Mycn Stimulates Glycine Decarboxylase (Gldc)

MYCN STIMULATES GLYCINE DECARBOXYLASE (GLDC)

IN NEUROBLASTOMA TO PROMOTE TUMORIGENESIS

Ahmet Alptekin

Submitted to the Faculty of the Graduate School

of Augusta University in partial fulfillment

of the Requirements of the Degree of

Doctor of Philosophy in

Biochemistry and Cancer Biology

May

2019

ACKNOWLEDGMENTS

First and foremost, I would like to acknowledge my PI and mentor Dr. Han-Fei

Ding. He taught me how to be an experimental scientist. While he was mentoring me, he always listened to my ideas despite his extensive knowledge in the area and always gave me the chance and freedom to explore in order to be independent. On the other hand, when needed, his intervention was in a way that did not break my feeling of independence. His mentoring method will be one of the important lessons of my Ph.D. education and I hope to be a good mentor in the future like him. I thank the members of my committee; Drs

Satyanarayana Ande, Zheng Dong, Hasan Korkaya and Honglin Li for providing critiques and valuable advice that shaped my project. I am grateful to Dr. Jan van Riggelen for reading and reviewing my dissertation. I am also thankful to all the people who were involved in my doctoral education in the Graduate School, Biochemistry and Cancer

Biology Department, and the Georgia Cancer Center of the Augusta University.

I am thankful for the past and present members of the Ding Laboratory: Jane Ding,

Bingwei Ye, and Yajie Yu. I also acknowledge my classmates in the departments of

Biochemistry and Cancer Biology, my “peer teachers”, they were very helpful in sharing their knowledge, experiment protocols, materials and their precious time. I thank the

Cancer Biology program director Dr. Balakrishna Lokeshwar. He was not only a program director but was also a friend to cherish and encouraged students, a thing which we need sometimes the most. I also thank Mrs. Kimberly Lord for patiently and smoothly handling all the documents to keep us on track.

I am grateful to my parents, who always prioritized their children and always supported them beyond their power. I want to acknowledge all my teachers from elementary school to graduate school. Lastly, I want to acknowledge my wife Gamze. She made the most sacrifice and always extended her support. As of end word, I am grateful to

Allah (God) for being created and being amazed by his magnificent creations that we do research on.

ABSTRACT

AHMET ALPTEKIN MYCN stimulates glycine decarboxylase (GLDC) in neuroblastoma to promote tumorigenesis Under the direction of DR. HAN-FEI DING

Genomic amplification of the oncogene MYCN is a major driver in the development of high-risk neuroblastoma, a pediatric cancer with poor prognosis. Given the challenge in targeting MYCN directly for therapy, we sought to identify MYCN-dependent metabolic vulnerabilities that can be targeted therapeutically. Here, we report that the gene encoding glycine decarboxylase (GLDC), which catalyzes the first and rate-limiting step in glycine breakdown with the production of one-carbon unit 5,10-methylenetetrahydrofolate, is a direct transcriptional target of MYCN. GLDC expression is markedly elevated in MYCN-amplified neuroblastoma tumors and cell lines. This transcriptional upregulation of GLDC expression is of functional significance, as GLDC depletion by

RNA interference inhibits the proliferation and tumorigenicity of MYCN-amplified neuroblastoma cell lines. Metabolomic profiling reveals that GLDC knockdown disrupts purine and central carbon metabolism and reduces citrate production, leading to a decrease in the steady-state levels of cholesterol and fatty acids, which are essential to sustain cell proliferation. In addition, microarray gene expression profiling shows that GLDC silencing downregulates genes involved in regulation of cell proliferation. These findings suggest that GLDC is a key player in MYCN-regulated cancer metabolism and a potential drug target in the treatment of high-risk neuroblastoma.

KEYWORDS: Neuroblastoma, glycine cleavage, cancer metabolism, glycine decarboxylase, GLDC, one-carbon metabolism, serine metabolism.

Table of Contents

Chapter Page

I. INTRODUCTION ...... 6

A. Statement of Problem ...... 6

B. Review of the Literature ...... 9

1. Cancer Metabolism ...... 9

2. Serine-Glycine-One-Carbon Metabolism ...... 11

2.1. Utilization of serine in cancer ...... 14

2.2. SGOC metabolism supports cell growth in cancer ...... 14

2.3. SGOC metabolism supports redox potential ...... 15

2.4. Targeting the SGOC metabolic pathway in cancer treatment 16

2.5. Glycine is an end-product of the SGOC metabolic pathway that

needs to be eliminated ...... 17

3. Glycine Cleavage System ...... 18

3.1. Glycine Cleavage System in normal physiology ...... 18

3.2. Functions of GLDC in cancer metabolism ...... 21

II. MATERIALS AND METHODS ...... 23

1. Ethics statement ...... 23

2. Cell lines and maintenance ...... 23

3. Patient data analysis ...... 24

4. Overexpression and RNA interference of target genes...... 24

5. Quantitative reverse-transcription PCR (qRT-PCR) ...... 25

6. Immunoblotting...... 28

7. Chromatin Immunoprecipitation (ChIP) ...... 28

8. Microarray...... 29

9. Cell cycle ...... 29

10. Xenograft assay ...... 30

11. Metabolomics ...... 30

12. Statistical analysis ...... 31

III. RESULTS ...... 32

1. GLDC expression correlates with poor prognosis in neuroblastoma .... 32

2. GLDC overexpression in neuroblastoma cell lines has no significant

effect on cell proliferation ...... 34

3. Using shRNA to downregulate GLDC expression in neuroblastoma cell

lines ...... 37

4. GLDC knockdown inhibits cell proliferation in cell culture ...... 39

5. GLDC knockdown inhibits the tumorigenesis of SMS-KCNR in

xenograft assay...... 42

6. Rescue experiments in GLDC knockdown cells ...... 44

7. GLDC knockdown broadly impacts cellular metabolism ...... 48

8. GLDC knockdown represses the expression of cell cycle genes ...... 55

9. GLDC knockdown induces cell cycle arrest ...... 57

10. MYCN is a transcriptional activator of GLDC expression in

neuroblastoma ...... 60

11. MYC activates GLDC expression ...... 64

12. Transcriptional Regulation of GLDC expression ...... 66

13. GLDC expression is independent of glycine levels ...... 68

14. Excess glycine is toxic to neuroblastoma cells ...... 69

IV. DISCUSSION ...... 71

Metabolic function of GLDC in neuroblastoma ...... 71

Rescue experiments of cell proliferation in GLDC knockdown ...... 73

Therapeutic potential of targeting the SGOC metabolic pathway and GLDC

...... 77

V. SUMMARY ...... 79

VI. REFERENCES ...... 81

APPENDIX A: ABBREVIATIONS...... 89

List of Tables

Table 1. shGLDC and shMYCN sequences...... 26

Table 2. Specific primers used in qRT-PCR...... 27

Table 3: Metabolites used in rescue experiments of impaired SGOC metabolism...... 46

List of Figures

Figure 1. SGOC metabolic pathway...... 13

Figure 2. Glycine cleavage system...... 20

Figure 3. SGOC metabolic pathway and glycine cleavage system...... 22

Figure 4. GLDC expression correlates with poor prognosis in neuroblastoma...... 33

Figure 5. GLDC and MYCN protein expression in neuroblastoma cell lines...... 35

Figure 6. GLDC overexpression does not induce proliferation in neuroblastoma...... 36

Figure 7. GLDC knockdown system...... 38

Figure 8. GLDC knockdown impairs cell growth in neuroblastoma...... 40

Figure 9. GLDC is required in neuroblastoma cells...... 41

Figure 10. GLDC knockdown impairs tumorigenesis in neuroblastoma xenograft...... 43

Figure 11. Rescue experiments of GLDC knockdown...... 47

Figure 12. GLDC regulates amino acid and purine metabolism...... 50

Figure 13. GLDC regulates central carbon metabolism...... 53

Figure 14. GLDC knockdown reduces the expression of cell cycle genes...... 56

Figure 15. GLDC knockdown induces G1 phase arrest...... 59

Figure 16. GLDC is highly expressed in MYCN-amplified neuroblastoma...... 61

Figure 17. GLDC is a direct transcriptional target of MYCN...... 63

Figure 18. MYC upregulates GLDC expression...... 65

Figure 19. KLF15 modestly regulates GLDC expression...... 67

Figure 20. Excess glycine is toxic to neuroblastoma cells...... 70

I. INTRODUCTION

A. Statement of Problem

Neuroblastoma is a developmental tumor of the sympathetic nervous system in

children that originates from the neural crest and is found in the sympathetic ganglia and

adrenal glands (Brodeur, 2003). The median age for diagnosis is 18 months and it is the

leading cause of cancer-related deaths in ages one-to-five (Louis & Shohet, 2015; Maris,

Hogarty, Bagatell, & Cohn, 2007). For prognostic purposes and treatment stratification, it

has been categorized into 4 risk groups (very low, low, intermediate and high-risk groups)

based on clinical (age, stage, histologic category, grade of tumor differentiation) and

genetic (MYCN status, DNA ploidy, chromosomal aberrations) features (Cohn et al.,

2009). In selected very-low-risk cases, only observation could be sufficient. Low and

intermediate risk groups have a good response to surgery, chemotherapy, and radiotherapy.

These non-high-risk groups comprise almost two-thirds of disease, with more than 90%

overall survival rates (Whittle et al., 2017). However, treatment of high-risk groups is

challenging and their cure rate is low. The treatment of high-risk disease consists of

chemotherapy, surgery, myeloablative therapy, and autologous stem cell transplantation,

radiotherapy and recently included retinoic acid treatment and immunotherapy. Most high-

risk group patients receive all these treatments and despite the increase in the survival with

recent immunotherapy and biological agents, the overall survival rates are still around 50%

(Pinto et al., 2015). Due to inadequate therapy response to current treatment, it is of

paramount importance to better understand the molecular features of high-risk disease to

reveal potential drug targets.

One of the earliest genetic alterations reported in neuroblastoma was the minute chromatin bodies (Cox, Yuncken, & Spriggs, 1965), which indicates gene amplification.

In 1976, Biedler et al. showed homogenously stained regions in neuroblastoma cell lines, which is another indicator of gene amplification (Biedler & Spengler, 1976). Schwab et al. hypothesized that this amplified gene could be homologous to transforming genes of RNA viruses, which have been renamed oncogenes. Using sequences homologous to 14 retroviral transforming genes, they found a partial homology, a fragment of DNA hybridized to v-myc (viral homolog of c-myc) in the Southern blot of neuroblastoma cells, but not in normal fibroblasts (Schwab et al., 1983). Further studies confirmed an amplified gene that is homologous to c-myc and v-myc oncogenes, indicating the existence of a novel gene that was named as N-myc (MYCN) (Kohl et al., 1983). Subsequently, MYCN amplification in tumor samples was found to be highly correlated with advanced stages of neuroblastoma (Schwab et al., 1984).

MYCN, along with c-MYC and MYCL, is a member of the MYC oncogene family of transcription factors. MYCN dimerizes with MAX, binds to E-Box motifs of gene promoters and regulates gene expression (Murphy et al., 2009; Ruiz-Perez, Henley, &

Arsenian-Henriksson, 2017). Physiologically, MYCN is transiently expressed in neural crest of avian embryos and regulates migration and differentiation of neural crest cells

(Wakamatsu, Watanabe, Nakamura, & Kondoh, 1997). In mouse, it is required in embryonic development, and homozygous mutation of MYCN is embryonically lethal

(Stanton, Perkins, Tessarollo, Sassoon, & Parada, 1992). On the other hand, forced MYCN expression is found to be tumorigenic. Targeted expression of MYCN in the sympathetic nervous system using the rat tyrosine hydroxylase promoter generated neuroblastoma in

7 the mouse, demonstrated the tumor-initiating capability of MYCN amplification in neuroblastoma (Weiss, Aldape, Mohapatra, Feuerstein, & Bishop, 1997).

MYCN amplification denotes having at least 5 copies of MYCN gene (Ambros et al., 2009) but usually indicates more than 100 copies of the gene (Kohl et al., 1983). It is the most important genetic indicator of poor prognosis in neuroblastoma. While patients with MYCN non-amplified tumors have 82% five-year survival rates, patients with

MYCN-amplified have 34% five-year survival (Cohn et al., 2009). Inhibiting MYCN would significantly improve survival rates in neuroblastoma. It has been shown that introducing MYCN-antisense oligo DNA into the MYCN transgenic mice significantly reduced tumor formation (Burkhart et al., 2003), indicating that targeting MYCN would be a viable treatment strategy in neuroblastoma. Unfortunately, this could not be translated into the clinic because of the technical challenge of designing inhibitors against MYCN due to the absence of a potential binding site of MYCN.

Different approaches to target MYCN are under investigation, e.g., targeting

MYCN gene transcription, protein stability, dimerization, and transcriptional activity and immune therapies against MYCN (Ruiz-Perez et al., 2017). An alternative strategy is to inhibit MYCN-driven tumorigenesis by targeting MYCN-target genes. MYCN induces gene expression and activates biosynthetic pathways in order to meet the metabolic demand for cell proliferation. Identifying and targeting these MYCN-target genes could be promising for neuroblastoma treatment. This approach has been proven to be effective.

Hogarty et al. previously found that Ornithine Decarboxylase (ODC) was an MYCN target gene and silencing it in transgenic mouse model reduced tumor growth (Hogarty et al.,

2008). Follow-up studies on these findings identified difluoromethylornithine (DFMO) as

an ODC inhibitor, and the recent clinical trial showed that DFMO has increased the

survival in high-risk neuroblastoma patients (Sholler et al., 2018).

In search of novel MYCN-target genes that are important in neuroblastoma

metabolism, we performed cluster analysis of previously published MYCN-amplification

gene signature consisting of upregulated genes in neuroblastoma tumors (Puissant et al.,

2013). The analysis revealed that glycine decarboxylase (GLDC) and other Serine-

Glycine-One-Carbon (SGOC) pathway genes were among the MYCN target genes.

Previous studies reported that the SGOC metabolic pathway is important for cancer

metabolism (Yang & Vousden, 2016) and GLDC has a supporting role in SGOC

metabolism in glioblastoma (Kim et al., 2015). Furthermore, it has been reported that

GLDC has a key role in nucleotide synthesis during embryogenesis in mice (Pai et al.,

2015) and in the metabolism of tumor-initiating cells in Non-small cell lung cancer

(NSCLC) (W. C. Zhang et al., 2012). These findings suggest that GLDC could be an

important target of MYCN in driving high-risk neuroblastoma.

This thesis aims to investigate whether GLDC is required for the neuroblastoma

tumor growth, its function in tumor metabolism and whether it is regulated by MYCN.

B. Review of the Literature

1. Cancer Metabolism

Metabolism is the sum of chemical processes that are conducted by the cell to

maintain life. Main functions of metabolism are nutrient utilization for energy production

and building block generation, and elimination of the waste products. Cancer cells have

9 increased demand for nutrients, mainly glucose and glutamine, to meet the demand for building blocks and energy that are needed in cell proliferation (Boroughs & DeBerardinis,

2015). As observed by Otto Warburg about a century ago and studied intensively in the last decade, tumors have increased glucose consumption and lactate release, even under normal oxygen conditions (Otto, 2016). With this aerobic glycolysis, cancer cells benefit from ATP synthesis and generate substrates to meet the increased demands for the biosynthesis reactions of the pentose phosphate pathway and de-novo serine synthesis

(Liberti & Locasale, 2016).

The altered metabolism of cancer cells creates an opportunity to exploit these altered pathways in cancer diagnosis and treatment. For instance, using a radionuclide analog of glucose, fluorodeoxyglucose, we can detect the high-glucose uptake areas in the body, which is useful in detecting metastasis or evaluating the treatment response.

Moreover, it could be exploited in cancer treatment. By taking advantage of the dependence of cancer cells on folate, Sidney Farber used antifolate agent aminopterin for the first time in cancer treatment that resulted in improved survival of patients (Farber & Diamond,

1948).

Today, with the development of techniques to study metabolism, we are able to better characterize the landscape of tumor metabolism, identify the alterations of cancer metabolism from the normal, and use this knowledge in developing novel treatments. With the recent advancement of metabolomics techniques, there has been an increased interest in cancer metabolism. Activities of different metabolic pathways in cancer have been examined with the help of the metabolomics. And among cellular metabolic pathways,

Serine-glycine-one-carbon (SGOC) metabolic pathway was found to be significantly

altered in most of the cancers.

2. Serine-Glycine-One-Carbon Metabolism

The SGOC metabolic pathway consists of two associated metabolic pathway, de-

novo serine synthesis and folate-mediated one-carbon metabolism. De-novo serine

synthesis describes the synthesis of serine in the cytoplasm from 3-Phosphoglycerate (3-

PG), an intermediate metabolite in the glycolysis pathway (Fig. 1a). In one carbon

metabolism, serine donates a carbon to tetrahydrofolate (THF) to be utilized in purine and

thymidine synthesis, and methylation reactions. Of note, methylation reactions also

indicate a transfer of carbon groups but they are independent of folate. Our study focused

on folate-mediated transfer of one-carbon units, henceforth named as one-carbon

metabolism. One-carbon metabolism can take place in the cytoplasm or in the

mitochondria, depending on the isoenzymes used (Fig 1b). Because of its role in nucleotide

synthesis, the SGOC metabolic pathway becomes vital for cell proliferation.

Figure 1. SGOC metabolic pathway.

A De-novo serine synthesis pathway. 3-PG, 3-phosphoglycerate; NAD+, nicotinamide adenine dinucleotide (oxidized); NADH, nicotinamide adenine dinucleotide (reduced); 3-PHP, 3-phosphohydroxypyruvate; 3PS, 3-Phosphoserine; Glu, glutamate; a-KG, alpha-ketoglutarate; PHGDH, phosphoglycerate dehydrogenase; PSAT1, phosphoserine aminotransferase; PSPH, phosphoserine phosphatase. B (folate-mediated) one- carbon metabolic pathway. THF, tetrahydrofolate; 5,10-CH2-THF, methylene THF; CH-THF, methenyl THF; 10-CHO-THF, formyl THF; NADP+, nicotinamide adenine dinucleotide phosphate (oxidized); NADPH, nicotinamide adenine dinucleotide phosphate (reduced); SHMT1, serine hydroxymethyltransferase 1 (cytoplasmic); SHMT2, serine hydroxymethyltransferase 2 (mitochondrial); MTHFD2, methylenetetrahydrofolate dehydrogenase 2 (mitochondrial); MTHFD2L, methylenetetrahydrofolate dehydrogenase 2 like; MTHFD1L, methylenetetrahydrofolate dehydrogenase 1 like; MTHFD1, methylenetetrahydrofolate Dehydrogenase.

2.1. Utilization of serine in cancer

The earliest observation of the utilization of serine in cancer was reported by Dr.

Keith Snell and his colleagues in 1980s. It was reported that the activity of the first enzyme

in the de-novo synthesis of serine pathway, Phosphoglycerate Dehydrogenase (PHGDH),

was increased in rat hepatomas (Snell & Weber, 1986) and human colon carcinoma (Snell,

Natsumeda, Eble, Glover, & Weber, 1988). Furthermore, this team reported that the

activity of PHGDH was in parallel with the incorporation of radioactive serine into

nucleotides (Snell, Natsumeda, & Weber, 1987), indicating utilization of serine in

nucleotide synthesis. Two decades after these initial observations, scientists were able to

conduct mechanistic studies to reveal its function in cancer using high throughput

technologies.

2.2. SGOC metabolism supports cell growth in cancer

In an effort to identify essential genes for the tumorigenesis of breast cancer,

Possemato et al. identified 133 metabolic genes that are differentially expressed in

aggressive breast cancer and found that the expression of these genes is associated with

stemness. Targeting these genes using shRNA revealed that PHGDH, the first and rate-

limiting step in the SGOC pathway, is essential for tumorigenesis in MDA-MB-468 cells

(Possemato et al., 2011). About the same time, Locasale et al. reported that PHGDH is

amplified in a number of tumors. They found that PHGDH diverts the metabolite flux from

glucose into the serine synthesis by metabolizing significant amount of 3-PG in the

glycolysis (Locasale et al., 2011).

These reports that revealed the novel and significant role of PHGDH in cancer

metabolism were followed by other studies that focused on the other members of SGOC

metabolic pathway. Kim et al. found that the mitochondrial serine

hydroxymethyltransferase (SHMT2), which transfers a one-carbon unit of serine to THF,

is required for nucleotide synthesis in glioblastoma multiforme and is indispensable for

cell proliferation (Kim et al., 2015). Another comprehensive study utilized 2000 tumor

samples from 19 different cancer types and reported that genes of the mitochondrial one-

carbon metabolic pathway are the most markedly elevated among all metabolic genes

(Nilsson et al., 2014). These findings confirmed by Ducker et al. in a comprehensive

mechanistic study. In this study, they found that mitochondrial one-carbon pathway is

consistently overexpressed in cancer and is required for nucleotide synthesis, and targeting

SGOC metabolic pathway can reduce tumor growth (Ducker et al., 2016).

2.3. SGOC metabolism supports redox potential

In addition to providing building blocks for biosynthetic reactions, SGOC

metabolism also plays a critical role in regulating the redox potential of the cell. Reactive

Oxygen Species (ROS) are generated in normal metabolic processes, particularly in the

oxidative phosphorylation reactions of the mitochondria. While excess ROS is toxic and

could lead to apoptosis, antioxidant systems in the cells reduce the ROS levels to

physiological levels to avoid the toxicity and apoptosis (Circu & Aw, 2010). The main

antioxidant in the cell is the reduced glutathione (GSH), which is regenerated by

nicotinamide adenine dinucleotide phosphate (NADPH).

Fan et al. showed that one-carbon metabolism generates a significant amount of

NADPH in proliferating cells, which is comparable to that generated by the pentose

phosphate pathway, the main NADPH provider of the cell (Fan et al., 2014). In accordance

with these findings, PHGDH knockdown decreased the antioxidant capacity (reduced

glutathione (GSH)/oxidized glutathione (GSSG) ratio) of the MDA-MB-231 and MCF7

cell lines, and significantly decreased the mammosphere forming capacity of MDA-231

cells (Samanta et al., 2016). SHMT2 knockdown in Kelly neuroblastoma cells increased

ROS levels and cell death upon hypoxia, which was rescued by introduction of N-Acetyl-

L-Cysteine, (NAC, a precursor of glutathione) (Ye et al., 2014). These findings further

confirm the role of SGOC in the antioxidant system, which is vital for the cell to maintain

the ROS in physiologic levels and avoid ROS-induced toxicity.

2.4. Targeting the SGOC metabolic pathway in cancer treatment

In 1947, Sidney Farber and colleagues tested a folic acid conjugate on various final

stage cancer patients based on previous observations that reported the anti-tumorigenic

features of the compound. They did not see a significant improvement in survival (Farber

et al., 1947) but they observed an acceleration in the leukemic process of leukemia patients.

These findings and previous observations on the effect of folic acid deficiency in anemia

led them to test a folic acid antagonist, Aminopterin, for treatment of acute leukemia.

Aminopterin was able to prolong the survival of patients and this was the first successful

treatment of cancer with anti-metabolite (Farber & Diamond, 1948). Another closely

related antifolate, methotrexate, replaced Aminopterin with better therapeutic index and is

currently in clinical use for many types of cancers. These were followed by fluorouracil, a

pyrimidine analog anti-metabolite, which also targets the SGOC metabolic pathway

(Curreri, Ansfield, Mc, Waisman, & Heidelberger, 1958). Recently, a new folic acid

antagonist, Pemetrexed, has been developed and is currently in use for treatment of various

cancers (Jackman & Calvert, 1995).

With the renewed interest and recent findings on SGOC metabolism in cancer,

small molecule inhibitors have been developed to target different members of this pathway.

Pacold et al. discovered a small molecule inhibitor targeting PHGDH and showed that it

reduced MDA-MB-468 cells xenograft tumor growth (Pacold et al., 2016). Another key

member of this pathway, SHMT, was targeted with small molecule inhibitors, which were

found to be effective in inhibiting the growth of diffuse large B-cell lymphoma cell lines.

This inhibition of cell growth could be reversed by supplementation of glycine and formate

(a one-carbon unit), which are generated by SHMT (Ducker et al., 2017). These findings

validate the vital function of the SGOC pathway in sustaining cancer cell growth and the

strategy in targeting this pathway for cancer therapy.

Dietary modifications that aim to reduce SGOC activity have also been shown to

be successful in pre-clinical models. Maddocks et al. showed that serine and glycine free

diet reduced tumor growth and increased animal survival in xenograft (O. D. Maddocks et

al., 2013) and transgenic mouse models (O. D. K. Maddocks et al., 2017).

2.5. Glycine is an end-product of the SGOC metabolic pathway that needs to be

eliminated

SHMT in the SGOC metabolic pathway transfers a one-carbon group of serine to

THF for its utilization in one-carbon metabolism accompanied by glycine production (Fig.

1b). Unlike serine, which is generated and consumed in the SGOC pathway, an increase in

the glycine level was found to be toxic in HCT116, RKO, SW480 and A549 cancer cell

lines (Labuschagne, van den Broek, Mackay, Vousden, & Maddocks, 2014). It has been

hypothesized that excess glycine could reverse the reaction in favor of serine generation

(5,10-CH2-THF + Glycine = THF + Serine), thereby reducing the levels of one-carbon

units (5,10-CH2-THF) needed for nucleotide synthesis. A different mechanism of glycine

toxicity was demonstrated by Kim et al. They showed that the increased activity of SHMT

in LN229 glioblastoma multiforme cells generates excess glycine, which is then converted

to the toxic metabolites aminoacetone and methylglyoxal, leading to the inhibition of cell

growth (Kim et al., 2015). In summary, an excess amount of glycine is generated when the

SGOC metabolic pathway is activated. Thus, the flux of the pathway depends on the

elimination of excess glycine. The only biochemical reaction that eliminates glycine in the

cell is catalyzed by the glycine cleavage system in the mitochondria.

3. Glycine Cleavage System

3.1. Glycine Cleavage System in normal physiology

Glycine is metabolized in the mitochondria by the glycine cleavage system (GCS),

which generates CO2 and NH3 while donating one-carbon unit to THF to produce 5,10-

+ CH2-THF) and converting NAD to NADH to increase the redox potential. GCS consists

of four enzymes; P-protein (glycine decarboxylase), T-protein (aminomethyltransferase),

L-protein (dihydrolipoamide dehydrogenase) and H-protein (carrier protein). Glycine

decarboxylase (GLDC) catalyzes the first and rate-limiting step of the glycine cleavage

reaction (Kikuchi, Motokawa, Yoshida, & Hiraga, 2008) (Fig. 2). Mutations in GCS (86%

of mutations occur in GLDC) impair the glycine metabolism and leads to glycine

accumulation in the central nervous system and non-ketotic hyperglycinemia (NKH). Its

toxicity in the central nervous system can be explained by the hypothesis that glycine

18 accumulation causes neurotoxicity through its receptor N-methyl-D-aspartate (NMDA).

Kure et al. treated NKH patients with a NMDA antagonist, and this treatment ameliorated the symptoms (Kure, Tada, & Narisawa, 1997). These findings suggest that GCS has a key role in regulation of glycine levels in the central nervous system.

Intriguingly, GLDC mutations were also found in some neural tube defect (NTD) patients (Narisawa et al., 2012). Given that the most common reason for NTD is a defect in one-carbon supply for nucleotide synthesis, these mutations suggest an unconventional role for GLDC. These findings, indeed, reflect the one-carbon generating function of GCS

(Fig. 3). Recent studies confirmed this observation and showed that Gldc knockdown caused NTD and hyperglycinemia in the mouse embryo. Using isotope tracing and metabolomic analysis, Pai et al. showed that GCS provides one-carbon units for nucleotide synthesis. Furthermore, formate (a one-carbon unit) supplementation was able to rescue the NTD phenotype, but not hyperglycinemia. These findings reveal a novel function of

GCS as one-carbon source for nucleotide synthesis (Leung et al., 2017; Pai et al., 2015) and GLDC is a crucial part of this system.

Figure 2. Glycine cleavage system.

GLDC, glycine decarboxylase; GCSH, Glycine Cleavage System H Protein; AMT, Aminomethyltransferase (Glycine Cleavage System Protein T); DLD, dihydrolipoamide dehydrogenase.

3.2. Functions of GLDC in cancer metabolism

There are a few, yet insightful, studies on the role of GLDC in cancer metabolism.

Zhang et al. found that the tumor-initiating cells of non-small cell lung cancer have higher

GLDC expression compared to the rest of the tumor tissue and normal lung fibroblasts.

They showed that GLDC overexpression induced tumorigenicity in NIH/3T3 fibroblast

cells and its silencing reduced tumorigenicity of Caco-2 cancer cells in a xenograft model.

Furthermore, thymidine levels, but not other pyrimidines, were reduced in cells with GLDC

knockdown, indicating a failure in one-carbon supply for nucleotide synthesis (W. C.

Zhang et al., 2012). Another study demonstrated a vital role of GLDC in supporting SGOC

metabolism in cancer. Kim et al. reported that SHMT2 is vital in LN229 cancer cells by

providing one-carbon units in a GLDC dependent fashion. They showed that SHMT2

activity generates glycine and this glycine could be converted to toxic metabolites unless

eliminated by GCS (Kim et al., 2015). This indicates there is a requirement to eliminate

excess glycine created upon increased SGOC activity, which makes GLDC indispensable.

Furthermore, a surprising aspect of glycine cleavage has been discovered by Fan J et al.,

who reported that NADPH is generated in downstream reactions of glycine cleavage

system in the mitochondria by stable isotope tracing (Fan et al., 2014). This generation of

NADPH is instrumental in GSH regeneration for reducing ROS levels. These reports

indicate a potential employment of GLDC in supporting SGOC metabolism in cancer by

metabolizing excess glycine to eliminate its toxicity, generating one-carbon units for

nucleotide synthesis and providing the redox potential for the antioxidant system (Fig. 3).

Figure 3. SGOC metabolic pathway and glycine cleavage system.

3-PG, 3-phosphoglycerate; PHGDH, phosphoglycerate dehydrogenase; PSAT1, phosphoserine aminotransferase; PSPH, phosphoserine phosphatase; GLDC, glycine decarboxylase; THF, tetrahydrofolate;

5,10-CH2-THF, methylene THF; SHMT1, serine hydroxymethyltransferase 1 (cytoplasmic); SHMT2, serine hydroxymethyltransferase 2 (mitochondrial).

II. MATERIALS AND METHODS

1. Ethics statement

All the experiments were performed according to the National Institutes of Health

(NIH) guidelines and regulations. The Institutional Animal Care and Use Committee

(IACUC) of Augusta University (animal protocol # 2008-0120) approved all the animal

experimental protocols before starting the study. All animals were kept under regular

barrier conditions, and food and water were provided ad libitum. CO2 followed by cervical

dislocation were used to euthanize animals at the end of the study.

2. Cell lines and maintenance

Neuroblastoma cell lines BE(2)-C (CRL-2268), IMR32 (CCL-127), SH-SY5Y

(CRL2266), SK-N-AS (CRL-2137), SK-N-DZ (CRL-2149), and SK-N-FI (CRL-2141)

were obtained from ATCC (Manassas, VA), LA1-55n (06041203) from Sigma-Aldrich

(St. Louis, MO), and CHLA-90, LA-N-5, LA-N-6, SMS-KANR, and SMS-KCNR from

Children’s Oncology Group Cell Culture and Xenograft Repository at the Texas Tech

University Health Sciences Center. IMR5 was a gift from J.K. Cowell (Augusta

University), NLF from M.C. Simon (University of Pennsylvania), SHEP1 from V.P.

Opipari (University of Michigan), and P493-6 from J. van Riggelen (Augusta University).

IMR32, SHEP1, SH-SY5Y, and SK-N-AS were cultured in DMEM (HyClone SH30022,

Thermo Fisher Scientific, Hampton, NH), BE(2)-C in DME/F-12 1:1 (HyClone SH30023),

and all others in RPMI 1640 (HyClone SH30027). All culture media were supplemented

with 10% FBS (Atlanta Biologicals S11050, Flowery Branch, GA) and 2 mM GlutaMAX

(Gibco 35050-061, Thermo Fisher Scientific).

For cell rescue experiments, MitoTEMPO (Sigma-Aldrich SML0737), N-Acetyl-

L-cysteine (Sigma-Aldrich A9165), L-glutathione reduced (Sigma-Aldrich G6013),

sodium formate (Sigma-Aldrich 71539), thymidine (Sigma-Aldrich T1895), hypoxanthine

(Sigma Aldrich H9636), serine (Sigma-Aldrich S4311), Dimethyl 2-oxoglutarate (Sigma-

Aldrich 349631), ribose (Sigma-Aldrich R9629), betaine (Sigma-Aldrich 61962), glycine

(Sigma-Aldrich G6600), folinic acid (Sigma-Aldrich F7878), and EmbryoMax

Nucleosides (100X, Millipore Sigma ES-008-D) were purchased and used at indicated

concentrations. Cell viability and proliferation were determined by trypan blue exclusion

assay. Cell images were acquired using an EVOS FL imaging system (Thermo Fisher

Scientific).

3. Patient data analysis

Patient data used in this study were described previously (Kocak et al., 2013;

Valentijn et al., 2012; W. Zhang et al., 2015). All analyses of the neuroblastoma patient

data were conducted using R2: Genomics Analysis and Visualization Platform

(http://r2.amc.nl), and the resulting figures and p values were downloaded.

4. Overexpression and RNA interference of target genes

The Retro-X Tet-Off Advanced Inducible System (Clontech 632105, Takara Bio,

Kusatsu, Shiga Prefecture, Japan) and the lentiviral vector pCDH-CMV-MCS-EF1-puro

(CD510B-1, SBI System Biosciences, Palo Alto, CA) were used to generate neuroblastoma

cell lines with inducible (in the absence of doxycycline) and constitutive MYCN

expression, respectively. MYCN coding sequence was generated by PCR using pBabe-

hygro/N-Myc (Cui, Li, & Ding, 2005) as a template. The sequence was verified by

sequencing and subcloned into the expression vectors. Lentiviral pLKO.1 shRNA

constructs against GLDC and MYCN were purchased from Sigma-Aldrich (Table 1).

The shGLDC-01 and shGLDC-99 sequences were cloned into the lentiviral pLKO-

teton plasmid (Wiederschain et al., 2009) for inducible expression of shRNAs against

GLDC in the presence of doxycycline. shRNA against GFP (shGFP) was used as control.

Retroviruses were produced in 293FT cells using the packaging plasmids pHDM-G and

pMD.MLVogp (gifts from R. Mulligan at MIT), and lentiviruses were produced in 293FT

cells using the packaging plasmids pLP1, pLP2, and pLP/VSVG (Thermo Fisher Scientific

K497500). Retroviral and lentiviral infections of cells were conducted according to

standard procedures.

5. Quantitative reverse-transcription PCR (qRT-PCR)

Cells were lysed using TRIzol (Thermo Fisher Scientific 15596026) for total RNA

isolation. Reverse transcription of RNA to cDNA was conducted using iScript Advanced

cDNA Synthesis Kit (Bio-Rad 172-5038, Hercules, CA). qRT-PCR was done using a 2X

SYBR green qPCR master mix (Bimake B21202, Houston, TX) on Real-Time PCR system

CFX96 (Bio-Rad). Gene-specific primers are described in Table 2. All qRT-PCR reactions

were conducted as duplicate, presented as mean ± standard error of mean (SEM) and

normalized to β2 microglobulin (B2M) mRNA levels.

Gene-targeted Construct name shRNA sequence

shGLDC-00 TRCN0000036600 CCTGCCAACATCCGTTTGAAA

shGLDC-01 TRCN0000036601 CCACGGAAACTGCGATATTAA

shGLDC-02 TRCN0000036602 GCCACT GGGAAAGAAGTGTAT

shGLDC-71 TRCN0000303371 TGTAATCTCTGTCAAGGTAAA

shGLDC-99 TRCN0000036599 CGAGCCTACTTAAACCAGAAA

shMYCN-94 TRCN0000020694 GCCAGTATTAGACTGGAAGTT

shMYCN-95 TRCN0000020695 CAGCAGCAGTTGCTAAAGAAA

Table 1. shGLDC and shMYCN sequences.

Primer set Forward Reverse

GLDC CCAGACACGACGACTTCGC CAATTCATCAATGCTCGCCAG

MYCN ACCACAAGGCCCTCAGTACCTC TGACAGCCTTGGTGTTGGAGGA

CCNA2 CTCTACACAGTCACGGGACAAAG CTGTGGTGCTTTGAGGTAGGTC

CCNB1 GACCTGTGTCAGGCTTTCTCTG GGTATTTTGGTCTGACTGCTTGC

CDK1 GGAAACCAGGAAGCCTAGCATC GGATGATTCAGTGCCATTTTGCC

CDK2 ATGGATGCCTCTGCTCTCACTG CCCGATGAGAATGGCAGAAAGC

CCND1 TCTACACCGACAACTCCATCCG TCTGGCATTTTGGAGAGGAAGTG

MCM5 ATGTCGGGATTCGACGATCCT CCAGGTTGTAATGCCGCTTG

POLE2 ACAACCAAAATCGGAGATGCG GGGCTCAGTGGGTGGAAAT

CCNE1 TGTGTCCTGGATGTTGACTGCC CTCTATGTCGCACCACTGATACC

CCNE2 CTTACGTCACTGATGGTGCTTGC CTTGGAGAAAGAGATTTAGCCAGG

MYC CCTGGTGCTCCATGAGGAGAC CAGACTCTGACCTTTTGCCAGG

B2M TGCTGTCTCCATGTTTGATGTATCT TCTCTGCTCCCCACCTCTAAGT

ATF5 CTGTCTTGGATACTCTGGACTTG CGACTTGTTCTGGTCTCTCTTC

KLF15 GACAGCATCTTGGACTTCCTATT CCATGTTCTCCTCCAGAAACTC

ZNF263 AGGTGGGACAAGGAGGAAAGCT GCATACAGACGGAACACCTTCC

Table 2. Specific primers used in qRT-PCR.

6. Immunoblotting

Cells were lysed using SDS sample buffer and protein concentrations were

measured using Bio-Rad Protein Assay Dye (Bio-Rad 5000006). Proteins (~50 μg per

sample) were separated on SDS-polyacrylamide gels, transferred to nitrocellulose

membrane (LI-COR Biosciences, Lincoln, NE, 926-31092). Proteins were detected using

rabbit anti-GLDC (Sigma-Aldrich, HPA002318), mouse anti-MYCN (Millipore Sigma,

clone OP13), mouse anti-KLF15 (Santa-Cruz Biotech, Dallas, TX, sc-271675), goat anti-

Myc tag (Abcam, Cambridge, MA, ab9132), mouse anti-V5-probe (Santa Cruz Biotech,

sc-81594), rabbit anti-MYC (Santa Cruz Biotech, sc-764), rabbit anti-GAPDH (Santa Cruz

Biotech, sc-25778), and mouse anti-α-tubulin (Sigma-Aldrich, T5168) antibodies.

Horseradish Peroxidase-conjugated goat anti-mouse (Santa Cruz Biotech, sc-2005) and

goat anti-rabbit IgG (Santa Cruz Biotech, sc2004) were used as secondary antibodies to

visualize proteins by chemiluminescence. For visualization with the Odyssey system (LI-

COR Biosciences), goat anti-mouse IRDye 800 (926-32210) or 680 (926-32220) and anti-

rabbit IRDye 800 (926-32234) or 680 (926-68021) from LI-COR Biosciences were used

as secondary antibodies.

7. Chromatin Immunoprecipitation (ChIP)

Approximately 5x107 LA1-55n and SMS-KCNR neuroblastoma cells were

collected for each ChIP assay according to the Young Lab protocol

(http://younglab.wi.mit.edu/hESRegulation/ChIP.html). Briefly, following cross-linking,

cells were lysed, chromatins sonicated, and DNA-protein complexes immunoprecipitated

using Dynabead (Thermo Fisher Scientific 10004D) and antibody complexes. Mouse anti-

MYCN (B8.4.B, Santa Cruz Biotech sc-53993) was used to capture MYCN associated

genomic DNA fragments with mouse IgG (Santa Cruz Biotech sc2025) as control,

followed by qPCR analysis using primers that cover the GLDC promoter 5’ region (-732

to -639; TSS +1) and the first intron (+578 to +654): -700 forward,

TTTCCGAAAGTGTTGCGATTAC; -700 reverse, CTCCTCCTTCTTCTCACTCTCT;

1st intron forward, GTGGCGTGGAAGCAAATTC; 1st intron reverse,

CTGGGTGCAGGAGAAAGTAAG.

8. Microarray

Total RNA was isolated using TRIzol from three independent samples of LA1-55n

teton-shGFP and -shGLDC-99 cells cultured in the presence of 0.5 μg/ml doxycycline for

7 days. RNA quality was analyzed using a 2100 Bioanalyzer (Agilent Technologies, Santa

Clara, CA). Affymetrix Human Gene 2.0 ST Array chips were used for microarray analysis

(Affymetrix, Santa Clara, CA). Data were normalized, significance was determined by

ANOVA, and differential gene expression was calculated using the Partek Genomics Suite.

Differentially expressed genes ( ±1.4-fold, p < 0.05) were subjected to Gene Ontology

(GO) analysis using DAVID (Huang, Sherman, & Lempicki, 2008). Gene Set Enrichment

Analysis (GSEA) (Subramanian et al., 2005) was conducted using GSEA desktop

application software with annotated gene sets of Molecular Signature Database v6.2.

9. Cell cycle

LA1-55n teton-shGFP and -shGLDC-99 cells were cultured in the presence of 0.5

µg/ml doxycycline for various times, collected, washed with PBS, and fixed in 70% cold

ethanol for at least 30 minutes. After fixation, cells were treated with 100 μg/ml

ribonuclease A (Qiagen 19101, Hilden, Germany) for 30 minutes at 37˚C and stained with

5 g/ml propidium iodide (P3566, Thermo Fisher Scientific) for 10 minutes. Flow

cytometry data were acquired using a FACSCanto system (BD Biosciences, San Jose, CA)

and analyzed with FlowJo v10 software (FlowJo, Ashland, OR).

10. Xenograft assay

NOD.SCID mice were obtained from the Jackson Laboratory (Bar Harbor, ME).

SMS-KCNR cells expressing either shGFP or shGLDC-99 were suspended in 100 µl

Hanks’ Balanced Salt Solution and injected subcutaneously into both flanks of 6-week-old

male and female NOD.SCID mice at 5 x 106 cells per injection site. Mice were examined

every other day and tumor growth was monitored by measurement with a caliper. Tumor

volume was calculated with the equation volume = 1/2 x (length x width2). Mice were

euthanized when their tumors reach ~2.0 cm in diameter or show sign of necrosis. The

experiment was terminated 9 weeks after injection. The animal study was approved by the

Institutional Animal Care and Use Committee of Medical College of Georgia, Augusta

University.

11. Metabolomics

LA1-55n teton-shGFP and -shGLDC-99 cells were cultured in the presence of 1

µg/ml doxycycline for 9 days and collected by scraping and centrifugation. Cell pellets

were washed once with ice-cold PBS, snap-frozen in liquid nitrogen, and stored at -80oC

until analysis. Six biological replicate samples (~5 x 106 cells/sample) were analyzed for

each group. Metabolite extraction and metabolomic analysis were carried out at the NIH

West Coast Metabolomics Center at the University of California Davis. Data were acquired

using Agilent 6890 Gas Chromatography-Leco Pegasus IV Time of Flight Mass

Spectrometry and processed using ChromaTOF software version 2.32 (Leco, St. Joseph,

MI) and the BinBase algorithm as described previously (Fiehn et al., 2008).

12. Statistical analysis

Quantitative data from qRT-PCR and cell proliferation assays were presented as

mean ± SEM and metabolomics as mean ± standard deviation (SD). The data were

analyzed for statistical significance using unpaired, two-tailed Student’s t-test or two-way

analysis of variance (ANOVA) to compare two or more groups, respectively. GraphPad

Prism version 7.03 for Windows (GraphPad Software, Inc., San Diego, CA) was used to

perform the statistical analysis. Differences with p-values less than 0.05 were considered

statistically significant (ns: not significant, *p<.05, **p<.01, ***p<.001, ****p<.0001).

III. RESULTS

1. GLDC expression correlates with poor prognosis in neuroblastoma

To investigate the relevance of GLDC in neuroblastoma, we examined the

expression of GLDC in previously published gene expression profiling data of

neuroblastoma tumors. In the SEQC (W. Zhang et al., 2015), KOCAK (Kocak et al., 2013)

and VERSTEEG (Molenaar et al., 2012) cohorts, across 1235 neuroblastoma patients, high

levels of GLDC expression were significantly correlated with poor survival outcomes (Fig.

4a). GLDC expression levels were also found to be correlated with stages of the disease;

advanced stages of neuroblastoma tumors have higher GLDC expression (Fig. 4b). The

most useful prognostic factor in neuroblastoma is the risk group, and patients with high-

risk neuroblastoma have the worst prognosis. GLDC expression was significantly higher

in patients of the high-risk group (Fig. 4c). The prognosis in relapsed neuroblastoma cases

are significantly poor and five-year survival is around 20% (London et al., 2011) and

GLDC expression was also found to be higher in relapsed disease (Asgharzadeh et al.,

2006) (Fig. 4d). These findings suggest that GLDC may have an important role in the

prognosis of the neuroblastoma.

Figure 4. GLDC expression correlates with poor prognosis in neuroblastoma.

A Kaplan-Meier survival curves for three independent cohorts of neuroblastoma patients based on GLDC mRNA expression. B-D Box plots of GLDC mRNA expression in relation to neuroblastoma stages (B), risk group (C), and disease relapse (D) using datasets from indicated neuroblastoma cohorts.

2. GLDC overexpression in neuroblastoma cell lines has no significant effect on cell

proliferation

Previously, GLDC overexpression has been shown to increase cell growth and soft

agar colony formation of mouse NIH/3T3 fibroblasts and human lung fibroblast HLF cells,

and to increase the tumorigenicity of lung tumor cells in patient-derived lung cancer

xenograft assay (W. C. Zhang et al., 2012), indicating an oncogenic potential of the GLDC

gene. To test whether GLDC has a tumorigenic function in neuroblastoma, we planned to

overexpress GLDC in neuroblastoma cell lines that display no detectable levels of GLDC

protein and to monitor the cell proliferation. First, we performed immunoblot analysis of

neuroblastoma cell lines to determine GLDC expression, which showed no detectable

levels of GLDC protein in SK-N-AS and SHEP1 neuroblastoma cell lines (Fig. 5). We then

overexpressed GLDC in these cell lines. After confirming GLDC overexpression by RT-

PCR (mRNA) and immunoblotting (protein) (Fig. 6a, b), we cultured control (pCDH) and

GLDC overexpressing cells and monitored cell proliferation. After 10 days of cell culture,

we found that GLDC overexpression did not increase cell proliferation in both cell lines

(Fig. 6c, d). Thus, at least in the cell lines examined, GLDC overexpression alone is not

sufficient to promote cell proliferation.

Figure 5. GLDC and MYCN protein expression in neuroblastoma cell lines.

Western blot of neuroblastoma cells to determine GLDC protein expression and MYCN amplification. Most of the MYCN-amplified cells have GLDC expression.

Figure 6. GLDC overexpression does not induce proliferation in neuroblastoma.

A, B qRT-PCR (A) and immunoblot (B) analyses of GLDC expression in SHEP1 and SK-N-AS cells. Error bars (A) represent SEM (n = 2). Data were analyzed by unpaired, two-tailed Student’s t-test. **p < 0.01. C Cell growth assay of GLDC overexpressed cell lines. Error bars represent SEM (n = 5). D Microscopic image of SK-N-AS cells without or with GLDC overexpression.

3. Using shRNA to downregulate GLDC expression in neuroblastoma cell lines

Although GLDC overexpression is not sufficient to increase cell proliferation in

neuroblastoma cells, GLDC might be required for their proliferation. To test this

hypothesis, we planned to knockdown the expression of GLDC and then monitor cell

proliferation. We obtained five pLKO.1 plasmids having different shRNA sequences

targeting the GLDC gene. We tested these plasmids in LA1-55n cells and found shGLDC-

01 and shGLDC-99 constructs were the most effective in reducing GLDC protein levels

(Fig. 7a). Thus, we chose these constructs in GLDC knockdown experiments. Because of

the unavailability of GLDC inhibitor, we needed to conduct all experiments with shRNA

to silence GLDC expression and to inhibit GLDC activity. In order to have a versatile

knockdown system that we can silence GLDC expression with a drug (doxycycline) in an

established cell line, we cloned shGLDC DNA from the constitutively active pLKO.1

plasmid to the tetracycline inducible pLKO teton plasmid. In the pLKO teton plasmid,

shRNA is repressed by Tet repressor protein (TetR) in the absence of tetracycline or

doxycycline (Fig 7b). Upon doxycycline treatment, TetR is released from the tet operator

and shRNA is expressed. We introduced inducible shGFP, shGLDC-01, and shGLDC-99

expressing plasmids into LA1-55n cells and established stable cell lines. We confirmed

GLDC knockdown in time and dose-dependent fashion in the presence of doxycycline (Fig.

7c). Compared to parental and inducible shGFP controls, GLDC protein levels in the cells

with inducible shGLDC construct were significantly reduced upon doxycycline treatment,

and the shGLDC-99 construct was the most effective (Fig. 7d).

Figure 7. GLDC knockdown system.

A Immunoblot analysis of GLDC protein expression in LA1-55n cells infected with lentiviruses expressing shRNA sequences against different regions of the GLDC gene, with shGFP as control. B Simplified map of plasmids with constituent (pLKO.1) and inducible (pLKO teton) shRNA. TetR, tet repressor protein. C Immunoblot analysis of GLDC protein expression in LA1-55n teton shGLDC-99 cells in the absence and presence of different doses and duration of doxycycline. D Immunoblot analysis of GLDC protein expression in the presence of 0.5μg/ml doxycycline for 7 days.

4. GLDC knockdown inhibits cell proliferation in cell culture

To test whether GLDC is required for the proliferation of neuroblastoma cells, we

first silenced GLDC expression using constitutive shRNA constructs and monitored cell

proliferation with trypan blue exclusion assay. We found that constitutive GLDC

knockdown significantly reduced cell proliferation in neuroblastoma cell lines with

significant levels of endogenous GLDC expression, including SMS-KCNR, SK-N-DZ,

NLF, IMR5, IMR32, and LA1-55n neuroblastoma cell lines (Fig. 8a, Fig. 9). The shGLDC-

99 construct was most effective in inhibiting cell proliferation, which was also most

effective in reducing GLDC protein levels (Fig. 7a, d). We next tested inducible GLDC

knockdown in LA1-55n teton cells. Upon inducing shRNA expression in the presence of

doxycycline, we found a significant decrease in cell proliferation in shGLDC-01 and

shGLDC-99 cells compared to parental and shGFP cells (Fig. 8b). These results indicate

that GLDC is required for the proliferation of neuroblastoma cell lines with high levels of

endogenous GLDC expression.

Figure 8. GLDC knockdown impairs cell growth in neuroblastoma.

A Cell growth assay of neuroblastoma cell lines without (shGFP) or with GLDC knockdown (shGLDC). B Cell growth assay of LA1-55n (parental), with inducible shGFP and shGLDC cells in the presence of 0.5μg/ml doxycycline. Error bars represent SEM (n = 5). *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 9. GLDC is required in neuroblastoma cells.

Microscopic Images of NLF, IMR5 and LA1-55n cells without (shGFP) or with GLDC knockdown.

5. GLDC knockdown inhibits the tumorigenesis of SMS-KCNR in xenograft assay

To investigate whether GLDC is required for the tumorigenicity of neuroblastoma

cells in vivo, we performed tumorigenicity assay. We used the SMS-KCNR neuroblastoma

cell line, which is tumorigenic in immunodeficient mice (Khanna, Jaboin, Drakos, Tsokos,

& Thiele, 2002). We infected SMS-KCNR cells with pLKO.1-based lentiviruses with

constitutive expression of either shGFP or shGLDC-99. We then implanted the cells

subcutaneously to NOD.SCID mice (n=4 each group, on the basis of abovementioned

reference). All mice in the shGFP group formed measurable tumors after 9 days, whereas

no tumor growth detected in the shGLDC group until 40 days after implantation (Fig. 10a,

b). Tumors in the shGFP group reached to the endpoint (either tumor diameter reaches to

2cm diameter or tumor becomes necrotic) after 26 days and mice were sacrificed. On the

other hand, only one mouse with tumor in the shGLDC group reached the end-point after

55 days and sacrificed. The experiment was terminated after 66 days, with the remaining

three mice developing tumors but none reaching the end-point. Thus, GLDC knockdown

significantly impaired the tumorigenicity of SMS-KCNR cells and increased mouse

survival in this xenograft model (Fig. 10c).

Figure 10. GLDC knockdown impairs tumorigenesis in neuroblastoma xenograft.

A-B Images (A) and tumor growth curves (B) of SMS-KCNR xenografts without or with GLDC knockdown. Data were analyzed by multiple t-test (Prism). C Survival curves for mice carrying SMS-KCNR xenografts without or with GLDC knockdown. Log-rank test p-value is shown.

6. Rescue experiments in GLDC knockdown cells

GLDC knockdown impairs the glycine cleavage system and inhibits glycine

metabolism (Kim et al., 2015) and generation of products (Pai et al., 2015), which

decreases cell proliferation. To shed light on the function of GLDC in cell proliferation,

we conducted a series of rescue experiments in GLDC-knockdown cells by supplementing

with potential downstream products of the glycine cleavage system (Fig. 11a). We used

inducible LA1-55n teton-shGLDC-99 cells for the rescue experiments. Upon addition of

doxycycline, we observed a significant decrease in cell proliferation in shGLDC-99 cells

but not in control shGFP cells (Fig. 11b).

To rescue the proliferation-defective phenotype induced by GLDC knockdown, we

supplemented the cells with one-carbon units and potential downstream products to

substitute for 5,10-CH2-THF (formate, methionine, thymidine, hypoxanthine, nucleotides),

antioxidant metabolites to substitute for NADH (MitoTEMPO, NAC, GSH), serine to

supply SGOC metabolism, and folinic acid to supply folate (Fig. 11a, Table 3). For rescue

experiments, equal number of LA1-55n teton shGLDC-99 cells were cultured in the

absence of doxycycline (-Dox, without GLDC knockdown), in the presence of doxycycline

(GLDC knockdown) without supplemental metabolites (vehicle) or in the presence of

doxycycline with the indicated supplemental metabolites for about 10 days. Cell

proliferation was measured using trypan blue exclusion assay.

We found that only hypoxanthine or a combination of folinic acid, GSH and

formate showed a modest rescuing effect, as evidenced by the increased cell number

compared to vehicle (Fig. 11c). Although this increase in cell number was statistically

significant, it was only a modest rescue compared to the cells without GLDC knockdown

(-Dox). These findings suggest that GLDC knockdown in LA1-55n cells may have a broad effect on cellular metabolism that cannot be completely rescued by the supplementation of a limited number of metabolites.

Metabolite Used for the rescue of: Reference: Nucleosides, PHGDH inhibition (Reid et al., 2018) ribose, a-KG Formate MTHFD2 and SHMT1 knockdown (Ducker et al., 2016) Formate + NAC SHMT2 knockdown (Ducker et al., 2016) GSH, NAC MTHFD2 knockdown (Ducker et al., 2016) Formate, GLDC knockout (gene trap) (Leung et al., 2017) methionine Pyruvate, GSH Serine-glycine free diet (O. D. Maddocks et al., 2013) Sarcosine GLDC knockdown (W. C. Zhang et al., 2012) Formate MTHFD2 knockdown (Ben-Sahra, Hoxhaj, Ricoult, Asara, & Manning, 2016) Hypoxanthine + Knockdown of various genes in the one-carbon (Burgos-Barragan et al., thymidine metabolic pathway 2017) Metabolite The rationale to use in the rescue experiment: Reference: Folinic acid Targeting one-carbon metabolism leaves THF (Zheng et al., 2018) unsubstituted with one-carbon units. Unsubstituted THF is vulnerable to damage and degradation NAC, SGOC metabolism provides NADPH. Inhibiting (Fan et al., 2014) MitoTEMPO, GSH MTHFD1 and MTHFD2 increases ROS levels in the cells and they become vulnerable to ROS induced cell death. NAC, PHGDH knockdown increases mitochondrial ROS (Samanta et al., 2016) MitoTEMPO, GSH levels, decreases GSH/GSSG ratios and reduces mammosphere formation and metastasis.

Table 3: Metabolites used in rescue experiments of impaired SGOC metabolism.

Figure 11. Rescue experiments of GLDC knockdown.

A Substrates, immediate and downstream-products of glycine cleavage system. B Trypan blue exclusion assay of LA1-55n teton shGFP and shGLDC-99 cells cultured in the absence or presence of 0.5 µg/ml doxycycline (Dox) for 15 days. C Trypan blue exclusion assay of LA1-55n teton shGLDC-99 cells cultured around 10 days in the absence (-Dox) or presence of 0.5 µg/ml doxycycline (Dox) without (vehicle) or with supplementation of MitoTEMPO (10 µM), NAC (N-acetyl-L-cysteine, 100 µM), GSH (glutathione reduced, 250 µM), formate (2 mM), methionine (0.5 mM), thymidine (30 µM), hypoxanthine (30 µM), betaine (10mM), nucleosides (30 µM for each nucleoside), serine (0.5 mM), fol (folinic acid, 50μM), ribose (15mM), and a-KG (alpha ketoglutarate, 3mM). Error bars represent SEM (n = 5). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

7. GLDC knockdown broadly impacts cellular metabolism

To further elucidate the function of GLDC in cancer metabolism, we performed

metabolomic profiling of LA1-55n cells with inducible GLDC knockdown using gas

chromatography-mass spectrometry. We compared the metabolite profile of LA1-55n cells

with teton-shGFP (control) and teton-shGLDC (knockdown) in the presence of

doxycycline. The profiling revealed a total of 81 identified metabolites whose levels were

significantly changed (p < 0.05) as a result of GLDC knockdown. Pathway analysis (Xia

and Wishart, 2011) of the increased metabolites showed significant enrichment (p < 0.01,

FDR < 0.05) of metabolic pathways involved in serine, glycine and threonine metabolism,

as well as other amino acid metabolism (Fig. 12a).

Although we observed no significant increase in the intracellular level of glycine,

GLDC knockdown resulted in an accumulation of serine, threonine, methionine, and

aspartic acid (Fig 12b). In addition to the direct link between serine and glycine, an analysis

of the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (Kanehisa & Goto,

2000) revealed metabolic pathways that connect glycine with threonine, aspartic acid, and

methionine (Fig. 12c). Moreover, the levels of some intermediates of these pathways,

including cystathionine and homoserine, were also significantly increased following

GLDC knockdown (Fig. 12d). These findings suggest that interference with glycine

breakdown by silencing GLDC expression might disrupt the flux of these metabolic

pathways, leading to increased levels of these amino acids.

Figure 12. GLDC regulates amino acid and purine metabolism.

A Network-based pathway analysis of metabolites increased by GLDC knockdown. B Relative levels of amino acids in LA1-55n teton shGFP and shGLDC-99 cells cultured in the presence of 1 µg/ml doxycycline for 9 days. C Schematic of KEGG metabolic pathways linking glycine and serine, threonine, methionine, and aspartic acid. D Relative levels of the amino acids and intermediates of the KEGG pathways. E Network- based pathway analysis of metabolites decreased by GLDC knockdown. F Relative levels of nucleosides and bases. Error bars represent SD (n = 6). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

Pathway analysis of the decreased metabolites revealed that the most significantly enriched (p < 0.05, FDR < 0.25) metabolic pathways are related to purine metabolism, the

TCA cycle, and lipid synthesis (Fig. 12e). The metabolomic profiling did not identify any nucleotides except for AMP, which showed a modest increase following GLDC knockdown. On the other hand, GLDC knockdown resulted in a significant decrease in the levels of the purine nucleosides guanosine and inosine, and of the purine bases hypoxanthine and xanthine (Fig. 12f). The reduced levels of purine nucleosides and bases in LA1-55n cells with GLDC knockdown are consistent with the role of GLDC in producing one-carbon units for the purine backbone synthesis. The decrease in hypoxanthine levels also offers an explanation for the modest rescuing effect of supplemental hypoxanthine (Fig. 11c).

Surprisingly, GLDC knockdown had also a significant impact on central carbon metabolism (glycolysis and the TCA cycle) and lipid (sterol and fatty acid) synthesis (Fig.

13a). Within the glycolysis pathway, the levels of glucose-6-phosphate (glucose-6-P), 3- phosphoglycerate (3-PG), and lactate were markedly decreased (Fig. 13b, Glycolysis), suggesting that GLDC knockdown reduced glycolysis. Within the TCA cycle, GLDC knockdown significantly increased the levels of succinate, fumarate, and malate but decreased the levels of alpha-ketoglutarate (a-KG), aconitate, and citrate (Fig. 13c, TCA cycle). The accumulation of succinate, fumarate, and malate could result from a block in the downstream reaction that catalyzes the condensation of acetyl-CoA with oxaloacetate

(OAA) to form citrate as suggested by the decreased citrate level and/or from an increase in malate production from aspartate via the malate-aspartate shuttle as suggested by the markedly increased aspartate level (Fig. 12b).

Figure 13. GLDC regulates central carbon metabolism.

A Schematic of central carbon metabolism, glutaminolysis, and lipid synthesis. PPP, pentose phosphate pathway. B-D, Relative levels of metabolites in glycolysis; PEP, phosphoenolpyruvate; 3-PG, 3- phosphoglycerate (B), TCA cycle; a-KG, alpha-ketoglutarate (C), and sterol-fatty acid synthesis pathways (D). Error bars represent SD (n = 6). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

The marked reduction in the levels of a-KG, aconitate, citrate, and glutamine (Fig.

13c, TCA cycle) suggests that GLDC knockdown might inhibit the glutamine-dependent reductive carboxylation pathway. Glutaminolysis and reductive carboxylation have a key role in supporting acetyl-CoA production for lipid synthesis in cancer cells (Altman, Stine,

& Dang, 2016; DeBerardinis, Lum, Hatzivassiliou, & Thompson, 2008; Metallo et al.,

2012; J. Zhang, Pavlova, & Thompson, 2017). Consistent with this model, we found that

GLDC knockdown significantly reduced the levels of sterols and fatty acids (Fig. 13d,

Sterols-Fatty acids). Given the critical role of the mevalonate-cholesterol synthesis pathway in sustaining neuroblastoma cell proliferation and tumorigenicity (Liu et al.,

2016), the decreased production of cholesterol suggests an additional mechanism for the growth-inhibitory effect of GLDC knockdown.

Collectively, our metabolomic profiling revealed that GLDC knockdown resulted in alterations in multiple metabolic processes critical for supplying building blocks and energy to meet biosynthetic demands of cell proliferation.

8. GLDC knockdown represses the expression of cell cycle genes

To elucidate the mechanism for the growth-inhibitory effect of GLDC knockdown,

we compared the gene expression profile of LA1-55n cells with inducible teton-shGFP

(control) and teton-shGLDC (knockdown) in the presence of doxycycline (Fig. 7d). A total

of 1094 genes were dıfferently expressed (≥ ±1.40 fold, p < 0.05) resulted from GLDC

knockdown, with 571 genes being upregulated and 523 genes downregulated (The NCBI

Gene Expression Omnibus (GEO) accession number for the microarray data is

GSE129807).

Gene ontology (GO) analysis of the upregulated genes using The Database for

Annotation, Visualization and Integrated Discovery (DAVID) functional annotation tool

(Huang da, Sherman, & Lempicki, 2009a, 2009b) revealed no significant enrichment of

GO terms (FDR < 0.05), whereas GO analysis of the downregulated genes showed

significant enrichment for GO terms associated with DNA replication, G1 to S transition,

and cell proliferation (Fig. 14a). We obtained essentially the same results with Gene Set

Enrichment Analysis (GSEA) (Subramanian et al., 2005) of the microarray data, which

showed significant downregulation of gene sets involved in cell proliferation (Fig. 14b),

including those of the MCM pathway (unwinding DNA prior to replication), DNA

replication, cell cycle progression, and E2F targets (Fig. 14c). We confirmed the

microarray data by qRT-PCR, showing that GLDC knockdown led to a significant

reduction in mRNA expression of MCM5, cyclins (CCNs) and cyclin-dependent kinase

(CDKs) (Fig. 14d). These findings suggest that knockdown of GLDC metabolic enzyme,

beyond metabolic alteration, induces genetic alteration that impedes DNA replication, cell

cycle progression and eventually cell proliferation.

Figure 14. GLDC knockdown reduces the expression of cell cycle genes.

A-C GO (A) and GSEA (B-C) analyses of genes regulated by GLDC knockdown showing significant enrichment of gene sets involved in DNA replication and cell cycle progression. D qRT-PCR of the expression of indicated cell cycle genes in LA1-55n cells without or with GLDC knockdown. Error bars represents SEM (n = 2). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

9. GLDC knockdown induces cell cycle arrest

Based on the microarray data, we examined the effect of GLDC knockdown on cell

cycle progression using the LA1-55n cell line with inducible GLDC knockdown. Culturing

the cells in the presence of doxycycline resulted in time-dependent downregulation of

GLDC protein expression (Fig. 7c), which was accompanied by a gradual increase in the

fraction of cells in the G1 phase and a concomitant decrease in the S phase population (Fig.

15a-b). This G1 arrest in cell cycle was most pronounced in the LA1-55n cells expressing

shGLDC-99, which was most effective in silencing GLDC expression (Fig. 7d). In parallel

to these findings, RNA microarray of GLDC-silenced LA1-55n cells showed a significant

decrease in expression of genes related to G1/S transition, which also suggest a block of

cell cycle in G1 phase (Fig. 14a).

Previous studies showed that inhibiting SGOC metabolic pathway with NCT-503,

PHGDH inhibitor, induced G1 arrest in cell cycle in MDA-MB-468 cells (Pacold et al.,

2016). We also observed G1 arrest in cell cycle upon treatment with NCT-503 in BE(2)-C

cells (Xia et al., 2019). These findings suggest that inhibiting SGOC metabolic pathway

via targeting different members causes G1 arrest in cell cycle.

Figure 15. GLDC knockdown induces G1 phase arrest.

A Time-course cell cycle analysis of LA1-55n cells without (teton-shGFP) or with inducible GLDC knockdown (teton-shGLDC-01 and shGLDC-99) in the presence of doxycycline (Dox). B Cell cycle distribution (mean  SEM) of LA1-55n cells without or with inducible GLDC knockdown using data from (A). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

10. MYCN is a transcriptional activator of GLDC expression in neuroblastoma

We sought to identify the transcription factor(s) responsible for the elevated GLDC

expression in neuroblastoma tumors of advanced stages. MYCN is a transcription factor

and MYCN amplification is significantly associated with advanced stages of the disease

(Brodeur, 2003; Maris et al., 2007). To determine whether there is a correlation between

MYCN amplification and GLDC expression, we performed a hierarchical cluster analysis

of the MYCN-amplification gene signature consisting of 143 genes upregulated in

neuroblastoma tumors with MYCN amplification relative to those without MYCN

amplification (Janoueix-Lerosey et al., 2008; Puissant et al., 2013). This analysis revealed

co-upregulation of GLDC with other SGOC pathway genes in MYCN-amplified

neuroblastoma tumors (Fig. 16a). As a complementary approach, we analyzed the SEQC,

KOCAK and VERSTEEG datasets for a correlation between GLDC mRNA expression

and the MYCN amplification status. We found that GLDC expression was significantly

higher in tumors with MYCN amplification compared to tumors without MYCN

amplification (Fig. 16b). It is particularly interesting to note that while GLDC expression

was significantly higher in high-risk tumors with MYCN amplification there was no

significant difference in GLDC mRNA expression levels between low-risk tumors and

high-risk tumors without MYCN amplification (Fig. 16c). In agreement with the findings

from tumors, most MYCN-amplified neuroblastoma cell lines we examined showed higher

levels of GLDC protein in comparison with non-MYCN-amplified cell lines (Fig. 5).

Together, these results indicate that high GLDC expression is associated with MYCN

amplification in neuroblastoma.

Figure 16. GLDC is highly expressed in MYCN-amplified neuroblastoma.

A Heatmap showing coordinated upregulation of SGOC pathway genes in MYCN-amplified neuroblastoma tumors using SEQC-498 dataset. B Box plot of GLDC mRNA expression in relation to MYCN amplification status using three datasets. C Box plot of GLDC mRNA expression in relation to neuroblastoma risk groups and MYCN amplification status using the SEQC dataset.

We next investigated whether MYCN is able to induce GLDC expression. MYCN overexpression in non-MYCN-amplified SK-N-AS and SHEP1 neuroblastoma cell lines markedly increased GLDC mRNA (Fig. 17a) and protein expression (Fig. 17b).

Conversely, knockdown of MYCN expression in MYCN-amplified NLF, LA1-55n, and

LA-N-5 cell lines by two different shRNA constructs consistently downregulated the expression of GLDC mRNA (Fig. 17c) and protein (Fig. 17d). Thus, MYCN is a transcriptional activator of GLDC expression and is essential for maintaining GLDC expression in MYCN-amplified neuroblastoma cell lines.

MYCN and other members of the MYC family bind to the consensus E-box sequence CANNTG. Sequence examination revealed potential MYCN recognition sequences in the GLDC promoter region and first intron. To determine whether MYCN binds directly to the GLDC gene, we conducted chromatin immunoprecipitation and quantitative PCR (ChIP-qPCR) assays using MYCN-amplified SMS-KCNR and LA1-55n neuroblastoma cell lines. We detected significant levels of endogenous MYCN associated with the GLDC promoter and the first intron (Fig. 17e), suggesting that MYCN directly targets GLDC for transcriptional activation.

Figure 17. GLDC is a direct transcriptional target of MYCN.

A-B qRT-PCR (A) and immunoblot (B) analyses of GLDC expression in non-MYCN-amplified SK-N-AS cells with inducible MYCN expression in the absence of doxycycline (Doxy) and SHEP1 cells with constitutive MYCN expression. C-D qRT-PCR (C) and immunoblot (D) analyses of GLDC expression in MYCN-amplified cell lines with MYCN knockdown. E ChIP-qPCR showing endogenous MYCN binding to the GLDC promoter and 1st intron in MYCN-amplified cell lines. Error bars (A, C, E) represent SEM (n = 2). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

11. MYC activates GLDC expression

In light of the above findings, we wondered if GLDC is also a MYC target gene

given the recent findings on MYC regulation of SGOC metabolism (Dang, 2013; Dejure

& Eilers, 2017). We examined the P493-6 cell line, a lymphoid cell line with inducible

MYC expression in the absence of doxycycline (tetoff-MYC) (Pajic et al., 2000;

Schuhmacher et al., 1999). Repression of MYC expression by doxycycline resulted in a

marked reduction in GLDC mRNA expression (Fig. 18a), leading to decreased GLDC

protein expression (Fig. 18b), whereas MYC induction by removing doxycycline resulted

in a significant increase in GLDC protein expression (Fig. 18c). These findings suggest

that GLDC is MYC target gene. Our findings are consistent with previous observation that

GLDC mRNA expression is induced upon oncogenic KRASG12D, PI3KCAE545K, MYC or

MYCT58A transformation in MCF10A cells (W. C. Zhang et al., 2012).

Figure 18. MYC upregulates GLDC expression.

A-B qRT-PCR (A) and immunoblot (B) analyses of MYC and GLDC expression in the lymphoid cell line P493-6 in which MYC expression is repressed by doxycycline (Doxy) and induced by Dox removal. Error bars represent SEM (n = 2).

12. Transcriptional Regulation of GLDC expression

In addition, we investigated whether GLDC expression in neuroblastoma cells is

regulated independent of MYCN. We hypothesized that GLDC and its transcription factors

should exhibit a positive correlation in gene expression, we investigated co-expression

patterns in SEQC neuroblastoma dataset. Based on the correlation of gene expression with

GLDC, we selected Activating Transcription Factor 5 (ATF5), Kruppel Like Factor 15

(KLF15), and Zinc Finger Protein 263 (ZNF263) as candidate transcriptional regulators of

GLDC expression (Fig. 19a). We overexpressed these genes in neuroblastoma cell lines.

We found that ATF5 overexpression had no significant effect on GLDC expression in

BE2C and SK-N-DZ cell lines (Fig. 19b). ZNF263 overexpression modestly induced

GLDC mRNA expression (Fig. 19c) but GLDC protein levels did not increase in SHEP1

cells (Fig. 19d). KLF15 induced GLDC mRNA expression only in SK-N-AS and LA1-55n

cell lines (Fig. 19e) and protein levels only in SK-N-AS cells (Fig.19f) but had no effect in

other cell lines tested (data not shown). Thus, KLF15 induces GLDC in some cell lines but

not others, suggesting that it may not be a major transcriptional regulator of GLDC

expression.

Figure 19. KLF15 modestly regulates GLDC expression.

A Correlation co-efficient graph of transcription factors in SEQC dataset. R=correlation coefficient. B, C, E RT-qPCR of ATF5 (B), ZNF263 (C) and KLF15 (E) overexpressing cells. D, F Immunoblot of ZNF263 (D), and KLF15 (F) overexpressing cells. Error bars (B, C, E) represent SEM (n = 2). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

13. GLDC expression is independent of glycine levels Glycine is a substrate of GLDC and we wondered whether different levels of

glycine affect GLDC expression. We cultured LA1-55n and IMR5 neuroblastoma cells

lines and U-2 OS osteosarcoma in RPMI medium (contains 0.13 mM glycine) and HeLa

cells in DMEM medium (contains 0.4 mM glycine). After culturing the cells in the

presence of different concentrations of glycine (0.5 mM to 5 mM) at different time points

(0.5 hours to 7 days), we analyzed GLDC expression using RT-qPCR. We found that

GLDC mRNA levels were altered but these were inconsistent and not replicable (data are

not shown). We conclude that glycine might not play a direct and a major role in

transcriptional regulation of GLDC expression.

14. Excess glycine is toxic to neuroblastoma cells

While we were investigating whether alterations in glycine levels change GLDC

expression, we observed cytotoxicity with glycine supplementation in neuroblastoma cells.

Previously, excess glycine is reported to be cytotoxic to cancer cells (Kim et al., 2015;

Labuschagne et al., 2014). However, there was also a contradictory report showing that

glycine supplementation increases cell growth in rapid-proliferating cancer cells (Jain et

al., 2012). To test whether glycine is toxic in IMR5, LA1-55n, and NLF neuroblastoma

cell lines, we added 2 mM glycine to the regular growth medium (RPMI, which contains

0.13 mM glycine). We observed that 2 mM glycine inhibited cell proliferation in these cells

(Fig. 20a, b). We also found that this toxicity was dose-dependent as shown in LA1-55n

cells (Fig. 20c). These findings indicate that excess glycine is toxic to neuroblastoma cells

and need to be eliminated when generated in excess.

Figure 20. Excess glycine is toxic to neuroblastoma cells.

A Cell growth assay of cell lines without or with 2mM glycine. Error bars represent SEM (n = 5). B Microscopic image of NLF and IMR5 cells without or with 2mM glycine for five days. C Trypan blue exclusion assay of LA1-55n cells cultured in the absence or presence of glycine for 7 days. Error bars represent SEM (a, n = 5). Data were analyzed by unpaired, two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

IV. DISCUSSION

Our study uncovers an important role of GLDC in MYCN-amplified neuroblastoma, a subgroup of high-risk neuroblastoma with extremely poor prognosis. We show that

GLDC expression is elevated in neuroblastoma tumors and cell lines with MYCN- amplification. We provide further evidence that MYCN directly targets the GLDC gene for transcriptional activation and that GLDC is a common transcriptional target of the MYC family of oncogenes. At the functional level, we show that high GLDC expression is required for the proliferation and tumorigenicity of neuroblastoma cells.

Metabolic function of GLDC in neuroblastoma

Previous studies reported various metabolic functions of GLDC, including providing one-carbon units (Leung et al., 2017; Pai et al., 2015; W. C. Zhang et al., 2012), eliminating excess glycine that otherwise toxic (Kim et al., 2015), and increasing the redox potential via generating NADPH (Fan et al., 2014). These studies demonstrate a multifaceted role of GLDC in cellular metabolism in cell- and context-dependent manner.

We sought to determine the metabolic function of GLDC in the pediatric cancer neuroblastoma. Although we did not observe a significant increase in glycine levels following GLDC knockdown in LA1-55n cells, we speculate that the observed serine accumulation may be caused by a reduction in the conversion of serine to glycine as a result of decreased glycine breakdown.

We found that GLDC knockdown also altered central carbon metabolism with reduced glycolysis (as indicated by the lower level of lactate) and decreased citrate production. Glucose-derived acetyl-CoA is condensed with oxaloacetate (OAA) to form citrate. In addition, glutamine-derived -KG can generate citrate via either oxidation in

71 which -KG-derived OAA condenses with acetyl-CoA to form citrate or reductive carboxylation in which -KG is reduced to citrate via NADPH consumption (Fig. 13a)

(Comte, Vincent, Bouchard, Benderdour, & Rosiers, 2002; D’Adamo A.F & Haft, 1965;

Des Rosiers et al., 1995; Holleran, Briscoe, Fiskum, & Kelleher, 1995; Newsholme,

Crabtree, & Ardawi, 1985; Yoo, Antoniewicz, Stephanopoulos, & Kelleher, 2008). Our metabolomic profiling data are consistent with a decrease in reductive carboxylation of -

KG in neuroblastoma cells with GLDC knockdown as these cells showed increased levels of succinate, fumarate, and malate but decreased levels of aconitate. Thus, the decrease in citrate levels in neuroblastoma cells with GLDC knockdown may be caused by a combination of reduction in glycolysis and reductive carboxylation of -KG.

Mitochondrial citrate is exported to the cytosol and cleaved by ATP citrate lyase to produce cytosolic acetyl-CoA for sterol and fatty acid synthesis. In agreement with the observed reduction in citrate production, neuroblastoma cells with GLDC knockdown had significantly lower levels of sterols (e.g., lanosterol and cholesterol) and fatty acids (e.g., palmitate and myristic acid) in comparison with control cells. Cholesterol is a structural component of the cellular membrane (Harayama & Riezman, 2018) and thus, its continuous production is required for the replication of cancer cells. Recent studies also suggest an important role for other sterol products in cancer development. For example, oxysterols (products of cholesterol oxygenation) promote cancer cell proliferation by functioning as ligands of nuclear receptors (Kloudova, Guengerich, & Soucek, 2017;

Nelson et al., 2013; Silvente-Poirot, Dalenc, & Poirot, 2018).

Similarly, cancer cell proliferation requires fatty acids for the synthesis of membranes and signaling molecules (Currie, Schulze, Zechner, Walther, & Farese, 2013). We have shown

72 previously that blocking cholesterol synthesis using statins inhibits the proliferation and tumorigenicity of neuroblastoma cells (Liu et al., 2016). Thus, the reduced production of sterols and fatty acids may also contribute to the growth-inhibitory effect of GLDC knockdown. The molecular mechanism by which GLDC knockdown alters central carbon metabolism remains to be determined. In this regard, it is interesting to note that a recent study reports that inhibition of phosphoglycerate dehydrogenase (PHGDH), the first enzyme of the SGOC pathway, disrupts the mass balance within central carbon metabolism. Thus, disruption of the carbon flux in the SGOC pathway can simultaneously alter the activity of multiple related metabolic pathways (Reid et al., 2018). It is possible that GLDC knockdown acts through a similar mechanism.

Interestingly, The Human Protein Atlas shows that GLDC and AMT detected in the nucleus but not GCSH and DLD from the glycine cleavage system. This may indicate an unknown function of GLDC in the nucleus since there is no functional glycine cleavage system. Previously, it has been reported that some metabolic enzymes may have non- canonical functions and they can be found in the nucleus to regulate gene expression

(Huangyang & Simon, 2018). Since the only known function of GLDC is in the glycine cleavage system, we do not rule out that the presence of GLDC in the nucleus may indicate a novel function beyond its enzymatic activity.

Rescue experiments of cell proliferation in GLDC knockdown

To reveal the metabolic function of GLDC in cancer metabolism, we conducted a series of rescue experiments by providing metabolites that may rescue cell proliferation in

GLDC knockdown LA1-55n neuroblastoma cells. Previously, Pai et al. reported that

GLDC gene knockout caused neural tube defects in mouse embryo by reducing one-carbon

73 carrying folate levels. Maternal formate (a one-carbon unit) supplementation restored embryonic growth and prevented the neural tube defect in GLDC-deficient mouse embryo

(Pai et al., 2015). They found that methionine supplementation also rescued the neural tube defect in these mutant mouse embryo because one-carbon metabolism and methionine cycle is connected and one-carbon units can be transferred between them (Leung et al.,

2017). We supplemented GLDC-knockdown cells with formate and methionine but could not observe a rescuing effect on cell proliferation. This could indicate that GLDC’s function is context dependent, and the cells in embryonic development may rely more on

GLDC for one-carbon generation than neuroblastoma cells.

The SGOC metabolic pathway is important in nucleotide synthesis and inhibiting this pathway decreases cell proliferation. Reid et al. supplemented the PHGDH-inhibited cells with nucleosides and rescued the cell proliferation (Reid et al., 2018). In a similar manner, hypoxanthine (to supplement purine synthesis) and thymidine were used to rescue cell proliferation in SHMT1 mutant HAP1 cells (Burgos-Barragan et al., 2017). We used hypoxanthine, thymidine, and nucleosides for the rescue of the cell proliferation in GLDC knockdown neuroblastoma cells and observed a modest rescuing effect only with hypoxanthine.

In 2012, Zhang et al. reported that GLDC expression is elevated in the tumor- initiating cells of non-small cell lung cancer and silencing GLDC reduced cell proliferation.

In metabolomics study, they found that glycine-related metabolites sarcosine, betaine aldehyde, and serine were decreased. They showed that sarcosine supplementation moderately rescued cell proliferation in GLDC knockdown A549 cells (W. C. Zhang et al.,

2012). We supplemented GLDC knockdown cells with betaine aldehyde, which could be metabolized to sarcosine, but observed no rescue of cell proliferation.

An interesting study by Fan et al. showed that the glycine cleavage system contributes NADPH generation in the cell. In an effort to trace the sources of NADPH, they found that one-carbon units generated by glycine cleavage is further metabolized by

Aldehyde Dehydrogenase 1 Family Member L2 (ALDH1L2) mediated reaction (10-CHO-

+ + THF + NADP + H2O = THF + CO2 + NADPH + H ) to generate NADPH in the mitochondria (Fan et al., 2014). The redox potential in NADPH is required to regenerate the main antioxidant GSH (reduced glutathione) from GSSG (oxidized glutathione).

Another study by Ducker et al. showed that targeting the mitochondrial one-carbon metabolic pathway enzymes SHMT2 and MTHFD2 reduced GSH levels. They found that supplementation with either NAC (glutathione precursor) or GSH rescued the cell proliferation in MTHFD2 mutant HEK293T cells (Ducker et al., 2016). In agreement with this finding, Samanta et al. observed that PHGDH knockdown reduced NADPH levels and

GSH/GSSG ratios (reduced glutathione/oxidized glutathione, indicates cellular antioxidant levels), along with decreased mammosphere formation in MDA-MB-231 and MCF-7 cell lines (Samanta et al., 2016). These findings show that the glycine cleavage system and

SGOC metabolic pathway increases the redox potential, which is important for cell survival. We supplemented cells with NAC, GSH, and MitoTEMPO (mitochondria- specific antioxidant) to enhance the redox potential in order to rescue the cell proliferation in GLDC knockdown cells. Although we did not observe a rescue of cell proliferation with these metabolites individually, we found a modest rescue with a combination of GSH, folinic acid, and formate.

A recent study reported that small molecule-mediated PHGDH inhibition disrupted not only SGOC metabolism but also other related metabolic pathways, including pentose phosphate pathway and TCA cycle, and reduced cell proliferation. The study also showed that ribose (to supplement the pentose phosphate pathway) and -KG (to supplement the

TCA cycle) supplementation rescued the cell proliferation (Reid et al., 2018). Our findings in metabolomics study of GLDC knockdown also implicate a broad metabolic alteration, including a decrease in the levels of ribose and purines, and altered TCA cycle metabolism.

However, we observed no rescuing effect on the cell proliferation of GLDC knockdown cells by supplying ribose and -KG.

Tetrahydrofolate (THF), a member of the folate family, is employed in the transfer of one-carbon units from serine (and glycine) to nucleotide synthesis reactions, and deficiency of folates impairs cell proliferation. Zheng et al. showed that knockdown of

SHMT2 and MTHFD1L, members of SGOC metabolic pathway, reduces one-carbon unit generation and leaves THF unsubstituted with one-carbon units. They showed that unsubstituted THF is vulnerable to oxidative damage and degradation (Zheng et al., 2018).

To test whether THF left unsubstituted in GLDC knockdown cells and folate supplementation could rescue proliferation, we supplemented the GLDC knockdown cells with folinic acid and in combinations with one-carbon substitutes, antioxidant, and nucleoside. We observed only a modest rescue with the folinic acid, GSH and formate combination but not with other supplementations.

GLDC knockdown inhibits the glycine cleavage system and causes glycine accumulation in the cell. As a result, excess glycine could be converted to serine in a reverse reaction by SHMT enzymes (glycine + 5,10-CH2-THF = serine + THF) at the

76 expense of one-carbon units as it has been shown previously (Labuschagne et al., 2014).

To counterbalance the excess glycine and prevent reversal of the reaction that may drain one-carbon units, we supplemented GLDC knockdown cells with serine. We did not find a rescue in cell proliferation but we observed that excess serine (up to 5 mM), unlike glycine, does not inhibit cell proliferation (data not shown).

Overall, we found a modest rescue with hypoxanthine, folinic acid, formate, and

GSH, which indicates that GLDC has a function in support of SGOC and redox metabolism. However, we did not observe a robust and/or complete rescue of cell proliferation in our experiments as previously reported in the dramatic rescue of neural tube defect phenotype with formate supplementation in GLDC-silenced mouse embryo.

Metabolomics study with GLDC knockdown cells suggests that it has also impact on multiple metabolic processes beyond the SGOC metabolic pathway. This offers an explanation of why supplementation with a set of metabolites could not compensate for the extensive metabolic alterations caused by GLDC knockdown and completely rescue the cell proliferation.

Therapeutic potential of targeting the SGOC metabolic pathway and GLDC

Chemotherapy is one of the approaches in the multimodal treatment of high-risk neuroblastoma and it includes vincristine, vindesine, etoposide, cisplatin, carboplatin, dacarbazine, doxorubicin, cyclophosphamide, ifosfamide, and topotecan (Matthay et al.,

2016). Antimetabolites that targeting the SGOC metabolic pathway (e.g.; methotrexate, pemetrexed, 5-fluorouracil) are not used due to failure of a clinical trial with methotrexate

(Ablin et al., 1978) and with pemetrexed (Warwick et al., 2013). However, recent findings underlined the vital function of the SGOC metabolic pathway in neuroblastoma and other

77 cancers. These studies revealed the importance of key enzymes of the SGOC metabolic pathway in cancer and discovered new drugs to target these enzymes.

Our laboratory discovered a novel epigenetic regulation of de-novo serine synthesis and demonstrated the importance of this mechanism in the tumorigenesis of neuroblastoma

(Ding et al., 2013). Another report showed that MYCN amplified neuroblastoma cells are more sensitive to methotrexate toxicity (Lau et al., 2015), indicating functional importance of the SGOC metabolic pathway in neuroblastoma. A recent clinical trial on relapsed neuroblastoma used methotrexate in multi-drug combination (previous clinical trial used methotrexate as a single agent) and found that this treatment prevented disease progression

(Heng-Maillard et al., 2019). Although they also observed significant hematologic adverse effects, they were tolerated. Together, these findings suggest the benefit of targeting the

SGOC metabolic pathway in neuroblastoma treatment. However, the toxicity and inadequacy of the classic chemotherapeutics create a barrier for successful treatment.

In the last 5 years, new small molecule inhibitors have been discovered that target

PHGDH, SHMT1/2, and MTHFD2 enzymes in the SGOC metabolic pathway and they are in pre-clinical trials in various types of cancer. Here, we demonstrated that targeting

GLDC could be an effective strategy in treatment of high-risk neuroblastoma. GLDC expression is detected in a few organs as shown in The Human Protein Atlas, which may suggest fewer potential side effects. Moreover, a GLDC inhibitor was recently discovered by the research group in the Experimental Therapeutics Centre in Singapore (personal communication) and currently it is being tested in neuroblastoma. Employment of this kind of novel inhibitors that target GLDC in high-risk neuroblastoma might improve the survival of patients.

V. SUMMARY

 Neuroblastoma is a major cause of cancer-related deaths in young children and

current treatments are inadequate to improve patient survival.

 MYCN amplification is particularly associated with poor prognosis and its

downstream target genes favor aggressive properties.

 Directly targeting MYCN is challenging due to its structure. An alternative

approach is to target MYCN-regulated genes for neuroblastoma treatment.

 We identified a novel MYCN target gene, GLDC, and showed that MYCN directly

binds to GLDC gene and induces its expression.

 We found that GLDC expression correlates with poor prognosis of neuroblastoma.

GLDC knockdown decreased the expression of proliferation-related genes, induced

cell cycle arrest, and inhibited cell proliferation and tumorigenesis in

neuroblastoma.

 Metabolomic profiling and rescue experiments suggest that GLDC is required for

SGOC metabolism for nucleotide synthesis and for the maintenance of the redox

potential.

 We found that GLDC knockdown disrupts nutrient utilization (glycolysis and

glutaminolysis), TCA cycle, and reductive carboxylation metabolisms, and

decreases cholesterol and lipid synthesis.

In summary, our study demonstrates that GLDC is transcriptionally regulated by

MYCN. We present evidence that GLDC regulates neuroblastoma metabolism and is required for the tumorigenesis and proliferation of neuroblastoma cells. We conclude that inhibiting GLDC may be an effective strategy in the treatment of high-risk neuroblastoma.

VI. REFERENCES

Ablin, A. R., Bleyer, W. A., Finklestein, J. Z., Hartmann, J. R., Leikin, S., & Hammond, G. D. (1978). Failure of moderate-dose prolonged-infusion methotrexate and citrovorum factor rescue in patients with previously treated metastatic neuroblastoma--a phase II study. Cancer Treat Rep, 62(7), 1097-1099. Altman, B. J., Stine, Z. E., & Dang, C. V. (2016). From Krebs to clinic: glutamine metabolism to cancer therapy. Nat Rev Cancer, 16(10), 619-634. doi: 10.1038/nrc.2016.71 Ambros, P. F., Ambros, I. M., Brodeur, G. M., Haber, M., Khan, J., Nakagawara, A., . . . Maris, J. M. (2009). International consensus for neuroblastoma molecular diagnostics: report from the International Neuroblastoma Risk Group (INRG) Biology Committee. Br J Cancer, 100(9), 1471-1482. doi: 10.1038/sj.bjc.6605014 Asgharzadeh, S., Pique-Regi, R., Sposto, R., Wang, H., Yang, Y., Shimada, H., . . . Seeger, R. C. (2006). Prognostic significance of gene expression profiles of metastatic neuroblastomas lacking MYCN gene amplification. J Natl Cancer Inst, 98(17), 1193-1203. doi: 10.1093/jnci/djj330 Ben-Sahra, I., Hoxhaj, G., Ricoult, S. J. H., Asara, J. M., & Manning, B. D. (2016). mTORC1 induces purine synthesis through control of the mitochondrial tetrahydrofolate cycle. Science, 351(6274), 728-733. doi: 10.1016/j.devcel.2016.02.01210.1126/science.aad0489 Biedler, J. L., & Spengler, B. A. (1976). A novel chromosome abnormality in human neuroblastoma and antifolate-resistant Chinese hamster cell lives in culture. J Natl Cancer Inst, 57(3), 683-695. Boroughs, L. K., & DeBerardinis, R. J. (2015). Metabolic pathways promoting cancer cell survival and growth. Nat Cell Biol, 17(4), 351-359. doi: 10.1038/ncb3124 Brodeur, G. M. (2003). Neuroblastoma: biological insights into a clinical enigma. Nat Rev Cancer, 3(3), 203-216. doi: 10.1038/nrc1014 Burgos-Barragan, G., Wit, N., Meiser, J., Dingler, F. A., Pietzke, M., Mulderrig, L., . . . Patel, K. J. (2017). Mammals divert endogenous genotoxic formaldehyde into one-carbon metabolism. Nature, 548(7669), 549-554. doi: 10.1038/nature23481 Burkhart, C. A., Cheng, A. J., Madafiglio, J., Kavallaris, M., Mili, M., Marshall, G. M., . . . Haber, M. (2003). Effects of MYCN antisense oligonucleotide administration on tumorigenesis in a murine model of neuroblastoma. J Natl Cancer Inst, 95(18), 1394-1403. Circu, M. L., & Aw, T. Y. (2010). Reactive oxygen species, cellular redox systems, and apoptosis. Free Radic Biol Med, 48(6), 749-762. doi: 10.1016/j.freeradbiomed.2009.12.022 Cohn, S. L., Pearson, A. D., London, W. B., Monclair, T., Ambros, P. F., Brodeur, G. M., . . . Matthay, K. K. (2009). The International Neuroblastoma Risk Group (INRG) classification system: an INRG Task Force report. J Clin Oncol, 27(2), 289-297. doi: 10.1200/jco.2008.16.6785

Comte, B., Vincent, G., Bouchard, B., Benderdour, M., & Rosiers, C. D. (2002). Reverse flux through cardiac NADP+-isocitrate dehydrogenase under normoxia and ischemia. American Journal of Physiology-Heart and Circulatory Physiology, 283(4), H1505-H1514. doi: 10.1152/ajpheart.00287.2002 Cox, D., Yuncken, C., & Spriggs, A. I. (1965). MINUTE CHROMATIN BODIES IN MALIGNANT TUMOURS OF CHILDHOOD. Lancet, 1(7402), 55-58. Cui, H., Li, T., & Ding, H. F. (2005). Linking of N-Myc to death receptor machinery in neuroblastoma cells. J Biol Chem, 280(10), 9474-9481. Curreri, A. R., Ansfield, F. J., Mc, I. F., Waisman, H. A., & Heidelberger, C. (1958). Clinical studies with 5-fluorouracil. Cancer Res, 18(4), 478-484. Currie, E., Schulze, A., Zechner, R., Walther, T. C., & Farese, R. V., Jr. (2013). Cellular fatty acid metabolism and cancer. Cell Metab, 18(2), 153-161. doi: 10.1016/j.cmet.2013.05.017 D’Adamo A.F, J., & Haft, D. E. (1965). An Alternate Pathway of Alpha-Ketoglutarate Catabolism in the Isolated, Perfused Rat Liver. I. Studies with Dl-Glutamate-2- and -5-14c. J Biol Chem, 240, 613-617. Dang, C. V. (2013). MYC, metabolism, cell growth, and tumorigenesis. Cold Spring Harb Perspect Med, 3(8), a014217. doi: 10.1101/cshperspect.a014217 DeBerardinis, R. J., Lum, J. J., Hatzivassiliou, G., & Thompson, C. B. (2008). The biology of cancer: metabolic reprogramming fuels cell growth and proliferation. Cell Metab, 7(1), 11-20. doi: 10.1016/j.cmet.2007.10.002 Dejure, F. R., & Eilers, M. (2017). MYC and tumor metabolism: chicken and egg. Embo j, 36(23), 3409-3420. doi: 10.15252/embj.201796438 Des Rosiers, C., Di Donato, L., Comte, B., Laplante, A., Marcoux, C., David, F., . . . Brunengraber, H. (1995). Isotopomer analysis of citric acid cycle and gluconeogenesis in rat liver. Reversibility of isocitrate dehydrogenase and involvement of ATP-citrate lyase in gluconeogenesis. J Biol Chem, 270(17), 10027-10036. Ding, J., Li, T., Wang, X., Zhao, E., Choi, J. H., Yang, L., . . . Ding, H. F. (2013). The histone H3 methyltransferase G9A epigenetically activates the serine-glycine synthesis pathway to sustain cancer cell survival and proliferation. Cell Metab, 18(6), 896-907. doi: 10.1016/j.cmet.2013.11.004 Ducker, G. S., Chen, L., Morscher, R. J., Ghergurovich, J. M., Esposito, M., Teng, X., . . . Rabinowitz, J. D. (2016). Reversal of Cytosolic One-Carbon Flux Compensates for Loss of the Mitochondrial Folate Pathway. Elife, 23(6), 1140-1153. doi: 10.7554/eLife.1057510.1016/j.cmet.2016.04.016 Ducker, G. S., Ghergurovich, J. M., Mainolfi, N., Suri, V., Jeong, S. K., Hsin-Jung Li, S., . . . Rabinowitz, J. D. (2017). Human SHMT inhibitors reveal defective glycine import as a targetable metabolic vulnerability of diffuse large B-cell lymphoma. Proc Natl Acad Sci U S A, 114(43), 11404-11409. doi: 10.1073/pnas.1706617114 Fan, J., Ye, J., Kamphorst, J. J., Shlomi, T., Thompson, C. B., & Rabinowitz, J. D. (2014). Quantitative flux analysis reveals folate-dependent NADPH production. Nature, 510(7504), 298-302. doi: 10.1038/nature13236 Farber, S., Cutler, E. C., Hawkins, J. W., Harrison, J. H., Peirce, E. C., 2nd, & Lenz, G. G. (1947). The Action of Pteroylglutamic Conjugates on Man. Science, 106(2764), 619-621. doi: 10.1126/science.106.2764.619

Farber, S., & Diamond, L. K. (1948). Temporary remissions in acute leukemia in children produced by folic acid antagonist, 4-aminopteroyl-glutamic acid. N Engl J Med, 238(23), 787-793. doi: 10.1056/nejm194806032382301 Fiehn, O., Wohlgemuth, G., Scholz, M., Kind, T., Lee, D. Y., Lu, Y., . . . Nikolau, B. (2008). Quality control for plant metabolomics: reporting MSI-compliant studies. Plant J, 53(4), 691-704. doi: 10.1111/j.1365-313X.2007.03387.x Harayama, T., & Riezman, H. (2018). Understanding the diversity of membrane lipid composition. Nat Rev Mol Cell Biol, 19(5), 281-296. doi: 10.1038/nrm.2017.138 Heng-Maillard, M. A., Verschuur, A., Aschero, A., Dabadie, A., Jouve, E., Chastagner, P., . . . Andre, N. (2019). SFCE METRO-01 four-drug metronomic regimen phase II trial for pediatric extracranial tumor. Pediatr Blood Cancer, e27693. doi: 10.1002/pbc.27693 Hogarty, M. D., Norris, M. D., Davis, K., Liu, X., Evageliou, N. F., Hayes, C. S., . . . Haber, M. (2008). ODC1 is a critical determinant of MYCN oncogenesis and a therapeutic target in neuroblastoma. Cancer Res, 68(23), 9735-9745. doi: 10.1158/0008-5472.CAN-07-6866 Holleran, A. L., Briscoe, D. A., Fiskum, G., & Kelleher, J. K. (1995). Glutamine metabolism in AS-30D hepatoma cells. Evidence for its conversion into lipids via reductive carboxylation. Mol Cell Biochem, 152(2), 95-101. Huang da, W., Sherman, B. T., & Lempicki, R. A. (2009a). Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res, 37(1), 1-13. doi: 10.1093/nar/gkn923 Huang da, W., Sherman, B. T., & Lempicki, R. A. (2009b). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc, 4(1), 44-57. doi: 10.1038/nprot.2008.211 Huang, D. W., Sherman, B. T., & Lempicki, R. A. (2008). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protocols, 4(1), 44-57. doi: http://www.nature.com/nprot/journal/v4/n1/suppinfo/nprot.2008.211_S1.html Huangyang, P., & Simon, M. C. (2018). Hidden features: exploring the non-canonical functions of metabolic enzymes. Dis Model Mech, 11(8). doi: 10.1242/dmm.033365 Jackman, A. L., & Calvert, A. H. (1995). Folate-based thymidylate synthase inhibitors as anticancer drugs. Ann Oncol, 6(9), 871-881. doi: 10.1093/oxfordjournals.annonc.a059353 Jain, M., Nilsson, R., Sharma, S., Madhusudhan, N., Kitami, T., Souza, A. L., . . . Mootha, V. K. (2012). Metabolite profiling identifies a key role for glycine in rapid cancer cell proliferation. Science, 336(6084), 1040-1044. doi: 10.1126/science.1218595 Janoueix-Lerosey, I., Lequin, D., Brugieres, L., Ribeiro, A., de Pontual, L., Combaret, V., . . . Delattre, O. (2008). Somatic and germline activating mutations of the ALK kinase receptor in neuroblastoma. Nature, 455(7215), 967-970. doi: 10.1038/nature07398 Kanehisa, M., & Goto, S. (2000). KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res, 28(1), 27-30.

Khanna, C., Jaboin, J. J., Drakos, E., Tsokos, M., & Thiele, C. J. (2002). Biologically relevant orthotopic neuroblastoma xenograft models: primary adrenal tumor growth and spontaneous distant metastasis. In Vivo, 16(2), 77-85. Kikuchi, G., Motokawa, Y., Yoshida, T., & Hiraga, K. (2008). Glycine cleavage system: reaction mechanism, physiological significance, and hyperglycinemia. Proc Jpn Acad Ser B Phys Biol Sci, 84(7), 246-263. Kim, D., Fiske, B. P., Birsoy, K., Freinkman, E., Kami, K., Possemato, R. L., . . . Sabatini, D. M. (2015). SHMT2 drives glioma cell survival in ischaemia but imposes a dependence on glycine clearance. Nature, 520(7547), 363-367. doi: 10.1038/nature14363 Kloudova, A., Guengerich, F. P., & Soucek, P. (2017). The Role of Oxysterols in Human Cancer. Trends Endocrinol Metab, 28(7), 485-496. doi: 10.1016/j.tem.2017.03.002 Kocak, H., Ackermann, S., Hero, B., Kahlert, Y., Oberthuer, A., Juraeva, D., . . . Fischer, M. (2013). Hox-C9 activates the intrinsic pathway of apoptosis and is associated with spontaneous regression in neuroblastoma. Cell Death Dis, 4, e586. doi: 10.1038/cddis.2013.84 Kohl, N. E., Kanda, N., Schreck, R. R., Bruns, G., Latt, S. A., Gilbert, F., & Alt, F. W. (1983). Transposition and amplification of oncogene-related sequences in human neuroblastomas. Cell, 35(2 Pt 1), 359-367. Kure, S., Tada, K., & Narisawa, K. (1997). Nonketotic hyperglycinemia: biochemical, molecular, and neurological aspects. Jpn J Hum Genet, 42(1), 13-22. doi: 10.1007/bf02766917 Labuschagne, C. F., van den Broek, N. J., Mackay, G. M., Vousden, K. H., & Maddocks, O. D. (2014). Serine, but not glycine, supports one-carbon metabolism and proliferation of cancer cells. Cell Rep, 7(4), 1248-1258. doi: 10.1016/j.celrep.2014.04.045 Lau, D. T., Flemming, C. L., Gherardi, S., Perini, G., Oberthuer, A., Fischer, M., . . . Ashton, L. J. (2015). MYCN amplification confers enhanced folate dependence and methotrexate sensitivity in neuroblastoma. Oncotarget, 6(17), 15510-15523. doi: 10.18632/oncotarget.3732 Leung, K. Y., Pai, Y. J., Chen, Q., Santos, C., Calvani, E., Sudiwala, S., . . . Greene, N. D. E. (2017). Partitioning of One-Carbon Units in Folate and Methionine Metabolism Is Essential for Neural Tube Closure. Cell Rep, 21(7), 1795-1808. doi: 10.1016/j.celrep.2017.10.072 Liberti, M. V., & Locasale, J. W. (2016). The Warburg Effect: How Does it Benefit Cancer Cells? Trends Biochem Sci, 41(3), 211-218. doi: 10.1016/j.tibs.2015.12.001 Liu, M., Xia, Y., Ding, J., Ye, B., Zhao, E., Choi, J. H., . . . Ding, H. F. (2016). Transcriptional Profiling Reveals a Common Metabolic Program in High-Risk Human Neuroblastoma and Mouse Neuroblastoma Sphere-Forming Cells. Cell Rep, 17(2), 609-623. doi: 10.1016/j.celrep.2016.09.021 Locasale, J. W., Grassian, A. R., Melman, T., Lyssiotis, C. A., Mattaini, K. R., Bass, A. J., . . . Vander Heiden, M. G. (2011). Phosphoglycerate dehydrogenase diverts glycolytic flux and contributes to oncogenesis. Nat Genet, 43(9), 869-874. doi: 10.1038/ng.890

London, W. B., Castel, V., Monclair, T., Ambros, P. F., Pearson, A. D., Cohn, S. L., . . . Matthay, K. K. (2011). Clinical and biologic features predictive of survival after relapse of neuroblastoma: a report from the International Neuroblastoma Risk Group project. J Clin Oncol, 29(24), 3286-3292. doi: 10.1200/jco.2010.34.3392 Louis, C. U., & Shohet, J. M. (2015). Neuroblastoma: molecular pathogenesis and therapy. Annu Rev Med, 66, 49-63. doi: 10.1146/annurev-med-011514-023121 Maddocks, O. D., Berkers, C. R., Mason, S. M., Zheng, L., Blyth, K., Gottlieb, E., & Vousden, K. H. (2013). Serine starvation induces stress and p53-dependent metabolic remodelling in cancer cells. Nature, 493(7433), 542-546. doi: 10.1038/nature11743 Maddocks, O. D. K., Athineos, D., Cheung, E. C., Lee, P., Zhang, T., van den Broek, N. J. F., . . . Vousden, K. H. (2017). Modulating the therapeutic response of tumours to dietary serine and glycine starvation. Nature, 544(7650), 372-376. doi: 10.1038/nature22056 Maris, J. M., Hogarty, M. D., Bagatell, R., & Cohn, S. L. (2007). Neuroblastoma. Lancet, 369(9579), 2106-2120. doi: 10.1016/S0140-6736(07)60983-0 Matthay, K. K., Maris, J. M., Schleiermacher, G., Nakagawara, A., Mackall, C. L., Diller, L., & Weiss, W. A. (2016). Neuroblastoma. Nat Rev Dis Primers, 2, 16078. doi: 10.1038/nrdp.2016.78 Metallo, C. M., Gameiro, P. A., Bell, E. L., Mattaini, K. R., Yang, J., Hiller, K., . . . Stephanopoulos, G. (2012). Reductive glutamine metabolism by IDH1 mediates lipogenesis under hypoxia. Nature, 481(7381), 380-384. doi: 10.1038/nature10602 Molenaar, J. J., Koster, J., Zwijnenburg, D. A., van Sluis, P., Valentijn, L. J., van der Ploeg, I., . . . Versteeg, R. (2012). Sequencing of neuroblastoma identifies chromothripsis and defects in neuritogenesis genes. Nature, 483(7391), 589-593. doi: 10.1038/nature10910 Murphy, D. M., Buckley, P. G., Bryan, K., Das, S., Alcock, L., Foley, N. H., . . . Stallings, R. L. (2009). Global MYCN transcription factor binding analysis in neuroblastoma reveals association with distinct E-box motifs and regions of DNA hypermethylation. PLoS One, 4(12), e8154. doi: 10.1371/journal.pone.0008154 Narisawa, A., Komatsuzaki, S., Kikuchi, A., Niihori, T., Aoki, Y., Fujiwara, K., . . . Kure, S. (2012). Mutations in genes encoding the glycine cleavage system predispose to neural tube defects in mice and humans. Hum Mol Genet, 21(7), 1496-1503. doi: 10.1093/hmg/ddr585 Nelson, E. R., Wardell, S. E., Jasper, J. S., Park, S., Suchindran, S., Howe, M. K., . . . McDonnell, D. P. (2013). 27-Hydroxycholesterol links hypercholesterolemia and breast cancer pathophysiology. Science, 342(6162), 1094-1098. doi: 10.1126/science.1241908 Newsholme, E. A., Crabtree, B., & Ardawi, M. S. (1985). The role of high rates of glycolysis and glutamine utilization in rapidly dividing cells. Biosci Rep, 5(5), 393-400. Nilsson, R., Jain, M., Madhusudhan, N., Sheppard, N. G., Strittmatter, L., Kampf, C., . . . Mootha, V. K. (2014). Metabolic enzyme expression highlights a key role for MTHFD2 and the mitochondrial folate pathway in cancer. Nat Commun, 5, 3128. doi: 10.1038/ncomms4128

Otto, A. M. (2016). Warburg effect(s)-a biographical sketch of Otto Warburg and his impacts on tumor metabolism. Cancer Metab, 4, 5. doi: 10.1186/s40170-016- 0145-9 Pacold, M. E., Brimacombe, K. R., Chan, S. H., Rohde, J. M., Lewis, C. A., Swier, L. J., . . . Sabatini, D. M. (2016). A PHGDH inhibitor reveals coordination of serine synthesis and one-carbon unit fate. Nat Chem Biol, 12(6), 452-458. doi: 10.1038/nchembio.2070 Pai, Y. J., Leung, K. Y., Savery, D., Hutchin, T., Prunty, H., Heales, S., . . . Greene, N. D. (2015). Glycine decarboxylase deficiency causes neural tube defects and features of non-ketotic hyperglycinemia in mice. Nat Commun, 6, 6388. doi: 10.1038/ncomms7388 Pajic, A., Spitkovsky, D., Christoph, B., Kempkes, B., Schuhmacher, M., Staege, M. S., . . . Eick, D. (2000). Cell cycle activation by c-myc in a burkitt lymphoma model cell line. Int J Cancer, 87(6), 787-793. Pinto, N. R., Applebaum, M. A., Volchenboum, S. L., Matthay, K. K., London, W. B., Ambros, P. F., . . . Cohn, S. L. (2015). Advances in Risk Classification and Treatment Strategies for Neuroblastoma. J Clin Oncol, 33(27), 3008-3017. doi: 10.1200/JCO.2014.59.4648 Possemato, R., Marks, K. M., Shaul, Y. D., Pacold, M. E., Kim, D., Birsoy, K., . . . Sabatini, D. M. (2011). Functional genomics reveal that the serine synthesis pathway is essential in breast cancer. Nature, 476(7360), 346-350. doi: 10.1038/nature10350 Puissant, A., Frumm, S. M., Alexe, G., Bassil, C. F., Qi, J., Chanthery, Y. H., . . . Stegmaier, K. (2013). Targeting MYCN in neuroblastoma by BET bromodomain inhibition. Cancer Discov, 3(3), 308-323. doi: 10.1158/2159-8290.CD-12-0418 Reid, M. A., Allen, A. E., Liu, S., Liberti, M. V., Liu, P., Liu, X., . . . Locasale, J. W. (2018). Serine synthesis through PHGDH coordinates nucleotide levels by maintaining central carbon metabolism. Nat Commun, 9(1), 5442. doi: 10.1038/s41467-018-07868-6 Ruiz-Perez, M. V., Henley, A. B., & Arsenian-Henriksson, M. (2017). The MYCN Protein in Health and Disease. Genes (Basel), 8(4). doi: 10.3390/genes8040113 Samanta, D., Park, Y., Andrabi, S. A., Shelton, L. M., Gilkes, D. M., & Semenza, G. L. (2016). PHGDH Expression Is Required for Mitochondrial Redox Homeostasis, Breast Cancer Stem Cell Maintenance, and Lung Metastasis. Cancer Res, 76(15), 4430-4442. doi: 10.1158/0008-5472.CAN-16-0530 Schuhmacher, M., Staege, M. S., Pajic, A., Polack, A., Weidle, U. H., Bornkamm, G. W., . . . Kohlhuber, F. (1999). Control of cell growth by c-Myc in the absence of cell division. Curr Biol, 9(21), 1255-1258. Schwab, M., Alitalo, K., Klempnauer, K. H., Varmus, H. E., Bishop, J. M., Gilbert, F., . . . Trent, J. (1983). Amplified DNA with limited homology to myc cellular oncogene is shared by human neuroblastoma cell lines and a neuroblastoma tumour. Nature, 305(5931), 245-248. Schwab, M., Ellison, J., Busch, M., Rosenau, W., Varmus, H. E., & Bishop, J. M. (1984). Enhanced expression of the human gene N-myc consequent to amplification of DNA may contribute to malignant progression of neuroblastoma. Proc Natl Acad Sci U S A, 81(15), 4940-4944.

Sholler, G. L. S., Ferguson, W., Bergendahl, G., Bond, J. P., Neville, K., Eslin, D., . . . Kraveka, J. M. (2018). Maintenance DFMO Increases Survival in High Risk Neuroblastoma. Sci Rep, 8(1), 14445. doi: 10.1038/s41598-018-32659-w Silvente-Poirot, S., Dalenc, F., & Poirot, M. (2018). The Effects of Cholesterol-Derived Oncometabolites on Nuclear Receptor Function in Cancer. Cancer Res, 78(17), 4803-4808. doi: 10.1158/0008-5472.CAN-18-1487 Snell, K., Natsumeda, Y., Eble, J. N., Glover, J. L., & Weber, G. (1988). Enzymic imbalance in serine metabolism in human colon carcinoma and rat sarcoma. Br J Cancer, 57(1), 87-90. Snell, K., Natsumeda, Y., & Weber, G. (1987). The modulation of serine metabolism in hepatoma 3924A during different phases of cellular proliferation in culture. Biochem J, 245(2), 609-612. Snell, K., & Weber, G. (1986). Enzymic imbalance in serine metabolism in rat hepatomas. Biochem J, 233(2), 617-620. Stanton, B. R., Perkins, A. S., Tessarollo, L., Sassoon, D. A., & Parada, L. F. (1992). Loss of N-myc function results in embryonic lethality and failure of the epithelial component of the embryo to develop. Genes Dev, 6(12A), 2235-2247. Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., . . . Mesirov, J. P. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 102(43), 15545-15550. doi: 10.1073/pnas.0506580102 Valentijn, L. J., Koster, J., Haneveld, F., Aissa, R. A., van Sluis, P., Broekmans, M. E., . . . Versteeg, R. (2012). Functional MYCN signature predicts outcome of neuroblastoma irrespective of MYCN amplification. Proc Natl Acad Sci U S A, 109(47), 19190-19195. doi: 10.1073/pnas.1208215109 Wakamatsu, Y., Watanabe, Y., Nakamura, H., & Kondoh, H. (1997). Regulation of the neural crest cell fate by N-myc: promotion of ventral migration and neuronal differentiation. Development, 124(10), 1953-1962. Warwick, A. B., Malempati, S., Krailo, M., Melemed, A., Gorlick, R., Ames, M. M., . . . Blaney, S. M. (2013). Phase 2 trial of pemetrexed in children and adolescents with refractory solid tumors: a Children's Oncology Group study. Pediatr Blood Cancer, 60(2), 237-241. doi: 10.1002/pbc.24244 Weiss, W. A., Aldape, K., Mohapatra, G., Feuerstein, B. G., & Bishop, J. M. (1997). Targeted expression of MYCN causes neuroblastoma in transgenic mice. Embo j, 16(11), 2985-2995. doi: 10.1093/emboj/16.11.2985 Whittle, S. B., Smith, V., Doherty, E., Zhao, S., McCarty, S., & Zage, P. E. (2017). Overview and recent advances in the treatment of neuroblastoma. Expert Rev Anticancer Ther, 17(4), 369-386. doi: 10.1080/14737140.2017.1285230 Wiederschain, D., Wee, S., Chen, L., Loo, A., Yang, G., Huang, A., . . . Benson, J. D. (2009). Single-vector inducible lentiviral RNAi system for oncology target validation. Cell Cycle, 8(3), 498-504. doi: 10.4161/cc.8.3.7701 Xia, Y., Ye, B., Ding, J., Yu, Y., Alptekin, A., Thangaraju, M., . . . Fiehn, O. (2019). Metabolic reprogramming by MYCN confers dependence on the serine-glycine- one-carbon biosynthetic pathway. doi: 10.1158/0008-5472.can-18-3541 Yang, M., & Vousden, K. H. (2016). Serine and one-carbon metabolism in cancer. Nat Rev Cancer, 16(10), 650-662. doi: 10.1038/nrc.2016.81

Ye, J., Fan, J., Venneti, S., Wan, Y. W., Pawel, B. R., Zhang, J., . . . Thompson, C. B. (2014). Serine catabolism regulates mitochondrial redox control during hypoxia. Cancer Discov, 4(12), 1406-1417. doi: 10.1158/2159-8290.CD-14-0250 Yoo, H., Antoniewicz, M. R., Stephanopoulos, G., & Kelleher, J. K. (2008). Quantifying reductive carboxylation flux of glutamine to lipid in a brown adipocyte cell line. J Biol Chem, 283(30), 20621-20627. doi: 10.1074/jbc.M706494200 Zhang, J., Pavlova, N. N., & Thompson, C. B. (2017). Cancer cell metabolism: the essential role of the nonessential amino acid, glutamine. Embo j, 36(10), 1302- 1315. doi: 10.15252/embj.201696151 Zhang, W., Yu, Y., Hertwig, F., Thierry-Mieg, J., Zhang, W., Thierry-Mieg, D., . . . Fischer, M. (2015). Comparison of RNA-seq and microarray-based models for clinical endpoint prediction. Genome Biol, 16, 133. doi: 10.1186/s13059-015- 0694-1 Zhang, W. C., Shyh-Chang, N., Yang, H., Rai, A., Umashankar, S., Ma, S., . . . Lim, B. (2012). Glycine decarboxylase activity drives non-small cell lung cancer tumor- initiating cells and tumorigenesis. Cell, 148(1-2), 259-272. doi: 10.1016/j.cell.2011.11.050 Zheng, Y., Lin, T. Y., Lee, G., Paddock, M. N., Momb, J., Cheng, Z., . . . Cantley, L. C. (2018). Mitochondrial One-Carbon Pathway Supports Cytosolic Folate Integrity in Cancer Cells. Cell, 175(6), 1546-1560 e1517. doi: 10.1016/j.cell.2018.09.041

APPENDIX A: ABBREVIATIONS

10-CHO-THF Formyl THF

3-PG 3-phosphoglycerate

5,10-CH2-THF Methylene THF a-KG Alpha-ketoglutarate

ALDH1L2 Aldehyde dehydrogenase 1 family member l 2

AMT Aminomethyltransferase

ATF5 Activating transcription factor 5

CH-THF Methenyl THF

DLD Dihydrolipoamide dehydrogenase

Dox Doxycycline

GCS Glycine cleavage system

GCSH Glycine cleavage system h protein

GLDC Glycine decarboxylase

GSH Reduced glutathione

GSSG Oxidized glutathione

KEGG Kyoto Encyclopedia of Genes and Genomes

KLF15 Kruppel like factor 15

MTHFD1 Methylenetetrahydrofolate dehydrogenase

MTHFD1L Methylenetetrahydrofolate dehydrogenase 1 like

MTHFD2 Methylenetetrahydrofolate dehydrogenase 2 (mitochondrial)

MTHFD2L Methylenetetrahydrofolate dehydrogenase 2 like

NAC N-acetyl-l-cysteine

NAD+ Nicotinamide adenine dinucleotide (oxidized)

NADH Nicotinamide adenine dinucleotide (reduced)

NADP+ Nicotinamide adenine dinucleotide phosphate (oxidized)

NADPH Nicotinamide adenine dinucleotide phosphate (reduced)

NH3 Ammonia

NKH Non-ketotic hyperglycinemia

NTD Neural tube defect

OAA Oxaloacetate

PEP Phosphoenolpyruvate

PHGDH Phosphoglycerate dehydrogenase

PSAT1 Phosphoserine aminotransferase

PSPH Phosphoserine phosphatase

ROS Reactive oxygen species

SEM Standard error of mean

SGOC Serine-glycine-one-carbon

SHMT1 Serine hydroxymethyltransferase 1 (cytoplasmic)

SHMT2 Serine hydroxymethyltransferase 2 (mitochondrial)

THF Tetrahydrofolate

ZNF263 Zinc finger protein 263