amount of data generated. Our clustering method will add a means of determining the significance of similar levels of expression changes as found by NGS, which may, in due course, lead to the determination of molecules that induce these global or cluster changes. The ultimate goal of these studies was to identify molecules that could be exploited to modulate sensitivity of breast cancer cells to Tam. 104
Although our study was focused on two breast cancer cell lines, these results will help with future studies with patient tumors that have the potential to identify new targets that are universally dysregulated in tamoxifen resistance. In particular, as total smRNA on an NGS platform has not previously been used to characterize Tam resistant tumors, further research in this specific area of study would help to identify which smRNAs are universally dysregulated. In light of the large number of gene expression changes found, it is conceivable that some alterations are the driving changes leading to Tam resistance
Our findings reveal three areas that are modified on multiple levels in our Tam- resistant cells. First, proliferation signaling is modified with changes in cell cycle control and ESR1 down-stream genes that permit unregulated proliferation of the Tam-resistant cells (Fig. 2.6C1-3). Second, mitochondria and oxidative phosphorylation are affected
(Fig. 2.6C4). Several different types of units in the electron transport chain are altered that may permit new and more efficient means of energy production for the breast cancer cells. Finally, this study indicates that gene expression regulation is dramatically altered from changes in transcriptional control to adjustments in post-translational modifications and protein degradation (Fig 2.6C5-9). Within this area, we find that snoRNA could play a major role in Tam resistance. Independently, each of these areas could be investigated for therapeutic targets, and further exploration of the changes in snoRNA may lead to new diagnostic tests for Tam resistance. Together, our results exemplify the need for personalized medicine as the large number of genetic changes in Tam resistance can be overwhelming, but patterns of dysregulation may emerge as a patient’s own genetic signature is compared to samples of known resistant phenotypes. Using NGS and
105
clustering methods, therapies may be developed that target proteins or genes that are found to regulate these global changes, sensitizing more breast cancers to the anti- proliferative effects of Tam.
106
Chapter 3
Phosphorylation of elongation factor-2 kinase and the stability of the enzyme under various stress conditions
3.1. Abstract
Eukaryotic elongation factor-2 kinase (eEF-2K) is a CaM-dependent enzyme involved in regulation of protein synthesis. It inhibits translation by phosphorylating eukaryotic elongation factor-2 (eEF-2), which is responsible for ribosomal translocation from the A to P-site in eukaryotes, resulting in termination of peptide elongation. Studies show that eEF-2K plays a role in cell survival through this inhibition of protein synthesis and that its protein levels are increased in cancer. Post-translational modification of translation machinery is important for its regulation and could be critical for survival of cancer cells encountering stress. Thus, the purpose of our study is to examine the regulation of eEF-2K during stress with a focus on the phosphorylation status and stability of eEF-2Kprotein in cancer cells.
Using two human glioma cell lines (T98G and LN229), we have found a 2-5 fold increase in eEF-2K expression and activity under stress conditions of nutrient deprivation and hypoxia. mRNA levels are only transiently increased and shortly return to normal, while eEF-2K protein levels continue to increase after further exposure to stress. This
107
result could be explained by decreased turnover of eEF-2K protein, which has a normal half-life of ~ 6-8 hours in glioma cells, so cycloheximide experiments were used to examine the effect of stress on eEF-2K protein stability. A seemingly paradoxical decrease in eEF-2K stability (t1/2 = 2-4h) was found when glioma cells were subjected to stress despite increased protein expression. Phosphorylation may play a role in this altered protein stability as eEF-2K has multiple phosphorylation sites that are phosphorylated by the mTOR/S6 kinases (Ser78 and Ser366) and AMPK (Ser398), pathways which would be affected by stress. Therefore, phosphorylation-defective mutants of eEF-2K were made to examine the effect of phosphorylation at these sites on eEF-2K protein stability. We discovered that the AMPK site was pivotal to protein stability as the S398A mutant half-life increased to greater than 24 hours under both normal and stress conditions. Mutating the mTOR pathway sites made eEF-2K protein more stable under normal conditions (t1/2 > 24h) but decreased to normal levels under stress conditions (t1/2 = 8h). Inhibiting the mTOR pathway with rapamycin treatment increased protein expression ~ 5 fold and increased eEF-2K stability in these mutants under all culture conditions. These data indicate that eEF-2K is regulated at multiple levels with phosphorylation playing an important role in protein turnover. The unexpected decrease in eEF-2K protein stability during stress may be a compensatory mechanism for an additional level of regulation at the post-transcriptional level that increases eEF-2K translation. Due to the importance of translation regulation during stress, it is reasonable to have increased translation of a regulator of protein synthesis while decreasing the same protein’s stability in order to quickly adapt to changing
108
nutrient levels. Further studies will examine the post-transcriptional regulation of eEF-
2K during stress as these data demonstrate its complex and tight regulation.
Understanding the regulation of eEF-2K could lead to therapeutics targeting eEF-2K that could potentially render cancer cells intolerant to stress and susceptible to current treatments.
3.2. Introduction
3.2.1. Eukaryotic elongation factor-2 kinase (eEF-2K)
Elongation factor-2 kinase (eEF-2K) or CaM kinase III is a member of the calcium/calmodulin-dependent kinase (CaMK) family of serine/threonine protein kinases.
Other members of this family include CaMs I, II, IV; myosin light chain kinase; phosphorylase kinase; and an additional CaMK which phosphorylates and activates
CaMKs I and IV [264,265]. Members have a highly conserved catalytic domain next to a calmodulin binding domain. Even though eEF-2K is related to the other CaMKs, it has a distinct amino acid sequence sharing no homology with other CamKs and represents its own category; in fact, eEF-2K does not even share homology with other members of the eukaryotic protein kinase superfamily [266]. It does, however, share about 40% homology with Dictyostelium myosin heavy chain kinases A and B (MHCK) around the eEF2K catalytic domain, although MHCK is not a CaMK [267].
109
It has been suggested that the MHCKs and eEF-2K, along with channel kinases –
TRPM6 and 7 (“chanzymes” or channel-kinase fusions) [268] – and other uncharacterized kinases, represent a distinct protein kinase class called the α-kinases
[266] [269]. All conventional eukaryotic kinases share similar catalytic domains which have highly conserved motifs like DXXXN and DFG; eEF-2K has none of these domains
[270]. These α-kinases are named because of predicted α-helices in the C-terminus which are similar to SEL1 repeats (SLR) motif that are often found in protein-protein interactions [271]; this catalytic domain is thought to recognize amino acids that are located within other α-helices. The catalytic domain is not found in prokaryotic organisms, unicellular eukaryotes, or plant proteins [266]. Additionally, these alpha kinases seem to phosphorylate threonines which have previously been shown to destabilize α-helices [272], which would cause conformational changes in the target protein’s structure affecting its function.
Figure 3.1: Proposed structure of eEF-2K (partially modeled from Pigott et al, 2011) [273]. The N- terminus of eEF-2K contains phosphorylation (P) site Ser78 followed closely by a calmodulin (CaM) binding domain. Immediately adjacent is the catalytic domain responsible for eEF-2 phosphorylation at Thr56. After the catalytic domain is a linker section which contains various P-sites that affect eEF-2K activity and possibly conformation followed by a series of SEL1-like repeats (SLRs) that are proposed to stabilize a conformation of eEF-2K that promotes its catalytic activity. Finally, at the C-terminus is the eEF-2 binding domain. 110
3.2.1.1. eEF-2K structure
The C-terminus of eEF-2K contains a binding site for its substrate (eEF-2) and this same C-terminal end contains several phosphorylation sites (Fig. 3.1). The N- terminus contains the CaM binding site, which is adjacent to its catalytic domain
[269,274]. Deletion mutation experiments were performed to determine domains necessary for eEF-2K activity. Its catalytic domain does have some sequence homology to bacterial histidine kinases [266], which lead to the development of an inhibitor of eEF-
2K, NH125, during a bacterial histidine kinase pharmacological screen [275].
A new study by Pigott et al [273] goes into depth of the domains. It finds that eEF-2K only weakly binds CaM without calcium and that acidic pH allows for additional binding of CaM without calcium. The ATP binding site of eEF-2K is still accessible without Ca/CaM, but their presence does induce a conformational change that allows for
ATP to come closer to the catalytic domain. The C-terminus (15 residues) is essential for eEF2 phosphorylation – especially tyrosine residues. Residues from 1-75 (N-terminus might be important for its inhibition – if you remove it, it enhances autophosphorylation and activity) and 357-477 (these are linker site where phosphorylation could be important) are not required for function. In the middle of the protein from amino acids
76-356 exist autophosphorylation sites. eEF-2K also appears to have residues that are necessary for zinc binding and eEF-2K activity; zinc binding is also critical for other α- helix proteins ChaK1 and MHCKA activity[273]. Thus, additional unknown metal or mineral ions might regulate eEF-2K activity as well.
111
3.2.1.2. Regulation of protein synthesis by eEF-2K
eEF-2K is responsible for the phosphorylation of eukaryotic elongation factor 2
(eEF-2) [276]. This 725 amino acid protein inhibits protein synthesis by phosphorylating the 100 kDa protein elongation factor-2 (eEF-2) at Thr56 in its GTP-binding domain
[277,278]. eEF-2 promotes ribosomal translocation of mRNA from the A to P-site in eukaryotes by , and its phosphorylation decreases its affinity for the ribosome results in termination of peptide elongation [279,280]. eEF-2K has no other known in vivo substrates [281]. While eEF-2 is the only known substrate for eEF-2K in vivo, eEF-2K can also phosphorylate a peptide known as MH-1 (which is the peptide sequence in
Dictyostelium for myosin heavy chains, similar to eEF-2k) in vitro [269]. eEF-2 can be dephosphorylated by protein phosphatase 2A (PP2A) [282]
3.2.1.3 Regulation of eEF-2K
3.2.1.3.1. Calcium/Calmodulin and autophosphorylation
eEF-2K is a CaM kinase, so it is unsurprising that its activity would depend on calcium and CaM. eEF-2K is always dependent on calcium for its activity. The reliance of eEF-2K on calcium was first discovered by studies that revealed the importance of calcium on protein synthesis and eEF-2 activity. Increases in calcium levels can increase eEF-2 phosphorylation during transitions to proliferation [283] and mitosis [284], which in turn, decreases translation [285,286]. Drugs that increase intracellular levels of calcium (veratridine, thrombin, and histamine) all lead to increased phosphorylation of
112
eEF-2 and reduced translation [287,288]. Conditions like skeletal and heart muscle contractions associated with rapid rises in calcium levels have shown quick increases in pEF-2 levels [289]. The dependence of eEF-2K on CaM is not as strong. While CaM- dependent autophosphorylation of eEF-2K can lead to eEF-2K autonomy [290], eEF-2K is able to autophosphorylate itself without the presence of CaM. Mitsui’s group showed that autophosphorylation of eEF-2K resulted in partial activity (30%) of EF-2K that was partially Ca/CaM-independent; while CaM was not necessary, it appeared that some calcium was needed for eEF-2K activity during autophosphorylation.
The ability of eEF-2K to autophosphorylate itself was first discovered when it was shown that de-phosphorylated eEF-2K could be phosphorylated in vitro (test tube) in the absence of added kinases [291]. Activity of eEF-2K is decreased after such phosphorylation, indicating that autophosphorylation may turn off the kinase.
Autophosphorylation of other α-kinases like TRPM7 [292] and MHCKA [293] have been shown to affect their activation. Redpath and Proud have previously shown that calcium and CaM lead to eEF-2K autophosphorylation, but no sites were named [290]. Ser78 is only a minor autophosphorylation site. Thr-348 is a major site of autophosphorylation which affects minor phosphorylation at Ser61, Ser78, Ser491; mutating this site causes a decrease in eEF-2K activity. Thr348 is immediately off the C-terminal side of the catalytic domain. It has a low basal level, priming eEF-2K for activity by allowing a favorable conformation for the catalytic domain. Ser366 is another major autophosphorylation site. They argue that it is unexpected for Ser78 and Ser366 to be autophosphorylated when they are normally phosphorylated by the mTOR pathway
113
[294]. Studies also revealed that Ser78 phosphorylation impairs CaM binding [295] while Ser366 impairs calcium activation [296]. Redpath and Proud hypothesize that their autophosphorylation might be able to turn off eEF2K after it is activated by cellular stress; this is done through desensitization of eEF-2K to calcium and CaM which would allow elongation to continue under chronic stress conditions [294].
3.2.1.3.2. cAMP-dependent protein kinase (PKA) regulation of eEF-2K
cAMP-dependent protein kinase (PKA, also known as cAPK) can regulate eEF-
2K through increasing levels of the second messenger cyclic-AMP, which causes PKA to phosphorylate and activate eEF-2K in a calcium-independent manner [297]; cAMP phosphorylation of eEF-2K causes it to be semi-autonomous from calcium concentrations
– eEF-2K can function at about 40% of its activity in the absence of calcium [297].
It has also been shown that increases of cAMP caused by isoproterenol or CPT-
AMP activate eEF-2K while inhibiting protein synthesis [298,299]. Since glucagon and other hormone agents that increase cAMP can inhibit protein synthesis, it is hypothesized that eEF-2K might be responsible for their decrease in translation [300]. An initial study indicated that Ser152 may be the phosphorylation site of PKA [297], but no other studies have corroborated this finding. eEF-2K phosphorylation sites Ser499 and Ser365 (both by catalytic domain in C-terminus) were found to be phosphorylated by PKA both in vitro and in vivo, independently of one another (even when one was mutated); mutation of either reduced eEF-2K activity and autonomy in vitro, but only mutation of Ser499 affected autonomy in vivo [301]. The study also showed low levels of phosphorylation 114
on Ser434 by PKA, but the effect of Ser434 phosphorylation was not studied [301].
Thus, Ser499 is considered the primary target of cAMP, while Ser365 and Ser434 are minor sites.
Interestingly, the same study found it hard to express phosphomimetics of certain eEF-2K sites, which had been noted in previous studies. The authors could not express
S499D and hypothesized that it was because this site activates eEF-2K, which in turn, phosphorylate eEF-2 enough such that it inhibits more of its own production.
Phosphomimetic mutants could be highly constitutively active, which leads to inhibition of mRNA translation and therefore none of its own expression; the study tested both luciferase and beta-galactoside reporters, neither of which allowed for S499D expression
[301].
3.2.1.3.3. mammalian target of rapamycin (mTOR) regulation of eEF-2K
Mitogenic pathways in the cell are often regulated by mTOR. mTOR is signaled and activated by the presence of insulin, growth factors, and amino acids; in this way, the mTOR pathway is an important sensor of nutrients availability [302]. mTOR can associated with other proteins in two ways, mTOR complexes 1 (mTORC1 which includes Raptor) and 2 (mTORC2 which includes Rictor). mTORC1 is signaled by nutrients and activates downstream targets eukaryotic initiation factor 4E binding protein
(4E-BP1) and p70-S6kinase (S6K1), which are involved in cellular proliferation signaling. mTORC2 is also regulated by nutrients, but is less sensitive to rapamycin inhibition [303]. It can phosphorylate Akt, which in turn can activate mTORC1 [302]. 115
mTOR can phosphorylate eEF-2K through ribosomal S6 kinase 1 (S6K). This protein phosphorylate eEF-2K at Ser366, which lies in the C-terminus catalytic domain
[296]. Also in this pathway, studies have shown the mTOR or an mTOR target can phosphorylate eEF-2K at Ser78, which is immediately adjacent to the CaM binding site
[295]. Some studies suggest that Ser78 inhibits calmodulin binding and that is how it affects EF-2K activity [295]. These two phosphorylations of the enzyme, Ser78 and
Ser366, are inhibitory, which is reasonable since cells would continue protein synthesis during times of plentiful nutrients. Upstream signals of mTOR, such as insulin, decrease eEF-2 binding to eEF2K, which can be blocked through treatment with rapamycin.
Interestingly, insulin cannot phosphorylate eEF2K under amino acid deprivation, suggesting additional phosphorylation control of the kinase. Unexpectedly, in vivo,
AMPK, a stress signal, can phosphorylate eEF-2k at Ser78 as well while direct depletion of ATP cannot [295]. While mTOR is the primary phosphorylator of Ser78 and Ser366, the sites obviously are influenced by additional factors such as AMPK and autophosphorylation.
Studies in mouse cells, where Ser77 is the equivalent of Ser78 in eEF-2K, have shown phosphorylation at this site to be affected by Mg2+ levels and TRPM7 [304]. P-ef-
2 levels increased after Mg2+ was taken away and cellular growth correlated positively with the amount of Mg2+. TRPM7, which restores intracellular Mg2+ levels and growth
[305], mediates the phosphorylation of eEF-2 under low Mg2+ through eEF-2K at Ser77; this suggests cooperation between α-kinases. De-phosphorylated mutants of S77A or D are no longer sensitive to Mg2+, indicating that S78 is responsible for Mg2+ of eEF-2K.
116
Even more interesting is that low Mg2+ caused an increase in phosphorylated eEF-2K and total eEF-2K protein. S77A mutants had lower levels of eEF-2K. So, it is possible that mouse Ser77, and thus human Ser78, might have a positive effect on eEF-2K levels and possibly stability.
3.2.1 3.4. Adenosine monophosphate-activated protein kinase (AMPK) pathway regulation of
eEF-2K
AMPK is a major player in energy homeostasis of the cell [306]. The enzyme is able to detect fluctuations in the AMP/ATP ratio of the cell and is activated through phosphorylation at Thr-172 by its aptly named kinase, AMPK kinase, as AMP levels rise with ATP depletion. AMPK has many target proteins that it activates through phosphorylation which increase cellular energy by decreasing energy consumption. In general, AMPK decreases anabolic pathways and stimulates catabolic ones. Activation of AMPK leads to phosphorylation of EF2 and inhibition of protein synthesis [307].
Browne and Proud discovered that AMPK phosphorylated eEF-2K at Ser398
[308]. This phosphorylation of eEF-2K increases p-EF-2 levels. As previously state, they were also able to show that AMPK also phosphorylates Ser78 [307], despite the fact that this is a known mTOR phosphorylation site. In vitro (test tube), they found that
AMPK can phosphorylate Ser78, Ser366, and Ser398. Contrary to previous studies, they found that neither Ser78 nor Ser366 is a site of autophosphorylation since neither becomes phosphorylated when eEF-2K is incubated with CaM and calcium without
AMPK; thus, they determined that Ser78 and Ser366 can be direct targets of AMPK 117
[307]. However, this study was performed in cardiomyocytes while the previous study was done in HEK293 cells. This discrepancy between studies highlights the importance of different cellular systems and the complex regulation of eEF-2K phosphorylation.
It also appears that phosphorylation of one of these sites affects the phosphorylation of the other sites. When Ser398 is mutated to be phosphorylation- defective, phosphorylation of Ser78 and especially Ser366 increases. This regulation of phosphorylation is complicated by the fact that decreasing cellular glucose levels with 2- deoxyglucose decreases phosphorylation of Ser78 and Ser366 while increasing Ser398; this would be expected with the decrease in mTOR signaling but is not in line with the mutation studies. Therefore, this indicates that other factors and possibly other phosphorylation or regulatory sites may play a role in the phosphorylation end result of eEF-2K. Rapamycin treatment (inhibition of mTOR) does not increase Ser398 indicating that it is not phosphorylated just because mTOR is inhibited [307].
3.2.1.3.5. Multiple stress response pathways regulation of eEF-2K
Multiple additional pathways exist within the cell to help combat various stresses encountered by cells during their lifetime. The SAPK (JNK) and p38 MAPKs are all part of the mitogen-activated protein kinase (MAPK) protein family that includes extracellular-signal-related kinases (ERKs) as well [309]. While ERKs are normally activated by mitogens, the majority of MAPKs are like the SAPKs and p38MAPK, which are stress-activated kinases that react to stress stimuli such as radiation, oxidative stress, and inflammation [310]. The p38 MAPK pathway activates transcription that stops cell 118
growth and can affect other pathways like NF-κB [309]. The SAPK pathways can activate apoptosis and interacts significantly with p38 [309].
Recent studies have shown that eEF-2K is regulated by these stress pathways through multiple phosphorylation sites. Upstream MAPK kinases (MKKs) are especially important in this regulation. Despite the fact that it is a stress-activated protein,
SAPK4/p38δ is also downstream of insulin-like growth factor 1 (IGF1), and this enzyme has been shown to phosphorylate eEF-2K at Ser359 [311]. This finding was corroborated by another study that found Ser359 levels to have basal phosphorylation that was increased upon mitogen, IGF and EGF, treatment [312]. However, this same study found that protein synthesis inhibitor anisomycin can phosphorylate eEF-2K at Ser359, a phosphorylation which is not affected by inhibition of mTOR, MAPK, or SAPK2. This was the first report that eEF-2K can be inactivated, rather than activated, by stress;
Ser359 phosphorylation by stress decreased p-eEF-2 levels [312]. SAPK4 was also found to have a second, less intense phosphorylation was found at Ser396 [313].
Other SAPKs were studied for their effect on eEF-2K [314]. A SAPK2 inhibitor can induce phosphorylation of eEF-2K at Ser396, as well, both in vitro and in vivo; this phosphorylation decreases eEF-2K activity by increasing p-EF2 levels. This suggests that both SAPK2 and SAPK4 can affect Ser396 phosphorylation. MKK2 is able to induce phosphorylate eEF-2K at Ser-377 through SAPK2a/p38 [311], an effect which has been shown to be independent of AMPK signaling. However, phosphorylation at this site does not appear to regulate eEF-2K activity. Inhibition of protein synthesis by anisomycin cannot effect Ser-396 phosphorylation at low levels, but is able to increase
119
Ser359 and Ser377 phosphorylation during serum and amino acid deprivation. eEF-2K is therefore regulated by both SAPK2 and SAPK4 which affect multiple phosphorylation sites of eEF-2K [311].
While the previous study had shown that amino acid starvation decreases Ser359 phosphorylation, this could not be exclusively the result of activation by SAPK4δ since this enzyme is not normally active during basal levels in all cells (such as HeLa) and is not regulated by amino acid stress [315]. Therefore, a different team of researchers found that cdc2-cyclin B co-immunoprecipitated with eEF-2K phosphorylated at Ser359.
Mutation of the site still led to some phosphorylation of eEF-2K by cdc2 indicating that it has other minor phosphorylation sites on eEF-2K, but they do not affect activity. Ser359 phosphorylation of eEF-2K peaked at the same time as G2/M, as did cyclin B (cdc2 levels remained constant). p-eEF-2 was high after cell cycle block, low during G2/M, and rose at G1/S. Proteasome inhibitor MG132 actually decreased Ser359 levels, which was not due to degradation of eEF-2K. This study revealed that eEF-2K is inactive during mitosis due to phosphorylation by cdc2-cylinB. They also say that eEF-2 may stay active to allow IRES-driven mRNA’s to be translated [315]. However, this might not be true in all systems as other studies indicate that eEF-2 is phosphorylated and inactive during mitosis in heart [284] and other cells [316].
eEF-2K is also affected by cellular pH stress through unknown pathways. eEF-
2K activity is low at normal pH levels of 7.2-7.4, but eEF2K is activated when acidity increases at pH levels of 6.6-6.8 [317]. Since acidification that occurs during hypoxia
120
and ischemia has been shown to lead to global inhibition of translation [318], this might explain part of that inhibition of protein synthesis.
3.2.2. Stability of eEF-2K protein
Protein turnover is necessary in cells to conserve the integrity of the cellular proteins. Normal functioning of the cell produces reactive oxygen species and other matter that can adversely affect protein structure and function. Continual turnover of cellular proteins creates a homeostasis in protein amounts and activity, and a delicate balance exists between breakdown of proteins and their synthesis. Various proteolytic systems exist that hydrolyze proteins into their constituent amino acids. While some proteins have half-lives of days or weeks in the case of cytoskeletal proteins like actin, others, such as some regulatory enzymes, have half-lives of only a few minutes [319].
All systems of degradation break down proteins into their amino acids.
Lysosomal digestion of proteins within an acidic vacuole containing acid-optimal proteases is the preferred method for endocytosed extracellular and surface proteins and autophagic vacuoles[320], Caspases, cysteine proteases, can cleave proteins after aspartic acid during apoptosis [321], and the calcium-activated proteolytic system involves another set of cysteine proteases called calpains which function when intracellular calcium rises during cellular injury [322]. The major system of degradation of intracellular proteins is the ubiquitin-proteasome pathway [323].
This system involves the covalent attachment of ubiquitin through an acyl modification of a residue of the selected protein in order to target the protein for protein 121
degradation by the multi-catalytic protease complex, the 26S proteasome [324]. The ubiquitin-proteasome pathway involves three enzymes: ubiquitin-activating enzyme (E1), ubiquitin conjugating enzyme (E2), and ubiquitin ligase (E3) [325]. This system either mono- or poly-ubiquitinates its target protein. Poly-ubiquitination normally signals for protein degradation while a mono-ubiquitination can be a signaling marker on the protein
[326].
Protein turnover is necessary in a variety of situations ranging from the need for rapid termination of cellular processes under different physiologic conditions, to regulate gene transcription; quality control of damaged or mis-folded proteins; antigen presentation; and amino acid shortage [327].
eEF-2K has been shown to be degraded by the ubiquitin-proteasome pathway.
The enzyme is poly-ubiquitinated and degradation can be inhibited with the treatment of the proteasome inhibitor MG132. The half-life of endogenous eEF-2K in human glioma cells is approximately 6 hours. Disruption of the complex [328] with eEF-2K chaperone protein heat shock protein 90 (Hsp90) by treatment with Hsp90 inhibitor geldanamycin results in increased ubiquitination and turnover of eEF-2K; the half-life of eEF-2K decreased to less than 2 hours [329]. Thus, it has been previously shown that eEF-2K degradation can be regulated by its interaction with Hsp90 and its ubiquitination state.
Since rapid changes in eEF-2K protein expression have previously been implicated in regulating crucial processes within the cell [284,330,331], studying the effects of phosphorylation on the enzyme’s stability will be critical for understanding how eEF-2K works within the cell.
122
3.2.3. eEF-2K expression
Increased eEF-2K activity has been seen in a variety of circumstances including cellular differentiation [332], cell cycle progression [315], cell growth [283], and neuronal functions [333,334]. It is rapidly increased when cells are stimulated with growth factors [296],while decreasing quickly when growth stops [335]. eEF-2K expression and activity was found to be increased in aging rat livers, which could be mitigated slightly by a calorie-restricted diet (which has been shown to reduce free radicals and hence stress) [336]. eEF-2K and Ser366 eEF-2K were increased in
Alzheimer’s disease patients compared to control brains, however, levels of p-eEF-2 were not correlated with neurofibrillary degeneration [337]; the authors believe this is through activation of p70S6K which leads to increase Tau expression.
3.2.3.1. eEF-2K expression in cancer
eEF-2K is normally expressed at varying levels in all human and vertebrate tissues [269]. Some evidence exists for different isoforms being found in different tissues, which would confuse comparative analyses between studies in different tissues
[338]; however, numerous cancer cell lines were recognized with the rabbit antibody for eEF-2K including HeLA, MCF-7, and C6 cells. It has been shown to have increased expression and activity in human breast cancer tumors as compared to adjacent normal tissue [339]. Studies also indicate that there may be decreased levels of phosphatase activity of p-EF-2 in even the normal tissue. While basal eEF-2K levels are very low in normal tissues, increased levels of EF-2K activity have been found in DCIS, indicating 123
that it could be an early marker of invasive breast cancer [339]. The same study determined that eEF-2K was significantly increased in cell lines including T98G
(gliomablastoma), OVCAR-3 (ovarian adenocarcinoma), MCF-7 (E2 positive breast cancer), MDA-MB-231 (E2-negative) in both protein expression and activity (p-EF-2).
The study also showed that serum deprivation decreased eEF-2K activity and proliferation, while treatment with IGF-1 or EGF rescued this effect [339]. Other cancer studies have shown increases in other malignant cell lines and human cancers [335,340].
3.2.3.2. Correlation with stress and cellular energy
Altered protein synthesis can be an advantage to cancer cells due to its global effects on gene expression. Studies show that eEF-2K plays a role in cell survival through to this inhibition of protein synthesis and that its protein levels are increased in cancer.
Expression of eEF-2K has been shown to be increased in cancer, including in patient tumor samples. Association with cell survival was first seen in hibernating squirrels, where pEF-2 and eEF-2K were increased in their brain and liver samples during stressful environmental conditions that reduced oxygen profusion and nutrient availability [341].
Decreased protein synthesis makes metabolic sense under these conditions, and eEF-2K was shown to play a role in energy homeostasis.
Autophagy is also often involved in regulating cellular energy, and eEF-2K has been shown to induce autophagy under a variety of circumstances, such as during ER- stress [342] and amino acid deprivation [343]. Cycloheximide (protein synthesis inhibitor) treatment blocks eEF-2 phosphorylation [342] and is sufficient to trigger
124
autophagy [344]. Since the role of eEF-2 is to hydrolyze GTP to GDP [276], providing energy during translation, it is unsurprising that its inhibitor, eEF-2K, might be involved in energy-depleting situations. CHX, which inhibits protein synthesis, inhibits general protein synthesis but can cause translation of certain mRNAs [345] and prevent degradation of others [346]. Autophagy is regulated by mTOR [347] and withdrawal of growth factors [79], both of which effect eEF-2K activity; in fact, autophagy has been shown to be regulated by eEF-2K in human cancer cell lines. Cancer cells have altered energy statuses because of their reliance on aerobic glycolysis for ATP, even in the presence of oxygen; this is known as the Warburg effect [24]. Interfering with energy metabolism has been hypothesized to be a possible anti-cancer mechanism [348].
Nutrient deprivation increases autophagy and eEF-2K expression and activity in human glioma cells which was stopped by inhibition of eEF-2K either pharmacologically or genetically [343]. Thus, inhibition of eEF-2K can inhibit autophagy and lead to cell death during metabolic stress.
Inhibiting eEF-2K is linked to cancer cell death. Rotterlin can inhibit eEF-2K, which killed glioma cell lines [340], while eEF-2K and calmodulin siRNA were able to kill glioma cell lines as well [349]. Nutrient deprivation (serum) decreased Ser366 levels and increased eEF-2K and p-eEF-2 levels in MCF-7 and T98G cells [350]. Inhibiting glucose uptake by means of 2-deoxy-glucose (2-DG) was able to induce autophagy and activate eEF-2K in human glioma cells; silencing of eEF-2K lead to cell death [351]. It is also cytoprotective to breast cancer cells that are subjected to growth factor inhibition
125
through increasing autophagy [352]. Exposing glioma cells to eEF-2K inhibition lead to a sensitization to TRAIL, allowing cells to undergo apoptosis [353].
eEF-2K also has cytoprotective effects in normal cells. eEF-2K protects cardiomyocytes during hypoxia and has been proposed as an energy sensor as it protects cardiomyocytes against stress through its AMPK activation [354]. If you increase AMPK in cells, then cardiomyocytes have inhibited hypertrophic growth [355]. Macrophages can use oxLDL (oxidized low density lipoprotein) to increase calcium influx to activate eEF-2K (reducing protein synthesis) in order to block apoptosis. This leads to further recruitment of macrophages which is necessary for initiation and progression of artherosclerosis [356]. eEF-2K activity is necessary for the reduction in protein synthesis during skeletal muscle contractions, which interestingly enough, were unrelated to
AMPK activity [357].
3.3. Rationale
Elongation factor-2 (EF-2) kinase is a CaM-dependent enzyme involved in regulation of protein synthesis. It inhibits translation by phosphorylating elongation factor-2 (EF-2), which is responsible for ribosomal translocation from the A to P-site in eukaryotes, resulting in termination of peptide elongation. Studies show that EF-2K plays a role in cell survival through this inhibition of protein synthesis and that its protein levels are increased in cancer. Modulation of translation is critical for survival when cancer cells encounter stress. Thus, our lab is interested in the effect of metabolic stress 126
on EF-2K regulation in cancer cells. Studies show that EF-2K plays a role in cell survival through this inhibition of protein synthesis and that its protein levels are increased in cancer. Post-translational modification of translation machinery is important for its regulation and could be critical for survival of cancer cells encountering stress.
Thus, the purpose of our study is to examine the regulation of EF-2K during stress with a focus on the phosphorylation status and stability of EF-2K protein in cancer cells.
3.4. Experimental design
3.4.1. Cell lines and culture.
The human glioma cell lines T98G and LN299 were purchased from American
Type Culture Collection (Manassas, VA, USA). T98G(-EF2K) and LN299(-EF2K) cells were previously stably transfected with shRNA against EF-2K in a pcDNA 3.1 vector.
T98G cells were cultured in Ham’s F-10:DMEM (10:1), while LN229 cells were cultured in DMEM. Cell cultures were supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 ug/mL streptomycin. Cells were maintained at 37ºC in a humidified atmosphere containing 5% CO2/95% air. All cultures were monitored routinely and found to be free of contamination by mycoplasma or fungi. All cell lines were discarded after three months and new cell lines propagated from frozen stocks.
127
3.4.2. Reagents and antibodies
All cell culture media and reagents were purchased from Hyclone (Invitrogen,
Carlsbad, CA, USA). Cycloheximide was purchased from Sigma Aldrich. Antibodies to
EF2K, phospho-EF2K (Ser 366), S6 kinase, phospho-S6 kinase (Thr), AMPK, phospho-
AMPK (Thr), and β-actin were all purchased from Cell Signaling Technologie (Danvers,
MA; USA) while phospho-EF2K (Ser 78) was supplied by Santa Cruz. EF-2K mutants in pcDNA 3.1 vectors were purchased from Genewiz.
3.4.3. Stress Conditions
Serum conditions contained 10% fetal bovine serum (FBS). For transient transfections with expression vectors containing phosphorylation mutants EF-2K at
Ser78/Ser266 and Ser398, 2 x 10^5 T98G(-EF2K) cells were transfected with 1 ug of
Qiagen-purified DNA and 5 ul of Roche Fugene 6 as previously described (*). Stress experiments were conducted 24 hours after plating at of 2 x 10^5 cells or 24 hours after transfection. Media was changed to serum-free, glutamine-free, or oxygen-depleted media for serum, amino acid, and oxygen deprivation experiments, respectively. Oxygen deprivation was conducted in a hypoxia chamber at 1% oxygen.
3.4.4. Real time RT-PCR
Total RNA from T98G cells were extracted using Trizol system according to manufacuturer instructions. cDNA was made from harvested total RNA of T98G and
128
LN229 cells (Roche). The products were tested for purity using spectrophotometry
(Aligent Nanodrop). RT-PCR was performed using TaqMan Gene Expression Assays
(Applied Biosystems) on a Statagene Mx3005P (Aligent Technologies). GAPDH was used to normalize samples for comparison.
3.4.5. EF-2K phosphorylation-defective mutants
Mutation variants of the three eEF-2K phosphorylation sites were created using
Genewiz site-directed mutagenesis. De-phosphorylation mutants were created exchanging serine coding sequences with alanine coding sequences. eEF-2K Ser78
(TCC) was converted to Ala78 (GCC), Ser366 (TCT) was converted to Ala366 (GCT), and Ser398 (TCT) was converted to Ala398 (GCT). Mutations were confirmed by sequencing with SeqMan. eEF-2K mutant sequences were cloned into Invitrogen pcDNA3.1(+) plasmid vectors. DH5-T1 E.coli were then transformed with eEF-2K mutant vectors and control vector. Transformants were selected by ampicillin treatment.
DNA plasmids were extracted with Qiagen Maxi Prep DNA kit. Electrophoresis was performed to verify the final product.
3.4.6. Preparation of cellular extracts and Western blot analysis
Cells were lysed with M-PER mammalian protein extraction reagent (Pierce
Biotechnology) which was supplemented with protease inhibitor cocktail (Roche) and a phosphatase inhibitor cocktails 1 and 2 (Sigma) followed by centrifugation at 14,000 x g for 10 minutes. Cell lysates were collected and protein concentrations measured (Aligent 129
Nanodrop). Protein (30 µg) was resolved by SDS-PAGE and transferred to PDVF membrane (Bio Rad). Blots were blocked with 5% mile/TBST at room temperature and then incubated with primary antibodies in 1% BSA/TBST overnight at 4°C. Secondary antibody incubation was carried out at room temperature for 1 hr. Protein signals were detected by ECL method (Perkin Elmer).
3.4.7. Mining of mRNA functional elements
Web-based tool, RegRNA – A Regulatory RNA Motifs and Elements Finder, was used to find predicted regulatory RNA motifs and elements in eEF-2K [358]. eEF-2K gene accession number NM_013302 was used for database query. 5’- and 3’-UTR regulatory elements of eEF-2K were examined.
3.5. Results
3.5.1. eEF-2K levels are increased by metabolic stress in glioma cells
To determine how cellular stresses affect EF-2K activity and expression, we investigated if nutrient or oxygen deprivation could alter the expression EF-2K mRNA or protein levels. Glioma cells were subjected to various stresses for 48 hours. Figure 3.2A shows that incubation of human glioma cells lines, T98G and LN229, with media
130
131
deprived of serum or glutamine as well as oxygen deprivation increases EF-2K protein expression. Cellular stress conditions increased eEF-2K activity levels as well as indicated by increased phosphorylation of its substrate, EF-2, as determined in by increased phospho-eEF-2 levels.
A time course of serum deprivation demonstrates that EF-2K protein levels continued to increase from 24 to 48 hours (Fig. 3.2B). Analysis of EF-2K mRNA expression revealed that while mRNA levels transiently increased at 24 hours, mRNA levels reverted to baseline by 48 hours (Fig. 3.2C). However, eEF-2K protein levels continued to increase at 48 hours despite having a previously reported half-life of ~8 hours in T98G glioma cells. This indicated that eEF-2K protein stability could be responsible for the increased protein expression during metabolic stress.
3.5.2. eEF-2K protein turnover is increased by metabolic stress
Next we examined the effects of various stresses on eEF-2K protein turnover in human glioma cells. Surprisingly, eEF-2K turnover was increased when glioma cells were exposed to various stresses, with half-lives averaging from 2-4 hours as opposed to
~8-12 hours under normal culture conditions (Fig. 3.3A). This result led us to examine the effects of stress on upstream signaling of EF-2K.
Multiple pathways are known to signal and phosphorylate eEF-2K. The mTOR pathway is responsible for deactivating eEF-2K during nutrient-rich conditions. mTOR pathway protein S6K1 can directly phosphorylate eEF-2K at Ser366 which inhibits eEF-
2K activity, this protein and its activation was measured. Energy sensor AMPK is also a 132
major regulator of eEF-2K, although it plays an opposite role from the mTOR pathway.
Because AMPK phosphorylates eEF-2K at Ser398 which activates eEF-2K during cellular stress, its activity was examined as well. Western blot analysis showed that cellular stress decreased mTOR signaling and thus S6K1 phosphorylation while it increased AMPK activation (Fig. 3.3B). S6K1 become dephosphorylated at its activation 133
site, Thr. Conversely, AMPK was activated as determined by its phosphorylation at
Thr172. Thus, the major upstream signaling pathways of eEF-2K functioned as expected during metabolic stress.
3.5.4. Phosphorylation sites differentially regulate eEF-2K turnover
Since both mTOR/S6K and AMPK are regulators of eEF-2K through its phosphorylation, we next created mutants for their respective phosphorylation sites on eEF-2K, S78A/S366A and S398A, to examine the effects of eEF-2K phosphorylation on the protein’s stability during stress. Mutations of serine to alanine produce residues that cannot be phosphorylated, so the mutants are permanently de-phosphorylated at those sites. eEF-2K mutants were over-expressed in cell lines that were stably transfected with knock-downs of eEF-2K; thus, the majority of eEF-2K in the cell was the mutated form.
Mutation of the mTOR/S6 kinase sites (S78/366A) resulted in increased stability (t1/2 >
24 hrs) under normal culture conditions. Turnover rates measured during cellular stress decreased to the basal level seen for non-mutant eEF-2K under normal culture conditions
(t1/2 ~ 8 hrs) (Fig. 3.4A). AMP kinase phosphorylation-site mutants (S398A) of eEF-2K also show increased stability under normal conditions (t1/2 > 24 hrs) and this stability of eEF-2K continued under all stress conditions (t1/2 > 24 hrs) (Fig. 3.4B). Therefore, we determined that phosphorylation of eEF-2K at various sites differentially affects the protein’s turnover in glioma cells.
134
3.5.5. Effects of inhibition of upstream signaling cascades on eEF-2K stability
To verify the roles of upstream signaling pathway phosphorylation of eEF-2K on the enzyme’s stability, we used pharmacological inhibitors of these cascades to compare with the results of mutating their respective phosphorylation sites. Figure 3.5A 135
136
shows that the mTOR inhibitor rapamycin decreased phosphorylation of S6K under all culture conditions including stress. Rapamycin treatment differed from the mTOR/S6 kinase phosphorylation-site mutants of eEF-2K as it decreased the turnover of eEF-2K
137
under both normal and stress conditions (t1/2 > 24 hrs) (Fig. 3.5B). Pharmacological inhibition of AMPK activation by Compound C decreased phosphorylation of AMPK at
Thr172 (Fig. 3.6A). eEF-2K stability was decreased under all culture conditions with
Compound C treatment (t1/2 < 4 hrs), which was the opposite result of mutating the AMP kinase phosphorylation-sites of eEF-2K (Fig. 3.6B). These results indicate that eEF-2K is regulated not only by their known phosphorylation targets but by additional mechanisms.
3.5.6. Determination of RNA elements important for translation of eEF-2K
To elucidate the mechanism by which cellular stress increases EF-2K expression, which cannot be fully accounted for in EF-2K mRNA levels, we examined the eEF-2K mRNA transcript sequence for RNA functional elements. Using RNA regulatory motifs and element predictions, we discovered that eEF-2K’s 5-UTR region is predicted to contain a terminal oligopyrimidine tract (TOP). TOPs are common to all ribosomal proteins and translation elongation factors which can coordinate to repress translation
[359]. Thus, this finding is expected as eEF-2K plays a role in repressing protein synthesis.
Several other 5’-UTR mRNA elements were predicted in eEF-2K that affect mRNA translation. eEF-2K contains multiple internal ribosome entry sites (IRESs).
IRESs are responsible for the initiation of translation by internal ribosome binding of the mRNA when 5’-cap-dependent translation is inhibited [360]. Multiple upstream open reading frames (uORFs) were predicted as well in the 5’UTR of eEF-2K; uORFs can 138
initiate and enhance translation during specific conditions [361]. These two types of
RNA functional elements may elucidate the mechanism behind increased eEF-2K protein levels despite a decrease in mRNA levels.
The other end of the protein contains a 3’-UTR motif found in alcohol dehydrogenase (Adh) mRNA which down-regulates Adh gene expression through reduction of its mRNA stability [362]. Additionally, K- and GY-boxes were predicted at multiple sites in the 5’-UTR of eEF-2K. These sites are found in proteins, particularly repressor proteins, that have decreased transcript levels due to stability [363]. Predictions of such motifs in eEF-2K could explain why transcript levels are not increased for extended periods of time despite an initial increase in mRNA under stress conditions.
3.6. Discussion
Here, we report that post-translational modification of eEF-2K, a significant regulator of protein synthesis, is pivotal in regulating the enzyme’s protein levels during periods of cellular stress.
To explore the effects of stress on eEF-2K, we compared the enzyme’s mRNA and protein expression levels under various stressful culture conditions including nutrient and oxygen deprivation. We found that while eEF-2K mRNA levels only transiently increased, followed by a return to basal levels, eEF-2K protein expression continued to increase further above basal levels over the same time period (Fig. 3.2C), suggesting the cellular stress might cause an increase in the protein’s half-life. Thus, the stability of 139
eEF-2K was then examined to determine the effects of stress conditions on the enzyme’s turnover. Unexpectedly, all cellular stresses caused a decrease in eEF-2K stability (Fig.
3.3A). This surprising result led us to investigate the role of eEF-2K phosphorylation- status on its stability.
Figure 3.7: eEF-2K phosphorylation sites.
Previous studies have reported that eEF-2K activity is regulated by multiple upstream signaling cascades (Fig. 3.7). The mTOR/S6 kinase pathway phosphorylates eEF-2K on S78 and S366, respectively, to inhibit the enzyme’s activity [295,296], while the AMP kinase pathway phosphorylates it at Ser398 activating eEF-2K [307]. These pathways tie into cellular stress, as the mTOR/S6 kinase signals during nutrient-rich conditions while AMP kinase signals during times of low ATP and stress. Therefore, 140
when we examined these signaling cascades under the various stress conditions, we found mTOR/S6 kinase signaling to be decreased and AMP kinase signaling to be increased along with activation of eEF-2K, as expected (Fig. 3.3B). However, mutating the known phosphorylation-sites of these signaling cascades resulted in unexpected results.
Pharmacological inhibition of upstream signaling of eEF-2K demonstrated varying results from mutation of their phosphorylation sites on eEF-2K. mTOR/S6 kinase inhibitor, rapamycin, increased half-life to greater than 24 hours for eEF-2K under all conditions including stress (Fig. 3.5), while the S78A/S366A mutation resulted in decreased turnover under nutrient-conditions and only had a slight stabilization effect under stress where half-life remained around 8 hours instead of dropping to ~ 2-4 hours
(Fig. 3.4A). Rapamycin treatment was more effective at stabilization of eEF-2K than mutation of the known mTOR pathway phosphorylation sites on the enzyme. Treatment of glioma cells with Compound C, the AMP kinase inhibitor, resulted in decreased stability of eEF-2K under all conditions (t1/2 ~2-4 hrs) (Fig. 3.6), which was the exact opposite of the S398A eEF-2K mutant which extended half-life to greater than 24 hours under all conditions (Fig. 3.4BC).
Actual results from the pharmacological inhibition of the upstream pathway were more in line with the original hypothesis that the mTOR/pathway destabilized eEF-2K and inhibition of the pathway would result in decreased turnover, while the AMP kinase pathway signaling would increase stability of eEF-2K and blocking the cascade, would increase degradation of eEF-2K. However, further studies of phosphorylation and eEF-
141
2K upstream regulation are necessary in order to elucidate the mechanism behind the pharmacological and mutation discrepancy.
Phosphorylation control of eEF-2K is a complex process, with a variety of signaling pathways converging on eEF-2K in a seemingly paradoxical way. Sites like
Ser78 are phosphorylated by mTOR, but stress proteins like AMPK can also phosphorylate this site. In fact, AMPK phosphorylation of its main site, Ser398, can actually induced phosphorylation of mTOR sites Ser78 and Ser398. In some cell systems, autophosphorylation of eEF-2K seems to play a role, while in others, autophosphorylation is overshadowed by stronger phosphorylation signals. These findings highlight the fact that cell system matters tremendously in eEF-2K studies as do culture conditions and an intricate balance of phosphorylation sites.
Our results revealed that negative mutations of mTOR phosphorylation sites,
Ser78 and Ser366 [296], increased stability of the protein under normal conditions; however, stress conditions decreased the stability of eEF-2K back to normal levels. Thus, it appears that phosphorylation at the mTOR sites decreases eEF-2K stability under normal conditions, but has no effect on eEF-2K under stress conditions. When the stress regulator protein AMPK’s phosphorylation sites on eEF-2K are mutated (S398A), stability of eEF-2K is increased; this stability is not affected by culture conditions including stress. These results indicate that the AMPK site, Ser398, is pivotal to eEF-2K stability, regardless of conditions; phosphorylation at this site appears to decrease stability of eEF-2K. The mTOR mutants might still be reactive to stress, due to the fact that AMPK can still affect the phosphorylation of the kinase at Ser398; this might explain
142
the mTOR-site mutants reduced stability under stress conditions. Because we show that all of our stress conditions activate AMPK, and AMPK is known to phosphorylate Ser78 and Ser366 in addition to Ser398 [308], our results showing that the stability of S398A mutants are not affected during stress indicate that Ser78 and Ser398 phosphorylations are not as important as phosphorylation at Ser398. This reduced stability of eEF-2K under stress signaling by Ser398 is unexpected, however, but quick turnover of a regulatory protein is not unheard of.
In fact, strict regulation of the stability of translational factors is common. For instance, in two proteins involved in translation termination, eRF1 and eRF3, the stability and degradation of the proteins is dependent on their interaction with one another [364].
It was suggested that protein degradation was the mechanism by which the termination portion of translation was tightly regulated. The two proteins had long half lives when not associated with one another. The tight regulation would allow termination to vary with the fluctuations in global translation rate.
The unmatched levels of mRNA, protein levels, and protein turnover are uncommon, but discordant mRNA levels and protein stability have been found before in cells. During differentiation, c-myc translation is enhanced from relatively moderate levels of c-myc protein; however, there was a concomitant decreased in c-myc stability
[365]. This increase in enhanced translation of certain transcripts even occurs for mRNAs that are considered unstable and have high turnover rates. One study that examined mRNA stability and translation during ER-stress found that both stabilized and destabilized mRNAs were amongst those that were translationally induced – that is, there
143
translation was enhanced. Even more interesting was that those that were stabilized seemed to be mRNA associated with reduction of global translation [366]. They suggested that there was the possibility of RNA binding proteins that chaperone between genome and proteome which allows enhanced translation but rapid turnover of superfluous mRNA, which has previously been indicated for other transcripts [367].
These findings could be key to understanding the seemingly paradoxical changes in eEF-
2K protein levels and stability.
The unexpected decrease in EF-2K protein stability during stress may be a compensatory mechanism for an additional level of regulation at the post-transcriptional level that increases EF-2K translation. Due to the importance of translation regulation during stress, it is reasonable to have increased translation of a regulator of protein synthesis while decreasing the same protein’s stability in order to quickly adapt to changing nutrient levels. Upon mining databases of RNA functional elements, we found that eEF-2K contains multiple predicted IRESs in its 5’-UTR. Thus, while global translation of regular 5’-capped mRNAs is inhibited, eEF-2K and other mRNA with
IRESs might still be able to undergo protein synthesis. eEF-2K also had a number of uORFs which would further regulate the rate of translation of eEF-2K protein.
Additional 3’-UTR elements predicted in eEF-2K mRNA could also explain the lack of increase in eEF-2K mRNA as these sequences are associated with decreased mRNA stability. Thus, it is possible that eEF-2K mRNA and protein levels are carefully regulated by the cell, which may quickly and efficiently increase eEF-2K translation during stress while destabilizing both the protein and its mRNA in order to allow for fast
144
resumption of global protein synthesis when favorable conditions return. Further studies will examine the post-transcriptional regulation of EF-2K during stress as these data demonstrate its complex and tight regulation.
The differential regulation of eEF-2K stability and phosphorylation in response to upstream signaling should be carefully considered when targeting eEF-2K for cancer therapy. Modulation of eEF-2K has been proposed several times as a novel adjuvant cancer therapy target [352,368]. The results presented here indicate that eEF-2K stability and phosphorylation are delicately balanced by several upstream pathways. The action of these upstream signaling cascades might have functional consequences on eEF-2K phosphorylation and therefore protein conformation. As proposed by Pigott et al, the majority of eEF-2K phosphorylation sites, including pivotal Ser398, are in a linker region of eEF-2K that is thought to affect the conformation of the kinase [273] (Fig. 3.1).
Differential phosphorylation of eEF-2K in response to upstream signaling and stress could therefore alter conformation of the protein, which might shield or enhance interaction of the catalytic domain with pharmacological inhibitors. This could lead to therapies that work under certain conditions while failing under others. We propose that further study of the effect of upstream signaling inhibition on eEF-2K stability and phosphorylation, along with their functional consequence on eEF-2K pharmacological inhibition, be undertaken in order to effectively target eEF-2K for adjuvant and combination treatments of cancer.
145
Chapter 4: Discussion of the current Studies; Altered signaling and gene expression in cancer cells under stress
4.1. Preamble
Solid cancers are the 2nd leading cause of death in adults in the United States.
Understanding the mechanisms by which cancer cells survive under stress is pivotal to decreasing this mortality statistic. The preceding chapters sought to introduce the concept that cancer cells are constantly under stress, a factor which needs to be taken into consideration during the treatment of solid cancers. Both intrinsic and extrinsic stresses impact the development and progression of neoplasms, down to the level of individual proteins and genes. Stresses like nutrient deficiency, hypoxia, acidity, and the immune response are present during tumor growth, development, and treatment. Tumor cells that survive these stresses are more adept at surviving hostile conditions and are more resistant to current therapies. Stress, therefore, shapes the tumor cell population. Gene expression alterations and modified signaling are the driving features behind the adaptive ability of cancer cells to these stresses. Thus, careful examination of these gene expression changes must be undertaken in order to develop effective therapies for solid cancers.
146
4.2. Tam resistance in breast cancer
4.2.1. Clinical Studies
As one-third of women treated with Tam will have a recurrence of breast cancer within 15 years of treatment [203], determining mechanisms of resistance to Tam has been an important clinical research aim for breast cancer researchers. Numerous clinical studies of Tam resistance have showcased the variety of gene alterations in Tam-resistant breast cancer samples. The primary mechanism of intrinsic resistance appears to be a lack of expression of ERα [184], although recent clinical studies have shown some breast cancers that do not express ERα have been successfully treated with tamoxifen [199].
This indicates that the mechanism of Tam is still not fully understood. The other major intrinsic mechanism of Tam resistance is the CYP2D6 mutation, which clinical studies have determined is carried by 8% of Caucasian women, preventing the conversion of
Tam into its active metabolites [204]. These two intrinsic mechanisms of Tam resistance highlight the importance of gene expression at the foundation of breast cancer therapy.
As the majority of breast cancer patients do not have either of these intrinsic gene expression changes and positively respond to Tam treatment for several years, acquired resistance must account for the greater part of Tam resistance. Large microarray studies on patients’ normal and cancerous breast tissues have shown the diversity of gene expression patterns that are present in Tam resistant samples. Alterations in expression and splicing of different ER isoforms play a large role in resistance [206,207,208], as does amplification of growth factor receptors such as Her2neu (ErbB2) [211]. Gene 147
expression of cell cycle regulators are particularly affected in Tam resistance. Over- expression of cyclins like cyclin D1 and cyclin E1 along with MYC, can lead to inactivation of the tumor suppressor RB; this can lead to the progression of the cell cycle despite additional genetic insults [182]. Alterations in Anti-apoptotic and pro-survival genes help these cancer cells survive Tam treatment [253]. Many miRNAs including mir-221/222 and mir-21 have altered expression, which can lead to modification of other gene transcripts [104]. All these small genetic changes can add up to lead to Tam resistance.
Gene expression profiles have highlighted the extreme amount of gene expression changes in Tam-resistant samples. Many of these alterations might be the result of driver mutations, but the sheer number of gene expression changes is astounding. A meta- analysis of the largest Tam-resistant microarrays (GSE6532, GSE9195, GSE9893) on human breast cancer samples found little overlap between the studies despite the fact that these studies found 275, 130, and 252 genes, respectively, differentially expressed between Tam-sensitive and Tam-resistant samples [262]. The fact that so few genes were found in multiple studies indicates the complexity of gene expression regulation in breast cancer cells as the develop resistance to Tam. Epigenetic studies of Tam-resistant samples have shown altered methylation status of 177 genes that is predictive for response to Tam [369]. A study of comparative genomic hybridization was performed on clinical samples, and the results indicated that numerous genomic alterations occurred from copy number changes to chromosomal rearrangements [370]. Taken together, these
148
findings exemplify the large scale of gene expression changes that occur in Tam resistance.
4.2.2. Potential biomarkers of Tam resistance
As genetic and immunohistological tests have become faster and more cost effective, the demand for biomarkers that predict treatment success has grown. Breast cancer specialists and even family physicians have readily adapted to the use of genetic screens to find BRCA1 and BRCA2, which interestingly, in addition to their prognostic value in determining a patient’s risk of developing breast cancer, these tests can predict sensitivity to DNA-damaging agents as well. The adoption of so many physicians to a genetic test indicates the readiness of the medical community to easy, predictive tests of therapy response. Already in practice, a common occurrence in breast cancer treatment is to screen for the presence of ER, PR, and ErbB2 in biopsy samples which predict sensitivity to tamoxifen.
While numerous biomarkers have been discovered that predict tamoxifen resistance, most of these rely on tumor biopsies. Because biopsies can often be traumatic and painful for patients, the medical community has expressed a desire for less invasive techniques for predicting therapy resistance. The development of a simple blood or urine test is the considered the ultimate goal for a diagnostic or predictive test. A prime example is a new bladder cancer detection test, Cxbladder® by Pacific Edge Limited, that predicts both the presence and grade of bladder cancers by detecting mRNA levels with a non-invasive urine test. 149
Already, initial studies are indicating that blood tests might be possible for certain miRNA and other protected DNA/RNA transcripts in the plasma. Free circulating mRNAs of cyclin D1 in breast cancer patient plasma have been correlated with poor outcome and non-responsiveness to Tam treatment [256]. As previously mentioned, a clinical trial examining miRNA expression in patients serum is underway to determine their use as a predictive marker for Tam resistance, as is another trial which is looking at proteins and genetic material in patients serum that are treated with Tam. As more information about the genetic changes behind tamoxifen and other therapy resistance is assembled, the greater the potential for finding a simple blood screen which would indicate what therapy to give a patient, saving thousands of dollars and often years of a patient’s life.
4.2.3. Current findings: major gene expression changes found by next generation sequencing
The current study used next generation to develop of profile of genes and gene clusters that were differentially expressed in Tam-resistant breast cancer cells. The new analytical clustering method set up a framework for determining genes that had similar levels of change in expression while taking into account the different treatment environments. The hope in the future is that our clustering system could be used for larger clinical sample samples to help differentiate driver genetic changes from passenger changes. By examining the type of genes that cluster together at the sample level of
150
changes in gene expression, new candidates for passenger genetic changes might be found which control entire clusters or genes within clusters.
4.2.3.1. The importance of analytical methods in determining differential gene expression.
Our development of a new analytical model for discovering differential gene expression across different treatment groups is critical for producing an accurate gene expression profile for tamoxifen resistance. Some previous models have used standard z- tests or Poisson distributions to determine differential gene expression that do not take into account the large sample sizes or the nuances of single gene comparison. Our comparison of these methods with our FET method for determination of differential gene expression revealed dramatically different results for the two methods. While the
Poisson distribution found only 667 mRNAs that were differentially expressed between
TamS and TamR cell lines, FET found 1215 genes; only 150 genes were found by both methods indicating that the method of determining differential gene expression can alter results dramatically. These findings should be taken into account when examining previous microarray data on Tam resistance.
Traditional approaches for pattern identification are based on cluster analysis for gene expression in one replicate [224], without considering the mechanisms behind differential expression. The environment of a Tam-resistant cancer cell treated with
Tam-media is different from the non-stressful environment of a Tam-sensitive cancer cell in control-media. Previous studies have not accounted for this difference in environment.
151
Gene clusters have been determined according to their known functions into a predetermined number of clusters, which does not take into account phenotypic plasticity or how genes respond differently to different environmental signals. Our new model thus was able to determine clusters of genes that were regulated according to the different environments in which they were found without the bias of grouping genes according to similar functions.
The clusters revealed that most genes were moderately differentially expressed between TamS and TamR cells (Fig. 2.3-2.4). The difference model was found to be better able to account for transcripts that were only expressed in one cell line, which were missed by the ratio expression model, and was able to help visualize the extreme changes in expression in a few genes. Drawbacks of the difference model included its inability to reveal changes in genes that were lowly expressed. The ratio model was revealed to be the preferred method for such genes.
4.2.3.2. Important ontological groups altered by Tam resistance
Gene ontology and pathway analysis of our NGS differentially-expressed gene clusters revealed numerous gene expression changes that had been previously found in other Tam resistant studies (Table 2.2). Cluster 1 contained genes that were down- regulated in TamR cells, and ontological analysis revealed a high level of modification of mitochondrial oxidative phosphorylation and gene expression regulation mechanisms.
Slicesome genes, such as U2AF1 and U2AF2, were decreased as were transcription factors such as JUNB. Cluster 2 contained genes that were moderately over-expressed in
152
TamR, and once again, the primary transcripts in this cluster related to genes involved in the expression of transcripts and proteins. Transcripts from processes involved in epigenetics, such as histone-associated genes (HINT1) to those involved in protein synthesis, such as ribosomal proteins (RPSs and RPLs in addition to MRPLs and
MPRSs) and initiation/ elongation factors (eIFs and eEF1E1) to genes involved post- translational modifications, such as the proteasome (PMSs), were all increased in TamR cells revealing the multiple levels of gene expression regulation affected by Tam resistance. Cell cycle molecules (CDK1, CKD3, CCNB2, CCNC, RB1) and oxidative phosphorylation proteins (NDUFs and COXs) were also part of this cluster. Finally,
Cluster 3 included genes that were highly up-regulated in TamR, and again, mitochondrial proteins, such as were ATP synthases prominently dysregulated. Previous markers of drug resistance (CAV2) and proliferation (ESR1, HRAS, MAPK1) were also in
Cluster 2. Overall, the clusters indicated a general dysregulation in mitochondrial proteins and genes involved in gene expression.
4.2.3.3. Importance of small RNA in Tam resistance
The most interesting finding in the smRNA study was that the majority of differentially-expressed smRNAs were small nucleolar RNAs and other non-coding
RNAs. This category of smRNA has become increasingly more prevalent in the literature as additional functions of snoRNAs have been discovered. The original discovered role for snoRNAs in the cell was to chemically modify other RNA molecules, especially rRNA and tRNA [237]. As snoRNA has been studied further, it was revealed
153
that many snoRNAs can act as miRNA [238]. The large number of differentially expressed snoRNA in TamR cells indicate that they might play an important role in the overall difference in gene expression.
While numerous smRNAs were found in the NGS study, mapping to accurately annotated regions was an obstacle. While miRNA databases, such as miRBase, exist, these databases only catalogue known or predicated miRNA. Many of our smRNA transcripts mapped to the exons of annotated genes, but miRNA and other ncRNA studies have revealed that most ncRNA do not act on the genes that they are derived from. Thus, while we know where many of our smRNA transcripts come from, we were limited by database information in our designation of their function and classification.
The other large category of smRNA transcripts were those related to gene expression. We found transcripts related to acetyltransferases (MYST4) and methylators
(MBD1), which were prime examples of expression changes that can lead to the alteration of transcription by modification of histone acetylation and DNA methylation. Specific regulators of gene expression were found in the miRNAs. Some miRNAs identified by our study were previously characterized in other cancers such as mir-93, mir-1974, mir-
21, and mir-125A [241,242]. Changes in expression of these miRNAs have been implicated in the development of more invasive and metastatic phenotypes.
Overall, the existence of so many snoRNAs, miRNAs, and smRNA transcripts related to gene expression (histone modification, mitochondrial transcription, etc.) implicate the intricate regulation of a large set of gene expression changes in the development of Tam resistance. Epigenetic mechanisms like the histone and DNA
154
modifications genes we found, in addition to the other small RNA, may play a large role in the diversity of gene expression alterations found in Tam-resistant clinical samples.
4.2.5. Directions for future studies
The results presented in our NGS study have laid the groundwork both for future large-scale NGS projects and for focused mechanism studies. Our study highlights the ability of NGS to profile and characterize transcriptome changes in Tam-resistant breast cancer. The development of an analytical clustering model which accurately takes into account treatment differences could be used for any type of cancer or disease treatment study. Continued use of this model would validate its role in gene expression studies.
Further exploration of tamoxifen resistance will not only verify the results presented here and in previous global studies but will also be critical for the development of therapies that can circumvent the resistance mechanisms of Tam. The vast number of differentially-expressed genes indicates the importance of gene expression regulation mechanisms in Tam resistance, and finding that has been found but not highlighted by previous studies [262]. Whether they are driver or passenger changes, the range of alterations in Tam-resistant cells cannot be ignored. More and more evidence is being presented that Tam resistance is the result of global changes (Fig. 2.6) in gene expression and is not just the result of a change in an individual molecule or pathway. Thus, future therapeutic approaches to breast cancer therapy must keep this in mind.
In order to sort passenger from driver changes in gene expression, it has been suggested that a better database system that includes all results cited within PubMed 155
should be developed. As large-scale microarray studies continue to be used and next- generation sequencing becomes more common place, an enormous amount of data will be accumulated which could shed light on all biological processes. While not all information might be relevant to the objectives of individual scientist’s studies, the extraneous results they obtain about genetic changes may be vital for other researchers.
Thus, a central depository of searchable data from these large-scale studies would allow researchers to choose appropriate areas to focus their mechanistic studies. As more gene expression data from breast cancer cell lines and patient samples is accrued in a central location, the possibility of being able to decipher driver changes from passenger changes becomes possible.
First, it would allow researchers to see what genetic changes are consistently found in Tam resistant samples compared to other breast cancer samples. The addition of clinical data to the database such as tumor stage, treatments, and demographics would allow for more nuanced clustering studies that examine the genetic changes which are common amongst different types of breast cancer patients. A compilation of these common genes would allow for a better meta-analysis of the published studies.
However, this would still leave a large number of genetic changes. Thus, secondly, a compilation of the networks, pathways, and gene ontologies should be added to the database. This would involve a more public system for the determination of gene ontology than the individual programs like IPA and GeneGo. Looking at the pathways and types of genes that are changed might make it possible to find that genes that are driving the processes. For instance, a change in gene expression of a transcription factor
156
might explain the change in genes that are induced by such a factor, which would mean that the transcription factor alterations are the driver change for that set of genes. A more thorough, complete database would start to make the determination of driver genetic changes manageable. Finally, such a database might make it easier to determine which gene changes occur in acquired versus intrinsic Tam resistance. By including information about the time at which samples were taken, such as at diagnosis or at recurrence, a better picture of the progression of genetic changes would be developed.
Therefore, the ideal database would include all microarray and NGS studies of breast cancer tumors, control tissue, serum, and cell lines. It would have search criteria that discriminate for demographics, treatments, stage of breast cancer, recurrences, and gene ontologies. The ability to search by when samples were taken would give an additional level of depth to the data that would allow researchers to study the intrinsic mechanisms and progression Tam resistance and breast cancer leading to better diagnostic and therapeutic predictive biomarkers and therapeutic targets.
4.2.5.1. Moving from preclinical to clinical studies: the use of NGS in Tam resistant breast cancer tumor samples
The next step in validating our NGS results would be to use the analytical framework developed from our study in cell lines and use it to analyze clinical specimens with NGS. As previously mentioned, large microarray studies have examined Tam resistance in clinical breast cancer samples and focused NGS has been used to examine target genes found by an shRNA screen [223]. None of these large-scale studies
157
examined smRNA, which we believe to be playing a vital role in the gene expression changes seen in Tam resistant cell lines and tumors.
In an ideally examined experiment, clinical samples from a variety of patient populations would be used. Comparison of expression data among those treated with
Tam and those treated by other methods would help to elucidate mechanisms that are specific to Tam resistance and those that are characteristic of recurrences and advanced cancers regardless of treatments. Ideally, clinical samples would be collected before treatment and after the recurrence of the tumor in patients, in order to determine which gene expression changes were present from the start of treatment and which alterations occurred as treatment and resistance progressed. The clustering method could be used not only to determine novel biological relationships between genes, but to validate the predictive value of current Tam resistance biomarkers. It would also be interesting to see if samples from current prognostic groups, such as triple-negative breast cancer
(ER/PR/ErbB2-negative), would cluster together or if their expression would be more similar to other types of breast cancer.
While it might appear to be a backwards approach, pre-clinical models could then be undertaken in order to examine differences in clinical samples from samples obtained from in vivo animal studies. Prior knowledge of relevant NGS gene expression changes could be used to investigate the importance of gene expression changes found in animal studies. Tumor microenvironment could have a profound effect on gene expression of tam resistant cells in vivo, and cataloguing these differences will be helpful in accurately determining the significance of data gathered from animal studies. The use of animal and
158
a preclinical model for Tam resistance cannot be bypassed as it necessary for developing new therapies. Characterization of the differences in gene expression between clinical samples and mouse models will be vital in determining incidental findings from clinically relevant ones. The main established pre-clinical model uses xenografts of MCF-7 cells in ovariectomized young athymic mice supplemented with estrogen pellets until tumors develop, and are then placed into treatment categories which include E2, E2-withdrawal,
E2-withdrawal + Tam, and vehicle [371]. Both ATTC MCF-7 cells and characterized
Tam-resistant cell lines have been used previously in this model. One allows for examination of Tam development while the other allows for investigations into the changes in established resistant tumors.
As the goal of further NGS studies of Tam resistant tumors is to develop better therapies that can overcome or circumvent Tam resistance, NGS characterization of gene expression in breast cancer cell lines should be performed concurrently. Then, with cell culture, pre-clinical, and clinical results, the full picture of Tam resistance’s effect on gene expression will become more clear, allowing for better and more effective methods of treatment discovery and development.
4.2.5.2. SIRT3 findings and future experiments
The finding that human sirtuin SIRT3 levels could experimentally modulate sensitivity to Tam indicated that NGS is a valuable tool for the discovery of Tam resistance mechanisms. The increases seen in SIRT3 levels as TamS cells were exposed to Tam reveal that elevations of SIRT mRNA and protein levels in TamR cells are the
159
result of an early change during chronic exposure to Tam (Fig. 2.7). As patients are treated with Tam over a 5 year period, particular attention should be given to early events that are modulated by Tam treatment. Our preliminary results suggest that SIRT3 could play in important role in Tam resistance; however, the mechanism by which it affects sensitivity to Tam remains to be explored.
Since Tam is mostly a cytostatic drug at therapeutic levels, examination of SIRT3 effects on cell cycle might reveal potential mechanisms of Tam resistance through
SIRT3. Tam-resistant cells must overcome the cell cycle inhibition. Flow cytometry results of SIRT3 knockdown or over-expression studies might show changes in cell cycle. Initial hypotheses pointed towards the cooperation between cyclin D1and SIRT3 since cyclin D1 is increased in resistant tumors [254]and our cell line[231], but preliminary results (data not shown) indicate that alterations in SIRT3 levels have no effect on cyclin D1.
SIRT3 likely affects Tam resistance by multiple mechanisms as SIRT3 is a
NAD+ mitochondrial deacetylase with multiple known targets [372]. As a deacetylase,
SIRT3 may have a direct effect on gene expression of molecules that interact with estrogen receptor pathways. Over-expression of SIRT3 leads to increases in PPAR-γ- coactivator-1 (PGC-1) expression, a transcriptional co-activator, which could be the result of SIRT3’s ability to phosphorylate CREB that then activates the PGC-1 promoter.
PGC-1 was also found to be necessary for SIRT3 increases in expression of other genes
[373]. Interestingly, PGC-1 is a co-activator of ERα, which can shuttle between the mitochondria and nucleus [374]. It has been shown to interact with SRC-1 to promote
160
agonist activity of Tam, such that Tam treatment actually increases cancer cell proliferation through increased ERE transcriptional activity [375]. Thus, exploring the effect of SIRT3 and PGC-1 interactions on ERα in Tam-sensitive and Tam-resistant tumor cells might elucidate the mechanism by which SIRT3 affects Tam resistance in breast cancer.
4.2.5.3. Mitochondria and energy metabolism
Our global NGS study of gene expression has revealed a pattern of mitochondria and energy metabolism dysregulation in Tam resistant cells. Additionally, our focused study on SIRT3 implicated its importance in Tam resistance, and SIRT3 could affect energy metabolism as a mitochondrial deacetylase. “Reprogrammed energy metabolism” has been proposed as one of the emerging hallmarks of cancer [61]. Ever since Dr.
Warburg discovered that cancer cells prefer aerobic glycolysis over oxidative phosphorylation [24], researchers have been aware of the altered energy metabolism that cancer cells use in order to survive and grow. However, in recent years, a resurgence in cancer energy metabolism studies has led to a renewed interest in how cancer cells produce energy. New studies have actually found that while some cells still depend on glucose for energy, others cancer cells use their neighbor’s secreted lactate to create pyruvate which is then used by the Kreb’s cycle for oxidative phosphorylation [61].
Our clustering of NGS expression genes found numerous mitochondrial proteins that were up- or down-regulated (Table 2.2). A closer examination of the impact of these energy-related genes on Tam resistance might reveal new mechanisms of the regulation
161
of energy production, which are associated with new survival mechanisms of Tam resistant cells. It is possible that the up-regulation of SIRT3 found in TamR cells is responsible for such alterations in mitochondrial proteins. Many proteins involved in oxidative phosphorylation are on the mitochondrial genome, and expression of this genome would be affected by a mitochondrial deacetylase such as SIRT3. SIRT3 is also known to deacetylate and activate proteins such as acetyl coenzyme A synthetase which are involved directly in energy metabolism [373]. SIRT3 could have an effect on multiple enzymes related to metabolism, and these effects could alter resistance to Tam as our Tam resistance NGS study revealed a dysregulation of energy metabolism.
Further study of SIRT3 and other metabolic enzymes might elucidate novel mechanisms of resistance to Tam.
4.3. Regulation of eEF-2K in response to metabolic stress
eEF-2K is an important kinase involved in the regulation of translation elongation. Proteins levels of eEF-2K are increased in tumor samples, and it has been shown to play a role in cell survival through inhibition of protein synthesis. Post- translational modification of translation machinery is important for its regulation and could be critical for survival of cancer cells encountering stress. Thus, we examined how
EF-2K is regulated during stress, focusing on how the phosphorylation status of eEF-2K affects the enzyme’s turnover in cancer cells.
162
4.3.1. Protein synthesis in cancer cells
Protein synthesis is a dynamic process in cancer cells. As tumors encounter stresses throughout their progression, these stresses can actually affect protein synthesis, allowing for translation of pro-survival proteins while inhibiting the translation of other proteins. Cancer cells have developed a myriad of mechanisms to alter protein synthesis in such a way that it aids in cancer cell survival. These include modifications in the cellular mRNA, synthesis of miRNA and other non-coding RNA that prevent translation of pro-apoptotic proteins, and changes in RNA-binding proteins [376]. Alterations in the translational machinery and the enzymes that regulate protein synthesis are common.
The increase of eEF-2K seen in cancer cells ultimately affects protein synthesis in a manner that ensures survival under a variety of conditions including cancer treatment
[352]. Even indirect events, such as activation of many growth factor and survival pathways, can ultimately lead to altered protein synthesis [61]. The end goal of this alteration in translation is to favor cancer cells survival under the existing conditions they are experiencing and to be able to readily adapt to new environmental changes that occur throughout tumor progression.
4.3.2. Metabolic stress effects on protein phosphorylation
The cell’s multiple stress pathways lead to a variety of protein phosphorylation during different stress conditions. General stress responders in the SAPK/JNK pathway phosphorylate other MAPKs in a signaling cascade during stresses such as hypoxia, radiation, oxidative stress, and drug therapy; the end result is [39]. Hypoxia can also 163
cause an increase in phosphorylation of HIF-1 which leads to increased transcription of genes involved in cell survival and metabolism [52]. Activation of transcription factor
NF-κB occurs after phosphorylation of its inhibitory regulators which then releases NF-
κB, leading to rapid the rapid transcription of important genes involved with the stress response [53]. Phosphorylation of eEF-2K by SAPK pathways and AMPK leads to activation of the kinase which prevents translational elongation during stresses [308,313].
Thus, these examples represent the variety of signaling cascades and the corresponding changes in phosphorylation of the targets enzymes which play an important part in the stress response of cells.
4.3.3. Current findings: Phosphorylation at specific sites of eEF-2K differentially modulates its turnover
The study presented here reveals that phosphorylation plays an important role in the stability of eEF-2K protein. The phosphorylation of eEF-2K is dependent on upstream signaling pathways that determine which sites will be phosphorylated under different cellular conditions. Understanding the circumstances under which eEF-2K is phosphorylated and which sites are pivotal to the enzyme’s stability will lead to a more complete picture of how protein synthesis is regulated and the role of eEF-2K as a mediator of cell signaling cascades in the regulation of translation.
164
4.3.3.1. Cellular stress alters eEF-2K protein stability
eEF-2K was previously shown to have a half-life of ~ 8 hours under normal culture conditions [329]; this finding was corroborated by this study. Since it is known that activation of eEF-2K can induce autophagy [344,350], which occurs when cells are under stress, we were interested how stress affects eEF-2K levels. Under hypoxic and nutrient-deprived conditions, eEF-2K protein levels were increased (Fig. 3.2). This increase in eEF-2K protein levels was not accounted for in the mRNA levels, as eEF-2K mRNA only transiently increases, quickly falling back to basal levels of expression (Fig.
3.2C). We hypothesized that metabolic stress, therefore, increased eEF-2K stability.
This hypothesis agreed with the speculation that cells decrease the energy-consuming elongation process in order to survive stress. However, we found that eEF-2K turnover was increased under all stress conditions, resulting in half-life of only 2 hours (Fig. 3.3).
This unexpected result lead us to examine the impact of phosphorylation on eEF-2K, as stress is known to signal through multiple pathways that could affect eEF-2K activity
[308,314]. Decreased stability of eEF-2K under stress would affect the efficacy of any drug that would target this protein and might prevent a barrier for targeted therapy.
4.3.3.2. Identification of upstream signaling pathways and phosphorylation sites affecting eEF-2K protein turnover
Multiple upstream pathways are known to affect eEF-2K activity, including the mTOR and AMPK signaling cascades. mTOR pathway signaling decreases eEF-2K activity by phosphorylating eEF-2K on Ser78 and Ser366 [296], while AMPK increases
165
eEF-2K activity by phosphorylating Ser398[308]. In order to determine the impact of these upstream pathway-regulated phosphorylations of eEF-2K on the enzyme’s stability, we generated phosphorylation-defective mutants. We found that mutating the mTOR sites, S78/366A, caused an increased stability of eEF-2K under normal conditions; however, under stress conditions, eEF-2K stability reverted back to normal levels (t ½ ~ 8 hrs) (Fig. 3.4A). On the other hand, mutation of the AMPK site, S398A, stabilized eEF-
2K protein under all conditions, including stress (t ½ > 24 hrs) (Fig. 3.4B). These results indicated the importance of the Ser398 site on EF-2K stability. The role of the upstream signaling cascades, however, was less clear.
Results from pharmacological inhibition of the mTOR and AMPK pathways did not agree with the mutation findings. Treatment with rapamycin, an mTOR inhibitor, led to a decrease in eEF-2K turnover, stabilizing the protein under all conditions (t ½ > 24 hrs) (Fig. 3.5). AMPK inhibition with Compound C caused a decrease of eEF-2K under all conditions (t½ ~ 4 hrs) (Fig. 3.6). Pharmacological effects on eEF-2K protein turnover did not seem to be affected by stress. Thus, it appears that these inhibitors might be affecting other factors that regulate eEF-2K stability and possibly the phosphorylation of additional sites on eEF-2K. These findings are significant because they have implications for unplanned modulation of protein synthesis by targeting upstream pathways such as mTOR which have been suggested as therapeutic targets.
166
4.3.4. Directions for further studies
The results presented in our eEF-2K study indicate that eEF-2K is a tightly regulated molecule with a variety of phosphorylation sites that can affect both its stability and activity. Determining the significance of its tight regulation could result in a greater understanding of protein translation in cancer cells, both under normal conditions and in response to stress. Cancer cells rely on the translation of certain proteins that promote their survival, and discovering a way to circumvent their control on translation could lead to new therapeutics or greater efficacy of current treatments. Since so many signaling pathways can promote protein synthesis, regulation of translation might be an efficient target for new cancer therapies as it appears to be an area of convergence these pathways.
Therapies directed towards only one of these upstream pathways could still signal through other pathways ultimately leading to the same changes in protein synthesis.
Taking advantage of the altered protein synthesis in cancer cells might prove to be a valuable treatment strategy. A kinase such as eEF-2K which is increased in cancer cells and is known to promote survival should be examined further for its role in protein synthesis under different conditions as it may lead to new therapeutic targets.
As eEF-2K is expressed in all human tissues, there is concern that targeting eEF-
2K alone would be detrimental to the healthy cells of the body. eEF-2K has primarily been proposed as a therapeutic target to be used in combination with other treatments.
Knocking down of eEF-2K makes cells more susceptible to stress but does not outright kill the majority of cells. Therefore, using it in combination with inhibitors of other over- expressed pathways in cancers would lead to the death of cells reliant on both eEF-2K 167
and the up-regulated growth cascades, while leaving other cells relatively unharmed.
Additionally, by determining which sites on eEF-2K are phosphorylated by individual stresses and the role such phosphorylation plays on eEF-2K stability, such findings might allow researchers to develop drugs against eEF-2K that only work on specific phosphorylated forms of eEF-2K that are the result of the stressful tumor environment while leaving the healthy, unstressed tissue alone.
4.3.4.1. Identification of additional upstream pathways that phosphorylate eEF-2K under stressful conditions
The primary upstream pathways examined in this study were the mTOR and
AMPK pathways. Since pharmacological inhibition of these upstream pathways produced results that were more in line with our original hypotheses, it would be interesting to see if genetic silencing of the various signaling components would have the same effect. Many pharmacological inhibitors are known to have off-target effects, and as eEF-2K has demonstrated that it has phosphorylation sites from a variety of different upstream pathways, it is possible that the drug inhibitors are affecting additional proteins which have an effect on eEF-2K or other upstream molecules.
Additionally, other major signaling pathways should be examined for their effect on eEF-2K phosphorylation and stability. As glioma is known to have increased activity of EGFR and PDGFR [13], inhibition of these growth factor receptors with gefitinib and imatinib might affect eEF-2K phosphorylation. Since a decrease in growth factor signaling would be an added type of stress on tumor cells, it would be interesting to see if
168
the effect on eEF-2K phosphorylation and turnover would be different when growth factor inhibition was added on top of metabolic stresses such as hypoxia and nutrient deprivation.
4.3.4.2. Examination of discordant mRNA and protein levels
The fact that stress caused increased levels of eEF-2K protein despite a decrease in the enzyme’s stability and no increase in mRNA levels raises the question of whether eEF-2K mRNA is translated more efficiently. Our results show that eEF-2K mRNA levels are not increased for extending periods of time upon exposure of cancer cells to stress (Fig. 3.2), which indicates that the increase in eEF-2K protein is probably not a result of increased eEF-2K mRNA stability. However, since eEF-2K levels did not decrease below baseline, it is feasible that some eEF-2K mRNA might be stabilized while other eEF-2K transcripts are translated. This would hold a reserve of eEF-2K mRNA, which could either be translated or degraded, depending on how the cellular conditions changed.
We did find that eEF-2K mRNA contains predicted IRESs. This would allow eEF-2K to be translated by internal ribosome entry and initiation even when global protein synthesis is reduced. Since reduction of global protein synthesis at the level of elongation can be caused by increases in eEF-2K and its phosphorylation of eEF-2, it is possible that cell would need to have continued translation of eEF-2K to extend inhibition of general translation.
169
Further examination of eEF-2K mRNA could help to elucidate the mechanism behind the discordant mRNA and protein levels. Additional work on eEF-2K RNA elements might provide insight into how eEF-2K is translated. Reporter constructs of eEF-2K mRNA could be made to investigate if eEF-2K is translated during periods of stress and during global protein synthesis inhibition. Fluorescent microscopy would be an efficient method to examine if eEF-2K mRNA is clustered in particular areas of the cell either at ribosomes or perhaps in cytoplasmic stress granules [377].
4.3.4.3. Effect of eEF-2K phosphorylation on specific versus global protein synthesis
The delicate balance of the phosphorylation sites in eEF-2K and the complex regulation of the enzyme’s stability seem to suggest that the cell needs to have the ability to tightly regulate eEF-2K activity. Together, with the discordant data on the levels of eEF-2K mRNA and protein levels, this could indicate that eEF-2K might be able to regulate different types of protein synthesis, such as situation-specific or global translation. Perhaps the phosphorylation of eEF-2K can affect the rate of elongation and thus its activity might help determine which transcripts are translated at certain times and under different conditions. An investigation into the effect of eEF-2K phosphorylation and stability on general translation might reveal new mechanisms of protein synthesis regulation. Reporter constructs of different types of mRNAs that have been shown to be translated under different conditions might elucidate what type protein synthesis is regulated by eEF-2K, and eEF-2K phosphorylation mutants could be used to examine if the phosphorylation of eEF-2K effects these different types of translation. The mRNA
170
transcripts could be examined for functional element and sequence similarities between transcripts that are translated under similar conditions regulated by eEF-2K.
Understanding how eEF-2K affects general or specific protein synthesis might be key to finding a novel target for therapies that block dysregulated translation in cancer cells.
4.4. Epilogue: The importance of stress response in cancer research
While the use of clinical samples in cancer research has grown in the last few decades, the limitations in access to such samples by many institutions means that basic bench and cell culture studies are still the foundation of cancer research. The results from cell culture studies are foten used to direct further pre-clinical studies. Drug discovery is performed in cell culture models before experiment progress into animal-based models.
One of the dangers of this practice highlighted by this thesis is that cell culture is not the
“natural” cancer cell environment. Cancer cells in the body and within tumors are actually in a stressful environment where very rarely are nutrients in abundant supply and the body is actively trying to rid itself of these cancer cells. However, our cell culture experiments were performed under ideal growth conditions for cancer cells. Carefully balanced media is used in cell culture, which is full of necessary nutrients, while the cells are then left to grow in incubators calibrated perfectly for temperature and oxygen content. What many of these studies do not take into consideration is the fact that this
171
environment is not what cancer cells usually encounter; it is not the typical cancer cell environment.
Studies on the various cellular processes are often carried out in immortalized cell lines of non-cancerous tissues. Transcription and translation factors were discovered using such cell systems, and their role in cancer was verified in tissue-culture using cancer cell lines. These studies are used for the basis of animal and clinical studies.
However, they do not take into account that cancer cells in animals and clinical studies are under many more stresses than the perfectly balanced conditions of cell culture.
Some of these basic cellular processes might perform differently under stress, as was exemplified by our study of how stress affects eEF-2K, a molecule important in the regulation of protein synthesis. Protein synthesis is so fundamental to cancer cell survival, that differences in its function under stress could have broad implications for other cancer studies. The majority of cancer cell culture studies only examine the role of their protein or molecule of interest under ideal culture conditions, and these studies would benefit from examination of their molecule under more natural cancer cell conditions that include stress. The addition of such experiments to studies might increase the significance and the rate of validation in later pre-clinical and clinical models, thus increasing our chances of discovering more effective cancer therapies.
Our NGS study exemplifies the complexity of alterations caused by exogenous stresses on cancer cells. The wide variety of gene expression changes from genes involved in proliferation and survival to those that regulate energy metabolism and post- transcriptional indicate that stress selection can alter the overall functioning of the cancer
172
cell. This study represents a new era of genomics that explore the global gene expression changes in cancer cells as they develop, progress, and are subjected to treatments. Such technology could and should be used to explore these alterations in cell culture systems and to investigate how diversely different stresses can affect a cell population.
Understanding how gene expression and proteins are altered in cancer cells by different intrinsic and extrinsic stresses will help current and future researchers to develop novel and more effective methods of overcoming cancer cell survival and resistance to current treatments.
173
REFERENCES
1. Society AC (2012) Cancer Facts & Figures 2012. Atlanta: American Cancer Society. 2. Nomura K (2005) Cancer Statistics in Japan-2005. Tokyo. 3. Sherry L. Murphy JX, Kenneth D. Kochanek (2012) Deaths: Preliminary Data for 2010. 1-68 p. 4. Society AC (2011) Breast Cancer Facts & Figures 2011-2012. Atlanta: American Cancer Society, Inc. 5. Siegel R, Ward E, Brawley O, Jemal A (2011) Cancer statistics, 2011: the impact of eliminating socioeconomic and racial disparities on premature cancer deaths. CA Cancer J Clin 61: 212-236. 6. Barres BA (2008) The mystery and magic of glia: a perspective on their roles in health and disease. Neuron 60: 430-440. 7. Overstreet LS (2005) Quantal transmission: not just for neurons. Trends Neurosci 28: 59-62. 8. Grossman SA, Batara JF (2004) Current management of glioblastoma multiforme. Semin Oncol 31: 635-644. 9. Hou LC, Veeravagu A, Hsu AR, Tse VC (2006) Recurrent glioblastoma multiforme: a review of natural history and management options. Neurosurg Focus 20: E5. 10. Davis FG, Freels S, Grutsch J, Barlas S, Brem S (1998) Survival rates in patients with primary malignant brain tumors stratified by patient age and tumor histological type: an analysis based on Surveillance, Epidemiology, and End Results (SEER) data, 1973-1991. J Neurosurg 88: 1-10. 11. Ohgaki H, Dessen P, Jourde B, Horstmann S, Nishikawa T, et al. (2004) Genetic pathways to glioblastoma: a population-based study. Cancer Res 64: 6892- 6899. 12. Karlbom AE, James CD, Boethius J, Cavenee WK, Collins VP, et al. (1993) Loss of heterozygosity in malignant gliomas involves at least three distinct regions on chromosome 10. Hum Genet 92: 169-174. 13. Watanabe K, Tachibana O, Sata K, Yonekawa Y, Kleihues P, et al. (1996) Overexpression of the EGF receptor and p53 mutations are mutually exclusive in the evolution of primary and secondary glioblastomas. Brain Pathol 6: 217-223; discussion 223-214. 14. Nakamura M, Watanabe T, Klangby U, Asker C, Wiman K, et al. (2001) p14ARF deletion and methylation in genetic pathways to glioblastomas. Brain Pathol 11: 159-168. 15. Louis DN (2006) Molecular pathology of malignant gliomas. Annu Rev Pathol 1: 97-117.
174
16. Huse JT, Holland EC (2010) Targeting brain cancer: advances in the molecular pathology of malignant glioma and medulloblastoma. Nat Rev Cancer 10: 319-331. 17. Brandes AA, Tosoni A, Franceschi E, Sotti G, Frezza G, et al. (2009) Recurrence pattern after temozolomide concomitant with and adjuvant to radiotherapy in newly diagnosed patients with glioblastoma: correlation With MGMT promoter methylation status. J Clin Oncol 27: 1275-1279. 18. Ichimura K, Ohgaki H, Kleihues P, Collins VP (2004) Molecular pathogenesis of astrocytic tumours. J Neurooncol 70: 137-160. 19. Wrensch M, Lee M, Miike R, Newman B, Barger G, et al. (1997) Familial and personal medical history of cancer and nervous system conditions among adults with glioma and controls. Am J Epidemiol 145: 581-593. 20. Relling MV, Rubnitz JE, Rivera GK, Boyett JM, Hancock ML, et al. (1999) High incidence of secondary brain tumours after radiotherapy and antimetabolites. Lancet 354: 34-39. 21. Wen PY, Kesari S (2008) Malignant gliomas in adults. N Engl J Med 359: 492-507. 22. Gimbrone MA, Jr., Leapman SB, Cotran RS, Folkman J (1973) Tumor angiogenesis: iris neovascularization at a distance from experimental intraocular tumors. J Natl Cancer Inst 50: 219-228. 23. Vajkoczy P, Menger MD (2004) Vascular microenvironment in gliomas. Cancer Treat Res 117: 249-262. 24. Warburg O (1956) On respiratory impairment in cancer cells. Science 124: 269- 270. 25. Vaupel P, Kallinowski F, Okunieff P (1989) Blood flow, oxygen and nutrient supply, and metabolic microenvironment of human tumors: a review. Cancer Res 49: 6449-6465. 26. Chiche J, Brahimi-Horn MC, Pouyssegur J (2010) Tumour hypoxia induces a metabolic shift causing acidosis: a common feature in cancer. J Cell Mol Med 14: 771-794. 27. Brat DJ, Van Meir EG (2004) Vaso-occlusive and prothrombotic mechanisms associated with tumor hypoxia, necrosis, and accelerated growth in glioblastoma. Lab Invest 84: 397-405. 28. Li JL, Sainson RC, Shi W, Leek R, Harrington LS, et al. (2007) Delta-like 4 Notch ligand regulates tumor angiogenesis, improves tumor vascular function, and promotes tumor growth in vivo. Cancer Res 67: 11244-11253. 29. Vaupel P, Schlenger K, Knoop C, Hockel M (1991) Oxygenation of human tumors: evaluation of tissue oxygen distribution in breast cancers by computerized O2 tension measurements. Cancer Res 51: 3316-3322. 30. Liu L, Wise DR, Diehl JA, Simon MC (2008) Hypoxic reactive oxygen species regulate the integrated stress response and cell survival. J Biol Chem 283: 31153-31162.
175
31. Tan TT, Degenhardt K, Nelson DA, Beaudoin B, Nieves-Neira W, et al. (2005) Key roles of BIM-driven apoptosis in epithelial tumors and rational chemotherapy. Cancer Cell 7: 227-238. 32. Qian BZ, Pollard JW (2010) Macrophage diversity enhances tumor progression and metastasis. Cell 141: 39-51. 33. Egeblad M, Nakasone ES, Werb Z (2010) Tumors as organs: complex tissues that interface with the entire organism. Dev Cell 18: 884-901. 34. Grivennikov SI, Greten FR, Karin M (2010) Immunity, inflammation, and cancer. Cell 140: 883-899. 35. Colotta F, Allavena P, Sica A, Garlanda C, Mantovani A (2009) Cancer-related inflammation, the seventh hallmark of cancer: links to genetic instability. Carcinogenesis 30: 1073-1081. 36. Xie K, Huang S (2003) Regulation of cancer metastasis by stress pathways. Clin Exp Metastasis 20: 31-43. 37. Harrison LB, Chadha M, Hill RJ, Hu K, Shasha D (2002) Impact of tumor hypoxia and anemia on radiation therapy outcomes. Oncologist 7: 492-508. 38. Dal-Pizzol F, Ritter C, Klamt F, Andrades M, da Frota ML, Jr., et al. (2003) Modulation of oxidative stress in response to gamma-radiation in human glioma cell lines. J Neurooncol 61: 89-94. 39. Benhar M, Dalyot I, Engelberg D, Levitzki A (2001) Enhanced ROS production in oncogenically transformed cells potentiates c-Jun N-terminal kinase and p38 mitogen-activated protein kinase activation and sensitization to genotoxic stress. Mol Cell Biol 21: 6913-6926. 40. Miligkos V, Tiligada E, Papamichael K, Ypsilantis E, Delitheos A (2000) Anticancer drugs as inducers of thermotolerance in yeast. Folia Microbiol (Praha) 45: 339-342. 41. Nadin SB, Vargas-Roig LM, Cuello-Carrion FD, Ciocca DR (2003) Deoxyribonucleic acid damage induced by doxorubicin in peripheral blood mononuclear cells: possible roles for the stress response and the deoxyribonucleic acid repair process. Cell Stress Chaperones 8: 361-372. 42. Nagy JA, Chang SH, Shih SC, Dvorak AM, Dvorak HF (2010) Heterogeneity of the tumor vasculature. Semin Thromb Hemost 36: 321-331. 43. Kultz D (2005) Molecular and evolutionary basis of the cellular stress response. Annu Rev Physiol 67: 225-257. 44. Feder ME, Hofmann GE (1999) Heat-shock proteins, molecular chaperones, and the stress response: evolutionary and ecological physiology. Annu Rev Physiol 61: 243-282. 45. Calderwood SK, Khaleque MA, Sawyer DB, Ciocca DR (2006) Heat shock proteins in cancer: chaperones of tumorigenesis. Trends Biochem Sci 31: 164-172. 46. Goetz MP, Toft DO, Ames MM, Erlichman C (2003) The Hsp90 chaperone complex as a novel target for cancer therapy. Ann Oncol 14: 1169-1176.
176
47. Ciocca DR, Calderwood SK (2005) Heat shock proteins in cancer: diagnostic, prognostic, predictive, and treatment implications. Cell Stress Chaperones 10: 86-103. 48. Herr I, Debatin KM (2001) Cellular stress response and apoptosis in cancer therapy. Blood 98: 2603-2614. 49. Maki RG, Old LJ, Srivastava PK (1990) Human homologue of murine tumor rejection antigen gp96: 5'-regulatory and coding regions and relationship to stress-induced proteins. Proc Natl Acad Sci U S A 87: 5658-5662. 50. Sanchez-Prieto R, Rojas JM, Taya Y, Gutkind JS (2000) A role for the p38 mitogen-acitvated protein kinase pathway in the transcriptional activation of p53 on genotoxic stress by chemotherapeutic agents. Cancer Res 60: 2464- 2472. 51. Szabo D, Keyzer H, Kaiser HE, Molnar J (2000) Reversal of multidrug resistance of tumor cells. Anticancer Res 20: 4261-4274. 52. Jiang BH, Rue E, Wang GL, Roe R, Semenza GL (1996) Dimerization, DNA binding, and transactivation properties of hypoxia-inducible factor 1. J Biol Chem 271: 17771-17778. 53. Garg A, Aggarwal BB (2002) Nuclear transcription factor-kappaB as a target for cancer drug development. Leukemia 16: 1053-1068. 54. Hjelmeland AB, Wu Q, Heddleston JM, Choudhary GS, MacSwords J, et al. (2011) Acidic stress promotes a glioma stem cell phenotype. Cell Death Differ 18: 829-840. 55. Bolhuis H, van Veen HW, Poolman B, Driessen AJ, Konings WN (1997) Mechanisms of multidrug transporters. FEMS Microbiol Rev 21: 55-84. 56. Huang Y, Anderle P, Bussey KJ, Barbacioru C, Shankavaram U, et al. (2004) Membrane transporters and channels: role of the transportome in cancer chemosensitivity and chemoresistance. Cancer Res 64: 4294-4301. 57. Chen KG, Valencia JC, Lai B, Zhang G, Paterson JK, et al. (2006) Melanosomal sequestration of cytotoxic drugs contributes to the intractability of malignant melanomas. Proc Natl Acad Sci U S A 103: 9903-9907. 58. Hicks KO, Pruijn FB, Baguley BC, Wilson WR (2001) Extravascular transport of the DNA intercalator and topoisomerase poison N-[2- (Dimethylamino)ethyl]acridine-4-carboxamide (DACA): diffusion and metabolism in multicellular layers of tumor cells. J Pharmacol Exp Ther 297: 1088-1098. 59. Sladek NE (1999) Aldehyde dehydrogenase-mediated cellular relative insensitivity to the oxazaphosphorines. Curr Pharm Des 5: 607-625. 60. Zhang J, Tian Q, Chan SY, Duan W, Zhou S (2005) Insights into oxazaphosphorine resistance and possible approaches to its circumvention. Drug Resist Updat 8: 271-297. 61. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100: 57-70. 62. Houghton J, Morozov A, Smirnova I, Wang TC (2007) Stem cells and cancer. Semin Cancer Biol 17: 191-203. 177
63. Kuci S, Kuci Z, Latifi-Pupovci H, Niethammer D, Handgretinger R, et al. (2009) Adult stem cells as an alternative source of multipotential (pluripotential) cells in regenerative medicine. Curr Stem Cell Res Ther 4: 107-117. 64. Polyak K, Hahn WC (2006) Roots and stems: stem cells in cancer. Nat Med 12: 296-300. 65. Jordan CT (2009) Cancer stem cells: controversial or just misunderstood? Cell Stem Cell 4: 203-205. 66. Pallini R, Ricci-Vitiani L, Banna GL, Signore M, Lombardi D, et al. (2008) Cancer stem cell analysis and clinical outcome in patients with glioblastoma multiforme. Clin Cancer Res 14: 8205-8212. 67. Stingl J, Caldas C (2007) Molecular heterogeneity of breast carcinomas and the cancer stem cell hypothesis. Nat Rev Cancer 7: 791-799. 68. Liu G, Yuan X, Zeng Z, Tunici P, Ng H, et al. (2006) Analysis of gene expression and chemoresistance of CD133+ cancer stem cells in glioblastoma. Mol Cancer 5: 67. 69. Badve S, Nakshatri H (2012) Breast-cancer stem cells-beyond semantics. Lancet Oncol 13: e43-48. 70. Scharpfenecker M, Kruse JJ, Sprong D, Russell NS, Ten Dijke P, et al. (2009) Ionizing radiation shifts the PAI-1/ID-1 balance and activates notch signaling in endothelial cells. Int J Radiat Oncol Biol Phys 73: 506-513. 71. Kang MK, Kang SK (2007) Tumorigenesis of chemotherapeutic drug-resistant cancer stem-like cells in brain glioma. Stem Cells Dev 16: 837-847. 72. White E, DiPaola RS (2009) The double-edged sword of autophagy modulation in cancer. Clin Cancer Res 15: 5308-5316. 73. Kuma A, Hatano M, Matsui M, Yamamoto A, Nakaya H, et al. (2004) The role of autophagy during the early neonatal starvation period. Nature 432: 1032- 1036. 74. Abeliovich H, editor (2003) Regulation of Autophagy by the Target of Rapamycin (Tor) Proteins. University of Michigan: Landes Bioscence. 75. Paglin S, Hollister T, Delohery T, Hackett N, McMahill M, et al. (2001) A novel response of cancer cells to radiation involves autophagy and formation of acidic vesicles. Cancer Res 61: 439-444. 76. Kanzawa T, Germano IM, Komata T, Ito H, Kondo Y, et al. (2004) Role of autophagy in temozolomide-induced cytotoxicity for malignant glioma cells. Cell Death Differ 11: 448-457. 77. Harhaji-Trajkovic L, Vilimanovich U, Kravic-Stevovic T, Bumbasirevic V, Trajkovic V (2009) AMPK-mediated autophagy inhibits apoptosis in cisplatin-treated tumor cells. J Cell Mol Med. 78. Abedin MJ, Wang D, McDonnell MA, Lehmann U, Kelekar A (2007) Autophagy delays apoptotic death in breast cancer cells following DNA damage. Cell Death Differ 14: 500-510.
178
79. Takeuchi H, Kanzawa T, Kondo Y, Kondo S (2004) Inhibition of platelet-derived growth factor signalling induces autophagy in malignant glioma cells. Br J Cancer 90: 1069-1075. 80. Yang ZJ, Chee CE, Huang S, Sinicrope FA (2011) The role of autophagy in cancer: therapeutic implications. Mol Cancer Ther 10: 1533-1541. 81. Perona R (2006) Cell signalling: growth factors and tyrosine kinase receptors. Clin Transl Oncol 8: 77-82. 82. Hynes NE, MacDonald G (2009) ErbB receptors and signaling pathways in cancer. Curr Opin Cell Biol 21: 177-184. 83. Yuan TL, Cantley LC (2008) PI3K pathway alterations in cancer: variations on a theme. Oncogene 27: 5497-5510. 84. Mack PC, Holland WS, Burich RA, Sangha R, Solis LJ, et al. (2009) EGFR mutations detected in plasma are associated with patient outcomes in erlotinib plus docetaxel-treated non-small cell lung cancer. J Thorac Oncol 4: 1466-1472. 85. Davies MA, Samuels Y (2010) Analysis of the genome to personalize therapy for melanoma. Oncogene 29: 5545-5555. 86. Jiang BH, Liu LZ (2009) PI3K/PTEN signaling in angiogenesis and tumorigenesis. Adv Cancer Res 102: 19-65. 87. Pohlmann PR, Mayer IA, Mernaugh R (2009) Resistance to Trastuzumab in Breast Cancer. Clin Cancer Res 15: 7479-7491. 88. Heidorn SJ, Milagre C, Whittaker S, Nourry A, Niculescu-Duvas I, et al. (2010) Kinase-dead BRAF and oncogenic RAS cooperate to drive tumor progression through CRAF. Cell 140: 209-221. 89. O'Reilly KE, Rojo F, She QB, Solit D, Mills GB, et al. (2006) mTOR inhibition induces upstream receptor tyrosine kinase signaling and activates Akt. Cancer Res 66: 1500-1508. 90. Lowe SW, Cepero E, Evan G (2004) Intrinsic tumour suppression. Nature 432: 307-315. 91. Coley HM (2008) Mechanisms and strategies to overcome chemotherapy resistance in metastatic breast cancer. Cancer Treat Rev 34: 378-390. 92. Burkhart DL, Sage J (2008) Cellular mechanisms of tumour suppression by the retinoblastoma gene. Nat Rev Cancer 8: 671-682. 93. Dimri GP (2005) What has senescence got to do with cancer? Cancer Cell 7: 505- 512. 94. Bergers G, Hanahan D (2008) Modes of resistance to anti-angiogenic therapy. Nat Rev Cancer 8: 592-603. 95. Crawford Y, Kasman I, Yu L, Zhong C, Wu X, et al. (2009) PDGF-C mediates the angiogenic and tumorigenic properties of fibroblasts associated with tumors refractory to anti-VEGF treatment. Cancer Cell 15: 21-34. 96. Xiong YQ, Sun HC, Zhang W, Zhu XD, Zhuang PY, et al. (2009) Human hepatocellular carcinoma tumor-derived endothelial cells manifest increased angiogenesis capability and drug resistance compared with normal endothelial cells. Clin Cancer Res 15: 4838-4846. 179
97. Francia G, Emmenegger U, Kerbel RS (2009) Tumor-associated fibroblasts as "Trojan Horse" mediators of resistance to anti-VEGF therapy. Cancer Cell 15: 3-5. 98. Law LW (1952) Origin of the resistance of leukaemic cells to folic acid antagonists. Nature 169: 628-629. 99. Knutsen T, Mickley LA, Ried T, Green ED, du Manoir S, et al. (1998) Cytogenetic and molecular characterization of random chromosomal rearrangements activating the drug resistance gene, MDR1/P-glycoprotein, in drug-selected cell lines and patients with drug refractory ALL. Genes Chromosomes Cancer 23: 44-54. 100. Baylin SB, Herman JG, Graff JR, Vertino PM, Issa JP (1998) Alterations in DNA methylation: a fundamental aspect of neoplasia. Adv Cancer Res 72: 141-196. 101. Baylin SB, Jones PA (2011) A decade of exploring the cancer epigenome - biological and translational implications. Nat Rev Cancer 11: 726-734. 102. Lafarge S, Sylvain V, Ferrara M, Bignon YJ (2001) Inhibition of BRCA1 leads to increased chemoresistance to microtubule-interfering agents, an effect that involves the JNK pathway. Oncogene 20: 6597-6606. 103. Haber DA, Beverley SM, Kiely ML, Schimke RT (1981) Properties of an altered dihydrofolate reductase encoded by amplified genes in cultured mouse fibroblasts. J Biol Chem 256: 9501-9510. 104. Miller TE, Ghoshal K, Ramaswamy B, Roy S, Datta J, et al. (2008) MicroRNA- 221/222 confers tamoxifen resistance in breast cancer by targeting p27Kip1. J Biol Chem 283: 29897-29903. 105. Garofalo M, Romano G, Di Leva G, Nuovo G, Jeon YJ, et al. (2012) EGFR and MET receptor tyrosine kinase-altered microRNA expression induces tumorigenesis and gefitinib resistance in lung cancers. Nat Med 18: 74-82. 106. Gottesman MM (2002) Mechanisms of cancer drug resistance. Annu Rev Med 53: 615-627. 107. Kornblihtt AR, de la Mata M, Fededa JP, Munoz MJ, Nogues G (2004) Multiple links between transcription and splicing. RNA 10: 1489-1498. 108. Reed R, Hurt E (2002) A conserved mRNA export machinery coupled to pre- mRNA splicing. Cell 108: 523-531. 109. Danckwardt S, Kaufmann I, Gentzel M, Foerstner KU, Gantzert AS, et al. (2007) Splicing factors stimulate polyadenylation via USEs at non-canonical 3' end formation signals. EMBO J 26: 2658-2669. 110. Buratowski S (2005) Connections between mRNA 3' end processing and transcription termination. Curr Opin Cell Biol 17: 257-261. 111. Shepard PJ, Choi EA, Lu J, Flanagan LA, Hertel KJ, et al. (2011) Complex and dynamic landscape of RNA polyadenylation revealed by PAS-Seq. RNA 17: 761-772. 112. Mayr C, Bartel DP (2009) Widespread shortening of 3'UTRs by alternative cleavage and polyadenylation activates oncogenes in cancer cells. Cell 138: 673-684. 180
113. Caput D, Beutler B, Hartog K, Thayer R, Brown-Shimer S, et al. (1986) Identification of a common nucleotide sequence in the 3'-untranslated region of mRNA molecules specifying inflammatory mediators. Proc Natl Acad Sci U S A 83: 1670-1674. 114. Barreau C, Paillard L, Osborne HB (2005) AU-rich elements and associated factors: are there unifying principles? Nucleic Acids Res 33: 7138-7150. 115. Ule J, Stefani G, Mele A, Ruggiu M, Wang X, et al. (2006) An RNA map predicting Nova-dependent splicing regulation. Nature 444: 580-586. 116. Wang Z, Burge CB (2008) Splicing regulation: from a parts list of regulatory elements to an integrated splicing code. RNA 14: 802-813. 117. Martin KC, Ephrussi A (2009) mRNA localization: gene expression in the spatial dimension. Cell 136: 719-730. 118. Knowles RB, Sabry JH, Martone ME, Deerinck TJ, Ellisman MH, et al. (1996) Translocation of RNA granules in living neurons. J Neurosci 16: 7812-7820. 119. Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, et al. (2001) Initial sequencing and analysis of the human genome. Nature 409: 860-921. 120. Raghavan A, Ogilvie RL, Reilly C, Abelson ML, Raghavan S, et al. (2002) Genome-wide analysis of mRNA decay in resting and activated primary human T lymphocytes. Nucleic Acids Res 30: 5529-5538. 121. Gray NK, Wickens M (1998) Control of translation initiation in animals. Annu Rev Cell Dev Biol 14: 399-458. 122. Dever TE (2002) Gene-specific regulation by general translation factors. Cell 108: 545-556. 123. Wilkie GS, Dickson KS, Gray NK (2003) Regulation of mRNA translation by 5'- and 3'-UTR-binding factors. Trends Biochem Sci 28: 182-188. 124. Hamilton TL, Stoneley M, Spriggs KA, Bushell M (2006) TOPs and their regulation. Biochem Soc Trans 34: 12-16. 125. Jackson RJ, Hellen CU, Pestova TV (2010) The mechanism of eukaryotic translation initiation and principles of its regulation. Nat Rev Mol Cell Biol 11: 113-127. 126. Meyer S, Temme C, Wahle E (2004) Messenger RNA turnover in eukaryotes: pathways and enzymes. Crit Rev Biochem Mol Biol 39: 197-216. 127. Muckenthaler MU, Galy B, Hentze MW (2008) Systemic iron homeostasis and the iron-responsive element/iron-regulatory protein (IRE/IRP) regulatory network. Annu Rev Nutr 28: 197-213. 128. Firoz Ahmed VAB, and Patrick Xuechun Zhao (2011) Mining functional elements in messenger RNAs: overview, challenges, and perspectives. frontiers in plant science 2: 1-18. 129. Cressina E, Chen L, Moulin M, Leeper FJ, Abell C, et al. (2011) Identification of novel ligands for thiamine pyrophosphate (TPP) riboswitches. Biochem Soc Trans 39: 652-657. 130. Kurdistani SK, Tavazoie S, Grunstein M (2004) Mapping global histone acetylation patterns to gene expression. Cell 117: 721-733. 181
131. Mann M, Ong SE, Gronborg M, Steen H, Jensen ON, et al. (2002) Analysis of protein phosphorylation using mass spectrometry: deciphering the phosphoproteome. Trends Biotechnol 20: 261-268. 132. Johnson LN, Lewis RJ (2001) Structural basis for control by phosphorylation. Chem Rev 101: 2209-2242. 133. Walsh CT, Garneau-Tsodikova S, Gatto GJ, Jr. (2005) Protein posttranslational modifications: the chemistry of proteome diversifications. Angew Chem Int Ed Engl 44: 7342-7372. 134. Khorasanizadeh S (2004) The nucleosome: from genomic organization to genomic regulation. Cell 116: 259-272. 135. Roth J (2002) Protein N-glycosylation along the secretory pathway: relationship to organelle topography and function, protein quality control, and cell interactions. Chem Rev 102: 285-303. 136. Giles NM, Watts AB, Giles GI, Fry FH, Littlechild JA, et al. (2003) Metal and redox modulation of cysteine protein function. Chem Biol 10: 677-693. 137. Paetzel M, Karla A, Strynadka NC, Dalbey RE (2002) Signal peptidases. Chem Rev 102: 4549-4580. 138. Seemuller E, Lupas A, Baumeister W (1996) Autocatalytic processing of the 20S proteasome. Nature 382: 468-471. 139. Sanger F, Nicklen S, Coulson AR (1977) DNA sequencing with chain- terminating inhibitors. Proc Natl Acad Sci U S A 74: 5463-5467. 140. Consortium IHG (2004) Finishing the euchromatic sequence of the human genome. Nature 431: 931-945. 141. Ross JS, Cronin M (2011) Whole cancer genome sequencing by next-generation methods. Am J Clin Pathol 136: 527-539. 142. Haimovich AD (2011) Methods, challenges, and promise of next-generation sequencing in cancer biology. Yale J Biol Med 84: 439-446. 143. Mardis ER (2008) Next-generation DNA sequencing methods. Annu Rev Genomics Hum Genet 9: 387-402. 144. Margulies M, Egholm M, Altman WE, Attiya S, Bader JS, et al. (2005) Genome sequencing in microfabricated high-density picolitre reactors. Nature 437: 376-380. 145. Braslavsky I, Hebert B, Kartalov E, Quake SR (2003) Sequence information can be obtained from single DNA molecules. Proc Natl Acad Sci U S A 100: 3960- 3964. 146. Moore B, Hu H, Singleton M, Reese MG, De La Vega FM, et al. (2011) Global analysis of disease-related DNA sequence variation in 10 healthy individuals: implications for whole genome-based clinical diagnostics. Genet Med 13: 210-217. 147. Wilhelm BT, Marguerat S, Watt S, Schubert F, Wood V, et al. (2008) Dynamic repertoire of a eukaryotic transcriptome surveyed at single-nucleotide resolution. Nature 453: 1239-1243.
182
148. Li H, Ruan J, Durbin R (2008) Mapping short DNA sequencing reads and calling variants using mapping quality scores. Genome Res 18: 1851-1858. 149. Kohonen-Corish MR, Al-Aama JY, Auerbach AD, Axton M, Barash CI, et al. (2010) How to catch all those mutations--the report of the third Human Variome Project Meeting, UNESCO Paris, May 2010. Hum Mutat 31: 1374- 1381. 150. Poinar HN, Schwarz C, Qi J, Shapiro B, Macphee RD, et al. (2006) Metagenomics to paleogenomics: large-scale sequencing of mammoth DNA. Science 311: 392-394. 151. Oh JD, Kling-Backhed H, Giannakis M, Xu J, Fulton RS, et al. (2006) The complete genome sequence of a chronic atrophic gastritis Helicobacter pylori strain: evolution during disease progression. Proc Natl Acad Sci U S A 103: 9999- 10004. 152. Noonan JP, Coop G, Kudaravalli S, Smith D, Krause J, et al. (2006) Sequencing and analysis of Neanderthal genomic DNA. Science 314: 1113-1118. 153. Shah SP, Morin RD, Khattra J, Prentice L, Pugh T, et al. (2009) Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature 461: 809-813. 154. Navin N, Kendall J, Troge J, Andrews P, Rodgers L, et al. (2011) Tumour evolution inferred by single-cell sequencing. Nature 472: 90-94. 155. Talkowski ME, Ernst C, Heilbut A, Chiang C, Hanscom C, et al. (2011) Next- generation sequencing strategies enable routine detection of balanced chromosome rearrangements for clinical diagnostics and genetic research. Am J Hum Genet 88: 469-481. 156. Hampton OA, Den Hollander P, Miller CA, Delgado DA, Li J, et al. (2009) A sequence-level map of chromosomal breakpoints in the MCF-7 breast cancer cell line yields insights into the evolution of a cancer genome. Genome Res 19: 167-177. 157. Berger MF, Lawrence MS, Demichelis F, Drier Y, Cibulskis K, et al. (2011) The genomic complexity of primary human prostate cancer. Nature 470: 214-220. 158. Campbell PJ, Stephens PJ, Pleasance ED, O'Meara S, Li H, et al. (2008) Identification of somatically acquired rearrangements in cancer using genome-wide massively parallel paired-end sequencing. Nat Genet 40: 722- 729. 159. Stephens PJ, Greenman CD, Fu B, Yang F, Bignell GR, et al. (2011) Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell 144: 27-40. 160. Greif PA, Yaghmaie M, Konstandin NP, Ksienzyk B, Alimoghaddam K, et al. (2011) Somatic mutations in acute promyelocytic leukemia (APL) identified by exome sequencing. Leukemia 25: 1519-1522. 161. Ding L, Ellis MJ, Li S, Larson DE, Chen K, et al. (2010) Genome remodelling in a basal-like breast cancer metastasis and xenograft. Nature 464: 999-1005.
183
162. Stephens PJ, McBride DJ, Lin ML, Varela I, Pleasance ED, et al. (2009) Complex landscapes of somatic rearrangement in human breast cancer genomes. Nature 462: 1005-1010. 163. Boyle AP, Davis S, Shulha HP, Meltzer P, Margulies EH, et al. (2008) High- resolution mapping and characterization of open chromatin across the genome. Cell 132: 311-322. 164. Lu C, Kulkarni K, Souret FF, MuthuValliappan R, Tej SS, et al. (2006) MicroRNAs and other small RNAs enriched in the Arabidopsis RNA-dependent RNA polymerase-2 mutant. Genome Res 16: 1276-1288. 165. Ruby JG, Jan C, Player C, Axtell MJ, Lee W, et al. (2006) Large-scale sequencing reveals 21U-RNAs and additional microRNAs and endogenous siRNAs in C. elegans. Cell 127: 1193-1207. 166. Greif PA, Eck SH, Konstandin NP, Benet-Pages A, Ksienzyk B, et al. (2011) Identification of recurring tumor-specific somatic mutations in acute myeloid leukemia by transcriptome sequencing. Leukemia 25: 821-827. 167. Berger MF, Levin JZ, Vijayendran K, Sivachenko A, Adiconis X, et al. (2010) Integrative analysis of the melanoma transcriptome. Genome Res 20: 413- 427. 168. Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, et al. (2006) The consensus coding sequences of human breast and colorectal cancers. Science 314: 268- 274. 169. Jones SJ, Laskin J, Li YY, Griffith OL, An J, et al. (2010) Evolution of an adenocarcinoma in response to selection by targeted kinase inhibitors. Genome Biol 11: R82. 170. Beck J, Urnovitz HB, Mitchell WM, Schutz E (2010) Next generation sequencing of serum circulating nucleic acids from patients with invasive ductal breast cancer reveals differences to healthy and nonmalignant controls. Mol Cancer Res 8: 335-342. 171. Marguerat S, Wilhelm BT, Bahler J (2008) Next-generation sequencing: applications beyond genomes. Biochem Soc Trans 36: 1091-1096. 172. Mikkelsen TS, Ku M, Jaffe DB, Issac B, Lieberman E, et al. (2007) Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448: 553-560. 173. Jones PA, Baylin SB (2007) The epigenomics of cancer. Cell 128: 683-692. 174. Bird A (2002) DNA methylation patterns and epigenetic memory. Genes Dev 16: 6-21. 175. Saito Y, Liang G, Egger G, Friedman JM, Chuang JC, et al. (2006) Specific activation of microRNA-127 with downregulation of the proto-oncogene BCL6 by chromatin-modifying drugs in human cancer cells. Cancer Cell 9: 435-443. 176. Hansen KD, Timp W, Bravo HC, Sabunciyan S, Langmead B, et al. (2011) Increased methylation variation in epigenetic domains across cancer types. Nat Genet 43: 768-775. 184
177. Johnson DS, Mortazavi A, Myers RM, Wold B (2007) Genome-wide mapping of in vivo protein-DNA interactions. Science 316: 1497-1502. 178. Robertson G, Hirst M, Bainbridge M, Bilenky M, Zhao Y, et al. (2007) Genome- wide profiles of STAT1 DNA association using chromatin immunoprecipitation and massively parallel sequencing. Nat Methods 4: 651-657. 179. Barski A, Cuddapah S, Cui K, Roh TY, Schones DE, et al. (2007) High-resolution profiling of histone methylations in the human genome. Cell 129: 823-837. 180. Taylor AH, Al-Azzawi F (2000) Immunolocalisation of oestrogen receptor beta in human tissues. J Mol Endocrinol 24: 145-155. 181. Brzozowski AM, Pike AC, Dauter Z, Hubbard RE, Bonn T, et al. (1997) Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 389: 753- 758. 182. Riggins RB, Schrecengost RS, Guerrero MS, Bouton AH (2007) Pathways to tamoxifen resistance. Cancer Lett 256: 1-24. 183. Hall JM, McDonnell DP (1999) The estrogen receptor beta-isoform (ERbeta) of the human estrogen receptor modulates ERalpha transcriptional activity and is a key regulator of the cellular response to estrogens and antiestrogens. Endocrinology 140: 5566-5578. 184. Musgrove EA, Sutherland RL (2009) Biological determinants of endocrine resistance in breast cancer. Nat Rev Cancer 9: 631-643. 185. Tsai MJ, O'Malley BW (1994) Molecular mechanisms of action of steroid/thyroid receptor superfamily members. Annu Rev Biochem 63: 451- 486. 186. Bunone G, Briand PA, Miksicek RJ, Picard D (1996) Activation of the unliganded estrogen receptor by EGF involves the MAP kinase pathway and direct phosphorylation. EMBO J 15: 2174-2183. 187. Arnold SF, Obourn JD, Jaffe H, Notides AC (1994) Serine 167 is the major estradiol-induced phosphorylation site on the human estrogen receptor. Mol Endocrinol 8: 1208-1214. 188. Le Romancer M, Treilleux I, Leconte N, Robin-Lespinasse Y, Sentis S, et al. (2008) Regulation of estrogen rapid signaling through arginine methylation by PRMT1. Mol Cell 31: 212-221. 189. Razandi M, Oh P, Pedram A, Schnitzer J, Levin ER (2002) ERs associate with and regulate the production of caveolin: implications for signaling and cellular actions. Mol Endocrinol 16: 100-115. 190. Yang Z, Barnes CJ, Kumar R (2004) Human epidermal growth factor receptor 2 status modulates subcellular localization of and interaction with estrogen receptor alpha in breast cancer cells. Clin Cancer Res 10: 3621-3628. 191. Harvey JM, Clark GM, Osborne CK, Allred DC (1999) Estrogen receptor status by immunohistochemistry is superior to the ligand-binding assay for predicting response to adjuvant endocrine therapy in breast cancer. J Clin Oncol 17: 1474-1481. 185
192. Anderson E, Clarke RB (2004) Steroid receptors and cell cycle in normal mammary epithelium. J Mammary Gland Biol Neoplasia 9: 3-13. 193. Speirs V, Skliris GP, Burdall SE, Carder PJ (2002) Distinct expression patterns of ER alpha and ER beta in normal human mammary gland. J Clin Pathol 55: 371-374. 194. Forster C, Makela S, Warri A, Kietz S, Becker D, et al. (2002) Involvement of estrogen receptor beta in terminal differentiation of mammary gland epithelium. Proc Natl Acad Sci U S A 99: 15578-15583. 195. Ramakrishnan R, Khan SA, Badve S (2002) Morphological changes in breast tissue with menstrual cycle. Mod Pathol 15: 1348-1356. 196. Brueggemeier RW (2006) Update on the use of aromatase inhibitors in breast cancer. Expert Opin Pharmacother 7: 1919-1930. 197. Howell A (2006) Pure oestrogen antagonists for the treatment of advanced breast cancer. Endocr Relat Cancer 13: 689-706. 198. Osborne CK, Wakeling A, Nicholson RI (2004) Fulvestrant: an oestrogen receptor antagonist with a novel mechanism of action. Br J Cancer 90 Suppl 1: S2-6. 199. Jaiyesimi IA, Buzdar AU, Decker DA, Hortobagyi GN (1995) Use of tamoxifen for breast cancer: twenty-eight years later. J Clin Oncol 13: 513-529. 200. Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, et al. (1998) Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst 90: 1371- 1388. 201. Cummings SR, Eckert S, Krueger KA, Grady D, Powles TJ, et al. (1999) The effect of raloxifene on risk of breast cancer in postmenopausal women: results from the MORE randomized trial. Multiple Outcomes of Raloxifene Evaluation. JAMA 281: 2189-2197. 202. Howell SJ, Johnston SR, Howell A (2004) The use of selective estrogen receptor modulators and selective estrogen receptor down-regulators in breast cancer. Best Pract Res Clin Endocrinol Metab 18: 47-66. 203. (EBCTCG) EBCTCG (2005) Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 365: 1687-1717. 204. Ruchelman AL, Singh SK, Wu X, Ray A, Yang JM, et al. (2002) Diaza- and triazachrysenes: potent topoisomerase-targeting agents with exceptional antitumor activity against the human tumor xenograft, MDA-MB-435. Bioorg Med Chem Lett 12: 3333-3336. 205. Hoskins JM, Carey LA, McLeod HL (2009) CYP2D6 and tamoxifen: DNA matters in breast cancer. Nat Rev Cancer 9: 576-586. 206. Gutierrez MC, Detre S, Johnston S, Mohsin SK, Shou J, et al. (2005) Molecular changes in tamoxifen-resistant breast cancer: relationship between estrogen receptor, HER-2, and p38 mitogen-activated protein kinase. J Clin Oncol 23: 2469-2476. 186
207. Shi L, Dong B, Li Z, Lu Y, Ouyang T, et al. (2009) Expression of ER-{alpha}36, a novel variant of estrogen receptor {alpha}, and resistance to tamoxifen treatment in breast cancer. J Clin Oncol 27: 3423-3429. 208. Honma N, Horii R, Iwase T, Saji S, Younes M, et al. (2008) Clinical importance of estrogen receptor-beta evaluation in breast cancer patients treated with adjuvant tamoxifen therapy. J Clin Oncol 26: 3727-3734. 209. Santen RJ, Song RX, Zhang Z, Kumar R, Jeng MH, et al. (2005) Long-term estradiol deprivation in breast cancer cells up-regulates growth factor signaling and enhances estrogen sensitivity. Endocr Relat Cancer 12 Suppl 1: S61-73. 210. Carroll JS, Lynch DK, Swarbrick A, Renoir JM, Sarcevic B, et al. (2003) p27(Kip1) induces quiescence and growth factor insensitivity in tamoxifen-treated breast cancer cells. Cancer Res 63: 4322-4326. 211. Arpino G, Wiechmann L, Osborne CK, Schiff R (2008) Crosstalk between the estrogen receptor and the HER tyrosine kinase receptor family: molecular mechanism and clinical implications for endocrine therapy resistance. Endocr Rev 29: 217-233. 212. Doisneau-Sixou SF, Sergio CM, Carroll JS, Hui R, Musgrove EA, et al. (2003) Estrogen and antiestrogen regulation of cell cycle progression in breast cancer cells. Endocr Relat Cancer 10: 179-186. 213. Mandlekar S, Kong AN (2001) Mechanisms of tamoxifen-induced apoptosis. Apoptosis 6: 469-477. 214. Riggins RB, Bouton AH, Liu MC, Clarke R (2005) Antiestrogens, aromatase inhibitors, and apoptosis in breast cancer. Vitam Horm 71: 201-237. 215. Zhu Y, Singh B, Hewitt S, Liu A, Gomez B, et al. (2006) Expression patterns among interferon regulatory factor-1, human X-box binding protein-1, nuclear factor kappa B, nucleophosmin, estrogen receptor-alpha and progesterone receptor proteins in breast cancer tissue microarrays. Int J Oncol 28: 67-76. 216. Murphy LC, Watson PH (2006) Is oestrogen receptor-beta a predictor of endocrine therapy responsiveness in human breast cancer? Endocr Relat Cancer 13: 327-334. 217. Riggins RB, Lan JP, Zhu Y, Klimach U, Zwart A, et al. (2008) ERRgamma mediates tamoxifen resistance in novel models of invasive lobular breast cancer. Cancer Res 68: 8908-8917. 218. Zhou Y, Yau C, Gray JW, Chew K, Dairkee SH, et al. (2007) Enhanced NF kappa B and AP-1 transcriptional activity associated with antiestrogen resistant breast cancer. BMC Cancer 7: 59. 219. Osborne CK, Bardou V, Hopp TA, Chamness GC, Hilsenbeck SG, et al. (2003) Role of the estrogen receptor coactivator AIB1 (SRC-3) and HER-2/neu in tamoxifen resistance in breast cancer. J Natl Cancer Inst 95: 353-361.
187
220. Gutteridge E, Agrawal A, Nicholson R, Leung Cheung K, Robertson J, et al. (2010) The effects of gefitinib in tamoxifen-resistant and hormone- insensitive breast cancer: a phase II study. Int J Cancer 126: 1806-1816. 221. Osborne CK, Neven P, Dirix LY, Mackey JR, Robert J, et al. (2011) Gefitinib or placebo in combination with tamoxifen in patients with hormone receptor- positive metastatic breast cancer: a randomized phase II study. Clin Cancer Res 17: 1147-1159. 222. Hurd PJ, Nelson CJ (2009) Advantages of next-generation sequencing versus the microarray in epigenetic research. Brief Funct Genomic Proteomic 8: 174-183. 223. Mendes-Pereira AM, Sims D, Dexter T, Fenwick K, Assiotis I, et al. (2011) Genome-wide functional screen identifies a compendium of genes affecting sensitivity to tamoxifen. Proc Natl Acad Sci U S A. 224. Gao D, Kim J, Kim H, Phang TL, Selby H, et al. (2010) A survey of statistical software for analysing RNA-seq data. Hum Genomics 5: 56-60. 225. Li Y, Alvarez OA, Gutteling EW, Tijsterman M, Fu J, et al. (2006) Mapping determinants of gene expression plasticity by genetical genomics in C. elegans. PLoS Genet 2: e222. 226. Smith EN, Kruglyak L (2008) Gene-environment interaction in yeast gene expression. PLoS Biol 6: e83. 227. Akaike H (1981) Citation Classic - a New Look at the Statistical-Model Identification. Current Contents/Engineering Technology & Applied Sciences: 22-22. 228. Schwarz G (1978) Estimating the dimension of a model. Annals of Statistics 6: 461–464. 229. Kilker RL, Hartl MW, Rutherford TM, Planas-Silva MD (2004) Cyclin D1 expression is dependent on estrogen receptor function in tamoxifen-resistant breast cancer cells. J Steroid Biochem Mol Biol 92: 63-71. 230. Morgan D (2010) Statistical Significance for Basic Adverse Impact Analysis. Washington, DC: DCI Consulting. 1-7 p. 231. Kilker RL, Planas-Silva MD (2006) Cyclin D1 is necessary for tamoxifen-induced cell cycle progression in human breast cancer cells. Cancer Res 66: 11478- 11484. 232. Butte AJ, Ye J, Haring HU, Stumvoll M, White MF, et al. (2001) Determining significant fold differences in gene expression analysis. Pac Symp Biocomput: 6-17. 233. Huang W, Umbach DM, Vincent Jordan N, Abell AN, Johnson GL, et al. (2011) Efficiently identifying genome-wide changes with next-generation sequencing data. Nucleic Acids Res 39: e130. 234. Cordero F, Botta M, Calogero RA (2007) Microarray data analysis and mining approaches. Brief Funct Genomic Proteomic 6: 265-281.
188
235. Berthois Y, Katzenellenbogen JA, Katzenellenbogen BS (1986) Phenol red in tissue culture media is a weak estrogen: implications concerning the study of estrogen-responsive cells in culture. Proc Natl Acad Sci U S A 83: 2496-2500. 236. Clarke R, Skaar TC, Bouker KB, Davis N, Lee YR, et al. (2001) Molecular and pharmacological aspects of antiestrogen resistance. J Steroid Biochem Mol Biol 76: 71-84. 237. Dieci G, Preti M, Montanini B (2009) Eukaryotic snoRNAs: a paradigm for gene expression flexibility. Genomics 94: 83-88. 238. Ender C, Krek A, Friedlander MR, Beitzinger M, Weinmann L, et al. (2008) A human snoRNA with microRNA-like functions. Mol Cell 32: 519-528. 239. Ofir M, Hacohen D, Ginsberg D (2011) MiR-15 and miR-16 are direct transcriptional targets of E2F1 that limit E2F-induced proliferation by targeting cyclin E. Mol Cancer Res 9: 440-447. 240. Fang L, Deng Z, Shatseva T, Yang J, Peng C, et al. (2011) MicroRNA miR-93 promotes tumor growth and angiogenesis by targeting integrin-beta8. Oncogene 30: 806-821. 241. Cowden Dahl KD, Dahl R, Kruichak JN, Hudson LG (2009) The epidermal growth factor receptor responsive miR-125a represses mesenchymal morphology in ovarian cancer cells. Neoplasia 11: 1208-1215. 242. Bandiera S, Ruberg S, Girard M, Cagnard N, Hanein S, et al. (2011) Nuclear outsourcing of RNA interference components to human mitochondria. PLoS One 6: e20746. 243. Ozata DM, Caramuta S, Velazquez-Fernandez D, Akcakaya P, Xie H, et al. (2011) The role of microRNA deregulation in the pathogenesis of adrenocortical carcinoma. Endocr Relat Cancer 18: 643-655. 244. Dumont JA, Bitonti AJ, Wallace CD, Baumann RJ, Cashman EA, et al. (1996) Progression of MCF-7 breast cancer cells to antiestrogen-resistant phenotype is accompanied by elevated levels of AP-1 DNA-binding activity. Cell Growth Differ 7: 351-359. 245. Cardoso CM, Custodio JB, Almeida LM, Moreno AJ (2001) Mechanisms of the deleterious effects of tamoxifen on mitochondrial respiration rate and phosphorylation efficiency. Toxicol Appl Pharmacol 176: 145-152. 246. Shah MA, Schwartz GK (2001) Cell cycle-mediated drug resistance: an emerging concept in cancer therapy. Clin Cancer Res 7: 2168-2181. 247. Jia L, Kelsey SM, Grahn MF, Jiang XR, Newland AC (1996) Increased activity and sensitivity of mitochondrial respiratory enzymes to tumor necrosis factor alpha-mediated inhibition is associated with increased cytotoxicity in drug- resistant leukemic cell lines. Blood 87: 2401-2410. 248. Sugai M, Umezu H, Yamamoto T, Jiang S, Iwanari H, et al. (2008) Expression of hepatocyte nuclear factor 4 alpha in primary ovarian mucinous tumors. Pathol Int 58: 681-686. 249. Kamiyama Y, Matsubara T, Yoshinari K, Nagata K, Kamimura H, et al. (2007) Role of human hepatocyte nuclear factor 4alpha in the expression of drug- 189
metabolizing enzymes and transporters in human hepatocytes assessed by use of small interfering RNA. Drug Metab Pharmacokinet 22: 287-298. 250. Elsheikh SE, Green AR, Rakha EA, Samaka RM, Ammar AA, et al. (2008) Caveolin 1 and Caveolin 2 are associated with breast cancer basal-like and triple-negative immunophenotype. Br J Cancer 99: 327-334. 251. Lavie Y, Fiucci G, Liscovitch M (1998) Up-regulation of caveolae and caveolar constituents in multidrug-resistant cancer cells. J Biol Chem 273: 32380- 32383. 252. Masri S, Phung S, Wang X, Wu X, Yuan YC, et al. (2008) Genome-wide analysis of aromatase inhibitor-resistant, tamoxifen-resistant, and long-term estrogen- deprived cells reveals a role for estrogen receptor. Cancer Res 68: 4910- 4918. 253. Dorssers LC, Van der Flier S, Brinkman A, van Agthoven T, Veldscholte J, et al. (2001) Tamoxifen resistance in breast cancer: elucidating mechanisms. Drugs 61: 1721-1733. 254. Lange CA, Yee D (2011) Killing the second messenger: targeting loss of cell cycle control in endocrine-resistant breast cancer. Endocr Relat Cancer 18: C19-24. 255. Clarke R (2011) Cannibalism, cell survival, and endocrine resistance in breast cancer. Breast Cancer Res 13: 311. 256. Garcia V, Garcia JM, Pena C, Silva J, Dominguez G, et al. (2008) Free circulating mRNA in plasma from breast cancer patients and clinical outcome. Cancer Lett 263: 312-320. 257. Liu M, Tang Q, Qiu M, Lang N, Li M, et al. (2011) miR-21 targets the tumor suppressor RhoB and regulates proliferation, invasion and apoptosis in colorectal cancer cells. FEBS Lett 585: 2998-3005. 258. Loi S, Haibe-Kains B, Desmedt C, Wirapati P, Lallemand F, et al. (2008) Predicting prognosis using molecular profiling in estrogen receptor-positive breast cancer treated with tamoxifen. BMC Genomics 9: 239. 259. Jansen MP, Foekens JA, van Staveren IL, Dirkzwager-Kiel MM, Ritstier K, et al. (2005) Molecular classification of tamoxifen-resistant breast carcinomas by gene expression profiling. J Clin Oncol 23: 732-740. 260. Ma XJ, Wang Z, Ryan PD, Isakoff SJ, Barmettler A, et al. (2004) A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 5: 607-616. 261. Ioannidis JP, Khoury MJ (2011) Improving validation practices in "omics" research. Science 334: 1230-1232. 262. Huang L, Zhao S, Frasor JM, Dai Y (2011) An integrated bioinformatics approach identifies elevated cyclin E2 expression and E2F activity as distinct features of tamoxifen resistant breast tumors. PLoS One 6: e22274. 263. Ashraf N, Zino S, Macintyre A, Kingsmore D, Payne AP, et al. (2006) Altered sirtuin expression is associated with node-positive breast cancer. Br J Cancer 95: 1056-1061. 190
264. Soderling TR (1999) The Ca-calmodulin-dependent protein kinase cascade. Trends Biochem Sci 24: 232-236. 265. Matsushita M, Nairn AC (1998) Characterization of the mechanism of regulation of Ca2+/ calmodulin-dependent protein kinase I by calmodulin and by Ca2+/calmodulin-dependent protein kinase kinase. J Biol Chem 273: 21473-21481. 266. Ryazanov AG, Ward MD, Mendola CE, Pavur KS, Dorovkov MV, et al. (1997) Identification of a new class of protein kinases represented by eukaryotic elongation factor-2 kinase. Proc Natl Acad Sci U S A 94: 4884-4889. 267. Futey LM, Medley QG, Cote GP, Egelhoff TT (1995) Structural analysis of myosin heavy chain kinase A from Dictyostelium. Evidence for a highly divergent protein kinase domain, an amino-terminal coiled-coil domain, and a domain homologous to the beta-subunit of heterotrimeric G proteins. J Biol Chem 270: 523-529. 268. Montell C (2003) Mg2+ homeostasis: the Mg2+nificent TRPM chanzymes. Curr Biol 13: R799-801. 269. Pavur KS, Petrov AN, Ryazanov AG (2000) Mapping the functional domains of elongation factor-2 kinase. Biochemistry 39: 12216-12224. 270. Middelbeek J, Clark K, Venselaar H, Huynen MA, van Leeuwen FN (2010) The alpha-kinase family: an exceptional branch on the protein kinase tree. Cell Mol Life Sci 67: 875-890. 271. Mittl PR, Schneider-Brachert W (2007) Sel1-like repeat proteins in signal transduction. Cell Signal 19: 20-31. 272. Szilak L, Moitra J, Krylov D, Vinson C (1997) Phosphorylation destabilizes alpha-helices. Nat Struct Biol 4: 112-114. 273. Pigott CR, Mikolajek H, Moore CE, Finn SJ, Phippen CW, et al. (2012) Insights into the regulation of eukaryotic elongation factor 2 kinase and the interplay between its domains. Biochem J 442: 105-118. 274. Diggle TA, Seehra CK, Hase S, Redpath NT (1999) Analysis of the domain structure of elongation factor-2 kinase by mutagenesis. FEBS Lett 457: 189- 192. 275. Arora S, Yang JM, Kinzy TG, Utsumi R, Okamoto T, et al. (2003) Identification and characterization of an inhibitor of eukaryotic elongation factor 2 kinase against human cancer cell lines. Cancer Res 63: 6894-6899. 276. Proud CG (1994) Peptide-chain elongation in eukaryotes. Mol Biol Rep 19: 161- 170. 277. Ryazanov AG (2002) Elongation factor-2 kinase and its newly discovered relatives. FEBS Lett 514: 26-29. 278. Ryazanov AG, Shestakova EA, Natapov PG (1988) Phosphorylation of elongation factor 2 by EF-2 kinase affects rate of translation. Nature 334: 170-173.
191
279. Ryazanov AG, Davydova EK (1989) Mechanism of elongation factor 2 (EF-2) inactivation upon phosphorylation. Phosphorylated EF-2 is unable to catalyze translocation. FEBS Lett 251: 187-190. 280. Nairn AC, Palfrey HC (1987) Identification of the major Mr 100,000 substrate for calmodulin-dependent protein kinase III in mammalian cells as elongation factor-2. J Biol Chem 262: 17299-17303. 281. Nairn AC, Bhagat B, Palfrey HC (1985) Identification of calmodulin-dependent protein kinase III and its major Mr 100,000 substrate in mammalian tissues. Proc Natl Acad Sci U S A 82: 7939-7943. 282. Gschwendt M, Kittstein W, Mieskes G, Marks F (1989) A type 2A protein phosphatase dephosphorylates the elongation factor 2 and is stimulated by the phorbol ester TPA in mouse epidermis in vivo. FEBS Lett 257: 357-360. 283. Palfrey HC, Nairn AC, Muldoon LL, Villereal ML (1987) Rapid activation of calmodulin-dependent protein kinase III in mitogen-stimulated human fibroblasts. Correlation with intracellular Ca2+ transients. J Biol Chem 262: 9785-9792. 284. Celis JE, Madsen P, Ryazanov AG (1990) Increased phosphorylation of elongation factor 2 during mitosis in transformed human amnion cells correlates with a decreased rate of protein synthesis. Proc Natl Acad Sci U S A 87: 4231-4235. 285. Fan H, Penman S (1970) Regulation of protein synthesis in mammalian cells. II. Inhibition of protein synthesis at the level of initiation during mitosis. J Mol Biol 50: 655-670. 286. Whitaker M (2006) Calcium microdomains and cell cycle control. Cell Calcium 40: 585-592. 287. Cahill AL, Applebaum R, Perlman RL (1988) Phosphorylation of elongation factor 2 in the rat superior cervical ganglion. Neurosci Lett 84: 345-350. 288. Mackie KP, Nairn AC, Hampel G, Lam G, Jaffe EA (1989) Thrombin and histamine stimulate the phosphorylation of elongation factor 2 in human umbilical vein endothelial cells. J Biol Chem 264: 1748-1753. 289. Rose AJ, Broholm C, Kiillerich K, Finn SG, Proud CG, et al. (2005) Exercise rapidly increases eukaryotic elongation factor 2 phosphorylation in skeletal muscle of men. J Physiol 569: 223-228. 290. Redpath NT, Proud CG (1993) Purification and phosphorylation of elongation factor-2 kinase from rabbit reticulocytes. Eur J Biochem 212: 511-520. 291. Nygard O, Nilsson A, Carlberg U, Nilsson L, Amons R (1991) Phosphorylation regulates the activity of the eEF-2-specific Ca(2+)- and calmodulin- dependent protein kinase III. J Biol Chem 266: 16425-16430. 292. Matsushita M, Kozak JA, Shimizu Y, McLachlin DT, Yamaguchi H, et al. (2005) Channel function is dissociated from the intrinsic kinase activity and autophosphorylation of TRPM7/ChaK1. J Biol Chem 280: 20793-20803. 293. Crawley SW, Gharaei MS, Ye Q, Yang Y, Raveh B, et al. (2011) Autophosphorylation activates Dictyostelium myosin II heavy chain kinase A 192
by providing a ligand for an allosteric binding site in the alpha-kinase domain. J Biol Chem 286: 2607-2616. 294. Redpath NT, Price NT, Severinov KV, Proud CG (1993) Regulation of elongation factor-2 by multisite phosphorylation. Eur J Biochem 213: 689-699. 295. Browne GJ, Proud CG (2004) A novel mTOR-regulated phosphorylation site in elongation factor 2 kinase modulates the activity of the kinase and its binding to calmodulin. Mol Cell Biol 24: 2986-2997. 296. Wang X, Li W, Williams M, Terada N, Alessi DR, et al. (2001) Regulation of elongation factor 2 kinase by p90(RSK1) and p70 S6 kinase. EMBO J 20: 4370-4379. 297. Redpath NT, Proud CG (1993) Cyclic AMP-dependent protein kinase phosphorylates rabbit reticulocyte elongation factor-2 kinase and induces calcium-independent activity. Biochem J 293 ( Pt 1): 31-34. 298. Diggle TA, Redpath NT, Heesom KJ, Denton RM (1998) Regulation of protein- synthesis elongation-factor-2 kinase by cAMP in adipocytes. Biochem J 336 ( Pt 3): 525-529. 299. Hovland R, Eikhom TS, Proud CG, Cressey LI, Lanotte M, et al. (1999) cAMP inhibits translation by inducing Ca2+/calmodulin-independent elongation factor 2 kinase activity in IPC-81 cells. FEBS Lett 444: 97-101. 300. Ayuso-Parrilla MS, Martin-Requero A, Perez-Dias J, Parrilla R (1976) Role of glucagon on the control of hepatic protein synthesis and degradation in the rat in vivo. J Biol Chem 251: 7785-7790. 301. Diggle TA, Subkhankulova T, Lilley KS, Shikotra N, Willis AE, et al. (2001) Phosphorylation of elongation factor-2 kinase on serine 499 by cAMP- dependent protein kinase induces Ca2+/calmodulin-independent activity. Biochem J 353: 621-626. 302. Hay N, Sonenberg N (2004) Upstream and downstream of mTOR. Genes Dev 18: 1926-1945. 303. Sarbassov DD, Ali SM, Sengupta S, Sheen JH, Hsu PP, et al. (2006) Prolonged rapamycin treatment inhibits mTORC2 assembly and Akt/PKB. Mol Cell 22: 159-168. 304. Perraud AL, Zhao X, Ryazanov AG, Schmitz C (2011) The channel-kinase TRPM7 regulates phosphorylation of the translational factor eEF2 via eEF2-k. Cell Signal 23: 586-593. 305. Schmitz C, Perraud AL, Johnson CO, Inabe K, Smith MK, et al. (2003) Regulation of vertebrate cellular Mg2+ homeostasis by TRPM7. Cell 114: 191-200. 306. Hardie DG, Scott JW, Pan DA, Hudson ER (2003) Management of cellular energy by the AMP-activated protein kinase system. FEBS Lett 546: 113-120. 307. Horman S, Browne G, Krause U, Patel J, Vertommen D, et al. (2002) Activation of AMP-activated protein kinase leads to the phosphorylation of elongation factor 2 and an inhibition of protein synthesis. Curr Biol 12: 1419-1423.
193
308. Browne GJ, Finn SG, Proud CG (2004) Stimulation of the AMP-activated protein kinase leads to activation of eukaryotic elongation factor 2 kinase and to its phosphorylation at a novel site, serine 398. J Biol Chem 279: 12220-12231. 309. Lewis TS, Shapiro PS, Ahn NG (1998) Signal transduction through MAP kinase cascades. Adv Cancer Res 74: 49-139. 310. Tibbles LA, Woodgett JR (1999) The stress-activated protein kinase pathways. Cell Mol Life Sci 55: 1230-1254. 311. Knebel A, Haydon CE, Morrice N, Cohen P (2002) Stress-induced regulation of eukaryotic elongation factor 2 kinase by SB 203580-sensitive and - insensitive pathways. Biochem J 367: 525-532. 312. Knebel A, Morrice N, Cohen P (2001) A novel method to identify protein kinase substrates: eEF2 kinase is phosphorylated and inhibited by SAPK4/p38delta. EMBO J 20: 4360-4369. 313. Goedert M, Cuenda A, Craxton M, Jakes R, Cohen P (1997) Activation of the novel stress-activated protein kinase SAPK4 by cytokines and cellular stresses is mediated by SKK3 (MKK6); comparison of its substrate specificity with that of other SAP kinases. EMBO J 16: 3563-3571. 314. Gum RJ, McLaughlin MM, Kumar S, Wang Z, Bower MJ, et al. (1998) Acquisition of sensitivity of stress-activated protein kinases to the p38 inhibitor, SB 203580, by alteration of one or more amino acids within the ATP binding pocket. J Biol Chem 273: 15605-15610. 315. Smith EM, Proud CG (2008) cdc2-cyclin B regulates eEF2 kinase activity in a cell cycle- and amino acid-dependent manner. EMBO J 27: 1005-1016. 316. Sivan G, Kedersha N, Elroy-Stein O (2007) Ribosomal slowdown mediates translational arrest during cellular division. Mol Cell Biol 27: 6639-6646. 317. Dorovkov MV, Pavur KS, Petrov AN, Ryazanov AG (2002) Regulation of elongation factor-2 kinase by pH. Biochemistry 41: 13444-13450. 318. Kao R, Rannels DE, Morgan HE (1976) Effects of anoxia and ischemia on protein synthesis in perfused rat hearts. Circ Res 38: I124-130. 319. Mitch WE, Goldberg AL (1996) Mechanisms of muscle wasting. The role of the ubiquitin-proteasome pathway. N Engl J Med 335: 1897-1905. 320. Baehrecke EH (2005) Autophagy: dual roles in life and death? Nat Rev Mol Cell Biol 6: 505-510. 321. Salvesen GS, Dixit VM (1997) Caspases: intracellular signaling by proteolysis. Cell 91: 443-446. 322. Goll DE, Thompson VF, Li H, Wei W, Cong J (2003) The calpain system. Physiol Rev 83: 731-801. 323. Glickman MH, Ciechanover A (2002) The ubiquitin-proteasome proteolytic pathway: destruction for the sake of construction. Physiol Rev 82: 373-428. 324. Ciechanover A, Orian A, Schwartz AL (2000) Ubiquitin-mediated proteolysis: biological regulation via destruction. Bioessays 22: 442-451.
194
325. Lecker SH, Goldberg AL, Mitch WE (2006) Protein degradation by the ubiquitin-proteasome pathway in normal and disease states. J Am Soc Nephrol 17: 1807-1819. 326. Weissman AM (2001) Themes and variations on ubiquitylation. Nat Rev Mol Cell Biol 2: 169-178. 327. Muratani M, Tansey WP (2003) How the ubiquitin-proteasome system controls transcription. Nat Rev Mol Cell Biol 4: 192-201. 328. Yang J, Yang JM, Iannone M, Shih WJ, Lin Y, et al. (2001) Disruption of the EF-2 kinase/Hsp90 protein complex: a possible mechanism to inhibit glioblastoma by geldanamycin. Cancer Res 61: 4010-4016. 329. Arora S, Yang JM, Hait WN (2005) Identification of the ubiquitin-proteasome pathway in the regulation of the stability of eukaryotic elongation factor-2 kinase. Cancer Res 65: 3806-3810. 330. Severinov KV, Melnikova EG, Ryazanov AG (1990) Downregulation of the translation elongation factor 2 kinase in Xenopus laevis oocytes at the final stages of oogenesis. New Biol 2: 887-893. 331. Brady MJ, Nairn AC, Wagner JA, Palfrey HC (1990) Nerve growth factor-induced down-regulation of calmodulin-dependent protein kinase III in PC12 cells involves cyclic AMP-dependent protein kinase. J Neurochem 54: 1034-1039. 332. End D, Hanson M, Hashimoto S, Guroff G (1982) Inhibition of the phosphorylation of a 1000,000-dalton soluble protein in whole cells and cell- free extracts of PC12 pheochromocytoma cells following treatment with nerve growth factor. J Biol Chem 257: 9223-9225. 333. Scheetz AJ, Nairn AC, Constantine-Paton M (2000) NMDA receptor-mediated control of protein synthesis at developing synapses. Nat Neurosci 3: 211-216. 334. Weatherill DB, McCamphill PK, Pethoukov E, Dunn TW, Fan X, et al. (2011) Compartment-specific, differential regulation of eukaryotic elongation factor 2 and its kinase within Aplysia sensory neurons. J Neurochem 117: 841-855. 335. Bagaglio DM, Cheng EH, Gorelick FS, Mitsui K, Nairn AC, et al. (1993) Phosphorylation of elongation factor 2 in normal and malignant rat glial cells. Cancer Res 53: 2260-2264. 336. Riis B, Rattan SI, Palmquist K, Nilsson A, Nygard O, et al. (1993) Elongation factor 2-specific calcium and calmodulin dependent protein kinase III activity in rat livers varies with age and calorie restriction. Biochem Biophys Res Commun 192: 1210-1216. 337. Li X, Alafuzoff I, Soininen H, Winblad B, Pei JJ (2005) Levels of mTOR and its downstream targets 4E-BP1, eEF2, and eEF2 kinase in relationships with tau in Alzheimer's disease brain. FEBS J 272: 4211-4220. 338. Hait WN, Ward MD, Trakht IN, Ryazanov AG (1996) Elongation factor-2 kinase: immunological evidence for the existence of tissue-specific isoforms. FEBS Lett 397: 55-60.
195
339. Parmer TG, Ward MD, Yurkow EJ, Vyas VH, Kearney TJ, et al. (1999) Activity and regulation by growth factors of calmodulin-dependent protein kinase III (elongation factor 2-kinase) in human breast cancer. Br J Cancer 79: 59-64. 340. Parmer TG, Ward MD, Hait WN (1997) Effects of rottlerin, an inhibitor of calmodulin-dependent protein kinase III, on cellular proliferation, viability, and cell cycle distribution in malignant glioma cells. Cell Growth Differ 8: 327-334. 341. Chen Y, Matsushita M, Nairn AC, Damuni Z, Cai D, et al. (2001) Mechanisms for increased levels of phosphorylation of elongation factor-2 during hibernation in ground squirrels. Biochemistry 40: 11565-11570. 342. Boyce M, Py BF, Ryazanov AG, Minden JS, Long K, et al. (2008) A pharmacoproteomic approach implicates eukaryotic elongation factor 2 kinase in ER stress-induced cell death. Cell Death Differ 15: 589-599. 343. Wu H, Yang JM, Jin S, Zhang H, Hait WN (2006) Elongation factor-2 kinase regulates autophagy in human glioblastoma cells. Cancer Res 66: 3015-3023. 344. Py BF, Boyce M, Yuan J (2009) A critical role of eEF-2K in mediating autophagy in response to multiple cellular stresses. Autophagy 5: 393-396. 345. Walden WE, Thach RE (1986) Translational control of gene expression in a normal fibroblast. Characterization of a subclass of mRNAs with unusual kinetic properties. Biochemistry 25: 2033-2041. 346. Cleveland DW, Yen TJ (1989) Multiple determinants of eukaryotic mRNA stability. New Biol 1: 121-126. 347. Takeuchi H, Kondo Y, Fujiwara K, Kanzawa T, Aoki H, et al. (2005) Synergistic augmentation of rapamycin-induced autophagy in malignant glioma cells by phosphatidylinositol 3-kinase/protein kinase B inhibitors. Cancer Res 65: 3336-3346. 348. Pan JG, Mak TW (2007) Metabolic targeting as an anticancer strategy: dawn of a new era? Sci STKE 2007: pe14. 349. Prostko CR, Zhang C, Hait WN (1997) The effects of altered cellular calmodulin expression on the growth and viability of C6 glioblastoma cells. Oncol Res 9: 13-17. 350. Hait WN, Wu H, Jin S, Yang JM (2006) Elongation factor-2 kinase: its role in protein synthesis and autophagy. Autophagy 2: 294-296. 351. Wu H, Zhu H, Liu DX, Niu TK, Ren X, et al. (2009) Silencing of elongation factor- 2 kinase potentiates the effect of 2-deoxy-D-glucose against human glioma cells through blunting of autophagy. Cancer Res 69: 2453-2460. 352. Cheng Y, Li H, Ren X, Niu T, Hait WN, et al. (2010) Cytoprotective effect of the elongation factor-2 kinase-mediated autophagy in breast cancer cells subjected to growth factor inhibition. PLoS One 5: e9715. 353. Zhang Y, Cheng Y, Zhang L, Ren X, Huber-Keener KJ, et al. (2011) Inhibition of eEF-2 kinase sensitizes human glioma cells to TRAIL and down-regulates Bcl- xL expression. Biochem Biophys Res Commun 414: 129-134.
196
354. Terai K, Hiramoto Y, Masaki M, Sugiyama S, Kuroda T, et al. (2005) AMP- activated protein kinase protects cardiomyocytes against hypoxic injury through attenuation of endoplasmic reticulum stress. Mol Cell Biol 25: 9554- 9575. 355. Chan AY, Soltys CL, Young ME, Proud CG, Dyck JR (2004) Activation of AMP- activated protein kinase inhibits protein synthesis associated with hypertrophy in the cardiac myocyte. J Biol Chem 279: 32771-32779. 356. Libby P, Theroux P (2005) Pathophysiology of coronary artery disease. Circulation 111: 3481-3488. 357. Rose AJ, Alsted TJ, Jensen TE, Kobbero JB, Maarbjerg SJ, et al. (2009) A Ca(2+)- calmodulin-eEF2K-eEF2 signalling cascade, but not AMPK, contributes to the suppression of skeletal muscle protein synthesis during contractions. J Physiol 587: 1547-1563. 358. Huang HY, Chien CH, Jen KH, Huang HD (2006) RegRNA: an integrated web server for identifying regulatory RNA motifs and elements. Nucleic Acids Res 34: W429-434. 359. Kato S, Sekine S, Oh SW, Kim NS, Umezawa Y, et al. (1994) Construction of a human full-length cDNA bank. Gene 150: 243-250. 360. Le SY, Maizel JV, Jr. (1997) A common RNA structural motif involved in the internal initiation of translation of cellular mRNAs. Nucleic Acids Res 25: 362-369. 361. Meijer HA, Thomas AA (2002) Control of eukaryotic protein synthesis by upstream open reading frames in the 5'-untranslated region of an mRNA. Biochem J 367: 1-11. 362. Parsch J, Stephan W, Tanda S (1999) A highly conserved sequence in the 3'- untranslated region of the drosophila Adh gene plays a functional role in Adh expression. Genetics 151: 667-674. 363. Lai EC (2002) Micro RNAs are complementary to 3' UTR sequence motifs that mediate negative post-transcriptional regulation. Nat Genet 30: 363-364. 364. Chauvin C, Jean-Jean O (2008) Proteasomal degradation of human release factor eRF3a regulates translation termination complex formation. RNA 14: 240-245. 365. Spotts GD, Hann SR (1990) Enhanced translation and increased turnover of c- myc proteins occur during differentiation of murine erythroleukemia cells. Mol Cell Biol 10: 3952-3964. 366. Kawai T, Fan J, Mazan-Mamczarz K, Gorospe M (2004) Global mRNA stabilization preferentially linked to translational repression during the endoplasmic reticulum stress response. Mol Cell Biol 24: 6773-6787. 367. Keene JD (2001) Ribonucleoprotein infrastructure regulating the flow of genetic information between the genome and the proteome. Proc Natl Acad Sci U S A 98: 7018-7024. 368. Hait WN, Jin S, Yang JM (2006) A matter of life or death (or both): understanding autophagy in cancer. Clin Cancer Res 12: 1961-1965. 197
369. J Martens IN, T Koenig, M Look, N Harbeck, F Model. Epigenetic signature predicts failure of endocrine therapy in patients with recurrent breast cancer.; 2003. pp. Abstract 313. 370. Han W, Han MR, Kang JJ, Bae JY, Lee JH, et al. (2006) Genomic alterations identified by array comparative genomic hybridization as prognostic markers in tamoxifen-treated estrogen receptor-positive breast cancer. BMC Cancer 6: 92. 371. Osborne CK, Coronado-Heinsohn EB, Hilsenbeck SG, McCue BL, Wakeling AE, et al. (1995) Comparison of the effects of a pure steroidal antiestrogen with those of tamoxifen in a model of human breast cancer. J Natl Cancer Inst 87: 746-750. 372. Law IK, Liu L, Xu A, Lam KS, Vanhoutte PM, et al. (2009) Identification and characterization of proteins interacting with SIRT1 and SIRT3: implications in the anti-aging and metabolic effects of sirtuins. Proteomics 9: 2444-2456. 373. Shi T, Wang F, Stieren E, Tong Q (2005) SIRT3, a mitochondrial sirtuin deacetylase, regulates mitochondrial function and thermogenesis in brown adipocytes. J Biol Chem 280: 13560-13567. 374. Tcherepanova I, Puigserver P, Norris JD, Spiegelman BM, McDonnell DP (2000) Modulation of estrogen receptor-alpha transcriptional activity by the coactivator PGC-1. J Biol Chem 275: 16302-16308. 375. Kressler D, Hock MB, Kralli A (2007) Coactivators PGC-1beta and SRC-1 interact functionally to promote the agonist activity of the selective estrogen receptor modulator tamoxifen. J Biol Chem 282: 26897-26907. 376. Spriggs KA, Bushell M, Willis AE (2010) Translational regulation of gene expression during conditions of cell stress. Mol Cell 40: 228-237. 377. Ben-Ari Y, Brody Y, Kinor N, Mor A, Tsukamoto T, et al. (2010) The life of an mRNA in space and time. J Cell Sci 123: 1761-1774.
198
Curriculum Vitae Kathryn J. Huber-Keener
Education 2005 – Present MD/PhD candidate, Pharmacology The Pennsylvania State University College of Medicine 2000 – 2004 B.A. Physics, Pre-med Saint Olaf College, Northfield, MN
Selected Publications (5 of 10) Huber-Keener KJ, Evans BR, Ren X, Cheng Y, Zhang Y, & Yang JM. (2012) Phosphorylation of elongation factor-2 kinase differentially regulates the enzyme's stability under stress conditions. Biochem Biophys Res Commun. 2012 Jun 27. [Epub ahead of print] PMID: 22749997
Huber-Keener KJ, Liu X, Wang Z, Freeman W, Wu S, Silva-Planas M, Ren X, Cheng Y, Zhang Y, Vrana K, Liu CG, Yang JM, & Wu R. (2012) Differential gene expression in tamoxifen- resistant breast cancer cells revealed by a new analytical model of RNA-Seq data. (Accepted by PLoS One.)
Huber-Keener KJ & Yang JM. (2011) The impact of metabolic and therapeutic stresses on glioma progression and therapy. In CC Chen (Ed.), Advances in the Biology, Imaging, and Therapies for Glioblastoma. (pp. 23-52). Rijeka, Croatia: Intech Open Access Publisher. ISBN 979-953-307-197-7.
Zhang Y, Cheng Y, Zhang L, Ren X, Huber-Keener KJ, S Lee, J Yun, HG Wang, JM Yang. (2011) Inhibition of eEF-2 kinase sensitizes human glioma cells to TRAIL and down-regulates Bcl-xL expression. Biochem Biophys Res Commun 414: 129-134. PMID: 21945617
Cheng Y, Zhang Y, Zhang L, Ren X, Huber-Keener KJ, X Liu, L Zhou, J Liao, H Keihack, L Yan, E Rubin, JM Yang. (2012) MK-2206, a novel allosteric inhibitor of Akt, synergizes with gefitinib against malignant glioma via modulating both autophagy and apoptosis. Mol Cancer Ther 11: 154-164.PMID: 22057914
Selected Abstracts and Poster Presentations (2 of 17) Huber-Keener KJ, Evans BR, & Yang JM. “Phosphorylation-dependent Regulation of EF-2 Kinase in Response to Stress in Human Glioma Cells.” American Association of Cancer Researchers Annual 102nd Annual Meeting. April 2011.
Huber-Keener KJ, Liu X, Wang Z, Freeman W, Wu S, Ren X, Liu CG, Wu R, & Yang JM. “Transcriptome analysis of tamoxifen-resistant breast cancer cells using next generation sequencing” American Association of Cancer Researchers Annual 102nd Annual Meeting. April 2011. Late-breaking abstract.