Interrogation of the RP-- P53 Axis in Human

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Raiser, David Michael. 2015. Interrogation of the RP-MDM2-P53 Axis in Human Ribosomopathies. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:17467484

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA

Interrogation of the RP-MDM2-p53 Axis in Human Ribosomopathies

A dissertation presented

by

David Michael Raiser

to

The Division of Medical Sciences

in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Biological and Biomedical Sciences

Harvard University

Cambridge, Massachusetts

May 2015

© 2015 David Michael Raiser

All Rights Reserved. Dissertation Advisor: Dr. Benjamin Ebert David Michael Raiser

Interrogation of the RP-MDM2-p53 Axis in Human Ribosomopathies

Abstract

Ribosomopathies such as Diamond-Blackfan anemia (DBA) and 5q- syndrome are human diseases characterized by heterozygous loss or mutation of -associated .

Hallmarks of these diseases include a macrocytic anemia in an otherwise normocellular bone marrow and, in some cases, developmental defects such as craniofacial or thumb abnormalities.

Haploinsufficiency of ribosome-associated genes leads to dysfunctional ribosome biogenesis, which ultimately results in inhibition of MDM2 and consequent aberrant activation of the p53 pathway. Multiple genetic models have shown that p53 loss partially or completely rescues phenotypes associated with ribosomal (RP) haploinsufficiency, implicating p53 as the primary driver of those phenotypes in these diseases. However, the details of the molecular cascade leading to p53 activation – the RP-MDM2-p53 axis – are not fully understood. The thesis work presented here aimed to better characterize the molecular players downstream of ribosome dysfunction involved in modulation of -associated phenotypes, as well as to identify novel therapeutic opportunities.

The proteomics screen characterized the protein binding partner profiles of MDM2 in cells with and without ribosome dysfunction. We found several RPs, including RPL5, RPL11,

RPL23, and RPL38, to be commonly associated with MDM2, though RP deficiency may

iii enhance the binding of RPL5 and RPL11. We also identified IGF1R as associated with MDM2 and selectively degraded in RP-deficient hematopoietic stem and progenitor cell (HSPC) cultures, and we showed that this protein loss contributes to defective erythropoiesis. In addition, a chemical screen identified calmodulin inhibitors as effective in rescuing ribosomopathy- associated phenotypes in , mouse, and primary human HSPC models. We showed that these compounds (including the antipsychotic trifluoperazine) modulated p53 activity by inhibiting its nuclear localization, probably through inhibition of calmodulin-dependent CHK2.

Lastly, we found that heterozygous RP deletion is a common feature of many human cancers. We further showed that silencing of a number of RPs frequently deleted in cancer results in p53 pathway activation, and that RP-deleted cancers have defects in ribosomal RNA processing. Together, this work adds novel insight to several aspects of ribosomopathy pathology and the RP-MDM2-p53 axis, and it further provides foundational evidence for novel therapeutic approaches to both ribosompathies and RP-deficient cancers.

iv Acknowledgements

To me, this dissertation represents so much more than simply a body of scientific work. It represents a culmination of a long and winding journey that took me away from graduate school and back again, and on which I lived, laughed, loved, and learned with the people who have made me the scientist and person I am today. I owe so much to so many, and I want to acknowledge some of those who have helped me get here.

To my exceptional advisor, Ben – I am so grateful to have had the opportunity to work with you and learn from your example. I have tremendous respect for your ability to effectively lead such a large and diverse group of mentees, and I share this respect with everyone who works with you. More than anything, your support in all my pursuits - scientifically, professionally, and personally - has meant the world to me.

To all my trench buddies in the Ebert Lab – thank you for making the lab a fun and comfortable place to come to work every day for the last four years. I have benefitted from the scientific mentorship and friendship of many who have come through the Ebert lab, but special thanks go to Damien, Anu, Allegra, Rishi, Lisa, Alyse, Slater, Marie, Esther, and Rebekka for shaping my lab experience and making it memorable.

To my phenomenal collaborators, Alison & Liz – things rarely go as planned in science, but having people to work with who can roll with the punches and continue to do great work despite the setbacks (and do it with genuine enthusiasm) make the experience a pleasure. My experience with you as collaborators has spoiled me terribly – you’ve set a high bar. Thanks for being awesome.

To the members of my former lab, Sima, Kerstin, Steve, Alison, Raffaela, and Dave – thank you for teaching me so much about how to do science, and for being so supportive during the twists and turns of my early graduate career. My success in the Ebert lab would not have been possible without the time I spent working with and learning from all of you, and I remember our time in the lab together fondly.

To my parents and siblings – thank you for your constant love and support, and for your understanding over the many years and directions of my graduate school journey. I know that it has not always been easy to do, and I do not take it for granted.

v To my San Francisco family, Locoya, Richard, Brooks, Jose, and Quentin – I left Boston not knowing what I would find, and I count my lucky stars for having found each of you. Thank you for showing me the city you love, and for teaching me to love it as well. I learned so much and lived so much life in just that one year, but your friendships and lessons in being unapologetically myself are what will stay with me always.

To my many Boston friends over the years, but in particular to Jed, Omar, Perry, Marco, and Archie – I have not always loved Boston, but the time spent with all of you here has made it feel like home. Thank you for your friendship – each of you is an inspiration to me.

To my partner in crime in lab and in life, Becca – from Jazzercisers, to a figure skating pair, to Uncle Sam and Lady Liberty, we’ve yet to pass up an opportunity to make complete fools of ourselves – and I’ve loved every minute of it. I cherish the time when I was lucky enough to come to lab each day and sit inches away from my best friend, and it has simply never been the same without you. When I grow up, I want to be as cool as you. I love you, Flava!

To my former partner, Zach – you saw me across the country and back, and you were my shoulder to lean (and cry) on during some of the most tortuous legs of my graduate school journey. You taught me so much about the man I want to be, and I will be forever grateful for the time you joined me on my path. Tycho is the luckiest dog alive, and I’m pretty sure he knows it.

To my Redline brothers – on even my most exhausted and defeated days in the lab, a Redline rehearsal never failed to brighten my spirits and provide me with the escape I needed. Matt, Josh, (and JB) – thank you for helping me build something meaningful and fun, and for loving Redline just as much as I do. Making music with all of you has been an honor.

To my dear (and longest-standing on this acknowledgments list!) friend, Austin – there is something to be said for knowing and loving someone through thirteen (!) years of…well, growing up. You have been a sometimes roommate, a frequent confidant, and an always friend. Through all the roles you have played in my life, thank you for always doing so with genuine compassion and wisdom beyond your years.

To my dear friend, Jaime – from what seems like the very beginning, you have known me so much better than I know myself. Thank you for all the adventures, and for all the formative late nights of wine and waxing philosophical. Most of all, thank you for always (always) believing in me.

vi

To my boyfriend, AJ – your support has been a critical factor in me reaching this PhD finish line. During a veritable rollercoaster of highs and lows, your patience has been heroic and your love has kept me smiling. I am lucky to call you my partner.

And finally, to my best friend and business partner, Iain – I simply would not have made it this far without you. Without a doubt, you are the person who has had the single most significant impact on my development as a scientist, entrepreneur, and friend. At my worst, best, and everywhere in between, you have been there to hold me up. You have already done more for me than I can ever repay, and I am forever grateful for your friendship. Thank you for letting me come along with you on our Aldatu adventure – it is a privilege. Here’s to becoming something different, and to making up the future as we go.

vii Table of Contents

Abstract ...... iii!

Acknowledgements ...... v!

: Introduction ...... 1! Attributions ...... 2! Ribosomopathies ...... 3! Treating Ribosomopathies in the Clinic ...... 12! Modeling Ribosomopathies in the Laboratory ...... 13! The Emerging Importance of p53 in Ribosomopathies ...... 18! The Heme / Globin Balance: an additional mechanism of anemia ...... 21! Translating Cancer: Ribosome Dysfunction and Malignancy ...... 22! Summary ...... 24!

: Results I ...... 25! Selective IGF1R degradation mediates dyserythropoiesis in the context of RPS14 deficiency! Attributions ...... 26! Summary ...... 27! Introduction ...... 28! Materials and Methods ...... 29! Results ...... 35! Discussion ...... 52!

: Results II ...... 57! Calmodulin inhibition rescues the effects of deficiency by modulating p53 activity in models of Diamond-Blackfan anemia! Attributions ...... 58! Summary ...... 59! Introduction ...... 59! Materials and Methods ...... 61! Results ...... 67! Discussion ...... 80!

viii : Results III ...... 85! Ribosomal gene haploinsufficiency is a common feature of human cancers! Attributions ...... 86! Summary ...... 87! Introduction ...... 88! Materials and Methods ...... 89! Results ...... 92! Discussion ...... 101!

...... 104! Summary ...... 105! The RP-MDM2 Interaction: New Insights ...... 106! IGF1R: A Novel Player in the Ribosomopathy Phenotype ...... 109! Calmodulin Inhibitors and p53: Old Dogs, New Tricks ...... 115! Ribosomal Protein Gene Deletions in Cancer: an Achilles Heel? ...... 117! Concluding Remarks ...... 120!

Appendix: Supplementary Tables ...... 143!

ix !

Introduction

Attributions

I wrote a review article entitled “The emerging importance of ribosomal dysfunction in the pathogenesis of hematologic disorders” (Raiser DM, Narla A, Ebert BL. Leuk Lymphoma. 2014

Mar;55(3):491-500). The text and accompanying figures from this review are included here with edits.

2 Ribosomopathies

More than a decade has passed since the initial identification of ribosomal protein gene mutations in patients with Diamond-Blackfan anemia (DBA), a hematologic disorder that became the founding member of a class of diseases known as ribosomopathies. In these diseases, genetic abnormalities that result in defective ribosome biogenesis cause strikingly tissue-specific phenotypes in patients, specifically bone marrow failure, craniofacial abnormalities, and skeletal defects. Several animal models and numerous in vitro studies have demonstrated that the p53 pathway is central to the ribosomopathy phenotype. Additionally, there is mounting evidence of a link between the dysregulation of components of the translational machinery and the pathology of various malignancies. The importance of the role of ribosomal dysfunction in the pathogenesis of hematologic disorders is becoming clearer, and elucidation of the underlying mechanisms could have broad implications for both basic cellular biology and clinical intervention strategies.

Known and candidate ribosomopathies include Diamond-Blackfan anemia, 5q- syndrome,

Shwachman-Diamond syndrome, X-linked dyskeratosis congenita, cartilage-hair hypoplasia, and

Treacher-Collins syndrome. All are characterized by a distinct set of clinical features, including bone marrow failure and/or developmental abnormalities. Despite the heterogeneity of accompanying clinical features, this convergence highlights cell types that may be particularly sensitive to ribosome dysfunction, despite the universal requirement for ribosome function in all cells. Furthermore, several ribosomopathies are associated with a significant cancer predisposition, highlighting a potentially significant role for translational control in malignant transformation.

3 Diamond-Blackfan Anemia

Diamond-Blackfan Anemia (DBA) is a congenital bone marrow failure syndrome that was originally described by Josephs in 1936 and later characterized further by Diamond and

Blackfan in 1938 1, 2. DBA presents in infancy, most commonly with pallor and lethargy, at an estimated incidence of 4 to 5 cases per million live births, and there is often a family history of the disease 3. Primarily characterized by a hypoplastic anemia, other prominent hematologic characteristics of DBA include macrocytosis, reticulocytopenia, elevated levels of adenosine deaminase in red blood cells, presence of fetal membrane antigen “i”, and a selective paucity or complete absence of erythroid precursors in an otherwise normocellular bone marrow 4. In addition to these symptoms, roughly half of DBA patients also present with physical abnormalities including short stature, thumb abnormalities, heart defects, craniofacial abnormalities, and cleft lip and/or palate 4; some of these defects have been associated with specific genetic lesions (see below).

At present, the standard of therapy for DBA consists of corticosteroids or chronic red blood cell transfusions. Bone marrow transplantation is the only definitive treatment 5, 6.

Interestingly, there is a spontaneous remission rate of approximately 20% in DBA patients by age 25 7; however, despite a growing body of genetic mutation and phenotype data (see below), remission has not been associated with any specific genotype or physical abnormality.

The initial description of RPS19 mutations in DBA patients in 1999 was followed by the identification of mutations and deletions in several other ribosomal protein genes. These mutations and deletions are universally heterozygous. The current list of DBA genes includes

RPS7, RPS10, RPS17, RPS19, RPS24, RPS26, RPS29, RPL5, RPL11, RPL26, and RPL35A 8, 9.

Polymorphisms in DBA patients have been identified in genes encoding RPS15, RPS27A, and

4 RPL36, but their significance has not been definitively established 10, 11. Newer SNP- or quantitative PCR-based copy number variation assay techniques have enabled the identification of previously unknown deletions in RPL5, RPS17, RPS19, RPS26, and RPL35A in 17-25% of patients without an identified RP gene mutation 12-14. At present, more than 50% of patients have an identified mutation in a gene encoding ribosomal .

In a recent report, exome sequencing was used to identify mutations in two siblings who were diagnosed with DBA but who did not harbor mutations in ribosomal protein genes. In both siblings, a mutation was identified at a splicing site of the GATA-1 gene, which encodes a hematopoietic transcription factor 15. While these patients met nearly all of the main diagnostic criteria for DBA, subtle phenotypic differences, such as normal erythrocyte adenosine deaminase levels in one patient, and the lack of ribosomal protein gene mutations with autosomal dominant inheritance, may place patients with GATA1 mutations in a category distinct from DBA. Indeed, genetics may aid in the development of more precise diagnostic categories for patients with

Diamond Blackfan anemia and related congenital bone marrow failure syndromes.

Within DBA patients, mutations in RPL5 have been associated with a higher frequency of physical abnormalities, while mutations in RPL11 have been found to be more likely to have an isolated thumb abnormality than patients with RPS19 mutations 16. An additional study based on the Czech DBA registry also showed that RPS19-mutant patients did not have thumb abnormalities, whereas patients with either RPL5 or RPL11 mutations did 17. Additionally, recent in vitro work by Moniz and colleagues using DBA patient samples suggests that the erythroid phenotype of patients with RPS19 mutations is distinct from that seen in patients with RPL11 mutations 18.

5 RP gene mutations have known consequences in ribosome biogenesis. Systematic knockdown of each of the small subunit protein genes in mammalian cells revealed clear pre- rRNA processing functions for each of the individual ribosomal protein components 19.

Mutations in two DBA disease genes, RPS19 and RPS24, have been shown to cause defects in the pre-rRNA processing of 18S rRNA, which in turn leads to impaired and decreased production of the 40S small ribosomal subunit 20-23. Several other studies have shown that, in general, depletion of components of the 40S subunit (RPS proteins) reduce the levels of the small subunit, depletion of components of the 60S subunit (RPL proteins) reduce levels of the large subunit, and either type of depletion results in decreased levels of mature 80S 24-26.

5q- Syndrome

An independent subtype of myelodysplastic syndrome (MDS) first described in 1974 27, the 5q- syndrome is an acquired disorder involving a single cytogenetic abnormality in the form of a large interstitial deletion of the long arm of 5. Patients with the 5q- syndrome are predominately female and commonly have a severe macrocytic anemia, thrombocytosis with hypolobulated micromegakaryocytes, and a lower rate of progression to acute myeloid leukemia

(AML) relative to other subtypes of MDS 28, 29. The present standard of care for 5q- patients is lenalidomide, a thalidomide derivative, that shows high efficacy in these patients. In one phase 2 trial wherein lenalidomide was administered to low-risk MDS patients with del(5q), a reduction in transfusion dependence was seen in 76% of cases, whereas a complete cytogenetic response was seen in an impressive 61% of patients 30. In a subsequent phase 3 trial, patients treated with lenalidomide at a dose of 10 mg per day had a 50% cytogenetic response rate and an improved survival compared with patients treated with 5 mg per day 31. While lenalidomide has several

6 known biological functions including promotion of erythropoiesis 32-34, the mechanism of its efficacy in 5q- syndrome remains poorly understood.

An RNA interference screen targeting each individual gene contained within the common deleted region of 5q revealed that knockdown of RPS14 recapitulated the erythroid defect characteristic of the 5q- syndrome 32. Patients with the 5q- syndrome possess only a single allele of RPS14, and the gene is expressed at haploinsufficient levels in samples from these patients. In vitro studies involving overexpression of the RPS14 cDNA in 5q- syndrome patient samples demonstrated a rescue in the erythropoietic defect 32. Like many of the DBA-associated RP genes, depletion of RPS14 results in defective pre-rRNA processing and decreased 40S ribosome biogenesis 26, 32. While other genetic components within the common deleted region (such as miR-145, miR146a, and EGR1 35-38) may contribute to the phenotype of 5q- syndrome, RPS14 haploinsufficiency has been clearly shown to be responsible for the erythropoietic phenotype, which is the most severe aspect of the disease.

Shwachman-Diamond Syndrome

First reported in 1964 by Shwachman and colleagues, Shwachman-Diamond Syndrome

(SDS, or sometimes Shwachman-Bodian-Diamond Syndrome) has an incidence of approximately 1 in 50,000 live births 39. SDS is an autosomal recessive disorder that presents in infancy and results in bone marrow failure and exocrine pancreatic dysfunction. The hematologic features of SDS include depression of myeloid lineages, intermittent or persistent neutropenia, anemia, and thrombocytopenia 40. SDS patients are also at a significantly higher risk of developing leukemia 41. In addition to the hematologic and pancreatic phenotypes, patients also present with short stature and other skeletal abnormalities with variable penetrance 42. At present, bone marrow transplantation is the only definitive treatment for SDS, although patients are often

7 managed with supportive treatments including pancreatic enzymes, transfusions, antibiotics, and

G-CSF.

The gene responsible for SDS was identified in 2003 to be SBDS, which was found to be mutated as a result of gene conversion in ~90% of SDS patients analyzed, 67% of which carried two mutant alleles 43. These mutant alleles carried sequence changes similar to those present in a

SBDS (SBDSP) residing in a locally duplicated genomic segment, the transcript of which shares 97% homology with SBDS but results in a truncated form of the protein.

Studies in yeast and mice have elucidated a role for SBDS, in conjunction with EFL1, in the release of eIF6 from the 60S ribosome, a critical step in late cytoplasmic 60S ribosomal subunit maturation and activation 44-46. In bone marrow cells from SDS patients, a number of genes involved in ribosome biogenesis and rRNA processing are abnormally expressed, including decreased expression of a number of RP genes 47. Recent work in lymphoblasts and bone marrow stromal cells from patients with SBDS has confirmed that 40S and 60S ribosomal subunit association is impaired 48. In addition, SBDS has been shown to play a role in stabilization of the mitotic spindle, which may affect proliferation and/or chromosome segregation, which in turn may contribute to chromosomal instability, cancer predisposition, and potentially the SDS bone marrow failure phenotype 49, 50.

Dyskeratosis Congenita

Dyskeratosis congenita (DC), another bone marrow failure syndrome, is predominantly a disorder of telomere dysfunction, but some of the disease-causing mutations also alter ribosome function. DC was described more than century ago as a triad of diagnostic criteria: reticular skin pigmentation, nail dystrophy, and mucosal leukoplakia. Subsequently, bone marrow failure was

8 described as a distinguishing clinical feature and the major cause of death in DC patients 51, 52. In addition, the disorder is associated with a high risk of developing leukemia, solid tumors, and pulmonary fibrosis. The disease usually presents early in life, with skin pigmentation and nail changes occurring within the first decade. DC is diagnosed in roughly 1 in 1 million people, predominately affecting males. While bone marrow transplantation is a viable curative treatment for DC patients, pulmonary complications dramatically reduce the success rate of this procedure53.

All of the causative mutations for DKC identified to date are harbored in genes whose protein products function in telomerase activity, telomerase complex assembly, or telomere integrity 52, 54, 55. Telomerase, which functions to add nucleotide repeats to the ends of that are shortened as a result of DNA replication, is a ribonucleoprotein (RNP) complex that includes a reverse transcriptase (TERT), a catalytic RNA component (TERC), and dyskerin, a protein implicated in X-linked DC (the most prevalent form of the disorder).

Dyskerin is encoded by DKC1 and interacts with small nucleolar (snoRNAs) in a particular subtype of snoRNP complex involved in rRNA modification, specifically pseudouridylation.

Defects in rRNA modification resulting from DKC1 mutations are thought to contribute to impaired ribosome biogenesis, but the implications of this process in the pathogenesis of

DKC1 are still under investigation. Mutation analysis of DC patients has shown that disruption of the telomerase complex alone may be sufficient to impair hematopoiesis 56, though all known

DKC1 disease mutations have been shown to result in lowered levels of TERC and telomere shortening 57, 58. An intriguing mouse model involving a hypomorphic Dkc1 allele showed pseudouridylation defects and subsequent presentation of some DC clinical features in first and

9 second generation animals, though telomere shortening was only evident in later generations 59.

Additionally, studies of DC patients with the same DKC1 mutations and widely varying disease penetrance highlight the contribution of other genetic and environmental factors to DC pathogenesis 55.

Cartilage Hair Hypoplasia

Cartilage hair hypoplasia (CHH) was first described by McKusick and colleagues in 1965 in an Amish population 60, and it was later observed with relative frequency in patients of

Finnish decent 61; outside of these populations, incidence is rare and not well documented. An inherited autosomal recessive disorder, CHH presents with short-limbed dwarfism and fine, sparse, light-colored hair. Variable additional phenotypes are frequently observed and include defective cellular immunity affecting T-cell response, mild to severe anemia, gastrointestinal malabsorption or Hirschsprung’s disease, and a predisposition to malignancies such as non-

Hodgkin’s lymphoma and basal cell carcinoma 42, 62. Like several other bone marrow failure syndromes described here, stem cell transplantation is the only curative treatment for the hematopoietic phenotypes seen in CHH, and it is particularly indicated in patients with severe T- cell immunodeficiency 63.

An untranslated gene, RMRP, has been identified as the causative gene for CHH 64, and more than 100 mutations with variable phenotypic consequences have been identified since the initial description 65. The gene product of RMRP is the RNA component of the mitochondrial

RNA processing complex, which is classified as a snoRNA and is involved in the processing and maturation of the 5.8S rRNA as well as tRNA precursor maturation 66. Cells expressing the founder CHH RMRP mutation exhibited increased levels of cyclin B2 transcripts, which may contribute to chromosomal instability (and ultimately bone marrow dysfunction) through mitotic

10 spindle checkpoint perturbations 67, 68. Furthermore, the RMRP gene product may have some connection with the reverse transcriptase component of telomerase (TERT), and thus RMRP mutations may contribute to the CHH phenotype by mechanisms similar to those seen in DC (see above) 69.

Treacher Collins syndrome

Treacher Collins syndrome (TCS) is an autosomal dominant disorder characterized by distinct craniofacial abnormalities that arise from symmetrical diminished growth of the structures derived from the first and second pharyngeal arch, groove, and pouch 70, 71. TCS has a prevalence of approximately 1 in 50,000 live births and is often accompanied by complications including abnormal brain development and airway, swallowing, and hearing issues. Supportive care for TCS patients is complex and typically involves physicians from a variety of clinical disciplines 70.

TCOF1, encoding the protein treacle, is mutated in the vast majority of TCS cases.

Treacle localizes to the and is a constituent of a preribosomal RNP complex, and it has been shown to be required for ribosomal DNA transcription. Additionally, treacle may function in rRNA methylation 72. In a mouse model of TCS, loss of one allele of Tcof1 results in craniofacial abnormalities similar to those seen in the human form of the disease. Furthermore, these mice have reduced ribosome production, which is consistent with the essential role of

Tcof1 in ribosome biogenesis.

In a small subset of patients in which a causative mutation in TCOF1 has not been identified, deletions in a gene encoding a subunit of RNA polymerase I and III (POLR1D) and

11 mutations in both alleles of a second alpha-related subunit of RNA polymerase I and III

(POLR1C) have been discovered 73.

Notably, the craniofacial defects that are hallmarks of TCS are strikingly similar to those observed in some DBA patients, but TCS patients do not appear to have an associated hematologic phenotype. Nonetheless, studying TCS could provide specific insight into the skeletal and developmental effects of ribosome dysfunction observed in a number of ribosomopathies.

Treating Ribosomopathies in the Clinic

Bone marrow transplantation is the only definitive therapy for the bone marrow failure characteristic of many ribosomopathies. In addition, pharmacologic therapies exist for the management of several of these disorders. Corticosteroids are efficacious in some patients with

DBA. Lenalidomide, is the standard treatment for patients with the 5q- syndrome. Leucine, an known to stimulate global protein translation through the mTOR pathway, is also being explored as a treatment option for some ribosomopathy patients 74, 75; the success of this treatment also suggests that other compounds that stimulate the mTOR pathway may have clinical utility. Lastly, as the causative mutations in ribosomopathies become more well-defined, gene therapy remains an appealing therapeutic alternative that is worthy of further investigation76.

12 Modeling Ribosomopathies in the Laboratory

Primary Human CD34+ HSPCs

Primary human CD34+ hematopoietic stem and progenitor cells (HSPCs) isolated from human cord blood or adult bone marrow have provided a valuable model system for studying the effects of reduced ribosomal protein gene levels in human cells in vitro. When these cells were isolated from patients with DBA, higher than normal levels of apoptosis were observed 77, along with an impaired ability to proliferate in response to erythropoietin stimulation or to generate erythroid colonies in methylcellulose colony assays 78. When HSPCs were isolated from normal individuals, RNAi-mediated reduction of expression levels of RPS19 – the gene most frequently lost or mutated in DBA – resulted in a severe erythroid defect, impairing both proliferation and differentiation 79, 80. Impaired erythropoiesis was improved by treatment with the corticosteroid dexamethasone 79, the primary pharmacologic therapy for DBA. Furthermore, genetic overexpression of RPS19 in RPS19-deficient HSPCs isolated from DBA patients resulted in improved erythropoiesis 81 and improved engraftment in a mouse xenograft model 82, 83.

Similarly, overexpression of RPS14 in CD34+ cells from patients with the 5q- syndrome resulted in a rescue of erythropoiesis. These studies illustrate the utility of cultured primary human

HSPCs for the study of disorders of ribosome dysfunction.

Zebrafish

Additional insight into the consequences of ribosomal protein haploinsufficiency has been provided by studies performed in zebrafish. Several groups have created Rps19 morphants, wherein morpholinos are used to reduce Rps19 levels in developing embryos by a mechanism

13 comparable to RNA interference 84-86. Reduced expression of Rps19 caused impaired erythropoiesis, characterized by increased erythrocyte volume, decreased hemoglobin levels, and red blood cells that appeared to be incompletely differentiated; necrotic brain tissue and a mispositioned heart were also noted in these animals. All of these phenotypes were rescued with injection of Rps19 mRNA. Rps14 morphants have also been studied, and these animals exhibited a profound erythroid anemia along with developmental defects including short body length, abnormal bronchial arch development, and brain and cardiac edema 85. Injection of Rps14 mRNA rescued these phenotypes.

Another ribosomal protein gene mutated in human DBA, Rpl11, has also been targeted by morpholinos. Interestingly, these morphants exhibited defects analogous to the craniofacial abnormalities sometimes observed in DBA 87. Similar to the rps19 morphants, injection of rpl11 mRNA was able to rescue the observed defects in rpl11 morphants.

An rps29 insertional mutant strain of zebrafish was also recently characterized 88.

Though homozygous mutants did not survive beyond 5 days post-fertilization (which is consistent with embryonic lethality of homozygous loss of any RP studied to date), a DBA-like phenotype consisting of defective erythropoiesis and excessive apoptosis in the head region was observed even in very early embryogenesis, providing an additional example of how ribosome protein gene loss can exert strikingly tissue-specific effects. Notably, mutations in RPS29 have very recently been identified in a multiplex DBA family, adding this RP to the list of bona fide

DBA genes 11.

Ribosomal gene haploinsufficiency was also linked to cancer predisposition in an insertional mutagenesis screen performed in zebrafish. This screen focused on genes that are embryonic lethal when inactivated homozygously, but that also result in an increased

14 predisposition to tumor formation with heterozygous inactivation 89, 90. Remarkably, out of the 12 genes identified in this screen that fit these criteria, 11 were ribosomal protein genes 91.

Combined with the clinical observation that several diseases of ribosome dysfunction are associated with a higher risk of developing various cancers 4, 40, 92-95, some research has been directed toward elucidating the genetic and molecular mechanisms by which ribosomal protein gene haploinsufficiency contributes to tumorigenesis (discussed later in this chapter).

Mice – DBA and 5q- Models

Murine models with altered expression or mutation of ribosomal protein genes have provided in vivo functional evidence for the role of ribosomal genes in hematopoiesis, aided in the elucidation of the biology of ribosome dysfunction, and provided models for the investigation of novel therapies. The spectrum of murine models is summarized in 24, 59, 96-109.

In a model of one of the recurrent RPS19 mutations in DBA patients, an inducible transgenic mouse was generated that expresses a missense mutation in Rps19 (R26W). This model phenocopies DBA when expressed at late timepoints in development 110. Because the endogenous Rps19 alleles were still intact and expressed at normal levels, the investigators posited that Rps19R26W acts in a dominant negative manner. Mice with heterozygous missense mutations in Rps19 and Rps20 were also generated in a mutagenesis screen 104. These mutations, likely hypomorphic alleles, cause a mild anemia along with growth retardation and other tissue- specific effects. Another mouse model with a spontaneous mutation in Rpl24 exhibits growth defects 106. An Rps19 knockout mouse has been generated 103 that results in embryonic lethality with homozygous inactivation. The heterozygous animals, however, did not have an observable

15

References 2004 al., Matsson et 2008 al., McGowan et 2011 al., et Jaako 2004; al., et Oliver 2009 al., et Barkic 2000; al., et Volarevic 2005; al., et Sulic 2009; al., et Fumagalli 2011 al., McGowan et 2010 al., et Barlow al., Kondrashov et 2011 2007 al., Anderson et 2008 al., Jones et 2006 al., et Zhang 2003 al., et Ruggero 2008 al., Gu et

p53 Involvement n/a p53 null with rescue background p53 null with rescue background p53 null with rescue background p53 with rescue null knockdown and/or background p53 null with rescue background by p53 dose unaffected p53 null with rescue background p53 null with rescue background n/a tested) (not p53 with rescue partial background null

cell -

T αβ

failure related related -

transcript or transcript

mice are viable are mice

- phenotypes /

-

hypoplasia

dependent growth dependent -

: proliferation defect upon defect proliferation : : failure to regenerate after partial partial after regenerate to failure : : macrocytic anemia, erythroid erythroid anemia, macrocytic : cell lymphopenia, blockade of blockade lymphopenia, cell cells - - Phenotypes no lethality; homozygous hets in decrease protein growth anemia, macrocytic mild retardation marrow bone anemia, severe growth defects Liver hepatectomy T activation BM dysplasia megakaryocyte hypoplasia, dysplasia, erythroid anemia, macrocytic megakaryocytes hypolobulated Hox various gene short tail, developmental T development; craniofacial phenotype no het lethality, homozygous telomere defects; rRNA modifcation generations later in shortening telomerase disadvantages

models ofdysfunction models ribosome

models of ribosome dysfunctionmodels of ribosome

Murine Murine Molecular Approach Molecular knockout germline missense pathogenic mutation siRNA inducible mutation spontaneous knockout conditional of deletion conditional deleted common region knockout germline knockout germline knockout germline knockout germline allele hypomorphic pathogenic knockin mutations : : 1 . 1 . 1

1

Table Table Gene Rps19 Rps19, Rps20 Rps19 Rpl24 Rps6 Rps14 Rpl38 Rpl22 Tcof1 Sbds Dkc1 Dkc1 Table

1516 DBA phenotype, but Rps19 mRNA and protein levels were nearly identical between control and heterozygous mutant animals. In novel mouse model wherein Rps19 down-regulation is controlled by inducible transgenic expression of Rps19 shRNA, Rps19 deficiency resulted in a

DBA-like phenotype, including a macrocytic anemia leading to bone marrow failure. The hematopoietic phenotype was rescued by both Rps19 gene transfer and genetic inactivation of p53 100.

A conditional knockout of Rps6 has been very useful experimentally, although mutations in this gene have not been identified in human disease. When Rps6 was heterozygously deleted from the T-cell lineage in this model, the T-cell pool was largely unaffected, except a minor decrease in overall cell survival; however, upon T-cell activation, the mutant T-cells had a severe proliferation defect 107. Furthermore, when Rps6 was deleted homozygously from the liver, mutant animals failed to regenerate liver tissue following partial hepatectomy, despite relatively normal growth in response to nutrients after fasting 108. These data suggest that cells with ribosomal haploinsufficiency are able to tolerate reduced levels of ribosome biogenesis in periods of normal growth, but may be particularly sensitive to these defects during periods of high proliferative demand. As embryonic development and adult erythropoeisis are both processes defined by high rates of proliferation, the DBA phenotype is consistent with this hypothesis.

To model the 5q- syndrome, a mouse was generated wherein a large DNA segment syntenic to the commonly deleted region on human chromosome 5q (including Rps14) was conditionally deleted from the hematopoietic stem cell compartment 98. The conditionally deleted region included the Rps14 gene, the gene identified by RNA interference screen to be responsible for the disease phenotype in 5q- syndrome 32. Heterozygous loss of this region

17 resulted in a macrocytic anemia, prominent erythroid dysplasia, and monolobulated megakaryocytes, thereby recapitulating many aspects of the 5q- syndrome.

Murine models have also been generated bearing the key genetic lesions in Treacher-

Collins Syndrome, Shwachman-Diamond Syndrome, and X-linked Dyskeratosis Congenita. A knockout of Tcof1 resulted in craniofacial hypoplasia, reminiscent of the TCS phenotype 101.

Homozygous loss of Sbds, the affected gene in SDS, resulted in early embryonic lethality, wherein heterozygous loss resulted in no observable phenotype 109, perhaps underscoring the importance of gene dosage in SDS disease penetrance. In two different models of DC wherein

Dkc1 was perturbed, a clear role for telomere dysfunction in DC was demonstrated; however, defects in rRNA modification 59 and growth disadvantages 99 were also noted and suggest a role for ribosome biogenesis defects in DC as well.

The Emerging Importance of p53 in Ribosomopathies

The p53 pathway acts to monitor ribosome function as well as genome stability 111. In several animal models of ribosomal dysfunction, inactivation of p53 rescues the ribosomopathy phenotypes. Specifically, in the rps29 zebrafish mutant, crossing of mutant animals with p53- null animals led to a complete rescue of the head apoptosis and hematopoietic phenotypes 88.

Similarly, mice with a germline hypomorphic mutation in Rps19 had a p53 allele dosage- dependent rescue of red blood cell count and mean corpuscular volume (MCV) with p53 heterozygous and homozygous null animals 104. Mice with conditional deletion of 5q genes including RPS14 had impaired erythropoiesis that was also rescued in mice lacking p53 98.

Interestingly, in the Tcof1 heterozygous mouse, a model of TCS, chemical or genetic inhibition of p53 also rescued the craniofacial defects 101. A zebrafish model of DC with a hematopoietic stem cell deficiency phenotype showed rescue with loss of p53 function as well 112.

18 The mechanism of p53 activation in response to ribosome dysfunction, though not yet fully elucidated, appears to be mediated by the consequences of increased levels of free (non- ribosome-bound) ribosomal proteins. Evidence from several groups has shown that disruption of ribosome biogenesis leads to an increase in the levels of free ribosomal proteins, and that, when freely dispersed in the cell, these RPs can have extraribosomal functions 24, 113. A subset of ribosomal proteins – namely RPL5, RPL11, RPL23, RPS7, and RPL26 – have the ability to physically interact with murine double-minute 2 (MDM2, or HDM2 in humans), and this interaction modulates the inhibitory activity of MDM2 on p53 114-123. Thus, when ribosome biogenesis is perturbed, either through RP gene depletion-mediated impairment of ribosome subunit biogenesis or through chemical disruption of the nucleolus, the resulting increase in free

RPs leads to MDM2 inhibition and subsequent p53 stabilization and activation of downstream cell cycle inhibition and apoptotic pathways 25 (Figure 1.1). Underscoring the role of MDM2 inhibition in ribsosome dysfunction-associated p53 accumulation, a recent mouse model in which a mutation in MDM2 impairs MDM2-RP binding showed that this interaction is necessary for the induction of p53 following perturbation of ribosome biogenesis 124.

19

Figure 1.1: Putative mechanism of p53 activation in ribosomopathies.

In cells with unperturbed ribosome biogenesis (left), 40S and 60S ribosomal subunits are assembled in the nucleolus and join together in the where they assemble on mRNA transcripts. In normal cells, p53 is maintained at very low levels by MDM2-mediated ubiquitination and subsequent proteosomal degradation. In the case of ribosomal dysfunction (right), ribosomal subunit assembly is inhibited by an absence of one or more specific ribosomal protein (RP) or accessory components, resulting in a decrease in mature 80s ribosomes. Perturbed ribosome biogenesis leads to an increase in the pool of some free, non-ribosome-bound RPs (such as L5 and L11, depicted here), which can bind to MDM2 and inhibit its negative regulation of p53, in turn resulting in p53 protein stabilization and transcriptional activation of cell cycle arrest and cell death pathway genes.

Uniquely, RPL26 has been shown both to bind MDM2, thereby leading to p53 accumulation, and to bind p53 mRNA transcripts, recruiting them to heavy polysomes and increasing their translation 119, 121, 125. Though these findings have been in the context of the DNA

20 damage response, RPL26 provides an intriguing example of other mechanisms by which ribosomal proteins may have specific extraribosomal functions in response to cellular stress.

Increased free ribosomal proteins may be the consequence of abnormal ribosome stoichiometry in the context of ribosome dysfunction, but may also be caused by altered production of specific ribosomal proteins. Disruption of ribosome biogenesis has been shown to increase translation of RPL11, resulting from a global de-repression of 5’-TOP mRNA translation 24. Because the tumor suppressor function of p53 probably precludes its direct targeting for inhibition, this mechanism provides an interesting alternative therapeutic possibility, wherein targeting of those molecular components responsible for increased 5’-TOP mRNA translation may prevent ribosome dysfunction-induced p53 accumulation without altering the monitoring of genome stability.

The Heme / Globin Balance: an additional mechanism of anemia

The erythroid specificity of the DBA phenotype raises the possibility that the unique requirements of erythroid cells for coordinated globin and heme synthesis are related to the pathophysiology. Erythroid precursors have a profoundly high proliferation rate and have an especially high demand for ribosome biogenesis and protein production 126. Cells deficient in the biogenesis of mature ribosomes are less efficiently able to meet the immense demand for globin production and creates a scenario in which heme (normally bound by globin complexes) is present in excess in these cells. Free excess heme is poorly tolerated by erythroid precursors, and this heme/globin imbalance leads to apoptosis and, ultimately, anemia. It is thought that this toxicity is a result of iron-mediated generation of free radicals, and could also activate p53 127.

21 The negative effects of excess heme on erythroid cells was demonstrated by a mouse model in which a heme export protein, FLVCR, is conditionally inactivated 128. Analysis of these mice showed that this protein functions to prevent heme toxicity and is essential in erythropoiesis; conditional inactivation of FLVCR causes accumulation of intracellular heme and a severe macrocytic anemia reminiscent of that seen in human ribosomopathies. Further analysis revealed that FLVCR-null mice also exhibited some developmental defects in utero comparable to those observed in DBA, suggesting a possible extra-erythropoietic role for FLVCR 127. No mutations in the FLVCR gene have been identified to date in DBA or any other human ribosomopathy, nor has heme accumulation been rigorously examined in ribosomopathy patient samples.

Translating Cancer: Ribosome Dysfunction and Malignancy

Several ribosomopathies have been associated with an increased predisposition to cancer.

Patients with SDS have as high as a 36% risk of developing myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML) 129-131. DBA patients have a more modest predisposition to cancer, with particular predilection to MDS, AML, and osteosarcoma 95. Patients with the 5q- syndrome, a subtype of MDS, do progress to AML, but at a lower rate than patients with many other subtypes of MDS 29.

Nucleophosmin (NPM1), a protein that is critical for ribosome function, is a commonly somatically mutated gene in AML. NPM1 is a multifunctional nucleolar protein with RNA- binding properties, a role in pre-ribosomal assembly, and the ability to bind to MDM2 132. NPM1 is mutated in roughly 30-60% of AML cases with normal karyotype and may be involved in other hematologic disorders such as anaplastic large cell lymphoma, MDS, and acute

22 promyelocytic leukemia 133, 134. Intriguingly, heterozygous loss of Npm1 in a mouse model leads to the development of an MDS-like hematologic syndrome 135.

Recently, recurrent somatic mutations have been reported in RPL5 and RPL10 in approximately 7% of cases of T-cell acute lymphoblastic leukemia (T-ALL) 136. Both of these genes are required for ribosome biogenesis. While other pathogenic mutations in ribosomal protein genes result in haploinsufficiency, RPL10 is located on the X-chromosome, and the majority of T-ALL patients with mutations are male. Female patients with RPL10 mutations have expression of only the mutant allele, indicating that T-ALL patients with RPL10 mutations do not retain expression of the normal RPL10 protein. Recurrent heterozygous deletions of

RPL22 have also been reported in T-ALL 137. In a murine model of T cell lymphoma driven by transgenic expression of MyrAkt2 in T-cell progenitors, heterozygous inactivation of Rpl22 significantly accelerated lymphomagenesis. Mechanistically, Rpl22 haploinsufficiency caused increased NF-kB signaling and induction of Lin28B, a factor that governs the stem cell state.

Overall, these data provide direct genetic evidence of somatic mutations in ribosomal genes that are oncogenic.

The mechanisms underlying malignant transformation in the setting of ribosome dysfunction have not been established, but several hypotheses have been proposed. In RP- haploinsufficient zebrafish, which are prone to developing malignant peripheral nerve sheath tumors (MPNSTs), tumors have a marked reduction in p53 protein expression, suggesting that loss of p53 may be critical for oncogenesis in these models of ribosome dysfunction 138.

Aberrant ribosome function leads to differential translation of specific transcripts and the use of alternative translation initiation sites 139, 140. It is possible that translation of oncogenes and tumor suppressors is altered in such a way that the proteome of cells with certain types of

23 ribosome dysfunction promotes malignant transformation. The study of human tumor samples, as well as model organisms, is likely to inform the molecular basis of the link between ribosome dysfunction and cancer.

Summary

Despite the universal requirement of ribosomes for protein translation, germline and somatic mutations that alter ribosome biogenesis or function are associated with specific human diseases with distinct clinical phenotypes. Bone marrow failure, developmental abnormalities, and cancer predisposition are common themes in the human disorders and animal models of ribosome dysfunction, perhaps indicating particularly high demand for intact ribosome function in specific tissues and in the prevention of malignant transformation. The p53 pathway plays a critical role in the induction of cell cycle arrest and apoptosis of cells with aberrant ribosome function. In addition, altered translation of mRNA transcripts, including 5’ TOP mRNAs, may further alter the clinical phenotype. Insights into the biology of ribosomopathies have the potential to improve therapies for both the rare congenital disorders of ribosome function and more common malignancies associated with ribosomal dysfunction.

24 !: Results I

Selective IGF1R degradation mediates dyserythropoiesis in the context of RPS14 deficiency

Attributions

In advance of the proteomics experiment, I designed and generated the MDM2-V5 expression vector and generated the clonal MDM2-V5-expressing A549 cell line. I designed and executed the proteomics screen, in collaboration with Monica Schenone, Emily Hartmann, and Steve Carr at the Broad Institute. I also generated all the RP over-expression vectors used in this work. I performed all IGF1R validation experiments, with some technical assistance from Marie

McConkey. Iain MacLeod assisted with the qPCR array work. I conducted all differentiation experiments in CD34+ HSPCs.

Compilation of the SILAC IP/MS results and statistical analyses of the proteomics were performed by Monica Schenone and Emily Hartmann. Liz Macari injected the Igf1r morpholinos and analyzed the zebrafish embryos.

Note: We are currently writing a manuscript for publication based on the work presented in this chapter.

26 Summary

Ribosomopathies such as 5q- syndrome and Diamond-Blackfan anemia are broadly characterized by haploinsufficiency of one or more ribosome-associated proteins leading to an aberrant activation of the p53 pathway. Increased binding of a subset of free ribosomal proteins

(including RPL5, RPL11, RPL23, and RPS7) to MDM2 is believed to abrogate the negative regulatory control of MDM2 leading to p53 stabilization and activation. In order to elucidate changes in the binding partner profile of MDM2 as a consequence of ribosome dysfunction, we performed a SILAC-based IP/MS proteomics screen to identify proteins that were differentially bound to MDM2 with and without shRNA-mediated knockdown of RPS14, a haploinsufficient gene in 5q- syndrome. We found that RPL5, RPL11, RPL23, and RPS7 were strongly associated with MDM2 under control conditions. However, RPS14 knockdown significantly abrogated the binding of RPS7 to MDM2 in contrast to RPL5, RPL11, and RPL23, which remained abundant in MDM2 complexes. These data corroborate previous findings that a subset of 60S subunit ribosomal proteins can physically associate with MDM2, while suggesting that this binding may typically occur even in the absence of ribosomal haploinsufficiency.

We also identified IGF1R, a known substrate of MDM2, as selectively associated with

MDM2 specifically in the context of RPS14 knockdown. IGF1R protein is selectively degraded following RPS14 knockdown in primary human hematopoietic stem and progenitor cells

(HSPCs), and IGF1R deficiency alone was sufficient to impair erythroid differentiation of

HSPCs in a liquid culture system and to impair hemoglobinization in zebrafish embryos. Taken together, this work identifies IGF1R as a uniquely regulated protein in the context of RPS14 deficiency, and further suggests that modulation of IGF1R levels contributes to the dyserythropoiesis phenotype observed in many ribosomopathies.

27 Introduction

Heterozygous loss or mutation of specific ribosome-associated genes leads to ribosome dysfunction and consequent hematopoietic and/or developmental defects in a number of human disorders, collectively called ribosomopathies. In 5q- syndrome, a subtype of myelodysplastic syndrome (MDS), haploinsufficiency of the RPS14 gene has been shown to be the primary cause of the macrocytic anemia commonly observed in patients, who present with a paucity of erythroid precursors in an otherwise normocellular bone marrow 79. A number of in vitro and animal studies have implicated aberrant p53 activation as the driver of this phenotype 88, 113, 123,

141.

In the context of ribosomal protein deficiency, p53 pathway activity is modulated by the

RP-MDM2-p53 axis. Disrupted ribosomal protein stoichiometry results in dysregulated ribosomal subunit assembly and the increased availability of “free” ribosomal proteins that are not part of an intact ribosome. Protein-protein interaction studies have shown that a subset of these proteins, including RPL5, RPL11, RPL23, and RPS7, can physically associate with MDM2, thereby abrogating the steady-state negative regulatory control of MDM2 on p53 114, 115, 117, 118. In hematopoietic cells, the consequent stabilization of p53 and transcription of p53 target genes (e.g. p21) is largely restricted to cells of the erythroid lineage and results in defective erythropoiesis113.

These aforementioned studies have implicated a small number of candidate proteins in the MDM2-mediated modulation of p53 activity, but a broader understanding of the role played by MDM2 in the context of ribosome dysfunction could lead to additional insights into the molecular pathogenesis of ribosomopathies. In the present study, we performed a quantitative proteomic analysis of proteins that interact with MDM2 in the setting of ribosome dysfunction induced by gene knockdown of RPS14. Our findings confirm that a subset of ribosomal proteins,

28 including RPL5, RPL11, RPL23, RPL26, and RPL38, are specifically associated with MDM2, though ribosomal protein deficiency may enhance the association of RPL5 and RPL11. We also identified IGF1R, a known ubiquitination target of MDM2, as a binding partner of MDM2 specifically in the context of ribosomal protein deficiency. Follow-up experiments in primary human hematopoietic stem and progenitor cells (HSPCs) showed that IGF1R protein is selectively degraded in RPS14-deficient cells, and that IGF1R gene knockdown impairs erythropoiesis in both in vitro human and in vivo zebrafish models. Taken together, these findings confirm the ability of some ribosomal proteins to physically interact with MDM2, and further implicate IGF1R degradation as a novel MDM2-mediated mechanism contributing to dyserythropoiesis in the context of ribosomopathies.

Materials and Methods

Cell line and primary human hematopoietic cell culture

A549 cells were cultured in F-12K medium supplemented with 10% fetal calm serum and

1% penicillin/streptomycin. Primary human CD34+ cells were isolated fresh from umbilical cord blood using the AutoMACS™ Separation System according to manufacturer protocols. Adult

CD34+ G-CSF-mobilized peripheral blood stem cells were obtained from the Fred Hutchinson

Cancer Research Center. Cells were maintained in culture in erythroid differentiation media

(EDM): IMDM (Cellgro), 1% L-glutamine (Life Technologies), 2% penicillin/streptomycin

(Life Technologies), 330 µg/mL holo-human transferrin (Sigma), 10 µg/mL recombinant human insulin (Sigma), 2 IU/mL heparin (Sigma), 5% inactivated human plasma (Rhode Island Blood

Center), and 3 IU/mL erythropoietin (Amgen). From day 1 to day 7 (Phase I), EDM was supplemented with 10-6M hydrocortisone (Sigma), 5 ng/mL interleukin-3 (R&D Biosystems),

29 and 100 ng/mL stem cell factor (SCF)(R&D Biosystems). From day 8 to day 11 (Phase II), EDM was supplemented only with SCF. From day 12 to day 14 (Phase III), cells were cultured in

EDM with no additional supplements. All cells were maintained at 37°C and 5% CO2.

Lentiviral vectors and infection

Lentiviral shRNAs in the pLKO.1 vector were obtained from the Broad Institute of

Harvard and MIT. Sequences targeted by each shRNA are as follows: luciferase (LUC) [5’-

ACACTCGGATATTTGATATGT-3’]; RPS14 [5’- CCGAGATGAATCCTCACCATA-3’];

IGF1R #1 [5’- GAGACAGAGTACCCTTTCTTT-3’]; IGF1R #2 [5’-GCGGTGTCCAATA

ACTACATT-3’]. For lentiviral overexpression constructs, ORF clones for RPL5, RPL11,

RPL23, RPS7, and ARF were obtained from the ORFeome collection at the Broad Institute of

Harvard and MIT 142. These ORFs were Gateway®-cloned into pCDH vector backbones

(SystemBio) with GFP or dTomato expression cassettes. Lentivirus was produced in 293T cells as described previously 143. Cells were infected with lentivirus in the presence of 8 µg/mL polybrene (Sigma) and selected 24 hours later with 2 µg/mL puromycin (Sigma).

SILAC media labeling

We followed all standard SILAC media preparation and labeling steps as previously described with the addition of light proline to prevent the conversion of to proline 144, 145.

Briefly, L-methionine and 200mg/L of L-proline were added to base media according to standard formulations for RPMI (Caisson Labs). This base media was divided into three and to each

13 14 added L-arginine (R0) and L-lysine (K0) (“light”), C6 N4-L-arginine (R6) and 4,4,5,5-D4-L-

13 15 13 15 lysine (K4) (“medium”), or C6 N4-L-arginine (R10) and C6 N2-L-lysine (K8) (“heavy”) to

30 generate the three SILAC labeling mediums. Each medium, with the full complement of amino acids at the standard concentration for each media, was sterile filtered through a 0.22 µm filter

(Milipore). Each cell line was grown in the corresponding labeling media, prepared as described above, supplemented with 2 mM L-glutamine (Gibco), and 10% dialyzed fetal bovine serum

(Sigma) plus 1% penicillin / streptamycin (Gibco), in a humidified atmosphere with 5% CO2 in air at 37°C. Cells were grown for at least ten cell doublings in labeling media prior to use in

IP/MS experiments.

Coimmunoprecipitation

Cells for protein analysis were harvested and resuspended in IP lysis buffer (Pierce) with

HALT protease/phosphatase inhibitor (Pierce). 0.5 mg lysate (1.5 mg for SILAC experiments) was incubated with 1 µg anti-V5 antibody per 1 mg lysate (Invitrogen) at 4°C overnight and subsequently with Sepharose®-G beads (GE Healthcare) for 3 hours. Beads were washed 1-2 times with IP lysis buffer before elution in 2x Laemmli SDS sample buffer (Bio-Rad) with β- mercaptoethanol at 95°C and loading for analysis by gel electrophoresis.

Mass spectrometry

20 µL LDS was added to washed and combined beads, and affinity enriched proteins were with 2 µL 500 mM DTT for 30 minutes and alkylated with 4 µL 500 mM Iodoacetamide for 45 minutes in the dark. Eluted proteins were separated by SDS-PAGE at 200 volts for 30 minutes. Proteins were visualized by SimplyBlue SafeStain for 1 hour and destained in water overnight. Gel bands were cut into 8 fractions, diced, destained, and dehydrated. 0.7 µg of trypsin was added to each gel band, and digestion was carried out overnight at 37°C. Digestion

31 was quenched by adding 10% FA to a final concentration of 1%. Digested peptides were extracted from gel pieces by adding 50 µL 60% ACN .1% FA twice and fully dehydrated using

ACN. Extracted peptides were dried, resuspended in 40 µL 3% ACN .1% FA, and StageTipped according to published protocols. Eluted peptides were transferred into HPLC vials, dried, and analyzed by LC-MS. Peptides were separated on a ~24 cm column of sub 2 um C18 resin with a

110 minute gradient. Eluted peptides were analyzed on a Q Exactive using a Top12 method.

Raw MS files were extracted, searched against a Uniprot human database, using open sourced software MaxQuant. Carbamidomethylation was selected as a fixed modification.

Following export, contaminants and reverse positive proteins, as well as proteins with only one peptide, were removed, and remaining proteins were median normalized. A moderated T test was performed to identify significant protein interactors. A protein was deemed significant if its p value was less than 0.01 and its average ratio was in the 95th percentile or higher.

Flow cytometry

Cells were incubated for 30 minutes on ice with CD71-FITC (1:100) and CD235a

(Glycophorin A)-PE (1:400) (BD Pharmingen) in 2% FBS/PBS to assess early and late erythroid cell differentiation, respectively. Flow cytometry was performed on a FACSCanto cytometer

(BD Biosciences), and data were analyzed using FlowJo software (Version 10.0.7, TreeStar).

Quantitative real-time RT-PCR

TaqMan assays: RNA isolation and cDNA synthesis were performed in tandem on a MultiMACS™ Separation Unit, using MultiMACS™ mRNA and cDNA modules, respectively (Miltenyi Biotec). qPCR was performed using TaqMan© gene expression assays

32 and master mix (Applied Biosystems) on the ABI PRISM 7900HT Sequence Detection System.

GAPDH and/or RPLP0 were used as internal controls. Relative mRNA expression was quantified using the ΔΔCq method. TaqMan gene expression assays used were as follows: IGF1R

(Hs00609566_m1); RPS14 (Hs00852033_g1); GAPDH (Hs99999905_m1); CDKN1A

(Hs99999142_m1); and RPLP0 (Hs_99999902_m1).

Hematopoiesis-focused gene expression profiling: The Human Hematopoiesis RT²

Profiler™ PCR Array (SABiosciences) profiles the expression of 84 genes related to the development of blood-cell lineages from HSCs through progenitor stem cells. Each sample of cDNA was mixed with the RT² SYBR Green/ROX qPCR Master Mix (SABiosciences) according to the directions and 10µl was loaded into 96-wells of a 384-well plate such that four samples could be run simultaneously. qPCR cycling was performed using the ABI PRISM

7900HT Sequence Detection System: 10 minutes at 95°C, followed by 40 cycles of 15 seconds at 95°C, and 1 minute at 60°C during which fluorescence was detected.

Data analyses were performed using GenEx (http://genex.gene-quantification.info/) following the MIQE guidelines 146. Differential expression of genes associated with

147 hematopoiesis was determined by the ΔΔCq method, as per the manufacturer’s instructions .

Briefly, ΔCq values for each gene were determined by normalizing the Cq to reference genes that were selected based on their relative stability between biological replicates and treatment groups, as calculated by NormFinder 148. Fold-change in gene expression with associated 95% confident intervals, was calculated by the ΔΔCq method using average ΔCq of six replicates for the knockdown versus negative control groups. Normalized Cq values were used to determine statistical significance by the Student’s unpaired t test with p < 0.05 considered significant. All negative controls (NTCs), and those to determine the presence of contamination from genomic

33 DNA, were either undetectable or detected at above the threshold at which a gene was considered to be sufficiently expressed for accurate quantification (35 cycles).

Gel electrophoresis and Western blot

Western blots were performed using 25-50 µg of protein per lane and proteins were resolved on 4%-15% Bis-Tris SDS-PAGE gels (Invitrogen, NuPAGE) and transferred to polyvinylidene difluoride membranes (Millipore) for immunoblotting. Membranes were probed with the following primary antibodies resuspended in 5% BSA/PBS: MDM2 (Santa Cruz

(SMP14), 1:200), p53 (Santa Cruz (DO-1), 1:500), RPL5 (Santa Cruz (D-20), 1:200), RPL11

(Santa Cruz (N-17), 1:200), RPL23 (Abcam, 1:200), IGF1Rβ (Santa Cruz (C-20), 1:200), β-actin

(Santa Cruz (C4), 1:3000), and V5-HRP (Invitrogen, 1:5000). ECL™ anti-mouse and anti-rabbit

HRP-conjugated secondary antibodies (GE Healthcare) and anti-goat HRP-conjugated secondary antibody (Jackson ImmunoResearch) were used at 1:20000 dilution. Proteins were visualized using enhanced chemiluminescence reagents (Pierce).

Morpholino knockdown in zebrafish

Zebrafish were maintained under approved laboratory conditions. Morpholinos (MO) targeting the translational start site of igf1ra and igf1rb were ordered from Gene Tools, LLC

(Philomath, OR), Igf1ra MO: 5’-TCGCTGTTCCAGATCTCATTCCTAA-3’ and Igf1rb MO: 5’-

TGTTTGCTAGACCTCATTCCTGTAC-3’. The specificity of these MOs have been previously confirmed using multiple assays and described in Schlueter, PJ 2006. Stock solutions of MOs were diluted with molecular grade water and various concentrations ranging from 1-3ng of each

MO were injected into wild type (Tübingen) embryos at the one-cell stage. At 40 hours post

34 fertilization (hpf), embryos were stained for hemoglobin using benzidine, as previously described 149.

Results

A highly sensitive, quantitative SILAC-based proteomics screen identifies context-specific

MDM2 binding partners

To understand how ribosomal protein deficiency might alter the protein binding partner profile of MDM2, we utilized stable isotope labeling of amino acids in culture (SILAC) in combination with mass spectrometry as a proteome-wide approach to identify proteins bound to

MDM2 with and without RPS14 knockdown (Figure 2.1A). We chose to perform these experiments in the A549 cell line, which has an intact p53 pathway and has been used in many in vitro studies of ribosomopathy biology. We engineered the A549 cell line to express a V5-tagged

MDM2, and subsequently labeled the proteins in this clonal cell line as well as in the parental

A549 cell line with “light” (R0K0), “medium” (R6K4), or “heavy” (R10K8) amino acids over ten cell doublings in culture (>95% peptide labeling, data not shown). A forward and reverse labeling strategy was employed to prevent analytical bias by proteins whose MDM2 binding behavior may be affected by differential isotope labeling. Cells were then infected with lentivirus carrying shRNA constructs targeting either RPS14 or luciferase (LUC) as indicated in Figure 1A, and selected for stable hairpin expression with puromycin for 48 hours. Lysates were collected and MDM2 complexes were immunoprecipitated with an anti-V5 antibody. Immunoprecipitates from forward and reverse replicates were each mixed at a 1:1:1 (light:medium:heavy) ratio, and eluted proteins were then separated by SDS-PAGE and analyzed by LC-MS/MS. We then determined the relative abundance of proteins present in each condition and calculated the

35 median log2 ratios for proteins identified in each pairwise condition comparison using

MaxQuant.

Three separate pairwise comparisons were performed, allowing us to independently determine the full set of proteins found in MDM2 complexes with RPS14 knockdown (condition

B vs. condition A; Figure 2.1B, Figure 2.2A, and Supplementary Table 2.1), the full set of proteins found in MDM2 complexes without RPS14 knockdown (condition C vs. condition A;

Figure 2.1C, Figure 2.2B, and Supplementary Table 2.2), and those proteins preferentially bound to MDM2 with or without RPS14 knockdown (condition B vs. condition C; Figure 2.2C-

D, and Supplementary Tables 2.3 and 2.4).!A moderated T test was performed to identify significant protein interactors. MDM2 was abundantly present in all conditions analyzed, indicating successful purification of MDM2 complexes with the anti-V5 pull-down strategy.

A central premise of the putative ribosomopathy phenotype mechanism involves context- specific binding availability of a subset of cellular proteins. Because a goal of these experiments was to identify protein binding partners specific to RPS14 deficiency, RPS14 was knocked down in the parental A549 cells used for the “background” condition (condition A). In this way, we could accurately recapitulate the pool of proteins available for binding in this biological context, and exclude from downstream analysis those proteins in this specific pool that associated nonspecifically with the Sepharose® beads or anti-V5 antibody.

36 600 600 400 400 mean=−0.05 mean=−0.05 sd=0.49 sd=0.49 200 200 B C

0 A A 0 Trypsinized no%tag MDM2/V5 MDM2/V5 peptides

● ● 600

RPS14 RPS14 Luc 600 L 4 Replicate 4 M α"V5%IP H 1 L M H 400 α"V5%IP 400

2 M mean=H−0.05 L mean=−0.05 intensity ● ● 2 2 6 sd=0.49 sd=0.49 200 200 m/z ● rpsKD_ctrl 6 ● ● ● ● ● lucKD_ctrl 0 sh>RPS14 0 sh>LUC

0 B C MDM2 0 TP53 4 ● ● MDM2 4

4 MDM2 ● P53 ● 4 ● ● ● ● ●● ●● MDM2 rpsKD vs ctrl rep2 rpsKD vs ctrl rep2 H/M Rep2 2 2 ● ● ● ● L/M Rep2 L/M − P53 MDM4 ● − ● HIST1H4A MDM4 2 ● ●

2 DNAJC12 2 2 RPL5 RPL22L1 MFGE8

RPL11 RPL11 ● RPL23 ● ME3 ● RPL11 ● 4 FHL1 mean=0.02 sd=0.36 4 RPL23 ● ● mean=0.02 sd=0.36 − IGF1R − RPL5

[Replicate626(B6vs.6A)] RPS7 0 [Replicate626(C6vs.6A)] 0 0 0 2 4 6 0 0 2 4 6

M/L Rep1 H/L Rep1 −4 −2 0 2 ● 4 −40 200−2 400 6000 2 ● 4 0 200 400 600 ● ● ● ● [Replicate616(B6vs.6A)]●● [Replicate616(C6vs.6A)]●● rpsKD vs ctrl rep2 rpsKDrpsKD vs ctrl rep2 vs Ctrl rep1 rpsKD vs Ctrl rep1 2 2 ● ● ● ● − − ● ● D E 0 0.5 1 1.5 Gene( p.value( p.value( Protein(Name 4 mean=0.02 sd=0.36 4 Namemean=0.02 sd=0.36 (B(vs.(C) (B(vs.(A) − RPL5 − Cellular'tumor' TP53 <'0.001 <'0.001 RPL11 antigen'p53 −4 −2 0 2 4 0 −2004 400−2 600 0 2 4 0 200 400 600 RPL23 Insulin9like'growth' rpsKD vs Ctrl rep1 IGF1R <'0.01rpsKD vs<'0.01 Ctrl rep1 factor'1'receptor RPL26 Luc RPS14 RPL38

Figure 2.1. A SILAC-based proteomics screen identifies MDM2 binding partners with and without RPS14 knockdown.

(A) Schematic of proteomics screen. Parental A549 cells or A549 cells stably expressing a V5-tagged MDM2 were stably labeled with “light” (R0K0), “medium” (R6K4), or “heavy” (R10K8) amino acids and subsequently infected with lentiviral shRNA constructs targeting luciferase or RPS14, as shown. Lysates were collected after 48 hours of puromycin selection and MDM2 complexes were immunoprecipitated with anti-V5 antibody. Precipitated proteins in each condition were pooled in pairs and analyzed by mass spectrometry. (B) and (C) Proteins co-precipitating

with MDM2 with (B) and without (C) RPS14 knockdown. Log2 ratios for individual proteins are shown, with biological replicates are plotted on separate axes, and log2 ratiosare represented by single dots. (D) Abundance of

large subunit ribosomal proteins in MDM2 complexes. Average log2 ratios for two biological replicates are shown. (E) Partial list of proteins associated with MDM2 specifically in the context of RPS14 knockdown. Shown are proteins present on both [B vs. A] and [B vs. C] lists, with a p-value < 0.01.

37

A sh7RPS14 vs.*background B sh7LUC vs.*background 300

mean=0 sd=0.5 100 0 4 2 0 LucKD vs ctrlrep2 2 − mean=0.13 sd=0.58 [Replicate*2*(B*vs.*A)] [Replicate*2*(C*vs.*A)] 4 −

−4 −2 0 2 4 0 100 300 [Replicate*1*(B*vs.*A)] [Replicate*1*(C*vs.*A)]Luc KD vs Ctrl rep1

C sh7RPS14 vs.*sh7LUC D 400

mean=−0.03 200 sd=0.37 RPL5 P53 0 RPL11 4

2 IGF1R 0 rps vs ctrl rep2 2 − 0.13 − mean= sd=0.55 4 − [Replicate*2*(B*vs.*C)]

−4 −2 0 2 4 0 200 400

rps KD vs Ctrl rep1 [Replicate*1*(B*vs.*C)]

Figure 2.2. MDM2 complex proteins identified by mass spectrometry.

Full scatter plots of all proteins identified in each condition comparison are shown: (A) sh-RPS14 vs. background; (B) sh-LUC vs. background; (C) sh-RPS14 vs. sh-LUC. (D) is an enlargement of the upper right quadrant of (C), showing proteins preferentially bound to MDM2 in the context of RPS14 deficiency. Log2 ratios for individual proteins (represented by single dots) are shown, with biological replicates plotted on separate axes. Individual proteins are listed in Supplementary Tables 2.1-2.4.

38 Ribosomal proteins RPL5, RPL11, and RPL23 are present in MDM2 complexes with and without

RPS14 deficiency

Of those ribosomal proteins previously implicated in the RP-MDM2-p53 axis, RPL5,

RPL11, and RPL23 were among the most abundant proteins associated with MDM2 in the context of RPS14 deficiency, which is consistent with previous reports describing their MDM2- binding activity 123 (Figure 2.1B). Similarly, these three ribosomal proteins were also found to be abundantly associated with MDM2 in the control knockdown condition (Figure 2.1C). Direct comparison of these two conditions revealed that RPL5 and RPL11 exhibited a slight preference for MDM2 binding with RPS14, but this was not significant (RPL5 p = 0.076; RPL11 p = 0.033)

(Figure 2.2C and Figure 2.2D). There was no significant difference between the two conditions in the binding affinities for MDM2 by RPL23 (p = 0.901), RPL38 (p = 0.675) or RPL26, another ribosomal protein implicated in p53 activity modulation through MDM2 binding (p = 0.500) 119.

Validation of the IP/MS results was accomplished by subjecting whole cell lysates and

V5 immunoprecipitates from A549-MDM2-V5 cells to Western blot analysis for the indicated proteins (Figure 2.3). This analysis confirmed that the anti-V5 antibody efficiently enriched for

MDM2 complexes in the SILAC IP/MS experiment, and further confirmed that RPS14 knockdown readily induced p53 protein stabilization, as previously described 32, 113. The increased abundance of p53 in purified MDM2 complexes after RPS14 knockdown likely results from the marked increase in total p53 protein present in the whole cell lysate. The activation of p21, a downstream p53 target gene, indicated that the stabilized p53 was transcriptionally active and not aberrantly sequestered by the over-expressed MDM2.

39 Western blot analysis confirmed similar levels of RPL5, RPL11, and RPL23 proteins in

MDM2 complexes purified from luciferase and RPS14 knockdown cells. As the sensitivity of

SILAC IP/MS is significantly greater than that of Western blot analysis, small magnitude

10%$input IP:$V5

shRNA: luc S14 luc S14 MDM2

p53

RPL5

RPL11

RPL23

V5

p21

β7actin

Figure 2.3. Validation of ribosomal protein binding partners of MDM2 with and without RPS14 knockdown in A549 cells.

A549 cells stably expressing MDM2-V5 were infected with lentiviral shRNA constructs targeting luciferase or RPS14, and lysates were collected after 48 hours of puromycin selection. MDM2 complex members were immunoprecipitated from 1mg whole cell lysate input using an anti-V5 antibody, separated by SDS-PAGE, and probed with the indicated antibodies by Western blot. changes in MDM2 complex-member abundance detected by SILAC IP/MS may not be readily observable by Western blot. Nevertheless, taken together, these data confirm the presence of a subset of ribosomal proteins in MDM2 complexes in the context of ribosomal protein gene

40 deficiency, and further suggest that these proteins may commonly be present in MDM2 complexes without the prerequisite of ribosome dysfunction.

RPS7, another ribosomal protein with MDM2 binding activity and a putative role in the activation of p53 in response to ribosomal protein gene deficiency, was found to be highly abundant in MDM2 complexes in the luciferase knockdown condition (Figure 2.1C), but not in the RPS14 knockdown condition (Figure 2.1B). A strong preference for MDM2 binding in the absence of RPS14 knockdown was observed in the direct comparison of the two conditions (p =

0.018) (Figure 2.2C and Supplementary Table 2.4). Interestingly, this same preference was observed for all 40S subunit ribosomal proteins (Figure 2.4).

Luc S14

1.4

1.2

1

0.8

0.6

0.4 Average'log2'ratio 0.2

0 RPS2 RPS3 RPS5 RPS6 RPS7 RPS8 RPS9 RPSA RPS10 RPS11 RPS12 RPS13 RPS14 RPS15 RPS16 RPS17 RPS18 RPS19 RPS20 RPS21 RPS23 RPS24 RPS25 RPS26 RPS27 RPS28 RPS29 RPS30 RPS3A RPS4X RPS27L RPS15A

Figure 2.4. Abundance of 40S subunit ribosomal proteins in MDM2 complexes.

Average log2 ratios and standard error for two biological replicates are shown.

41 Over-expression of ribosomal proteins is insufficient for p53 pathway induction in cell lines and primary hematopoietic cells

Binding of ribosomal proteins to MDM2 in the context of RP deficiency is thought to be a result of increased availability of certain RPs in a “free”, non-ribosome bound pool. To determine whether an overabundance of ribosomal proteins alone – in the absence of ribosomal dysfunction – was sufficient to induce p53 pathway activity, we over-expressed human RPL5,

RPL11, RPL23, and RPS7 in A549 cells and primary human HSPCs using lentiviral vectors expressing full-length cDNAs for each RP. As a positive control, we also generated a vector to over-express ARF, a protein known to induce p53 pathway activity by binding and inhibiting

MDM2 150. RP over-expression was confirmed by qPCR (Figure 2.5).

RPL5 RPL11 4 5

4 3 3 2 2 1 1 Relative Expression Relative Expression

0 0

RPL5 Empty Empty RPL11

RPL5/RPL11RPL5/RPL23 RPL5/RPS7 RPL11/RPL5 RPL11/RPL23RPL11/RPS7 Overexpressed cDNA(s) Overexpressed cDNA(s)

RPL23 RPS7 4 15

3 10

2

5 1 Relative Expression Relative Expression

0 0

Empty RPL23 Empty RPS7

RPL23/RPL5 RPS7/RPL5 RPL23/RPL11RPL23/RPS7 RPS7/RPL11RPS7/RPL23 Overexpressed cDNA(s) Overexpressed cDNA(s)

Figure XX. Ribosomal protein over-expression in A549 cells. Figure 2.5. Ribosomal protein over-expression in A549 cells. RPL5, RPL11, RPL23, and RPS7 were over-expressed individually and in each pairwise RPL5, RPL11, RPL23, and RPS7combination were in over A549- expressedcells, and qPCR individuallywas performed and to assess in each the expression pairwise of combination each indicated in A549 cells, RP. Each graph shows the relative expression of the indicated RP in each condition in which that and qPCR was performed toRP assess was over the- expressed,expression as compared of each to indicated an empty vector RP. control. Each Errorgraph bars shows indicate the standard relative expression of the indicated RP in each conditiondeviation in between which technical that RP replicates. was over -expressed, as compared to an empty vector control. Error bars indicate standard deviation between technical replicates.

42 A p21 2.5

2.0

1.5

1.0

0.5 Relative Expression

0.0

ARF RPL5 Empty RPL11 RPL23 RPS7

RPL5/RPL11RPL5/RPL23RPL5/RPS7 RPL11/RPL23RPL11/RPS7RPL23/RPS7 Overexpressed cDNA(s) B Empty ARF RPL5 RPL11 RPL23 RPS7 L5/L11 L5/L23 L5/S7 L11/L23 L11/S7 L23/S7 p53

p21

β2actin

C p21 D p21 Flow

1000 100 800

80 600

5 MFI 4 400

Fold Change 3 2 200 1 0 0

ARF ARF Empty Empty

RPL5/RPS7 RPL5/RPL11RPL5/RPL23RPL5/RPS7 RPL5/RPL11RPL5/RPL23 RPL11/RPS7RPL23/RPS7 RPL11/RPL23RPL11/RPS7RPL23/RPS7 RPL11/RPL23 Overexpressed cDNA(s) Overexpressed cDNA(s)

Figure XX. Over-expression of ribosomal proteins is insufficient for p53 Figure 2.6. Overpathway-expression induction of inribosomal A549 cells proteins and primary is insufficient human HSPCs. for p53 pathway induction in A549 cells and primary human HSPCs. RPL5, RPL11, RPL23, and RPS7 were over-expressed individually and in each pairwise combination. (A-B) p53 pathway activity was assessed in A549 cells by p21 qPCR(A) and RPL5, RPL11, RPL23, and RPS7 were over-expressed individually and in each pairwise combination. (A-B) p53 p53/p21 Western blot (B). (C-D) p53 pathway was assessed in human HSPCs by p21 qPCR(C) pathway activity andwas intracellular assessed in flow A549 cytometry cells by forp21 p21 qPCR (D). (A) and p53/p21 Western blot (B). (C-D) p53 pathway was assessed in human HSPCs by p21 qPCR (C) and intracellular flow cytometry for p21 (D).

43 ARF over-expression readily induced p53 pathway activity in both A549 cells and primary human HSPCs, as measured by p53 protein levels, and p21 transcript and protein induction (Figure 2.6). In contrast, over-expression of either individual RPs or pairs of RPs did not result in p53 pathway activation in either cell type. These data suggest that, in the absence of ribosomal dysfunction induced by RP deficiency, overabundance of MDM2-interacting RPs may not be sufficient for inhibition of MDM2 and/or p53 pathway activation.

RPS14 deficiency induces MDM2-IGF1R binding and IGF1R protein loss in primary human hematopoietic cells

To identify MDM2 binding partners specific to the context of ribosomal protein gene deficiency, we selected candidates that were present on both lists of MDM2-associated proteins available from the pairwise comparisons with p < 0.01 between replicates (Supplementary

Tables 2.1 and 2.3). After exclusion of nonspecific associations, only two proteins met the candidate criteria: p53 and IGF1R (Figures 2.1B and 2.1E). While IGF1R is a known binding partner and ubiquitination target of MDM2 151, 152, this is the first example of IGF1R interacting with MDM2 in the context of ribosomal protein gene deficiency.

To investigate this context-specific MDM2-IGF1R interaction in primary human hematopoietic cells, where ribosomopathies most commonly manifest phenotypically, we depleted RPS14 in primary HSPCs with a gene-targeting shRNA and assessed IGF1R protein levels by Western blot and gene transcript levels by quantitative real-time PCR (qPCR) (Figure

2.7).

44 A C Luc S14 shRNA: luc S14 p53 Day: 6 8 10 12 14 6 8 10 12 14

p21 IGF1R

IGF1R&β MDM2

actin p53 p21

β&actin

B IGF1R 5 D IGF1Rigf1r cdkn1ap21 n.s. 3 15 4 sh-luc sh-luc sh-rps14 sh-rps14 3 2 10

2

Fold Change 1 5 Fold Change 1 Fold Change

0 0 0 sh-LUC sh-RPS14 6 8 10 12 14 6 8 10 12 14 Days in liquid culture Days in liquid culture

Figure 2.7. IGF1R is down-regulated at the protein level in response to RPS14 knockdown in primary human HSPCs.

(A-B) CD34+ hematopoietic stem and progenitor cells (HSPCs) were isolated from adult human peripheral blood and cultured in vitro in media permissive of erythroid precursor proliferation (see Methods). Cells were infected with lentiviral shRNA constructs targeting luciferase or RPS14 and subsequently collected and processed for protein and RNA analysis after 48 hours of puromycin selection. (A) Whole cell lysates were subjected to SDS-PAGE and probed with the indicated antibodies. (B) cDNA was synthesized from isolated RNA and subjected to qPCR analysis for IGF1R transcript expression levels. Each reaction was performed in triplicate, and error bars indicate variation across three biological replicates.

(C-D) CD34+ hematopoietic stem and progenitor cells (HSPCs) were isolated from human umbilical cord blood and cultured in vitro in media permissive of erythroid proliferation and differentiation (see Methods). Cells were expanded in vitro for four days before being infected with lentiviral shRNA constructs targeting luciferase or RPS14. Cells were collected and processed for protein and RNA analysis every two days after infection. (C) Whole cell lysates from each time point were subjected to SDS-PAGE and probed with the indicated antibodies. (D) cDNA was synthesized from RNA isolated at each time point and subjected to qPCR analysis for IGF1R (left) or p21 (right) transcript expression levels. Transcript expression levels were normalized to that of the luciferase knockdown sample at each time point. Each reaction was performed in triplicate, and error bars indicate standard deviation across three biological replicates.

45 Strikingly, IGF1R protein levels were significantly decreased by three days in RPS14- depleted HSPC cultures despite no significant reduction in IGF1R transcript. This protein decrease persisted for at least seven days, and the timing and degree of IGF1R protein loss was similar in HSPCs isolated from adult peripheral blood (Figure 2.7A-B) and umbilical cord blood

(Figure 2.7C-D). These data show a clear protein-level effect on IGF1R regulation in the context of ribosomal protein gene deficiency.

Dyserythropoiesis is induced in HSPCs lacking RPS14 or IGF1R

As IGF1R ligands (including IGF-1, IGF-2, and insulin) have previously been shown to promote in vitro erythroid differentiation of HSPCs in various liquid culture systems 153, 154, we hypothesized that IGF1R protein down-regulation in response to RPS14 knockdown might lead to dysregulation of in vitro erythroid differentiation of primary human HSPCs. We evaluated the consequences of IGF1R downregulation on the differentiation of primary HSPCs into committed erythroid cells by targeting IGF1R for depletion using lentiviral shRNA constructs (Figure

2.9D), and assessing their differentiation capacity with an in vitro liquid culture system promoting erythroid lineage differentiation. Erythroid differentiation was assessed by flow cytometry for the early and late erythroid differentiation markers CD71 and glycophorin A

(GlyA), respectively, on days 8, 11, and 14 of liquid culture. We also separately depleted RPS14 in order to have a positive control for dyserythropoiesis, as depletion of RPS14 is known to result in an erythroid differentiation block in primary HSPCs 113.

On day 8 of liquid culture (4 days after stable shRNA expression), RPS14-depleted cultures contained significantly fewer cells with markers of both early (CD71+/GlyA-) and late

(CD71+/GlyA+) erythroid differentiation as compared to the negative control (Figure 2.8A-B).

46 A sh#LUC sh#RPS14 sh#IGF1R

CD71 GlyA

B Surface MarkerDay.8 Expression Surface MarkerDay.11 Expression 80 100 *** n.s. sh-LUC n.s. sh-LUC *** sh-RPS14 *** sh-RPS14 *** sh-IGF1R 1 80 ** sh-IGF1R 1 60 sh-IGF1R 2 sh-IGF1R 2 60 *** 40 *** n.s. 40 % Live Cells % Live Cells ** *** *** 20 20

0 0 CD71+/GlyA- CD71+/GlyA+ CD71+/GlyA- CD71+/GlyA+

C D p21 15 sh-RPS14 * 1.0 sh-IGF1R #1 sh-LUC 10 sh-IGF1R #2 sh-RPS14 sh-IGF1R *** 0.5 5 *** Fold Change Cell count (x 10^6)

0.0 0 1 4 Day 8 Day 11 Day

Figure 2.8. IGF1R loss results in dyserythropoiesis in primary human HSPCs.

CD34+ hematopoietic stem and progenitor cells (HSPCs) were expanded in vitro for four days before being infected with lentiviral shRNA constructs targeting luciferase, RPS14, or IGF1R. Cells were collected and processed for flow cytometry and RNA analysis four days after infection (“day 8”). (A) Representative flow plots of cells infected with each hairpin and analyzed on day 8 for cell surface expression of CD71 and glycophorin A. (B) Graphical representation of flow data depicted in part in (A). Error bars indicate standard deviation across three biological replicates. (C) Total cell proliferation was assessed by cell counting over 4 days in culture. Cell numbers were normalized on day 7 of the in vitro culture protocol, after 48 hours of puromycin selection (“day 1” in this panel), and counted 3 days later (“day 4”). (D) p21 gene expression was assessed on days 8 and 11 of in vitro culture.

47 In contrast, IGF1R-depleted cell cultures contained normal levels of early erythroid cells but had fewer late erythroid cells, with levels of the latter more similar to those observed in RPS14- depleted cells. Similar trends were observed on day 11 of liquid culture, but to a lesser degree.

By day 14 of liquid culture, >95% of cells in all conditions expressed markers of late erythroid differentiation (data not shown), suggesting that prolonged exposure to these erythroid cytokine- rich culture conditions can eventually drive erythroid differentiation despite the delay caused by either RPS14 or IGF1R depletion. These results indicated that IGF1R depletion impairs late erythroid differentiation, but to a lesser extent than RPS14 depletion. Total cell proliferation, which was significantly affected in RPS14-deficient cells, did not appear to be affected by

IGF1R depletion (Figure 2.8C). p21 expression was strongly induced in RPS14-deficient cells, but not IGF1R-deficient cells (Figure 2.8D).

RPS14 and IGF1R deficiency alters the expression of genes involved in HSPC differentiation

To further characterize the effects of IGF1R depletion on erythroid differentiation, we isolated RNA from cells collected on days 8 and 11 of in vitro culture and performed qPCR to profile the expression of 84 genes related to the development of blood-cell lineages from HSPCs.

On day 8 both IGF1R- and RPS14-deficient cells had significantly higher mRNA expression of markers associated with more primitive hematopoietic cell stages such as CD34, CD11a, and

CD27 (Table 2.1 and Figure 2.9B). These data corroborate the cell surface marker expression

FACS analysis and further illustrate an erythroid differentiation defect caused by IGF1R depletion that is similar to, but less severe than, that caused by RPS14 depletion.

Additionally, a number of transcription factors and regulators of hematopoiesis, GATA-1,

LMO2, TAL-1, and TRIM10, were all significantly down-regulated by day 8 of in vitro culture as

48 A GATA-1 LMO2 TAL-1 TRIM10 0 0 0 0 sh-RPS14 sh-rps14 sh-rps14 sh-rps14 sh-IGF1R #1 -20 sh-igf1r #1 sh-igf1r #1 sh-igf1r #1 -5 -1 -1 sh-IGF1R #2 sh-igf1r #2 sh-igf1r #2 sh-igf1r #2 -40 -10 -60 -2 -2 Fold Change Fold Change Fold Change -15 Fold Change -80

-3 -3 -20 -100

B CD34 CD11a CD27 400 250 50 sh-RPS14 sh-RPS14 sh-RPS14 200 40 sh-IGF1R #1 sh-IGF1R #1 sh-IGF1R #1 300 sh-IGF1R #2 sh-IGF1R #2 sh-IGF1R #2 150 30 200 100 20 Fold Change Fold Change Fold Change 100 50 10

0 0 0

C D HPRT1 MMP9 RPS14 IGF1R 3 1.5 1.5 sh-LUC sh-LUC ** 0 sh-LUC sh-LUC * sh-IGF1R #1 sh-IGF1R #1 sh-RPS14 sh-IGF1R #1 sh-IGF1R #2 sh-IGF1R #2 2 1.0 1.0 sh-IGF1R #2 -10 * ** 1 0.5 0.5 Fold Change

Fold Change -20 % Expression % Expression

0 -30 0.0 0.0

Figure 2.9. Gene expression analysis of RPS14- and IGF1R-deficient HSPCs. cDNA was synthesized from RNA isolated on day 8 of in vitro culture and subjected to qPCR array analysis for a panel of hematopoiesis genes. Expression levels relative to sh-LUC infected control are shown. Each reaction was performed in triplicate, and error bars indicate standard error across three biological replicates. Panels are grouped as follows: (A) erythroid genes; (B) cell surface markers; (C) other genes of interest; (D) knockdown validation. All values in panels A-C are also displayed in Table 2.1.

49 p p 0.065 0.003 0.047 0.004 0.030 0.147 0.004 0.350 0.874 0.564 0.001 0.215 0.593 0.024 1.2 1.3 1.4 3.2 1.5 1.9 1.1 1.1 ------26.8 45.0 12.2 16.1 Upper 237.9 Upper #2 #2 1130.3 SEM SEM IGF1R IGF1R IGF1R - - sh sh 8 - 1.3 1.5 2.2 1.7 1.2 2.4 6.2 1.8 2.5 1.0 - - - 26.5 - - - 34.5 51.4 - Lower Lower 1.2 1.4 1.7 5.1 1.5 2.6 1.6 1.1 2.4 8.9 1.1 ------Fold Fold 12.8 197.6 110.6

- p p 0.066 0.007 0.481 0.024 0.002 0.196 0.064 0.990 0.489 0.971 0.686 0.283 0.506 0.490

RPS14 1 1.5 1.7 1.1 3.6 6.5 9.5 9.1 1.1 1.2 1.5 - - - - 73.5 24.9 Upper 667.3 Upper #1 #1 SEM SEM Day 8 IGF1R IGF1R Day 11 - - targeting shRNA. targeting sh sh - 1.2 2.1 2.2 2.3 9.9 1.6 1.2 1.3 1.6 1.0 ------10.4 - - - 29.7 19.1 610.9 - Lower Lower - 1.7 1.9 1.6 1.1 1.1 1.2 2.3 1.0 3.9 1.0 1.2 - - - 46.9 - - - Fold Fold 37.5 - 140.8 p p 0.096 0.002 0.027 0.017 0.061 0.029 0.087 0.629 0.132 0.432 0.689 0.007 0.577 0.076 f select hematopoiesis genes in genes hematopoiesis select HSPC f 1.4 1.3 2.8 2.1 1.2 1.4 8.2 1.1 1.2 ------46.3 38.2 99.2 Upper 242.2 Upper 1146.8 RPS14 RPS14 SEM SEM - - sh sh 27 2.5 3.2 3.2 1.1 1.2 1.4 3.1 - 9.1 4.0 1.2 - - 31.9 - - - - - 14.8 58.6 - Lower Lower 1.9 2.0 8.8 8.2 1.3 1.2 2.1 3.5 2.7 1.0 ------Fold Fold 20.6 19.9 130.1 119.1 deficient HSPC cultures. cultures. HSPC deficient - Gene expression analysis o Gene expression . 1 . 2 IGF1R Cell Surface Markers Cell CD34 CD11A CD27 and Regulators Erythropoiesis Factors Transcription GATA1 LMO2 TAL1 TRIM10 Surface Markers Cell CD34 CD11A CD27 and Regulators Erythropoiesis Factors Transcription GATA1 LMO2 TAL1 TRIM10 Table Table and change Fold in shownexpressionis relative of expressing to that cells luciferase

50 compared to the luciferase control knockdown cells (Table 2.1 and Figure 2.9A). In IGF1R- deficient cells, these same erythroid genes were generally downregulated to a lesser extent than in RPS14-deficient cells, which correlates with the more profound erythroid differentiation defect phenotype seen in RPS14-deficient cells. These genes were also down-regulated on day 11 of in vitro culture, but less profoundly than at day 8 (Table 2.1), which correlates with the cell surface marker expression. Consistent with on-target effects of our IGF1R shRNAs, MMP9 was also found to have decreased expression in IGF1R-deficient CD34+-derived cell cultures, which has been shown previously to be down-regulated in response to IGF1R silencing 155(Figure

2.9C). In relation to changes in nucleotide that have been seen in ribosomopathies 156,

HPRT1, a nucleoside salvage enzyme gene recently found to be up-regulated and potentially correlated with p53 pathway activation in zebrafish models of DBA 157, was also more highly expressed in IGF1R-deficient cells (Figure 2.9C).

Taken together, these data show that IGF1R depletion independently impairs erythroid differentiation in primary human HSPCs, and further suggests that the selective loss of IGF1R in the context of RPS14 deficiency contributes to the dyserythropoiesis phenotype.

IGF1R deficiency impairs erythroid development in an in vivo zebrafish model

Hematopoiesis in zebrafish is similar to that of humans, occurring in two distinct waves, with the first occurring ~12-24 hours post fertilization, and gives rise to a transient populations of progenitors that undergo differentiation into erythrocytes. To test the effects of IGF1R loss on in vivo hematopoiesis, we injected previously validated Igf1r morpholinos into one-cell zebrafish embryos and assessed terminal erythroid differentiation by performing a o-dianisidine

(benzidine) stain for hemoglobin at 48 hours post-fertilization (hpf) 88. Staining was visibly

51 decreased in morpholino-injected embryos, demonstrating hemoglobinization impairment as a result of IGF1R loss. A dose-dependent response was seen with the severity of impairment correlating with the morpholino dosage: at 1ng of Igf1r morpholinos 7 out of 19 embryos (37%) were affected (p<0.001), with 6 out of 13 embryos (46%) with significantly decreased staining using 3ng of Igf1r morpholinos (p<0.001) (Figure 2.10). These data support the in vitro findings in primary human HSPCs and identify IGF1R as an important contributor in vivo to erythroid development.

Uninjected igf1r&α/β&MO

7/19 6/13

Figure 2.10. IGF1R loss results in dyserythropoiesis in zebrafish.

One-cell zebrafish embryos were injected with 1ng (middle panel) or 3ng (right panel) each of morpholino targeting both alpha and beta isoforms of igf1r and subjected to benzidine staining at 40hpf.

Discussion

Here we show that a number of ribosomal proteins, including RPL5, RPL11, RPL23,

RPL26, and RPL38, are commonly associated with MDM2, though ribosomal protein deficiency may enhance the association of RPL5 and RPL11. Furthermore, we showed that RPS14 deficiency in primary HSPCs strongly induces binding of MDM2 to IGF1R, which is subsequently degraded. Moreover, genetic silencing of IGF1R in HSPCs resulted in defective in

52 vitro erythropoiesis, suggesting that MDM2-mediated IGF1R loss in RPS14-deficient cells contributes to the erythroid differentiation block observed in 5q- syndrome and other ribosomopathies such as Diamond-Blackfan anemia.

Recent work has focused more closely on the roles of RPL5 and RPL11 in the inhibition of MDM2-mediated regulation of p53, potentially through an inhibitory RPL5/RPL11/5S rRNA ribonucleoprotein (RNP) complex 158, 159. Our data provide compelling evidence that both RPL5 and RPL11 are strongly associated with MDM2 in the context of RPS14 deficiency, as they ranked among the most abundant associated proteins in this condition. However, it was somewhat unexpected that both RPL5 and RPL11 also ranked highly among those proteins associated with MDM2 without RPS14 deficiency, albeit less abundantly. These data suggest that abundant interaction of RPL5 and RPL11 with MDM2 is not unique to the context of ribosomal protein deficiency.

Over-expression of ribosomal proteins was technically difficult to achieve (requiring the spleen focus-forming virus (SFFV) promoter, which drives high levels of expression in mammalian cells), perhaps because ribosomal proteins are already expressed at very high levels in proliferating cells. It is possible that modest RP over-expression (ranging from 2- to 10-fold increases in our experiments, depending on the RP) is not sufficient to recapitulate the levels of free RPs present in cells with ribosomal dysfunction, and that more substantial over-expression could result in p53 pathway induction.

The necessity of RPL5/RPL11-MDM2 binding in the ribosomal perturbation-induced p53 response has been demonstrated through the characterization of mice harboring a cancer- associated mutant MDM2 that lacks the ability to bind RPL5 and RPL11 124. However, the sufficiency of RPL5/RPL11-MDM2 binding in p53 pathway induction without ribosomal

53 dysfunction remains to be rigorously shown. Taken together, the finding that RPL5 and RPL11 are commonly and abundantly associated with MDM2 along with the failure of RP over- expression to induce p53 pathway activity raise the possibility that increased binding of these two RPs is part of a larger set of signals that contribute to MDM2 activity modulation as a result of ribosome dysfunction.

Utilizing a SILAC-based approach, these data provide the first highly quantitative measure of relative RP-MDM2 binding changes in the context of ribosomal protein deficiency, and thereby provide a resolution perhaps not achievable by other previously-employed methods.

This may also explain why differences in RPL5 and RPL11 binding with and without RPS14 deficiency were not observed by IP/Western analysis, but were detectable in the IP/MS analysis.

Our data further shows that RPS7 also binds to MDM2, but that this binding is decreased in RPS14-deficient HSPCs. In the studies that initially characterized the RPS7-MDM2 binding interaction, like many of the other studies that characterized other RP-MDM2 binding interactions, ribosome dysfunction was induced through nucleolar disruption, which broadly affects ribosome biogenesis but is not seen clinically in ribosomopathy patients. In our study, ribosomal haploinsufficiency is modeled by shRNA-mediated knockdown of a single ribosomal protein gene (RPS14) to roughly 50% of wild-type levels, which more closely mimics the heterozygous loss observed in 5q- syndrome patients. However, our data do not rule out the possibility that RPS7-MDM2 binding may be important in other ribosomopathies where other

RPS or RPL proteins are mutated or lost.

We also identified IGF1R as a novel context-specific binding partner of MDM2, and further showed that IGF1R protein is lost following shRNA-mediated depletion of RPS14. As

IGF1R is a known ubiquitination substrate of MDM2 151, we propose that RPS14

54 haploinsufficiency leads to the MDM2-mediated degradation of IGF1R. In support of this, our proteomics screen also showed that beta-arrestins, which have been shown to be required in complex with MDM2 for ubiquitination of various surface receptor substrates 151, 160, 161, were associated with MDM2 in HSPCs that had reduced expression of RPS14 (Suppplementary

Tables 2.1-2.4).

The role of IGF1R in erythropoiesis has remained somewhat unclear, with conflicting reports regarding the efficacy of IGF1R ligands (including IGF-1, IGF-2, and insulin) in stimulating erythropoiesis, which is likely attributable to inconsistencies in the cytokine cocktails employed in different experimental systems. Although IGF1R is not easily detectable on the surface of primitive (CD34+) hematopoietic cells (Raiser DM unpublished results) 162, IGF-1 has been shown to stimulate proliferation of erythroid progenitors 163, 164, and is commonly included in in vitro culture media to promote expansion of erythroid precursors 153, 165.

Our data support a role for IGF1R in erythropoiesis, as demonstrated by the erythroid differentiation defect observed in primary hematopoietic cells with shRNA-mediated silencing of

IGF1R. Further, our data suggest that IGF1R loss as a result of ribosomal protein deficiency, modeled here by RPS14 knockdown, contributes to the dyserythropoiesis characteristics of many ribosomopathies. As evidenced by a lack of p21 induction in response to IGF1R depletion, it is possible that these effects are independent of p53. In order to pursue the provocative therapeutic possibilities of targeting IGF1R in the treatment of ribosomopathies, it will be important to tease out the specific effects of IGF1R loss on erythropoiesis in the context of ribosomal protein deficiency versus those effects more generally associated with ribosome dysfunction.

Interestingly, recent clinical trials of cixutumumab, a monoclonal antibody therapy targeting

IGF1R for the treatment of pediatric refractory solid tumors, reported that the most common

55 hematologic toxicity observed in patients was anemia, supporting our hypothesis that reduction in IGF1R results in dyserythropoiesis 166.

The role of IGF1R in and embryonic development is fairly well-established

167-169, and our findings are the first to demonstrate an effect on IGF1R as a result of ribosomal protein deficiency. Another intriguing possibility raised by these findings is a potential role for

MDM2-mediated IGF1R loss in the developmental abnormalities observed in ribosomopathies such as DBA, though further investigation is warranted to probe this possible relationship.

56 !: Results II

Calmodulin inhibition rescues the effects of ribosomal protein deficiency by modulating p53 activity in models of Diamond-Blackfan anemia

Attributions

I performed the in vitro work in close collaboration with Alison Taylor in Leonard Zon’s laboratory. I prepared virus and infected A549 cells and HSPCs. I performed immunofluorescence staining, Western blots, and some of the qPCR. I also performed the immunoprecipitations for the IP/MS experiment. Liz Macari assisted with qPCR and CHK2 inhibitor experiments in zebrafish. Chemical treatments, RNA/protein collection, and data analysis were performed jointly with Alison and Liz.

The chemical screen and initial validation in zebrafish was performed by Alison Taylor with technical assistance from Jessica Humphries. Compilation of the SILAC IP/MS results and statistical analyses of the proteomics were performed by Monica Schenone and Emily Hartmann at the Broad Institute. Mouse experiments were performed by Kavitha Sita in Johan Flygare’s laboratory at Lund University.

Note: We are currently writing a manuscript for publication based on the work presented in this chapter, with Alison Taylor and I as equally contributing co-lead authors.

58 Summary

Ribosomal protein (RP) mutations cause Diamond Blackfan anemia (DBA), characterized by ribosomal stress and p53 activation. To identify compounds that rescue ribosomal stress, we performed a chemical suppressor screen in a zebrafish model of DBA with p53-dependent anemia, and found that calmodulin inhibitors rescued all phenotypes. In a human cell line, calmodulin inhibition attenuated nuclear accumulation and transcription factor activity of p53 upon RP knockdown. Inhibition of CHK2, a calmodulin-dependent kinase, phenocopies calmodulin inhibition. In both murine and human hematopoietic models of DBA, calmodulin inhibitors rescue the erythroid differentiation block caused by RP deficiency. Our work describes a novel calmodulin-dependent mechanism of p53 regulation and suggests that calmodulin or

CHK2 inhibition could be therapeutic in diseases of ribosome dysfunction.

Introduction

Diamond Blackfan anemia (DBA) is a congenital anemia that presents in young children7.

The primary phenotype is a block in erythroid differentiation, and some patients also have craniofacial anomalies, short stature, and thumb abnormalities. Ribosomal protein S19 (RPS19) was the first gene found mutated in DBA patients 170. The sequencing of patient samples has identified mutations in a set of both large and small subunit ribosomal proteins, including RPS29, in over 50% of patients 171, 172. Patients are universally heterozygous for these mutations, always maintaining a wildtype copy of the affected ribosomal protein gene.

In both yeast and human cells, RPS19 knockdown or mutation causes decreased 40S subunit assembly and aberrant processing of 18S rRNA 80, 173. In vivo, ribosomal protein mutants

59 have stunted growth and are homozygous lethal, as seen in several drosophila minute mutants and zebrafish mutants 91, 174. However, evidence suggests that ribosomal protein mutations affect blood cell development in a p53-dependent manner. Ribosomal protein deficiency leads to an increase of free ribosomal proteins, a subset of which bind MDM2 and putatively inhibit its p53 ubiquitination activity, leading to increased p53 levels in the context of ribosomal protein deficiency 24, 116, 117, 175. Evidence has also been shown that p53 translation can be specifically upregulated as a result of ribosomal protein binding to p53 transcripts 125.

p53 activation has been shown to be a critical mediator of hematopoietic defects in many

DBA models. In human CD34+ cells, RPS19 deficiency leads to an erythroid-specific accumulation of p53, and differentiation defects can be rescued by p53 inhibition 113. Mouse models of Rps19 mutation or knockdown have hematopoietic defects that can be rescued by p53 mutation 100, 104. Targeting of rps19 by morpholino causes hematopoietic defects in zebrafish embryos, which some studies have shown is dependent on p53 84. Hematopoietic defects seen in rpl11 mutant zebrafish are also rescued by p53 knockdown 176, 177. These studies establish that ribosomal stress requires p53 to cause many tissue-specific effects.

We previously characterized zebrafish rps29 mutants that have hematopoietic and endothelial defects 88. Rps29-/- embryos have a defect in arterial specification, leading to decreased HSCs and decreased flk1 expression in the intersegmental vessels at 24 hours post fertilization (hpf). Primitive erythropoiesis is also affected, as seen in rps29-/- embryos by lower levels of hemoglobin. These embryos also have increased apoptosis, particularly in the head region, and die by five days post fertilization (dpf). p53 pathways are activated in the embryo, and p53 mutation rescues all hematopoietic and apoptotic phenotypes.

60 Improving treatment options for DBA patients is a primary goal of research being conducted using these disease models. Standard therapy includes regular blood transfusions and/or steroids 7. Although patients may undergo a spontaneous remission, patients who remain on treatment have significant side effects such as iron overload and other complications.

Currently, the only known cure for DBA is a hematopoietic stem cell transplant, which carries its own risks. Recent studies suggest that lenalidomide and leucine may be useful therapies, and clinical trials with these drugs are ongoing 6, 75.

In an effort to identify small molecules that could rescue defects caused by ribosomal protein deficiency, we performed a chemical screen on rps29-/- embryos. Several calmodulin inhibitors were found to rescue mutant phenotypes, including flk1 expression and hemoglobin levels. In a human cell model where RPS19 knockdown activates p53, calmodulin inhibitors attenuated p53 activity due to a defect in nuclear localization of p53. Inhibition of calmodulin- dependent CHK2 also resulted in reduced p53 nuclear localization and activity. Inhibitors of either calmodulin or CHK2 rescued erythroid differentiation defects in murine and primary human CD34+ cell assays. Our studies elucidate a mechanism for calmodulin-dependent modulation of p53 activity, and provide early evidence of therapeutic potential in patients with

DBA.

Materials and Methods

Embryo manipulation and screening

Fish were maintained under approved laboratory conditions. Studies were performed on

AB wildtype strains and hi2903, an insertional mutant in the first of ribosomal protein s29

61 (rps29) 91. Irradiation was performed on 24 hpf AB embryos, one 10Gy dose. For morpholino injections, 125-375 picograms of chk2 morpholino 178 were injected into 1-2 cell embryos.

Embryonic Chemical Treatment

For the chemical treatments, rps29+/- fish were incrossed, and embryos were collected for treatment at bud stage (10 hpf). Embryos were treated from bud to 24 or 48 hpf with compounds of known bioactivity. For screening, chemicals from two libraries were tested at 1:300 dilutions

(in E3) from library stock: BIOMOL 480 (Enzo Life Sciences, Farmingdale, NY) and Sigma

Lopac1280 (Sigma-Aldrich, St. Louis, MO). Compounds were tested in two independent experiments of twenty embryos each, so approximately ten mutant embryos were scored per chemical. The following chemicals were diluted in DMSO or water and tested in doses from 5-

50 µg/mL: A-3 (Enzo Life Sciences), W-7 (Tocris Bioscience, Minneapolis, MN), A-5 (Tocris

Bioscience), W-5 (Enzo Life Sciences), CGS-9343B (Tocris Biosceince), and BML-277 (Enzo

Life Sciences), and trifluoperazine (TFP) (Enzo Life Sciences). YS-035 (Sigma-Aldrich) was diluted in water and treated in doses from 0.8 to 8 µg/mL.

In situ hybridization, phospho-H3, and benzidine staining

Whole-mount in situ hybridization was performed as described 179, 180. Antisense probes were synthesized from digested plasmid. Flk1 staining was counted as rescued if most of the intersegmental vessels had flk1 expression by ISH. Phospho-H3 antibody was used to identify proliferating cells as previously described 181. Benzidine staining was performed as described previously 149. Head morphology was counted as rescued if the mutant embryos could no longer be distinguished from wildtype embryos at 24hpf.

62 Cell culture and drug treatment

A549 cells were cultured in F-12K medium supplemented with 10% fetal calm serum and

1% penicillin/streptomycin. Unless otherwise noted, drugs were added one day post-infection

(described below), and media with or without drug was changed daily for the course of the experiment. After 2-5 days of drug treatment, cells were trypsinized and collected for mRNA expression, Western blot, or immunofluorescence analysis. Culture conditions for CD34+ hematopoietic stem and progenitor cells are described below. All cells were maintained at 37°C and 5% CO2.

Lentiviral vectors and infection

Lentiviral shRNAs in the pLKO.1 vector were obtained from the Broad Institute of

Harvard and MIT. The lentiviral vectors used in this work have been characterized previously 113.

Lentivirus was produced in 293TL cells as described previously 143. Cells were infected with lentivirus in the presence of 8 µg/mL polybrene (Sigma-Aldrich) and selected 24 hours later with

2 ug/mL puromycin (Sigma-Aldrich) for at least 48 hours prior to analysis.

Flow cytometry and immunofluorescence

For flow cytometry based measurement of intracellular protein levels, cells were fixed in

2% paraformaldehyde for 20 minutes at 37°C, and methanol was added for overnight incubation at 4°C. Cells were incubated for one hour in 1:100 diluted p21 primary antibody (rabbit polyclonal, 12D1 Cell Signaling, Danvers, MA) followed by one hour in AlexaFluor 750- conjugated anti-rabbit secondary antibody (Molecular Probes, Eugene, OR) and 1:50 diluted p53-conjugated antibody (mouse monoclonal, 1C12 Cell Signaling). For flow cytometry

63 measuring cell surface markers, differentiated CD34+ cells were incubated for 30 minutes with

PE-Cy5-conjugated anti-CD71 (BD Biosciences PharMingen, San Jose, CA).

Immunofluorescence staining was performed as previously described using an antibody against p53 (mouse monoclonal, Sigma Aldrich DO-1, 1:100) 113.

Methods for Quantification of Localization

Images were quantified in two ways. Individual cells were blindly scored and counted based on degree of nuclear intensity compared to intensity throughout the cell. Cells with a stronger nuclear intensity compared to cytoplasmic intensity were considered to have nuclear p53 localization. If the fluorescence intensity was diffuse and equal in the cytoplasm and nucleus, cells were not considered to have nuclear localization of p53. In addition, for 10-15 cells per condition, average nuclear and cytoplasmic intensity was determined using ImageJ.

Stable isotope labeling by amino acids in cell culture (SILAC) and immunoprecipitation

A549 cells were cultured in RPMI medium containing “light” amino acids (R0, K0) or

“heavy” amino acids (R8, K10) for a minimum of 10 cell doublings to ensure complete labeling of the proteome. Cells were infected with a lentiviral vector targeting RPS19 one day after plating and treated 24 hours later with TFP or vehicle for 48 hours. Cells were then trypsinized and whole cell lysates were collected. 1 mg of protein was incubated with 1 µg of anti-p53 antibody (DO-1, Santa Cruz Biotechnologies, Santa Cruz, CA) overnight at 4°C and with

Sepharose G beads (GE Healthcare Life Sciences, Pittsburgh, PA) for 4 hours at 4°C the following day. Beads were washed with lysis buffer, and bound protein was eluted.

64 Immunoprecipitates from light and heavy conditions were combined, separated by SDS-PAGE, and analyzed by LC-MS.

Quantitative PCR

RNA was isolated from cells using the RNeasy kit according to manufacturer instructions

(Qiagen, Valencia, CA). cDNA was synthesized with equal amounts of RNA using iScript cDNA synthesis kit (Biorad, Hercules, CA). Real-time PCR was performed with SsoFast

Evagreen Supermix (Biorad) and gene expression was calculated relative to GAPDH according to methods previously described 182, 183.

Statistics

For each zebrafish embryo experiment and all cell counts, p-value is taken from binomial distribution calculations. All other p-values were calculated using the Student’s T Test. * = p<0.05, ** = p<0.01, *** = p<0.001.

Mice

Mice expressing inducible Rps19 shRNA have been described previously 74. Wildtype

C57BL/6 recipient mice were irradiated with 900 cGy and transplanted with 2x106 unfractionated bone marrow cells isolated from donor mice harboring two copies each of the

Rps19 inducible hairpin construct and the doxycycline-responsive M2-rtTA element. Cells were allowed to engraft for 7 weeks prior to treatment. Beginning on day 1 of each experiment, mice were given doxycycline (2 mg/mL in drinking water, with 10 mg/mL sucrose) to induce the

Rps19-deficiency phenotype. Starting on day 2, TFP (5 mg/kg) or saline vehicle was injected

65 intraperitoneally every other day for two weeks. On day 15, peripheral blood was collected and analyzed for red blood cell counts and hemoglobin levels, and bone marrow was harvested for

RNA isolation and qPCR analysis.

Culture Conditions for CD34+ Cells

CD34+ hematopoietic stem and progenitor cells (HSPCs) were purified from human umbilical cord blood and maintained in liquid culture. For p21 expression experiments, cells were cultured in medium supportive of erythroid differentiation: serum-free expansion medium

(StemCell Technologies, Vancouver, British Columbia), 100 U/mL penicillin/streptomycin, 2 mM glutamine, 10 g/mL lipids (Sigma-Aldrich), 100 ng/mL stem cell factor (SCF), 10 ng/mL interleukin-3 (IL-3), and 0.5 U/mL erythropoietin (EPO). On day 7 of liquid culture, the concentration of EPO was increased to 3 U/mL. In experiments evaluating differentiation, 15 ng/mL granulocyte colony-stimulating factor (G-CSF) (Neupogen; Amgen, Cambridge, MA) and 40 ng/mL FLT3 ligand were added to support myeloid differentiation as well. Cells were allowed to expand for 4 days prior to infection. After infection (see below), drug (A-3, TFP, or

BML-277) or vehicle was added to the media. Media with or without drug was changed daily for the course of the experiment. Cells were harvested for flow cytometric and/or gene expression analysis after 10 days of liquid culture.

66 Results

Chemical screen finds calmodulin inhibitors rescue rps29-/- defects

We performed a screen to identify chemicals that could rescue the endothelial and morphological defects of the rps29-/- mutant embryo (Figure 3.1). We chose to screen for rescue of morphological and angiogenesis defects as these readouts were more readily detectable than the rescue of anemia. 600 bioactive chemicals were screened in duplicate. Rps29-/- embryos were scored for rescue of head morphology, and subsequently fixed at 24 hpf for in situ hybridization of flk1 and rps29. Embryos without rps29 staining (rps29-/- mutants) were scored for rescue of flk1 intersegmental vessel staining. One of the compounds identified in the screen to rescue flk1 expression was W-7, a naphthalenesulfonamide that inhibits calmodulin (Figure 3.2A). We tested other naphthalenesulfonamides known to inhibit calmodulin, including A-7 and W-5, and they also rescued the vasculature defect (Figure 3.2B). To verify that calmodulin was the relevant target, we tested known calmodulin inhibitors of different structural classes. Several structurally dissimilar calmodulin inhibitors rescued flk1 expression, including CGS-9343B and members of the phenothiazine family such as trifluoperazine (TFP), as well as calcium channel blockers (Figure 3.1, Figure 3.2B). There are six genes for calmodulin in the zebrafish genome that are identical at the amino acid sequence, precluding a genetic analysis of calmodulin deficiency 184. Taken together, the effects of these chemicals establish that calmodulin inhibition is capable of rescuing ribosomal protein deficiency of flk1 expression.

67

Figure 3.1. Zebrafish in vivo screen identifies calcium channel blockers that rescue vasculature defect in the rps29-/- embryo.

(A) Chemical screen design. Rps29+/- fish were incrossed, and embryos were collected for treatment at bud stage (10 hpf). Embryos were treated from bud to 24 hpf with compounds of known bioactivity. At 24 hpf, mutant embryos were scored for decreased cell death in the head. Embryos were then fixed for whole mount in situ hybridization (ISH), and stained for both flk1 and rps29. (B) Flk1 expression by ISH at 24 hours post fertilization (hpf).

Figure 3.2. Calmodulin inhibitors rescue vasculature defect in the rps29-/- embryo.

(A) and (B) Flk1 expression by whole mount in situ hybridization (ISH) at 24 hours post fertilization (hpf). Arrows denote location of intersegmental vessels. Numbers denote embryos that resembled the representative image.

68 Calmodulin inhibition rescues head and hemoglobin defects in rps29-/- embryos

Rps29-/- mutants have increased apoptosis visible in the head of the embryo. Of the 600 chemicals screened, only A-3 rescued the morphology of the rps29-/- mutant head (Figure 3.3A).

A-3 is a structural derivative of W-7 and a known calmodulin inhibitor. Our previous work has shown that irradiation of zebrafish embryos activates p53 similarly to ribosomal protein deficiency 88, so we tested A-3 in irradiated embryos. Irradiation of wildtype embryos at 24 hpf leads to decreased proliferation as measured by phospho-H3 staining. A-3 treatment of irradiated embryos can mitigate the decreased proliferation levels caused by irradiation (Figure 3.3B).

Treatment with A-3 or W-7 can also rescue hemoglobin levels in the rps29 mutant embryos

(Figure 3.3C), establishing that calmodulin inhibition rescues RP deficiency in hematopoietic tissues.

Calmodulin inhibition attenuates p21 protein upon RPS19 knockdown

We next tested the effect of calmodulin inhibitors in a human cell line (A549), in which p53 is not mutated and is induced upon ribosomal protein knockdown 24. A549 cells were infected with lentiviruses expressing RPS19 or control shRNAs and treated for 3 to 6 days with

A-3 or trifluoperazine (TFP). RPS19 knockdown alone increases p21 protein levels (Figure 3.4)

113. In contrast, both A-3 and TFP caused a significant inhibition of p21 protein induction as measured by flow cytometry and Western blot, whereas p53 protein levels remained relatively unaffected by drug treatment (Figure 3.4A-C). The mRNA levels for p53 target genes p21,

GADD45a, and NOXA were also decreased in the presence of calmodulin inhibitors (Figure

3.4D-F, Figure 3.5). Studies with shorter treatments showed that p21 protein levels decrease

69 within 6 hours of treatment (Figure 3.5). These data demonstrate that chemical inhibition of calmodulin inhibits p53 activation and attenuates expression of its targets.

Figure 3.3. A-3 rescues morphological and hemoglobin defects in the rps29 mutant embryo.

(A) Brightfield images of embryos at 24 hpf. Arrowheads mark regions of the head with increased apoptotis in the homozygous mutant embryo. (B) 24hpf wildtype embryos were irradiated at 10 Gy and stained for phospho-H3 at 27 hpf. (C) 40 hpf embryos from an rps29+/- incross were stained with benzidine to measure hemoglobin levels. Location of arrowheads demarcates red blood cells circulating over the yolk of the embryo.

70

Figure 3.4. Calmodulin inhibition decreases p21 upon RPS19 knockdown in A549 cells.

(A-B) p21protein levels, as determined by flow cytometry, in A549 cells with shRNA targeting luciferase or RPS19 and treated with A-3 (A) or trifluoperazine (TFP) (B). (C) p53 and p21 protein levels, as measured by Western blot, in A549 cells with shRNA targeting luciferase or RPS19 and treated with 50uM A-3 or 20uM TFP. (D-F) qPCR measuring expression levels of p21 (D), GADD45 (E), or NOXA (F) in A549 cells with shRNA targeting luciferase or RPS19 and treated with TFP.

Nuclear localization of p53 and p21 is blocked by calmodulin inhibitors

We hypothesized that calmodulin inhibitors affect p53 function by disrupting its binding partners after ribosomal protein knockdown. To interrogate this possibility, we utilized stable isotope labeling by amino acids in cell culture (SILAC) 185 in combination with quantitative mass spectrometry analysis to compare the binding partner profiles of p53 in RPS19- deficient A549 cells in the presence or absence of TFP (Figure 3.6). A number of proteins showed differential binding to p53 upon TFP treatment (Supplementaary Tables 3.1 and 3.2).

Proteins associated with p53 in the absence of TFP included nuclear proteins such as CDK4 and

STAT3, consistent with the nuclear localization of p53 when activated by ribosomal stress. In

71 contrast, these nuclear proteins were absent from the p53 binding partner list in the presence of

TFP (Figure 3.7A). These data led us to investigate whether calmodulin inhibitors could be affecting the sub-cellular localization of p53 upon ribosomal protein knockdown.

Figure 3.5. A-3 attenuates p53 pathway activation upon RPS19 knockdown in A549 cells.

(A) p21 levels, as determined by flow cytometry, in a timecourse of A-3 treatment. (B-D) p21, GADD45a, and NOXA mRNA levels upon A-3 treatment, as measured by qPCR.

72

Figure 3.6. p53 SILAC IP/MS schematic.

“Light” (blue) represents A549 cells with RPS19 knockdown but no drug. “Heavy” (red) represents A549 cells with RPS19 knockdown and 20uM TFP. Adapted from http://www.broadinstitute.org/scientificcommunity/science/ platforms/proteomics/silac

Immunofluorescence staining for p53 was performed on A549 cells with and without

RPS19 knockdown (Figure 3.7B-D). In the absence of calmodulin inhibitors, RPS19 knockdown resulted in a clear increase in nuclear p53 staining. In accordance with the binding partner analysis, treatment with TFP (or A-3) resulted in a dramatic attenuation of nuclear p53 levels.

We found this localization effect to be significant using two different blinded scoring methods

(Figure 3.7B-C).

Our initial observations regarding attenuation of p53 nuclear accumulation in response to

TFP treatment did not distinguish between a block of nuclear import and an increase in nuclear export. To test the latter possibility, we knocked down RPS19 in A549 cells to induce p53 nuclear translocation and treated them with the nuclear export inhibitor, leptomycin B (LMB).

In the absence of TFP, inhibition of nuclear export resulted in p53 residing exclusively in cell

73 D

C

Figure 3.7. Treatment with A-3 or trifluoperazine inhibits p53 nuclear localization upon RPS19 knockdown.

(A) Localization of top 20 proteins differentially bound to p53 in A549 cells with RPS19 knockdown, with and without TFP treatment, as determined by SILAC/LC-MS analysis. In the presence of TFP, nuclear binding partners are absent from this list. A full list of differentially bound proteins is provided in Supplementary Table 2.1. (B) Localization of p53 in cells treated with TFP or A-3 upon RPS19 knockdown was determined by analysis of immunoflourescence images and co-staining with DAPI. Cells in each condition were scored for intense nuclear staining compared to equal levels of nuclear and cytoplasmic staining. Graph shows percentage of cells scored as having nuclear staining in each condition. (C) Alternative method of image scoring, depicting nuclear:cytoplasmic intensity ratios. (D) Left column - DAPI staining; middle column - immunofluorescence staining of p53 with GFP; right column - merged images. All images are 200x.

74 nuclei (Figure 3.8). In contrast, treatment with TFP attenuated nuclear accumulation of p53 as observed in the previous experiments, suggesting that p53 is prevented from entering the nucleus in the presence of TFP. Together, these data suggest that calmodulin inhibitors modulate the activity of p53 by attenuating its nuclear accumulation in response to ribosomal stress.

S19$+$LMB S19$+$LMB no$TFP TFP

p53

Figure 3.8. Trifluoperazine treatment prevents p53 nuclear translocation in response to RPS19 knockdown and is unaffected by nuclear export inhibition.

Immunofluorescence images of p53 staining in A549 cells expressing RPS19-targetng shRNAs and treated with leptomycin B (20nM) and either DMSO or TFP (20uM). All images are 200x.

Calmodulin inhibitors increase red blood cell numbers in a mouse model of DBA

Based on the effects of calmodulin inhibition on p53 activity, we hypothesized that calmodulin inhibitors might rescue the p53-mediated erythroid defect in DBA. We used a mouse model of DBA to test the effect of calmodulin inhibitors on hematopoiesis in a mammalian animal model. Wildtype mice were irradiated and transplanted with bone marrow from mice harboring a knock-in allele that expresses a doxycycline-inducible Rps19 shRNA 100.

Following engraftment of the transplanted cells, mice were administered doxycycline in drinking water and received an IP injection of 5 mg/kg TFP every other day (Figure 3.9A). After two weeks, blood was analyzed for red blood cell (RBC) counts and hemoglobin levels, and bone

75 marrow was collected for qPCR analysis. In vehicle-treated mice, knockdown of RPS19 causes anemia and an induction of p53 target genes in hematopoietic tissues. Bone marrow collected from TFP-treated animals showed reduced p21 RNA expression (Figure 3.9B). Additionally, both red blood cell counts and hemoglobin levels are partially rescued by TFP treatment (Figure

3.9C-D). Notably, treatment with calmodulin inhibitors in this model achieves therapeutic effects similar to those seen with dietary leucine treatment 74.

Figure 3.9. Calmodulin inhibition rescues hematopoietic phenotypes in murine models of DBA.

(A) Schematic of mouse transplantation and drug treatment. Unfractionated bone marrow from inducible Rps19 shRNA donor mice was transplanted into irradiated wildtype recipients. After engraftment, hairpin expression was induced with doxycycline and mice were treated with TFP or vehicle for two weeks. (B) RNA was isolated from bone marrow collected from recipient mice after two weeks of treatment. (C-D) Peripheral blood samples from recipient mice and analyzed using a Hemavet hematology system for red blood cell counts (C) and hemoglobin levels (D).

76 Human CD34+ model of DBA responds to calmodulin inhibition

To test the effects of calmodulin inhibition on human hematopoietic cells in the context of ribosomal protein deficiency, primary human CD34+ hematopoietic stem and progenitor cells

(HSPCs) were treated with calmodulin inhibitors after shRNA-mediated RPS19 knockdown.

CD34+ cells were expanded in vitro and infected with an RPS19 hairpin, and were subsequently maintained with or without TFP in liquid culture with growth factors promoting myeloid and erythroid differentiation (Figure 3.10A).

A

B C

Figure 3.10. Calmodulin inhibition rescues hematopoietic phenotypes in human cell models of DBA.

(A) Schematic of CD34+ HSPC differentiation Cells were expanded for 4 days prior to infection with luciferase- or RPS19-targeting shRNAs. Infected cells were selected with puromycin beginning on day 5, and drug was added to the media beginning on day 7. Cells were processed for RNA/expression analysis or flow cytometry on day 10. (B) p21 RNA levels, as measured by qPCR, in CD34+ cells cultured for 10 days in medium supportive of erythroid differentiation. (C) Percentage of CD71+ erythroid cells, as measured by flow cytometry, in CD34+ cells cultured for 10 days in medium supportive of both erythroid and myeloid differentiation, and treated with TFP.

77 As in A549 cells, RPS19 knockdown induced p21 mRNA levels in CD34+ HSPCs. TFP treatment blocked this induction in a dose-dependent manner (Figure 3.10B). Consistent with previous reports, RPS19 knockdown also decreased the percentage of erythroid precursors upon liquid culture differentiation as measured by CD71 expression 113. When cultured in the presence of TFP, CD71+ cell percentages approached those observed in cultures without RPS19 knockdown, suggesting that calmodulin inhibitors are effective in reversing the ribosomal dysfunction-mediated erythroid differentiation block (Figure 3.10C). Our data demonstrate that calmodulin can rescue erythroid differentiation defects caused by ribosomal protein knockdown in human CD34+ HSPCs.

Inhibition of calmodulin-dependent CHK2 phenocopies calmodulin inhibition and is sufficient for rescue

Calmodulin serves as a cofactor for many calmodulin-dependent enzymes, and we hypothesized that calmodulin inhibitors might be affecting p53 activity by indirectly inhibiting calmodulin-dependent enzyme activity. Using the human CD34+ HSPC differentiation assay described above, we performed a small screen of calmodulin-dependent enzymes and found that a chemical inhibitor of the calmodulin-dependent kinase CHK2, BML-277 186, rescued CD71+ cell percentages in a manner similar to TFP (Figure 3.11A-B). Similar to calmodulin inhibitors

A-3 and TFP, BML-277 prevented nuclear accumulation of p53 and decreased p21 protein levels in response to RPS19 knockdown in A549 cells (Figure 3.11C-F). In zebrafish embryos, a chk2 morpholino rescued hemoglobin levels in the rps29-/- embryo (Figure 3.11G). These data demonstrate that CHK2 inhibition phenocopies calmodulin inhibition in our assays, and further

78 suggest that blocking CHK2 is the mechanism by which calmodulin inhibition disrupts p53 nuclear localization and attenuates its activity (Figure 3.12).

A B C

E D

F Luc RPS19 G DMSO TFP BML DMSO TFP BML p53 p21

actin

Figure 3.11. Inhibition of calmodulin-dependent CHK2 rescues aspects of RPS19 deficiency.

(A) Various inhibitors of calmodulin-dependent kinases were used in the human CD34+ HSPC differentiation assay described in Figure 2.10. Control cells expressed a luciferase-targeting shRNA; RPS19 was knocked down in all other drug testing conditions. (B) Percentage of CD71+ erythroid cells, as measured by flow cytometry, in CD34+ cells cultured for 10 days in medium supportive of both erythroid and myeloid differentiation, and treated with BML-277. (C) Percentage of cells with nuclear staining of p53 upon treatment with BML-277. (D) Representative immunofluoresence images (quantified in (C)) showing attenuated nuclear translocation of p53 (green); DAPI (nuclei; blue). (E) p21 protein levels, as determined by intracellular flow cytometry, in A549 cells with RPS19 knockdown. (F) p53 and p21 protein levels, as measured by Western blot, in A549 cells with shRNA targeting luciferase or RPS19 and treated with 20uM TFP or 50uM BML-277. (G) Chk2 morpholino was injected into rps29- /- mutant embryos. Benzidine staining for hemoglobin levels was performed at 40 hours post fertilization.

79

Figure 3.12. Model for the mechanism of p53 activity modulation by calmodulin inhibitors.

Ribosomal protein deficiency leads to activation of p53 and its translocation to the nucleus. Calmodulin inhibitors block calmodulin-dependent CHK2, a kinase involved in promoting p53 nuclear translocation. Inhibition of either calmodulin or CHK2 attenuates nuclear accumulation of p53, which diminishes its ability to activate its transcriptional targets.

Discussion

Ribosomal protein mutations in patients with DBA cause hematopoietic specific defects that have been modeled in a variety of organisms by knockdown or mutation of ribosomal proteins. We performed a chemical screen for rescue of the rps29 zebrafish mutant phenotype and found two structurally-related calmodulin inhibitors, A-3 and W-7, that could rescue hematopoietic, endothelial, and apoptotic aspects of the phenotype. Our results in zebrafish, mice, human cell lines, and primary human blood cells show that calmodulin inhibition blocks p53- mediated induction of p21 by preventing p53 nuclear accumulation upon RPS19 knockdown.

Inhibition of CHK2, a calmodulin-dependent kinase, also rescues zebrafish embryonic phenotypes and attenuates p53 transcription factor activity in multiple models. The hematopoietic phenotype (including RBC counts and hemoglobin levels) in a mouse model of

80 DBA is improved by calmodulin inhibition. Both calmodulin and CHK2 inhibitors function in primary human hematopoietic cells as well, as treatment of RPS19-deficient CD34+ HSPCs results in attenuated p21 induction and rescues erythroid differentiation as shown by an increase in CD71+ erythroid precursors. These data suggest that calmodulin inhibition can rescue the majority of phenotypes associated with RP deficiency, particularly those associated with the activation of p53.

These data propose a role for calmodulin in the nuclear localization of p53. Calmodulin has been shown to play a role in nuclear transport 187. Classically, proteins are transported into the nucleus by a GTP-dependent cycle requiring Ran and importin. There are also importin- independent mechanisms of regulating nuclear import, including calmodulin-dependent mechanisms 188, as well as evidence for calmodulin binding to block phosphorylation and subsequent nuclear localization 189. A calcium flux is required for p53 phosphorylation and activation 190; however, a connection between calmodulin-dependent p53 phosphorylation and nuclear localization has not previously been described.

Our data suggest that calmodulin’s effect on p53 is through the activity of the calmodulin-dependent kinase, CHK2. CHK2, as well as CHK1, are both known to phosphorylate p53 and lead to an increase of p53 transcriptional activity 191. It is possible that our inhibitors work directly by preventing p53 phosphorylation. Another possibility is that checkpoint kinase activity under conditions of ribosomal stress indirectly activate p53, and our drugs block this downstream activation. Since both CHK1 and CHK2 are calmodulin dependent, both are likely inhibited by our calmodulin inhibitors. BML-277 is a highly specific inhibitor for

CHK2 186. However, we cannot rule out the possibility that CHK1 is also affected at the doses we tested. Experiments in a Chk2 knockout mouse have demonstrated its requirement for

81 transcription of p53 targets upon irradiation 192. Although our data support a role for CHK2 in this regulation, it does not preclude a role for CHK1. Nonetheless, to our knowledge, this is the first time that a checkpoint kinase has been shown to be important for p53 nuclear localization.

A key finding of this work is in the characterization of a novel mechanism of p53 modulation by inhibiting its nuclear localization. In the case of ribosomal protein deficiency, p53 is believed to be activated as a consequence of an increase in the pool of free, non-ribosome bound ribosomal proteins. Some of these free ribosomal proteins can bind to and inhibit MDM2 or directly bind the p53 UTR for increased translation 24, 125. Our results suggest that the inhibition of p53 localization is not specific to these mechanisms of p53 activation, as calmodulin inhibitors can also prevent irradiation-induced phenotypes. It will be interesting to determine whether calmodulin inhibition affects localization of p53 in all cases of activation, or only in response to specific p53-activating cellular stresses. Although calmodulin inhibitors can inhibit p53 nuclear localization, they do not completely block all p53 from entering the nucleus.

As a result, background levels of p53 are present in the nucleus, albeit at lower levels than without drug treatment. In addition, p53 has cytoplasmic functions that are likely unaffected 193.

These calmodulin inhibitors provide a unique way of mediating nuclear p53 activity without completely blocking its function.

Current treatments for DBA patients are not specific to diseases caused by ribosomal protein mutations nor do they specifically target p53 activation, and they are associated with serious side effects. Some efficacy was observed in a patient treated with leucine, currently in clinical trials, which is believed to work by increasing overall translation in the cell 75.We have shown that TFP, a calmodulin inhibitor, has effects similar to leucine in a mouse model of DBA

74. Our data identify calmodulin inhibition as a potential therapeutic approach for attenuating

82 aberrant p53 activity. In primary human blood cells, inhibiting a component of the p53 activation pathway can attenuate some of the defects resulting from ribosomal protein deficiency, such as p21 induction and a block in erythroid differentiation. In this way, calmodulin inhibition provides a method for fine-tuning some consequences of p53 activity.

There is justifiable concern with targeting proteins as multifaceted as calmodulin and, more importantly, p53. Schizophrenia patients treated with TFP and other phenothiazines have experienced side effects associated with other antipsychotics that may preclude its use as a therapy for DBA 194, particularly in children. No inhibition of p53 function has been described in patients receiving TFP, and there is no documented increase in tumor incidence. There is a theoretical risk to long term attenuation of p53 nuclear activity, but the risks of calmodulin inhibition are not yet characterized in the context of ribosomal dysfunction.

Mutations in a number of ribosomal protein genes have been identified in more than 50% of DBA patients 172. The compounds tested here are effective when different ribosomal proteins are deficient, including rps29 in the zebrafish and RPS19 in human cell lines and primary human

HSPCs, suggesting that they may be effective in DBA patients with different ribosomal protein mutations. If calmodulin inhibition is a successful means of p53 inhibition in DBA, these drugs may be beneficial to patients with other diseases as well. DBA is one of a group of diseases called ribosomopathies, where patients exhibit ribosome dysfunction caused by a mutation in or loss of a ribosomal protein or related gene. Ribosomopathies are nearly all thought to involve aberrant p53 activity 141, so inhibiting p53 function could be therapeutic in most all of these diseases. For example, 5q- MDS patients exhibit haploinsufficient loss of RPS14, and this is thought to cause aberrant p53 activity and hematopoietic defects by a mechanism similar to that seen in DBA patients 32, 113. Patients with Shwachman-Bodian-Diamond syndrome, who have a

83 mutation in a ribosome biogenesis gene, also exhibit hematopoietic symptoms 46. Dyskeratosis congenita and Treacher Collins syndrome patients have mutations in genes involved in rRNA modification which also induce p53 activation 101, 112. Other bone marrow failure syndromes are also thought to result from aberrant p53 activity, including Fanconi’s anemia 195. A therapy developed to target the p53 pathway could be effective in any ribosomopathy, as well as in other diseases associated with aberrant p53 activation.

Ribosomopathies, including DBA, are associated with increased predisposition to various types of cancers, though the underlying mechanisms are poorly understood. It has been shown in zebrafish models that cancers arising in RP-deficient settings down-regulate or inactivate p53 to support a hyperproliferative state 138. Together with the observation that p53-deficient cells are more sensitive to inhibition of irradiation-induced checkpoints mediated by CHK proteins 196, our data suggests that CHK2 inhibition may actually be selectively toxic to cells that have decreased p53 in patients with ribosomal dysfunction. CHK inhibitors are currently being tested in clinical trials for their efficacy in treating various blood and solid tumors 197. Further investigation is warranted to gain a better understanding of the role of CHK2 in the modulation of p53 activity in normal and disease contexts. Here, we have shown how an in vivo screen in the zebrafish embryo identified compounds that mediate the p53 pathway under ribosomal stress in human cells. Further understanding of how calmodulin inhibition affects the p53 pathway may prove it a useful therapeutic approach for DBA and other diseases characterized by aberrant p53 activity.

84 !: Results III

Ribosomal gene haploinsufficiency is a common feature of human cancers

Attributions

I performed the in vitro functional work in A549 cells including all knockdown experiments and qPCR and Western blot analysis; Bernd Boidol assisted with the protein work. I also coordinated all aspects of the ALL patient sample analysis and purified expanded tumor cells from mouse spleens and bone marrow.

All other work, including analysis of copy number microarrays, whole-exome sequencing datasets, and RNA sequencing datasets, was performed by Björn Nilsson and Magnus Jöud.

Mouse xenografts and tissue isolation was performed by Amanda Christie and Alexandra

Christodoulou in David Weinstock’s laboratory. rRNA processing analysis was performed by

Steve Ellis.

Note: The work presented here is part of a larger manuscript currently in preparation, with selections presented to provide context for my contributions.

86 Summary

Haploinsufficiency for ribosomal protein genes (RPGs) has been associated with both ribosomopathies (such as Diamond Blackfan anemia and 5q- syndrome) and specific malignancies. To determine the prevalence of RPG lesions in cancer, we performed a large-scale analysis of copy-number microarray (n=9,701), whole-exome sequencing (n=4,675), and RNA- sequencing (n=4,919) data from 28 tumor types. We found that hemizygous RPG deletions were strikingly common in human cancers, detected in 43% of specimens and in all tumor types analyzed. shRNA-mediated knockdown of a panel of RPGs showed that RPG loss broadly results in p53 pathway activation; together with the observation that RPG deletions occurred more commonly in tumors that also harbored TP53 mutations / deletions, this suggests that RPG loss is associated with selective pressure against wild-type TP53. We also observed rRNA processing defects in patient ALL samples with RPS6 deletions, suggesting that ribosome biogenesis is broadly affected in RP-deficient cancers. Lastly, we propose that RP deficiency may provide a therapeutic opportunity in the context of human cancer.

87 Introduction

The human ribosome consists of an rRNA scaffold and approximately 80 proteins, divided into a large (60S) and a small (40S) subunit 198. Mutations in ribosomal protein genes

(RPGs), and other genes required for ribosome biogenesis, cause a set of diseases collectively known as ribosomopathies 141, 199. These diseases include Diamond-Blackfan anemia (DBA), which is caused by congenital, heterozygous, inactivating mutations or deletions in RPS19 and other RPGs 9, 10, 16, 170, 200. In the 5q- syndrome, a subtype of myelodysplastic syndrome, somatic hemizygous deletions on chromosome 5 lead to haploinsufficiency for RPS14 32.

Increased translation is a central feature of rapidly proliferating tumor cells, with amplification of MYC and activation of AKT. While inherited ribosomopathies are primarily characterized by bone marrow failure, growth defects, and developmental abnormalities, several lines of recent evidence have also linked RPG haploinsufficiency to cancer development.

Patients with inherited ribosomopathies are known to have a higher risk of developing malignancies 95, 141, 199. A similar propensity to develop acute leukemia is seen in patients with

5q- syndrome. Furthermore, recurrent somatic mutations in RPGs have been associated with specific tumor types, including heterozygous mutations in RPL5, RPL10, and RPL22 136, 137 that lead to T-cell acute lymphoblastic leukemia, and microsatellite-unstable endometrial and gastric cancer as a result of heterozygous mutations in RPL22 201, 202. Lastly, eleven RPGs have been identified as tumor suppressor genes in zebrafish, where hemizygous inactivation of these genes cause malignant peripheral nerve sheath tumors with high penetrance 91, 136-138, 203.

Thus, haploinsufficiency for RPGs could contribute to cancer development, yet given the central role of the ribosome in cellular metabolism, RPG deletion may be selected against in rapidly dividing cell populations. To better understand the profile of RPG deletions in cancer, we

88 undertook a large-scale analysis of RPG copy-number with respect to other cellular genes, and in conjunction with whole-exome sequencing data to understand differential patterns of RPG stability in the context of concomitant p53 genetic aberrations. We found that hemizygous loss of

RPGs was strikingly common in human cancers, and that this loss frequently co-occurred with loss or mutation of TP53. We also found that, in a TP53 wild-type background, RPG silencing broadly induces p53 pathway activity, which likely explains the co-occurrence of RPG and TP53 loss. Lastly, we identified rRNA processing defects in primary RPS6-deficient ALL patient samples, demonstrating additional consequences of RPG loss on ribosome biogenesis in human cancer.

Materials and Methods

Genomics data sets and analysis

We retrieved DNA copy-number profiles of 7,225 primary tumor samples belonging to

24 tumor types from TCGA (Affymetrix and Agilent microarrays). We also retrieved matched whole-exome sequencing data (somatic mutation calls; available for 4,675 samples) and RNA- sequencing data (available for 4,919 samples). As a replication cohort, we used DNA copy- number profiles of 2,476 primary tumor samples belonging to 13 tumor types from the

Tumorscape compendium (Affymetrix microarrays). To delineate copy-number changes, we used the program Ultrasome204. To call hemizygous and homozygous deletions, we used copy- number thresholds of 1.6 (log2 ratio -0.3) and 0.8 (log2 ratio -1.3), respectively. Gene-wise copy-numbers were calculated by integrating the copy-number signal from gene start to gene end and dividing by gene length. To visualize copy-number data, we used the program Integrative

Genomics Viewer205.

89 Lentiviral vectors and infection

Lentiviral shRNAs in the pLKO.1 or pLKO_TRC005 vector were obtained from the

Broad Institute of Harvard and MIT. Sequences targeted by each shRNA are listed in

Supplemental Table 4.1. Lentivirus was produced in 293TL cells as described previously

(CITATION). A549 cells were infected with lentivirus one day after plating in the presence of 8

µg/mL polybrene (Sigma-Aldrich) and selected 24 hours later with 2 µg/mL puromycin (Sigma-

Aldrich) for at least 48 hours before sample processing.

Western blot

Western blots were performed using 25-50µg of protein and primary antibodies against p53 (DO-1; Santa Cruz Biotechnology) at 1:500 dilution and β-actin (C4; Santa Cruz

Biotechnology) at 1:3000 dilution. HRP-conjugated anti-mouse secondary antibody (GE

Healthcare) was used at 1:10000 dilution. Immunoreactive proteins were visualized using

SuperSignal® West Pico Chemiluminescence Substrate (Thermo Scientific).

Quantitative real-time PCR

RNA was purified using TRIzol (Invitrogen). First-strand cDNA was generated using

100-200ng of total RNA and oligo dT primers with the Superscript III reverse transcription kit

(Invitrogen). Quantitative RT-PCR was performed using TaqMan® Gene Expression Master Mix and gene-specific expression assays (Applied Biosystems) or SYBR® Green PCR Master Mix

(Applied Biosystems) on an ABI Prism 7900HT Sequence Detection System (Applied

Biosystems). For TaqMan® assays, β-actin (ABI part #401846) was used as an internal control.

For SYBR® Green assays, the following primers were used: P21 forward 5’-

90 GCTCTGCTGCAGGGGACAGC-3’; P21 reverse 5’-GCCGCCGTTTTCGACCCTGA-3’;

ACTB forward: 5’-AGCGAGCATCCCCCAAAGTT-3’; and ACTB reverse: 5’-

GGGCACGAAGGCTCATCATT-3’.

Primary Tumor Xenografts

Banked human ALL tumor cells, which had been previously characterized by copy number variation microarray, were generously provided by Andrew Kung and Scott Armstrong.

Tumor cells were transplanted into immunodeficient mice (NOD scid gamma) and expanded in vivo for 6 months. When the injected animals appeared moribund, the mice were sacrificed and tumor cells were harvested from the spleens and bone marrow. Harvested cells were stained with anti-human CD45 (Miltenyi Biotec) and sorted by FACS on a BD Aria to purify the tumor cells, from which RNA was isolated for rRNA maturation analysis.

Northern blot analysis of rRNA species

Total RNA was isolated using RNeasy kits (Qiagen). RNA was fractionated on a 1.5% formaldehyde-agarose gel and ethidium bromide stained gels were visualized. For Northern analysis, gel-fractionated RNA was transferred to zeta-probe membranes (Bio-Rad). An oligonucleotide probe 5′-CCTCGCCCTCCGGGCTCCGTTAATGATC-3′ (complementary to sequences 5520-5547 spanning the boundary between 18S rRNA and ITS1) was labeled with 32P using T4 polynucleotide kinase. The probe was hybridized overnight with membrane-bound

RNA at 37° C in ULTRAHyb-Oligonucleotide hybridization buffer (Ambion). Membranes were washed at 37° C with 6XSSC and subjected to phosphorimage analysis.

91 Results

Ribosomal protein gene deletions are highly prevalent in numerous tumor types

Based on previous studies that have linked RPG haploinsufficiency with oncogenesis, we profiled the occurrence of RPG lesions in a broad spectrum of cancers. Using data from the The

Cancer Genome Atlas, (TCGA)206, we first analyzed DNA copy-number profiles of 7,225 cancer specimens belonging to 24 tumor types. Remarkably, we detected somatic deletions affecting

RPGs in approximately 43% of specimens and in all tumor types analyzed (Figure 4.1A and

Table 4.1). Interestingly, deletions affecting RPGs were segmental deletions covering multiple genes, and the most frequently deleted RPGs are located close to (and are co-deleted with) key tumor suppressors. For example, RPL26 (17p13) is co-deleted with TP53; RPS6 (9p21) with

CDKN2A; and RPL21 (13p12) with RB1.Almost all (>99%) detected RPG deletions had copy- numbers between 1 and 2, which is consistent with hemizygous loss. Validation in an independent cohort of 2,476 copy-number profiles from the Tumorscape compendium 207 identified the same RPGs as frequently deleted, and all RPG deletions had amplitudes consistent with hemizygous loss (Figure 4.1B) 136, 137, 201, 202. Taken together, these results show that RPG haploinsufficiency – primarily on the basis of hemizygous deletions – occurs in a high percentage of many different tumor types.

92 showed

ency),whereas amplitudes compatiblewith homozygous deletion were exceptionally of of RPGs in TCGA and Tumorscape compendium TCGA and datasets. RPGs of Tumorscape of in

mber analysis ble with heterozygous loss (haploinsuffici loss heterozygous with ble . nu Copy 1 . 4 Analysis in an independent cohort of 2,476 copy number profiles from the Tumorscape compendium. Tumorscape from the profiles number 2,476copy of cohort independent an in Analysis

(B) Analysis of 7,275 DNA copy number profiles from The Cancer Genome Atlas (TCGA) compendium. Deletions affecting RPGs affecting Deletions (TCGA)compendium. GenomeAtlas fromThe Cancer profiles number DNA copy of 7,275 Analysis Figure (A) amplitudes compati rare. " " Figure 1 a b A B

93 Table 4.1. List of the most frequently deleted RPGs in The Cancer Genome Atlas dataset.

The RPGs listed were also found to be frequently deleted in additional datasets, including the Cancer Cell Line Encyclopedia (CCLE) and Global Copy-number Map (GCM).

TCGA CCLE GCM Gene Chromosome Start (bp) End (bp) n % n % n % RPL26 17 8,280,833 8,286,565 1752 24.2 340 32.6 310 12.5 RPL13 16 89,627,064 89,633,237 1430 19.8 204 19.6 222 9.0 RPS6 9 19,376,253 19,380,235 1411 19.5 373 35.8 269 10.9 RPL17 18 47,014,850 47,018,935 1335 18.5 393 37.7 173 7.0 RPL29 3 52,027,643 52,029,958 1283 17.8 340 32.6 229 9.2 RPL3 22 39,708,886 39,715,670 1205 16.7 263 25.2 152 6.1 RPL14 3 40,498,782 40,503,863 1133 15.7 292 28.0 209 8.4 RPSA 3 39,448,203 39,454,032 1129 15.6 284 27.2 202 8.2 RPL21 13 27,825,691 27,830,702 1121 15.5 328 31.4 224 9.0 RPS15 19 1,438,362 1,440,492 1108 15.3 264 25.3 170 6.9

RPG haploinsufficiency is selected against in tumors harboring wild-type TP53

Activation of the p53 pathway, which occurs as a result of defective ribosome biogenesis, has been shown to arise as a result of RPG haploinsufficiency, including as a result of RPS14 and

RPS19 loss. The promotion of p53 activation would ultimately be detrimental during oncogenesis, and as such would reduce the likelihood of those cell populations continuing to expand. Due to this potential for negative selection, we hypothesized that RPG deletions would be less prevalent in tumor samples that retained TP53 genetic integrity as compared to those with

TP53 point mutations.

To evaluate this, we used 4,675 TCGA samples where both copy-number microarray and whole-exome sequence data were available. This allowed us to correlate the presence and absence of TP53 genetic aberrations with the copy-number distributions of both RPGs and other

94 cellular genes. We found that TP53-wildtype tumors (n=2,210) contained a median of 1 RPG deletion (IQR: 0-4), which was significantly fewer compared to a median of 7 deletions (IQR: 2-

13) in TP53-mutant tumors (n=2,362) (Figure 4.2A). Strikingly, TP53-wildtype tumors showed fewer deletions with copy-numbers between 1 and 2 in RPGs than in other genes (p<0.001), whereas no difference was observed with TP53-mutant tumors (Figure 4.2B). These observations support the hypothesis that RPG haploinsufficiency is incompatible with the presence of an intact p53 pathway, which in turn leads to the negative selection of RPG hemizygous cell populations in a TP53-dependent manner.

95 Figure 2 Figure 2

a b A" B" No or silent TP53 mutation a b

−51 p < 2.2e−16 10 −1 × P<1.6 (cdf) 10

log −2

Figure 2 All genes RPGs −3 0.5 1 1.5 2 Gene copy number a b TP53 mutation or deletion c d

n.s. Deleted ribosomal genes (n) −1 (cdf) 0 5 10 15 20 25 30 35 10

No or silent Non−silent TP53 log −2 TP53 mutation mutation or deletion (n=2210) (n=2362) All genes RPGs −3 c d 0.5 1 1.5 2 Gene copy number

Figure 4.2. Haploinsufficiencyc for RPGs is subjectd to TP53-dependent negative selection.

(A) Cases without TP53 mutations showed significantly fewer RPG deletions than cases with mutated TP53. (B) Cumulative DNA copy-number distributions for RPGs (solid blue) vs all genes (dashed black) for the 4,675 cases from the TCGA having both copy-number microarray and whole-exome sequence data. Top panel: cases with no detectable TP53 mutation/deletion show a right-shifted copy-number distribution in RPGs compared to other genes in the haploinsufficiency interval (copy-number 1.0 to 2.0). Bottom panel: in cases with TP53 mutations/deletions, RPGs showed the same copy number distributions as other genes.

Ribosomal protein transcriptional knockdown activates the p53 pathway

To determine whether RPG haploinsufficiency alone is sufficient for p53 activation, we

mimiced the effects of heterozygous RPG deletion through shRNA-mediated knockdown of

eight ribosomal proteins that are components of the large and small subunits and were found to

be commonly deleted in the primary tumor samples (Figure 4.3). In the lung adenocarcinoma

96 cell line A549, which is TP53-wildtype, knockdown of any one of the eight ribosomal proteins by shRNA resulted in elevated p53 protein levels (Figure 4.3A). We also detected a

p53 corresponding increase in p21mRNAtranscript levels withinL26 β-Actin the same cells, which is indicative

p53 of p53 nuclear translocation and transcriptional activityS6 (Figureβ-Actin 4.3B).

p53

L13 β-Actin

A p53

L21 β-Actin

p53 p53 L29 L26 β-Actin L26 β-Actin

p53 p53 p53 L14 S6 S6 β-Actin β-Actin β-Actin

p53 p53 SA L13 β-Actin L13 β-Actin

p53 p53 p53

S12 β-Actin L21 β-Actin L21 β-Actin

p53 p53 L29 β-Actin L29 β-Actin RPL26 RPS6 RPL13 RPL21

B p53 p53 160 90 40 9 L14 β-Actin L14 β-Actin 8 140 80 35 70 7

120 p53 30 p53 60 6

SA 25

100 SA β-Actin 50 β-Actin 5 80 20 40 4

15 60 p53 30 p53 3 40 10

S12 20 2 β-Actin S12 β-Actin

eaie p21(expression Relative( 20 10 5 1 0 0 0 0 shLuc shRNA/shRNA/ shLuc shRNA/shRNA/ shLuc shRNA/shRNA/ shLuc shRNA/shRNA/ #1 #3 #2 #3 #2 #3 #2 #3

RPL29 RPL14 RPSA RPS12

12 60 45 120 40 10 50 100 35 8 40 30 80 25 6 30 60 20 4 20 15 40 10 2 10 20 eaie p21(expression Relative( 5 0 0 0 0 shLuc shRNA/shRNA/ shLuc shRNA/shRNA/ shLuc shRNA/ shRNA/ shLuc shRNA/ shRNA/ #1 #3 #1 #3 #2 #3 #2 #3 Figure 4.3. p53 pathway induction as a result of RPG loss.

We used shRNA-mediated knockdown of commonly deleted RPGs in A549 cells (TP53 wild-type). (A) p53 protein levels as analyzed by Western blot (β-actin as loading control), and (B) p21 transcript levels as analyzed by qPCR (β-actin as normalization control) after 4 or 6 days of shRNA expression. Error bars indicate standard deviation between technical replicates.

97 Cellular dependency on ribosomal proteins limits the extent of homozygous RPG deletions in cancer

The selection bias in TP53 wild-type tumors to maintain RPG integrity may act as a blockade to prevent the homozygous loss of proximal genes. Examining the copy-number data from TCGA and Tumorscape, we detected numerous events with hemizygous deletions that included both CDKN2A and RPS6 loci at 9p21. Deletions involving the 9p21 region are common in acute leukemia and other tumor types, and focus at the tumor suppressor gene CDKN2A. In contrast, events with homozygous deletions that included CDKN2A reached all the way up to

RPS6 but never included this gene (Figure 4.4A). This finding identifies RPS6 as a gene that shapes cancer genomes by constraining the extent of homozygous deletions. This finding agrees with the generally held view that complete loss or inactivation of an RPG would prohibit protein synthesis, and conflict with the maintenance of cellular viability.

Despite the low frequency with which homozygous RPG deletions were detected, we did identify recurrent deep deletions in RPL13 and RPL25 in acute myeloid leukemia and ovarian cancer (data not shown), suggesting that total loss of certain RPGs may actually be compatible with continued protein synthesis and cell survival. Such RPGs have not previously been identified.

RPS6 deletion causes rRNA maturation defects in human ALL

Previous in vitro work has shown that shRNA-mediated knockdown of RPS6 results in a defined defect in the processing of pre-ribosomal RNA species into mature rRNA 19. To see if these defects are observed in primary human cancer samples with RP deletions, we obtained pediatric acute lymphoblastic leukemia (ALL) samples that had been characterized previously by

98 copy-number microarray with known CDKN2A and RPS6 copy number status and interrogated rRNA maturation in tumors with and without RPS6 haploinsufficiency (Figure 4.4B). Viable tumor cells were xenografted into NOD scid gamma (NSG) immunodeficient mice, expanded in vivo, and purified from murine bone marrow by sorting human CD45-expressing cells by flow cytometry. RNA from purified tumor cells was then assessed by Northern blot. RPS6 haploinsufficient cells exhibited a reduction in 41S species and an accumulation of 30S species, indicating defective pre-rRNA cleavage at sites 1, E, and 3 (Figure 4.4C-D). These rRNA maturation defects were similar to those seen previously in shRNA-targeted RPS6-deficient cells

19, and demonstrate a functional consequence of RPS6 loss on ribosomal RNA biogenesis in human ALL. Because defined rRNA maturation defects have been observed in vitro with knockdown of each 40S ribosomal protein gene 19, these types of defects may also be common features of RP-deleted human cancers.

99 Figure 3

A a chr9

p24.2 p23 p22.3 p22.1 p21.3 p21.2 p21.1 p13.3 p13.1 p11.2 q11 q12 q13 q21.12 q21.2 q21.32 q22.1 q22.31 q22.33 q31.1 q31.2 q32 q33.1 q33.2 q33.3 q34.11 q34.2

10000 1000 100

10 RPS6 CDKN2A Cases w. deletion 0 16 17 18 19 20 21 22 23 24 25 chr9 position (million bp)

10000 1000 100

10 RPS6

Cases w. deletion 0 18.7 18.8 18.9 19 19.1 19.2 19.3 19.4 19.5 19.6 19.7 ADAMTSL1→ ←SCARNA8chr9 position (millionDENND4C bp) → ←SLC24A2 ←FAM154A ←PLIN2 ←RPS6 RRAGA→ ←AK094196 TRNA_Met→ ←HAUS6 ←DQ572382 ACER2→

CDKN2A: -/- -/- RPS6: +/+ +/-

b CDKN2A: -/- -/- c d RPS6: +/+ +/- 1 45S 45S B C 41S 45S 5’RPS6$ETS del/kd ITS10.8 ITS2 3’$ETS 45S CDKN2A: $/$ 41S $/$ 30S 30S 41S 45S RPS6 del/kd RPS6: +/+ +/$ A0 1 0.6 2 30S 30S 41S {A0+1} 21S {2} 0.4 21S 21S RPS6 del/kd 18S-E 30S 41S RPS6 del/kd 18S-E 21S 45S 18S 0.2 18S-E {2} All genes {A0+1} Cumulative distribution 18S 18S-E Genes in set 41S 0 21S-4 -2 0 2 4 {E} E Gene score (t-statistic) 30S 18S)E Normal%Maturation {3} 3 18S

D 5’$ETS ITS1 ITS2 3’$ETS 21S 45S A0 1 2

18S$E RPS6%KD 30S

Figure 4.4. RPS6 is a barrier for homozygous deletions and is required for proper rRNA maturation in human ALL.

(A) Schematic view of human , with enlarged views of the 9p16-9p25 (top) and 9p19 (bottom) regions. Deletions detected in TCGA and Tumorscape samples are indicated in light gray (hemizygous) and dark gray (homozygous); number of cases with each type of deletion are on the vertical axis on a log scale. (B) Northern blot analysis of rRNA from ALL patient samples with the indicated genotypes. Schematic of normal (C) and RPS6-/- perturbed (D) rRNA maturation as observed in (B) is shown. [Adapted from O’Donohue et al., 2010]

100 Discussion

Taken together, our results show that RPG haploinsufficiency should no longer be regarded as restricted to rare disease entities, but as a highly recurrent phenomenon in common cancers. This finding raises several intriguing questions.

The finding that RPG haploinsufficiency is common is surprising. Numerous studies aimed at understanding the molecular mechanisms of ribosomopathies (particularly DBA and the

5q- syndrome) have identified p53 as a central mediator of the clinical features of these diseases

141, 199. Impaired ribosome biogenesis leads to increased concentrations of free ribosomal proteins, which bind MDM2 and inhibit its ability to target p53 for proteasomal degradation 24, 123, 208. In vitro and in animal studies have shown that the phenotypic effects of RPG haploinsufficiency can be alleviated by genetic or pharmacological inhibition of p53 24, 98, 100, 107, 108, 113. Although previous studies have focused on certain RPGs (primarily RPS6, RPS14 and RPS19), our findings show that RPG haploinsufficiency activates p53 broadly, which likely leads to negative selection in evolving cancer cells.

The molecular consequences of RPG haploinsufficiency in the context of cancer await further exploration. Our results support that, while RPG haploinsufficiency activates p53 broadly,

TP53 mutations allow cells to escape negative selection and make RPG deletions permissible.

Nevertheless, the oncogenic effects of RPG haploinsufficiency are poorly understood. While our data cannot discern RPG deletion as a driver or passenger event in oncogenesis, it is plausible that RPG haploinsufficiency activates additional oncogenic pathways or enhances the effects of other lesions. For example, a link to the Myc pathway has been reported 124, and the RPL26- mRNA is required for optimal translation of the TP53-mRNA by binding a dsRNA region containing complementary sequences of the 5’ and 3’ UTR9, 209. Furthermore, abnormal

101 ribosomal protein stoichiometry leads to inefficient ribosome assembly and increased concentrations of free ribosomal proteins, which may have regulatory effects similar to how

RPL5 and RPL11 modulate MDM2 activity 116, 118, 123.

The fact that cells can sustain damage to a component as critical to cellular function as the ribosome and still proliferate with a clonal advantage is noteworthy. This observation – along with the links to ribosomopathies and human tumors – suggests that RPG haploinsufficiency may be associated with additional lesions that allow cells to overcome dysregulated ribosome biogenesis and support a hyperproliferative state in spite of it. While our observation that RPG loss often co-occurs with loss of tumor suppressor genes provides some explanation for escape from cell cycle arrest or apoptotic pathways, it does not offer any insight into how the translational demands of rapidly proliferating cancer cells are met without properly functioning ribosomes. Differential translation of specific transcripts has been observed in cells with RPG haploinsufficiency 139, so it is possible that oncogene and/or tumor suppressor translation is altered in malignancies to compensate for reduced ribosome biosynthesis.

Lastly, independently of other molecular effects – and independently of their status as driver or passenger events – the relative absence of homozygous RP gene deletions in human cancers suggests that RPG haploinsufficiency could introduce cancer cell-specific therapeutic vulnerabilities. Indeed, given what has been learned from the study of ribosomopathies, there are justifiable concerns regarding disruption of ribosome biogenesis as a therapeutic strategy.

However, the predominantly hematopoietic phenotype observed in DBA patients with ubiquitous lesions in RPGs suggests that tolerance for ribosomal dysfunction may vary widely between tissues, and may be particularly low in rapidly proliferating cells (such as normal erythroid

102 progenitors, or malignant tumor cells). 210-212Future studies will uncover the oncogenic roles and targetability of RPG haploinsufficiency in human cancer.

103 !

Discussion

Summary

Ribosomopathies are an intriguing set of human diseases, jointly set apart in their characteristic defects in ribosome-associated genes, but distinct in their phenotypic manifestations ranging from curiously lineage-specific anemia, to developmental abnormalities, to cancer predisposition. The last decade of research in this field has shed light on some of the downstream events that result from defective or haploinsufficient ribosome-associated gene expression, but many of the mechanistic details of these events have yet to be elucidated, and little has been learned with respect to useful therapeutic targets.

In this thesis work, I set out to better understand the biological consequences of defective ribosome biogenesis by interrogating the ribosomal protein (RP)-MDM2-p53 axis thought to be central to ribosomopathy pathology. By performing the first highly quantitative proteome-wide screen for MDM2-interacting proteins in the context of RP deficiency, we were able to identify and validate known and novel context-specific changes in the MDM2 binding partner profile relevant to ribosomopathy biology. The findings of this screen corroborate previous work identifying a subset of RPs as MDM2 binding partners, but they also call into question the context-specific nature of these interactions and further contribute to a growing skepticism in the field regarding the sufficiency of RP-MDM2 binding in the activation of p53. Additionally, this screen led us to the first characterization of IGF-1R and its selective loss in the context of RP deficiency as a novel contributing player in the dyserythropoiesis phenotype characteristic of ribosomopathies such as Diamond-Blackfan anemia (DBA) and 5q- syndrome. We also leveraged our expertise in in vitro and in vivo DBA models to screen for and identify compounds that rescue aspects of the DBA phenotype through modulation of p53 activity, which both lays the groundwork for the exploration of the therapeutic potential of calmodulin inhibitors and/or

105 CHK2 inhibitors in DBA as well as provides novel insight into the regulation of p53 more broadly. Lastly, we discovered that heterozygous RP gene loss is a common feature of human cancers, and that these losses are broadly p53-activating and cause defects in rRNA processing.

Consequently, RP gene deletion is associated with a strong selection pressure for mutation or loss of p53 or other tumor suppressor genes in many cancers. Together, this work adds novel insight to several aspects of ribosomopathy pathology and the RP-MDM2-p53 axis, and it further provides foundational evidence for novel therapeutic approaches to both ribosompathies and RP- deficient cancers.

The RP-MDM2 Interaction: New Insights

The proteomics work described in Chapter 2 is the first to comprehensively and quantitatively assess the protein binding partner profile of MDM2 in a biological context relevant to ribosomopathy disease pathology – namely, the context of ribosomal haploinsufficiency. Moreover, the design of the experiment and the use of SILAC-based IP/MS allowed us to assess four distinct MDM2 binding partner profiles: (1) proteins normally associated with MDM2 (in unperturbed cells); (2) proteins associated with MDM2 with ribosomal protein gene haploinsufficiency (as modeled with shRNA-mediated knockdown of

RPS14, a haploinsufficient gene in 5q- myelodysplastic syndrome, to roughly 50% of wild-type levels); and proteins preferentially associated with MDM2 either with (3) or without (4) ribosomal haploinsufficiency.

Many proteins, including the ribosomal proteins RPL5, RPL11, RPL23, and RPS7 have been shown in individual studies to interact with MDM2 and affect p53 pathway activity. We successfully identified all of these proteins, as well as other known interacting proteins p53 and

MDM4 (MDMX), as abundantly associated with MDM2 in one or more of the binding partner

106 profiles enumerated above. This provided a great deal of confidence in the validity of our approach, and further provides strong confirmatory evidence of previously identified RP-MDM2 interactions.

The SILAC-based approach used here is extraordinarily useful for the direct comparison of protein binding interactions in distinct biological settings. Because SILAC IP/MS is highly quantitative, we were able to show not only that RPL5 and RPL11 were bound to MDM2 in the absence of ribosomal stress, but also that enhanced binding with ribosomal stress is relatively subtle. Further, it allowed to us to appreciate broad shifts in the MDM2 binding partner profile – such as the association of all 40S RPs with MDM2, abrogated by RPS14 deficiency – that may not have been observable with other approaches. One caveat of our experimental design is that the “background” condition, included to eliminate those proteins that nonspecifically associated with either the Sepharose® beads or the anti-V5 antibody, was one in which RPS14 was knocked down. Thus, proteins that only bind nonspecifically in the context of ribosomal stress (but that might bind specifically to MDM2 otherwise) may be underrepresented in the binding partner profile of MDM2 with luciferase knockdown, which was determined relative to the background condition. However, we suspect that there are few proteins that fall into this category.

RPL5 and RPL11

At first blush, the finding that RPL5 and RPL11 are abundantly bound to MDM2 both with and without ribosomal dysfunction was somewhat unexpected, as a substantial body of work has provided insight into the increasingly complex role of these two proteins, as well as the

RPL5/RPL11/5S rRNA ribonucleoprotein (RNP), in the regulation of MDM2 in the ribosomal stress response. However, these data are consistent with a growing skepticism in the field with

107 respect to the sufficiency of RP-MDM2 binding in p53 pathway induction in the context of ribosomopathies.

The requirement for RPL5/RPL11-MDM2 binding in the ribosomal stress-associated p53 response has been shown by several groups, but perhaps the most compelling evidence of this has been demonstrated in studies of mice expressing the cancer-associated Mdm2C305F mutant, which is deficient in RPL5/RPL11 binding 124. These mice exhibit a normal p53 response to

DNA damage, but lack the p53 response to ribosomal stress. Intriguingly, more recent studies using these same mice have revealed a requirement for RPL5/RPL11 binding in normal hematopoiesis, with mice expressing two copies of Mdm2C305F developing mild macrocytic anemia with reticulocytosis 213. These data support the notion that RPL5/RPL11-MDM2 binding is necessary for proper modulation of MDM2 activity even in the absence of ribosomal stress or p53 pathway activation, and together with our data support the idea that this binding is not peculiar to scenarios of ribosomal dysfunction.

An often-neglected inconsistency in the RP-MDM2-p53 pathway hypothesis of ribosomopathy pathogenesis is the existence of DBA patients with mutations in or loss of RPL5 or RPL11 10. These mutations are unique among the many that have been identified in dozens of

RP genes in DBA patients, in that there are significant genotype-phenotype correlations in these patients: RPL5 mutations are correlated with craniofacial defects, whereas RPL11 mutations are correlated with thumb abnormalities 214. However, these patients develop a macrocytic anemia similar to that seen in DBA patients with other RP gene defects. One possible explanation for this is that the mutations present in RPL5 and RPL11 in DBA patients are deleterious with respect to their ribosomal functions, but do not affect extra-ribosomal functions such as MDM2- binding ability. This, however, has not been shown.

108 Other groups have performed over-expression experiments in an attempt to show the sufficiency of RP overabundance and increased RP-MDM2 binding in p53 pathway activation.

These studies have utilized transient transfection over-expression approaches to achieve extremely high temporal levels of RP expression, and have occasionally observed dramatic p53 pathway activation. It is unclear how clinically relevant these approaches are, as these levels of non-ribosome associated RPs are unlikely to be seen in patients, especially in the case of DBA where genetic RP lesions are ubiquitous and only result in p53 pathway activation a minority of the time. Our over-expression approach, which involved stable RP cDNA expression, did not result in p53 pathway activation. Despite expression driven by strong promoters, we were unable to achieve more than modest fold increases in RP expression. This may be a finding in and of itself, consistent with the idea that RP expression at both the gene and protein level is tightly regulated in cells, and that sustained over-expression is simply technically difficult to achieve.

IGF1R: A Novel Player in the Ribosomopathy Phenotype

IGF1R in erythropoiesis

Prior to this work, it was known that IGF1R ligands (including insulin and IGF-1) stimulate proliferation of erythroid progenitors in culture 153, 154. Much of what is else known about IGF-1 signaling in erythroid development was learned decades ago from studying patients with polycythemia vera (PV), a myeloproliferative neoplasm characterized by overproduction of red blood cells. Erythroid progenitors from PV patients are hypersensitive (~100x more sensitive) to IGF-1 as compared to normal erythroid progenitors, and further exhibit increased sensitivity and responsiveness of IGF1R to IGF-1 stimulation 215, 216. Moreover, PV patients express higher levels of IGF-1 binding protein 1 (IGFBP1) 217, which greatly stimulates erythroid

109 burst formation in the presence of IGF-1 218. It is important to note that erythroid progenitors from PV patients are also hyposensitive to the canonical erythroid stimulatory factor erythropoietin (EPO), perhaps as a consequence of expressing only low-affinity EPO receptor

(EpoR), and that normal erythroid progenitors do not exhibit such exquisite sensitivity to IGF-1.

Nonetheless, these data strongly support a role for IGF-1 signaling through IGF1R in erythroid cell development, and further show that signaling through either IGF1R or EpoR is able to promote erythropoiesis, though the signaling events downstream of these receptors have yet to be fully elucidated.

Our findings, too, support a clear role for IGF1R in erythropoiesis. Chapter 2 describes the discovery of IGF1R as a binding partner and likely degradation target of MDM2 in the context of RPS14 deficiency, and further shows that depletion of IGF1R results in an erythroid differentiation defect similar to, but less severe than, that caused by RPS14 deficiency.

Interestingly, this defect was observed even without IGF-1 present in the culture media. Insulin was present in the media and has low affinity for IGF1R, so it is possible that IGF1R signaling was achieved through insulin binding, though this was not tested directly. In any case, it will be interesting to test the ability of IGF1R over-expression to rescue erythroid differentiation defects in RPS14-deficient cells. To achieve this, it may be necessary to express a degradation-resistant

IGF1R (such as the Δ1245 C-terminal truncation mutant 152), as MDM2 may effectively target ectopically expressed wild-type IGF1R for degradation as well.

In vitro erythroid culture protocols have been optimized over the years to maximize expansion and differentiation of erythroid cells, and most achieve this with extremely high, superphysiologic concentrations of EPO. The EDM media used in the work described in Chapter

2 also includes high concentrations of EPO (3U/mL). As the studies in PV erythroid progenitors

110 would suggest, any ability to observe IGF-1-dependent effects on erythroid development would likely be obscured by saturated signaling through EPO/EpoR in these culture conditions. In order to utilize in vitro culture systems to better understand the importance of signaling through IGF1R in normal cells and in ribosomopathies, establishment of in vitro culture conditions in which erythroid expansion and/or differentiation is responsive to IGF-1 dosage will be advantageous.

Indeed, some of the initial work describing the effects of IGF-1 on erythroid cell development could inform such a protocol.

The idea of using IGF-1, IGFBP1, or other IGF1R-stimulatory factors as therapeutics in ribosomopathy patients is provocative, perhaps made moreso by the fact that, as in PV, EPO treatment has long been known to be ineffective in DBA 219. Despite a reduction in IGF1R protein, it is possible that high doses of IGF-1 may be effective in stimulating IGF1R signaling.

However, it will first be important to determine if and how IGF1R levels are affected in patients with ribosomopathies such as 5q- syndrome and DBA, which could be accomplished by immunohistochemistry staining for IGF1R on bone marrow aspirates and/or blood smears from these patients. Additionally, transcriptional profiling could be performed to assess changes in

IGF1R signaling gene signatures. Should decreases in IGF1R levels be observed, it would suggest that the effects we observed in primary human HSPCs are also present in human disease.

Even if IGF1R levels are not significantly reduced, an IGF1R hyperstimulation strategy could still be explored as a therapeutic approach to anemia in ribosomopathies.

The RP-MDM2-IGF1R axis

Our work adds additional color to an already complex relationship between IGF1R,

MDM2, and p53. MDM2 can ubiquitinate IGF1R and target it for degradation by the proteasome

111 151, 152, a process which requires the presence of beta-arrestins in complex with MDM2 and

IGF1R 160, 161, 220. IGF-1 is considered to be an anti-apoptotic factor, with signaling through

IGF1R most commonly leading to cell proliferation, growth, and survival, and neoplastic transformation in some tissues 221. MDM2-mediated degradation of IGF1R has been shown to occur even in the absence of p53, and thus has been described as a p53-independent pro- apoptotic function of MDM2 222. MDM2 loss results in increased levels of IGF1R protein, but it should be noted that in our studies stable over-expression of a V5-tagged MDM2 alone did not result in IGF1R protein loss, suggesting that additional signals (e.g. association of beta-arrestins with IGF1R) affecting MDM2 activity and/or localization are required for effective targeting of

IGF1R for degradation.

Our studies show that IGF1R loss alone does not result in p53 pathway activation (as shown by a lack of p21 induction), suggesting that the contribution of MDM2-mediated IGF1R loss to the erythroid differentiation block in response to ribosomal stress is independent of p53. It will be interesting to test whether IGF1R is similarly lost in p53-deficient cells experiencing ribosomal stress, and if this loss still results in erythroid defects. Further, it is unclear from our work what role RP-MDM2 binding plays in the selective targeting of IGF1R in the context of ribosomal protein deficiency, and whether or not this binding is required for IGF1R loss. The

Mdm2C305F mouse model, which utilizes a mutant MDM2 deficient in RP-binding, could be of particular use in testing this.

Different studies have reported different effects of IGF1R signaling on MDM2 subcellular localization and p53 modulating activity. Both nuclear import and nuclear export of

MDM2 can be regulated by IGF1R signaling, though the directionality of MDM2 localization depends on whether signaling results in downstream activation of the phosphatidylinositol 3-

112 kinase (PI3K)/Akt pathway (nuclear import) or the mitogen-activated protein kinase

(MAPK)/p90Rsk pathway (nuclear export) 223, 224. MDM2-mediated degradation of p53 can take place in either the nucleus or the cytoplasm 225, 226. Inhibition of MDM-mediated degradation of p53 can occur (1) by physical sequestration of MDM2 away from p53 [as accomplished by p19(ARF), which sequesters MDM2 in the nucleolus] 150; (2) by directly interfering with the E3 ubiquitin ligase activity of MDM2 toward p53 [as putatively accomplished by the

RPL5/RPL11/5S rRNA RNP] 159; or (3) by other indirect means (as accomplished by PICT1- mediated sequestration of RPL11 in the nucleolus) 227.

From these previous studies, together with the novel discovery of ribosomal protein deficiency leading to rapid loss of IGF1R protein in hematopoietic cells and the observation that

RPL5 and RPL11 are commonly associated with MDM2, a model begins to emerge (Figure 5.1).

Induction of ribosomal dysfunction results in inhibition of MDM2-mediated degradation of p53, due in part to increased binding of RPL5 and RPL11, but probably involving other additional as- yet unidentified cellular signals. IGF1R is targeted for degradation by MDM2 at the cell membrane, perhaps as a consequence of increased nuclear export of MDM2, and pro-erythroid signals are attenuated. A potential feedback loop, wherein reduced IGF1R results in reduced

IGF-1 signaling and consequent dysregulation of MDM2 subcellular localization, could promote the physical sequestration of MDM2 in the cytoplasm, allowing p53 activate its effector genes in the nucleus.

To be sure, many aspects of this model remain to be shown experimentally, including whether IGF1R regulation in hematopoietic cells has direct or indirect consequences on p53 activity (if any), and whether MDM2 localization is indeed affected by ribosomal protein deficiency. The role of accessory proteins (such as the beta arrestins) and/or post-translational

113 modifications on IGF1R in MDM2-mediated targeting also needs to be explored further, specifically in this biological context. Nonetheless, our work implicates IGF1R as a novel player in the ribosomopathy phenotype, and lays the groundwork for further exploration of the IGF1R-

MDM2 relationship in hematologic disease.

A B

MDM2 L5 p53 IGF-1R MDM2 IGF-1R cyto L11 L5 cyto L11 nuc nuc

Cell cycle MDM2 PI3K L5 p53 arrest PI3K L11 Cell growth p53

p21 MAPK p21 MAPK

Figure 5.1. Model of IGF1R regulation in RP-deficient hematopoietic cells.

(A) With normal ribosome biogenesis, IGF1R signaling promotes cell growth and expansion of erythroid progenitors, and also influences nucleocytoplasmic shuttling of MDM2 through PI3K and MAPK signaling pathways. p53 is degraded by MDM2. (B) Ribosomal protein deficiency results in inhibition of the negative regulatory control of MDM2 on p53, which functions as a transcription factor in the nucleus. MDM2 associates with beta-arrestins on IGF1R at the cell membrane and targets it for degradation, resulting in reduced signaling and defective erythropoiesis.

114 Calmodulin Inhibitors and p53: Old Dogs, New Tricks

Mechanism of p53 modulation

In Chapter 3 of this work, we used a zebrafish model of DBA to screen for compounds that could rescue ribosomal protein haploinsufficiency, and we successfully identified calmodulin inhibitors as regulators of p53. This novel finding was extremely intriguing and somewhat unexpected, given that calmodulin is able to interact with and activate quite a wide range of cellular proteins 228. Upon the initial finding that calmodulin inhibitors could rescue rps29-/- phenotypes in the zebrafish, we considered several hypotheses for their activity. As some calmodulin-dependent enzymes have been shown to function in apoptosis 229, it was possible that calmodulin inhibition was effectively blocking p53-induced apoptosis. We also considered the possibility that, since flk1 expression was rescued in zebrafish embryo vasculature, calmodulin inhibition may be affecting signaling in a tissue specific manner, independent of p53. The discovery that p53 localization was affected by calmodulin inhibition was not anticipated, and was the first to associate calmodulin with p53 in this way. This discovery will be of interest to not only the DBA field, but to the p53 research community as a whole.

Our work has made great strides in elucidating the mechanism of p53 modulation by calmodulin inhibition, including the identification of CHK2 as at least one calmodulin-dependent enzyme important for p53 nuclear localization, but additional work is needed to understand the exact mechanism by which p53 nuclear translocation is attenuated. CHK2 phosphorylates p53 at a number of different sites 230, and it is possible that direct phosphorylation of p53 by CHK2 is required for nuclear translocation. This could be tested using a targeted CHK2 phosphorylation site mutagenesis approach, or by performing post-translational modification analysis on p53

115 isolated from RPS19-deficient cells with and without calmodulin inhibitor treatment. Our data do not rule out the possibility that calmodulin and/or CHK2 act indirectly on p53 through modulation of other p53-modifying enzymes or other proteins required for p53 nuclear import.

Additionally, previous studies have shown that calmodulin can function in nuclear import when cellular calcium levels are high and importin function inhibited 187; it is possible that calcium levels are affected by ribosomal stress, though this has not yet been tested.

The majority of our work was performed in the context of ribosomal stress, and we have clearly shown a role for calmodulin inhibitors in the attenuation of p53 activity under these conditions. However, we also showed that calmodulin inhibitors rescue irradiation-induced apoptosis in zebrafish embryos, suggesting that calmodulin inhibition may be functional in modulation of p53 when induced by other stimuli as well (e.g. DNA damage, oxidative stress).

Therapeutic implications of calmodulin inhibition

We have demonstrated the p53 modulatory effects of calmodulin inhibition in two in vivo animal models, as well as in an in vitro model using primary human hematopoietic cells.

Specifically, we have shown that calmodulin inhibition can rescue hemoglobinization in ribosomal protein deficient zebrafish, anemia in Rps19-deficient mice, and an erythroid differentiation block in primary human HSPCs. These data suggest that inhibition of calmodulin

(and possibly of CHK2) may be a viable therapeutic strategy for the treatment of hematopoietic defects in patients with DBA. Intriguingly (and somewhat fortuitously), trifluoperazine (TFP), the calmodulin inhibitor studied in the most depth in our work, is currently being used in the clinic for the treatment of schizophrenia 194. While both the intended tissue of action (brain vs. blood) and target patient population (mostly adult for schizophrenia, pediatric for DBA) differ

116 substantially in these two indications, its previous approved use in humans shows that TFP is acceptably safe and tolerated. Further, the fact that schizophrenia patients receiving TFP have not been reported to have a higher cancer incidence may assuage some fears regarding the potential cancer-associated risks of using of p53 attenuating molecules in patients. Intriguingly, at least one report has shown that TFP actually inhibits cancer cell growth 231.

Together, our data provide strong justification for the testing of TFP in the clinic.

Additionally, we have established several in vitro and in vivo systems for the testing of additional calmodulin inhibitors. In our studies, the effective doses of TFP used in our studies were quite high, and probably not achievable in humans without significant toxic side effects. Medicinal chemistry and lead optimization should be performed on TFP and other phenothiazines to see if the efficacy and toxicity profiles of these molecules can be improved, improving their likelihood of therapeutic success in the clinic.

Ribosomal Protein Gene Deletions in Cancer: an Achilles Heel?

Searching for compensatory mechanisms in RP-deficient cells

The work presented in Chapter 4 contributes novel insight into the role of ribosomal protein gene (RPG) deletions in human cancers. While our observation that p53 and other tumor suppressor genes are often co-deleted with RPGs provides an explanation for how RP-deficient tumors can evade ribosomal stress-induced p53 pathway activation, a major question remains: how do cells with dysfunctional ribosomes support a hyperproliferative state?

The study of ribosomopathies may ultimately provide the answer. Studies utilizing ribosomopathy models have already shown that RPG haploinsufficiency leads to altered

117 translation in DBA 139, and other studies have provided some foundational evidence that ribosomal stress provides a selective pressure for acquisition of mutations that compensate for defective ribosome biogenesis 232, 233. Our broad analysis of RPG deletions in a wide range of cancer and tissue types will be useful in future analyses seeking to identify translational changes or mutation patterns associated with RPG loss. Genetic models involving RPGs such as RPL5 and RPL11, which are both DBA genes also associated with T-ALL 136, 234, may be of particular use in gaining a better understanding of the events that occur as a cell with ribosomal stress transitions from a hypo- to a hyper-proliferative state.

Targeting RP-deficient cancers

The concept of targeting genes that are already reduced in levels and/or function in cancer is not entirely novel. A landmark study revealing vulnerabilities in cancers with so-called

CYCLOPS (Copy number alterations Yielding Cancer Liabilities Owing to Partial losS) genes has drawn much attention to this idea, and demonstrated that suppression of cancer-specific haploinsufficient genes is an effective means of achieving targeted toxicity 235; notably, the

RPGs RPS9, RPS11, RPS15 were included in the CYCLOPS gene list, and RPS15 was one of the

RPGs most commonly deleted in human cancers. In another study, analysis of homozygous passenger deletions of genes within functionally redundant families of genes carrying out essential cellular functions exposed another vulnerability. Suppression of a functionally redundant counterpart of one such cancer-deleted gene resulted in selective toxicity to cells harboring the deletion, suggesting that this may be a viable approach to targeting these cancers with minimal toxicity to normal cells 236.

118 The relative absence of homozygous RPG deletions – along with the observation that

RPS6 constrains the extent of homozygous deletions around the tumor suppressor locus

CDKN2A – indicates that there could be a therapeutic window for drugs that exploit ribosome defects, or altered protein synthesis, to selectively kill cancer cells. Considering the high frequency of RPS6 deletions in cancer, it may be an interesting target, particularly as mTORC inhibitors rapamycin and its analogues modulate RPS6 through RPS6-kinase. Hence, RPG haploinsufficiency could introduce cancer cell-specific vulnerabilities. A number of drugs that modulate ribosomal function exist, including rapamycin, other inhibitors (”rapalogs”) of the mammalian target of rapamycin (mTOR)210, and lenalinomide (Revlimid™; active in the 5q- syndrome) 211, 212. These drugs are active against subsets of human tumors, but their exact therapeutic scope is unknown. That their activity could depend on the genotype of the tumor, and more specifically whether or not RPG deletions are present, is an intriguing possibility.

Our analysis revealed that certain RPGs are lost more frequently in particular tumor types, such as RPS6 in ALL and glioblastoma, RPL17 in colon and rectal cancers, and RPL26 in liver cancers. These associations may provide some clues as to which RPGs to target in which cancer types. Strikingly, ovarian adenocarcinomas had at least moderately high deletion frequency of roughly half of the RPGs analyzed our work. Given the relatively high number of ovarian cancer cell lines with well-characterized oncogene and tumor suppressor gene status 237, this cancer type may provide an ideal in vitro model for the study of both specific and broadly-acting agents that suppress RPGs.

RPG deletions are likely to be rare in normal tissues, as we and others have demonstrated that deficiency of many individual RPs results in p53 pathway activation. Moreover, when these hemizygous genetic lesions are present, as they are ubiquitously in DBA, they appear to be well

119 tolerated in most tissues – though the hematopoietic system seems to be the marked exception here. That said, the potential oncogenic selective pressure of defective ribosome biogenesis discussed above will be a major concern of RPG-targeting therapies. Nonetheless, it will be interesting to learn whether pharmacological suppression of RPGs can be tolerated by normal cells, and thus whether there is viability in this approach to targeting RP-deficient cancers.

Concluding Remarks

Collectively, this work contributes multiple novel insights into various aspects of the increasingly complex RP-MDM2-p53 axis in human ribosomopathies. We have implicated

IGF1R as a novel player in the ribosomopathy phenotype, and further raised questions regarding the sufficiency of RP-MDM2 binding in ribosomal stress-induced p53 activation. Further investigation is warranted to identify additional signals and molecular players that may contribute to this process. We have also laid the groundwork for new therapeutic strategies in both ribosomopathies and RP-deficient cancers, and the models developed in these studies will be useful in further characterizing these therapeutic approaches before bringing them to the clinic.

120

References

1. Diamond LK, Blackfan KD. Hypoplastic anemia. AmJDisChild. 1938;56:464.

2. Josephs HW. Anaemia of infancy and early childhood. Medicine. 1936;15:307.

3. Orfali KA, Ohene-Abuakwa Y, Ball SE. Diamond Blackfan anaemia in the UK: clinical and genetic heterogeneity. Br J Haematol. 2004;125(2):243-252.

4. Lipton JM, Ellis SR. Diamond-Blackfan anemia: diagnosis, treatment, and molecular pathogenesis. Hematol Oncol Clin North Am. 2009;23(2):261-282.

5. Vlachos A, Muir E. How I treat Diamond-Blackfan anemia. Blood. 2010;116(19):3715- 3723.

6. Narla A, Ebert BL. Translational medicine: ribosomopathies. Blood. 2011;118(16):4300- 4301.

7. Vlachos A, Ball S, Dahl N, et al. Diagnosing and treating Diamond Blackfan anaemia: results of an international clinical consensus conference. Br J Haematol. 2008;142(6):859-876.

8. Ellis SR, Gleizes PE. Diamond Blackfan anemia: ribosomal proteins going rogue. Semin Hematol. 2011;48(2):89-96.

9. Gazda HT, Preti M, Sheen MR, et al. Frameshift mutation in p53 regulator RPL26 is associated with multiple physical abnormalities and a specific pre-ribosomal RNA processing defect in diamond-blackfan anemia. Hum Mutat. 2012;33(7):1037-1044.

10. Gazda HT, Sheen MR, Vlachos A, et al. Ribosomal protein L5 and L11 mutations are associated with cleft palate and abnormal thumbs in Diamond-Blackfan anemia patients. Am J Hum Genet. 2008;83(6):769-780.

11. Mirabello L, Ballew BJ, Neelam G, et al. RPS29 is Mutated in a Multi-Case Diamond Blackfan Anemia Family. American Society of Hematology Annual Conference. Atlanta, GA; 2012.

121 12. Farrar JE, Vlachos A, Atsidaftos E, et al. Ribosomal protein gene deletions in Diamond- Blackfan anemia. Blood. 2011;118(26):6943-6951.

13. Kuramitsu M, Sato-Otsubo A, Morio T, et al. Extensive gene deletions in Japanese patients with Diamond-Blackfan anemia. Blood. 2012;119(10):2376-2384.

14. Quarello P, Garelli E, Brusco A, et al. High frequency of ribosomal protein gene deletions in Italian Diamond-Blackfan anemia patients detected by multiplex ligation-dependent probe amplification assay. Haematologica. 2012;97(12):1813-1817.

15. Sankaran VG, Ghazvinian R, Do R, et al. Exome sequencing identifies GATA1 mutations resulting in Diamond-Blackfan anemia. J Clin Invest. 2012;122(7):2439-2443.

16. Gazda HT, Grabowska A, Merida-Long LB, et al. Ribosomal protein S24 gene is mutated in Diamond-Blackfan anemia. Am J Hum Genet. 2006;79(6):1110-1118.

17. Cmejla R, Cmejlova J, Handrkova H, Petrak J, Pospisilova D. Ribosomal protein S17 gene (RPS17) is mutated in Diamond-Blackfan anemia. Hum Mutat. 2007;28(12):1178-1182.

18. Moniz H, Gastou M, Leblanc T, et al. Primary hematopoietic cells from DBA patients with mutations in RPL11 and RPS19 genes exhibit distinct erythroid phenotype in vitro. Cell Death Dis. 2012;3:e356.

19. O'Donohue MF, Choesmel V, Faubladier M, Fichant G, Gleizes PE. Functional dichotomy of ribosomal proteins during the synthesis of mammalian 40S ribosomal subunits. J Cell Biol. 2010;190(5):853-866.

20. Choesmel V, Bacqueville D, Rouquette J, et al. Impaired ribosome biogenesis in Diamond-Blackfan anemia. Blood. 2007;109(3):1275-1283.

21. Choesmel V, Fribourg S, Aguissa-Toure AH, et al. Mutation of ribosomal protein RPS24 in Diamond-Blackfan anemia results in a ribosome biogenesis disorder. Hum Mol Genet. 2008;17(9):1253-1263.

22. Flygare J, Aspesi A, Bailey JC, et al. Human RPS19, the gene mutated in Diamond- Blackfan anemia, encodes a ribosomal protein required for the maturation of 40S ribosomal subunits. Blood. 2007;109(3):980-986.

122 23. Idol RA, Robledo S, Du HY, et al. Cells depleted for RPS19, a protein associated with Diamond Blackfan Anemia, show defects in 18S ribosomal RNA synthesis and small ribosomal subunit production. Blood Cells Mol Dis. 2007;39(1):35-43.

24. Fumagalli S, Di Cara A, Neb-Gulati A, et al. Absence of nucleolar disruption after impairment of 40S ribosome biogenesis reveals an rpL11-translation-dependent mechanism of p53 induction. Nat Cell Biol. 2009;11(4):501-508.

25. Fumagalli S, Ivanenkov VV, Teng T, Thomas G. Suprainduction of p53 by disruption of 40S and 60S ribosome biogenesis leads to the activation of a novel G2/M checkpoint. Genes Dev. 2012;26(10):1028-1040.

26. Robledo S, Idol RA, Crimmins DL, Ladenson JH, Mason PJ, Bessler M. The role of human ribosomal proteins in the maturation of rRNA and ribosome production. RNA. 2008;14(9):1918-1929.

27. Van den Berghe H, Cassiman JJ, David G, Fryns JP, Michaux JL, Sokal G. Distinct haematological disorder with deletion of long arm of no. 5 chromosome. Nature. 1974;251(5474):437-438.

28. Ebert BL. Molecular dissection of the 5q deletion in myelodysplastic syndrome. Semin Oncol. 2011;38(5):621-626.

29. Vardiman JW, Harris NL, Brunning RD. The World Health Organization (WHO) classification of the myeloid neoplasms. Blood. 2002;100(7):2292-2302.

30. List A, Dewald G, Bennett J, et al. Lenalidomide in the myelodysplastic syndrome with chromosome 5q deletion. N Engl J Med. 2006;355(14):1456-1465.

31. Fenaux P, Giagounidis A, Selleslag D, et al. A randomized phase 3 study of lenalidomide versus placebo in RBC transfusion-dependent patients with Low-/Intermediate-1-risk myelodysplastic syndromes with del5q. Blood. 2011;118(14):3765-3776.

32. Ebert BL, Pretz J, Bosco J, et al. Identification of RPS14 as a 5q- syndrome gene by RNA interference screen. Nature. 2008;451(7176):335-339.

33. Moutouh-de Parseval LA, Verhelle D, Glezer E, et al. Pomalidomide and lenalidomide regulate erythropoiesis and fetal hemoglobin production in human CD34+ cells. J Clin Invest. 2008;118(1):248-258.

123 34. Narla A, Dutt S, McAuley JR, et al. Dexamethasone and lenalidomide have distinct functional effects on erythropoiesis. Blood. 2011;118(8):2296-2304.

35. Joslin JM, Fernald AA, Tennant TR, et al. Haploinsufficiency of EGR1, a candidate gene in the del(5q), leads to the development of myeloid disorders. Blood. 2007;110(2):719-726.

36. Kumar M NA, et al. Abstract 947: Coordinate Loss of a MicroRNA Mir 145 and a Protein-Coding Gene RPS14 Cooperate in the Pathogenesis of 5q- Syndrome Oral Session presented at Annual Meeting of the American Society of Hematology December 8, 2009 New Orleans, LA.

37. Min IM, Pietramaggiori G, Kim FS, Passegue E, Stevenson KE, Wagers AJ. The transcription factor EGR1 controls both the proliferation and localization of hematopoietic stem cells. Cell Stem Cell. 2008;2(4):380-391.

38. Starczynowski DT, Kuchenbauer F, Argiropoulos B, et al. Identification of miR-145 and miR-146a as mediators of the 5q- syndrome phenotype. Nat Med. 2010;16(1):49-58.

39. Shwachman H, Diamond LK, Oski FA, Khaw KT. The Syndrome of Pancreatic Insufficiency and Bone Marrow Dysfunction. J Pediatr. 1964;65:645-663.

40. Burroughs L, Woolfrey A, Shimamura A. Shwachman-Diamond syndrome: a review of the clinical presentation, molecular pathogenesis, diagnosis, and treatment. Hematol Oncol Clin North Am. 2009;23(2):233-248.

41. Donadieu J, Leblanc T, Bader Meunier B, et al. Analysis of risk factors for myelodysplasias, leukemias and death from infection among patients with congenital neutropenia. Experience of the French Severe Chronic Neutropenia Study Group. Haematologica. 2005;90(1):45-53.

42. Ganapathi KA, Shimamura A. Ribosomal dysfunction and inherited marrow failure. Br J Haematol. 2008;141(3):376-387.

43. Boocock GR, Morrison JA, Popovic M, et al. Mutations in SBDS are associated with Shwachman-Diamond syndrome. Nat Genet. 2003;33(1):97-101.

44. Finch AJ, Hilcenko C, Basse N, et al. Uncoupling of GTP hydrolysis from eIF6 release on the ribosome causes Shwachman-Diamond syndrome. Genes Dev. 2011;25(9):917-929.

124 45. Menne TF, Goyenechea B, Sanchez-Puig N, et al. The Shwachman-Bodian-Diamond syndrome protein mediates translational activation of ribosomes in yeast. Nat Genet. 2007;39(4):486-495.

46. Wong CC, Traynor D, Basse N, Kay RR, Warren AJ. Defective ribosome assembly in Shwachman-Diamond syndrome. Blood. 2011;118(16):4305-4312.

47. Rujkijyanont P, Adams SL, Beyene J, Dror Y. Bone marrow cells from patients with Shwachman-Diamond syndrome abnormally express genes involved in ribosome biogenesis and RNA processing. Br J Haematol. 2009.

48. Burwick N, Coats SA, Nakamura T, Shimamura A. Impaired ribosomal subunit association in Shwachman-Diamond syndrome. Blood. 2012;120(26):5143-5152.

49. Austin KM, Gupta ML, Coats SA, et al. Mitotic spindle destabilization and genomic instability in Shwachman-Diamond syndrome. J Clin Invest. 2008;118(4):1511-1518.

50. Orelio C, Verkuijlen P, Geissler J, van den Berg TK, Kuijpers TW. SBDS expression and localization at the mitotic spindle in human myeloid progenitors. PLoS One. 2009;4(9):e7084.

51. Dokal I, Vulliamy T. Dyskeratosis congenita: its link to telomerase and aplastic anaemia. Blood Rev. 2003;17(4):217-225.

52. Mason PJ, Bessler M. The genetics of dyskeratosis congenita. Cancer Genet. 2011;204(12):635-645.

53. Savage SA, Alter BP. Dyskeratosis congenita. Hematol Oncol Clin North Am. 2009;23(2):215-231.

54. Vulliamy T, Beswick R, Kirwan M, et al. Mutations in the telomerase component NHP2 cause the premature ageing syndrome dyskeratosis congenita. Proc Natl Acad Sci U S A. 2008;105(23):8073-8078.

55. Vulliamy TJ, Marrone A, Knight SW, Walne A, Mason PJ, Dokal I. Mutations in dyskeratosis congenita: their impact on telomere length and the diversity of clinical presentation. Blood. 2006;107(7):2680-2685.

125 56. Marrone A, Walne A, Tamary H, et al. Telomerase reverse-transcriptase homozygous mutations in autosomal recessive dyskeratosis congenita and Hoyeraal-Hreidarsson syndrome. Blood. 2007;110(13):4198-4205.

57. Mochizuki Y, He J, Kulkarni S, Bessler M, Mason PJ. Mouse dyskerin mutations affect accumulation of telomerase RNA and small nucleolar RNA, telomerase activity, and ribosomal RNA processing. Proc Natl Acad Sci U S A. 2004;101(29):10756-10761.

58. Wong JM, Collins K. Telomerase RNA level limits telomere maintenance in X-linked dyskeratosis congenita. Genes Dev. 2006;20(20):2848-2858.

59. Ruggero D, Grisendi S, Piazza F, et al. Dyskeratosis congenita and cancer in mice deficient in ribosomal RNA modification. Science. 2003;299(5604):259-262.

60. McKusick VA, Eldridge R, Hostetler JA, Ruangwit U, Egeland JA. Dwarfism in the Amish. Ii. Cartilage-Hair Hypoplasia. Bull Johns Hopkins Hosp. 1965;116:285-326.

61. Makitie O. Cartilage-hair hypoplasia in Finland: epidemiological and genetic aspects of 107 patients. J Med Genet. 1992;29(9):652-655.

62. Thiel CT, Rauch A. The molecular basis of the cartilage-hair hypoplasia-anauxetic dysplasia spectrum. Best Pract Res Clin Endocrinol Metab. 2011;25(1):131-142.

63. Berthet F, Siegrist CA, Ozsahin H, et al. Bone marrow transplantation in cartilage-hair hypoplasia: correction of the immunodeficiency but not of the chondrodysplasia. Eur J Pediatr. 1996;155(4):286-290.

64. Ridanpaa M, van Eenennaam H, Pelin K, et al. Mutations in the RNA component of RNase MRP cause a pleiotropic human disease, cartilage-hair hypoplasia. Cell. 2001;104(2):195-203.

65. Hermanns P, Tran A, Munivez E, et al. RMRP mutations in cartilage-hair hypoplasia. Am J Med Genet A. 2006;140(19):2121-2130.

66. Welting TJ, van Venrooij WJ, Pruijn GJ. Mutual interactions between subunits of the human RNase MRP ribonucleoprotein complex. Nucleic Acids Res. 2004;32(7):2138-2146.

126 67. Sarafan-Vasseur N, Lamy A, Bourguignon J, et al. Overexpression of B-type cyclins alters chromosomal segregation. Oncogene. 2002;21(13):2051-2057.

68. Thiel CT, Horn D, Zabel B, et al. Severely incapacitating mutations in patients with extreme short stature identify RNA-processing endoribonuclease RMRP as an essential cell growth regulator. Am J Hum Genet. 2005;77(5):795-806.

69. Maida Y, Yasukawa M, Furuuchi M, et al. An RNA-dependent RNA polymerase formed by TERT and the RMRP RNA. Nature. 2009;461(7261):230-235.

70. Sakai D, Trainor PA. Treacher Collins syndrome: unmasking the role of Tcof1/treacle. Int J Biochem Cell Biol. 2009;41(6):1229-1232.

71. Treacher-Collins E. Case with symmetrical congenital notches in the outer part of each lower lid and defective development of the malar bones. Trans Opthalmol Soc UK. 1900;20:90.

72. Valdez BC, Henning D, So RB, Dixon J, Dixon MJ. The Treacher Collins syndrome (TCOF1) gene product is involved in ribosomal DNA gene transcription by interacting with upstream binding factor. Proc Natl Acad Sci U S A. 2004;101(29):10709-10714.

73. Dauwerse JG, Dixon J, Seland S, et al. Mutations in genes encoding subunits of RNA polymerases I and III cause Treacher Collins syndrome. Nat Genet. 2011;43(1):20-22.

74. Jaako P, Debnath S, Olsson K, Bryder D, Flygare J, Karlsson S. Dietary L-leucine improves the anemia in a mouse model for Diamond-Blackfan anemia. Blood. 2012;120(11):2225-2228.

75. Pospisilova D, Cmejlova J, Hak J, Adam T, Cmejla R. Successful treatment of a Diamond-Blackfan anemia patient with amino acid leucine. Haematologica. 2007;92(5):e66-67.

76. Sjogren SE, Flygare J. Progress towards mechanism-based treatment for Diamond- Blackfan anemia. ScientificWorldJournal. 2012;2012:184362.

77. Miyake K, Utsugisawa T, Flygare J, et al. Ribosomal protein S19 deficiency leads to reduced proliferation and increased apoptosis but does not affect terminal erythroid differentiation in a cell line model of Diamond-Blackfan anemia. Stem Cells. 2008;26(2):323- 329.

127 78. Ohene-Abuakwa Y, Orfali KA, Marius C, Ball SE. Two-phase culture in Diamond Blackfan anemia: localization of erythroid defect. Blood. 2005;105(2):838-846.

79. Ebert BL, Lee MM, Pretz JL, et al. An RNA interference model of RPS19 deficiency in Diamond-Blackfan anemia recapitulates defective hematopoiesis and rescue by dexamethasone: identification of dexamethasone-responsive genes by microarray. Blood. 2005;105(12):4620- 4626.

80. Flygare J, Kiefer T, Miyake K, et al. Deficiency of ribosomal protein S19 in CD34+ cells generated by siRNA blocks erythroid development and mimics defects seen in Diamond- Blackfan anemia. Blood. 2005;105(12):4627-4634.

81. Hamaguchi I, Ooka A, Brun A, Richter J, Dahl N, Karlsson S. Gene transfer improves erythroid development in ribosomal protein S19-deficient Diamond-Blackfan anemia. Blood. 2002;100(8):2724-2731.

82. Flygare J, Olsson K, Richter J, Karlsson S. Gene therapy of Diamond Blackfan anemia CD34(+) cells leads to improved erythroid development and engraftment following transplantation. Exp Hematol. 2008;36(11):1428-1435.

83. Zivny J, Jelinek J, Pospisilova D, Plasilova M, Necas E, Stopka T. Diamond blackfan anemia stem cells fail to repopulate erythropoiesis in NOD/SCID mice. Blood Cells Mol Dis. 2003;31(1):93-97.

84. Danilova N, Sakamoto KM, Lin S. Ribosomal protein S19 deficiency in zebrafish leads to developmental abnormalities and defective erythropoiesis through activation of p53 protein family. Blood. 2008;112(13):5228-5237.

85. Payne EM, Virgilio M, Narla A, et al. L-Leucine improves the anemia and developmental defects associated with Diamond-Blackfan anemia and del(5q) MDS by activating the mTOR pathway. Blood. 2012;120(11):2214-2224.

86. Uechi T, Nakajima Y, Chakraborty A, Torihara H, Higa S, Kenmochi N. Deficiency of ribosomal protein S19 during early embryogenesis leads to reduction of erythrocytes in a zebrafish model of Diamond-Blackfan anemia. Hum Mol Genet. 2008;17(20):3204-3211.

87. Chakraborty A, Uechi T, Higa S, Torihara H, Kenmochi N. Loss of ribosomal protein L11 affects zebrafish embryonic development through a p53-dependent apoptotic response. PLoS One. 2009;4(1):e4152.

128 88. Taylor AM, Humphries JM, White RM, Murphey RD, Burns CE, Zon LI. Hematopoietic defects in rps29 mutant zebrafish depend upon p53 activation. Exp Hematol. 2012;40(3):228-237 e225.

89. Amsterdam A, Nissen RM, Sun Z, Swindell EC, Farrington S, Hopkins N. Identification of 315 genes essential for early zebrafish development. Proc Natl Acad Sci U S A. 2004;101(35):12792-12797.

90. Golling G, Amsterdam A, Sun Z, et al. Insertional mutagenesis in zebrafish rapidly identifies genes essential for early vertebrate development. Nat Genet. 2002;31(2):135-140.

91. Amsterdam A, Sadler KC, Lai K, et al. Many ribosomal protein genes are cancer genes in zebrafish. PLoS Biol. 2004;2(5):E139.

92. Alter BP, Giri N, Savage SA, Rosenberg PS. Cancer in dyskeratosis congenita. Blood. 2009;113(26):6549-6557.

93. Ebert BL. Deletion 5q in myelodysplastic syndrome: a paradigm for the study of hemizygous deletions in cancer. Leukemia. 2009;23(7):1252-1256.

94. Montanaro L. Dyskerin and cancer: more than telomerase. The defect in mRNA translation helps in explaining how a proliferative defect leads to cancer. J Pathol. 2010;222(4):345-349.

95. Vlachos A, Rosenberg PS, Atsidaftos E, Alter BP, Lipton JM. Incidence of neoplasia in Diamond Blackfan anemia: a report from the Diamond Blackfan Anemia Registry. Blood. 2012;119(16):3815-3819.

96. Anderson SJ, Lauritsen JP, Hartman MG, et al. Ablation of ribosomal protein L22 selectively impairs alphabeta T cell development by activation of a p53-dependent checkpoint. Immunity. 2007;26(6):759-772.

97. Barkic M, Crnomarkovic S, Grabusic K, et al. The p53 tumor suppressor causes congenital malformations in Rpl24-deficient mice and promotes their survival. Mol Cell Biol. 2009;29(10):2489-2504.

98. Barlow JL, Drynan LF, Hewett DR, et al. A p53-dependent mechanism underlies macrocytic anemia in a mouse model of human 5q- syndrome. Nat Med. 2010;16(1):59-66.

129 99. Gu BW, Bessler M, Mason PJ. A pathogenic dyskerin mutation impairs proliferation and activates a DNA damage response independent of telomere length in mice. Proc Natl Acad Sci U S A. 2008;105(29):10173-10178.

100. Jaako P, Flygare J, Olsson K, et al. Mice with ribosomal protein S19 deficiency develop bone marrow failure and symptoms like patients with Diamond-Blackfan anemia. Blood. 2011;118(23):6087-6096.

101. Jones NC, Lynn ML, Gaudenz K, et al. Prevention of the neurocristopathy Treacher Collins syndrome through inhibition of p53 function. Nat Med. 2008;14(2):125-133.

102. Kondrashov N, Pusic A, Stumpf CR, et al. Ribosome-mediated specificity in Hox mRNA translation and vertebrate tissue patterning. Cell. 2011;145(3):383-397.

103. Matsson H, Davey EJ, Draptchinskaia N, et al. Targeted disruption of the ribosomal protein S19 gene is lethal prior to implantation. Mol Cell Biol. 2004;24(9):4032-4037.

104. McGowan KA, Li JZ, Park CY, et al. Ribosomal mutations cause p53-mediated dark skin and pleiotropic effects. Nat Genet. 2008;40(8):963-970.

105. McGowan KA, Pang WW, Bhardwaj R, et al. Reduced ribosomal protein gene dosage and p53 activation in low-risk myelodysplastic syndrome. Blood. 2011;118(13):3622-3633.

106. Oliver ER, Saunders TL, Tarle SA, Glaser T. Ribosomal protein L24 defect in belly spot and tail (Bst), a mouse Minute. Development. 2004;131(16):3907-3920.

107. Sulic S, Panic L, Barkic M, Mercep M, Uzelac M, Volarevic S. Inactivation of S6 ribosomal protein gene in T lymphocytes activates a p53-dependent checkpoint response. Genes Dev. 2005;19(24):3070-3082.

108. Volarevic S, Stewart MJ, Ledermann B, et al. Proliferation, but not growth, blocked by conditional deletion of 40S . Science. 2000;288(5473):2045-2047.

109. Zhang S, Shi M, Hui CC, Rommens JM. Loss of the mouse ortholog of the shwachman- diamond syndrome gene (Sbds) results in early embryonic lethality. Mol Cell Biol. 2006;26(17):6656-6663.

130 110. Devlin EE, Dacosta L, Mohandas N, Elliott G, Bodine DM. A transgenic mouse model demonstrates a dominant negative effect of a point mutation in the RPS19 gene associated with Diamond-Blackfan anemia. Blood. 2010;116(15):2826-2835.

111. Talos F, Moll UM. Role of the p53 family in stabilizing the genome and preventing polyploidization. Adv Exp Med Biol. 2010;676:73-91.

112. Pereboom TC, van Weele LJ, Bondt A, MacInnes AW. A zebrafish model of dyskeratosis congenita reveals hematopoietic stem cell formation failure resulting from ribosomal protein-mediated p53 stabilization. Blood. 2011;118(20):5458-5465.

113. Dutt S, Narla A, Lin K, et al. Haploinsufficiency for ribosomal protein genes causes selective activation of p53 in human erythroid progenitor cells. Blood. 2011;117(9):2567-2576.

114. Chen D, Zhang Z, Li M, et al. Ribosomal protein S7 as a novel modulator of p53-MDM2 interaction: binding to MDM2, stabilization of p53 protein, and activation of p53 function. Oncogene. 2007;26(35):5029-5037.

115. Dai MS, Lu H. Inhibition of MDM2-mediated p53 ubiquitination and degradation by ribosomal protein L5. J Biol Chem. 2004;279(43):44475-44482.

116. Dai MS, Zeng SX, Jin Y, Sun XX, David L, Lu H. Ribosomal protein L23 activates p53 by inhibiting MDM2 function in response to ribosomal perturbation but not to translation inhibition. Mol Cell Biol. 2004;24(17):7654-7668.

117. Jin A, Itahana K, O'Keefe K, Zhang Y. Inhibition of HDM2 and activation of p53 by ribosomal protein L23. Mol Cell Biol. 2004;24(17):7669-7680.

118. Lohrum MA, Ludwig RL, Kubbutat MH, Hanlon M, Vousden KH. Regulation of HDM2 activity by the ribosomal protein L11. Cancer Cell. 2003;3(6):577-587.

119. Ofir-Rosenfeld Y, Boggs K, Michael D, Kastan MB, Oren M. Mdm2 regulates p53 mRNA translation through inhibitory interactions with ribosomal protein L26. Mol Cell. 2008;32(2):180-189.

120. Zhang X, Wang W, Wang H, Wang MH, Xu W, Zhang R. Identification of ribosomal protein S25 (RPS25)-MDM2-p53 regulatory feedback loop. Oncogene. 2012.

131 121. Zhang Y, Wang J, Yuan Y, et al. Negative regulation of HDM2 to attenuate p53 degradation by ribosomal protein L26. Nucleic Acids Res. 2010;38(19):6544-6554.

122. Zhang Y, Wolf GW, Bhat K, et al. Ribosomal protein L11 negatively regulates oncoprotein MDM2 and mediates a p53-dependent ribosomal-stress checkpoint pathway. Mol Cell Biol. 2003;23(23):8902-8912.

123. Zhou X, Hao Q, Liao J, Zhang Q, Lu H. Ribosomal protein S14 unties the MDM2-p53 loop upon ribosomal stress. Oncogene. 2013;32(3):388-396.

124. Macias E, Jin A, Deisenroth C, et al. An ARF-independent c-MYC-activated tumor suppression pathway mediated by ribosomal protein-Mdm2 Interaction. Cancer Cell. 2010;18(3):231-243.

125. Takagi M, Absalon MJ, McLure KG, Kastan MB. Regulation of p53 translation and induction after DNA damage by ribosomal protein L26 and nucleolin. Cell. 2005;123(1):49-63.

126. Lajtha LG, Oliver R. A kinetic model of the erythron. Proc R Soc Med. 1961;54:369-371.

127. Keel SB, Doty RT, Yang Z, et al. A heme export protein is required for red blood cell differentiation and iron homeostasis. Science. 2008;319(5864):825-828.

128. Quigley JG, Yang Z, Worthington MT, et al. Identification of a human heme exporter that is essential for erythropoiesis. Cell. 2004;118(6):757-766.

129. Alter BP, Giri N, Savage SA, et al. Malignancies and survival patterns in the National Cancer Institute inherited bone marrow failure syndromes cohort study. Br J Haematol. 2010;150(2):179-188.

130. Freedman MH, Alter BP. Risk of myelodysplastic syndrome and acute myeloid leukemia in congenital neutropenias. Semin Hematol. 2002;39(2):128-133.

131. Taskinen M, Kananen K, Valimaki M, et al. Risk factors for reduced areal bone mineral density in young adults with stem cell transplantation in childhood. Pediatr Transplant. 2006;10(1):90-97.

132 132. Grummitt CG, Townsley FM, Johnson CM, Warren AJ, Bycroft M. Structural consequences of nucleophosmin mutations in acute myeloid leukemia. J Biol Chem. 2008;283(34):23326-23332.

133. Naoe T, Suzuki T, Kiyoi H, Urano T. Nucleophosmin: a versatile molecule associated with hematological malignancies. Cancer Sci. 2006;97(10):963-969.

134. Patel JP, Gonen M, Figueroa ME, et al. Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med. 2012;366(12):1079-1089.

135. Grisendi S, Bernardi R, Rossi M, et al. Role of nucleophosmin in embryonic development and tumorigenesis. Nature. 2005;437(7055):147-153.

136. De Keersmaecker K, Atak ZK, Li N, et al. Exome sequencing identifies mutation in CNOT3 and ribosomal genes RPL5 and RPL10 in T-cell acute lymphoblastic leukemia. Nat Genet. 2013;45(2):186-190.

137. Rao S, Lee SY, Gutierrez A, et al. Inactivation of ribosomal protein L22 promotes transformation by induction of the stemness factor, Lin28B. Blood. 2012;120(18):3764-3773.

138. MacInnes AW, Amsterdam A, Whittaker CA, Hopkins N, Lees JA. Loss of p53 synthesis in zebrafish tumors with ribosomal protein gene mutations. Proc Natl Acad Sci U S A. 2008;105(30):10408-10413.

139. Gazda HT, Kho AT, Sanoudou D, et al. Defective ribosomal protein gene expression alters transcription, translation, apoptosis, and oncogenic pathways in Diamond-Blackfan anemia. Stem Cells. 2006;24(9):2034-2044.

140. Ruggero D, Pandolfi PP. Does the ribosome translate cancer? Nat Rev Cancer. 2003;3(3):179-192.

141. Raiser DM, Narla A, Ebert BL. The emerging importance of ribosomal dysfunction in the pathogenesis of hematologic disorders. Leuk Lymphoma. 2013.

142. Yang X, Boehm JS, Yang X, et al. A public genome-scale lentiviral expression library of human ORFs. Nat Methods. 2011;8(8):659-661.

133 143. Moffat J, Grueneberg DA, Yang X, et al. A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen. Cell. 2006;124(6):1283-1298.

144. Ong SE, Mann M. A practical recipe for stable isotope labeling by amino acids in cell culture (SILAC). Nat Protoc. 2006;1(6):2650-2660.

145. Bendall SC, Hughes C, Stewart MH, Doble B, Bhatia M, Lajoie GA. Prevention of amino acid conversion in SILAC experiments with embryonic stem cells. Mol Cell Proteomics. 2008;7(9):1587-1597.

146. Bustin SA, Benes V, Garson JA, et al. The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem. 2009;55(4):611-622.

147. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods. 2001;25(4):402-408.

148. Andersen CL, Jensen JL, Orntoft TF. Normalization of real-time quantitative reverse transcription-PCR data: a model-based variance estimation approach to identify genes suited for normalization, applied to bladder and colon cancer data sets. Cancer Res. 2004;64(15):5245- 5250.

149. Paffett-Lugassy NN, Zon LI. Analysis of hematopoietic development in the zebrafish. Methods Mol Med. 2005;105:171-198.

150. Tao W, Levine AJ. P19(ARF) stabilizes p53 by blocking nucleo-cytoplasmic shuttling of Mdm2. Proc Natl Acad Sci U S A. 1999;96(12):6937-6941.

151. Girnita L, Shenoy SK, Sehat B, et al. {beta}-Arrestin is crucial for ubiquitination and down-regulation of the insulin-like growth factor-1 receptor by acting as adaptor for the MDM2 E3 ligase. J Biol Chem. 2005;280(26):24412-24419.

152. Sehat B, Andersson S, Vasilcanu R, Girnita L, Larsson O. Role of ubiquitination in IGF- 1 receptor signaling and degradation. PLoS One. 2007;2(4):e340.

153. Miyagawa S, Kobayashi M, Konishi N, Sato T, Ueda K. Insulin and insulin-like growth factor I support the proliferation of erythroid progenitor cells in bone marrow through the sharing of receptors. Br J Haematol. 2000;109(3):555-562.

134 154. Ratajczak J, Zhang Q, Pertusini E, Wojczyk BS, Wasik MA, Ratajczak MZ. The role of insulin (INS) and insulin-like growth factor-I (IGF-I) in regulating human erythropoiesis. Studies in vitro under serum-free conditions--comparison to other cytokines and growth factors. Leukemia. 1998;12(3):371-381.

155. Ma Z, Dong A, Kong M, Qian J. Silencing of the type 1 insulin-like growth factor receptor increases the sensitivity to apoptosis and inhibits invasion in human lung adenocarcinoma A549 cells. Cell Mol Biol Lett. 2007;12(4):556-572.

156. Fargo JH, Kratz CP, Giri N, et al. Erythrocyte adenosine deaminase: diagnostic value for Diamond-Blackfan anaemia. Br J Haematol. 2013;160(4):547-554.

157. Danilova N, Bibikova E, Covey TM, et al. The role of the DNA damage response in zebrafish and cellular models of Diamond Blackfan anemia. Dis Model Mech. 2014;7(7):895- 905.

158. Donati G, Peddigari S, Mercer CA, Thomas G. 5S ribosomal RNA is an essential component of a nascent ribosomal precursor complex that regulates the Hdm2-p53 checkpoint. Cell Rep. 2013;4(1):87-98.

159. Sloan KE, Bohnsack MT, Watkins NJ. The 5S RNP couples p53 homeostasis to ribosome biogenesis and nucleolar stress. Cell Rep. 2013;5(1):237-247.

160. Zhang T, Shen H, Dong W, Qu X, Liu Q, Du J. Antitumor effects and molecular mechanisms of figitumumab, a humanized monoclonal antibody to IGF-1 receptor, in esophageal carcinoma. Sci Rep. 2014;4:6855.

161. Shenoy SK, McDonald PH, Kohout TA, Lefkowitz RJ. Regulation of receptor fate by ubiquitination of activated beta 2-adrenergic receptor and beta-arrestin. Science. 2001;294(5545):1307-1313.

162. Ratajczak MZ, Kuczynski WI, Onodera K, et al. A reappraisal of the role of insulin-like growth factor I in the regulation of human hematopoiesis. J Clin Invest. 1994;94(1):320-327.

163. Frostad S, Bjerknes R, Abrahamsen JF, Olweus J, Bruserud O. Insulin-like growth factor- 1 (IGF-1) has a costimulatory effect on proliferation of committed progenitors derived from human umbilical cord CD34+ cells. Stem Cells. 1998;16(5):334-342.

164. Aron DC. Insulin-like growth factor I and erythropoiesis. Biofactors. 1992;3(4):211-216.

135 165. Dorn I, Lazar-Karsten P, Boie S, et al. In vitro proliferation and differentiation of human CD34+ cells from peripheral blood into mature red blood cells with two different cell culture systems. Transfusion. 2008;48(6):1122-1132.

166. Malempati S, Weigel B, Ingle AM, et al. Phase I/II trial and pharmacokinetic study of cixutumumab in pediatric patients with refractory solid tumors and Ewing sarcoma: a report from the Children's Oncology Group. J Clin Oncol. 2012;30(3):256-262.

167. Schlueter PJ, Peng G, Westerfield M, Duan C. Insulin-like growth factor signaling regulates zebrafish embryonic growth and development by promoting cell survival and cell cycle progression. Cell Death Differ. 2007;14(6):1095-1105.

168. Girnita L, Worrall C, Takahashi S, Seregard S, Girnita A. Something old, something new and something borrowed: emerging paradigm of insulin-like growth factor type 1 receptor (IGF- 1R) signaling regulation. Cell Mol Life Sci. 2014;71(13):2403-2427.

169. Agrogiannis GD, Sifakis S, Patsouris ES, Konstantinidou AE. Insulin-like growth factors in embryonic and fetal growth and skeletal development (Review). Mol Med Rep. 2014;10(2):579-584.

170. Draptchinskaia N, Gustavsson P, Andersson B, et al. The gene encoding ribosomal protein S19 is mutated in Diamond-Blackfan anaemia. Nat Genet. 1999;21(2):169-175.

171. Mirabello L, Ballew, B.J., Giri, N., Jacobs, K.B., Yeager, M., Boland, J.F., Hicks, B.D., Burdett, L., Hutchinson, A.A., Alter, B.P., et al. RPS29 is Mutated in a Multi-Case Diamond Blackfan Anemia Family. American Society of Hematology; 2012.

172. Vlachos A, Muir E. How I treat Diamond Blackfan anemia. Blood. 2010;116(19):3715- 3723.

173. Leger-Silvestre I, Caffrey JM, Dawaliby R, et al. Specific Role for Yeast Homologs of the Diamond Blackfan Anemia-associated Rps19 Protein in Ribosome Synthesis. J Biol Chem. 2005;280(46):38177-38185.

174. Marygold SJ, Roote J, Reuter G, et al. The ribosomal protein genes and Minute loci of Drosophila melanogaster. Genome Biol. 2007;8(10):R216.

175. Zhu Y, Poyurovsky MV, Li Y, et al. Ribosomal protein S7 is both a regulator and a substrate of MDM2. Mol Cell. 2009;35(3):316-326.

136 176. Torihara H, Uechi T, Chakraborty A, Shinya M, Sakai N, Kenmochi N. Erythropoiesis failure due to RPS19 deficiency is independent of an activated Tp53 response in a zebrafish model of Diamond-Blackfan anaemia. Br J Haematol. 2011;152(5):648-654.

177. Danilova N, Sakamoto KM, Lin S. Ribosomal protein L11 mutation in zebrafish leads to haematopoietic and metabolic defects. Br J Haematol. 2011;152(2):217-228.

178. Sidi S, Sanda T, Kennedy RD, et al. Chk1 suppresses a caspase-2 apoptotic response to DNA damage that bypasses p53, Bcl-2, and caspase-3. Cell. 2008;133(5):864-877.

179. Imamura S, Uchiyama J, Koshimizu E, et al. A non-canonical function of zebrafish telomerase reverse transcriptase is required for developmental hematopoiesis. PLoS One. 2008;3(10):e3364.

180. Thisse C, Thisse B. High-resolution in situ hybridization to whole-mount zebrafish embryos. Nat Protoc. 2008;3(1):59-69.

181. Murphey RD, Stern HM, Straub CT, Zon LI. A chemical genetic screen for cell cycle inhibitors in zebrafish embryos. Chem Biol Drug Des. 2006;68(4):213-219.

182. Larionov A, Krause A, Miller W. A standard curve based method for relative real time PCR data processing. BMC Bioinformatics. 2005;6:62.

183. Macari ER, Lowrey CH. Induction of human fetal hemoglobin via the NRF2 antioxidant response signaling pathway. Blood. 2011;117(22):5987-5997.

184. Friedberg F, Taliaferro L. Calmodulin genes in zebrafish (revisited). Mol Biol Rep. 2005;32(1):55-60.

185. Ong SE, Blagoev B, Kratchmarova I, et al. Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol Cell Proteomics. 2002;1(5):376-386.

186. Arienti KL, Brunmark A, Axe FU, et al. Checkpoint kinase inhibitors: SAR and radioprotective properties of a series of 2-arylbenzimidazoles. J Med Chem. 2005;48(6):1873- 1885.

137 187. Sweitzer TD, Hanover JA. Calmodulin activates nuclear protein import: a link between signal transduction and nuclear transport. Proc Natl Acad Sci U S A. 1996;93(25):14574-14579.

188. Wagstaff KM, Jans DA. Importins and beyond: non-conventional nuclear transport mechanisms. Traffic. 2009;10(9):1188-1198.

189. Rodriguez-Vilarrupla A, Jaumot M, Abella N, et al. Binding of calmodulin to the carboxy-terminal region of p21 induces nuclear accumulation via inhibition of protein kinase C- mediated phosphorylation of Ser153. Mol Cell Biol. 2005;25(16):7364-7374.

190. Can G, Akpinar B, Baran Y, Zhivotovsky B, Olsson M. 5-Fluorouracil signaling through a calcium-calmodulin-dependent pathway is required for p53 activation and apoptosis in colon carcinoma cells. Oncogene. 2012.

191. Ou YH, Chung PH, Sun TP, Shieh SY. p53 C-terminal phosphorylation by CHK1 and CHK2 participates in the regulation of DNA-damage-induced C-terminal acetylation. Mol Biol Cell. 2005;16(4):1684-1695.

192. Takai H, Naka K, Okada Y, et al. Chk2-deficient mice exhibit radioresistance and defective p53-mediated transcription. EMBO J. 2002;21(19):5195-5205.

193. Mihara M, Erster S, Zaika A, et al. p53 has a direct apoptogenic role at the mitochondria. Mol Cell. 2003;11(3):577-590.

194. Marques LO, Lima MS, Soares BG. Trifluoperazine for schizophrenia. Cochrane Database Syst Rev. 2004(1):CD003545.

195. Ceccaldi R, Parmar K, Mouly E, et al. Bone marrow failure in Fanconi anemia is triggered by an exacerbated p53/p21 DNA damage response that impairs hematopoietic stem and progenitor cells. Cell Stem Cell. 2012;11(1):36-49.

196. MacLaren A, Slavin D, McGowan CH. Chk2 protects against radiation-induced genomic instability. Radiat Res. 2009;172(4):463-472.

197. Furgason JM, Bahassi EM. Targeting DNA repair mechanisms in cancer. Pharmacol Ther. 2013;137(3):298-308.

138 198. Cech TR. Structural biology. The ribosome is a ribozyme. Science. 2000;289(5481):878- 879.

199. Narla A, Ebert BL. Ribosomopathies: human disorders of ribosome dysfunction. Blood. 2010;115(16):3196-3205.

200. Farrar JE, Nater M, Caywood E, et al. Abnormalities of the large ribosomal subunit protein, Rpl35a, in Diamond-Blackfan anemia. Blood. 2008;112(5):1582-1592.

201. Nagarajan N, Bertrand D, Hillmer AM, et al. Whole-genome reconstruction and mutational signatures in gastric cancer. Genome Biol. 2012;13(12):R115.

202. Novetsky AP, Zighelboim I, Thompson DM, Jr., Powell MA, Mutch DG, Goodfellow PJ. Frequent mutations in the RPL22 gene and its clinical and functional implications. Gynecol Oncol. 2013;128(3):470-474.

203. Lai K, Amsterdam A, Farrington S, Bronson RT, Hopkins N, Lees JA. Many ribosomal protein mutations are associated with growth impairment and tumor predisposition in zebrafish. Dev Dyn. 2009;238(1):76-85.

204. Nilsson B, Johansson M, Al-Shahrour F, Carpenter AE, Ebert BL. Ultrasome: efficient aberration caller for copy number studies of ultra-high resolution. Bioinformatics. 2009;25(8):1078-1079.

205. Robinson JT, Thorvaldsdottir H, Winckler W, et al. Integrative genomics viewer. Nat Biotechnol. 2011;29(1):24-26.

206. Cancer Genome Atlas Research N, Genome Characterization C, Chang K, et al. The Cancer Genome Atlas Pan-Cancer analysis project. Nat Genet. 2013;45(10):1113-1120.

207. Beroukhim R, Mermel CH, Porter D, et al. The landscape of somatic copy-number alteration across human cancers. Nature. 2010;463(7283):899-905.

208. Miliani de Marval PL, Zhang Y. The RP-Mdm2-p53 pathway and tumorigenesis. Oncotarget. 2011;2(3):234-238.

209. Chen J, Kastan MB. 5'-3'-UTR interactions regulate p53 mRNA translation and provide a target for modulating p53 induction after DNA damage. Genes Dev. 2010;24(19):2146-2156.

139 210. Easton JB, Houghton PJ. mTOR and cancer therapy. Oncogene. 2006;25(48):6436-6446.

211. Ebert BL, Galili N, Tamayo P, et al. An erythroid differentiation signature predicts response to lenalidomide in myelodysplastic syndrome. PLoS Med. 2008;5(2):e35.

212. List A, Dewald G, Bennett J, et al. Lenalidomide in the myelodysplastic syndrome with chromosome 5q deletion. N Engl J Med. 2006;355(14):1456-1465.

213. Kamio T, Gu B-w, Olson TS, Zhang Y, Mason PJ, Bessler M. Mice with a Mutation in the Mdm2 Gene That Interferes with MDM2/Ribosomal Protein Binding Develop Bone Marrow Failure. Blood. 2014;124(21):4381-4381.

214. Quarello P, Garelli E, Carando A, et al. Diamond-Blackfan anemia: genotype-phenotype correlations in Italian patients with RPL5 and RPL11 mutations. Haematologica. 2010;95(2):206-213.

215. Correa PN, Eskinazi D, Axelrad AA. Circulating erythroid progenitors in polycythemia vera are hypersensitive to insulin-like growth factor-1 in vitro: studies in an improved serum-free medium. Blood. 1994;83(1):99-112.

216. Weinberg RS. In vitro erythropoiesis in polycythemia vera and other myeloproliferative disorders. Semin Hematol. 1997;34(1):64-69.

217. Michl P, Spoettl G, Engelhardt D, Weber MM. Alterations of the insulin-like growth factor system in patients with polycythemia vera. Mol Cell Endocrinol. 2001;181(1-2):189-197.

218. Mirza AM, Ezzat S, Axelrad AA. Insulin-like growth factor binding protein-1 is elevated in patients with polycythemia vera and stimulates erythroid burst formation in vitro. Blood. 1997;89(6):1862-1869.

219. Tsai PH, Arkin S, Lipton JM. An intrinsic progenitor defect in Diamond-Blackfan anaemia. Br J Haematol. 1989;73(1):112-120.

220. Girnita L, Shenoy SK, Sehat B, et al. Beta-arrestin and Mdm2 mediate IGF-1 receptor- stimulated ERK activation and cell cycle progression. J Biol Chem. 2007;282(15):11329-11338.

221. Samani AA, Yakar S, LeRoith D, Brodt P. The role of the IGF system in cancer growth and metastasis: overview and recent insights. Endocr Rev. 2007;28(1):20-47.

140 222. Froment P, Dupont J, Christophe-Marine J. Mdm2 exerts pro-apoptotic activities by antagonizing insulin-like growth factor-I-mediated survival. Cell Cycle. 2008;7(19):3098-3103.

223. Mayo LD, Donner DB. A phosphatidylinositol 3-kinase/Akt pathway promotes translocation of Mdm2 from the cytoplasm to the nucleus. Proc Natl Acad Sci U S A. 2001;98(20):11598-11603.

224. Jackson MW, Patt LE, LaRusch GA, Donner DB, Stark GR, Mayo LD. Hdm2 nuclear export, regulated by insulin-like growth factor-I/MAPK/p90Rsk signaling, mediates the transformation of human cells. J Biol Chem. 2006;281(24):16814-16820.

225. Joseph TW, Zaika A, Moll UM. Nuclear and cytoplasmic degradation of endogenous p53 and HDM2 occurs during down-regulation of the p53 response after multiple types of DNA damage. FASEB J. 2003;17(12):1622-1630.

226. O'Keefe K, Li H, Zhang Y. Nucleocytoplasmic shuttling of p53 is essential for MDM2- mediated cytoplasmic degradation but not ubiquitination. Mol Cell Biol. 2003;23(18):6396-6405.

227. Sasaki M, Kawahara K, Nishio M, et al. Regulation of the MDM2-P53 pathway and tumor growth by PICT1 via nucleolar RPL11. Nat Med. 2011;17(8):944-951.

228. Shen X, Valencia CA, Szostak JW, Dong B, Liu R. Scanning the human proteome for calmodulin-binding proteins. Proc Natl Acad Sci U S A. 2005;102(17):5969-5974.

229. Bialik S, Kimchi A. The death-associated protein kinases: structure, function, and beyond. Annu Rev Biochem. 2006;75:189-210.

230. Meek DW, Anderson CW. Posttranslational modification of p53: cooperative integrators of function. Cold Spring Harb Perspect Biol. 2009;1(6):a000950.

231. Yeh CT, Wu AT, Chang PM, et al. Trifluoperazine, an antipsychotic agent, inhibits cancer stem cell growth and overcomes drug resistance of lung cancer. Am J Respir Crit Care Med. 2012;186(11):1180-1188.

232. Sulima SO, Patchett S, Advani VM, De Keersmaecker K, Johnson AW, Dinman JD. Bypass of the pre-60S ribosomal quality control as a pathway to oncogenesis. Proc Natl Acad Sci U S A. 2014;111(15):5640-5645.

141 233. De Keersmaecker K, Sulima SO, Dinman JD. Ribosomopathies and the paradox of cellular hypo- to hyperproliferation. Blood. 2015;125(9):1377-1382.

234. Tzoneva G, Perez-Garcia A, Carpenter Z, et al. Activating mutations in the NT5C2 nucleotidase gene drive chemotherapy resistance in relapsed ALL. Nat Med. 2013;19(3):368-371.

235. Nijhawan D, Zack TI, Ren Y, et al. Cancer vulnerabilities unveiled by genomic loss. Cell. 2012;150(4):842-854.

236. Muller FL, Colla S, Aquilanti E, et al. Passenger deletions generate therapeutic vulnerabilities in cancer. Nature. 2012;488(7411):337-342.

237. Domcke S, Sinha R, Levine DA, Sander C, Schultz N. Evaluating cell lines as tumour models by comparison of genomic profiles. Nat Commun. 2013;4:2126.

142 Appendix: Supplementary Tables

Supplementary Table 2.1. SILAC IP/MS proteomics results for sh-RPS14 vs. background.

Condition B vs. condition A as indicated in Figure 1.1. The top 100 MDM2-interacting proteins, ranked by log2 MQ values, are included in the table. Additional proteins referred to in the text of Chapter 1 are listed at the bottom of the table.

Log2 MQ Log2 MQ Norm M/L Norm H/M p- Rank Gene Name Protein Name Rep1 Rep2 value 1 TP53 Cellular tumor antigen p53 3.7248 4.3072 0.0003 2 MDM2 E3 ubiquitin-protein ligase Mdm2 2.1603 3.7723 0.0095 3 MDM4 Protein Mdm4 1.7640 2.1607 0.0009 4 RPL11 60S ribosomal protein L11 0.9548 1.2422 0.0026 5 FHL1 Four and a half LIM domains protein 1 0.8910 0.8359 0.0017 NADP-dependent malic enzyme, 6 ME3 mitochondrial 0.8281 0.9431 0.0019 7 MAGT1 Magnesium transporter protein 1 0.8279 0.3876 0.0294 8 RPL5 0.7898 1.1846 0.0063 Phosphoribosyltransferase domain-containing 9 PRTFDC1 protein 1 0.7125 0.8062 0.0027 10 RPL23 60S ribosomal protein L23 0.6760 0.9629 0.0060 Tetratricopeptide repeat protein 19, 11 TTC19 mitochondrial 0.6337 0.5210 0.0064 12 ISYNA1 Inositol-3-phosphate synthase 1 0.6253 0.6615 0.0037 13 SH3PXD2B SH3 and PX domain-containing protein 2B 0.6176 0.4141 0.0137 Putative ribosome-binding factor A, 14 RBFA mitochondrial 0.5812 0.6260 0.0046 15 FKBP11 Peptidyl-prolyl cis-trans isomerase FKBP11 0.5715 0.2603 0.0423 Bifunctional arginine demethylase and lysyl- 16 JMJD6 hydroxylase JMJD6 0.5380 0.3333 0.0218 17 SDPR Serum deprivation-response protein 0.5283 0.4874 0.0074 18 AKAP12 A-kinase anchor protein 12 0.5211 0.7906 0.0101 Growth hormone-inducible transmembrane 19 GHITM protein 0.5156 1.5417 0.0496 20 HMGB3 High mobility group protein B3 0.5105 0.3639 0.0162 21 MID1 Midline-1 0.5044 0.8007 0.0119 22 NXN Nucleoredoxin 0.4989 0.3963 0.0126 23 SART1 U4/U6.U5 tri-snRNP-associated protein 1 0.4830 0.8952 0.0185 24 IGF1R Insulin-like growth factor 1 receptor 0.4763 0.5529 0.0078 25 MVD Diphosphomevalonate decarboxylase 0.4678 0.3596 0.0159 26 DPYSL4 Dihydropyrimidinase-related protein 4 0.4677 0.7212 0.0122 Major facilitator superfamily domain- 27 MFSD10 containing protein 10 0.4670 0.2542 0.0372 Probable DNA dC->dU-editing enzyme 28 APOBEC3C APOBEC-3C 0.4616 0.2442 0.0402

144 Supplementary Table 2.1 (continued)

Log2 MQ Log2 MQ Norm M/L Norm H/M p- Rank Gene Name Protein Name Rep1 Rep2 value 29 CTNNB1 Catenin beta-1 0.4586 0.3345 0.0191 30 PVRL2 Poliovirus receptor-related protein 2 0.4544 0.2550 0.0363 Chaperone activity of bc1 complex-like, 31 ADCK3;CABC1 mitochondrial 0.4502 0.7891 0.0172 32 RBMX RNA-binding motif protein, X chromosome 0.4286 0.3215 0.0207 33 HIST1H1B Histone H1.5 0.4242 0.5041 0.0105 34 GEMIN2 Gem-associated protein 2 0.4216 0.2703 0.0308 35 CBS Cystathionine beta-synthase 0.4194 0.4934 0.0108 36 CRNKL1 Crooked neck-like protein 1 0.4138 0.4893 0.0112 37 MPG;PIG16 DNA-3-methyladenine glycosylase 0.4101 0.2768 0.0290 Na(+)/H(+) exchange regulatory cofactor 38 SLC9A3R2 NHE-RF2 0.4058 0.6774 0.0173 39 DNAJC12 DnaJ homolog subfamily C member 12 0.4023 0.8234 0.0264 40 UBE2D1 Ubiquitin-conjugating enzyme E2 D1 0.4016 0.3889 0.0144 41 MYO1C Unconventional myosin-Ic 0.3991 0.3460 0.0179 Guanine nucleotide-binding protein subunit 42 GNA11 alpha-11 0.3955 0.4571 0.0126 43 BOP1;KM-PA-2 Ribosome biogenesis protein BOP1 0.3902 0.2385 0.0394 Probable E3 ubiquitin-protein ligase makorin- 44 MKRN2 2 0.3775 0.3098 0.0230 45 ELAVL1 ELAV-like protein 1 0.3740 0.3250 0.0210 46 ITGB5 Integrin beta-5;Integrin beta 0.3702 0.2618 0.0325 47 TAF7 Transcription TFIID subunit 7 0.3695 0.7241 0.0259 48 ADD2 Beta-adducin 0.3606 0.2177 0.0467 49 LEPREL1 Prolyl 3-hydroxylase 2 0.3601 0.3579 0.0186 50 MAPK3 Mitogen-activated protein kinase 3 0.3576 0.2890 0.0270 51 ARPC1A Actin-related protein 2/3 complex subunit 1A 0.3514 0.6197 0.0228 52 RNASEH2A Ribonuclease H2 subunit A 0.3487 0.2582 0.0339 53 TRIP6 Thyroid receptor-interacting protein 6 0.3463 0.7619 0.0334 54 PTRF Polymerase I and transcript 0.3456 0.2915 0.0271 55 ITPK1 Inositol-tetrakisphosphate 1-kinase 0.3446 0.9194 0.0456 56 SNX9 Sorting nexin-9 0.3434 0.6317 0.0251 57 FRG1 Protein FRG1 0.3391 0.3584 0.0203 Phospholysine phosphohistidine inorganic 58 LHPP pyrophosphate phosphatase 0.3379 0.3166 0.0240 59 PRNP Major prion protein 0.3354 0.3060 0.0256 60 FAM96A MIP18 family protein FAM96A 0.3333 0.2773 0.0304 61 ACAT2 Acetyl-CoA acetyltransferase, cytosolic 0.3331 0.3146 0.0247

145 Supplementary Table 2.1 (continued)

Log2 MQ Log2 MQ Norm M/L Norm H/M p- Rank Gene Name Protein Name Rep1 Rep2 value Deoxyuridine 5-triphosphate 62 DUT nucleotidohydrolase, mitochondrial 0.3326 0.2855 0.0290 63 GALM Aldose 1-epimerase 0.3308 0.3203 0.0242 64 MMTAG2 Multiple myeloma tumor-associated protein 2 0.3275 0.5335 0.0227 65 SF3B5 Splicing factor 3B subunit 5 0.3185 0.2358 0.0416 66 PPIL4 Peptidyl-prolyl cis-trans isomerase-like 4 0.3127 0.2563 0.0365 67 PROCR Endothelial protein C receptor 0.3108 0.2843 0.0310 68 TRIM28 Transcription intermediary factor 1-beta 0.3081 0.2574 0.0366 69 MYO9B Unconventional myosin-IXb 0.3066 0.6202 0.0325 70 EEF1A2 1-alpha 2 0.3052 0.2404 0.0413 71 SRI Sorcin 0.3003 0.2505 0.0391 72 ALDH3B1 Aldehyde dehydrogenase family 3 member B1 0.2977 0.3200 0.0280 73 HIST1H3A Histone H3.1 0.2941 0.3254 0.0281 74 CCDC165;SOGA2 Coiled-coil domain-containing protein 165 0.2929 0.4714 0.0270 75 EXOC7 Exocyst complex component 7 0.2922 0.6518 0.0388 76 HNRNPA3 Heterogeneous nuclear ribonucleoprotein A3 0.2910 0.2366 0.0439 77 SDCBP Syntenin-1 0.2909 0.2984 0.0314 78 EXOSC8 Exosome complex component RRP43 0.2752 0.2616 0.0399 79 C7orf50 Uncharacterized protein C7orf50 0.2744 0.4015 0.0295 80 ALCAM CD166 antigen 0.2708 0.3694 0.0304 81 CCDC88A Girdin 0.2702 0.2914 0.0358 82 ACTC1 Actin, alpha cardiac muscle 1 0.2682 0.4285 0.0315 83 NASP Nuclear autoantigenic sperm protein 0.2671 0.2610 0.0414 84 C3orf26 Uncharacterized protein C3orf26 0.2627 0.3428 0.0331 85 CDC42BPB Serine/threonine-protein kinase MRCK beta 0.2575 0.5249 0.0397 86 CHMP1A Charged multivesicular body protein 1a 0.2548 0.4796 0.0377 87 RPS27 40S ribosomal protein S27 0.2522 0.3007 0.0386 Synaptic vesicle membrane protein VAT-1 88 VAT1 homolog 0.2521 0.2553 0.0457 89 C15orf38 UPF0552 protein C15orf38 0.2501 0.4561 0.0379 Interferon-induced protein with 90 IFIT5 tetratricopeptide repeats 5 0.2451 0.3333 0.0381 40S ribosomal protein S27;40S ribosomal 91 RPS27L protein S27-like 0.2434 0.4664 0.0407 92 GET4 Golgi to ER traffic protein 4 homolog 0.2432 0.2675 0.0455 93 EXOSC3 Exosome complex component RRP40 0.2431 0.3752 0.0378 LIM and calponin homology domains- 94 LIMCH1 containing protein 1 0.2399 0.3884 0.0389

146 Supplementary Table 2.1 (continued)

Log2 MQ Log2 MQ Norm M/L Norm H/M p- Rank Gene Name Protein Name Rep1 Rep2 value Discoidin, CUB and LCCL domain-containing 95 DCBLD2 protein 2 0.2368 0.5274 0.0472 96 HNRNPH3 Heterogeneous nuclear ribonucleoprotein H3 0.2366 0.2990 0.0430 97 RPS25 40S ribosomal protein S25 0.2324 0.2693 0.0482 98 CNN2 Calponin-2 0.2275 0.3215 0.0442 99 HSDL2 Hydroxysteroid dehydrogenase-like protein 2 0.2274 0.3151 0.0446 Ubiquitin carboxyl-terminal hydrolase 100 UCHL3 isozyme L3 0.2258 0.2891 0.0475 411 RPL38 60S ribosomal protein L38 0.2714 0.8083 0.0583 1477 RPL26 60S ribosomal protein L26 0.0303 0.1701 0.3360

147 Supplementary Table 2.2. SILAC IP/MS proteomics results for sh-LUC vs. background.

Condition C vs. condition A as indicated in Figure 1.1. The top 100 MDM2-interacting proteins, ranked by log2 MQ values, are included in the table. Additional proteins referred to in the text of Chapter 1 are listed at the bottom of the table.

Log2 MQ Log2 MQ Norm H/L Norm L/M p- Rank Gene Name Protein Name Rep1 Rep2 value 1 SERPINB7 Serpin B7 4.0853 3.1791 0.0045 2 MDM4 Protein Mdm4 3.7172 2.6724 0.0082 3 RPL22L1 60S ribosomal protein L22-like 1 2.8288 1.8438 0.0145 4 MDM2 E3 ubiquitin-protein ligase Mdm2 2.3007 3.5572 0.0149 5 LEMD2 LEM domain-containing protein 2 1.8927 1.6034 0.0023 6 SPTLC2 Serine palmitoyltransferase 2 1.8387 1.0280 0.0279 7 MFGE8 Lactadherin;Lactadherin short form;Medin 1.8335 1.6929 0.0008 8 DNAJC12 DnaJ homolog subfamily C member 12 1.6713 2.0073 0.0027 NADP-dependent malic enzyme, 9 ME3 mitochondrial;Malic enzyme 1.6369 1.5047 0.0010 10 HIST1H3A Histone H3.1 1.5850 2.3691 0.0129 11 HIST1H4A Histone H4 1.5327 2.4427 0.0175 12 EMD Emerin 1.5281 1.4815 0.0007 13 AXL -protein kinase receptor UFO 1.5030 1.0815 0.0092 14 CD59 CD59 glycoprotein 1.4894 1.8200 0.0033 Heterogeneous nuclear ribonucleoproteins 15 HNRNPC C1/C2 1.4651 1.4885 0.0007 16 HMOX1 Heme oxygenase 1 1.4269 1.7043 0.0027 17 DPYSL5 Dihydropyrimidinase-related protein 5 1.3733 2.3485 0.0234 18 TP53 Cellular tumor antigen p53 1.3231 2.3803 0.0280 19 BANF1 Barrier-to-autointegration factor 1.2832 1.7748 0.0086 20 HIST1H2AG Histone H2A type 1 1.2553 2.1956 0.0255 Ribonucleoside-diphosphate reductase subunit 21 RRM2 M2 1.2485 1.1448 0.0015 22 LMNB1 Lamin-B1 1.2440 1.3580 0.0014 23 SH3PXD2B SH3 and PX domain-containing protein 2B 1.2261 0.8445 0.0125 Pyrroline-5-carboxylate reductase 2;Pyrroline- 24 PYCR2 5-carboxylate reductase 1.1407 0.8476 0.0085 25 RPS29 40S ribosomal protein S29 1.1388 0.9367 0.0043 26 GRN Granulins;Acrogranin 1.1364 1.1304 0.0012 Bifunctional methylenetetrahydrofolate 27 MTHFD2 dehydrogenase/cyclohydrolase, mitochondrial 1.1251 0.8592 0.0072 40S ribosomal protein S27;40S ribosomal 28 RPS27L protein S27-like 1.1176 1.4764 0.0067 29 GDI1 Rab GDP dissociation inhibitor alpha 1.1106 1.0640 0.0014 30 ADPRHL2 Poly(ADP-ribose) glycohydrolase ARH3 1.1104 0.8539 0.0069 Myosin regulatory light chain 12B;Myosin 31 MYL12B;MYL12A regulatory light chain 12A 1.1039 1.6918 0.0151 32 DDX24 ATP-dependent RNA helicase DDX24 1.0998 0.8268 0.0080 Guanine nucleotide-binding protein 33 GNG12 G(I)/G(S)/G(O) subunit gamma-12 1.0535 1.2758 0.0038 34 CPNE1 Copine-1 1.0454 0.8331 0.0057

148 Supplementary Table 2.2 (continued) Log2 MQ Log2 MQ Norm H/L Norm L/M p- Rank Gene Name Protein Name Rep1 Rep2 value Guanine nucleotide-binding protein G(i) 35 GNAI1 subunit alpha-1 1.0453 1.2253 0.0030 Guanine nucleotide-binding protein G(i) 36 GNAI2 subunit alpha-2 1.0143 1.1150 0.0020 Pyrroline-5-carboxylate reductase;Pyrroline-5- 37 PYCR1 carboxylate reductase 1, mitochondrial 1.0125 0.8759 0.0033 Transcriptional repressor protein 38 YY1;YY2 YY1;Transcription factor YY2 1.0069 0.8022 0.0059 39 HIST1H2BJ Histone H2B type 1-J 0.9933 1.8701 0.0329 40 CBX3 Chromobox protein homolog 3 0.9888 1.1836 0.0037 41 MYL6 Myosin light polypeptide 6 0.9846 1.6953 0.0246 42 GPRC5A Retinoic acid-induced protein 3 0.9835 1.3057 0.0073 43 LMNB2 Lamin-B2 0.9800 0.4747 0.0458 44 NUCB1 Nucleobindin-1 0.9688 0.9558 0.0017 Deoxyuridine 5-triphosphate 45 DUT nucleotidohydrolase, mitochondrial 0.9638 0.8244 0.0038 Serine beta-lactamase-like protein LACTB, 46 LACTB mitochondrial 0.9614 1.0469 0.0021 47 HIST1H2BK Histone H2B type 1-K 0.9600 2.0629 0.0469 48 RBMX RNA-binding motif protein, X chromosome 0.9595 1.0559 0.0022 49 YTHDF3 YTH domain family protein 3 0.9501 1.1051 0.0031 50 ABCF1 ATP-binding cassette sub-family F member 1 0.9326 0.8642 0.0024 51 CBS Cystathionine beta-synthase 0.9321 1.4850 0.0184 Mitotic spindle-associated MMXD complex 52 FAM96B subunit MIP18 0.9307 1.1134 0.0039 53 LAMC1 Laminin subunit gamma-1 0.9290 1.1163 0.0040 54 RPL23 60S ribosomal protein L23 0.9223 0.8961 0.0020 55 ISYNA1 Inositol-3-phosphate synthase 1 0.9182 0.8499 0.0025 56 ITGB5 Integrin beta-5;Integrin beta 0.9175 0.9651 0.0020 57 MYO10 Unconventionnal myosin-X 0.9081 1.0447 0.0031 58 RPS14 40S ribosomal protein S14 0.8988 0.7920 0.0035 CD97 antigen;CD97 antigen subunit 59 CD97 alpha;CD97 antigen subunit beta 0.8934 0.6916 0.0078 Tetratricopeptide repeat protein 19, 60 TTC19 mitochondrial 0.8913 0.8217 0.0027 61 MRPL4 39S ribosomal protein L4, mitochondrial 0.8822 0.7389 0.0049 62 MRTO4 mRNA turnover protein 4 homolog 0.8753 0.6005 0.0147 Phosphoribosyltransferase domain-containing 63 PRTFDC1 protein 1 0.8749 0.7871 0.0032 64 SNX9 Sorting nexin-9 0.8714 0.5425 0.0220 65 KLHL22 Kelch-like protein 22 0.8596 0.8699 0.0022 Guanine nucleotide-binding protein 66 GNB2 G(I)/G(S)/G(T) subunit beta-2 0.8577 0.8773 0.0022 MOB kinase activator 1A;MOB kinase 67 MOB1A;MOB1B activator 1B 0.8577 0.7110 0.0055 68 ACLY ATP-citrate synthase 0.8576 0.7773 0.0032 69 VIM Vimentin 0.8546 0.6096 0.0125 70 RPS8 0.8470 0.6801 0.0067 71 RPS23 40S ribosomal protein S23 0.8441 0.7512 0.0036 72 RPS13 40S ribosomal protein S13 0.8436 0.7622 0.0034

149 Supplementary Table 2.2 (continued) Log2 MQ Log2 MQ Norm H/L Norm L/M p- Rank Gene Name Protein Name Rep1 Rep2 value 73 RPS11 40S ribosomal protein S11 0.8384 0.8596 0.0023 74 POLDIP3;PDIP46 Polymerase delta-interacting protein 3 0.8368 0.9262 0.0028 75 RPS3 0.8360 0.7937 0.0027 76 RPS4X 40S ribosomal protein S4, X isoform 0.8336 0.7402 0.0038 77 RPS20 40S ribosomal protein S20 0.8333 0.7581 0.0033 78 RPS16 40S ribosomal protein S16 0.8307 0.7135 0.0046 79 POLD3 DNA polymerase delta subunit 3 0.8272 0.8674 0.0024 Guanine nucleotide-binding protein G(k) 80 GNAI3 subunit alpha 0.8229 0.7606 0.0031 81 NXN Nucleoredoxin 0.8211 0.8355 0.0024 82 RPS27 40S ribosomal protein S27 0.8078 0.9439 0.0039 83 RPS15A 40S ribosomal protein S15a 0.7996 0.7477 0.0031 84 SNX1 Sorting nexin-1 0.7949 0.6763 0.0052 85 FAU 40S ribosomal protein S30 0.7881 0.6869 0.0046 86 PHF6 PHD finger protein 6 0.7877 0.3863 0.0467 87 FLOT2 Flotillin-2 0.7843 0.4911 0.0226 88 KPNA2 Importin subunit alpha-2 0.7825 0.7839 0.0027 89 RPS9 0.7806 0.7411 0.0031 90 RPS3A 40S ribosomal protein S3a 0.7806 0.6817 0.0046 91 RPS18 40S ribosomal protein S18 0.7806 0.7602 0.0029 92 TRIM28 Transcription intermediary factor 1-beta 0.7798 0.6681 0.0051 93 RPS24 40S ribosomal protein S24 0.7780 0.6419 0.0063 94 USP11 Ubiquitin carboxyl-terminal hydrolase 11 0.7742 0.6454 0.0060 95 MID1 Midline-1 0.7716 0.6910 0.0041 96 PNO1 RNA-binding protein PNO1 0.7686 0.7790 0.0028 ;40S ribosomal 97 RPS5 protein S5, N-terminally processed 0.7677 0.6213 0.0072 98 RPS26 40S ribosomal protein S26 0.7640 0.6088 0.0078 99 LARP4B La-related protein 4B 0.7606 1.0476 0.0101 100 ERH Enhancer of rudimentary homolog 0.7586 0.6190 0.0070 115 RPL11 60S ribosomal protein L11 0.7154 0.6862 0.0036 138 RPL38 60S ribosomal protein L38 0.6408 0.6888 0.0044 150 RPL5 60S ribosomal protein L5 0.6149 0.5942 0.0050 159 RPS7 0.5987 0.6243 0.0049 1520 RPL26 60S ribosomal protein L26 0.0872 0.1934 0.1514

150 Supplementary Table 2.3. SILAC IP/MS proteomics results for sh-RPS14 vs. sh-LUC (upper right quadrant).

Condition B vs. condition C as indicated in Figure 1.1. The top 100 proteins preferentially bound to MDM2 in the context of RPS14 deficiency, ranked by log2 MQ values, are included in the table. Additional proteins referred to in the text of Chapter 1 are listed at the bottom of the table.

Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 1 TP53 Cellular tumor antigen p53 2.2443 2.2208 0.0002 2 NOC4L Nucleolar complex protein 4 homolog 1.7168 1.5737 0.0007 3 FTH1 Ferritin heavy chain;Ferritin 1.1976 1.5226 0.0032

4 FDXR NADPH:adrenodoxin oxidoreductase, mitochondrial 1.1873 0.9750 0.0030 5 DGKA Diacylglycerol kinase alpha 1.1648 0.5750 0.0319 6 TP53I3 Quinone oxidoreductase PIG3 1.0229 0.7884 0.0053 7 CTPS CTP synthase 1 1.0085 1.0644 0.0014 8 DPYSL4 Dihydropyrimidinase-related protein 4 0.9739 1.2096 0.0033 9 ACTN4 Alpha-actinin-4 0.9700 0.8082 0.0036 10 MAP7 Ensconsin 0.9642 0.7945 0.0039 11 DICER1 Endoribonuclease Dicer 0.9418 0.6942 0.0074 12 CAPZA2 F-actin-capping protein subunit alpha-2 0.9374 0.8307 0.0027 13 NCOR2 Nuclear receptor corepressor 2 0.9027 0.5920 0.0133 14 MOB2 MOB kinase activator 2 0.8994 0.5278 0.0205 15 CORO1C Coronin-1C 0.8987 0.8298 0.0024

16 ALDH3A1 Aldehyde dehydrogenase, dimeric NADP-preferring 0.8646 0.5129 0.0202 17 EDC4 Enhancer of mRNA-decapping protein 4 0.8398 0.5745 0.0118

18 RRM2B Ribonucleoside-diphosphate reductase subunit M2 B 0.8368 0.3698 0.0464 19 HSPA4L Heat shock 70 kDa protein 4L 0.8367 0.3612 0.0488 20 FBXO22 F-box only protein 22 0.8271 0.8085 0.0024 21 DDB2 DNA damage-binding protein 2 0.8116 0.4276 0.0304 22 CAPZB F-actin-capping protein subunit beta 0.7952 0.7589 0.0028 23 STK4 Serine/threonine-protein kinase 4 0.7927 0.5843 0.0090

24 ASCC3 Activating signal cointegrator 1 complex subunit 3 0.7595 0.5059 0.0145

25 APOBEC3C Probable DNA dC->dU-editing enzyme APOBEC-3C 0.7504 0.5806 0.0078 26 EFHD2 EF-hand domain-containing protein D2 0.7448 1.0229 0.0073 27 HEATR1 HEAT repeat-containing protein 1 0.7298 0.6060 0.0059 Serine hydroxymethyltransferase, cytosolic;Serine 28 SHMT1 hydroxymethyltransferase 0.7249 0.3943 0.0296

151

Supplementary Table 2.3 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 29 SRA1 Steroid receptor RNA activator 1 0.7230 0.5610 0.0081

30 DDX49 Probable ATP-dependent RNA helicase DDX49 0.7215 0.7950 0.0034 31 KLHDC10 Kelch domain-containing protein 10 0.7179 1.0022 0.0081 32 BAX Apoptosis regulator BAX 0.7054 0.5118 0.0112 33 ENDOG Endonuclease G, mitochondrial 0.6994 0.5785 0.0065 34 ITGB4 Integrin beta-4 0.6908 0.3886 0.0278 35 ACTN1 Alpha-actinin-1 0.6651 0.6602 0.0041 36 HSPA12A Heat shock 70 kDa protein 12A 0.6643 0.6775 0.0040 37 IGF1R Insulin-like growth factor 1 receptor 0.6624 0.5420 0.0075 Epidermal growth factor receptor kinase substrate 8- 38 EPS8L2 like protein 2 0.6500 0.5423 0.0072 39 TIGAR Probable fructose-2,6-bisphosphatase TIGAR 0.6447 0.5746 0.0058 40 PGM2L1 Glucose 1,6-bisphosphate synthase 0.6372 1.0996 0.0195 41 FRG1 Protein FRG1 0.6357 0.3136 0.0415 42 ARRB2 Beta-arrestin-2 0.6304 0.5127 0.0084 43 STOM Erythrocyte band 7 integral membrane protein 0.6088 0.3622 0.0263 44 CMBL Carboxymethylenebutenolidase homolog 0.6080 0.6630 0.0049 45 LMCD1 LIM and cysteine-rich domains protein 1 0.6015 0.6979 0.0055 46 PLXNB2 Plexin-B2 0.5966 0.4966 0.0086 47 TNKS1BP1 182 kDa tankyrase-1-binding protein 0.5951 0.4448 0.0127 48 THUMPD1 THUMP domain-containing protein 1 0.5880 0.6172 0.0053 49 TP53I11 Tumor protein p53-inducible protein 11 0.5776 0.3525 0.0257 50 CAPZA1 F-actin-capping protein subunit alpha-1 0.5695 0.4876 0.0086 51 NTPCR;C1orf57 Cancer-related nucleoside-triphosphatase 0.5590 0.6300 0.0062 52 TROVE2 60 kDa SS-A/Ro ribonucleoprotein 0.5501 0.3262 0.0294 53 CPD Carboxypeptidase D 0.5475 0.4541 0.0104

54 IGF2R Cation-independent mannose-6-phosphate receptor 0.5421 0.5090 0.0075 55 RDX Radixin 0.5331 0.6377 0.0074 56 IGBP1 Immunoglobulin-binding protein 1 0.5289 0.5559 0.0069 57 PDLIM1 PDZ and LIM domain protein 1 0.5287 0.4632 0.0095 58 DENND4C DENN domain-containing protein 4C 0.5268 0.3391 0.0246 59 RSU1 Ras suppressor protein 1 0.5141 0.3689 0.0184 60 HSPB8 Heat shock protein beta-8 0.4947 0.7838 0.0173 61 TPT1 Translationally-controlled tumor protein 0.4934 0.6333 0.0099 62 EEF1G Elongation factor 1-gamma 0.4880 0.4247 0.0116 63 EPB41L2 Band 4.1-like protein 2 0.4835 0.3372 0.0221

152

Supplementary Table 2.3 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 64 C11orf73 Protein Hikeshi 0.4818 0.9728 0.0334 65 SUPT5H Transcription elongation factor SPT5 0.4779 0.4287 0.0113 Transmembrane and ubiquitin-like domain-containing 66 TMUB1 protein 1 0.4759 0.2553 0.0437 Caspase-6;Caspase-6 subunit p18;Caspase-6 subunit 67 CASP6 p11 0.4718 0.5786 0.0098 68 KIAA1598 Shootin-1 0.4578 0.3631 0.0170 69 LACTB2 Beta-lactamase-like protein 2 0.4533 0.4106 0.0125 70 VPS13C Vacuolar protein sorting-associated protein 13C 0.4525 0.3731 0.0157 Ribosomal RNA small subunit methyltransferase 71 EMG1 NEP1 0.4495 0.4828 0.0101 72 EEF1B2 Elongation factor 1-beta 0.4450 0.4928 0.0103 73 PDIA4 Protein disulfide-isomerase A4 0.4435 0.3146 0.0243 74 GGPS1 Geranylgeranyl pyrophosphate synthase 0.4411 0.7162 0.0203

75 RBM4;RBM4B RNA-binding protein 4;RNA-binding protein 4B 0.4357 0.3052 0.0258 76 ERMP1 Endoplasmic reticulum metallopeptidase 1 0.4351 0.6931 0.0196 77 RAB11FIP1 Rab11 family-interacting protein 1 0.4293 0.3387 0.0196 78 TMX3 Protein disulfide-isomerase TMX3 0.4237 0.4223 0.0125 79 SUPT16H FACT complex subunit SPT16 0.4223 0.4315 0.0122

80 DRG2 Developmentally-regulated GTP-binding protein 2 0.4223 0.4355 0.0121 81 AP1M1 AP-1 complex subunit mu-1 0.4191 0.2817 0.0305 82 AK1 Adenylate kinase isoenzyme 1 0.4181 0.3914 0.0142

83 RBFA Putative ribosome-binding factor A, mitochondrial 0.4146 0.3716 0.0157 84 CDC42 Cell division control protein 42 homolog 0.4123 0.3765 0.0153 85 EIF2C2 Protein argonaute-2 0.4060 0.4510 0.0128 86 PARN Poly(A)-specific ribonuclease PARN 0.4002 0.3942 0.0145 87 PLIN2 Perilipin-2 0.3997 0.5409 0.0155 88 FAM114A2 Protein FAM114A2 0.3925 0.3816 0.0155

89 EIF3G initiation factor 3 subunit G 0.3924 0.2878 0.0278 90 CNN2 Calponin-2 0.3923 0.5381 0.0163 91 VCL Vinculin 0.3906 0.4527 0.0140 92 ERAP1 Endoplasmic reticulum aminopeptidase 1 0.3888 0.3456 0.0185 93 FUBP3 Far upstream element-binding protein 3 0.3879 0.2438 0.0402 94 PDCD2L Programmed cell death protein 2-like 0.3866 0.4102 0.0147 95 RAB8B Ras-related protein Rab-8B 0.3844 0.5285 0.0169 96 RGS10 Regulator of G-protein signaling 10 0.3841 0.3877 0.0156

153 Supplementary Table 2.3 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 97 ADA Adenosine deaminase 0.3841 0.2700 0.0319

98 EML1 Echinoderm microtubule-associated protein-like 1 0.3829 0.6616 0.0265 Integrin alpha-3;Integrin alpha-3 heavy chain;Integrin 99 ITGA3 alpha-3 light chain 0.3805 0.3513 0.0182 100 EIF4B Eukaryotic translation initiation factor 4B 0.3750 0.2528 0.0365 133 RPL11 60S ribosomal protein L11 0.3045 0.5302 0.0333 922 RPL5 60S ribosomal protein L5 0.2064 0.5544 0.0760 2461 RPL26 60S ribosomal protein L26 -0.1074 -0.0086 0.5005 2929 RPL38 60S ribosomal protein L38 -0.2950 0.1456 0.6749

154 Supplementary Table 2.4. SILAC IP/MS proteomics results for sh-RPS14 vs. sh-LUC (lower left quadrant).

Condition B vs. condition C as indicated in Figure 1.1. The top 100 proteins preferentially bound to MDM2 without RPS14 deficiency, ranked by log2 MQ values, are included in the table. Additional proteins referred to in the text of Chapter 1 are listed at the bottom of the table.

Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 1 AMY1A Alpha-amylase 1 -2.5517 -2.1152 0.0015 2 MFGE8 Lactadherin;Lactadherin short form;Medin -2.1340 -1.4178 0.0087 3 LEMD2 LEM domain-containing protein 2 -1.7634 -1.4458 0.0020 Ribonucleoside-diphosphate reductase subunit 4 RRM2 M2 -1.7054 -1.3456 0.0029 5 RPL22L1 60S ribosomal protein L22-like 1 -1.6907 -2.3478 0.0052 Polymeric immunoglobulin receptor;Secretory 6 PIGR component -1.6811 -3.1680 0.0226 7 HMOX1 Heme oxygenase 1 -1.6251 -1.3950 0.0014 8 EMD Emerin -1.6182 -1.7817 0.0007 9 HIST1H4A Histone H4 -1.5999 -2.1834 0.0047 10 DPYSL5 Dihydropyrimidinase-related protein 5 -1.4954 -2.2820 0.0093 11 HIST1H3A Histone H3.1 -1.4946 -1.9894 0.0040 12 CDSN Corneodesmosin -1.4917 -2.5273 0.0152 13 LCN1;LCN1P1 Lipocalin-1;Putative lipocalin 1-like protein 1 -1.4803 -1.5188 0.0005 14 AZGP1 Zinc-alpha-2-glycoprotein -1.4684 -1.5889 0.0007 15 SPTLC2 Serine palmitoyltransferase 2 -1.4624 -0.9577 0.0104 16 CST6 Cystatin-M -1.4264 -1.8327 0.0031 17 DSC3 Desmocollin-3 -1.4118 -2.6125 0.0214 18 BANF1 Barrier-to-autointegration factor -1.3167 -1.8562 0.0061 19 CD59 CD59 glycoprotein -1.2957 -1.2609 0.0008 20 DNAJC12 DnaJ homolog subfamily C member 12 -1.2677 -1.3129 0.0008 21 CALML5 Calmodulin-like protein 5 -1.2226 -2.6811 0.0359 Heterogeneous nuclear ribonucleoproteins 22 HNRNPC C1/C2 -1.2036 -1.5800 0.0040 23 HIST1H2BJ Histone H2B type 1-J -1.1783 -1.6020 0.0051 24 HIST1H2AG Histone H2A type 1 -1.1755 -2.0128 0.0163 25 LGALS7 Galectin-7 -1.0792 -2.5702 0.0439 26 CBX3 Chromobox protein homolog 3 -1.0313 -1.2116 0.0022 27 CNTN1 Contactin-1 -1.0220 -0.8305 0.0039 28 MYH10 Myosin-10 -1.0216 -2.2590 0.0370 29 SQSTM1 Sequestosome-1 -1.0089 -0.8147 0.0041 30 HIST1H2BK Histone H2B type 1-K -1.0078 -1.5135 0.0094

155

Supplementary Table 2.4 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 31 GDI1 Rab GDP dissociation inhibitor alpha -1.0040 -1.2769 0.0037 Pyrroline-5-carboxylate reductase 2;Pyrroline- 32 PYCR2 5-carboxylate reductase -1.0024 -0.4802 0.0361 33 NOB1 RNA-binding protein NOB1 -0.9993 -1.6912 0.0160 34 NUCB1 Nucleobindin-1 -0.9728 -1.1180 0.0021 35 CPNE1 Copine-1 -0.9607 -0.8483 0.0026 36 ORMDL2 ORM1-like protein 2 -0.9479 -1.1701 0.0033 37 GPRC5A Retinoic acid-induced protein 3 -0.9457 -0.7056 0.0069 38 IGHG2 Ig gamma-2 chain C region -0.9447 -1.9113 0.0293 Thyroid hormone receptor-associated protein 39 THRAP3 3 -0.9437 -0.8513 0.0024 40 RPS29 40S ribosomal protein S29 -0.9387 -0.3967 0.0492 41 MYH10 -0.9371 -2.0829 0.0377

42 LMNB1 Lamin-B1 -0.9141 -1.2891 0.0072 43 YTHDF3 YTH domain family protein 3 -0.9138 -1.0063 0.0020 Bifunctional methylenetetrahydrofolate 44 MTHFD2 dehydrogenase/cyclohydrolase, mitochondrial -0.9001 -0.8166 0.0026 Pyrroline-5-carboxylate reductase;Pyrroline-5- 45 PYCR1 carboxylate reductase 1, mitochondrial -0.8951 -0.8459 0.0022 Sodium/potassium-transporting ATPase 46 ATP1A1 subunit alpha-1 -0.8934 -0.7908 0.0029 Guanine nucleotide-binding protein G(k) 47 GNAI3 subunit alpha -0.8896 -0.5020 0.0235 48 ABCF1 ATP-binding cassette sub-family F member 1 -0.8879 -0.9096 0.0019 49 AXL Tyrosine-protein kinase receptor UFO -0.8832 -0.8076 0.0026 Guanine nucleotide-binding protein G(i) 50 GNAI1 subunit alpha-1 -0.8707 -1.3881 0.0130 51 ACLY ATP-citrate synthase -0.8692 -0.9359 0.0021 52 LARP4B La-related protein 4B -0.8637 -1.0539 0.0035 Guanine nucleotide-binding protein 53 GNG12 G(I)/G(S)/G(O) subunit gamma-12 -0.8532 -0.6236 0.0086 MOB kinase activator 1A;MOB kinase 54 MOB1A;MOB1B activator 1B -0.8408 -0.7978 0.0025 Guanine nucleotide-binding protein G(i) 55 GNAI2 subunit alpha-2 -0.8365 -0.4790 0.0232 56 USP5 Ubiquitin carboxyl-terminal hydrolase 5 -0.8362 -0.8182 0.0023 57 LAMC1 Laminin subunit gamma-1 -0.8350 -0.7873 0.0026 Patatin-like phospholipase domain-containing 58 PNPLA2 protein 2 -0.8152 -0.7041 0.0040 59 QDPR Dihydropteridine reductase -0.8115 -0.8344 0.0024 60 MBOAT7 Lysophospholipid acyltransferase 7 -0.8105 -0.7919 0.0025 61 POLD3 DNA polymerase delta subunit 3 -0.8020 -0.7648 0.0028

156

Supplementary Table 2.4 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value Mitotic spindle-associated MMXD complex 62 FAM96B subunit MIP18 -0.8009 -0.9032 0.0029 63 C19orf10 UPF0556 protein C19orf10 -0.8008 -0.7337 0.0032 64 PNO1 RNA-binding protein PNO1 -0.7934 -0.6698 0.0047 Endoplasmic reticulum-Golgi intermediate 65 ERGIC2 compartment protein 2 -0.7808 -0.4368 0.0260 66 MAPK9 Mitogen-activated protein kinase 9 -0.7800 -0.5854 0.0084 Large neutral amino acids transporter small 67 SLC7A5 subunit 1 -0.7697 -0.4525 0.0225 68 MRPL4 39S ribosomal protein L4, mitochondrial -0.7693 -0.6614 0.0046 69 AK2 Adenylate kinase 2, mitochondrial -0.7583 -0.5799 0.0081 70 RPS3 40S ribosomal protein S3 -0.7528 -0.6602 0.0044 71 LMAN2 Vesicular integral-membrane protein VIP36 -0.7503 -1.0796 0.0090 Guanine nucleotide-binding protein 72 GNB2 G(I)/G(S)/G(T) subunit beta-2 -0.7411 -0.6285 0.0052 73 EIF2A Eukaryotic translation initiation factor 2A -0.7300 -0.7751 0.0031 Guanine nucleotide-binding protein G(s) 74 GNAS subunit alpha isoforms Xlas -0.7268 -0.4507 0.0196 75 RPS8 40S ribosomal protein S8 -0.7190 -0.7150 0.0033 76 ITGB5 Integrin beta-5;Integrin beta -0.7184 -0.4629 0.0173 77 SH3PXD2B SH3 and PX domain-containing protein 2B -0.7092 -0.5187 0.0107 78 SFXN1 Sideroflexin-1 -0.7088 -0.4957 0.0129 79 CAB39 Calcium-binding protein 39 -0.7075 -1.2383 0.0197 26S proteasome non-ATPase regulatory 80 PSMD10 subunit 10 -0.7073 -0.6301 0.0047 81 MRPL44 39S ribosomal protein L44, mitochondrial -0.7073 -0.6448 0.0043 40S ribosomal protein S27;40S ribosomal 82 RPS27L protein S27-like -0.7067 -0.8799 0.0051 83 TNS3 Tensin-3 -0.7034 -0.6765 0.0038 84 MYO10 Unconventionnal myosin-X -0.7015 -0.8538 0.0047 85 SPTLC1 Serine palmitoyltransferase 1 -0.6932 -0.6412 0.0043 Sodium/potassium-transporting ATPase 86 ATP1B1 subunit beta-1 -0.6868 -0.7133 0.0036 87 RPS6KA3 Ribosomal protein S6 kinase alpha-3 -0.6849 -0.7969 0.0043 88 SNX1 Sorting nexin-1 -0.6808 -0.8323 0.0050 89 RPS10 40S ribosomal protein S10 -0.6748 -0.5888 0.0056 90 SLC3A2 4F2 cell-surface antigen heavy chain -0.6736 -0.5871 0.0057 91 UBAP2 Ubiquitin-associated protein 2 -0.6726 -0.5131 0.0097 92 RPS13 40S ribosomal protein S13 -0.6710 -0.6808 0.0039 93 SLC1A5 Neutral amino acid transporter B(0) -0.6698 -0.5981 0.0053

157

Supplementary Table 2.4 (continued) Log2 MQ Log2 MQ Norm M/H Norm H/L p- Rank Gene Name Protein Name Rep1 Rep2 value 94 RPS9 40S ribosomal protein S9 -0.6636 -0.7067 0.0040 95 MRTO4 mRNA turnover protein 4 homolog -0.6572 -0.5543 0.0068 96 DDX24 ATP-dependent RNA helicase DDX24 -0.6560 -0.8669 0.0071 97 RPS4X 40S ribosomal protein S4, X isoform -0.6465 -0.5729 0.0058 98 CBS Cystathionine beta-synthase -0.6411 -1.1196 0.0202 99 MRPL38 39S ribosomal protein L38, mitochondrial -0.6404 -0.6044 0.0050 100 UCK2 Uridine-cytidine kinase 2 -0.6377 -0.4811 0.0110 123 RPS7 40S ribosomal protein S7 -0.3925 -0.5802 0.0186

158 Supplementary Table 3.1. Top 20 p53-interacting proteins with RPS19 knockdown and DMSO treatment.

The top 20 proteins preferentially bound to p53 with DMSO treatment, ranked by log2 MQ values, are included in the table. Normal subcellular localization of interacting proteins is keyed as follows: C = cytoplasmic; N = nuclear; B = both cytoplasmic and nuclear.

Log2 L/H Log2 H/L MQ Norm MQ Norm Rank Gene name Protein name Rep1 Rep2 p-value Loc 1 EHD2 EH domain-containing protein 2 1.9627 1.4341 0.0063 C 2 PPT1 Palmitoyl-protein thioesterase 1 1.4715 1.2599 0.0041 C 3 CNN2 Calponin-2 1.1441 1.2693 0.0047 C 4 CTSZ Cathepsin Z 1.4236 0.9549 0.0130 C Nucleoside diphosphate-linked moiety X motif 19, 5 NUDT19 mitochondrial 1.3672 0.9921 0.0099 C 6 CDK4 Cyclin-dependent kinase 4 0.9006 1.2690 0.0119 N Probable DNA dC->dU-editing enzyme APOBEC- 7 APOBEC3C 3C 0.9857 1.0367 0.0066 N 8 CRIP2 Cysteine-rich protein 2 1.1332 0.8440 0.0118 N 9 PPP1R12A Protein phosphatase 1 regulatory subunit 12A 1.0380 0.9132 0.0080 C 10 UCC1;EPDR1 Mammalian ependymin-related protein 1 1.0009 0.9445 0.0074 C 11 PTRF Polymerase I and transcript release factor 0.8869 1.0539 0.0089 N 12 SPATS2L SPATS2-like protein 1.1190 0.7919 0.0144 N 13 FBXO22 F-box only protein 22 1.0298 0.8367 0.0104 C 14 EPHA2 Ephrin type-A receptor 2 0.9108 0.9059 0.0086 C 15 PLOD2 Procollagen-lysine,2-oxoglutarate 5-dioxygenase 2 0.8041 0.9998 0.0115 C 16 STAT3 Signal transducer and activator of transcription 3 0.9065 0.8943 0.0088 B 17 MAP7 Ensconsin 0.7179 0.9621 0.0156 C 18 TP53I3 Quinone oxidoreductase PIG3 0.7253 0.8821 0.0140 C 19 EIF4B Eukaryotic translation initiation factor 4B 0.7284 0.8196 0.0137 C 20 TNS3 Tensin-3 0.7485 0.7889 0.0132 C

159 Supplementary Table 3.2. Top 20 p53-interacting proteins with RPS19 knockdown and TFP treatment.

The top 20 proteins preferentially bound to p53 with 20uM TFP treatment, ranked by log2 MQ values, are included in the table. Normal subcellular localization of interacting proteins is keyed as follows: C = cytoplasmic; N = nuclear; B = both cytoplasmic and nuclear.

Log2 Log2 L/H MQ H/L MQ Norm Norm Rank Gene name Protein names Rep1 Rep2 p-value Loc 1 SCD Acyl-CoA desaturase -2.4746 -3.0085 0.0018 C 2 FCGBP IgGFc-binding protein -2.1618 -3.1733 0.0071 C 3 SDCBP Syntenin-1 -1.7973 -2.2242 0.0029 C 4 FMNL2 -1.7916 -2.1364 0.0024 C 5 STOM Erythrocyte band 7 integral membrane protein -1.7591 -2.1491 0.0028 C 6 PSAP Proactivator polypeptide; Saposin-A -1.9003 -1.9781 0.0013 C 7 QPCT Glutaminyl-peptide cyclotransferase -1.5039 -2.2641 0.0094 C Gamma-aminobutyric acid receptor-associated 8 GABARAP protein -1.6641 -2.0130 0.0029 C TMEM106 9 B Transmembrane protein 106B -1.9310 -1.7286 0.0019 C 10 FDFT1 Squalene synthase -1.5777 -2.0409 0.0044 C 11 NEU1 Sialidase-1 -1.4781 -2.0997 0.0073 C 12 NPC1 Niemann-Pick C1 protein -1.9920 -1.4933 0.0054 C 13 CD63 CD63 antigen -1.5434 -1.7699 0.0026 C 14 RAP1GAP Rap1 GTPase-activating protein 1 -1.9072 -1.3979 0.0064 C 15 SLC38A2 Sodium-coupled neutral amino acid transporter 2 -1.9862 -1.2623 0.0124 C 16 RDH10 Retinol dehydrogenase 10 -1.4141 -1.8316 0.0050 C 17 DHRS3 Short-chain dehydrogenase/reductase 3 -1.5971 -1.6347 0.0019 C 18 CYP51A1 Lanosterol 14-alpha demethylase -1.6845 -1.4729 0.0028 C 19 DHCR7 7-dehydrocholesterol reductase -1.6005 -1.4107 0.0030 C 20 FDPS Farnesyl pyrophosphate synthase -1.4223 -1.5868 0.0028 C

160 Supplementary Table 4.1. Sequences of ribosomal protein-targeting shRNAs.

Numbering of shRNAs corresponds with the labeling of Figure 4.2. All shRNAs were obtained from the RNAi Consortium (TRC) at the Broad Institute of Harvard and MIT.

shRNA Gene TRC Clone ID Number Target Sequence RPS6 TRCN0000010430 1 GATGAACGCAAACTTCGTACT RPS6 TRCN0000010431 2 TACTTTCTATGAGAAGCGTAT RPS6 TRCN0000040080 3 CGCAAACTTCGTACTTTCTAT RPL26 TRCN0000117402 1 CCAGGTTTACAGGAAGAAATA RPL26 TRCN0000117404 2 CCCACATTCGAAGGAAGATTA RPL26 TRCN0000117405 3 AGTCCAGGTTTACAGGAAGAA RPL13 TRCN0000117417 1 TGCTTTACTTTCTGTGTTGAA RPL13 TRCN0000117418 2 CCGCAGAACAGGATGTTGAAA RPL13 TRCN0000117420 3 GAGGAAGAGAAGAATTTCAAA RPL21 TRCN0000117619 1 CAAGATTCTTGCCAAGAGAAT RPL21 TRCN0000117621 2 GTTGGCATTGTTGTAAACAAA RPL21 TRCN0000117618 3 GCCACATATATGCGAATCTAT RPL29 TRCN0000072984 1 CCCGATCACAAAGATACGAAT RPL29 TRCN0000072985 2 GCACAGAAATGGTATCAAGAA RPL29 TRCN0000072986 3 CCCTACAAAGGCTTCAGAGTA RPL14 TRCN0000117490 1 CACTGATTTCATCCTCAAGTT RPL14 TRCN0000299968 2 CACCAGAAGTATGTCCGACAA RPL14 TRCN0000299967 3 GCGATTGTAGATGTTATTGAT RPSA TRCN0000029481 1 CGTGCAATTGTTGCCATTGAA RPSA TRCN0000029483 2 CCAGATGGAACAGTACATCTA RPSA TRCN0000029479 3 CCTGCTGATGTCAGTGTTATA RPS12 TRCN0000303317 1 ACTGTGATGAGCCTATGTATG RPS12 TRCN0000303318 2 CGAAGCTGCCAAAGCCTTAGA RPS12 TRCN0000074827 3 GACGTTAATACTGCTTTACAA Luciferase TRCN0000072253 n/a ACACTCGGATATTTGATATGT

161