Structural and Functional Studies of Proteins of Medical Relevance

! "# $ #%## & ' ( ) * ( + !%

, %- % . ( ( / % . ( % (/ %0 / / ( % &01 ( / / 2 3 %. (/ %- -/ &01 %. &01 / &01 ( %. / (/ (/ &01 ( / --%- --- &01 / &01 %4 !&01 / / / 5% &01 ( / -6%0 / / %- / &01 ( 6/ &01 % &01 " %&01 " ( 5/ % 6- &01 "( / %0 &01 " % 0 %0 ( / %1/( / ( %- 6--/ (78 9,"( (/ %0 ( %0 / %0/ / (/ %

!"!" #

"# $ :;; %%; < = : ::: >#!$$

-2?8$? 88?8#?8" -2?8$? 88?8#?$?

( #!?

STRUCTURAL AND FUNCTIONAL STUDIES OF PROTEINS OF MEDICAL RELEVANCE

Robert Gustafsson

Structural and functional studies of proteins of medical relevance

Protein-ligand complexes in cancer and novel structural folds in bacteria

Robert Gustafsson ©Robert Gustafsson, Stockholm University 2018

ISBN print 978-91-7797-097-2 ISBN PDF 978-91-7797-098-9

Cover picture by Robert Gustafsson and Elisabeth Westergren

Printed in Sweden by Universitetsservice US-AB, Stockholm 2018 Distributor: Department of Biochemistry and Biophysics, Stockholm University To my family, and Elisabeth, with all my love

List of Publications

The following publications and manuscripts are included in this thesis and are referred to by their roman numerals in the text.

Paper I: MTH1 inhibition eradicates cancer by preventing sanitation of the dNTP pool Gad, H.*, Koolmeister, T.*, Jemth, A.-S.*, Eshtad, S.*, Jacques, S. A.*, Ström, C. E.*, Svensson, L. M., Schultz, N., Lundbäck, T., Einarsdottir, B. O. , Saleh, A., Göktürk, C., Baranczewski, P., Svensson, R., Bernts- son, R. P.-A., Gustafsson, R., Strömberg, K., Sanjiv, K., Jacques- Cordonnier, M.-C., Desroses, M., Gustavsson, A.-L., Olofsson, R., Jo- hansson, F., Homan, E. J., Loseva, O., Bräutigam, L., Johansson, L., Höglund, A., Hagenkort, A., Pham, T., Altun, M., Gaugaz, F. Z., Vi- kingsson, S., Evers, B., Henriksson, M., Vallin, K. S. A., Wallner, O. A., Hammarström, L. G. J., Wiita, E., Almlöf, I., Kalderén, C., Axelsson, H., Djureinovic, T., Carreras Puigvert, J., Häggblad, M., Jeppsson, F., Mar- tens, U., Lundin, C., Lundgren, B., Granelli, I., Jenmalm Jensen, A., Ar- tursson, P., Nilsson, J. A., Stenmark, P., Scobie, M., Warpman Berglund, U., Helleday, T. Nature 508(7495), 215-221 (2014). DOI: 10.1038/nature13181

Paper II: Hypoxic signaling and the cellular redox tumor environment determine sensitivity to MTH1 inhibition Bräutigam, L., Pudelko, L., Jemth, A.-S., Gad, H., Narwal, M., Gus- tafsson, R., Karsten, S., Carreras Puigvert, J., Homan, E. J., Berndt, C., Warpman Berglund, U., Stenmark, P., Helleday, T. Cancer research 76(8), 2366-2375 (2016). DOI: 10.1158/0008-5472.CAN-15-2380

Paper III: Crystal structures and inhibitor interactions of mouse and dog MTH1 reveal species-specific differences in affinity Narwal, M.*, Jemth, A.-S.*, Gustafsson, R., Almlöf, I., Warpman Ber- glund, U., Helleday, T., Stenmark, P. Biochemistry, in press (2018). DOI: 10.1021/acs.biochem.7b01163 Paper IV: Fragment-based discovery and optimization of enzyme inhibitors by docking of commercial chemical space Rudling, A.*, Gustafsson, R.*, Almlöf, I., Homan, E. J., Scobie, M., Warpman Berglund, U., Helleday, T., Stenmark, P., Carlsson, J. Journal of Medicinal Chemistry 60(19), 8160-8169 (2017). DOI: 10.1021/acs.jmedchem.7b01006

Paper V: MutT homologue 1 (MTH1) catalyzes the hydrolysis of mutagenic O6-methyl-dGTP Jemth, A.-S., Gustafsson, R., Bräutigam, L., Henriksson, L. M., Carreras Puigvert, J., Homan, E. J., Warpman Berglund, U., Stenmark, P., Helleday, T. Manuscript

Paper VI: Crystal structure of the emerging cancer target MTHFD2 in complex with a substrate-based inhibitor Gustafsson, R., Jemth, A.-S., Gustafsson, N. M. S., Färnegårdh, K., Lo- seva, O., Wiita, E., Bonagas, N. Dahllund, L. Llona-Minguez, S., Hägg- blad, M., Henriksson, M., Andersson, Y., Homan, E. J., Helleday, T., Stenmark, P. Cancer Research 77(4), 937-948 (2017). DOI: 10.1158/0008-5472.CAN-16-1476

Paper VII: Crystal structures of OrfX2 and P47 from a Botulinum neurotoxin OrfX-type gene cluster Gustafsson, R.*, Berntsson, R. P.-A.*, Martínez-Carranza, M., El Tekle, G., Odegrip, R., Johnson, E. A., Stenmark, P. FEBS Letters 591(22), 3781-3792 (2017). DOI: 10.1002/1873-3468.12889

Reprints of the published articles were made with permission from the pub- lishers.

* These authors contributed equally to this work.

Additional Publications

Crystal structure, biochemical and cellular activities demonstrate sepa- rate functions of MTH1 and MTH2 Carter, M., Jemth, A.-S., Hagenkort, A., Page, B. D. G., Gustafsson, R., Griese, J. J., Gad, H., Valerie, N. C. K., Desroses, M., Boström, J., Warp- man Berglund, U., Helleday, T., Stenmark, P. Nature Communications 6, 7871 (2015). DOI: 10.1038/ncomms8871

Glycans confer specificity to the recognition of ganglioside receptors by botulinum neurotoxin A Hamark, C., Berntsson, R. P.-A., Masuyer, G., Henriksson, L. M., Gus- tafsson, R., Stenmark, P., Widmalm, G. Journal of the American Chemical Society 139(1), 218-230 (2017). DOI: 10.1021/jacs.6b09534

Contents

Introduction ...... 1 Cancer ...... 2 Clostridium botulinum neurotoxin ...... 3

Cancer and the cancer cell ...... 5 Somatic mutation theory ...... 6 Critiques of the somatic mutation theory ...... 7 Genetic instability and general cancer therapy ...... 7 Synthetic lethality ...... 9 Redox regulation and reactive oxygen species ...... 10 Generation of reactive oxygen species ...... 11 Reactive oxygen species in health and disease ...... 11 Hypoxia and reactive oxygen species ...... 12 Oxidised nucleotides ...... 13 Oxidative damage to nuclear and mitochondrial DNA leads to cell death ...... 14 Methylated nucleotides ...... 15

MutT homologue 1 (MTH1) ...... 17 The MutT protein from E. coli degrades 8-oxo-dGTP ...... 17 MTH1 can degrade several oxidised nucleotides ...... 17 Structures of MTH1 show a versatile binding site ...... 18 mth1 gene and isoforms ...... 19 MTH1 is part of the NUDIX hydrolase protein family ...... 20 MTH1 prevents damage caused by ROS ...... 21 Cancer phenotypic lethality: MTH1 as a drug target for cancer ...... 22 MTH1 controversy ...... 24 New MTH1 function: degradation of O6-methyl-dGTP ...... 24

De novo nucleotide synthesis and one-carbon metabolism ...... 25 One-carbon metabolism ...... 26 Methylene-tetrahydrofolate dehydrogenase family ...... 27 Outcomes from the one-carbon metabolism ...... 29 Drugs for one-carbon metabolism: antifolates and antimetabolites ...... 30 Importance of transport and polyglutamation of THF and drugs ...... 31 MTHFD2 is upregulated in cancers and is a potential target for cancer phenotypic lethality ...... 32 Structural information on the MTHFD family ...... 33

Clostridium botulinum ...... 35 Botulism ...... 36 Botox and medical use of botulinum neurotoxins ...... 36 Botulinum neurotoxin and mechanism of action ...... 37 BoNT gene clusters ...... 39 HA cluster ...... 40 OrfX cluster ...... 41 TULIP proteins ...... 41

Present investigations ...... 43 Paper I – MTH1 validated as a drug target for cancer and first-in-class drug candidates presented ...... 43 Paper II – Hypoxia and oxidative stress makes cells sensitive to MTH1 inhibition ..... 44 Paper III – Differences in MTH1 affinity between species and suggestions for pre- clinical studies ...... 46 Paper IV – Virtual fragment screening for MTH1 supported by X-ray crystal structures ...... 47 Paper V – MTH1 can degrade O6-methyl-dGTP ...... 48 Paper VI – The first MTHFD2 structure and inhibitor ...... 49 Paper VII – Proteins from the BoNT OrfX gene cluster are TULIP proteins which bind lipids ...... 49

Conclusion and outlook ...... 51

Popular science summary ...... 53

Populärvetenskaplig sammanfattning ...... 57

Acknowledgements ...... 61

References ...... 65

Abbreviations

2-OH-A 2-hydroxyadenine 2-OH-dATP 2-hydroxy-2’-deoxyadenosine triphosphate 8-oxo-dGTP 8-oxo-2’-deoxyguanosine triphosphate 8-oxoG 8-oxoguanine BoNT Botulinum neurotoxin BPI Bactericidal/permeability-increasing protein BRCA1/BRCA2 Breast cancer susceptibility gene 1 and 2 DC-domain Dehydrogenase and cyclohydrolase domain (of MTHFD1) DMOG Dimethyloxallyl glycine DNA Deoxyribonucleic acid ER Endoplasmatic reticulum Formyl-THF 10-formyl-tetrahydrofolate FPGS Folylpolyglutamate synthetase GSH Glutathione (reduced) GSSG Glutathione (oxidised) HA Haemagglutinin HIF-1 Hypoxia-induced factor 1 KRAS / HRAS Members of the Ras division of RAS superfamily L-PTC Large progenitor toxin complex M-PTC Minimal progenitor toxin complex Methenyl-THF 5,10-methenyl-tetrahydrofolate Methylene-THF 5,10-methylene-tetrahydrofolate MGMT O6-methylguanine-DNA methyltransferase mRNA Messenger ribonucleic acid MTH1 MutT homologue 1 MTHFD family Methylene-tetrahydrofolate dehydrogenase family MTHFD2 Methylene-tetrahydrofolate dehydrogenase and cyclohydrolase 2 MUTYH MutY homologue NAD+ Nicotinamide adenine dinucleotide (oxidised) NADH Nicotinamide adenine dinucleotide (reduced) NADP+ Nicotinamide adenine dinucleotide phosphate (oxidised) NADPH Nicotinamide adenine dinucleotide phosphate (reduced)

NTNH Non-toxic non-haemagglutinin protein O6-methyl-dGTP O6-methyl-2’-deoxyguanosine triphosphate OGG1 8-oxoG DNA-glycosylase OrfX Open-reading-frame-X PARP Poly (adenosine diphosphate [ADP]-ribose) polymerase Pi Inorganic phosphate PTC Progenitor toxin complex RAS Protein superfamily of small GTPases involved in signalling Ras Subfamily within the RAS superfamily of small GTPases RFC Reduced folate carrier RNA Ribonucleic acid ROS Reactive oxygen species SAM S-adenosyl-methionine SHMT Serine hydroxymethyltransferase shRNA Short hairpin ribonucleic acid siRNA Small interfering ribonucleic acid SMT Somatic mutation theory SNAP25 Synaptosomal-associated protein of 25 kDa SNARE Soluble N-ethylmaleimide-sensitive factor attachment protein receptor SV2 Synaptic vesicle protein 2 Syt I/II Synaptotagmin I/II THF Tetrahydrofolate TOFT Tissue organization field theory TULIP Tubular lipid-binding protein VAMP Vesicle-associated membrane protein VHL von Hippel-Lindau tumour suppressor protein

Introduction

“If we knew what we were doing, it would not be called research, would it?” – Albert Einstein

Proteins are the molecular machines inside every living organism, cell or virus. Built from a combination of only twenty amino acids, they produce an astonishingly large variety of proteins with different sizes, functions and structures. Some proteins are enzymes, catalysing chemical reactions that would otherwise be unfavourable. Others provide a scaffold upon which more proteins or other cellular components can be assembled and anchored. Another group includes receptors and transporters that ensure the cell re- ceives information and nutrients from its surroundings. For any of these proteins to perform their function, their amino acid building blocks need to be arranged in a way that is specific for that protein. This is the protein structure. The arrangement of certain amino acids in an active site will allow the protein to perform a certain function, for example, the hydrolysis of a covalent bond using water. Often the same arrangement of part of the protein structure is found in several proteins and they can be grouped into a family. Members of a protein family have similar functions but can differ in substrate preference or cellular location. The protein primary structure is the order of amino acids in the protein. The secondary structure is the local folding of the amino acid chain to form α-helices, β-strands and loops. The tertiary structure is the folding of the different secondary structure elements that are held together by interactions between parts that are distant in the primary structure. If more than one protein amino acid chain is present in the structure this is referred to as a complex of protein monomers, or oligomer. Proteins are far too small to be visualised using an ordinary light micro- scope and so alternative methods must be employed to determine their structures. To see the proteins in their full atomic detail in three dimensions X- ray crystallography is used, and this method was used for the projects of this thesis. The field of X-ray crystallography recently celebrated 100 years, although protein X-ray crystallography is a considerably younger technique. Crystals of purified protein are prepared and a single crystal is placed in front of an intense X-ray beam, resulting in a pattern of diffracted X-ray light. The patterns are recorded while the crystal is rotated, yielding a series

1 of images. The images undergo extensive computer processing to produce a map of the electrons in the crystal. This is because electrons are responsible for the diffraction of the X-ray light. The map of electrons, called an electron density map, can be used to build the structure of the protein. Ligands, inhibitors or substrates for the protein can be included in the crystallisation mixture, and if these molecules form strong interactions to the protein their binding modes can be seen. X-ray crystal structures can aid the optimisation of a drug binding to its target protein. This method is called structure-based drug design.

Cancer Cancer and cancer therapy are important and interesting study subjects. Ap- proximately one in three people in Sweden will receive a cancer diagnosis during their lifetime (1). Many cancer chemotherapies today focus on inducing levels of DNA damage and DNA replication errors that are too high for cancer cells to tolerate, forcing them into apoptosis (2). Serious side-effects arise as healthy cells, especially those with rapid replication rate, are also affected by this treatment (2). The enzyme MTH1 (MutT homologue 1) is proven to be important for the survival of cancer cells (3). Cancer cells produce reactive oxygen species (ROS) at a higher rate than healthy cells because of their high metabolic rate (4). ROS oxidise many of the components in the cell including the DNA nucleotides, giving rise to mutations and DNA replication errors if incorporated into DNA (4). MTH1 hydrolyses oxidised nucleotides from the triphosphate to the monophosphate form (5). As monophosphate nucleotides cannot be inserted into DNA, MTH1 prevents DNA damage and replication errors (5). This is highly valuable for a cancer cell and MTH1 is overexpressed in many cancers (3). We have studied MTH1 in several projects. Paper I describes the estab- lishment of MTH1 as a potential cancer target and the development of first- in-class inhibitors. In Paper II we show that cells with increased oxidative stress are more sensitive to MTH1 inhibition. Paper III describes our studies of MTH1 from potential model organisms including mouse and dog, to describe the small differences that can prove crucial in pre-clinical studies. Paper IV shows a virtual fragment screening campaign towards MTH1, validated with X-ray crystal structures. Paper V describes a novel function of MTH1, namely the hydrolysis of methylated nucleotides. Cancer cells have a high metabolic and division rate, which is responsible for their high level of oxidative stress. To be able to divide at this elevated rate, cancer cells need a constant production of new nucleotides (6, 7). Sev- eral steps in nucleotide base production require the addition of a single car-

2 bon atom. These carbon atoms, attached to the tetrahydrofolate (THF) molecule, need to be reduced and oxidised. This is facilitated by proteins from the methylene-tetrahydrofolate dehydrogenase family (MTHFD) (8). The MTHFD2 protein of this family is highly upregulated in cancer cells, but is not present in healthy adult cells (9). In Paper VI we describe the first inhibitor to MTHFD2 and present the first structure of MTHFD2. This will facilitate future efforts to develop cancer treatments based on MTHFD2 inhibition.

Clostridium botulinum neurotoxin The gram-positive bacterium Clostridium botulinum produces the world’s deadliest toxins, the botulinum neurotoxins (BoNTs) that cause the disease botulism (10). BoNTs have also been successfully used to treat neuromuscu- lar disorders by dampening or inactivating specific muscles (11). The botulinum neurotoxin consists of three domains, a receptor-binding domain, a translocation domain and a protease domain (12). The receptor-binding domain recognises the nerve cell receptors and the toxin is incorporated into an intracellular vesicle. The acidic pH in the vesicle allows the translocation domain to move the protease domain into the cytoplasm. In the cytoplasm the protease domain of BoNT finds the target proteins of the soluble N- ethylmaleimide-sensitive factor attachment protein receptor (SNARE) family responsible for vesicle fusion, and cleaves the specific target for the particular type of BoNT (12). This blocks the fusion of acetylcholine vesicles with the cell membrane and thus the release of acetylcholine, resulting in muscle paralysis. In the disease botulism, paralysis of the muscles operating the lungs can lead to suffocation (10). C. botulinum has several genes in close proximity to the bont gene (13). There are two different sets of genes present, and the bont gene is always associated to either of these gene clusters (13). The associated cluster de- pends on the strain of C. botulinum (13). One of the gene clusters, the HA cluster, is well studied and recently both the structure and function of the proteins transcribed by the genes in the cluster have been determined (14- 17). These proteins were shown to increase the bioavailability of the neurotoxins by protecting ingested toxin in the stomach (18), and to facilitate transport through the cell layer lining the gastrointestinal tract, allowing the toxin to reach the bloodstream (16, 17). The second type of gene cluster, the OrfX cluster, is not as well under- stood (19). In Paper VII we solved the first structures of P47 and OrfX2 from the gene cluster of the Clostridium botulinum strain Kyoto-F encoding BoNT subtype A2. We find that the proteins from this cluster bind to certain

3 lipids in vitro and that their structures are similar to tubular lipid-binding proteins, a protein family previously only seen in eukaryotes (20).

4 Cancer and the cancer cell

“understanding how the parts relate to each other is a precondition to understanding process and understanding process is the precursor of uncovering principles” – Patrick Bateson (21)

Cancer is a terrible disease, severely impacting those affected, as well as their family and friends. In 2012 cancer was the cause of death of 8.2 million people worldwide, with an estimated 14.1 million new cases (22). Lung cancer, breast cancer, colorectal cancer, prostate cancer, stomach cancer and liver cancer make up 55% of the total global incidence (22). In Sweden in 2012, 60524 people received a cancer diagnosis for a malignant tumour, of which 50072 cases had no previous cancer diagnosis (23). Of these cases, 52% were male and 48% were female (23). In the same year, 11692 persons died of malignant tumours in Sweden (24). It has been estimated that every third person in Sweden will receive a cancer diagnosis during their lifetime (1). Radiation and chemotherapy treatments are successful in many cases but can cause severe side-effects. Often these treatments are combined with surgery. Massive efforts are made for finding new treatments in hope that they will have greater selectivity towards cancer cells over healthy cells, and therefore reduce the risk of side-effects. Cancer develops when a population of cells show unrestricted growth compared to their healthy counterparts and continuously increase their numbers (25, 26). The malignant cells invade surrounding tissues, and may spread to secondary sites in a process called metastasis (25). This chapter gives an overview and critique of the dominant theory of cancer and cancer initiation (carcinogenesis). Current cancer treatments such as radiation therapy and chemotherapy, and the concept of synthetic lethality are presented. Several aspects of cancer cell biology have been studied that are related to the redox-deregulation of cancer cells and their rewired metabolism. The concepts and molecular targets for the structure-based drug design component of this study are presented here and in the coming two chapters.

5 Somatic mutation theory In the year 2000, a review on the hallmarks of cancer was published (27). The authors detailed six features of cancer cells, which are required for a malignant cancer tumour to develop. The hallmarks are, according to the somatic mutation theory (SMT), acquired by stepwise somatic mutations of a single cancer cell (27). The theory is based on cancer being a genetic disease and a case of “cells gone bad”. The SMT states that the cell acquires mutations affecting certain cellular functions, giving rise to the observed hallmarks (27). The hallmarks are: self-sufficiency in growth, insensitivity to anti-growth signals, evasion of apoptosis, limitless replicative potential, sustained angiogenesis and finally, tissue invasion and metastasis (Figure 1) (27). In 2011, the same authors expanded on their previous concepts and introduced four more hallmarks: genomic instability, tumour-promoting inflammation, deregulation of cellular energetics and avoiding immune de- struction (Figure 1) (28). They also specified the importance of the tumour microenvironment in cancer formation. The previous way of viewing a tumour as containing only one type of cell, the cancer cell in high numbers, is too simplistic (27, 28). According to SMT the cancer cell recruits several different cells to provide the necessary signals for the cancer cell to continue to proliferate (27, 28). Without these signals, the cancer cells remain in quiescence and proliferation does not occur (27, 28). SMT has been the dominant theory in cancer research during the last fifty years. It originated in 1914 when Theodor Boveri published Concerning the Origin of Malignant Tumours (29).

Figure 1: The hallmarks of cancer, adapted from Hanahan and Weinberg (27, 28).

Critiques of the somatic mutation theory The tissue organization field theory (TOFT) disputes the requirement of SMT for nucleotide mutations, and states that cancer arises from the disruption of interactions with adjacent tissue (30-33). It is a tissue-based “development gone awry” theory, rather than cell-based as suggested by SMT (30- 33). The development of tissue is influenced by gradients of certain molecules and proteins called morphostats, which are similar to the morphogens described for embryonic development (30, 34, 35). Disruptions of the levels and gradients of these morphostats result in cancer (30). One of the main premises is that the default state of the cell is proliferation, not quiescence (27, 28, 30-33, 36-39). A cell going into a proliferative state is returning to the default state without regulation of cell replication (30-33, 36-39). TOFT addresses inconsistencies and paradoxes that have arisen within the cancer research field that cannot be explained by SMT (36, 37, 39). Se- quencing of tumour cells show somatic mutations are not always present (30, 40) and some cancer tumours revert back to healthy tissue morphology (30, 41, 42). This could be explained by epigenetic changes such as DNA methylation or histone modification, which are non-mutational changes affecting gene regulation (28). TOFT considers mutations and genetic instability a by-product of cancer (30). However, cancers can arise by mutations that influence the levels of morphostats (30). In this thesis, results will be discussed based on the dominant somatic mutation theory. However, as our targets are based on phenotype, not genotype (43), the results are not in direct opposition to TOFT.

Genetic instability and general cancer therapy One of the hallmarks of cancer is genetic instability (27, 28). In a cell with a fully healthy genome and DNA repair system, acquiring the necessary mutations for cancer to arise would not be possible (27, 28). The DNA monitoring, repair systems and checkpoints ensure that the genome is intact and that mutations are rare (27, 28). With the low mutation rate in healthy cells, the acquisition of the mutations required for the described hallmarks of cancer could be slower than the average human lifespan (27). However, the genome is unstable and slowly deteriorates over time when affected by endogenous processes such as hydrolysis, oxidation and non-enzymatic methylation (44- 49). In a single mammalian cell, these factors combined result in more than ten thousand potentially mutagenic and cytotoxic attacks on double-stranded

7 DNA per day (49, 50). In addition, molecules and agents from the cell and environment can also damage the DNA (51-57). To repair DNA lesions and prevent mutation there are several types of repair systems; base-excision repair (58-62), mismatch repair (63-70) and nucleotide-excision repair (71- 76). The most dangerous DNA damage, double-stranded breaks, occur if other systems fail to repair errors correctly. These breaks also happen if a replication fork encounters a single-strand break, and needs to be repaired by either the homologous repair system or the non-homologous end joining system (77-79). There are one hundred and fifty human genes that have been identified as DNA repair genes (80, 81). The discovery of the fragility of DNA, DNA lesions and DNA repair systems was awarded the Nobel Prize in Chemistry in 2015 (50, 70, 72). Defects in the genome and DNA monitoring and maintenance systems could result in an increased rate of mutations (27, 28). The products of these caretaker genes are involved in: 1) genome and DNA monitoring and recruitment of repair enzymes, 2) enzymes responsible for genome and DNA repair, 3) enzymes that block or inactivate DNA damaging molecules before they reach the DNA (28). Cancer cells often have a higher rate of mutation than other cells (28). However, even in a cancer cell, high levels of DNA damage will inevitably force the cell to undergo apoptosis (2, 82). Radiation and traditional cancer chemotherapy are focused on inducing DNA damage and apoptosis in cancerous cells (2). The high proliferation rates of cancer cells makes them more susceptible to DNA damage since the risk of cell death increases when replicating damaged DNA (Figure 2) (2). Early studies with radiation showed its ability to inflict DNA damage and kill both healthy cells and cancer cells (83-85). Chemotherapeutic agents have several mechanisms of action: 1) directly reacting with DNA forming alkylations or DNA cross- links (2, 86), 2) acting as antimetabolites (such as nucleotide analogues) that when inserted into DNA cause mismatches, replication-stalling and hinder further strand elongation (2), 3) acting as antimetabolites to interfere with enzymes important for DNA or nucleotide synthesis, such as the folate analogues (2, 87), 4) interfering with DNA-protein binding for any important DNA process, e.g. topoisomerase poisons (2, 88). One of the limitations to these treatments is the development of resistance to the chemotherapeutic agents by the cancer cells (2, 89). Understanding the mechanisms of resistance to chemotherapeutic compounds can give rise to new treatment strategies (2, 89-91). Resistance can arise when a cell over- expresses transporters that can remove anticancer drugs from the target cell (90) or produces compounds that inactivate the anticancer drug, for example increased production of cellular thiols to counteract Cisplatin (91). Cells also develop resistance by increasing the amount of the molecular target, e.g. overexpression of dihydrofolate reductase provides resistance to Methotrex-

8 ate (2). Resistance to Doxorubicin, a topoisomerase poison, can arise from a decrease in topoisomerase-II expression and/or its catalytic activity (2, 89). As many cancer drugs were found to have effectiveness against cancer, only limited consideration was given to their toxicity to other tissues initially (2). Cisplatin is associated with high risk of nephrotoxicity of the kidneys (91-94) and toxicity to the brain, causing tinnitus and high frequency hearing loss (2, 91). Doxorubicin can cause cardiotoxicity, including cardiomyopa- thy and congestive heart failure (95, 96). In one study, 26% of patients re- ceiving a Doxorubicin dose of 550 mg/m2 experienced congestive heart failure (97). Doxorubicin has also been linked to side-effects in the brain, mediated by tumour necrosis factor-α, since Doxorubicin cannot cross the blood- brain-barrier (98). The oxidative stress induced by Doxorubicin use has also been linked to liver toxicity (99). Combination therapy can limit adverse side-effects as the dosage of each drug can be reduced (2). The timing of combinations can also relieve side-effects. Many patients are, for example, pre-treated with Tamoxifen to reduce the levels of toxic products from Dox- orubicin treatment (89). A recent review categorised current and new cancer therapies by which of the hallmarks of cancer they targeted (27, 28, 100). Classical cytotoxic drugs promote tumour cell death, while cell death is produced indirectly in targeted therapy by the drug acting on a particular target or pathway (100).

Synthetic lethality Cancer cells display high genomic instability and mutation rates (28), and must repair mutations and lesions as they arise to avoid apoptosis (2, 82). Defects in one or more DNA repair processes are common in cancerous cells (27, 28, 101-103). This creates a dependence on other repair mechanisms to resolve problems the lesions create (101-103). Synthetic lethality is a cancer treatment that seeks to exploit this dependence (101-103). The best example is the use of poly (adenosine diphosphate [ADP]-ribose) polymerase (PARP) inhibitors to target tumours lacking breast cancer susceptibility genes 1 and 2 (BRCA 1 and BRCA2) (104, 105). PARP and BRCA1/2 are both involved in DNA break repairs (102). PARP is involved in repairing naturally occurring or induced single-strand breaks (102). If the single-stranded break is not repaired before a replication fork arrives at the site, the replication fork will collapse and form a double-stranded break (102). A cell can survive the loss of one gene or gene product in a synthetic lethal pair, but loss of an entire pair results in cell death (Figure 2) (101-103). The repair of double-stranded breaks is impaired in BRCA1/2 deficient cells. Inhibition of PARP in these cells results in the accumulation of double-strand breaks and apoptosis (101- 103). As healthy cells are not BRCA1/2 deficient this treatment selectively

9 targets the BRCA1/2 deficient cancer cells, reducing the chance of adverse side-effects (101-103). Drugs targeting PARP and other synthetic lethal pathways are in clinical trials (101-103), with some PARP inhibitors recently approved for clinical use (106). However, several possible synthetic lethal targets are not yet well established and no selective inhibitors for them exist (103). There are likely multiple routes to cancer within a tumour due to the ran- dom nature of mutations arising in individual cells, resulting in considerable tumour heterogeneity (27, 28, 30, 31, 33). Therefore, the synthetic lethal approach may not be effective against all cells of a tumour and combination therapy may prove necessary to treat a single tumour (101), and to combat drug resistance (103). As not all cancers have the mutations or defects needed for a synthetic lethal approach, our approach has been to elucidate the common features of cancer cell biology (Paper I-VI).

Figure 2: Different cancer treatments. Non-essential protein are represented by un- filled symbols, essential proteins are shown with filled symbols. Adapted from Helleday (43).

Redox regulation and reactive oxygen species Reactive oxygen species (ROS) is the collective name for oxygen radicals - such as superoxide (O2 ), hydroxyl (OH ), alkoxyl (RO ) and other oxygen species that are oxidising agents and/or are easily converted into radicals such as hypochlorous acid (HOCl), ozone (O3) and hydrogen peroxide (H2O2) (4, 55, 56, 107, 108). Reactive nitrogen species are radicals of nitrogen oxides such as the nitric oxide radical (NO), peroxynitrite (ONOO-) and the nitrogen dioxide radical (NO2 ) (56, 107). Reactivity differs between - species: H2O2, O2 and NO react quickly with a small number of substrates, while OH reacts quickly with a high number of substrates. (56).

10 Generation of reactive oxygen species ROS are very important agents in the cell and several signalling pathways for cell growth, proliferation, survival and motility are influenced by them (4). The majority of ROS production is thought to happen during oxidative - phosphorylation in the electron chain of mitochondria (4, 109, 110). The O2 formed from Complex I and Complex III when the ADP concentration is low (4, 109-112) is also converted to H2O2 and the highly reactive OH radical (4, 110, 112). H2O2 is more stable and can diffuse through membranes, possibly acting as a signalling molecule (4, 110). ROS are also formed in the endoplasmic reticulum (ER), when protein disulphide isomerases reduce and oxidise disulphide bonds during protein folding (4, 113, 114). It is estimated that 25% of ROS formation in a healthy cell comes from the ER, and increased ROS formation in this organelle can cause Ca2+ leakage that stimulates mitochondrial ROS production (113). Reduced glutathione is a major antioxidant in the cell (113). The ratio of reduced glutathione (GSH) to oxidised glutathione (GSSG) is seen as an index of the cellular redox state (113). The ratio of GSH to GSSG is between 30:1 and 100:1 in cytoplasm, but is 3:1 in ER, which favours the oxidation of thiols in the ER (113, 115). ROS are also generated by NADPH oxidases in phagocytes for killing in- vading microbes (4, 107, 108), the oxidation of xanthine to uric acid by xanthine oxidase (4, 116-118), ultraviolet light and other sources of ionizing radiation (119), metals (120) and the metabolism of toxic compounds by the cytochrome P450 family (121).

Reactive oxygen species in health and disease There are several antioxidant systems in the cell such as the glutathione redox buffer system, superoxide dismutases, peroxiredoxins and catalases (4, 78). They act by reducing ROS to less harmful species (4, 78). ROS can cause oxidative damage to cells by the oxidation of lipids, proteins, DNA and nucleotides (4, 56, 78). Several proteins in the cell have sensors for oxidative stress, for example, free cysteines that are oxidised to disulphate bridges (4). A modest increase in ROS can be counteracted by the antioxidant systems of the cell, but more severe oxidative stress can initiate carcinogenesis (4, 56). High levels of ROS result in DNA damage and apoptosis (4, 56). Cancer cells have an increased production of ROS (4, 122-124). Cancer cells may switch their metabolism from oxidative phosphorylation to aerobic glycolysis (4, 27, 28, 125) during hypoxia (4, 122, 123), but main- tain the mitochondrial structure and capacity for mitochondrial ATP production, and thus ROS production (126). Even when oxygen levels are low and aerobic glycolysis is the main pathway for energy generation, the mitochondria still produce ROS (122, 123). In addition, several oncogenes such as

11 KRAS influence mitochondrial ROS generation (122, 123). Increased ROS production also increases mitochondrial biogenesis, the formation of new mitochondria, and also increases mitochondrial DNA copy numbers, which further amplifies mitochondrial ROS generation (4, 123). Cancer cells can overexpress antioxidant proteins to counteract excessive ROS production, in order to avoid apoptosis (4, 56, 78).

Hypoxia and reactive oxygen species Low oxygen levels (hypoxia) are often observed in tumours (4, 27, 28, 122, 123, 127, 128). The rapid proliferation of cancer cells means that the available blood vessels cannot supply enough oxygen (27, 28, 122, 123, 127, 128). Hypoxia induces ROS formation in the mitochondrial electron transport chain (122, 123). Hypoxia-induced factors, especially HIF-1, are upregulated in cancers (129) and stabilised by ROS generation (122, 123, 130, 131). HIF-1 regulates the expression of genes responsible for angiogenesis, the metabolic switch between oxidative phosphorylation in mitochondria to aerobic glycolysis in the cytoplasm, and other cell cycle events (4, 122, 123, 128). The HIF-1 dimer contains two distinct monomers, HIF-1α and HIF-1β (4, 122, 123). HIF-1β is expressed in the cell in an O2-independent manner (4). Under normal oxygen levels HIF-1α is degraded rapidly by the hydrox- ylation of proline residues in its structure. This is carried out by proline hydroxylase domain proteins, which enable von Hippel-Lindau tumour suppressor protein (VHL) to bind HIF-1α and add ubiquitin, targeting HIF-1α for proteasomal degradation (127, 132-135). Low oxygen levels inhibit the O2-dependent proline hydroxylase domain proteins, and HIF-1α can move to the nucleus and bind HIF-1β (127, 132-135). Compounds such as dimethyloxallylglycine (DMOG) can also inhibit proline hydroxylase domains and cause pseudo hypoxic behaviour (123), as was used to test the influence of hypoxia on drug sensitisation in Paper II. As ROS induces HIF-1 activation by stabilising HIF-1α, and HIF-1 leads to increased mitochondrial ROS production, this highlights the need for the overexpression of antioxidant systems in cancer cells in order to avoid apoptosis from ROS-related damage (Figure 3) (4, 56, 78, 122, 123).

12 Figure 3: Hypoxia induces ROS production in mitochondria, leading to changes that may cause tumours (123). DMOG, dimethyloxallylglycine, PHD, proline hydroxylase domain, HIF-1α, hypoxia-induced factor 1α.

Oxidised nucleotides ROS cause oxidative damage to DNA bases and nucleotides in free nucleotide pools (55, 56, 136-139), the most commonly occurring oxidised lesion being 8-oxoguanine (8-oxoG) (55, 56). 8-oxoG is formed by the direct oxidation of guanine within a DNA molecule or by the oxidation of free dGTP to form 8-oxo-2’-deoxyguanosine triphosphate (8-oxo-dGTP) that is then incorporated into DNA (55, 56). In addition to cancer, increased 8-oxoG in DNA has also been implicated in Huntington’s disease (140). 8-oxoG is mutagenic and within the DNA double helix it can base pair with adenine in addition to cytosine (Figure 4) (55, 56, 141-148). Other mutated nucleotides such as 2-hydroxyadenine (2-OH-A), are also frequently found in DNA (56, 137, 141, 146, 147, 149-153). 2-OH-A can base pair with guanine as well as with thymidine (137, 141, 146, 147, 149-153). 8-oxoG DNA-glycosylase (OGG1) is responsible for removing the 8-oxoG base from DNA and initiat- ing the base-excision pathway (78, 142, 154-160). OGG1 preferentially removes the 8-oxoG bound to cytosine within the DNA double helix (81, 154, 158, 160). Several studies have found mutations of OGG1 in cancer cells (161, 162). An isoform of OGG1 is targeted to mitochondria and is the main enzyme responsible for removal of 8-oxoG in mitochondrial DNA (154, 163). The MutY homologue (MUTYH) enzyme removes adenine bound to 8-oxoG within DNA and is important for removal of 2-OH-A bound to guanine (81, 146, 164).

13 Figure 4: 8-oxo-dG forms Watson-Crick or Hoogsteen base pairs (165). Hydrogen bonds are shown as dashes.

The oxidation of bases in DNA occurs frequently (49), however, the free nucleotide pools are 190-13000 times more susceptible to modification than double-stranded DNA (136-139). Removal of these oxidatively damaged nucleotides before their insertion into DNA is extremely important for cell survival (143, 145, 147, 152, 166, 167).

Oxidative damage to nuclear and mitochondrial DNA leads to cell death The mitochondrial DNA is more susceptible to oxidative damage from ROS than nuclear DNA (4, 143, 147, 152, 154, 168). This could be due to the close proximity of mitochondrial DNA to the electron transport chain where ROS production occurs, or the vulnerability of mitochondrial DNA due to the absence of histone proteins (4, 56, 152, 154). Levels of oxidised bases are also higher in mitochondrial DNA (56, 154, 168). It has been shown that accumulation of 8-oxoG in DNA, both nuclear and mitochondrial, leads to cell death. However, the pathways to cell death at these sites differ (144). Oxidatively damaged nuclear DNA causes PARP-dependent translocation of apoptosis-inducing factor to the nucleus, by the accumulation of single- stranded breaks (144). Damaged mitochondrial DNA also accumulates single-stranded breaks, causing mitochondrial DNA depletion, dysfunction and Ca2+ release which activates calpain, resulting in cell death (144). Oxidative damage to mitochondrial DNA is responsible for the majority of cases of cell death (4, 56, 143, 152, 154). The importance of mitochondrial DNA is also indicated by the existence of OGG1, MUTYH, and MutT homologue 1 (MTH1) isoforms targeted to mitochondria (152, 154, 159, 160, 163, 164, 169, 170).

14 Methylated nucleotides DNA mutations can occur following non-enzymatic methylation by naturally occurring S-adenosyl-methionine (SAM) (48, 49, 171), environmental methylation factors, or by chemotherapeutic drugs such as Temozolomide (172). Alkylation of DNA may also occur (173). Epigenetic methylation however, is important for the regulation of gene expression and occurs at the C5 position of cytosine, in the presence of specific methyltransferases (174, 175). Free nucleotides are more susceptible to methylation than double- stranded DNA, as they are more accessible to the mediators of methylation (136). Different methylating agents preferentially form different methylated nucleotides (176). Nucleotides damaged by methylation can be inserted into DNA and cause mutations (171, 177), most commonly the O6-methyl-dGTP nucleotide formed from the methylation of dGTP (136, 177). The incorporated base O6-methyl-dG causes mismatch-repair-dependent single-stranded breaks in DNA that lead to the collapse of replication forks and thus the formation of double-stranded DNA breaks (178). These single-stranded breaks are induced by the mismatch repair system as the thymine paired with O6-methyl-dG is removed instead of O6-methyl-dG, after which thymine is erroneously reincorporated (173). This creates a futile mismatch-repair cycle causing tertiary lesions in DNA, double-strand breaks and apoptosis (173). O6-methyl-guanosine in RNA slows ribosomal translation and affects ribosomal accuracy (179). As GTP levels are 100-fold higher than dGTP levels, the methylation of RNA precursors is highly relevant (180). Several DNA- glycosylases work to remove the methylated bases (181-184). Interestingly, O6-methyl-dG is primarily removed from DNA not by removal of the base or the nucleotide, but by direct repair of the methylation itself (173). The methyl group on O6-methyl-dG is transferred to a cysteine in O6- methylguanine-DNA methyltransferase (MGMT) in a suicide reaction that forms an inactive protein that requires degradation by the proteasome (173). Mice with no MGMT were found to be sensitive to methylating and alkylating agents that are used in cancer therapy (185, 186). The mice with a defective mgmt gene developed more spontaneous tumours after treatment with methylating agents than mice with functioning MGMT (187). Tumour cells that lacked MGMT were very sensitive to methylating agents (188). In glio- blastoma patients with an epigenetically methylated promoter for MGMT that silences the gene, the patients responded better to treatment with the methylating agent Temozolomide (189).

16 MutT homologue 1 (MTH1)

“while a high cellular antioxidant capacity tends to protect DNA from oxidative damage and related mutagenesis, antioxidant activity may also (ironical- ly) protect ‘initiated’ cells from ROS-mediated killing” – Helen Wiseman and Barry Halliwell (56)

The MutT protein from E. coli degrades 8-oxo-dGTP The protein in Escherichia coli with 8-oxo-dGTPase activity was found to be the product of the mutT gene (190). MutT degrades 8-oxo-dGTP to its monophosphate form, 8-oxo-dGMP, which is not incorporated into DNA (190). MutT has a high specificity for 8-oxo-dGTP over all non-oxidised nucleotides (dGTP, dATP, dCTP and dTTP) (190). The first mention of a factor in E. coli that can affect the rate of mutation is from 1954 (191). After its initial characterization as a 8-oxo-dGTPase (190), MutT has also been shown to be able to degrade 8-oxo-GTP, 8-oxo- dGDP and 8-oxo-GDP (192). This shows that MutT is also important for the removal of oxidised RNA building blocks (192, 193). The removal of oxidised diphosphates is important for the integrity of DNA as they can be converted to triphosphates by nucleoside diphosphate kinases and incorporated into DNA (192). The first structure of MutT was solved in solution by NMR (194). The X-ray crystal structure of MutT was solved in the apo form and also in complex with 8-oxo-dGMP and Mn2+ (Figure 5) (195). However, magnesium is the preferred metal ion for MutT hydrolysis activity and should bind in the same position as in the manganese bound structure (195, 196).

MTH1 can degrade several oxidised nucleotides When a protein in humans and other eukaryotes was found with the same 8- oxo-dGTPase activity as MutT (5, 139), it was named MutT homologue 1 (MTH1) (3). MTH1 is able to degrade several different oxidised nucleotides, such as 2-OH-dATP (141), 8-oxo-GTP (197), 2-OH-ATP (197), 8-oxo- dATP (141), 8-oxo-ATP (197) and numerous other nucleotide analogues (198). The activity of MTH1 is inhibited by ribo- and deoxyribonucleoside

17 diphosphates (NDPs and dNDPs) (199). At physiological concentrations, around 15% of MTH1 8-oxo-dGTPase activity would be inhibited (199). In addition, 8-halogenated-7-deaza-dGTP compounds can inhibit MTH1 by competitively binding to the active site (200). Before the structure of MTH1 was available, several parts of the MTH1 sequence were investigated for substrate binding (201). It was shown that residues W117 and D119 were important for binding of both 2-OH-dATP and 8-oxo-dGTP, with the W117A mutant showing no activity (201). The 8-oxo-dGMP formed from MTH1 cannot be reactivated by human guanylate kinase, which is usually responsible for dGMP to dGDP conver- sion, as it is completely inactive with 8-oxo-dGMP (202, 203). Therefore, MTH1 effectively stops the oxidised base from being incorporated into DNA (202). The 8-oxo-dGMP is degraded to 8-oxodeoxyguanosine by a nucleotidase, and is then excreted from the cell, and then from the body in the urine (202).

Structures of MTH1 show a versatile binding site The first structural information on MTH1 was provided by NMR, which showed differences in the substrate-binding pocket between MTH1 and MutT, despite the enzymes having a similar fold (204). The first crystal structures of MTH1 were solved in 2011 in both the apo form and in complex with 8-oxo-dGMP (Figure 5) (205). These data showed that 8-oxo- dGMP binds MTH1 in the anti conformation (205), while it binds to MutT in the syn conformation (195). This was later confirmed by binding studies with substrate analogues (198). Later it was shown that 8-oxo-dGMP is more stable than dGMP in the anti conformation, which may explain the specificity of MTH1 for oxidised nucleotides over non-oxidised ones (206). The residues N33, D119 and D120 make hydrogen bonds to the nitrogenous base of the nucleotide (Figure 5) and W117 forms a pi-stacking interaction with the nitrogenous base of the nucleotide (205). There was no direct hydrogen bonding that could be seen with the oxidised part of the ligand. It was hypothesised that the O8 of 8-oxo-dGTP influences the keto-enol tau- tomerization at the O6 position, increasing the stability of the nucleotide binding (205). The 2-hydroxyl group of ribonucleotides would be directed into a hydrophobic pocket, which is reflected by the lower activity of MTH1 towards ribonucleotides than deoxyribonucleotides (197, 205). Later a structure of MTH1 with 8-oxo-ATP, an oxidised ribonucleotide, was solved. This showed ligand binding in the anti conformation, but with the base and ribose rotated 180° in the binding site (207). The terminal pyrophosphate of 8-oxo- ATP could be seen, as was the case in another structure solved with 8-oxo- dGTP in complex with MTH1. It was hypothesised that hydrolysis did not

18 occur as the authors used soaking instead of co-crystallization for the complexes (205, 207). A recent study provided high-resolution crystal structures of MTH1 with 8-oxo-dGTP and 2-OH-dATP (208), using a G2K mutant previously shown to give high-resolution crystals (209). In this case it is likely that cleavage did not occur, as Mg2+ was not present during crystallization (208, 209). They showed 2-OH-dATP binding in a similar manner to 8-oxo-dGTP, but with a slight movement of the base resulting in a different hydrogen-bonding pattern (208). D120 was also shown to be important for specificity, with mutants D120N and D120A being selective towards 2-OH- dATP over 8-oxo-dGTP (208). Previously it was shown that the W117Y mutant only has significant activity towards 2-OH-dATP, while the D119A mutant predominantly degrades 8-oxo-dGTP (201).

Figure 5: MTH1 (left), MTH1 binding to 8-oxo-dGMP (middle) and MutT (right) shown as cartoons. The Nudix box is coloured pink. Amino acid side-chains and 8- oxo-dGMP are shown as sticks. In the MutT structure the manganese ion is shown as a sphere. Hydrogen bonds are shown as black dashes. PDB codes 3ZR0 (MTH1) and 3A6U (MutT) (195, 205). Figure made using PyMol (210). mth1 gene and isoforms The mth1 gene is located on chromosome 7p22 and consists of five exons. Because of alternative splicing, the mth1 gene can produce seven types of messenger RNAs (mRNAs). For four of the mRNAs types (Type 1, 2A, 3A and 4A) only isoform p18 MTH1 can be formed. However, due to the inclusion of other segments in the other three mRNAs (Type 2B, 3B and 4B), two more AUG codons are in frame, and isoforms p18, p21 and p22 MTH1 can be formed. The B-type mRNAs with a polymorphic alteration (GU GC) turns one stop codon UGA into CGU and an AUG further upstream becomes available, yielding p26 MTH1 (211, 212). The p18 MTH1 isoform is considered to be the “standard” MTH1, and is predominantly localised in the cytoplasm with about 5% located in the mitochondrial matrix (169). Computational methods suggest that p18 and p26 can be imported into mitochondria, and the 18 amino acid leader sequence

19 of p26 MTH1 has been suggested to be a mitochondrial targeting sequence (211). Very little is known regarding the isoforms other than p18, but both p21 and p22 were expressed in cell cultures along with p18, and all four have been expressed in vitro (212). A polymorphism in the mth1 gene corresponding to position 83 in the MTH1 protein has been found, where either valine or methionine is present (211-213). The p18 form of Met83-MTH1 is thermally less stable than Val83-MTH1 (213). Later it was shown that the V83M mutation is correlat- ed to mutations in the p53 tumour suppressor gene, and is more frequent in gastric cancer patients than in healthy controls (214).

MTH1 is part of the NUDIX hydrolase protein family MTH1 belongs to the NUDIX hydrolase protein family, and is also called NUDT1 (3). NUDIX stands for nucleoside diphosphate linked to another moiety, x. The family has a common function in that they hydrolyse substrates that include a nucleoside diphosphate linked to something else. Sub- strates include (d)NTPs, oxidised (d)NTPs, NADH, ADP-ribose and capped mRNAs (3). The protein family has been extensively reviewed (3, 215-219). The Nudix domain involved in metal binding and catalytic activity contains the Nudix box sequence (Gx5Ex5[UA]xREx2EExGU), where U is a hydrophobic residue (3, 196). The Nudix box, or Nudix sequence, which forms a loop-helix-loop motif, is present in all Nudix protein structures determined to date (Figure 5) (3). Most NUDIX proteins require magnesium for their catalytic activity (196, 215). In general the amino acids in the Nudix sequence are not involved in substrate binding but in hydrolysis. This means that the residues responsible for substrate binding may differ significantly between family members, thus explaining the variety of substrates within the protein family (3, 215). Several NUDIX proteins have been suggested to have analogous functions to MTH1 (3, 215, 216). NUDT15 was originally characterised as having 8-oxo-dGTPase activity and was thus named MTH2 (220). However, further investigation showed that the enzyme is more active towards 6-thio- dGTP, 6-thio-GTP and dGTP (221). As 6-thio-dGTP and 6-thio-GTP are the active metabolites of the thiopurine cancer drugs, this makes differences in NUDT15 activity important for drug tolerability and the development of serious side-effects (221-223). Several polymorphisms of NUDT15 occur in the general population that can affect NUDT15 activity (222). The most important appears to be the R139C mutant (222). The position of this residue was shown to be close to 6-thio-dGMP in a recent NUDT15 complex structure (223) The presence of a cysteine at this position results in the for-

20 mation of a disulphide bridge with C140, which influences the binding site (221, 223). NUDT18 was found to have activity towards several oxidised nucleoside diphosphates including 8-oxo-dGDP, 8-oxo-GDP, 8-oxo-dADP and 2-OH- dADP, and was named MTH3 (224). However, it has been shown to be much less active towards traditional MTH1 substrates (221). By degrading diphosphates, NUDT18 prevents nucleoside diphosphate kinase converting them to triphosphates for DNA and RNA insertion (192, 203). It has also been demonstrated that the oxidised nucleoside diphosphates are inhibitors of MTH1 (141), as are canonical (d)NTPs (199). In a recent substrate activity assay it was shown that while MTH1, NUDT15 and NUDT18 have similar substrates, there are large differences in their activities (3). It has been suggested that the name MTH should only be used for enzymes that are more closely related to MutT in their main function (3, 216, 221). NUDT5 was found to cleave 8-oxo-dGDP (225) and may have a role in removing the damaged nucleoside diphosphates that inhibit MTH1 activity. However, the activity of the enzyme was low and only present at high pH (199, 225). NUDT5 can also degrade 8-oxo-GDP, but the activity was found to be comparable to the low activity of MTH1 towards the substrate (226). NUDT5 was later shown to have activity towards a range of damaged nucleoside diphosphates (227).

MTH1 prevents damage caused by ROS Several studies have shown that MTH1 can reduce the amount of 8-oxoG in DNA and also lower the DNA mutation rate (143, 147, 152, 228, 229). MTH1 can protect cells from death via oxidative stress and high 8-oxo- dGTP and 2-OH-dATP levels (143). Although only 5% of p18 MTH1 is directed to mitochondria (169), MTH1 still effectively prevents cell death caused by oxidised mitochondrial nucleotide pools (143). Mouse fibroblast cells lacking the mth1 gene are more susceptible to cell death from oxidised nucleotides, with H2O2 treated cells showing pyknotic nuclei (with the con- densation of chromatin) and degenerate mitochondria. This condition is avoided by expression of MTH1 (147, 152). MTH1 mutants W117Y or D119A have only 2-OH-dATPase and 8-oxo-dGTPase activity, respectively, and only partially suppress the condition when expressed, suggesting that both 8-oxoG and 2-OH-A contribute to the condition and cell death (147, 152). Knockdown of MTH1 in 293T cells increased the number of A:T G:C mutations (230, 231). Suppression of MTH1 induced double-stranded DNA breaks and senescence (a state of cell crisis with proliferation arrest) showing the importance of removing oxidised nucleotides from the cellular

21 nucleotide pool (166). These studies indicate that MTH1 deficiency increases the susceptibility of the cell to ROS damage (232). Ribonucleotide reductase is unable to use 8-oxo-GTP as a substrate (203), meaning that oxidation of the ribonucleotide pool could lead to a depletion of the deoxyribonucleotide pool (233). As the NTP levels are ten to several hundred-fold larger than dNTP levels (234), and adequate levels of deoxyribonucleotides are needed for base-excision repair and mismatch repair, this could be one reason for mutations in DNA after MTH1 inhibition (233). Mice lacking MTH1 have an increased incidence of spontaneous tumours (36%) found at 18 months, compared to 11% observed for MTH1-competent wild-type mice. However, MTH1 knockout does not affect survival rates (153, 235, 236) and an increase in A:T G:C transversion mutations is not observed in MTH1-/- mice (235), in contrast to MutT deficient E. coli which show a 1000-fold increase in mutations (152, 237). However, there is a marked increase in single-base frameshifts, indicating that mismatch repair is used to repair the increased number of mutations (235). Mice overexpress- ing MTH1 had lower levels of 8-oxoG in their nuclear and mitochondrial DNA, and lived significantly longer than wild-type mice (238). These mice also showed reduced anxiety and enhanced investigation of environmental and social cues (238). Mismatch repair is important for the removal of damaged bases, and MTH1 was shown to be important for the survival of mismatch repair- deficient cells (145).

Cancer phenotypic lethality: MTH1 as a drug target for cancer MTH1 clearly has a function in healthy cells, even though MTH1-/- mice only show mild phenotypic changes (153, 235, 236), possibly from overlap- ping functionality of other proteins (3). The reason for MTH1 being an interesting drug target is that cancer cells depend on MTH1, while healthy cells do not. Tumour cells have elevated ROS levels and oxidative stress due to oncogene activation, mitochondrial aberrations and metabolic dysfunction (124, 233, 239-242), and malignant tumours upregulate redox protective proteins to counteract the effects of ROS (233, 243, 244). The state of increased oxidative stress in cancerous cells creates potential targets for anticancer therapy (245). The strategy is to generate high levels of DNA damage to selectively kill cancer cells without affecting healthy cells (82). Cancer phenotypic lethality allows selective and specific elimination of cancerous cells by targeting proteins that are essential for cancer cells but not healthy cells (Figure 2) (43, 82). As the concept of cancer phenotypic lethality has poten-

22 tial to work regardless of genotype, which is not the case for synthetic lethality, it is possible that treatment of a wide range of tumours could benefit from this strategy (43, 82). MTH1 is overexpressed in renal-cell carcinomas (246), lung cancer (247), brain tumours (248), breast cancer (249) and several other tumours (3). In small-cell lung cancer patients, it was shown that MTH1 has more influence on the levels of 8-oxoG in DNA than OGG1 (250). The 8-oxo-G levels are lower in DNA from the lung cancer cells compared to DNA from surrounding healthy lung tissue, with higher activity of MTH1 in lung cancer cells (250). In mice lacking both OGG1 and MTH1 genes and proteins (152, 155) tumours do not develop, despite increased 8-oxoG levels in DNA. However, mice deficient in either OGG1 or MTH1 have an increased formation of tumours (152, 153, 155, 236). Loss of both MTH1 and OGG1 results in an increase of oxidised nucleotides, leading to cell death rather than cancer (152, 155). Both MTH1 expression and accumulation of 8-oxoG in DNA correlate to malignancy in brain tumours, indicating a higher level of oxidative stress in these tumours and the need for defence against oxidative stress (152, 248). Cancer cells need increased expression of MTH1 to cope with their phenotype of oxidative stress (152). In contrast, OGG1 is not overexpressed in cancers and is inactivated during oxidative conditions (152, 251). Incorporation of 8-oxoG at telomeres stops telomerase elongation giving rise to drastically shortened telomeres and inducing cell death (252). How- ever, telomerase is able to extend telomeres with a pre-existing 8-oxoG terminus, since it disrupts a G-quadruplex structure enabling telomerase load- ing (252). Depletion of MTH1 increases telomere dysfunction and cell death in cancer cells with short telomeres (252). Telomerase is often reactivated in cancer cells to achieve the “limitless replicative potential” hallmark of cancer (27, 28, 253). Approximately 20-25% of tumours contain mutations in the RAS family of oncogenes (167, 254). Oncogenic RAS is increased in highly malignant tumours, leading to increased ROS production and oncogene-induced senescence (167). Overexpression of MTH1 prevents the DNA damage response and RAS-induced senescence in RAS-transformed cells (167). This indicates that the nucleotide pools are critical targets of ROS produced by oncogenic RAS and RAS-transformed cells require MTH1 expression to proliferate (167). MTH1 may also participate in maintaining the ROS-associated tumour-promoting mechanisms and aid in the acquisition of malignant traits (254). The inhibition of MTH1 increases 8-oxoG content leading to senescence and cell death but also inhibits survival-signalling and the epithelial- to-mesenchymal transition (254). MTH1 is required for the maintenance of several oncogenes such as KRAS and HRAS (255, 256). It has been shown that OGG1 excision of 8-oxoG can activate Ras-GTPase (142), so MTH1 may also have a role in RAS activation/signalling.

23 MTH1 controversy Compounds that inhibit MTH1 are currently being developed by several groups for use as anti-cancer drugs (257-260) following the validation of MTH1 as a cancer target in Paper I and (261). These selective and potent MTH1 inhibitors did have a cellular target engagement but have not displayed cytotoxicity raising the question of the validity of MTH1 as a cancer target (257, 259, 260). MTH1 inhibitors TH287 and TH588 (Paper I) were found to also inhibit tubulin polymerization, a frequent target for anti- tumour treatment (258). This inhibitory effect could not be seen in a prote- ome-wide cellular thermal shift assay, however cannot be excluded (262). Compound IACS-4759 (260) and compounds 15, 19 and 24 (259) inhibit MTH1 but do not increase 8-oxoG levels in DNA nor kill cancer cells (262). Increased 8-oxoG levels in DNA of cancer cell lines was one factor used to develop the original inhibitors TH287 and TH588 (Paper I), and off-target effects cannot be excluded (262). Microinjection of 8-oxo-dGTP and 2-OH- dATP into zebrafish eggs is only toxic to them if the MTH1 inhibitor TH588 is present (Paper II). An SW480 MTH1-/-/- clone was shown to be healthy (259), however this could be related to increased dependence on other proteins with similar activities (Paper I). Suppression of MTH1 expression by siRNA or shRNA could reproduce the cytotoxic effect (263), which was not observed in other studies (259, 260). The use of different model systems was proposed to influence the results (263), and also that all cancer cells might not depend on MTH1 for their survival (262). Clearly we do not yet have the full answer as to how MTH1 biology works, and why the developed inhibitors are cytotoxic (262, 263).

New MTH1 function: degradation of O6-methyl-dGTP Recently it was found that MTH1 also degrades the mutagenic O6-methyl- dGTP in vitro and in vivo, showing that MTH1 is also important for the removal of methylated nucleotides from nucleotide pools (Paper V). In addition, O6-methyl-GTP could also be degraded by MTH1 (Paper V). MTH1 inhibition could then be synthetic lethal to cancer cells with a si- lenced mgmt gene, or otherwise inactivated MGMT (PaperV).

24 De novo nucleotide synthesis and one-carbon metabolism

A dividing cell requires replication of its DNA, to provide a copy for each of the two daughter cells. To replicate the six billion base-pair human genome (264), twelve billion deoxyribonucleotides are needed (7). Nucleotide synthesis requires one-carbon units supplied by one-carbon metabolism (265- 268). One-carbon units are also required to produce methionine and the methylation cofactor S-adenosyl-methionine used for epigenetic methylation (266, 268). Of the five carbon atoms that comprise a purine, two carbons come from an incorporated glycine, two from the one-carbon donor 10-formyl- tetrahydrofolate (formyl-THF) and one carbon from a CO2 molecule (Figure 6) (265, 267, 269). This is the case regardless of whether the purine produced in de novo purine synthesis is GMP or AMP, as both are produced as inosine monophosphate (267). The base is then further differentiated to GMP or AMP, from which the triphosphates and deoxyribonucleotides can be produced (267).

Figure 6: Origin of atoms in purine synthesis, and methyl group in thymidylate synthesis (267, 269, 270).

5,10-methylene-tetrahydrofolate (methylene-THF) donates a methyl group to dUMP by thymidylate synthase to generate dTMP (Figure 6 and Figure 7) and dihydrofolate (87, 270-273). Dihydrofolate is reduced back to tetrahy-

25 drofolate (THF) by dihydrofolate reductase, which is the target of the anticancer drug Methotrexate (87, 270-273). Embryonic cells and cancer cells are rapidly dividing and have a high demand for a steady supply of one-carbon units to support their proliferation (7, 9, 274, 275). Production of these one-carbon donors occurs through one- carbon metabolism (Figure 7) (266, 268, 276). This metabolism is upregulated in cancer (277) and in embryonic development (278).

Figure 7: One-carbon metabolism. Important outcomes are underlined. Reactions performed by: 1-3, MTHFD1; 2m and 3m, MTHFD2/2L; 1m, MTHFD1L; 4 and 4m, SHMT1/2; 5, glycine cleavage system; 6, thymidylate synthase; 7, dihydrofolate reductase; 8, ALDH1L2 (also occur in cytosol with ALDH1L1, not shown); 9, me- thionyl-tRNA-formyltransferase. Adapted from Tibbetts and Appling (8).

One-carbon metabolism One-carbon metabolism is a folate-dependent pathway, where the reduced form of folate (tetrahydrofolate, THF) carries the donated carbon groups through the steps of the cycle (Figure 7) (266, 268, 276). In mammals one- carbon metabolism is localised to the mitochondria and cytosol (8, 272, 279, 280). The one-carbon donors serine, glycine and formate can be transported between these compartments across the mitochondrial membrane (8, 266, 268, 276). One-carbon-loaded folate intermediates cannot cross the mitochondrial membrane (266, 281), possibly because of polyglutamation (8). The direction of metabolic flow is determined by the different redox states and cofactor dependencies in the cytosol and the mitochondria (280, 282).

294). In addition, the rapid turnover rate of the intermediate methenyl-THF without release from the protein indicated substantial substrate channelling between the dehydrogenase and cyclohydrolase activity (288, 295, 296). The rate-limiting step for the DC-domain is the cyclohydrolase step (297). MTHFD1 uses NADP+ for the dehydrogenase step, yielding NADPH (292). The cytoplasmic redox state however favours the reverse reaction as the GSH to GSSG ratio is between 30:1 to 100:1 (113, 115). The NADP+/NADPH ratio in the cytosol is usually in the range of 0.001 to 0.02 (298-300). This causes MTHFD1 to use formate generated in mitochondria or from other sources to produce the formyl-THF needed for purine synthesis, and after reduction also the methylene-THF needed for thymidylate synthesis (280, 282). All tissues express MTHFD1 (301), however mRNA levels might vary (293). MTHFD1 is an essential gene and its disruption results in embryonic lethality (302). In the mitochondria there are three proteins of the MTHFD family present: MTHFD2, MTHFD2L and MTHFD1L (8, 280, 282). MTHFD2 and MTHFD2L have dehydrogenase and cyclohydrolase functions (303), whereas MTHFD1L is a mono-functional formyl-THF-synthetase (304, 305). MTHFD2 was first found in Erlich tumour cells in 1960 (306) and was later characterized as a bi-functional dehydrogenase and cyclohydrolase present in mitochondria (307-310). MTHFD2 uses NAD+ for the dehydro- 2+ genase function (311), and requires inorganic phosphate (Pi) and Mg (307, 309, 310, 312), but can use NADP+ with lower activity (307, 312). With + 2+ NADP as cofactor, MTHFD2 only requires Mg , and Pi instead inhibits the reaction (312). It has been suggested that MTHFD2 has evolved from a tri- functional enzyme through the loss of the synthetase domain and has + + 2+ changed cofactor dependence from NADP to NAD , Pi and Mg (307, 313). As MTHFD2 uses NAD+, the redox potential of NAD+ (8, 288) and the oxidative environment of the mitochondria drive these reactions towards the formation of formyl-THF (8, 288, 298) rather than methylene-THF (8, 288). The free NAD+/NADH ratio in mitochondria is between 3-5 according to one source (298) and around 7 according to another source (314, 315). The ratio of free NADP+/NADPH in mitochondria is between 5-8 (298), but another measure of the total NADP+/NADPH gives a ratio in mitochondria of 0.033, while the ratio of total NAD+/NADH is 2.2 (315). MTHFD2 is expressed by embryonic cells (311, 316) and is required for embryonic development (317). MTHFD2 knockout is embryonic lethal for mice (317). MTHFD2 is expressed in cancer cells, but not in healthy adult cells (9, 311, 316). A low level of mRNA is still present in adults, but is not translated into protein (9, 318). MTHFD2 is needed for glycine production since MTHFD2-null mutant fibroblasts need glycine supplementation to survive (319), but added glycine is not required when a methylene-THF- dehydrogenase-cyclohydrolase construct is expressed in mitochondria (320).

28 MTHFD2-Like (MTHFD2L) is translated in all adult tissues (303) and embryonic cells, and is 64% identical to MTHFD2 (321). It is a methylene- THF dehydrogenase and cyclohydrolase (303) that can use either NAD+ or + 2+ NADP for its catalytic dehydrogenase function (321). Pi and Mg enhance + 2+ the activity of MTHFD2L when NAD is the cofactor but neither Pi or Mg alone gives the same enhancement (321). When NADP+ is involved in the 2+ reaction, MTHFD2L activity is enhanced by Mg , but addition of Pi removes this enhanced activity (321). When assayed under physiological con- 2+ centrations of methylene-THF, Pi and Mg , MTHFD2L preferentially uses NAD+ as a cofactor, however NADP+ is used when the methylene-THF concentration is low (321). MTHFD2L is considered to be a housekeeping enzyme as it is constitutively expressed throughout all tissues (303, 321). MTHFD2L has significantly lower activity than MTHFD2, which is responsible for providing the one-carbon units required by the rapidly dividing embryonic and cancerous cells (303). MTHFD1-Like (MTHFD1L) is expressed in embryonic cells and all adult tissues (305, 322) and is responsible for formyl-THF-synthetase function in mitochondria (282, 304, 305). MTHFD1L has 61 % homology to cytosolic MTHFD1 excluding the mitochondrial targeting sequence (282, 304, 305). It contains the DC-domain and synthetase domain as MTHFD1, but critical mutations in the active site of the DC-domain renders it a mono-functional synthetase (304, 305). MTHFD1L deletion causes neural tube defects and is embryonic lethal (323). Upregulation of MTHFD1L mRNA transcription was found in human colon adenocarcinoma (324).

Outcomes from the one-carbon metabolism Formyl-THF and glycine are required for de novo purine synthesis in the cytosol (267). As formyl-THF cannot be transported across the mitochondrial membrane (281), one-carbon units produced in mitochondria are transported as formate, and bound to THF in the cytosol by MTHFD1 (8, 268, 276, 280, 282). This leads to ATP production by MTHFD1L in mitochondria and ATP consumption in cytosol by MTHFD1 (8, 268) driven by the high ADP/ATP ratio in mitochondria (325), and the high ATP/ADP ratio in the cytosol (8, 326). The formyl-THF in mitochondria has at least two other fates. The formyl group is required for the production of formyl-methionine-tRNA, which is the starting amino acid needed for protein synthesis in mitochondria (327). One-carbon metabolism thus supports protein production in mitochondria and mitochondrial biogenesis (327), however this may only use a small per- centage of the one-carbon units present in mitochondria (8). Initiation of mitochondrial protein production does not strictly require formyl-methionine

29 (328). Another fate of the formyl-THF in mitochondria is “waste” of the formyl group as CO2, regenerating THF while producing NADPH in a reaction catalysed by ALDH1L2, a formyl-THF dehydrogenase (8, 268, 329), which may be a mechanism to consume excess formate (329). The same reaction occurs in the cytosol catalysed by ALDH1L1 (8, 268, 329). ALDH1L1 is downregulated in several cancer types, while ALDH1L2 is overexpressed in a number of cancers (268). This indicates that the formation of the antioxidant NADPH is important in mitochondria under the conditions of oxidative stress present in cancerous cells (268). Methylene-THF in the cytosol is mainly formed by reduction of formate by MTHFD1 (280, 282). Methylene-THF can be used by thymidylate synthase to methylate dUMP forming dTMP and generating dihydrofolate that is reduced to THF by dihydrofolate reductase (87, 270-273). Methylene-THF is also reduced by methylene-THF reductase to methyl- THF, used to form methionine from homocysteine (268, 276, 280). Methio- nine is used in protein production, and the formation of S-adenosyl- methionine (SAM) (268, 276, 280), a potent methylating agent used in epi- genetics (175, 268, 276). SAM is also one reason for the non-enzymatic formation of other methylated nucleotides (48, 49). ATP is required for SAM production, and homocysteine is used for the production of the antioxidant glutathione (GSH). Therefore the flow of one-carbon units can influence both ATP and GSH levels (268). As several reactions in the one-carbon metabolism either generate or consume NADP+/NADPH/NAD+/NADH, the redox level of the cytoplasm and mitochondria and their ability to reduce harmful ROS is influenced by the metabolic flow through this cycle (266, 268, 276).

Drugs for one-carbon metabolism: antifolates and antimetabolites The first antifolate developed was Aminopterin, which proved very successful in treatment of acute leukaemia in children (330). However, the drug also produced serious side-effects, including stomatitis, pharyngitis, diarrhoea, gastro-intestinal haemorrhage and bone marrow aplasia (331). Methotrexate (originally Amethopterin) was the first anticancer drug to cure a patient with a solid tumour (332) and is routinely used to treat acute lymphoblastic leukaemia and several other forms of cancer (87, 270, 331, 332). Methotrexate is not as effective as Aminopterin, but causes less serious side-effects and thus has a better therapeutic window (87, 331). Methotrexate inhibits dihydrofolate reductase, the enzyme responsible for reducing dihydrofolate to THF (87, 332). If dihydrofolate reductase is not functioning, thymidylate synthase will consume all available methylene-THF generating dihydro-

30 folate, causing a depletion of THF pools as the dihydrofolate accumulates (87). This leads to the inhibition of thymidylate and purine synthesis resulting in apoptosis (332). More recently, the antifolates Pemetrexed, Pralatrexate and Raltitrexed have been developed that optimise drug uptake and polyglutamation (87). Thymidylate synthase is targeted by the commonly used antimetabolite drug 5-fluorouracil (270, 333, 334), which leads to a decrease in thymine and genomic instability with incorporation of uracil instead of thymine (333, 334). 5-fluorouracil also forms 5-fluoro-dUTP that can be incorporated into DNA and together with increased incorporation of dUTP leads to DNA damage-induced cell death (333, 334). It has serious side-effects to healthy cells but is still used for the treatment of colorectal cancers (333, 334). Drugs that target the MTHFD family are not currently available for clinical use, however several folate analogues inhibit MTHFD1 in vitro (294). One of these inhibitors, LY345899 (294)(referred to as LY354899 in reference (335)), has anti-proliferative and cell cycle effects in leukaemia cells, and caused apoptosis after long incubation times (335). LY345899 is not specific for MTHFD1 as low levels of inhibition of several other folate- requiring enzymes were found (335). Several new strategies to develop clinical drugs by inhibition of metabolic enzymes are currently in development (270).

Importance of transport and polyglutamation of THF and drugs The major route for folate entry into cells is the Reduced folate carrier (RFC) (87). Antifolates also enter tumour cells via the RFC, and uptake is important for drug cytotoxicity (87). In cells, folates and antifolates are polyglutamated by the action of folylpolyglutamate synthetase (FPGS), which can add up to eight glutamates on the glutamyl tail of the folate (8, 87). Pol- yglutamated folates and drugs are retained within the cell and are not as good substrates for export carriers (8, 87). The retention of drugs in the cell after the antifolate levels in the blood drops is a major factor for the dosage times of the drugs (87). The polyglutamated drug may be a more potent inhibitor of the target protein, and can inhibit additional targets. For example, polyglutamated Methotrexate was found to be an inhibitor of thymidylate synthase as well as AICAR, one of the enzymes in the de novo purine synthesis pathway (87). Also, levels of polyglutamated Methotrexate are lower in healthy cells than in cancer cells during treatment, which may explain why these cells survive better than the cancer cells (87). Both drug uptake and polyglutamation were optimised when developing the new antifolates Raltitrexed, Pemetrexed and Pralatrexate (87). RFC and

31 FPGS have greater activity with Pralatrexate than with Methotrexate, which leads to higher levels of inhibition when Pralatrexate is used (87). FPGS has greater activity with Raltitrexed than with Methotrexate, and the polyglutamated form of Raltitrexed is a potent thymidylate synthetase inhibitor (87). Pemetrexed is also a potent inhibitor of thymidylate synthetase when polyglutamated. Polyglutamation of Pemetrexed enhances transport into the cell through the proton-coupled folate transporter, which makes Pemetrexed a good substitute for other antifolates in cancers defective in RFC (87). The folate-dependent enzymes may have different activity and affinity of the polyglutamated folates compared to the monoglutamed folates, as shown for MTHFD2L (8, 321) and MTHFD2 (336). Recently it has been shown that at physiological concentrations using polyglutamated substrates, MTHFD2 can use NAD+ and NADP+ with almost the same efficiency, which could influence the redox state of the cell by supplying NADPH (336). Folates are transported into mitochondria by carrier proteins as monoglutamated species, but become polyglutamated and are retained in the mitochondria (8). This may be one reason why transport between the cytosol and mitochondria of one-carbon-charged folates is low (8, 281).

MTHFD2 is upregulated in cancers and is a potential target for cancer phenotypic lethality Cancer cells activate the embryonic-specific MTHFD2 in order to drive production of one-carbon units for purine and thymidylate synthesis (317, 337). Both MTHFD2 mRNA and protein are upregulated in cancers and overexpression is linked to tumour cell proliferation (9, 274, 338). High levels of MTHFD2 is linked to cancer cell migration and invasion in breast cancer (277, 339-341), with poor prognosis in breast cancer patients (9, 342). Depletion of MTHFD2 mRNA by RNA interference can decrease cancer cell proliferation independent of tissue of origin (9). Several polymorphisms in MTHFD2 are associated with a higher risk of bladder cancer (343), and sensitivity of cancer cells to the antimalarial drug Artesunate is determined by MTHFD2 expression (344). Nuclear MTHFD2 supports cancer cell proliferation independent of its dehydrogenase activity, and therefore has a nucleic “moonlighting” function (274). MTHFD2 was in this case co-expressed with proteins involved in the S, G2 and M phases of cell proliferation (274). This group of proteins was also overexpressed in tumours (274). The mechanistic target of Rapamycin complex 1 activates the ATF4 transcription factor in response to growth signals (269). This in turn stimulates the expression of MTHFD2, the de novo serine synthesis pathway and other mitochondrial THF cycle proteins (269). Thus the production of one-carbon

32 units required for de novo purine synthesis increases in response to growth signals (269). As MTHFD2 is overexpressed in cancer cells compared to healthy cell types (338), including cells with high proliferation (9), and promotes the proliferation of cancer cells (274) it may be targeted for a cancer phenotypic lethal approach. Here the decrease of nucleotides in cancer cells would cause further genomic instability (345) and increase the possibility of cell death. Combination therapy with conventional chemotherapy drugs may be benefi- cial (338). Many of the common cancer drugs such as Methotrexate are also toxic to non-cancerous cell types with high proliferation rates such as immune cells, hair follicles and intestinal fibroblasts (9). A target that is more specifically expressed in cancer cells would be preferred and side-effects of current treatments could be avoided (338).

Structural information on the MTHFD family Within the human MTHFD family, only the structures of the MTHFD1 DC- domain in complex with NADP+ (292) and several folate analogues (294) had, prior to this work (Paper VI), been solved. A homology model of MTHFD2 was presented (307). This model was based on the DC-domain of the human MTHFD1 structure bound to the folate-analogue inhibitor LY345899 (294) and also the structures of homologues from E. coli (346) and S. cerevisiae (347). The homology model was used to elucidate potential phosphate and magnesium binding residues (307). Recently a MTHFD2L homology model was also presented (338), based on the human MTHFD1 DC-domain structure (292). I have recently solved the crystal structure of MTHFD2L to a resolution of 2.0Å. The MTHFD2L structure is highly similar to the crystal structure of MTHFD2 presented in Paper VI, but display significant structural differences. The structures of methylene-THF dehydrogenase and cyclohydrolases from other species have also been determined. These include structures from the bacteria Pseudomonas aeruginosa and Acinetobacter baumannii, and the parasites Leishmania major and Trypanosoma brucei (348-351).

34 Clostridium botulinum

“BoNT is an astonishing modular nanomachine that unites recognition, traf- ficking, unfolding, translocation, refolding, and catalysis in a single entity.” – Mauricio Montal (12)

Clostridium botulinum is a gram-positive rod-shaped bacterium, which produces the botulinum neurotoxins (BoNTs) and is the causative agent of the deadly disease botulism (10, 352-355). The name botulinum comes from the Latin word for sausage - botulus - as the toxin was first extracted from in- fected sausages (11). It grows strictly anaerobically (10), and under stress the bacteria forms spores that can survive standard cooking and food- processing procedures (10, 355). These spores are found ubiquitously in soil and in sediments (10, 355), and therefore often found in food (356). The spores are regularly ingested by humans, however the necessary conditions for spore germination and formation of vegetative bacteria are rarely achieved in food and the spores pass through the gut without any harm to the individual (355). The toxin itself is inactivated by heating at 85°C for 5 min (355), as opposed to the bacterial spores (357). The “botulinum cook” with heat treatment at 121°C for 2.4 minutes, has been developed for inactivation of spores during preparation of canned food (357). The strains of C. botulinum are divided into four groups with different characteristics (10, 357). Groups I and II are the cause of human disease (352, 357), while group III causes mostly animal disease (352). Group IV has been found in soil samples and have not been implicated in disease (352). C. botulinum strains are also divided by the neurotoxin serotype they produce (10). Eight serotypes of neurotoxin have been descibed, and these are further subdivided based on the gene sequence (10). The most recently dis- coved serotype, BoNT/X, is not detected by the developed antisera, and cleaves several unique substrates (358). Chimeric toxins can also form, for example the CD and DC toxins (10). BoNT is the most lethal toxin known to humankind, with intravenous LD50 values for humans of approximately 1 ng/kg (359). One gram of purified toxin could potentially kill 1 million people if administered effectively (354). The toxin is considered a significant bioterrorism hazard, and several government programs, groups and extremists have tried to develop BoNT as a bioweapon, but it has fortunately not been used (354, 355, 360).

35 Botulism The C. botulinum neurotoxin inhibits neurotransmitter release at peripheral nerve terminals causing flaccid paralysis in muscles and resulting in respiration failure and death (355). Hospital treatment involves respiration assis- tance for several weeks (355), and administration of antitoxin to counter neurotoxin present in the blood, but the efficacy of the antitoxin diminishes with disease progression (352). Animal botulism occurs more frequently than human botulism cases and is part of the bacterial lifecycle, as the bacterium can proliferate anaerobically in the decomposing carcass releasing toxin (10). Toxin-insensitive inver- tebrates such as maggots feed on the carcass and are in turn eaten by other animals thus spreading the disease (10). Several types of human botulism exist: foodborne, infant, wound, inhalational and iatrogenic (355). Foodborne botulism originates from food with preformed botulinum neurotoxin (10, 352, 355). Often the source is homemade canned food or mini- mally treated food that has been improperly stored (10, 352, 355). Infant botulism occurs in children under twelve months of age (10, 352, 355, 356, 361) and is the most frequent kind of human botulism (355). The gut flora of the child is not fully developed and ingested spores of C. botulinum can germinate and produce toxin (10, 352, 355, 356, 361). The human derived antitoxin BabyBIG has been able to reduce the length of hospitalisa- tion for infants with botulism by 50% since approved for clinical use in 2003 (361). Spores are often present in honey, which is therefore not recommended for children under twelve months (356). An adult version of infant botulism exists where the bacteria grow in the gut of adults after abdominal surgery, usage of antimicrobials that remove Clostridia competitors or in the presence of gastrointestinal wounds (352, 355). Wound botulism is seen in intravenous drug users, and comes from con- tamination of the wound by C. botulinum spores that can germinate in the anaerobic milieu of the abscess (352, 355). Inhalational botulism does not occur naturally, but one case study described inhalation of aerosolised BoNT by German laboratory workers in 1962 (352, 355). The injection of BoNT for cosmetic or therapeutic reasons can lead to iatrogenic botulism if the recommended dose is exceeded (355).

Botox and medical use of botulinum neurotoxins BoNT inhibits acetylcholine release from peripheral nerve terminals, thus relaxing muscles, and was suggested to be used for treatment of different types of movement disorders as early as 1817 (11). Today, four drugs con-

36 taining botulinum neurotoxin are available: Botox, Dysport and Xeomin use BoNT/A1, while NeuroBloc uses BoNT/B (11). Botox is widely used for the cosmetic treatment of wrinkles, this effect was discovered during treatment for blepharospasm (abnormal contraction or twitch of the eyelid) (353). BoNT is used in medical treatments for disorders with abnormal, excessive or involuntary muscle contractions, and proven effective against blepharospasm, cervical dystonia, upper-limb spasticity, lower-limb spasticity and chronic migraine (11). BoNT treatment of chronic migraine was discovered when migraine-sufferers treated with Botox to target facial wrinkles reported a decrease in the frequency and magnitude of migraine episodes (353). BoNT has also been used off-label to treat a number of other disorders (11).

Botulinum neurotoxin and mechanism of action BoNT is produced by the C. botulinum bacterium as a single polypeptide chain of 150 kDa forming three structural domains (Figure 9) (10, 12, 362). Activation by proteolytic cleavage of the chain produces two fragments, held together by a disulphide bridge (10, 12). The 50 kDa light chain is a zinc-dependent protease that can cleave the soluble N-ethylmaleimide- sensitive factor attachment protein receptor (SNARE) protein target: vesicle- associated membrane protein (VAMP), synaptosomal-associated protein of 25 kDa (SNAP25), or syntaxin (10, 12, 363). SNARE proteins are responsible for fusing vesicles containing acetylcholine neurotransmitter to the plasma membrane during nerve signalling (10, 12). The SNARE complex does not form if a SNARE protein is cleaved by BoNT, and vesicle fusion and neurotransmitter release is blocked, giving rise to paralysis (10, 12). The 100 kDa heavy chain of BoNT consists of two functional domains, each performing one function (10, 12). Two subdomains comprise the C-terminal segment of the heavy chain, HC (10, 12). HC binds to two different receptors on the nerve cell membrane, giving maximal affinity and specificity, in a dual receptor model (364). Firstly, HC binds to the terminal sugars of a ganglioside receptor, a complex sugar moiety with lipid tails that reside in the membrane in abundant numbers (10, 12, 365, 366). The HC also binds a protein receptor such as Synaptic vesicle protein 2 (SV2) (BoNT/A, BoNT/D, BoNT/E and BoNT/F (367-372)) or Synaptotagmin I and II (SytI and SytII, used by BoNT/B, BoNT/G and BoNT/DC (373-377)) giving serotype specificity (Table 1) (10, 12). BoNT binding to ganglioside (365, 378) or protein (379-384) receptors have been displayed in several crystal structures using various serotypes (10, 12). Following receptor binding the toxin is endocytosed into the nerve cell, and the acidic pH of the endocytosed vesicle causes structural changes in the BoNT structure (10, 12). The N- terminal heavy chain domain, HN, is highly α-helical, with one segment

37 known as the belt region wrapping around the light chain to act as a chaper- one and inhibitor of the light chain (385) before it reaches the cytosol of the target nerve cell (10, 12). The HN translocates the light chain into the neu- ronal cytoplasm (10, 12). A beltless HN creates an ion-conductive channel (386). The exact procedure for the translocation event is not known, but it is proposed that acidic pH causes the HN to become incorporated into the membrane, creating a pore (12). The light chain unfolds and moves through the pore into the cytoplasm (10, 12). The reductive environment of the cytosol results in cleavage of the disulphide bridge and the free light chain refolds and cleaves its substrate SNARE protein (10, 12).

Figure 9: BoNT structure and dual receptor binding. Prepared from BoNT/A (PDB 3BTA; LC, HN) and BoNT/B (PDB 4KBB; HC, GD1a, Syt II) (362, 383). Figure made using PyMol (210).

38 Table 1: The BoNT serotypes Serotype Discov- Discovered Uses gang- Protein Cleavage Gene Clus- ered in lioside? receptor substrate(s) ter A 1897 Human Yes SV2 SNAP25 HA or OrfX B 1904 Human Yes Syt VAMP HA C 1922 Animals Yes unknown SNAP25, Syntaxin HA D 1929 Animals Yes SV2 VAMP HA E 1934 Human, Fish Yes SV2 SNAP25 OrfX F 1960 Human Yes SV2 VAMP OrfX G 1970 Soil Yes Syt VAMP HA X 2017 Human Likely* unknown VAMP OrfX

* BoNT/X ganglioside binding has not yet been tested, but it is proposed to bind gangliosides based on the canonical binding sequence being present.

BoNT gene clusters In all strains of C. botulinum, the bont gene is located next to several other genes that form a cluster (387, 388). This occurs regardless of whether the bont gene is located in a plasmid or in the bacterial chromosome (387, 388). The haemagglutinin (HA) gene cluster is present for BoNT/B, BoNT/C, BoNT/D, BoNT/G and some subtypes of BoNT/A, including the most clini- cally relevant BoNT/A1 (387, 388). The open-reading-frame-X (OrfX) gene cluster is found in BoNT/E, BoNT/F, BoNT/X and several subtypes of BoNT/A (387, 388), for example BoNT/A2 used for the work presented in this thesis. The closely related BoNT/A1 and BoNT/A2 (90% identity (389)) are thus found in different types of gene clusters. BoNT is produced by C. botulinum as part of a protein complex called the progenitor toxin complex (PTC) (19). The PTC comes in different sizes, with the minimal PTC (M-PTC) at 300 kDa, and the large PTC (L-PTC) at up to 900 kDa (19). The size of the PTC varies between drugs containing BoNT (11). Both BoNT gene clusters contain BoNT as well as the non-toxic non- haemagglutinin protein (NTNH) gene in addition to several genes specific for that gene cluster (387, 388). The BoNT and NTNH proteins are the components of the M-PTC (18, 19, 390-393), which is found in cultures from all C. botulinum strains (19). NTNH is structurally similar to BoNT with a similar size of approximately 150 kDa (18, 391). However, several critical residues needed for activity and receptor binding are lost in NTNH (18, 19). Structures of the M-PTC from BoNT/A (18) and BoNT/E (391) have shown that NTNH and BoNT interact in a tight complex resembling an interlocked handshake. This interaction can protect BoNT from degradation by the low

39 pH and proteases present in the stomach, and NTNH increases the oral toxicity of BoNT (18, 19).

Figure 10: The L-PTC protective complex. M-PTC is NTNH and BoNT only. Pre- pared from pdb 3WIN (HA-complex) and pdb 3V0A (M-PTC) (14, 18). Figure made using PyMol (210).

HA cluster The HA cluster contains genes for three haemagglutinin proteins, together with BoNT and NTNH, named HA17, HA33 and HA70 due to their sizes (19). The cluster also contains a gene for the regulatory protein BotR, which is an alternative sigma factor responsible for regulation of toxin production in C. botulinum (394). The HA proteins build up a large triskelion complex (14) which also bind NTNH in the M-PTC, forming the larger PTCs (15). The crystal structure of the HA complex has been solved (14), and the HA proteins assist the toxin in crossing the epithelial barrier of the stomach to reach the blood stream (14, 16, 17, 395, 396). The HA complex binds carbo- hydrates and E-Cadherin to relax the tight coupling of the epithelial cell-to- cell adhesion, allowing BoNT to enter the bloodstream, increasing the oral toxicity of BoNT (14, 16, 17, 395, 396). A low-resolution cryo-electron microscopy structure of the full L-PTC (~760 kDa) has been solved (Figure 10) where the BoNT/NTNH complex is situated on top of the triskelion complex of HA proteins, with NTNH interacting with the HA70 center component of the HA complex (15).

40 OrfX cluster The OrfX cluster contains genes for the proteins OrfX1, OrfX2, OrfX3 and P47, in addition to BoNT and NTNH (387, 388, 397). Previously only a limited initial characterization of the translation and expression of these proteins was performed (388, 398-401), and nothing was known regarding their structure and function (19, 387). The gene cluster also contains botR, having the same regulatory function in these C. botulinum strains as in the HA cluster strains (394). In strains with the OrfX cluster, previously only the M-PTC containing BoNT and NTNH had been identified (401-403). It has been proposed that the OrfX cluster proteins also form a complex with BoNT under certain conditions, however the complex was easier to disrupt than the HA cluster complex (402). OrfX2, OrfX3 and P47 of the OrfX cluster were expressed from Clostridium botulinum A2 cultures (401). The recently published structure of P47 from BoNT/E resembles that of the bactericidal/permeability-increasing protein (BPI) that belongs to the tubular lipid-binding (TULIP) protein family (404). The structural resemblance of the OrfX proteins to TULIP proteins was also one conclusion from Paper VII.

TULIP proteins The tubular lipid-binding (TULIP) protein superfamily has a similar fold based on the TULIP domain: a long α-helix wrapped in a β-sheet (20, 405). The first member of the family to be discovered was bactericidal/permeability-increasing protein (BPI), whose structure contains two such domains (Figure 11) (406, 407). Later the structures of several other eukaryotic proteins where found to have the same fold (20, 405-421). The protein superfamily members were determined by sequence homology (20, 405). Several structures of TULIP family proteins display bound lipids or other long hydrophobic compounds in a cavity between the central α-helix and the surrounding β-sheet (Figure 11) (406, 408, 410-412, 416-418). The TULIP resemblance of many TULIP family members was first seen by structural studies, and not from sequence alignment, and further family members may yet be discovered in this way (422). Only eukaryotic proteins have been annotated as TULIP family members (20, 405), however the structure of lipoprotein YceB from E. coli was recently solved (PDB 3L6I), showing the structure of a TULIP protein (422). In DALI structural comparison it is structurally similar to BPI (422). Therefore TULIP proteins also exist in bacteria, and evolutionary the TULIP proteins must have evolved earlier than previously believed (422).

41 Figure 11: Crystal structures of several TULIP proteins, along with N-terminal domains of OrfX2 (residues 6-168) and P47 (residues 35-175) from Paper VII, shown in cartoon mode highlighting secondary structure. Bound lipids are shown as green sticks. PDB codes: 1BP1, 2RCK, 4M4D, 3UV1, 2OBD, 4P42, 6EKT and 6EKV ((406, 408, 413, 416, 417, 419) and Paper VII). Figure made using PyMol (210).

42 Present investigations

Table 2: Structures solved by me for the projects in this thesis. Protein* Ligands used in Unique Published Reference crystallization trials structures structures** hsMTH1*** 54 23 5 Paper I, Paper IV, Paper V drMTH1 2 2 2 Paper II, Paper V mmMTH1 1 1 1 Paper III MTHFD2*** 11 6 1 Paper VI OrfX2 0 1 1 Paper VII * hs, homo sapiens, human; dr, danio rerio, zebrafish; mm, mus musculus, mouse ** Published or included in manuscripts in this thesis. *** For MTH1 and MTHFD2 most structures are used for structure-based drug design and will not be published.

Paper I – MTH1 validated as a drug target for cancer and first-in-class drug candidates presented Several key experiments in this publication validated MTH1 as a good cancer target. Using small interfering RNA (siRNA) the MTH1 protein was depleted in several cancer cell lines. This resulted in reduced survival and viability of the cancer cells, which could be reversed by the expression of MTH1 from a RNA interference-resistant expression construct. The expression of an MTH1 mutant (E56A) with no catalytic activity did not rescue the cancer cells. The MTH1 knockdown also resulted in the accumulation of DNA damage, measured as an increase of 8-oxo-dG and p53 binding protein 1 foci, indicating an increased number of double-stranded breaks. Several parts of the apoptosis pathway were activated when MTH1 was depleted. A non-malignant cell line VH10 was not affected by the MTH1 knockdown. By reducing MTH1 expression with induction of short hairpin RNA (shRNA), mice that carried a human derived tumour survived longer and the cancer tumour volume decreased. Since the effect of the MTH1 knockdown could be reversed with expression of a functional MTH1, but not with a catalytic dead mutant (E56A), it was proposed that MTH1 inhibition by small molecules could achieve the same effect as an MTH1 knockdown. A good starting point for a small molecule inhibitor was found by screening compound libraries. The identified

43 compounds were optimised by organic chemistry to achieve even better potency and stability. Several crystal structures, many contributed by me (Table 2), of these molecules in complex with MTH1 greatly aided the structure-based drug design process. This resulted in the potent inhibitors TH287 and TH588, which had IC50 values of 0.8 nM and 5.0 nM, respectively. TH588 is an improved version of TH287 that increases the metabolic stability of the compound. X-ray crystal structures of both TH287 and TH588 in complex with MTH1 were solved to a resolution of 1.6Å, with the TH588 structure being my contribution to Paper I (Figure 12). The addition of the cyclopropane group in TH588 forced the dichlorobenzene moiety to rotate 180 degrees to fit into the binding pocket of MTH1. Targeting MTH1 with inhibitors was found to be effective against cancer cells in vitro by inducing DNA damage. These inhibitors were also able to kill cancer cell lines but had no effect on non-malignant cell lines. In xeno- graft mice with tumours of different origins (melanoma, colorectal cancer and breast cancer) MTH1 inhibitors also reduced the tumour volume. Overall it was demonstrated that the cancer phenotypic lethal approach of targeting MTH1 could take advantage of the deregulated metabolism and high ROS levels in cancer, and cause toxic DNA damage to these cells. It was also shown that MTH1 was upregulated in most cancer types compared to healthy cells. After this study several other studies have opposed this view, while others have supported it. For more on this topic, see the section “MTH1 controversy” presented earlier in this thesis.

Paper II – Hypoxia and oxidative stress makes cells sensitive to MTH1 inhibition This paper studied the redox environment in the cell and the influence of hypoxia on MTH1 sensitisation. In Paper I, TH588 was determined to be non-toxic to BJ hTERT cells, which are a non-cancer cell line. Pre-treatment of the cells with KBrO3, a non-alkylating oxidation reagent, increased the levels of oxidative stress, making the cells more sensitive to MTH1 inhibition with TH588. Similar results were obtained in other cell lines. In a similar manner, treating cells from the cancer cell line U2OS with the antioxidant N-acetyl-L-cysteine made them more resistant to MTH1 inhibition. This resulted in greater cell survival, with less 8-oxo-dG being incorporated into DNA.

44 Figure 12: The crystal structures of MTH1 prepared for the papers and manuscripts in the thesis. Protein structures are shown as cartoon models, with the amino acid residues N33, W117, D119 and D120 shown as sticks. Bound compounds are presented as sticks and hydrogen bonds are shown as black dashed lines. Human MTH1 (green), zebrafish MTH1 (yellow), mouse MTH1 L116M mutant (light blue), TH588 (cyan), compound 2 (magenta), compound 1a (grey), compound 4b (orange), O6-methyl-dGMP (light brown). PDB codes: A) 4N1U, B) 5HZX, C) 6EHH, D) 5NGR, E) 5NGS, F) 5NGT, G) 5OTM, H) 5OTN (Paper I-V). Figure made using PyMol (210).

Zebrafish was established as a good model system to study MTH1 biology. The structure of zebrafish MTH1 in complex with TH588 indicated a similar binding mode to human MTH1. Furthermore, the enzyme activity and inhibition by TH588 were also similar. A colleague and I determined the crystal structure of zebrafish MTH1 bound with TH588 (Figure 12). I also aided in structure and sequence evaluation. Analysis of this structure confirmed that no residues in the active site interfered with TH588 binding.

45 As zebrafish eggs can be injected with nucleotides, another key experiment could be performed. 8-oxo-dGTP and 2-OH-dATP were injected into zebrafish eggs, which survived the increased nucleotide pool oxidation. When MTH1 activity was inhibited by TH588, most of the eggs did not survive. This showed that MTH1 is needed to remove oxidised nucleotides from zebrafish nucleotide pools. VHL is involved in the cellular degradation of hypoxia-induced factor 1α (HIF-1α) (127). Zebrafish are useful for studying VHL knockouts as their embryos survive for 7 days after fertilization, while VHL knockout mice die in utero. The loss of functional VHL, and thus stabilisation of HIF-1α dra- matically sensitises the embryos to MTH1 inhibition. In addition, overexpression of HIF-1α, which is seen in many cancers, could be replicated in the zebrafish embryos by the addition of DMOG. These embryos became dependent on MTH1 for their survival, however they could also be rescued by treating the cells with various antioxidants. It was concluded that oxidative stress, or the induction of hypoxia, which increases ROS production, sensitises the cells to MTH1 inhibition. Since these factors are often present in cancer cells, MTH1 inhibition could be a means of inducing cancer cell death selectively, as these cells depend on MTH1 for their survival.

Paper III – Differences in MTH1 affinity between species and suggestions for pre-clinical studies As drugs are progressed towards clinical studies, it is important to assess their toxicity in several model systems. These model systems can include dog, mouse, rat and pig. In Paper III, MTH1 protein from the aforemen- tioned species, in addition to zebrafish and human MTH1 were expressed and purified. The inhibition of TH588 towards the MTH1 orthologues was tested and the IC50-values for human, dog, rat, pig and zebrafish were found to be similar. However, TH588 was a considerably weaker inhibitor of mouse MTH1 as evidenced by the 20-fold higher IC50-value. Structures of mouse MTH1 with TH588, mouse MTH1 with 8-oxo-dGTP and dog MTH1 gave insights into the small differences that occur between species. A methionine residue positioned close to the active site was found to be a determi- nant for TH588 inhibition. This amino acid is conserved across all species, with the exception of mouse MTH1, which has a leucine at this position. Mutation M116 to leucine in human MTH1 increased the IC50-value of TH588, making it a worse inhibitor. In contrast, changing the leucine to a methionine in mouse MTH1 made it more similar to human MTH1. The structure of the mouse MTH1 L116M mutant shows that the methionine superimposes well with the methionine in human MTH1. This could explain

46 the more human-like behaviour of the mouse MTH1-mutant. My contribution to this project was to determine the structure of mouse MTH1 L116M mutant in complex with TH588. In addition, I assisted in sequence and structure comparisons and evaluations. Overall Paper III showed that small changes close to the MTH1 active site between species could cause differences in enzyme inhibition by TH588. Due to the higher IC50-value of TH588 displayed against the mouse enzyme, the MTH1 proteins from rat, dog, pig or zebrafish would be better model systems for human MTH1 biology. It is possible that the MTH1 orthologues could behave differently in the presence of other inhibitors. This should be taken into account before the choice of model organisms when progressing MTH1 inhibitors towards clinical trials.

Paper IV – Virtual fragment screening for MTH1 supported by X-ray crystal structures In Paper IV a different approach was tried for developing MTH1 inhibitors. The original inhibitors for MTH1 were found using large compound screening efforts. In Paper IV, 0.3 million of commercially available compounds from the ZINC database were tested by docking to the MTH1 crystal structure in silico. The results were ranked, and 22 of the highest-ranking hits were experimentally tested, with 12 of the compounds inhibiting MTH1. Commercial analogues with high docking scores for five of the fragments were purchased and tested. In some cases optimisation of the original fragment into the analogues increased MTH1 inhibition by several 100-fold. The fragment-based docking strategy was thereby an effective method of finding and optimising new MTH1 inhibitors with minimum need for organic chemistry. My contribution was the determination of several crystal structures of MTH1 with the bound fragment or optimised analogue, in order to validate the docked position. The positions of the fragments were investigated by X-ray crystal structures (Figure 12). In one case a structure from the recent literature provided validation for the binding mode of the related fragment. In three other cases the crystal structures of the original fragment or optimised inhibitors bound to MTH1 were solved. Of these three structures, two were in agreement with the docking mode determined in silico, whereas the third (compound 2) bound in a different manner. This was due to the position of N33 in the binding site, which was shown to have moved when compared to the original crystal structure (MTH1 with TH588). As the docking procedure does not allow for protein movements, this likely resulted in the incorrect positioning of the fragment. When our new crystal structure was used for docking, the mode identified did match that of the fragment in the crystal structure.

47 This methodology could provide a means of screening large numbers of fragments for a target in silico and developing hits further. However, a drawback of this method is the need for a good starting structure for docking.

Paper V – MTH1 can degrade O6-methyl-dGTP In this study MTH1 was found to have a novel function, the degradation of methylated nucleotides. In Paper V, MTH1 was tested against different methylated deoxyribonucleotides, which were methylated using methyl me- thane sulfonate. MTH1 was shown to be active against methylated dGTP, dATP and was slightly active towards methylated dCTP. O6-methyl-dGTP, N2-methyl-dGTP and O6-methyl-GTP was tested in an MTH1 activity assay. MTH1 degraded O6-methyl-dGTP and O6-methyl-GTP better than non-methylated dGTP and GTP. MTH1 is the only NUDIX protein that showed high levels of this activity, but NUDT17 and NUDT18 displayed some O6-methyl-dGTP degradation at high protein concentrations. The O6- methyl-dGTP activity of MTH1 was shown to be present in humans, mouse, rat, pig, dog, zebrafish and plants (the AtNUDT1 of Arabidopsis thaliana), but not in the bacterial homologue E. coli MutT. This activity must have therefore evolved after the separation of the bacterial and eukaryotic branch- es of the phylogenetic tree during evolution. Similar to the studies in Paper II, when O6-methyl-dGTP was injected in- to zebrafish eggs in combination with TH588 (a potent zebrafish MTH1inhibitor), the eggs died. When no MTH1 inhibitor or an inhibitor that does not target zebrafish MTH1 (TH1579) was added, the zebrafish eggs survived the increased levels of methylated nucleotides. This shows that MTH1 activity towards O6-methyl-dGTP is important for cell survival. A colleague and I solved the structures of zebrafish and human MTH1 with the cleaved product O6-methyl-dGMP (Figure 12), and I was also in charge of structure evaluation. In both structures the methyl group points towards a highly hydrophobic pocket, and the nitrogenous base is moved 1.1Å towards this pocket, giving rise to a new hydrogen-bonding pattern. Zebrafish MTH1 in complex with O6-methyl-dGMP was determined at 0.99Å resolution. This structure featured several calcium ions close to the Nudix box in the active site, where one would normally expect the magnesium ions required for MTH1 activity to bind. This shows that MTH1 has evolved several functions for the removal of damaged nucleotides from free nucleotide pools.

48 Paper VI – The first MTHFD2 structure and inhibitor Prior to this study the structure of MTHFD2 had not been solved, and no inhibitors for the protein had been developed. In Paper VI, the inhibitor LY345899, which had previously been shown to inhibit MTHFD1, was found to also inhibit MTHFD2. LY345899 is a substrate analogue for the natural substrates of both MTHFD1 and MTHFD2. The IC50-values of Ly345899 were 662 nmol/L for MTHFD2 and 96 nmol/L for MTHFD1. The inhibitor was shown to interact with MTHFD2 in vitro using differential scanning fluorimetry, as well as in lysates of U2OS cancer cells as determined using cellular thermal shift assays and drug affinity responsive target stability tests. There was no target engagement observed in intact cells, which was speculated to be due to the poor cell permeability of the inhibitor. I determined the X-ray crystal structure of MTHFD2 with bound inhibitor + and the cofactors NAD and Pi (Figure 13). The protein fold was similar to MTHFD1 DC-domain (292, 294) and was shown to be a symmetric dimer with inhibitor and cofactors bound to both monomers. The binding site for the inhibitor was similar to that of MTHFD1, however the cofactor site dif- fered in several ways. An insertion loop from one MTHFD2 monomer is assisting in binding the free inorganic phosphate in the other monomer. The phosphate rests close to the ribose moiety of NAD+, where the phosphate of NADP+ would be positioned. One of the residues that bonds to the free phosphate is an aspartate, with the hydrogen bond distance being short (only 2.4Å). This has previously been shown to be important for binding hydrogen phosphate (the most common form of phosphate at physiological pH), instead of phosphate or sulphate (423). Drugs targeting MTHFD2 are attractive as the enzyme is overexpressed in cancer cells but is not present in healthy cells. Therefore it should be possible to selectively kill cancer cells while sparing healthy cells, avoiding the side-effects observed for other antimetabolite drugs (9, 87). The MTHFD2 crystal structure with the first identified inhibitor provides the starting point for developing potent and selective MTHFD2 inhibitors.

Paper VII – Proteins from the BoNT OrfX gene cluster are TULIP proteins which bind lipids The proteins of the OrfX type gene cluster were of unknown function and structure. In Paper VII we attempted to determine the crystal structures of these proteins, and to determine what their functions are. The sequences of the proteins themselves did not give any information on their structures and functions, as no similar proteins with known structures or functions could be found from sequence comparisons. For this project I expressed and purified

49 most proteins of the gene cluster from C. botulinum strain Kyoto-F and performed differential scanning fluorimetry in buffers of two different pH’s. The inactivated BoNT/A2 and NTNH did form a minimal PTC as expected. In initial trials no other protein interactions could be seen amongst the proteins of the gene cluster, and none interacted with the minimal PTC or an E- Cadherin construct. The OrfX3 protein did interact with the membrane when expressed as full-length, while a shorter construct without the putative membrane insertion region could be purified as a soluble protein. Crystal structures of P47 and OrfX2 were solved using single-wavelength anomalous diffraction. I was responsible for the OrfX2 crystal structure (Figure 13) and also produced the selenomethionine labelled OrfX2 required for structure solution. I also performed the structure comparisons. Both structures displayed domains with a fold similar to that of TULIP proteins (Figure 11). Using the DALI search engine it was shown that the OrfX2 and the P47 protein structure are related to the TULIP proteins. As TULIP proteins often bind lipids, a lipid-binding assay using lipid strips was performed for OrfX1 and OrfX2. This showed that these proteins could bind phosphati- dylinositolphosphate lipids, depending on their headgroup. Therefore, the OrfX gene cluster proteins are amongst the first bacterial proteins with the TULIP fold, previously only found in eukaryotes. It is clear however, that much remains to be discovered regarding these proteins.

Figure 13: Structure of MTHFD2 (left) and OrfX2 (right), shown in cartoon mode. For MTHFD2 one monomer is coloured rainbow with blue at N-terminus going to red at C-terminus. The other monomer in the dimer is coloured grey. LY345899 is + coloured cyan, NAD in purple and Pi in standard colours, all shown in sticks. OrfX2 coloured in rainbow with blue at N-terminus going to red at C-terminus. PDB codes: 5TC4 (MTHFD2) and 6EKV (OrfX2) from Paper VI and Paper VII. Figure made using PyMol (210).

50 Conclusion and outlook

The major conclusions from the work presented here: • MTH1 has been validated as a cancer target. • First-in-class inhibitors for MTH1 are presented. • Inhibitors for MTH1 can reduce tumour volume, kill cancer cells and spare healthy cells. • Hypoxia or oxidative stress sensitises cells to MTH1 inhibition. • The mouse is not a good model organism for TH588 inhibition. However, the inhibition of the model organism orthologue is valuable prior to pre- clinical trials. • Virtual fragment screening can provide scaffolds for drug development in a cost-effective way and optimise them into potent inhibitors. • MTH1 can perform a new function in addition to known functions, the degradation of methylated nucleotides. • The first structure of MTHFD2 with the first inhibitor can provide the starting point required for future drug development targeting this interesting protein. • The BoNT OrfX gene cluster encodes several bacterial TULIP proteins that can bind lipids.

Some prospects for the future: • The MTH1 inhibitors have been further optimised since the publication of Paper I, and one inhibitor has been included in a clinical phase I trial, in- volving a dose escalation study in cancer patients with advanced solid malignancies, performed at Karolinska Universitetssjukhuset, Sweden (www.clinicaltrial.gov/MASTIFF). Hopefully the results will be promis- ing and the drug candidate continued for further clinical studies. • The method of virtual fragment screening can be applied to other targets. The inclusion of chemical modifications of the analogues by organic chemistry could expand the toolbox even further. • The influence of MTH1 breakdown of methylated nucleotides in cancer and healthy cells requires further studies. • The development of MTHFD2 inhibitors will be continued and the effects of inhibiting MTHFD2 need to be further elucidated. In order to develop selective inhibitors for MTHFD2, inhibitors should be tested against a

51 wide range of folate-dependent enzymes, including MTHFD1 and MTHFD2L. • The possibility of the OrfX cluster proteins forming a complex with BoNT and NTNH is still an open question that requires more studies. The structures of the OrfX1 and OrfX3 proteins could at least in part resemble the TULIP fold based on the presented results. Solving these structures would be very helpful. Further lipid-binding studies are also needed. • The OrfX cluster proteins should be tested for effects on the oral toxicity of BoNT, and by what mechanism these effects arise. The purpose of the lipid-binding ability of OrfX cluster proteins needs to be elucidated.

52 Popular science summary

Proteins are small components of all living things such as humans, animals, plants and bacteria as well as non-living viruses. Proteins have many different functions, appearances and sizes. Similar to how the catalyst in a car speeds up the breakdown of harmful substances in the exhaust fumes, proteins can speed up processes necessary for life to be possible, such as the breakdown of certain substances and speed up the build up of others. All proteins are made up of a string of connected amino acids, which is folded into a final structure. The positioning of the amino acids close to one another in the active site, determines the function of the protein. X-ray crystallography is one method used to give detailed three- dimensional images of a protein, by determining the position of all atoms in the protein. This information can be used to create drugs that bind to a certain part of the protein, often the protein active site, rendering the protein non-functional. A detailed picture of the protein can also give insights as to what the function of the protein is. In this thesis X-ray crystallography has been used for both purposes, in several projects of medical relevance.

Cancer and the search for new ways to defeat it A cancer tumour is a group of cells in the body that grow and divide uncon- trolled manner. The Swedish Cancer Society has calculated that every third person in Sweden will develop cancer during their lifetime. A cancer diagnosis weighs heavy on the person affected, as well as their family and friends. Many forms of cancer are treatable today, but many chemotherapy drugs have serious side-effects as they are not specific enough. Unfortunate- ly these drugs can harm and kill normal healthy cells in addition to cancer cells. One way to develop new treatments is to find a protein upon which the cancer cells depend for their survival, but which is not necessary for healthy cells. A drug specifically targeting such a protein would kill cancer cells, while healthy cells would be spared, greatly reducing the side-effects. In this thesis two such proteins, MTH1 and MTHFD2, are described. MTH1 is an enzyme that can breakdown damaged DNA building blocks. These damaged building blocks can give rise to DNA damage and lead to cell death if they are not removed before their incorporation into DNA. Can- cer cells produce more of the chemical substances that harm DNA building

53 blocks. Cancer cells are therefore dependent on MTH1 for their survival, whereas healthy cells are not. The first paper of this thesis showed that cancer cells are dependent on MTH1 for survival, while healthy cells can survive in the absence of MTH1. Tumours implanted into mice, decreased in volume when MTH1 was removed or inhibited by different drug candidates. Several X-ray crystal structures of MTH1 with bound drug candidates were determined, which aided the development of effective inhibitors. Paper two showed that if cells have many damaged DNA building blocks they are also more sensitive to MTH1 inhibition. In addition, in several cell conditions that lead to the development of harmful chemicals able to damage the DNA building blocks the cells were more sensitive to MTH1 inhibition. Paper three proved that MTH1 from different animal species such as mice, dog and zebrafish are very similar to human MTH1. Small differences between these proteins in their active sites, may have a huge impact in how well a drug targeting MTH1 can bind and inhibit the enzyme. This has an impact on the choice of animal species used for animal trials to determine drug toxicity before human trials. Due to the small changes near the active site of mouse MTH1, it was concluded that other species such as rat are more suitable for animal trials, as their MTH1 is more similar to human MTH1. Paper four displayed the result of a new study to develop effective inhibitors for MTH1. A computer program tested hundreds of thousands of molecules to see which could potentially bind MTH1. When tested in the laboratory, several of these molecules where good inhibitors of MTH1. Structures of MTH1 with several of the inhibitors were determined with X-ray crystallography. In several (but not all) cases, the experimental binding mode matched the binding mode predicted by the computer program. Paper five showed that in addition to taking care of the breakdown of DNA building blocks damaged by oxidation, MTH1 can also breakdown DNA building blocks that have been methylated, meaning that a carbon- group have been added to the building block. This is a previously unknown function for MTH1, which may be related to its importance for cancer cell survival. The protein MTHFD2 is important for the cancer cell to be able to form new DNA building blocks, which are needed for the cell to be able to divide. MTHFD2 is found in cancer cells but not in healthy cells, making the enzyme an attractive drug target. The inhibition of MTHFD2 could kill cancer cells without affecting healthy cells, reducing side-effects during treatment. In paper six the first inhibitor for MTHFD2 was described, and the structure of MTHFD2 in complex with the inhibitor was determined by X-ray

54 crystallography. This information will help the development of drugs targeting MTHFD2.

The world´s most poisonous toxin and the proteins that help it The bacterium Clostridium botulinum produces the most potent poison in the world, the botulinum neurotoxin, of which one gram could kill one million people. The toxin severs the connection between nerve cells and muscles by stopping the release of the neurotransmitter acetylcholine. The dangerous disease botulism is characterised by paralysed muscles, and can lead to death by suffocation if the patient does not receive respiratory aid. Despite its toxicity the botulinum neurotoxin is, in very small doses, an important drug for the treatment of squint, extreme sweating and spasms amongst others. The toxin is used for treatment of more than 80 different diseases. The most famous drug is Botox, most known for its use in cosmetic treatment of wrinkles. The bacteria also produces several proteins which attach to the toxin, helping it through the stomach to the bloodstream, after which the toxin can find nerve cells to attacks. The genes of these proteins are situated right next to the gene of the toxin forming what is called a cluster. In paper seven proteins from one of the two clusters of proteins believed to help the toxin, have been studied. The appearance and function of these proteins have previously been unknown. The structures of two proteins, P47 and OrfX2, from this group of proteins from Clostridium botulinum were determined. These protein structures were similar to a group of proteins previously not found in bacteria. The structures indicated that the proteins could potentially bind fatty acids. Following experiments showed that they do in fact bind to several kinds of fatty acids.

The results of the work presented in this thesis have contributed to the continuous struggle to find new treatments for cancer. The work in this thesis has also given new knowledge on the structure and function of the proteins that help the botulinum toxin.

56 Populärvetenskaplig sammanfattning

Proteiner är små delar av allt levande, såväl människor, djur, växter och bakterier men också icke-levande virus. De kan ha många olika funktioner, utseenden och storlekar. Liksom katalysatorn på en bil påskyndar nedbrytningen av skadliga ämnen i avgaserna, så kan proteiner påskynda reaktioner som behöver ske för att livet ska vara möjligt, till exempel nedbrytningen av vissa ämnen och uppbyggnaden av andra ämnen. Alla proteiner består av en lång rad aminosyror som är sammansatta och därefter ihop-veckade till sin slutgiltiga form. Proteinets struktur bestämmer vilken funktion det har, genom att olika viktiga aminosyror kommer nära varandra och bildar en aktiv yta. Röntgenkristallografi är en metod som ofta används för att ge en detaljerad tredimensionell bild av hur ett protein ser ut, genom att bestämma var alla atomer sitter. Informationen kan användas för att försöka skapa läkeme- del som kan binda in till en speciell del av proteinet, ofta på proteinets aktiva yta, och få det att sluta fungera. I andra fall kan en detaljerad bild av proteinet visa vad ett protein med okänd funktion gör. Röntgenkristallografi har i denna avhandling använts i båda dessa syften till flera olika projekt med medicinsk relevans.

Cancer och sökandet efter nya sätt att besegra den En cancertumör är en grupp av cancerceller i kroppen som växer okontrolle- rat och inkräktar på andra vävnader, och skadar deras funktion. Cancerfon- den har beräknat att var tredje person i Sverige kommer att drabbas av cancer någon gång under sin livstid. En cancerdiagnos är tung för den drabbade, men även familj och vänner drabbas hårt av ett cancerbesked. Många can- cerformer går idag att behandla, men mycket ofta har de cytostatika (cellgif- ter) som används svåra biverkningar eftersom de inte är så specifika. Cy- tostatikan kan skada och slå ut inte bara cancerceller utan även vanliga friska celler. Att hitta ett protein som cancercellerna behöver för sin överlevnad, men som friska celler inte behöver, är ett sätt att kunna utveckla nya behandlingar av cancer. Om ett sådant protein slås ut med ett läkemedel kan cancercellerna dödas, och biverkningar undvikas. I denna avhandling beskrivs två av dessa proteiner, kallade MTH1 och MTHFD2.

57 MTH1 är ett enzym som kan bryta ner skadade byggstenar till cellens DNA. De skadade byggstenarna ger upphov till DNA-skador och leder till cellens död om de inte tas om hand i DNA eller innan de sätts in i DNA. Cancerceller bildar extra många kemiska föreningar som kan skada DNA- byggstenarna och är därför ofta beroende av MTH1 för att överleva, medan friska celler inte alls har ett lika stort behov av att MTH1 fungerar. Den första artikeln i avhandlingen bekräftar bland annat att cancerceller behöver MTH1 för att överleva, medan friska celler överlever om MTH1 tas bort. Tumörer som inplanterats i möss krymper i volym när MTH1 tas bort eller hämmas med olika molekyler. Flera strukturer av MTH1 med olika hämmare bestämdes med röntgenkristallografi vilket hjälpte utvecklingen av de effektivare varianterna av hämmare. Artikel två visar att om cellen har många skadade DNA-byggstenar så är de mycket känsligare för om MTH1 tas bort eller inte. Även vid tillstånd i cellen som leder till bildandet av många skadliga föreningar är cellerna känsligare för om MTH1 hämmas. Artikel tre bevisar att MTH1 från olika djurarter som möss, hund och zebrafisk är mycket lika människans MTH1. Små skillnader, kanske bara en utbytt aminosyra i strukturen för MTH1 från de olika djurarterna, kan ha stor betydelse för hur effektivt ett läkemedel kan binda in. Det har stor betydelse för vilken djurart som används i studier av läkemedlens giftighet innan de sätts in i människoförsök. På grund av just en sådan minimal skillnad nära aktiva ytan av musens MTH1 borde till exempel råttor användas istället för möss för tester av läkemedel, då råttans MTH1 är mer likt människans. Artikel fyra förevisar resultatet från en ny studie för att få fram bra hämmare till MTH1, där ett datorprogram testat hundratusentals olika möj- liga molekyler för att undersöka vilka som passar bäst att binda till MTH1. När molekylerna sedan testades i laboratoriet var flera av dem bra hämmare för MTH1. Strukturer av MTH1 med flera av hämmarna kunde erhållas med röntgenkristallografi, och i flera (men inte alla) fall stämde de överens med hur datorprogrammet förutspått att de binder in till MTH1. Artikel fem visar att MTH1 inte bara kan ta hand om och bryta ned byggstenar till DNA som är skadade av olika syreföreningar, utan även genom så kallad metylering, då en kolgrupp satts på byggstenen. MTH1 har alltså en tidigare okänd funktion som mycket väl kan vara en del i hur viktigt proteinet är i cancerceller. MTHFD2 är ett protein som är viktigt för att cancercellen ska kunna bilda nya byggstenar till sitt DNA, vilket behövs för att den ska kunna dela sig. MTHFD2 finns i cancerceller men inte i friska celler. Att MTHFD2 enbart finns i cancerceller gör att MTHFD2 är attraktivt att kunna utveckla läkemedel mot. Att stoppa MTHFD2 skulle potentiellt sett skada en process som är viktig i cancerceller utan att påverka friska celler.

58 I artikel sex beskrivs den första hämmaren till MTHFD2, och strukturen av proteinet tillsammans med hämmaren har bestämts med röntgenkristallo- grafi. Den här informationen kan hjälpa utvecklingen av läkemedel som har MTHFD2 som mål.

Världens giftigaste gift och proteinerna som hjälper det Bakterien Clostridium botulinum tillverkar världens giftigaste gift, Botulinumtoxinet, där ett gram gift kan döda en miljon människor. Giftet bryter kopplingen mellan nerver och muskelceller genom att stoppa frisätt- ningen av signalsubstansen acetylkolin vid en nervsignal. Den farliga sjuk- domen botulism kännetecknas av förlamade muskler, som trots behandling kan leda till döden genom kvävning om patienten inte får hjälp med respira- tor. Trots sin giftighet är botulinumtoxinet i mycket små doser ett viktigt lä- kemedel för behandling av skelningar, extrema svettningar och spasmer bland mycket annat. Giftet används nu för att behandla mer än 80 olika sjukdomar. Det mest kända läkemedlet är Botox, vilket är mest känt för sin användning inom kosmetisk behandling mot rynkor. Bakterien tillverkar också flera andra proteiner som sitter ihop med toxinet och hjälper det igenom magen till blodet vid en infektion, så toxinet däref- ter kan hitta de nervceller det attackerar. Generna för dessa proteiner sitter precis bredvid toxinets gen vilket kallas för ett kluster. I artikel sju har proteiner studerats som ingår i ett av de två olika kluster med proteiner som man tror hjälper toxinet. Utseendet och funktionen för proteinerna från det här klustret har man tidigare inte känt till. Strukturen har bestämts för två proteiner, P47 och OrfX2, från den här gruppen ifrån Clo- stridium botulinum. Proteinernas strukturer liknar en grupp av proteiner som man tidigare aldrig sett i bakterier. Strukturerna indikerade att proteinerna skulle kunna binda till fettsyror. Efterföljande experiment har visat att de kan binda till flera sorters fettsyror.

Resultatet av arbetet i denna avhandling har bidragit och fortsätter att bidra till kampen för att finna nya behandlingar och läkemedel mot cancer. Arbe- tet i avhandlingen har också gett ny kunskap om hur de proteiner som hjäl- per botulinumtoxinet ser ut och har för uppgifter.

60 Acknowledgements

First of all I wish to thank my supervisor Pål Stenmark, for allowing me to come to Stockholm to do my PhD in your research group. Thank you for being a good-hearted, caring and devoted scientist. The projects you have given me have been perfect for me and allowed me to develop my skills in the lab and perhaps most importantly in crystallography. I feel very grateful for your support and continuous guidance. I feel honoured to have been your first PhD student to finish their studies in your lab, and I certainly hope I will not be the last one. Thanks to Martin Högbom, my co-supervisor. Thank you for excellent help when I have had questions, and for all your expert opinions at progress reports. Thank you for making all of Pål´s group feel welcome into your lab and thanks to both Pål and Martin for making us all feel as one big group. Lena Mäler, thanks for being my appointed mentor during my PhD studies. I have luckily not had many issues to discuss, but it has been helpful to know there is someone to talk to if needed. Thanks to Stefan Nordlund for your support and careful control of my progress as director of PhD studies at DBB, and thanks to Pia Ädelroth for excellently taking over the responsibilities as director of PhD studies towards the end of my studies. I want to thank Stefan Nordlund, Gunnar von Heijne and Peter Brzezinski for being my committee for the Big Exam, and to Stefan and Peter for directing my TIBS exam. I actually, after a while, have decided I liked our discussions. Thanks to all collaborators and co-authors for sharing your knowledge, time and effort together with me to reach further and achieve more. Especially thanks to Thomas Helleday for carefully directing the collab- orations on MTH1 and MTHFD2 along with your group members. I also want to greatly thank Ann-Sofie Jemth for all the experiments, protein production, purifications, activity assays and inhibition curves you have produced over the years for our common projects. Many times your experiments have been the starting point of new discoveries and research focuses. Also thanks for taking your time to read through this thesis so carefully. Thanks to Evert Homan for putting all my crystal structures to good use, and for being so keen on my opinions regarding them.

61 I want to, with all my heart, thank all former and current members of the StenHög lab for making it one big family. Thanks to Ronnie for helping me during my first time in the lab, and for working together for a long time as the only members of Påls group. I wish you all the best for the future and I hope that your own research group thrives up there in the cold harsh north. Thanks to Megan, Mohit and Magnus, who was part of the Stenmark group for a long time and gave so much laughter and good discussions. I want to thank the former Högbom group members Henryk, Charlotta, Karin, Julia, Matthew and Catrine for all discussions at the lunch table, and all help in the lab that you offered in both small and large issues. Thanks to the lunch-table crew for all your discussions on everything, absolutely everything: Michael, Markel, Dan, Camilla, Riccardo, Daniel, Emma and Geoffrey. Thanks to the late crew, Hugo, Vivek, Ben and Kristine, for being there when lab work consumed my normal lunch time. Thanks to Linda, first of all for doing much of the foundation for the MTH1 project before my arrival, and more recently for being my companion in the office. You will do great, I know it. Thanks to Ram, Juliene, Oskar, Sarah and Sara for coming into the lab and fill it with new blood and life. For careful reading of this thesis I want to specially thank Geoffrey, Emma and Sarah. Especially thanks to Mohit and Parul, and Kristine and Sudarsan for inviting me to your respective weddings in India. My trips to India are defi- nitely some of my most beloved moments. Thanks to all my project students over the years for your help in the projects: Markel, Nina and Julia, and all other students passing through the lab. I wish you all the best in your future endeavours. I wish to thank all the people at DBB that make the department really run, the secretariat and technical staff, and especially Charlotta, Maria, Alex- ander, Håkan and Matthew for all your help with various small and large issues. Thanks also to the rest of our small piece of the world, the A4 corri- dor people and the rest of DBB for all conversations and laughter at pubs, courses and in the lunchroom. Lina, Sara, Theresa and Mattias from the scientific communications course, I hope for many more lunches together. Jag vill här rikta ett speciellt stort tack till alla vänner från Linköping som hängt kvar efteråt och förhoppningsvis kommer göra det för evigt. Listan är för lång för att skrivas ut, men jag vill speciellt tacka Urfadderiet 2011. Martin, Rebecka, Jessika, Tom, Linnea, Linnéa, Karin, Shad, Ulrika och Mark, ni är för alltid mina dinosaurier, Gilla! Tomas, Amélie, Dick, Daniel, Rosemarie, Anna, Jesper, Helena och Henric, jag är för evigt tacksam för att få kalla mig er vän. Alla medlemmar och vänner i Röda Arméns Gosskör och QQQ, ni är alltid mina bästa sångarkompisar. Tack till alla vänner från Linköpings Shotokan Karate-Do Klubb, för ert stöd och er hjälp i min träning och utveckling över alla dessa år. Tack till

62 Per Holmén för att du har fortsatt att tro på mig och varit min tränare, moti- vation och piska under så lång tid. Tack för att din styrka kunnat bära så många andra, och även mig, och hjälpt mig utvecklas till en starkare och bättre människa. Tack till Tobias Holmén, för att förlusten av dig fick mig att välja att forska om cancer. Tack till Patrik, Karl-Johan och Sara för att ni ville köra på järnet mot det andra svarta bältet med mig. Också tack till Simo och alla andra från Zanshin Kase-Ha Karate-Do Stockholm för att jag fick träna med er ett tag. Till min familj vill jag också rikta ett oerhört stort tack. Ni har alltid funnits där när jag har behövt er, och ni har alltid stöttat mig i mina vägval och hjälp mig mer än ni någonsin kan ana. Mamma och pappa, jag säger det kanske inte så ofta, men jag älskar er. Rikard, Sandra och Linda, jag öns- kar er all lycka och framgång i livet. Till min älskade Elisabeth, som funnits vid min sida i vått och torrt, på långt avstånd eller som sambo, i sju års tid, vill jag bara säga att jag hoppas på många fler år tillsammans. Tack för att du stöttat mig under min dokto- randtid, och speciellt under skrivandet av den här avhandlingen, trots att du själv haft din egen master-rapport att tänka på. Du är i mitt hjärta för all framtid.

64 References

"If I have seen further, it is by standing upon the shoulders of giants" – Isaac Newton

1. Cancerfonden (2017) Cancerfondsrapporten 2017. (Stockholm). 2. Cheung-Ong K, Giaever G, & Nislow C (2013) Chemistry and Biology 20(5):648-659. 3. Carreras-Puigvert J, Zitnik M, Jemth AS, et al. (2017) Nature communications 8(1):1541. 4. Zhang Y, Du Y, Le W, et al. (2011) Antioxidants & Redox Signaling 15(11):2867-2908. 5. Sakumi K, Furuichi M, Tsuzuki T, et al. (1993) Journal of Biological Chemistry 268(31):23524-23530. 6. Tennant DA, Duran RV, Boulahbel H, et al. (2009) Carcinogenesis 30(8):1269-1280. 7. DeBerardinis RJ, Lum JJ, Hatzivassiliou G, et al. (2008) Cell Metabolism 7(1):11-20. 8. Tibbetts AS & Appling DR (2010) Annual Review of Nutrition 30:57-81. 9. Nilsson R, Jain M, Madhusudhan N, et al. (2014) Nature communications 5:3128. 10. Rossetto O, Pirazzini M, & Montecucco C (2014) Nature Reviews. Microbiology 12(8):535-549. 11. Jankovic J (2017) Movement Disorders 32(8):1131-1138. 12. Montal M (2010) Annual Review of Biochemistry 79:591-617. 13. Hill KK & Smith TJ (2013) Current Topics in Microbiology and Immunology 364:1-20. 14. Amatsu S, Sugawara Y, Matsumura T, et al. (2013) Journal of Biological Chemistry 288(49):35617-35625. 15. Lee K, Gu S, Jin L, et al. (2013) PLoS Pathogens 9(10):e1003690. 16. Lee K, Zhong X, Gu S, et al. (2014) Science 344(6190):1405-1410. 17. Sugawara Y, Matsumura T, Takegahara Y, et al. (2010) Journal of Cell Biology 189(4):691-700. 18. Gu S, Rumpel S, Zhou J, et al. (2012) Science 335(6071):977-981. 19. Gu S & Jin R (2013) Current Topics in Microbiology and Immunology 364:21- 44. 20. Kopec KO, Alva V, & Lupas AN (2011) Biochemical Society Transactions 39(4):1033-1038. 21. Bateson P (2006) Biology & Philosophy 21(4):553-558. 22. Ferlay J, Soerjomataram I, Dikshit R, et al. (2015) International Journal of Cancer 136(5):E359-386.

65 23. Socialstyrelsen (2015) Cancerincidens i Sverige 2014 – Nya diagnosticerade cancerfall år 2014. (Stockholm). 24. Socialstyrelsen (2015) Dödsorsaker 2014 - Causes of Death 2014. (Stockholm). 25. Clark WH (1991) British Journal of Cancer 64(4):631-644. 26. Clark WH, Jr. (1991) Cancer and Metastasis Reviews 10(2):83-88. 27. Hanahan D & Weinberg RA (2000) Cell 100(1):57-70. 28. Hanahan D & Weinberg RA (2011) Cell 144(5):646-674. 29. Hansford S & Huntsman DG (2014) Journal of Pathology 234(2):142-145. 30. Baker SG (2015) Journal of the National Cancer Institute 107(2). 31. Sonnenschein C & Soto AM (2013) Journal of Biosciences 38(3):651-663. 32. Sonnenschein C & Soto AM (2016) Progress in Biophysics and Molecular Biology 122(1):70-76. 33. Sonnenschein C, Soto AM, Rangarajan A, et al. (2014) Journal of Biosciences 39(2):281-302. 34. Lewis EB (1978) Nature 276(5688):565-570. 35. Nusslein-Volhard C & Wieschaus E (1980) Nature 287(5785):795-801. 36. Baker SG (2012) Disruptive Science and Technology 1(2):100-107. 37. Baker SG (2013) Disruptive Science and Technology 1(3):143-148. 38. Soto AM & Sonnenschein C (2011) Bioessays 33(5):332-340. 39. Soto AM & Sonnenschein C (2013) Disruptive Science and Technology 1(3):154-156. 40. Greenman C, Stephens P, Smith R, et al. (2007) Nature 446(7132):153-158. 41. Haas D, Ablin AR, Miller C, et al. (1988) Cancer 62(4):818-825. 42. Horii R, Akiyama F, Kasumi F, et al. (2005) Breast Cancer 12(2):140-144. 43. Helleday T (2014) Annals of Oncology 25(7):1253-1255. 44. Kuraoka I, Robins P, Masutani C, et al. (2001) Journal of Biological Chemistry 276(52):49283-49288. 45. Lindahl T & Karlstrom O (1973) Biochemistry 12(25):5151-5154. 46. Lindahl T & Nyberg B (1972) Biochemistry 11(19):3610-3618. 47. Lindahl T & Nyberg B (1974) Biochemistry 13(16):3405-3410. 48. Rydberg B & Lindahl T (1982) EMBO Journal 1(2):211-216. 49. Lindahl T (1993) Nature 362(6422):709-715. 50. Lindahl T (2016) Angewandte Chemie. International Ed. In English 55(30):8528-8534. 51. Kao YT, Saxena C, Wang L, et al. (2007) Cell Biochemistry and Biophysics 48(1):32-44. 52. Sancar A (2003) Chemical Reviews 103(6):2203-2237. 53. Tan C, Liu Z, Li J, et al. (2015) Nature communications 6:7302. 54. Beckman KB & Ames BN (1998) Physiological Reviews 78(2):547-581. 55. Marnett LJ (2000) Carcinogenesis 21(3):361-370. 56. Wiseman H & Halliwell B (1996) Biochemical Journal 313 ( Pt 1):17-29. 57. Hix S, Morais Mda S, & Augusto O (1995) Free Radical Biology and Medicine 19(3):293-301. 58. Breimer LH & Lindahl T (1985) Mutation Research 150(1-2):85-89. 59. Dianov G & Lindahl T (1994) Current Biology 4(12):1069-1076. 60. Lindahl T (1976) Nature 259(5538):64-66.

66 61. Lindahl T (2000) Mutation Research 462(2-3):129-135. 62. Nilsen H, Lindahl T, & Verreault A (2002) EMBO Journal 21(21):5943-5952. 63. Au KG, Welsh K, & Modrich P (1992) Journal of Biological Chemistry 267(17):12142-12148. 64. Constantin N, Dzantiev L, Kadyrov FA, et al. (2005) Journal of Biological Chemistry 280(48):39752-39761. 65. Dzantiev L, Constantin N, Genschel J, et al. (2004) Molecular Cell 15(1):31- 41. 66. Iyer RR, Pluciennik A, Burdett V, et al. (2006) Chemical Reviews 106(2):302- 323. 67. Lahue RS, Au KG, & Modrich P (1989) Science 245(4914):160-164. 68. Lu AL, Clark S, & Modrich P (1983) Proceedings of the National Academy of Sciences of the United States of America 80(15):4639-4643. 69. Modrich P (1994) Science 266(5193):1959-1960. 70. Modrich P (2016) Angewandte Chemie. International Ed. In English 55(30):8490-8501. 71. Sancar A (1996) Annual Review of Biochemistry 65:43-81. 72. Sancar A (2016) Angewandte Chemie. International Ed. In English 55(30):8502-8527. 73. Sancar A, Lindsey-Boltz LA, Gaddameedhi S, et al. (2015) Biochemistry 54(2):110-123. 74. Sancar A, Lindsey-Boltz LA, Unsal-Kacmaz K, et al. (2004) Annual Review of Biochemistry 73:39-85. 75. Sancar A & Reardon JT (2004) Advances in Protein Chemistry 69:43-71. 76. Wakasugi M, Kawashima A, Morioka H, et al. (2002) Journal of Biological Chemistry 277(3):1637-1640. 77. Burma S, Chen BP, & Chen DJ (2006) DNA Repair (Amst) 5(9-10):1042-1048. 78. Luo M, He H, Kelley MR, et al. (2010) Antioxidants & Redox Signaling 12(11):1247-1269. 79. Sung P & Klein H (2006) Nature Reviews: Molecular Cell Biology 7(10):739- 750. 80. Wood RD, Mitchell M, & Lindahl T (2005) Mutation Research 577(1-2):275- 283. 81. Wood RD, Mitchell M, Sgouros J, et al. (2001) Science 291(5507):1284-1289. 82. Puigvert JC, Sanjiv K, & Helleday T (2016) The FEBS journal 283(2):232- 245. 83. Puck TT (1958) Proceedings of the National Academy of Sciences of the United States of America 44(8):772-780. 84. Puck TT & Marcus PI (1956) Journal of Experimental Medicine 103(5):653- 666. 85. Puck TT, Morkovin D, Marcus PI, et al. (1957) Journal of Experimental Medicine 106(4):485-500. 86. Fischhaber PL, Gall AS, Duncan JA, et al. (1999) Cancer Research 59(17):4363-4368. 87. Visentin M, Zhao R, & Goldman ID (2012) Hematology/Oncology Clinics of North America 26(3):629-648, ix.

67 88. Froelich-Ammon SJ & Osheroff N (1995) Journal of Biological Chemistry 270(37):21429-21432. 89. Sawicka M, Kalinowska M, Skierski J, et al. (2004) Journal of Pharmacy and Pharmacology 56(9):1067-1081. 90. Gottesman MM (2002) Annual Review of Medicine 53:615-627. 91. Wang D & Lippard SJ (2005) Nature Reviews. Drug discovery 4(4):307-320. 92. Manohar S & Leung N (2017) Journal of Nephrology. 93. Pabla N & Dong Z (2008) Kidney International 73(9):994-1007. 94. Yao X, Panichpisal K, Kurtzman N, et al. (2007) American Journal of the Medical Sciences 334(2):115-124. 95. Minotti G, Menna P, Salvatorelli E, et al. (2004) Pharmacological Reviews 56(2):185-229. 96. Olson RD & Mushlin PS (1990) FASEB Journal 4(13):3076-3086. 97. Swain SM, Whaley FS, & Ewer MS (2003) Cancer 97(11):2869-2879. 98. Tangpong J, Cole MP, Sultana R, et al. (2006) Neurobiology of Disease 23(1):127-139. 99. Kalender Y, Yel M, & Kalender S (2005) Toxicology 209(1):39-45. 100. Bailon-Moscoso N, Romero-Benavides JC, & Ostrosky-Wegman P (2014) Tumour Biology 35(5):3981-3995. 101. McLornan DP, List A, & Mufti GJ (2014) New England Journal of Medicine 371(18):1725-1735. 102. Shaheen M, Allen C, Nickoloff JA, et al. (2011) Blood 117(23):6074-6082. 103. Helleday T, Petermann E, Lundin C, et al. (2008) Nature Reviews: Cancer 8(3):193-204. 104. Bryant HE, Schultz N, Thomas HD, et al. (2005) Nature 434(7035):913-917. 105. Farmer H, McCabe N, Lord CJ, et al. (2005) Nature 434(7035):917-921. 106. Cerrato A, Morra F, & Celetti A (2016) Journal of Experimental and Clinical Cancer Research 35(1):179. 107. Bedard K & Krause KH (2007) Physiological Reviews 87(1):245-313. 108. Lambeth JD (2004) Nature Reviews: Immunology 4(3):181-189. 109. Cadenas E, Boveris A, Ragan CI, et al. (1977) Archives of Biochemistry and Biophysics 180(2):248-257. 110. Lenaz G (2001) IUBMB Life 52(3-5):159-164. 111. Boveris A, Oshino N, & Chance B (1972) Biochemical Journal 128(3):617- 630. 112. Cadenas E & Davies KJ (2000) Free Radical Biology and Medicine 29(3- 4):222-230. 113. Malhotra JD & Kaufman RJ (2007) Antioxidants & Redox Signaling 9(12):2277-2293. 114. Tu BP & Weissman JS (2004) Journal of Cell Biology 164(3):341-346. 115. Hwang C, Sinskey AJ, & Lodish HF (1992) Science 257(5076):1496-1502. 116. Amaya Y, Yamazaki K, Sato M, et al. (1990) Journal of Biological Chemistry 265(24):14170-14175. 117. Harrison R (2004) Drug Metabolism Reviews 36(2):363-375. 118. Hille R & Nishino T (1995) FASEB Journal 9(11):995-1003. 119. Morgan WF, Day JP, Kaplan MI, et al. (1996) Radiation Research 146(3):247- 258.

68 120. Valko M, Morris H, & Cronin MT (2005) Current Medicinal Chemistry 12(10):1161-1208. 121. Zangar RC, Davydov DR, & Verma S (2004) Toxicology and Applied Pharmacology 199(3):316-331. 122. Hielscher A & Gerecht S (2015) Free Radical Biology and Medicine 79:281- 291. 123. Sabharwal SS & Schumacker PT (2014) Nature Reviews: Cancer 14(11):709- 721. 124. Szatrowski TP & Nathan CF (1991) Cancer Research 51(3):794-798. 125. Warburg O (1956) Science 123(3191):309-314. 126. Fantin VR, St-Pierre J, & Leder P (2006) Cancer Cell 9(6):425-434. 127. Maxwell PH, Wiesener MS, Chang GW, et al. (1999) Nature 399(6733):271- 275. 128. Semenza GL (2010) Oncogene 29(5):625-634. 129. Zhong H, De Marzo AM, Laughner E, et al. (1999) Cancer Research 59(22):5830-5835. 130. Chandel NS, McClintock DS, Feliciano CE, et al. (2000) Journal of Biological Chemistry 275(33):25130-25138. 131. Chandel NS, Maltepe E, Goldwasser E, et al. (1998) Proceedings of the National Academy of Sciences of the United States of America 95(20):11715- 11720. 132. Epstein AC, Gleadle JM, McNeill LA, et al. (2001) Cell 107(1):43-54. 133. Huang LE, Gu J, Schau M, et al. (1998) Proceedings of the National Academy of Sciences of the United States of America 95(14):7987-7992. 134. Ivan M, Kondo K, Yang H, et al. (2001) Science 292(5516):464-468. 135. Jaakkola P, Mole DR, Tian YM, et al. (2001) Science 292(5516):468-472. 136. Topal MD & Baker MS (1982) Proceedings of the National Academy of Sciences of the United States of America 79(7):2211-2215. 137. Kamiya H & Kasai H (1995) Journal of Biological Chemistry 270(33):19446- 19450. 138. Kasai H & Nishimura S (1984) Nucleic Acids Research 12(4):2137-2145. 139. Mo JY, Maki H, & Sekiguchi M (1992) Proceedings of the National Academy of Sciences of the United States of America 89(22):11021-11025. 140. De Luca G, Russo MT, Degan P, et al. (2008) PLoS Genetics 4(11):e1000266. 141. Fujikawa K, Kamiya H, Yakushiji H, et al. (1999) Journal of Biological Chemistry 274(26):18201-18205. 142. German P, Szaniszlo P, Hajas G, et al. (2013) DNA Repair (Amst) 12(10):856- 863. 143. Ichikawa J, Tsuchimoto D, Oka S, et al. (2008) DNA Repair (Amst) 7(3):418- 430. 144. Oka S, Ohno M, Tsuchimoto D, et al. (2008) EMBO Journal 27(2):421-432. 145. Russo MT, Blasi MF, Chiera F, et al. (2004) Molecular and Cellular Biology 24(1):465-474. 146. Ushijima Y, Tominaga Y, Miura T, et al. (2005) Nucleic Acids Research 33(2):672-682. 147. Yoshimura D, Sakumi K, Ohno M, et al. (2003) Journal of Biological Chemistry 278(39):37965-37973.

69 148. Shibutani S, Takeshita M, & Grollman AP (1991) Nature 349(6308):431-434. 149. Inoue M, Kamiya H, Fujikawa K, et al. (1998) Journal of Biological Chemistry 273(18):11069-11074. 150. Kamiya H & Kasai H (2000) Nucleic Acids Research 28(7):1640-1646. 151. Murata-Kamiya N, Kamiya H, Muraoka M, et al. (1997) Journal of Radiation Research 38(2):121-131. 152. Nakabeppu Y, Tsuchimoto D, Furuichi M, et al. (2004) Free Radical Research 38(5):423-429. 153. Tsuzuki T, Egashira A, & Kura S (2001) Mutation Research 477(1-2):71-78. 154. de Souza-Pinto NC, Eide L, Hogue BA, et al. (2001) Cancer Research 61(14):5378-5381. 155. Sakumi K, Tominaga Y, Furuichi M, et al. (2003) Cancer Research 63(5):902- 905. 156. Arai K, Morishita K, Shinmura K, et al. (1997) Oncogene 14(23):2857-2861. 157. Boiteux S & Radicella JP (2000) Archives of Biochemistry and Biophysics 377(1):1-8. 158. Bruner SD, Norman DP, & Verdine GL (2000) Nature 403(6772):859-866. 159. Radicella JP, Dherin C, Desmaze C, et al. (1997) Proceedings of the National Academy of Sciences of the United States of America 94(15):8010-8015. 160. Rosenquist TA, Zharkov DO, & Grollman AP (1997) Proceedings of the National Academy of Sciences of the United States of America 94(14):7429- 7434. 161. Chevillard S, Radicella JP, Levalois C, et al. (1998) Oncogene 16(23):3083- 3086. 162. Lu R, Nash HM, & Verdine GL (1997) Current Biology 7(6):397-407. 163. Croteau DL, ap Rhys CM, Hudson EK, et al. (1997) Journal of Biological Chemistry 272(43):27338-27344. 164. Ohtsubo T, Nishioka K, Imaiso Y, et al. (2000) Nucleic Acids Research 28(6):1355-1364. 165. Poderoso JJ (2016) Cancer and Mitochondria. Biochemistry of Oxidative Stress: Physiopathology and Clinical Aspects, eds Gelpi RJ, Boveris A, & Poderoso JJ (Springer International Publishing, Cham), pp 409-423. 166. Rai P, Onder TT, Young JJ, et al. (2009) Proceedings of the National Academy of Sciences of the United States of America 106(1):169-174. 167. Rai P, Young JJ, Burton DG, et al. (2011) Oncogene 30(12):1489-1496. 168. Hudson EK, Hogue BA, Souza-Pinto NC, et al. (1998) Free Radical Research 29(6):573-579. 169. Kang D, Nishida J, Iyama A, et al. (1995) Journal of Biological Chemistry 270(24):14659-14665. 170. Nishioka K, Ohtsubo T, Oda H, et al. (1999) Molecular Biology of the Cell 10(5):1637-1652. 171. Loechler EL, Green CL, & Essigmann JM (1984) Proceedings of the National Academy of Sciences of the United States of America 81(20):6271-6275. 172. Newlands ES, Stevens MF, Wedge SR, et al. (1997) Cancer Treatment Reviews 23(1):35-61. 173. Kaina B, Christmann M, Naumann S, et al. (2007) DNA Repair (Amst) 6(8):1079-1099.

70 174. Goll MG & Bestor TH (2005) Annual Review of Biochemistry 74:481-514. 175. Struck AW, Thompson ML, Wong LS, et al. (2012) Chembiochem: A European Journal of Chemical Biology 13(18):2642-2655. 176. Beranek DT (1990) Mutation Research 231(1):11-30. 177. Topal MD, Hutchison CA, 3rd, & Baker MS (1982) Nature 298(5877):863- 865. 178. Mojas N, Lopes M, & Jiricny J (2007) Genes and Development 21(24):3342- 3355. 179. Hudson BH & Zaher HS (2015) RNA 21(9):1648-1659. 180. Traut TW (1994) Molecular and Cellular Biochemistry 140(1):1-22. 181. Aas PA, Otterlei M, Falnes PO, et al. (2003) Nature 421(6925):859-863. 182. Duncan T, Trewick SC, Koivisto P, et al. (2002) Proceedings of the National Academy of Sciences of the United States of America 99(26):16660-16665. 183. O'Brien PJ & Ellenberger T (2004) Journal of Biological Chemistry 279(11):9750-9757. 184. Rubinson EH, Christov PP, & Eichman BF (2013) Biochemistry 52(42):7363- 7365. 185. Glassner BJ, Weeda G, Allan JM, et al. (1999) Mutagenesis 14(3):339-347. 186. Tsuzuki T, Sakumi K, Shiraishi A, et al. (1996) Carcinogenesis 17(6):1215- 1220. 187. Sakumi K, Shiraishi A, Shimizu S, et al. (1997) Cancer Research 57(12):2415- 2418. 188. Day RS, 3rd, Ziolkowski CH, Scudiero DA, et al. (1980) Nature 288(5792):724-727. 189. Hegi ME, Diserens AC, Gorlia T, et al. (2005) New England Journal of Medicine 352(10):997-1003. 190. Maki H & Sekiguchi M (1992) Nature 355(6357):273-275. 191. Treffers HP, Spinelli V, & Belser NO (1954) Proceedings of the National Academy of Sciences of the United States of America 40(11):1064-1071. 192. Ito R, Hayakawa H, Sekiguchi M, et al. (2005) Biochemistry 44(17):6670- 6674. 193. Taddei F, Hayakawa H, Bouton M, et al. (1997) Science 278(5335):128-130. 194. Abeygunawardana C, Weber DJ, Gittis AG, et al. (1995) Biochemistry 34(46):14997-15005. 195. Nakamura T, Meshitsuka S, Kitagawa S, et al. (2010) Journal of Biological Chemistry 285(1):444-452. 196. Mildvan AS, Xia Z, Azurmendi HF, et al. (2005) Archives of Biochemistry and Biophysics 433(1):129-143. 197. Fujikawa K, Kamiya H, Yakushiji H, et al. (2001) Nucleic Acids Research 29(2):449-454. 198. Kamiya H, Yakushiji H, Dugue L, et al. (2004) Journal of Molecular Biology 336(4):843-850. 199. Bialkowski K & Kasprzak KS (2003) Free Radical Biology and Medicine 35(6):595-602. 200. Yin Y, Sasaki S, & Taniguchi Y (2016) Chembiochem: A European Journal of Chemical Biology 17(7):566-569.

71 201. Sakai Y, Furuichi M, Takahashi M, et al. (2002) Journal of Biological Chemistry 277(10):8579-8587. 202. Hayakawa H, Taketomi A, Sakumi K, et al. (1995) Biochemistry 34(1):89-95. 203. Hayakawa H, Hofer A, Thelander L, et al. (1999) Biochemistry 38(12):3610- 3614. 204. Mishima M, Sakai Y, Itoh N, et al. (2004) Journal of Biological Chemistry 279(32):33806-33815. 205. Svensson LM, Jemth AS, Desroses M, et al. (2011) FEBS Letters 585(16):2617-2621. 206. Fujiwara S, Sawada K, & Amisaki T (2014) Journal of Molecular Graphics and Modelling 51:158-167. 207. Nissink JW, Bista M, Breed J, et al. (2016) PLoS One 11(3):e0151154. 208. Waz S, Nakamura T, Hirata K, et al. (2017) Journal of Biological Chemistry 292(7):2785-2794. 209. Koga Y, Inazato M, Nakamura T, et al. (2013) Acta Crystallographica. Section F: Structural Biology and Crystallization Communications 69(Pt 1):45-48. 210. Schrödinger, LLC (The PyMOL Molecular Graphics System, Version MacPyMOL). 211. Nakabeppu Y (2001) Mutation Research 477(1-2):59-70. 212. Oda H, Taketomi A, Maruyama R, et al. (1999) Nucleic Acids Research 27(22):4335-4343. 213. Yakushiji H, Maraboeuf F, Takahashi M, et al. (1997) Mutation Research 384(3):181-194. 214. Kimura Y, Oda S, Egashira A, et al. (2004) Journal of Medical Genetics 41(5):e57. 215. McLennan AG (2006) Cellular and Molecular Life Sciences 63(2):123-143. 216. McLennan AG (2013) Cellular and Molecular Life Sciences 70(3):373-385. 217. Bessman MJ, Frick DN, & O'Handley SF (1996) Journal of Biological Chemistry 271(41):25059-25062. 218. Gabelli SB, Bianchet MA, Bessman MJ, et al. (2001) Nature Structural Biology 8(5):467-472. 219. McLennan AG (1999) International Journal of Molecular Medicine 4(1):79- 89. 220. Cai JP, Ishibashi T, Takagi Y, et al. (2003) Biochemical and Biophysical Research Communications 305(4):1073-1077. 221. Carter M, Jemth AS, Hagenkort A, et al. (2015) Nature communications 6:7871. 222. Moriyama T, Nishii R, Perez-Andreu V, et al. (2016) Nature Genetics 48(4):367-373. 223. Valerie NC, Hagenkort A, Page BD, et al. (2016) Cancer Research 76(18):5501-5511. 224. Takagi Y, Setoyama D, Ito R, et al. (2012) Journal of Biological Chemistry 287(25):21541-21549. 225. Ishibashi T, Hayakawa H, & Sekiguchi M (2003) EMBO Reports 4(5):479- 483. 226. Ishibashi T, Hayakawa H, Ito R, et al. (2005) Nucleic Acids Research 33(12):3779-3784.

72 227. Kamiya H, Hori M, Arimori T, et al. (2009) DNA Repair (Amst) 8(10):1250- 1254. 228. Nakabeppu Y, Kajitani K, Sakamoto K, et al. (2006) DNA Repair (Amst) 5(7):761-772. 229. Ventura I, Russo MT, De Luca G, et al. (2010) Mutation Research 703(1):59- 65. 230. Hori M, Satou K, Harashima H, et al. (2010) Free Radical Biology and Medicine 48(9):1197-1201. 231. Kamiya H (2010) Mutation Research 703(1):32-36. 232. Nakabeppu Y, Oka S, Sheng Z, et al. (2010) Mutation Research 703(1):51-58. 233. Rai P (2010) Mutation Research 703(1):71-81. 234. Ferraro P, Franzolin E, Pontarin G, et al. (2010) Nucleic Acids Research 38(6):e85. 235. Egashira A, Yamauchi K, Yoshiyama K, et al. (2002) DNA Repair (Amst) 1(11):881-893. 236. Tsuzuki T, Egashira A, Igarashi H, et al. (2001) Proceedings of the National Academy of Sciences of the United States of America 98(20):11456-11461. 237. Tajiri T, Maki H, & Sekiguchi M (1995) Mutation Research 336(3):257-267. 238. De Luca G, Ventura I, Sanghez V, et al. (2013) Aging Cell 12(4):695-705. 239. Irani K, Xia Y, Zweier JL, et al. (1997) Science 275(5306):1649-1652. 240. Toyokuni S, Okamoto K, Yodoi J, et al. (1995) FEBS Letters 358(1):1-3. 241. Klaunig JE, Wang Z, Pu X, et al. (2011) Toxicology and Applied Pharmacology 254(2):86-99. 242. Sosa V, Moline T, Somoza R, et al. (2013) Ageing Research Reviews 12(1):376-390. 243. Karihtala P & Soini Y (2007) APMIS 115(2):81-103. 244. Lincoln DT, Ali Emadi EM, Tonissen KF, et al. (2003) Anticancer Research 23(3B):2425-2433. 245. Gorrini C, Harris IS, & Mak TW (2013) Nature Reviews. Drug discovery 12(12):931-947. 246. Okamoto K, Toyokuni S, Kim WJ, et al. (1996) International Journal of Cancer 65(4):437-441. 247. Kennedy CH, Pass HI, & Mitchell JB (2003) Free Radical Biology and Medicine 34(11):1447-1457. 248. Iida T, Furuta A, Kawashima M, et al. (2001) Neuro-Oncology 3(2):73-81. 249. Coskun E, Jaruga P, Jemth AS, et al. (2015) DNA Repair (Amst) 33:101-110. 250. Speina E, Arczewska KD, Gackowski D, et al. (2005) Journal of the National Cancer Institute 97(5):384-395. 251. Jaiswal M, LaRusso NF, Nishioka N, et al. (2001) Cancer Research 61(17):6388-6393. 252. Fouquerel E, Lormand J, Bose A, et al. (2016) Nature Structural & Molecular Biology 23(12):1092-1100. 253. Kim NW, Piatyszek MA, Prowse KR, et al. (1994) Science 266(5193):2011- 2015. 254. Rai P (2012) Small GTPases 3(2):120-125. 255. Giribaldi MG, Munoz A, Halvorsen K, et al. (2015) Oncotarget 6(13):11519- 11529.

73 256. Patel A, Burton DG, Halvorsen K, et al. (2015) Oncogene 34(20):2586-2596. 257. Ellermann M, Eheim A, Rahm F, et al. (2017) ACS Chemical Biology 12(8):1986-1992. 258. Kawamura T, Kawatani M, Muroi M, et al. (2016) Scientific Reports 6:26521. 259. Kettle JG, Alwan H, Bista M, et al. (2016) Journal of Medicinal Chemistry 59(6):2346-2361. 260. Petrocchi A, Leo E, Reyna NJ, et al. (2016) Bioorganic and Medicinal Chemistry Letters 26(6):1503-1507. 261. Huber KV, Salah E, Radic B, et al. (2014) Nature 508(7495):222-227. 262. Warpman Berglund U, Sanjiv K, Gad H, et al. (2016) Annals of Oncology 27(12):2275-2283. 263. Samaranayake GJ, Huynh M, & Rai P (2017) Cancers 9(5). 264. Venter JC, Adams MD, Myers EW, et al. (2001) Science 291(5507):1304- 1351. 265. Baggott JE & Tamura T (2015) Advances in Nutrition 6(5):564-571. 266. Ducker GS & Rabinowitz JD (2017) Cell Metabolism 25(1):27-42. 267. Pedley AM & Benkovic SJ (2017) Trends in Biochemical Sciences 42(2):141- 154. 268. Yang M & Vousden KH (2016) Nature Reviews: Cancer 16(10):650-662. 269. Ben-Sahra I, Hoxhaj G, Ricoult SJ, et al. (2016) Science 351(6274):728-733. 270. Vander Heiden MG (2011) Nature Reviews. Drug discovery 10(9):671-684. 271. Fu TF, Rife JP, & Schirch V (2001) Archives of Biochemistry and Biophysics 393(1):42-50. 272. Nijhout HF, Reed MC, Lam SL, et al. (2006) Theoretical Biology & Medical Modelling 3:40. 273. Woeller CF, Anderson DD, Szebenyi DM, et al. (2007) Journal of Biological Chemistry 282(24):17623-17631. 274. Gustafsson Sheppard N, Jarl L, Mahadessian D, et al. (2015) Scientific Reports 5:15029. 275. Tennant DA, Duran RV, & Gottlieb E (2010) Nature Reviews: Cancer 10(4):267-277. 276. Locasale JW (2013) Nature Reviews: Cancer 13(8):572-583. 277. Vazquez A, Tedeschi PM, & Bertino JR (2013) Cancer Research 73(2):478- 482. 278. Tedeschi PM, Markert EK, Gounder M, et al. (2013) Cell Death & Disease 4:e877. 279. Appling DR (1991) FASEB Journal 5(12):2645-2651. 280. Christensen KE & MacKenzie RE (2006) Bioessays 28(6):595-605. 281. Anderson DD, Quintero CM, & Stover PJ (2011) Proceedings of the National Academy of Sciences of the United States of America 108(37):15163-15168. 282. Christensen KE & Mackenzie RE (2008) Vitamins and Hormones 79:393-410. 283. Schirch V & Szebenyi DM (2005) Current Opinion in Chemical Biology 9(5):482-487. 284. Lamers Y, Williamson J, Gilbert LR, et al. (2007) Journal of Nutrition 137(12):2647-2652. 285. Lamers Y, Williamson J, Ralat M, et al. (2009) Journal of Nutrition 139(3):452-460.

74 286. Lamers Y, Williamson J, Theriaque DW, et al. (2009) Journal of Nutrition 139(4):666-671. 287. Burgos-Barragan G, Wit N, Meiser J, et al. (2017) Nature 548(7669):549-554. 288. Pelletier JN & MacKenzie RE (1995) Biochemistry 34(39):12673-12680. 289. Gregory JF, 3rd, Cuskelly GJ, Shane B, et al. (2000) American Journal of Clinical Nutrition 72(6):1535-1541. 290. Herbig K, Chiang EP, Lee LR, et al. (2002) Journal of Biological Chemistry 277(41):38381-38389. 291. Pasternack LB, Laude DA, Jr., & Appling DR (1992) Biochemistry 31(37):8713-8719. 292. Allaire M, Li Y, MacKenzie RE, et al. (1998) Structure 6(2):173-182. 293. Thigpen AE, West MG, & Appling DR (1990) Journal of Biological Chemistry 265(14):7907-7913. 294. Schmidt A, Wu H, MacKenzie RE, et al. (2000) Biochemistry 39(21):6325- 6335. 295. Pawelek PD, Allaire M, Cygler M, et al. (2000) Biochimica et Biophysica Acta 1479(1-2):59-68. 296. Smith DDS & MacKenzie RE (1983) Canadian Journal of Biochemistry and Cell Biology 61(11):1166-1171. 297. Pawelek PD & MacKenzie RE (1998) Biochemistry 37(4):1109-1115. 298. Sarau HM, Foley JJ, Moonsammy G, et al. (1977) Journal of Biological Chemistry 252(23):8572-8581. 299. Veech RL (2006) Biochemistry and Molecular Biology Education 34(3):168- 179. 300. Veech RL, Eggleston LV, & Krebs HA (1969) Biochemical Journal 115(4):609-619. 301. Peri KG & MacKenzie RE (1991) FEBS Letters 294(1-2):113-115. 302. MacFarlane AJ, Perry CA, Girnary HH, et al. (2009) Journal of Biological Chemistry 284(3):1533-1539. 303. Bolusani S, Young BA, Cole NA, et al. (2011) Journal of Biological Chemistry 286(7):5166-5174. 304. Christensen KE, Patel H, Kuzmanov U, et al. (2005) Journal of Biological Chemistry 280(9):7597-7602. 305. Prasannan P, Pike S, Peng K, et al. (2003) Journal of Biological Chemistry 278(44):43178-43187. 306. Scrimgeour KG & Huennekens FM (1960) Biochemical and Biophysical Research Communications 2(3):230-233. 307. Christensen KE, Mirza IA, Berghuis AM, et al. (2005) Journal of Biological Chemistry 280(40):34316-34323. 308. Mejia NR & MacKenzie RE (1988) Biochemical and Biophysical Research Communications 155(1):1-6. 309. Mejia NR, Rios-Orlandi EM, & MacKenzie RE (1986) Journal of Biological Chemistry 261(20):9509-9513. 310. Rios-Orlandi EM & MacKenzie RE (1988) Journal of Biological Chemistry 263(10):4662-4667. 311. Mejia NR & MacKenzie RE (1985) Journal of Biological Chemistry 260(27):14616-14620.

75 312. Yang XM & MacKenzie RE (1993) Biochemistry 32(41):11118-11123. 313. Patel H, Christensen KE, Mejia N, et al. (2002) Archives of Biochemistry and Biophysics 403(1):145-148. 314. Vina J, Saez GT, Gambini J, et al. (2016) Archives of Biochemistry and Biophysics 595:176-180. 315. Williamson DH, Lund P, & Krebs HA (1967) Biochemical Journal 103(2):514-527. 316. Smith GK, Banks SD, Monaco TJ, et al. (1990) Archives of Biochemistry and Biophysics 283(2):367-371. 317. Di Pietro E, Sirois J, Tremblay ML, et al. (2002) Molecular and Cellular Biology 22(12):4158-4166. 318. Peri KG & MacKenzie RE (1993) Biochimica et Biophysica Acta 1171(3):281- 287. 319. Patel H, Pietro ED, & MacKenzie RE (2003) Journal of Biological Chemistry 278(21):19436-19441. 320. Patel H, Di Pietro E, Mejia N, et al. (2005) Archives of Biochemistry and Biophysics 442(1):133-139. 321. Shin M, Bryant JD, Momb J, et al. (2014) Journal of Biological Chemistry 289(22):15507-15517. 322. Pike ST, Rajendra R, Artzt K, et al. (2010) Journal of Biological Chemistry 285(7):4612-4620. 323. Momb J, Lewandowski JP, Bryant JD, et al. (2013) Proceedings of the National Academy of Sciences of the United States of America 110(2):549-554. 324. Sugiura T, Nagano Y, Inoue T, et al. (2004) Biochemical and Biophysical Research Communications 315(1):204-211. 325. Campanella M, Parker N, Tan CH, et al. (2009) Trends in Biochemical Sciences 34(7):343-350. 326. Veech RL, Raijman L, & Krebs HA (1970) Biochemical Journal 117(3):499- 503. 327. Tucker EJ, Hershman SG, Kohrer C, et al. (2011) Cell Metabolism 14(3):428- 434. 328. Li Y, Holmes WB, Appling DR, et al. (2000) Journal of Bacteriology 182(10):2886-2892. 329. Brosnan ME, MacMillan L, Stevens JR, et al. (2015) Biochemical Journal 472(2):135-146. 330. Farber S, Diamond LK, Mercer RD, et al. (1948) New England Journal of Medicine 238(23):787-793. 331. Farber S (1949) Blood 4(2):160-167. 332. Chabner BA & Roberts TG, Jr. (2005) Nature Reviews: Cancer 5(1):65-72. 333. Hu CM, Yeh MT, Tsao N, et al. (2012) Cancer Cell 22(1):36-50. 334. Stover PJ & Weiss RS (2012) Cancer Cell 22(1):3-4. 335. Tonkinson JL, Habeck LL, Toth JE, et al. (1998) Journal of Pharmacology and Experimental Therapeutics 287(1):315-321. 336. Shin M, Momb J, & Appling DR (2017) Cancer & Metabolism 5(1):11. 337. Di Pietro E, Wang XL, & MacKenzie RE (2004) Biochimica et Biophysica Acta 1674(1):78-84.

76 338. Tedeschi PM, Vazquez A, Kerrigan JE, et al. (2015) Molecular Cancer Research 13(10):1361-1366. 339. Lehtinen L, Ketola K, Makela R, et al. (2013) Oncotarget 4(1):48-63. 340. Selcuklu SD, Donoghue MT, Rehmet K, et al. (2012) Journal of Biological Chemistry 287(35):29516-29528. 341. Xu X, Qiao M, Zhang Y, et al. (2010) Proteomics 10(7):1374-1390. 342. Liu F, Liu Y, He C, et al. (2014) Tumour Biology 35(9):8685-8690. 343. Andrew AS, Gui J, Sanderson AC, et al. (2009) Human Genetics 125(5- 6):527-539. 344. Sertel S, Eichhorn T, Sieber S, et al. (2010) Chemico-Biological Interactions 185(1):42-52. 345. Bester AC, Roniger M, Oren YS, et al. (2011) Cell 145(3):435-446. 346. Shen BW, Dyer DH, Huang JY, et al. (1999) Protein Science 8(6):1342-1349. 347. Monzingo AF, Breksa A, Ernst S, et al. (2000) Protein Science 9(7):1374- 1381. 348. Eadsforth TC, Cameron S, & Hunter WN (2012) Molecular and Biochemical Parasitology 181(2):178-185. 349. Eadsforth TC, Gardiner M, Maluf FV, et al. (2012) PLoS One 7(4):e35973. 350. Eadsforth TC, Maluf FV, & Hunter WN (2012) The FEBS journal 279(23):4350-4360. 351. Eadsforth TC, Pinto A, Luciani R, et al. (2015) Journal of Medicinal Chemistry 58(20):7938-7948. 352. Lindstrom M & Korkeala H (2006) Clinical Microbiology Reviews 19(2):298- 314. 353. Montecucco C & Molgo J (2005) Current Opinion in Pharmacology 5(3):274- 279. 354. Shukla HD & Sharma SK (2005) Critical Reviews in Microbiology 31(1):11- 18. 355. Sobel J (2005) Clinical Infectious Diseases 41(8):1167-1173. 356. Tanzi MG & Gabay MP (2002) Pharmacotherapy 22(11):1479-1483. 357. Dahlsten E, Lindstrom M, & Korkeala H (2015) Research in Microbiology 166(4):344-352. 358. Zhang S, Masuyer G, Zhang J, et al. (2017) Nature communications 8:14130. 359. Gill DM (1982) Microbiological Reviews 46(1):86-94. 360. Dembek ZF, Smith LA, & Rusnak JM (2007) Disaster Medicine and Public Health Preparedness 1(2):122-134. 361. Rosow LK & Strober JB (2015) Pediatric Neurology 52(5):487-492. 362. Lacy DB, Tepp W, Cohen AC, et al. (1998) Nature Structural Biology 5(10):898-902. 363. Rossetto O & Montecucco C (2008) Handbook of Experimental Pharmacology (184):129-170. 364. Rummel A (2017) Current Topics in Microbiology and Immunology 406:1-37. 365. Stenmark P, Dupuy J, Imamura A, et al. (2008) PLoS Pathogens 4(8):e1000129. 366. Hamark C, Berntsson RP, Masuyer G, et al. (2017) Journal of the American Chemical Society 139(1):218-230.

77 367. Dong M, Liu H, Tepp WH, et al. (2008) Molecular Biology of the Cell 19(12):5226-5237. 368. Dong M, Yeh F, Tepp WH, et al. (2006) Science 312(5773):592-596. 369. Fu Z, Chen C, Barbieri JT, et al. (2009) Biochemistry 48(24):5631-5641. 370. Mahrhold S, Rummel A, Bigalke H, et al. (2006) FEBS Letters 580(8):2011- 2014. 371. Peng L, Tepp WH, Johnson EA, et al. (2011) PLoS Pathogens 7(3):e1002008. 372. Rummel A, Hafner K, Mahrhold S, et al. (2009) Journal of Neurochemistry 110(6):1942-1954. 373. Dong M, Richards DA, Goodnough MC, et al. (2003) Journal of Cell Biology 162(7):1293-1303. 374. Nishiki T, Kamata Y, Nemoto Y, et al. (1994) Journal of Biological Chemistry 269(14):10498-10503. 375. Peng L, Berntsson RP, Tepp WH, et al. (2012) Journal of Cell Science 125(Pt 13):3233-3242. 376. Rummel A, Eichner T, Weil T, et al. (2007) Proceedings of the National Academy of Sciences of the United States of America 104(1):359-364. 377. Rummel A, Karnath T, Henke T, et al. (2004) Journal of Biological Chemistry 279(29):30865-30870. 378. Zhang S, Berntsson RP, Tepp WH, et al. (2017) Nature communications 8(1):1637. 379. Chai Q, Arndt JW, Dong M, et al. (2006) Nature 444(7122):1096-1100. 380. Jin R, Rummel A, Binz T, et al. (2006) Nature 444(7122):1092-1095. 381. Benoit RM, Frey D, Hilbert M, et al. (2014) Nature 505(7481):108-111. 382. Benoit RM, Scharer MA, Wieser MM, et al. (2017) Scientific Reports 7:43588. 383. Berntsson RP, Peng L, Svensson LM, et al. (2013) Structure 21(9):1602-1611. 384. Yao G, Zhang S, Mahrhold S, et al. (2016) Nature Structural & Molecular Biology. 385. Brunger AT, Breidenbach MA, Jin R, et al. (2007) PLoS Pathogens 3(9):1191- 1194. 386. Fischer A, Sambashivan S, Brunger AT, et al. (2012) Journal of Biological Chemistry 287(3):1657-1661. 387. Hill KK, Xie G, Foley BT, et al. (2009) BMC Biology 7:66. 388. Smith TJ, Hill KK, Foley BT, et al. (2007) PLoS One 2(12):e1271. 389. Hill KK, Smith TJ, Helma CH, et al. (2007) Journal of Bacteriology 189(3):818-832. 390. East AK & Collins MD (1994) Current Microbiology 29(2):69-77. 391. Eswaramoorthy S, Sun J, Li H, et al. (2015) Scientific Reports 5:17795. 392. Kitamura M (1970) Biochemical and Biophysical Research Communications 40(4):925-&. 393. Kitamura M, Sakaguchi S, & Sakaguchi G (1968) Biochimica et Biophysica Acta 168(2):207-217. 394. Connan C, Deneve C, Mazuet C, et al. (2013) Toxicon 75:90-100. 395. Ito H, Sagane Y, Miyata K, et al. (2011) FEMS Immunology and Medical Microbiology 61(3):323-331. 396. Sugawara Y & Fujinaga Y (2011) Cell Adhesion & Migration 5(1):34-36.

78 397. Li B, Qian X, Sarkar HK, et al. (1998) Biochimica et Biophysica Acta 1395(1):21-27. 398. Chen Y, Korkeala H, Linden J, et al. (2008) Applied and Environmental Microbiology 74(19):6132-6137. 399. Dineen SS, Bradshaw M, Karasek CE, et al. (2004) FEMS Microbiology Letters 235(1):9-16. 400. Jacobson MJ, Lin G, Raphael B, et al. (2008) Applied and Environmental Microbiology 74(9):2778-2786. 401. Lin G, Tepp WH, Pier CL, et al. (2010) Applied and Environmental Microbiology 76(1):40-47. 402. Kalb SR, Baudys J, Smith TJ, et al. (2017) Toxins (Basel) 9(6). 403. Lin G, Tepp WH, Bradshaw M, et al. (2015) Applied and Environmental Microbiology 81(2):481-491. 404. Lam KH, Qi R, Liu S, et al. (2017) Toxicon (In press). 405. Kopec KO, Alva V, & Lupas AN (2010) Bioinformatics 26(16):1927-1931. 406. Beamer LJ, Carroll SF, & Eisenberg D (1997) Science 276(5320):1861-1864. 407. Kleiger G, Beamer LJ, Grothe R, et al. (2000) Journal of Molecular Biology 299(4):1019-1034. 408. Eckert JK, Kim YJ, Kim JI, et al. (2013) Immunity 39(4):647-660. 409. Garland AL, Walton WG, Coakley RD, et al. (2013) Proceedings of the National Academy of Sciences of the United States of America 110(40):15973- 15978. 410. Hamiaux C, Basten L, Greenwood DR, et al. (2013) Journal of Structural Biology 182(3):259-263. 411. Hamiaux C, Stanley D, Greenwood DR, et al. (2009) Journal of Biological Chemistry 284(6):3496-3503. 412. Jeong H, Park J, & Lee C (2016) EMBO Reports 17(12):1857-1871. 413. Kolodziejczyk R, Bujacz G, Jakob M, et al. (2008) Journal of Molecular Biology 377(3):870-881. 414. Liu S, Mistry A, Reynolds JM, et al. (2012) Journal of Biological Chemistry 287(44):37321-37329. 415. Mueller GA, Edwards LL, Aloor JJ, et al. (2010) Journal of Allergy and Clinical Immunology 125(4):909-917 e904. 416. Qiu X, Mistry A, Ammirati MJ, et al. (2007) Nature Structural & Molecular Biology 14(2):106-113. 417. Schauder CM, Wu X, Saheki Y, et al. (2014) Nature 510(7506):552-555. 418. Suzuki R, Fujimoto Z, Shiotsuki T, et al. (2011) Scientific Reports 1:133. 419. Tan KW, Jobichen C, Ong TC, et al. (2012) PLoS One 7(9):e44850. 420. Vance SJ, McDonald RE, Cooper A, et al. (2013) Journal of the Royal Society, Interface 10(85):20130453. 421. Walton WG, Ahmad S, Little MS, et al. (2016) Biochemistry 55(21):2979- 2991. 422. Wong LH & Levine TP (2017) Biochimica et Biophysica Acta 1864(9):1439- 1449. 423. Luecke H & Quiocho FA (1990) Nature 347(6291):402-406.