THE FUNCTION OF CUL-4 UBIQUITIN LIGASE COMPLEXES IN DNA REPLICATION AND ENHANCERS OF CULLIN-INHIBITOR CAND-1 IN C. ELEGANS JIHYUN KIM

Transcription

1 THE FUNCTION OF CUL-4 UBIQUITIN LIGASE COMPLEXES IN DNA REPLICATION AND ENHANCERS OF CULLIN-INHIBITOR CAND-1 IN C. ELEGANS by JIHYUN KIM (Under the Direction of Edward T. Kipreos) ABSTRACT Ubiquitin-mediated proteolysis plays a central role in specific and selective degradation of cell cycle regulators. E3 ubiquitin ligases determine the specificity of targets in the pathway. CUL-4, a member of the cullin-ring ubiquitin ligase (CRL) family, has a major function in restricting DNA replication in C. elegans. Genome stability is a fundamental issue for survival and is achieved by precise duplication of DNA. The temporal separation of the assembly of pre-replicative complexes from initiation of DNA replication prevents DNA re-replication. The replication licensing factors Cdc6 and Cdt1 participate in the formation of pre-replicative complexes during G1 phase, but they are regulated in different ways. Metazoan CUL4 mediates Cdt1 degradation and vertebrate Cdc6 is exported from the nucleus during S phase. However, because residual Cdc6 remains in the nucleus during S phase, it has been unclear whether Cdc6 translocation prevents re-replication. We show that C. elegans CDC-6 translocates from the nucleus during S phase in a multiple-phosphorylation-dependent manner. CUL-4 promotes CDC-6 phosphorylation and its subsequent nuclear export via negatively regulating the CDK-inhibitor CKI-1. Coexpression of non-exportable CDC-6 with non-degradable CDT-1 induces re-replication, but re-replication is not observed upon expression of deregulated CDT-1 or CDC-6,

2 indicating that CDC-6 translocation and CDT-1 degradation redundantly prevent rereplication. CUL-4 independently controls both replication-licensing factors. CRL activity is regulated by Nedd8 conjugation that enhances CRL activity and also by CAND-1, a cullin-inhibitor. Although CAND-1 blocks assembly of CRL complexes, the absence of cand-1 leads to inactivation of CRLs, suggesting that CAND1 is required for proper CRL activity. However, the mechanism by which CAND1 promotes CRL activity is ambiguous. To understand the roles of CAND-1, the cand-1 enhancers dynein and hpk-1 (Homeodomain interacting Protein Kinase) are analyzed. Inactivation of dynein causes a reduction of CUL-2 deneddylation. However, CUL-2 is still functional and dynein(rnai) does not lead to cul-2 mutant phenotypes. Combining the cand-1(tm1683) mutant with hpk-1(rnai) causes embryonic lethality but is not obviously defective in CRL function or cell fate. Further investigation is necessary to understand the causes of embryonic lethality when both cand-1 and hpk-1 are coinactivated. INDEX WORDS: C. elegans, ubiquitin-mediated proteolysis, cell cycle, DNA replication, replication licensing, CUL-4, CDC-6, re-replication, cullin-ring ubiquitin ligase (CRL), CAND-1, Nedd8, cand-1 enhancer, hpk-1, dynein

3 THE FUNCTION OF CUL-4 UBIQUITIN LIGASE COMPLEXES IN DNA REPLICATION AND ENHANCERS OF CULLIN-INHIBITOR CAND-1 IN C. ELEGANS by JIHYUN KIM BS, Korea University, Korea, 2001 A Dissertation Submitted to the Graduate Faculty of The University of Georgia in Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY ATHENS, GEORGIA 2009

4 2009 Jihyun Kim All Rights Reserved

5 THE FUNCTION OF CUL-4 UBIQUITIN LIGASE COMPLEXES IN DNA REPLICATION AND ENHANCERS OF CULLIN-INHIBITOR CAND-1 IN C. ELEGANS by JIHYUN KIM Major Professor: Committee: Edward T. Kipreos Marcus Fechheimer Jacek Gaertig Michael McEachern Electronic Version Approved: Maureen Grasso Dean of the Graduate School The University of Georgia May 2009

6 DEDICATION This dissertation is dedicated to my parents for their unconditional love. Without their encouragement, support, love, and discipline, it would be impossible for me to make all achievements. My parents made me who I am. Their deep love and sacrifice to make me able to carry though all the ups and downs in my life. iv

7 ACKNOWLEDGEMENTS I express my deepest gratitude to Dr. Edward T. Kipreos, whose intelligence is beyond my words and whose advice and assistance have been invaluable for me through this program. Without his tremendous guidance, it would be impossible for me to complete the program. I thank my committee members Drs. Marcus Fechheimer, Jacek Gaertig, and Michael McEachern for their patience and mentorship. Special thanks to Marian Thomas and everyone in the Cellular Biology department at the University of Georgia for their patience, advice and help. I would like thank current and former members in Kipreos lab: Dimple Bosu, Chris Dowd, Hui Feng, Cass Heighington, Ji Liu, Kyoengwoo Min, Mohammad Rhaman, Fernando Santiago, Natalia Starostina, Katie Willams, and Srividya Vasudevan for excellent collaborations and for being wonderful friends. Thanks to members in the Gaertig lab for being helpful neighbors and good friends. Thanks to all my friends and loved ones for their inspiration and the happiness that they bring to my life. v

8 TABLE OF CONTENTS Page ACKNOWLEDGEMENTS...v LIST OF FIGURES...vi CHAPTER 1 GENERAL INTRODUCTION AND LITERATURE REVIEW...1 EUAKRRYOTIC CELL CYCLE...1 UBIQUITIN-MEDIATED PROTEOLYSIS...8 REGULATION OF CRL (CULLIN-RING UBIQUITIN LIGASE) C. ELEGANS CUL-4 PROMOTES CDC-6 NUCELAR EXPORT BY A CKI-1- DEPENDENT PATHWAY TO MAINTAIN GEMOME INTEGRITY...29 BACKGROUND...30 RESULTS...33 DISCUSSION...42 MATERIALS AND METHODS...46 ACKNOWLEDGEMENTS ANALYSIS OF C. ELEGANS CAND-1 ENHANCERS IDENTIFIED BY A GENOME-WIDE RNAI SCREENING...72 BACKGROUND...72 RESULTS...80 DISCUSSION...90 MATERIALS AND METHODS...92 ACKNOWLEDGEMENTS GENERAL DISCUSSION vi

9 REFERENCES vii

10 LIST OF FIGURES Page Figure 1-1: The formation of pre-replicative complexes (pre-rcs)...21 Figure 1-2: Ubiquitin-mediated proteolysis...23 Figure 1-3: Structure of multisubunit CRL complexes...25 Figure 1-4: Activated and Inactivated forms of CRL complexes...27 Figure 2-1: CDC-6 is exported from the nucleus during S phase in a CUL-4-dependent manner...52 Figure 2-2: CDC-6::GFP localization throughout the cell cycle...54 Figure 2-3: Nuclear localization of a CDC-6 mutant with intact CDK phosphorylation sites56 Figure 2-4: CDC-6 is phosphorylated on CDK sites in a CUL-4-dependent manner during S phase...58 Figure 2-5: The subcellular localization of the CDC-6mCDK::GFP mutant during G1 and S/G2 phase...60 Figure 2-6: The subcellular localization of CDC-6 CDK-phosphorylation site mutants...62 Figure 2-7: CDT-1 perdurance does not affect CDC-6 nuclear export...64 Figure 2-8: cki-1 RNAi rescues the failure of CDC-6 nuclear export in cul-4 mutant cells..66 Figure 2-9: Expression of deregulated CDT-1 and CDC-6 causes DNA re-replication...68 Figure 2-10: Model for the regulation of the CDT-1 and CDC-6 replication licensing factors by the CUL-4 ubiquitin ligase in C. elegans...70 Figure 3-1: hpk-1 and dynein do not affect CAND-1 proteins but dynein affects CUL-2 proteins...95 Figure 3-2: dynein does not regulate CUL-2 or CAND-1 subcellular localization...98 Figure 3-3: CUL-2 or CUL-4 substrates do not accumulate in dynein loss of function Figure 3-4: HPK-1::GFP expression viii

11 Figure 3-5: cand-1(tm1683) mutants with hpk-1(rnai) have hypodermis, gut, and pharyngeal cells Figure 3-6: CAND-1 and HPK-1 do not regulate UNC-37 in C. elegans ix

12 CHAPTER 1 GENERAL INTRODUCTION and LITERAURE REVIEW 1 1 Control of the Cdc6 replication licensing factor in metazoa: the role of nuclear export and the CUL4 ubiquitin ligase. Kim J, Kipreos ET. Cell Cycle Jan 15;7(2): Epub 2007 Nov 3. Reprinted here with permission of Cell Cycle. 1

13 EUKARYOTIC CELL CYCLE The eukaryotic cell cycle is composed of Gap phase 1 (G1), DNA synthesis (S), Gap phase 2 (G2), and Mitosis (M). In G1 phase, cells grow and prepare for entry into S phase. Cells receive information from the extracellular environment and determine whether to proliferate or to adopt an alternate fate. During S phase, DNA synthesis occurs. Cells carefully restrict the duplication of their genomic DNA, so that it occurs only once during each cell cycle. G2 phase is an interval between S and M to prepare for M phase. During M phase, dramatic cellular changes occur and two genetically identical daughter cells are generated. Non-proliferative cells exit the cell cycle from G1 phase and enter a state of quiescence called G0 phase. Mitosis itself is composed of prophase, prometaphase, metaphase, anaphase, telophase and cytokinesis [1]. In prophase, chromosomes are condensed and visible. In the end of prophase and the beginning of prometaphase, the nuclear envelope is broken down. During prometaphase, chromosomes become attached to and positioned on the mitotic spindle. In metaphase, all chromosomes align near the spindle equator to form the metaphase plate. At anaphase, the two sister chromatids that comprise each replicated chromosome separate and move towards the opposite spindle poles. During telophase, each set of sister chromatids begins to decondense upon reaching opposite spindle poles, the nuclear envelope re-forms, a microtubule-based midbody assembles near the original spindle equator, and the mitotic spindle disassembles. During anaphase and telophase, cells begin to divide. This process is called cytokinesis and it distributes a copy of each chromosome to each genetically identical new daughter cell. Control of the eukaryotic cell cycle Cyclin-Dependent Kinase (CDK) 2

14 CDKs are serine and threonine kinases, and their actions are dependent on association with their activating subunits, cyclins. CDK-cyclin complexes play an essential role in controlling progression through the eukaryotic cell cycle [2]. The activity of CDK-cyclin complexes oscillates throughout the cell cycle. CDK activity is low in G1 phase but increase during S and G2 phases. High levels of CDK activity subsequently drive cells into mitosis. Afterward, CDK activity decreases substantially to allow cells to exit mitosis, thereby enabling cells to begin another round of the cell cycle. CDKs are required for critical cell cycle events: the G1-to-S transition, the initiation of DNA replication, the G2-to-M transition, and the initiation of multiple mitotic events. In S. cerevisiae and S. pombe, a single CDK (Cdc28 and Cdc2, respectively) catalyzes all major cell cycle transitions [3]. In higher eukaryotes, a number of CDKs regulate cell cycle progression. CDK4, CDK6, and CDK3 regulate G1 phase progression and entry into S phase. CDK2 is required for entry into S phase. CDK1 (CDC2) controls mitosis [3]. In mammalian cells, CDK4 or CDK6 activated by D-type cyclins (D1, D2, and D3) in combination with CDK2 activated by E-type cyclins (E1 and E2) together promote the G1-to-S transition. In G1 phase, the phosphorylation of retinoblastoma tumor suppressor protein (prb) by CDK4/6-cyclin D reverses the repressive effects of prb on E2F family transcription factors which allows for the transcription of a number of genes required for exit from G1 and initiation of S phase [4, 5]. CDK2-cyclin A/E are required for progression through S phase to prevent further pre-replicative complex (pre-rc) assembly or loading, thereby ensuring that cells replicate their genomic DNA only once per cell cycle. CDK1-cyclin A/B plays a major role in several mitotic events: regulating centrosome separation, nuclear envelope breakdown, chromosomal condensation, assembly of the mitotic spindle, and cytokinesis [5]. For example, CDK1 phosphorylates 3

15 condensin components that are essential for chromosome assembly and segregation, and this phosphorylation promotes chromosome condensation [6]. Regulation of CDK activity Proper oscillation of CDK activity is essential to promote cell cycle progression. CDK activity is controlled at multiple levels, such as by phosphorylation and dephosphorylation, and by the inhibitory binding of CDK inhibitors (CKIs); however, the activity of CDKs is primarily regulated by their interactions with activating cyclins. The abundance of cyclins is strictly regulated during the cell cycle both by transcriptional control and by ubiquitin-mediated protein degradation that is dependent on the cell cycle [7, 8]. Phosphorylation and dephosphorylation of CDKs regulate CDK activity. Mitotic CDK1 is predominantly controlled by phosphorylation and dephosphorylation [9]. Phosphorylation of CDK1 by Myt1 and Wee1 kinases at Thr14 and Tyr15 inhibits its activity. Phosphorylation of CDK1 mediated by a CDK-activating kinase (CAK) at Thr167 in S. pombe (Thr169 in S. cerevisiae) and its dephosphorylation by Cdc25 phosphatase at Thr14 and Tyr15 stimulate its activity [9]. Binding of CKIs is one of the major regulatory mechanisms to restrict CDK activity. Metazoans have two families of CKIs: the INK4 gene family and the CIP/KIP family. p16 INK4a, p15 INK4c, p18 INK4c, and p19 INK4d belong to INK4 gene family [10, 11]. Members of the INK4 protein family interact with CDK4 and CDK6. The INK4 proteins bind exclusively to and form tight binary complexes with either CDK4 or CDK6 but not with the CDKs complexed to cyclin D in vivo. INK4 proteins bind adjacent to the ATPbinding site of the catalytic cleft and opposite to the cyclin-binding site in CDK4/6, inducing a conformational change that distorts the catalytic cleft and allosterically alters the cyclin binding site [12-14]. INK4 proteins inhibit CDK4/6 activity both by interfering 4

16 with ATP binding and by interfering with their binding to D-type cyclins. p21 Cip/Waf1/Sdi1, p27 Kip1, and p57 Kip2 belong to the CIP/KIP family of CKIs. In contrast to the INK4 family, the CIP/KIP family binds to both CDK and cyclin subunits and modulates the activities of CDK-cyclin D/E/A/B complexes [15, 16]. CIP/KIP proteins form ternary complexes with CDK1 cyclin B, CDK2 cyclin A, CDK2 cyclin Es, CDK4 cyclin Ds, and CDK6 cyclin Ds. For example, p27 inhibits cyclina CDK2 efficiently: not only does it rearrange and destabilize the catalytic cleft, it mimics ATP to fill up the catalytic cleft, and it also binds to a peptide-binding groove in cyclin A. As a result, the binding of substrates is blocked [17]. The inhibitory potential of CKIs is dependent on cellular context and regulated via phosphorylations and protein-protein interactions. Protein degradation during the cell cycle Besides oscillation of CDK activity, destruction of key cell-cycle regulators plays a major role in controlling cell cycle progression [8]. Critical regulators in cell cycle progression are selectively degraded exclusively through ubiquitin-mediated proteolysis (see below). Control of DNA replication Genome stability is essential for cell survival. Faithful duplication of the genome is of fundamental importance for ensuring this stability. Over-replication or underreplication will quickly destabilize the genome. To prevent replicated DNA from becoming re-licensed and re-replicated within a single S phase, replication origins temporally alternate between two distinct states: prereplicative in mitosis or early G1 phase and postreplicative in S phase [18]. The first step is licensing replication origins in mitosis or early G1 phase. Formation of pre-replicative Complex (pre-rc) involves the ordered assembly of the 5

17 six-member Origin Recognition Complex (ORC), the replication licensing factors Cdt1 (Chromatin licensing and DNA replication factor 1) and Cdc6 (Cell Division Cycle 6), and the DNA helicase MCM2-7 (MiniChromosome Maintenance 2-7) complex. ORC binds to replication origins and then recruits the replication licensing factors Cdt1 and Cdc6 that are required to load MCM2-7 complexes onto replication origins for initiation of DNA replication (Fig 1-1). This process is called replication licensing ; the formation of pre- RCs allows the MCM2-7 complex, which functions as a helicase that unwinds DNA ahead of replication fork, to load onto replication origins prior to S phase. Regulation of replication licensing refers to mechanisms that restrict DNA replication to occur only once each cell cycle [18-20]. These mechanisms stop the loading of new MCM2-7 complexes onto replication origins before entry into S phase. This can be achieved by down-regulating the activity of the ORC-Cdc6-Cdt1 loading machinery without affecting MCM2-7 helicase activity. The next step is the initiation of DNA replication and the inactivation of replication origins in S phase. Those events are triggered by increasing S-phase CDK activity, which occurs at the G1-to-S transition. Increased CDK activity prevents new MCM2-7 complexes from loading onto replication origins. In addition, this leads to the conversion of each pre-rc into a pre-ic (pre-initiation Complex). Activation of MCM2-7 helicase activity is a key event here. The MCM2-7 complex at replication origins is activated during initiation of DNA replication by the action of two kinases: S phase CDKs and Dbf4-Cdc7 kinase referred to as DDK (Dbf4-dependent kinase) [21, 22]. Those kinases promote loading of Cdc45 (Cell Division Cycle 45) and the Go-Ichi-Ni-San, or , complex of Sld5-Psf1-Psf2-Psf3 (GINS) that together convert a MCM2-7 complex from an assembly factor into a replicative helicase [23, 24]. A tight complex of MCM2-7 and Cdc45 unwinds DNA during replication initiation and elongation, and progresses DNA replication forks [24-27]. Once replication origins are unwound, DNA polymerases α and 6

18 δ, proliferating cell nuclear antigen (PCNA), and replication factor C (RFC) are recruited to synthesize DNA [18]. CDKs play dual roles in regulating DNA replication, being required not only to trigger replication initiation but also to limit DNA replication to a single round of duplication per cell cycle. The replication licensing factor Cdc6 Cdc6 was first identified in the screen for S. cerevisiae mutants with changes in the cell division cycle [28]. Cdc6 plays a critical role in initiating DNA replication, not elongating DNA [29, 30]. It is a member of the AAA+ ATPase family and is related to ORC1 [31]. Cdc6 has a Walker A motif which is involved in ATP binding and a Walker B motif which is the ATP hydrolysis domain. These motifs are important for Cdc6 function: the recruitment of MCM2-7 complexes [32, 33]. After Cdc6 associates with ORC, a MCM2-7 complex is escorted to replication origins via Cdt1. ATP hydrolysis by Cdc6 causes a conformational change in MCM2-7, which induces tight interaction of MCM2-7 with DNA and displacement of Cdt1. Finally, ATP hydrolysis by ORC launches a new round of Mcm2-7 loading [34]. The prevention of DNA re-replication in Yeast and Metazoans The mechanisms that prevent excessive DNA replication focus on preventing replication origins from reusing pre-rc components within the same cell cycle. In S. pombe, the degradation of the Cdc6 homolog Cdc18 and Cdt1 ensures that DNA re-replication does not occur during S phase [35]. Phosphorylation of Cdc18 by CDC2 during S phase leads to its ubiquitin-mediated degradation via the SCF Pop1/Pop2 ubiquitin ligase complex [36-38]. Cdt1 is degraded during S phase via the CRL4- DDB1 CDT2 ubiquitin ligase complex [39]. Overexpression of Cdc18 is sufficient to induce extensive DNA re-replication [30]. In contrast to Cdc18, overexpression of Cdt1 by itself 7

19 does not cause DNA re-replication [35, 40]. However, co-overexpressing Cdt1 and Cdc18 enhances the DNA re-replication phenotype relative to overexpression of Cdc18 alone, indicating that Cdt1 increases the efficiency of replication licensing [35, 40]. In S. cerevisiae, four mechanisms are known to block DNA re-replication: the phosphorylation of ORC2 and ORC6, the degradation of Cdc6, the nuclear export of Cdt1, and the nuclear export of MCM2-7 [41-44]. The deregulation of any one pre-rc component is not sufficient to induce DNA re-replication [42-45]. The simultaneous deregulation of two components induces limited origin re-firing in a subset of replication origins; however, deregulation of ORC, Cdc6, and MCM2-7 together induces more substantial DNA re-replication than deregulation of any combination of two components alone [42, 44, 45]. The simultaneous deregulation of Cdt1 with the other three components has not been reported. In vertebrates, the degradation of ORC1, the nuclear export of Cdc6, and the inactivation of Cdt1 by degradation or through binding to the Cdt1-inhibitor geminin have been implicated in preventing DNA re-replication. The degradation of ORC1 may not be a consistent regulatory feature, as it is observed in some cells but not in others [46-48]. The importance of Cdc6 nuclear export has been controversial due to residual nuclearlocalized Cdc6 during S phase [49-54]. My research as recorded in this dissertation clarifies the significance of CDC6 nuclear export during S phase in preventing DNA rereplication. Overexpression of a non-degradable Cdt1 produces limited DNA rereplication (significantly less than in S. pombe) [55, 56]. Overexpression of both Cdt1 and Cdc6 to levels several orders of magnitude higher than the endogenous proteins induces limited DNA re-replication in human cells deficient in the p53 checkpoint pathway [57]. Additionally, the inactivation of the metazoan-specific Cdt1-inhibitor geminin produces limited DNA re-replication in Xenopus and human cells [58-60]. It 8

20 appears that the regulation of coordinated replication licensing factors rather than a dominant single inhibitory mechanism restricts DNA replication in vertebrates. UBIQUITIN-MEDIATED PROTEOLYSIS Selective proteolysis by the ubiquitin-proteasome system allows specific and efficient removal of key regulators to control diverse cellular pathways, including the cell cycle, signal transduction, and transcription [61]. Ubiquitin-mediated proteolysis requires the cascade action of three enzymes (E1 ubiquitin-activating enzyme, E2 ubiquitinconjugating enzyme, and E3 ubiquitin ligases) to transfer mono- or multiple-ubiquitins to specific target proteins (Fig 1-2). Substrates that are polyubiquitinated in the correct conformation are recognized and degraded by the 26S proteasome [62]. Ubiquitin The conserved 76-amino acid protein ubiquitin was initially discovered as the heat-stable polypeptide of the ATP-dependent proteolytic system [63]. Ubiquitin is an extremely stable protein because ubiquitin folds tightly in β-grasp fold, wherein four β- sheets pack around a central α-helix. Later studies found that ubiquitin is a distinct signal in proteasomal proteolysis [64]. The C-terminus of ubiquitin becomes isopeptidelinked to the ε-amino group of lysine within target proteins. Ubiquitin itself is then ubiquitylated on a lysine residue (K48, K63, K11 or K29), leading to the assembly of ubiquitin chains. Polyubiquitin (the K48 linkage) is usually recognized by the 26S proteasome [65]. K63-linked ubiquitin chains are involved in non-proteolytic signaling such as endocytosis and signal transduction through nuclear factor-κb [66]. 9

21 E1 ubiquitin-activating enzyme E1 enzymes activate the C-terminal Gly residue of ubiquitin in an ATP-dependent manner [67]. In this step, an intermediate ubiquitin adenylate is formed with the release of PPi and ubiquitin then forms a thiolester linkage to a Cys residue of E1 with the release of AMP [62]. The E1-ubiquitn transfers the ubiquitin moiety to an E2 ubiquitinconjugating enzyme. It was previously believed that all organisms have only a single dedicated E1 ubiquitin-activating enzyme for ubiquitin. Insects, worms, fungi, and plants, for example, have only a single E1 enzyme UBE1 (UBA1 in yeasts and C. elegans). However, a recent study reported that vertebrates and the echinoderm sea urchin have additional E1 enzyme, UBA6 [68]. E2 ubiquitin-conjugating enzyme There is a large family of E2s dedicated to ubiquitin, comprising 11 E2 enzymes in S. cerevisiae, 22 in C. elegans, and many more in higher organisms [64, 69]. Some E2s (for example, Ubc4) can act with more than one E3 enzyme, and some E3s can act with several E2s. All E2s share a conserved globular domain of ~150 residues [70, 71]. The E2 active site cysteine, which is positioned within a highly conserved sequence, sits in a shallow cleft on the protein surface. E2 enzymes accept ubiquitin from E1 through thioester bond exchange, resulting in a charged E2~Ubiquitin that can transfer the donor ubiquitin to a lysine residue of a target protein, forming an isopeptide bond between the C-terminus of ubiquitin and the lysine residue of a target protein. Most E2 enzymes contact their cognate E3 ubiquitin ligases through side chains at the C-terminal end of E2 helix 1, the loop connecting β-strands 1 and 2, and the distal end of the active site loop. 10

22 E3 ubiquitin ligase E3 ubiquitin ligases play key roles in determining the selectivity of ubiquitinmediated protein degradation. E3 ubiquitin ligases provide specificity for substrate selection [72]. Different targets are recognized by unique E3 ubiquitin ligase complexes. In contrast to E1 and E2 enzymes, each organism has a huge number of E3 ubiquitin ligases or E3 components that allow the specific recognition of a diverse range of targets [66]. For example, in mammals several percentages of the whole genome encode E3s or E3 complex components [73]. E3 ubiquitin ligases recruit an E2 enzyme and a substrate together, and then promote conjugation by bringing a charged E2~Ubiquitin and a substrate into close proximity. E3 ubiquitin ligases can be classified into three families: HECT (Homologous to E6AP Carboxy Terminus), RING (Really Interesting New Gene), and U-box (UFD2 homology) proteins [64]. The HECT domain of E3 ligases is directly involved in ubiquitin transfer by forming covalent ubiquitin-thioester conjugates. The RING domain of RING E3 ligases binds to the charged E2 enzyme, which then transfers the activated ubiquitin to substrates. The RING finger motif-containing E3 ligases can be further classified into two groups: single-subunit RING E3 ligases and multisubunit RING E3 ligases [66, 74]. Single-subunit RING E3 ligases contain both a RING finger domain and the substraterecognition site in the same molecule [74]. This class of E3 ligases can bind E2 enzymes to catalyze ubiquitylation on the proteins with which they associate, or on themselves. Most of RING finger motif-containing E3 ligases are composed of multiple subunits. The largest known class of RING finger motif-containing ubiquitin ligases is that of the Cullin-based E3 ligases. E3 ubiquitin ligases, which have an E2-binding domain called the U-box, are a relatively small family of E3 ligases. The U-box is structurally similar to the RING finger motif; however, Ufd2 (yeast U-box containing E3 11

23 ligase) does not have its own substrate and instead promotes the polyubiquitylation of substrates of other E3 ubiquitin ligases, so that some researchers classify it as an E4 ubiquitin ligase [75]. Cullin-RING ubiquitin ligases In cullin-ring ubiquitin ligases (CRLs), which are the largest superfamily of E3 ubiquitin ligases, multiple subunits are assembled on a cullin scaffold that acts as a rigid backbone. Cullins are characterized by a conserved 150 amino acid cullin domain, which mediates interaction with Rbx1 (RING-box protein 1) / Roc1 (regulator of cullins 1) / Hrt1 (Hairy/enhancer-of-split related with YRPW motif protein 1) [61]. The APC2 subunit of the anaphase promoting complex/cyclosome (APC/C) has a cullin-homology domain [76, 77] but the APC/C is clearly distinct from other CRLs in its structure and regulation [61]. Composition and structure of cullin-ring ubiquitin ligases Cullin homologs have been found across eukaryotic phyla; for example, there are seven known members of the cullin gene family in H. sapiens, six in C. elegans, three in S. cerevisiae, and three in S. pombe [78, 79]. They have a conserved family of E3 components and share a common set of general principles [61]. The metazoan CUL1-based CRL complex called SCF (Skp1, CUL1/Cdc53, F- box protein) is well studied. The crystal structure of the human SCF Skp2 complex shows that the cullin functions as a rigid backbone for the assembly of the complex (Fig 1-3A) [80]. The CUL1 C-terminus binds to Rbx1 (a RING finger protein) that facilitates the recruitment of an E2 enzyme to an E3 complex. The CUL1 N-terminus interacts with the adaptor Skp1 that associates with the F-box protein Skp2 through the F-box motif. Skp2 is one of the F-box proteins, which are Substrate Recognition Subunits (SRSs) that bind 12

24 to and position substrates for ubiquitylation by an E2 enzyme. Different F-box proteins bind to the same core SCF components, and the combination of distinct SRSs creates unique complexes that bind distinct sets of substrates [80]. Their specificity for targets is determined by the SRS in each CRL. CUL2-based CRL complexes are structurally similar to that of the SCF complex (Fig 1-3B). The C-terminus of CUL2 binds to Rbx1 and the CUL2 N-terminus interacts with the Skp1-related adaptor protein, elongin C, which binds the complex in combination with elongin B, which contains a ubiquitin-like domain [81, 82]. A VHL-box motif on SRSs mediates interaction between the SRS and elongin C [83]. CUL5-based CRL complexes have a very similar structure to CRL2 complexes in the way that elongin C functions as the adaptor protein (Fig 1-3C). SRSs of CRL5 have a conserved motif that mediates interaction with elongin C. This motif is known as a SOCS-box domain which is related to but distinct from the VHL-box motif [83]. Although CUL2 and CUL5 share the same adaptor protein, they have different classes of SRSs. CRL5 complexes uniquely utilize the RING finger protein Rbx2/Roc2 rather than Rbx1 [84]. CUL3-based CRL complexes have a slightly different structure from that of other CRL complexes in the way that a single BTB/POZ domain protein functions as both the substrate recognition subunit and adaptor [85] (Fig 1-3D). Mammals have hundreds of BTB proteins, suggesting the potential for a large number of distinct CRL3 complexes. CUL4-based CRL complexes contain Rbx1 in the C-terminus and the adaptor DDB1 (Damaged DNA-Binding protein 1) in the N-terminus (Fig 1-3E). DDB1 has three WD-repeat-based β-propeller motifs. DDB1 interacts with SRSs through a subclass of WD repeats denoted WDxR or DxR on SRSs [86, 87]. 13

25 Function of CRL complexes in C. elegans Members of the cullin family and aspects of cell cycle regulatory mechanisms mediated by ubiquitin-mediated proteolysis are well conserved throughout yeasts and metazoans. In my research documented here, the nematode C. elegans is used as a model system. C. elegans is a multicellular organism which has many advantages for use in genetic analysis. For example, its genes can be studied in a developmental context, its gene functions are highly conserved with mammals, and it has a relatively simple and rapid development [88]. Molecular and genetic approaches for studies in C. elegans have been well developed such as robust-performing RNA-mediated interference (RNAi) to study loss of gene functions [89]. The phenotypic analysis of cullins in C. elegans provides important insights into CRL functions. The invariant pattern of cell divisions, cell positions, and cell fates of C. elegans has facilitated this analysis [90]. The phenotypes of CUL-1 ~ CUL-4 have been characterized in C. elegans. Inactivation of cul-5 or cul-6 by RNAi does not produce any phenotypes, suggesting that CUL-5 and CUL-6 may be dispensable in C. elegans [91, 92]. In this section, the functions of each cullin in C. elegans are described. CUL-1-based CRL complexes The functions of CUL1-based CRL (SCF) complexes were characterized initially, before any other CRLs had been studied. SCF is required for cell cycle exit and distal tip cell (DTC) migration in C. elegans [78]. In cul-1 mutants, cells undergo extra rounds of cell division before they exit the cell cycle, resulting in hyperplasia in all tissues. The inactivation of F-box proteins (lin-23 and skp-2) or Skp1 homologs (skr-1 and skr-2) by RNAi shows shared cul-1 phenotypes [78, 92]. Biochemical studies show that CUL-1 interacts with Skp1 homologs [91, 92]. These studies suggest that they form SCF complexes similar to those in yeast and other metazoans. 14

26 SCF complexes have non-cell cycle related functions as well. For examples, SCF SEL-10 plays roles in the Notch pathway and vulva development by targeting LIN- 12/Notch and SEL-12/Presenilin for degradation, and in sex-determination by mediating the turnover of the male-promoting proteins FEM-1 and FEM-3 [93-95]. CUL-2-based CRL complexes CUL-2 regulates several key steps in cell division and embryonic development. The absence of cul-2 causes severe defects in meiotic and mitotic cell cycles. CUL-2 is required for the G1-to-S phase transition by negatively regulating the CDK inhibitor CKI- 1 to allow cell-cycle progression from G1 to S phase [96]. CUL-2 is essential for other cell cycle processes: the Meiosis II metaphase-to-anaphase transition, establishment of anterior-posterior polarity, cyclin B degradation [97, 98], chromosome condensation, and mitotic progression. CRL2 complexes have other cellular process functions rather than cell cycle regulation: proper cytoplasmic organization [96], and degradation of CCCH proteins [99]. CUL-2 is also required for sex determination [100]. CRL2 FEM-1 promotes male development by ubiquitin-mediated degradation of TRA-1, a DNA-binding Zinc finger protein, which promotes hermaphrodite development by repressing the expression of male-specific genes in hermaphrodites [ ]. CUL-3-based CRL complexes CRL-3 complexes regulate microtubule dynamics and spindle function at the meiosis-to-mitosis transition. CUL-3 associates with MEL-26 (Maternal Effect Lethal 26) as a SRS; and CRL-3 MEL-26 targets MEI-1/katanin (microtubule-severing protein defective meiosis 1) for degradation [ ]. Accumulation of MEI-1 by cul-3 RNAi treatment causes failure in mitotic spindle formation and in chromosome segregation [106, 107]. 15

27 CUL-4-based CRL complexes The first study of CRL4 complexes was conducted in C. elegans. Since the discovery of CUL-4 functions in C. elegans, several studies in H. sapiens, Xenopus, and S. pombe have showed that CUL4 indeed forms CRL4 complexes with DDB1 and CDT2 to target CDT1, and the CDK inhibitor CKI-1 and p21 CIP1 for degradation and shares the conserved functions of CRL4 complexes throughout the evolution of yeast and metazoans [39, 79, 86, ]. As with other cullin paralogs, CUL-4 also regulates important cell cycle processes and development in C. elegans. CUL-4/DDB-1 CDT-2 is required to maintain genome stability by restraining the activity of the DNA replication-licensing factor CDT-1 [116]. Upon depletion of cul-4 with RNAi, animals arrest at the L2 larval stage, and mitotically dividing cells arrest at S phase while undergoing severe DNA re-replication. The enlarged mitotically dividing cells can have up to 50 times more DNA content, relative to normal cells which have a 2C DNA content. One of the reasons for the severe DNA re-replication phenotype in cul-4 mutants is due to the accumulation of the replication-licensing factor CDT-1 which normally undergoes degradation at the beginning of S phase to prevent extra rounds of DNA replication. A further study shows that CRL4 CDT-2 directly targets CDT-1 for ubiquitin-mediated degradation [117]. My research project documented in this dissertation explains a second pathway whose inactivation in cul-4 mutant contributes to the DNA re-replication phenotype. REGULATION OF CULLIN-RING UBIQUITIN LIGASES (CRL) All CRL complexes transit through similar stages of assembly and disassembly, resulting in switching between an active form and inactive forms. All CRL complexes are subject to global regulatory mechanisms that control their activity: for example, the covalent attachment of Nedd8 (neural-precursor-cell-expressed developmentally down- 16

28 regulated protein 8) to cullins to activate them; and removal of Nedd 8 by CSN (COP9 Signalosome) and inhibitory binding by CAND-1 (Cullin-Associated and Neddylated- Dissociated) to inactivate cullins (Fig 1-4) [118]. Neddylation Nedd8 (Rub1 in S. cerevisiae) is a post-translational protein modification that is most closely related to ubiquitin; it is a ubiquitin-like protein, sharing 59% identity [119]. Nedd8 is conjugated to a conserved lysine residue of the C-terminal region of the cullin in a process known as neddylation; using an E1-E2-E3 enzyme cascade similar to ubiquitylation [119]. The E1, E2, and E3 enzymes that catalyze neddylation have analogs to those in the ubiquitylation reaction. Nedd8 is activated by the heterodimeric E1 activating enzyme APP-BP-1 (Amyloid Precursor Protein-Binding Protein 1) / Uba3 in an ATP-dependent manner. Activated Nedd8 is then transferred to the E2 conjugating enzyme UBC12 which shuttles activated Nedd8 to the E3 ligase Rbx1 and Dcn1 (Defective in Cullin Neddylation 1) that then ensures specific conjugation of Nedd8 to its cullin complexes. Neddylation enhances activity of CRL complexes [ ]. Based on studies in mammalian systems, three mechanisms are proposed. In vitro studies have found that Nedd8 functions as an E2 docking site [120, 125]. Based on the interaction of E2 with RING finger domains such as Rbx1, it has been proposed that both Nedd8 and Rbx1 form an interface for loading an E2 enzyme. In vivo study leads to another proposal that Nedd8 conjugation facilitates CRL dimerization that is presumably the more active form [126]. A recent study in the crystal structure of Nedd8~CUL1-Rbx1 proposes the third mechanism that Nedd8 conjugation induces a conformational change of CRL complexes to promote ubiquitylation of a substrate. Upon Nedd8 conjugation, cullin and Rbx1 subdomains are reoriented; a CAND1-binding site is eliminated; and a substrate is 17

29 juxtaposed to the activated E2 enzyme which is loaded on Rbx1 region in a CRL complex. As result of it, a substrate is easily ubiquitylated [127]. CSN (COP9 Signalosome) CSN (COP9 Signalosome) is a conserved multiprotein complex that is typically composed of eight subunits. CSN interacts with neddylated CRLs and removes Nedd8 from CRLs by the isopeptidase activity of the metalloprotease CSN5/Jab1 subunit of CSN in a process called deneddylation [ ]. In CSN loss-of-function mutants, neddylated cullins accumulate in vivo [131, 132]. However, loss of CSN causes autoubiquitylation of SRSs and leads to decreasing CRL activity [ ]. Those studies suggest that the deneddylation by CSN is required for stabilization of CRL components [134, ]. CSN is essential for proper CRL activity in vivo [ ]. CAND-1 (Cullin-Associated and Neddylated-Dissociated) CAND-1, which is composed of multiple HEAT repeats, is a CRL inhibitor that binds to cullin-rbx1 complexes without adaptors and Nedd8, and prevents formation of an active CRL complex [140, 141]. The crystal structure of human CAND1 bound to a CUL1-Rbx1 complex shows that CAND1 wraps around the cullin complex [142]. The CAND1 N-terminus binds to the cullin C-terminus and the CAND1 C-terminus interacts with the cullin N-terminus. As a result, CAND1 completely blocks both the adaptordocking site and the Nedd8 conjugation site, making it impossible to assemble active CRL complexes. In human cells, CAND1 can interact with all cullins [141, 143]. However, CAND1 has a preference in associating with certain members of the cullin family depending on the cell type. In human HEK293T cells, CAND1 predominantly interacts with CUL1, and 18

30 interaction with CUL2 or CUL3 is not observed [141]. In human HeLa cells, CAND1 interacts with CUL1, CUL2, CUL3, and CUL4A [143]. In C. elegans, CAND1 binds strongly to CUL-2 and CUL-4 [100](unpublished data), but no interaction with CUL-3 was detected [144]. It can be assumed that CAND1 is released from cullin-rbx1 at certain points. An in vivo study found that CAND1 is indeed able to dissociate from the CUL1-Rbx1 complex [141]. However, how the dissociation of CAND1 from CUL1-Rbx1 is regulated is not yet clear. When CAND1 was first discovered, it was expected that the absence of CAND1 would increase CRL activity. However, the inactivation of CAND1 causes the inactivation of SCF complexes in humans and Arabidopsis, and the CRL3 complex in humans due to autoubiquitylation of SCF complexes and a yet-unknown mechanism for CRL3 complexes [140, ]. CAND1 is required for proper CRL activity but how CAND1 controls CRL activity is still a mystery. Shuttling between the activated and inactivated forms of CRL CRL complexes continuously switch between an active state of full assembly with SRSs and Nedd8 conjugation, and an inactive state with no Nedd8 conjugation, missing CRL components, or bound to the CAND-1 inhibitor (Fig 1-4). Activated CRL complexes are neddylated and enable ubiquitylation of targets. Without substrates, SRSs either undergo autoubiquitylation or degradation by other CRL complexes [148]. Once the SRS is degraded, the core CRL components are able to be deneddylated by CSN [149]. A deneddylated adaptor-cullin-rbx1 complex can then either reform a CRL complex or undergo sequestration by CAND1, in which the adaptor is stripped away from cullin- Rbx1 in the process of CAND1 binding. CAND1 is released from cullin-rbx1 by yetundefined mechanisms. Once cullin-rbx1 is available, an adaptor and an SRS bind to 19

31 cullin-rbx and reconstitute a CRL complex. Upon binding of a substrate, a CRL complex is neddylated and fully active [150]. 20

32 Figure 1-1. The formation of pre-replicative complexes (pre-rcs) In late mitosis or early G1 phase, the licensing factors Cdt1 and Cdc6 load onto replication origins via interaction with the Origin Recognition Complex (ORC). Cdt1 and Cdc6 together recruit DNA helicase MCM2-7 complexes onto replication origins. During S phase, the MCM2-7 complex unwinds DNA to allow the initiation of DNA replication, and then moves with the replication machinery at replication forks (not shown). At the same time, in a number of species, Cdt1 and Cdc6 are inactivated to prevent the reloading of MCM2-7 complexes to replication origins. (Diagram from [151]) 21

33 22

34 Figure 1-2. Ubiquitin-mediated proteolysis E1 ubiquitin-activating enzyme activates ubiquitin. E2 ubiquitin-conjugating enzyme accepts ubiquitin from an E1 and transfers it to a target with the assistance of E3 ubiquitin ligase. Polyubiquitinated substrate are recognized and degraded by the 26S proteasome. (Diagram provided by E.T. Kipreos) 23

35 24

36 Figure 1-3. Structure of multisubunit CRL complexes Diagrams of the CRL1 (A), CRL2 (B), CRL5 (C), CRL3 (D), and CRL4 (E) complexes. Components in the complexes are labeled. (Diagram provided by E.T. Kipreos [118]) 25

37 26

38 Figure 1-4. Activated and Inactivated forms of CRL complexes CRL complexes can shift between an active form and an inactive form lacking Nedd8 by a CNS complex (top). CRL complexes which do not have a substrate can either bind to CSN or lose their SRS due to autoubiquitylation or degradation by another E3 ubiquitin ligase (right). The absence of SRSs allows a CSN complex to deneddylate a CRL complex. The deneddylated adaptor-cullin-rbx1 complex can either reform a CRL complex by binding a SRS or be sequestered by a cullin-inhibitor CAND-1 (bottom). CAND-1 is dissociated from cullin-rbx1 by yet-undefined mechanisms and released cullin-rbx1 is able to reconstitute an active state of a CRL complex (left). (Diagram modified from [118]) 27

39 28

40 CHAPTER 2 C. elegans CUL-4 promotes CDC-6 nuclear export by a CKI-1-dependent pathway to maintain genome integrity 2 2 C. elegans CUL-4 prevents rereplication by promoting the nuclear export of CDC- 6 via a CKI-1-dependent pathway. Kim J, Feng H, Kipreos ET. Curr Biol Jun 5;17(11): Epub 2007 May 17. Reprinted here with permission of Current Biology 2 Control of the Cdc6 replication licensing factor in metazoa: the role of nuclear export and the CUL4 ubiquitin ligase. Kim J, Kipreos ET. Cell Cycle Jan 15;7(2): Epub 2007 Nov 3. Reprinted here with permission of Cell Cycle 29

41 BACKGROUND To maintain genome stability, the duplication of genomic DNA in eukaryotic cells is tightly regulated and coordinated with events in the cell cycle. The initiation of DNA replication is temporally regulated in two distinct steps to ensure that replication origins fire only once in each cell cycle [ ]. The first temporal step is the acquisition of replication competence during late mitosis or G1 phase when pre-replicative complexes (pre-rcs) assemble at replication origins. The pre-rc forms by the sequential binding of conserved proteins: the six-member origin recognition complex (ORC); the replication licensing factors Cdt1 and Cdc6; and the MCM2-7 helicase complex. ORC binds the DNA of the replication origin and recruits the licensing factors Cdt1 and Cdc6, which together load the MCM2-7 complex onto replication origins. The MCM2-7 complex is the replicative helicase that is presumed to unwind the replication origin DNA and may promote replication fork movement [155]. Upon loading of the MCM2-7 complex, the origin is considered to be licensed to allow the initiation of DNA replication [153, 156]. The second temporally-regulated step of DNA replication occurs upon entry into S phase. Two classes of protein kinases, Cdc7-Dbf4 (DDK) and S phase cyclin/cyclindependent kinase (CDK), phosphorylate pre-rc components to allow additional replication factors to load onto the replicative complex, leading to origin unwinding, polymerase loading, and the initiation of DNA replication. During S phase, specific pre- RC components are regulated to prevent the re-formation of pre-rcs until the next cell cycle. In a wide range of eukaryotes, the licensing factors Cdt1 and Cdc6 are tightly regulated to prevent their participation in pre-rc formation during S phase [153, 157]. The temporal segregation of pre-rc assembly (during late M or G1 phase) and origin firing (during S phase) effectively prevents replication origins from initiating DNA replication more than once per cell cycle. 30

42 The licensing factor Cdc6 is essential for pre-rc formation and replication initiation [153, 156]. Cdc6 is a member of the AAA+ ATPase family, which contains Walker A and B ATPase domains that are required for the loading of MCM2-7 onto chromatin [32, 34, 158, 159]. In a wide range of eukaryotes, Cdc6 activity is restricted during S phase to prevent the reloading of MCM2-7 onto chromatin, although the mode of regulation differs between yeast and metazoan species. In budding yeast and fission yeast, the Cdc6 ortholog is phosphorylated by the major cell cycle CDK and undergoes ubiquitin-mediated proteolysis by a E3 ubiquitin ligase, SCF cdc4 or SCF Pop-1/Pop-2, respectively [37, ]. In mammals and Xenopus, Cdc6 is not degraded during S phase, but is instead exported from the nucleus in response to phosphorylation near two nuclear localization signals (NLSs) in the Cdc6 N-terminus [51, 52, 165, 166]. Ectopically expressed mammalian Cdc6 is primarily exported from the nucleus during S phase [51, 52, 167]. In contrast, endogenous Cdc6 is only partially exported, with a substantial percentage remaining in the nucleus throughout S phase [50, 51, 53, 54, 168, 169]. The function of the residual nuclear Cdc6 is not known; and the role of Cdc6 nuclear export in preventing excessive DNA replication, if any, has not been established [153]. In several eukaryotes, overexpression of the licensing factors Cdt1 and Cdc6 can induce DNA re-replication, in which replication origins fire more than once per cell cycle. In fission yeast, overexpression of Cdc18 induces significant levels of DNA rereplication, while co-expression of Cdt1 with Cdc18 enhances DNA re-replication [30, 35, 40, 170]. In budding yeast, overexpression of Cdc6 delays the initiation of mitosis but does not affect the extent of DNA replication [171]. Limited re-replication is observed if a non-degradable Cdc6 mutant is expressed in combination with ORC and MCM2-7 components that lack CDK phosphorylation sites [42]. In mammals, overexpression of Cdt1 and Cdc6 (to levels several orders of magnitude higher than the endogenous 31

43 proteins) induces limited DNA re-replication in cells lacking the p53 checkpoint pathway [57]. We previously found that inactivation of the cul-4 gene in C. elegans does not affect the timing of S phase entry, however, once cells enter S phase they arrest in that stage and undergo severe DNA re-replication, accumulating up to 50-fold higher levels of genomic DNA [116]. CUL-4 is a member of the cullin-ring ubiquitin ligase family, and acts as a scaffold within multisubunit ubiquitin ligase complexes, which conjugate ubiquitin to substrates thereby promoting their degradation by the 26S proteasome [61]. In wild-type C. elegans, the licensing factor CDT-1 is degraded upon entry into S phase [116]. However, in cul-4(rnai) animals, CDT-1 is not degraded and instead accumulates in the S phase-arrested, re-replicating cells [116]. Genetic experiments demonstrated that the failure to degrade CDT-1 directly contributes to the DNA rereplication [116]. Subsequently, CRL4 complexes in C. elegans, humans and Xenopus were shown to bind and target Cdt1 for ubiquitin-mediated proteolysis in a PCNAdependent manner [108, , 117]. The observation of dramatic DNA re-replication upon loss of C. elegans CUL-4 is curious given that replication licensing mechanisms are generally redundant [157], and so de-regulation of CDT-1 by itself should not lead to DNA re-replication. This led us to ask whether the loss of CUL-4 deregulates additional licensing mechanisms that would normally prevent DNA re-replication. In this study we show that C. elegans CDC-6 translocates from the nucleus to the cytoplasm during S phase in response to phosphorylation. We demonstrate that CUL-4 is required for the phosphorylation of CDC-6 during S phase and its subsequent nuclear export via the negative regulation of the CDK-inhibitor CKI-1. Finally, we present the first evidence that CDC-6 nuclear export is an important safeguard to prevent the re- 32

44 initiation of DNA replication. In C. elegans, CDC-6 nuclear export works redundantly with CDT-1 degradation to prevent DNA re-replication. RESULTS Changes in CDC-6 subcellular localization through the cell cycle We followed the dynamics of CDC-6 during the first cell division of the V1-V6 hypodermal seam cells. The timing of S phase entry for the seam cells was determined by following Prnr-1-driven expression of GFP [172] or tdtomato fluorescent proteins. The different seam cells were found to vary in their timing (post-hatching) of S phase entry: min post-hatch for V5, min for V2-V4; min for V6; and min for V1 (data not shown). S phase entry for the seam cell V4 was similar to that previously reported [116]. To enable detailed analysis of nuclear export, we created a transgenic system in which ectopically expressed CDC-6 was fused to GFP or tdtomato proteins. Initial attempts to express CDC-6 from its own promoter produced lethality. However, by restricting expression to hypodermal cells using the wrt-2 promoter [173], we were able to achieve stable expression of CDC-6. At min post-hatch, the V1-V6 seam cells began expression of CDC-6 in the nucleus; and by min post-hatch, nuclear CDC-6 was present at high levels in the seam cells (Fig 2-1 and 2-2A, C, D). As the seam cells enter into S phase, the level of nuclear CDC-6 dropped gradually and cytoplasmic CDC-6 increased. Beyond 180 min post-hatch, cytoplasmic CDC-6 remained at high levels in the seam cells (Fig 2-1 and 2-2A, C, D). These data indicate that CDC-6::GFP expressed in the V1-V6 seam cells underwent translocation from the nucleus to the cytoplasm during S phase (Fig 2-1 and 2-2A, C, D). A higher percentage of the exogenously-expressed CDC-6::GFP protein translocated from the nucleus compared to endogenous CDC-6, similar to what 33

45 is observed in other metazoa (Fig 2-1; data not shown) [51, 167]. CDC-6::GFP was present throughout the cell cycle (Fig 2-2A). During mitosis, CDC-6::GFP localized to the condensed chromatin (Fig 2-2A, B). To determine if the cytoplasmic localization of CDC-6::GFP during S phase is due in part to the degradation of CDC-6::GFP in the nucleus, we expressed a mutant CDC-6::GFP that was nuclear-localized throughout S phase. Nuclear localization was achieved by substituting alanine residues for two conserved leucines (434 and 436) in the predicted CDC-6 nuclear-export sequence (NES) to inactivate the NES [174, 175], and adding two SV40 large T antigen NLS sequences to C-terminus. We observed that the CDC-6 m nls mutant protein was nuclear-localized throughout the cell cycle and its level did not noticeably change before, during or after S phase (Fig 2-3), thereby confirming that CDC-6 is not regulated by degradation during S phase. To determine the time course of CDC-6 nuclear export during S phase, we investigated CDC-6::GFP or CDC-6::tdTomato localization at set time points after hatching (Fig 2-2C). At min post-hatch, CDC-6::GFP remained nuclear in most of the seam cells except V5. At 200 min post-hatch, the V2-V6 seam cells had cytoplasmic expression of CDC-6::GFP (Fig 2-2C). The differences in timing of CDC-6 translocation correlated with the timing of S phase entry for the seam cells, and indicated that CDC-6 translocation to the cytoplasm occurred after entry into S phase. CUL-4 is required for CDC-6 nuclear export and its phosphorylation in S/G2 phase In animals that lack CUL-4, larval blast cells have normal S phase entry but arrest in S phase and undergo extensive DNA re-replication, which is associated with a failure to degrade the licensing factor CDT-1 [116]. We sought to address whether CUL- 4 also regulates the licensing factor CDC-6 during S phase. Endogenous CDC-6 localization was followed in cul-4(rnai) larvae by immunofluorescence using affinity 34

46 purified anti-cdc-6 antibody. Prior to S phase, CDC-6 nuclear expression in cul- 4(RNAi) seam cells increased with a time course similar to that of wild-type seam cells (data not shown). However, in S phase cul-4(rnai) seam cells, CDC-6 remained nuclear with no appreciable increase in cytoplasmic staining (data not shown). A similar requirement for CUL-4 was observed for the nuclear export of ectopically expressed CDC-6::GFP, which remained nuclear-localized in S phase cul-4(rnai) seam cells that expressed the S phase-marker Prnr-1::tdTomato (Fig 2-1). These results indicate that CDC-6 nuclear export during S phase requires the CUL-4 ubiquitin ligase. In humans, Cdc6 nuclear export is triggered by phosphorylation of three residues that are located near two N-terminal NLSs [51, 52] (Fig 2-4A). C. elegans CDC-6 has three putative NLSs in the N-terminus: one bipartite NLS and two simple NLSs [176] (Fig 2-4A). There are two sites in CDC-6 that match the consensus for CDK phosphorylation (S/T-P-X-K/R) and seven additional sites that have the minimum requirement for CDK phosphorylation (S/T-P) [177]. Of these potential CDK phosphorylation sites, six are located in the N-terminal region, near the NLSs (Fig 2-4A). Additionally, five consensus sequences for cyclin binding (RXL) are present in CDC-6 with one located in the N- terminus near the NLSs at position 88 [178]. We were interested in whether phosphorylation near the NLSs promotes C. elegans CDC-6 nuclear export. To examine this, we generated an anti-phospho-specific CDC-6 antibody against the predicted CDK phosphorylation site residue threonine-131, which is located between the second and third NLS sequences. Affinity purified antiphospho-t 131 -CDC-6 antibody recognized the endogenous CDC-6 protein on a western blot but could not recognize CDC-6 after treatment with phosphatase, thereby demonstrating the specificity of this antibody (data not shown). We examined the status of CDC-6 phosphorylation in G1 and S phase by immunofluorescence using the anti-phospho-t 131 -CDC-6 antibody. In wild-type larvae, 35

47 no detectable anti-phospho-t 131 -CDC-6 signal was observed in V1-V6 seam cells before S phase (Fig 2-4B). After cells entered S phase ( min post-hatch), antiphospho-t 131 -CDC-6 signal was detected in both the nucleoli and cytoplasm (Fig 2-4B). Therefore, at least a subset of CDC-6 that undergoes nuclear export is phosphorylated on position T 131. Within the nucleus, the observation of a discrete nucleolar anti-t CDC-6 signal contrasted with the more uniform nuclear signal with anti-cdc-6 immunofluorescence, suggesting that a higher percentage of nucleolar CDC-6 is phosphorylated on the T 131 site. In contrast to the pattern in wild type, anti-phospho-t CDC-6 stain was not detected in cul-4(rnai) larvae either before or after S phase, indicating that CUL-4 is required for the phosphorylation of T 131 during S phase (Fig 2-4B, C). Phosphorylation triggers CDC-6 nuclear export during S phase To determine the functional relevance of phosphorylation for CDC-6 translocation, we generated transgenic strains expressing CDC-6::GFP with site-directed mutations of potential CDK phosphorylation sites. Serine and threonine residues in the predicted N-terminal CDK phosphorylation sites (Fig 2-4A) were replaced by alanine, so that the sites were incapable of being phosphorylated. All of the CDC-6::GFP CDK-site mutants had normal nuclear localization in G1 phase (Fig 2-5; data not shown). Subcellular localization during S phase was assessed in the V1-V6 seam cells at min post-hatch, a time when wild-type CDC-6 is predominantly localized to the cytoplasm (Fig 2-1 and 2-2A, C, D). Single mutations of T 48, S 93, T 131, or T 145 each caused a modest retention of CDC-6::GFP in the nucleus during S phase relative to wild type, with the T 145 mutation having the greatest effect (Fig 2-6A, B). However, each of these single-site mutant CDC-6 proteins still predominantly translocated to the cytoplasm. Double mutations of 36

48 T 48 and T 131 impeded CDC-6 translocation more than the respective single mutations. Combining T 48, S 109, T 131, and T 145 mutations gave approximately the same translocation rate as the T 145 mutation alone (Fig 2-6A, B). However, mutating all six of the N-terminal CDK phosphorylation sites (T 48, S 93, S 109, T 128, T 131 and T 145 ) completely blocked the nuclear export of CDC-6 during S phase (Fig 2-6A, B). This analysis implies that the phosphorylation of multiple sites is required for CDC-6 nuclear export. The CDC-6 mutant with all six N-terminal CDK-phosphorylation residues converted to alanines (CDC-6mCDK) was localized exclusively to the nucleus throughout the cell cycle, from hatch until seam cells entered mitosis (Fig 2-5; data not shown). No obvious defects in DNA replication or cell cycle progression were observed in transgenic animals expressing CDC-6mCDK. The CDC-6 m nls mutant, which is nuclear localized through a phosphorylation-independent means, was similarly expressed in the nucleus throughout the cell cycle and did not induce DNA replication or cell cycle defects (Fig 2-3; data not shown). These results suggest that continuous CDC-6 nuclear localization during S phase does not, by itself, deregulate DNA replication. CDC-6 nuclear export is independent of CDT-1 degradation In fission yeast and mammals, Cdt1 physically interacts with Cdc6 [40, 179]. This allows a potential scenario in which CDT-1 degradation is a necessary precedent for CDC-6 nuclear export, perhaps by allowing kinases to gain access to CDC-6 phosphorylation sites. In this model, the failure to observe CDC-6 nuclear export in cul- 4(RNAi) animals could be a secondary consequence of the failure to degrade CDT-1. To test this hypothesis, we wanted to express a non-degradable CDT-1 mutant to determine if its perdurance in S phase would prevent CDC-6 nuclear export. 37

49 In Drosophila and humans, mutation of CDK-phosphorylation sites on Cdt1 can partially stabilize Cdt1 protein [ ]. Additionally, the PCNA-interacting protein (PIP) box sequence of human Cdt1, which allows it to interact with Proliferating Cell Nuclear Antigen (PCNA), is essential for its degradation [108, ]. In order to stabilize C. elegans CDT-1, we substituted alanine residues for three conserved amino acids in the PIP box motif that mediates interaction with PCNA and for serine and threonine residues in the five N-terminal CDK consensus sites to create the CDT- 1mCDK+PIP(3A) mutant (Fig 2-7A). Both wild-type and mutant CDT-1::GFP were present in G1 phase seam cells (Fig 2-7B). As expected, wild-type CDT-1::GFP was not observed in S/G2 phase seam cells ( min post-hatch) (Fig 2-7B). However, CDT-1mCDK +PIP(3A)::GFP was present in the majority of S/G2 phase seam cells at either high or low levels (28% or 55%, respectively, n = 18), indicating that the mutations partially stabilize CDT-1 (Fig 2-7B). To test if CDC-6 is retained in the nucleus during S phase in the presence of stabilized CDT-1, we investigated the localization of CDC-6::tdTomato in animals expressing the stabilized CDT-1mCDK+PIP(3A)::GFP. We observed that CDC- 6::tdTomato was still exported to the cytoplasm during S phase despite the presence of stabilized CDT-1::GFP (Fig 2-7C). It should be noted that the stabilized CDT-1 is functional and can promote DNA replication (see below). This result suggests that CDC- 6 nuclear export is regulated independently of CDT-1 degradation. CKI-1 inactivation rescues CDC-6 nuclear export in cul-4 mutants Recent studies in Drosophila and humans have reported that CUL4 negatively regulates the stability of CDK inhibitors of the CIP/KIP family [ ]. We have found that C. elegans CUL-4 negatively regulates the CIP/KIP-family member CKI-1 [117]. CKI-1 accumulates in enlarged re-replicating cells of cul-4(rnai) arrested larvae (data not 38

50 shown) [117]. RNAi depletion of cki-1 reduces the size of seam cell nuclei and DNA levels in cul-4(gk434) larvae, indicating that cki-1(rnai) can partially suppress the rereplication phenotype (Fig 2-8A) [117]. cki-1 RNAi does not suppress CDT-1 accumulation in cul-4(gk434) deletion mutant, indicating that its role in preventing DNA re-replication is independent of the failure to degrade CDT-1 [117]. Our results indicate that the phosphorylation of CDK sites is required to promote CDC-6 nuclear export. We hypothesize that CUL-4 promotes CDC-6 nuclear export by negatively regulating CKI-1 levels, and that in the absence of CUL-4, CKI-1 accumulation prevents CDK(s) from initiating CDC-6 nuclear export. To test whether inactivation of CKI-1 is required for the nuclear export of CDC-6, we investigated the localization of Pwrt-2::CDC-6::GFP upon inactivation of both cul-4 and cki-1. CDC-6 is predominantly nuclear-localized in 73% of cul-4(gk434) mutant V2-V6 seam cells during S phase (n = 30) (the failure to observe full penetrance for nuclear localization is likely the effect of cul-4 maternal product [116]). cki-1 RNAi in the cul-4(gk434) mutant significantly increases the percentage of seam cells with cytoplasm-localized CDC- 6::GFP (83% with cki-1 RNAi, n = 29, vs. 27% with no RNAi), consistent with our model (Fig 2-8B). A prediction of the model is that cki-1(rnai) will restore the phosphorylation of CDC-6 in cul-4(gk434) mutants during S phase. The status of CDC-6 phosphorylation was assessed by immunofluorescence using the anti-phospho-t 131 -CDC-6 antibody. In S phase, anti-phospho-t 131 -CDC-6 signal was detected in a relatively small percentage of cul-4(gk434) seam cells (33% for cul-4(gk434), n=30 vs. 80% for wild-type, n=41) (Fig 2-4B, C). However, cki-1(rnai) in cul-4(gk434) increased the percentage of seam cells with phospho-t 131 signal to the level seen in wild type (82% n=45) (Fig 2-4C). If cki-1 RNAi induces CDC-6 nuclear export in cul-4 mutants by permitting CDC-6 phosphorylation, then it would be expected that a CDC-6 mutant lacking phosphorylation sites would not undergo nuclear export upon cki-1 RNAi depletion. We observed that 39

51 the non-phosphorylatable CDC-6mCDK mutant protein did not undergo nuclear export in cki-1(rnai), cul-4(gk434) seam cells (n = 47) (Fig 2-8C). These results indicate that the presence of CKI-1 is essential for the block on the phosphorylation-dependent nuclear export of CDC-6 that is observed in cul-4 mutant cells. Deregulation of both CDT-1 and CDC-6 can induce re-replication We investigated the biological significance of the regulation of CDT-1 and CDC-6 in S phase by overexpressing wild-type and deregulated proteins using the heat shock promoters hsp16-2 and hsp16-41 [186]. Wild-type CDT-1 and CDC-6, or stabilized CDT-1 [CDT-1mCDK+PIP(6A)] and non-exportable CDC-6 [CDC-6m5CDK], were expressed individually or in combination (Fig 2-9A). Expression of wild-type CDT-1 or CDC-6 individually produced a low level of lethality (29% and 33% arrested embryos or L1 larvae, respectively). Expression of deregulated CDT-1 or CDC-6 produced higher levels of lethality (75% and 77%, respectively). Expressing combinations of CDT-1 and CDC-6 produced more lethality than individually expressing the proteins. The lethality associated with expression of deregulated CDC-6 was consistently greater than the lethality associated with deregulated CDT-1, and 100% lethality was obtained only when deregulated CDC-6 was combined with either wild-type CDT-1 or deregulated CDT-1 (Fig 2-9A). The increased lethality associated with deregulated CDC-6 indicates that a failure to phosphorylate CDC-6 and thereby induce its translocation is incompatible with viability. Analysis of the arrested embryos showed that 47% (n = 36) of the embryos expressing non-exportable CDC-6 with stabilized CDT-1 contained cells with increased DNA levels (21.5 ± 2.4 C DNA content, n = 63, vs. 2.4 ± 0.3 C, n = 13, for asynchronous early embryo interphase cells) (Fig 2-9B). In contrast, other combinations of wild-type and/or deregulated CDT-1 and CDC-6 did not produce noticeably increased DNA levels 40

52 within the arrested embryos (Fig 2-9B; data not shown). The increased DNA content in the embryos expressing both deregulated licensing factors could arise either from DNA re-replication, in which replication initiates continuously throughout S phase, or from failed mitosis, in which cells with duplicated genomes re-enter S phase without segregation of DNA to daughter cells. These two mechanisms can be distinguished by analyzing centrosome numbers. The failed mitosis mechanism is associated with the presence of extra centrosomes in cells, due to the duplication of centrosomes in subsequent cell cycles [96]; while cells that undergo DNA re-replication within one S phase will have only two centrosomes. We analyzed centrosome number in cells expressing deregulated CDT-1 and CDC-6 by immunofluorescence with anti-spd-2 antibody, which stains centrioles and pericentriolar material [187]. We observed that cells with excessive DNA levels had only two centrosomes (Fig 2-9B). This implies that the increased ploidy is not due to a failure of mitosis, but rather arises from DNA rereplication within a single S phase. Therefore, the stabilized CDT-1 and non-exportable CDC-6 are functional, and can act in concert to induce the re-initiation of DNA replication within an S phase. DISCUSSION CDC-6 nuclear export is regulated by phosphorylation on multiple CDK sites In humans and Xenopus, Cdc6 is exported from the nucleus to the cytoplasm during S phase [51, 52, 166, 167]. In contrast, Drosophila Cdc6 has been reported to remain nuclear throughout the cell cycle [188]. We found that C. elegans CDC-6 is exported from the nucleus during S phase, suggesting that this is a conserved mechanism for the regulation of Cdc6 among metazoans. In vertebrates, Cdc6 subcellular localization is controlled by balancing nuclear import (mediated by N-terminal NLSs) and nuclear export (mediated by C-terminal 41

53 NESs) [52, 165]. During G1 phase, nuclear import is dominant and the majority of Cdc6 is nuclear localized. During S phase, nuclear import is disrupted by phosphorylation near the NLSs, and Cdc6 undergoes nuclear export. C. elegans CDC-6 has three NLSs located in the N-terminus and six potential N-terminal CDK phosphorylation sites. Using phospho-specific antibody, we showed that one CDK site (T 131 ) is not phosphorylated during G1 phase but becomes phosphorylated during S phase. During S phase, phospho-t 131 CDC-6 signal was present in the cytoplasm as well as in the nucleolus. Because a distinct nucleoli signal was not detected by immunofluorescence with the regular anti-cdc-6 antibody, the phosphorylation of T 131 appears to be enriched in the nucleolar sub-fraction of CDC-6. This suggests that the phosphorylation of particular CDC-6 sites may direct specific subcellular localization. Our studies suggest that multiple phosphorylation sites in CDC-6 are required to direct nuclear export, with some sites more important than others. Mutation of all six N- terminal CDK phosphorylation sites completely blocks the translocation of CDC-6 from the nucleus to the cytoplasm during S phase. There are conflicting reports on whether the expression of unphosphorylatable human Cdc6 inhibits the initiation of DNA replication [51, 52]. We did not observe any significant effects on either DNA replication or cell cycle progression upon expression of unphosphorylatable CDC-6 in C. elegans, consistent with other reports in humans and Xenopus [51, 166]. These results suggest that the phosphorylation of CDC-6 is not required for DNA replication and that constitutively nuclear-localized CDC-6 is compatible with DNA replication. C. elegans CDC-6 protein levels are not regulated by phosphorylation In humans, Cdc6 has been variously reported to have relatively stable protein levels throughout the cell cycle [52, 189] or to be degraded in early G1 phase cells [168, 190]. The APC/Cdh1 ubiquitin ligase degrades human Cdc6 in G1 phase [190]. This 42

54 degradation has been reported to be inhibited by phosphorylation on CDK consensus sites, and mutating these sites to non-phosphorylatable alanine residues increases degradation while mutation to aspartic acid residues blocks degradation [191]. However, it has also been reported that altering the CDK phosphorylation sites in human Cdc6 by mutation to alanine (mimic non-phosphorylatable form) or aspartic acid residues (mimic a constitutively phosphorylated form) has no effect on APC/C-mediated degradation [190]. In C. elegans post-hatch larvae, CDC-6 is not expressed during their quiescent period. We observed that after cells leave quiescence, CDC-6::GFP levels remain constant throughout the cell cycle. In contrast to what has been reported for human Cdc6, mutation of the CDK phosphorylation sites did not destabilize CDC-6, which remains nuclear localized throughout the cell cycle. CDC-6 that was nuclear-localized through inactivation of the NES and addition of NLS sequences also remained nuclear and stable throughout the cell cycle. These observations suggest that C. elegans CDC- 6 levels are not regulated by degradation to any appreciable extent during the cell cycle. Deregulation of both C. elegans CDC-6 and CDT-1 is required for re-replication In S. pombe, the overexpression of Cdc18 (Cdc6 ortholog) is sufficient to induce significant DNA re-replication [30, 170]. While overexpression of S. pombe Cdt1 does not induce DNA re-replication, co-overexpression of Cdt1 and Cdc18 produces more efficient DNA re-replication than the expression of Cdc18 alone [35, 40]. In S. cerevisiae, Drosophila, and humans, overexpressing Cdc6 does not induce DNA rereplication [57, 188, 192]. In humans, co-overexpression of wild-type Cdt1 and Cdc6 produces modest DNA re-replication in a subset of cells that lack a cell cycle checkpoint pathway [57]. 43

55 In our studies, co-expression of deregulated CDT-1 that could not be degraded and deregulated CDC-6 that could not be exported from the nucleus produced DNA rereplication in the early embryo. In contrast, overexpression of deregulated CDT-1 or CDC-6 in combination with wild-type CDC-6 or CDT-1, respectively, did not induce DNA re-replication. This indicates that redundant regulation of CDT-1 and CDC-6 prevents the re-initiation of DNA replication within the same cell cycle. However, only approximately half of the embryos with co-expression of both deregulated licensing factors exhibited DNA re-replication. This may point to further redundancy in the replication licensing system, perhaps other regulatory mechanisms that have been implicated in vertebrate licensing such as inhibition of Cdt1 by Geminin or inactivation of ORC1 by ubiquitylation or degradation [153]. We did observe embryonic lethality upon expression of CDT-1 and/or CDC-6 that was not associated with increased DNA levels. The reason for this lethality is unclear but could arise from changes in the timing of DNA replication, which is known to produce developmental defects and embryonic arrest [193] or from disruption of cellular functions that unrelated to replication licensing, as have been reported for Cdc6 orthologs [194, 195]. CDC-6 nuclear export acts to prevent DNA re-replication In humans and Xenopus, ectopically expressed CDC-6 is largely exported to the cytoplasm during S phase [51, 52, 54, 167]. In contrast, a substantial fraction of endogenous CDC-6 remains nuclear localized during S phase [49-54]. Consistent with this, a substantial fraction of endogenous C. elegans CDC-6 remains nuclear during S phase, while ectopic CDC-6 is predominantly in the cytoplasm during S phase with hardly any remaining in nuclei. In no system is it understood why endogenous and ectopically expressed Cdc6 have different localization patterns during S phase. 44

56 The presence in mammals of nuclear-localized Cdc6 during S phase has led to the proposal that the nuclear export of Cdc6 is not important for restraining DNA replication licensing activity [54]. During S phase, nuclear Cdc6 in human cells has been reported to localize to a chromatin fraction based on the observation that Cdc6 is not solubilized with Triton X-100 detergent and it pellets upon centrifugation [53, 54, 168, 169]. However, another study reported that Cdc6 is localized to non-chromatin nuclear foci during S phase that cannot be solubilized with DNase treatment [196]. Regardless, of whether the nuclear Cdc6 is chromatin bound, there is currently no evidence for a functional role of nuclear export in restraining DNA replication licensing [153]. In this study, we observed that constitutively nuclear-localized CDC-6 could synergize with deregulated CDT-1 to induce DNA re-replication, thereby implicating the nuclear export of CDC-6 as a potent safeguard mechanism to prevent the re-initiation of DNA replication. This provides the first evidence of a significant functional role for the nuclear export of CDC-6 in any organism. CUL-4 regulates both CDC-6 and CDT-1 licensing factors RNAi depletion of CUL-4 produces dramatic levels of DNA re-replication, with up to 50-fold increases in ploidy, which is associated with a failure to degrade CDT-1 [116]. However, overexpressing Cdt1 in yeast does not induce DNA re-replication, and overexpressing human Cdt1 several log-fold higher than the endogenous protein produces only modest re-replication in a subset of cells [35, 57, 197]. Given this, it was hard to reconcile the substantial DNA re-replication in cul-4(rnai) animals with simply a failure to degrade CDT-1 during S phase. Our work reveals that the CDC-6 licensing factor is also deregulated in cul- 4(RNAi) animals. CDC-6 is not phosphorylated on CDK-sites in cul-4(rnai) animals, and this phosphorylation is essential for CDC-6 nuclear export. One potential 45

57 mechanism to explain the lack of CDC-6 translocation in cul-4 mutants would be if the failure to degrade CDT-1 led to the failure to phosphorylate CDC-6 and thereby prevented its subsequent translocation. However, expression of a stabilized CDT-1 did not block CDC-6 nuclear export during S phase, suggesting that this model is not valid. Rather, our data suggest that failure of CDC-6 nuclear export in cul-4 mutants, results from a loss of the CUL-4-mediated negative regulation of the CDK-inhibitor CKI-1, and the accumulation of CKI-1 prevents the phosphorylation of CDC-6 and its subsequent translocation (Fig 2-10). RNAi depletion of cki-1 restored CDC-6 nuclear export to cul-4(gk434) mutant cells, indicating that CKI-1 is essential for the prevention of CDC-6 nuclear export in cul-4 mutants. Our results lead to the conclusion that CUL-4 is a master regulator that independently regulates two replication licensing factors during S phase: directly mediating the ubiquitin-dependent proteolysis of CDT-1 and indirectly promoting the nuclear export of CDC-6 via the negative regulation of CKI-1 (Fig 2-10). MATERIALS AND METHODS Nematode strains C. elegans strains were cultured as previously described [198]. The following strains were used: N2, wild type; PS3729, unc-119(ed4) III, syls78[ajm-1::gfp, unc- 119(+)]; VT765, unc-36(e251) III; mais103[prnr-1::gfp+unc-36(+)] [172]; ET268, ekex13[pwrt-2::cdc-6::tdtomato + prf4, which expresses the marker rol-6(su1006) [199]];PhistoneH1::GFP; ET237, ekis4[pwrt-2::cdc-6::gfp + Prnr-1:: tdtomato + prf4]; ET246, ekex14[pwrt-2::cdc-6m48::gfp + prf4]; ET292, ekex20[pwrt-2:: CDC- 6m93::GFP + Prnr-1::tdTomato+pRF4]; ET240, ekis5[pwrt-2::cdc-6m131::gfp + prf4]; ET256, ekis6[pwrt-2::cdc-6m145::gfp + Prnr-1::tdTomato + prf4]; ET224, ekex15[pwrt-2:: CDC-6m48.131::GFP + prf4]; ET251, ekex16[pwrt-2::cdc-6 m ::gfp + Prnr-1:: tdtomato + prf4]; ET241, ekis7[pwrt-2::cdc- 46

58 6m ::GFP + Prnr-1::tdTomato + prf4]; ET277, ekex17[pwrt-2::cdc- 6m ::GFP + Prnr-1::tdTomato + prf4]; ET252, ekex18[pwrt- 2::CDC6m ::GFP (CDC-6mCDK::GFP) + Prnr-1:: tdtomato + prf4]; ET296, ekex21[pwrt-2::cdc-6m nls::gfp + Prnr-1::tdTomato + prf4]; VC1033, cul-4(gk434)/min1[mis14 dpy-10(e128)]ii; ET291, cul-4(gk434)/min1, ekis4; ET318, cul-4(gk434)/min1, ekex22[pwrt-2::cdc6mcdk::gfp + Prnr-1::tdTomato + prf4]; and a strain containing an integrated transgene in which the histone H1 promoter drives GFP expression, the kind gift of Sudhir Nayak. Plasmid constructions Plasmid ppd95.75/pwrt-2::cdc-6::gfp was created by cloning 1371 base pairs of the wrt-2 promoter in front of the cdc-6 genomic coding sequence in the vector ppd95.75, which contains an in-frame GFP coding region. To create ppd95.75/pwrt- 2::CDC-6::tdTomato, the GFP cdna sequence was replaced with tdtomato cdna [200]. In order to create pprnr-1::tdtomato, the GFP sequence in pvt501(containing Prnr- 1::GFP) [172] was replaced with the tdtomato cdna. The plasmid ppd95.75/pnhr- 168::CDT-1::GFP was created by cloning 3334 bps of the nhr-168 promoter region upstream of the cdt-1 genomic coding sequence in the vector ppd Site-directed mutations were introduced into the coding sequences of ppd95.75/pwrt-2::cdc-6::gfp, ppd95.75/pwrt-2::cdt-1::gfp, or ppd95.75/pnhr-168::cdt-1::gfp with either the QuikChange Site Directed Mutagenesis kit (Stratagene) or an overlap extension mutagenesis method [201]. Constructs were sequenced to ensure the introduction of correct amino acid substitutions. To express wild-type or mutant cdt-1 and cdc-6 genes under the control of heat-shock promoters, cdt-1 and cdc-6 cdnas were cloned into plasmids ppd49.78 and ppd49.83, which contain the hsp16-2 and hsp16-41 promoters, 47

59 respectively [202]. The deregulated CDC-6 expressed by the heat-shock promoters, CDC-6m5CDK, contains alanine substitutions of T 93, S 109, T 128, T 131 and T 145. Transgenic lines Extragenic transgene lines were generated by microinjection according to standard methods [203]. Plasmids containing cdc-6 or cdt-1 ( µg/ml) were coinjected either with pprnr-1::tdtomato (66.6 µg/ml) and prf4 (66.6 µg/ml), or with prf4 (100 µg/ml) alone. Extrachromosomal arrays were integrated using 4000 rad from a 137 Cs gamma radiation source [203]. Expression from the wrt-2 promoter was predominantly in the seam cells H1, H2, V1-V6, and T, as well as the QL and QR neuroblasts. Pwrt-2-driven expression was consistently weaker in V6 than in other seam cells. F1 progeny of injected adult hermaphrodites were analyzed in the experiments presented in Figs 2-7B, 2-7C, and 2-9. For the experiments of Fig. 2-7B and C, plasmids were injected at the concentrations specified above. For the experiments of Fig 2-9, hermaphrodites were co-injected with combinations of CDT-1; CDT- 1mCDK+PIP(6A); CDC-6; and CDC-6m5CDK with each gene in both ppd49.78 and ppd49.83 vectors (40 µg/ml for each plasmid) together with prf4 (40 µg/ml) and pbs SK- (added to bring the total DNA concentration to 200 µg/ml). Injected hermaphrodites were allowed to lay eggs for 12 hrs at 25ºC, then the eggs were heat-shocked at 33ºC for 30 min, and harvested 12 hrs later. The percentage of viable transgenic progeny was calculated as the number of roller progeny divided by the total number of presumed transgenic progeny (the number of roller progeny plus the number of progeny that arrested as eggs or L1 larvae). The total number of transgenic progeny for the conditions listed in the Fig 2-9A graph (from top to bottom): 643; 378; 334; 51; 26; 116; 261; 200; and

60 RNAi RNA-mediated interference (RNAi) was accomplished either by injecting dsrna at a concentration of mg/ml or by feeding L4 larvae or young adults with HT115 bacteria that expressed dsrna [89, 204]. cul-4, and cki-1 dsrna for injection were synthesized from cdna clones yk34c8, and yk490e9 respectively. Sense and antisense RNAs were created with T3 and T7 MegaScript kits (Ambion) and annealed to generate dsrna. F1 progeny were analyzed for RNAi phenotypes. Antibodies and immunofluorescence We attempted to generate anti-phospho-specific CDC-6 antibodies against three potential phosphorylation sites: Ser 109, Thr 131, and Thr 145. Sera were generated in rabbits against the following phospho-peptides conjugated to mcklh: KLEISS 109 PRARGRC; TPRT 131 PEKQSRSRSC; and GSQGT 145 PEKKSC (with the numbered residues phosphorylated). The anti-phospho-cdc-6 antibodies were precleared against CDC-6 peptide lacking phosphorylation residues and affinity purified with CDC-6 peptide containing phospho-residues. As shown in Fig 2-4B and C, affinity purified anti-phospho-t 131 -CDC-6 specifically recognized phosphorylated CDC-6. The other two antibodies (against phospho-s 109 and phospho-t 145 ) did not specifically recognize phosphorylated CDC-6 on western analysis or by immunofluorescence (data not shown). Other primary antibodies used were: anti-spd-2 [187]; anti-cki-1 [96]; and anti-ajm-1 (MH27, Developmental Studies Hybridoma Bank) [205]. Secondary antibodies used were: anti-mouse Rhodamine (Cappel); and anti-rabbit Alexa Fluor 488 and 546 (Molecular Probes). Immunofluorescence was performed with the freezecrack method, followed by methanol and acetone fixation [206]. DNA was stained with 1 µg/ml 4,6-diamidino-2-phenylindole (DAPI). To analyze timed CDC-6 expression, 49

61 pretzel-stage embryos were collected, and hatched larvae were transferred every 15 min to plates with OP50 bacteria, as described [116]. Microscopy Animals were observed with a Zeiss Axioskop microscope equipped for differential interference contrast (DIC) and fluorescence microscopy. Images were captured with a Hamamatsu ORCA-ER digital camera with Openlab software (Improvision). Images were processed using Adobe Photoshop 7.0. Matched images were taken with the same exposure and processed identically. In Fig 2-2C, the following number of V1-V6 cells were observed: min, n=6; min, n=16; min, n=11; min, n=16-17; min, n=12-19; min, n=4; min, n=5-7. In Fig 2-6, the following number of V1-V6 seam cells were observed for strains expressing CDC-6::GFP transgenes: wild-type CDC-6::GFP, n=21-30; mutant CDC-6m48::GFP, n=27-31; m93, n=28-37; m131, n=16-26; m145, n=24-26; m48.131, n=26-35; m , n=39-47; m , n=53-65; m , n=9-20; and m , n= The scale of nuclear vs. cytoplasmic localization in Fig 2-2C and 2-6B was derived by assessing for each cell whether there was predominantly nuclear localization (assigned a score of -5), equivalent nuclear and cytoplasmic localization (a score of 0), or predominantly cytoplasmic (a score of +5); the average score for all cells examined was placed on the graph with a scale from -5 to +5. Quantitation of DNA content was performed as described [207, 208]. The DNA contents of nuclei were standardized to the DNA content in 2n cells (2C DNA content) in the same images, either post-mitotic nuclei or telophase nuclei. 50

62 ACKNOWLEDEGMENTS I would like to thank H. Feng for initiating the researches of CDC-6 tranlocalization by CUL-4, generating CDC-6 and phospho-t 131 -CDC-6 antibodies, and contributing to the immunofluorescence results presented in Fig 2-4B. We are grateful to Yuji Kohara for cdna clones; Roger Y. Tsien for tdtomato cdna; Richard Roy and Andrew Z. Fire for plasmids; Sudhir Nayak for the strain expressing histone H1::GFP; Kevin F. O Connell for anti-spd-2 antibody; Chris Dowd for technical assistance; and the Caenorhabditis Genetics Center for strains. This work was supported by grant R01GM from the National Institute of General Medical Sciences (NIGMS) (National Institutes of Health [NIH]) to E.T.K. 51

63 Figure 2-1. CDC-6 is exported from the nucleus during S phase in a CUL-4- dependent manner. CDC-6 translocates from the nucleus to the cytoplasm during S phase in wild-type larvae (top panels), but remains nuclear in cul-4(rnai) larvae (bottom panels). CDC-6:: GFP expressed from the wrt-2 promoter was observed at min post-hatch (G1 phase; top right panels) and at min post-hatch (S/G2 phase; lower right panels). Prnr- 1::tdTomato was used as an S phase marker (left panels). Scale bars, 10 µm. Figure from [209] 52

64 53

65 Figure 2-2. CDC-6::GFP localization throughout the cell cycle. (A) Dynamic localization of CDC-6::GFP through the cell cycle. Representative pictures of seam cells at the indicated cell cycle phases are presented. Scale bar, 10 µm. (B) CDC-6 localizes with mitotic chromosomes. Epifluorescence images of CDC- 6::tdTomato and histone H1::GFP in a metaphase cell. Scale bar, 5 µm. (C) Time course of CDC-6 translocation from the nucleus to the cytoplasm in V1-V6 seam cells. Graph of nuclear vs. cytoplasmic localization of CDC-6::GFP or CDC- 6::tdTomato at times post-hatch. Error bars reflect standard error of the mean (SEM). See Material and Methods section for the number of cells analyzed. (D) CDC-6::GFP translocation to the cytoplasm in the V1-V6 seam cells. Epifluorescence images of CDC-6::GFP in V1-V6 seam cells that are in G1 phase ( min post-hatch) or S/G2 phase ( min post-hatch). The QR cell and its descendents (QR.a and QR.p) have different cell cycle timing. Scale bar, 10 µm. Figure from [209] 54

66 55

67 Figure 2-3. Nuclear localization of a CDC-6 mutant with intact CDK phosphorylation sites. Pwrt-2::CDC-6 m nls::gfp is retained in the nucleus in both G1 phase ( min post-hatch; top left panel) and S/G2 phase ( min post-hatch; bottom left panel). Prnr-1::tdTomato is presented as an S phase marker (right panels). Scale bar, 10 µm. Figure from [209] 56

68 57

69 Figure 2-4. CDC-6 is phosphorylated on CDK sites in a CUL-4-dependent manner during S phase. (A) Schematic representation of human and C. elegans CDC-6 orthologs. Potential CDK phosphorylation sites are shown as black bars labeled S (serine) or T (threonine); NLSs as yellow boxes; and Walker A and B domains as green boxes. In humans, the phosphorylation of S 54, S 74, and S 106 (red) near the two N-terminal NLSs are essential for Cdc6 nuclear export [51, 52]. C. elegans CDC-6 has two consensus CDK sites (S/T-P- X-K/R; purple lettering) and seven minimum CDK phosphorylation sites (S/T-P; blue). Scale bar, 100 amino acids. (B) CDC-6 T 131 -phosphorylation in wild-type and cul-4(rnai) larvae in G1 and S/G2 phase. Wild-type (top) and cul-4(rnai) (bottom) larvae were observed in G1 phase (upper panels) and in S phase (lower panels). Animals were stained with anti-phospho- T 131 -CDC-6 (green), DAPI (blue), and anti-ajm-1 (red overlay, staining adhesion junctions to indicate seam cell boundaries [205]). Scale bar, 10 µm. (C) CDC-6 T 131 -phosphorylation in cul-4(gk434) (top) and cul-4(gk434), cki-1(rnai) (bottom) seam cells in S phase ( min post-hatch). cki-1(rnai) in cul-4(gk434) restored the phosphorylation of CDC-6 in S phase. Figure from [209] 58

70 59

71 Figure 2-5. The subcellular localization of the CDC-6mCDK::GFP mutant during G1 and S/G2 phase. Pwrt-2::CDC-6mCDK::GFP remains nuclear-localized in G1 phase ( min posthatch; top panels) and S/G2 phases ( min post-hatch; bottom panels). Prnr- 1::tdTomato is shown as an S phase marker (right panels). Due to the use of the wrt-2 promoter, CDC-6mCDK::GFP is only observed in the V1-V6 seam cells and Q cells, while the rnr-1 promoter has a wider expression in all S phase cells. Scale bar, 10 µm. Figure from [209] 60

72 61

73 Figure 2-6. The subcellular localization of CDC-6 CDK-phosphorylation site mutants. (A) Subcellular localization of wild-type and CDK-site mutant CDC-6::GFP proteins expressed from the wrt-2 promoter. Localization in V1-V6 seam cells during S/G2 phase ( min post-hatch) is plotted on a continuum from nuclear localization to cytoplasmic localization. Numbers on the left side indicate the location of serine or threonine residues replaced with alanine in the mutants. Error bars represent SEM. See Materials and Methods for the number of cells analyzed. (B) Epifluorescence images of CDK-site mutant CDC-6::GFP proteins in V1-V6 seam cells at min post-hatch (S/G2 phase). Numbers in the upper left corner indicate the location of serine or threonine residues replaced with alanine. In the CDC- 6m image, an inset shows a longer exposure of the V6 seam cell, whose expression was too weak to view with the normal exposure. Scale bar, 10 µm. Figure from [209] 62

74 63

75 Figure 2-7. CDT-1 perdurance does not affect CDC-6 nuclear export. (A) Schematic representation of C. elegans CDT-1. CDK consensus phosphorylation sites are shown as black bars; cyclin binding domain as purple bars; and the PIP box as a red box, which is predicted to bind to PCNA. The sequences of the PIP box motif in wild type, and 3A and 6A mutants are shown below the red box. In the wild-type PIP box sequence, red letters represent critical residues that are conserved between C. elegans and vertebrate PIP box sequences [108, ]. In the mutant PIP box sequences, substituted alanine residues are colored blue. (B) CDT-1mCDK+PIP(3A)::GFP is stabilized during S phase. Pwrt-2::CDT-1::GFP (left panels) and Pwrt-2:: CDT-1 mcdk+pip(3a)::gfp (right panels) were each expressed in a strain that contains AJM-1::GFP, which highlights seam cell boundaries. Larvae were observed at min post-hatch (G1 phase, upper panels) and at min posthatch (S phase, lower panels). (C) CDC-6::tdTomato is exported in the presence of stabilized CDT-1::GFP. A seam cell at min post hatch (S/G2 phase) that expresses Pwrt-2::CDC-6::tdTomato (red) and Pnhr-168::CDT-1mCDK+PIP(3A) ::GFP (green); AJM-1:: GFP (green) marks seam cell boundary. Scale bars, 10 µm. Figure from [209] 64

76 65

77 Figure 2-8. cki-1 RNAi rescues the failure of CDC-6 nuclear export in cul-4 mutant cells. (A) cki-1(rnai) suppresses the re-replication phenotype of cul-4(gk434) mutants. DIC images of the lateral hypodermis of: wild-type L2 larvae; cul-4(gk434) 2.5-day arrested L2 larvae; and cul-4(gk434), cki-1(rnai) 2.5-day arrested L2 larvae. Arrows indicate seam cells. (B) cki-1(rnai) restores CDC-6 nuclear export in cul-4(gk434). Pwrt-2::CDC-6::GFP was observed in cul-4(gk434) (upper panels) and cul-4(gk434), cki-1(rnai) (lower panels) at min post-hatch (S/G2 phase). For both genetic backgrounds, the expression of Pwrt-2::CDC-6::GFP was nuclear in G1 phase (data not shown). Scale bars, 10 µm. (C) The subcellular localization of CDC-6mCDK::GFP in cul-4(gk434) and cul-4(gk434), cki-1(rnai) seam cells. Pwrt-2::CDC-6mCDK::GFP remains nuclear-localized in cul- 4(gk434) (top panels) and cul-4(gk434), cki-1(rnai) (bottom panels) at min post-hatch (S phase). Prnr-1::tdTomato is shown as an S phase marker. In both genotypes, CDC-6mCDK remained nuclear during S phase in 100% of seam cells [n = 90 for cul-4(gk434), and n = 47 for cul-4(gk434), cki-1(rnai)]. Scale bar, 10 µm. Figure from [209] 66

78 67

79 Figure 2-9. Expression of deregulated CDT-1 and CDC-6 causes DNA rereplication. (A) Graph of viability upon overexpression of wild-type or deregulated CDT-1 and CDC-6 transgenes. CDT-1 wt and CDC-6 wt are wild-type genes; CDT-1 mut is CDT- 1mCDK+PIP(6A); and CDC-6 mut is CDC-6m5CDK. Each CDC-6 or CDT-1 gene was expressed under the control of both the hsp16-41 and hsp16-21 heat-shock promoters. Embryos from injected hermaphrodites were incubated at 25 C (a semi-permissive temperature for hsp expression [210]); and heat-shocked for 30 min at 33 C 12 hrs prior to harvest. The percentages of viable progeny are listed to the right of each bar in the graph. (B) Expression of non-degradable CDT-1 and non-exportable CDC-6 induces DNA rereplication. Epifluorescence images of a wild-type embryo and transgenic embryos expressing either CDT-1mCDK+PIP(6A) plus wild-type CDC-6 or CDT-1mCDK+PIP(6A) plus CDC-6m5CDK. Animals were stained with DAPI (blue) and anti-spd-2 (red), which highlights centrosomes [187]. Arrows indicate pairs of centrosomes in cells with increased DNA content (DNA content for the cell on the left is 13.9 C and the right cell is 11.1 C). Scale bar, 10 µm. Figure from [209] 68

80 69

81 Figure Model for the regulation of the CDT-1 and CDC-6 replication licensing factors by the CUL-4 ubiquitin ligase in C. elegans Upon entry into S phase, the CUL-4/DDB-1 ubiquitin ligase complex directly targets CDT-1 for ubiquitin-mediated proteolysis. The CUL-4/DDB-1 complex indirectly stimulates CDC-6 nuclear export by negatively regulating the levels of the CDK-inhibitor CKI-1. A reduction of CKI-1 levels allows the activity of CDK-cyclin complexes, which phosphorylate CDC-6 at multiple N-terminal CDK sites to inactivate NLSs and trigger CDC-6 nuclear export. The two pathways are redundant, and the deregulation of both CDT-1 and CDC-6 pathways is required to allow cells to undergo DNA re-replication. Figure from [151] 70

82 71

83 CHAPTER 3 ANALYSIS OF C. ELEGANS CAND-1 ENHANCERS INDENTIFIED BY A GENOME-WIDE RNAI SCREEN 3 3 Jihyun Kim and Edward T. Kipreos To be submitted 72

84 BACKGROUND The ubiquitin-mediated proteolysis pathway is an enzyme cascade that allows selective protein degradation and involves the covalent transfer of ubiquitin to a target protein. Cullins function as a rigid backbone that brings the E2 ubiquitin-conjugating enzyme together with substrates. Through the process known as neddylation, cullins are post-translationally modified by Nedd8, which is a ubiquitin-like protein [211]. Nedd8 conjugation enhances cullin-ring ubiquitin ligase (CRL) activity through the recruitment of the E2 conjugating enzyme to the complex, possibly facilitating cullin heterodimer formation, and also a conformational change of the CRL-E2 complex promoting transfer of ubiquitin to substrates [120, ]. The Nedd8 conjugation modification is reversible via the process called deneddylation. The COP9 Signalosome (CSN), composed of at least eight subunits, is a conserved multiple subunit complex. CSN removes Nedd8 conjugation from cullins through the isopeptidase activity of the metalloprotease CSN5/Jab subunit of CSN [130, 131, 212]. Cullin-Associated and Neddylation-Dissociated (CAND1) is a CRL inhibitor that interacts with unneddylated cullin-rbx1 complexes lacking adaptor proteins and Substrate-Recognition Subunits (SRSs) [140, 141, 143, 213]. Human CAND1 is a 136- kda protein composed of multiple HEAT repeats. The crystal structure shows that human CAND1 wraps around the CUL1-Rbx1 complex, with its N-terminus binding to the C-terminus of the cullin and its C-terminus binding to the N-terminus of the cullin [142]. Binding of CAND1 to CUL1 precludes adaptor proteins from interacting with the N-terminus of CUL1 and Nedd8 from attaching to the C-terminus. Initial biochemical studies propose that CAND1 acts as a negative regulator of the CRL complex assembly by sequestering cullins. However, genetic approaches and in vivo studies show that CAND1 functions as a positive regulatory element of CRL activity. In genetic studies, cand1 mutants reduced SCF activity in Arabidopsis and SCF and CRL3 activity in 73

85 humans [146, 147, 214, 215]. It was shown that SCF substrates accumulate in cand1 mutants because substrate recognition subunits (SRSs) undergo autoubiquitylation [146, 214]. Biochemical and genetic studies suggest that CAND1 is required for optimal CRL function in vivo. How CAND1 regulates CRL activity is still mysterious and multiple aspects of CAND1 activity are not clear. Endogenous CUL1 can be dissociated from CAND1 upon Nedd8 conjugation [141]. It was proposed that neddylation serves as a signal for CAND1 dissociation from cullin-rbx1 and subsequent assembly of the CRL complexes [141], although the precise mechanisms leading to CAND1 release remain ill-defined. More recent biochemical and structural studies suggest that Nedd8 conjugation occurs after SCF complex assembly and the binding of CAND1 to cullin-rbx1 blocks neddylation of cullins [142, 216], raising new questions about the regulation of CAND1/cullin-Rbx1 interactions. Another study suggested that an adaptor-srs complex (Skp1-Skp2), but not the adaptor alone, triggers dissociation of CAND1 from endogenous CUL1 [216]. However, the precise mechanism of CAND1 dissociation from cullin-rbx1 is still unknown. CAND1 is released from cullin-rbx1 upon neddylation or binding of an adaptor-srs complex only in vivo experiments, suggesting that additional modification or an additional factor may be required for dissociation of CAND1 from cullin-rbx1 [118]. How cullin-rbx1 is released from CAND1 is a major open question in understanding the regulation of CRL activity. To understand how CAND1 regulates CRL activity in C. elegans, our lab has been studying CAND-1, the C. elegans homolog of CAND1. Consistent with other studies in vertebrates, C. elegans CAND-1 physically interacts with a substantial percentage of CUL-2::FLAG and CUL-4::FLAG in vivo (unpublished data). The yeast two hybrid system shows that CAND-1 can directly bind to all six C. elegans cullins (D.R. Bosu and ETK unpublished). Immunofluorescence with a CAND-1 antibody reveals that CAND-1 is expressed in dividing cells of the early 74

86 embryo and in the proliferating tissues of larvae, which is consistent with the cell cycle regulatory functions of CUL-1, -2, and -4 (H. Feng and ETK unpublished). We obtained the cand-1(tm1683) deletion allele from the National Bioresource Project in C. elegans (Japan) to further investigate the functions of CAND-1. cand- 1(tm1683) mutants are viable and can be maintained as homozygotes. RT-PCR and western analysis show that cand-1(tm1683) mutants have low levels of CAND-1 proteins and that this allele is a hypomorphic mutant (D.R. Bosu and ETK unpublished). Western analysis reveals that levels of unneddylated CUL-2 and CUL-4 decrease in cand-1 mutants (D.R. Bosu and ETK unpublished). Some other phenotypes of the cand- 1(tm1683) allele include adults that produce approximately 70% less eggs than wild type as well as a developmental arrest phenotype. 25% of cand-1(tm1683) mutants arrest as mid-to-late stage embryos, and 10% arrest as L2-stage larvae, with the remaining ~65% becoming adults. A certain percentage of cand-1(tm1683) mutants produce protruding vulva and defective tail morphologies. Additionally, the timing of development is lengthened for cand-1(tm1683) mutants relative to wild type (D.R. Bosu and ETK unpublished). Since the cand-1(tm1683) allele is hypomorphic, cand-1(tm1683) mutants fed with cand-1 RNAi have more severe phenotypes: a higher percentage of these animals undergo embryonic or larval arrest. However, cand-1(tm1683) mutants treated with cand-1 RNAi are still viable, indicating that cand-1 is not an essential gene for viability (D.R. Bosu and ETK unpublished). We want to identify factors involved in the cand-1 regulator pathway or cand-1 parallel pathway to regulate CRL activity that do not physically interact, so that a genome-wide RNAi screen was performed. An RNAi screen is less labor intensive than other approaches and C. elegans genes are easily inactivated by feeding bacteria expressing dsrna (feeding RNAi), but its major advantage is to identity the interacting gene immediately. The Ahringer RNAi feeding bacteria library, which contains 75

87 approximately clones (86% of the C. elegans genome) has been used for the RNAi screen [217]. cand-1 screening with this library has been done in liquid format in 96-well plates by D.R. Bosu [218]. Another graduate student (D.R. Bosu) completed an RNAi screen of three of the six C. elegans chromosomes: III, IV, and X. This screen covered 7182 genes, which is 43% of the RNAi library. We identified 18 enhancers whose RNAi inactivation reproducibly negatively impacts cand-1 mutants without affecting wild-type animals. I have been investigating two enhancers (dli-1 and hpk-1) among the eighteen enhancers of cand-1(tm1683) for further analysis by certain criteria: 1) CAND-1 protein levels; 2) enhancer specificity and phenocopies of cullins; and 3) levels of CUL-2 neddylation. The absence of dynein in wild-type backgrounds reduces unneddylated CUL-2 proteins and phenocopies cul-2 phenotypes: defects in pronuclei migration; multinuclei; meiotic defects; and cytoplasmic extensions. hpk-1(pk1393) null mutants are viable without noticeable defects in morphology or development [219]. However, cand-1(tm1683) hypomorphic mutants treated with hpk-1(rnai) produce approximately 100% embryonic lethality. Dynein Cytoplasmic dynein is the major minus-end-directed microtubule motor of eukaryotic cells. Cytoplasmic dynein is composed of a dimer of heavy chains (HCs), along with several accessory chains (ACs: intermediate, light intermediate, and light chains) [220]. Dynein serves a role in retrograde axonal transport and is involved in diverse cellular processes including spindle assembly and function, neuronal transport, and organelle positioning [220]. In C. elegans, dhc-1 encodes a cytoplasmic dynein 1 HC, while 11 other genes encode five classes of predicted dynein ACs [221, 222]. 76

88 DHC-1 (Dynein Heavy Chain) dhc-1 RNAi causes defects in female meiotic divisions, migration of the oocyte and sperm pronuclei after fertilization, and centrosome separation during mitotic spindle assembly [223]. A study with dhc-1 temperature-sensitive (ts) mutants shows that dhc-1 is required for chromosome congression to the metaphase plate during mitosis, as well as for mitotic spindle positioning [224]. DLI-1 (Dynein Light Intermediate chain) DLI-1 is required for multiple dynein-dependent functions: pronuclear migration, centrosome separation, and meiotic and mitotic spindle function [225]. dli-1 mutants exhibit dhc-1-like phenotypes similar to a weak allele of dhc-1. For example, dhc-1 mutants exhibit numerous female pronuclei but this defect is not observed in dli-1 mutants [225]. DLC-1 (Dynein Light Chain) dlc-1 encodes a dynein light chain 1. An RNAi screen of early embryogenesis showed that the dlc-1 RNAi phenotype is similar to that of dli-1 RNAi [226]. dlc-1 is not well characterized. DYLT-1 (DYnein Light chain [Tctex type]) dylt-1 encodes a Tctex1-type dynein light chain and dylt-1 null mutants are viable. It was shown that dylt-1 RNAi could suppress the viability of three ts dhc-1 mutants. The mechanism for this suppression is not known [227]. It is also not clear if dylt-1 functions as a positive or negative regulatory element for the dynein complex. 77

89 DYRB-1 (DYnein light chain, RoadBlock type) dyrb-1 is required for completion of meiosis and pronuclear migration [227]. dyrb-1(rnai) in wild-type backgrounds does not cause any visible defects, but dyrb-1 RNAi suppresses embryonic lethality in all three ts dhc-1 mutants [227]. How dyrb- 1(RNAi) suppresses dhc-1 viability is not known. A recent study of a dyrb-1 deletion mutant shows that the dyrb-1 deletion phenocopies a weak loss of dynein activity and plays a role in the regulation of spindle positioning [228]. It is proposed that dyrb-1 plays a positive role in the regulation of dynein activity along with other regulators. DYCI-1 (DYnein Chain, light Intermediate) dyci-1 is required for pronuclear migration [226, 227]. It was shown that dyci-1 RNAi does not suppress dhc-1 ts viability in contrast to dynein light chains such as dyrb- 1 [227]. dyci-1 is not well characterized. LIS-1 (human lissencephaly gene related) lis-1 interacts with cytoplasmic dynein (dynein-interacting protein) and is required for all known dynein-dependent processes: pronuclear migration, centrosome separation and spindle assembly, as well as dynein-mediated transport of yolk granule cargo in C. elegans one-cell-stage embryos [229]. HPK-1 (Homeodomain-interacting Protein Kinase-1) C. elegans HPK-1 is a homolog of the mammalian homeodomain-interacting protein kinases HIPK1 ~ HIPK4 that belong to a novel family of serine - threonine kinases. HIPK2 has been the most extensively studied. HIPK2 is a multifunctional coregulator that either interacts with homeobox proteins and act as corepressors, or interacts with other types of transcription factors and act as coactivators or corepressors, 78

90 depending on the promoters and the cellular context [230]. HIPK2 regulates gene expression by phosphorylation of transcription factors and accessory components of the transcription machinery in response to morphogenic signals or DNA-damaging agents [230, 231]. HIPK2 plays roles in transcriptional regulation, chromatin remodeling, and key components of signaling pathways such as development- or damage-induced apoptosis (the p53 activation pathway) and TGFβ/BMP and Wnt signaling. HIPK2 regulates transcription during development HIPK2 activates the apoptotic function of p53 and antagonizes the activity of the anti-apoptotic transcription factor Brn3a [230, 231]. HIPK2 regulates Pax6, c-myb, AML1 (acute myeloid leukemia 1), and the histone acetyltransferase p300 in response to developmental signals [232, 233]. HIPK2 phosphorylates the activation domain of Pax6 and increases its transcriptional activity [232, 233]. Together with NLK (nemo-like kinase), HIPK2 inhibits c-myb transcriptional activity by promoting its phosphorylation upon stimulation of the Wnt-1 signaling pathway [232, 233]. HIPK2 promotes hematopoietic gene transcription by phosphorylating the transcription factor AML1 and the histone acetyltransferase p300 [232, 233]. Drosophila has a single HIPK ortholog which regulates the global corepressor Groucho during development [234]. Direct phosphorylation of Groucho by HIPK leads to Groucho being subsequently released from the corepressor complex. Dissociation of Groucho relieves transcriptional repression of Groucho target genes [234, 235]. HIPK2 regulates transcription in the cellular response to DNA damage HIPK2 exerts its effects on DNA damage response by binding and/or phosphorylating a large array of transcription factors and coregulators. HIPK2 regulates p53 localization, phosphorylation, acetylation, and transcriptional activity [232, 233]. 79

91 HIPK2 phosphorylates p53 and enhances the p53-mediated transcriptional activation of pro-apoptotic factors such as PIG3, BAX, NOXA, and p53aip1, as well as the repression of anti-apoptotic factor Galectin-3. Additionally, HIPK2 modulates the transcriptional activity of the p53 family members p63 and p73 [236]. HIPK2 regulates the stability of CtBP (transcriptional corepressor C-terminal binding protein), which functions in antiapoptosis. Upon UV irradiation, activated HIPK2 phosphorylates CtBP at Ser422 and targets CtBP for proteasomal degradation [237]. This elimination of CtBP by proteolysis also plays a role in promoting apoptosis. RESULTS hpk-1 and dli-1 are cand-1 enhancers Seventeen out of eighteen enhancers identified by a genome-wide RNAi screen have visible RNAi phenotypes in wild-type or a RNAi-hypersensitive genetic background (D.R. Bosu and ETK unpublished). hpk-1 is the only enhancer that does not exhibit any noticeable phenotypes in either genetic background. However, hpk-1 RNAi in a cand- 1(tm1683) background causes 100% late-stage embryonic lethality. Embryos of cand- 1(tm1683) mutants fed with hpk-1 RNAi arrest as post-lima bean stage in which morphogenesis starts and the first muscle twitches are observed at 430 min after first cell cleavage (between 1.5 and 2-fold stages) embryos with abnormal morphology. dli-1, a dynein component, was identified as another enhancer. C. elegans dynein is composed of DHC-1, DLC-1, DLI-1, DYCI-1, DYRB-1, and DYLT-1 and has several dynein-interacting proteins, such as LIS-1 (human lissencephaly gene related). All of the dynein components and LIS-1 have similar phenotypes that phenocopy cul-2 mutants: defects in pronuclei migration, multinuclei, meiotic defects, and cytoplasmic extensions, [221, 222, 226, 227]. Since dli-1 was identified as an enhancer, I examined if other dynein components were cand-1 enhancers. cand-1(tm1683) mutants combined 80

92 with any dynein component RNAi exhibited a more severe phenotype than either cand- 1(tm1683) or dynein RNAi alone but are otherwise indistinguishable from each other. cand-1(tm1683) mutants treated with dynein RNAi grow more slowly and have enlarged germ cells that are not observed in cand-1(tm1683) or dynein RNAi itself (data not shown). HPK-1 and dynein do not affect CAND-1 protein levels The simplest theoretical explanation for cand-1(tm1683) enhancers is that they are required for CAND-1 protein stability. Such enhancers would not provide further insights into how CRL activity is regulated and would not be especially promising genes to study. To test if cand-1 enhancers regulate CAND-1 protein stability, CAND-1 protein levels were examined by western analysis after RNAi-treatment of the eighteen genes identified as cand-1 enhancers. Depletion of any enhancers, including components of dynein and hpk-1, by RNAi does not change endogenous CAND-1 protein levels (Fig 3-1A; data not shown). Depletion of dynein or hpk-1 by RNAi does not alter interaction between CUL-2 and CAND-1 Although it was shown how CAND-1 blocks the assembly of CRL complexes, the precise mechanism of CAND-1 dissociation from cullin-rbx1 complexes is still mysterious. One of the primary purposes of the cand-1 RNAi screen is to figure out how cullin-rbx1 is released from CAND-1. It is possible that additional post-translational modifications of the CAND-1/CUL-Rbx-1 complex or additional components trigger dissociation of CAND-1. If enhancers identified by RNAi screening are involved in the process of CAND-1 dissociation, inactivation of these enhancers would cause a severe 81

93 phenotype in cand-1(tm1683) backgrounds and strengthen interaction between CAND-1 and cullins in wild-type backgrounds. To examine if hpk-1 or dynein regulates association of CAND-1 with cullins, we used two different genotypes of transgenic worms expressing either CUL-2::FLAG or CUL-4::FLAG. We immunoprecipitated CUL-2::FLAG or CUL-4::FLAG proteins both in wild-type and dynein- or hpk-1-depleted backgrounds with FLAG antibodies and observed how much endogenous CAND-1 associated with CUL-2::FLAG or CUL- 4::FLAG by western analysis with CAND-1 antibodies. Upon dhc-1 or hpk-1 RNAi, no significant difference of CAND-1 binding to CUL-2::FLAG or CUL-4::FLAG was observed, indicating that dynein and hpk-1 do not affect the steady-state interaction between CAND-1 and either cullin (Fig 3-1B; date not shown). Dynein, not hpk-1, changes levels of CUL-2 neddylation We want to know how CRL activity is controlled and if additional undiscovered pathways regulate CRL activity. One criteria for characterizing enhancers is if they affect CRL activation. Neddylation stimulates the activity of CRL [ ] and cand- 1(tm1683) mutants have reduced levels of unneddylated CUL-2 relative to wild-type animals (D.R. Bosu and ETK unpublished). To study how enhancers affect the activity of CRL complexes, cullin neddylation levels were examined via western analysis. Enhancers that regulate the activity of CRL independently of CAND-1 would be expected to affect cullin neddylation levels in both wild type and cand-1(tm1683) backgrounds upon treatment with enhancer RNAi. Upon inactivation of the enhancers required for optimal CAND-1 activity, the proportion of neddylated cullin in cand-1(tm1683) mutant backgrounds would increase, reflecting additional loss of CAND-1 activity. In the case that any effects these enhancers have on neddylation are not visible, a wild-type and a RNAi-hypersensitive genetic background, such as rrf-3 mutant would 82

94 be used to test levels of Nedd8 upon inactivation of enhancers by RNAi. Effects of enhancers on the neddylation levels of endogenous CUL-2 in wild-type and cand- 1(tm1683) backgrounds were examined by western analysis with CUL-2 antibodies. Most enhancers, including hpk-1, do not change the levels of CUL-2 protein or the levels of neddylated CUL-2 (D.R. Bosu and ETK unpublished; data not shown). This suggests that hpk-1 may be involved in another CRL regulatory pathway rather than the Nedd8 conjugation pathway. Upon inactivation of dli-1 (dynein light intermediate chain) by RNAi, levels of neddylated CUL-2 vary but the apparent reduction of unneddylated CUL-2 is similar in the presence or absence of CAND-1 (wild-type and rrf-3 backgrounds) (Fig 3-1C; data not shown). C. elegans dynein components (DHC-1, DLC-1, DLI-1, DYCI-1, DYRB-1, and DYLT-1) and dynein-interacting proteins, such as lis-1, share dynein-dependent pathways and have similar phenotypes to cul-2 mutants [221, 222, 226, 227]. Levels of unneddylated CUL-2 in the absence of all dynein components or a dynein-interacting protein were examined by western analysis. Upon RNAi treatment of any dynein component or lis-1, levels of unneddylated CUL-2 decrease as seen in dli-1(rnai) (Fig 3-1C, D). This may be due to protein degradation of unneddylated CUL-2 or decreased gene expression, and either mechanism would imply that the dynein complex regulates the abundance of CUL-2. Dynein does not regulate ectopically expressed CUL-4 protein Upon dynein RNAi, a reduction of unneddylated CUL-2 was observed, suggesting the possibility that dynein complexes may play a global role in regulating cullin stability. We want to know if dynein regulates other cullins besides CUL-2. Other cullin antibodies are not available, so I used transgenic worms expressing C-terminal 83

95 FLAG fused to CUL-4. Levels of exogenously expressed CUL-4 protein were tested by western analysis with a FLAG antibody. In contrast to CUL-2, levels of neddylated or unneddylated CUL-4 were not noticeably changed upon inactivation of dynein components (Fig 3-1E). In contrast to western analysis with CUL-2 antibody, we observed several bands in our anti-flag western analysis due to phosphorylation (unpublished data). CUL-4 may have an additional post-translational modification or this pattern may be specific to ectopically expressed cullins. Dynein does not alter CUL-2 or CAND-1 subcellular localization Dynein complexes serve a role in retrograde transport [220]. We hypothesized that dynein may regulate only a subset of CUL-2 protein (i.e. nuclear localized CUL-2 or cytoplasmic localized CUL-2). Although overall CAND-1 protein levels do not change in dynein(rnai), dynein may alter the subcellular distribution of CAND-1 which may in turn affect CUL-2 stability. To determine the subcellular distribution of endogenous CUL-2 or CAND-1, embryos of wild type and dynein mutants were examined by immunofluorescence. The dhc-1(or195ts) mutant is used to eliminate maternal effects at non-permissive temperature. Homozygous dhc-1(or195ts) mutants produce healthy embryos at 15ºC but generate 100% dead embryos at 26ºC due to loss of dynein function [238]. Immunostaining and quantification of nuclear versus cytoplasmic CUL-2 or CAND-1 staining in wild type or dhc-1(or195ts) mutants shows strong nuclear staining for both genes and the staining patterns were not changed in the different genetic backgrounds (Fig 3-2A, B), suggesting that dynein does not regulate a subset of CUL-2 or CAND-1 proteins. To determine whether the absence of cul-2 alters the subcellular distribution of dynein, LIS-1 localization was examined by immunofluorescence with LIS-1 antibody (a 84

96 gift from Dr. P. Gonczy) [229]. As I expected, cul-2(rnai) does not change LIS-1 subcellular localization (date not shown). Depletion of dynein does not cause a failure to degrade CUL-2 substrates Inactivation of dynein components phenocopies cul-2 mutants in the one-cell stage embryo. I wanted to clarify if dynein phenotypes are due to reduced levels of CUL-2 activity. If decreased levels of CUL-2 activity cause dynein phenotypes, then we would expect to observe increased substrates of CUL-2 or other known cul-2 phenotypes (beyond the one-cell stage embryo phenotypes). In wild-type embryos, the mitotic cyclin CYB-1 undergoes degradation after the one-cell stage. In embryos of cul-2 or zyg-11 mutants, CYB-1 remains at high levels after the one-cell stage, suggesting that CYB-1 is a CUL-2 substrate [97, 98, 239]. Expression of CYB-1::GFP in embryos was examined to verify if dynein functions in a CUL-2-dependent manner. CYB-1::GFP expression disappears after the one-cell stage in dhc-1(rnai) and wild-type embryos [97, 98] (Fig 3-3A), suggesting that CUL-2 is still functional in dynein mutants. dynein(rnai) does not cause an accumulation of known CUL-4 substrates Although visible changes of ectopic CUL-4 levels were not detected in western analysis, it is possible that the CUL-4 protein is less stable and is not functional in the absence of dynein. Further, western analysis was performed with ectopically expressed CUL-4, whose behavior may be different from endogenous CUL-4. I wanted to test if the absence of dynein causes decreased CUL-4 activity, so that known CUL-4 substrates such as CDT-1 and CKI-1 would accumulate in dynein deletion mutants [ ]. Some of dyrb-1(tm2645) homozygotes from homozygote parents develop to late-stage larva or adults, while other alleles of dynein mutants produce arrested embryos. Embryonic defects in cul-4 mutants have not been reported. Therefore, I used dyrb- 85

97 1(tm2645) mutants to analyze the CRL4 CDT-2 substrates CDT-1 and CKI-1 levels in larvae. In larval animals of wild type or dyrb-1(tm2645) backgrounds, endogenous levels of CDT-1 and CKI-1 were analyzed by immunofluorescence. In dyrb-1(tm2645) mutants, CDT-1 or CKI-1 does not accumulate as seen in wild-type animals, suggesting that the loss of dynein does not affect CUL-4 function (Fig 3-3B; data not shown). dhc-1 suppressors do not restore levels of CUL-2 protein O'Rourke and colleagues [227] performed a genome-wide RNAi screen initially with a temperature-sensitive (ts) allele of dhc-1 at a semipermissive temperature. They identified 20 genes that suppress conditional dynein HC mutants but not other conditional mutants with unrelated defects [227]. Six deletion alleles available from the 20 specific suppressor genes are homozygous viable. ufd-2 (Ubiquitin Fusion Degradation [yeast UFD homolog]), rab-10 (RAB family), cua-1 (CU [copper] ATPase), dylt-1 and dyrb-1 are among them. ufd-2 encodes an E4 ubiquitin conjugation factor orthologous to S. cerevisiae Ufd2p which catalyzes multiubiquitin chain assembly [75]. rab-10 encodes a Rab-like GTPase that is a member of the Ras superfamily of small GTPases and is a key regulator of the endocytic recycling pathway [240]. cua-1 encodes a copper-transporting E1-E2 ATPase ortholog. dylt-1 (encoding a Tctel1-type light chain) and dyrb-1 (encoding a roadblock-type light chain) are two dynein components that suppress embryonic lethality in all three ts dhc-1 mutants [227]. These suppressors encode evolutionally conserved proteins that are linked to human disease. One suppressor was already shown to be involved in ubiquitin-mediated proteolysis pathways in other systems. This suggests the possibility that some suppressors may function in CRL regulation. The study of dhc-1 suppressors might give us insight into how dynein complexes regulate CUL-2 stability. 86

98 We obtained the available viable null dynein suppressor mutants and tested the levels of CUL-2 proteins in these mutants by western analysis. If a dhc-1 suppressor is involved in CUL-2 regulation, CUL-2 protein would be expected to remain at the same levels as wild type in a dhc-1 suppressor fed with dynein RNAi. However, all dhc-1 suppressor null mutants fed with dhc-1 RNAi still have reduced levels of unneddylated CUL-2 (data not shown), suggesting that dhc-1 suppressors impact DHC-1 functions through a pathway that is independent of DHC-1 regulation. HPK-1 expression pattern In mammals, HIPK2 has been the best-studied gene among HIPK1 ~ HIPK4. Drosophila has only one member of the HIPK family, HIPK, which has been characterized. C. elegans also has only one member of this family, HPK-1, but this gene is uncharacterized. Therefore, at first, we want to examine HPK-1 expression patterns in different genetic backgrounds. The HPK-1 antibody is not available, so we generated transgenic worms expressing C-terminal GFP fused to HPK-1 protein whose plasmid is a gift from Dr. O. Hobert [219]. As previously reported [219], HPK-1::GFP was broadly expressed during embryogenesis and larval stages, and nuclear puncta staining of HPK- 1::GFP was observed (Fig 3-4A, B, and C). At postembryonic stages, HPK-1::GFP is widely expressed especially in head and tail neural cells, and vulval cells (Fig 3-4A, B; data not shown). Most of cells that express HPK-1::GFP have intense nuclear puncta expression (Fig 3-4A, B). As animals develop, the number, intensity, and nuclear puncta are greatly reduced. The meaning of nuclear puncta has not been studied, but it was suggested that intensity and nuclear puncta might be linked to the activity of HPK-1 [219]. Since hpk-1 was identified as a cand-1 enhancer, we want to address the question if HPK-1 expression is changed in a cand-1(tm1683) background. HPK-1::GFP 87

99 expression in this background is not altered, suggesting that CAND-1 does not regulate HPK-1 expression (Fig 3-4C, D). Although CAND-1 does not directly regulate HPK-1 activity, it is possible that CRL complexes regulate HPK-1 expression due to a role of CAND-1 in CRL activity. We want to know if CRL complexes affect HPK-1 expression. HPK-1::GFP expression was examined upon RNAi depletion of cul-2 or ddb-1, which is a CUL-4 adaptor and has the same phenotype as cul-4 mutants. cul-2(rnai) or ddb- 1(RNAi) does not change the HPK-1::GFP expression pattern, suggesting that CRL complexes do not have an effect on HPK-1 expression (Fig 3-4D). cand-1(tm1683) mutants treated with hpk-1(rnai) do not exhibit cell fate changes in arrested embryos Embryos of cand-1(tm1683) hermaphrodites fed with hpk-1 RNAi arrested at a late stage of embryogenesis. We would like to determine if cell fates are defective and if these arrested embryos have normal tissues. Several types of antibody markers were used. Several bodywall muscle cell markers used to identify muscle tissues are: MH2 (which recognizes perlecan); MH4 (intermediate filament subunit); MH24 (vinculin); MH25 (integrin beta); and MH46 (myostatin) [241, 242]. MH5 is a hypodermal marker [241]. MH27 is a standard marker for studying cell morphology during C. elegans development that recognizes cell-junctions, and is used to examine cell membranes of the hypodermis, gut, and pharyngeal cells [241]. Using these antibody markers, tissues in cand-1(tm1683) embryos with hpk-1(rnai) were examined via immunofluorescence. It appears that prior to the bean stage (cell number remains at ~560 cells), early stage embryos in cand-1(tm1683) fed with hpk-1 RNAi have no significant morphology defects based on immunofluorescence with antibody markers. At late stage embryogenesis, arrested embryos have all tissues although these tissues are not properly organized (Fig 3-5; data not shown). This suggests that cell fate is not 88

100 defective and cell-fate change is not the reason for the embryonic lethality of cand- 1(tm1683) with hpk-1(rnai). HPK-1 and CAND-1 do not regulate the levels of UNC-37 in C. elegans Drosophila has only one member of the HIPK family, HIPK, which prevents the function of the global co-repressor Groucho by inhibitory phosphorylation [235]. Groucho regulates many conserved signaling pathways, including TGFβ, Wnt, and Notch through transcriptional repression in invertebrates and vertebrates [243]. UNC-37 is a Groucho homolog in C. elegans. Although UNC-37 has been characterized, HPK-1 has not been studied in C. elegans, and the role of HPK-1 in UNC-37 activity has not been reported. HPK-1 and UNC-37 are conserved proteins, so it is possible that C. elegans HPK-1 regulates UNC-37 as shown in Drosophila, and CAND-1 may be involved in this regulation. If both CAND-1 and HPK-1 independently inhibit the activity of UNC-37, the absence of both cand-1 and hpk-1 may cause the continuous repression of gene transcription by UNC-37. Constitutively active UNC-37 may lead to embryonic lethality in cand-1(tm1683) with hpk-1(rnai). If this is true, a hypomorphic allele of unc-37, unc- 37(e265) should rescue the embryonic lethality in cand-1(tm1683) fed with hpk-1 RNAi. cand-1(tm1683); unc-37(e265) double mutants are severely sick: they grow significantly slower and produce fewer progeny that either single mutant. cand-1(tm1683); unc- 37(e265) double mutants fed with hpk-1 RNAi exhibit the same percentage of embryonic lethality as cand-1(tm1683) with hpk-1(rnai), both are 100% embryonic lethal, suggesting that embryonic lethality in cand-1(tm1683) mutants with hpk-1(rnai) may be independent of UNC-37 activity. Although the hypomorphic allele of unc-37(e265) cannot suppress the embryonic lethality of cand-1(tm1683) with hpk-1(rnai) due to the complexity of UNC-37 roles, it is 89

101 possible that CAND-1 and HPK-1 regulate UNC-37 stability. UNC-37 antibodies (a gift from Dr. Miller) are available [244]. We examined the endogenous UNC-37 levels in arrested and wild-type embryos by western analysis and immunofluorescence. If hpk-1 inhibits UNC-37 activity by inhibitory phosphorylation, this modification of UNC-37 would not be detected in western analysis of hpk-1(pk1393) null mutants. In western analysis, two bands were observed in both wild type and hpk-1(pk1393) mutants, and the additional post-translational modifications of UNC-37, such as phosphorylation, were not observed even in wild type (Fig 3-6A). In western analysis and immunofluorescence, UNC-37 protein levels in wild type, cand-1(tm1683), and cand-1(tm1683) with hpk- 1(RNAi) were not significantly changed [3.3±0.6 for cytoplasm, 10.1±0.2 for nucleus (n=11) in wild-type versus 3.0±0.4 for cytoplasm, 0.8±0.04 for nucleus (n=23) in cand- 1(tm1683)] (Fig 3-6A, B). As shown by immunofluorescence, cand-1(tm1683) with hpk- 1(RNAi) does not alter the subcellular localization of UNC-37 either (Fig 3-6B). These results suggest that HPK-1 and CAND-1 do not regulate the levels of co-repressor UNC- 37 proteins. DISCUSSION Dynein is required for maintenance of unneddylated CUL-2 but a reduction of unneddylated CUL-2 does not cause the dynein mutant phenotype dynein RNAi causes a reduction of unneddylated CUL-2. However, I did not observe any influences on the stability of CUL-4. dynein mutants seem to phenocopy cul-2 phenotypes, suggesting that dynein may function through CUL-2. However, they do not share any pathways for regulating cell cycle progression. The reduced CUL-2 protein observed in dynein mutants is not severe enough to cause a loss of cul-2 function. Although CUL-2 levels decrease, CUL-2 still functions properly. Dynein phenotypes are not due to reduced CUL-2 protein but instead due to a failure of spindle 90

102 assembly and retrograde transport. The dynein mutant phenotype is not dependent on a reduction of CUL-2 protein. The decreased CUL-2 protein observed may only be a secondary consequence of the absence of dynein or dynein may have an indirect effect on CUL-2 levels. Dynein is not required for subcellular localization of CUL-2 or CAND-1 protein How dynein regulates CUL-2 stability is not clear. Although unneddylated CUL-2 protein decreases in the absence of dynein as shown by western analysis, significant changes in the subcellular distribution of CUL-2 were not detected. Although the dynein complex was identified as a cand-1 enhancer, inactivation of dynein does not alter subcellular localization of CAND-1 or the interaction between CAND-1 and cullins. It is likely that an increase in CUL-2 neddylation would be a secondary consequence of dynein inactivation and therefore that dynein does not have significant functions linked to CUL-2 stability. Absence of hpk-1 and cand-1 causes late stage embryonic lethality without affecting cullin stability hpk-1 RNAi itself does not cause any visible phenotypes, but hpk-1 RNAi in cand-1(tm1683) mutants causes late stage embryonic lethality. This suggests that hpk-1 is generally a dispensable gene, but it is essential for cand-1(tm1683) viability. Embryos of cul-4 mutants do not exhibit visible defects, but both cul-2 and cul-3 mutants arrest as a very early stage embryo [96, 105, 116, 245]. I did not observe any known cullin phenotypes in embryos of cand-1(tm1683) with hpk-1(rnai) or changes in cullin protein levels. The pattern of HPK-1 expression is not changed in the absence of either cand-1 or cullins. HPK-1 may not directly regulate CAND-1 or CRL activity. 91

103 Absence of hpk-1 and cand-1 does not change cell fate cand-1(tm1683) mutants with hpk-1(rnai) causes late rather than early stage embryonic lethality, suggesting that occurrences in early embryogenesis, such as pronuclei migration and meiosis that is dependent on CRL2 or CRL3 complexes, may not be defective. Studies in other systems have shown that the HIPK family is involved in regulation of signaling pathways, including TGFβ and Wnt signaling which are required for cell fate determination [230]. However, cell fates of arrested cand- 1(tm1683) embryos treated with hpk-1(rnai) are not changed, suggesting that C. elegans HPK-1 and CAND-1 have novel roles in embryogenesis. MATERIALS AND METHODS Nematode strains C. elegans strains were cultured as previously described [198]. The following strains were used: N2, wild type; cand-1(tm1683); EU828 dhc-1(or195ts) [238]; EK273 hpk-1(pk1393) [219]; ET382 unc-37(e262); cand-1(tm1683); NL2099 rrf-3(pk1426); ET342 him-8(e1489); ekex19[cul-2::flag(f8.6), prf4, rol-6(+)]; ET361 unc-119(ed3); ekis9[ppd49.75/pcul-4::cul-4::flag]; ET113 unc-119(ed3); ekis[ppd3.01b/cyb-1]; ET359 dyrb-1(tm2645); CB262 unc-37(e262); ET366 hpk-1(pk1393); ekis10[pbr132, prf4]; and ET362 cand-1(tm1683); ekex24[pbr132, prf4] RNA-mediated interference (RNAi) RNA-mediated interference (RNAi) was accomplished by feeding L1 larvae, L4 larvae, or young adults with HT115 bacteria that expressed dsrna [89, 204]. Ahringer RNAi feeding bacteria library was used for feeding worms [217]. Transgenic worms 92

104 Extragenic transgene lines were generated by microinjection according to standard methods [203]. pbr132 (HPK-1::GFP) which is the gift from Dr. Hobert [219] was co-injected with prf4 (100 µg/ml) into hpk-1(pk1393) null mutants. Extrachromosomal arrays were integrated using 4000 rad from a 137 Cs gamma radiation source [203]. To generate ET362 cand-1(tm1683); ekex24[pbr132, prf4], male progenies from ET343 (hpk-1(pk1393); ekex24[pbr132, prf4]) crossed with wild type were re-crossed with cand-1(tm1683). Immunofluorescence and microscopy Anti-CUL-2, anti-cand-1, anti-cdt-1, anti-cki-1, anti-lis-1 and anti-unc-37 were used as described previously [96, 209, 229, 244](D.R. Bosu and ETK unpulished). MH2, MH4, MH5, MH24, MH25, MH27, and MH46 were obtained from the Developmental Studies Hybridoma Bank. 1:400 mouse anti-α-tubulin (DM1A; Sigma) was used as another primary antibody. Secondary antibodies used were: anti-mouse rhodamine (Cappel); and anti-rabbit Alexa Fluor 488 (Molecular Probes). Immunofluorescence was performed with the freeze-crack method, followed by methanol and acetone fixation [206]. DNA was stained with 1 µg/ml 4,6-diamidino-2- phenylindole (DAPI). Animals were observed with a Zeiss Axioskop microscope equipped for differential interference contrast (DIC) and fluorescence microscopy. Images were captured with a Hamamatsu ORCA-ER digital camera with Openlab software (Improvision). Images were processed using Adobe Photoshop 7.0. Matched images were taken with the same exposure and processed identically. Western analysis and Immunoprecipitation 93

105 Western blots were performed with the SuperSignal West Femto kit (Pierce). Worm lysate was prepared by asynchronized worm culture according to standard procedures and resuspended in SDS loading buffer and boiled for 5 minutes prior to freezing. Immunoprecipitation using anti-flag antibodies was performed as previously described [100, 115]. ACKNOWLEDGEMENTS I would like to thank D.R. Bosu for the initial cand-1 screen (chromosome I, IV, and V) and KR. Williams for re-testing cand-1 enhancers and technical help. We are grateful to Yuji Kohara for cdna clones; Oliver Hobert and Andrew Z. Fire for plasmids; Pierre Gonczy and David M. Miller III for antibodies; and S. Mitani and the National Bioresource Project for the Experimental Animal Nematode C. elegans (Japan) and the Caenorhabditis Genetics Center for strains. 94

106 Figure 3-1. hpk-1 and dynein do not affect CAND-1 proteins but dynein affects CUL-2 proteins (A) hpk-1 and dynein do not regulate CAND-1 protein levels. Wild-type or cand- 1(tm1683) animals fed with no RNAi, hpk-1 RNAi, or dhc-1 RNAi were used to test levels of CAND-1 proteins. Anti-CAND-1 western analysis is shown in the top panel and antiα-tubulin is used as loading control (bottom panel). Depletion of hpk-1 or dhc-1 (one of the dynein components) does not alter CAND-1 protein levels. (B) hpk-1 RNAi does not affect the interaction between CAND-1 and cullins. Immunoprecipitation (IP) with FLAG antibodies was performed in transgenic worms expressing either CUL-2::FLAG (right) or CUL-4::FLAG (left). Western analysis with CAND-1 antibodies is shown. Note that in immunoprecipitations of transgenic worms expressing CUL-4::FLAG, hpk-1(rnai) animals was loaded much more compared to no RNAi animals. A significant difference of CAND-1 binding to either CUL-2 or CUL-4 was not observed upon hpk-1 RNAi. (C) Inactivation of dynein components causes a reduction of unneddylated CUL-2. Western analysis was performed in wild-type or in the RNAi-sensitive strain rrf-3 to further inactivate dynein components. Anti-CUL-2 western analysis is shown in the top panel and anti-α-tubulin is used as a loading control (bottom panel). Upon RNAi of dynein components (dlc-1 or dli-1), unneddylated CUL-2 proteins decrease. (D) The dynein-interacting protein, LIS-1, also regulates CUL-2 neddylation. Anti-CUL-2 western analysis is shown in the top panel and anti-α-tubulin is used as loading control (bottom panel). Experiments were performed similarly to panel C. As seen in dynein component RNAi, depletion of lis-1 causes a decrease of unneddylated CUL-2 proteins. 95

107 (E) Dynein does not regulate ectopically expressed CUL-4 protein. Transgenic worms expressing CUL-4::FLAG fed with no RNAi or dynein RNAi is used to test levels of CUL- 4 neddylation. Anti-FLAG western analysis is shown in the top panel and anti-α-tubulin is used as a loading control (bottom panel). The inactivation of dynein components does not change any post-translationally modified CUL-4 proteins. 96

108 97

109 Figure 3-2. dynein does not regulate CUL-2 or CAND-1 subcellular localization (A) CUL-2 localization is not changed in dhc-1(or195ts) mutants. Embryos of wild type and dhc-1(or195ts) mutants were stained with anti-cul-2 (green, left panels), DAPI (blue, middle panels) to visualize DNA, and anti-α-tubulin (red, right panels). In wild type and dhc-1(or195ts) at 16ºC, CUL-2 mainly localizes to the nucleus as previously reported [96]. In dhc-1(or195ts) at 26ºC (complete loss of dynein function), CUL-2 still localizes to the nucleus. Note that multinuclei are observed in dhc-1(or195ts) at 26ºC due to loss of dynein function as previously reported. (B) dhc-1(or195ts) mutants do not alter the subcellular localization of CAND-1. Embryos of dhc-1(or195ts) mutants were stained with anti-cand-1 (green, left panels), DAPI (blue, middle panels) to visualize DNA, and anti-α-tubulin (red, right panels). In dhc- 1(or195ts) at 16ºC, CAND-1 mainly localizes to the nucleus as previously seen (ETK unpublished). In dhc-1(or195ts) at 26ºC (complete loss of dynein function), CAND-1 still localizes to the nucleus. 98

110 99

111 Figure 3-3. CUL-2 or CUL-4 substrates do not accumulate in dynein loss of function (A) Cyclin B1 is degraded during meiosis in dhc-1(rnai) embryos. Epifluorescence (top) and DIC (bottom) images of cul-2(rnai) and dhc-1(rnai) gravid adults expressing a CYB-1::GFP transgene. CYB-1::GFP signals are absent in dhc-1(rnai) pronuclei stage embryo (immediately after meiosis) (right panels) but perdure in cul-2(rnai) embryos (left panels) as previously reported [97, 98]. (B) CDT-1 undergoes degradation in dyrb-1(tm2645) mutants. dyrb-1(tm2645) and wildtype larvae were stained with anti-cdt-1 (green), DAPI (blue), and anti-ajm-1 (red overlay, staining adhesion junctions to indicate seam cell boundaries [205]). Anti-CDT-1 signals are absent in wild-type (right panels) and dyrb-1(tm2645) larvae (left panels) 100

112 101

113 Figure 3-4. HPK-1::GFP expression HPK-1::GFP is broadly expressed in all tissues and primarily localizes to nuclei. (A-B) wild-type larvae expressing HPK-1::GFP in head regions (A) and vulva regions (B). DIC (top) and epifluorescence (bottom) images of larvae expressing a HPK-1::GFP transgene. (C) Embryos expressing a HPK-1::GFP transgene. Epifluorescence images of wild-type (left) and cand-1(tm1683) (right) embryos expressing a HPK-1::GFP transgene. Wildtype and cand-1(tm1683) embryos have the same expression patterns of HPK-1::GFP. (D) Embryos of cullin RNAi or cand-1(tm1683) expressing HPK-1::GFP. Epifluorescence images of cul-2(rnai) (top), ddb-1(rnai) (middle), and cand-1(tm1683) (bottom) larvae expressing a HPK-1::GFP transgene in vulva regions. Inactivation of cul-2, ddb-1 or cand-1 does not change the expression pattern of HPK-1::GFP. 102

114 103

115 Figure 3-5. cand-1(tm1683) mutants with hpk-1(rnai) have hypodermis, gut, and pharyngeal cells Late stage embryos in wild type (top), cand-1(tm1683) (middle), and cand-1(tm1683) with hpk-1(rnai) (bottom) were stained with MH27. Left panels show a lateral view and right panels present a perpendicular view. Although embryos of cand-1(tm1683) with hpk-1(rnai) have hypodermis, gut, and pharyngeal cells, the tissues are not properly organized. 104

116 105

117 Figure 3-6. CAND-1 and HPK-1 do not regulate UNC-37 in C. elegans (A) Wild-type, cand-1(tm1683), and hpk-1(pk1393) animals have doublet UNC-37 bands in western analysis. The modification patterns of endogenous UNC-37 were examined in wild-type, cand-1(tm1683) or hpk-1(pk1393). Anti-UNC-37 western analysis is shown in the top panel and anti-α-tubulin is used as a loading control (bottom panel). Wild-type animals do not have additional post-translational modifications of UNC-37. (B) Subcellular localization of UNC-37 is not changed in cand-1(tm1683) or cand- 1(tm1683) with hpk-1(rnai). Embryos of wild type (left), cand-1(tm1683) (middle), and cand-1(tm1683) with hpk-1(rnai) (right) were stained with anti-unc-37 (green, top panels), DAPI (blue, middle panels), and anti-α-tubulin (red, bottom panels). UNC-37 levels as well as subcellular localizations remain the same in wild type, cand-1(tm1683), and cand-1(tm1683) with hpk-1(rnai). 106

118 107