Characterizing the novel GTPase function of folliculin and its role in tumourigenesis. Sebastien Latapie. Department of Biochemistry

Transcription

1 Characterizing the novel GTPase function of folliculin and its role in tumourigenesis Sebastien Latapie Department of Biochemistry McGill University, Montreal Submitted December 2011 A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Master of Science Sebastien Latapie 2011

2 Abstract Birt-Hogg-Dube (BHD) is an autosomal dominant neoplasia syndrome that increases the risk of developing kidney cancer. Skin hamartomas, lung cysts, pneumothorax and a wide variety of kidney tumours characterize BHD syndrome. Germline inactivation of the gene, which encodes folliculin protein (FLCN), often results in early protein truncation suggesting that loss of function is responsible for development of tumours. The molecular and cellular mechanisms by which FLCN suppresses tumourigenesis remain unclear. Previously, our laboratory demonstrated that FLCN harbored a GTPase activity that was lost in common BHD patient mutations. In this report, we revisit these experiments and conclude that GTPase activity is much weaker than previously observed and is not lost in a panel of BHD mutants. Additionally, GTPase experiments with recombinant GST- FLCN fusion proteins expressed in a bacterial system show little to no activity. Taken together, these data suggest that FLCN does not harbor GTPase activity but may be acting as a regulator of GTPase activity. 2

3 Résumé Birt-Hogg-Dube (BHD) est un syndrome néoplasique de transmission autosomique dominante qui accroît le risque de développer un cancer du rein. Le syndrome BHD est caractérisé par le développement de harmatomes de la peau, kyste pulmonaire, pneumothorax est de diverses tumeurs rénales. Les mutations germinales de BHD, codant pour la protéine folliculine (FLCN), ont pour conséquence une troncation de la protéine ce qui suggère une perte de fonction contribuant au développement de tumeurs. Les mécanismes d actions cellulaires et moléculaires utilisé par FLCN pour empêcher le développement de tumeurs demeure irrésolu. Précédemment, notre laboratoire de recherche a démontre que FLCN procédait une activité GTPase qui était perdue naturellement par des mutations affectants des patients atteints du syndrome BHD. Cette thèse revisite les essaies expérimentaux antérieurement effectués et conclue que l activité GTPase et beaucoup plus faible que précédemment observé et n est pas perdue par des mutations observée chez les patients atteints du syndrome BHD. De plus, des expériences de GTPase avec des protéines recombinant de fusion GST-FLCN exprimées dans un hôte bactérienne possède peu voire aucune activité GTPase. En conclusion, les résultats présentés suggères que FLCN ne possède pas d activité GTPase mais pourrait jouer un rôle dans la régulation de son activité. 3

4 Acknowledgments First and foremost I would like to thank my supervisor Arnim Pause, for giving me the opportunity to work in his laboratory, providing valuable feedback, and helping me achieve my goals in academia. I would also like to thank past members of the lab Abbas Ghazi and Anders Dydensborg. Abbas, thank you for your help with all aspects of my research and believing in me. Your advice, guidance, and most importantly positive attitude motivated me to stay on track, not give up, and pursue my goal. Anders, thank you for taking me under your wing. You were a mentor and inspiring person to work with. I want to acknowledge all current members of the Pause lab for their invaluable help in all the projects I dabbled in. Thank you: Yann for your help and unrivaled maxi prep skills, Ming for your assistance and support when sub-cloning became a true puzzle, Elite for her energy and encouraging conversations, Zhara for being a great bench partner and her subtle humor, Sanaz for her integrity and abundant kindness, Vlad for his miniprep skills and uplifting humor, Marie-Claude for believing in me and her abundant help and guidance when I just joined the lab, and finally Dmitri for his great advice, unsurpassed western skills, and friendship. I wish you all the best in the year to come. Finally, I would like to thank my friends and family for their consistent support since the day I started my research. 4

5 Preface-Contribution of Authors All the work presented in this thesis is entirely my own except for the following: In part III, all the cloning, protein expression, purification and crystallization of the recombinant FLCN CT was performed by Dr. Ravi K. Nookala (University of Cambrdige UK); the vector constructs used for the first GTPase assay were provided by Abbas Ghazi; Dr. Marie-Claude Gingras generated the R527X FLCN mutant in pcdna3.1 used in all assays; Ming Yan independently performed part of the second GTPase assay. All work was done under the supervision of Dr. Arnim Pause. The thesis was written solely by me, with revisions by Arnim Pause. 5

6 Table of contents Abstract... 2 Resume... 3 Acknowledgments... 4 Preface-Contribution of Authors... 5 Table of contents... 6 List of figures... 8 Introduction... 9 Renal Cell Carcinoma... 9 Birt-Hogg-Dube Syndrome... 9 Rationale and objective of this study Part I: Establishing BHD as a Tumour Suppressor Gene Discovery of Oncogenes and Tumour Suppressor Genes Oncogenes Tumour Suppressor Genes Knudson Two-Hit Hypothesis Haplo-insufficiency models Birt-Hogg-Dube Tumour Suppressor Gene Part II: Cellular functions of BHD gene product folliculin AKT-AMPK-TSC-mTORC pathway FLCN involved in AMPK-mTORC1 signaling pathway The FLCN controversy Alternate pathways affected by FLCN Part III: Exploring folliculin as a GTPase P-loop GTPases Crystal structure of folliculin C-terminal Results and Discussion

7 Immuno-precipitated FLCN WT and K485A GTPase assays Immuno-precipitated FLCN mutant GTPase assays GST tagged FLCN WT and FLCN K485A GTPase assay Conflicting functional role of FLCN Materials and Methods pcdna3.1 2X FLAG FLCN constructs pgex-2tk FLCN constructs Expression and purification of FLAG constructs Expression and purification of GST constructs GTPase assay Conclusion References

8 List of figures Figure 1: Mutational spectrum of germline BHD Figure 2: FLCN CT crystal structure Figure 3: GTPase assays with FLCN WT and FLCN K485A constructs Figure 4: GTPase assay with range of FLCN mutants Figure 5: GTPase assay comparing GST purified FLCN WT and FLCN K485A to FLAG purified FLCN WT

9 Introduction Renal Cell Carcinoma In 2010, kidney cancer affected 4,800 Canadians and is estimated to affect an additional 5,100 Canadians in 2011 [1]. An estimated 1 in 141 males and 1 in 231 females will die from kidney cancer in Canada. The incidence of kidney cancer is on the rise [2] and is the 13 th leading cause of cancer related death in Canada. This is in part due to the lack of early warning signs, making it difficult to detect the cancer in its early stages, which inherently decreases the effectiveness of its treatment. Kidney cancer is not a single disease but a wide variety of different types of cancer that occur in the kidney. Similarly to other cancers such as prostate, colon and breast cancer, kidney cancer occurs both in sporadic (non-familial) and hereditary forms. Although the majority of renal cell carcinomas (RCC) are sporadic in nature, it is estimated that 1-4% of cases are due to inherited predispositions [3]. Hereditary RCCs are distinct from sporadic RCCs in several aspects. Generally, hereditary RCCs manifest themselves with the development of bilateral, multifocal lesions while sporadic RCCs typically only affect a single kidney [4]. Additionally, hereditary RCCs tend to manifest themselves earlier than sporadic RCCs. RCCs are histologically diverse lesions characterized as five distinct subtypes: clear cell (conventional) renal carcinoma, papillary type I and II RCC, chromophobe renal carcinoma, oncocytoma, and collecting duct carcinoma [2]. Over the past 25 years, families with inherited RCCs have been extensively studied. This has led to the identification and characterization of seven 9

10 autosomal-dominant inherited renal cancer syndromes for which the predisposing genes have been identified. The first gene identified for hereditary RCC was the von Hippel-Lindau (VHL) gene, which using positional cloning strategies was identified on the human chromosome 3p26 [5]. The VHL gene encodes a protein that forms a stable complex with elongin B, elongin C, and cullin 2 [6]. Together this complex targets hypoxia-inducible factors HIF1-α and HIF2- α for ubiquitin mediated degradation [7]. This oxygen dependent mechanism helps control the expression levels of down stream genes, such as vascular endothelial growth factor (VEGF), glucose transport 1 (GLUT1), and epidermal growth factor receptor (EGFR). Germline mutations of the VHL gene have been found in almost all families affected by this hereditary syndrome [8]. Inactivation of the gene occurs through mutations, methylations and chromosomal loss [9]. These mutations lead to a loss of function, which is characteristic of tumour suppressor genes. The tumour suppressor gene follows the Knudson two hit hypothesis [10]. In the absence of properly expressed and functional VHL protein, HIF factors accumulate in the cells leading to an up-regulation of the downstream proteins VEGF, GLUT1, and EGFR. These proteins, together with additional factors, all contribute to the development of clear-cell RCCs as well as retinal and central nervous system hemangioblastomas, pheochromocytomas, pancreatic cysts, neuroendocrine tumours, endolymphatic-sac tumours, epididymal and broad-ligament cystadenomas [2]. Following this discovery, additional genes were identified and linked to the different observable kidney cancer phenotypes. The hepatocyte growth factor receptor, proto-oncogene MET, was identified as the gene responsible for hereditary papillary type 10

11 1 renal carcinoma (HPRC) [11]. Mutations to the tyrosine kinase domain of the protein cause ligand-independent activation of the kinase and deregulate downstream effectors that are tied into the AMPK-mTOR energy-sensing pathway. Hereditary leiomyomatosis renal cell carcinoma (HLRCC) has been linked to the fumarate hydratase gene. This gene encodes the mitochondrial fumarate hydratase enzyme responsible for taking fumarate to malate in the tricarboxylic acid cycle. Deficiency in the enzyme modifies the response of downstream effectors and eventually leads to the accumulation of HIF- α. This aggressive form of cancer is characterized by papillary type 2 RCCs, as well as collecting-duct carcinomas [12]. Germline mutations to the succinate dehydrogenase (SDH) genes have also been identified as a predisposing factor for the development of RCCs. These genes encode for succinate dehydrogenase, another enzyme of the tricarboxylic acid cycle. Not only do mutations to these genes lead to the development of RCCs, but they also predispose to the development of paraganglioma, pheochromocytoma, and carotid body tumours. It is expected that SDH mutants follow a similar mechanism of tumourigenesis as FH mutants [13]. Tuberous sclerosis complex 1 and 2 (TSC) encode two proteins involved in the AMPK mtor energy-sensing pathway and mutations to either of these two genes have been identified in patients with TSC disorder. These mutants predispose the affected individuals to the development of cutaneous angiofibromas, pulmonary lymphangiomyomatosis, renal angiomyolipoma lesions, and renal cell carcinoma. Birt-Hogg-Dube Syndrome Birt-Hogg-Dube syndrome (BHD) is an inherited neoplasia syndrome that was first reported by three Canadian physicians in 1977[14]. The Canadian physicians were 11

12 the first to describe the observable dermatological features benign fibrofolliculomas made up of fibroepithelial polyps and skin tags, on the neck and face. These skin hamartomas were inherited in autosomal dominant manner. Further studies have revealed that those affected by BHD syndrome are also predisposed to develop pulmonary cysts, spontaneous pneumothorax, and renal neoplasms. The risk of spontaneous pneumothorax for affected individuals is 50 times greater than non-affected individuals, and the risk for developing renal tumours is 7 times greater [15]. Unlike renal tumours present in patients with other inherited diseases, renal tumours found in individuals affected by BHD are bilateral, multifocal and have a varied histology including chromophobe RCC (34%), clear cell carcinoma (9%), and papillary RCC (3%) [16]. The most commonly observed tumours are oncocytic hybrid tumours (50%), which contain zones that are histologically consistent with oncocytomas, zones consistent with chromophobe RCC, and mixed zones [16]. Using fibrofolliculomas as a marker for determining affected individuals, a linkage assay was performed and the BHD locus was mapped to the human chromosome 17p11 [17]. Further studies using mutation analysis and positional cloning strategies pinpointed the locus to chromosome 17p11.2 and identified a new gene BHD [18]. The human BHD gene contains 14 exons (11 coding and 3 non-coding) and the BHD full length mrna is 3674 nucleotides. The BHD cdna encodes two isoforms of a novel cytoplasmic protein (FLCN), named after the fibrofolliculomas observed on affected individuals. The first isoform produces a 579 amino acid protein with an estimated molecular weight of 64 kda, while the second isoform produces a shorter 342 amino acid protein with a distinct C-terminus. FLCN shows no significant sequence or functional 12

13 homology with any other known protein, yet is highly conserved across species (90% identical with cow, monkey, dog, rat, and mouse homologs), suggesting an important biological function. 13

14 Rationale and objective of this study Although the genetic basis of BHD syndrome is well understood, the cellular and molecular mechanisms involving the BHD encoded protein folliculin (FLCN) remain unclear. For example, multiple reports propose contradictory cellular functions for FLCN and its mechanism of function is poorly studied. Therefore, the overall objectives of this study are to review the establishment of BHD as a tumour suppressor gene, present the currently debated cellular functions of FLCN, and further characterize FLCN biological functions. Part I: Establishing BHD as a Tumour Suppressor Gene Part II: Cellular functions of BHD gene product folliculin Part III: Exploring folliculin as a GTPase 14

15 Part I: Establishing BHD as a Tumour Suppressor Gene Discovery of Oncogenes and Tumour Suppressor Genes The German biologist, Theodor Boveri, originally established the concept of tumour suppressor genes in He suggested that there existed chromosomes that were able to promote the growth of malignant tumours, and others that were able to suppress the development of these tumours [19]. Boveri s observations led him to theorize that cancer starts from a cell in which the genetic makeup has been scrambled, allowing the cells to proliferate uncontrollably. While his work was long overlooked, this insight was the foundation that led to the identification of both tumour suppressor genes and oncogenes. Oncogenes In 1960, with his discovery of the Ph chromosome, Peter Nowell was the first to observe evidence of a genetic abnormality causing cancer. Nowell observed the chromosome abnormality while studying chronic myelogenous leukemic cells, a cancer which affects white blood cells [20]. This observation, however, was met with significant skepticism, and it was revealed that these specific chromosomal changes were not characteristic of other malignancies. In 1973, however, Janet Rowley showed that the observed abnormality was the result of a genetic recombination between chromosomes 9 and 22 [21]. More importantly, Rowley s observation that this recombination was present in almost all bone marrow cells from CML patients strongly suggested that the genetic recombination was an initiation factor of the cancer, and not a result of the malignancy. 15

16 This discovery provided strong credibility to the idea that modified genetic material is a cause of cancer, which helped propel the field of cancer genetics. The discovery of c-src by J. Michael Bishop and Harold E. Varmus in 1979 [22] was paramount in furthering our understanding of cancer as a disease related to the deregulation of normal functioning genes within cells. In 1911, Francis Peyton Rous first proposed that cancer was related to viruses. He showed that upon grinding a malignant tumour from a chicken and exposing the cell free lysate to healthy chicks, the healthy chicks would eventually develop sarcomas. The virus responsible for the infection was called rous sarcoma virus (RSV) and contained the v-src gene, gene encoding a constitutively activated tyrosine kinase, which was essential for the development of the sarcomas. The discovery that there was a normal gene within healthy chicken cells that had high homology to the viral counterpart was ground breaking. As it turns out, the cancer causing gene v-src was a co-opted and mutated version of c-src. These observations demonstrated that certain normal cellular genes when mutated are able to cause cancer, validating one of Boveri s theories. This led to the discovery of numerous other cellular oncogenes and proto-oncogenes, which activated through mutations or translocations, lead to the development of tumours. Tumour Suppressor Genes Although an increase in dosage or activity of specific genes is a recognized cause of tumourigenesis, the concept of decreased gene expression and protein activity as a mechanism towards tumourigenesis has taken longer to gain acceptance. These second types of genes, known as tumour suppressor genes, were identified in the 1980s and 16

17 validated the second theory proposed by Boveri. These genes inhibit cancer development within the cell and oppose oncogene function. A tight and regulated balance between tumour suppressor genes and oncogenes is essential to maintain homeostasis and carefully regulated cell proliferation. An imbalance to either of these types of genes can lead to the development of cancer. Henry Harris was the first to demonstrate proof that tumour suppressor genes existed with his somatic cell fusion experiments in 1969 [23]. Further experiments demonstrated that the conservation of specific chromosomes could suppress the malignant phenotype. When fusing malignant mouse fibroblasts to normal mouse fibroblasts, the resultant could fully suppress the malignancy if both copies of chromosome 4 were kept, and nearly fully suppress the malignancy if only a single copy of the chromosome 4 was kept [24]. These experiments demonstrated that certain genes could repress tumour development. Knudson Two-Hit Hypothesis Cancer susceptibility syndromes are usually inherited in a dominant fashion, yet the tumour suppressor gene model suggests a recessive mode of inheritance for cancer susceptibility, where both genes need to be inactive in order to produce a cancerous phenotype. This contradiction was reconciled by Alfred G. Knudson in 1971, with his statistical analysis of retinoblastoma [10]. The disease primarily affects children and arises from fetal retinoblast cells that normally differentiate into post-mitotic retinal photoreceptor cells and neurons. In diseased patients, retinoblast cells fail to differentiate and continue to go through the cell cycle, which can eventually spread and metastasize. Knudson s main intrigue was his observation that certain individuals with a heritable predisposition that never developed a tumour gave birth to children afflicted by 17

18 retinoblastoma, demonstrating that the parent carried the germline mutation. This implied that a single mutation was not sufficient for tumourigenesis. Further statistical analysis of 48 patients revealed to Knudson that the number of tumours per heritable case followed a Poisson distribution. He used this information to infer that 5 percent of the individuals with a heritable pre-disposition would not develop tumours, a percentage that matched the observations. Additionally, he observed that the distribution of bilateral fraction that had not been diagnosed showed linear decline. This is in accordance to a one hit model of tumourigenesis By contrast, unilateral cases with no hereditary form followed a two hit model, with both mutations being somatic. In both the hereditary and sporadic forms of retinoblastoma, two hits are required to initiate tumourigenesis. In the hereditary form the first hit is a germline mutation and second hit is somatic, while both are somatic in the non-hereditary form of retinoblastoma. Whether these two hits were dominant mutations affecting two distinct genes, or if they were recessive mutations to the same gene on each allele remained unclear until the 1980s brought technological advancements and the development of restriction length fragment polymorphism mapping technology. The application of this technique allowed the identification of a locus on chromosome 13 that was consistently deleted in tumours, as well as the eventual cloning and characterization of the first human tumour suppressor gene RB1 [25-29]. These studies clearly showed that families with hereditary retinoblastoma contained mutations or deletions of RB1 in the germline and that retinoblastoma tumours almost always contained the mutation or deletion of the second RB1 gene. 18

19 Together, these observations validated the two hit model and helped explain how recessive tumour suppressor genes could account for a dominantly inherited cancer susceptibility phenotype. Due to the high chance of a second hit leading to tumourigenesis, the cancer susceptibility phenotype is inherited dominantly, however tumourigenesis requires two hits and is therefore recessive. Additionally, this discovery helped establish a strategy to study and uncover new tumour suppressor genes. This framework consisted of a) identifying a deleted chromosomal region in tumours or patients with a hereditary cancer syndrome, and b) subsequently searching for the gene in that region that was mutated in other allele or completely missing in homozygosity. This method was successfully implemented in the identification of VHL [30, 31], BRCA1 [32, 33], BRCA2 [34], and APC [35] tumour suppressor genes. This approach was essential in identifying p53 as a tumour suppressor gene rather than an oncogene, as it was initially established [31, 36-38]. Haplo-Insufficiency Models Applying the two hit model to sporadic cancers was not as successful as anticipated. Although researchers identified numerous areas of chromosomal deletions in various cancers (a characteristic indicative of a tumour suppressor gene), many of these consistently deleted regions were not linked to an obvious mutation or deletion on the other allele. This led many to believe that the presumed tumour suppressor of that region had either not been found, or simply did not exist (even if many genes in these loci had been shown to have tumour suppressor properties in vitro and in vivo). As a consequence, it was necessary to revisit and refine the mechanism of tumorigenesis after the inactivation of a tumour suppressor gene. One proposed model was that of haplo- 19

20 insufficiency of tumour suppressor function. This implied that a single-copy loss of a tumour suppressor gene is sufficient to deregulate its expected function and promote cancer. Unlike the two hit tumour suppressor gene model, there is no single approach that can definitively prove the tumour suppressor gene s involvement in tumourigenesis. This model requires an integrated approach involving genetic analysis, murine modeling, as well as in vitro and in vivo functional assays in order to fully implicate a gene in haploinsufficient tumour suppressor model. One such gene, among many others, is the p53 tumour suppressor gene [39]. The p53 gene is involved in many critical cellular functions such DNA damage repair, cellular cycle arrest, and apoptosis as a stress-responsive transcription factor [40]. Although p53 may sometimes fit the two hit-hypothesis [38], the first evidence that suggested it exhibited haplo-insufficiency came from the analysis of individuals affected by the Li- Fraumeni Syndrome. This rare cancer susceptibility syndrome is characterized by a high susceptibility to a variety of cancers caused by a germline mutation of the p53 gene [41, 42]. It was initially expected that the majority of tumours present in individuals with Li- Fraumeni Syndrome would select for loss of heterozygosity and fit the two hit model. In actuality, only 60% of the tumours exhibited loss of heterozygosity at the p53 locus [42]. The question still remained whether the p53 mutations acted as a dominant negative or that loss of a single allele was sufficient to contribute to the development of cancer. In response to this question, an analysis of mice with targeted deletion of p53 was used. Heterozygous p53(+/-) mice were shown to have an intermediate survival rate compared to homozygous p53(-/-) mice and p53 wild type (+/+) mice [43]. Analysis of the tumours revealed that the heterozygous p53(+/-) mice did not always lose or contain a mutated 20

21 form of the remaining p53 allele [43]. Additionally, it was shown that p53 heterozygous thymocytes had intermediate and partial resistance to apoptosis induced by ionization [44]. Further studies in isogenic heterogeneous p53-deleted HCT116 cell lines showed a 4 fold reduction in mrna and protein expression, as well as an impairment of induction of p53 target genes and apoptosis after being exposed to ultra-violet radiation [45]. Taken together, these studies provide strong evidence that a reduction of p53 dosage and function is sufficient for tumorigenesis, and strongly support haplo-insufficiency of tumour suppressor genes as a mechanism that promotes cancer. Similarly to p53, encoding phosphatase and tensin homolog (PTEN) is a key tumour suppressor gene that plays an important role in suppressing many cancers. It is among the most frequently inactivated tumour suppressor genes in sporadic cancers, including breast cancer, colon cancer, prostate cancer, thyroid cancer, and endometrial cancer [46]. The importance of PTEN as a tumour suppressor is further enforced by the study of PTEN germline mutations found in autosomal dominant syndromes collectively known as PTEN hamartoma tumour syndromes [47]. Analysis of human and mouse tumours with germline PTEN mutation does not always exhibit loss of the wild type PTEN allele, supporting the notion that PTEN is a halpoinsufficient tumour suppressor gene [47]. As is the case with p53, tumour suppressor gene haplo-insufficiency is capable of initiating and contributing to tumorigenesis, while homozygous loss of the tumour suppressor gene furthers tumorigenesis. Interestingly, depending on the context, complete loss of Pten does not further induce tumorigenesis [48] but instead triggers a p53- dependent fail-safe senescence mechanism called Pten-loss-induced cellular senescence [48], which inhibits cancer progression. This, along with many other genes [49, 50], 21

22 defines a novel mechanism of tumour suppression known as obligate haploinsufficiency, in which the most tumourigenic phenotype lies at a midpoint of protein expression, rather than at a complete loss. This mechanism of action is highly tissue-specific and contextdependent. In the case where p53 is mutated or deleted, Pten-loss-induced cellular senescence cannot be activated and complete loss of Pten further enhances tumourigenesis [51]. The field of cancer genetics has substantially evolved over the past century. As technology advances, the complexity of cancer genetics is becoming increasingly evident. It is now believed that non-hereditary sporadic cancers require approximately 4 or more distinct mutational events affecting key cellular signaling pathways, including phopshoinositide-3-kinase, p53, or RB. Malignancy results from the iterative process of somatic mutations, followed by clonal expansion. Additionally, a deeper understanding of the subtle and complex functions of such genes is slowly overshadowing the discrete step-wise models used to identify the involvement of tumour suppressor genes in tumorigenesis. A focus on the cellular and molecular context as well as the profound effect of slight dosage variations is essential to further our understanding of tumour suppressor genes. For example, in one specific tissue, there may be additional proteins that mask the expression of a cancerous phenotype caused by heterozygosity of a tumour suppressor gene, yet in another tissue, these proteins may not be expressed and the heterozygosity of the tumour suppressor gene will exhibit itself as a tissue-specific cancer. This underlines the importance of using more complex experiments, such as mouse models and functional studies, to help determine the functional relevance of heterozygous loss towards tumourigenesis. It is clear that precise regulation of tumour suppressor gene 22

23 expression and function is critical for proper function. It is therefore also imperative to explore the mechanism of regulation and function of tumour suppressor genes with regards to tumour suppression. Birt-Hogg-Dube Tumour Suppressor Gene As previously mentioned, BHD syndrome is an autosomal dominate inherited neoplasia syndrome characterized by a predisposition to hamartomatous skin lesions, pulmonary cysts, and renal neoplasms [15]. Using 9 families identified with BHD skin lesions, including the original Canadian family described by Birt Hogg and Dube [14], Schmidt et al. performed an initial genome wide scan and 2 point linkage analysis to attempt to identify a disease gene locus for BHD [17] This analysis pointed to chromosome 17p11.2, a highly gene rich chromosomal region encoding many proteins involved in regulation of cell proliferation and cell differentiation. Further analysis by recombinant mapping narrowed the BHD disease locus and led to the eventual identification of a novel gene named BHD encoding folliculin [18]. More importantly the study identified a number of germline mutations in the BHD locus. Further screening of 60 afflicted families revealed 22 novel germline mutations in the BHD locus [52]. These mutations included 16 insertion/deletion events leading to a frame shift mutation and premature truncation of the protein, 3 nonsense mutations, and 3 putative splice-site mutations. More than half of all germline mutations had a cytosine insertion or deletion in the mononucleotide tract of eight cytosines in exon 11 [52]. Mononucleotide tracts are often susceptible to DNA polymerase slippage during replication, which leads to frame shift mutations and subsequently protein truncations. Hyper mutable mononucleotide tracts have been observed in other heritable disorders[53], and suggest a loss of function 23

24 mechanism leading to tumour development. To date, the BHD mutational spectrum includes a total of 55 germline mutations (Figure 1) [54]. The identification of a high frequency of inactivating germline mutations at the BHD locus suggested that BHD might act as a tumour suppressor. Figure 1. Mutational spectrum of germline BHD BHD germline mutations reported by the NCI and other investigators. Figure taken from Toro, J.R., et al. [54]. 24

25 Genetic studies of VHL, hereditary papillary renal carcinoma, and hereditary leiomyomatosis, have led to the identification of a number of kidney cancer specific genes [11, 13, 30]. These genes are often cell-type specific and have also been found in the sporadic counterparts of the same tumour type. Unlike renal tumours in patients with other inherited diseases, BHD patients see a wide variety of neoplasms [16]. In a study of sporadic cases of renal cell carcinomas, somatic inactivation and loss of heterozygosity of BHD locus was identified in renal cell carcinoma of papillary clear cell, chromophobe and oncocytoma subtypes [55]. The Sprague-Dawley rat model, also known as the Nihon rat, is an example of Mendelian dominantly inherited predisposition of renal carcinoma. It carries a single gene mutation predisposing it to clear cell type renal carcinoma. This gene was mapped to rat chromosome 10, a region homologous to human chromosome 17p11.2, and subsequently identified as a rat BHD homolog [56]. Loss of heterozygosity at the BHD locus was examined in 11 tumours from primary renal carcinomas in male Nihon rats. Ten out of the 11 tumours showed loss of heterozygosity at the BHD locus [56]. In addition to the genetic studies, the research group also evaluated the protein levels and found them comparable to normal tissue levels. Further analysis revealed cytosine insertions in the cdna of the renal carcinoma cell lines, leading to a frame shift mutation and premature termination of the protein [56]. A third research group analyzed 77 renal tumours from BHD patients by direct DNA sequence analysis [57]. The tumours and matched normal samples were taken from 12 individuals. Using PCR and DNA sequencing they identified and confirmed the germline mutations in each tumour. More importantly they identified a second hit somatic mutation or loss of heterozygosity in at least 1 tumour from each of the 12 25

26 patients. In total, 54 of the examined tumours showed loss of heterozygosity, while no mutation was observed in the normal tissue. Unlike the Nihon rat model, the somatic mutations were distributed throughout the BHD coding region. The frequency of loss of heterozygosity or secondary somatic hit varied by tumour type, but the overall frequency of a second hit was great than 50% [57]. In addition to renal tumour analysis, two research groups explored the involvement of BHD gene inactivation in the other BHD syndrome phenotype manifestations: fibrofolliculomas and spontaneous pneumothorax. As previously mentioned, individuals affected by BHD syndrome have a 50 fold increased risk for the development of pneumothorax over individuals unaffected by BHD syndrome. Although the role of lung cysts in the mechanism leading to spontaneous pneumothorax in BHD syndrome has not been established, lung cysts appear to be a precursor lesion leading to such afflictions [54, 58]. In their study, the research group identified lung cysts in 84% of patients with germline BHD mutations. Another study aimed to investigate the status of BHD in fibrofolliculomas. Upon analysis of fibrofolliculomas from three patients, the research group identified germline mutations in all patients. They then proceeded to search for loss of heterozygosity or second hit mutations at the DNA level, but comparative sequence analysis with healthy peripheral blood lymphocytes showed no sign of sequence deviations [59]. To further determine the inactivation profile of BHD locus in fibrofolliculomas, laser capture micro-dissection was performed on tumour tissue followed by DNA analysis in multiple tumour fields. Once again, there was no sign of second hit mutations or loss of heterozygosity [59]. 26

27 Taken together this data shows that BHD syndrome is characterized by germline mutation often leading to a truncated protein. Analysis of kidney tumours from rat models, sporadic renal carcinomas, and BHD patients reveal inactivation of both copies of BHD alleles in a large number of these tumours. Analysis from lung cysts and fibrofolliculomas suggest that inactivation of a single copy is sufficient to initiate tumour formation. This data confirms BHD as a tumour suppressor gene and, depending on the specific tissue, suggests conformity to the Knudson two hit tumour suppressor model [54-57] or haploinsufficiency model [54, 58, 59]. 27

28 Part II: Cellular functions of BHD gene product folliculin Mutational analyses of BHD tumours from patients, as well as from a number of animal models of BHD syndrome, have identified a high frequency and wide spectrum of second hits, strongly establishing BHD as a tumour suppressor. The observation of biallelic inactivation of BHD in several histological types of renal cancer suggests the involvement of BHD at early stages of kidney oncogenesis. While the genetic basis of BHD syndrome is well understood, the molecular and cellular mechanisms by which BHD suppresses tumourigenesis remain unclear. Research in the past decade has provided evidence linking several hamartoma syndromes to the convergent energy and nutrient sensing pathways involved in mammalian target rapamycin (mtorc1) regulation. These autosomal dominant inherited syndromes include Cowden syndrome, Peutz-Jeghers syndrome, and tuberous sclerosis complex (TSC). Cowden syndrome, a subset of PTEN hamartoma tumour syndrome, is characterized by a germline mutation of the PTEN tumour suppressor gene [60]. The germline mutation has been reported in 80% of affected individuals. Affected individuals develop hamartomas of the hair follicle, the mucocutaneous membranes, breast, thyroid, and intestinal tissues [61]. Peutz-Jeghers syndrome is characterized by a germline mutation of the LKB1 tumour suppressor gene identified in 80-94% of afflicted individuals [62, 63]. Patients with Peutz-Jeghers syndrome can develop mucocutaneous hyperpigmentation of the hand and lips, vascular hamartomas throughout the gastrointestinal tract, and are at an increased risk of developing malignancies such as gastrointestinal tumours, lung tumours, and breast cancer [63]. TSC is the manifestation 28

29 of a germline mutation affecting one of two tumour suppressor genes: TSC1 or TSC2 [64]. TSC patients develop a number of hamartomas that can affect the brain, skin, lung, kidney, liver, and heart. TSC patients also develop a number of renal neoplasms, such as renal cell carcinoma, as well as pulmonary lymphoangiomyomatosis [64]. BHD syndrome is also an autosomal dominant inherited syndrome and shares clinical manifestations with the three hamartoma disorders described above, with particular similarity to TSC. These phenotypic similarities strongly suggest that BHD may be involved in functions universal to the three syndromes. AKT-AMPK-TSC-mTORC Pathway Protein kinase B (also known as Akt), adenosine mono phosphate activated kinase (AMPK), tuberous sclerosis complex (TSC), and mammalian target of rapamycin complex 1 (mtorc1) signaling pathway play a central role in regulating cell metabolism and cell growth by integrating information regarding cellular energy, nutrient availability, oxygen status, and presence of growth factors. If the cell environment is favorable, this pathway will stimulate growth. In the case of nutrient starvation and low energy, this pathway will alter the cell s metabolism to increase metabolite availability and avoid death. Dysregulation of this pathway is implemented in a number of diseases where growth and homeostasis are compromised, such as cancer. LKB1 (liver kinase B1) is a ubiquitously expressed and evolutionary conserved serine-threonine kinase [65]. It functions as a heterotrimer with STRAD (sterile-20- related adaptor) and MO25 (mouse protein-25). LKB1 is normally found in an inactive state in the nucleus and is activated by its translocation to the cytosol, where it 29

30 phosphorylates 14 members of the AMPK serine/threonine kinase family [65]. Its most studied substrate is AMPK, a heterotrimer made of a catalytic subunit AMPKα, and two regulatory subunits AMPKβ and AMPKγ [66]. When activated, LKB1 can activate the AMPKα subunit by phosphorylating it. In response to an increase in the AMP/ATP ratio, whether due to excess ATP consumption or decreased ATP production, AMPKγ binds AMP causing a conformational that allows phosphorylation of AMPKα subunit [66]. To help restore energy balance in the cell, activated AMPK suppresses energy consumption such as lipid synthesis, while up-regulating energy producing pathways such as glycolysis [65]. An important downstream substrate of AMPK is TSC2. TSC1 and TSC2 encode hamartin and tuberin, which physically interact to form a dimer [64]. These proteins are widely expressed in human tissues, including the liver [67]. Tuberin contains a C- terminal GTPase activating protein (GAP) domain. Though hamartin does not contain any catalytic domains, it plays an important role in preventing ubiquitin-mediated degradation of tuberin. Phosphorylation of hamartin inhibits the complex, while phosphorylation of tuberin can have an activating or inhibiting function. AMPK can activate tuberin by phosphorylating at serine 1345 and threonine 1227 [67]. In response to growth factors, such as insulin and various cytokines, TSC1:TSC2 complex can be inactivated by phosphorylation of tuberin mediated by extracellular signal-regulated kinase (Erk) and Akt [64]. Erk is a downstream kinase of the Ras-Raf-MAPK singling pathway. Akt binds phosphoinositides PIP 2, which is phosphorylated to PIP 3 by phosphoinositide 3-kinase (PI3K) in response to growth factors. Phosphorylation of PIP 2 to PIP 3 causes a conformational change allowing proper orientation of Akt for activating 30

31 phosphorylation by phosphoinositide dependent kinase 1 (PDK1) [46]. PTEN possesses a lipid phosphatase activity that opposes the function of PI3K, dephosphorylating PIP 3 to PIP 2. PTEN is therefore negatively regulating Akt, and as a consequence, promoting TSC2 activation [46]. The three pathways described above, LKB1-AMPK, Ras-Raf-MAPK, and PI3K- Akt all channel their signals to the TSC1:TSC2 complex. This complex s main cellular function is the regulation of mtorc1 complex through Rheb, a small G protein that activates mtorc1 when bound to GTP[64]. Therefore, when TSC1:TSC2 is activated, TSC2 will help catalyze the GTP hydrolysis of Rheb, which will reduce its activating interaction with mtorc1, thereby negatively regulating mtorc1[64]. mtorc1 is a multi-protein complex composed of a catalytic mtor subunit, as well as defining regulatory-associated protein of mtor (RAPTOR). It is also a 40kDa Pro-rich Akt substrate (PRAS40), mammalian lethal with SEC13 protein 8 (mlst8), and DEP domain-containing mtor interacting protein (DEPTOR) [68]. In addition to mtorc1, the catalytic mtor domain along with DEPTOR, mlst8, rapamycin-insensitive companion of mtor (RICTOR), proteins observed with RICTOR (PROTOR), and map kinase interacting protein 1 (msin1) form another complex known as mtorc2. This complex positively regulates Akt, which in turn further up-regulates mtorc1 [68]. In addition to TSC1:TSC2 complex, mtorc1 is regulated by Akt by phosphorylating PRAS40 that prevents it from inhibiting mtorc1. mtorc1 signaling is also regulated by the presence of amino acids, as well as by chemical inhibitors of glycolysis and mitochondrial function [68]. 31

32 The major mechanism by which mtorc1 controls cell growth is by regulating protein synthesis. Ribosomal S6 kinase 1 (S6K1) and eukaryotic initiation factor 4E binding protein 1 (4EBP1) are the most extensively characterized downstream substrates of mtorc1. When activated, mtorc1 phosphorylates S6K and 4EBP1. Phosphorylation of S6K selectively increases the translation of mrnas by phosphorylating and binding multiple proteins such as eukaryotic elongation factor 2 (eef2k), 80 kda nuclear cap-binding protein (CBP80), and eukaryotic translation factor 4B (eif4b) which all contribute towards enhancing the translational capacity of the cell. Phosphorylation of 4EBP1 causes dissociation from eif4e, which can then recruit translation initiation factor eif4g to the 5 cap of mrna[68]. One consequence of this protein synthesis up-regulation is the induction of angiogenesis, an important contributing factor for the development of tumours, by inducing hypoxia-inducible factor 1α (HIF1α) which in turn increases the expression of vascular endothelial growth factor (VEGF). Autophagy is the catabolic process facilitating the recycling of organelles to increase the availability of nutrients within the cell. When activated, mtorc1 inhibits this process by phosphorylating ATG13 and ULK1/2 [68] mtorc1 plays a critical roll in integrating both the extracellular and intracellular signals of cell growth. Strict regulation of this protein is paramount for a coherent signal response to prevent energy imbalance and aberrant growth. With this pathway in place, we can clearly understand how the germline mutations affecting Cowden syndrome, Peutz-Jeghers syndrome, and TSC can easily dysregulate mtorc1 signaling, contributing to improper cell growth. 32

33 FLCN involved in AMPK-mTORC1 signaling pathway In addition to the phenotypic link observed between BHD syndrome and other hamartoma syndromes, studies in S. pombe fission yeast further suggested a role of FLCN in the AMPK-mTORC1 signaling pathway [69]. Mutational studies involving yeast BHD and TSC homologs revealed that in S. pombe, BHD and TSC have opposing roles in regulating amino acid permease expression and intracellular levels of specific amino acids, two downstream events modulated by yeast mtorc1 homolog [69]. The fundamental discovery associating FLCN to the AMPK-mTORC1 pathway was the identification of a novel human protein named folliculin interacting protein 1 (FNIP1) [70]. As the name suggests, this 130 kda protein binds FLCN at the C-terminal and shows similar expression patterns to FLCN, with moderate expression in both lung and kidney tissues [70]. Interestingly, analysis of germline C-terminal mutations found in BHD patients prevented the interaction of FNIP1 and FLCN [70]. The significance of this discovery stems from the fact that FNIP1 also binds to AMPK, hence tying FLCN to the AMPK-mTORC1 signaling pathway [70]. Following this discovery, two research groups independently identified an uncategorized human FNIP1 homolog, subsequently named FNIP2 [71, 72]. FNIP2 shows high sequence homology to FNIP1 at both the N-terminal and a 152 block of the C-terminal. Phylogenetic studies suggest that FNIP2 is an evolutionary divergence of FNIP1. Similarly to FNIP1, FNIP2 can bind FLCN at the C- terminal and can also interact with AMPK [71, 72]. Biochemical studies of this interaction in UOK257 cell line, a FLCN-null cell line established from a BHD patient with germline insertion mutation, revealed that FNIP1 and FLCN could be phosphorylated by AMPK, that FCLN phosphorylation is facilitated by FNIP1 in an 33

34 AMPK and mtorc dependent manner, and that FLCN can be phosphorylated at serine residue 62 and serine residue 302 [70, 73, 74]. Together, the interaction of FLCN with FNIP1, FNIP2 and AMPK strongly suggests a role for FLCN FNIP1, FNIP2 in the AMPK-mTORC1 signaling pathway. The FLCN controversy The exact role of FLCN in regulating the mtorc1 signaling pathway is highly controversial and has yet to be fully resolved. On the one hand, multiple studies performed by Baba et al. and Hasumi et al. suggest that FLCN acts as a negative regulator of the mtroc1 signaling pathway. This research group developed a mouse model containing a conditional BHD allele in cadherin 16(KSP)-Cre transgene to target BHD inactivation in the kidney [75]. The purpose of the mouse model was to study the homozygous deletion of FLCN in the kidney. When completely knocked out, the mice were afflicted by abrupt and uncontrolled renal cell proliferation leading to the development of cysts and death at 3 weeks. Analysis of the mouse kidneys revealed an increased expression and activation of both Raf-ERK1/2 and PI3K-AKT-mTORC pathways accompanied by a large increase of phosphorylated mtorc1 and phosphorylated S6R (key indicators of mtorc1 activity) [75]. Treatment with rapamycin (mtor inhibitor) decreased the growth rate and size of the BHD-null kidneys, and was accompanied by complete loss of phosphorylated S6R. Treatment with rapamycin, however, did not fully reverse the BHD-null phenotype, suggesting that other pathways were contributing to the observed phenotype [75]. This could include both the Raf-Erk and the PIK3-AKT pathways, which are not exclusively funneled through mtorc1. Together this data suggests that loss of FLCN may activate an upstream 34

35 effector of the Raf-Erk and PIK3-AKT pathways. Following this study, Hasumi et al developed a heterozygous BHD mouse model [76]. Within 10 months, these mice developed complex cyst and kidney tumours similar to those observed in BHD patients. Analysis of the tumours revealed that both copies of BHD had been deleted and that Akt was highly elevated and activated in kidney tumours, while the mrna level of Akt maintained the same [76]. They also identified the increased phosphorylation of mtor at serine 2448 and serine 2481, which activated both the mtorc1 and mtorc2 complex. The increase in mtorc1 activity was confirmed by an increase in phosphorylation of its downstream effector S6K. The increase in mtor phosphorylation was accompanied by an increase in RICTOR protein expression, allowing for increased mtorc2 complex formation and mtorc2 activity. When active, mtorc2 can phosphorylate Akt and activate it, potentially explaining the observed increase in Akt phosphorylation and activity. This increase in activity was accompanied by an increase of FOXO1 and FOXO3 phosphorylation, two transcription factors which enhance apoptosis. Phosphorylation of these transcription factors prevents translocation to the nucleus thereby inhibiting apoptosis. In addition to heterozygous mouse kidney analysis, the group studied tumours from BHD patients and found elevated levels of phospho-s6k, phospho-s6r, and phospho-akt, all suggesting that BHD loss results in the increased signaling activity of the Akt-mTORC1 pathway. Together, these two research studies suggest that the phosphorylation of FLCN is regulated by mtorc1, and through a negative feedback loop, suppresses the PI3K-AKT-mTORC1 signaling pathway [70, 72, 75, 76]. This data strongly favors an activation of Akt signaling involved in driving kidney tumorigenesis in BHD patients. 35

36 In contrast to the aforementioned reports, multiple studies have suggested an opposite function of FLCN in mtorc1 regulation. Using RNAi knockdown of FLCN, FNIP1 and FNIP2 in HeLa cell lines, Takagi et al showed a decrease in phospho S6K (an indicator of mtorc1 activity) as well as a decrease in the total levels of mtor [71]. Following this study, Hartman et al developed a heterozygous BHD mouse model to further characterize its role in the mtorc1 pathway [77]. The mouse model was generated by using a gene trap βgeo cassette inserted between exon 8 and 9 of BHD, leading to the expression of a truncated FLCN fusion protein. The mice developed small kidney cysts, microscopic cysts, and tumours in the kidney. Analysis of these tumours displayed 50% BHD mrna expression levels. Phosphorylation of S6 was lowered in the three cell lines studied (U251, HEK 293, and HK-2) and in all the heterozygous mouse tumours analyzed [77]. This suggests that the loss of FLCN leads to mtorc1 pathway inhibition, suggesting a positive regulation of mtorc1 by FLCN. These contradictory reports suggest that FLCN is more complex than initially imagined. Rather than merely positively or negatively regulating a specific target, FLCN may be capable of demonstrating both activating and inhibitory functions. This could potentially explain the differences observed in the above studies. There is the possibility that truncating mutation removes a C-terminal region responsible for mtorc1 suppression leading to a dominant negative phenotype. On the other hand, the complete loss of the protein results in an absence of the N-terminal required for the activation of mtorc1. This possibility of demonstrating both activating and inhibitory functions is supported by two studies which modified both phosphorylation residues of FLCN and monitored their affinity for AMPK [73, 74]. In this study, serine residues were either 36

37 modified to alanine, preventing phosphorylation, or aspartic acid, to mimic phosphorylation. FLCN S62A showed less affinity for the AMPKα1 subunit, while FLCN S62D showed a higher affinity for the AMPKα1 subunit compared to FLCN WT. Interestingly, FLCN S302A showed a higher affinity for AMPKα1 while FLCN S302D shower less affinity for AMPKα1 compared to WT. This suggests that phosphorylating FLCN at serine 302 inhibits its interaction with AMPK, while phosphorylating FLCN at serine 62 promotes interaction with AMPK [74]. This framework could provide a mechanism for modulating FLCN in a context dependent manner. Additionally, FLCN regulation of the mtroc1 signaling pathway may be dependent on specific cell and tissue types. Indeed, our laboratory analyzed both tumours from both the BHD heterozygous mouse model and multiple renal cancer cell lines for mtorc1 activity and found varying results [78]. An analysis revealed that solid tumours and normal kidneys showed a decrease in phospho-s6 upon decreased FLCN expression; in vivo experiments showed phospho-s6 to be elevated or negative in renal cyst, and in vitro experiments with renal cell carcinoma cell lines showed no difference in phospho-s6, regardless of growth conditions [78]. Together, these results suggest that FLCN may exhibit a complex regulation of the mtorc1 signaling pathway that is highly specific to the cell type and context. Alternate pathways affected by FLCN Although the interaction of FLCN with AMPK through FNIP1 and FNIP2 strongly suggests the involvement of FLCN in energy and nutrient sensing through mtorc1, the tumourigenesis of FLCN may be mediated through an alternative pathway. 37

38 In multiple studies and as previously discussed, BHD-null cell lines and tumours showed the activation of both the PIK3-ARK and RAS-RAF-MAPK pathways [75, 76, 78]. The PIK3-AKT pathway plays a critical role in mediating the escape from apoptosis cell cycle progression and cell proliferation, all factors that can contribute to tumorigenesis [46]. The RAS-RAF-MAPK pathway can also regulate cellular transcription and translation independently of mtorc1 [46]. A study in Drosophila showed that the BHD homolog regulates male germ-line stem cell maintenance in the fly by interacting with both JAK-STAT and decapentaplegic (dpp), a tumour growth factor β (TGF-β) family member, signaling pathways [79]. This suggests that FLCN may potentially control tumour development either by regulating stem cell development or through regulating the JAK/STAT or TGF-β pathways. Recently, two studies have provided further evidence supporting the function of FLCN in the TGF-β signaling pathway. Microarray studies in the UOK257 cell line revealed that the loss of FLCN dysregulated the expression of TGF-β signaling components including TGFB2, INHBA, SMAD7, and GREM1 [80]. A second study addressed this possibility by studying BHD-null ES cells derived from the BHD mouse model. In their study, they determined that FLCN acts downstream of TGF-β and affects Smad target gene promoters [81]. While these two studies suggest different mechanisms of action, they both propose that FLCN exerts its tumour suppressor effect through the TGF-β mediated signaling pathway. Taken together, it is clear that the role of FLCN in tumorigenesis is much more complex than initially believed. The regulation of the mtorc1 pathway by FLCN 38

39 remains a controversial topic and there is increasing evidence that the tumour suppressive function of FLCN may not involve a single pathway. 39

40 Part III: Exploring folliculin as a GTPase As previously discussed, FLCN and its interacting partners FNIP1 and FNIP2 were shown to form a complex with AMPK and interact with mtorc1, a master regulator of mammalian cell growth. Conflicting evidence of FLCN s role in the mtorc1 pathway has been reported and it remains unclear whether FLCN is directly involved or acts independently of mtorc1. Additionally, increasing evidence suggests that FLCN may perform a majority of its tumour suppressing function in alternate pathways. Studying the molecular structure of FLCN will help better understand its function and understand its mechanism of tumour suppression. P-loop GTPases Proteins that bind and hydrolyze nucleoside triphosphates are crucial for almost all aspects of life. These proteins are involved in many cellular functions and are globally categorized based on their specific chain folds. These include the dinucleotide binding fold, the tubulin/ftsz fold, the mononucleotide binding fold (also known as the P-loop NTPase fold), the protein kinase fold, the histidine kinase/hsp90/topoii fold, and the HSP70/RNAse H fold [82]. P-loop NTPases are the most common protein fold in most organisms and make up an estimated 10-18% of all gene products [18, 82, 83]. Seven monophyletic lineages of the P-loop NTPase fold can be delineated based on distinct sequence motifs and structural features. These lineages are RecA F1/F0-related ATPases, nucleic acid-dependent ATPases, AAA+ ATPases, MJ/PH/AP/NACHT NTPases, ABC- PilT ATPases, nucleotide kinases, and GTPases [82]. P-loop NTPase fold proteins are characterized, at the secondary structure level, as α/β proteins. These are made up of 40

41 recurring α-β units, with the β-strands forming a sheet which is flanked on both sides by α-helices [82]. To facilitate secondary structure comparisons, β-strands are often sequentially assigned a letter starting from the N-terminus. Sequentially, the β-strands order would read: A-B-C-D-E (and so forth), but upon secondary structure determination the spatial order would read differently, in a manner such as E-D-A-B-C. At the sequence level, P-loop NTPase fold proteins are characterized by an N-terminal Walker A motif, a flexible loop between a β-strand and α-helix which assumes the following amino acid sequence pattern: DxxxxGK[S/T] (single letter code) [84]. This structure plays an important role in properly aligning the triphosphate portion of nucleoside triphosphates. They are also characterized by a Walker B motif containing a highly conserved aspartate residue [84]. Guanine nucleotide-binding proteins and GTP hydrolysis, a fundamental reaction, control a number of essential cell processes including protein biosynthesis, signal transduction, growth control and differentiation, sensory perception, transport processes, cytoskeleton reorganization and exocytosis to name a few [85]. These proteins are precisely engineered molecular switches that can change their affinity for macromolecules based on two different states: GTP-bound on state and GDP-bound off state. The GTP-bound on state is characterized by two hydrogen bonds formed between the oxygen molecules of the γ phosphate in GTP and the main chain amino groups in switch I and switch II of the guanine nucleotide-binding protein [86]. The conformational change is best described as a loaded spring mechanism where the irreversible hydrolysis of GTP to GDP and release of inorganic phosphate allows the two switch regions to relax into the GDP-bound off state [86]. This highly versatile switch 41

42 mechanism ensures that high-affinity binding with specific effector molecules is optimized in the GTP-bound state. These GTPases are one of seven monophyletic superclasses within the P-loop NTPase fold by specific sequence patterns, structural features, and hydrolysis mechanisms. The GTPase superclass is characterized by five G-motifs [87], some of which are shared with other P-loop NTPase superclasses. The G1 motif is characterized by the DxxxxGK[S/T] amino acid sequence, also known as the Walker A motif or P- loop[84, 87, 88]. The G2 motif contains a conserved threonine involved in Mg 2+ cation binding and consists of the N-terminus of the second β-strand and the loop that precedes it. The G3 motif is characterized by the DxxG amino acid sequence, also known as the Walker B motif [84]. The aspartic acid residue plays a crucial role in binding the nucleotide associated Mg 2+ cation through H 2 0, while the conserved glycine is involved in forming hydrogen bonds with the terminal γ phosphate oxygen atom of GTP [87]. The conserved glycine residue is a specific form of the Walker B motif that differentiates GTPases from other P-loop NTPases [82]. Additionally, both the G2 and G3 motifs and the conserved residues are part of the switch I and switch II elements responsible for binding to GTP[86]. The G4 motif is characterized by the [N/T]KxD amino acid sequence which interacts with the guanine ring of GTP, conferring specificity over the other bases, and is another conserved region not found in other P-loop NTPases [82, 87]. Finally, the G5 motif is characterized by the hhe[a/c/s/t]sa[k/l] amino acid sequence (where h indicates a hydrophobic amino acid) which indirectly associates with GTP and is less well conserved within the GTPase superclass [87]. Whereas other P-loop NTPases rely on a conserved glutamate residue to serve as a general base in abstracting a proton 42

43 from the catalytic water molecule for NTP hydrolysis, GTPases rely on the γ-phosphate group itself to act as the general base for abstracting a water proton [82]. This generates a nucleophilic hydroxide ion that attacks the protonated γ-phosphate group to form a pentacovalent reaction intermediate and subsequently causes the release of the γ-phosphate. The GTPase superclass is further subdivided into two classes. The first of these is the TRAFAC class (TRAnslation FActor-related Class), and includes a majority of the well known GTPases, such as septins, elongation factors, the all important Ras superfamily, and heterotrimeric G proteins [82]. This class is characterized at the sequence level by a highly conserved threonine or serine residue in the G2 motif that makes a hydrogen bond to the Mg 2+ cation required for GTP hydrolysis and by a conserved serine residue in the G5 motif that interacts with the guanine ring [87]. The distinctive structural feature of this class is the six-stranded β-sheet containing one antiparallel β-strand flanking the Walker B motif. The second class of GTPases is the SIMIBI class, named after its three largest subgroups (signal recognition particle, MinD, and BioD). At the sequence level, this class is defined by a specific variation of the Walker A motif that includes an additional glycine residue GxxGxGK[S/T], a conserved aspartate residue in the second β-strand between the G2 and G3 motifs, a higher variability of the Walker B motif, and a higher variability of the [N/T]KxD amino acid sequence of the G4 motif [87]. This class is structurally defined by an exclusively parallel β-sheet. Additionally, SIMIBI GTPases tend to form dimers, as opposed to TRAFAC GTPases that typically function as monomers or in large complexes. 43

44 To help regulate their biological function, GTPase superclass proteins are assisted by a number of additional proteins. Cycling between the GTP-bound on state and the GDP-bound off state is regulated at different levels to ensure that no GTP is unnecessarily consumed. In order for the GTPase to enter its active state, it needs to swap GDP for GTP. Both GTP and GDP have an affinity in the lower nano- to pico-molar range. This high affinity for the GTPase dictates a slow dissociation rate with the half-life on the order of hours. Yet in biological systems, the exchange rate of GDP for GTP is on the order ranging from minutes to seconds. Proteins known as guanine exchange factors (GEFs) accelerate this exchange rate by several orders of magnitude. All GEFs function via a similar mechanism to catalyze the dissociation of the nucleotide from the GTPase. GEFs bind GTPase proteins causing a conformation change in the switch region of the GTPase P-loop. This sterically occludes the binding of the coordinating magnesium ion, a required molecule for the high affinity binding of GDP to the GTPase. Modification of the nucleotide-binding site decreases the affinity of GDP for the GTPase and subsequently accelerates the release of GDP. Although GDP and GTP generally have similar affinities for the same GTPase, GTP is more often rebounded to the GTPase than GDP because the cellular concentration of GTP is about ten times higher than that of GDP. Even though GEFs share a similar mechanism to accelerate the exchange rate, these proteins are highly varied and structurally unrelated. Once the GTP is in its active state it can bind its effector protein and perform its biological function. The nature and duration of the interaction with the effector protein determines the biological function of the protein. In general, the binding of the GTPase and its effector protein and GTP hydrolysis are mutually exclusive events. The rate of 44

45 hydrolysis of GTP determines the lifetime of the GTP-bound on state of the GTPase. A majority of GTPase proteins are inefficient enzymes and the hydrolysis reaction is inherently slow. This reaction can be accelerated by several orders of magnitude with help from GTPase activating proteins (GAPs). Mechanistically, GTP hydrolysis is an S N 2 nucleophilic substitution. GAPs are thought to accelerate this reaction by two main mechanisms. The first mechanism suggests that GTPases contain the whole catalytic machinery. Therefore, GAPs main contributions would be to stabilize the catalytically competent conformation. The second mechanism suggests that GAPs contribute a catalytic residue to the active site of the GTPases. Both of these mechanisms accelerate the hydrolysis of GTP to GDP and the return of the active GTPase to its inactive GDPbound state. As with GEFs, GAPs are highly varied and structurally unrelated. An alternative to the GAP and GEF mediated GTPase cycle is the G-proteins activated by nucleotide-dependent dimerization (GADs) mediated GTPase cycle. Rather than relying on external proteins to regulate and facilitate their biological functions, GADs are self-activating GTPases that contain all the elements required for a nucleotideregulated switching cycle. GADs dimerize in a nucleotide dependent manner, which leads to a biologically active form, with GTP binding moderating the interaction between the G-domains. For proper GTP hydrolysis, an efficient catalytic mechanism is assembled by the dimerization of the active site of one promoter with that of another. Due to GAD proteins low affinity, they do not require GEFs for activation. Unlike other GTPases, the interaction of GADs with the effector proteins and GTP hydrolysis are not mutually exclusive. Therefore, GADs coordinate their biological function with the 45

46 GTPase reaction. GTP hydrolysis to GDP leads to the dissociation of the dimer back to its inactive monomeric state. Crystal structure of folliculin C-terminal In order to investigate the molecular structure of FLCN, our laboratory collaborated with Dr. Ravi K. Nookala s laboratory at the University of Cambridge, UK. Their laboratory performed all the experiments in determining the crystal structure of FLCN carboxy-terminal (CT). The data and information that follows is taken from a manuscript in preparation. A bioinformatics analysis of the FLCN secondary structure was performed using the JPred and Disopred software. Aligning various homologous sequences of FLCN in ClustalW showed the conservation of the carboxy terminus in higher vertebrates. These analyses revealed two main structural domains for FLCN. The first was a globular amino terminus spanning from amino acid 89 to 290. The second was a globular carboxy terminus spanning from amino acid 340 to 579. A central region with little secondary structure linked the two globular domains. Although many attempts were made, the intrinsic instability of the full-length folliculin protein made it difficult to obtain a crystal structure for the entire protein. Instead, only the carboxy terminal region was crystallized. Considering that majority BHD locus mutations lead to early protein truncation and the subsequent loss of the carboxy terminus, determining its structure could provide valuable insight towards determining a potential functional role for FLCN carboxy terminal. The crystal structure of FLCN CT was determined by collecting multiplewavelength anomalous diffraction data from selenomethionine crystals by exploiting the 46

47 anomalous signal of the incorporated selenium. The three-dimensional structure of the apo-form of FLCN CT was resolved to a resolution of 1.9 Å (Figure 2). The overall structure of FLCN compromises an αβ fold with 6 α helices stacked on one side of the β- sheet followed by an additional α helical region. The β-sheet consisted of five β-strands organized as E-D-A-C-B. The structure was then searched for homology with other known proteins. Using DALI servers and NCBI s VAST search program, a database search returned a number of hits corresponding to several nucleoside triphosphatases (NTPases) with similar spatial arrangements of secondary structure. This information inspired the decision to perform UV mediated cross-linking experiments to determine if FLCN was capable of binding any nucleotides. This assay was performed to determine the direct interaction of NTPs with our protein of interest. The process consisted of incubating the purified FLCN protein with radioactive α- 32 P labeled nucleotide triphosphates under UV light. The data collected showed the binding of both ATP and GTP to the purified recombinant FLCN CT with a significantly increased affinity for GTP. Following this observation, FLCN CT was co-crystalized with a GTP nucleotide by mixing 5 mm GTP and 10mM MgCl 2 with a 5 mg/ml FLCN CT protein mixture. The structure of FLCN CT and the nucleotide complex was determined to 2.7 Å resolution via the molecular replacement technique, using the FLCN CT promoter from the apo-form crystals as the molecular replacement probe. The crystal structure revealed that FLCN CT forms a dimer with the nucleotide bound between the short helices of both promoters, forming a tight pocket to stabilize the nucleotide. Specifically, it was observed that lysine residue 485 (K485) from both promoters forms a hydrogen bond with the N7 of the 47

48 guanine ring. Interestingly, although the GTP nucleotide was incubated with the FLCN CT for co-crystallization, only GDP fit the electron density map. Figure 2: FLCN CT crystal structure. A cartoon representation of the crystal structure of FLCN CT determined to a resolution of 1.9 Å. The protein chain is rainbow colored from blue at the N-terminal to red at the C-terminal. The front view shows the five β-strands (labeled A-E) organized as a β-sheet with E-D-A-C-B arrangement. The side view shows the 10 alpha helices (labeled H1-H10) stacked to the side of the protein. 48

49 To investigate whether FLCN CT shared structural similarities with proteins of the GTPase superclass, the obtained 3D structure was compared with the 3D structure of the human ras protein and with the 3D structure of Thermus aquaticus signal sequence recognition Ffh protein, both prototypical GTPases of the TAFAC and SIMIBI subclasses respectively. The FLCN CT and ras protein showed similar β-strand arrangement (both E-D-A-C-B) but differed in the orientation of the E and B β-strands. While the B β-strand in ras and other TRAFAC GTPases is anti-parallel to the rest of the β-sheet and the E β- strand is parallel to the β-sheet, the E β-strand in FLCN CT is anti-parallel to the β-sheet and the B β-strand is parallel to the β-sheet. Structural comparisons with the Ffh protein showed fewer similarities with FLCN CT. Although the orientation of the B β-strand in both Ffh and FLCN CT were similar, Ffh contained 7 parallel β-strands with a different strand arrangement (E-A-D-B-C). Interestingly, though the full length FLCN did not contain the Walker A or Walker B motifs that are usually characteristic of P-loop GTPases, FLCN CT was shown to preferentially bind GTP. In all P-loop GTPases, the G4 [N/T]KxD motif interacted with the guanine ring and confers specificity over other nucleotides [82, 87]. The FLCN CT contained a putative G4 motif, NKIE, and the cocrystal structure revealed the dimerization of FLCN CT and the interaction of the conserved lysine residue from both promoters with the guanine ring. This mechanism of nucleotide dependent dimerization, wherein two identical active site residues present on the G4 motif from both proteins interact with the nucleotide, has been observed in the previously mentioned GADs and is specifically exemplified by two SIMIBI GTPases, Ffh and FtsY[89]. Additionally, the putative G4 motif forms a small helix as opposed to conventional GTPases where the motif exists as a loop. 49

50 This data provides the first structural and biochemical evidence that FLCN CT harbors a GTP binding site. It suggests that FLCN CT shares structural and sequence similarities with P-loop GTPases from both the TRAFAC and SIMIBI subfamilies, yet also suggests that FLCN doesn t belong to either of these two classes. It harbors an unusual GTP fold and undergoes dimerization upon GTP binding, similar to GADs. Taken together, these data suggest that FLCN may be a unique GTPase. Results and Discussion Although structural analysis and classification may help predict a protein s function, it does not allow for any conclusion to be drawn about said function. In order to verify whether FLCN possesses or does not posses GTPase activity, immune-precipitated FLAG tagged FLCN from HEK 293T cells was subjected to GTPase and ATPase assays. Although no ATP hydrolysis was observed, FLCN showed a dose response relationship with GTP hydrolysis. Following this observation, the GTPase assay was repeated with a FLCN mutant at tyrosine residue 463 found in BHD patients that led to a truncated protein [90], a FLCN mutant replacing the K485 residue involved in the guanine ring binding with a hydrophobic alanine, and with the recombinant FLCN CT. Both the patient mutant and the lysine substitution mutant showed no GTPase activity, while the recombinant FLCN CT showed weak GTPase activity compared to the full length protein. These data suggest that FLCN possesses a carboxy-terminal GTPase activity (potentially explaining the observation of GDP in the co-crystallization of GTP and FLCN CT) that is lost in a naturally occurring truncation mutant and is dependent on the full-length protein. 50

51 Abbas Ghazi, a former master s student of the Pause lab, performed these experiments before passing the project on to me with the objective of confirming these results and further characterizing the GTPase activity of FLCN and of FLCN mutants. Immuno-precipitated FLCN WT and K485A GTPase assays The first objective of the project was to repeat the experiment performed by Abbas Ghazi to confirm the previously observed results on GTPase activity. This experiment was performed using the three constructs made by Abbas Ghazi. These included FLAG tagged FLCN WT, FLCN K485A, and an empty vector construct all in pcdna3.1 plasmid. Each construct was expressed in HEK 293T cells, immunoprecipitated, and quantified by both the Bradford assay and SDS-PAGE coomassie gel staining. The resulting purified protein was incubated with radiolabelled γ - 32 P GTP for an hour at 37 degrees Celsius, separated by thin layer chromatography, and revealed by autoradiography. The results showed the complete hydrolysis of the radiolabelled GTP by the immuno-precipitated FLCN WT while immuno-precipitated FLCN K485A showed no hydrolysis of radiolabelled GTP (Figure 3A). These results confirmed what was previously observed: that FLCN possesses a C-terminal GTPase activity that is dependent on the Lysine residue found in the putative G4 NKIE motif. Immuno-precipitated FLCN mutant GTPase assays The next objective of the study was to further characterize FLCN GTPase activity. To this extent, a number of FLCN mutants were generated with the intent of subjecting them to a GTPase assay in order to assess the effect of the mutation on the GTPase activity. In total, 6 mutants were generated for the study. As previously mentioned, the 51

52 rationale behind the K485A mutant was to modify the highly conserved lysine amino acid residue, which has structurally been shown to interact with the guanine ring of GTP with a hydrophobic alanine; this should theoretically prevent the formation of the hydrogen bond. This mutant would help determine if the lysine was essential for binding GTP and necessary for GTP hydrolysis. In addition to the K485A mutant, a histidine to arginine H255R missense mutant was generated. This mutation was originally observed in German Shepard dogs with a hereditary predisposition for renal carcinoma, known as hereditary multifocal renal cystadenocarcinoma [91], and was more recently observed in human BHD patients. Another mutation tested by Abbas, but not initially repeated, was the T463X (X designates stop codon) nonsense mutant, which leads to a truncated protein. This mutation was selected since it has been reported and observed multiple times in tumours from BHD patients [18, 52, 54]. Similarly, R239C missense mutant, K508R missense mutant, and R527X nonsense mutants were generated to assess their GTPase activity. These mutations were selected because they have also been observed in the tumours of patients with the BHD syndrome [52, 54, 90]. Previous data, and repeated experiments strongly suggest that the biological function of FLCN WT is a GTPase. Additionally, BHD tumour suppressor gene appears to adhere to a Knudson-two hit model of tumourigenesis. Screening the second hit mutations identified in tumours of BHD patients for GTPase activity could link the observed biological function to the tumour suppressor function of BHD. Nonsense mutation leading to early translation termination and formation of a truncated protein could understandably modify or abolish the regular function of the protein. Although these mutants will provide valuable information, more revealing are the missense 52

53 mutations. If the missense mutations lead to a non-functional FLCN protein, this would strongly correlate FLCNs biological function with its tumour suppressing activity. Although our lab already had both a FLAG tagged FLCN K485A mutant and a FLAG tagged FLCN T463X mutant in a pcdna3.1 vector, we decided not to use these since the FLAG tag was positioned at the C-terminal for the K485A mutant and at the N- terminal for the FLCN T463X mutant. To keep every aspect of the experiment consistent, new constructs were created in the pcdna3.1 vector such that the expressed protein contained 2 FLAG tags at the N-terminal of the protein. This would guarantee the proper immuno-purification of all FLCN constructs regardless of whether the mutation encoded a simple amino acid substitution or a protein truncation. To generate these mutants, 2xFLAG FLCN WT was subcloned into the pcdna3.1 vector and used as the template for mutagenesis. Each mutant was constructed following the QuikChange II protocol and confirmed by sequence analysis. Before testing the entire spectrum of FLCN mutants, we wanted to confirm the previous results by repeating the GTPase assay with the new FLCN WT and FLCN K485A mutant constructs. Both constructs were expressed in HEK 293T cells, immunoprecipitated, and quantified before being used in the assay. However, the initial results were inconsistent. Therefore, the assay was repeated in triplicate using three independent batches of HEK 293T cell lines to express the different constructs. Interestingly, both the full-length wild type protein and the lysine substituted mutant protein displayed GTPase activity (Figure 3B). This result contradicts what was previously observed that the substitution of the lysine for the hydrophobic alanine completely abolished the GTPase activity. 53

54 Figure 3. GTPase assays with FLCN WT and FLCN K485A constructs A) Autoradiograph (40 minute exposure) of the GTPase assay showing full hydrolysis of GTP by FLCN WT whereas no hydrolysis is observable for FLCN K485A mutant. This suggests that the lysine residue K485 is essential for FLCN GTPase activity. B) Autoradiograph (3 hour exposure) of the GTPase assay with FLCN WT and FLCN K485A performed in triplicate. Each set was performed with protein expressed and harvested form independent HEK 293T cell lines. The results show that both FLCN K485A and FLCN WT have GTPase activity. 54

55 Following this observation, the GTPase assay was carried out on the entire spectrum of FLCN mutants. As with the other experiments, the constructs were expressed in HEK 293T cells, immuno-purified, and quantified before usage. In this experiment, all mutant proteins showed weak and similar activity to the wild type FLCN protein (Figure 4). Figure 4. GTPase assay with range of FLCN mutants Autoradiograph (60 minute exposure) of the GTPase assay of a range of FLCN mutants. The results show weak GTPase activity for all mutants, and comparable GTPase activity of the mutants and FLCN WT. 55

56 Together these results present an inconsistent view of FLCN GTPase activity. In the three GTPase assays presented, the level of GTP hydrolysis by immuno-precipitated FLCN WT varied considerably. This variation was also seen throughout the numerous repeat experiments performed in the lab. Although the GTPase activity was still observed, the overall trend suggests that the level of GTP hydrolysis is much lower than initially observed. Indeed, throughout all the GTPase experiments performed, none were able to replicate the total GTP hydrolysis observed in the first repeat experiment performed with the inherited FLCN WT construct. More importantly, the results obtained for the immuno-precipitated FLCN K485A mutant contradicted the previously obtained results. As previously mentioned, the cocrystal structure of FLCN with GTP shows a hydrogen bond formation between the lysine of the putative G4 motif and the guanine ring. In all other P-loop GTPases, this motif is crucial for selectively binding GTP over other nucleosides. Abolishing this interaction would theoretically abolish the GTPase activity as well. Yet, the data presented shows that the FLCN K485A mutant exhibits low levels of GTP hydrolysis, as opposed to no hydrolysis at all. In fact, the level of GTPase activity is often comparable if not identical to the GTPase activity observed for immuno-precipitated FLCN WT. This was also observed for the other immuno-precipitated FLCN mutant constructs. Overall, this data presents FLCN as protein with weak GTPase activity, regardless of the FLCN mutational status. A potential explanation for the weak GTPase activity of FLCN WT could be due to the 2 N-terminal FLAG tags. It is known that addition of a fusion tag to a protein can affect the protein s properties [92]. These tags could potentially obscure the putative 56

57 biological function of the protein by affecting protein stability or by disrupting proper post-translational modifications, which could prevent proper protein folding or inhibit the dimerization observed in crystal structure. Yet these modifications would not account for the low GTPase activity observed in the FLCN mutants. The crystal structure studies and previous GTPase assays with the FLCN CT suggest that the GTPase activity of FLCN is contained in the C-terminal of the protein. Consequently, any mutation leading to the early truncation of the protein should eliminate the GTPase activity. Yet both the R527X and T463X nonsense mutants tested for GTPase activity show weak GTP hydrolysis. This suggests that the hydrolysis mechanism may not be exclusively located at the C- terminal of the protein, but may in fact be contained towards the N-terminus of the protein. Since neither the R239C nor H255R missense mutations abolished the limited GTP hydrolysis, the theory of N-terminal GTPase activity fits all the data observed. As previously mentioned, FLCN has been shown to interact and bind various proteins. FLAG tagged immuno-precipitation of FLCN expressed in HEK 293T cells will not only pull down our tagged protein, but will also immuno-precipitate any proteins strongly interacting with FLCN in that cellular environment. This brings up a new possibility for the observed GTP hydrolysis. Rather than FLCN being functionally active as a GTPase, FLCN may be interacting with a GTPase protein that is being co-immuno-precipitated alongside FLCN. This mechanism could account for the variability of GTPase activity observed during repeat GTPase assays, but more importantly could explain why weak GTPase activity is observed in all the mutants screened. Unless the mutation effectively abolishes the interaction of FLCN and a hypothetical GTPase, the activity will continue to be observed. 57

58 GST tagged FLCN WT and FLCN K485A GTPase assay To further investigate whether the GTPase activity was a biological function of FLCN or was a result of another protein being pulled down alongside FLCN, we decided to repeat the experiment using a bacterial system for protein expression and purification. FLCN WT and FLCN K485A mutants were subcloned into a pgex-2tk vector and expressed in E. coli. This system was designed for the inducible, high-level intracellular expression of our protein of interest, FLCN, as a fusion with glutathione s-transferase (GST). Expressing and purifying the recombinant fusion protein in E. coli offers one main advantage over expressing and purifying our recombinant protein from HEK 293T cells. Since E. coli does not contain any known homologs to FLCN, we assume that any protein interactions occurring between FLCN and the human cellular environment of HEK 293T cells will not occur in the bacterial environment of E. coli. Therefore, the affinity-purified product should only contain our expressed FLCN protein and the observed GTPase activity will better assess the actual hydrolysis capacity of FLCN. By comparing the activity of the affinity purified FLCN WT expressed in E. coli to the activity of the purified FLCN WT expressed in HEK 293T cells, it is clear that the FLCN WT expressed in E. coli shows significantly less GTP hydrolysis than FLCN WT expressed in HEK 293T cells (Figure 5). In fact, the level of hydrolysis for 1 µg of GST FLCN WT was similar to both the EV control and the GST negative control. Increasing the concentration to 2 µg slightly increased the amount of hydrolysis; yet the amount of activity was still considerably inferior to the already weak GTP hydrolysis at 1 µg of FLCN WT expressed in HEK 293T cells. At both the 1 µg and 2 µg concentrations, the 58

59 FLCN K485A expressed in E. coli showed hydrolysis levels comparable to both controls. This new system suggests that FLCN WT alone has very little capacity for GTP hydrolysis, which is abolished by mutating the lysine residue of the NKIE motif. Similarly to the FLAG tagged proteins, the presence of the GST tag could interfere with protein stability and the formation of secondary structures, preventing FLCN from properly accomplishing its function. Additionally, expressing the protein in E. coli may not provide all the necessary machinery for the proper post-translational modification of FLCN, which could greatly affect the protein s function. 59

60 Figure 5. GTPase assay comparing GST purified FLCN WT and FLCN K485A to FLAG purified FLCN WT Autoradiograph (90 minute exposure) comparing the hydrolysis level of HEK 293T expressed FLCN WT with E. coli expressed FLCN WT and FLCN K485A. The results show that FLCN WT and K485A expressed and purified in a bacterial system has little to no GTPase activity compared to FLCN WT expressed and purified from HEK 293T. 60