Regulation of Ferroptosis by a novel NADPH phosphatase MESH1

by

Chien-Kuang Cornelia Ding

University Program in Genetics and Genomics Duke University

Date:______Approved:

______Jen-Tsan Ashley Chi, Supervisor

______Kris Wood, Chair

______Pei Zhou

______Hiroaki Matsunami

Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Program in Genetics and Genomics in the Graduate School of Duke University

2018

i

v

ABSTRACT

Regulation of Ferroptosis by a novel NADPH phosphatase MESH1

by

Chien-Kuang Cornelia Ding

University Program in Genetics and Genomics Duke University

Date:______Approved:

______Jen-Tsan Ashley Chi, Supervisor

______Kris Wood, Chair

______Pei Zhou

______Hiroaki Matsunami

An abstract of a thesis submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Program in Genetics and Genomics in the Graduate School of Duke University

2018

i

v

Copyright by Chien-Kuang Cornelia Ding 2018

Abstract

Ferroptosis is a form of regulated cell death featured by lipid peroxidation and breakage of cell membrane. However, the molecular mediators and regulators are not fully understood. Here, we identified the metazoan homologues of ppGpp hydrolase

(MESH1) is an efficient cytosolic NADPH phosphatase, an unexpected enzymatic activity that is captured by the crystal structure of the MESH1-NADPH complex.

Ferroptosis elevates MESH1, whose upregulation depletes NADPH and sensitizes cells to ferroptosis. Conversely, MESH1 depletion rescues ferroptosis by sustaining the levels of NADPH and GSH and by reducing lipid peroxidation. Importantly, the ferroptotic protection by MESH1 depletion is ablated by suppression of the cytosolic NAD(H) kinase, NADK, but not its mitochondrial counterpart NADK2. MESH1 depletion also triggers extensive transcriptional changes that are distinct from the canonical integrated stress response, but show striking similarity to the bacterial stringent response. MESH1 depletion also leads to dNTP depletion and inhibition of cell proliferation in tumor cells.

Finally, we established Mesh1 knockout mouse model to study the physiological relevance of MESH1.

iv

Contents

Abstract ...... iv

Contents ...... v

List of Tables...... ix

List of Figures ...... x

Acknowledgements ...... xi

1. Introduction ...... 1

1.1 Bacterial Stringent Response ...... 1

1.1.1 ppGpp ...... 1

1.1.2 RSH protein family ...... 3

1.2.3 Transcriptional response in bacterial stringent response ...... 4

1.2 Ferroptosis: an iron-dependent oxidative cell death ...... 5

1.3 NADPH ...... 9

1.4 Overview of chapters ...... 10

2. The Role of MESH1 as NADPH phosphatase in Ferroptosis Survival ...... 11

2.1 Introduction: Enzyme kinetics and structure of MESH1 as a NADPH phosphatase ...... 11

2.2 Methods ...... 15

2.2.1 Cell culture and ferroptosis induction ...... 15

2.2.2 Genetic depletion of MESH1 and other genes by RNAi ...... 16

2.2.3 Overexpression of MESH1 by lentivirus or plasmid ...... 16

2.2.4 Quantitative real-time PCR ...... 18

v

2.2.5 Protein lysates collection and Western Blots...... 18

2.2.6 Measurement of cellular NADPH phosphatase activity ...... 19

2.2.7 NADP(H) measurements ...... 19

2.2.8 Lipid Peroxidation measurement ...... 20

2.2.9 Cell viability and cell death assays ...... 20

2.2.10 Cell fractionation ...... 21

2.2.11 Metabolomic profiling of MESH1 silenced cells ...... 22

2.3 MESH1 is a major NADPH phosphatase in human cells ...... 23

2.4 Regulation of MESH1 during stresses ...... 26

2.5 MESH1 depletion lead to ferroptosis protection by preserving NADPH ...... 28

2.6 MESH1 is a cytosolic NADPH phosphatase ...... 35

3. The phenotypes of MESH1 depletion ...... 38

3.1 Introduction ...... 38

3.2 Methods ...... 40

3.2.1 Cell Culture ...... 40

3.2.2 Cell Viability assays ...... 40

3.2.3 Genetic depletion of MESH1 by siRNA or shRNA ...... 41

3.2.4 cDNA microarray analysis ...... 42

3.2.5 RNAseq analysis ...... 43

3.2.6 dNTP quantification...... 43

3.2.7 Xenograft of MESH1-depleted tumors ...... 44

3.2.8 Generation of transgenic Mesh1 knockout mice ...... 45

vi

3.2.9 Isoprotenerol challenge in mice to profile cardiac function...... 46

3.3.10 Uropathogenic E.coli (UPEC) infection challenge ...... 47

3.3 Phenotypes of MESH1-depleted cells ...... 47

3.3.1 Transcriptional response ...... 47

3.3.2 dNTP depletion in MESH1 silencing cells...... 55

3.3.3 Inhibition of cell growth ...... 57

3.4 Phenotypes of Mesh1 knockout mice ...... 59

3.4.1 Validation of knockout ...... 59

3.4.2 General phenotypes ...... 61

3.4.3 Response to isoproterenol challenge ...... 63

3.4.4 Response to E.coli Urinary tract infection ...... 65

4. Discussion ...... 67

4.1 Physiological relevance of MESH1 as a NADPH phosphatase ...... 67

4.1.1 MESH1 as the only known NADPH phosphatase ...... 67

4.1.2 The regulation of MESH1 expression ...... 69

4.1.3 Potential MESH1 phenotypes that links to NADPH regulation ...... 71

4.2 Potential therapeutic implications of MESH1inhibition ...... 72

4.2.1 Inhibition of tumor growth ...... 72

4.2.2 Resistance to ferroptosis stress or oxidative stress ...... 73

4.3 Future direction of phenotypic studies of MESH1 knockout mice ...... 74

4.3.1 Potential consideration to study phenotypes in C57BL/6J background ...... 74

4.3.2 Innate immune response against intracellular microbes ...... 75

vii

4.3.3 Response under metabolic stresses ...... 75

5. Conclusion ...... 76

Appendix A: RNAi used in this study ...... 78

Appendix B: primers used in this study...... 79

Appendix C: antibodies used in this study ...... 81

References ...... 82

Biography ...... 92

viii

List of Tables

Table 1 Top 10 significant Gene Sets of KEGG using Gene Set Enrichment Analysis for MESH1 silencing cells...... 51

Table 2 Top 10 significant Gene Sets of Gene Ontology using Gene Set Enrichment Analysis for MESH1 silencing cells...... 52

Table 3 Predicted regulatory elements of MESH1 by GeneHancer ...... 70

ix

List of Figures

Figure 1. A schematic model of Ferroptosis...... 8

Figure 2. MESH1 is a mammalian NADPH phosphatase...... 14

Figure 3. MESH1 is the major NADPH phosphatase in human cells...... 25

Figure 4 MESH1 regulation during stresses...... 27

Figure 5. MESH1 depletion lead to ferroptosis protection by preserving NADPH...... 30

Figure 6 The protective effect of MESH1 depletion during ferroptosis is validated in multiple cell lines...... 32

Figure 7. MESH1 regulates ferroptosis by preserving reduced glutathione ...... 33

Figure 8 MESH1 is a cytosolic NADPH phosphatase ...... 37

Figure 9 MESH1 silencing induces a distinct pattern of transcriptional response...... 50

Figure 10 MESH1 silencing induces a transcriptional response that is partially ATF4 mediated...... 53

Figure 11 dNTP regulation in MESH1 silencing and overexpression cells...... 56

Figure 12 Reversible cell growth inhibition in MESH1-silenced cells...... 58

Figure 13 Mesh1 knock out alleles and validation of knock out ...... 60

Figure 14 Abnormal findings in Mesh1 homozygous knockout mice ...... 62

Figure 15 Parameters of Pressure-volume loop analysis of Mesh1 knockout mice ...... 64

Figure 16. Uropathogenic E.coli infection in MEF cells and mouse bladder...... 66

x

Acknowledgements

The works in this dissertation is a collection of efforts and helps from many people.

First, I would like to thank the Chi lab, especially my advisor Ashley Chi. Ashley is a very supportive mentor who guides and helps me on my way to pursue a career in research since 2009, even before I join the lab. He exemplifies hard working and critical thinking, and has been very helpful in supporting my research and my career. Xiaohu

Tang is also a supportive mentor who trained me with first hand experiences to build up my project when I just joined the lab. Jianli Wu is essential in conducting the kinome- wide screen and helping me to complete the MESH1 project. I also appreciate the helps and supports from Alice Sun, Wen-Hsuan Yang and Po-Han Chen for my work, as well as advice and inputs from Greg LaMonte, Jennifer Doss, Melissa Keenan, Jerry Lin and

Katie Walzer. I also want to thank undergraduates and high school students who have worked and helped me: Rachel Clark, Jadesoela Akinwuntan, Emily Xu, Steven Wu, and

Sarah Tian.

I would also like to thank my two essential collaborators, Pei Zhou and Josh Rose.

Without their input, my project would not have been the story it is. Pei guided me in exploring the biochemistry characters and physiological relevance of MESH1 as well as writing the manuscript, which I grew and learned a lot scientifically and personally.

xi

Josh’s hard work resulted in the resolution of crystallography structure of MESH1-

NADPH complex and the development of MESH1 enzymology assay, which is the essence of the story of MESH1 project.

In addition, I want to thank many other people who had helped my project at

Duke. So Young Kim was very helpful while we designed and conducted the RNAi screens. Mikhail Kovtun and Kai-Yuan Chen helped for the computational analysis of screens and sequencing projects. Ziqiang Guan and Ivan Spasojevic helped for the mass spectrometry analysis of metabolites. Rachael Rempel and Everardo Macias taught me how to do animal work. Cheryl Bock and Gary Kucera helped me to establish the Mesh1 knock out mice. Jeffrey Everitt performed the autopsy of Mesh1 knockout mice. Lan Mao and Dennis Abraham from Rockman lab helped for cardiac function analysis of Mesh1 knock out mice. Byron Hayes from Abraham lab helped to measure the response to

E.coli challenge during urinary tract infection. For people outside of Duke, I want to thank Beak Kim for dNTP quantification in cells.

I would like to thank all my committee members for guiding my research and supporting my career development throughout my PhD: Ashley Chi, Kris Wood, Pei

Zhou, and Hiroaki Matsunami. Their scientific inputs were critical for my work and the resulting publication. I really appreciate their time and advice.

I would like to thank the supports from UPGG, especially DGS Allison Ashley-

Koch and Simon Gregory, as well as DGSAs Elizabeth Labriola, Carolyn Weinbaum, and

xii

Leslie Mavengere. As a foreign graduate student I would have been so lost without you answering my questions and helping me to establish my life and career in Durham.

Personally, I want to thank my friends and family who support me throughout this process. Special thanks to my husband Sean Chang, whose unconditional love walk me through the ups and downs of graduate school. Many thanks to my friends from

Duke Chapel Choir, especially my best friend Karen Rivers, my lunch buddies Nicole

Dalzell, Natasha Parikh, and Rossie Clark-Cotton. I am so grateful to have you in this amazing journey.

xiii

1. Introduction

1.1 Bacterial Stringent Response

Stringent response is the main strategy for bacteria to cope with fluctuating nutrient supplies and metabolic and oxidative stresses(1, 2). During this process, alterations in transcriptional and metabolic profiles rapidly redirect energy from cell proliferation toward stress survival by reduction of biosynthesis, conservation of ATP, and blockage of GTP production(3). The stringent response is triggered by the accumulation of the bacterial “alarmone” (p)ppGpp (guanosine tetra- or penta- phosphate) through the regulation of ppGpp synthetases and hydrolases in the

RelA/SpoT Homologue (RSH) family(2). Recent studies suggest that the stringent response may also function in metazoans, as metazoan genomes encode a homologue of bacterial SpoT -- MESH1 that hydrolyzes ppGpp in vitro and functionally complements

SpoT in E. coli(4). Furthermore, Mesh1 deletion in Drosophila displays impaired starvation resistance and extensive transcriptional reprogramming(4). Despite these supporting lines of evidence, neither ppGpp nor its synthetase has been discovered in metazoans, mystifying the genuine function and the relevant substrate(s) of MESH1 in mammalian cells.

1.1.1 ppGpp

ppGpp and pppGpp are synthesized by adding a pyrophosphate group from

ATP to either GDP or GTP, respectively. (p)ppGpp are often referred to “magic spot” or

1

“alarmone”, which is found to accumulated in bacterial cells while exposed to stresses, such as nutrient deprivation, oxidative stresses or heat shock(5) . As a signaling molecule, (p)ppGpp executes its function via direct or indirect mechanisms, depends on the different organisms. For example, in E.coli, (p)ppGpp directly binds to RNA polymerase (RNAP) and regulates the transcription(6), and inhibits protein biosynthesis by directly binding to translation initiation factor 2 (IF2)(7). In B. subtilis, massive synthesis of (p)ppGpp indirectly depletes the pool of GTP, thus changing the preference of available pool of initiating nucleotides and further influencing the promoter preference of RNAP(8). The depletion of GTP in B. subtilis also inhibits a transcriptional repressor CodY, which requires bindings to GTP and branched amino acids and thus influences the gene expression of its target gene networks(9).

While bacteria are in exponential growth phase in the absence of stresses,

(p)ppGpp is presented in a low level and functions as the regulator of bacterial growth rate(10) and metabolism(11). However, when there is a limitation of nutrient, the accumulation of (p)ppGpp redirects the energy from cellular growth to support general metabolism(12); the accumulation of (p)ppGpp inhibits bacterial growth rate by inhibiting generation of ribosomal RNA, but promotes transcription of genes for amino acid biosynthesis(2), and binds to enzymes that involved in nucleotide metabolism to regulate nucleotide biosynthesis and uptake(13, 14). (p)ppGpp is also required for the

2

production and function of the alternative sigma factor S, which redirect the bacterial cells to stationary phase and lead to survival in stress conditions(15).

(p)ppGpp and the stringent response are also known to play an important role in the pathogenesis of microbes. For example, the virulence of Enterococcus faecalis in infective endocarditis is influenced by basal level of (p)ppGpp(16). (p)ppGpp also impacts the survival of S. aureus after phagocytosis(17) and its virulence(18). In addition, bacterial stringent response is also reported as a determinant of antibiotics resistance(19) and persistence(19).

In addition to bacteria, (p)ppGpp can also be found in the chloroplast of the plants. It is reported to regulate nuclear and chloroplast transcription, chloroplast translation, plant development, and stress response(20). However, neither (p)ppGpp nor the synthetase were ever identified in animal kingdom.

1.1.2 RSH protein family

RelA/SpoT Homologue (RSH) proteins are a superfamily of enzymes that synthesize and/or hydrolize (p)ppGpp(21). The classical long RSH proteins such as RelA and SpoT in E.coli comprise a (p)ppGpp synthesis domain (SYNTH), a (p)ppGpp hydrolysis domain (HD), a TGS domain (Threonyl-tRNA synthetase, GTPase, and SpoT proteins) with unknown function, and an ACT domain (Aspartate kinase, chorismate mutase and TyrA) that is widely found in a range of metabolic enzymes that are regulated by amino acid concentration(22). Typically either the SYNTH domain or HD

3

domain dominates the function of the enzyme. For example, SpoT is predominantly a

(p)ppGpp hydrolase that degrades (p)ppGpp to either GTP or GDP with pyrophosphate.

The long RSH family can be widely found in bacteria, archaea, and chloroplast of the plants. However, in opisthokont eukaryotes (fungi and animals), only the HD domain, which is dedicated only for (p)ppGpp hydrolyzing can be found, which is called Mesh1 .

Mesh1 are also found in some proteobacteria, but have not been reported in bacteria or plants. There are three hypotheses for the origin of Mesh1 in opisthokonts. First, it was directly inherited by eukaryotes as RSH homologues. Second, Mesh1 could enter eukaryote cells with mitochondrion. Third, Mesh1 could enter eukaryotes genome through horizontal gene transfer directly from bacteria. As most archaea lost Mesh1, and there is no mitochondrial localization or regulation peptide sequence found in Mesh1, the hypothesis of horizontal gene transfer is the most favorable hypothesis in present(21).

1.2.3 Transcriptional response in bacterial stringent response

One of the hallmarks of bacterial stringent response is the transcription response.

For example, in E.coli, treatment of serine hydroxymate mimic amino acid starvation and induced the stringent response, which altered the transcriptional profiles within minutes(23). The majority of differentially regulated genes were up-regulated, which included the induction of amino acid biosynthetic operons, genes involved in transcriptional regulation and alternative sigma factor S target genes. The down- regulated genes included tRNA-encoding genes and genes in glycolytic/gluconeogenic

4

pathways(23). Many of the up-regulated genes have the so called “stringent promoters”, which contained a GC-rich region between the -10 hexamer and the start of transcription site(24). In another study, the stringent response is induced by DL-norvaline in B. subtilis(25). The down-regulated genes were correlated to cell growth and reproduction, such as ribosome synthesis, DNA synthesis, and cell wall synthesis. On the other hand, the up-regulated genes include enzymes involved in amino acid biosynthesis, small molecule transporters, and regulators of sporulation/stationary phase(25). Although the detailed profiles and regulatory mechanism could varied from species to species, the transcriptional features of the bacterial stringent response are usually featured by the redistribution of the resources from growth and proliferation to biosynthesis pathways.

1.2 Ferroptosis: an iron-dependent oxidative cell death

Ferroptosis is a form of non-apoptotic, regulated cell death. It can be induced in specific types of cancer cells by depletion of intracellular glutathione, or inhibition of phospholipid hydroperoxidase GPX4 (Figure 1). Depletion of glutathione or inhibition of GPX4 impairs the lipid peroxide repair capacity, lead to oxidation of polyunsaturated fatty acid (PUFA)-containing phospholipids, and the accumulation of these lipid peroxides leads to the final cell death(26). Thus, ferroptosis stands at the intersection of metabolism disturbance and oxidative stress for the cells.

GPX4 uses reduced glutathione (GSH) to convert lipid hydroperoxides to lipid alcohol, thus eliminate the accumulation of lipid peroxides and prevent ferroptosis(26).

5

There are multiple ways to deplete intracellular glutathione to induce ferroptosis. One of the effective ways is to directly inhibit the enzymatic activity of GPX4 by compounds such as RSL3 and ML162 (Figure 1). By this approach, one can bypass glutathione metabolism pathways to directly induce lipid peroxidation and ferroptosis. Another effective way is to inhibit cystine imports by compounds such as erastin or sulfasalazine targeting the glutamate-cystine antiporter xCT (Figure 1). After import into cells, cystine is reduced to amino acid cysteine and is the rate limiting material for the biosynthesis of

GSH(27). By these methods of cysteine depletion, the intracellular pool of GSH is depleted, thus impairing the cellular capacity to mitigate lipid peroxides. This can also be achieved by depletion of extracellular cystine or addition of excessive extracellular glutamate to inhibit the xCT transporter system.

GPX4 generates lipid alcohol and oxidized glutathione (GSSG). GSSG is then recycled by glutathione reductase using NADPH, which is reviewed in section 1.3.

At the transcriptional level during ferroptosis, endoplasmic reticulum (ER) stress response is induced, including the branch of ATF3 (Activating Transcription Factor 3),

ATF4 (Activating Transcription Factor 4), and DDIT3 (DNA Damage Inducible

Transcript 3; CHOP)(28), which is a commonly seen gene expression signature during amino acid deprivation(29). Notably, CHAC1 (ChaC, cation transport regulator homolog 1), an enzyme catalyzes the cleavage of glutathione into 5-oxo-L-proline and a

Cys-Gly dipeptide, is dramatically upregulated in ferroptotic cells, which CHAC1

6

mRNA level can serve as a transcriptional marker of ferroptosis(28). In addition, CHAC1 has been shown to be critical for the GSH depletion during ferroptosis (30).

7

Figure 1. A schematic model of Ferroptosis.

Ferroptosis can be triggered by extracellular depletion of cystine, direct inhibition of cystine import by targeting xCT inhibitor, or inhibit GPX4 that utilize GSH to mitigate lipid peroxides (lipid ROS) accumulation.

8

1.3 NADPH

NADPH (Nicotinamide adenine dinucleotide phosphate) is the major reduced electron carrier in the human cells. As NADPH is required to regenerate GSH from

GSSG, which is the major metabolism pathway to maintain redox homeostasis, it is found to be a predictive biomarker to define ferroptosis sensitivity of the different cell types(31).

In mammalian cells, NADPH is compartmentalized in either nuclear/cytosolic or mitochondrial pools. In mitochondria, there is a direct trans-hydrogenation between

NADH and NADP+, which result in net generation of NADPH(32). In nucleus/cytosol,

NADPH is reduced from NADP+ majorly by G6PD (glucose-6-phosphate dehydrogenase) in pentose phosphate pathway, as well as by IDP (isocitrate dehydrogenase), ME (malic enzyme), and ALDH (aldehyde dehydrogenase)(32). The only known reaction to synthesize NADP+ is phosphorylation of NAD+ by NAD kinases(33). There are two known NAD kinases in human, NADK and NADK2, that are predominantly located in the cytosol or mitochondria(33, 34), respectively. However, there is no known NADP(H) phosphatase reported in mammalian cells, although such activities has been reported in rat liver lysates(35).

NAD+, the material for NADP+, can be generated by two pathways. The de novo biosynthesis of NAD+ utilized L-tryptophan. In contrast, the salvage pathway utilizes nicotinic acid or nicotinamide, which are degrading product of NAD+, to resynthesize

9

NAD+. Compared to the de novo biosynthesis pathway, the salvage pathway is more important for the overall NAD+ synthesis(32). While NADP(H) and NAD(H) are considered compartmentalized and not permeable across mitochondria membranes, there is new evidence that NAD+ might be permeable to mitochondrial membrane in mammalian cells(36).

1.4 Overview of chapters

In this chapter, Chapter 1, I introduced several topics that are relevant to the thesis. In Chapter 2, I described the biochemistry characters of human MESH1 as a

NADPH phosphatase and how it mediates the cell response in ferroptosis. This is a collaborative work with Dr. Joshua Rose in Dr. Pei Zhou lab (MESH1 structure and enzymology) and Dr. Ziqiang Guan (mass spectrometry validation of MESH1 substrate), which I contributed to the major functional and mechanistic studies of MESH1. In chapter 3, I focused on the phenotypic studies of MESH1 in human cells and in animal models, which is collaborative work with Dr. Kai-Yuan Chen (RNAseq data analysis),

Dr. Everardo Macias (xenograft), and Duke Transgenic Mouse Facility (MESH1 knockout mice). In Chapter 4, I summarized the findings of this research and described the potential future directions.

10

2. The Role of MESH1 as NADPH phosphatase in Ferroptosis Survival

2.1 Introduction: Enzyme kinetics and structure of MESH1 as a NADPH phosphatase

In this chapter, we elucidate the novel enzymatic function of human MESH1 as a

NADPH phosphatase, and how MESH1 plays a role as a NADPH phosphatase during ferroptosis. The biochemistry aspects of MESH1 are described by Dr. Joshua Rose in Dr.

Pei Zhou’s lab and in collaboration Dr. Ziqiang Guan, which laid the foundation of this research and is briefly introduced here.

Because there is no RelA or ppGpp synthetase homologue found in mammalian genomes, and ppGpp is not detectable in human cells (data not shown), we reasoned that MESH1 may function through different substrate(s) from ppGpp in human cells.

We first did metabolomic profiling of MESH1-silenced RCC4 cells and H1975 cells

(human lung adenocarcinoma cell lines), and generated a list of accumulating metabolites in MESH1-silenced cells that shared similar structure with ppGpp.

Collaborating with Joshua Rose in Zhou lab and Guan lab, we tested purified, recombinant human MESH1 against these metabolites, including UDP-glucose (uridine diphosphate glucose), GDP (guanosine diphosphate), CDP (cytidine diphosphate), creatine phosphate, GDP-fucose, thiamine pyrophosphate and various forms of inositol phosphates, but failed to detect any activity (data not shown). We then examined

NADPH, which shares many structural similarities with ppGpp but differs in that it

11

contains a 2’-phosphate instead of a 3’-pyrophosphate in ppGpp (Figure 2a). Bacterial

SpoT catalyzes the hydrolysis of the 3’-pyrophosphate group of ppGpp, but the crystal structure of the bifunctional RelA/SpoT homologue from Streptococcus dysgalactiae subsp. equisimilis captured an unusual ppGpp derivative, GDP-2’,3’-cyclic monophosphate, in the active site of the hydrolase domain(37), suggesting that the enzyme may accommodate a 2’-substituted phosphate group(37). Based on this observation, we reasoned that MESH1 could similarly hydrolyze the 2’-phosphate group of NADPH to yield NADH and an inorganic phosphate (Figure 2a).

Indeed, incubation of NADPH with recombinant human MESH1 readily released inorganic phosphate, which can be detected in the malachite green assay(38) (Figure 2b).

Importantly, the phosphate accumulation was linear over time (Figure 2c), reflecting continuous enzymatic turnover of NADPH by MESH1. The product is further validated by mass spectrometry, which revealed a peak at m/z 664.141 as [M-H]- as NADH (Figure

2d). Steady state kinetics indicates that human MESH1 is an efficient NADPH phosphatase in vitro (Figure 2e); its KM value of 0.12 ± 0.01 mM, which is comparable with other reported cellular enzymes that utilize NADPH, such as human phagocytic

NADPH oxidase (KM value of 0.11 mM(39)), supporting the role of MESH1 as a physiologically relevant NADPH phosphatase in human cells. Although MESH1 also catalyzes hydrolysis of 2’-phosphate group towards NADP+ in vitro, it is ~10-fold less efficient (kcat/KM = 1.4 ± 0.1 s-1mM-1), and its KM value of 0.43 ± 0.04 mM is significantly

12

higher than the estimated cytosolic concentration of NADP+ (4.9 ± 0.2μM in human colorectal carcinoma cell line(40)), rendering MESH1 an ineffective phosphatase for

NADP+ in cells (Figure 2e).

Previous biochemical and structural analysis has revealed MESH1 as a Mn2+- dependent enzyme(4). Accordingly, we found that the NADPH phosphatase activity of

MESH1 is also Mn2+-dependent, which the activity was largely depleted when Mn2+ was substituted with Zn2+ (Figure 2f). Likewise, mutations of MESH1 residues near the Mn2+ ion in the active site severely compromised the NADPH phosphatase activity by >103- to

104-folds in enzymatic assays (Figure 2f).

In order to visualize the molecular details of the NADPH recognition, Joshua

Rose in Zhou lab further determined the co-crystal structure of the hMESH1-NADPH complex, which the coordinate is available at the PDB databank (accession number:

5VXA). Briefly, the NADPH ribose group and its 5’-pyrophosphate group are extensively recognized by MESH1, with same catalytic amino acids in the active site as for ppGpp (E65 and D66). Notably, MESH1 lacks the ability to differentiate the adenine nucleotide of NADPH from the guanine nucleotide as was found in ppGpp. These structural observations reinforce the notion that MESH1 is a bona fide NADPH phosphatase(41).

13

Figure 2. MESH1 is a mammalian NADPH phosphatase. a, Chemical similarity between NADPH and ppGpp and the proposed chemical reaction of MESH1. b, Detection of the hMESH1-catalyzed phosphate release from NADPH by the colorimetric malachite green assay. The malachite green reagent changes from yellow to green in the presence of inorganic phosphate. c, Linear product accumulation of the NADPH dephosphorylation reaction catalyzed by hMESH1 (n=3, error bar indicates s.d.). d, Validation of the formation of NADH by mass spectrometry analysis. e, Enzymatic characterization of hMESH1 toward NADPH and NADP+ (mean ± s.d., n=4). f, Effects of Zn2+ substitution and active site mutation on the specific activity of purified recombinant hMESH1 (mean ± s.d., n=3). (Joshua Rose, Pei Zhou and Ziqiang Guan)

14

2.2 Methods

2.2.1 Cell culture and ferroptosis induction

The RCC4 cell line was provided by Denise Chan (University of California, San

Francisco, San Francisco, CA) and was authenticated by DDC (DNA Diagnostics Center)

Medical using the short tandem repeat method in November 2015. HEK-293T, H1975,

MDA-MB-231, 786-O, PC3, HT1080, A673, and PANC1 were obtained from the Duke

Cell Culture Facility and tested negative for mycoplasma. All cells were cultured in

DMEM with 4.5 g/L glucose and 4 mM Glutamine (11995-DMEM, ThermoFisher

Scientific) and 10% heat-inactivated fetal bovine serum (Hyclone # SH30070.03HI) in a humidified incubator, at 37 °C with 5% CO2.

Ferroptosis was induced by drug treatment with erastin (Cayman #17754), sulfasalazine (Sigma S0883), ML-162 (Cayman #20455), or RSL3 (Cayman #19288) at the concentrations indicated in the figure for one day, or by cystine deprivation. Cystine deprived media were prepared using the DMEM, high glucose, no glutamine, no methionine, no cystine media (ThermoFisher Scientific, 21013024) with 10% dialyzed

Fetal Bovine Serum (Sigma, F0392), L-Methionine (Sigma, M5308, final concentration 30 mg/L), L-glutamine (ThermoFisher, #25030-081, final concentration 4mM), and with added L-Cystine (Sigma, C6727) at 200 μM for regular full media or at the concentrations indicated in the figures.

15

2.2.2 Genetic depletion of MESH1 and other genes by RNAi

The siRNAs used in this study are listed in Table of appendix A. The efficacies of these siRNA were accessed by rt-qPCR and/or Western blots. For enzymology, 8 x 105

RCC4 cells were seeded in a 100 mm plate, and transfected after one day of growth with

600 pmole of siRNA and 40 μL Lipofectamine RNAiMAX (ThermoFisher Scientific,

#13778150) for 72 hours before the collection of cell lysates. For NADP(H) measurement,

8 x 104 cells were seeded per well of 6-well plate with 40 pmole of siRNA and transfected using 3μL of Lipofectamine RNAiMAX for 48 hours before drug treatment for one day.

For viability assays, 2,800 RCC4 cells were seeded per well on 96-well plates with 5 pmole of siRNA and 0.4 μL Lipofectamine RNAiMAX for at least 48 hours before drug treatment for one day. To collect RNA or protein, 105 RCC4 cells were seeded in a well of a 6-well plate with 40 pmole of siRNA and 3μL of Lipofectamine RNAiMAX for 72 hours before collection.

2.2.3 Overexpression of MESH1 by lentivirus or plasmid

Lentivral constructs of MESH1-WT were generated by cloning human MESH1

(RefSeq: NM_001286451.1) into lentiviral vector pLX302(42) (gift from David Root,

Addgene plasmid #25896) without expression of the V5 tag. MESH1-E65A and MESH1-

D66A constructed were generated based on MESH1-WT construct using QuikChange

XL Site-Directed Mutagenesis Kit (Agilent, #200517) and the online primer design tool

(https://www.genomics.agilent.com/primerDesignProgram.jsp retrieved on September

16

2018). The primer sequences can be found in Appendix B. These constructed were used to generate cell lines that are stably overexpressing MESH1.

Briefly, virus was generated by transfecting HEK-293T cells with a 0.1:1:1 ratio of pMD2.G: psPAX2 : pLX302 with TransIT-LT1 Transfection Reagent (Mirus, MIR2305). pMD2.G and psPAX2 were gifts from Didier Trono (Addgene plasmid #12259 and #

12260, respectively). Virus was collected 48 hours after transfection. Stable cell lines were generated by adding 200 μL virus to a 60 mm dish of RCC4 or other cells with 8 μg/mL polybrene. 1 μg/mL puromycin was used for selection. Complete death in blank infection dishes was used to determine the success of infection and the length of puromycin selection. The efficiency of overexpression was determined by Western blotting.

Transient overexpression of human MESH1-WT, MESH1-E65A or MESH1-D66A was achieved using the pCMV6-neo vector (OriGene, SC334209). Transient overexpression of mouse MESH1-WT and MESH1-D66N was achieved using the pcDNA4/myc-His A vector (ThermoFisher, V86320) without expressing the tag by cloning the cDNA with stop codon. For NADP(H) content assay or enzymology assay, 2 x 105 HEK-293T cells were seeded in one well of a 6-well plate for 24 hours, and then 1

μg of plasmid (empty vector or with MESH1-WT or MESH1-E65A) was transfected with

TransIT-LT1 Transfection Reagent (Mirus) for additional 48 hours before collection.

17

2.2.4 Quantitative real-time PCR

RNA was extracted using the RNeasy kit (Qiagen, 74104) following manufacturer’s instructions. 500 ng of total RNA with or without reverse transcriptase were prepared using a SensiFast cDNA synthesis kit (Bioline, BIO-65054) for real-time

PCR comparison with Power SYBRGreen Mix (ThermoFisher Scientific, 4367659).

Primers were designed across exons whenever possible using PrimerBot! developed by

Jeff Jasper at Duke University. The product of PCR was validated for specificity by DNA electrophoresis. Please refer to Appendix B for primer sequences used in this study.

2.2.5 Protein lysates collection and Western Blots

Cells lysates were collected in buffer with 50 mM Tris pH 8.0 and 200 mM NaCl

(for enzymology, Section 4.2.7), NADP(H) lysis buffer (for NADP(H) measurement, ATT

Bioquest #15272, described in Section 4.2.8), cell fractionation buffer (Section 4.2.10) or

Radioimmunoprecipitation assay (RIPA) buffer (Sigma, R0278) if not specified with protease inhibitors (Roche, 11836170001). Lysates are quantified by Bicinchoninic acid

(BCA) assay. For Western blots, 15-50 μg of lysates were loaded on SDS-PAGE gels, semi-dry transferred to PDVF membrane, blocked with 5% milk in TBST with 0.1%

Tween-20, then incubated with primary antibodies overnight at 4°C. After washing, the membrane was incubated with secondary antibody at room temperature for one hour.

The images were then developed by SuperSignal West Pico PLUS Chemiluminescent substrate (ThermoFisher, #34577) or Amersham ECL Prime Western Blotting Detection

18

Reagent (GE Healthcare Life Sciences, RPN2232) and exposed in ChemiDoc imaging system (Biorad). Please refer to Appendix C for antibodies used in this study.

2.2.6 Measurement of cellular NADPH phosphatase activity

Enzymatic assays for cellular MESH1 for NADPH were performed in a buffer containing 50 mM Tris pH 8, 200 mM NaCl, and 1 mM MnCl2. Specific activities of cell lysates were determined using 1 mM NADPH with the reaction time varying from 15-90 minutes. Assays were carried out using 3μg/μL cell protein lysates. Reactions were stopped by the addition of formic acid (3 M final concentration). The amount of released phosphate was assessed using the malachite green reagent as previously described(38) and established by Joshua Rose.

2.2.7 NADP(H) measurements

Amplite colorimetric NADPH assay kit (#15272, ATT Bioquest) was used to measure NADP(H) (both NADP+ and NADPH) in this study. Similar enzymatic approach were compared with multiple protocols and showed compatible NADPH preservation as fresh fixation of the cells with mass spectrometry measurement(40).

Briefly, cells were seeded into a 6-well plate with siRNA or plasmid as described.

NADP(H) was collected using 100 μL of the provided lysis buffer. NADP(H) content was measured and normalized by protein content of the lysate, quantified by BCA assay.

In cells treated with erastin or cystine deprivation, the NADP(H) change is normalized with NADP(H) measured in DMSO-treated cells.

19

2.2.8 Lipid Peroxidation measurement

Lipid ROS levels were determined using 10 µM of C11-BODIPY dye (D3861,

ThermoFisher Scientific) according to the manufacturer’s instructions. Cells were seeded and reverse transfected with siRNAs in six-well plates for two days, then the culture medium was replaced with 1µM erastin treatment for overnight. The next day, the medium was replaced to 10 µM C11-BODIPY containing medium for 1 hour. Later, the cells were harvested by trypsin and washed three times with ice-cold PBS followed by re-suspending in PBS plus 1% BSA. The amount of ROS within cells was examined by flow cytometry analysis (FACSCantoTM II, BD Biosciences).

2.2.9 Cell viability and cell death assays

Multiple viability and cell death assays were used in this study. The cell viability was measured by CellTiter-Glo assay (Promega #G7570), which quantified intracellular

ATP by luminescence as an indicator of viability, or Crystal Violet that stain for the attached cells and quantify by dissolving in 10% acetic acid and measure the absorbance at 570nm. Cell death under erastin with multiple time points was evaluated by CellTox-

Green Cytotoxicity assay (Promega, #G8741), which stain for DNA in dead cells. The cell death was further confirmed with CytoTox-Fluor assay (Promega #G9260), which quantified protease release from broken cells in the cell culture media. The fluorescence intensity, absorbance, and luminescence intensity were measured by plate reader

(FLUOstar Optima, BMG lab tech).

20

2.2.10 Cell fractionation

Cell fractionation to isolate protein from nucleus, cytosol, or mitochondria were described in details in Clayton et al(43). Briefly, RCC4 cells or HEK-293T cells were cultured in 150mm plate and reverse transfected with siRNA for 3 days to reach 80% confluence before fractionation. After washing twice in ice-cold PBS, the cells were collected by scraping and resuspended in 1mL RSB hypo buffer (10mM NaCl, 1.5mM

MgCl2, 10mM Tris-HCl pH7.5) on ice for 10 mins. The suspension was transferred to glass dounce tissue grinder (Wheaton #357542) with tight pestle (0.05mm clearance). The cells were broken by 30 strokes. 0.667mL of 2.5X MS homogenization buffer (525mM mannitol, 175mM sucrose, 12.5mM Tris-HCl pH 7.5, 2.5mM EDTA) were added, mixed and transferred to a new centrifuge tube for differential centrifugation. Lysates were centrifuged in 1300g for 5 min, 4°C to get nuclei in pellet. supernatant were centrifuged in 1300g, 5min for 4°C for two more time to get rid of nuclei, unbroken cells, and large membrane fragments, then centrifuged in 16,600g for 15min, 4°C to get mitochondria in pellet. Cytosolic protein in the supernatant was concentrated using 3kD filter (EMD

Millipore, UFC500396). Nuclei in pellet were washed with 0.1% Nonidet P-40 in PBS twice. Mitochondria in pellet were washed by 1X MS buffer (210mM Mannitol, 70mM sucrose, 5mM Tris-HCl pH 7.5, 1mM EDTA) once. Nuclei or mitochondria pellet were then resuspend in RIPA buffer and break open by sonication (Bioruptor, Diagenode) in

High setting, 30 seconds on / 30 seconds off, 15 min, 4°C

21

2.2.11 Metabolomic profiling of MESH1 silenced cells

The sample preparation process for metabolite quantification was carried out using an automated MicroLab STAR system (Hamilton). After removal of the protein fraction, the extract was divided into two fractions; one for analysis by LC (GSH, GSSG) and one for analysis by GC (cysteine). The LC/MS portion of the platform is based on a

Waters ACQUITY UPLC and a Thermo-Finnigan LTQ mass spectrometer, which consists of an electrospray ionization (ESI) source and linear ion-trap (LIT) mass analyzer. The sample extract was analyzed using acidic positive ion optimized conditions. Extracts reconstituted in acidic conditions were gradient eluted using water and methanol both containing 0.1% formic acid, while the basic extracts, which also used water/methanol, contained 6.5mM ammonium bicarbonate. The MS analysis alternated between MS and data-dependent MS2 scans using dynamic exclusion. The samples destined for GC/MS analysis were re-dried under vacuum desiccation for a minimum of

24 hrs prior to being derivatized under nitrogen using bistrimethyl-silyl- triflouroacetamide (BSTFA). The GC column was 5% phenyl and the temperature ramp was from 40° to 300°C in a 16 min period. Samples were analyzed on a Thermo-

Finnigan Trace DSQ fast-scanning single-quadrupole mass spectrometer using electron impact ionization. The information output from the raw data files was automatically extracted as discussed below. For ions with counts greater than 2 million, an accurate mass measurement could be performed. Accurate mass measurements could be made

22

on the parent ion as well as fragments. The typical mass error was less than 5 ppm.

Identification of known chemical entities was based on comparison to metabolomic library entries of purified standards. The combination of chromatographic properties and mass spectra gave an indication of a match to the specific compound or an isobaric entity.

2.3 MESH1 is a major NADPH phosphatase in human cells

In order to determine whether MESH1 is a significant contributor to the cellular

NADPH phosphatase activity, we measured the enzymatic activity in cell lysates extracted from two human cell lines, RCC4 cells (a human clear cell renal carcinoma cell line) and HEK-293T cells (a human embryonic kidney cell line). We found that lysates from MESH1-silenced cells had a substantial decrease of the NADPH phosphatase activity (Figure 3 a, b). Conversely, overexpression of wild-type hMESH1, but not the catalytically deficient E65A or D66A hMESH1 mutant, significantly enhanced the cellular NADPH phosphatase activity (Figure 3 c, d). Taken together, these observations verify hMESH1 as a significant contributor to the NADPH phosphatase activity in human cells. It is worth noting that the siRNA of MESH1 did not suppress NADPH phosphatase activity completely, indicating the possible existence of another enzyme that also contributes to the NADPH phosphatase activity in mammalian cells. One possible candidate is the NADP(H) phosphatase activity previously observed in rat liver lysates(35), though the reported enzymatic activity favors NADP+ and the identity of the

23

enzyme has remained unknown in animal cells(32). Therefore, to the best of our knowledge, the NADPH phosphatase activity of MESH1 is distinct from the previously observed activity and represents the first description of an NADPH phosphatase in human cells.

24

Figure 3. MESH1 is the major NADPH phosphatase in human cells.

NADPH phosphatase activity of cells was estimated by incubating NADPH with cell lysates with different MESH1 status; phosphate release over time was measured by malachite green assays and normalized by protein (μmol/min/mg total lysate protein). (a-b) hMESH1 is a significant contributor to the NADPH phosphatase activity in a, RCC4 lysates, b, HEK-293T lysates. (c-d)There is an increase of cellular NADPH phosphatase activity with overexpression of WT MESH1, but not with overexpression of enzymatically deficient MESH1 (c, MESH1-E65A or d, MESH1-D66A). Black dots represent siNT samples or empty vector control samples, blue dots represent siMESH1 samples, red dots represent MESH1-WT overexpression samples, and grey dots represent MESH1 mutant samples. Samples were analyzed by Western blot to confirm knock-down or overexpression efficiency. n=3-4 for each condition. Error bar indicates s.d. One-way ANOVA with Tukey HSD post-hoc test, *P<0.05, **P<0.01.

25

2.4 Regulation of MESH1 during stresses

After establishing hMESH1 as a significant contributor to the cellular NADPH phosphatase activity, we investigated whether MESH1 itself is regulated under different starvation stresses as in stringent response in bacteria. We exposed the cells under a series of nutrient deprivation, including different amino acids (cystine, glutamine, total amino acid) deprivations, glucose deprivation, and serum deprivation. We found that

MESH1 is up-regulated in all of these stresses in different degree (Figure 4a). As cystine deprivation also induce ferroptosis, we tested MESH1 level under erastin treatment and also found MESH1 was up-regulated. (Figure 4b). The regulation seems to be transcriptional, as RNA levels were up-regulated in two different cell lines. (Figure 4c, d).

Furthermore, overexpression of wild-type hMESH1 (MESH1-WT), but not an enzymatically deficient mutant (MESH1-E65A), significantly lowered intracellular

NADP(H) level (Figure 4e, f) and sensitizes cells to ferroptosis upon erastin treatment

(Figure 4g). Although the mechanism of MESH1 transcriptional regulation is yet to be explored, the data implicates a potential functional role of MESH1 in nutrient starvation and ferroptosis.

26

Figure 4 MESH1 regulation during stresses. a, MESH1 protein level is upregulated in stresses. b, MESH1 protein level is upregulated during ferroptosis. RCC4 Cells were treated by erastin for 1μM for 8h (by T. Sun). c-d, MESH1 RNA level is upregulated in stresses in c, RCC4 cells (by T. Sun), d, MDA-MB- 231 cells. F and Full: full media control; -C and CysDep: cystine deprivation; -Q and GlnDep: glutamine deprivation; -G and GlcDep: glucose deprivation; LS: low (0.05%) serum; -AA and no AA: total amino acids deprivation. e-f, Overexpression of WT MESH1, but not the MESH1 E65A mutant, significantly reduced the intracellular NADP(H) content in both RCC4 (e) and HEK-293T (f) cells. g, overexpression of MESH1-WT, but not MESH1-E65A, sensitize the cell death to low dose (0.625 μM) erastin treatment in RCC4 cells. Black dots represent siNT or empty vector samples, blue dots represent siMESH1 samples, red dots represent MESH1-WT samples, and grey dots represent MESH1-E65A samples. Error bar indicates s.d. (n=3 per condition for panel e- g). Statistical analysis: one-way ANOVA with Tukey HSD post-hoc test, *P<0.05, **P<0.01.

27

2.5 MESH1 depletion lead to ferroptosis protection by preserving NADPH

In order to further assess the MESH1 function in ferroptosis, we investigated whether depletion of MESH1 and its associated NADPH phosphatase activity could mitigate lipid peroxidation through the accumulation of NADPH and alteration of the glutathione redox potential, and alleviate ferroptotic death. Erastin treatment dramatically reduced the level of NADP(H) (Figure 5a) and increased degree of lipid peroxidation in control (siNT) cells (Figure 5c). While MESH1 silencing did not increase intracellular NADP(H) level (Figure 5b), it sustained a significantly higher level of

NADP(H) under erastin (Figure 5a). In addition, MESH1 silencing also markedly reduced the level of lipid peroxidation (Figure 5c). As a result of NADPH preservation erastin-induced death of RCC4 cells using CellTox-Green assay that measures cell death by DNA staining (Figure 5d). The ferroptosis protection of MESH1 silencing was further validated by other types of viability and death assays (Figure 5e-g). Such a ferroptosis- rescuing effect appeared to be general, as MESH1 depletion similarly promoted resistance to erastin-triggered ferroptosis in multiple cell lines, including HEK-293T,

H1975, MDA-MB-231, PC3, HT1080, A673, PANC-1, and 786-O (Figure 6). Importantly, the resistance to erastin-induced ferroptosis is directly correlated with the loss of the catalytic activity of MESH1, as the ferroptotic survival phenotype in MESH1-depletion cells was abolished when WT MESH1, but not the catalytically deficient E65A mutant of

MESH1 was expressed (Figure 5h).

28

Besides erastin, ferroptosis can be also triggered by sulfasalazine, another xCT inhibitor as well as cystine deprivation (Figure 1). In addition, various inhibitors of

GPX4 can also trigger ferroptosis downstream of NADPH/GSH (Figure 1). Therefore, we determined the ability of MESH1 silencing to rescue ferroptosis triggered by different means. Similar to erastin, cystine deprivation also depletes intracellular NADP(H), whereas the level of NADP(H) was sustained at a significantly higher level in MESH1- silenced (siMESH1) cells than in control cells (Figure 7a). MESH1-depletion also conferred resistance to ferroptosis induced by cystine deprivation and sulfasalazine

(Figure 7b, c) that trigger ferroptosis by depleting intracellular GSH. However, MESH1 depletion did not rescue ferroptosis induced by FIN56, ML162, or RSL3, different GPX4 inhibitors (Figure 7d-f). These observations indicate that MESH1-depletion rescues ferroptosis by NADPH-driven repletion of reduced glutathione (GSH). Indeed, although no significant effect on the levels of intracellular cysteine or oxidized glutathione (GSSG) were observed in MESH1-depletion cells, (Figure 7g, h), we observed a significantly increase of the GSH and GSH/GSSG ratio (Figure 7i, j). According to these data, we proposed that MESH1 plays a role during ferroptosis as a NADPH phosphatase (Figure

7k).

29

Figure 5. MESH1 depletion lead to ferroptosis protection by preserving NADPH.

30

a, The percentage change of NADP(H) after one day of 20μM erastin when compared to vector control (DMSO) treatment in RCC4 cells with non-targeting (NT) siRNA or 2 distinct MESH1-targeting siRNAs. MESH1 knock-down efficiency was validated by Western blot. b, No significant difference of NADP(H) content between MESH1-silenced and siNT RCC4 cells in the absence of stress. c, The lipid peroxidation of RCC4 cells with or without one day treatment of erastin (1μM) was accessed by C11-BODIPY and flow cytometry in cells transfected with non-targeting (NT) siRNA or 2 distinct MESH1- targeting siRNAs. (by W.-H. Yang) d, Time course of cell death in RCC4 cells with non- targeting (NT) siRNA or 2 distinct MESH1-targeting siRNAs under erastin (0.625μM) treatment up to one week. e-g, Erastin-induced cell death is rescued by MESH1 silencing in RCC4 cells using multiple indicated assays: Cytotox-Flour, measured protease release in the media for cell death (e); crystal violet staining for attached viable cells (f), and CellTiter-Glo for intracellular ATP and viability (g). h, Percentage of cell viability of RCC4 cells after erastin 5μM treatment for one day. Viability were compared between MESH1-WT or MESH1-E65A overexpression over MESH1-silenced cells and showed the restoration of sensitivity in ferroptosis only in MESH1-WT overexpressed cells.

31

Figure 6 The protective effect of MESH1 depletion during ferroptosis is validated in multiple cell lines.

Erastin-induced ferroptotic cell death of indicated cell lines is rescued by MESH1 silencing as accessed by CellTiter-Glo: a, RCC4 cells exposed to erastin 2.5 μM for one day; b, HEK-293T cells exposed to erastin 2.5 μM for one day, c, lung adenocarcinoma cell line H1975 with erastin 10 μM for two days, d, breast cancer cell line MDA-MB-231 with erastin 2.5 μM for one day, e, prostate cancer cell line PC3 with erastin 10 μM for one day, f, fibrosarcoma cell line HT1080 with erastin 2.5 μM for one day, g, Ewing’s sarcoma cell line A673 with erastin 2.5 μM for one day, h, pancreatic carcinoma cell line PANC1 with erastin 2.5 μM for one day, and i,clear cell renal carcinoma cell line 786-O with erastin 5 μM for one day. Black dots represent siNT samples, blue dots represent siMESH1 samples. Error bar indicates s.d. (n=3 per condition). Statistical analysis: one- way ANOVA with Tukey HSD post-hoc test, *P<0.05, **P<0.01

32

Figure 7. MESH1 regulates ferroptosis by preserving reduced glutathione a, The percentage change of NADP(H) after one day of cystine deprivation in RCC4 cells transfected with non-targeting (NT) siRNA or 2 distinct MESH1-targeting siRNAs. b-f, 33

The viability of RCC4 cells transfected with siNT (black) or two independent siMESH1 (blue) and treated with different concentrations of indicated ferroptosis-inducing means/agents for one day: (b) cystine deprivation; (c) sulfasalazine; (d) FIN56; (e) ML162; (f) RSL3. Each dot represents one sample; there are three independent samples (n=3) in each group. g-j, intracellular metabolites of RCC4 with non-targeting (NT) siRNA or MESH1-targeting siRNA were quantified by mass spectrometry and normalized by DNA content, (g) cysteine, (h) reduced glutathione (GSH), (i) oxidized glutathione (GSSG), and normalized intensity ratio of GSH/GSSG (j). k, schematic model of MESH1 regulation of glutathione during ferroptosis. Black dots represent siNT or empty vector samples, blue dots represent siMESH1 samples. Error bar indicates s.d. (n=3 per condition). Statistical analysis: ANOVA with Tukey HSD post-hoc test (a-f), Student’s t- test (g-j), *P<0.05, **P<0.01.

34

2.6 MESH1 is a cytosolic NADPH phosphatase

As the cytosolic and mitochondrial pools of NADP(H) and NAD(H) are compartmentalized in mammalian cells due to the impermeability of these molecules across the mitochondrial membranes(44), we reasoned that removal of the NAD kinase either from the cytosol (NADK) or mitochondria (NADK2) would reduce the distinct pools of NADP(H) and compromise the ferroptosis survival phenotypes of MESH1- silenced cells. Indeed, when the gene encoding the cytosolic enzyme NADK was silenced, the survival benefit of silencing MESH1 was largely eliminated (Figure 8 a, b).

In contrast, silencing the gene encoding the mitochondrial enzyme NADK2 did not affect the MESH1-mediated ferroptosis survival (Figure 8 a,b). Consistent with this notion, we found the cellular NADP(H) was significantly diminished when NADK was silenced, whereas silencing of NADK2 has little effect on cellular NADP(H) (Figure 8c), suggesting a limited role of NADK2 in this process. Corroborating these observations, cellular fractioning experiments(43) indicated that MESH1 was predominately enriched in the cytosolic, but not the mitochondrial/nuclear pools (Figure 8d)

Importantly, we also examined the gene expression level of MESH1 compared to

NADK, NADK2 and other genes involved in NADPH metabolism, including NNT

(NAD(P) Transhydrogenase, Mitochondrial), G6PD (Glucose-6-Phosphate

Dehydrogenase), PGD (Phosphogluconate Dehydrogenase), GSR (Glutathione reductase), and NOX1 (NADPH oxidase 1) by RNAseq. We found the expression levels

35

are compatible (within 0.5-4 fold compared to MESH1) (Figure 8e), suggest a MESH1 expression level is in an acceptable range compared to other NADPH metabolism genes.

36

Figure 8 MESH1 is a cytosolic NADPH phosphatase a, Knockdown of the cytosolic enzyme NADK, but not mitochondrial enzyme NADK2, eliminated the ferroptosis protection in MESH1 silenced cells. b, Normalized mRNA levels of MESH1, NADK and NADK2 in cells treated with indicated siRNAs. The mRNA abundance was quantified by rt-qPCR, normalized to that of β-actin, and presented in fold change (2∆∆CT). c, Co-silencing of NADK, but not NADK2, reduced the NADP(H) levels accomplished with MESH1 silencing. d, Cell fractionation separating cytosol from nucleus and mitochondria, showing that MESH1 is specifically enriched in the cytosolic compartment. e, Three plates of RCC4 cells were sequenced and the relative expression levels of the indicated genes were calculated by RPKM (Reads per kilo base per million mapped reads). If unspecified, siMESH1 indicates siMESH1-CDS.

37

3. The phenotypes of MESH1 depletion

3.1 Introduction

Stringent response induced by ppGpp accumulation in bacteria lead to a prominent transcriptional response(1), and the depletion of Drosophila Mesh1 also leads to profound transcriptional changes reminiscent of bacterial stringent response without ppGpp accumulation(4). Thus, in this chapter, we examined whether a similar transcriptional change can be observed upon silencing of human MESH1 in human cells.

Specifically, we profiled the transcriptional changes of MESH1-silenced cells to determine if the transcriptional signatures of this oxidative stress survival phenotype share any similarity with other types of stress response pathways, such as the mammalian integrated stress response (ISR)(29) and the bacterial stringent response.

In addition to transcriptional response, GTP homeostasis is also reported to be regulated during stringent response in Bacillus subtilis, which (p)ppGpp directly inhibits the activities of multiple GTP biosynthesis enzymes and lead to a decrease of GTP levels in cells(3). GTP is the precursor of dGTP, thus the downregulation of GTP during stringent response may interfere DNA replication and thus cell growth(3). Although human cells do not have ppGpp, we observed a transcriptional inhibition of both subunits of the dNTP generating enzyme ribonucleotide reductase (encodes by RRM1,

Ribonucleotide Reductase Catalytic Subunit M1 and RRM2, Ribonucleotide Reductase

Regulatory Subunit M2) during transcriptional profiling (Section 3.3.1), thus we

38

hypothesized that silencing human MESH1 may also lead to such dysregulation of nucleotides and deoxynucleotides , which will be discussed in Section 3.3.2.

From the analysis of the transcriptional response and the profiling of dNTPs in human cells during MESH1 silencing, we found there is an inhibition of cell cycle regulating pathways at transcriptional level (Section 3.3.1) and down regulation of dNTP (Section 3.3.2). In bacterial stringent response, ppGpp also contributes to growth rate inhibition by various mechanisms(10, 45). Thus, we hypothesized that there could also be a similar inhibition of cell proliferation during MESH1 silencing in human cells, which will be discussed in Section 3.3.3.

To understand the importance of MESH1 in organism level, we also looked at various Mesh1 manipulation in other model systems and generate a mouse knockout model by ourselves. The Drosophila Mesh1 knockout model is reported to have a transcriptional response and impaired starvation response in larvae stage.(4) In C. elegans, RNAi inhibition of Mesh1 homolog ZK909.3 was reported to cause moderate reduction of fat storage in the worm(46), which implicates a possible connection of

MESH1 and fat metabolism that might fit into the role of MESH1 as a NADPH phosphatase. Silencing ZK909.3 did not lead to embryonic lethality, larval arrest, slower growth, or maternal sterile or sterile progeny(47-49). The zebrafish mesh1 (hddc3) knockout phenotypes were not determined yet (Zebrafish mutation project(50), phenotypes searched on November 2018). The homozygous Mesh1 knockout mice

39

(Hddc3tm1.1(KOMP)Vlcg)generated by IMPC (International Mouse Phenotypeing Consortium) was reported to be viable with several abnormal morphology in a variety of organ system, including enlarged heart and spleen, abnormal kidney, liver, and pancreas morphology(51) (database searched on November 2018). Overall, in these animal models, silencing or knocking out MESH1 are not reported to be lethal. However, its physiological importance need to be further studied, especially under stresses. The generation of Mesh1 knockout mice and the phenotypes will be discussed in section 3.4.

3.2 Methods

3.2.1 Cell Culture

The RCC4 cell line was provided by Denise Chan (University of California, San

Francisco, San Francisco, CA) and was authenticated by DDC (DNA Diagnostics Center)

Medical using the short tandem repeat method in November 2015. All cells were cultured in DMEM with 4.5 g/L glucose and 4 mM Glutamine (11995-DMEM,

ThermoFisher Scientific) and 10% heat-inactivated fetal bovine serum (Hyclone #

SH30070.03HI) in a humidified incubator, at 37 °C with 5% CO2.

3.2.2 Cell Viability assays

The cell viability was measured by the CellTiter-Glo assay (Promega #G7570), which quantifies intracellular ATP by luminescence as an indicator of viability, or by

Crystal Violet staining of the attached live cells, which is quantified by dissolving in 10% acetic acid and measured by the absorbance at 570nm. The absorbance and luminescence

40

intensity were measured by using a plate reader (FLUOstar Optima, BMG lab tech). For clonogenic assay, 4,000 cells were seeded in each well of 6-well cell culture dish.

Doxycycline were added 24 hours after cells were seeded and attached to the plate and were replace every 72 hours.

3.2.3 Genetic depletion of MESH1 by siRNA or shRNA

Transient silencing of MESH1 is achieved by siRNA targeting MESH1 (siMESH1-

CDSs, siMESH1-3’UTRs) and compared to non-targeting siRNA (siNT), and the targeting sequence are specified in Appendix A. The efficacies of these siRNA were accessed by rt-qPCR, Western blots or both. For dNTP quantification, 1.5 x 106 RCC4 cells were seeded in a 150 mm plate, and transfected after one day of growth with 900 pmole of siRNA and 60 μL Lipofectamine RNAiMAX (ThermoFisher Scientific,

#13778150) for 72 hours before the collection of cell lysates. For viability assays, 2,800

RCC4 cells were seeded per well on 96-well plates with 5 pmole of siRNA and 0.4 μL

Lipofectamine RNAiMAX for at least 48 hours before drug treatment for one day. To collect RNA or protein, 105 RCC4 cells were seeded in a well of a 6-well plate with 40 pmole of siRNA and 3μL of Lipofectamine RNAiMAX for 72 hours before collection.

Induction of shRNA silencing against MESH1 were achieved by stably express doxycycline-inducible shRNA using lentiviral system Tet-pLKO-puro (Addgene plasmid #21915), which is a gift from Dmitri Wiederschain(52). 3 independent sequences

(clone ID: TRCN0000243218, TRCN0000243217, TRCN0000243216 from Mission human

41

shRNA library) targeting MESH1 were cloned into Tet-pLKO-puro plasmid. Virus were generated by transfecting HEK-293T cells with a 0.1:1:1 ratio of pMD2.G: psPAX2: Tet- pLKO-puro with TransIT-LT1 Transfection Reagent (Mirus, MIR2305). pMD2.G and psPAX2 were gifts from Didier Trono (Addgene plasmid #12259 and # 12260, respectively). Virus was collected 48 hours after transfection. Stable cell lines were generated by adding 200 μL virus to a 60 mm dish of RCC4 cells or H1975 cells with 8

μg/mL polybrene. 1 μg/mL puromycin was used for selection. Complete death in blank infection dishes was used to determine the success of infection and the length of puromycin selection. The efficiency of overexpression was determined by Western blotting.

3.2.4 cDNA microarray analysis

RNA samples were collected by RNeasy Mini Kit (Qiagen, #74104) according to manufacturer’s instructions. RNA qualities of each sample were assessed by Agilent

BioAnalyzer to insure the RNA integrity. cDNA were synthesized from 200ng RNA with the Ambion® MessageAmp™ Premier RNA amplification (Life Technologies). The gene expression level were interrogated with Affymetrix U133A gene chips and normalized by the RMA (Robust Multi-Array) algorithm. cDNA synthesis and microarray interrogation were performed by the Duke Microarray Core Facility. The influence of silencing MESH1 and ATF4 on gene expression was derived by the zero transformation process, in which we compared the gene expression level with siMESH1 or

42

siMESH1+siATF4 to the average of gene expression level in control (siNT) samples.

Significant differentially-expressed genes (t test, P<0.05) were identified and a selection of genes were validated by quantitative rt-PCR.

3.2.5 RNAseq analysis

Total RNA was isolated as detailed above. RNA quality was assessed using an

Agilent BioAnalyzer (Agilent). For RNAseq, 1 μg of total RNA was used to generate a cDNA library with Illumina TruSeq Stranded mRNA LT Sample Prep Kit – Set A

(Illumina, RS-122-2101) according to the manufacturer’s instructions. The library was pooled and sequenced using Illumina HiSeq 4000 with single-end 50 bp read length at

The Sequencing and Genomic Technologies Shared Resource of Duke Cancer Institute.

The data was analyzed using TopHat2(53) and HTSeq(54) with USCS hg19 as the reference genome. The differential analysis was performed using DESeq2(55). The genes that are differentially expressed (adjusted p value <0.05) in both siMESH1-CDS and siMESH1-3’UTR were selected and analyzed using MSigDB (Molecular Signatures

Database) database v6.1 and web site v6.3 with C5 (GO gene(56)) for Gene Ontology analysis.

3.2.6 dNTP quantification

2 x 106 cells were washed twice with cold PBS, than resuspended in 100μL 60%

Methanol. After vortex vigorously, the samples were heated at 95°C for 3 minutes, than centrifuge at 12,000 x g for 30 seconds. The supernatants were collected and dried under

43

vacuum using SpeedVac with medium heat. The dried pellets were then shipped to Dr.

Beak Kim’s lab at Emory University on dry ice. The pellets were subsequently resuspended in dNTP buffer (50mM Tris-HCl, pH 8.0 and 10mM MgCl2; 200μL for 2 x

106 cells.) and used for single nucleotide incorporation reaction(57). The quantification of dNTP in each samples were normalized by pmole/1x106 cells. The dNTP quantification were performed by Ms. Caitlin Shepard in Kim lab and was funded by R01 GM104198 and R01 AI049781-0.

3.2.7 Xenograft of MESH1-depleted tumors

H1975 cells with stably expression of Tet-pLKO-puro (empty vector) or MESH1- targeting shRNA were selected under puromycin; single cell colonies were screened by rt-qPCR for the best silencing effect achieved upon adding doxycycline for 3 days. 106 transduced H1975 cells were then injected as subcutaneous xenografts in 0.1mL of a 1:1 media to Matrigel (Corning #354234) solution and allows the tumors become palpable

(0.2cm3, anticipate 2 – 3 weeks). Then the mice were fed by doxycycline-containing food to induce the expression of control or MESH1-targetting shRNAs. Tumor volumes will be measured twice weekly using digital calipers. Mice and tumor growth will be monitored until tumors reach 1.5-2cm3, at which time mice were euthanized using carbon dioxide. The tumoral tissue were collected for weight measurement and further analysis.

44

3.2.8 Generation of transgenic Mesh1 knockout mice

We obtained a targeting construct (IKMC ID 92733) from EUCOMM (The

European Conditional Mouse Mutagenesis Program). The targeting construct contained

Mesh1tm1a(EUCOMM)Hmgu allele (abbreviated as Mesh1a). Briefly, the allele contains Mesh1 genomic sequence, a 6950bp long cassette (FRT – LacZ – loxP – neo – FRT) on intron 1 of

Mesh1, and two loxP sites franking critical exon 2 and exon 3. The LacZ reporter is flanked by an upstream splice acceptor (SA) and a downstream transcriptional termination sequence (polyA), terminating Mesh1 transcription while expressing LacZ.

Thus Mesh1a is a loss-of-function allele (“knock-out first”)(Figure 13a). In collaboration with Duke Transgenic Mouse Facility, the construct were cloned, linearlized, and transfected to mouse Embryonic Stem cells (ES cells) with 129/B6 background. The resulting homologous recombination leads to the incorporation of this “knock-out first” allele into the genome and thus disrupt Mesh1 expression. The recombination and incorporation of Mesh1a allele in ES cells were validated by Southern Blot (data not shown). Several clones of ES cells were injected into embryos of 129S1/SvlmJ mice and result in chimera mice.

The allele could be further cross with FLP deleter mouse, which removes the lacZ cassette, and leaving the intact Mesh1 with loxP sites franking exon 2 and 3

(Mesh1tm1c(EUCOMM)Hmgu, abbreviated as Mesh1fl), thus could be utilized for future cre- mediated conditional mutagenesis to generate tissue specific expression of

45

Mesh1tm1d(EUCOMM)Hmgu, abbreviated as Mesh1- (Figure 13a). In this project, we back cross the chimera mice carried Mesh1a to C57BL/6 mice for five generations for phenotypic studies.

3.2.9 Isoprotenerol challenge in mice to profile cardiac function

The animal in this study were handled according to Duke IACUC approved protocol (A223-15-08). Genetically engineered 8- to 12-wk-old Mesh1 knockout mice and wild type littermate in C57/B6 and 129S1/SvlmJ hybrid mice were used in this study. The procedure was conducted by Lan Mao in Dr. Rockman lab. Mice were anesthetized with a mixture of ketamine (100mg/kg) and xylazine (2.5mg/kg), than placed in the supine position on a heated operation board and a midline cervical incision was made to expose the trachea for endotracheal intubation and connected to a rodent ventilator. In vivo pressure-volume (P-V) analysis was performed as previously described(58, 59). Briefly, after bilateral vagotomy, the heart was exposed and a 1.4-Fr pressure-conductance catheter (Millar Instrument) was inserted into the right carotid artery and advanced retrograde into the left ventricle to record hemodynamics. A polyethylene-50 catheter was inserted into the right external jugular vein for isoprotenerol (ISO) infusion. Steady- state pressure and volume measurements were recorded at baseline and after 3-min Iso infusion (20pgg-1min-1). Pressure and volume measurements were recorded during the increase of the afterload, which is generated by transiently constrict the aorta by suture.

46

Parallel conductance (Vp) was determined by 10μL injection of 15% saline into the right jugular vein and was used to correct the P-V loop data.

3.3.10 Uropathogenic E.coli (UPEC) infection challenge

For bacterial infection assay in MEF (mouse embryonic fibroblast) cells, MEF cells were collected in our lab from female mouse pregnant for 14 days as previously described(60). Within two passages of expansion, 40,000 MEF cells were seeded into 12- well plates for one day and send to Abraham lab for following assays. The cells were infected with uropathogenic E.coli (UPEC) for one hour, and then the extracellular bacteria were killed with 100μg/mL gentamicin. This point was counted as time zero (0 h.p.i.), and following time points were collected from independent wells of cells. The bacteria load were quantified as C.F.U. (Colony-Forming Unit) as previously described(61).

For bacterial infection assay in vivo, UPEC were transurethrally inoculated into the bladder for one hour. Then gentamicin was inoculated to kill the un-invaded bacteria. The infected mice were kept overnight and then euthanized; the bladder were weight and homogenized to quantify the bacterial load.

3.3 Phenotypes of MESH1-depleted cells

3.3.1 Transcriptional response

Our transcriptome analysis using two independent siRNAs directed at MESH1 revealed that MESH1-silencing triggered extensive transcriptional responses in RCC4

47

cells. Gene Ontology (GO) analysis of differentially expressed genes showed that

MESH1-silencing repressed pathways relevant to cell cycle progression and DNA replication, such as the E2F1 (E2F transcription factor 1) and RRM2 (Ribonucleotide

Reductase Regulatory Subunit M2) genes (Figure 9 a-c), with a striking similarity to those observed in the bacterial stringent response(23, 45) and in Mesh1 deficient

Drosophila(4). The repression of DNA replication and cell cycle genes was further confirmed by GSEA (Gene Set Enrichment Analysis) analysis of GO and KEGG (Table 1,

Table 2). In addition, MESH1-silencing was also associated with the depletion of gene sets in sugar metabolism/catabolism and ribosome and mitochondrial translation (Table

1, Table 2). Importantly, the transcriptional changes of a subset of the genes affected by

MESH1 depletion, such as ATF3 (Activating transcription factor 3), SERPINE1 (Serpin

Family E Member 1), and VEGFA (Vascular Endothelial Growth Factor A), can be reversed by co-depletion of NADK, implicating a regulatory role of NADPH in mediating their transcriptional responses (Figure 9 d-f).

Interestingly, MESH1-silencing induced the expression of ATF4 (Activating transcription factor 4, also known as cAMP-responsive element binding protein 2)

(Figure 10a)—a key regulator of the integrated stress response—and its target genes, such as ATF3 (Figure 10b) and CTH (Cystathionine gamma-lyase, Figure 10c)(62), supporting the notion that MESH1-silencing also triggered the mammalian integrated stress response. However, ATF4-silencing did not alter the ferroptosis survival

48

phenotype of MESH1-silenced cells (Figure 10d). Furthermore, while ATF4-silencing abolished the induction of ATF3 and CTH (Figure 10b, c), it did not affect the transcriptional profile of other MESH1-silenceing responsive genes such as TPM1

(Tropomyosin 1) and PP1R12A (Protein Phosphatase 1 Regulatory Subunit 12A) (Figure

10e, f). In general, only 31% (294 out of 980genes) of the MESH1-silencing signature was affected by simultaneous ATF4 silencing (Figure 10g), suggesting that the MESH1- silencing-induced cell survival under lipid peroxidative stress has a distinct transcriptional profile and regulatory mechanism that functions independently of the integrated stress survival pathway.

To further investigate whether the transcriptional response induced by MESH1 silencing is functionally relevant in the context of ferroptosis survival, we examined the level of CHAC1 (ChaC Glutathione Specific Gamma-Glutamylcyclotransferase 1), a prominent biomarker essential for ferroptosis(28, 30) (Figure 10h). Although CHAC1 was reported to be an ATF4-response gene(63), its upregulation during ferroptosis- inducing erastin treatment was largely abolished by MESH1 silencing (Figure 10h). As

CHAC1 degrades glutathione(63), the mitigation of the erastin-triggered CHAC1 induction by MESH1 silencing reflects a distinct transcriptional response that may contribute to the ferroptosis protection phenotype.

49

Figure 9 MESH1 silencing induces a distinct pattern of transcriptional response. a, Top 10 repressed Gene Ontology processes of the transcriptome of MESH1-silenced RCC4 cells. MESH1 was silenced by siMESH1-CDS or siMESH1-3’UTR and profiled by RNA-seq. b-c, mRNA abundance of E2F1 and RRM2 validated by rt-qPCR. d-f, mRNA abundance of VEGFA, SERPINE1, and ATF3 tested by rt-qPCR. NADK-silencing reversed the inhibition of VEGFA by MESH1 silencing (d), but inhibited the induction of SERPINE1 and ATF3 by MESH1 silencing (e,f) mRNA abundance was normalized by β- actin, and presented in fold change (2∆∆CT) (b-f).

50

Table 1 Top 10 significant Gene Sets of KEGG using Gene Set Enrichment

Analysis for MESH1 silencing cells.

NAME (Depleted Gene Sets)* p-val q-val BASE_EXCISION_REPAIR 0 0.023 DNA_REPLICATION 0.001 0.112 RIG_I_LIKE_RECEPTOR_SIGNALING_PATHWAY 0.005 0.123 CELL_CYCLE 0.001 0.150 BLADDER_CANCER 0.023 0.356 VEGF_SIGNALING_PATHWAY 0.025 0.325 OXIDATIVE_PHOSPHORYLATION 0.012 0.345 HUNTINGTONS_DISEASE 0.008 0.326 FRUCTOSE_AND_MANNOSE_METABOLISM 0.045 0.343 GLYCEROPHOSPHOLIPID_METABOLISM 0.041 0.378 *Note: No KEGG Gene Sets are enriched in MESH1 silencing cells.

51

Table 2 Top 10 significant Gene Sets of Gene Ontology using Gene Set

Enrichment Analysis for MESH1 silencing cells.

NAME (Enriched Gene Sets) p-val q-val

CILIUM_MOVEMENT 0 0 CILIUM_MORPHOGENESIS 0 0.087 MOTILE_CILIUM 0 0.060 NEGATIVE_REGULATION_OF_EPITHELIAL_CELL_PROLIFERATION 0 0.069 CILIARY_PLASM 0 0.068 CILIUM_ORGANIZATION 0 0.084 CELL_PROJECTION_ASSEMBLY 0 0.200 MICROTUBULE_BUNDLE_FORMATION 0.009 0.178 SPERM_FLAGELLUM 0 0.161 UBIQUITIN_LIKE_PROTEIN_CONJUGATING_ENZYME_BINDING 0.012 0.153 NAME (Depleted Gene Sets) p-val q-val

TRANSLATIONAL_TERMINATION 0 0.021 MITOCHONDRIAL_TRANSLATION 0 0.048 DNA_REPLICATION_INITIATION 0.002 0.362 CELLULAR_PROTEIN_COMPLEX_DISASSEMBLY 0 0.374 HEXOSE_CATABOLIC_PROCESS 0.002 0.531 ORGANELLAR_RIBOSOME 0.001 0.569 NUCLEAR_UBIQUITIN_LIGASE_COMPLEX 0.003 0.498 MONOSACCHARIDE_CATABOLIC_PROCESS 0.003 0.596 INTERMEDIATE_FILAMENT 0 0.538 BASE_EXCISION_REPAIR 0.005 0.501

52

Figure 10 MESH1 silencing induces a transcriptional response that is partially ATF4 mediated. a. Upregulation of ATF4 protein in MESH1-silenced RCC4 cells. ATF4-silencing abolished the induction of ATF4-mediated genes ATF3 (b) and CTH (c) by MESH1- silencing, but it did not affect the ferroptosis resistance phenotype (d) or the repression of non-ATF4-mediated genes TPM1 (e) and PPP1R12A (f) in MESH1-silenced RCC4 cells. Gene expression is determined by rt-qPCR. g. The transcriptional signature of MESH1 silenced cells shows a distinct pattern from that of the ATF4-mediated integrated stress response. Transcriptional changes were calculated from the expression profiling of MESH1-silenced, ATF4-silenced, MESH1-ATF4-silenced, and control (siNT) RCC4 cells using cDNA microarrays. h, CHAC1 induction upon erastin treatment is inhibited by MESH1 silencing, determined by rt-qPCR. Black dots represent siNT samples, and blue dots represent siMESH1 samples. Error bars indicate s.d. (n=3 for each condition). Statistical analysis: (b-c, e-f, h) two-way ANOVA with Tukey-HSD post-hoc, (d) one- way ANOVA with Tukey-HSD post-hoc, **P<0.01, ***P<0.005, N.S., not significant. (g) Transcriptional signature of MESH1 silencing (siMESH1 signature) is defined as

53

differential gene expression with fold change > 1.41 with t-test P<0.05. ATF4-mediated genes are defined as having a siMESH1 signature that are reversed upon simultaneous ATF4-silencing with fold change > 1.41 and t-test P<0.05 compared to MESH1-silencing toward the direction of siNT samples. Note that genes affected by siATF4 alone are identified as off-target effect (due to low expression level of ATF4 in unstressed RCC4 cells) and are excluded from the analysis.

54

3.3.2 dNTP depletion in MESH1 silencing cells

As two subunits of ribonucleotide reductase RRM1 (data not shown) and RRM2

(Figure 9b) are both depleted in MESH1-silencing cells, we predict that dNTPs, the product of ribonucleotide reductase, will also be depleted in MESH1 silenced cells.

Indeed, in MESH1-silenced RCC4 cells, all four dNTP are depleted (Figure 11a).

Interestingly, when we applied siMESH1 in cells stably overexpressed MESH1, dNTP level still dropped significantly in both siMESH1-CDS and siMESH1-3’UTR-2, but the level of reduction is less prominent, especially in siMESH1-3’UTR-2 samples (Figure

11b). Notably, overexpression of MESH1 doesn’t lead to an increase in the dNTP level

(Figure 11c); in fact, there is a drop of baseline dGTP and dTTP in MESH1 overexpressed cells (Figure 11c). Overall, the overexpression of MESH1 only partially rescued the dNTP depletion of MESH1 silenced cells, indicates other potential regulation mechanism. We further examined the gene expression level of RRM2 in RCC4 cells with either empty vector, MESH1-WT or MESH1-E65A stably overexpression, and we found that paradoxically both MESH1 silencing and MESH1 overexpression inhibit RRM2 expression; moreover, both WT and E65A mutant of MESH1 inhibit RRM2 expression

(Figure 11d), indicates the regulation of RRM2 expression is independent from the enzymatic activity.

55

Figure 11 dNTP regulation in MESH1 silencing and overexpression cells. a-b, Normalized dNTP level of (a) RCC4 cells with empty vector silenced by siMESH1s, (b) RCC4 cells with MESH1 overexpression silenced by siMESH1s. c, dNTP level of RCC4 cells with endogenous MESH1 (EV) or MESH1-WT overexpression (WT). d, RRM2 mRNA level is determined by cDNA microarray.

56

3.3.3 Inhibition of cell growth

In agreement of transcriptional change, we found a prominent cell growth inhibition in MESH1-silenced cells in both cell culture system (Figure 12a) and in vivo tumor growth using xenograft (Figure 12b). MESH1 silencing also inhibits clonogenic growth of RCC4 cells (Figure 12c). Importantly, the cell growth inhibition is reversible upon restoration of MESH1 level by doxycycline withdrawal (Figure 12d). This phenotype is reminiscent to growth inhibition in bacterial stringent response upon ppGpp accumulation(45), indicates an evolutionary conserved functional role of MESH1 in mammalian cells.

57

MESH1

β-tubulin

Figure 12 Reversible cell growth inhibition in MESH1-silenced cells.

MESH1 was silenced using doxycycline-inducible shRNA lentivirus in this figure. a, normalized cell count of RCC4 cells. b, tumor volume measured by digital caliber (left) and tumor weight (right) of xenograft H1975 cells c, clonogenic assay of RCC4 cells. d, estimated cell division times were measured by cell count of RCC4 cells; doxycycline were withdrawn on day 4 and the protein level of MESH1 was restored upon Day 13, accessed by Western blot.

58

3.4 Phenotypes of Mesh1 knockout mice

3.4.1 Validation of knockout

We used PCR primers targeting Exon 1 and the inserted cassette to genotype

Mesh1a allele, and used PCR primers targeting Exon 1 and 2 to genotype Mesh1+ allele.

An example of genotyping result is shown in Figure 13b. The heterozygous mice

(Mesh1+/a) were picked and then back crossed to C57BL/6J mice. After three generations, we breed Mesh1+/+, Mesh1+/a, and Mesh1a/a for Western blot validation of protein extraction.

Homogenate lysates from kidney, brain and heart were tested and the result is shown in

Figure 13c, which there are less MESH1 expression in Mesh1+/a mouse, and no detectable

MESH1 in Mesh1a/a mouse. beta-gal staining were performed on embryos of Mesh1+/+ and

Mesh1+/a embryos to validate the expression of lacZ reporter driven by Mesh1 promoter, which we found the staining is only visible on the amniotic membrane of Mesh1+/a embryo (data not shown), indicating low MESH1 expression in most tissue during embryonic stage.

59

Figure 13 Mesh1 knock out alleles and validation of knock out a, different Mesh1 alleles: Mesh1a as knock-out first allele; Mesh1fl as wild type allele with conditional knock-out potential, and Mesh1- as complete knock out. b, PCR genotyping result of Mesh1 wild type and heterozygous mouse. c, Western blot validation of MESH1 expression status in mice with different Mesh1 status in kidney, heart and brain.

60

3.4.2 General phenotypes

After three generations of backcross to BL6 background, the Mesh1a/a mice and their wild type and heterozygous litter mates were kept for more than two years for general phenotypes observations. The genotypes of pups from Mesh1+/a x Mesh1+/a followed the ratio of 1:2:1 (Mesh1+/+:Mesh1+/a::Mesh1a/a), which indicates no embryonic lethality of Mesh1 homozygous knockout mice. No obvious differences of body size or body weight were observed between litter mates (data not shown). Several common defects seen in wild type BL6 mice, such as microphthalmia, ocular infection, and anal prolapse, were reported when the mice were close to two year olds, and these mice were sacrificed for humane reasons. Necropsy were performed on one male and one female knockout mice and compared to their wild type litter mates at the age of two (Figure 14).

Interestingly, a hemorrhagic mass was found in the area of prostate in the Mesh1 knockout male mouse, and a paraovarian cyst was found in the Mesh1 knockout female mouse. These lesions were not found in their wild type littermates. Whether these lesions were correlated with the genetic knockout or coincidental findings related to aging still require further investigation.

61

Figure 14 Abnormal findings in Mesh1 homozygous knockout mice

Left panel, an abnormal cystic hemorrhagic mass were found in the area of prostate (arrow) in Mesh1 knockout male mouse. Right panel, a paraovarian cyst were found in Mesh1 knockout female mouse.

62

3.4.3 Response to isoproterenol challenge

In this experiment, 4 female wild type mice, 4 female knockout mice, 1 male wild type mouse and 1 male knockout mouse in age 20 wks were tested for their cardiac function using in vivo pressure-volume (P-V) analysis upon different dosage of isoproterenol challenge. Isoproterenol is a beta agonist that is known to increase heart rate and cardiac output. However, in this experiment, the dp/dtmax of left ventricle (a commonly used index of ventricular systolic function) and heart rate only increased until the maximum dosage of isoproterenol (Figure 15c-f), indicating a poor response in this batch of experiment in general. In addition to the heart rate and dp/dtmax, we also measured LVP (Left ventricle pressure), EDP (End-diastolic pressure), and dp/dtmin. The only difference was seen between male wild type and knock out mice for EDP (Figure

15a) but not in female (Figure 15b). There is no obvious difference in other parameters we measured. In consideration the poor response to isoproterenol in all the mice we tested and small n (n=1) for male mice, it is unclear whether the difference of EDP between wild type and knockout male mice is significant. Further testing with larger numbers will be needed to determine the significance and physiological relevance of this finding.

63

Figure 15 Parameters of Pressure-volume loop analysis of Mesh1 knockout mice

End Diastolic Pressure (EDP) of left ventricle of a, male or b, female mice; dp/dtmax of c, male or d, female mice; heart rate (HR) of e, male or f, female mice. ISO50, isoproterenol 50pg; ISO500, isoproterenol 500pg; ISO1000, isoproterenol 1000pg, ISO5000, isoproterenol 5000pg.

64

3.4.4 Response to E.coli Urinary tract infection

In this experiment, we first tested whether there is a difference of clearance of intracellular UPEC infection of MEF cells between different genotypes. We found that the MEF cells with heterozygous Mesh1 have the best UPEC clearance, especially in 18 hours post infection (Figure 16a, b), which there were less intracellular UPEC remained in the heterozygous MEF cells. To test whether the heterozygous mice actually have better UPEC clearance in vivo, we challenge 3 wild type and 3 heterozygous Mesh1 female mice at the age of 8 weeks for UPEC infection (Figure 16c). The result suggests a lower CFU in the heterozygous Mesh1 mice, but not statistically significant. Therefore, the inconclusive results may need further testing with increased animal numbers will be required to determine the significance of this finding.

65

Figure 16. Uropathogenic E.coli infection in MEF cells and mouse bladder. a-b, The remained intracellular Uropathogenic E.coli (UPEC) is measured in 18 and 40 hour post infection (Hrpi) in two batches of independent MEF cells with different genotypes. c, UPEC infection is measured in urinary bladder of Mesh1 wild type and heterozygous knockout mice.

66

4. Discussion

In this dissertation, I described the discovery of the novel enzymatic activity of human MESH1 as a NADPH phosphatase and its physiological relevance as a regulator of ferroptosis survival. In addition, I also described the cellular and organism phenotypes of MESH1 depletion in human cell line models and mouse model. Here, I discuss the physiological relevance of MESH1 as a NADPH phosphatase at section 4.1.

In section 4.2, I discuss the potential therapeutic implications of MESH1 inhibition based on the cellular phenotypic findings in Chapter 3. In section 4.3, I discuss the future directions to study MESH1 phenotypes at organism level using animal models. I end this chapter with final conclusion in Section 4.4.

4.1 Physiological relevance of MESH1 as a NADPH phosphatase

4.1.1 MESH1 as the only known NADPH phosphatase

NADPH phosphatase has been identified in plants(64, 65), bacteria(66) and archaea(67). In animals, such activity was found in rat liver Golgi apparatus and mitochondria, predominantly against NADP+(35, 68), but the identity of the enzyme has never been revealed. Thus, to our knowledge, human MESH1 is the first NADPH phosphatase identified in mammalian cells. In chapter 2, we provided compelling evidence that human MESH1 is a bona fide NADPH phosphatase in vivo. However, it is important to note that even with good silencing efficiency by RNAi, we can’t eliminate all NADPH phosphatase activity in human cell line models by silencing MESH1 alone

67

(Figure 3 a, b). In fact, we also tried CRISPR-Cas9 approach to knockout both alleles of human MESH1 and tested the remaining cellular NADPH phosphatase activity, but some remaining activities can still be observed (data not shown).

One potential explanation of this data is that there might be nonspecific phosphatase activities from other enzymes that contribute to the NADPH phosphatase activity of this assay. One potential contributors is the Nudix hydrolase superfamily, which is capable to cleave (p)ppGpp in bacteria(69) and plants(70). It is known to cleave a wide range of organic pyrophosphates, and one of the member AtNUDX19 in

Arabidopsis thaliana is a known NADPH-recycling enzyme(71), but the activity against 2’- phosphate group of NADPH has not been reported.

As NADPH is compartmentalized, it won’t be surprised that there might be another NADPH phosphatase in mitochondria or other organelle. We tested human

RCC4 cells and HEK-293T cells by fractionating cytosol, nucleus and mitochondria and tested their individual NADPH phosphatase activity. Although the distributions of

NADPH phosphatase activity among organelles are different between these two cell line models, we did not identify a strong NADPH phosphatase activity in mitochondria

(data not shown). Therefore, we did not find a strong evidence to indicate the mitochondria counterpart of MESH1.

68

4.1.2 The regulation of MESH1 expression

We observed an upregulation of MESH1 in many stresses (Figure 4) at both RNA and protein level. This indicates that transcriptional regulation might be the major route of MESH1 expression control in human cells. In this study, we manipulated the MESH1 expression by RNAi, but we have not covered how MESH1 itself is regulated under these stressors. One possibility is to look at the regulatory elements near the regulatory regions of MESH1. In human genome, MESH1 allele is located on chromosome 15 and is very close to UNC45A (Unc-45 Myosin Chaperone A). These two genes shared several regulatory elements, including several promoter/enhancer regions. Numerous of transcription factors are predicted to be able to bind to these regions, a few examples were summarized in Table 3. These are information from a collection of databases from the Encyclopedia of DNA elements (ENCODE), the Ensembl regulatory build, the functional annotation of the mammalian genome (FANTOM) project and the VISTA

Enhancer Browser by GeneHancer(72). Further experimental validation is required to elucidate the molecular mechanism of MESH1 regulation.

69

Table 3 Predicted regulatory elements of MESH1 by GeneHancer

GeneHancer type GH score TSS distance Examples of TF binding identifier (kb) sites GH15J090928 promoter/enhancer 2.3 +4.6 E2F1, MYC, ETV4 GH15J090934 promoter/enhancer 2 -0.2 HDGF, MYC, E2F1 GH15J090933 enhancer 0.5 +1.8 HMBOX1, EBF1 GH15J090835 promoter/enhancer 2.2 +95.9 HDGF, ATF4, HDAC1, GH15J090829 enhancer 1.4 +104.6 HDGF, EBF1, CEBPB

70

4.1.3 Potential MESH1 phenotypes that links to NADPH regulation

NADPH is an important coenzyme for many biochemical pathways(32). In chapter 2, we focus on how MESH1 influence the reduction of GSH from GSSG by regulating NADPH level during ferroptosis, and how this reduction influences the survival during ferroptosis. In addition to GSH regeneration, NADPH is also known to provide reducing power to thioredoxin system that is also able to scavenge reactive oxygen species and thus maintaining redox homeostasis(73). NADPH is also an important coenzyme to generate oxidative burst by NADPH oxidase enzyme such as

NOX1 in phagocytes in animals. As this is an important mechanism to eliminate pathogens, it might be valuable to test whether the MESH1 knockout animals might have a stronger immunity against catalase positive microbes such as S. aureus.

Moreover, NADPH is also involved in the reductive biosynthesis such as fatty acid synthesis. Fatty acid synthetase (FAS) utilizes acetyl-CoA to synthesize fatty acid while consuming NADPH as a reducing agent. As a result, we predicted that MESH1 silenced cells might have an increase of fatty acid synthesis. We indeed observed an accumulation of fatty acids metabolites, including multiple polyunsaturated fatty acids

(PUFA) and long chain fatty acids in the MESH1-silenced RCC4 cells (data not shown).

The enrichment analysis of metabolites also shown an enrichment of fatty acid metabolism (p<0.05, Enrichment score = 1.5). In ZK903.3 (ortholog of human MESH1 in C.

71

elegans) silenced worm, it is also reported to have increased fatty acid content in body composition(46), which is compatible with our finding.

Although NADPH itself is never reported as a signaling molecule that regulates transcription, we found a subset of gene expression that is altered by MESH1 silencing and can be reversed while co-silencing with NADK (Figure 9d-f), which is upstream of

NADP+ synthesis and thus block NADPH preservation by MESH1 silencing. One potential mechanism is through regulation of NAD+ as a non-canonical initiating nucleotide(74, 75). This is a non-traditional RNA 5’-cap that could regulate transcription initiation by RNA polymerase (RNAP) and RNA stability that was first discovered in bacteria but also recently reported in yeast(76) and human cells(77). Whether these

“NADPH-regulated genes” or their upstream regulators are directly capped by NAD+ could be further validated in the future using techniques such as NAD CaptureSeq(78).

4.2 Potential therapeutic implications of MESH1inhibition

4.2.1 Inhibition of tumor growth

As shown in Figure 12, silencing MESH1 leads to an inhibition of cell growth for tumor cells in both cell culture model and xenograft model. Using CRISPR-Cas9 approach, we also observed a slower cell growth in tumor cells with both MESH1 alleles being knockout (data not shown). Despite such a strong growth inhibition, MESH1 deletion is not lethal in animal models, including C. elegans(47), Drosophila(4), and mouse models by ourselves and others(51). No major abnormality or body size difference was

72

observed in normal growth condition in these animal models. Thus, it is possible that the growth inhibition phenotypes might be specific in malignant tumor cells, including

VHL-deficient renal cell clear cell carcinoma and lung adenocarcinoma, which are the models we tested. This implicates that inhibiting MESH1 by genetic or chemical means could a feasible strategy to stop tumor growth in vivo. This possibility will be further explored in future.

4.2.2 Resistance to ferroptosis stress or oxidative stress

As shown in Figure 6, the protective effect of MESH1 silencing against ferroptosis is validated in many cell types. Ferroptosis is a form of cell death that is seen in some pathological settings such as cell death during acute renal injury(79). Thus, inhibition of MESH1 could potentially reduce the cell death in these pathological conditions. Moreover, as a regulator of NADPH, MESH1 silencing could have potential therapeutic application in other disease settings associated with oxidative stress such as hypoxic injury or ischemic-reperfusion injury of the brain(80).

A couple of (p)ppGpp analogues have been developed to inhibit bacterial stringent response(81-83). However, its efficacy against human MESH1 is unknown. We also tested a few compound based on prediction of molecular docking in silico against human MESH1, but the IC50 are all in a range of mM level. Further screening is required to develop an efficient MESH1 inhibitor.

73

4.3 Future direction of phenotypic studies of MESH1 knockout mice

4.3.1 Potential consideration to study phenotypes in C57BL/6J background

We generated our Mesh1 knockout mice in a hybrid background and have been back crossed to the C57BL/6J strain, which is one of the most commonly used mouse strain in the field. However, it is important to notice that C57BL/6J strain has a deletion of exons 7-11 of the Nnt gene(84), which leads to non-formation of the Nicotinamide nucleotide transhydrogenase (NNT) protein. NNT is a mitochondrial protein that convert NADH and NADP+ to NADPH and NAD+, thus overall promotes the reduction of NADP+ to NADPH. Lacking of NNT lead to several phenotypic consequences in

C57BL/6J strain, such as difference in response to hypoxic-ischemic brain injury(85), hypertension development(86), and exacerbation of atherosclerosis(87). Although

MESH1 is a cytosolic protein that regulates cytosolic NADPH, it is possible that the loss of Nnt gene in mitochondria could interfere the NADPH metabolism in general and influence the phenotypic read out of Mesh1 knockout mice. Thus, it is important to perform the phenotypic studies only by comparing the mice with similar genetic background (C57BL/6J), ideally their littermates. Alternatively, the MESH1 deficient phenotypes should be analyzed in other mouse strains to avoid this potential confounding factor.

74

4.3.2 Innate immune response against intracellular microbes

In our transcription analysis, we identified a few genes that are related to innate immune response are upregulated in MESH1-silencing human cells (data not shown), including GBP1 (Guanylate Binding Protein 1) and several other interferon inducible genes. This could be one of the mechanisms of better intracellular UPEC clearance of heterozygous Mesh1 MEF cells (Figure 16), although the gene expression profiles of these MEF cells need to be further validated. Another possibility is that mammalian

MESH1 could potentially regulate (p)ppGpp of the intracellular pathogens, thus influence the virulence and host-pathogen interaction. This can be validated by comparing the (p)ppGpp level of bacterial pathogens in host cells with different Mesh1 status. Similar idea can also be tested in other intracellular pathogens which stringent response plays an important role in pathogenesis, such as mycobacterium(88). In summary, the potential different responses to intracellular pathogens need to be further validated with different models and larger number of animals.

4.3.3 Response under metabolic stresses

As reported in Drosophila, Mesh1 knockout larvae are smaller and have slower cell proliferation during starvation(4). Thus, it is possible that the Mesh1 knockout phenotype will only be prominent under specific diet or under starvation. A possible experiment could be done using metabolic cages that record physiological parameters of

75

the mice during normal feeding condition and compare to a short-term starvation condition.

As MESH1 is a NADPH phosphatase, it is possible that Mesh1 deficient mice may have a better preservation of NADPH during oxidative stress condition such as ischemic-reperfusion injury. We tested protein lysates of the heart from animals with

LAD (left anterior descending artery) ligated and released as a model of ischemic- reperfusion injury of the heart, and we found that MESH1 is significantly upregulated after the injury (data not shown). Thus, with different status of Mesh1, the extent of injury and the resulting cardiac function could be different in these mice and could be a measurable phenotype.

5. Conclusion

In summary, this dissertation described the identification of human MESH1 as a novel cytosolic NADPH phosphatase. By silencing MESH1, cells could better preserve

NADPH, increase the pool of reduced glutathione, and thus better survive from ferroptosis. In addition to ferroptosis survival phenotype, MESH1 silencing also induces a transcriptional response that is partially mediated by ATF4 with functional similarity as bacterial stringent response, lead to dNTP depletion by transcriptional inhibition of

RRM1 and RRM2, and lead to a reversible cell growth inhibition in tumors. To further study the physiological relevance of MESH1, we establish a Mesh1 knockout mouse model. There is no major growth defect or body size difference among Mesh1 wild type

76

and knockout mice, and other phenotypes such as response to intracellular pathogens, isoproterenol challenge of cardiac function and starvation response requires further investigation. There are several questions unanswered, including other potential substrates of MESH1 and their physiological function, which requires future studies.

77

Appendix A: RNAi used in this study

Name Target Sequence Company Catalog # siNT (non- Qiagen SI03650318 targeting) siMESH1-CDS GGGAAUCACUGACAUUGUG Dharmacon D-031786-01 siMESH1-3’UTR CTGAAGGTCTCCTGCTAACTA Qiagen SI04167002 siMESH1-02 GACAAGCUGUACAAUCUGA Dharmacon D-031786-02 siMESH1-10 GGACAGGAUUCAUACGCCA Dharmacon J-031786-10 siMESH1-12 CCCAGAUAUUAGAGGCCAA Dharmacon J-031786-12 siMESH1-3’UTR- CCGGACAGGATTCATACGCCA Qiagen SI00642173 2 siNADK UGAAUGAGGUGGUGAUUGA Dharmacon M-006318-01 CGCCAGCGAUGAAAGCUUU GAAGACGGCGUGCACAAU CCAAUCAGAUAGACUUCAU siNADK2 GCAAUUGCUUCGAUGAUGA Dharmacon M-006319-01 GAGAAUUGGUAGAGAAAGU UGUGAAAGCUGGACGGAUA UUUGAUUACUGGCGAGAAU siATF4 GAUCAUUCCUUUAGUUUAG Dharmacon M-005125-02 CAUGAUCCCUCAGUGCAUA GUUUAGAGCUGGGCAGUGA CUAGGUACCGCCAGAAGAA Clone ID Tet-shMESH1-16 TGAGGTGGAGCTACACTTTGG TRCN0000243216 Tet-shMESH1-17 TGGTGGAGGAGGTAACAGATG TRCN0000243217 Tet-shMESH1-18 TCCATCCTTCCCAGATATTAG TRCN0000243218

78

Appendix B: primers used in this study

Name Sequence Purpose MESH1-E65A-F ccatgacacggtggcggacacagacacca Site-directed mutagenesis MESH1-E65A-R tggtgtctgtgtccgccaccgtgtcatgg for human MESH1 MESH1-D66A-F ggtggtgtctgtggcctccaccgtgtc MESH1-D66A-R gacacggtggaggccacagacaccacc MESH1-R24A-F tcggggtccttcgcccgctgctgccg MESH1-R24A-R cggcagcagcgggcgaaggaccccga mMESH1-D66N-F cagggtagtatctgtgttttctactgtgtcatggagc Site-directed mutagenesis mMESH1-D66N-R gctccatgacacagtagaaaacacagatactaccctg for mouse MESH1 51211BetaActinF GGGGTGTTGAAGGTCTCAAA Courtesy of G.LaMonte 51211BetaActinR GGCATCCTCACCCTGAAGTA MESH1-rt-1F GAGGCGGGAATCACTGACAT Target CDS of MESH1 MESH1-rt-1R TTGTGCCCCAAAGTGTAGCT MESH1-rt-5F CTCCCTCCATCCTTCCCAGA Target 3’UTR of MESH1 MESH1-rt-5R ACATCTGATTCCCACCACGC RRM2-rt-1F TGGGGACAAAGAGGCTACCT RRM2-rt-1R CCAGGCATCAGTCCTCGTTT CDK1-rt-1F CAGGTCAAGTGGTAGCCATGA CDK1-rt-1R ACCTGGAATCCTGCATAAGC NADK-rt-1F CACAATGGGCTGGGTGAGAA NADK-rt-1R TTGGACAGGTAGGAGGAGGG NADK2-rt-1F GCTCTACAGTCCGGAAGAACC NADK2-rt-1R GCATCCCAACAACGAGAACG ATF3-rt-1F GTCCATCACAAAAGCCGAGG ATF3-rt-1R GGCACTCCGTCTTCTCCTTC VEGFA-rt-1F CTTGCCTTGCTGCTCTACCT VEGFA-rt-1R AGCTGCGCTGATAGACATCC CTH-rt-1F CTCACTGTCCACCACGTTCA CTH-rt-1R GCCACTGCTTTTTCAAGGCA PPP1R12A-rt-1F ACCACTACATGCAGCAGCTT PPP1R12A-rt-1R GCCTCCTCCTCCGCAATATC mMESH1-rt-1F CTCACAAACACCGACAGCAG Mouse genotyping mMESH1-rt-1R GCCTCATGGGTTAGGATCCG LAR3 CACAACGGGTTCTTCTGTTAGTCC mActb-rt-1F GGCTGTATTCCCCTCCATCG rt-qPCR for mouse cells mActb-rt-1R CCAGTTGGTAACAATGCCATGT mMESH1-rt-4F CCGCTCACAAACACCGACA

79

mMESH1-rt-4R GCGGCCTGTAACACCACAA

80

Appendix C: antibodies used in this study

Gene Protein size Company Catalog # Note (kDa) MESH1 (HDDC3) 22 Abcam ab118325 Western Blot MESH1 (HDDC3) 22 Western Blot MESH1 (HDDC3) Sigma Immunohistochemistry GAPDH 37 Santa Cruz sc-25778 β-tubulin 54 CST 2128 Citrate Synthetase 52 GeneTex GTX110624 Histone H3 17 CST 4499 ATF4 (CREB-2) 37 Santa Cruz sc-200 anti-mouse-IgG CST 7076 Western Blot secondary HRP Anti-rabbit-IgG CST 7074 Western Blot secondary HRP

81

References

1. Srivatsan A, Wang JD. Control of bacterial transcription, translation and replication by (p)ppGpp. Curr Opin Microbiol. 2008 Apr;11(2):100-5. PubMed PMID: 18359660.

2. Potrykus K, Cashel M. (p)ppGpp: still magical? Annu Rev Microbiol. 2008;62:35-51. PubMed PMID: 18454629.

3. Kriel A, Bittner AN, Kim SH, Liu K, Tehranchi AK, Zou WY, et al. Direct regulation of GTP homeostasis by (p)ppGpp: a critical component of viability and stress resistance. Mol Cell. 2012 Oct 26;48(2):231-41. PubMed PMID: 22981860. Pubmed Central PMCID: PMC3483369.

4. Sun D, Lee G, Lee JH, Kim HY, Rhee HW, Park SY, et al. A metazoan ortholog of SpoT hydrolyzes ppGpp and functions in starvation responses. Nat Struct Mol Biol. 2010 Oct;17(10):1188-94. PubMed PMID: 20818390.

5. Hauryliuk V, Atkinson GC, Murakami KS, Tenson T, Gerdes K. Recent functional insights into the role of (p)ppGpp in bacterial physiology. Nature reviews Microbiology. 2015 May;13(5):298-309. PubMed PMID: 25853779. Pubmed Central PMCID: 4659695.

6. Zuo Y, Wang Y, Steitz TA. The mechanism of E. coli RNA polymerase regulation by ppGpp is suggested by the structure of their complex. Mol Cell. 2013 May 9;50(3):430-6. PubMed PMID: 23623685. Pubmed Central PMCID: 3677725.

7. Milon P, Tischenko E, Tomsic J, Caserta E, Folkers G, La Teana A, et al. The nucleotide-binding site of bacterial translation initiation factor 2 (IF2) as a metabolic sensor. Proc Natl Acad Sci U S A. 2006 Sep 19;103(38):13962-7. PubMed PMID: 16968770. Pubmed Central PMCID: 1599896.

8. Krasny L, Gourse RL. An alternative strategy for bacterial ribosome synthesis: Bacillus subtilis rRNA transcription regulation. The EMBO journal. 2004 Nov 10;23(22):4473-83. PubMed PMID: 15496987. Pubmed Central PMCID: 526457.

9. Geiger T, Wolz C. Intersection of the stringent response and the CodY regulon in low GC Gram-positive bacteria. International journal of medical microbiology : IJMM. 2014 Mar;304(2):150-5. PubMed PMID: 24462007.

10. Potrykus K, Murphy H, Philippe N, Cashel M. ppGpp is the major source of growth rate control in E. coli. Environmental microbiology. 2011 Mar;13(3):563-75. PubMed PMID: 20946586. Pubmed Central PMCID: 4556285. 82

11. Gaca AO, Kajfasz JK, Miller JH, Liu K, Wang JD, Abranches J, et al. Basal levels of (p)ppGpp in Enterococcus faecalis: the magic beyond the stringent response. mBio. 2013 Sep 24;4(5):e00646-13. PubMed PMID: 24065631. Pubmed Central PMCID: 3781836.

12. Scott M, Gunderson CW, Mateescu EM, Zhang Z, Hwa T. Interdependence of cell growth and gene expression: origins and consequences. Science. 2010 Nov 19;330(6007):1099-102. PubMed PMID: 21097934.

13. Gallant J, Irr J, Cashel M. The mechanism of amino acid control of guanylate and adenylate biosynthesis. J Biol Chem. 1971 Sep 25;246(18):5812-6. PubMed PMID: 4938039.

14. Hochstadt-Ozer J, Cashel M. The regulation of purine utilization in bacteria. V. Inhibition of purine phosphoribosyltransferase activities and purine uptake in isolated membrane vesicles by guanosine tetraphosphate. J Biol Chem. 1972 Nov 10;247(21):7067-72. PubMed PMID: 4343167.

15. Gentry DR, Hernandez VJ, Nguyen LH, Jensen DB, Cashel M. Synthesis of the stationary-phase sigma factor sigma s is positively regulated by ppGpp. Journal of bacteriology. 1993 Dec;175(24):7982-9. PubMed PMID: 8253685. Pubmed Central PMCID: 206978.

16. Colomer-Winter C, Gaca AO, Chuang-Smith ON, Lemos JA, Frank KL. Basal levels of (p)ppGpp differentially affect the pathogenesis of infective endocarditis in Enterococcus faecalis. Microbiology. 2018 Oct;164(10):1254-65. PubMed PMID: 30091695.

17. Geiger T, Francois P, Liebeke M, Fraunholz M, Goerke C, Krismer B, et al. The stringent response of Staphylococcus aureus and its impact on survival after phagocytosis through the induction of intracellular PSMs expression. PLoS pathogens. 2012;8(11):e1003016. PubMed PMID: 23209405. Pubmed Central PMCID: 3510239.

18. Geiger T, Goerke C, Fritz M, Schafer T, Ohlsen K, Liebeke M, et al. Role of the (p)ppGpp synthase RSH, a RelA/SpoT homolog, in stringent response and virulence of Staphylococcus aureus. Infection and immunity. 2010 May;78(5):1873-83. PubMed PMID: 20212088. Pubmed Central PMCID: 2863498.

19. Poole K. Bacterial stress responses as determinants of antimicrobial resistance. The Journal of antimicrobial chemotherapy. 2012 Sep;67(9):2069-89. PubMed PMID: 22618862. 83

20. Boniecka J, Prusinska J, Dabrowska GB, Goc A. Within and beyond the stringent response-RSH and (p)ppGpp in plants. Planta. 2017 Nov;246(5):817-42. PubMed PMID: 28948393. Pubmed Central PMCID: 5633626.

21. Atkinson GC, Tenson T, Hauryliuk V. The RelA/SpoT homolog (RSH) superfamily: distribution and functional evolution of ppGpp synthetases and hydrolases across the tree of life. PLoS One. 2011;6(8):e23479. PubMed PMID: 21858139. Pubmed Central PMCID: 3153485.

22. Chipman DM, Shaanan B. The ACT domain family. Current opinion in structural biology. 2001 Dec;11(6):694-700. PubMed PMID: 11751050.

23. Durfee T, Hansen AM, Zhi H, Blattner FR, Jin DJ. Transcription profiling of the stringent response in Escherichia coli. Journal of bacteriology. 2008 Feb;190(3):1084- 96. PubMed PMID: 18039766. Pubmed Central PMCID: 2223561.

24. Travers A. Modulation of RNA polymerase specificity by ppGpp. Molecular & general genetics : MGG. 1976 Aug 19;147(2):225-32. PubMed PMID: 787760.

25. Eymann C, Homuth G, Scharf C, Hecker M. Bacillus subtilis functional genomics: global characterization of the stringent response by proteome and transcriptome analysis. Journal of bacteriology. 2002 May;184(9):2500-20. PubMed PMID: 11948165. Pubmed Central PMCID: 134987.

26. Dixon SJ, Stockwell BR. The Hallmarks of Ferroptosis. Annual Review of Cancer Biology. 2019;3. Epub October 24, 2018.

27. White CC, Viernes H, Krejsa CM, Botta D, Kavanagh TJ. Fluorescence-based microtiter plate assay for glutamate-cysteine ligase activity. Anal Biochem. 2003 Jul 15;318(2):175-80. PubMed PMID: 12814619.

28. Dixon SJ, Patel DN, Welsch M, Skouta R, Lee ED, Hayano M, et al. Pharmacological inhibition of cystine-glutamate exchange induces endoplasmic reticulum stress and ferroptosis. Elife. 2014 May 20;3:e02523. PubMed PMID: 24844246. Pubmed Central PMCID: 4054777.

29. Pakos-Zebrucka K, Koryga I, Mnich K, Ljujic M, Samali A, Gorman AM. The integrated stress response. EMBO reports. 2016 Oct;17(10):1374-95. PubMed PMID: 27629041. Pubmed Central PMCID: 5048378.

30. Chen MS, Wang SF, Hsu CY, Yin PH, Yeh TS, Lee HC, et al. CHAC1 degradation of glutathione enhances cystine-starvation-induced necroptosis and ferroptosis in

84

human triple negative breast cancer cells via the GCN2-eIF2alpha-ATF4 pathway. Oncotarget. 2017 Dec 29;8(70):114588-602. PubMed PMID: 29383104. Pubmed Central PMCID: 5777716.

31. Shimada K, Hayano M, Pagano NC, Stockwell BR. Cell-Line Selectivity Improves the Predictive Power of Pharmacogenomic Analyses and Helps Identify NADPH as Biomarker for Ferroptosis Sensitivity. Cell chemical biology. 2016 Feb 18;23(2):225- 35. PubMed PMID: 26853626. Pubmed Central PMCID: 4792701.

32. Pollak N, Dolle C, Ziegler M. The power to reduce: pyridine nucleotides--small molecules with a multitude of functions. Biochem J. 2007 Mar 1;402(2):205-18. PubMed PMID: 17295611. Pubmed Central PMCID: PMC1798440.

33. Pollak N, Niere M, Ziegler M. NAD kinase levels control the NADPH concentration in human cells. J Biol Chem. 2007 Nov 16;282(46):33562-71. PubMed PMID: 17855339.

34. Ohashi K, Kawai S, Murata K. Identification and characterization of a human mitochondrial NAD kinase. Nat Commun. 2012;3:1248. PubMed PMID: 23212377. Pubmed Central PMCID: PMC3535332.

35. Navas P, Minnifield N, Sun I, Morré DJ. NADP phosphatase as a marker in free- flow electrophoretic separations for cisternae of the Golgi apparatus midregion. Biochim Biophys Acta. 1986;881(1):9.

36. Davila A, Liu L, Chellappa K, Redpath P, Nakamaru-Ogiso E, Paolella LM, et al. Nicotinamide adenine dinucleotide is transported into mammalian mitochondria. Elife. 2018 Jun 12;7. PubMed PMID: 29893687. Pubmed Central PMCID: PMC6013257.

37. Hogg T, Mechold U, Malke H, Cashel M, Hilgenfeld R. Conformational antagonism between opposing active sites in a bifunctional RelA/SpoT homolog modulates (p)ppGpp metabolism during the stringent response [corrected]. Cell. 2004 Apr 2;117(1):57-68. PubMed PMID: 15066282.

38. Baykov AA, Evtushenko OA, Avaeva SM. A malachite green procedure for orthophosphate determination and its use in alkaline phosphatase-based enzyme immunoassay. Anal Biochem. 1988 Jun;171(2):266-70. PubMed PMID: 3044186.

39. Petheo GL, Demaurex N. Voltage- and NADPH-dependence of electron currents generated by the phagocytic NADPH oxidase. Biochem J. 2005 Jun 1;388(Pt 2):485- 91. PubMed PMID: 15689187. Pubmed Central PMCID: PMC1138955. 85

40. Lu W, Wang L, Chen L, Hui S, Rabinowitz JD. Extraction and Quantitation of Nicotinamide Adenine Dinucleotide Redox Cofactors. Antioxid Redox Signal. 2018 Jan 20;28(3):167-79. PubMed PMID: 28497978. Pubmed Central PMCID: PMC5737638.

41. Ding C-KC, Rose J, Wu J, Sun T, Chen K-Y, Chen P-H, et al. Mammalian stringent- like response mediated by the cytosolic NADPH phosphatase MESH1. BioRxiv. 2018. Epub 5/17/2018.

42. Yang X, Boehm JS, Yang X, Salehi-Ashtiani K, Hao T, Shen Y, et al. A public genome-scale lentiviral expression library of human ORFs. Nat Methods. 2011 Jun 26;8(8):659-61. PubMed PMID: 21706014. Pubmed Central PMCID: PMC3234135.

43. Clayton DA, Shadel GS. Isolation of mitochondria from tissue culture cells. Cold Spring Harb Protoc. 2014 Oct 1;2014(10):pdb prot080002. PubMed PMID: 25275104.

44. Lewis CA, Parker SJ, Fiske BP, McCloskey D, Gui DY, Green CR, et al. Tracing compartmentalized NADPH metabolism in the cytosol and mitochondria of mammalian cells. Mol Cell. 2014 Jul 17;55(2):253-63. PubMed PMID: 24882210. Pubmed Central PMCID: 4106038.

45. Nazir A, Harinarayanan R. (p)ppGpp and the bacterial cell cycle. Journal of biosciences. 2016 Jun;41(2):277-82. PubMed PMID: 27240988.

46. Ashrafi K, Chang FY, Watts JL, Fraser AG, Kamath RS, Ahringer J, et al. Genome- wide RNAi analysis of Caenorhabditis elegans fat regulatory genes. Nature. 2003 Jan 16;421(6920):268-72. PubMed PMID: 12529643.

47. Fraser AG, Kamath RS, Zipperlen P, Martinez-Campos M, Sohrmann M, Ahringer J. Functional genomic analysis of C. elegans chromosome I by systematic RNA interference. Nature. 2000 Nov 16;408(6810):325-30. PubMed PMID: 11099033.

48. Sonnichsen B, Koski LB, Walsh A, Marschall P, Neumann B, Brehm M, et al. Full- genome RNAi profiling of early embryogenesis in Caenorhabditis elegans. Nature. 2005 Mar 24;434(7032):462-9. PubMed PMID: 15791247.

49. Rual JF, Ceron J, Koreth J, Hao T, Nicot AS, Hirozane-Kishikawa T, et al. Toward improving Caenorhabditis elegans phenome mapping with an ORFeome-based RNAi library. Genome research. 2004 Oct;14(10B):2162-8. PubMed PMID: 15489339. Pubmed Central PMCID: 528933.

86

50. Kettleborough RN, Busch-Nentwich EM, Harvey SA, Dooley CM, de Bruijn E, van Eeden F, et al. A systematic genome-wide analysis of zebrafish protein-coding gene function. Nature. 2013 Apr 25;496(7446):494-7. PubMed PMID: 23594742. Pubmed Central PMCID: 3743023.

51. Brown SDM, Holmes CC, Mallon AM, Meehan TF, Smedley D, Wells S. High- throughput mouse phenomics for characterizing mammalian gene function. Nature reviews Genetics. 2018 Jun;19(6):357-70. PubMed PMID: 29626206.

52. Wiederschain D, Wee S, Chen L, Loo A, Yang G, Huang A, et al. Single-vector inducible lentiviral RNAi system for oncology target validation. Cell Cycle. 2009 Feb 1;8(3):498-504. PubMed PMID: 19177017.

53. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome biology. 2013 Apr 25;14(4):R36. PubMed PMID: 23618408. Pubmed Central PMCID: 4053844.

54. Anders S, Pyl PT, Huber W. HTSeq--a Python framework to work with high- throughput sequencing data. Bioinformatics. 2015 Jan 15;31(2):166-9. PubMed PMID: 25260700. Pubmed Central PMCID: 4287950.

55. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome biology. 2014;15(12):550. PubMed PMID: 25516281. Pubmed Central PMCID: 4302049.

56. Liberzon A, Subramanian A, Pinchback R, Thorvaldsdottir H, Tamayo P, Mesirov JP. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011 Jun 15;27(12):1739-40. PubMed PMID: 21546393. Pubmed Central PMCID: 3106198.

57. Diamond TL, Roshal M, Jamburuthugoda VK, Reynolds HM, Merriam AR, Lee KY, et al. Macrophage tropism of HIV-1 depends on efficient cellular dNTP utilization by reverse transcriptase. J Biol Chem. 2004 Dec 3;279(49):51545-53. PubMed PMID: 15452123. Pubmed Central PMCID: 1351161.

58. Perrino C, Naga Prasad SV, Mao L, Noma T, Yan Z, Kim HS, et al. Intermittent pressure overload triggers hypertrophy-independent cardiac dysfunction and vascular rarefaction. The Journal of clinical investigation. 2006 Jun;116(6):1547-60. PubMed PMID: 16741575. Pubmed Central PMCID: 1464895.

59. Esposito G, Rapacciuolo A, Naga Prasad SV, Takaoka H, Thomas SA, Koch WJ, et al. Genetic alterations that inhibit in vivo pressure-overload hypertrophy prevent 87

cardiac dysfunction despite increased wall stress. Circulation. 2002 Jan 1;105(1):85- 92. PubMed PMID: 11772881.

60. Durkin ME, Qian X, Popescu NC, Lowy DR. Isolation of Mouse Embryo Fibroblasts. Bio-protocol. 2013 Sep 20;3(18). PubMed PMID: 27376106. Pubmed Central PMCID: 4928858.

61. Miao Y, Bist P, Wu J, Zhao Q, Li QJ, Wan Y, et al. Collaboration between Distinct Rab Small GTPase Trafficking Circuits Mediates Bacterial Clearance from the Bladder Epithelium. Cell host & microbe. 2017 Sep 13;22(3):330-42 e4. PubMed PMID: 28910634. Pubmed Central PMCID: 5659305.

62. Consortium EP. The ENCODE (ENCyclopedia Of DNA Elements) Project. Science. 2004 Oct 22;306(5696):636-40. PubMed PMID: 15499007.

63. Crawford RR, Prescott ET, Sylvester CF, Higdon AN, Shan J, Kilberg MS, et al. Human CHAC1 Protein Degrades Glutathione, and mRNA Induction Is Regulated by the Transcription Factors ATF4 and ATF3 and a Bipartite ATF/CRE Regulatory Element. J Biol Chem. 2015 Jun 19;290(25):15878-91. PubMed PMID: 25931127. Pubmed Central PMCID: 4505494.

64. Wells GN, Hageman RH. Specificity for nicotinamide adenine dinucleotide by nitrate reductase from leaves. Plant physiology. 1974 Aug;54(2):136-41. PubMed PMID: 16658848. Pubmed Central PMCID: 541519.

65. Gallais S, de Crescenzo MA, Laval-Martin DL. Evidence of active NADP(+) phosphatase in dormant seeds of Avena sativa L. Journal of experimental botany. 2000 Aug;51(349):1389-94. PubMed PMID: 10944152.

66. Reidl J, Schlor S, Kraiss A, Schmidt-Brauns J, Kemmer G, Soleva E. NADP and NAD utilization in Haemophilus influenzae. Molecular microbiology. 2000 Mar;35(6):1573-81. PubMed PMID: 10760156.

67. Kawai S, Murata K. Structure and function of NAD kinase and NADP phosphatase: key enzymes that regulate the intracellular balance of NAD(H) and NADP(H). Bioscience, biotechnology, and biochemistry. 2008 Apr;72(4):919-30. PubMed PMID: 18391451.

68. Richter C. NADP+ phosphatase: a novel mitochondrial enzyme. Biochemical and biophysical research communications. 1987 Jul 15;146(1):253-7. PubMed PMID: 3038107.

88

69. Ooga T, Ohashi Y, Kuramitsu S, Koyama Y, Tomita M, Soga T, et al. Degradation of ppGpp by nudix pyrophosphatase modulates the transition of growth phase in the bacterium Thermus thermophilus. J Biol Chem. 2009 Jun 5;284(23):15549-56. PubMed PMID: 19346251. Pubmed Central PMCID: 2708851.

70. Ito D, Kato T, Maruta T, Tamoi M, Yoshimura K, Shigeoka S. Enzymatic and molecular characterization of Arabidopsis ppGpp pyrophosphohydrolase, AtNUDX26. Bioscience, biotechnology, and biochemistry. 2012;76(12):2236-41. PubMed PMID: 23221701.

71. Corpas FJ, Aguayo-Trinidad S, Ogawa T, Yoshimura K, Shigeoka S. Activation of NADPH-recycling systems in leaves and roots of Arabidopsis thaliana under arsenic-induced stress conditions is accelerated by knock-out of Nudix hydrolase 19 (AtNUDX19) gene. Journal of plant physiology. 2016 Mar 15;192:81-9. PubMed PMID: 26878367.

72. Fishilevich S, Nudel R, Rappaport N, Hadar R, Plaschkes I, Iny Stein T, et al. GeneHancer: genome-wide integration of enhancers and target genes in GeneCards. Database : the journal of biological databases and curation. 2017 Jan 1;2017. PubMed PMID: 28605766. Pubmed Central PMCID: 5467550.

73. Nathan C, Ding A. SnapShot: Reactive Oxygen Intermediates (ROI). Cell. 2010 Mar 19;140(6):951- e2. PubMed PMID: 20303882.

74. Vvedenskaya IO, Bird JG, Zhang Y, Zhang Y, Jiao X, Barvik I, et al. CapZyme-Seq Comprehensively Defines Promoter-Sequence Determinants for RNA 5' Capping with NAD(). Mol Cell. 2018 May 3;70(3):553-64 e9. PubMed PMID: 29681497. Pubmed Central PMCID: 5935523.

75. Cahova H, Winz ML, Hofer K, Nubel G, Jaschke A. NAD captureSeq indicates NAD as a bacterial cap for a subset of regulatory RNAs. Nature. 2015 Mar 19;519(7543):374-7. PubMed PMID: 25533955.

76. Walters RW, Matheny T, Mizoue LS, Rao BS, Muhlrad D, Parker R. Identification of NAD+ capped mRNAs in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A. 2017 Jan 17;114(3):480-5. PubMed PMID: 28031484. Pubmed Central PMCID: 5255579.

77. Jiao X, Doamekpor SK, Bird JG, Nickels BE, Tong L, Hart RP, et al. 5' End Nicotinamide Adenine Dinucleotide Cap in Human Cells Promotes RNA Decay through DXO-Mediated deNADding. Cell. 2017 Mar 9;168(6):1015-27 e10. PubMed PMID: 28283058. Pubmed Central PMCID: 5371429.

89

78. Winz ML, Cahova H, Nubel G, Frindert J, Hofer K, Jaschke A. Capture and sequencing of NAD-capped RNA sequences with NAD captureSeq. Nature protocols. 2017 Jan;12(1):122-49. PubMed PMID: 27977022.

79. Friedmann Angeli JP, Schneider M, Proneth B, Tyurina YY, Tyurin VA, Hammond VJ, et al. Inactivation of the ferroptosis regulator Gpx4 triggers acute renal failure in mice. Nat Cell Biol. 2014 Dec;16(12):1180-91. PubMed PMID: 25402683. Pubmed Central PMCID: 4894846.

80. Choi JH, Pile-Spellman J. Reperfusion Changes After Stroke and Practical Approaches for Neuroprotection. Neuroimaging clinics of North America. 2018 Nov;28(4):663-82. PubMed PMID: 30322601.

81. Syal K, Flentie K, Bhardwaj N, Maiti K, Jayaraman N, Stallings CL, et al. Synthetic (p)ppGpp Analogue Is an Inhibitor of Stringent Response in Mycobacteria. Antimicrobial agents and chemotherapy. 2017 Jun;61(6). PubMed PMID: 28396544. Pubmed Central PMCID: 5444170.

82. Wexselblatt E, Kaspy I, Glaser G, Katzhendler J, Yavin E. Design, synthesis and structure-activity relationship of novel Relacin analogs as inhibitors of Rel proteins. European journal of medicinal chemistry. 2013;70:497-504. PubMed PMID: 24189495.

83. Wexselblatt E, Oppenheimer-Shaanan Y, Kaspy I, London N, Schueler-Furman O, Yavin E, et al. Relacin, a novel antibacterial agent targeting the Stringent Response. PLoS pathogens. 2012 Sep;8(9):e1002925. PubMed PMID: 23028324. Pubmed Central PMCID: 3447753.

84. Mekada K, Abe K, Murakami A, Nakamura S, Nakata H, Moriwaki K, et al. Genetic differences among C57BL/6 substrains. Experimental animals. 2009 Apr;58(2):141-9. PubMed PMID: 19448337.

85. Wolf S, Hainz N, Beckmann A, Maack C, Menger MD, Tschernig T, et al. Brain damage resulting from postnatal hypoxic-ischemic brain injury is reduced in C57BL/6J mice as compared to C57BL/6N mice. Brain research. 2016 Nov 1;1650:224- 31. PubMed PMID: 27620649.

86. Leskov I, Neville A, Shen X, Pardue S, Kevil CG, Granger DN, et al. Nicotinamide nucleotide transhydrogenase activity impacts mitochondrial redox balance and the development of hypertension in mice. Journal of the American Society of Hypertension : JASH. 2017 Feb;11(2):110-21. PubMed PMID: 28087333.

90

87. Vozenilek AE, Vetkoetter M, Green JM, Shen X, Traylor JG, Klein RL, et al. Absence of Nicotinamide Nucleotide Transhydrogenase in C57BL/6J Mice Exacerbates Experimental Atherosclerosis. Journal of vascular research. 2018;55(2):98-110. PubMed PMID: 29455203.

88. Prusa J, Zhu DX, Stallings CL. The stringent response and Mycobacterium tuberculosis pathogenesis. Pathogens and disease. 2018 Jul 1;76(5). PubMed PMID: 29947752.

91

Biography

Chien-Kuang attended National Yang-Ming University, Taiwan for a combined program of undergraduate and medical education, and graduated as a Doctor of

Medicine in June, 2012. Upon graduation, Chien-Kuang enrolled into University

Program in Genetics and Genomics at Duke University and joined Chi lab in 2013.

During her time at Duke, Chien-Kuang authored and co-authored a number of research manuscripts and reviews:

WH Yang, Chien-Kuang C. Ding, et al. (2018) “The Hippo pathway effector TAZ regulates ferroptosis in renal cell carcinoma.” Manuscript in preparation. PH Chen, Chien-Kuang C. Ding, et al. (2018). “Kinome screen of ferroptosis reveals a novel role of ATM in regulating iron metabolisms.” Under revision in Cell Death and Differentiation. Chien-Kuang C. Ding*, J Rose*, et al.(2018). “Mammalian stringent-like response mediated by the cytosolic NADPH phosphatase MESH1.” under revision in Nature. Preprint available in BioRxiv, https://doi.org/10.1101/325266 *Contributed equally X Tang, Chien-Kuang Ding, et al. (2017). “Cystine addiction of triple-negative breast cancer associated with EMT augmented death signaling.” Oncogene, 36, 4235-4242 X Tang, J Wu, Chien-Kuang Ding, et al. (2016). “Cystine Deprivation Triggers Programmed Necrosis in VHL-Deficient Renal Cell Carcinomas.” Cancer Research, 76:1892-1903 EL LaGory, C Wu, CM Taniguchi, Chien-Kuang C. Ding, et al. (2015). “Suppression of PGC-1α is critical for reprogramming oxidative metabolism in renal cell carcinoma” Cell Report, 12(1), 116-127 MM Keenan*, Chien-Kuang C. Ding*, and JT Chi (2014). “An Unexpected Alliance between Stress Responses to Drive Oncogenesis” Breast Cancer Research 16:471 *Contributed equally CA Markunas, E Lock, K Soldano, H Cope, Chien-Kuang C. Ding, et al. (2014). “Identification of Chiari Type I Malformation subtypes using whole genome expression profiles and cranial base morphometrics.” BMC medical genomics 7(1):39

92

G LaMonte, X Tang, JLY Chen, J Wu, Chien-Kuang C. Ding, et al (2013). “Acidosis induces reprogramming of cellular metabolism to mitigate oxidative stress.” Cancer & Metabolism 1(1):23.

While in graduate school, Chien-Kuang received Chancellor’s Scholar award, which awarded to outstanding biomedical graduate students and provided an one-time stipend on top of full financial support for the first two years of graduate study. She completed her PhD research in 2018, December. Chien-Kuang is applying for residency training in Pathology and hopes to continue to contribute the biomedical research field as a clinical scientist in her future.

93