Hypoxia Influences Polysome Distribution of Human Ribosomal Protein S12 and Alternative Splicing of Ribosomal Protein Mrnas

Hypoxia Influences Polysome Distribution of Human Ribosomal Protein S12 and Alternative Splicing of Ribosomal Protein mRNAs

by Andrea Brumwell

A Thesis presented to The University of Guelph

In partial fulfillment of requirements for the degree of Doctor of Philosophy in Molecular and Cellular Biology

Guelph, Ontario, Canada © Andrea Brumwell, June 2020 ii

ABSTRACT

HYPOXIA INFLUENCES POLYSOME DISTRIBUTION OF HUMAN RIBOSOMAL PROTEIN S12 AND ALTERNATIVE SPLICING OF RIBOSOMAL PROTEIN MRNAS

Andrea Brumwell Advisor: University of Guelph, 2020 Dr. James Uniacke

Ribosomes were once considered static in composition because of their essential role in protein synthesis and kingdom-wide conservation. This view is changing however, as mutations in certain ribosomal proteins are tolerated by cells, albeit with disease phenotypes known as

“ribosomopathies”. Heterogeneity in the protein composition of eukaryotic ribosomes is an emerging concept with evidence that different pools of ribosomes exist with transcript-specificity, although evidence in human cells is severely lacking. Furthermore, the influence of a physiological stressor on human ribosomal proteins has yet to be studied. We show that the polysome association of human RPS12 (eS12) is altered by low oxygen (hypoxia), a feature of the tumor microenvironment. Our data suggest that RPS12 (eS12) is enriched in hypoxic monosomes, which increases the heavy polysome association of structured transcripts APAF-1 and XIAP. We also show that alternative splicing of RPS24 (eS24) is substantially altered in cell culture models of tumor hypoxia (spheroids), which may be partly influenced by hypoxia and acidosis. Since APAF-

1 and XIAP play opposing roles in apoptosis, these data may help to further understand cell death under stress. Additionally, alternative splicing of RPS24 changes the coding sequence, thus could provide heterogeneity to ribosomes as an adaptation to the spheroid/tumor microenvironment. Our data suggest that features of the tumor microenvironment, including hypoxia, may influence regulation of the human ribosome through changes in RP incorporation and the production of stress-specific RP isoforms.

iii

ACKNOWLEDGEMENTS

First and foremost, I would like to thank my advisor, Dr. Jim Uniacke, for his continued support and guidance over the past 5 years. His wealth of novel ideas and constant positivity were a welcome reprieve during times of struggle and exhaustion. I am sincerely grateful for the opportunities he has given me to present my research, network, and develop effective writing skills that will continue to benefit me in the future. I would like to thank my undergraduate project students for all their hard work and dedication to this project: Lorian Fay, Michael Rosen, Lindsay Obress, and Leslie Fell. I thank the members of my advisory committee, Dr. John Vessey, Dr. Richard Mosser, and Dr. Jonathan LaMarre, for their helpful feedback and insights throughout the course of my degree. I would also like to thank the many agencies that have funded my research and conference travel: Natural Sciences and Engineering Research Council, Government of Ontario, Canadian Cancer Society, and RiboClub. Thank you to my fellow labmates for troubleshooting advice, frequent discussions, and mental support. I thank Mathieu Durand, Philippe Thibault, and Jonathan Krieger for help with interpretation of ASPCR and mass-spectrometry techniques. I acknowledge that some of this data has first been published in RNA Journal (Brumwell et al., 2020). Finally, I wish to thank my family for their constant support. Although they do not understand the research, they were always more than willing to listen and offer advice as best they could.

TABLE OF CONTENTS

ABSTRACT ...... ii ACKNOWLEDGEMENTS ...... iii TABLE OF CONTENTS ...... iv LIST OF TABLES ...... vi LIST OF FIGURES ...... vii LIST OF SYMBOLS, ABBREVIATIONS, AND NOMENCLATURE ...... ix CHAPTER 1 – HYPOXIC REGULATION OF RIBOSOMAL PROTEIN EXPRESSION, POLYSOME INVOLVEMENT, AND ALTERNATIVE SPLICING ...... 1 1.1 Introduction ...... 1 1.1.1 Origin of the Genetic Code ...... 1 1.1.2 The Ribosome ...... 2 1.1.3 Ribosomopathies ...... 5 1.1.4 Translation Regulation During Stress ...... 8 1.1.5 Ribosome Heterogeneity and Specialized Translation ...... 11 1.1.6 Hypoxia ...... 14 1.2 Methods...... 15 1.2.1 Cell Culture ...... 15 1.2.2 RNA Isolation, RP RT-qPCR, ASPCR...... 16 1.2.3 Polysome Fractionation, Isolation of Protein and RNA ...... 16 1.2.4 Tandem Mass Tags Mass Spectrometry of Polysome Fractions ...... 17 1.2.5 Transient Transfection, RNA Isolation, and RT-PCR ...... 18 1.2.6 Total RNA Extraction and RT-qPCR ...... 18 1.2.7 Western Blot ...... 19 1.2.8 Statistical Analyses ...... 19 1.3 Results ...... 20 1.3.1 Hypoxia decreases the expression of 20% of ribosomal protein mRNAs in HEK293 ...... 20 1.3.2 Hypoxia influences the participation of four RPs in actively translating ribosomes of HEK293 LLLLcells ...... 23 1.3.3 RPS12 overexpression increases heavy polysome association of APAF-1 and XIAP mRNAs 27 1.3.4 Hypoxia induces alternative splicing of five RP mRNAs in HEK293 ...... 35 1.3.5 Hypoxia induces alternative splicing events in RP mRNAs in cancer cell lines ...... 39 1.4 Discussion ...... 41

CHAPTER 2 – ALTERNATIVE SPLICING OF RP MRNAS AND THE TUMOR MICROENVIRONMENT ...... 48 2.1 Introduction ...... 48 2.1.1 Alternative Splicing ...... 48 2.1.2 Alternative Splicing in Cancer ...... 49 2.1.3 Tumor Microenvironment ...... 50 2.1.4 RPS24 ...... 53 2.2 Methods...... 54 2.2.1 ASPCR and RT-qPCR of Origene Human Prostate Tumor cDNA Panels ...... 54 2.2.2 Spheroid Formation ...... 55 2.2.3 Cloning and Generation of RPS24 Stable Cell Lines ...... 55 2.2.4 Acidosis ...... 56 2.2.5 Lactic Acidosis ...... 56 2.2.6 Cell Confluency ...... 57 2.2.7 Statistical Analyses ...... 57 2.3 Results ...... 57 2.3.1 Exon inclusion within RPS24 and an unpredicted amplicon in RPL22L1 correlate with LLLLhypoxia in prostate tumors ...... 57 2.3.2 The long variant of RPS24 is significantly upregulated in spheroids of four cell lines ...... 61 2.3.3 The unpredicted amplicon in RPL22L1 was not detected in cell lines or other tumor types ... 63 2.3.4 Endogenous RPS24 short and long transcript variants, and exogenous protein isoforms LLLLassociate with polysomes ...... 66 2.3.5 RPS24 switches splice variant abundance over 7 days of spheroid growth ...... 69 2.3.6 Acidosis has modest influence on RPS24 alternative splicing in normoxia...... 71 2.3.7 Lactic acidosis does not affect RPS24 splice variant expression in U87MG ...... 73 2.3.8 Increasing cell confluency of a monolayer does not increase RPS24 long expression in LLLLU87MG under 21% O2 ...... 74 2.4 Discussion ...... 75 CONCLUSION ...... 81 BIBLIOGRAPHY ...... 84 APPENDIX ...... 97

LIST OF TABLES

Appendix Table 1: TMT-MS Ribosomal Protein Identification...... 101

Appendix Table 2: TMT-MS Ribosomal Protein Polysome Quantification...... 102-103

Appendix Table 3: HEK293 ASPCR Data...... 104

Appendix Table 4: Cell Line ASPCR Data...... 105

Appendix Table 5: Prostate Tumor RPS24 and RPL22L1 Data...... 106-109

Appendix Table 6: List of Primers...... 110-113

vii

LIST OF FIGURES

Figure 1: Hypoxia decreases the expression of 20% of ribosomal protein mRNAs in

HEK293...... 22

Figure 2: Hypoxia influences the participation of four RPs in actively translating ribosomes of

HEK293...... 25

Figure 3: Confirmation of mass spectrometry data by western blot...... 27

Figure 4: Knockdown of RPS12 impairs 40S ribosome biogenesis...... 28

Figure 5: Exogenous RPS12 associates with polysome fractions...... 30

Figure 6: RPS12 overexpression increases heavy polysome association of APAF-1 and XIAP mRNAs...... 32-33

Figure 7: RPS12 overexpression does not augment p53 levels...... 35

Figure 8: Hypoxia induces alternative splicing of five RP mRNAs in HEK293...... 36

Figure 9: Splicing maps of alternative splicing events with hypoxic ∆PSI > 10%...... 38

Figure 10: Hypoxia induces alternative splicing events in RP mRNAs in cancer cell lines...... 41

Figure 11: Exon inclusion within RPS24 and an unpredicted amplicon in RPL22L1 correlate with hypoxia in prostate tumors...... 60

Figure 12: The long variant of RPS24 is significantly upregulated in spheroids of four cell lines...... 63

Figure 13: The unpredicted amplicon of RPL22L1 was not detected in cell lines or other tumor types...... 66

Figure 14: Endogenous RPS24 short and long transcript variants, and exogenous protein isoforms associate with polysomes...... 68

viii

Figure 15: RPS24 switches splice variant abundance over 7 days of spheroid growth...... 70

Figure 16: Acidosis has modest influence on RPS24 alternative splicing in normoxia...... 72

Figure 17: Lactic acidosis does not affect RPS24 splice variant expression in U87MG...... 74

Figure 18: Increasing cell confluency of a monolayer does not increase RPS24 long expression in

U87MG under 21% O2...... 75

Appendix Figure 1: Hypoxic exposure confirmed by expression of hypoxia inducible factors....99

Appendix Figure 2: SLC2A1 transcripts encoding GLUT1 are increased in hypoxic cells and spheroids used for ASPCR...... 100

LIST OF SYMBOLS, ABBREVIATIONS, AND NOMENCLATURE

4E-BP – eIF4E-binding protein 5’ TOP – 5’ terminal oligopyrimidine ANOVA – analysis of variance APAF-1 – apoptotic protease activating factor 1 ASE – alternative splicing event ASPCR – alternative splicing PCR ATP – adenosine triphosphate B2M – β2 microglobulin BLAST - Basic Local Alignment Search Tool bp – base pair CAIX – carbonic anhydrase 9 cDNA – complementary DNA CRISPR – clustered regularly interspaced short palindromic repeats Ct – cycles to threshold C-terminus – carboxy-terminus DHX29 – DExH-box helicase 29 EDTA – ethylenediaminetetraacetic EGFR – epidermal growth factor receptor eIF – eukaryotic initiation factor GAPDH – glyceraldehyde 3-phosphate dehydrogenase GDP – guanosine diphosphate GTP – guanosine triphosphate H – heavy polysomes HIF – hypoxia-inducible factor IRES – internal ribosome entry site kDa – kilodalton L – light polysomes M – monosomes

x mTOR – mechanistic target of rapamycin N-terminus – amino-terminus PCR – polymerase chain reaction PDGFRA – platelet-derived growth factor receptor A pH – potential hydrogen PIC – pre-initiation complex PSI – percent splicing index rDNA – ribosomal DNA RNA-Seq – RNA-Sequencing RNP – ribonucleoprotein RP – ribosomal protein RT-PCR – reverse transcription PCR RT-qPCR – quantitative reverse transcription PCR S (40S, 60S, 18S) – Svedberg sedimentation coefficient s.e.m. – standard error of the mean SDS – sodium dodecyl sulfate siRNA – small interfering RNA SLC2A1/GLUT1 – solute carrier family 2, facilitated glucose transporter member 1 /glucose transporter 1 snoRNA – small nucleolar RNA SR proteins – serine arginine proteins tRNA – transfer RNA uORF – upstream open reading frame UTR – untranslated region VEGF-A – vascular endothelial growth factor A XIAP – X-linked inhibitor of apoptosis protein

CHAPTER 1 – HYPOXIC REGULATION OF RIBOSOMAL PROTEIN EXPRESSION, POLYSOME INVOLVEMENT, AND ALTERNATIVE SPLICING

1.1 Introduction

1.1.1 Origin of the Genetic Code

In the beginning there was… RNA or protein? Proposed separately by Francis Crick, Leslie Orgel, and Carl Woese, the RNA World Hypothesis postulates that RNA was likely the primordial molecule out of the three basal components of the genetic code: DNA, RNA, and protein (Crick,

1968; Orgel, 1968; Woese, 1967). It’s been suggested that RNA likely emerged before DNA since deoxyribose is a modified version of ribose and its synthesis requires catalysis by a protein enzyme.

Additionally, ribose can be synthesized from formaldehyde which was likely abundant in the primordial atmosphere. DNA likely evolved as a more stable storage form of RNA suitable for the inheritable genetic material, albeit RNA is still the genome of many viruses (Alberts B, Johnson

A, 2002). Therefore, the prevailing hypotheses are that RNA or protein were the primordial molecules. Currently, proteins (as enzymes) are mostly responsible for carrying out catalytic activity within cells, but RNA is known to possess catalytic “ribozyme” activity, most notably in the peptidyl transferase center of the ribosome which catalyzes polypeptide synthesis (Nissen et al., 2000; Noller et al., 1992). Polypeptides lack the innate ability to self-replicate, a required skill of the primordial molecule and one possessed by nucleic acids due to complementary base-pairing.

It has also been proposed that proto-ribosomes were comprised only of RNA, a possible explanation for the paradox of how a machine that synthesizes proteins could itself be made of protein (Alberts B, Johnson A, 2002).

Despite the prevalence of the RNA World Hypothesis, it has been scrutinized for many reasons, such as the inherent instability of RNA, the fact that ribonucleotides are not spontaneously formed, and RNA’s catalytic capabilities are limited. The alternative theory is the protein interaction world hypothesis, where RNA arose either co-evolutionarily or at a later point through rudimentary reverse transcription from proteins and acted as a memory molecule (Andras and

Andras, 2005; Lacey et al., 1999). Ultimately, proposed theories on the origin of the genetic code are imperfect and may never solve the existential “chicken or egg” dilemma, but provide useful insight that will be essential to fully understand the increasing complexity of gene expression regulation.

1.1.2 The Ribosome

An ancient structure, essential to the genetic code, the ribosome is comprised of 4 rRNAs and 80 ribosomal proteins (including Receptor for activated C kinase 1) in humans. The large 60S subunit contains the 5S, 5.8S, and 28S rRNA and 47 ribosomal proteins whereas the small 40S subunit contains the 18S rRNA and 33 ribosomal proteins. The two subunits join to form the 80S ribosome, and structural landmarks are described with terms such as body, head, beak, platform, stalks, and feet (Khatter et al., 2015). At its innermost core the ribosome contains remnants suggesting it emerged ~4 billion years ago, in the last universal common ancestor (Fox, 2010; Petrov et al.,

2015). The peptidyl transferase center in the ribosome large subunit is comprised solely of rRNA

(Nissen et al., 2000; Noller et al., 1992) and acts as the catalytic component of the ribosome. The small subunit contains the decoding center, ensuring accurate codon-anticodon pairing (Khatter et al., 2015). Ribosomes are universally distributed and highly conserved, as evidenced by the use of

16S rRNA in prokaryotic phylogenetic studies to estimate evolutionary timelines and identify species in biological samples (Ludwig et al., 1998). Ribosomal RNA is typically encoded in

hundreds of copies of rDNA operons, and early evidence into rRNA sequence variation in humans suggested there was no difference in rDNA operon copies within an individual due to homogenization (Kuo et al., 1996). Contrarily, other studies have reported that both rDNA copy number and rRNA sequence variations exist in different cell lines, tissue-types, and between individuals (Leffers and Andersen, 1993; Parks et al., 2018). Adding another layer of complexity,

~2% of rRNA is modified by small nucleolar RNAs (snoRNAs), with approximately 100 of each of the two predominant rRNA modifications present in humans that can also be subject to regulation: 2’-O-ribose methylation and pseudouridylation (Sloan et al., 2017). The purpose of rRNA modifications likely involves translation regulation as they are highly concentrated in important functional regions of the ribosome. Indeed, rRNA modifications are important for proper translational fidelity (Sloan et al., 2017).

Despite accounting for 30-70% of the total mass of ribosomes, ribosomal proteins do not directly catalyze protein synthesis (Bernhardt and Tate, 2015). Ribosomal proteins are required for ribosome biogenesis, can stabilize rRNA, and even possess functions outside of the ribosome, although specific functions of most ribosomal proteins are still unknown (Bernhardt and Tate,

2015; Wool, 1996). Each ribosomal protein is encoded by a single gene, excluding RPS4 which is encoded by a gene on the X-chromosome (RPS4X) and two genes on the Y-chromosome (RPS4Y1 and RPS4Y2). The coding sequences of ribosomal protein genes have a high degree of kingdom- wide conservation, displaying 63% conservation on average across Homo sapiens, Drosophila melanogaster, Caenorhabditis elegans, and Saccharomyces cerevisiae (Yoshihama et al., 2002).

Ribosomal protein genes also possess the highest number of pseudogenes of any gene class, with over 2 000 having been discovered thus far (Zhang et al., 2002). Unlike other eukaryotes such as yeast and flies, human ribosomal protein genes are not duplicated or present in paralogous pairs,

although 8 human paralogs are known to exist. Alternative splicing has been observed in 14 ribosomal protein mRNAs, and rRNA-modifying snoRNAs are found in many introns of ribosomal protein mRNAs (Gupta and Warner, 2014; Hoeppner et al., 2009).

Ribosomal proteins are small proteins (<50 kDa), that are nearly all basic/positively charged in order to facilitate binding to negatively charged rRNA. Apart from RPLP1 and RPLP2 which are present as dimers, crystallization of ribosome structure has demonstrated the stoichiometry of the ribosomal protein complement is a single copy (Razi and Ortega, 2017). Of the 80 ribosomal proteins in human ribosomes, 33 are universal proteins that are found across all domains of life, 35 are shared by eukaryotes and archaea, whereas 12 are specific to eukaryotes

(Ban et al., 2014). A new nomenclature was proposed to avoid confusion between naming of ribosomal proteins from prokaryotes and eukaryotes, which begins with a “b” for bacterial, “u” for universal, or “e” for eukaryotic but the old nomenclature system is still used when only studying one species (Ban et al., 2014). The eukaryote-specific ribosomal proteins are found on the periphery of the ribosome, whereas the universal proteins are buried deeper in the ribosome core, indicating essentiality of universal ribosomal proteins is likely due to contacts with rRNA and proximity to the decoding and peptidyl transferase centers (Khatter et al., 2015; Piir et al.,

2014).

Ribosomal proteins are also subject to post-translational modifications, which could alter the function of ribosomal proteins or potentially ribosome-interacting proteins. The acidic proteins

RPLP0, RPLP1, and RPLP2 are phosphorylated, and phosphorylation of RPS6 typically occurs in response to stimuli such as growth factors and nutritional sufficiency. Other post-translational modifications of ribosomal proteins include acetylation, ubiquitylation, and methylation (Simsek and Barna, 2017; Van De Waterbeemd et al., 2018). Considering the unbelievable complexity

involved in regulating the behemothic ribosome, there are theoretically infinite combinations of ribosomal components, suggesting that the view of rigid and unchanging ribosomes is rather archaic.

1.1.3 Ribosomopathies

Although all ribosomal proteins may not be necessary for basic translation, their expression is usually required for ribosome biogenesis, and thus homozygous ribosomal protein mutants are not viable. Heterozygotic inactivation or haploinsufficiency of ribosomal protein genes however is somewhat tolerated by cells, albeit with abnormal and usually dysfunctional phenotypes.

Dominant haploinsufficiency of ribosomal proteins in embryos of Drosophila melanogaster results in the “Minute” phenotype, characterized by prolonged development, thin and shortened bristles, and reduced viability and fertility (Marygold et al., 2007). Similarly, ribosome-based diseases (ribosomopathies) in humans can be caused by mutations in ribosomal proteins or non- ribosomal proteins and RNAs involved in ribosome biogenesis. Ribosomopathies present as a spectrum of clinical effects involving defects in hematopoiesis or erythropoiesis, anemia, bone marrow failure, immunological impairment, growth retardation, skeletal malformations, and increased susceptibility to cancer (Nakhoul et al., 2014). The most well-known human ribosomopathy, Diamond Blackfan Anemia, is mostly caused by heterozygous mutations in ribosomal proteins, with 60% of mutations being found in RPL11, RPL15, RPL26, RPL31,

RPL36A, RPL5, RPS7, RPS10, RPS17, RPS19, RPS24, or RPS26 (Boria et al., 2010). Clinically similar to Diamond Blackfan Anemia, 5q-syndrome has erythroid defects that can be reproduced by shRNA knockdown of RPS14, and rescued by RPS14 overexpression (Ebert et al., 2008).

The etiology of other ribosomopathies may be due to defects in rRNA processing and ribosome biogenesis. Unlike autosomal dominant and recessive forms of dyskeratosis congenita

which are caused by mutations in exclusively telomere-associated genes, X-linked dyskeratosis congenita is caused by mutation in the X-linked DKC1 gene which encodes the protein dyskerin

(Savage and Alter, 2009). Although also a component of the telomerase complex, dyskerin is associated with H/ACA box snoRNAs which are responsible for pseudouridylation of rRNAs in the nucleolus (He et al., 2002). Despite being considered a ribosomopathy, it has not been conclusively determined whether the disease is caused by impaired ribosome biogenesis or defective telomere maintenance (Gu et al., 2015; Ruggero et al., 2003; Thumati et al., 2013).

Cartilage-hair hypoplasia is caused by mutation in the RMRP gene, which is a snoRNA. However, like with DKC1, the RMRP RNA also functions with telomerase reverse transcriptase, so it is still unknown if disease phenotypes are directly caused by impaired rRNA processing (Maida et al.,

2009; Ridanpää et al., 2001). Treacher-Collins Syndrome is caused by mutations in the TCOF1 gene, and the protein from TCOF1 participates in rRNA transcription and processing (Gonzales et al., 2005). Shwachman-Diamond Syndrome is an autosomal recessive disorder where 90% of patients have mutations in SBDS gene, which encodes an essential protein involved in ribosome biogenesis. The extremely rare Bowen-Conradi Syndrome is caused by mutations in pseudouridine methyltransferase gene EMG1 and usually leads to death at a young age (Armistead et al., 2009).

Although the pathology of the above ribosomopathies are characteristic of decreased cell proliferation, they paradoxically are associated with a heightened susceptibility to increased cell proliferation found in cancer, often described as “Dameshek’s Riddle” (Dameshek, 1967). Further to this, cancer itself can be a ribosomopathy. Mutations of ribosomal proteins (eg. point mutations, deletions, copy number variations) occur in many different types of cancer. In T-cell acute lymphoblastic leukemia (T-ALL), mutations have been found in ribosomal proteins such as RPL5,

RPL10, RPL11, and RPL22. In fact, ribosomal protein mutations are found in 15-20% of all

pediatric T-ALL cases. Mutations of different ribosomal proteins have also been observed in cancer types such as melanoma, chronic lymphocytic leukemia, glioblastoma, breast cancer, uterine cancer, and stomach cancer (Sulima et al., 2017).

The frequent mutation of ribosomal proteins in cancer, as well as high tumor incidence in zebrafish with heterozygous ribosomal protein mutations, suggest that ribosomal proteins may function as tumor suppressors (Amsterdam et al., 2004). Ribosomal proteins, such as RPL11, can inhibit expression and functioning of the proto-oncogene c-MYC, thus their inactivation can drive c-MYC expression in cancer (Dai et al., 2007). Another proposed explanation for ribosomal proteins acting as tumor suppressors relates to the extra-ribosomal interactions of many ribosomal proteins with p53-regulator MDM2. Normally, MDM2 binds to, and negatively regulates, p53.

Nucleolar stress and defects in ribosome biogenesis and assembly lead to an excess of free ribosomal proteins. These proteins, including but not limited to RPL5, RPL11, and RPL23, bind to MDM2 and prevent it from inhibiting p53, leading to subsequent p53 accumulation and protective growth arrest (Zhang and Lu, 2009). However, if these ribosomal proteins are mutated, p53 is not stabilized through this pathway, and the growth-arresting response to defective ribosome assembly is not mounted, thus leading to hyper-proliferation and potentially transformation.

Cancer is characterized by uncontrolled and rapid cell growth, which demands high rates of protein synthesis and therefore ribosome biogenesis. Since rapid and uncontrolled protein synthesis can lead to skipping of essential quality control checkpoints, ribosomal protein alterations may simply arise as errors that bypass surveillance mechanisms. However, it is also a possibility that dysregulation of ribosomal components may bestow cells with a selective advantage to survive and proliferate despite the stressful environment created by rapidly dividing

cells. Thus, the possibility remains that ribosome dysregulation may actually be a cause of tumorigenesis and metastasis, rather than a consequence of malignancy.

1.1.4 Translation Regulation During Stress

Solid tumors accumulate stressors due to rapid and dense growth, such as nutrient deprivation, low pH, oxidative stress, and low oxygen tension (hypoxia). Cancer cells are also characterized by their ability to survive under these harsh environmental conditions (Giampietri et al., 2015).

Protein synthesis (translation) is the largest energy-consuming process occurring in growing cells, consuming 30 % of ATP in mammalian cells (Buttgereit and Brand, 1995). During a stress such as hypoxia, cellular respiration switches from aerobic metabolism (via oxidative phosphorylation) to anaerobic metabolism (via glycolysis). As oxidative phosphorylation provides the vast majority of ATP to a cell, energy is extremely limited during hypoxia (Solaini et al., 2010). Translation is globally repressed during stress in order to conserve precious energy stores, yet other processes exist that allow for translation of stress response transcripts to be maintained (Holcik and

Sonenberg, 2005). Unlike transcription, regulation of translation during stress provides a more rapid cellular response to changing conditions.

The majority of translation regulation occurs at the rate-limiting step of initiation. The canonical pathway of translation initiation is the cap-dependent scanning model and requires a host of eukaryotic initiation factors (eIFs). In this model, the ternary complex (eIF2-GTP-Met- initiator tRNA) assembles to form the 43S pre-initiation complex (PIC) by joining with the 40S small ribosomal subunit and eIFs 1, 1A, 3 and 5. Next the eIF4F cap-binding complex is assembled onto the 5’ 7-methylguanosine cap of mRNAs. The eIF4F complex is comprised of eIF4E cap- binding protein, eIF4A helicase, and eIF4G scaffolding protein. Upon cap-binding, eIF4A unwinds the 5’ untranslated region (UTR), along with eIF4B and eIF4H. Poly-A binding protein allows for

mRNA circularization by association with the eIF4F complex, and mRNAs are loaded onto 43S

PIC via eIF3. Scanning of the 43S PIC begins, and when an AUG is encountered in the context of the Kozak sequence, eIF1 leaves, and GTP is exchanged for GDP on eIF2 (facilitated by eIF5).

The initiator tRNA then delivers methionine to the P site of the 40S small subunit, and initiation factors are removed as the 60S large subunit joins and elongation begins (Aitken and Lorsch, 2012;

Sonenberg and Hinnebusch, 2009).

Global translation repression under stress occurs primarily through inhibition of canonical- cap dependent translation initiation as well as reduced translation of components of the protein synthesis machinery. Phosphorylation of Ser51 on the eIF2α subunit inhibits the exchange of GDP for GTP and therefore prevents ternary complex formation (Liu et al., 2006). Phosphorylation also regulates the activity of eIF4E-binding proteins (4E-BPs), which in turn mediates cap-dependent initiation by the eIF4F complex. Under normal conditions, 4E-BP is hyperphosphorylated and has a reduced affinity for eIF4E. Under stressful conditions, 4E-BP is hypophosphorylated and binds eIF4E, outcompeting eIF4G, and thus preventing formation of the eIF4F complex (Sonenberg and

Hinnebusch, 2009). Phosphorylation of 4E-BP occurs via the mechanistic target of rapamycin

(mTOR). During stress, mTOR is inhibited, leaving 4E-BP in a hypophosphorylated state.

Inhibition of mTOR also results in reduced translation of 5’ terminal oligopyrimidine (5’TOP) mRNAs, mainly comprised of ribosomal protein and translation factor mRNAs, thus repressing translation through a process that is heavily dependent on La-related protein 1 (Philippe et al.,

2020).

How are these mechanisms circumvented to continue synthesis of mRNAs that aid in cell survival and recovery during periods where canonical cap-dependent initiation is inhibited?

Upstream open reading frames (uORFs) often inhibit translation of the downstream main ORF of

certain mRNAs, but this repression is lifted during stress where scanning ribosomes bypass the uORF due to increased phosphorylated eIF2α (Vattem and Wek, 2004). An alternative cap-binding protein and eIF4E homolog, eIF4E2, has been discovered to function during hypoxic stress. This cap-binding protein binds mRNAs (such as EGFR and PDGFRA) containing an RNA hypoxia response element in their 3’ UTR (Uniacke et al., 2012). Further, delivery of the Met-initiator tRNA during hypoxia (when eIF2α is phosphorylated) was later proposed to occur via eIF5B (Ho et al., 2018).

Mechanisms of cap-independent translation initiation have also been discovered. First identified in viruses, highly-structured elements known as Internal Ribosome Entry Sites (IRES), have also been identified in cellular RNAs (such as VEGF-A, APAF-1, and XIAP) (Holcik and

Sonenberg, 2005). Although the exact mechanism of cellular IRES-mediated initiation has yet to be elucidated, translation of these transcripts requires various IRES trans-acting factors and occurs independent of the 5’ 7-methylguanosine gap and can continue to be translated during periods where cap-dependent initiation is inhibited, such as stress. One such IRES trans-acting factor is death-associated protein 5 (DAP5), also known as eIF4G2, which lacks the eIF4E- and poly-A binding protein- interacting domains of eIF4G but appears to promote translation of classic cellular

IRES during cell stress (Yoffe et al., 2016). Although numerous IRES (>100) are believed to be present in eukaryotic mRNAs, their existence is embroiled in controversy due to lack of reliability of assays used to validate their cap-independent function. The most common, bicistronic assays, are often confounded by alternative splicing or the presence of cryptic promoters (Jackson, 2013).

As a result, other mechanisms have been proposed to explain cap-independent translation initiation including cap-independent translation enhancers, and 5’end-dependent translation where initiation on an “uncapped” construct occurs only in the first cistron of a bicistronic transcript

(Andreev et al., 2012; Lacerda et al., 2017). A modification found in RNA, N6-methyladenosine

(m6A) has also been shown to recruit ribosomes directly to the 5’UTR independent of eIF4E, in what are known as m6A-induced ribosome engagement sites (MIRES) (Lacerda et al., 2017). Since cells under stress undergo such profound regulation of translation initiation, it is tempting to speculate that other components of the translation machinery may also be subject to regulation.

1.1.5 Ribosome Heterogeneity and Specialized Translation

The perception of the ribosome, the nexus of protein synthesis within the cell, has changed dramatically throughout the past couple of decades. No longer viewed as a completely rigid and passive structure, mounting evidence suggests that cells maintain a heterogeneous pool of ribosomes. The abundance of core ribosomal proteins in actively translating ribosomes is not altered in cancerous prostate cells compared to normal cells, although heterogeneity in ribosomal protein association was observed upon strong mTOR inhibition (Reschke et al., 2013). Altered stoichiometry of ribosomal proteins has been detected in both ribosomes of wildtype yeast and mouse embryonic stem cells (Shi et al., 2017; Slavov et al., 2015). Further, the presence or absence of certain ribosomal proteins in ribosomes may specify transcript selection. Two proteins identified as substoichiometric in mouse embryonic stem cells, RPL10A and RPS25, bind different categories of mRNAs when present in ribosomes, such as those involved in extracellular matrix organization and the vitamin B12 pathway, respectively (Shi et al., 2017). The recent use of advanced mass spectrometric techniques to measure ribosomal protein stoichiometry identified minimal heterogeneity in wildtype human 40S subunits, albeit RPS25 was again identified as substoichiometric (Van De Waterbeemd et al., 2018). Minimal heterogeneity in ribosome composition is unsurprising in unstressed cells, but it is plausible that physiological stressors may regulate gene expression through changes in ribosomal protein incorporation into the ribosome.

Indeed, ribosome regulation during stress is not a novel concept. The stress-induced

Escherichia coli toxin MazF, an endoribonuclease that cleaves mRNAs, was shown to generate leaderless mRNAs and 16S rRNA lacking anti-Shine Dalgarno sites (Vesper et al., 2011). The researchers found that leaderless mRNAs, which included stress-response transcripts, were preferentially translated by ribosomes containing cleaved 16S rRNA and postulated that MazF generated adaptive and specialized “stress ribosomes”. This paper was later refuted however as a subsequent in-depth study demonstrated that MazF produced very few leaderless mRNAs, and these leaderless transcripts were not preferentially translated (Culviner and Laub, 2018). Culviner and Laub (2018) also observed that MazF did not cleave mature 16S rRNA to produce specialized ribosomes but rather preferentially cleaved premature rRNA and ribosomal protein mRNAs to globally downregulate ribosome biogenesis.

Other rRNA sequence variations have been observed, however, which may function to regulate the stress response in E. coli. Actively translating ribosomes from nutrient-deprived E. coli are enriched with rRNA encoded by a specific rDNA operon. Nine out of ten variant nucleotides of this rDNA operon are located in the beak region of the ribosome, and these ten variant nucleotides are sufficient to alter the expression and translation of genes associated with nutrient stress (Kurylo et al., 2018).

Heterogeneity in rRNA modifications may also bestow specialized translation capabilities to ribosomes. Both pseudouridylation and methylation of rRNA have been implicated in the translation of IRES-containing mRNAs (Sloan et al., 2017). Loss of methylation of a single rRNA nucleotides in yeast led to preferential recruitment of oxidative stress-related mRNAs to ribosomes

(Schosserer et al., 2015). Further, oxidative damage to rRNA nucleotides by reactive oxygen species interferes with translational capacity (Willi et al., 2018).

The majority of ribosomal protein genes are duplicated in yeast, which produce paralogs that can have amino acid sequence variations. The ratio of paralog incorporation into ribosomes during high salt and hygromycin stress is altered, suggesting that yeast may modulate their ribosome composition in response to stress, similar to other studies showing alterations to ribosomal protein composition occur when changing carbon sources (Ghulam et al., 2019; Samir et al., 2018; Slavov et al., 2015). Exposure of yeast to high salt and high pH stress induced RPS26- deficient ribosomes that preferentially translated mRNAs bearing Kozak sequence variations

(Ferretti et al., 2017).

Similar to RPS26, the presence of RPL38 and RPL13A in mammalian ribosomes also affected the ability of ribosomes to translate IRES-containing transcripts while having no effect on cap-dependent translation initiation (Chaudhuri et al., 2007; Kondrashov et al., 2011; Xue et al., 2015). Depletion of RPL13A did not perturb ribosome biogenesis or global translation but indirectly reduced translation of cellular IRES-containing transcripts through reduced rRNA methylation (Chaudhuri et al., 2007). Tail short mice mutants are characterized by homeotic transformations caused by mutation of RPL38. It was discovered that RPL38 is required for assembly of ribosomes on a subset of Hox mRNAs, many of which contain IRES-elements

(Kondrashov et al., 2011; Xue et al., 2015). Despite amassing evidence of ribosomes with transcript specificity in bacteria, yeast, and mice, data is severely lacking in human cells, presumably due to reproducibility issues and the fact that the human ribosome was only recently crystallized. Furthermore, there has yet to be a study into the effect of a physiological stressor on the regulation of human ribosomes.

1.1.6 Hypoxia

Hypoxia, or low oxygen tension, is a key feature of solid tumors and one of many physiological stressors encountered by cells. Cell stressors such as heat shock, nutrient deprivation and pH fluctuations lead to induction of an adaptive stress response, promoting the survival of viable cells and removal of defective ones (Fulda et al., 2010). The cellular stress response is characterized by the synthesis of stress response proteins such as those involved in DNA damage repair, redox regulation, and energy metabolism (Kültz, 2004). Though mainly studied for its role in ischemic diseases and cancer, hypoxia is also implicated in normal physiological processes such as physical exercise, embryonic development, and the adaptation to life at high altitudes (Dunwoodie, 2009;

Hanaoka et al., 2012; Hopkins, 2006). Additionally, the measurement of oxygen tension at capillary ends of different human organs have demonstrated human tissues are in a physiological oxygen range that is much closer to hypoxia than the atmospheric 21% O2 commonly used in cell culture (Carreau et al., 2011).

Cells respond primarily to hypoxia through regulation of hypoxia inducible factors (HIFs) which regulate gene expression by acting as transcription factors. HIFs are heterodimeric complexes comprised of an oxygen-regulated α subunit and constitutive β subunit (HIF1β).

Activation and nuclear accumulation of HIFα subunits (HIF1α, HIF2α, and HIF3α) occurs during hypoxia, which allows them to bind HIF1β, and subsequently activate transcription of target genes possessing hypoxia response elements in their promoters such as VEGF-A, CAIX, and SLC2A1

(Kaluz et al., 2009; Poon et al., 2009).

As oxygen’s diffusion limit is 100-200 μm, tumors exceeding 1 mm in diameter (virtually undetectable through palpation) develop regions of hypoxia (Naumov et al., 2006). Mammalian cells require oxygen, thus as a tumor grows angiogenesis is stimulated to provide growing cells

with nutrients and oxygen through blood vessels. Rapidly synthesized tumor blood vessels lack the protective regulation of normal vessels, resulting in irregular, poorly constructed vessels with reduced effectiveness, subsequently leading to hypoxic regions within the tumor despite increased angiogenesis (Weis and Cheresh, 2011). Hypoxia in tumors is associated with metastasis, poor prognosis, and resistance to therapeutics, therefore it is an important focus of cancer research where discovery of hypoxic biomarkers could aid in early tumor detection and critical hypoxic processes are promising targets of tumor-specific treatments (Comerford et al., 2002; Liu et al.,

2014; Swinson et al., 2003).

Fulfilling the need for studies into human ribosome regulation during stress, I will describe our maiden exploration into the influence of hypoxia on a broad scale, focusing on ribosomal protein (RP) mRNA abundance, incorporation into actively translating ribosomes, and alternative splicing. We show that hypoxia decreases 20% of RP mRNAs and alters the translating ribosome- association of select ribosomal proteins (RPs), one of which may lead to more efficient translation of two stress-related transcripts. We also investigate hypoxic regulation of RP alternative splicing, which will be explored further in Chapter 2.

1.2 Methods

1.2.1 Cell Culture

Cell lines were obtained from the ATCC. HEK293, U87MG, and HCT116 cell lines were maintained in DMEM (Dulbecco’s Modified Eagle Medium), whereas PC3 were maintained in

RPMI (Roswell Park Memorial Institute) 1640, both media containing 7.5% (v/v) fetal bovine serum and 1% (v/v) penicillin-streptomycin. Cells were cultured at 37°C, 21% O2, and 5% CO2.

Cells were exposed to hypoxia in a HypoxyStation H35 for 24 h at 1% O2 and 5% CO2. Hypoxic exposure for samples analyzed by mass spectrometry was confirmed by induction of HIF1α or

HIF2α protein levels. Spheroids were generated using the liquid overlay method with brief manual rotation: Cells (50 000) were seeded into 96-well plates coated with 1.5% low-melting agarose

(Fisher) followed by quick rotation by hand for 60 sec and incubation for 5 days. Hypoxic exposure in cells used for RP mRNA expression and alternative splicing PCR (ASPCR) was confirmed by a >2-fold increase in SLC2A1/GLUT1 mRNA level compared to normoxia.

1.2.2 RNA Isolation, RP RT-qPCR, ASPCR

Cells were lysed using RiboZol (VWR) according to the manufacturer’s protocol. RNA was purified using PureLink RNA Mini Kit (Life Technologies) with on-column DNaseI treatment

(BioBasic). RNA integrity was confirmed using an Agilent Bioanalyzer or gel electrophoresis prior to reverse transcription. RP gene expression was performed in technical triplicate and three biological replicates. Primers (Appendix Table 6) for RP RT-qPCR amplified all known variants of the genes, and each biological replicate was normalized to the two most stable reference transcripts of RPLP0, GAPDH, and B2M using qBase software. Alternative splicing PCR

(ASPCR) was used to quantify a pool of 68 ASEs within RP mRNAs (Klinck et al., 2008). End- point PCR using primers flanking splicing sites (Appendix Table 6) was followed by capillary electrophoresis to separate and quantify amplicons. Using the most predominant splicing event, the PSI (percent splicing index) was calculated as the ratio of long amplicon over the sum of short

Long (nM) and long amplicons: PSI = ( ∗ 100). Short+Long (nM)

1.2.3 Polysome Fractionation, Isolation of Protein and RNA

Cells were grown in 4 x 150 mm plates until 80% confluency. Before harvesting, 0.1 mg/ml cycloheximide was added to media and incubated at 37°C for 10 min. Cells were lysed in 1X

Phosphate Buffered Saline (PBS) (140 mM NaCl, 3 mM KCl, 10 mM Na2HPO4, 15 mM KH2PO4)

containing 0.1 mg/ml cycloheximide. Pellets were combined and resuspended in RNA Lysis

Buffer (15 mM Tris-HCl pH 7.4, 15 mM MgCl2, 0.3 M NaCl, 1% Triton X-100, 0.1 mg/ml cycloheximide, 125 units RNaseOut (ThermoFisher), and EDTA-free protease inhibitor cocktail

(Cell Signaling Technologies)). Polysomes were dissociated by adding 20 mM EDTA pH 8.0 to lysis buffer and sucrose gradients in the absence of cycloheximide (Chapter 2). Equal RNA

(quantified as optical density (OD)/A260) was loaded over 7-47% sucrose gradients and ultracentrifuged for 1.5 h at 39 000 RPM at 4°C in a SW40 or SW41 Ti rotor (Beckman Coulter).

Sucrose gradients were fractionated and analyzed using Brandel BR-188 Density Gradient

Fractionation System to collect 1 ml fractions or manually pooled for mass spectrometry (and confirmation), as the absorbance at 254 nM was continuously measured. For protein analysis,

100% trichloroacetic acid was added to fractions at a ratio of 1:4 (250 μl per 1 ml fraction) and, following centrifugation at 4°C, pellets were washed twice with cold acetone. For western blot analysis, pellets were resuspended in SDS Lysis Buffer (4% SDS in 1x PBS) and 4X Laemmli

Buffer (250 mM Tris-HCl pH 6.8, 8% w/v SDS, 0.04% w/v Bromophenol Blue, 40% v/v Glycerol,

8% v/v 2-mercaptoethanol). For RNA extraction, polysome fractions were subjected to proteinase

K treatment by adding 50 μl of proteinase k solution (7.5% SDS, 75 mM EDTA, 1 μl Glycoblue, and 1.6 mg/ml proteinase K) and incubating at 55°C for 1 h, followed by extraction with phenol- chloroform-isoamyl alcohol (pH 6.7) and ethanol precipitation.

1.2.4 Tandem Mass Tags Mass Spectrometry of Polysome Fractions

Polysome fractions were pooled into monosome (M; single ribosomes), light polysomes (L; 2-4 ribosomes/mRNA), and heavy polysomes (H; ≥ 5 ribosomes/mRNA) prior to protein isolation.

Protein abundances in L and H were made relative to M in normoxia and hypoxia. Mass spectrometry analyses were carried out in three biological replicates. TCA-precipitated proteins

were reduced, alkylated, and digested with Trypsin/Lys-C, followed by labeling with TMT-10plex

(ThermoFisher). Each biological replicate (i.e. All 6 samples - M, L, H in 21% and 1% O2) was performed in a single run. Detailed protocol can be found in the appendix.

1.2.5 Transient Transfection, RNA Isolation, and RT-PCR

Cells (5 x 106) were seeded into 150 mm plates and transiently transfected (15 μg DNA) 24 h later using polyethylenimine (26 μg/mL final concentration) with either the empty plasmid control

(pMyc-N1 was a gift from Lei Lu (Addgene plasmid # 85759 ; http://n2t.net/addgene:85759 ;

RRID:Addgene_85759)) or RPS12 (Myc-DDK C-terminal tagged) Human Tagged ORF clone

(Origene RC214733). ONTARGETPlus SMARTPOOL Human RPS12 siRNA and

ONTARGETPlus Non-targeting Control Pool were used to transiently knockdown RPS12

(Dharmacon) using GenMute siRNA Transfection Reagent (SignaGen). Cells were lysed 48 h post-transfection and polysome fractionation and RNA isolation was performed as described above. Reverse transcription was optimized for structured transcripts (i.e. APAF-1 and XIAP) by incubating equal volumes of RNA with random primers for 5 min at 65°C, followed by incubation for 1 min on ice before adding the remaining kit components (Applied Biosystem High Capacity cDNA Reverse Transcription Kit). Endpoint PCR and agarose gel electrophoresis (1X TAE – 40 mM Tris, 20 mM acetic acid, 1 mM EDTA pH 8.0) were performed, band intensity was measured using Image Lab (BioRad), and the % mRNA was calculated for each fraction on the gradient out of the total band intensity from all fractions.

1.2.6 Total RNA Extraction and RT-qPCR

For total mRNA abundance measurements on whole cell lysates (Total APAF-1 and XIAP and all qPCR in Chapter 2), cells were lysed using RiboZol (VWR) according to the manufacturer’s

protocol. RNA integrity was confirmed by separating 1 μg of RNA on a 1.5 % agarose gel and observation of intact 18S and 28S rRNA bands. Reverse transcription of 2 μg of RNA was performed as per manufacturer’s instructions (Applied Biosystems). Primers were designed spanning exon junctions (Appendix Table 6). Quantitative PCR performed using SsoAdvanced

Universal SYBR Green Supermix (BioRad) using RPL13A and/or RPLP0 as reference.

1.2.7 Western Blot

Standard western blotting procedure was used. Primary antibodies (all antibodies diluted at 1/1

000 – 1/5 000, except for β-Actin at 1/20 000): anti-FLAG (F1804; Sigma), anti-β-Actin (GT5512;

GeneTex), anti-RPL37 (GTX104688; GeneTex), anti-RPS24 (A303-842A; Bethyl), anti-RPS12

(ab175219; Abcam), anti-RPL7 (A300-741A; Bethyl), anti-RPS21 (A305-070A; Bethyl), anti-

RPL24 (ab126172; Abcam), anti-RPSA (ab137388; Abcam), anti-RPL5 (ab137617; Abcam), anti-

HIF1α (NB100-123; Novus), anti-HIF2α (NB100-122; Novus), anti-p53 (sc-126; Santa Cruz), and anti-c-Myc (sc-40, Santa Cruz). Secondary antibodies diluted at 1/5 000 or 1/10 000 were HRP- conjugated anti-rabbit and anti-mouse (Promega). Densitometry was performed using Image Lab

(BioRad). Experiments used GAPDH or β-Actin as loading control, where applicable, excluding total ribosomal protein measurements which were normalized to Amido Black whole membrane staining (0.1% (w/v) amido black, 45% (v/v) methanol and 10% (v/v) acetic acid).

1.2.8 Statistical Analyses

All statistical analyses were performed using GraphPad Prism 7.0. Experiments containing biological replicates were analyzed for statistical significance using unpaired, two-tailed Student’s t-test on mean fold changes. One-sample t-tests were performed for RP RT-qPCR (i.e. was the mean fold change statistically different from no change = 1.0). Two-sample t-tests were performed

on mean abundance values for each RP in light/heavy polysomes compared to mean abundance in monosome fraction of corresponding oxygen concentration. P-value <0.05 was considered statistically significant.

1.3 Results

1.3.1 Hypoxia decreases the expression of 20% of ribosomal protein mRNAs in HEK293

Given the activation of the hypoxia-inducible transcription factor program and global repression of protein synthesis under low oxygen stress, we first investigated whether hypoxia influenced the total levels of RP mRNAs. The relative change in RP mRNA levels between normoxia (21% O2) and hypoxia (1% O2) was measured in HEK293 cells for 87 RP mRNAs (79 core RPs and 8 paralogs; RPS4Y1 and RPS4Y2 were not detected) using RT-qPCR in three biological replicates.

We chose HEK293 cells as our model in order to avoid the effects of cancer cell line-specific mutations. Primer pairs recognized all known transcript variants of a RP mRNA. The normalized relative expression was calculated, normalizing to the two most stable reference mRNAs in each replicate (out of RPLP0, B2M, and GAPDH), and making hypoxic normalized expression relative to normoxia.

The expression of 17 of the 87 RP genes decreased significantly, with 14 decreasing greater than 2-fold (Figure 1A). The expression of RPL7 was decreased the most by hypoxia, decreasing over 5-fold. Although hypoxia is associated with a global repression of translation (Liu and Simon,

2004), the majority of RP mRNAs were only slightly (non-significantly) decreased or unchanged with hypoxic treatment. As to be expected with such highly expressed genes, the RP mRNAs were not increased in hypoxia, except for RPL3L which was increased in 2 out of 3 replicates but had overall low amplification (high Ct values). Since RPs are considered housekeeping genes, the number of RP mRNAs affected by hypoxia was more than expected. We also measured the whole-

cell protein abundance of select RPs to determine if there was any agreement with total mRNA levels (Figure 1B). Of the four RPs that displayed significant transcript reduction (RPL37, RPL7,

RPL5, and RPS24), none were similarly decreased on the protein level. Thus from the small subset of total RPs measured, except for RPS12, the mRNA level did not appear to dictate total protein level.

Figure 1. Hypoxia decreases the expression of 20% of ribosomal protein mRNAs in HEK293. (A) The hypoxic abundance of 87 ribosomal protein (RP) mRNAs relative to normoxia was measured via RT-qPCR. Data (n = 3) were normalized to the two most stable reference mRNAs of RPLP0, B2M, and GAPDH. (B) The whole-cell abundance of select RPs in normoxia and hypoxia was measured using western blot. Quantification used whole- membrane Amido Black staining as loading control. Changes in protein and transcript abundance (taken from A) are compared. Data are mean (n ≥ 8) ± s.e.m. * = P < 0.05, ** = P < 0.01 using one-sample two-tailed t-test on mean fold change. Dotted lines represent no change relative to normoxia. 22

1.3.2 Hypoxia influences the participation of four RPs in actively translating ribosomes of HEK293 cells

As hypoxic RP mRNA levels did not dictate total protein levels (for the RPs measured), we decided to next investigate the effects of hypoxia on the association of RPs with actively translating ribosomes (polysomes). To measure polysome association of RPs, we cultured HEK293 cells in normoxia and hypoxia for 24 h and used sucrose density gradient centrifugation to separate polysomes from less-actively translating monosomes. Molecular hypoxic response was confirmed by detection of HIF1α or HIF2α protein by western blot (Appendix Figure 1). The absorbance at

254 nm, corresponding to RNA (mostly rRNA) of the sucrose gradient is shown (Figure 2A). We isolated 80S monosomes (M; single ribosomes), light polysomes (L; 2-4 ribosomes/transcript), and heavy polysomes (H; ≥ 5 ribosomes/transcript) via sucrose density gradient fractionation of an equal number of cells. M, L, and H fractions in both normoxia and hypoxia were tagged with separate isobaric labels for a single injection (per biological replicate) for tandem mass tags-mass spectrometry. Among the 3,232 unique proteins identified across three biological replicates

(~1,000 of which were present in all 18 samples), we detected 78 of the 80 canonical RPs and three paralogs with an average of 11 unique peptides. RPL39 and RPL41 were not detected perhaps due to small size (Appendix Table 1).

We calculated abundance ratios of RPs in L and H fractions relative to M, obtaining L/M and H/M ratios in normoxia and hypoxia (Figure 2B). The M fraction was used as reference within each condition to control for oxygen-dependent differences in ribosome biogenesis, ribophagy, and translation capacity. In accordance with a higher ribosome density in the polysome fractions, most RPs are more abundant in L and H relative to M (positive Log2 fold change; Figure 2B). In normoxia, 68/81 and 81/81 RPs displayed L/M and H/M ratios ≥1, respectively. Similarly, 50/81 and 80/81 RPs displayed hypoxic L/M and H/M ratios ≥1, respectively. Since our abundance ratios

were mostly ≥1, we divided hypoxic L/M or H/M by normoxic L/M or H/M to highlight hypoxia- specific changes in polysome association. We then applied a 20% cutoff to this value (dashed line;

Figure 2C) and statistical significance in at least hypoxic L/M or H/M (solid bars; Figure 2C).

Given that eukaryotic cells usually contain ~1 – 10 million ribosomes (Shi et al., 2017), even a

20% difference could correspond to changes in potentially hundreds of thousands of ribosomes.

Since hypoxia reduces global translation (Liu and Simon, 2004), it was not surprising that most

RPs were less associated with polysomes in hypoxia compared to normoxia. Indeed, 56/81 and

68/81 ratios had a negative Log2 fold change (Figure 2C). Interestingly, from the RPs that had a positive Log2 fold change, only three (RPL7L1, RPL8, and RPL27A) significantly increased by greater than 20% in hypoxic L/M or H/M relative to normoxia. Strikingly, RPS12 was the only RP to be more abundant in M relative to H in hypoxia (H/M < 1) and was also more abundant in hypoxic M relative to L (Figure 1B). Furthermore, RPS12 decreased 2.2-fold (i.e. 0.44 fold change) in hypoxic relative to normoxic H/M (Figure 1C). These data suggest that within a heterogeneous pool of ribosomes, RPL7L1, RPL8, and RPL27A are more likely, whereas RPS12 is less likely, to be incorporated into hypoxic polysomes.

Figure 2. Hypoxia influences the participation of four RPs in actively translating ribosomes of HEK293 cells. (A) Fractionated ultracentrifuged cell lysates from HEK293 in 21% O2 (normoxia) and 1% O2 (hypoxia) were pooled into 80S monosomes (M; within dotted line), light polysomes (L), and heavy polysomes (H). (B) Volcano plot of the abundance ratio using the mean abundances from three biological replicates to calculate the L/M and H/M

ratios. Dotted line represents P = 0.05 using two-sample two-tailed unpaired t-test. (C) The ratio of hypoxic L/M or H/M to normoxic L/M or H/M was calculated and represented as Log2 fold change. Dashed line represents cut-off value of ± 0.2 non-Log2 transformed change in hypoxic L/M or H/M relative to normoxic (L/M or H/M). Solid bars represent proteins above this cut-off that were statistically significant in hypoxic polysomes relative to monosomes.

We attempted to confirm the mass spectrometry data using RPS12 and RPL5, resulting in fold changes of 0.66 and 0.89, respectively (Figure 3A and B). We did not attempt to validate

RPL7L1, RPL8, and RPL27A by western blot as a less than two-fold change would not be detectable. Consistently, RPS12 was weakly associated with heavy polysomes despite a relatively strong level of detection in whole-cell lysates. A 2-fold difference for RPS12 was not fully achieved, likely stemming from the difficulty of quantifying both very intense and weak bands on the same image. Nonetheless, RPS12 was confirmed to have a greater decrease in hypoxic relative to normoxic H/M ratio compared to RPL5, but was mainly due to an increase in its abundance in the monosome fraction rather than a decrease in heavy polysomes (Figure 3A-C). The antibody against RPS12 recognizes its C-terminal region, indicating that changes in abundance detected via mass spectrometry were not due to nascent polypeptides. The mass spectrometry data are somewhat discordant with our gene expression data (Figure 1). Despite having increased hypoxic polysome association, the total mRNA of RPL7L1 and RPL27A was decreased in hypoxia.

Although we observed a hypoxic decrease in RP mRNA expression and polysome association, except for RPS12, none of the remaining RPs seem to be individually regulated externally to coordinated global translation repression. Since the polysome/monosome ratio of RPS12 is decreased by hypoxia (more so than all other RPs), this suggests that actively translating ribosomes

may be less likely to incorporate RPS12 or RPS12 may be more frequently associated with monosomes.

Figure 3. Confirmation of mass spectrometry data by western blot. (A) Representative western blot of RPS12 and RPL5 levels in 80S monosomes (M) and heavy polysomes (H) of HEK293 cells in 21% O2 (normoxia) and 1% O2 (hypoxia). (B) Band intensities were scaled based on individual images and average scaled intensities are shown for n = 8 ± s.e.m. * = P < 0.05, *** = P < 0.001. The ratio of L/M and H/M in hypoxia relative to normoxia is shown. (C)Western blot of RPS12 and RPL5 on whole polysome gradients (unpooled fractions).

1.3.3 RPS12 overexpression increases heavy polysome association of APAF-1 and XIAP mRNAs

While the mass spectrometry data indicated a reduction of RPS12 in hypoxic heavy polysomes relative to normoxia, there was also an enrichment of RPS12 in monosomes. Indeed, when validated via western blot, the enrichment of RPS12 in hypoxic monosomes was more evident than its reduction in heavy polysomes (Figure 3). We next sought to determine whether the hypoxic regulation of RPS12 in polysomes had any possible functional implications. Structural studies place RPS12 in the beak of the ribosome where it interacts with DExH-box helicase 29 (DHX29)

(Hashem et al., 2013). Since DHX29 has been linked to the translation initiation of structured transcripts (Parsyan et al., 2009; Pisareva and Pisarev, 2016), we investigated whether RPS12 had a similar role. We initially regarded the data as a reduction of the polysome/monosome ratio of

RPS12 in hypoxia, we attempted knocking down RPS12 to reproduce and even exaggerate the effect of hypoxia (Figure 4).

Figure 4. Knockdown of RPS12 impairs 40S ribosome biogenesis. Western blots from polysome fractions are shown for HEK293 transiently transfected with either non-targeting control siRNA pool or RPS12-targeting siRNA pool under 21% and 1% O2. RPL5 and RPS24 are blotted for as representation of the large and small ribosomal subunits, respectively.

However, although RPS12 is a bonafide ribosomal protein, its abundance was constantly weaker in polysomes than other RPs (such as RPL5) and most enriched in monosomes under both normoxia and hypoxia. Since RPS12 was mostly knocked down from the monosome fractions, this did not reproduce the hypoxic effect (i.e. where RPS12 was enriched in hypoxic monosomes compared to normoxia). Additionally, it was evident that siRNA knockdown of RPS12 impaired

40S ribosome biogenesis; The polysome profile displayed a large 60S peak with reduced 40S and

80S peaks. Western blot on the polysome fractions showed enrichment of 60S-member RPL5 in corresponding monosome fractions and an overall decrease of 40S-member RPS24 in all fractions

(Figure 4). This has been previously shown with siRNA-mediated depletion of other ribosomal proteins (O’Donohue et al., 2010; Robledo et al., 2008).

Since depletion of RPS12 impaired 40S ribosome biogenesis, we explored RPS12 overexpression, as the reduced polysome/monosome ratio was also characterized by an increased association with monosome/earlier polysome fractions. Transient expression of exogenous C- terminal Myc-tagged RPS12 enriched specifically in monosomes and light polysomes (Figure 5) so this provided a condition that exaggerated the hypoxic perturbation. In order to investigate whether RPS12 may be involved with the translation of mRNAs with highly structured 5’UTRs, we chose to measure the polysome association of two mRNAs with considerable secondary structures in their 5’UTRs, Apoptotic protease activating factor 1 (APAF-1) and X-linked inhibitor of apoptosis protein (XIAP). We compared their polysome association with that of two mRNAs with minimal secondary structures (β-Actin and GAPDH) (Figure 6).

Figure 5. Exogenous RPS12 associates with polysome fractions. Polysome profiles for normoxic (21% O2) and hypoxic (1% O2) HEK293 cells overexpressing RPS12 (Myc-DDK- RPS12) or empty vector control (pMyc-N1) are shown. Western blot for the Myc tag of exogenous RPS12 demonstrates polysome association. RPL5 was used to indicate protein presence in polysome fractions. Endogenous RPS12 is included for 21% O2.

Indeed, APAF-1 mRNA, a transcript with a very highly structured (ΔG = -289.1 kcal/mol via mfold) and long (587 nt) 5’ UTR, associated significantly more with hypoxic heavy polysomes in RPS12 overexpressing cells relative to the empty vector control (Figure 6A). We also tested

XIAP mRNA, a DHX29-dependent transcript (Parsyan et al., 2009) with a moderately structured

(ΔG = -65.0 kcal/mol) and shorter (159 nt) 5’ UTR, which associated more with hypoxic heavy polysomes in RPS12 overexpressing cells albeit less so than APAF-1 (Figure 6A). GAPDH and

β-Actin mRNAs, two transcripts with short 5’ UTRs (< 100 nt) containing weak secondary structures, displayed no change in their polysome association between RPS12 overexpressing cells and controls. Overexpression of RPS12 does not appear to alter global translation (polysome profiles; Figure 5). Crystal structures of the ribosome indicate that the C-terminus of RPS12 is solvent-exposed (Anger et al., 2013), so the tag should not interfere with ribosome incorporation and indeed exogenous RPS12 associated with the polysome fractions to a similar degree as the endogenous RP (Figure 5). However, likely due its overexpression, a considerable portion of exogenous RPS12 associated with free, non-ribosome pool (fraction 1; Figure 5). It’s therefore possible that the effects observed could be due to extra-ribosomal RPS12, although an increase in translational efficiency (rather than a decrease) may suggest ribosome involvement.

Figure 6. RPS12 overexpression increases heavy polysome association of APAF-1 and XIAP mRNAs. Semi-quantitative RT-PCR was used to measure the abundance of (A) APAF- 1, (B) XIAP, (C) GAPDH, and (D) β-Actin mRNAs in polysome fractions of normoxic (21% O2) and hypoxic (1% O2) HEK293 cells overexpressing RPS12 (eS12) with the Myc-DDK- RPS12 ORF vector (red line) or control cells expressing the empty vector pMyc-N1 (blue line). The fraction of mRNA in each lane was calculated based on total band intensity of the transcript across the entire polysome gradient. Dashed line denotes the start of the polysome fractions. Representative gel images are shown. (E) Total levels of APAF-1 and XIAP mRNA were measured in normoxic and hypoxic HEK293 control cells and those overexpressing RPS12 using RT-qPCR. The Δ∆Ct method was used, normalizing to reference mRNA RPL13A, and fold-change made relative to empty vector control. Data (n = 3), mean ± s.e.m. * = P < 0.05 Two-tailed unpaired t-tests were performed.

Some free-floating RPs can stabilize p53 (Zhang and Lu, 2009) and APAF-1 is a transcriptional target of p53 (Robles et al., 2001), thus we were concerned that enhanced polysome-association of APAF-1 may have been due to p53 accumulation and increased APAF-1 transcription. Upon further investigation however, we determined that p53 (Figure 7A-B) and total

APAF-1 mRNA (Figure 6B) levels did not change between RPS12-overexpressing and control cells. Total XIAP mRNA was also unchanged with RPS12-overexpression, thus the increase in heavy polysome association of APAF-1 and XIAP suggests higher translation efficiency of these structured mRNAs rather than being simply due to higher overall expression. Normoxic overexpression of RPS12 did not increase the translation efficiency of APAF-1 and XIAP mRNAs suggesting that other hypoxia-induced factors could be required.

Figure 7. RPS12 overexpression does not augment p53 levels. (A) Western blots for p53 levels using β-Actin as loading control from normoxic (21% O2) and hypoxic (1% O2) HEK293 cells overexpressing RPS12 (Myc-DDK-RPS12) or empty vector control (pMyc-N1). (B) Quantification of western blots from (A), normalizing band intensity of p53 to β-Actin and making relative to empty vector control. Data (n = 3), mean ± s.e.m. Two tailed unpaired t-test was performed.

1.3.4 Hypoxia induces alternative splicing of five RP mRNAs in HEK293

Another way that hypoxia could influence the regulation of RPs is through alternative splicing events (ASEs) which could introduce regulatory elements or produce alternative RP isoforms. We examined a panel of 68 ASEs in RP mRNAs in normoxic and hypoxic HEK293 cells by end-point

PCR. Primer pairs were designed based on the AceView database and flanked specific ASEs which produced long amplicons that included a certain exon or short amplicons that either excluded the exon or included an alternative shorter exon. The expression ratio of the long amplicon (L) relative to the short (S) was quantified by capillary electrophoresis and expressed as percent splicing index

푙표푛𝑔 (PSI (%) = ∗ 100). Molecular hypoxic response was confirmed by measuring the (푠ℎ표푟푡+푙표푛𝑔) expression of HIF-target SLC2A1/GLUT1 mRNA (Appendix Figure 2).

Only 6/68 ASEs displayed ≥ 10% PSI difference between normoxia and hypoxia (ΔPSI), while the majority (40/68) of ASEs had 0% ΔPSI (Figure 8). A ΔPSI of ≥ 10% has been previously used as a threshold for validation (Klinck et al., 2008). Of the six ASEs that displayed a ΔPSI of

Figure 8. Hypoxia induces alternative splicing of five RP mRNAs in HEK293. A heat map 푙표푛𝑔 representing the percent splicing index (푃푆퐼 (%) = ∗ 100 ) for 68 alternative (푠ℎ표푟푡+푙표푛𝑔) splicing events (ASEs) in HEK293. Five ASEs (boxed) had a ΔPSI (|hypoxic PSI – normoxic PSI|) ≥ 10%. Nineteen ASEs had a ΔPSI < 10%, 40 had a ΔPSI = 0%, and the ∆PSI was unable to be calculated for three ASEs due to lack of detection or undefined short/long variant.

≥ 10%, we chose to investigate five events: RPL10 e1, RPL17 e2, RPL22L1, RPS24 e2, and RPS9 e3, eliminating RPLP0 due to low detection. These five events can be described as follows: 1) An

ASE in RPL10 displayed a complete shift toward inclusion of an alternate 163-bp exon where the

PSI increased from 76.1% in normoxia to 100% in hypoxia; 2) An ASE in RPL17 displayed a reduction in the inclusion of a 17-bp-longer alternative first exon from 40% in normoxia to 26.2%

in hypoxia; 3) the ASE in RPL22L1 displayed a shift toward inclusion of a 67-bp-longer alternative third exon where the PSI increased from 15.7% in normoxia to 30.8% in hypoxia; 4) An ASE in

RPS24 displayed a shift toward inclusion of a 22-bp cassette exon where the PSI increased from

42.3% in normoxia to 56.5% in hypoxia; 5) An ASE in RPS9 displayed a shift toward exclusion of a 1,372-bp exon rather than three shorter exons where the PSI decreased from 43.8% in normoxia to 17.9% in hypoxia (Figure 9). These data demonstrate that hypoxia has minimal influence on AS in RP mRNAs in HEK293, but that the affected genes could be part of a hypoxia splicing signature. Of these five events, those in RPL10, RPL22L1, and RPS9 do not correspond to the canonical curated variants found on NCBI and were from direct submissions and cDNA library clones. The events in RPL10, RPL22L1, and RPS9 do appear to alter the coding sequence however, and the long variant in RPL22L1 is a predicted target of nonsense-mediated decay. Both the events in RPL17 and RPS24 affect the canonical curated variants (NCBI), but the coding sequence is only altered by the event in RPS24.

Figure 9. Splicing maps of alternative splicing events with hypoxic ∆PSI > 10%. Shown are the variants amplified by the primer set (harpoons) for each splicing event. The relative length of the colored/patterned bars represents the proportion of that particular amplicon, with the length in base pairs displayed in the legend. Long vs. short nomenclature is based on the size of the amplicon and not the total transcript length. White exons are common to both variants and color-coded exons are variant specific. Maps designed based on AceView.

1.3.5 Hypoxia induces alternative splicing events in RP mRNAs in cancer cell lines

We next investigated whether the hypoxia-induced ASEs in HEK293 RP mRNAs are conserved in other cell lines. We chose three cancer cell lines (U87MG glioblastoma, HCT116 colorectal carcinoma, and PC3 prostate carcinoma) due to the relevance of hypoxia in cancer and tumor biology. In addition to cell monolayers, we measured these ASEs in spheroids, avascular models of tumor hypoxia, to identify possible hypoxic tumor biomarkers. Molecular hypoxic response was confirmed by SLC2A1/GLUT1 mRNA expression in both the monolayers and spheroids

(Appendix Figure 2).

From the data in HEK293 (Figure 8), we selected five ASEs that did not change (0% ΔPSI) between normoxia and hypoxia and the five that had a ΔPSI ≥ 10% to determine their reproducibility in the cancer cell lines. Among the ASEs that displayed 0% ΔPSI in HEK293, the

PSI were also unchanged in all three cancer cell lines, regardless of oxygen, although the ASE in

RPL11 could not be detected in hypoxic PC3 cells (Figure 10). Overall, the splicing events that changed by > 10% in HEK293 displayed some variability across the cancer cell lines, but some commonalities were identified. The amplicons for RPL10 e1 were often below the detection threshold of 5 nM, likely due to this ASE not being in the canonical RPL10 mRNAs and thereby making this a poor candidate biomarker. As in hypoxic HEK293, the PSI of RPS9 e3 was decreased in hypoxic HCT116, as well as both U87MG and HCT116 spheroids, although the PSI in PC3 cells did not follow this trend. Splicing of RPL17 e2 had the same trend in all three cancer

cell lines, where the PSI decreased in spheroids but increased slightly in hypoxia, unlike HEK293 where the PSI decreased in hypoxia. The splicing event in RPL22L1 was reproduced in PC3 cells where, as with HEK293, compared to normoxia the PSI increased in hypoxia and then increased even further (by 38%) in PC3 spheroids. Similar to RPL22L1, RPS24 e2 had an increase in PSI in only hypoxic PC3 cells (as HEK293), but had a dramatic shift towards exon inclusion in spheroids of all three cancer cell lines, increasing by 48-90 % compared to normoxia. Based on these data, we chose to investigate the ASEs in RPL22L1 and RPS24 as candidate markers of hypoxia in prostate tumors, as hypoxia alone increased exon inclusion in PC3 cells, and spheroids induced even greater exon inclusion.

Figure 10. Hypoxia induces alternative splicing events in RP mRNAs in cancer cell lines. Ten alternative splicing events were chosen based on ∆PSI values from Figure 8 of HEK293 cells: Five ASEs with ∆PSI = 0% and five ASEs with ∆PSI ≥ 10% were measured in three

cancer cell lines in 21% O2, 1% O2, and spheroids.

1.4 Discussion

Recent studies have highlighted that eukaryotic ribosomes are amenable to alterations in their RP composition, and that RPs may play specialized roles in transcript selection (Chaudhuri et al.,

2007; Ferretti et al., 2017; Shi et al., 2017; Slavov et al., 2015). Since the function of many RPs is unclear in humans, it is possible that some could regulate ribosome specificity to alter gene expression in response to various stimuli, such as hypoxia. In support of ribosome heterogeneity in cancer, translational control in general has been extensively shown to be deregulated in cancer

and is currently a major target of chemotherapeutics (Bhat et al., 2015). Translation regulation has thus become fertile ground to understanding hypoxic gene expression.

In this chapter, I show that hypoxia represses 20% of RP mRNAs, alters polysome- association of select RPs, and affects alternative splicing of RP mRNAs. Although the induction of HIF-responsive genes is most studied, transcriptional repression under hypoxia does occur, albeit the mechanism is largely unknown but may involve certain transcriptional repressors

(Cavadas et al., 2017). Decreased abundance of a portion of RP mRNAs was observed under hypoxia in HEK293 cells, but this does not appear to determine total protein levels nor the association of these RPs with polysomes, except for possibly RPS12. The instability of RP mRNA expression is beginning to be recognized, especially with observed tissue-specificity (Thorrez et al., 2008). Thus our data demonstrates that use of RP mRNAs as reference in hypoxic gene expression studies must be accompanied by measurements to first ensure their stability and suitability for use.

Given the reduction in translation that occurs during hypoxia, decreased expression of RP mRNAs is unsurprising, but the fact that total RP levels did not similarly decrease conflicts traditional views of RP regulation during stress. Since virtually all RP mRNAs contain 5’TOP sequences (except for UBA52/RPL40) (Philippe et al., 2020), their translation should be decreased under hypoxia, a condition which inhibits mTOR activity (Vadysirisack and Ellisen, 2012).

Although the normalized relative expressions of RPs measured were not appreciably decreased in hypoxia, the individual measurements were somewhat variable, but a larger number of replicates were used to account for this variability. Although HEK293 are commonly used in mTOR studies

(Thoreen et al., 2009; Zhao et al., 2015), hyperactivation of mTOR due to mutation is frequent in cancer (Tian et al., 2019). We attempted to keep cell passage numbers low, but it’s possible that

cells could acquire mutations that could alter mTOR activity, since we did not specifically measure mTOR status. However, the reduction in polysome-abundance of RPs under hypoxia in our mass spectrometry data is evident, suggesting mTOR was most likely inhibited. Differences in total protein abundance likely stem from the fact that a whole-cell lysate was used (rather than cytoplasmic) to measure total RP abundance and is therefore influenced by fluctuations in ribosome biogenesis, ribophagy, and the inherent stability of ribosomal proteins (Belle et al.,

2006). The discordance between RP mRNA, total RP abundance, and polysome association in our data suggest that ribosomal protein regulation likely is not determined by transcription of RP mRNAs, a similar observation of other studies where RPs appeared to be regulated at the level of decay (Tsay et al., 1988).

Our data also highlights the potential usage of RP paralogs by human cells. Although most

RP mRNAs were decreased, an expected result given the already high expression of RPs, a RP paralog mRNA encoding RPL3L had a >2-fold increase in 2 out of 3 biological replicates.

Although not statistically significant, it warrants further exploration as RPL3L is paralogous to

RPL3, one of the first RPs to assemble into the ribosome core and has even been shown to affect peptidyl transferase function (Petrov et al., 2004). Furthermore, unlike the ubiquitous expression of RP genes, expression of RPL3L was previously determined to be tissue-specific (highly expressed in cardiac and skeletal muscle) (Van Raay et al., 1996). Due to the relevance of hypoxia in the pathology of ischemic heart disease, it would be an intriguing avenue to investigate how hypoxia might regulate RPL3L. We also detected RPL7L1, a paralog of RPL7, a universal RP which has previously been suggested to have a regulatory role in translation (Neumann et al.,

1995). Thus RPL7L1 could incorporate into ribosomes in place of RPL7, a phenomenon that may be more frequent in hypoxia based on our data.

We show that the monosome to polysome ratio of RPS12 is increased in hypoxia compared to normoxia, suggesting that this RP may be involved in regulation of hypoxic translation initiation

(Coudert et al., 2014). RPS12 is located on the outer surface in the beak of the ribosome, near the mRNA entry channel (Khatter et al., 2015; Rabl et al., 2011) and where translation initiation factors eIF3 and DHX29 bind to cooperate in scanning of structured mRNAs (Pisareva and

Pisarev, 2016). Overexpression of RPS12 in monosomes caused a greater association of APAF-1 and XIAP mRNAs with hypoxic heavy polysome fractions relative to control. RPS12 overexpression in monosomes minimally influenced the polysome distribution of these mRNAs in normoxia, suggesting a requirement for other hypoxia-induced pathways.

APAF-1 mRNA has a long, highly structured 5’ UTR that allows its translation to be cap- independent, highly eIF4A-dependent, and resistant to mTOR suppression (Lyabin and

Ovchinnikov, 2016). An IRES has been identified in the 5’ UTR of APAF-1 (Coldwell et al.,

2000), however 5’ end-dependent scanning has also been shown to occur independent of the IRES

(Andreev et al., 2012). Notably, proximity proteomics places RPS12 near eIF4A1 (a helicase that unwinds the 5’ UTR during scanning), suggesting that RPS12 is more frequent in initiating ribosomes (Padrón et al., 2019). We observed a lesser increase, only in hypoxia, of XIAP mRNA association with heavy polysomes when RPS12 was overexpressed in monosomes. This lesser increase could be due to degree of structure within the transcripts, since according to the curated mRNA variants (NCBI), the 5’ UTR of XIAP is less structured than APAF-1. Even though APAF-

1 (apoptosis) and XIAP (survival) play opposing roles in the cell, they both contain an IRES which maintains translation efficiency in hypoxia (Holcik and Sonenberg, 2005). Therefore, it is not surprising that both are translated in hypoxia and this could be part of a larger feedback loop in the tug-of-war between cell survival and death under stress. Our data suggest that RPS12 is enriched

in hypoxic monosomes to increase the translation efficiency of select transcripts. These transcripts could be selected based on the degree of structure within their 5’ UTR, although future work will be required to fully understand how hypoxic initiating ribosomes select these transcripts via ribosome-associated factors and/or additional features within mRNAs.

The exact function of RPS12 in ribosomes is not known, but it has recently been studied in

Drosophila melanogaster for its role in cell competition, a process which eliminates mutant cells via apoptosis. RPS12 is one of only a few ribosomal proteins that are not haploinsufficient in

Minute loci (Marygold et al., 2007). Following chemically-induced mutagenesis, a mutation was found in a region of RPS12 that prevented competitive elimination of cells with heterozygotic RP inactivation (minute mutants) (Kale et al., 2018). Despite this mutation being close to the rRNA binding loop of RPS12, it did not alter the ability of RPS12 to incorporate into polysomes, thus suggesting RPS12’s role in cell competition could possibly be extra-ribosomal. Further, RPS12 in yeast is non-essential and although lack of RPS12 sensitized cells to the growth-inhibiting effects of cycloheximide, it did not impair polysome assembly (Ferreira-Cerca et al., 2005). Similar to our findings where overexpression of RPS12 increased the heavy polysome association of APAF-1, the researchers also found that increasing the copy number of wildtype RPS12 caused increased apoptosis in minute mutant mosaics (Kale et al., 2018).

Despite the possibility that our results stem from RPS12 in ribosomes, we must acknowledge that they could also be due to extra-ribosomal roles of RPS12 in other RNPs of similar mass. In our data, the polysome association of RPS12 is much weaker in comparison to other RPs. It is interesting to note the predicted isoelectric point of RPS12 protein (7.21 -

Appendix Table 1), which would make RPS12 negatively charged at physiological pH (~7.4).

Nearly all RPs are highly positively charged to facilitate binding to negatively charged rRNA. The

well-known exceptions to this rule are the acidic phosphorylated proteins RPLP0, RPLP1, and

RPLP2. Out of the remaining proteins, just two other RPs are predicted to be negative at physiological pH (7.4): RPSA and RPS12. RPSA is well-studied for its extra-ribosomal roles, evidenced by its aliases such as the 67-kda laminin receptor and Multidrug Resistance-Associated

Protein (Digiacomo and Meruelo, 2016; Sun et al., 2014). Could RPS12 be weakly associated with ribosomes due to its negative charge? Would this facilitate its movement between ribosomes and free-floating forms in the cytoplasm? Its position on the periphery of the ribosome, and the fact that the beak undergoes remodeling in the cytoplasm suggests this could be possible (Panse and

Johnson, 2010).

It was not surprising that most ASEs examined in RP mRNAs were conserved between normoxia and hypoxia due to the important role of the ribosome. When investigated in cancer cell lines, the genes with complete exon inclusion/exclusion (i.e. PSI = 0 or 100%, ∆PSI = 0) retained the same degree of splicing, suggesting these genes are tightly regulated and the cancer cell lines investigated did not have mutations that altered their splicing. There were some events that showed

< 10% ∆PSI which could possibly be attributed to erroneous or stochastic splicing. Except for

RPL10, the non-curated variants measured for RPS9 and RPL22L1 were detected at relatively high amplification, providing additional evidence to the existence of these variants. There was variation in the five ASEs when investigated in cancer cell lines, but there are a few events that warrant further investigation; The event in RPL17 displayed decreased exon inclusion in spheroids of all three cancer cell lines, but this effect was not reproduced in hypoxic monolayers suggesting it might not be caused by hypoxia specifically. Although the event in RPL17 does not alter the coding sequence, it could still alter the expression of RPL17 protein through modulation of cis-regulatory sequences in the 5’UTR. The event in RPL22L1 was reproduced only in PC3, where there was a

modest increase in exon inclusion in hypoxia (as HEK293) and further increase in spheroids. The long variant of RPL22L1 is a predicted target of nonsense-mediated decay, thus exon inclusion could function to repress RPL22L1 levels. Further, RPL22L1 has been proposed to compensate for its paralog RPL22 in surprisingly viable RPL22-null mice, as it can incorporate into polysomes

(O’Leary et al., 2013). Alternative splicing of RPS24 similarly had an increase in exon inclusion in hypoxic PC3 cells, as well as spheroids of all three cancer cell lines. Based on the cell line data, the alternative splicing events in RPL22L1 and RPS24 have potential to be biomarkers of hypoxia in prostate cancer and will be measured in human prostate normal tissue and tumors. Additionally, the alternative splicing event in RPS24 was the only to affect the canonical curated mRNA variants as well as alter the coding sequence, and is therefore more likely to have a functional role.

How is it that splicing of ribosomal proteins could be altered by hypoxia? Ribosomal proteins are known to regulate splicing of their own pre-mRNA in an autoregulatory feedback loop

(Ivanov et al., 2006). Addition of recombinant ribosomal proteins to nuclear extracts of yeast and

HeLa extracts led to intron retention by binding of the proteins to their own pre-mRNA. Many of the splicing events in pre-mRNAs mediated by their own ribosomal protein products resulted in nonsense-mediated decay or inactive mRNAs (Ivanov et al., 2006). It was proposed that when ribosomal proteins are produced in excess, they can bind to their respective pre-mRNAs to modulate splicing and decrease expression as a compensatory feedback mechanism (Fewell and

Woolford, 1999). Hypoxia-specific splicing factors may exist as well, such as SRSF3, CLK1, and

CLK3 (Bowler et al., 2018; Brady et al., 2017). Further investigation to identify exact regulatory mechanisms of these alternative splicing events in RPL22L1 and RPS24 will be essential to understand whether they provide selective advantage to hypoxic cells or are simply a consequence of the cellular environment.

CHAPTER 2 – ALTERNATIVE SPLICING OF RP MRNAS AND THE TUMOR MICROENVIRONMENT

2.1 Introduction

2.1.1 Alternative Splicing

Discovered over 40 years ago, splicing of pre-mRNAs is a critical step in co-transcriptional gene regulation (Berget et al., 1977). Alternative splicing, the production of multiple mature mRNAs from a single pre-mRNA, is now estimated to occur in ~95% of multiexon human genes and is considered a main effector of increasing genomic complexity in higher eukaryotes (Pan et al.,

2008). Alternative splicing also contributes to proteomic diversity, as 75% of exon-skipping events were found to be engaged in translation with ribosome profiling experiments (Weatheritt et al.,

2016).

Eukaryotic mRNAs undergo constitutive splicing, RNA processing which removes introns and join exons co-linearly, whereas alternative splicing includes 7 main types: cassette/alternative exons (most common form in vertebrates), mutually exclusive exons, alternative 3’ splice site, alternative 5’ splice site, intron retention, alternative promoters, and alternative polyadenylation

(Blencowe, 2006). The process of splicing is facilitated by the spliceosome, a large and complex molecular machine that in humans contains 5 small nuclear RNAs (U1, U2, U4, U5, and U6) and

>300 proteins which recognize short consensus sequences that define intron-exon boundaries to recognize and splice pre-mRNA (Jurica and Moore, 2003). The consensus sequences include the

5’ splice site (GU beginning intron), 3’ splice site (AG ending intron), branch point (invariant A found 20-50 bases upstream of 3’splice site), and a pyrimidine tract (10 – 12 bases long) near the

3’ end of the intron. The GU-AG rule is followed by 99% of pre-mRNAs in humans. The process of splicing occurs via two transesterification reactions: The first reaction joins the 5’ splice site G to the branch point A, forming a lariat of the intron, and the second joins the 3’ end of exon 1 to

the 5’ end of exon 2, resulting in removal of the intron lariat (Lodish et al., 2012). The spliceosome is formed by complementary binding of U1 to the 5’ splice site, Splicing Factor 1 to the branch point, and U2 Associated Factor to both the pyrimidine tract and the 3’ splice site. U2 then binds to the branch following release of SF1. U4/U6/U5 form a heterotrimeric complex which then binds to form the spliceosome. Subsequent rearrangement steps allow for activation and catalysis of the two transesterification reactions (Lodish et al., 2012). Following splicing, heterogeneous nuclear

RNPs remain bound to exon-exon junctions, acting as the exon-junction complex which are recognized during the quality control pioneer round of translation. Proper recognition of the exon junction complex regulates the process of nonsense-mediated decay, a protective process to eliminate the synthesis of truncated proteins (Maquat et al., 2010).

Alternative splicing decisions are mediated by cis-regulatory sequences which act as binding sites for regulatory proteins. For example, exonic splicing enhancers promote exon inclusion via binding of SR proteins. Named for having domains enriched in serine (S) and arginine

(R), there are 9 known SR protein genes in humans: SRSFs 1-7, SRSF9, and SRSF11 (Shepard and Hertel, 2009). In addition, many other SR-family and SR-related proteins exist that can also bind exonic splicing enhancers or act as co-activators (Blencowe, 2000). Other cis-regulatory sequences exist to fine-tune alternative splicing and include exonic splicing silencers, intronic splicing enhancers, and intronic splicing silencer (Chen and Manley, 2009).

2.1.2 Alternative Splicing in Cancer

Despite the tightly regulated and intricate process of alternative splicing, aberrant splicing is heavily implicated in the pathology of multiple diseases, including cancer. A comprehensive study of nearly 9 000 cancer patients and 32 cancer types showed that tumors have 20-30% more alternative splicing events than normal tissues (Kahles et al., 2018). Mutations in splice sites, the

introduction of cryptic splice sites, mutations in exonic splicing enhancers, and deregulation of splicing factors have all been implicated as causative mechanisms for changes to alternative splicing in cancer (El Marabti and Younis, 2018; Venables, 2004). For example, splice site selection is known to be altered in a multitude of well-known cancer-related genes such as TP53,

BRCA1, VEGF-A, and Bcl-x (Liu and Cheng, 2013).

Although they observed more alternative splicing events and novel exon-exon junctions in tumors, Kahles and colleagues postulated that a large proportion of the signal was likely due to stochastic noise stemming from errors in splicing machinery, rather than a driving force of malignancy (Kahles et al., 2018). Apparent random splicing of the progesterone and estrogen receptor α observed in breast cancer supports this stochastic model, but even if not the driving force, aberrantly spliced transcripts can still provide selective advantages to tumors and therefore be targeted therapeutically (Venables, 2004). Further, although determining the functional implications of alternative splicing in cancer is a more onerous task, accumulating evidence shows that alternative splicing events may be suitable as cancer biomarkers. Alternative splicing events have been effectively used to differentiate between normal and cancerous tissue in blind studies, and also display associations with tumor grade and survival times (Klinck et al., 2008; Venables et al., 2008; Wu et al., 2019). Due to the heterogeneity of cancer, more biomarkers are always needed, particularly if they can be linked to tumor features which have poor prognostic outcomes, such as hypoxia.

2.1.3 Tumor Microenvironment

The tumor microenvironment is an incredibly complex network of multiple cell types, interactions, and stressors which facilitates tumor growth and metastasis. The concept of a tumor microenvironment was first described by Stephen Paget over a century ago; “When a plant goes

to seed, its seeds are carried in all directions; but they can only live and grow if they fall on congenial soil.”(Paget, 1889). Paget’s hypothesis contradicted the previous standing theory that metastasis was simply based on distribution through the vasculature system when he noticed that the frequency of metastases from breast cancer did not correlate with relative blood flow in the site of metastasis. He postulated that the features of the original organ’s microenvironment dictated location of metastasis and these sites were not completely random (Fidler and Poste, 2008). We now have a better appreciation for the importance of interactions of cancer cells with surrounding stroma, immune cells, extra-cellular matrix, inflammatory cells, blood vessel cells, and cancer- associated fibroblasts (Mueller and Fusenig, 2004).

Solid tumors also have an accumulation of stressors that contribute to the aggressiveness of cancer cells. As oxygen’s diffusion limit is 100-200 μm (Naumov et al., 2006), even the smallest tumors develop regions of hypoxia. Rapidly dividing cancer cells outgrow their vasculatures and although hypoxia stimulates angiogenesis via induction of VEGF-A, newly formed blood vessels are usually dysfunctional and leaky (Weis and Cheresh, 2011). As a result, delivery of oxygen and nutrients to tumor cells is severely impaired. A switch from aerobic respiration to anaerobic glycolysis leads to the production of lactic acid. Reduced waste efflux in tumors means lactic acid accumulates and this, in conjunction with hydration of CO2 in oxygenated regions producing carbonic acid, decreases the extracellular pH in solid tumors (Corbet and Feron, 2017). This is further augmented by the Warburg effect, where cancer cells ferment glucose into lactate even in the presence of oxygen (Warburg, 1925). While the intracellular pH of tumors is virtually unchanged from that of normal tissues, electrode measurements of the extracellular pH of tumors showed that tumor pH was almost always equal to or lower than that of adjacent normal tissue.

The mean pH difference of tumors was a decrease of 0.4 compared to normal, although both the

inter- and intra-tumoral pH was quite variable, the latter likely due to differing distances from vasculatures and accompanying hypoxia (Gerweck and Seetharaman, 1996). For example, glioblastoma tumors measured pH values ranging from 6.4 – 7.3 whereas normal brain tissue ranged from 6.6 – 7.6. Astrocytomas measured from 5.8 – 7.2, compared to 6.3 – 7.8 in normal brain tissue (Pampus, 1963). In addition to decreasing pH, lactate itself regulates inflammation and can even act as a signaling molecule by binding and activating G-protein coupled receptor

GPR81 (Sun et al., 2017).

The acidic and hypoxic tumor microenvironment confers multidrug resistance to tumor cells, reducing chemotherapy effectiveness. Acidic extracellular pH alters the activity of ionizable drugs, reducing the uptake of weak base therapies such as the anthracycline doxorubicin

(Wojtkowiak et al., 2011). Activation of the HIF-pathway in tumors stimulates the expression of drug efflux pumps and inhibition of apoptosis, allowing for cells to circumvent chemotherapy- mediated cytotoxicity (Rohwer and Cramer, 2011). Additionally, poorly vascularized tumors prevent drug delivery to hypoxic regions. Despite the negative impact of the tumor microenvironment on drug efficacy, it provides a unique niche that can be exploited with targeted therapies that should theoretically spare normal tissue. Hypoxia-activated prodrugs (such as evofosfamide) are inert under normoxic conditions and become cytotoxic under low oxygen, and delivery via nanoparticles can overcome reduced tumor vascularization (Huang et al., 2018; Wang et al., 2019).

Important cellular tools often used in cancer drug discovery are spheroids, 3-dimensional

(3D) cell culture models of tumor hypoxia. Due to its simplicity, monolayer (2D) cell culture is usually the starting point of basic research, but cells in a single flat layer on polystyrene is hardly reflective of the cellular environment in vivo. Unlike monolayers, spheroids are more

physiologically representative of many facets of solid tumors, mimicking 3D cell contacts and the gradients of nutrients, oxygen, and metabolites found in solid tumors (Mehta et al., 2012).

Spheroids are homogenous avascular tumor models, so they do not contain multiple cell types, but they do have tumor features such as hypoxia, acidosis, enhanced cadherin-mediated cell-cell interactions, increased extra-cellular matrix deposition, and presence of cancer stem cells. Since all of these features are often associated with malignancy, spheroids are effective tools for studying tumor growth kinetics, migration and invasion, and anticancer drug delivery (Nunes et al., 2019).

2.1.4 RPS24

RPS24/eS24 is a 15-kDa component of the 40S ribosome small subunit. Often studied for its mutation in Diamond-Blackfan Anemia (Gazda et al., 2006), RPS24 is required for ribosome biogenesis but an exact function in the ribosome has not yet been identified (Choesmel et al.,

2008). RPS24 is located on the periphery of the ribosome, in proximity to the mRNA entry channel

(Khatter et al., 2015). RPS24 is alternatively spliced to produce 3 main splice variants which differ in their C-termini: short mRNA variant C (…VGAGKKPKE), long mRNA variant A

(…VGAGKK), and extra-long mRNA variant B (…VGAGKKK) (Gazda et al., 2006; Gupta and

Warner, 2014). The inclusion of a 22-bp cassette exon in the long mRNA variant introduces a slightly premature stop codon (insufficient for nonsense-mediated decay) and pushes the end of the coding sequence into the 3’ UTR, truncating the C-terminus by 3 amino acids. The extra-long variant, which includes an 18-bp cassette exon in addition to the 22-bp exon, truncates the C- terminus and adds an additional lysine. The C-terminus is exposed on the surface of the ribosome, but it is not highly conserved across species so the C-terminus is likely not critical to the function of RPS24 (Gupta and Warner, 2014).

RPS24 exhibits tissue-specific alternative splicing, with the long variant being most- predominant in the brain, the extra-long variant in heart and skeletal muscle, and the short variant in liver and kidneys (Gazda et al., 2006; Gupta and Warner, 2014; Xu and Roufa, 1996). Further, an investigation into intron retention within ribosomal protein genes found RPS24 as the sole RP to display tissue-specific splicing (Gupta and Warner, 2014). Splicing of RPS24 is also associated with cell differentiation, where exon inclusion (i.e. Increased expression of long variant) increased within differentiated neurons compared to stem cells (Gonatopoulos-Pournatzis et al., 2020; Song et al., 2017). To what extent the microenvironment of different tissues affects RPS24 splicing has yet to be investigated.

In this chapter, I show how RPS24 and RPL22L1 splicing are altered in human prostate tumors and measured the impact of different spheroid features on RPS24 splicing. I show that

RPS24 is the better candidate biomarker and warrants further study due to the dramatic and reproducible increase in expression of the long RPS24 variant in spheroids of multiple cell lines, although additional features besides hypoxia may be involved in its regulation. In preliminary data from work that is still ongoing, I show that acidosis may contribute to the RPS24 long induction seen in spheroids, and it is likely not due to lactic acid acting as a signaling molecule or simply increased cell confluency.

2.2 Methods

2.2.1 ASPCR and RT-qPCR of Origene Human Prostate Tumor cDNA Panels

Splicing of RPS24 and RPL22L1 in human prostate tumors was measured using Origene

TissueScan prostate tumor cDNA panels II and III. Since cDNA was loaded into wells based on

β-actin, the expression was already normalized and we directly correlated the Ct value of CAIX with the percent splicing index of RPS24 and RPL22L1 obtained from ASPCR. The extra-long

variant of RPS24 was more abundant in the human tissue samples than in the cell lines, so two PSI were calculated, each reflecting the inclusion of a single exon. The molarity (nM) of an unpredicted

(~150 bp) amplicon from RPL22L1 was directly correlated with the Ct value of CAIX as a PSI could not be calculated.

2.2.2 Spheroid Formation

Spheroids were generated using the liquid overlay method with brief agitation. Cells (50 000) were seeded in round-bottom, low attachment 96-well plates (Corning), quickly rotated by hand for 60 s, and grown for 5 days for validation of the ASPCR, or 1-7 days to monitor RPS24 long induction over spheroid growth. Approximately 36 spheroids were combined per sample, after adding

Ribozol the samples were passed through a 27 gauge needle to break up spheroids.

2.2.3 Cloning and Generation of RPS24 Stable Cell Lines

PCR-based cloning was used to insert the coding sequence of the RPS24 short or long variant into p3XFLAG-CMV-10puroE- with N-terminal tag. The coding sequences of the short and long

RPS24 variants were amplified from cDNA from HEK293 cells using forward and variant-specific reverse primers with restriction enzyme sites attached (Appendix Table 6) and Q5 High Fidelity

DNA Polymerase (New England Biolabs). Restriction enzyme digest was followed by separation of fragments by gel electrophoresis. Inserts were excised and purified using a gel purification kit

(BioBasic), followed by ligation using T4 DNA Ligase (New England Biolabs). Successful cloning of the two variants was confirmed via sequencing. Stable cell lines were generated in

U87MG by individual transfection of empty plasmid backbone, FLAG-RPS24 Short (referring to mRNA variant length, not protein), and FLAG-RPS24 Long (mRNA variant length) with

Lipofectamine 2000 (Invitrogen). Selection in 1 μg/ml puromycin, followed by limited dilution to generate single cell clones allowed for expansion and generation of homogenous single colonies.

2.2.4 Acidosis

Cells were treated with acidosis-permissive media, which when added to cells is at neutral pH, but decreasing buffer concentrations allow for a gradual decrease in media pH resulting from the cells’ own metabolic processes (modified from Mekhail et al., 2004). Over 24 h, this resulted in a pH gradient where the media self-adjusted to a final pH as follows: 5.8, 6.0, 6.3, 6.5, 6.8, and 7.0.

Powdered media was made with four concentrations of NaHCO3 (16 mM, 22 mM, 34 mM, and 40 mM) in two batches, one of which was adjusted to pH 6.0 (AP) and the other to pH 7.2 (SD). Air was bubbled into the pH 6.0 AP media until it increased to pH 7.2, as well as briefly bubbled into pH 7.2 SD media to stabilize the pH. The media were then sterilized by vacuuming through 0.2

μm bottle top filters. Serum (7.5% fetal bovine serum and 1% penicillin-streptomycin) was added to media following filtration. Use of the following media resulted in a pH gradient after 24 h of incubation with cells: 16 mM AP (pH 5.8), 34 mM AP (pH 6.0), 40 mM AP (pH 6.3), 22 mM SD

(pH 6.5), 34 mM SD (pH 6.8), and 40 mM SD (pH 7.0). Cells (350 000) were seeded into 6-well plates 24 h before treatment, followed by subsequent addition of acidosis-permissive media and incubation in normoxia or hypoxia for 24 h before harvesting.

2.2.5 Lactic Acidosis

Cells were grown in 100 mm dishes until 80% confluent. Prior to treatment, cells were maintained in serum-free media for 24 h. Cells were then treated for 24 h with 20 mM Lactic Acid pH 7.1

(Fisher Chemicals), 20 mM Sodium Lactate pH 7.4 (Fisher Chemicals) or 25 mM Sodium

Oxamate (Alfa Aesar) (Seliger et al., 2013). Media pH was measured using pH strips 2 h after commencing treatment and prior to lysis to ensure neutral pH and adjusted if necessary.

2.2.6 Cell Confluency

Cells (500 000) were seeded into 5 x 100 mm plates. On each of the 5 following days, a single plate was lysed, and the media was refreshed on the remaining plates to remove dead cells and reduce waste buildup. Brightfield microscopy images were taken to visualize confluency.

2.2.7 Statistical Analyses

All statistical analyses were performed using GraphPad Prism 7.0. RPS24 RT-qPCR experiments were analyzed for statistical significance using one-way ANOVA with Tukey’s HSD test on the

∆Ct values from qPCR data. Prostate tumor samples were analyzed by correlating CAIX Ct value with PSI using Spearman’s correlation coefficient. Tumor stage and PSI were compared using

Kruskal-Wallis test (One-way ANOVA) followed by Dunn’s multiple comparisons test. P-value

<0.05 was considered statistically significant.

2.3 Results

2.3.1 Exon inclusion within RPS24 and an unpredicted amplicon in RPL22L1 correlate with hypoxia in prostate tumors

Since both RPS24 and RPL22L1 had increased PSI (∆PSI >10%) in hypoxic monolayers, and an even greater PSI increase in PC3 spheroids, we chose to investigate these events in a panel of human prostate tumors as possible hypoxic biomarkers. We used ASPCR to measure the events in

RPS24 and RPL22L1, and RT-qPCR to measure the expression of established hypoxic marker carbonic anhydrase-9 (CAIX) in 96 cDNA samples from human prostate tumors or normal tissue

(Figure 11). In total: 17 normal samples, 40 stage II, 30 stage III, 4 stage IV, and 5 stage not

reported. To determine the relationship between hypoxia and alternative splicing, we measured the correlation between CAIX expression (Ct values, as cDNA was normalized to β-Actin) and the

PSI obtained from ASPCR. For RPS24, a less abundant extra-long (XL) amplicon (includes not only the 22-bp cassette exon but also an 18-bp cassette exon) that was detected in HEK293 (Figure

9) was a major amplicon in the tumor samples. Therefore, we calculated two PSI for RPS24: L/S+L and XL/L+XL (Figure 11A). Overall, the normal tissue samples (blue dots) cluster fairly tightly, whereas the tumor samples are more variable, consistent with reports of aberrant splicing in tumors

(Kahles et al., 2018). For both RPS24 and RPL22L1, the PSI was weakly negatively correlated with the CAIX Ct value, albeit only significantly with RPS24 (Figure 11A and C). This suggests that hypoxia favors exon inclusion as the more hypoxic samples (i.e. Lower CAIX Ct) had higher

PSI values, similarly to the cell line data. Conversely, both PSI in RPS24 and the PSI in RPL22L1 decreased significantly in stage II and III tumors compared to normal tissue, suggesting that exon exclusion is favored (Figure 11B and D). Hypoxia does not always correlate with stage (Höckel et al., 1996), and indeed 13/17 of the hypoxic samples (i.e. Ct < 31) were from stage II tumors.

Therefore, it was not surprising that the exon inclusion we observed in hypoxia was not more common at later cancer stages.

An unpredicted amplicon in RPL22L1 (~150-bp) was detected through ASPCR, and this amplicon was strongly associated with the degree of hypoxia within the tumors. In fact, this unpredicted amplicon only appeared in samples that had >2-fold increase in CAIX expression compared to normal. If the correlation was measured only in the “hypoxic” tumors (i.e. >1 Ct value lower than normal tissue value of ~31), the correlation was even stronger (r = -0.9118, P =

<0.0001). Although the cDNA from the plates were quality controlled and supposed to be free from genomic DNA contamination, initial measurements of CAIX (prior to RPL22L1 ASPCR)

using primers that did not flank exon-exon junctions resulted in a second amplicon in the hypoxic samples that was sequenced and found to correspond to genomic DNA (data not shown). Since the

ASPCR primers did not flank exon-exon junctions in order to pick up alternative exon inclusion, it was essential that samples were free from DNA as genomic contaminants could be detected if primers matched to regions of the chromosome. CAIX primers were re-designed to flank exon junctions (primers used in this thesis – Figure 11), and new plates (with associated quality control assurance) were ordered before performing ASPCR for RPL22L1. The new plates were of the same lot as those with potential genomic DNA contamination. Curiously, the relative expression of CAIX was not changed, thus the hypoxic samples as well as those containing RPL22L1 were mostly the same as those that appeared to contain genomic DNA contamination before.

Nonetheless, due to the very strong correlation between this unpredicted amplicon of RPL22L1 and CAIX, proper validation was required to determine the origin of the unpredicted amplicon.

These data show that exon inclusion in RPS24 and an unpredicted amplicon in RPL22L1 have potential as biomarkers of hypoxia in human prostate tumors. We next sought to validate these data in spheroids and use cell lines to investigate the regulation and functional implications of these splicing events.

Figure 11. Exon Inclusion within RPS24 and an Unpredicted Amplicon in RPL22L1 Correlate with Hypoxia in Prostate Tumors (A) An alternative splicing event in RPS24 produces three amplicons in human prostate normal and tumor cDNA samples: short (S), long (L), and extra-long (XL). Because the extra-long amplicon was more abundant than in the cell lines (Figure 9), two PSI were calculated (L/S+L and XL/L+XL) each reflecting the inclusion of a single exon. Each PSI was correlated with the Ct (cycles to threshold in PCR) value of the hypoxia marker CAIX. The Spearman's correlation coefficient (r) and associated p-value is shown for 94 samples: normal tissue (n=16), stage II (n=40), stage III (n=29), stage IV (n=4), cancer stage not reported (n=5). (B) Box plots compare RPS24 PSI with tumor stage using one- way ANOVA (Kruskal-Wallis) with Dunn's multiple comparisons test. + represents mean. (C) The PSI of RPL22L1 was correlated with CAIX Ct values as with RPS24. A 150 bp unpredicted amplicon was detected in hypoxic tumors and was also correlated with CAIX Ct value. (D) The PSI of RPL22L1 was compared across tumor stage as described for RPS24. * = P < 0.05, ** = P < 0.01, *** = P < 0.001 2.3.2 The long variant of RPS24 is significantly upregulated in spheroids of four cell lines

We next sought to validate the levels of RPS24 short and long splice variants using variant-specific primers and quantitative PCR in hypoxia and spheroids of the four cell lines from Figure 10.

Although a third extra-long variant of RPS24 exists, the inclusion of only the 22-bp cassette exon of the long variant was by far the most predominant splicing event and used to calculate the PSIs for Figures 9 and 10. The extra-long variant was only detected in appreciable amounts in HEK293

(Figure 9), but not the other cell lines, so we chose to focus only on RPS24 short and long for experiments in cell lines.

Equal numbers of cells were seeded and spheroids were harvested 5 days post-seeding.

Spheroids were all ≥500 μm in diameter (large spheroids that are more reflective of tumors) and cell lines were varied in their ability to form dense, compact spheroids (Figure 12A) (Zanoni et al., 2016). HCT116 and U87MG formed the most dense and compact spheroids, whereas PC3 formed looser spheroids. We also included spheroids of HEK293 (not included in ASPCR screen), which although are not cancer cells, formed spheroids comparable in size to spheroids formed using the cancer cell lines (Figure 12A).

The long variant of RPS24 containing a 22-bp cassette exon increased by 1.4 to 2.1-fold in hypoxic samples of all cell lines (Figure 12B). The long variant of RPS24 was not detected in U87 monolayers with ASPCR, but was consistently present upon validation using RT-qPCR.

Consistent with the ASPCR data (Figure 10), the long variant of RPS24 increased ≥ 4-fold in spheroids of all cell lines (Figure 12B). Although hypoxia modestly increases expression of

RPS24 long, it is < 2-fold, thus ultimately it may be regulated by a combination of other features of the spheroid microenvironment (alone or in conjunction with hypoxia). Of note, the short variant of RPS24 decreased in hypoxia relative to normoxia in U87MG and PC3 monolayers and spheroids, further increasing the ratio of long to short RPS24 variants. Therefore, the RPS24 long/RPS24 short variant ratio could be a better predictor of hypoxia and/or the tumor microenvironment than the induction of the RPS24 long variant alone. Although U87MG had the lowest increase in RPS24 long of all the cell lines, it had the most significant and substantial concomitant decrease in RPS24 short under both hypoxia and spheroids. This would further increase the ratio of long to short, a switch in alternative splicing, and in conjunction with RPS24 long being the most predominant variant in the brain (Gupta and Warner, 2014) we chose U87MG as the initial cell line of study for specific assays and making stable cell lines.

Figure 12. The long variant of RPS24 is significantly upregulated in spheroids of four cell lines. (A) Spheroids from four cell lines were imaged (4x magnification) and harvested 5 days following cell seeding. (B) RT-qPCR using variant-specific RPS24 primers were used to confirm the ASPCR data. The ∆∆Ct method was used, normalizing to reference genes RPLP0 and RPL13A, and the normalized expression made relative to normoxia. Leslie Fell performed the RT-qPCR for HCT116. Data (n = 8), mean normalized relative expression ± s.e.m. One-way ANOVA with Tukey’s HSD test was performed on the ∆Ct values. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001. Scale bar, 500 μm.

2.3.3 The unpredicted amplicon in RPL22L1 was not detected in cell lines or other tumor types

Since the ~150 bp unpredicted amplicon detected in human prostate tumors was so strongly associated with hypoxia, we attempted to detect it in cell lines so we could identify this amplicon

and further study its regulation. The primers used in ASPCR match to the processed pseudogene

RPL22P13, which would produce a 143-bp amplicon (~150) (Figure 13A). Although pseudogenes are often regarded as inert, processed pseudogenes occur due to mRNA retrotransposition and, although lacking protein-coding potential, they are often transcribed and can affect the regulation of other genes (Harrison et al., 2005).Very little information is available regarding RPL22P13, and despite lacking histone marks (found near active regulatory elements), its transcription has been detected via RNA-Seq of some cell lines (ENCODE).

This unpredicted amplicon was not previously detected in any of the samples we used for

ASPCR (i.e. HEK293, U87MG, HCT116, or PC3 cell lines), so we first obtained 3 additional prostate cancer cell lines (DU145, LNCaP, and 22RV1), exposed them to 21% and 1% O2, and performed endpoint Taq-based RT-PCR using the ASPCR primers for RPL22L1 (Figure 13). It is interesting to note that the long variant of RPL22L1 does not appear in LNCaP and 22RV1 cells, and when it is detected in DU145 it is a doublet, indicating an additional unpredicted band ~228 bp. We did not detect a 143-bp band in any of the samples measured (Figure 13B). Spheroids were made of DU145 cells, the only one of the three additional prostate cancer cell lines expressing the long variant, but RNA could not be obtained from them (data not shown). In case the unpredicted amplicon was characteristic of human tissue samples, we also attempted to detect it in a normoxic breast tumor, and hypoxic colorectal and lung tumors (indicated by 25- and 170-fold increased

CAIX expression, respectively) (Figure 13C - obtained from Ontario Institute for Cancer

Research). Interestingly, these data demonstrate potentially tissue-specific alternative splicing of

RPL22L1, as the long variant was not detected in the colorectal tumor sample. Ultimately, we were unable to detect the 143-bp band in these three tumor samples, thus we were unable to confirm its origin and whether it was indeed RPL22P13.

Since we were trying to resolve a very small difference between amplicon lengths, we wanted to confirm lack of detection was not due to technical issues regarding gel electrophoresis.

If the unpredicted amplicon were indeed RPL22P13, it would be 18-bp smaller than the 161-bp short amplicon, and as shown with ASPCR primers detecting RPS24 short, long, and other PCR products of similar band differences, as small as 17-bp difference in ~100-150-bp amplicons would have been resolved by a 3% agarose gel (Figure 13D). The unpredicted amplicon was ~150 bp so if it was not in fact RPL22P13, I would likely not have separated an 11-bp difference and it would be worth repeating experiments and using a more sensitive technique such as capillary electrophoresis followed directly by sequencing. But the possibility remains that RPL22P13 was not transcribed and was an artifact due to genomic contamination of the cDNA as the RPL22L1

ASPCR primers do not flank exon-exon junctions. Further, although cell line RNA samples were not DNase-treated, the ASPCR primers are in two separate exons so the lack of a 324-bp amplicon

(i.e. containing 163-bp intron between these two exons) indicates cell lines did not have genomic

DNA carry-over and could explain lack of RPL22P13 detection. Since we were unable to detect the unpredicted amplicon that was so strongly associated with hypoxia, we could not confirm its identity (as RPL22P13) nor study its regulation in cell lines, so we chose to focus only on RPS24 moving forward.

Figure 13. The unpredicted amplicon of RPL22L1 was not detected in cell lines or other tumor types. (A) BLAST alignment of RPL22P13 with RPL22L1. Primers are underlined, the region of RPL22L1 that RPL22P13 is lacking is highlighted in red, and the yellow box represents a SNP that can be found as G instead of A which would allow for complete primer binding. RPL221 short amplicon is 161-bp whereas RPL22P13 amplicon would be 143-bp. (B) Amplification of RPL22L1 (ASPCR primers) in three human prostate cancer cell lines exposed to 21% and 1% O2. (C) Amplification of RPL22L1 in three human tumor cDNA samples. (D) Amplification of RPS24 short and long (ASPCR primers RPS24 e2 - HEK293) where the amplicons are 112- and 134-bp, respectively. RPS24 XL was not detected, and primers producing a single PCR product of similar sizes to RPS24 L and XL are included to demonstrate that the upper band is RPS24 L and not XL, and that a 17-bp difference can be resolved.

2.3.4 Endogenous RPS24 short and long transcript variants, and exogenous protein isoforms associate with polysomes

In order to justify moving further with studies into RPS24, we first wanted to confirm that both endogenous transcripts are actually translated into protein and also that protein isoforms are

functional ribosomal proteins. To determine if endogenous transcripts were translated into protein, we isolated polysomes in the presence of cycloheximide or EDTA, extracted RNA from the fractions, and performed semi-quantitative RT-PCR for the two endogenous RPS24 variants.

EDTA chelates magnesium required for ribosome stability, so it is used to dissociate ribosomes.

Puromycin is often used for this purpose, but the stable cell lines were resistant to puromycin, so an alternative was required. Typically EDTA is not specific to actively translating ribosomes and also dissociates other RNPs (i.e. Monosomes) and would not be ideal for this experiment, but from the polysome profiles obtained (Figure 14A-B), the use of EDTA with regular magnesium- containing Basic solution left 40S, 60S, and 80S particles intact. Treatment with EDTA caused both RPS24 short and long to shift out of polysomes towards the earlier monosome fractions, indicating that the two transcripts were being actively translated by polysomes into protein (Figure

14A).

Since the short and long variants produce protein isoforms that differ by only 3 amino acids, and the long is simply lacking 3 amino acids found in the short, it is difficult to differentiate between the two endogenous protein isoforms. Thus we generated U87MG cell lines stably expressing N-terminal FLAG-tagged versions of the two variants. The protein isoforms are always referred two by the length of the mRNA variant and not the protein length. Upon treatment with

EDTA, both isoforms shift out of polysomes and into monosomes, suggesting ribosome incorporation (Figure 14B).

Figure 14. Endogenous RPS24 short and long transcript variants, and exogenous protein isoforms associate with polysomes. (A) Using transcript variant-specific RPS24 primers, short and long variants were amplified via RT-PCR from wildtype U87MG polysome fractions with either cycloheximide, or EDTA as a control to dissociate polysomes. (B) Polysomes were isolated from two cell lines each stably expressing one of the FLAG-tagged RPS24 protein isoforms (eS24) using cycloheximide and EDTA (cell lines made by Lindsay Obress). Western blots were performed using anti-FLAG, or anti-β-Actin as a control protein not associated with polysomes. Experiments performed in U87MG glioblastoma in normoxia (21% O2).

2.3.5 RPS24 switches splice variant abundance over 7 days of spheroid growth

Although exon inclusion in RPS24 was significantly weakly correlated with hypoxia in human prostate tumors, we decided to return to cell lines and homogenous spheroids where the effect on

RPS24 splicing was very strong in order to study its regulation. We previously validated the

ASPCR data for RPS24 using spheroids grown for 5 days, but we were curious to monitor the induction of RPS24 long expression over 7 days of spheroid compaction.

We plated spheroids, as described previously, and harvested them on Days 1, 3, 5, and 7 post-plating, and compared the expression of the two RPS24 splice variants in spheroids to a normoxic monolayer (Figure 15). The data was highly reproducible within and between cell lines, where the expression of RPS24 long gradually increased over 7 days of spheroid compaction to levels 6- to 10-fold higher than in a normoxic monolayer. We also observed an effect that was missed previously by only picking up spheroids on Day 5, where the expression of RPS24 short also initially increased on Day 1, like RPS24 long, albeit to a lesser degree (Figure 15). The expression of RPS24 short then decreased back to similar levels as the monolayer for HEK293 and

HCT116, and below monolayer levels with U87MG and PC3 as in ASPCR validation samples.

These data highlight a gradual switch in RPS24 alternative splicing as a spheroid compacts and accumulates stressors.

Figure 15. RPS24 switches splice variant abundance over 7 days of spheroid growth. Spheroids from four different cell lines were lysed on days 1,3,5, and 7 post-seeding. The expression of the long and short mRNA variants of RPS24 was measured using qRT-PCR and variant-specific primers. The ∆∆Ct method was used, normalizing to reference genes RPLP0 and RPL13A, and the normalized expression made relative to normoxia. Leslie Fell performed the RT-qPCR for HCT116. Data (n ≥ 7), mean normalized relative expression ± s.e.m. One-way ANOVA with Tukey’s HSD test was performed on the ∆Ct values. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001.

2.3.6 Acidosis has modest influence on RPS24 alternative splicing in normoxia

Since it appears that hypoxia alone may not be the only stressor in spheroids responsible for inducing RPS24 long expression (Figure 12), we sought to measure the influence of other features of the tumor/spheroid microenvironment. As spheroids compact over 7 days, they reproduce a key feature of tumors, the buildup of metabolic byproducts, namely lactic acid and resulting acidosis.

Acidosis has previously been shown to alter the splicing of key tumor and angiogenic factor

VEGF-A (Elias and Dias, 2008).

We exposed cells in a monolayer to acidosis-permissive media, which due to varying concentrations of NaHCO3 and pH adjustments, are neutral pH when added to cells, but decrease in pH due to the cells’ own metabolic processes and reduced buffer capacity. After 24 h, a gradual pH gradient was achieved from 7.0-5.8. For reference, the pH of U87MG spheroid media on days

1, 3, 5 and 7 were as follows: 6.8, 6.5, 6.5-6.0, 6.0-5.5, respectively. We expected if acidosis were causing RPS24 long expression to increase, decreasing pH would result in increased RPS24 long, and/or decreased RPS24 short in the cell lines where this occurred as previously (U87MG and PC3

– Figure 12). In normoxia of HEK293, U87MG, and HCT116, we see a subtle gradual increase in

RPS24 long expression with decreasing pH, however all less than 2-fold (Figure 16), similar to the effects of hypoxia. We saw a slight decrease in RPS24 short in only HEK293 and PC3.

Surprisingly, when cells were given acidosis-permissive media and exposed to 1% O2, this ablated

any pH effect seen in normoxia. Much like hypoxia, these data suggest that acidosis may play a minor role in inducing RPS24 long expression but is likely not the main contributing factor.

Figure 16. Acidosis has modest influence on RPS24 alternative splicing in normoxia. Cells were treated with acidosis-permissive media containing varying concentrations of NaHCO3 that generated a pH gradient from 5.8 to 7.0. RT-qPCR was performed for the long and short variants of RPS24. The ∆∆Ct method was used, normalizing to reference genes RPLP0 and/or RPL13A, and the normalized expression made relative to pH 7.0. All samples and RT-qPCR were obtained and performed by Leslie Fell. Data (n = 6), mean normalized relative expression ± s.e.m. One-way ANOVA with Tukey’s HSD test was performed on the ∆Ct values.* = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001.

2.3.7 Lactic acidosis does not affect RPS24 splice variant expression in U87MG

Since tumor acidosis is primarily caused by lactic acid accumulation and lactic acid is now recognized as a signalling molecule in its own right (Sun et al., 2017), we next chose to investigate the effects of lactic acidosis independent of low pH. We treated cells with lactic acid and sodium lactate, and the lactate dehydrogenase- inhibitor sodium oxamate to remove lactate from the cells.

The proton of lactic acid dissociates at pH 7.35-7.45 (Gladden, 2004), therefore regions of higher pH in a tumor would have lactic acid found in the form of lactate. We treated cells with lactic acid and sodium lactate at pH 7.1 and 7.4 to keep in protonated or deprotonated state, respectively, as well as to observe the effects of lactic acid at neutral pH (i.e. independent of acidosis).

If lactic acid was responsible for inducing RPS24 long expression, we would have expected an increase in RPS24 long and/or decrease in RPS24 short upon treatment with lactic acid and lactate, and a decrease in RPS24 long and/or increase in RPS24 short with sodium oxamate. RPS24 long decreased, albeit not significantly, in hypoxic sodium oxamate treatment, but there was no significant changes in RPS24 long or short expression during lactate/lactic acid treatment (Figure

17). Additionally, the two variants display the same general trends, suggesting no differential regulation between the variants occurred. Although protocols from a previous publication were followed, it would be beneficial to confirm that optimal lactic acid concentrations were used by

measuring the expression of a target of lactic acid signalling to confirm activation of this pathway.

Nonetheless, since no trends were evident we did not repeat the experiments in other cell lines.

Figure 17. Lactic acidosis does not affect RPS24 splice variant expression in U87MG. Following 24 h of serum starvation, cells were incubated with lactic acid, sodium lactate, and lactate-dehydrogenase inhibitor sodium oxamate (all at neutral pH). RT-qPCR was performed for both variants of RPS24. The ∆∆Ct method was used, normalizing to reference genes RPLP0

and/or RPL13A, and the normalized expression made relative to untreated control. All samples and RT-qPCR were obtained and performed by Michael Rosen. Data (n = 5), mean normalized relative expression ± s.e.m. One-way ANOVA with Tukey’s HSD test was performed on the ∆Ct values. 2.3.8 Increasing cell confluency of a monolayer does not increase RPS24 long expression in U87MG under 21% O2

Since spheroids are forms of 3D cell culture, their formation is highly dependent on cadherin- mediated cell-cell junctions (Ivascu and Kubbies, 2007). Increasing cell monolayer confluency also increases cadherin expression (Conacci-Sorrell et al., 2003; Elisha et al., 2018), without the addition of hypoxia and acidosis, so we wondered whether increasing cell monolayer confluency would be sufficient to increase RPS24 long. We seeded U87MG cells, refreshed media daily to prevent waste accumulation, and harvested them on days 1-5 post-seeding, imaging before lysing to visualize confluency (Figure 18A). Over 5 days of monolayer growth and increasing cell confluency, both splice variants actually decreased. RPS24 long and short variants had nearly identical relative expressions over the 5 days, as with lactic acid, suggesting they were not

differentially regulated (Figure 18B). This decrease may be due to slowed growth rates at higher cell confluency, leading to a downregulation in ribosome biogenesis and thus decreased overall

RPS24 expression. We did not perform the experiments under 1% O2 as results from increasing confluency would be confounded by extended hypoxic exposure.

Figure 18. Increasing cell confluency of a monolayer does not increase RPS24 long expression in U87MG under 21% O2. (A) U87MG cells were plated and harvested on Days 1 - 5 following seeding. Scale bar is 500 μM. (B) RT-qPCR was performed for the long and short variants of RPS24. The ∆∆Ct method was used, normalizing to reference genes RPLP0 and/or RPL13A, and the normalized expression made relative to Day 1. All samples and RT-qPCR were obtained and performed by Leslie Fell. Data (n = 4), mean normalized relative expression ± s.e.m. One-way ANOVA with Tukey’s HSD test was performed on the ∆Ct values. * = P < 0.05.

2.4 Discussion

Providing genomic complexity to higher eukaryotes, alternative splicing of pre-mRNAs helps organisms respond to environmental insults by aiding in the stress response (Kucherenko and

Shcherbata, 2018). Likely due to their perceived static nature, RPs are less frequently studied for

their alternative splicing. Alternative splicing which alters RP mRNA stability or protein coding sequence could provide heterogeneity to ribosomes via substoichiometry of RPs or incorporation of alternative protein isoforms, and is a potential avenue for translation regulation.

We measured alternative splicing of RPL22L1 and RPS24 as possible biomarkers of hypoxia in human prostate tumors. Hypoxic cells and spheroids of prostate cancer cell line PC3 displayed increased exon inclusion of both RPL22L1 and RPS24, and a similar (albeit weaker) trend was observed with human prostate tumor samples. Alternative splicing of RPL22L1 in tumors has not previously been studied, but RPS24 has appeared in RNA-Seq studies as a differentially spliced gene in cancer, albeit without a specific link to hypoxia or other stressors.

We also observed that tumors of higher stage displayed slightly lower levels of exon inclusion compared to normal tissue, consistent with other reports on differential splicing of RPS24 in tumors (Danan-Gotthold et al., 2015; Munkley et al., 2019; Zhang et al., 2015). Hypoxia is not always associated with increased tumor stage (Höckel et al., 1996), which is evident in our pool of tumors in which the majority of hypoxic tumors were stage II (although we had very few stage

IV tumors).

Although the slightly lower level of exon inclusion in human prostate tumors compared to normal is contradictory to increased exon inclusion in spheroids, these are models which provide a more controlled environment to directly study the effects of certain stressors. In contrast to cell lines and spheroids, tumors are heterogeneous not only in oxygenation, but also in cell type, so a reduced effect was expected. Based on these data, and given the tissue-specific splicing of these

RPs, alternative splicing of RPS24 and RPL22L1 warrants further investigation in other tumor types to determine whether they could be used as general hypoxic biomarkers. Since the effect we see is mainly in stage II tumors, RPS24 could potentially be used as an early prognostic marker. If

the unpredicted amplicon in RPL22L1 could be identified and validated in other tissue types (i.e. non-artifactual) it would be the strongest biomarker candidate. Since hypoxia is associated with poor prognosis, metastasis, and resistance to chemotherapeutics (Muz et al., 2015), early detection of these biomarkers would be key. Extracellular vesicles (such as exosomes), contain RNAs

(microRNAs, long non-coding RNAs, circular RNAs, and mRNAs) which are secreted into the urine and blood, thus we have the potential to detect biomarkers much earlier than in a biopsy

(Antoury et al., 2018; Batagov and Kurochkin, 2013; Rahbarghazi et al., 2019).

We measured an ASE in RPL22L1 which may alter its regulation via targeting to the nonsense-mediated decay pathway upon exon inclusion. An unpredicted amplicon, potentially the pseudogene RPL22P13, was strongly associated with hypoxia in the prostate tumor samples measured. Unfortunately, we were unable to detect this extra amplicon in 8 cell lines

(normoxic/hypoxic) and a few other tumor types and were concerned about it possibly being an artifact of genomic DNA contamination. Although genomic DNA would explain unexpected detection of a pseudogene, the fact that the genomic DNA was found in mostly hypoxic samples makes one inclined to speculate it was not random contamination but perhaps evidence of intron retention. Intron retention is a frequent phenomenon in tumors, and has been shown to be induced by hypoxia (Brady et al., 2017; Dvinge and Bradley, 2015). RPL22P13 is located in the intron of a fusion gene FIP1L1-PDGFRA (ENSEMBL), the product of which is involved in the pathogenesis of hypereosinophilic syndrome (Cools et al., 2003). RPL22P13 is also upstream of eIF4E2-dependent mRNA PDGFRA (Uniacke et al., 2012). The unpredicted amplicon may not in fact have been RPL22P13, and could have been an entirely novel product arising from erroneous splicing, such as a back-spliced circular RNA (Chen and Yang, 2015) . Further, this unpredicted amplicon may have been produced by non-cancer cells found in tumors such as immune cells or

fibroblasts, which could explain the lack of detection in cell lines. Ultimately, a larger scale investigation of RPL22L1 alternative splicing in human tumors using RNA-Seq would provide higher resolution to validate and identify this unpredicted amplicon that was so strongly associated with hypoxia.

Our validation of RPS24 alternative splicing in four cell lines demonstrated that hypoxia modestly increased the levels of RPS24 long (< 2-fold), whereas spheroids had >4-fold induction of the long variant. Further, over 7 days of spheroid growth, both RPS24 variants initially increase, but then a gradual switch from short to long was observed, suggesting a delayed adaptive response.

Since hypoxic monolayers did not have as strong an induction of RPS24 long as spheroids, other features of spheroids (such as nutrient deprivation or necrosis), possibly in conjunction with hypoxia, are likely responsible for inducing exon inclusion. In ongoing work, we attempted to isolate features of the spheroid microenvironment and apply them to monolayers to see if we could reproduce the spheroid’s induction of RPS24 long. Administration of lactic acid (neutral pH) and increasing cell confluency did not influence RPS24 splicing. Acidosis did show a modest induction in the ratio of long/short RPS24 in all four cell lines under normoxia, but this effect was not seen in hypoxia. Acidosis under normal oxygenation has previously been shown to stabilize HIF1α by preventing its degradation (Mekhail et al., 2004). Perhaps normoxic acidosis induced a

“pseudohypoxic” activation of HIF1α, leading to similar induction of RPS24 long as previously seen in hypoxic (non-acidotic) monolayers. Loss of the effect in hypoxia could then be explained by the dominance of the other hypoxia inducible factor, HIF2α, during chronic hypoxia (i.e. 24 h)

(Melanson et al., 2017).

Since they have a modest effect on RPS24 long expression, hypoxia and acidosis could be further explored with longer treatment periods. Although 24 h of hypoxic and acidotic treatments

did not recapitulate the ≥ 3.5-fold induction of RPS24 long on day 1 of spheroids (i.e. 24 h post- plating), the accumulation and degree of stressors in day 1 spheroids may already be much greater than that in a 24 h-treated monolayer. Thus extending hypoxic and acidotic exposure, such as up to 5 days, may elicit a stronger RPS24 long induction. Longer and more stressful treatments will also likely increase cellular stress responses such as apoptosis, so this is another feature that should also be investigated using apoptosis-inducing drugs on monolayers. Ultimately, a singular spheroid feature may not be identifiable as the cause of the switch in alternative splicing of RPS24 and further experimentation should focus on the functional implications of this regulation.

Although RPS24 alternative splicing has been identified in a growing number of screens of different tissues, cancer types, or cells in different stages of differentiation, no studies have sought to investigate further for potential regulatory mechanisms or functional implications. We established that both variants are translated into protein and the isoforms produce functional ribosomal proteins that incorporate into polysomes, thus laying the groundwork for future functional studies from our research group. The longer splice variant of RPS24 produces a protein shortened by three amino acids at the C-terminus (Gupta and Warner, 2014). As with RPS12, proximity proteomics of eIF4A1 showed enrichment of RPS24, highlighting its position near the mRNA entry channel and potential importance in translation initiation (Padrón et al., 2019).

RPS24 isoforms could produce specialized hypoxic ribosomes through the recruitment of different ribosome-associated proteins since the C-terminus is accessible at the surface. Although a loss of three amino acids may seem minor, the endogenous RPS24 antibody (epitope is 50-amino acids to end of C-terminus) detects only FLAG-RPS24 Short (i.e. full length protein isoform) and not FLAG-RPS24 Long (data not shown). The shortened C-terminus produced by the long RPS24 splice variant loses a lysine residue, which is predicted to be acetylated, so the two protein isoforms

may be differentially modified which could modulate their stability or interaction with translation factors (Khatter et al., 2015).

Since the C-terminus of RPS24 is not highly conserved (Gupta and Warner, 2014), it may not be critical to RPS24’s role in the ribosome, but could rather be involved in mediating extra- ribosomal functions. Further, the C-terminus of RPS24 is disordered (PDB), and intrinsically disordered regions of proteins are often considered hot-spots for post-translational modifications

(Darling and Uversky, 2018). Indeed, altered post-translational modification of other ribosomal proteins caused them to leave ribosomes and act in an extra-ribosomal “moonlighting” role:

Delayed phosphorylation of RPL13A upon stimulation with interferon-γ led to RPL13A being found almost exclusively in the non-ribosome pool. Extra-ribosomal RPL13A was found to bind to a specific element in a transcript, silencing its translation, as a component of the IFN-gamma activated inhibitor of translation (GAIT) complex (Mazumder et al., 2003). DNA damage-induced phosphorylation of RPS3 stimulated its nuclear endonuclease activity to participate in DNA repair

(Kim et al., 2009). Ergo post-translational modifications can have a profound effect on the function of ribosomal proteins.

Differential regulation of splicing factors in spheroids could explain how RPS24 splicing is regulated. In prostate cancer, the ESRP2 splicing factor represses the inclusion of the 22-bp microexon of RPS24 and was induced upon androgen stimulation (Munkley et al., 2019). Further, the splicing factor nSR100/SRRM4 regulates the splicing of neuronal microexons (Irimia et al.,

2014), and RPS24 long is highly expressed in the brain (Gupta and Warner, 2014). It would be interesting to investigate whether hypoxia or spheroids regulate ESRP2 and SRRM4 or whether the hypoxia-regulated splicing factor SRSF3 acts on RPS24 in spheroids (Brady et al., 2017).

Additionally, the long noncoding RNA MVIH (microvascular invasion in hepatocellular

carcinoma) overlaps the 22-bp microexon in RPS24 (Zhang et al., 2015), thus perhaps if MVIH duplexes with RPS24 pre-mRNA it could alter its splicing (Romero-Barrios et al., 2018).

Our data demonstrate exon inclusion of RPS24 is modestly promoted by hypoxia and acidosis, but it is likely due to a complex combination of features within spheroids. Since RPS24 alternative splicing changes the coding sequence, this could potentially provide ribosome heterogeneity or extra-ribosomal functions to aid in the stress response in spheroids. Ultimately, further investigation is required to determine whether the switch in alternative splicing is a cause or consequence of spheroid formation, and to elucidate the exact mechanism regulating this phenomenon.

CONCLUSION

Despite the substantial body of evidence regarding heterogeneous ribosomes in mouse and yeast cells, the extent of ribosome regulation in human cells is unknown. Our study is the first to investigate the influence of an environmental stressor (hypoxia) on the regulation of human ribosomal proteins. As this was a pilot study, many of our conclusions are observational, albeit the wealth of data obtained may facilitate future studies and collaborations to investigate the mechanisms behind hypoxic regulation of RPS12 and the switch in alternative splicing of RPS24 in spheroids.

We show that overexpression of RPS12 increases heavy polysome association of APAF-1 and XIAP mRNAs, both of which have long, highly structured 5’UTRs and are subject to specialized translational regulation, usually independent of cap-dependent translation initiation.

Since APAF-1 and XIAP mRNA play opposing roles in apoptosis, but both seem to be translated more efficiently upon RPS12 overexpression, it’s possible that this may be a part of a larger feedback loop of the apoptotic response. Although APAF-1 and XIAP have highly structured 5’

UTRs, RPS12 may not be involved in the translation of all mRNAs with structured 5’ UTRs, since individual transcript regulation decreases the likelihood of a universal mechanism. Nonetheless, identification of RPS12-dependent transcripts by immunoprecipitation of RPS12-containing ribosomes followed by RNA-Seq (ribosome footprinting) is a pertinent next step to identify specific sequence elements that may facilitate transcript selection. Furthermore, investigating how hypoxia regulates the presence of RPS12 in monosomes/initiating ribosomes/RNPs, such as through post-translational modifications, will provide necessary insight on a specific function of

RPS12 in hypoxia.

We show that expression of the long variant of RPS24 is substantially increased in spheroids of four different cell lines, but we were thus far unable to determine a specific spheroid feature which could reproduce RPS24 long expression when applied to monolayers. Cell-sorting based on different molecular markers (eg. CAIX, apoptosis/necrosis, epithelial/mesenchymal phenotype) and measuring RPS24 short/long expression may provide more insight into a causative role if the change in splicing is derived from specific sub-populations of cells within spheroids. A

CRISPR-Cas9/RNAi screen of different splicing factors would be a beneficial high-throughput method to highlight specific splicing factors regulating RPS24 alternative splicing. Use of cell lines expressing tagged versions of both RPS24 isoforms can be used for many functional experiments. First, do the isoforms differ in stability or subcellular localization? Do ribosomes containing RPS24 short vs. long translate different mRNAs? These cell lines can also be used to determine whether RPS24 alternative splicing is a cause of spheroid formation, rather than simply a coincidence, by observing the effects of isoform overexpression on spheroid formation, compaction, and stability. If the long variant of RPS24 aids in spheroid formation, RPS24 splicing could potentially be targeted therapeutically using antisense oligonucleotides, which have found

great success in personalized medicine where they are used to treat various neuromuscular disorders (Sardone et al., 2017). Additionally, a CRISPR-guided cytidine deaminase has been shown to alter splice sites and promote exon skipping of RPS24, although the effects were very modest (Yuan et al., 2018).

A final avenue of study could be to repeat this investigation of ribosomal protein regulation

(using mass-spectrometry and high-throughput RNA-Seq) in physiologically relevant oxygen conditions (physioxia) (Keeley and Mann, 2019). Translation initiation undergoes a gradual switch in usage of cap-binding homologs over a gradient of 21% to 1% O2 (Timpano and Uniacke,

2016), thus it’s possible that changes could also occur with ribosome regulation. Indeed, this novel study is important in the context of human biology and gives a greater appreciation of the regulation of ribosomal proteins by stimuli involved in normal physiology and disease.

BIBLIOGRAPHY

Aitken, C.E., and Lorsch, J.R. (2012). A mechanistic overview of translation initiation in eukaryotes. Nat. Struct. Mol. Biol. 19, 568–576. Alberts B, Johnson A, L.J. (2002). The RNA World and the Origins of Life (New York: Garland Science). Amsterdam, A., Sadler, K.C., Lai, K., Farrington, S., Bronson, R.T., Lees, J.A., and Hopkins, N. (2004). Many ribosomal protein genes are cancer genes in zebrafish. PLoS Biol. 2, e139. Andras, P., and Andras, C. (2005). The origins of life - The “protein interaction world” hypothesis: Protein interactions were the first form of self-reproducing life and nucleic acids evolved later as memory molecules. Med. Hypotheses 64, 678–688. Andreev, D.E., Dmitriev, S.E., Zinovkin, R., Terenin, I.M., and Shatsky, I.N. (2012). The 5′ untranslated region of Apaf-1 mRNA directs translation under apoptosis conditions via a 5′ end- dependent scanning mechanism. FEBS Lett. 586, 4139–4143. Anger, A.M., Armache, J.-P., Berninghausen, O., Habeck, M., Subklewe, M., Wilson, D.N., and Beckmann, R. (2013). Structures of the human and Drosophila 80S ribosome. Nature 497, 80– 85. Antoury, L., Hu, N., Balaj, L., Das, S., Georghiou, S., Darras, B., Clark, T., Breakefield, X.O., and Wheeler, T.M. (2018). Analysis of extracellular mRNA in human urine reveals splice variant biomarkers of muscular dystrophies. Nat. Commun. 9, 3906. Armistead, J., Khatkar, S., Meyer, B., Mark, B.L., Patel, N., Coghlan, G., Lamont, R.E., Liu, S., Wiechert, J., Cattini, P.A., et al. (2009). Mutation of a Gene Essential for Ribosome Biogenesis, EMG1, Causes Bowen-Conradi Syndrome. Am. J. Hum. Genet. 84, 728–739. Ban, N., Beckmann, R., Cate, J.H.D., Dinman, J.D., Dragon, F., Ellis, S.R., Lafontaine, D.L.J., Lindahl, L., Liljas, A., Lipton, J.M., et al. (2014). A new system for naming ribosomal proteins. Curr. Opin. Struct. Biol. 24, 165–169. Batagov, A.O., and Kurochkin, I. V. (2013). Exosomes secreted by human cells transport largely mRNA fragments that are enriched in the 3′-untranslated regions. Biol. Direct 8, 12. Belle, A., Tanay, A., Bitincka, L., Shamir, R., and O’Shea, E.K. (2006). Quantification of protein half-lives in the budding yeast proteome. Proc. Natl. Acad. Sci. U. S. A. 103, 13004–13009. Berget, S.M., Moore, C., and Sharp, P.A. (1977). Spliced segments at the 5’ terminus of adenovirus 2 late mRNA. Proc. Natl. Acad. Sci. U. S. A. 74, 3171–3175. Bernhardt, H.S., and Tate, W.P. (2015). A Ribosome Without RNA. Front. Ecol. Evol 3, 129. Bhat, M., Robichaud, N., Hulea, L., Sonenberg, N., Pelletier, J., and Topisirovic, I. (2015). Targeting the translation machinery in cancer. Nat. Rev. Drug Discov. 14, 261–278. Blencowe, B.J. (2000). Exonic splicing enhancers: Mechanism of action, diversity and role in human genetic diseases. Trends Biochem. Sci. 25, 106–110. Blencowe, B.J. (2006). Alternative Splicing: New Insights from Global Analyses. Cell 126, 37–

47. Boria, I., Garelli, E., Gazda, H.T., Aspesi, A., Quarello, P., Pavesi, E., Ferrante, D., Meerpohl, J.J., Kartal, M., Da Costa, L., et al. (2010). The ribosomal basis of diamond-blackfan anemia: Mutation and database update. Hum. Mutat. 31, 1269–1279. Bowler, E., Porazinski, S., Uzor, S., Thibault, P., Durand, M., Lapointe, E., Rouschop, K.M.A., Hancock, J., Wilson, I., and Ladomery, M. (2018). Hypoxia leads to significant changes in alternative splicing and elevated expression of CLK splice factor kinases in PC3 prostate cancer cells. BMC Cancer 18, 355. Brady, L.K., Wang, H., Radens, C.M., Bi, Y., Radovich, M., Maity, A., Ivan, C., Ivan, M., Barash, Y., and Koumenis, C. (2017). Transcriptome analysis of hypoxic cancer cells uncovers intron retention in EIF2B5 as a mechanism to inhibit translation. PLoS Biol. 15, e2002623. Brumwell, A., Fell, L., Obress, L., and Uniacke, J. (2020). Hypoxia influences polysome distribution of human ribosomal protein S12 and alternative splicing of ribosomal protein mRNAs. RNA 26, 361–371. Buttgereit, F., and Brand, M.D. (1995). A hierarchy of ATP-consuming processes in mammalian cells. Biochem. J. 312, 163–167. Carreau, A., Hafny-Rahbi, B. El, Matejuk, A., Grillon, C., and Kieda, C. (2011). Why is the partial oxygen pressure of human tissues a crucial parameter? Small molecules and hypoxia. J. Cell. Mol. Med. 15, 1239–1253. Cavadas, M.A.S., Cheong, A., and Taylor, C.T. (2017). The regulation of transcriptional repression in hypoxia. Exp. Cell Res. 356, 173–181. Chaudhuri, S., Vyas, K., Kapasi, P., Komar, A.A., Dinman, J.D., Barik, S., and Mazumder, B. (2007). Human ribosomal protein L13a is dispensable for canonical ribosome function but indispensable for efficient rRNA methylation. RNA 13, 2224–2237. Chen, L.L., and Yang, L. (2015). Regulation of circRNA biogenesis. RNA Biol. 12, 381–388. Chen, M., and Manley, J.L. (2009). Mechanisms of alternative splicing regulation: Insights from molecular and genomics approaches. Nat. Rev. Mol. Cell Biol. 10, 741–754. Choesmel, V., Fribourg, S., Aguissa-Touré, A.H., Pinaud, N., Legrand, P., Gazda, H.T., and Gleizes, P.E. (2008). Mutation of ribosomal protein RPS24 in Diamond-Blackfan anemia results in a ribosome biogenesis disorder. Hum. Mol. Genet. 17, 1253–1263. Coldwell, M.J., Mitchell, S.A., Stoneley, M., MacFarlane, M., and Willis, A.E. (2000). Initiation of Apaf-1 translation by internal ribosome entry. Oncogene 19, 899–905. Comerford, K.M., Wallace, T.J., Karhausen, J., Louis, N.A., Montalto, M.C., and Colgan, S.P. (2002). Hypoxia-inducible factor-1-dependent regulation of the multidrug resistance (MDR1) gene. Cancer Res. 62, 3387–3394. Conacci-Sorrell, M., Simcha, I., Ben-Yedidia, T., Blechman, J., Savagner, P., and Ben-Ze’Ev, A. (2003). Autoregulation of E-cadherin expression by cadherin-cadherin interactions: The roles of β-catenin signaling, Slug, and MAPK. J. Cell Biol. 163, 847–857.

Cools, J., DeAngelo, D.J., Gotlib, J., Stover, E.H., Legare, R.D., Cortes, J., Kutok, J., Clark, J., Galinsky, I., Griffin, J.D., et al. (2003). A tyrosine kinase created by fusion of the PDGFRA and FIP1L1 genes as a therapeutic target of imatinib in idiopathic hypereosinophilic syndrome. N. Engl. J. Med. 348, 1201–1214. Corbet, C., and Feron, O. (2017). Tumour acidosis: From the passenger to the driver’s seat. Nat. Rev. Cancer 17, 577–593. Coudert, L., Adjibade, P., and Mazroui, R. (2014). Analysis of translation initiation during stress conditions by polysome profiling. J. Vis. Exp. 51164. Crick, F.H.C. (1968). The origin of the genetic code. J. Mol. Biol. 38, 367–379. Culviner, P.H., and Laub, M.T. (2018). Global Analysis of the E. coli Toxin MazF Reveals Widespread Cleavage of mRNA and the Inhibition of rRNA Maturation and Ribosome Biogenesis. Mol. Cell 70, 868–880. Dai, M.S., Sears, R., and Lu, H. (2007). Feedback regulation of c-Myc by ribosomal protein L11. Cell Cycle 6, 2735–2741. Dameshek, W. (1967). Riddle: what do aplastic anemia, paroxysmal nocturnal hemoglobinuria (PNH) and “hypoplastic” leukemia have in common? Blood 30, 251–254. Danan-Gotthold, M., Golan-Gerstl, R., Eisenberg, E., Meir, K., Karni, R., and Levanon, E.Y. (2015). Identification of recurrent regulated alternative splicing events across human solid tumors. Nucleic Acids Res. 43, 5130–5144. Darling, A.L., and Uversky, V.N. (2018). Intrinsic disorder and posttranslational modifications: The darker side of the biological dark matter. Front. Genet. 9, 158. Digiacomo, V., and Meruelo, D. (2016). Looking into laminin receptor: Critical discussion regarding the non-integrin 37/67-kDa laminin receptor/RPSA protein. Biol Rev Camb Philos Soc 91, 288–310. Dunwoodie, S.L. (2009). The Role of Hypoxia in Development of the Mammalian Embryo. Dev. Cell 17, 755–773. Dvinge, H., and Bradley, R.K. (2015). Widespread intron retention diversifies most cancer transcriptomes. Genome Med. 7, 45. Ebert, B.L., Pretz, J., Bosco, J., Chang, C.Y., Tamayo, P., Galili, N., Raza, A., Root, D.E., Attar, E., Ellis, S.R., et al. (2008). Identification of RPS14 as a 5q- syndrome gene by RNA interference screen. Nature 451, 335–339. Elias, A.P., and Dias, S. (2008). Microenvironment changes (in pH) affect VEGF alternative splicing. Cancer Microenviron. 1, 131–139. Elisha, Y., Kalchenko, V., Kuznetsov, Y., and Geiger, B. (2018). Dual role of E-cadherin in the regulation of invasive collective migration of mammary carcinoma cells. Sci. Rep. 8, 4986. Ferreira-Cerca, S., Pöll, G., Gleizes, P.E., Tschochner, H., and Milkereit, P. (2005). Roles of eukaryotic ribosomal proteins in maturation and transport of pre-18S rRNA and ribosome function. Mol. Cell 20, 263–275.

Ferretti, M.B., Ghalei, H., Ward, E.A., Potts, E.L., and Karbstein, K. (2017). Rps26 directs mRNA-specific translation by recognition of Kozak sequence elements. Nat. Struct. Mol. Biol. 24, 700–707. Fewell, S.W., and Woolford, J.L. (1999). Ribosomal Protein S14 of Saccharomyces cerevisiae Regulates Its Expression by Binding to RPS14B Pre-mRNA and to 18S rRNA. Mol. Cell. Biol. 19, 826–834. Fidler, I.J., and Poste, G. (2008). The “seed and soil” hypothesis revisited. Lancet Oncol. 9, 808. Fox, G.E. (2010). Origin and evolution of the ribosome. Cold Spring Harb. Perspect. Biol. 2, a003483. Fulda, S., Gorman, A.M., Hori, O., and Samali, A. (2010). Cellular stress responses: Cell survival and cell death. Int. J. Cell Biol. 2010, 214074. Gazda, H.T., Grabowska, A., Merida-Long, L.B., Latawiec, E., Schneider, H.E., Lipton, J.M., Vlachos, A., Atsidaftos, E., Ball, S.E., Orfali, K.A., et al. (2006). Ribosomal protein S24 gene is mutated in Diamond-Blackfan anemia. Am. J. Hum. Genet. 79, 1110–1118. Gerweck, L.E., and Seetharaman, K. (1996). Cellular pH gradient in tumor versus normal tissue: Potential exploitation for the treatment of cancer. Cancer Res. 56, 1194–1198. Ghulam, M.M., Catala, M., and Abou Elela, S. (2019). Differential expression of duplicated ribosomal protein genes modifies ribosome composition in response to stress. Nucleic Acids Res. 48, 1954–1968. Giampietri, C., Petrungaro, S., Conti, S., Facchiano, A., Filippini, A., and Ziparo, E. (2015). Cancer microenvironment and endoplasmic reticulum stress response. Mediators Inflamm. 2015, 417281. Gladden, L.B. (2004). Lactate metabolism: A new paradigm for the third millennium. J. Physiol. 558, 5–30. Gonatopoulos-Pournatzis T., Niibori R., Salter EW, Weatheritt RJ, Tsang B, Farhangmehr S, Liang X, Braunschweig U, Roth J, Zhang S, Henderson T, Sharma E, Quesnel-Vallières M, Permanyer J, Maier S, Georgiou J, Irimia M, Sonenberg N, Forman-Kay JD, Gingras, B.B. (2020). Autism-Misregulated eIF4G Microexons Control Synaptic Translation and Higher Order Cognitive Functions. Mol. Cell S1097-2765, 30006–x. Gonzales, B., Henning, D., So, R.B., Dixon, J., Dixon, M.J., and Valdez, B.C. (2005). The Treacher Collins syndrome (TCOF1) gene product is involved in pre-rRNA methylation. Hum. Mol. Genet. 14, 2035–2043. Gu, B.W., Apicella, M., Mills, J., Fan, J.M., Reeves, D.A., French, D., Podsakoff, G.M., Bessler, M., and Mason, P.J. (2015). Impaired telomere maintenance and decreased canonical WNT signaling but normal ribosome biogenesis in induced pluripotent stem cells from X-linked dyskeratosis congenita patients. PLoS One 10, e0127414. Gupta, V., and Warner, J.R. (2014). Ribosome-omics of the human ribosome. RNA 20, 1004– 1013.

Hanaoka, M., Droma, Y., Basnyat, B., Ito, M., Kobayashi, N., Katsuyama, Y., Kubo, K., and Ota, M. (2012). Genetic Variants in EPAS1 Contribute to Adaptation to High-Altitude Hypoxia in Sherpas. PLoS One 7, e50566. Harrison, P.M., Zheng, D., Zhang, Z., Carriero, N., and Gerstein, M. (2005). Transcribed processed pseudogenes in the human genome: An intermediate form of expressed retrosequence lacking protein-coding ability. Nucleic Acids Res. 33, 2374–2383. Hashem, Y., Des Georges, A., Dhote, V., Langlois, R., Liao, H.Y., Grassucci, R.A., Hellen, C.U.T., Pestova, T. V., and Frank, J. (2013). Structure of the mammalian ribosomal 43S preinitiation complex bound to the scanning factor DHX29. Cell 153, 1108–1119. He, J., Navarrete, S., Jasinski, M., Vulliamy, T., Dokal, I., Bessler, M., and Mason, P.J. (2002). Targeted disruption of Dkc1, the gene mutated in X-linked dyskeratosis congenita, causes embryonic lethality in mice. Oncogene 21, 7740–7744. Ho, J.J.D., Balukoff, N.C., Cervantes, G., Malcolm, P.D., Krieger, J.R., and Lee, S. (2018). Oxygen-Sensitive Remodeling of Central Carbon Metabolism by Archaic eIF5B. Cell Rep. 22, 17–26. Höckel, M., Schlenger, K., Aral, B., Mitze, M., Schäffer, U., and Vaupel, P. (1996). Association between tumor hypoxia and malignant progression in advanced cancer of the uterine cervix. Cancer Res. 56, 4509–4515. Hoeppner, M.P., White, S., Jeffares, D.C., and Poole, A.M. (2009). Evolutionarily Stable Association of Intronic snoRNAs and microRNAs with Their Host Genes. Genome Biol. Evol. 1, 420–428. Holcik, M., and Sonenberg, N. (2005). Translational control in stress and apoptosis. Nat. Rev. Mol. Cell Biol. 6, 318–327. Hopkins, S.R. (2006). Exercise induced arterial hypoxemia: The role of ventilation-perfusion inequality and pulmonary diffusion limitation. Adv. Exp. Med. Biol. 588, 17–30. Huang, Y., Tian, Y., Zhao, Y., Xue, C., Zhan, J., Liu, L., He, X., and Zhang, L. (2018). Efficacy of the hypoxia-activated prodrug evofosfamide (TH-302) in nasopharyngeal carcinoma in vitro and in vivo. Cancer Commun. (London, England) 38, 15. Irimia, M., Weatheritt, R.J., Ellis, J.D., Parikshak, N.N., Gonatopoulos-Pournatzis, T., Babor, M., Quesnel-Vallières, M., Tapial, J., Raj, B., O’Hanlon, D., et al. (2014). A highly conserved program of neuronal microexons is misregulated in autistic brains. Cell 159, 1511–1523. Ivanov, A. V., Malygin, A.A., and Karpova, G.G. (2006). Eukaryotic ribosomal proteins: Interactions with their own pre-mRNAs and their involvement in splicing regulation. Mol. Biol. 40, 640–649. Ivascu, A., and Kubbies, M. (2007). Diversity of cell-mediated adhesions in breast cancer spheroids. Int. J. Oncol. 31, 1403–1413. Jackson, R.J. (2013). The current status of vertebrate cellular mRNA IRESs. Cold Spring Harb. Perspect. Biol. 5, a011569.

Jurica, M.S., and Moore, M.J. (2003). Pre-mRNA splicing: Awash in a sea of proteins. Mol. Cell 12, 5–14. Kahles, A., Lehmann, K. Van, Toussaint, N.C., Hüser, M., Stark, S.G., Sachsenberg, T., Stegle, O., Kohlbacher, O., Sander, C., Caesar-Johnson, S.J., et al. (2018). Comprehensive Analysis of Alternative Splicing Across Tumors from 8,705 Patients. Cancer Cell 34, 211–224. Kale, A., Ji, Z., Kiparaki, M., Blanco, J., Rimesso, G., Flibotte, S., and Baker, N.E. (2018). Ribosomal Protein S12e Has a Distinct Function in Cell Competition. Dev. Cell 44, 42–55. Kaluz, S., Kaluzová, M., Liao, S.Y., Lerman, M., and Stanbridge, E.J. (2009). Transcriptional control of the tumor- and hypoxia-marker carbonic anhydrase 9: A one transcription factor (HIF- 1) show? Biochim. Biophys. Acta - Rev. Cancer 1795, 162–172. Keeley, T.P., and Mann, G.E. (2019). Defining physiological normoxia for improved translation of cell physiology to animal models and humans. Physiol. Rev. 99, 161–234. Khatter, H., Myasnikov, A.G., Natchiar, S.K., and Klaholz, B.P. (2015). Structure of the human 80S ribosome. Nature 520, 640–645. Kim, T.S., Kim, H.D., and Kim, J. (2009). PKCδ-dependent functional switch of rpS3 between translation and DNA repair. Biochim. Biophys. Acta - Mol. Cell Res. 1793, 395–405. Klinck, R., Bramard, A., Inkel, L., Dufresne-Martin, G., Gervais-Bird, J., Madden, R., Paquet, É.R., Koh, C.S., Venables, J.P., Prinos, P., et al. (2008). Multiple alternative splicing markers for ovarian cancer. Cancer Res. 68, 657–663. Kondrashov, N., Pusic, A., Stumpf, C.R., Shimizu, K., Hsieh, A.C., Xue, S., Ishijima, J., Shiroishi, T., and Barna, M. (2011). Ribosome-mediated specificity in Hox mRNA translation and vertebrate tissue patterning. Cell 145, 383–397. Kucherenko, M.M., and Shcherbata, H.R. (2018). miRNA targeting and alternative splicing in the stress response - Events hosted by membrane-less compartments. J. Cell Sci. 131, jcs202002. Kültz, D. (2004). Molecular and evolutionary basis of the cellular stress response. Annu. Rev. Physiol. 67, 225–257. Kuo, B.A., Gonzalez, I., Gillespie, D.A., and Sylvester, J.E. (1996). Human ribosomal RNA variants from a single individual and their expression in different tissues. Nucleic Acids Res. 24, 4817–4824. Kurylo, C.M., Parks, M.M., Juette, M.F., Zinshteyn, B., Altman, R.B., Thibado, J.K., Vincent, C.T., and Blanchard, S.C. (2018). Endogenous rRNA Sequence Variation Can Regulate Stress Response Gene Expression and Phenotype. Cell Rep. 25, 236–248. Lacerda, R., Menezes, J., and Romão, L. (2017). More than just scanning: the importance of cap- independent mRNA translation initiation for cellular stress response and cancer. Cell. Mol. Life Sci. 74, 1659–1680. Lacey, J.C., Cook, G.W., and Mullins, D.W. (1999). Concepts related to the origin of coded protein synthesis. Chemtracts 12, 398–418. Leffers, H., and Andersen, A.H. (1993). The sequence of 28S ribosomal RNA varies within and

between human cell lines. Nucleic Acids Res. 21, 1449–1455. Liu, L., and Simon, M.C. (2004). Regulation of transcription and translation by hypoxia. Cancer Biol. Ther. 3, 492–497. Liu, S., and Cheng, C. (2013). Alternative RNA splicing and cancer. Wiley Interdiscip. Rev. RNA 4, 547–566. Liu, L., Cash, T.P., Jones, R.G., Keith, B., Thompson, C.B., and Simon, M.C. (2006). Hypoxia- induced energy stress regulates mRNA translation and cell growth. Mol. Cell 21, 521–531. Liu, Y., Liu, Y., Yan, X., Xu, Y., Luo, F., Ye, J., Yan, H., Yang, X., Huang, X., Zhang, J., et al. (2014). HIFs enhance the migratory and neoplastic capacities of hepatocellular carcinoma cells by promoting EMT. Tumor Biol. 35, 8103–8114. Lodish, H., Berk, A., Kaiser, C.A., Krieger, M., Bretscher, A., Ploegh, H., Amon, A., and Scott, M.P. (2012). Molecular Cell Biology (New York: W.H. Freeman and Company). Ludwig, W., Strunk, O., Klugbauer, S., Klugbauer, N., Weizenegger, M., Neumaier, J., Bachleitner, M., and Schleifer, K.H. (1998). Bacterial phylogeny based on comparative sequence analysis. Electrophoresis 19, 554–568. Lyabin, D.N., and Ovchinnikov, L.P. (2016). Selective regulation of YB-1 mRNA translation by the mTOR signaling pathway is not mediated by 4E-binding protein. Sci. Rep. 6, 22502. Maida, Y., Yasukawa, M., Furuuchi, M., Lassmann, T., Possemato, R., Okamoto, N., Kasim, V., Hayashizaki, Y., Hahn, W.C., and Masutomi, K. (2009). An RNA-dependent RNA polymerase formed by TERT and the RMRP RNA. Nature 461, 230–235. Maquat, L.E., Tarn, W.Y., and Isken, O. (2010). The pioneer round of translation: Features and functions. Cell 142, 368–374. El Marabti, E., and Younis, I. (2018). The Cancer Spliceome: Reprograming of Alternative Splicing in Cancer. Front Mol Biosci 5, 80. Marygold, S.J., Roote, J., Reuter, G., Lambertsson, A., Ashburner, M., Millburn, G.H., Harrison, P.M., Yu, Z., Kenmochi, N., Kaufman, T.C., et al. (2007). The ribosomal protein genes and Minute loci of Drosophila melanogaster. Genome Biol. 8, R216. Mazumder, B., Sampath, P., Seshadri, V., Maitra, R.K., DiCorleto, P.E., and Fox, P.L. (2003). Regulated release of L13a from the 60S ribosomal subunit as a mechanism of transcript-specific translational control. Cell 115, 187–198. Mehta, G., Hsiao, A.Y., Ingram, M., Luker, G.D., and Takayama, S. (2012). Opportunities and challenges for use of tumor spheroids as models to test drug delivery and efficacy. J. Control. Release 164, 192–204. Mekhail, K., Gunaratnam, L., Bonicalzi, M.-E., and Lee, S. (2004). HIF activation by pH- dependent nucleolar sequestration of VHL. Nat. Cell Biol. 6, 642–647. Melanson, G., Timpano, S., and Uniacke, J. (2017). The eIF4E2-directed hypoxic cap-dependent translation machinery reveals novel therapeutic potential for cancer treatment. Oxid. Med. Cell. Longev. 2017, 6098107.

Mueller, M.M., and Fusenig, N.E. (2004). Friends or foes - Bipolar effects of the tumour stroma in cancer. Nat. Rev. Cancer 4, 839–849. Munkley, J., Li, L., Krishnan, S.R.G., Hysenaj, G., Scott, E., Dalgliesh, C., Oo, H.Z., Maia, T.M., Cheung, K., Ehrmann, I., et al. (2019). Androgen-regulated transcription of ESRP2 drives alternative splicing patterns in prostate cancer. Elife 8, e47678. Muz, B., de la Puente, P., Azab, F., and Azab, A.K. (2015). The role of hypoxia in cancer progression, angiogenesis, metastasis, and resistance to therapy. Hypoxia (Auckl) 3, 83–92. Nakhoul, H., Ke, J., Zhou, X., Liao, W., Zeng, S.X., and Lu, H. (2014). Ribosomopathies: Mechanisms of Disease. Clin Med Insights Blood Disord 7, 7–16. Naumov, G.N., Akslen, L.A., and Folkman, J. (2006). Role of angiogenesis in human tumor dormancy: Animal models of the angiogenic switch. Cell Cycle 5, 1779–1787. Neumann, F., Hemmerich, P., Von Mikecz, A., Peter, H. hatmut, and Krawinkel, U. (1995). Human ribosomal protein L7 inhibits cell-free translation in reticulocyte lysates and affects the expression of nuclear proteins upon stable transfection into jurkat T-lymphoma cells. Nucleic Acids Res. 23, 195–202. Nissen, P., Hansen, J., Ban, N., Moore, P.B., and Steitz, T.A. (2000). The structural basis of ribosome activity in peptide bond synthesis. Science. 289, 920–930. Noller, H.F., Hoffarth, V., and Zimniak, L. (1992). Unusual Resistance of Peptidyl Transferase to Protein Extraction Procedures. Source Sci. New Ser. Transplant. Bull. Ann. Surg. Br. Med. J 256, 1416–1419. Nunes, A.S., Barros, A.S., Costa, E.C., Moreira, A.F., and Correia, I.J. (2019). 3D tumor spheroids as in vitro models to mimic in vivo human solid tumors resistance to therapeutic drugs. Biotechnol. Bioeng. 116, 206–226. O’Donohue, M.F., Choesmel, V., Faubladier, M., Fichant, G., and Gleizes, P.E. (2010). Functional dichotomy of ribosomal proteins during the synthesis of mammalian 40S ribosomal subunits. J. Cell Biol. 190, 853–866. O’Leary, M.N., Schreiber, K.H., Zhang, Y., Duc, A.C.E., Rao, S., Hale, J.S., Academia, E.C., Shah, S.R., Morton, J.F., Holstein, C.A., et al. (2013). The Ribosomal Protein Rpl22 Controls Ribosome Composition by Directly Repressing Expression of Its Own Paralog, Rpl22l1. PLoS Genet. 9, e1003708. Orgel, L.E. (1968). Evolution of the genetic apparatus. J. Mol. Biol. 38, 381–393. Padrón, A., Iwasaki, S., and Ingolia, N.T. (2019). Proximity RNA Labeling by APEX-Seq Reveals the Organization of Translation Initiation Complexes and Repressive RNA Granules. Mol. Cell 75, 875–887. Paget, S. (1889). The distribution of secondary growths in cancer of the breast. Lancet 8, 98– 101. Pampus, F. (1963). Die Wasserstoffionenkonzentration des Hirngewebes bei raumfordernden intracraniellen Prozessen. Acta Neurochir. (Wien). 11, 305–318.

Pan, Q., Shai, O., Lee, L.J., Frey, B.J., and Blencowe, B.J. (2008). Deep surveying of alternative splicing complexity in the human transcriptome by high-throughput sequencing. Nat. Genet. 40, 1413–1415. Panse, V.G., and Johnson, A.W. (2010). Maturation of eukaryotic ribosomes: acquisition of functionality. Trends Biochem. Sci. 35, 260–266. Parks, M.M., Kurylo, C.M., Dass, R.A., Bojmar, L., Lyden, D., Vincent, C.T., and Blanchard, S.C. (2018). Variant ribosomal RNA alleles are conserved and exhibit tissue-specific expression. Sci. Adv. 4, eaao0665. Parsyan, A., Shahbazian, D., Martineau, Y., Petroulakis, E., Alain, T., Larsson, O., Mathonnet, G., Tettweiler, G., Hellen, C.U., Pestova, T. V., et al. (2009). The helicase protein DHX29 promotes translation initiation, cell proliferation, and tumorigenesis. Proc. Natl. Acad. Sci. U. S. A. 106, 22217–22222. Petrov, A., Meskauskas, A., and Dinman, J.D. (2004). Ribosomal protein L3: influence on ribosome structure and function. RNA Biol. 1, 59–65. Petrov, A.S., Gulen, B., Norris, A.M., Kovacs, N.A., Bernier, C.R., Lanier, K.A., Fox, G.E., Harvey, S.C., Wartell, R.M., Hud, N. V., et al. (2015). History of the ribosome and the origin of translation. Proc. Natl. Acad. Sci. U. S. A. 112, 15396–15401. Philippe, L., van den Elzen, A.M.., Watson, M.J., and Thoreen, C.C. (2020). Global analysis of LARP1 translation targets reveals tunable and dynamic features of 5′ TOP motifs. Proc. Natl. Acad. Sci. 117, 5319–5328. Piir, K., Tamm, T., Kisly, I., Tammsalu, T., and Remme, J. (2014). Stepwise splitting of ribosomal proteins from yeast ribosomes by LiCl. PLoS One 9, e101561. Pisareva, V.P., and Pisarev, A. V. (2016). DHX29 and eIF3 cooperate in ribosomal scanning on structured mRNAs during translation initiation. RNA 22, 1859–1870. Poon, E., Harris, A.L., and Ashcroft, M. (2009). Targeting the hypoxia-inducible factor (HIF) pathway in cancer. Expert Rev. Mol. Med. 11, e26. Van Raay, T.J., Connors, T.D., Klinger, K.W., Landes, G.M., and Burn, T.C. (1996). A novel ribosomal protein L3-like gene (RPL3L) maps to the autosomal dominant polycystic kidney disease gene region. Genomics 37, 172–176. Rabl, J., Leibundgut, M., Ataide, S.F., Haag, A., and Ban, N. (2011). Crystal structure of the eukaryotic 40S ribosomal subunit in complex with initiation factor 1. Science. 331, 730–736. Rahbarghazi, R., Jabbari, N., Sani, N.A., Asghari, R., Salimi, L., Kalashani, S.A., Feghhi, M., Etemadi, T., Akbariazar, E., Mahmoudi, M., et al. (2019). Tumor-derived extracellular vesicles: Reliable tools for Cancer diagnosis and clinical applications. Cell Commun. Signal. 17, 73. Razi, A., and Ortega, J. (2017). Ribosomal Proteins: Their Role in the Assembly, Structure and Function of the Ribosome. ELS 1–12. Reschke, M., Clohessy, J.G., Seitzer, N., Goldstein, D.P., Breitkopf, S.B., Schmolze, D.B., Ala, U., Asara, J.M., Beck, A.H., and Pandolfi, P.P. (2013). Characterization and Analysis of the

Composition and Dynamics of the Mammalian Riboproteome. Cell Rep. 4, 1276–1287. Ridanpää, M., Van Eenennaam, H., Pelin, K., Chadwick, R., Johnson, C., Yuan, B., VanVenrooij, W., Pruijn, G., Salmela, R., Rockas, S., et al. (2001). Mutations in the RNA component of RNase MRP cause a pleiotropic human disease, cartilage-hair hypoplasia. Cell 104, 195–203. Robledo, S., Idol, R.A., Crimmins, D.L., Ladenson, J.H., Mason, P.J., and Bessler, M. (2008). The role of human ribosomal proteins in the maturation of rRNA and ribosome production. RNA 14, 1918–1929. Robles, A.I., Bemmels, N.A., Foraker, A.B., and Harris, C.C. a. (2001). APAF-1 is a transcriptional target of p53 in DNA damage-induced apoptosis. Cancer Res. 61, 6660–6664. Rohwer, N., and Cramer, T. (2011). Hypoxia-mediated drug resistance: Novel insights on the functional interaction of HIFs and cell death pathways. Drug Resist. Updat. 14, 191–201. Romero-Barrios, N., Legascue, M.F., Benhamed, M., Ariel, F., and Crespi, M. (2018). Splicing regulation by long noncoding RNAs. Nucleic Acids Res. 46, 2169–2184. Ruggero, D., Grisendi, S., Piazza, F., Rego, E., Mari, F., Rao, P.H., Cordon-Cardo, C., and Pandolfi, P.P. (2003). Dyskeratosis congenita and cancer in mice deficient in ribosomal RNA modification. Science. 299, 259–262. Samir, P., Browne, C.M., Rahul, Sun, M., Shen, B., Li, W., Frank, J., and Link, A.J. (2018). Identification of Changing Ribosome Protein Compositions using Mass Spectrometry. Proteomics 18, e1800217. Sardone, V., Zhou, H., Muntoni, F., Ferlini, A., and Falzarano, M.S. (2017). Antisense oligonucleotide-based therapy for neuromuscular disease. Molecules 22, E563. Savage, S.A., and Alter, B.P. (2009). Dyskeratosis Congenita. Hematol. Oncol. Clin. North Am. 23, 215–231. Schosserer, M., Minois, N., Angerer, T.B., Amring, M., Dellago, H., Harreither, E., Calle-Perez, A., Pircher, A., Gerstl, M.P., Pfeifenberger, S., et al. (2015). Methylation of ribosomal RNA by NSUN5 is a conserved mechanism modulating organismal lifespan. Nat. Commun. 6, 6158. Seliger, C., Leukel, P., Moeckel, S., Jachnik, B., Lottaz, C., Kreutz, M., Brawanski, A., Proescholdt, M., Bogdahn, U., Bosserhoff, A.K., et al. (2013). Lactate-modulated induction of THBS-1 activates Transforming growth factor (TGF)-beta2 and migration of glioma cells in vitro. PLoS One 8, 1–14. Shepard, P.J., and Hertel, K.J. (2009). The SR protein family. Genome Biol. 10, 242. Shi, Z., Fujii, K., Kovary, K.M., Genuth, N.R., Röst, H.L., Teruel, M.N., and Barna, M. (2017). Heterogeneous Ribosomes Preferentially Translate Distinct Subpools of mRNAs Genome-wide. Mol. Cell 67, 71-83.e7. Simsek, D., and Barna, M. (2017). An emerging role for the ribosome as a nexus for post- translational modifications. Curr. Opin. Cell Biol. 45, 92–101. Slavov, N., Semrau, S., Airoldi, E., Budnik, B., and van Oudenaarden, A. (2015). Differential

Stoichiometry among Core Ribosomal Proteins. Cell Rep. 13, 865–873. Sloan, K.E., Warda, A.S., Sharma, S., Entian, K.D., Lafontaine, D.L.J., and Bohnsack, M.T. (2017). Tuning the ribosome: The influence of rRNA modification on eukaryotic ribosome biogenesis and function. RNA Biol. 14, 1138–1152. Solaini, G., Baracca, A., Lenaz, G., and Sgarbi, G. (2010). Hypoxia and mitochondrial oxidative metabolism. Biochim. Biophys. Acta - Bioenerg. 1797, 1171–1177. Sonenberg, N., and Hinnebusch, A.G. (2009). Regulation of Translation Initiation in Eukaryotes: Mechanisms and Biological Targets. Cell 136, 731–745. Song, Y., Botvinnik, O.B., Lovci, M.T., Kakaradov, B., Liu, P., Xu, J.L., and Yeo, G.W. (2017). Single-Cell Alternative Splicing Analysis with Expedition Reveals Splicing Dynamics during Neuron Differentiation. Mol. Cell 67, 148-161.e5. Sulima, S.O., Hofman, I.J.F., De Keersmaecker, K., and Dinman, J.D. (2017). How ribosomes translate cancer. Cancer Discov. 7, 1069–1087. Sun, L., Liu, L., Liu, X., Wang, Y., Li, M., Yao, L., Yang, J., Ji, G., Guo, C., Pan, Y., et al. (2014). MGr1-Ag/37LRP induces cell adhesion-mediated drug resistance through FAK/PI3K and MAPK pathway in gastric cancer. Cancer Sci. 105, 651–659. Sun, S., Li, H., Chen, J., and Qian, Q. (2017). Lactic acid: No longer an inert and end-product of glycolysis. Physiology 32, 453–463. Swinson, D.E., Jones, J.L., Richardson, D., Wykoff, C., Turley, H., Pastorek, J., Taub, N., Harris, A.L., and O’Byrne, K.J. (2003). Carbonic anhydrase IX expression, a novel surrogate marker of tumor hypoxia, is associated with a poor prognosis in non-small-cell lung cancer. J Clin Oncol 21, 473–482. Thoreen, C.C., Kang, S.A., Chang, J.W., Liu, Q., Zhang, J., Gao, Y., Reichling, L.J., Sim, T., Sabatini, D.M., and Gray, N.S. (2009). An ATP-competitive mammalian target of rapamycin inhibitor reveals rapamycin-resistant functions of mTORC1. J. Biol. Chem. 284, 8023–8032. Thorrez, L., Van Deun, K., Tranchevent, L.C., Van Lommel, L., Engelen, K., Marchal, K., Moreau, Y., Van Mechelen, I., and Schuit, F. (2008). Using ribosomal protein genes as reference: A tale of caution. PLoS One 3, e1854. Thumati, N.R., Zeng, X.L., Au, H.H.T., Jang, C.J., Jan, E., and Wong, J.M.Y. (2013). Severity of X-linked dyskeratosis congenita (DKCX) cellular defects is not directly related to dyskerin (DKC1) activity in ribosomal RNA biogenesis or mRNA translation. Hum. Mutat. 34, 1698– 1707. Tian, T., Li, X., and Zhang, J. (2019). mTOR signaling in cancer and mtor inhibitors in solid tumor targeting therapy. Int. J. Mol. Sci. 20, E755. Timpano, S., and Uniacke, J. (2016). Human cells cultured under physiological oxygen utilize two cap-binding proteins to recruit distinct mRNAs for translation. J. Biol. Chem. Tsay, Y.F., Thompson, J.R., Rotenberg, M.O., Larkin, J.C., and Woolford, J.L. (1988). Ribosomal protein synthesis is not regulated at the translational level in Saccharomyces

cerevisiae: balanced accumulation of ribosomal proteins L16 and rp59 is mediated by turnover of excess protein. Genes Dev. 2, 664–676. Uniacke, J., Holterman, C.E., Lachance, G., Franovic, A., Jacob, M.D., Fabian, M.R., Payette, J., Holcik, M., Pause, A., and Lee, S. (2012). An oxygen-regulated switch in the protein synthesis machinery. Nature 486, 126–129. Vadysirisack, D.D., and Ellisen, L.W. (2012). mTOR activity under hypoxia. Methods Mol. Biol. 821, 45–58. Vattem, K.M., and Wek, R.C. (2004). Reinitiation involving upstream ORFs regulates ATF4 mRNA translation in mammalian cells. Proc. Natl. Acad. Sci. U. S. A. 101, 11269–11274. Venables, J.P. (2004). Aberrant and alternative splicing in cancer. Cancer Res. 64, 7647–7654. Venables, J.P., Klinck, R., Bramard, A., Inkel, L., Dufresne-Martin, G., Koh, C.S., Gervais-Bird, J., Lapointe, E., Froehlich, U., Durand, M., et al. (2008). Identification of alternative splicing markers for breast cancer. Cancer Res. 68, 9525–9531. Vesper, O., Amitai, S., Belitsky, M., Byrgazov, K., Kaberdina, A.C., Engelberg-Kulka, H., and Moll, I. (2011). Selective translation of leaderless mRNAs by specialized ribosomes generated by MazF in Escherichia coli. Cell 147, 147–157. Wang, Y., Shang, W., Niu, M., Tian, J., and Xu, K. (2019). Hypoxia-active nanoparticles used in tumor theranostic. Int. J. Nanomedicine 14, 3705–3722. Warburg, O. (1925). The metabolism of carcinoma cells. J. Cancer Res. Van De Waterbeemd, M., Tamara, S., Fort, K.L., Damoc, E., Franc, V., Bieri, P., Itten, M., Makarov, A., Ban, N., and Heck, A.J.R. (2018). Dissecting ribosomal particles throughout the kingdoms of life using advanced hybrid mass spectrometry methods. Nat. Commun. 9, 2493. Weatheritt, R.J., Sterne-Weiler, T., and Blencowe, B.J. (2016). The ribosome-engaged landscape of alternative splicing. Nat. Struct. Mol. Biol. 23, 1117–1123. Weis, S.M., and Cheresh, D.A. (2011). Tumor angiogenesis: molecular pathways and therapeutic targets. Nat. Med. 17, 1359–1370. Willi, J., Küpfer, P., Eviquoz, D., Fernandez, G., Katz, A., Leumann, C., and Polacek, N. (2018). Oxidative stress damages rRNA inside the ribosome and differentially affects the catalytic center. Nucleic Acids Res. 46, 1945–1957. Woese, C.R. (1967). The genetic code (New York: Harper and Row). Wojtkowiak, J.W., Verduzco, D., Schramm, K.J., and Gillies, R.J. (2011). Drug resistance and cellular adaptation to tumor acidic pH microenvironment. Mol. Pharm. 8, 2032–2038. Wool, I.G. (1996). Extraribosomal functions of ribosomal proteins. Trends Biochem. Sci. 21, 164–165. Wu, P., Zhou, D., Wang, Y., Lin, W., Sun, A., Wei, H., Fang, Y., Cong, X., and Jiang, Y. (2019). Identification and validation of alternative splicing isoforms as novel biomarker candidates in hepatocellular carcinoma. Oncol. Rep. 41, 1929–1937.

Xu, W.B., and Roufa, D.J. (1996). The gene encoding human ribosomal protein S24 and tissue- specific expression of differentially spliced mRNAs. Gene 169, 257–262. Xue, S., Tian, S., Fujii, K., Kladwang, W., Das, R., and Barna, M. (2015). RNA regulons in Hox 59 UTRs confer ribosome specificity to gene regulation. Nature 517, 33–38. Yoffe, Y., David, M., Kalaora, R., Povodovski, L., Friedlander, G., Feldmesser, E., Ainbinder, E., Saada, A., Bialik, S., and Kimchi, A. (2016). Cap-independent translation by DAP5 controls cell fate decisions in human embryonic stem cells. Genes Dev. 30, 1991–2004. Yoshihama, M., Uechi, T., Asakawa, S., Kawasaki, K., Kato, S., Higa, S., Maeda, N., Minoshima, S., Tanaka, T., Shimizu, N., et al. (2002). The human ribosomal protein genes: Sequencing and comparative analysis of 73 genes. Genome Res. 12, 379–390. Yuan, J., Ma, Y., Huang, T., Chen, Y., Peng, Y., Li, B., Li, J., Zhang, Y., Song, B., Sun, X., et al. (2018). Genetic Modulation of RNA Splicing with a CRISPR-Guided Cytidine Deaminase. Mol. Cell 72, 380-394.e7. Zanoni, M., Piccinini, F., Arienti, C., Zamagni, A., Santi, S., Polico, R., Bevilacqua, A., and Tesei, A. (2016). 3D tumor spheroid models for in vitro therapeutic screening: A systematic approach to enhance the biological relevance of data obtained. Sci. Rep. 6, 19103. Zhang, Y., and Lu, H. (2009). Signaling to p53: Ribosomal Proteins Find Their Way. Cancer Cell 16, 369–377. Zhang, H., Ye, J., Weng, X., Liu, F., He, L., Zhou, D., and Liu, Y. (2015). Comparative transcriptome analysis reveals that the extracellular matrix receptor interaction contributes to the venous metastases of hepatocellular carcinoma. Cancer Genet. 208, 482–491. Zhang, Z., Harrison, P., and Gerstein, M. (2002). Identification and analysis of over 2000 ribosomal protein pseudogenes in the human genome. Genome Res. 12, 1466–1482. Zhao, J., Zhai, B., Gygi, S.P., and Goldberg, A.L. (2015). mTOR inhibition activates overall protein degradation by the ubiquitin proteasome system as well as by autophagy. Proc. Natl. Acad. Sci. U. S. A. 112, 15790–15797.

APPENDIX

Detailed TMT-MS Protocol The proteins were reduced in 1 mM DTT for 1 h at 56 °C and the free cysteine residues were alkylated using iodoacetamide. The proteins were precipitated with 5 volumes of acetone overnight at 20 °C and centrifugation. The proteins were digested according to the TMT 10-plex manual, but using Trypsin/LysC (Promega). 50 μg of protein from each condition was labeled using 0.4 mg of

TMT 10-plex (ThermoFisher) by incubating at room temperature for 1 h. The labeling reaction was stopped using 5% hydroxylamine. The peptides were mixed and the solvent removed under vacuum. The pellet was resuspended and were analyzed on an Orbitrap analyzer (Q-Exactive,

ThermoFisher, San Jose, CA) outfitted with a nanospray source and EASY-nLC nano-LC system

(ThermoFisher, San Jose, CA). Lyophilized peptide mixtures were dissolved in 0.1% formic acid and loaded onto a 75 μm x 50 cm PepMax RSLC EASY-Spray column filled with 2 μM C18 beads

(ThermoFisher San, Jose CA) at a pressure of 800 Bar. Peptides were eluted over 240 min at a rate of 250 nl/min using a stepwise gradient (0%-4% Acetonitrile containing 0.1% Formic Acid over

2 min; 4%-28% Acetonitrile containing 0.1% Formic Acid over 226 min, 28%-95% Acetonitrile containing 0.1% Formic Acid over 2 min, constant 95% Acetonitrile containing 0.1% Formic Acid for 10 min). Peptides were introduced by nano-electrospray into the Q-Exactive mass spectrometer

(Thermo-Fisher). The instrument method consisted of one MS full scan (525–1600 m/z) in the

Orbitrap mass analyzer with an automatic gain control (AGC) target of 1e6, maximum ion injection time of 120 ms and a resolution of 35 000 followed by 15 data-dependent MS/MS scans with a resolution of 35,000, an AGC target of 1e6, maximum ion time of 120 ms, and one microscan. The intensity threshold to trigger a MS/MS scan was set to an underfill ratio of 0.2%.

Fragmentation occurred in the HCD trap with normalized collision energy set to 30 V. The dynamic exclusion was applied using a setting of 40 s. The raw data was searched against the

Uniprot database using Proteome Discoverer (Version 2.2.0.388 ThermoFisher Scientific) which also extracted the quantitation data from the 10 TMT tags.

Quantification was carried out using SequestHT and Amanda 2.0 search engines. Unique and razor peptides were used, with a minimum peptide length of 6 amino acids, minimum 1 unique peptide, and average reporter signal to noise threshold of 10. False discovery rates of 0.01 (strict) and 0.05 (relaxed) for PSMs and peptides were used. Abundances were normalized to the same total peptide amount per channel and scaled so that the average abundance per protein and peptide was 100. The mean abundance was calculated for each protein from the three biological replicates, followed by the ratio between L/M and H/M in both normoxia and hypoxia. Two-tailed unpaired t-tests were performed on the L/M and H/M ratios within each oxygen concentration. The M fraction was used as reference in each condition because: 1) There should be stoichiometric amounts of 40S and 60S subunits; 2) To account for differences in total ribosome numbers/translation rates between conditions, we measured the distribution of RPs in L and H versus M (L/M and H/M) rather than the absolute amounts between hypoxia and normoxia (i.e.

Hhypoxia/Hnormoxia).

Appendix Figures

Appendix Figure 1. Hypoxic exposure confirmed by expression of hypoxia inducible factors. For the three biological replicates of TMT-MS, hypoxic exposure was confirmed before sending for analysis. Two 100 mm plates of cells were plated and incubated with each of the normoxic and hypoxic polysome samples for 24 h exposure. Whole cell lysates were harvested

in SDS lysis buffer, quantified, and western blot was performed for either HIF1α or HIF2α, and Amido Black or GAPDH were used as loading controls.

Appendix Figure 2. SLC2A1 transcripts encoding GLUT1 are increased in hypoxic cells and spheroids used for ASPCR. The expression of a hypoxia marker (GLUT1) was used to confirm hypoxia (1% O ) in cells and spheroid ASPCR samples via RT-qPCR. Data (n = 3 2 technical replicates) are mean ± s.e.m. Hypoxic GLUT1 mRNA expression normalized to RPLP0 and made relative to respective normoxic samples (dotted line at y=1).

100

Appendix Tables

101

102

103

104

105

106

107

108

109

(5’ → 3’) (5’ → 3’)

110

111

(5’ → 3’) (5’ → 3’)

112

(5’ → 3’) (5’ → 3’)

113