University of Alberta

Function and regulation of the FOXC transcription factors

by

Lijia Huang

A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements for the degree of

Doctor of Philosophy in Medical Sciences

Medical Genetics

©Lijia Huang Fall 2009 Edmonton, Alberta

Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms.

The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission. Library and Archives Bibliotheque et 1*1 Canada Archives Canada Published Heritage Direction du Branch Patrimoine de I'edition 395 Wellington Street 395, rue Wellington OttawaONK1A0N4 Ottawa ON K1A 0N4 Canada Canada Your file Votre ref6rence ISBN: 978-0-494-55816-4 Our file Notre reference ISBN: 978-0-494-55816-4

NOTICE: AVIS:

The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lntemet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distribute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.

••I Canada Examining Committee

Dr. Michael A. Walter, Medical Genetics

Dr. D. Moira Glerum, Cell Biology

Dr. Heather E. McDermid, Biological Sciences

Dr. Yves Sauve, Ophthalmology

Dr. Ted Allison, Biological Sciences

Dr. Judith West-Mays, Pathology and Molecular Medicine, McMaster University Abstract

FOXCl and FOXC2 are the two members of FOXC transcription factors.

Both of them play an important role in the embryonic development of multiple

systems including the anterior segment of the eye. Transcription factors such as

FOXCl and FOXC2 interact with other proteins to fulfill their function. I

screened a human trabecular meshwork cDNA library using yeast two-hybrid

methods in order to discover novel FOXCl or FOXC2 protein-protein interactions.

From these experiments, human p32 was identified as a FOXCl interacting

protein, and human Protein Inhibitors of Activated STAT 3 (PIAS3) was isolated

as a FOXC2 interacting protein. These interactions have functional consequences.

p32 is able to repress FOXCl transactivation in luciferase assays; whereas its

impaired interaction with the transactivation-deficient FOXCl Fl 12S suggests

that p32 may regulate FOXCl activity both positively and negatively. These

results also indicate that impaired protein-protein interaction may be an

underlying mechanism of Axenfeld-Rieger malformation. Study of the effect of

PIAS3 on FOXC2 transactivation in luciferase assays revealed that PIAS3 itself

induces the expression of the reporter gene in a FOXC-binding site dependent

manner. However, PIAS3 doesn't physically bind FOXC-binding sequence

indicating that PIAS3 can up-regulate endogenous transcription factor(s) which

may include FOXC2 to induce the expression of the reporter gene. Knocking

down FOXC2 results in a decreased PIAS3-induced expression of the reporter

gene, supporting the idea that PIAS3 can activate endogenous FOXC2. Further investigations are necessary to reveal the biological and clinical significance of

FOXC2/PIAS3 interaction.

In the last part of the thesis, I report that both FOXC transcription factors regulate the expression of FGF19 and FOXOlA in an ocular cell line, supporting the hypothesis that FOXC1 and FOXC2 have common target genes. Furthermore,

I discovered for the first time that FOXC1 and FOXC2 can physically interact and regulate each other's activity, indicating that FOXC1 and FOXC2 are not simply redundant. Instead, these two proteins have a complex relationship in how they co-regulate their target genes. These findings provide evidence of a novel regulatory mechanism to control the overall FOXC activity, which can be very critical for the normal development and cellular function. Acknowledgements

I would first and foremost like to acknowledge my supervisory committee for their time and guidance. In particular, I would like to thank my supervisor, Dr.

Michael Walter, for accepting me as his student and for his support and patience during my graduate studies. I would also like to thank the Glerum laboratory for their assistance in the yeast experiments.

It has been a great pleasure working with all of the members of the ocular genetics laboratory, both past and present, during my PhD training. I wish everyone the best. Thanks to: Dr. Fred Berry for all the intelligent suggestions and for being a role model for me in the field of research; Michael Sharp for training me when I first came to the lab - it was fun working with you; Tim Footz and

Farideh Mirzayans for technical support; Chanchal Birdi Larsen for your encouragement and help with my English study; Yoko Ito for your friendship and the delicious food from your Mom.

I would also like to thank my friends from the department and outside the department. Thanks to: Ming ye and Jason Bush for your help with my research projects; Hongying Zhao, Wei Wang, Weizheng Guo, Ninghe Hu and Kelly

Narine for being wonderful friends to me - it has been always enjoyable spending time with you guys.

Finally, I would like to express my deepest appreciation to the most important people in my life: my parents, M.Mtfr and MMii, and my husband, T ^tiTc. Without your unconditional love and endless support, I would not have made it this far. Table of Contents

CHAPTER 1. GENERAL INTRODUCTION 1

FORKHEAD DOMAIN STRUCTURE 2

FORKHEAD BOX TRANSCRIPTION FACTOR CI (FOXC1) 3 FOXC1 and Axenfeld-Rieger (AR) malformations 3 Domain Structures ofFOXCl and Molecular Analyses of Disease-causing FOXC1 Missense Mutations 5 FOXC1 Expression Pattern and Animal Models 6 Developmental Pathogenesis 12 Molecular Regulation ofFOXCl Transcription Activity 14

FORKHEAD BOX TRANSCRIPTION FACTOR C2 (FOXC2) 15 FOXC2 and Lymphedema-Distichiasis (LD) Syndrome 15 FOXC2 Expression Pattern and Animal Models 17 FOXC2 and Metabolism 20 FOXC2 and Cancer 22 FOXC2 and Its Direct Target Genes 22 FOXC1 AND FOXC2 23 Overlapping Expression ofFoxcl and Foxc2 23 Foxcl and Foxc2 Double Knockout Mouse Models * 24

RATIONALE OF MY THESIS RESEARCH 27

CHAPTER 2. HUMAN P32 IS A NOVEL FOXC1-INTERACTING PROTEIN THAT REGULATES FOXC1 TRANSCRIPTIONAL ACTIVITY IN OCULAR CELLS 67

INTRODUCTION 68

METHODS 71 Plasmids 71 Yeast two-hybrid screen 71 Mammalian cell culture and transfection 73 Immunoblot analysis 73 Ni2+-NTA pull-downs 74 Immunoprecipitation (IP) 75 Immunofluorescence 76 Transactivation assays 77 Electrophoretic mobility shift assay (EMSA) 78 Mutation screen ofp32 gene in AR patients 79 Realtime qPCR 79

RESULTS 81 Isolation ofp32 as a FOXC1 interacting protein by yeast two-hybrid (Y2H) screening 81 Confirmation of the interaction between FOXC1 andp32 82 Colocalization ofFOXCl andp32 82 The FOXC1 forkhead domain and intact p3 2 are required for the interaction between F0XC1 andp32 83 Mutation screen and detection of copy number variation ofp32 gene in AR patients 83 p32 inhibits FOXC1-mediated transactivation 84 p32 does not affect FOXC1 DNA binding ability 84 The FOXC1 carrying the patient mutation F112S displays an impaired interaction withp32 85

DISCUSSION 86

CHAPTER 3. ISOLATION AND ANALYSIS OF FOXC2 INTERACTING PROTEINS 108

INTRODUCTION 109

METHODS 112 Plasmids and reagents 112 Yeast two-hybrid screen 112 Mammalian cell culture and transfection 113 Immunoblot analysis 113 Immunoprecipitation (IP) 113 Immunofluorescence 114 Transactivation assays 114 siRNA transfection 115 Electrophoretic mobility shift assay (EMSA) 116

RESULTS 117 Isolation ofPIAS3 as a FOXC2 interacting protein by yeast two-hybrid screening 117 Confirmation of the interaction between F0XC2 and PIAS3 by immunoprecipitation 117 Colocalization ofFOXC2 and PIAS3 118 The effect ofPIAS3 on F0XC2-mediated trans activation 118 PIAS3 does not directly bind to the FOXC DNA binding sequence 119 PIAS3 is able to up-regulate the transcriptional activity of endogenous FOXC2 119

DISCUSSION 121 CHAPTER 4. ANALYSES OF THE REGULATION OF TARGET GENES BY FOXC1 AND FOXC2 144

INTRODUCTION 145

METHODS 150 Plasmids 150 Mammalian cell culture and transfection 150 Immunoblot analysis 151 Transactivation assays 151 Chromatin Immunoprecipitation (ChIP) 151

RESULTS 154 Both FOXC1 andFOXC2 bind to the FGF19 and FOXOlA promoter in vivo 154 Both FOXC1 and FOXC2 activate transcription from the FGF19 and FOXOl A promoters 155 FOXC1 and FOXC2 do not function synergistically or additively in transactivation assays 755 FOXC1 and FOXC2 are able to interact with each other 156

DISCUSSION 158 CHAPTER 5. GENERAL DISCUSSION AND FURTHER DIRECTIONS 171

FOXC1 AND FOXC2 INTERACTING PROTEINS CAN BE IDENTIFIED BY THE HTM

YEAST TWO-HYBRID SYSTEM 172

PROTEIN-PROTEIN INTERACTIONS CAN REGULATE THE TRANSACTIVITY OF FOXC

TRANSCRIPTION FACTORS 174

SIGNALING PATHWAYS IN THE EYE ASSOCIATED WITH FOXC TRANSCRIPTION

FACTORS AND THEIR INTERACTING PROTEINS 177

FUTURE DIRECTION 180 Further characterization oftheFOXC2/PIAS3 and FOXC 1/FOXC2 interaction 180 Identification of additional FOXC 1 or FOXC2 interacting proteins 183

BIBLIOGRAPHY 186

APPENDIX: ANTIBODY INFORMATION AND SUPPLIERS 210 List of Tables Table 2-1: primer sets for sequencingp32 gene 89 List of Figures Figure 1-1: Structure of the forkhead domain 29 Figure 1-2: Clinical manifestation of Axenfeld-Rieger malformations 31 Figure 1-3: Assignment of functional sub-domains to FOXC1 FHD 33 Figure 1-4: Summary of functional domains of FOXC1 35 Figure 1-5: Expression of Foxcl RNA and protein in the developing somites.... 37 Figure 1-6: Skeletal defects in Foxcl acZ homozygous mouse 39 Figure 1-7: Expression of Foxcl lacZ in the cardiovascular system 41 Figure 1-8: Cardiovascular defects in FoxcllacZ mutant mice 43 Figure 1-9: Expression of Foxcl in the developing kidney 45 Figure 1-10: Kidney and ureter defects in Foxcl0 homozygous mice 47 Figure 1-11: Expression of Foxcl ac in the developing eye 49 Figure 1-12: Axenfeld Rieger-like ocular defects in Foxcl0 heterozygotes 51 Figure 1-13: Clinical features of lymphedema-distichiasis syndrome 53 Figure 1-14: Expression of Foxc2 in 7.5 to 11.5 dpc mouse embryos 55 Figure 1-15: Aortic arch defects in Foxc2-mx\\ mice at birth 57 Figure 1-16: Defects in the skeletal development of Foxc2-mi\\ mice 59 Figure 1-17: Defects in lymph system of Foxc2+' heterozygous mice 61 Figure 1-18: Protein sequence alignment of human FOXC1 and human FOXC2 63 Figure 1-19: Ocular defects in Foxcl heterozygous, Foxc2 heterozygous as well as Foxcl and Foxc2 double heterozygous mice 65 Figure 2-1: Interaction phenotype displayed by colony 2F061 in the yeast two- hybrid library screen using GAL4DBD-FOXC1 fusion construct as a bait vector 90 Figure 2-2: Retransformation assay and test against other bait constructs 92 Figure 2-3: Confirmation of the interaction between FOXC1 and p32 94 Figure 2-4: Subcellular localization of FOXC1 and p32 96 Figure 2-5. Cellular localization of endogenous FOXC1 and p32 98 Figure 2-6: The FOXC1 forkhead domain is required for binding to p32 100 Figure 2-7. p32 impairs FOXC1-mediated transactivation 102 Figure 2-8: p32 does not affect FOXC1 DNA binding ability 104 Figure 2-9: The FOXC1 Fl 12S mutant is not able to interact with p32 106 Figure 3-1: Interaction phenotype displayed by colony 2F2021 in the yeast two- hybrid library screen using GAL4DBD-FOXC2 fusion construct as a bait vector 128 Figure 3-2: Retransformation assays 130 Figure 3-3: Confirmation of the interaction between FOXC2 and PIAS3 132 Figure 3-4: Subcellular localization of FOXC2 and PIAS3 134 Figure 3-5: The effect of PIAS3 on FOXC2 transactivation 136 Figure 3-6: FOXC binding site is required for the transactivation of PIAS3 138 Figure 3-7: PIAS3 does not bind to FOXC consensus binding sequence and does not affect FOXC2 DNA binding 140 Figure 3-8: PIAS3 can up-regulate the transcriptional activity of endogenous FOXC2 142 Figure 4-1: Both FOXC1 and FOXC2 bind to the FGF19 and FOXOIA promoter in vivo 163 Figure 4-2: Both FOXC1 and FOXC2 activate transcription from the FGF19 and FOXOIA promoters 165 Figure 4-3: FOXC1 and FOXC2 do not function synergistically or additively in transactivation assays 167 Figure 4-4: FOXC1 and FOXC2 interact with each other 169 Figure 5-1: Summary of the known interacting proteins of FOXC1 or FOXC2 184 List of symbols and abbreviations A deletion 3AT 3-amino-1, 2, 4-Triazole 5FOA 5-fluoroorotic acid AD transactivation domain AD1 N-terminal transactivation domain of FOXC1 AD2 C-terminal transactivation domain of FOXC1 Alx4 aristaless-like homeobox 4 Ang-2 angiogenic factor angiopoietin-2 AR Axenfeld-Rieger ASF/SF2 splicing factor, arginine/serine-rich 1 BMP bone morphogenetic protein bp base-pairs BSA bovine serum albumin cAMP cyclic adenosine monophosphate CBF CCAAT-binding factor CBP CREB binding protein cDNA complementary DNA CDC2L5 cell division cycle 2-like 5 C/EBPa CCAAT enhancer binding protein a ch congenital hydrocephalus ChIP chromatin immunoprecipitation Cpm counts per minute C-terminal carboxyl-terminal DAPI 4',6-diamidino-2-phenylindole DBD DNA binding domain DU1.D114 delta-like 1,4 dpc days post coitum E embryonic day EBD Evans blue dye ECM extracellular matrix EGF epidermal growth factor EGFP enhanced green fluorescent protein EGFR epidermal growth factor receptor EMSA electrophoretic mobility shift assay EMT epithelial-mesenchymal transition ERK extracellular signal-regulated kinase EST expressed sequence tag ETA endothelin receptor A Eyal Drosophila eyes absent gene 1 FFA free fatty acids FGF fibroblast growth factor FGFR4 fibroblast growth factor receptor 4 FHD Forkhead domain FLNA filamin A Fox forkhead box Foxa2, Foxa3 forkhead box transcription factor a2, a3 FOXCl,FOXC2 forkhead box transcription factor CI, C2 Foxd3 forkhead box transcription factor d3 FOXFl,FOXF2 forkhead box transcription factor Fl, F2 FOXG1 forkhead box transcription factor Gl FOXH2 forkhead box transcription factor H2 FOX01,FOX04 forkhead box transcription factor Ol, 04 FOXP2 forkhead box transcription factor P2 FREAC-3, FREAC-11 forkhead-related activator 3,11 Fsp27 fat specific protein 27 GAL4 galactose auxotrophy complementation group 4 GDF growth and differentiation factor Gdnf glial cell line-derived neurotrophic factor GFP green fluorescent protein Gfi-1 growth factor independent 1 GLUT4 glucose transporter type 4 GST glutathione S transferase Hes5 hairy and enhancer of split 5 Hey hairy/enhancer-of-split related with YRPW motif His histidine HNF3 hepatocyte nuclear factor 3 HPla heterochromatin protein 1 a HRP horseradish-peroxidase HTM human trabecular meshwork ID transcription inhibitory domain IL-6 interleukin-6 IFN interferon IOP intraocular pressure IP immunoprecipitation IPTG isopropyl P-D-l -thiogalactopyranoside IR insulin receptor IRF-1 interferon regulatory factor-1 IRS1.IRS2 insulin receptor substrate 1, 2 Itgb3 integrin p3 JAK Janus kinase Kb kilobase-pairs LacZ P-galactosidase LARII luciferase assay reagent II LB medium Luria-Bertani medium LD lymphedema-distichiasis L-dopa L-dihydroxyphenylalanine LEF1 lymphoid enhancer factor 1 leucine Lfng lunatic fringe Lhxl LIM homeobox 1 Lmxlb LIM homeobox transcription factor lb MAP kinase mitogen-activated protein kinase MBD1 methyl-CpG-binding domain protein 1 MEKs MAPK/ERK Kinases Mesp mesoderm posterior MITF microphthalmia transcription factor Msx2 msh homeobox 2

NF-KB nuclear factor-KB Ni2+ nickel NICD Notch intracellular domain NLS nuclear localization signal NMR nuclear magnetic resonance NPCE nonpigmented ciliary epithelial N-terminal amino-terminal Oct4 POU class 5 homeobox 1 gene product ORFs open reading frames Osrl odd-skipped related 1 P postnatal day PAGE polyacrylamide gel electrophoresis PAI-1 plasminogen activator inhibitor type I Paxl paired box 1 PBS phosphate buffered saline PBSX phosphate buffered saline plus Triton X-100 (0.05%) PBX1 pre-B-cell leukemia homeobox 1 PCR polymerase chain reaction PDE4 phosphodiesterase 4 PI3K phosphoinositide 3-kinases PIAS protein inhibitors of activated STAT PITX2 pituitary homeobox gene 2 PKA protein kinase A PMSF phenylmethanesulfonyl fluoride PPARy peroxisome proliferator-activated receptor y pRB retinoblastoma protein PSM presomitic mesoderm qPCR quantitative polymerase chain reaction RA retinoic acid Raldh retinaldehyde dehydrogenase RPE retinal pigment epithelium RT-PCR reverse transcriptase-polymerase chain reaction SAP scaffold attachment factor-A/B, apoptotic chromatin-condensation inducer in the nucleus (ACINUS), PIAS SC synthetic complete SCF stem cell factor SDS sodium dodecyl sulfate Ser serine SETDB1 SET domain, bifurcated 1 Shh sonic hedgehog SIM SUMO-1-interaction motif siRNA small interfering RNA SKIP ski-interacting protein Smad3, Smad4 SMAD family member 3, 4 SNP single nucleotide polymorphism STAT signal transducer and activator of transcription Su(H) suppressor of hairless SUMO small ubiquitin-like modifier TBS Tris-buffered saline TBST TBS plus Tween (0.05%) Tbxl T-box 1 TFIID transcription factor II D TGE Tris-glycine-EDTA TGFp transforming growth factor p Tgfbr2 TGFp receptor type 2 TNFcc tumor necrosis factor a TM trabecular meshwork Trp tryptophan Tyr tyrosinase Ucp-1 uncoupling protein-1 Ura uracil UTR untranslated region VEGF vascular endothelial growth factor VSD ventricular septae defects Wnt wingless-type ligand family WT wild type Y2H yeast two-hybrid YPAD yeast extract-peptone-dextrose plus Adenine -1 zonula occluden-1 Chapter 1. General Introduction

1 Forkhead box (Fox) transcription factors constitute a superfamily of proteins carrying an evolutionarily conserved 110-amino acid, monomelic DNA-binding domain named the Forkhead domain (FHD) or winged helix domain. Since the identification of the first member of this family, the Drosophila melanogaster gene Fork head (1), over 100 genes in species ranging from yeast to human have been identified, which are further delineated into 18 subfamilies (Fox A-R) using phylogenetic analysis based on the amino acid sequence of FHD (2). Members of the Fox family play key roles in a wide spectrum of biological processes, such as cell proliferation and cell fate specification, cell migration, cell cycle regulation, as well as cell signaling cascades (3, 4). Loss or gain of Fox function can promote aberrant biological events including developmental defects and tumorigenesis (2).

Forkhead Domain Structure The three dimensional structures of the FHDs have been resolved for several Fox members, including mouse Foxa3 (HNF3y), human FOXC2 (FREAC-11), mouse Foxd3 (Genesis) and human FOX04 using X-ray crystallography or NMR (nuclear magnetic resonance) analysis (5-8). A very similar protein folding of these FHD DNA-binding domains has been found. The FHD domain forms a winged-helix structure consisting of three amino-terminal (N-terminal) a-helices packed against a small three-stranded antiparallel p-sheet from which two large loops (wings) protrude (5) (Figure 1-1). DNA binding is mainly achieved through the helix 3, the "recognition helix", which inserts into the major groove of the DNA and makes sequence-specific contacts with the DNA. In addition, the contact of the second wing with the DNA phosphate backbone also contributes to the protein-DNA interaction (5). The FHDs share a core consensus RTAAAYA binding-sequence (R: A or G; Y: C or T), whereas they are different in their preference for 5' and 3' of the core as well as for nucleotides at the R and Y positions in the core (9). The DNA-binding specificity of the FHDs is also influenced by the relatively variable region immediately preceding the recognition helix, which modulates the presentation of the recognition helix (9-11). In vitro site selection assays found that FOXC1 bound a nine base pair sequence 5'-

2 GTAAATAAA-3' with high affinity. In addition, the FHD of FOXC1, when bound to DNA, bends the DNA double helix by approximately 90 degrees (9).

Forkhead Box Transcription Factor CI (FOXC1) FOXC1 is a member of the Fox family of transcription factor. FOXC1 is a single exon gene located at chromosome 6p25 with a 1659 bp-open reading frame encoding a protein 553 amino acid in length. FOXC1 is expressed in multiple fetal and adult tissues and plays essential roles in the development of multiple organ systems. Mutations in FOXC1 result in congenital malformations in the anterior segment of the eye. A detailed description and discussion of FOXC1 and its function will be included in the following sections.

FOXC1 and Axenfeld-Rieger (AR) malformations Axenfeld-Rieger malformation consists of a range of phenotypically and genetically heterogeneous ocular disorders prominently affecting the development of anterior segment of the eye. Historically AR malformation has been described as separated disorders with related conditions such as Axenfeld anomaly, Rieger anomaly, Axenfeld syndrome, Rieger syndrome(12). Axenfeld anomaly is diagnosed when patients present with iris strands connecting the iridocorneal angle to the trabecular meshwork and embryotoxon cornea posterius (a thickening and centrally displaced Schwalbe's line). Rieger anomaly is considered when patients present with defects involving the iris including iris hypoplasia, corectopia or polycoria. When these findings are accompanied with systemic defects, diagnosis of Axenfeld syndrome or Rieger syndrome is considered. Although Axenfeld anomaly and Rieger anomaly were initially considered to be distinct clinical disorders, the overlapping and similar iridocorneal angle dysgenesis as well as non-ocular defects present in both disorders lead to the proposal that both disorders are within the same phenotypic spectrum and should be grouped as a single, clinically variable disorder, Axenfeld-Rieger anomaly or Axenfeld-Rieger syndrome when systemic defects are presented (13). The term

3 Axenfeld-Rieger malformation is used as a general term representing both Axenfeld-Rieger anomaly and Axenfeld-Rieger syndrome. AR malformation is inherited in an autosomal dominant manner with high penetrance and variable expressivity. Patients with AR malformation receive most clinical attention due to their increased risk of developing glaucoma (13, 14), a morbid progression of visual-field loss starting from peripheral vision and eventually leading to blindness. The medical treatment of glaucoma in AR patients is similar to that of primary open angle glaucoma and focuses on decrease of intraocular pressure (IOP). The systemic phenotypes are seen with incomplete penetrance and variable expressivity including hypodontia, redundant periumbilical skin, sensory hearing loss, cardiovascular defects, hydrocephalus, skeletal limb anomalies as well as pituitary insufficiency (15-19). Some typical ocular and systemic manifestations of Axenfeld-Rieger malformation are shown in Figure 1-2. Three chromosomal loci: 4q25 (20-22), 6p25 (23, 24) and 13ql4 (25) have been identified to be associated with AR malformation. The gene at 13ql4 has not yet been identified. The homeobox transcription factor PITX2 gene mapping to 4q25 is the first gene that has been shown to be mutated in AR malformation (26). The distal end of chromosome 6 has been noticed to be important for ocular development since patients with 6p terminal deletions or ring chromosome 6 display ocular defects including iris hypoplasia and glaucoma (27-29). cDNA encoding human FOXC1 (FREAC-3) was first cloned from a craniofacial cDNA library (9) and the gene was mapped to the chromosome 6p25 subtelomeric region (30), making it a good positional candidate gene within the 6p25 AR locus (23, 31). Direct sequence analysis of FOXC1 by two independent groups identified mutations within families diagnosed with AR malformation (32, 33). So far over 40 AR-causing mutations in FOXC1 have been reported (34), constituting either frameshift mutations resulting in truncated proteins or missense mutations almost always within the FHD domain. Only one missense variant P297S identified in an AR patient was found downstream of the FHD (Fetterman C et al, in press). Moreover, duplication of a segment of the 6p region containing FOXC1 also leads

4 to AR malformation (35-37), demonstrating that neither haploinsufficiency nor increased gene dosage of FOXC1 can be tolerated. Therefore, a strict range of FOXC1 dose is required for the normal ocular development.

Domain Structures of FOXC1 and Molecular Analyses of Disease-causing FOXC1 Missense Mutations The FHD is the DNA-binding domain of the FOXC1 transcription factor and its 3-dimensional structure has been described in the preceding section. The FHD of FOXC1 occupies residues 69 to 178 of the protein. In addition to DNA binding, the FHD carries motifs that are required for correct nuclear localization. Amino acid residues 168-176 in the Carboxyl-terminal (C-terminal) portion of the FHD are required for the nuclear localization of FOXC1 and, when fused to a green fluorescent protein (GFP), are able to target the GFP reporter to the nucleus. Amino acid residues 78-93 in the N-terminus of the FHD provide additional nuclear localization signal (NLS) for efficient targeting of FOXC1 to the nucleus but are not sufficient for the nuclear localization of the GFP reporter (38). Moreover, the FHD of FOXC1 may contribute to the protein's transcriptional activity, since several patient missense mutations (P79T, P79L, Fl 12S, and G165R) within the FHD reduce transactivation but retain DNA-binding ability (39-41). Biochemical analyses of sixteen of the FOXC1 missense mutations located within the FHD has demonstrated that all of these mutations perturb FOXC1 activity (39-43). Several different mechanisms underlie the impairments to FOXC1 function, including reduction in protein stability, impairment in DNA- binding capacity, alteration in DNA-binding specificity, defects in nuclear localization and transactivation as well as formation of abnormal protein aggregates. Additionally, these molecular analyses of disease-causing missense mutations provide an insight into structure-function relationship of the FHD. In summary, mutations residing in the N-terminal region to helix 1 or within helix 1 reduce DNA binding, impair nuclear localization and transactivation; mutations in

5 helix 2 only impair transactivation; mutations in helix3 greatly impair DNA binding and specificity as well as affect nuclear localization; while mutations in wing 2 impair DNA binding and transactivation (39-45) (Figure 1-3). In addition to the FHD, other domain structures have been identified in FOXC1 which are of importance for the function of FOXC1. FOXC1 contains N- and C-terminal transactivation domains (AD), from amino acids 1-51 (AD1) and from 435-553 (AD2), respectively. Deletion of both transactivation domains from FOXC1 results in depletion of transactivation ability of FOXC1 in luciferase assays. However, when fused with the DNA-binding domain of galactose auxotrophy complementation group 4 (GAL4), AD2 induces reporter expression more efficiently than AD1, indicating that AD2 may have a greater degree of intrinsic activity (38). FOXC1 also carries a central transcription inhibitory domain (ID) spanning amino acid residues from 215-366. The inhibitory domain does not have intrinsic transcriptional repressor activity. Instead, it applies its function by inhibiting the ability of activation domains to stimulate transcription. The ID is also a target for phosphorylation by protein kinases including p44/p42 extracellular signal-regulated kinases (ERKs), indicating that FOXC1 is a phosphoprotein and is subjected to regulation by the mitogen-activated protein kinase (MAP kinase) pathway (38, 46). A summary of FOXC1 functional domains is shown Figure 1-4.

FOXC1 Expression Pattern and Animal Models Expression of FOXC1 has been detected in multiple fetal and adult tissues. A 3.9 Kb transcript is detected with abundance in adult , liver, skeletal muscle, peripheral blood leukocytes, prostate, iris, as well as in fetal kidney and craniofacial tissue. An additional transcript of 3.4 Kb, which may be derived from differential polyadenylation, is detected in fetal kidney as well (9, 32, 33). Expression profiling of FOXC1 in the adult human eye determined by quantitative RT-PCR shows the greatest expression of FOXC1 in the trabecular meshwork

6 (TM) followed by the optic nerve head, choroid/retinal pigment epithelium (RPE), ciliary body, cornea and iris (47). Foxcl is the murine homologue of human FOXC1. Human FOXC1 and murine Foxcl share an identical FHD at the amino acid level, while outside the FHD the two proteins share 91% sequence identity (48). Expression of mouse Foxcl is detected in heart, kidney, adrenal gland, and brain, but not in liver (33, 49). Foxcl expression during embryonic development has been extensively studied and will be described in the following paragraphs. Expression of Foxcl is first detected on both sides of the primitive streak (50). Between 7.5-9.0 days post coitum (dpc), FoxcFs expression is seen in the entire presomitic mesoderm (PSM) as well as in head mesenchymes derived from the neural crest and paraxial mesoderm, respectively (50). By 9.5 dpc, Foxcl protein is detected in the PSM forming a gradient with lower levels in the posterior PSM and higher levels in the anterior PSM (51) (Figure 1-5). As somitogenesis proceeds, Foxcl is expressed in cartilage primordia, which will form the future vertebra, ribs, limbs, sternum and skull (48, 49). Thus, it is evident that Foxcl is required for the skeletal development, which is supported by the skeletal defects found in mouse models with disturbed Foxcl. One such mouse model is the naturally occurred congenital hydrocephalus (ch) mouse carrying a point mutation (C376T) in Foxcl, which consequently results in a truncated protein without the FHD (Q123X) (48, 52). Another mouse model was generated by replacing part of Foxcl sequence with [5-galactosidase gene (LacZ) in frame with the translation start site (FoxcltacZ) (48). The phenotype of the two mouse models is identical. Foxcl'' homozygous null mice die pre- or perinatally due to respiratory failure with greatly enlarged and hemorrhagic cerebral hemispheres (48, 53). By 14.5 dpc, the homozygous null embryos can be easily distinguished from their wild type (WT) littermates by frontal bulging of the head resulting from an absence of the calvarial bones. In addition, the basioccipital, exoccipital and hyoid bones are smaller in size and the superoccipital bone is severely malformed. At the base of the skull, the presphenoid bones are absent and the basisphenoid is misshaped. Moreover, the squamosal, zygomatic and mandible

7 are abnormally shaped and overly ossified. The nasal septum is shorter, resulting in the short snout appearance in the mutants (48, 52). The severe facial bone developmental defects in the homozygous mice are reminiscent of the cranial facial abnormalities observed in some AR patients. The dorsal neural arches of the vertebrae, including the atlas and axia, are grossly malformed and fail to fuse; while the lateral arches and vertebral bodies are reduced in size. The ribs in the null embryos are thinner than normal. The rib cage is very unstable due to the absence of the sternum ossification centers, the weak attachment of the right and left costal cartilage and the fragmented xiphoid process (48, 52). The skeletal defects found in FoxcllacZ acZ homozygotes are shown in Figure 1-6. In addition to skeletal development, Foxcl is involved in cardiovascular development. Expression of Foxcl is first detected in the developing first branchial arches between 6.5-9.5 dpc (50). During this period, a weak expression signal is also seen in the endothelium of the heart tube. Later on as the valves and septae form at 10.5 -11 dpc, Foxcl is expressed in the dorsal portion of the , the mesenchyme surrounding the , the dorsal and the dorsal component of the developing septum primum (the future atrial septum) and endocardium (48, 54). From 11.5 dpc onward, Foxcl is expressed in the mesenchyme surrounding all arterial vessels, the mesenchyme of each leaflet of the semilunar valves, the future spiral septum of the outflow tract, the septum primum, the endocardial cushion tissue of the heart and the tricuspid and mitral valves (53). Expression is also observed in the aortic and pulmonary valves, the venous valves, and the smooth muscle of the pulmonary trunk, the endocardium and the trabeculated region of the ventricular wall but not in the (53, 54). By 15 dpc the expression levels of Foxcl decreases generally but there is a retained signal in the atrial septum, and the venous, mitral, tricuspid, aortic and pulmonary valves (54). Foxcl continues to be expressed in adult mouse cardiac tissues including the aorta, the pulmonary trunk, the endocardium and the smooth muscle and endothelium of the coronary vessels (53) (Figure 1-7). Foxcl homozygous null embryos also show disruptions to the cardiovascular system. Cardiovascular defects include patent (ductus arteriosus fails to

8 close after birth), Type B interruptions (an interrupted left aorta, the interruption occurs distal to the origin of the left common carotid artery), as well as clear coarctation or narrowing of the aortic arch. In addition, ventricular septae defects (VSD), aortic and pulmonary valve dysplasia and infundibular hypertrophy have also been observed (53) (Figure 1-8). Foxcl is also expressed in the developing urogenital system. At 8.5 dpc, expression of Foxcl is first observed in the presumptive intermediate mesoderm. At 9.5 dpc, when the nephrogenic cord develops and the Wolffian duct begins to elongate caudally in the intermediate mesoderm, Foxcl is expressed in the mesonephric mesenchyme alongside the Wolffian duct with a dorsoventral gradient. At 10.5 dpc, when the metanephric kidney begins to form with the outgrowth of the ureteric bud from the Wolffian duct into the metanephric mesenchyme, Foxcl is expressed in the metanephric mesenchyme and in the condensing mesenchyme of the kidney thereafter (55) (Figure 1-9). During the development of germ cells and gonads, expression of Foxcl is first observed in the paraxial and intermediate mesoderm, extending into the mesentery of hindgut and the somatic cells of the genital ridges at 10.5 dpc. Afterward, the expression is restricted in the somatic cells of the gonad and is persistent throughout gonad development. In the adult ovary, Foxcl is seen in the thecal and granulosa cells of the developing follicles. However, Foxcl is not expressed in the primordial germ cells and oocytes (56). The development of the urogenital system is affected in Foxcl homozygous null embryos, with duplex kidneys connected to double ureters, one of which is a hydroureter (The distention of the ureter with urine due to blockage) (55, 57). An ectopic ureteric bud is induced from the Wolffian duct more anteriorly to the normal ureteric bud in the homozygous mutants. It is believed that the ectopic ureteric bud induces the formation of an ectopic kidney, which fuses with the normal kidney (55, 57) (Figure 1-10). Although the sexual differentiation is normal, the gonads of both sexes in homozygous null mutants are more anteriorly displaced compared to wild-type mice, and are reduced in size with fewer germ cells and disorganized somatic tissue. The primordial germ cells

9 fail to complete the migration out of the hindgut and into the genital ridges (55- 57). In the head, Foxcl expression is detected in the head mesenchyme including those around the eye at 10.5 dpc (48). Afterward, it becomes evident that Foxcl is expressed in an extensive area of periocular mesenchyme, including cells between the corneal epithelium and the lens, within the optic cup, between the lens and the retina, the prospective trabecular meshwork and sclera, the future conjunctival epithelium as well as the ectoderm of the future inner eyelids. The Foxcl positive periocular mesenchyme cells have been shown to arise developmentally from both mesoderm and neural crest cells (58). In addition, a weak signal is detected in the developing cornea. As the corneal endothelium differentiates, Foxcl expression is down-regulated and is limited to the prospective trabecular meshwork cells, the sclera, and the conjunctival epithelium at 16.5 dpc (59) (Figure 1-11). Lacrimal gland, which produces lubricating secretions to protect the cornea, develops from an out budding of conjunctival epithelium starting from 13.5 dpc. Foxcl is expressed in both the gland epithelium and mesenchyme during the elongation and branching stages of lacrimal gland development (60). Ocular findings in Foxcl homozygous null embryos include open eyelids, iris hypoplasia, small lens, a failure of the anterior chamber to form due to the failure in the development of the corneal endothelium and the subsequent attachment of the lens to the cornea, a thickening of the corneal epithelium and a disorganized corneal stroma (48, 59). The thickened FoxcllacZ/lacZ homozygous corneal epithelium displays a persistence and expansion of LacZ expression, indicating that Foxcl regulates its own expression (59). There is an absence of organized zonula occludens junctions in corneal mesenchyme cells from the homozygous nulls, suggesting that Foxcl may be involved in regulating the expression of extracellular matrix (ECM) components (59). The lacrimal gland development is severely affected in the homozygous null mice with misshaped branches and less terminal buds, resulting in a smaller exorbital lobe and absence of the intra-orbital lobe (60). Examination of Foxcl heterozygotes discovers that they have ocular abnormalities similar to AR malformations (52, 61). The anterior segment

10 phenotypes observed in Foxcl heterozygotes include progressive corectopia, iridocorneal synechiae, iris tears and hypoplasia, posterior embryotoxon and sclerocornea and corneal opacities (61) (Figure 1-12). Histological and electron microscopy analyses of heterozygous mouse eyes reveal that the iridocorneal angle has morphologically normal and abnormal regions (61). In normal regions, the TM is composed of spaced trabecular beams with organized collagen and elastic tissue and the aqueous humor drains through the spaces separating the beams. In the affected region, small or absent Schlemm's canal, large blood vessels, iris strands, as well as hypoplastic or compact or absent trabecular meshwork are observed. Additionally, there are a general paucity of ECM components; cells resembling trabecular meshwork precursors that failed to differentiation sometimes occupy areas that are absent with TM (61). Interestingly, ocular features of the heterozygotes greatly vary dependent on genetic background raising the possibility that modifier loci may exist (52, 61). In zebrafish, two foxcl homologues,ybjcc7a and foxclb, have been found. Both genes have overlapping expression pattern in the unsegmented presomitic mesoderm, newly formed somites, adaxial cells, head mesoderm and periocular mesenchyme (62-64). Foxcl a is essential for the formation of somites in zebrafish. Knocking down foxcl a by morpholino antisense oligonucleotides inhibits the formation of morphologically distinct anterior somites and the segment boundaries (63). Knocking down both foxcla and foxclb leads to a small eye phenotype with arrested cell differentiation in periocular mesenchyme and delayed retinal lamination (65). Evidence from expression data, the range of defects observed in Foxcl mutant mice and zebrafish, as well as the ocular defects and the systemic defects found in patients with AR malformation demonstrate that FOXC1 is an important regulator of ocular, skeletal, cardiovascular and urogenital development.

11 Developmental Pathogenesis FOXC1 is a transcription factor, playing a key role in the normal development of multiple systems. To understand the pathogenesis underlying the human AR malformations and the developmental defects found in Foxcl knockout mice, it is necessary to identify FOXC1 direct target genes. T-box 1 {Tbxl) gene was the first identified FOXC1 target gene. The human homologue of Tbxl is located within the critical region of the 22ql 1.2 deletion syndrome (DiGeorge syndrome) and is likely a major contributor to cardiac and craniofacial phenotypes associated with DiGeorge syndrome (66-68). Foxcl, together with the other two forkhead transcription factor Foxa2 and Foxc2, induces the expression of Tbxl in the head mesenchyme and pharyngeal endoderm through a Fox- binding site upstream of Tbxl. Deletion of this Fox-binding site abolishes the expression of Tbxl in the head mesenchyme and pharyngeal endoderm (69). Sonic hedgehog (Shh) induces expression of Tbxl which is mediated by Foxcl and Foxc2, since Shh null mutants show a reduced expression of Foxcl and Foxc2 (69). Additional FOXC1 target genes have also been identified. Twenty-six genes have been identified as potential target genes regulated by FOXC1 in an ocular cell line, nonpigmented ciliary epithelial (NPCE) cells using a revised chromatin immunoprecipitation (ChIP) assay (70). One of these genes is the fibroblast growth factor 19 gene (FGF19). FOXC1 induces FGF19 expression through direct binding to a FOX binding-site in the regulatory region upstream of FGF19 (65). Knocking downyaxC7 expression in zebrafish embryos decreases fgfl 9 expression in the anterior periocular mesenchyme where foxCl and fgfl 9 are co- expressed. Loss of either foxCl or fgfl 9 during development results in similar, but distinct anterior segment dysgenesis (65). The MAP kinase cascade is activated when FGF19 binds to its receptor fibroblast growth factor receptor 4 (FGFR4). The fact that foxCl morphants show a reduced activation of MAP kinase pathway in the developing cornea leads to a model that FOXC1 regulates gene expression in the cornea through FGF19-FGFR4-ERK1/2 signaling pathway (65). In addition to FGF19, FOXC1 directly binds to a conserved FOX binding- site in the promoter of another forkhead transcription factor, FOXOIA and

12 induces its expression in the eye (71). FOXOIA expression is decreased in cultured trabecular meshwork cells and in the developing zebrafish eye when FOXC1 expression is knocked down (71). This study has revealed that FOXC1 is an important mediator of cell viability and resistance to oxidative stress in the eye through regulation of FOXOIA expression, which may underlie AR-glaucoma pathogenesis (71). Several studies have discovered genes that show altered expression levels in Foxcl homozygous null mice. Unlike the genes discussed above, these genes are downstream of FOXC1 but are not necessarily direct targets. During kidney development, glial cell line-derived neurotrophic factor (Gdnf) is expressed in the nephrogenic mesenchyme and is required for ureteric bud formation. Human homologue of the Drosophila eyes absent gene 1 {Eyal) is also an essential player for ureters budding through inducing the expression Gdnf. In the developing kidney of Foxcl homozygous embryos, Gdnf and Eyal expression are anteriorly expanded, which may lead to the ectopic anterior ureteric buds found in Foxcl mutants (55). Zonula Occluden-1 (ZO-1) protein is a key component of occluding junctions in the corneal mesenchyme cells. The absence of ZO-1 in homozygous Foxcl corneal endothelium results in the defect in the formation of tight occluding junctions between endothelial cells, demonstrating that function of FOXC1 in corneal development at least in part attributes to its transcriptional regulation of ECM components (59). Additionally, expression of msh homeobox 2 (Msx2) and aristaless-like homeobox 4 (Alx4) in the calvarial mesenchyme are reduced in Foxcl null mice, both of which play pivotal roles in the osteoprogenitor cell proliferation and differentiation in calvarial development (72). Bone morphogenetic protein 2 (BMP2) or bone morphogenetic protein 7 (BMP7) beads placed on the WT calvarial mesenchyme induce the expression of Msx2 and Alx4 and this process is mediated by Foxcl. The Foxcl mutant explants do not respond to BMP2/7 signal to induce Msx2 and Alx4 expression (72). The idea of Foxcl being a mediator for BMP signaling pathways during the skeletal development is also supported by the findings that the differentiation of prechondrogenic mesenchyme of the sternum in response to BMP2 or

13 transforming growth factor [31 (TGFpl) treatment is arrested in the Foxcl homozygote cultures (48). Although Foxcl is a mediator of BMP signaling, its own expression in the developing calvarial bone is under control of fibroblast growth factor 2 (FGF2) instead of BMP signaling, indicating that Foxcl integrates BMP and FGF pathways (73). In addition to the FGF2 signal as mentioned above, retinoic acid (RA) and TGF|32 signals are necessary for the normal expression of Foxcl in the developing eye (74, 75). RA is synthesized by the RA-synthesizing retinaldehyde dehydrogenase 1 and 3 (Raldhl/3), which are expressed in the developing retina, lens as well as corneal ectoderm. RA synthesized in these regions diffuses towards the periocular mesenchyme and induces the expression of Foxcl through activation of the RA receptors. Absence of either Raldhl/3 or RA receptors causes similar ocular abnormalities including anterior segment dysgenesis and reduced expression of Foxcl (74). Tgf/32 is expressed in the developing lens at 13.5-15 dpc and its receptor, TGFft receptor type 2 (Tgfbr2) is expressed in the neural crest-derived periocular mesenchyme. Tgfbr2 mutant embryos show microphthalmia with anomalies of the anterior segment, reminiscent of the defects found in AR malformation. Particularly, the expression of Foxcl is hardly detectable in Tgfbr2 mutant eyes (75). As mention above, pheno types of Foxcl mutant mice vary in severity among different genetic background, indicating the existence of modifier loci. Tyrosinase (Tyr) has been identified as a modifier of the ocular defects caused by Foxcl mutations (76). Tyrosinase catalyzes the conversion of tyrosine to L- dihydroxyphenylalanine (L-dopa) and the subsequent L-dopa to dopaquinone. Tyr'~ mice are albinos with mild ocular defects. Intriguingly, the ocular defects of the heterozygous Foxcl* ' mice are more severe on Tyf' background (76).

Molecular Regulation of FOXC1 Transcription Activity FOXC1 is a phosphoprotein and its protein levels and activity are regulated by phosphorylation (38, 46). The residue Ser272 is one of the ERK1/2 target sites

14 in FOXC1. Mutant FOXC1 S272A displays a significant decrease in the steady state protein level as compared to the WT protein (46). Activation of ERK1/2 MAP kinase pathway by epidermal growth factor (EGF) treatment elevates the stability and transcriptional activity of WT FOXC1 but has no effects on S272A. Moreover, inhibition of MAP kinase greatly reduces WT FOXC1 protein levels. These results indicate that ERK1/2 MAP kinase pathway is required for the maintenance of FOXC1 stability which is dependent on Ser272 (46). In addition to the post-translational modulations, other mechanisms such as protein-protein interactions are also involved in the regulation of FOXC1 transactivity. The actin- binding protein filamin A (FLNA) was identified as the interacting protein of FOXC1 (77). FLNA can negatively regulate FOXC1 's activity in melanoma cells by retargeting FOXC1 into heterochromatin protein la (HPla)-rich regions of nucleus. A homeodomain transcription factor, pre-B-cell leukemia homeobox 1 (PBX1) can also interact with FOXC1 and inhibit FOXC1 transactivation (77). Interestingly, FLNA is required for the nuclear import and subnuclear localization of PBX1. It is proposed that FLNA provides a medium in the nucleus for FOXC1 and PBX1 to form a transcriptionally inactive complex, which is targeted to the HPla heterochromatin-rich regions (77). The interaction between FOXC1 and PITX2, the two transcription factors associated with AR malformation, partially answers why both of genes cause the same disease (78). PITX2 interacts with FOXC1 and inhibits FOXC1 transactivation of its putative target genes. Mutant PITX2 proteins fail to repress FOXC1 activity leading to over-activated FOXC1, which is similar to FOXC1 duplication resulting in an increased level of FOXC1 protein and AR malformation (78).

Forkhead Box Transcription Factor C2 (FOXC2)

FOXC2 and Lymphedema-Distichiasis (LD) Syndrome FOXC2 is another member of the forkhead of transcription factors family. The FOXC2 gene is located on chromosome 16q24 and was first found to be responsible for the neonatal lymphedema in a patient with a balanced

15 translocation between chromosome 16 and the Y chromosome (t(Y; 16)(ql2;24.3)) (79, 80), as well as in patients with lymphedema-distichiasis syndrome in two unrelated families (79). LD is a rare, autosomal dominant inherited primary lymphedema associated with distichiasis (Figure 1-13). The lymphedema is of variable age of onset with peripubertal onset in majority of the cases. Distichiasis is a congenital condition with presentation of a second row of eyelashes developed from the meibomian gland instead of the normal origin from tarsal plate (81). Additional ocular anomalies found in LD patients include photophobia, exotropia, ptosis, congenital ectropion and congenital cataracts (82-85). Another commonly occurring anomaly is the early onset varicose veins (83). Mellor and his coworkers have found that the venous reflux due to the failure of the venous valves is always present in the subjects with FOXC2 mutations (86). LD is also associated with complications in a number of other organs such as heart defects, cleft palate, spinal extradural cysts, nephritis, skeletal and craniofacial abnormalities (82-85, 87). To date, over 70 FOXC2 mutations have been identified in LD patients. Interestingly, the majority of these mutations are frame shift or nonsense mutations that introduce premature stop codons and are predicted to produce premature truncated proteins (79, 82, 83, 85, 87-91). Only four missense mutations that cause amino acid substitutions in the FOXC2 coding region have been found, a marked contrast to the findings with FOXC1 mutations in Axenfeld- Rieger syndrome, where both missense mutations and frame shift/nonsense mutations are equally prevalent. Among these missense mutations, S125L, R121H and Wl 16R are within the DNA-binding FHD, whereas S235I is downstream of the FHD (83, 91). Interestingly, FOXC2 S125L and R121H, lying in the third a- helix of the FHD, have their paralogous AR-causing mutants, S131L and R127H in FOXC1 (45). Biochemical analyses of FOXC2 S125L and R121H found that they have impaired DNA binding and transactivation. FOXC2 R121H impedes nuclear localization of FOXC2 as well (45). When detailed ocular examinations were given to 18 individuals from 9 LD families, anterior segment dysgenesis were found in all individuals with FOXC2 missense mutations within the FHD or

16 mutations that introduce a stop codon in the FHD. In contrast, those individuals with mutations outside the FHD exhibited no such ocular phenotypes. The more severe ocular phenotypes resulting from FHD mutations suggests that alteration of this highly conserved DNA-binding domain has a greater impact on FOXC2 functions, at least during the development of the anterior segment of the eye (92).

FOXC2 Expression Pattern and Animal Models The expression pattern of Foxc2 during mouse embryonic development correlates with tissues affected in LD patients. Foxc2 is mainly expressed in mesoderm and neural crest derived tissues in a dynamic pattern during mouse embryogenesis (93-95) (Figure 1-14). Foxc2 is first expressed in the non- notochordal mesoderm surrounding the node and notochord and later in the presomitic mesoderm, somites and cephalic mesoderm, branchial arch, mesenchymal cells surrounding the eye, nasal processes, blood vessels, endocardium as well as the metanephros. As the development proceeds, Foxc2 is detected in the condensing mesenchyme in vertebrae, limbs, kidney and head (95, 96). Most Foxc2 homozygous null embryos die prenatally around El3.5, but some survive to birth then die shortly after with cardiovascular and skeletal defects (96). During cardiovascular development, Foxc2 is expressed in the 3 rd, 4th and 6th pairs of arch arteries as well as in the dorsal from E9.5 (96). In the absence of Foxc2, instead of forming the isthmus of the aortic arch together with the

th , the left 4 arch artery regresses, which might account for the type B interruption of the aortic arch found in Foxc2~' mice (96) ( Figure 1-15). Interestingly, endothelin receptor A (ETA) deficient mice display similar cardiovascular defects to those of Foxc2 nulls. A study of the interaction between the ETA and Foxc2 pathways in the developing cardiovascular system has found that ETA and Foxc2 function additively to direct the formation of the (97). In Foxc2~'~ homozygotes, craniofacial and vertebral defects are often observed. During neurocranium development, Foxc2 is initially expressed in the

17 head mesoderm and is subsequently restricted in the mesenchymal condensations giving rise to the skull. Lacking of Foxc2 results in reduced size, malformations or even absence of those skeletal elements and the ossicles of middle ear as well (95, 96) (Figure 1-16). Besides the craniofacial abnormalities, Foxc2" mice also display vertebral defects including small vertebral bodies, small or absence of ossification centers, misshaped cartilaginous condensations, and absence of spinous processes as well as fused ribs (95, 96). Paired box 1 (Paxl) is another essential transcription factor for the appropriate generation of ventral part of the vertebrae. There is an overlapping expression Foxc2 and Paxl and their expression is induced by the sonic hedgehog signals from the notochord (98). A more severely affected vertebrae development, such as loss of dorsomedial structures of vertebrae and subcutaneous myelomeningocoele as well as loss of the vertebral bodies and intervertebral discs, are found in Foxc2 and Paxl double homozygous null embryos. The underlying defects for these malformations appear to be a synergistic reduction of cell proliferation and/or migration in the sclerotome of Foxc2/Paxl double mutants (98). Taken together, these findings suggest an important role for Foxc2 in regulating or facilitating the proliferation in sclerotome-derived lineages, which is under control of the Shh signaling and is synergistically enhanced by Paxl. The fact that FOXC2 mutations cause primary lymphedema warrants an essential role of FOXC2 during . During mouse embryonic development, Foxc2 is expressed in endothelial cells of the cardinal vein and surrounding mesenchyme at E9.5 and El 0.5 when some of these endothelial cells start to bud off from cardinal vein to form lymphatic primordia (99). Later on, Foxc2 is continuously expressed throughout the entire lymphatic development. In adult mice, expression of Foxc2 is still observed in lymphatic vessels of the mesentery, skin and lymph nodes (99). Although no apparent peripheral edema are seen in Fox2+/~ mice, most of them display lymphatic hyperplasia consistent with the findings in LD patients (100), including a generalized increase in the number and size of lymphatic vessels and lymph nodes throughout the body (101). In addition, lymph reflux is observed within the hepatic hilum, mesentery and the

18 intestinal wall of Foxl ' mice (101) (Figure 1-17). The lymphatic vessels in Foxl' mice display an increased coverage of smooth muscle cells and lack of lymphatic valves, which lead to the dysfunction of lymphatic vessels and consequently lymph reflux (102). What is interesting about Foxc2 expression in the lymphatic development is that there is no accompanying expression of Foxcl in the lymphatic vessels (51, 99), which is co-expressed with Foxc2 in most target tissues during development (49, 51, 55, 61). The absence of Foxcl expression in the lymphatics indicates that no redundancy is provided by the presence of the two closely related transcription factors and thus appears to explain the higher penetrance of lymphedema in LD patients (99). However, a recent study has found that Foxcl and Foxcl are co-expressed in the mesenchymal cells surrounding the cardinal veins and together play a role in early sprouting of lymphatic endothelial cells, which may attribute to their induction of vascular endothelial growth factor C (VEGF-C) expression (103). Some LD patients have concomitant kidney diseases indicating the involvement of FOXC2 in the development and/or function of kidney. The regions of Foxc2 expression that are relevant to kidney development include the intermediate mesoderm and subsequent mesonephric and metanephric mesenchyme (55). As the development proceeds, Foxc2 is expressed in prospective podocytes before the appearance of other podocyte markers, suggesting a role in podocyte differentiation. Foxcl homozygous null mice have smaller kidneys and reduced numbers of glomeruli than normal kidneys (55, 104). The differentiation of mutant podocytes is arrested, indicated by the failure in formation of foot process and the slit diaphragm as well as the presence of immature adherence junctions between the podocytes (104). Foxcl deficiency in podocytes leads to changes in the expression of a combination of genes that are directly or indirectly regulated by Foxcl, which include some podocyte markers and certain subtypes of collagen IV in the mature glomerular basement membrane (104). During ocular development, Foxcl is expressed in the periocular mesenchyme that gives rise to the anterior segment structures, including the

19 corneal endothelium and stroma, ciliary muscle and stroma, iris stroma, trabecular meshwork, as well as Schlemm's canal (49, 58, 95, 105). The expression pattern of Foxc2 in the developing eye largely overlaps with that of Foxcl (49, 58). A LIM homeodomain transcription factor, LIM homeobox transcription factor lb (Lmxlb) directly or indirectly turns off Foxcl and Foxc2 expression in the mesenchyme cells that migrate into the presumptive cornea and differentiate into keratocytes (106). Histological examination of Foxc2+' eyes found small or absent Schlemm's canal, hypoplastic or absent trabecular meshwork and abnormally thin iris stroma and pigment epithelium (61).

FOXC2 and Metabolism In addition to its pleiotropic roles during embryonic development, FOXC2 has been found to be a key regulator of adipocyte metabolism during postnatal life. Transgenic mice overexpressing the human FOXC2 specifically in white and brown adipocytes are lean and hold resistance to high-fat diet induced insulin resistance (107). In the transgenic mice, the intra-abdominal white adipose tissue is reduced and acquires a brown-fat like property (107). Overexpressing FOXC2 has a broad impact on the expression profiles of genes involved in the adipocyte differentiation, metabolism, insulin action and adrenergic pathway (107). The induction of uncoupling protein-1 (Ucp-1), which is a marker of brown adipocytes and favors energy expenditure over storage, is in line with the transition of the white adipose tissue towards brown adipose tissue observed in the transgenic mice. The increased insulin sensitivity in adipocytes is reflected by the increased expression of insulin receptor (IR), insulin receptor substrate 1 (IRS1), insulin receptor substrate 2 (IRS2) and glucose transporter type 4 (GLUT4). The high sensitivity to insulin and the resistance to diet-induced obesity attribute to a hypersensitive P-adrenergic/cAMP/protein kinase A (PKA) pathway partially resulted from the up-regulation of the expression of the ^-adrenergic receptors and PKA subunit RIa{\01), and partially from the reduced expression and activity of phosphodiesterase 4 (PDE4) (108). A further study of the

20 mechanism of FOXC2-mediated up-regulation of the PKA RIahas found that FOXC2 is able to release a potential repressor from the promoter of PKA RIa in adipocytes (109). The hypersensitive (3-adrenergic pathway leads to an increased lipolysis whereas the resulting free fatty acids (FFA) from the lipolysis are kept at low level through the increased consumption of FFA by Ucp-1 induced heat production and increased fatty acid (3-oxidation (107). Because the fatty acids are largely metabolized by adipose tissue, the amount of fatty acid redistributing into muscle cells decreases, which may promote tyrosine phosphorylation of IRS 1 and as a result, increase the insulin sensitivity of the muscle cells (110). The expression level of FOXC2 in 3T3-L1 adipocytes is actively regulated by insulin and tumor necrosis factor a (TNFa) via Phosphoinositide 3-kinases (PI3K) and ERK1/2 dependent pathways (111, 112). Retinoblastoma protein (pRB) might function upstream of Foxc2 and directly or indirectly repress the expression of Foxc2 (113). When differentiate into adipocytes, ./^embryonic stem cells and embryonic fibroblasts display a higher level of Foxc2 expression, compared to the WT cells. Furthermore, the gene expression pattern in Rb deficient cells resembles that of FOXC2 overexpressing adipocytes (113). The phenotypes of Fat specific protein 27 (Fsp27) deficient mice are reminiscent of those of the Foxc2 overexpression mice. A significant elevated expression of Foxc2 and reduced expression of Rb have been observed in the white adipose tissue ofFsp2T'~ mice, indicating that Fsp27 acts somewhere upstream to Foxc2 and Rb signaling pathways (114). Several studies in humans have found that an inversed correlation between FOXC2 expression level and insulin resistance (112, 115). Those data appear to be consistent with the results from Foxc2 transgenic mice study and suggest that increased FOXC2 level may protect human from developing insulin resistance. A common FOXC2 polymorphism situated at the 5'UTR (-5120T) has been repeatedly found to be associated with obesity and dyslipidemia in various populations (116-118); whereas this polymorphism does not contribute to an increased risk for type 2 diabetes (117-119). Although allelic variation can lead to

21 differences in gene expression, studies still need to be performed to determine whether this polymorphism is of direct functional importance.

FOXC2 and Cancer During embryogenesis, epithelial-mesenchymal transition (EMT) is involved in gastrulation and neural crest formation, in which FOXC2 plays a role in specifying the mesenchymal cell fates (120). Carcinoma cells undergo EMT preceding invasion and metastasis. FOXC2 retains its ability to promote the mesenchymal differentiation of breast cancer as a downstream mediator of a variety of EMT-inducing factors including TGF[3l, Twist, Snail and Goosecoid. Foxc2 expression level is positively correlated with the aggressiveness of the mouse mammary tumor cell lines. In addition, FOXC2 is specifically overexpressed in highly aggressive basal-like breast cancers in human. Suppression of FOXC2 expression in highly metastatic cells results in reduced metastatic dissemination, whereas overexpression of FOXC2 in weakly metastatic cells promotes metastatic spread (120).

FOXC2 and Its Direct Target Genes Recently, a number of Foxc2 direct target genes have been identified. One of these genes is the plasminogen activator inhibitor type I gene {Pai-1), which is the primary physiological inhibitor of plasminogen activation and elevated plasma level of PAI-1 is a significant risk factor of ischemic heart disease (121). In endothelial cells, Foxc2 is able to directly induce the expression of Pai-1. FOXOl competed with Foxc2 for the insulin response element on the Pai-1 promoter and antagonized the activation of Pai-1 promoter by Foxc2 (121). In addition, Foxc2 cooperated with SMAD family member 3 (Smad3) and SMAD family member 4 (Smad4) to synergistically activate the Pai-1 expression through direct interaction with Smad3 and Smad4. TGFpl induced Pai-1 expression is impaired in heart and white adipose tissue of Foxc2+/~ mice. Taken together, Foxc2 appears to mediate the insulin and TGF(3 signaling to regulate Pai-1 expression (121).

22 However, the contribution of this event to the pathogenesis of ischemic heart disease and insulin resistance is still unclear. Foxc2 also directly regulates Integrin (53 (Itgb3), which is a pro-angiogenic factor and is up-regulated during tumor vascularization (122). Overexpression of Foxc2 enhances the adhesion and migration of endothelial cells, and enhances the sprouting and formation of microvessels in ex vivo aortic cultures; whereas these effects are blocked by Itgb3 neutralization antibody (122). In agreement with the overexpression experiment, Foxc2 knockout results in reduced Itgb3 expression, cell migration and outgrowth of microvessels from aortic cultures (122). These results indicate Foxc2 regulates by promoting Itgb3-mediated adhesion and migration of endothelial cells. Reexamining the Foxc2 transgenic mice has found that overexpression of Foxc2 in adipose tissues leads to altered vascular remodeling among adipose tissues due to the increased production of angiogenic factor angiopoietin-2 (Ang-2) by the adipocytes (123). Foxc2 is able to activate Ang-2 promoter directly in adipocytes and Foxc2 deficient mice show a lower level of Ang-2 expression. These findings imply that adipose tissue can adapt its vascularization to meet its metabolic demand by regulating Foxc2 expression (123).

FOXC1 and FOXC2 FOXCl and FOXC2 share 98% sequence identity of their forkhead domain with only two amino acid differences. FOXCl contains glutamic acid residues at positions 90 and 110, whereas FOXC2 carries aspartic acid residues at the corresponding positions in the FHD (45). Other than the FHD, FOXCl and FOXC2 share only about a 36% sequence identity (45) (Figure 1-18).

Overlapping Expression of Foxcl and Foxc2 FOXCl and FOXC2 not only show considerable sequence homology in their FHDs, but also share extensive overlap in their sites of expression during mouse development. Co-expression of Foxcl and Foxc2 has been observed in cartilage primordia of nasal capsule, nasal septum and meckel's cartilage, ribs and

23 developing vertebrae (49). Foxcl and Foxc2 are both transcribed in presomitic mesoderm and developing somites (51). In addition, Foxcl and Foxc2 have largely overlapping expression pattern in the periocular mesenchyme and in tissues derived from periocular mesenchyme (49, 58). Regions of co-expression in the developing cardiovascular system include the mesenchyme of the , the endothelium of the heart, the aortic and pulmonary valve leaflets and the mesenchyme surrounding the aortic arteries, the pulmonary trunk and the most anterior part of the descending aorta (53). Foxcl and Foxcl are also expressed in the ectomesenchyme and derived from cardiac neural crest cells, as well as in the proepicardium (124, 125). Regions of co-expression related to kidney development include the intermediate mesoderm, mesonephric mesenchyme and metanephric mesenchyme (55). Although the expression of Foxcl and Foxcl are largely overlapping, the two genes also exhibit some difference in their domains of expression. For example, in the developing vertebra, Foxcl is highly expressed in the vertebral body itself, whereas the expression of Foxcl is more restricted in domains other than vertebral body (49). No accompanying expression of Foxcl has been found in the developing cardinal veins from where the lymphatic vessels sprout (51, 99). In contrast to Foxcl, Foxcl transcripts are present at higher levels in the periphery layers of the developing cartilage primordia for nasal capsule and septum (49). Expression of Foxcl is lower than that of Foxcl in the ventral region of the mesonephric mesenchyme but higher around the ureteric bud (55).

Foxcl and Foxcl Double Knockout Mouse Models Foxcl and Foxcl deficient mice often demonstrate similar developmental defects in the tissues that they are co-expressed, indicating that these two closely related transcription factors play interactive roles during development. Foxcl ac "; Foxcl+ "compound heterozygotes have cardiovascular defects resembling those of Foxcllac2~'~ nulls and Foxcl'1' nulls, including Type B interruption, coarctation of the aortic arch, VSD, and valve dysplasia (53). Moreover, compound Foxcl and

24 Foxc2 homozygous null mutants die earlier with much more severe abnormalities than either single homozygote in cardiovascular development, affecting early stage of branchial arch remodeling, the remodeling of the primitive vascular plexus as well as the formation of the outflow tract, inflow tract and right ventricle (51, 124). In compound homozygous nulls, the epithelial somites and segmented paraxial mesoderm do not form, whereas individual homozygotes and compound homo/heterozygotes do have formation of epithelial somites, although they are misshaped and abnormally developed later (51). FoxcllacZ~'~ homozygous mutants have developmental defects in kidney and ureters including duplex kidneys and double ureters with hydronephrosis and hydroureter (55). The compound heterozygotes of Foxcl and Foxc2 (Foxcl ac +'; Foxc2im +") mutants have similar defects such as hydroureter and hypoplastic kidney, whereas the single heterozygotes are normal in kidney development (55). Taken together, the normal development of cardiovascular system, kidney and somites in each single heterozygote, the failure of compound heterozygote in complementation for the defects in these systems in each single homozygote, as well as the much more severe defects resulting from inactivation of all four alleles suggest that the Foxcl and Foxc2 play similar, dose-dependent roles in somitogenesis, cardiovascular and genitourinary development. However, there are phenotypic differences in the cardiovascular development between FoxcF"; Foxc2+' and Foxcl+'; Foxc2~'. For example, compound Foxcl*'; Foxc2~' mutant embryos have less severe but more extensive defects in the morphology and remodeling of the blood vessels in the head and branchial arches than Foxcl''; Foxc2+' embryos, indicating that the function of the two gene are not completely identical and/or the requirements for the onset of expression and the expression level of the two gene are different (51, 103, 124). Some studies have addressed the molecular mechanisms underlying the overlapping role of Foxc in heart and vascular development and have found that Foxcl and Foxc2 can regulate the expression of some common target genes (69, 126). One of these studies showed that Foxcl and Foxc2, in response to Shh signal, directly activate the transcription of Tbxl in the head mesenchyme through

25 a Foxc binding-element upstream of Tbxl (69). This study has discovered a Shh- Foxcl/Foxc2-Tbxl cascade during the aortic arch remodeling and identified Tbxl as the first known direct target of both Foxc proteins (69). In the process of arterial cell specification, VEGF (vascular endothelial growth factor) signaling activates the Delta-Notch pathway and consequently the downstream targets of Notch signaling including hairy/enhancer-of-split related with YRPW motif (Hey) transcription factors (126). Foxcl and Foxc2 play important roles in this process by directly inducing the expression of Delta-like 4 (D114), a ligand for Notch receptors, and Notch target gene hairy/enhancer-of-split related with YRPW motif 2 (Hey2) in endothelial cells (126). Phenotypical examinations of Foxcl heterozygous mice have found multiple abnormalities in the eye, including eccentric irregularly shaped pupils and displaced Schwalbe's line (61). Further histological analyses have discovered focally abnormal aqueous humor drainage structures, including small or absent Schlemm's canal and abnormal TM (61). Foxc2+ ' mice display similar abnormalities in anterior segment development to Foxcl+' heterozygotes. In addition to the similar irido-corneal angle manifestations, both Foxcl and Foxc2 heterozygotes exhibit comparable hypoplasia in iris stroma and iris pigment epithelium (61). In order to investigate the overlapping functions of the two genes, double heterozygotes have been examined and the same range and intermittent pattern of angle abnormalities as those seen in each single heterozygote have been found, suggesting that the two genes have common targets. However, double heterozygotes exhibit more severely affected iris than in either single heterozygotes (Figure 1-19), and the double heterozygotes have ciliary body defects, corneal vascularization and open eyelids that are not present in single heterozygotes, indicating a genetic interaction between the two genes and the overall level of both Foxcl and Foxc2 is important for the normal development of these tissues (61). Interestingly, the double heterozygotes do not show all the abnormalities found in Foxcl homozygous mutants, such as failure in the formation of corneal endothelium and anterior chamber. This suggests that Foxcl and Foxc2 have some unique functions in the anterior segment development.

26 In summary, the genetic interaction between Foxcl and Foxc2 during development is complex. From the sequence identity, overlapping expression pattern and the similar phenotypes of the gene knockout mouse models, it is evident that Foxcl and Foxc2 have very similar functions likely resulting from regulating the expression of a group of common target genes during embryonic development. Although the existence of two functionally similar Foxc proteins may provide redundancy for normal development, the overall level of Foxcl and Foxc2 appears also important in some situations. Furthermore, Foxcl and Foxc2 have some unique function, such as Foxcl in the formation of corneal endothelium and anterior chamber and Foxc2 in the development of lymphatic system, indicating they may have their unique targets or may activate some genes to different levels to allow the normal development to occur.

Rationale of My Thesis Research FOXCl and FOXC2 are two closely related transcription factors that play overlapping yet distinct roles during embryonic development. FOXCl and FOXC2 are both expressed in mesenchyme-derived tissues of the eye during development of the aqueous outflow tract. Transcription factors do not function alone but interact with other proteins to regulate gene expression. Specific protein-protein interactions can affect the stability, cellular localization as well as post-translational modifications of transcription factors so as to control transcriptional activity to an appropriate level. In addition, protein-protein interactions are also involved in assembly of transcriptional complexes, which may promote or repress transcriptional activation by altering local chromatin structure. Although protein-protein interactions are of importance in transcriptional regulation, the interacting partners of FOXC transcription factors have remained largely undefined. I hypothesize that FOXCl and FOXC2 transcription factors interact with other proteins to fulfill their normal function. Isolating FOXCl or FOXC2 interacting partners can aid in a better understanding of the role of FOXCl and FOXC2 in transcriptional regulation. In order to isolate interacting proteins for FOXCl and FOXC2, a human trabecular meshwork

27 cDNA library was screened by FOXC1 and FOXC2 respectively, in a yeast two- hybrid system. Putative interactions were confirmed in other experimental systems, and the effects of the interactions on FOXC were further characterized. Results from the above experiments are described in chapter 2 and 3 for FOXC1 and FOXC2, respectively. I discovered that human p32 is a FOXC1 interacting protein and human Protein Inhibitors of Activated STAT 3 (PIAS3) is a FOXC2 interacting protein. Human p32 and PIAS3 can regulate FOXC1 and FOXC2 transactivity, respectively. As well, the overlapping expression pattern in the eye, almost identical DNA binding domains and similar ocular phenotypes of Foxcl and Foxc2 mouse models lead me to the hypothesis that FOXC1 and FOXC2 share at least some common target genes in the eye. This hypothesis is explored in chapter 4. In this chapter, I also demonstrate for the first time that FOXC1 and FOXC2 can physically interact and that this interaction modulates the functions of FOXC 1 and FOXC2.

28 Figure 1-1: Structure of the forkhead domain.

A schematic diagram of the FOXA3 forkhead domain-DNA complex, a-helices are shown as coils. P-sheets are shown as arrows. HI, H2, H3: helices; SI, S2 and S3: p-sheets; Wl and W2: wings. Taken from Clark et al. 1993 (5).

29 ,wi

30 Figure 1-2: Clinical manifestation of Axenfeld-Rieger malformations.

(A) The underlying pupillary sphincter muscle is apparent due to iris hypoplasia. The pupil is elliptic shaped. Sclerocornea (9 o'clock). Protruding Schwalbe line (5, 11 o'clock). (B) Corectopia and hypoplastic iris with iris stromal tears. Sclerocornea is visible. (C) Redundant periumbilical skin. (D) Hypodontia. Taken from Lines et al. 2002 (15).

31 'A 1 X" f

1 i i I \x D

9 i « •.-=<>-x ) 1

32 Figure 1-3: Assignment of functional sub-domains to FOXC1 FHD.

The N-terminal portion of the FHD and helix 1 function in the organization of the FHD for transactivation, DNA binding and nuclear localization. Helix 2 contributes to FOXC1 transactivation. Helix 3 is involved in DNA-binding specificity of the FHD and also plays a role in nuclear localization and DNA- binding. The wing 2 region functions in the organization of the FHD for transactivation and DNA-binding. Taken from Berry et al. 2005 (45).

33 Wing

Transcriptional Activation

Nuclear Localization DNA-Binding Transcriptional Regulation CQ01

Nuclear Localization DNA-Binding & Speeifity DNA-Binding Wing 2 Transcriptional Activation

34 Figure 1-4: Summary of functional domains of FOXC1.

The FHD contains two nuclear localization signals (NLS-1 and NLS-2) are at the N- and C-terminal ends, respectively. Two transcriptional activation domains (AD-1 and AD2) are located at N- and C-terminal ends of FOXC1. A transcriptional inhibitory domain is downstream of the FHD. Taken from Berry et al. 2002 (38).

35 AD-1 AD-2

NLS-I NLS-2 Inhibitory Domain'

36 Figure 1-5: Expression of Foxcl RNA and protein in the developing somites.

Upper panel: Whole-mount in situ hybridization of 9.5 dpc embryos. The anterior PSM and somites have high expression level of Foxcl. Arrowhead: boundary between the newly formed (SI) and forming (SO) somites. Arrow: the region of highest Foxcl expression. Lower panel. Whole-mount immunohistochemistry of 9.5 dpc embryos detecting Foxcl protein levels. Taken from Kume et al. 2001 (51).

37 38 Figure 1-6: Skeletal defects in Foxclac homozygous mouse

(A) Arrow indicates the absence of calvarial bones in the mutant. Small arrow indicates thinner digits. Arrowhead indicates smaller ossification centers. (B) Lateral view of newborn skulls. The frontal (f), parietal (p), and interparietal (i) bones are missing in the homozygous mutant. The zygomatic process (z) is expanded and fused with the mandible (m). (C) Neural arch (na) of cervical vertebra is open dorsally in the Foxcl'1' mutant. (D and E) Ventral view of rib cages. Other than the manubrium (arrowhead) part, the sternum is absent in the mutant. The xiphoid process (x) is also misshaped in the mutant. Abbreviations: f, frontal bone; i, interparietal bone; m, mandible; mx, maxilla; na, neural arches; nc, nasal cartilage; p, parietal bone; z, zygomatic process; x, xiphoid process. Taken from Kume et al. 1998(48).

39 Newborn B Newborn «

+/+ -/- Q Newborn D Newborn £• „ Newborn na na tiff

+/+ -/-

40 Figure 1-7: Expression of Foxcl lacZ in the cardiovascular system.

(A, B) Whole-mount LacZ staining of 10.5 dpc embryos. (A) The single heterozygous (normal) embryo. (B) Homozygous mutant embryos. FoxcllacZ is expressed in the developing the branchial arches, outflow tract (arrowhead), and ventricle (v). (C, D) Dark-ground illumination views of sections of embryos in (A, B). LacZ staining is shown in pink. Foxcl'acZ is expressed in the branchial arches (numbered 3, 4, and 6). (E) Transverse section of an 11.5dpc FoxcllacZ/+ heart shows that LacZ expression in the endothelial cells of the ventricles. Arrowhead indicates the . (F) Transverse section of the 11.5dpc FoxcllacZ/+ outflow tract shows LacZ staining within the muscular wall. Arrowheads indicate a low level of LacZ expression in the endocardial cushion tissue of the future spiral septum. (G) Foxcl,acZ is highly expressed in the muscular wall of the pulmonary trunk (pt) and in the pulmonary valves (pv) of the normal newborn FoxclhcZ/+ heart. (H) Section of another normal newborn heart shows LacZ expression in the mitral (mv) and tricuspid (tv) valves. (I) Dark- ground illumination view of normal newborn pulmonary trunk region shows high levels of FoxcllacZ expression in the pulmonary trunk (pt) walls but lower expression in the closed ductus arteriosus (da). (J) High-power magnification of the tricuspid valves in H. (K-M) FoxcllacZ expression in the normal adult FoxcllacZ/+ heart. Foxcl,acZ expression is persistent in the aortic arch (aa) and pulmonary trunk (pt) and in the coronary vessels (cv) of adult heart. Scale bars: 500 mm in G and H. Abbreviations: aa, aortic arch; cv, coronary vessels; da, ductus arteriosus; mv, mitral valves; pt, pulmonary trunk; pv, pulmonary valves; tv, tricuspid valves; v, ventricle. Taken from Winnier et al. 1999 (53).

41 newborn

42 Figure 1-8: Cardiovascular defects in FoxcllacZ mutant mice.

(A) Ink injection of a wild-type heart reveals the aortic arch (aa), pulmonary trunk (pt), and descending aorta (ao) and the complete closure of the ductus arteriosus (da). (B) Normal FoxcllacZ/+ heart, showing decreased expression of FoxcllacZ in the closed ductus arteriosus. (C) FoxcllacZ/,acZ mutant displays patent ductus arteriosus and the defects of the aortic arch. (D) FoxcllacZ/lacZ mutant displays patent ductus arteriosus. Arrowheads indicate coarctation of the aortic arch between the left common carotid artery (lcca) and the left subclavian artery (lsa). (E, F) Sagittal section through the semilunar valves (va) of the pulmonary trunk of normal newborn heart and Foxc 1lacZ/lacZ mutant heart, showing severe dysplasia in the mutant (F). The mutant heart also shows decreased thickness of the myocardium. (G, H) Sections of wild-type and mutant hearts at 13.5 dpc. Arrowhead indicates the ventricular septal defect in the mutant (H). Abbreviations: aa, aortic arch; ao, decending aorta; as, ventricular septum; bca, brachiocephalic artery; da, ductus arteriosus; lcca, left common carotid artery; lsa, left subclavian artery; pt, pulmonary trunk; va, semilunar valves; s, ventricular septum. Scale bars: 210 mm in E and F; 160 mm in G and H. Taken from Winnier et al. 1999 (53).

43 ^ newborn newlorrt va. */+ H ./.

s

13,5 m&

44 Figure 1-9: Expression of Foxcl in the developing kidney.

(A) Expression of Foxcl within the posterior region of 8.5 dpc embryo viewed by in situ hybridization. Foxcl is expressed in the intermediate mesoderm (i), the presomitic mesoderm (p) and somites (s). (B) Transverse sections at the level of the anterior trunk of a 9.5 dpc embryo. Expression of Foxcl is within the mesonephric mesenchyme alongside the Wolffian duct (arrowheads) (C) Transverse sections through the presomitic mesoderm of 9.5 dpc embryos. Foxcl is expressed in the mesoderm with highest levels closest to the neural tube (n). (D) Sagittal sections through the metanephric mesenchyme at 10.5 dpc. Foxcl is expressed in the metanephric mesenchyme. (E) Transverse sections through the developing metanephric kidney at 12.5 dpc. Expression of Foxcl is seen in the condensing mesenchyme. Abbreviations: da, dorsal aorta; i, intermediate mesoderm; m, metanephric mesenchyme; n, neural tube; p, presomitic mesoderm; s, somites; u, ureteric bud; w, Wolffian duct. Taken from Kume et al. 2000 (55).

45 8 JS dpc

mmssv 1 B lm

9.5 dpc

^©'Mii F®£i£'J D

10.5 dpc

umj 12.5 dpc

mim$

46 Figure 1-10: Kidney and ureter defects in Foxclc homozygous mice

(A, B) Wild-type and mutant newborn kidneys. (A) Asterisks illustrate hydroureters in male mutant kidneys. Mutant testes (arrowheads) are located more anteriorly compared to wild-type testes (arrowheads). (B) Asterisks illustrate hydroureters in female mutant kidneys. White arrow indicates the normal ureter behind the hydroureter. Yellow arrows indicate the oviducts of mutant. Mutant ovaries are located more anteriorly (arrowheads) compared to the wild type. (C, D) Sections of newborn wild-type (C) and mutant (D) kidneys. Arrow indicates the border of the peripheral metanephrogenic mesenchyme of the duplex kidney in mutant. The hydroureter (asterisk) connects to the upper part of the duplex kidney. (E) Dorsal view of newborn mutant kidneys with double ureters. Arrows illustrate the normal ureters and asterisks illustrate ectopic hydroureters. Taken from Kume et al. 2000 (55).

47 r ^ V

I

r-^ \/> m k i/Ms,.^ yJm$4&k %r?g: j2%%jjg' If i!

I X

Fojfei^^asrff*^1' Figure 1-11: Expression of Foxclac in the developing eye.

(A) At 11.5 dpc FoxcllacZ is expressed in mesenchyme cells that are around the eye, between the surface ectoderm and the lens and between the lens and the retina close to the future hyaloid region (hy). (B) At 12.5 dpc, Foxcl,acZ expression level is reduced in the corneal mesenchyme. (C) High-power magnification view of the boxed region in (B) shows Foxcl is not expressed in the corneal epithelium (arrowhead) and corneal stroma. (D) Foxcl expression is restricted to the trabecular meshwork (tm) and epithelial conjunctival cells (en) of the fornix by 16.5 dpc. The anterior chamber is not visible due to the growth of the lens (open arrow). (E) The same expression pattern continues into 18.5 dpc. At this stage, the anterior chamber is visible. Abbreviations: cm, corneal mesenchyme; c, cornea; en, conjunctival cells; el, eyelid; hy, hyaloid; ir, iris; le, lens; r, retina; tm, trabecular meshwork. Adapted from Kidson et al. 1999 (59).

49 +/- em

I * -hy-V

tt.Si B ----- j > "*'*,. * ' •. •-• ^:i. ,,- •'• ; ,'ii "_J '* t' «*•*

* iBilfjEi - s

w*' *gtp->? , **r~*rrcTW .,-*

fc/£V**F7S, £%*

tm,

50 Figure 1-12: Axenfeld Rieger-Iike ocular defects in Foxcf heterozygotes

Ocular phenotypes are shown for mice of the indicated genetic backgrounds and genotypes (B6, C57BL/6J; 129, 129/SvEvTac; CH, CHMU/Le). (a, b) These eyes are grossly normal with small, central round pupils, (c) In this B6 Foxcl+ ' mouse, irregular shaped pupil is located to one side of the iris, where Schwalbe's line is anteriorly placed (arrow), (d) Elongated pupil is observed. Asterisks illustrate the displaced iris pigment epithelium (ectropian uveae). Arrowheads indicate peripheral irido-corneal attachments that appear to exert traction on the iris, (e, f) In both heterozygotes, the pupils are irregularly shaped and are not centrally positioned. Both eyes display scleralization of the (arrow). In (f), the attachment of the inferior iris to the posterior corneal surface produces an irregular corneal opacity (arrowheads). Taken from Smith et al. 2000 (61).

51 CH Foxc1ch/*

52 Figure 1-13: Clinical features of lymphedema-distichiasis syndrome.

(A) Bilateral lower limb lymphedema. (B) Distichiasis of lower eye lid reveals several eye lashes growing upwards and vertically. Taken from Connell et al. 2008 (127) and Traboulsi et al. 2002 (89).

53 lr~\\

Hfc h. 2*1

«.r-f,> -

54 Figure 1-14: Expression of Foxc2 in 7.5 to 11.5 dpc mouse embryos.

(A) Expression of Foxc2 in seen in paraxial mesoderm of a 7.5 dpc embryo. The node or notochord (arrowhead) lacks expression of Foxc2. (B) Strong expression of Foxcl is seen in the presomitic mesoderm (p), somites (s), and cephalic mesoderm (cm) of an 8.5 dpc embryo. (C) At 9.5-dpc, Foxc2 is expressed in the periocular region, the second and third branchial arches, and the presomitic mesoderm, somites, and cephalic mesoderm. The white arrowhead indicates the strong expression of Foxc2 in the posterior region of the somite. (D) At 10.5 dpc, expression level of Foxc2 is down-regulated in anterior somites, but still high in the posterior somites and tail bud. (E) At 11.5 dpc, external view of a hemisected embryo shows Foxc2 is expressed in the periocular mesenchyme (e), vertebral column, forelimbs (fb), and hindlimbs (hb), and in the branchial arches (b). (F) Internal view of the same embryo shows high level of expression in the nasal region (n) and blood vessels. Abbreviations: b, branchial arch; cm, cephalic mesoderm; fb, forelimb; h, heart; hb, hindlimb; mcv, mesenchymal condensation of vertebrae; n, node; p, presomitic mesoderm; s, somites. Taken from Winnier et al. 1997 (95).

55 56 Figure 1-15: Aortic arch defects in Foxc2-nul\ mice at birth.

(A) Normal development of aortic arch from a wild-type mouse. (B) In a Foxc2- null mouse, part of the aortic arch between LCCA and LSA does not exist (Type B interruption). Abbreviations: Ao, aorta; BCA, brachiocephalic artery; DA, ductus arteriosus; LCCA, left common carotid artery; PT, pulmonary truncus. Taken from Iida et al. 1997 (96).

57 'AT PBi

,G©©& s®\

•KM m ?>? S»J ..5 \ ^ r-J Figure 1-16: Defects in the skeletal development of Foxc2-null mice.

Wild type (A, C) and Foxc2-rm\\ (B, D) mice. (A, B) Ventral views of the skull base. (A) In the wild-type animal, the secondary palate consists of palatal shelves of the maxilla (small asterisks) and palatine (large asterisks). (B) In Foxc2-mx\\ mice, the palatal shelves of the maxilla (thin arrows), the palatine (bold arrows) and the pterygoid bone (bold open arrows) are laterally displaced. The basisphenoid bone (BS) was slightly shortened and a small defect was observed in the mediocaudal region (open arrow). (C, D) Whole-mount view of the vertebral column shows that the number of vertebrae is unchanged, but the vertebral column of the Foxc2-null mice is significantly shortened (D). In the cervical region of the mutant mouse, ossification centers in the vertebral bodies do not exist (D). Taken from Iida et al. 1997 (96).

59 WT F0XC2-A

1 K'ls

*--Ssi

m

I

60 Figure 1-17: Defects in lymph system of Foxc2 ' heterozygous mice.

Wild type (A, C) and Foxc2+/' (B, D) mice. (A, B) Evans blue dye (EBD) injections in the skin of the ear. (A) Normal drainage pattern of the wild type mouse. (B) Increased area of blue-stained draining lymphatics in Foxc2+'. (C, D) Intradermal hindpaw EBD injection. (C) Normal appearance of colorless lymphatics in the liver hilum of a wild type mouse. (D) Reflux of EBD-stained lymph from the cisterna chyli into dilated hepatic hilar lymphatics (arrows) in a Foxc2+/~ mouse. Taken from Kriederman et al. 2003 (101).

61 62 Figure 1-18: Protein sequence alignment of human FOXC1 and human FOXC2

The amino acid sequences for FOXC1 and FOXC2 are aligned using the ClustW program. The forkhead domain is boxed. Only two amino acid residues are different between FOXC1 and FOXC2 within the forkhead domain. Asterisk indicates identical amino acids. Colon indicates conservative substitutions. Dot indicates similar amino acid residues. Taken from Berry et al. 2005 (45).

63 FOXC1 MQARYSVSSPNSLGV^TYLGGEQSYYRAAAAAAGCGYTAMPAPMS^'SHPAHAEQYPGGM 60 FOXC2 MQARYSVSDPNALGVVPYLS-EQNYYRAAGSYGG MASPMGVYSG—HPEQYSAGM 52

F0XC1 ARAYGPYTPQP—QPKDM\KPPYSYIALIT)MIQNAPDKKITLNGIYQFIMDRFPFYRDN 118 FOXC2 GRSYAPYHHHQPAAPKDUKPPYSYIALIBIAIQNAPEKKITLNGIYQFIiiDRFPFYREN 112

3§t • ^ ${;$: * ^fc^Mc *

FOXC1 KQGWaNSIRH^SLNTCn^TRDDKKPGKGSYlTLDPDSYmiFENGSFLRRRRRFKKKD 78 F0XC2 KQGWaNSIRlM,SLNf:CRTaTRDDKKPGKGS\TTLDPDSYmiFENGSFLRRRRRFKKKD E72

F0XC1 AVKDKEEKDRLHLKEPPPPGRQPPPAPPEQADGNAPGPQPPPTOIQ-DIKTENGTCPSPP 237 F0XC2 VSKEKEE—RAHLKEPPPAASKGAPATPHLAD—APKEAEKKWIKSEAASPALPVITKV 228

FOXC1 QPLSPAAALGSGSAAAVPKIESPDSSSSSLSSGSSPPGSLPSARPLSLDGADSAPPPPAP 297 F0XC2 ETLSPESALQG SPRSAAS—TPAGSPDGSLP EH 259

FOXC 1 SAPPPHHSGGFSVDNIMTSLRGSPQSMAELSSGLLASAAASSRAG-IAPPLALGAYSPG 356 F0XC2 HAAAPNGLPGFSmMMT-LRTSP—PGGELSPG AGRAGLWPPLAL-PYAAA 308

FOXCl QSSLYSSPCSQTSSAGSSGGGGGGAGAAGGAGGAGTYHCNLQAMSLYAAGERGGHLQGAP 116 F0XC2 PPAAYGQPCAQGLEAGAAGG YQCSMRAMSLYTGAERPAHMCVPP 352

FOXCl GGAGGSA\T)DPLPDYSLPPVTSSSSSSLSHGGGGGGGGGGQEAGHH 462 FOXC2 ALDEALSDHPSGPTSPLSALMAAGQEGALAATGHHHQHHGHHHPQAPPPPPAP 406

FOXCl PAAHQGRLTSWVlNQAGGDLGHIASAAAAAAAAGYPGQQQNFHS\'REi!FESQR 515 FOXC2 QPQPTPQPGAAAAQAASmNHSG-DLNHLP GHTFAAQQQTFPNVREMFNSHR 4 58

FOXCl IGLNN SPVNGNSSCQKAFPSSQSLYRTSGAFVYDCSKF 553 FOXC2 LGIENSTLGESQVSGNASCQLPYRSTPPLYRHAAPYSYDCIKY 501

64 Figure 1-19: Ocular defects in Foxcl heterozygous, Foxc2 heterozygous as well as Foxcl and Foxc2 double heterozygous mice.

All sections are from 129BS mice. (A-D) Irido-corneal angle sections. (E-H) Iris sections. (I-L) Corneal sections. (A) In wild-type mice, the Schlemm's canal (SC) is within the region confined by the two arrowheads. The normal appearance of the trabecular meshwork (TM, arrows) is normal and the inter-trabecular spaces can be seen. (B) In this Foxcl+/~ heterozygote, SC and TM (arrows) are shorter than normal. (C) In this Foxc2+' heterozygote, SC is short than normal (between the two arrowheads) and the trabecular beams are hypoplastic. Arrow indicates an abnormally located ciliary process. (D) In this Foxcl+"; Foxc2+ ' double heterozygotes, no SC and TM are observed and there is a long irido-corneal attachment. Arrow indicates the malformed ciliary body. (E) Normal iris section shows that the iris pigment epithelium (arrowhead) is separated from the anterior iris stroma (S) by the dilator muscle (white arrows). (F, G) In either Foxcl+l~ or Foxc2+l~ heterozygotes, the iris stroma and iris pigment epithelium are thinner than normal. (H) In Foxcl and Foxc2 double heterozygote, the iris pigment epithelium is flattened and the iris stroma is almost completely absent. (I, J, K) In wild-type and single heterozygous eyes, normal appearance of corneal epithelium and stroma is seen. (L) In double heterozygote, abnormal stromal vascularization is observed. Scale bars are ~40 um in all cases. Taken from Smith et al. 2000 (61).

65 WT Foxrt*/- Foxc2+/- Foxrf*/-; Foxc2*/-

A 1298S + + B 129BS Foxcl** c D 129BS fowl" FOXC2"- > '•.,.„-«' : ,„ n ,-—-• * •' ' , . ^: , .. • ^^^ — ~^--—-.;-• -• m -* ,v \^o.- ; A „- • ' - }% U'.-V"^*-'- -~ V •' - -*,-, -,,. JJ$*-.= •• -'- ^*" tj?^SSife^ - v '"-'V ••' ' .-•)••' V

66 Chapter 2. Human p32 is a novel FOXCl-interacting protein that regulates FOXC1 transcriptional activity in ocular cells

This chapter contains work published in

Huang L, Chi J, Berry FB, Footz TK, Sharp MW, Walter MA. Human p32 is a novel FOXCl-interacting protein that regulates FOXC1 transcriptional activity in ocular cells. Invest Ophthalmol Vis Sci. 2008 Dec;49(12):5243-9.'

1 Chi. J helped with mutation screening of p32 gene; Berry, FB did immunofluorescence to detect the cellular localization of the endogenous p32 and FOXClin figure 2-5; Footz TK made deletion constructs ofp32; Sharp MW involved in making HTM cDNA library and helped with Y2H screening.

67 Introduction The forkhead family of transcription factors is characterized by a 110-amino- acid DNA binding domain termed the forkhead domain (FHD) (128). This DNA binding domain was first identified as a region of homology between the Drosophila melanogaster fork head protein (1) and the rat hepatocyte nuclear factor 3 protein (129). The FHD is highly conserved in a variety of species, from yeast to humans. The FHD forms a winged-helix-turn-helix structure consisting of a three-a helix bundle and two large loops that form 'wing' structures (5). Members of the forkhead family are necessary for a wide range of cellular and developmental processes, including cell migration and differentiation, organogenesis, and tumorigenesis (4, 130). FOXC1 is a member of the forkhead transcription factor family, and plays an important role in eye organogenesis. Mutations in FOXC1 cause Axenfeld-Rieger malformation, an autosomal dominant developmental disorder (32, 33). The ocular defects of AR patients consist of anomalies in the structure of the anterior chamber angle, including trabecular meshwork defects, adhesions of the iris and cornea, iris hypoplasia, corectopia and posterior embryotoxon. The anterior angle malformation in AR can lead to an increased intraocular pressure and consequently, glaucoma. Over half of AR patients suffer from an early onset secondary glaucoma. Systemic manifestations may also present in some patients, such as dental dysgenesis and redundant periumbilical skin (15). In rare cases, patients have congenital cardiac defects or hearing loss (17, 131). Foxcl+/- heterozygous mice have ocular defects similar to those found in patients with AR (61). Several different mechanisms underlie the impairments to FOXC1 function caused by the disease-causing mutations, including reduction in protein stability, alteration to DNA-binding specificity, alteration in nuclear localization, and defects in DNA-binding capacity and transactivation (39-42, 44). Since mutations that lead to reduced FOXC1 activity as well as chromosome duplications that result in an extra copy of FOXC1 both cause ocular defects (35-37), it appears

68 that there are strict requirements for upper and lower thresholds of FOXC1 activity for normal eye development and functions. The genetic pathways in which FOXC1 is involved in ocular developmental events are not fully understood. In the eye, recent work suggests that retinoic acid signaling may be upstream of FOXC1, as the expression of Foxcl in the periocular mesenchyme cells is reduced in mouse embryos with both retinoic acid -synthesizing retinaldehyde dehydrogenases (Raldhl/3) knocked-out (74). Moreover, the Raldhl/3 knockout mouse displayed ocular malformations affecting the anterior chamber. Recent experiments also revealed potential genes downstream of FOXC1 in ocular genetic pathways (65, 70, 132). These genes implicate FOXC1 in a variety of processes, including IOP regulation and eye organogenesis. Studying protein-protein interactions is also necessary to understand genetic pathways due to their important roles in regulation of the activities of transcription factors. Previous work from our laboratory found that the actin- binding protein Filamin A interacts with FOXC1 and directs FOXC1 into HP la heterochromatin-rich regions of the nucleus, resulting in the down-regulation of FOXC1 transactivation (77). Mutations in PITX2, encoding a homeodomain transcription factor also cause AR malformations (26). Very recent results indicate that FOXC1 and PITX2 can physically interact with each other, tying both proteins into a common pathway. PITX2 is a negative regulator of FOXC1, and PITX2 loss-of-function mutants lose their ability to inhibit FOXC1 (78). These studies indicate that FOXC1 activity is stringently controlled by protein- protein interactions. Here we identify human p32 as a binding partner for FOXC1 through a yeast two-hybrid (Y2H) experiment. The FHD of FOXC1 was identified as the p32 interaction domain. The proportion of p32 that resides in the nucleus colocalizes with FOXCl. Interestingly, the interaction with p32 results in reduced FOXC1 transactivation ability. Failure of p32 to interact with FOXCl containing the disease-causing Fl 12S mutation indicates that impaired protein interaction may

69 be a disease mechanism underlying AR, and further demonstrates that correct regulation of FOXC1 is necessary for normal ocular development and functions.

70 Methods

Plasmids N-terminally tagged 6xHisXpress::FOXCl fusion protein was expressed in pcDNA4 His-Max B. The construct was assembled by Ramsey Saleem and has been previously described (41). N-terminally tagged 6xHis::FOXCl bacterial expression vector, pET28-FOXCl, was built by Fred Berry and has been previously described (38). Enhanced green fluorescent protein (EGFP) FOXC1 fusion protein was expressed by subcloning FOXC1 into pEGFP-Nl (Clontech). This construct, FOXC1EGFP, was built by Ramsey Saleem and has been previously described (38). FOXC1 full length cDNA was amplified by PCR from pcDNA4-FOXCl and subcloned into pDEST32 (Invitrogen) in-frame to the GAL4 DNA-binding domain (GAL4DBD) or into the pcDNA3.1/nV5-DEST vector (Invitrogen) in-frame to the V5 epitope by gateway technology. FOXC1FHD and FOXC1 FID were amplified by PCR from pcDNA4-FOXCl and subcloned into pcDNA3.1/nV4-DEST vector by gateway technology. FOXC1AFHD was amplified by PCR from pcDNA4-FOXCl ABOX, which is kindly given by Fred Berry, and subcloned into pcDNA3.1/nV4-DEST vector by gateway technology. pCMV-p32 was purchased from Open Biosystems. p32 was generated by PCR amplification and subcloned into the pcDNA3.1/nV5-DEST by gateway technology or pET28b vector (Novagen) in-frame to the 6xHis tag. All newly built vectors were sequenced to confirm that no mutations were introduced into the cDNAs and the cDNAs were in-frame to the epitopes. FOXC1 mutants in pcDNA4 were also amplified by PCR, subcloned into pcDNA3.1/nV5-DEST vector and resequenced. The pGL3-TK-luciferase reporter vector under control of 354 bp fragment from FGF19 promoter (FGF19RE), which contains a FOXC1 binding site, was constructed by Yahya Tamimi and has been previously described (65).

Yeast two-hybrid screen A human trabecular meshwork (HTM) cDNAs library was custom built by Invitrogen. In brief, cDNAs were synthesized from mRNAs extracted from a

71 primary HTM cell line and subcloned into pEXP-AD502 (Invitrogen) in-frame to the GAL4AD. Before the library screen, self activation of bait FOXC1 protein and the concentration of 3-amino-l, 2, 4-Triazole (3AT; Sigma) used in subsequent yeast culture were determined as in the manufacturer's protocol (Invitrogen). lOug of the full length FOXC1 bait construct was co-transformed with lOug of isolated library DNA according to the manufacturer's protocol and plated onto 15cm SC-Leu-Trp-His+25mM 3AT agar plates. The plates were kept at 30°C for 7 days. Colonies growing on the plates were picked and streaked onto SC-Leu-Trp plates to isolate single purified colonies from the original. Four colony isolates per original colony were streaked onto master plates (SC-Leu-Trp) along with the yeast control strains supplied by the manufacturer. After an overnight incubation, master plates were replica plated onto four assay media, which are SC-Leu-Trp-Ura, SC-Leu-Trip-His+25mM 3AT, SC-Leu-Trp+0.2% 5- fluoroorotic acid (5FOA; Sigma) and YPAD containing a nitrocellulose membrane for an X-gal assay. Following the suggested incubation and replica cleaning, as well as the X-gal assay, colonies displaying a putative interaction phenotype were selected. FOXC1 bait plasmid and prey pEXP-AD502 cDNA plasmid were extracted from these colonies and bacterially transformed into ElectroMAX DH10B cells (Invitrogen) or XLIO-Gold Kan Ultracompetent cells (Stratagene). Transformations were plated on LB plates containing gentamicin (lOug/ml) to recover the bait plasmid or ampicillin (100ug/ml) to recover the prey plasmid. Prey plasmids were then partially sequenced using pEXP-AD502 forward primer at the DNA sequencing facility in the department. Prey cDNA insert sequences were compared with publicly available genome database using the BLASTn program at NCBI to reveal identity. A total of 1.9X10e6 library clones were screened for growth on selective media and assayed for 0- galactosidase activity. In order to rule out false positive results, the candidate prey plasmids were retransformed into yeast cells with the empty bait vector or FOXC1 bait construct or an irrelevant PITX2 bait construct provided by Mike Sharp, followed by the same selective procedures as mentioned above. Clones that reproduce the original phenotypes likely contained FOXC1 interacting proteins.

72 Mammalian cell culture and transfection HeLa cells, COS-7 cells and HTM cells were maintained in Dulbeco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum in a 37°C, humidified incubator under an atmosphere containing a constant 5% CO2. The day prior to transfections, plates were seeded (106 cells in 10ml media per 100mm Petri dish, 4xl05 cells in 4ml media for 60mm Petri dish, 2xl05 cells in 2ml media per 25mm well or 4x10 cells in 1ml media per well of 24 well plate) so as to reach approximately 30-50% confluence the following day. Transfections were performed using FuGene 6 (Roche) according to the manufacturer's protocol. 4ug, 1.6ug or 0.8ug of total DNA was used for transfections in 100mm, 60mm or 6 well dishes, respectively. 1: 3 (DNA/FuGene) ratio was used in COS-7 and HeLa cells transfection, while 1: 5 (DNA/FuGene) ratio was used in transfection of HTM cells. Transfected cells were subjected to immunofluorescence on the next day of transfection. Dual-luciferase assays (Promega) and EMSAs were performed 48 hours after transfection.

Immunoblot analysis Protein samples were loaded onto 10% polyacrylamide gel and subjected to electrophoresis on a Protean 3 minigel system (Bio-Rad) at a constant 150V for 1 hour in running buffer (25mM Tris, 200mM glycine, 0.2% SDS). A prestained protein size marker (New England Biolabs) was used to trace both electrophoresis and transfer. Proteins separated on gels were transferred onto nitrocellulose membrane (Bio-Rad) using a Protean 3 semidry transfer apparatus (Bio-Rad) for 1 hour at a constant 100V at 4°C in transfer buffer (25mM Tris, 200mM glycine, 20% methanol). Successful transfer was confirmed by Ponceau S stain (Sigma) and by the presence of the prestained protein marker lane on the membrane. Blots were blocked in 5% (W/V) skim milk in TBST (lOmM Tris-Hcl, pH7.5, 75mM NaCl, 0.05% Tween 20) for 30 minutes. Following the blocking, blots were incubated with primary antibodies (1:5000 for anti-Xpress and anti-V5, 1:10000 for anti-p32) in TBST, 5% milk at 4°C overnight or at room temperature for 1 hour and subsequently washed 3 times for 10 minutes in TBST. Then blots were

73 subjected to hybridization with secondary (HRP-conjugated) antibodies (1:5000) in TBST, 5% milk at room temperature for 1 hour and subsequent 3 washes in TBST, 10 minutes for each. HRP-conjugated secondary antibody signal was visualized by using SuperSignal West Pico chemiluminescent substrate (Pierce) and exposing to medical X-ray film (Fuji) for various intervals. Films were developed in an M35A X-OMAT automatic film processor (Kodak).

Ni2+-NTA pull-downs Part I Bacterial protein expression and purification BL21 Rosetta (Novagen) bacterial cells transformed with pET28 based constructs were cultured in 100ml LB liquid media containing 50ug/ml kanamycin (final concentration) and 34|ig/ml chloramphenicol (final concentration) at 37°C with vigorous shaking (250rpm/min). When cell cultures reached OD600-0.5 to 1, protein expression from the pET28 based plasmids was induced by ImM IPTG for 3 hours. Following the induction, bacteria cells were spin down at 5000 x g for 15 minutes at 4°C and stored at -80°C or continue with protein purification. Cell pellets were resuspended in 1ml binding buffer (50mM sodium phosphate, pH 8.0, 300 mM NaCl, lOmM Imidazole , 0.05% Tween 20) and treated with lmg/ml lysozyme (final concentration) on ice for 30 minutes. Sonication using a Sonic dismembrator 60 (Fisher Scientific) at setting 5 for three-6 second bursts was conducted to break the bacteria. The lysates were treated with 5 ug/ml Dnase I (final concentration) and 5 ug/ml Rnase I (final concentration) for 10 minutes on ice. Lysates were cleared of debris by microcentrifuging for 30 minutes at 10, 000 x g at 4°C. lOOul Ni2+-NTA beads (Qiagen) washed beforehand with binding buffer for 3 times and prepared individually for each sample were incubated with the lysates for 60 minutes at 4°C. Then the beads were collected and subjected to three-20minutes washes with wash buffer (50mM sodium phosphate, pH 8.0, 300mM NaCl, 50mM Imidazole, 0.05% Tween 20) as well as three-20 minutes subsequent washes with low pH buffer (50mM sodium phosphate, pH 8.0, 500mM NaCl, 50mM Imidazole, 0.05% Tween 20). Following the final wash, low pH buffer was removed as

74 completely as possible by using 1 ml syringe (BD Syringe) with 26G lA needle (Bection Dickinson). Finally the beads were resuspended in equal volume of MilliQ H20 and stored at 4°C. The success of the bacterial expression and the binding of 6xHis tagged proteins to the beads were examined using immunoblot analysis with anti-6xHis antibodies and coomassie blue (ICN Biomedicals) staining. Part II Ni2+-NTA pull-downs To preclear mammalian protein extract, 50 ul of Ni2+-NTA beads, prepared individually for each pull-downs and rinsed with binding buffer for 3 times were incubated with 300 ug of protein extract in 1 ml binding buffer for 1 hour at 4 °C. After preclear, Ni2+-NTA beads were spin down and the protein extract was transferred into a fresh 1.5 ml eppendorf tube. 20 ul of Ni2+- NTA beads bound with purified 6xHis tagged protein (from Part I) were then added into mammalian protein extract. A 1 hour incubation at 4°C was performed to allow mammalian proteins to interact with 6xHis tagged protein on Ni +- NTA beads, followed by four washes with wash buffer (46.6mM Na2HP04, 3.4mM NaH2P04, 300mM NaCl, 50mM Imidazol pH6.0, 0.05% Tween20). Protein complexes captured on the beads were eluted in 20 ul 2xSDS loading buffer (0.125M Tris-Cl pH6.8, 20% glycerol, 2% P-mercaptoethanol, 4% SDS, 0.2% Bromophenol Blue) and subjected to immunoblot analysis with antibody against the mammalian-expressed proteins. The input fraction represented 5% of the mammalian protein extracts used in the pull-downs.

Immunoprecipitation (IP) HTM cells were transfected with pcDNA3.1/nV5-DEST-FOXCl or the empty vector. 48 hours after the transfection, the cells were lysed in the lysis buffer (50mM Tris-Cl pH 8.0, 150mM NaCl, 0.5% Igepal). lOOul of protein lysates were added into 400ul of extra buffer (12.5mM Tris-Cl pH 8.0, 87.5mM NaCl, 0.125% Igepal) containing protein G-agarose (Sigma) and precleared for 2 hours at 4°C. The precleared cell lysates were incubated with 3 ug of anti-V5 antibody overnight at 4°C with rotation. 25 ul of protein G-agarose blocked

75 beforehand by 1 mg/ml bovine serum albumin (BSA; Sigma) in the wash buffer (20mM Tris-Cl pH8.0, lOOmM NaCl, 0.2% Igepal) were added and incubated with the precleared cell lysates for 4 hours at 4°C. Then the protein G-agarose was collected and washed four times with the wash buffer. Final elution of bound protein complexes from the beads was accomplished by adding 20ul of 2xSDS loading buffer, vortexing briefly, and heating at 95°C for 5 minutes. Samples were then analyzed by immunoblot with anti-p32 antibody. Input fraction represented 5% of the protein extracts used in the immunoprecipitation.

Immunofluorescence After one day of transfection with pcDNA-FOXCl plasmid, HTM cells plated onto coverslips were subjected to immunofluorescence. Plates were kept at room temperature through the entire protocol and covered by foil paper at all possible times in order to prevent photobleaching of fluorochrome-conjugated secondary antibodies. Cells were washed in PBS for 2 times and fixed for 20min in PBS containing 2% w/v paraformaldehyde. Two-5 minute washes with PBSX (PBS+0.05% Triton X-100) were conducted to permeate cell and organellar membranes, followed by incubating the coverslips in PBSX containing 5% (w/v) BSA for 15min to reduce nonspecific protein binding of antibodies. Coverslips were then incubated for 1 hour with the first primary antibodies to Xpress epitope (diluted 1:500 in lOOul PBSX containing 5% BSA), washed twice with PBSX, followed by the incubation with Cy2 conjugated anti-mouse secondary antibodies (diluted 1:500 in lOOul PBSX containing 5% BSA) for another 1 hour. To detect the endogenous p32, cells were treated in the similar fashion with sequential incubation for 1 hour with antibodies against p32 (diluted 1:500 in lOOul PBSX containing 5% BSA) and Alexa fluor 594 anti-goat secondary antibodies (diluted 1:500 in lOOul PBSX containing 5% BSA) for another 1 hour. With the aim of detecting endogenous FOXC1, an overnight incubation with the primary antibodies (diluted 1:500 in lOOul PBSX containing 5% BSA) was carried out in a humid chamber followed by fluor 594 anti-goat secondary antibodies treatment for 1 hour. 4',6-diamidino-2-phenylindole (DAPI) staining was performed by

76 incubating the coverslips with DAPI (5mg/ml solution diluted 1:500 in lOOul PBS) for 5 minutes. After twice-5 minutes final rinses with PBS, coverslips were mounted onto glass slides with mounting medium (90% Glycerol, 10% PBS containing lug/ml p-phenylenediamine) and sealed with nail polish. Slides were stored in dark at -20°C or subjected to imaging analysis immediately. The images were collected on a Zeiss LSM510 confocal microscope or a Leica DMR immunofluorescence microscope. Pixel intensities were obtained using LSM510 imaging system software to create line scan plots for Xpress-FOXCl and p32 immunofluorescence.

Transactivation assays Transactivation assays were performed using Promega Dual-luciferase reporter assay system according to the manufacturer's instructions. Subcultured HTM cells on 24-well tissue culture plates were transfected with pcDNA3.1/nV4- DEST -FOXC1 expression vectors, or pcDNA3.1/nV4-DEST-p32 expression vectors, or FOXC1 expression vectors and increasing amounts of p32 expression vectors along with pGL3-TK-FGF19RE-luciferase reporter, and pRL-CMV (Promega). The ratios of the doses of FOXC1 expression vectors to p32 expression vectors used in transfections are 1:1, 1:2 and 1:3. The total amount of transfected DNA was equalized with empty pcDNA3.1/nV4-DEST. 2 days after transfection, cells were washed with PBS twice. Cell lysates were obtained by vigorous shaking the cells in 100 ul of passive lysis buffer (Promega) per well of cells for 15 minutes at room temperature. Meanwhile lOOul Luciferase Assay Reagent II (LARII) was predispensed into the appropriate number of luminometer tubes. 5 ul of cell lysates were transferred into the luminometer tube containing LAR II, and mixed up by pipetting 3 times. The firefly luciferase activity was monitored for 10 seconds on a TD-20/20 luminometer (Turner Designs). Following this measurement, 100 ul of Stop & Glo reagent was added into the luminometer tube to quench the firefly luciferase luminescence and concomitantly activate renilla luciferase. After a quick mix-up by vortex, the tube was placed back on the luminometer and the renilla luciferase luminescence was

77 measured for 10 seconds. The luciferase assays were performed in triplicate and repeated 3 times. The data were analyzed and the bar graphs were generated using Microsoft Excel.

Electrophoretic mobility shift assay (EMSA) Part I Probe labeling and purification The double stranded DNA containing the FOXC1 binding site (forward: 5'- gatccaaagtaaataaacaacaga-3', reverse: 5'-gatctctgttgtttatttactttg-3') was labeled by incubating 4 ug of the DNA strand, 5xl0"3 uM each of dATP, dGTP, dTTP, 2 ul of lOx REACT2 buffer (Invitrogen), 0.5 ul of Klenow Fragment (Invitrogen) and 5 ul of a32-dCTP in a 20 ul total volume for 30 minutes at 30°C. Then the

78 loaded on to the gel and separated by electrophoresis at 105V for 50 minutes. Finally, the gel was dried on an 853 vacuum gel dryer (Bio-Rad) at 87°C for 30 minutes and then exposed to autoradiography film (Fuji) at -80°C overnight.

Mutation screen of p32 gene in AR patients This research adhered to the tenets of the Declaration of Helsinki. Six primer sets (Table 2-1) were designed using primer3 software to sequence the six exon coding regions of p32 gene and intron/exon boundaries in a panel of 50 AR patients and 50 unaffected controls. Each DNA sample was PCR amplified and visualized by electrophoresis in 1% agarose gel containing ethidium bromide. The targeted PCR amplicons were recovered from the gel by using QIAquick Gel Exaction kit according to the manufacturer's protocol. The concentration of the recovered PCR product was roughly determined by comparing the band intensity of 5 u.1 of the recovered PCR product to that of 5 ul of MassRuler DNA Ladder (low range, Fermentas), which is designed for DNA quantification. Following the DNA quantification, appropriate amount of PCR product was sent for sequencing with the same primer set used for PCR amplification at the DNA sequencing facility in the department.

Realtime qPCR A TaqMan Gene Expression PCR assay (Applied Biosystem) was used to quantitate the copy number of p32 gene in AR patients. PCR primers and probe were designed in the region of exon 2 of p32 gene by ABI's Primer Express software. BLAST analysis was performed to verify the specificity of the target amplicon from the primer set. The target amplicon was also checked for the SNP distributions along its sequence, in order to exclude primer or probe site nucleotide alteration. Each 15ul PCR reaction contained 14um of forward and reverse primers (GCATAAAACCCTCCCTAAGATGTC; ATTTCGCTTCTGTCCCATTCA), 3.8um of a dual labeled PCR (TaqMan, Applied Biosystem) probe (VIC-AGGTTGGGAGCTGGAA-TAMRA) and 75ng of DNA sample as well. A commercial mixture for quantification of human connexin 40 (CX40) locus on chromosome 1 (bearing 5'FAM and 3'MGB

79 fluorochrome labels, Applied Biosystem) was also included as an endogenous normalization control. The PCR reactions were conducted on the Applied Biosystems 9700HT Thermal Cycler. Each patient's DNA sample was amplified in triplicate. Each 384 well plate included triplicate reactions of two unrelated normal samples and one DNA-free control. Reaction progress was recorded and the data were analyzed by the comparative CT method. Samples with AACT value within 1±0.15 were considered to have 2 copies ofp32 gene.

80 Results

Isolation of p32 as a FOXC1 interacting protein by yeast two-hybrid (Y2H) screening To identify proteins that interact with FOXC1, we screened a human trabecular meshwork (HTM) cDNA yeast two hybrid library that was created by using mRNA extracted from primary HTM cell culture. The TM cDNA library inserts were cloned into the plasmid pEXP-AD502 with the open reading frames (ORFs) fused to the GAL4 activation domain (GAL4AD). Full-length FOXC1 was cloned into the vector pDEST-32 fused to the GAL4 DNA-binding domain (GAL4DBD). Yeast cells containing three reporter genes {HISS, URA3 and lacZ) were co-transformed with the GAL4AD-cDNA library and the GAL4DBD- FOXC1 plasmid. A standard yeast two-hybrid screen procedure was carried out (Invitrogen). Approximately 1.9xl0e6 transformants were subjected to the selection. Seventeen independent clones that fulfilled the criteria for interaction of the gene products were obtained (Figure 2-1). cDNAs were extracted from those clones and partially sequenced. All of the seventeen cDNAs were found to match the published sequence of human p32, which was previously co-purified with the pre-mRNA splicing factor from a HeLa cell extract (133). p32 protein is initially translated as a precursor which is proteolytically processed to produce the mature p32 by removing the N-terminal 73 amino acids (134). All of the cDNAs contained the full-length sequence of the mature protein and part of the sequence of the leader peptide. The specificity of the interaction between p32 and FOXC1 in the yeast two-hybrid system was confirmed by retransformation of the positive cDNA clone into yeast cells, together with vectors expressing GAL4 DNA- binding domain alone or with GAL4DBD-FOXC1 or the GAL4 DNA binding domain fused with a different protein, PITX2. Only the yeast cells that contained the positive cDNA plasmid and the GAL4DBD-FOXC1 plasmid displayed an interaction phenotype (Figure 2-2).

81 Confirmation of the interaction between FOXC1 and p32 Ni2+ pull-down assays were performed to confirm whether the interaction could also be observed in other experimental systems. The full-length wild type FOXC1 was cloned into a bacterial expression vector (pET28), allowing expression of FOXC1 as a 6xHis-tagged fusion protein. Whole cell lysates prepared from HeLa cells transfected with V5-tagged full-length p32 or V5- tagged mature p32 were incubated with 6xHisFOXCl bound to Ni -agarose beads or with Ni2+-agarose beads alone. Western blotting using an anti-V5 antibody showed that FOXC1 can interact with the precursor or mature p32 protein. The endogenous p32 in the cell lysate with the molecular weight of the mature form (135, 136) is also able to interact with FOXC1 (Figure 2-3A). We further confirmed this interaction between FOXC1 and p32 by immunoprecipitation. HTM cells were transfected with pcDNA3.1/nV5-DEST FOXC1 or empty pcDNA3.1/nV5-DEST constructs. Both cell lysates were incubated with anti-V5 antibody overnight. Western blotting using an anti-p32 antibody showed that FOXC1 was able to immunoprecipitate endogenous p32 in HTM cells (Figure 2-3B).

Colocalization of FOXC1 and p32 Immunofluorescence was used to study the cellular distribution of FOXC1 and p32 and to determine the compartment in which they colocalize. The vector pcDNA4-FOXCl expressing N-terminal Xpress epitope tagged FOXC1 (Xpress- FOXC1) was transfected into HTM cells. The Xpress-FOXCl was detected by an anti-Xpress antibody and an anti-mouse IgG coupled to Alexa Fluor 488. Endogenous p32 was detected by an anti-p32 antibody and an anti-goat IgG coupled to Alexa Fluor 594. FOXC1 is predominantly in the nucleus (Figure 2- 4A), as expected (41). p32 is localized predominantly in the cytoplasm (Figure 2- 4A) in a pattern consistent with a mitochondria location (136, 137). However, a proportion of the p32 signal can be observed in the nucleus and colocalized with FOXC1 (Figure 2-4A). Line scan data graphically represents the distribution of Xpress-FOXCl and p32 immunoactivity throughout the cell and indicates the

82 extent of colocalization between FOXC1 and p32 (Figure 2-4B). Because both our anti-FOXCl and anti-p32 antibodies were raised in goat, it was not possible to use co-immunofiuorescence to show the colocalization of endogenous FOXC1 and p32 in the nucleus. However, both endogenous FOXC1 and endogenous p32 are localized in the similar nuclear regions of the HTM cells (Figure 2-5).

The FOXC1 forkhead domain and intact p32 are required for the interaction between FOXC1 and p32 Ni2+ pull-down assays were performed to map the domains within FOXC1 and p32 that are required for their interaction. Wild type FOXC1 and a series of FOXC1 deletions (Figure 2-6A) were cloned into the plasmid pcDNA3.1/nV5- DEST fused to an N-terminal V5 epitope. These deletion constructs were expressed in HeLa cells and the cell lysates were incubated with Ni2+-agarose beads alone or beads bound with p32 fused to a 6xHis tag. As shown in Figure 2- 6B, all of the FOXC1 deletion constructs interacted with p32 except for FOXC1AFHD a FOXC1 construct lacking the forkhead domain. These results indicate that FOXC1 interacts with p32 through the FOXC1 forkhead domain. Similar experiments were performed to map the domain within p32 that is required for the interaction with FOXC1. p32 does not have any known functional domains. The crystal structure of the protein molecule exhibits a (3-sheet core, flanked by its N- and C-terminal a-helices that interact in a coiled-coil fashion (138). We designed two p32 deletion constructs by separating the p-sheet core from the N- and C-terminal a-helices. Neither of the deletion constructs of p32 interacted with FOXC1. It appears that the intact p32 structure is necessary for the interaction between p32 and FOXC1.

Mutation screen and detection of copy number variation of p32 gene in AR patients About 60% of AR patients do not have mutations in the two known AR malformation genes, FOXC1 and PITX2 (Mirzayans and Walter unpublished data). Proteins that interact with FOXC1 may be involved in the same genetic pathways as FOXC1. We hypothesized that FOXC 1-interacting proteins may also be

83 involved in AR and that mutations of the protein-coding genes might contribute to AR. Therefore a mutation screen of p32 was conducted in a panel of 50 AR patients, in whom no FOXC1 and P1TX2 mutations were found. No nucleotide alterations were found in the exon or splice site regions of the p32 gene in these patients. However, a 14 base pair deletion, from -44 to -31 upstream of exon 3, was detected in both AR patients and unaffected controls. This alteration is therefore unlikely to be pathogenic for AR. We also detected the copy number of p32 gene in AR patients using TaqMan realtime QPCR and did not find any deletion or duplication of the p32 gene in the patients. Our results therefore indicate that p32 mutations are unlikely to be a direct cause of AR malformations. p32 inhibits FOXC1 -mediated transactivation The role of p32 in regulating FOXC1 transactivation was studied by dual- luciferase assays. I used luciferase reporter containing an FGF19 promoter element (FGF19RE) that we have previously demonstrated to be regulated by FOXC1 (65). HTM cells were co-transfected with this luciferase reporter construct and FOXC1 expression vector along with an increasing amount of p32 expression vector (Figure 2-7). Equivalent levels of FOXC1 were expressed in the co-transfections (Figure 2-7 bottom). This experiment revealed that expression of p32 impairs transactivation by FOXC1 in a dose-dependent manner.

p32 does not affect FOXC1 DNA binding ability Given that p32 impaired FOXC1 transactivity (Figure 2-8), we determined whether p32 alters FOXC1-DNA binding. Electrophoretic mobility shift assays were performed to analyze FOXC1 DNA binding ability in the presence of p32. COS-7 cells were transfected with pcDNA3.1/nV5-DEST empty vector, V5- FOXC1 expression vector, or V5-p32 expression vector separately or co- transfected with FOXC1 expression vector along with increasing amounts of p32 expression vector. Each cell lysate was incubated with radio-labeled oligomers containing the FOXC1 consensus binding sequence (5'-GTAAATAAA-3'). FOXC1 showed consistent levels of DNA binding in the presence or absence of

84 p32 (Figure 2-8). This result demonstrated that the inhibition of FOXC 1 transactivation by p32 was not due to impaired DNA binding.

The FOXC1 carrying the patient mutation F112S displays an impaired interaction with p32 To date, all published missense AR patient mutations identified in FOXC1 are located in the FHD. Previous work from our lab showed that mutations in three residues, P79, Fl 12, and G165 impair FOXC1 transactivation, yet have near normal DNA-binding ability (40-42). Moreover, molecular modeling of the FOXC1 FHD predicts that the side chains of these three residues point away from the DNA and face opposite to the DNA-binding interface, suggesting the possibility that these three residues are not involved in DNA binding but rather protein-protein interactions (Figure 2-9A). The fact that p32 interacts with FOXC1 through the FHD yet does not interfere with DNA-binding led us to test the interaction of p32 with FOXC1 mutants in these three residues. Four recombinant FOXC1 proteins, harboring four patient mutations (P79L, P79T, Fl 12S, G165R) respectively, were analyzed by Ni -NTA pull-down assays for their ability to interact with p32. FOXC1 carrying the P79L, P79T or G165R mutations were able to interact with p32 (Figure 2-9B). However, the FOXC1- Fl 12S mutant was not able to interact with p32 (Figure2-9B), indicating that this FOXC1 mutation impairs the protein-protein interaction capacity of FOXC 1.

85 Discussion In this study, we sought to isolate proteins that interact with the transcription factor FOXC1 to elucidate the molecular mechanisms involved in FOXC1- mediated gene regulation. Human p32 protein was identified as a novel FOXC1 interacting protein by yeast two-hybrid screening of a human trabecular meshwork cDNA library. p32 is expressed in a variety of human tissues including the brain, ear, heart, liver, and eye (http://www.ncbi.nlm.nih.gov/UniGene/ESTProfileViewer.cgi?uglist=Hs.555866). In the eye, p32 is found in the iris, lens, cornea and retina (http://neibank.nei.nih.gov). The specific interaction between p32 and F0XC1 was confirmed by in vitro analysis and was observed in mammalian cells. Human p32 is a multi-compartmental and multi-functional protein, which is able to interact with a variety of cellular proteins in the mitochondrion, nucleus and with exogenous virus proteins under infectious conditions. Some of the interactions between p32 and its interacting partners have been shown to regulate gene expression. For example, p32 interacts with a splicing factor ASF/SF2 and inhibits its function (133). CDC2L5, a protein having a role in pre-mRNA splicing, is able to interact with p32 (139). p32 is also shown to be involved in transcription, as p32 was co-purified in a cellular transcription factor CBF complex from HeLa cell extracts and was found to specifically repress CBF- mediated transcription in an in vitro transcription reaction (140). In our study, the functional analysis using transactivation assays demonstrated that the interaction with p32 negatively regulates FOXC1 transactivation, which is consistent with the previously established role of p32 in regulating gene expression. Others have suggested that p32 might act as a co­ mpressor and/or a molecular barrier for the recruitment of transcription factors necessary for transactivation (133). However, the impaired interaction between p32 and the FOXC1 Fl 12S mutant, which is able to bind DNA but not transactivate genes, is consistent with our hypothesis that p32 is not simply a co­ mpressor and/or a molecular barrier for the recruitment of FOXC1. There is growing evidence showing that several transcription co-regulator complexes are

86 able to serve as either co-repressors or co-activators depending on the gene being regulated, or cell signaling switches (141-144). p32 could be a component in such a transcription co-regulator complex, helping the complex to be recruited to the promoter bound by FOXC1 in order to regulate FOXC1 transcription activity. The complex would either inhibit or promote FOXC1 transcription activation in different cell contexts. In our studies, overexpression of p32 results in inhibition of FOXC1 transactivation. We hypothesize that mutations of FOXC1, such as Fl 12S, that result in impaired protein-protein interaction with proteins such as p32 lose both negative and positive regulation by these interacting proteins, and that this deregulation deficiency underlies the transactivation deficits of these FOXC1 mutations. The FOXC1 Fl 12S mutant, which cannot interact with regulatory proteins such as p32, thus loses both the positive and negative regulation by such molecules and is unable to transactivate the reporter genes in transactivation assays. Alternatively, p32 may serve as a tether protein like SKIP to recruit both co-repressor and co-activator complex to bind to FOXC1 and help the switch of FOXC 1 between transcription repression and activation (145). Another possibility that cannot be completely ruled out is that the Fl 12 residue plays an important role in recruitment of both the positive and negative regulatory complex for FOXC1. Interestingly, clinical study of family members with a FOXC1 Fl 12S mutation found that these patients present with severe ocular manifestations including iris processes, iris hypoplasia, corectopia and posterior embryotoxon. In addition, some Fl 12S patients also have systemic findings including cardiac, facial and dental anomalies, which usually are not found in patients with FOXC1 mutations (146). The severe eye phenotype and the wide spectrum of clinical defects found in patients with FOXC1 Fl 12S mutations could be due to the lack of both positive and negative regulation of FOXC 1 mediated by proteins such as p32, but the small number of patients with Fl 12S mutations precluded unequivocal phenotype-genotype analysis. Mutations in the genes that are part of the same genetic pathway may cause the same genetic disease or similar syndromes. Since p32 interacts with FOXC1,

87 it is possible that p32 may also play a role in eye development and therefore mutations in p32 could also result in ocular disease like AR. However, we did not find any nucleotide changes in p32 except for a deletion within an intron in a panel of 50 AR patients. We observed a similar frequency of the intronic deletion in the AR patients and the unaffected controls. We also did not observe any copy number alterations of the p32 gene in AR patients. While these results indicate that mutations in the p32 gene are not a direct cause of AR, impaired FOXCl/p32 interaction, due to FOXC1 mutation, may have a role in AR pathogenesis. Previous work on our FOXC1 molecular model, which is based upon FOX FHD structures and an analysis of FOXC1 mutants (39-42), has suggested that this FOXC1 model has prediction value (45). Our current work strongly supports this suggestion. Mutation of one of the residues (Fl 12S) previously predicted to be involved in protein-protein interaction based on this model (Fig 2-9A) results in failure of FOXC1 Fl 12S to interact with p32. This result provides further evidence that the FOXC1 molecular model has the capacity to predict effects of mutations on FOXC1 function. Lastly, our data suggest that altered protein- protein interaction may be a disease causing-mechanism for AR malformations. Understanding the factors that regulate the activity of FOXC1 may allow modulation of the effects of FOXC1 mutations in cells.

88 Table 2-1: primer sets for sequencingp32 gene

Exon 1 Forward: 5'-GGC-CTT-AGG-TCG-TCA-GAG-3' Reverse: 5'-GTT-TAG-CCC-GCT GGT-TGC-3' Exon 2 Forward: 5 '-TTT-TTA-CCG-ATA-AGG-AAA-CAG-3' Reverse: 5'-TGA-ACT-CAA-GGC-TCT-TCT-GG-3' Exon 3 Forward: 5'-TAA-ACT-GAG-GCC-AGC-TTT-GG-3' Reverse: 5'-ACA-GCC-TGT-CAG-AAC-TGA-GG-3' Exon 4 Forward: 5'-TGA-GAT-CAA-GCT-CTT-CAT-TTG-G-3' Reverse: 5'-CTG-CTG-GGC-TGG-TCT-AAG-C-3' Exon 5 Forward: 5'-CCT-TGG-GAA-CTT-CTT-GTC-C-3' Reverse: 5'-CCT-TGA-ACT-AGG-ACC-CTT-CC-3' Exon 6 Forward: 5 '-CTG-GGT-GAG-TGC-TTG-ATA-AGG-3' Reverse: 5 '-ATT-TGT-TCA-CTG-GCC-AAA-GC-3'

89 Figure 2-1: Interaction phenotype displayed by colony 2F061 in the yeast two-hybrid library screen using GAL4DBD-FOXC1 fusion construct as a bait vector

Colony 2F061 in the black box grows on the selective medium without histidine (- HIS+3AT) or uracil (-URA), and shows positive X-Gal staining (X-GAL ASSAY), while the growth on medium containing 5FOA was inhibited, indicating a protein-protein interaction in the yeast. (The prey construct in 2F061 was sequenced later and found to encode human p32). A-E are control yeast strains supplied by the ProQuest Two-Hybrid System (Invitrogen) displaying a spectrum of interaction strength. A: No interaction; B: weak interaction; C: moderately strong interaction; D: strong interaction; E: very strong interaction. F and G are patches of experimental negative control yeast cells. F is the yeast cells co- transformed by the FOXCl-bait construct and the empty prey vector. G is the yeast cells co-transformed by both empty prey vector and bait vector. 2F021, 2F031, 2F032 are patches yeast cells containing FOXC1 bait construct and different library prey plasmids, which were tested for the interaction phenotype on the selective medium.

90 -HIS+3AT X-GAL ASSAY

-LIRA

91 Figure 2-2: Retransformation assay and test against other bait constructs.

The 2F061 prey construct was re-co-transformed into yeast with the original FOXC1 bait construct (FF) and the interaction phenotypes were confirmed. The 2F061 prey construct was also tested for interactions with the FOXC1FHD bait construct (FH) and a full length PITX2 construct (PF). The 2F061/FOXC1FHD showed interaction phenotypes while the 2F061/PITX2 did not. The 2F061 prey constructs were co-transformed with an empty bait construct to test for prey self- activation (DBleu). A-E are control yeast strains supplied by the ProQuest Two- Hybrid System (Invitrogen) displaying a spectrum of interaction strength. A: No interaction; B: weak interaction; C: moderately strong interaction; D: strong interaction; E: very strong interaction. F and G are patches of experimental negative control yeast cells. F is the yeast cells co-transformed by the FOXCl-bait construct and the empty prey vector. G is the yeast cells co-transformed by both empty prey vector and bait vector.

92 X-GAL ASSAY

93 Figure 2-3: Confirmation of the interaction between FOXC1 and p32.

(A). HeLa cell lysates were subjected to Ni2+ pull-down assays with the 6xHis tagged FOXC1 bound to the Ni2+-NTA beads (Ni-FOXCl) or the empty beads (Ni). Bound proteins were analyzed by SDS-PAGE followed by Western blot analysis using an anti-p32 antibody. (B). Co-immunoprecipitation of FOXC1 and p32 in HTM cells. V5-FOXC1 (+) or empty (-) plasmid was transfected into HTM cells. The cell lysates were immunoprecipitated with an anti-V5 antibody and immunoblotted with an anti-p32 antibody.

94 B

Ni pull-down Co-immunoprecipitation

o X Input IP: a-V5 ! o

- 32.5 KD

32.5 KD- IB: a-p32

a-p32

95 Figure 2-4: Subcellular localization of FOXC1 and p32.

(A). HeLa cells were transfected with Xpress-FOXCl expression vector. Endogenous p32 was stained with goat polyclonal anti-p32 antibody followed by Alexa Fluor 594-conjugated anti-goat secondary antibody. Xpress-FOXC 1 was visualized with mouse monoclonal anti-Xpress antibody followed by Alexa Fluor 488-conjugated mouse secondary antibody. The red staining corresponds to p32 protein while the green corresponds to Xpress-FOXCl. The merge shows an overlay of red and green staining. (B). Line scan plot shows the intensity of each fluorescence emission along the yellow line across the cell.

96 XpFOXCI p32 Merge

B

300 250 200 1S0 100 SO 0 MM •C tr> T- to oi r—

97 Figure 2-5. Cellular localization of endogenous FOXC1 and p32.

Upper panel: In the nucleus of a HTM cell, endogenous FOXC1 was stained with a goat polyclonal anti-FOXCl antibody followed by Alexa Fluor 594 conjugated anti-goat secondary antibody (left). The same nucleus was stained with DAPI (right). Lower panel: In the nucleus of a HTM cell, endogenous p32 was stained as described above (left). The same nucleus was stained with DAPI (right).

98 Q-F0XC1 DAPI

a-p32 DAPI

99 Figure 2-6: The FOXC1 forkhead domain is required for binding to p32.

(A). Schematic presentation of the FOXC1 constructs used in this study. (B). HeLa cells were transiently transfected with plasmids expressing the series of FOXC1 deletion constructs (show in A) tagged with the V5 epitope. The cell 9+ lysates were then used for Ni pull-down assays with the 6xHis tagged p32 bound to Ni2+-agarose beads (Ni-p32) or the empty beads (Ni). Bound FOXC1 deletions constructs were detected by Western blot analysis using an anti-V5 antibody.

100 Interact with p32

FOXC 1(1-553)

FOXC1FID(52-434)

FOXC1FHD(69-178)

FOXC1AFHD(A69-178) X

B WTFOXC1 FOXC1FID FOXC1FHD FOXC1AFHD

3 T3 T3 W h. K> S: I I s I I 1 * 83KD - i

62KD-

^ ^ 25KD-

a-V5

101 Figure 2-7. p32 impairs FOXCl-mediated transactivation.

Upper panel: HTM cells were transfected with FGF19RE luciferase reporter along with empty pcDNA4 vector; or FOXC1 expression vector, and increasing amounts of p32 expression vector. The mass ratios of the doses of FOXC 1 expression vectors to p32 expression vectors used in transfections are 1:1, 1:2 and 1:3. The total amount of transfected DNA was equalized with empty pcDNA4. Lower panel: Western blot showed that equal amount of FOXC 1 were expressed.

102 O Relative to empty vector W Is) X O -» M W ^ o H 1

I < +

o I I I Figure 2-8: p32 does not affect FOXC1 DNA binding ability.

Upper panel: EMS As were performed with 32P-labeled FOXC binding site probes. COS-7 cell lysates transfected with V5-FOXC1 expression vectors or lysates co- transfected with V5-FOXC1 and an increasing amount of V5-p32 expression vectors were tested in each EMSA reaction so as to compare the DNA binding ability of FOXC 1 in the presence or absence of p32. EMSA binding buffer alone (the first line) and COS-7 cell lysate (the second line) were also used in EMSA reactions as negative controls. Lower panel: Immunoblots showed the amounts of V5-FOXC1 and /or V5-p32 were used in each corresponding EMSA reaction.

104 FOXC1 + - + + +

p32

F0XC1/DNA complex

V5-FOXC1 V5-p32

105 Figure 2-9: The FOXC1 F112S mutant is not able to interact with p32.

(A) Molecular model of the FOXC1 FHD (based on FOXA3) shows the predicted outward orientation of the side chains of the amino acids of P79, Fl 12, and G165 that previously shown to decrease FOXC1 transactivation activity without affecting DNA binding or nuclear localization (13, 15, 17). (B) The FOXC1 Fl 12S mutant displays an impaired interaction with p32 in Ni2+ pull-down assays. The other three FOXC1 mutants, FOXC1 P79T, FOXC1 P79L, and FOXC1 G165R interacted with p32 similar to WT FOXC1. HeLa cells were transiently transfected with plasmids expressing V5 tagged WT FOXC1, FOXC1 P79T, FOXC1 P79L, FOXC 1F112S, or FOXC1 G165R, respectively. The cell lysates were then used for Ni2+pull-down assays with the 6xHis tagged p32 bound to Ni2+-agarose beads (Ni-p32) or the empty beads (Ni). Proteins bound to the beads were detected by Western blot analysis using an anti-V5 antibody.

106 A B

1 Z z

WTFOXC1

FOXC1P79L

FOXC1P79T

FOXC1F112S

FOXC1G165R

a-V5

107 Chapter 3. Isolation and analysis of FOXC2 interacting proteins

108 Introduction The right spatiotemporal regulation of gene transcription is crucial for the cell fate specification and differentiation during proper development of all organisms. Thus it is very common that many developmental disorders are caused by mutations in genes encoding transcription factors. The Forkhead Box transcription factor family comprises such developmentally important transcription factors that carry a characteristic Forkhead DNA-binding domain and are involved in the proper formation of a number of organs (128). Mutations in the FOXC2 transcription factor gene are associated with lymphedema-distichiasis syndrome (LD), an autosomal dominant inherited primary lymphedema and distichiasis (an accessory row of eyelashes) (79, 81). Other commonly occurring ocular findings in LD patients include photophobia, exotropia, ptosis, congenital ectropion and congenital cataracts (83). In addition to lymphatic and ocular phenotypes, some patients present with early onset varicose veins, cardiac defects, cleft palate, spinal extradural cysts, nephritis, skeletal and craniofacial abnormalities (79, 82, 83, 85-91). Most LD cases result from deletion, insertion or nonsense mutations in the coding region of FOXC2, which produce truncated FOXC2 proteins. Missense mutations affecting DNA binding of FOXC2 have also been identified. The mouse homologue Foxc2 transcription factor gene is expressed in neural crest and mesoderm derived mesenchyme during development. The expression pattern of Foxc2 correlates with organs affected in LD patients. Foxc2 is highly expressed in the developing lymphatic vessels as well as in lymphatic valves in adults (99, 102). In Foxc2'/' mice, the initial development of lymphatic vasculature seems normal, whereas, the specification of the lymphatic capillaries and collecting lymphatic vessels at later stages is affected. The valves in the collecting lymphatic vessels fail to develop. Ectopic coverage of the lymphatic capillaries by basal lamina components and smooth muscle cells resulting in lymphatic backflow is observed (102). Foxc2 is also expressed in the mesenchyme in the developing vertebrae, limbs, kidney, heart and surrounding vasculature and head (95, 96). Foxc2 homozygous deficient mice die perinatally

109 due to aortic arch malformations. Moreover, mice have skeletal defects, cleft palate and abnormal kidneys (55, 95, 96, 104). During ocular development, Foxc2 is expressed in the periocular mesenchyme and in periocular mesenchyme derived anterior segment structures including trabecular meshwork. Foxcl expression pattern in the eye is largely overlapped with Foxcl, mutations in which causing Axenfeld-Rieger syndrome in human with severe anterior segment dysgenesis in the eye (49, 95). In heterozygous Foxc2+' mice, eyes are small in size with absent Schlemm's canal, hypoplastic or absent trabecular meshwork, abnormally thin iris stroma and pigment epithelium, demonstrating that Foxcl directs ocular development in a similar fashion to Foxcl (61). Study of FOXC2 also illustrates that other than being a key transcription factor regulating the embryonic development of mesenchyme derived tissues; FOXC2 appears to be a master regulator of energy expenditure, energy storage and glucose metabolism (107). Transgenic overexpression of FOXC2 in mouse brown and white adipose tissues has a significantly pleiotropic effect on gene expression, which contributes to the conversion of white adipose to brown adipose tissues and the increase in lipolysis, mitochondrial content and oxygen consumption observed in the transgenic mice. As a result, blood levels of free fatty acids, triglycerides, glucose and insulin are decreased; total body fat content is also decreased; whereas insulin sensitivity is enhanced (107). In addition to its role in mesenchyme specification during embryonic development, FOXC2 can promote epithelial-mesenchymal transition and subsequent invasion and metastasis of breast cancer cells (120). How FOXC2 as a transcription factor controls gene expression in embryogenesis, metabolism, and tumor metastasis remains relatively unknown. It is very well known that transcription factors do not function alone and the enhancement or repression of their activities is often involved in recruiting and assembling transcription complexes through protein-protein interactions. Thus it is highly likely that FOXC2 interacts with other proteins to regulate the initiation of transcription. Thus far, the transcription factors Smad3/4, Suppressor of Hairless (Su(H)) have been shown to interact with FOXC2 by GST pull-down

110 assays (121, 126). Both Foxc2 and Smad proteins induce the expression ofPai-1 expression in endothelial cells through binding to their corresponding binding sites on Pai-1 promoter. Interaction between Foxc2 and the adjacent Smad3/4 leads to a synergistic activation of Pai-1 promoter, indicating that Foxc2 is a mediator of TGF[3 signaling (121). Foxc2 also directly regulate expression of Hey2 in endothelial cells. Su(H), a component of Notch transcriptional activation complex, physically interacts with Foxc2 and together they activate the expression of Hey2 synergistically, suggesting that Foxc2 specifies arterial cell fate through interacting with the VEGF-Notch signaling pathway (126). Although these studies further our understanding of Foxc2's function in vascular endothelial cells, the identity of FOXC2 interacting partners in other tissues has not been determined. Since FOXC1 and FOXC2 have overlapping expression pattern in the trabecular mesh work and we have successfully isolated interacting protein of FOXC 1 by Y2H screening of our HTM cDNA library (as described in Chapter 2), I hypothesized that FOXC2 interacting proteins could also be identified by the same method.

Ill Methods

Plasmids and reagents N-terminally tagged 6xHisXpress::FOXC2 fusion protein was expressed in pcDNA4 His-Max C. The construct was assembled by Fred Berry and has been previously described (45). FOXC2 cDNA was amplified by PCR from pcDNA4- FOXC2 and subcloned into pDEST32 (Invitrogen) in-frame to the GAL4DBD or into the pcDNA3.1/nV5-DEST vector (Invitrogen) in-frame to the V5 epitope by gateway technology. FOXC2 cDNA was again amplified by PCR from pcDNA- FOXC2 and then subcloned in-frame to the C-terminal enhanced green fluorescent protein (EGFP) into the vector pEGFPNl (Clontech). The PIAS3 cDNA was amplified by PCR from the positive cDNA clone screened out from the yeast two-hybrid screen. Then the PCR amplicon containing the full length of the PIAS3 cDNA was subcloned into the vector pcDNA4. All constructs were sequenced to confirm that no mutations were introduced into the cDNAs and the cDNAs were in frame with the epitopes of each vector. The two pGL3-TK- luciferase reporter constructs under control of a 354 bp fragment from FGF19 promoter were constructed by Yahya Tamimi and have been previously described (65). One of the reporter constructs contains a FOXC binding site (FGF19RE) and the other one have the FOXC binding site replaced by an irrelevant sequence (FGF19RE-)(65).

Yeast two-hybrid screen The same human trabecular meshwork (HTM) cDNAs library, which is described in Chapter 2, was used. Ten ug of the full length FOXC2 bait construct was co-transformed with lOug of isolated library DNA according to the manufacturer's protocol. The detailed description of Y2H screening can be found in Chapter 2. A total of 2 xl0e5 library clones were screened with FOXC2 for growth on selective media and assayed for P-galactosidase activity. In order to rule out false positive results, the candidate prey plasmids were retransformed into yeast cells with the empty bait vector or FOXC2 bait construct followed by the

112 same selective procedures. Clones that reproduce the original phenotypes likely contained FOXC2 interacting proteins.

Mammalian cell culture and transfection HeLa cells and COS-7 cells were maintained in Dulbeco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum in a 37°C, humidified incubator under an atmosphere containing a constant 5% CO2. The day prior to transfection, plates were seeded (106 cells in 10ml media per 100mm Petri dish, 2xl05 cells in 2ml media per 25mm well or 4xl04 cells in 1ml media per 15mm well) so as to reach approximately 30-50% confluence the following day. Transfections were performed using FuGene 6 (Roche) or TranlT-LTl transfection reagent (Minis bio) according to the manufacturer's protocol. Transfected cells were subjected to immunofluorescence on the next day of transfection. Dual-luciferase assays (Promega) and EMSAs were performed 48 hours after transfection.

Immunoblot analysis The detailed description can be found in Chapter 2. Primary antibodies and the amount of each antibody used are anti-Xpress (1:5000), anti-V5 (1:5000), anti-FOXC2 (1:1000), anti-PIAS3 (1:1000) and anti-TFIID (1:1000). HRP- conjugated secondary antibodies used are goat anti-mouse (1:5000), goat anti- rabbit (1:5000) and donkey anti-goat (1:5000).

Immunoprecipitation (IP) HeLa cells were co-transfected with pcDNA3.1/nV5-DEST-FOXC2 and pcDNA4-PIAS3 or empty pcDNA3.1/nV5-DEST vector and pcDNA4-PIAS3 as experimental control. Both cell lysates were immunoprecipitated using 3 ug of anti-V5 antibody. In the end, proteins immunoprecipitated by anti-V5 antibody were analyzed by immunoblot with anti-PIAS3 antibody (1:5000). The detailed description can be found in Chapter 2.

113 Immunofluorescence HeLa cells plated onto coverslips were co-transfected with pEGFPNl- FOXC2 plasmid and pcDNA4 plasmid or co-transfected with pcDNA4-PIAS3 plasmid and pEGFPNl plasmid; or co-transfected with pEGFPNl-FOXC2 plasmid and pcDNA4-PIAS3. One day after transfection, HeLa cells were subjected to immunofluorescence. Plates were kept at room temperature through the entire protocol and covered by foil paper at all possible times in order to prevent photobleaching of fluorochrome-conjugated secondary antibodies and EGFP. Cells were washed in PBS for 2 times and fixed for 20min in PBS containing 2% w/v paraformaldehyde. Two-5 minute washes with PBSX (PBS+0.05% Triton X-100) were conducted to permeate cell and organellar membranes, followed by incubating the coverslips in PBSX containing 5% (w/v) BSA for 15min to reduce nonspecific protein binding of antibodies. Coverslips were then incubated for 1 hour with the primary antibodies to Xpress epitope (diluted 1:500 in lOOul PBSX containing 5% BSA), washed twice with PBSX, followed by the incubation with Cy3 conjugated anti-mouse secondary antibodies (diluted 1:500 in lOOul PBSX containing 5% BSA) for another 1 hour. After two- 5 minute washes with PBSX, DAPI staining was performed by incubating the coverslips with DAPI (5mg/ml solution diluted 1:500 in lOOul PBS) for 5 minutes. After twice-5 minutes final rinses with PBS, coverslips were mounted onto glass slides with mounting medium (90% Glycerol, 10% PBS containing lug/ml p- phenylenediamine) and sealed with nail polish. Slides were stored in dark at - 20°C or subjected to imaging analysis immediately. The images were collected on a Leica DMR immunofluorescence microscope.

Transactivation assays Transactivation assays were performed using Promega Dual-luciferase reporter assay system according to the manufacturer's instructions. For testing the effect of PIAS3 on FOXC2 transactivity, subculrured HeLa cells on 24-well tissue culture plates were transfected with pcDNA3.1/nV4-DEST -FOXC2 expression vectors, or pcDNA4-PIAS3 expression vectors, or both expression vectors along

114 with pGL3-TK-FGF19RE luciferase reporter, and pRL-CMV (Promega). To study the mechanism of the transactivity of PIAS3, HeLa cells on 24-well tissue culture plates were transfected with pcDNA4-PIAS3 expression vectors along with pRL-CMV (Promega) and pGL3-TK-FGF19RE luciferase reporter or pGL3- TK-FGF19RE- luciferase reporter. The total amount of transfected DNA was equalized with empty vectors. The transactivation assays were performed two days after transfection. The detailed procedure can be found in Chapter 2. siRNA transfection Human FOXC2-specific siRNA was purchased from Dharmacon (D-008987- 02). The non-targeting control siRNA was purchased from Ambion (AM-4635). Twenty-four prior to transfections, HeLa cells were plated at a density of 106 cells per 100 mm tissue culture plate or 2 X105 cells per well of a six-well plate in serum supplemented DMEM medium. For transfection of cells in a 6-well plate, 50 nM of FOXC2-specific or non-targeting control siRNA was diluted in 500 p.1 of Opti-MEM I Reduced Serum Medium (Invitrogen). Meanwhile, 5 u.1 of Lipofectamine 2000 (Invitrogen) was diluted in another 500 ^1 of Opti-MEM I Reduced Serum and then incubated for 5 minutes at room temperature. For transfection of cells in 100 mm plate, 50 nM of FOXC2-specific or non-targeting control siRNA was diluted in 2.5 ml of Opti-MEM I Reduced Serum Medium. 12.5 \xl of Lipofectamine 2000 (Invitrogen) was diluted in 2.5 ml of Opti-MEM I Reduced Serum. After the 5 minute incubation, the diluted siRNA was combined with the diluted Lipofectamine 2000 and incubated for 20 minutes at room temperature to allow the siRNA:Lipofectamine 2000 complexes to form. After the 20 minutes incubation, the serum supplemented DMEM medium was removed from the cells and then the siRNA:Lipofectamine 2000 mixture was added to the cells. The siRNA:Lipofectamine 2000 mixture was replaced by regular growth medium after incubation of the cells for about 5 hours. Cells were subjected to plasmid transfection after 5 hours or immunoblot analysis after 48 hours.

115 Electrophoretic mobility shift assay (EMSA) Part I Probe labeling and purification The double stranded DNA containing the FOXC binding site (forward: 5'- gatccaaagtaaataaacaacaga-3'; reverse: 5'-gatctctgrtgtttatttactttg-3') was used. Probe labeling and purifications were done as in Chapter 2. Part II EMSA Protein extracts from COS-7 cells containing recombinant FOXC2 or FOXC2 and PIAS3 were equalized for the amount of recombinant FOXC2 using untransfected COS-7 cell lysate. The detailed description of the EMSA procedure can be found in Chapter 2.

116 Results

Isolation of PIAS3 as a FOXC2 interacting protein by yeast two-hybrid screening To identify proteins that interact with FOXC2,1 screened a HTM cDNA yeast two-hybrid library that was created by using mRNA extracted from a HTM primary cell culture. The HTM cDNA inserts were cloned into the plasmid pEXP- AD502 with the open reading frames fused to the GAL4 activation domain (GAL4AD). Full-length FOXC2 was subcloned into the vector pDEST-32 fused to the GAL4 DNA binding domain (GAL4DBD). Yeast cells containing three reporter genes (HIS3, URA3 and lacZ) were co-transformed with the GAL4AD- cDNA library and the GAL4DBD-FOXC2 plasmid. A standard yeast two-hybrid screen procedure was carried out (Invitrogen). Approximately 2 xl0e5 transformants were subjected to the selection. Two independent clones that fulfilled the criteria for interaction of the gene products were obtained (Figure 3- 1). cDNAs were extracted from those clones and partially sequenced. Comparison of the DNA sequences to the publicly available genome database revealed that both clones encode the full-length human Protein Inhibitors of Activated STAT 3 (PIAS3). The specificity of the interaction between PIAS3 and FOXC2 in the yeast two-hybrid system was confirmed by retransformation of the positive cDNA clone into yeast cells, together with vectors expressing GAL4 DNA-binding domain alone or with GAL4DBD-FOXC2. Only the yeast cells that contained the positive cDNA plasmid and the GAL4DBD-FOXC2 plasmid displayed an interaction phenotype (Figure 3-2).

Confirmation of the interaction between FOXC2 and PIAS3 by immunoprecipitation The interaction between FOXC2 and PIAS3 was further confirmed by immunoprecipitation. HeLa cells were co-transfected with pcDNA3.1/nV5-DEST FOXC2 (V5-FOXC2) and pcDNA4-PIAS3 (Xpress-PIAS3) or empty pcDNA3.1/nV5-DEST vector and Xpress-PIAS3 construct. Both cell lysates were incubated with anti-V5 antibody overnight. Western blotting using an anti-PIAS3

117 antibody showed that Xpress-PIAS3 was recovered in anti-V5 IP in a V5- FOXC2-dependant manner (Figure 3-3).

Colocalization of FOXC2 and PIAS3 Immunofluorescent microscopy was conducted to study the cellular distribution of FOXC2 and PIAS3 and to determine the cellular compartment in which they colocalize. The vector pEGFPNl-FOXC2 expresses C-terminal EGFP tagged FOXC2 (FOXC2EGFP). The vector pcDNA4-PIAS3 encodes N-terminal Xpress tagged PIAS3. HeLa cells were transfected with pEGFPNl-FOXC2 or pcDNA4-PIAS3, or co-transfected with both constructs. This analysis demonstrated that FOXC2EGFP and Xpress-PIAS3 are predominantly in the cell nucleus as expected and at least some of these two proteins are colocalized within the nucleus (Figure 3-4). Previous studies have reported that PIAS proteins can alter the subnuclear distribution of their interacting partners (147, 148). Comparison of the FOXC2 nuclear localization in the cells only transfected with FOXC2 and that in the cells co-transfected with both FOXC2 and PIAS3, no significant difference in the FOXC2 subnuclear distribution was observed.

The effect of PIAS3 on FOXC2-mediated transactivation Next, the dual-luciferase assays were performed to determine the effect of PIAS3 on FOXC2's transcription activity. I used a luciferase reporter under the control of a 354 bp FGF19 promoter element (FGF19RE) containing a FOXC binding site. HeLa cells were transfected respectively with empty vector, V5- FOXC2 expression construct, or Xpress-PIAS3 expression construct or co- transfected both V5-FOXC2 and Xpress-PIAS3 along with the FGF19RE luciferase reporter. Surprisingly, I found that PIAS3 is not only able to positively regulate FOXC2 transactivation of the reporter gene but is also to transactivate the luciferase gene under the control of FGF19 promoter without co-transfected FOXC2 (Figure 3-5). In order to understand the mechanism underlying the transactivation on FGF19RE luciferase reporter by PIAS3 without co-expression of FOXC2, another luciferase reporter (FGF19RE-) was used. FGF19RE- is derived from the FGF19RE luciferase reporter but has the FOXC binding site

118 replaced by an irrelevant sequence (65). In brief, Hela cells were co-transfected with Xpress-PIAS3 and FGF19RE luciferase reporter or co-transfected with Xpress-PIAS3 and FGF19RE- luciferase reporter and then the dual-luciferase assays were performed. Comparison of the transactivity of PIAS3 on these two reporter constructs demonstrated that ablating the FOXC binding site on the FGF19 promoter abolished the transactivation by PIAS3 (Figure 3-6). This result indicated that PIAS3's transactivation is dependent on the FOXC binding site. Therefore, PIAS3 may directly bind to FOXC site and subsequently induce the expression of the reporter gene or PIAS3 may be able to positively regulate certain unknown endogenous transcription factor(s) that can bind to the FOXC site. FOXC2 is an excellent candidate for being such an endogenous transcription factor.

PIAS3 does not directly bind to the FOXC DNA binding sequence Since PIAS3 itself can activate the expression of FGF19RE luciferase reporter through the FOXC binding site, electrophoretic mobility shift assays (EMSAs) were performed in order to determine if PIAS3 is able to directly bind to the FOXC binding sequence. COS-7 cells were transfected with pcDNA3.1/nV5-DEST empty vector, V5-FOXC2 expression vector, or Xpress- PIAS3 expression vector separately or co-transfected with FOXC2 and PIAS3 expression vectors. Each cell lysate was incubated with radio-labeled oligomers containing the FOXC consensus binding sequence (5'-GTAAATAAA-3'). These analyses demonstrated that PIAS3 is not able to bind to the DNA probe containing FOXC binding site and further that the presence of PIAS3 does not significantly affect FOXC2 DNA binding ability (Figure 3-7).

PIAS3 is able to up-regulate the transcriptional activity of endogenous FOXC2 In order to determine whether PIAS3 induced expression of luciferase reporter gene controlled by promoter containing FOXC binding sites is due to up- regulating of the endogenous FOXC2, siRNA knock down of FOXC2 followed by luciferase assays was performed. HeLa cells were first transfected with control

119 siRNA and F0XC2 siRNA (Dharmcon), respectively. Six hours after siRNA transfection, cells with control siRNA transfection as well as cells with FOXC2 siRNA transfection were co-transfected with empty pcDNA4 plasmid and FGF19RE luciferase reporter constructs or co-transfected with Xpress-PIAS3 and the FGF19RE luciferase reporter constructs. Parallel transfections were performed to determine the efficiency of siRNA knocking down of FOXC2 by immunoblot analysis. The nuclear protein TFIID was used as loading control for the immunoblot. The intensities of the FOXC2 bands and the TFIID bands shown in the immunoblot were quantified using software ImageJ. The quantity of FOXC2 in each sample was normalized with the quantity of TFIID. Comparison of the amount of FOXC2 among samples was conducted using the normalized values. The results showed that with 37% reduced FOXC2 protein levels, the induction of luciferase gene expression by PIAS3 decreased 20%. The difference between PIAS3-induced expression of the reporter gene with siRNA control transfection and that with siFOXC2 transfection is significant (Figure 3-8 A). In addition, overexpression of PIAS3 did not lead to increased protein level of endogenous FOXC2 (Figure 3-8 B). The results from the experiments described above support the idea that PIAS3 up-regulates the transcription activity of the endogenous FOXC2 rather than increasing FOXC2 protein levels. The relative small degree of reduction in the transactivity of PIAS3 after knocking down of FOXC2 is possibly due to the low efficiency of siFOXC2. Another possibility that cannot be fully excluded, however, is that FOXC2 is not the only endogenous transcription factor influenced by PIAS3 that is able to transactivate the luciferase reporter gene. Further optimization of the FOXC2 siRNA transfection to obtain maximum knock down and repetition of this experiment are necessary.

120 Discussion In this study, I isolated proteins that interact with the transcription factor FOXC2 to elucidate the molecular mechanisms involved in regulating FOXC2- mediated gene expression. Previously, only Smad3/4 and Su(H) were known to be able to interact with the adjacent Foxc2 on the promoters of target genes in endothelial cells (121, 126). The major finding of my study is the identification of the Protein Inhibitors of Activated STAT 3 (PIAS3) from HTM tissue as a novel FOXC2 interacting protein. When looking for a protein interaction, there are several criteria that need to be met. Here I have fulfilled these major criteria and confirmed that PIAS3 is a FOXC2 interacting protein. First of all, the FOXC2/PIAS3 interaction was identified from a yeast two-hybrid cDNA library screen, which is an in vivo genetic system for isolating protein-protein interactions in high throughput. As well, this interaction between FOXC2 and PIAS3 was reassessed within the same yeast two-hybrid system by the retransformation assays. I have isolated two cDNA clones encoding the full length PIAS3 prey protein as FOXC2 interactors from the library screen. The identification of multiple PIAS3 clones is evidence that FOXC2/PIAS3 interaction is real within the Y2H system, even before the further confirmation. I then successfully confirmed the interaction with a second biochemical system, co-immunoprecipitation. Another line of evidence for a physical interaction between FOXC2/PIAS3 came from the immunofluorescence study observing the two recombinant proteins in the cells. Both of these proteins reside in the nuclei of the cells and colocalize with each other in the nuclei. Their overlapping cellular localizations meet the fundamental spatial prerequisite for two proteins to interact and demonstrate that there are instances for FOXC2 and PIAS3 to meet and interact with each other within the nuclei. Searching through several databases, I have found considerable overlap of expression of these two proteins. The expression profiles of FOXC2 and PIAS3 suggested by analyses of ESTs count (UniGene) indicate overlapping expression in the eye, lung, pancreas and uterus in humans. In addition, FOXC2 is found in a retina foveal and macular cDNA library (NCBI AceView), while PIAS3 is found

121 in an optic nerve cDNA library and RPE cDNA library (NCBI Ace View). Obviously, PIAS3 is expressed in HTM as well, since it was isolated from our HTM Y2H library. The co-expression of FOXC2 and PIAS3 in the same tissues provides additional evidence that they can interact and is consistent with the idea that this interaction may have biological consequences. The FOXC2 interacting protein PIAS3 belongs to the Protein inhibitor of activated STATs (signal transducer and activator of transcription) protein family. The mammalian PIAS proteins consist of four members, PIAS1, PIAS3, PIASx and PIASy. PIAS proteins are evolutionarily conserved and the orthologues of mammalian PIAS proteins are also found in yeast and Drosophila. The acronym PIAS comes from the initial finding that members of the family are able to interact with and inhibit STAT transcription factors in Janus kinase (JAK)-STAT signaling pathway. Among them, PIAS1 and PIASy are inhibitors of STAT 1 signaling, whereas PIAS3 represses the activity of STAT3 and STAT5 (149-152). The role for PIAS proteins in JAK-STAT signaling was further confirmed by genetic experiments in Drosophila demonstrating that the Drosophila ortholog dPIAS and stat92E functionally interact to regulate blood cell and eye development (153). It is clear now that the interactions and functions of PIAS proteins go far beyond STATs. PIAS proteins have been identified in yeast two- hybrid screens as interacting partners for a broad spectrum of proteins with structural and functional diversities, most of which are transcription factors (154- 156), which is consistent with my findings. Several functional and structural conserved domains and motifs have been identified among the PIAS family members. A SAP domain (scaffold attachment factor-A/B, apoptotic chromatin-condensation inducer in the nucleus (ACINUS), PIAS) is located at the N-terminus of PIAS proteins. The SAP domain has been found in many chromatin-associated proteins and is able to bind A/T rich DNA sequences from nuclear scaffold-attachment regions, where together with nuclear- scaffold proteins SAP domain anchors chromatin to nuclear matrix and provides a microenvironment for transcription regulation (147, 156, 157). There is a LXXLL motif within the SAP domain. LXXLL motif of PIASy is required for its

122 transrepression effect on androgen receptor and Statl (149, 158). The PINIT motif is localized to the C-terminus of SAP domain within a highly conserved region of PIAS protein. Disruption of the PINIT motif of PIAS3 disturbed its exclusive nuclear localization, suggesting that the PINIT motif of PIAS proteins plays a role in their normal subcellular localization (159). Another important domain of PIAS proteins is known as Miz-zinc finger or Siz/PIAS-RTNG (SP-RTNG) domain (160, 161), which is related to the classical RING finger domain (162, 163). The typical RING finger domain is found commonly in ubiquitin E3 ligases, whereas the SP­ RING domain is required for the SUMO-E3 ligase activity of PIAS proteins (163, 164). The C-terminal region of PIAS proteins, which is the most variable region, contains a highly acidic glutamic/aspartic residue-rich domain and a serine/threonine rich region (S/T). There is a putative SUMO-1-interaction motif (SIM) within acidic glutamic/aspartic residue-rich domain (165). In my study, after confirming the physical interaction between FOXC2 and PIAS3,1 wanted to determine the biological consequence of the interaction between FOXC2 and PIAS3. To determine this, I performed luciferase assays to see if PIAS3 is able to regulate FOXC2 transcription activity and if it does, whether the regulation is positive or negative. Transfection of PIAS3 without FOXC2 was able to induce the expression of the luciferase reporter gene under the control of FGF19 promoter containing a FOXC-binding site (FGF19RE). Two other luciferase reporter constructs with different number of FOXC-binding sites and in different vector backbones were tested later and similar results were obtained (data not shown). Previous publications have showed that PIAS3 functions as a transcriptional co-regulator and either positively or negatively regulates the activity of various transcription factors (156, 166). This indicates that PIAS3 is not a general transcriptional co-activator, which increases transactivation independent of the promoter or transcription factor. Mutagenesis of the FOXC-binding site within the FGF19RE luciferase reporter construct abolished the transactivation observed when cells were transfected with PIAS3, revealing that PIAS3's transactivation is mediated by the FOXC-binding site. Since the SAP domains from PIAS1 and PIASy are capable of binding to the A/T

123 rich DNA sequence from nuclear scaffold-attachment regions and PIAS3 shares a high degree sequence identity in the SAP domain with other family members (147, 157), EMS As using a DNA oligomer containing a FOXC consensus binding sequence were performed to detect if PIAS3 can directly bind to the FOXC- binding sequence. My EMSAs results demonstrated that PIAS3 doesn't bind the FOXC-binding site and suggested that although PIAS3's transactivation requires the FOXC-binding sequence, it does not result from direct binding of PIAS3 to the site. Taken together, I hypothesized that PIAS3 has the capability of positively regulating certain endogenous transcription factor(s), which can bind to the FOXC site and subsequently induce the expression of the luciferase gene. These presumed endogenous transcription factors appear to include FOXC2. A number of previous studies support my hypothesis. One study has shown that PIAS1 interacts with p53. Expression of PIAS1 in the p53-negative cell lines (MEF or HI299) does not induced p53-dependent luciferase gene expression, while co- transfection of PIAS1 together with p53 increases the expression of p53- dependent transcription. Furthermore, PIAS1 expression alone in a p53-positive cell line (Tera-1) activates transcription of the same reporter construct, indicating that this effect is at least partially mediated by the endogenous p53 (167). Several studies of the effect of PIAS3 on cytokine signaling have shown that in the presence of the cytokine, PIAS3 up- or down-regulates its target transcription factors at the endogenous level (168, 169). Because all the human cell lines available in our laboratory express FOXC2,1 decided to test my hypothesis by knocking down endogenous FOXC2 and subsequently testing the transactivation by PIAS3. My results showed that with about 37% decrease in FOXC2 protein levels, there is a 20% statistically significant reduction in PIAS3-induced transcription. In addition, expression of PIAS3 does not increase the endogenous FOXC2 protein levels. Although these experiments still need to be repeated and the condition for knocking down FOXC2 still need to be optimized, my results strongly support the hypothesis that PIAS3 interacts with endogenous FOXC2 and up-regulate FOXC2-mediated transcription. Another experiment that can be conducted so as to avoid the influence from the endogenous FOXC2 is to make a

124 GAL4DBD fused F0XC2 and test the transcriptional activity of this fusion protein with or without PIAS3 on a GAL4 binding site-derived luciferase reporter construct. In this report, I have shown that PIAS3 can interact with FOXC2 and activate FOXC2-dependent transcription. The mechanism underlying the ability of PIAS3 to modulate FOXC2-dependent transcription is still unknown. PIAS3 interacts with more than 20 transcription factors and uses multiple modes to regulate their activity (155, 156, 166). Previous reports have showed that PIASy and PIASx can affect transcriptional activity by regulating the subnuclear localization of their interacting transcription factors (147, 165, 170). Therefore, I first observed the cellular localization of FOXC2 and PIAS3. Immunofluorescence was carried out with cells transfected with FOXC2 or PIAS3 respectively and cells co-transfected with both proteins. I compared the subnuclear localization of FOXC2 in the absence of PIAS3 with that in the presence of PIAS3. No apparent changing of FOXC2 subnuclear localization was found when PIAS3 was co-transfected. Also FOXC2 did not significantly influence the nuclear distribution of PIAS3 significantly. These results indicated that interaction between FOXC2 and PIAS3 does not alter their cellular distribution. PIAS3 has been showed to increases the transcription activity of SMAD3 by recruiting the CREB binding protein (CBP)/p300 co-activator complex. In this case, the SP-RING domain of PIAS3 links the SMAD3 and CBP/p300 to form a stable protein complex (171). Based on this information, the stimulatory effect of PIAS3 on FOXC2 is likely due to the recruitment of a co- activator complex, such as the CBP/p300, by PIAS3. Interestingly, Foxc2 has been shown to interact with adjacent Smad3 and Smad4 on Pai-1 gene promoter and cooperatively induce Pai-1 expression with Smad3 and Smad4 under TGFp signaling (121). Therefore, it is possible that FOXC2, SMAD3, PIAS3 and CBP/p300 are components of the same co-activator complex. Additionally, PIAS proteins can regulate transcription by affecting the DNA-binding ability of their interacting transcription factors. For example, PIAS3 was initially isolated through its inhibition of STAT3 DNA-binding (151). PIAS3 can also block the

125 DNA-binding of microphthalmia transcription factor (MITF) (172); whereas PIASxa interacts with Msx2 and subsequently enhances the DNA-binding affinity of Msx2 to a functionally important element on the rat osteocalcin promoter (173). I also used EMSA to analyze whether PIAS3 can affect FOXC2 DNA-binding ability. My results indicated that PIAS3 does not significantly affect FOXC2 DNA-binding activity in vitro. However, since the chromatin- association SAP domain of PIAS proteins is able to wedge into the minor groove of DNA double helix (156), PIAS3, after binding to FOXC2, might help with creating a more opened chromatin structure and thus facilitate FOXC2 transactivation. PIAS proteins have SUMO-E3 ligase activity and the SP-RING domain is required for their ligase activity (155, 161). SUMO is a member of the growing family of ubiquitin-like modifier that can be covalently attached to target protein via a multi-step pathway. The E3 ligase triggers the formation of isopeptide bond between SUMO and the target protein by recruiting SUMO- conjugating enzymes to the substrate proteins (155). In most cases the lysine residue of a target protein within a consensus sequence containing a characteristic \j/KxE motif (\|/: a large hydrophobic residue) is sumoylated (155). PIAS3 can regulate transcription through inducing SUMO modification of its interacting transcription factors (174). Some natural questions that arise from this information are whether FOXC2 is subjected SUMO modification and whether the interaction with PIAS3 can induce FOXC2 sumoylation. Using an online sumoylation prediction program (http://www.abgent.com/tools/sumoplot), I found that F0XC2 has four highly possible sumoylation sites, indicating that FOXC2 may be SUMO modified. It is very interesting that one of the predicted SUMO sites on F0XC2 is within the F0XC2 predicted transcription inhibitory domain. Recently studies of the role of PIAS3 in transcriptional regulation suggest some intriguing possibilities. MBD1 (methyl-CpG-binding domain protein 1), which forms a complex with SETDB1 histone methylase to silence transcription, can be sumoylated by PIAS3 leading to an impaired incorporation of the SETDB1 into MBD1 and subsequently relieving the repressive effect of MBD1 on transcription (175). Therefore, as for MBD1, SUMO modification within FOXC2's

126 transcription inhibitory domain by PIAS3 might impair the function of the inhibitory domain and result in a hyperactive FOXC2 molecule. Further experimental investigation of FOXC2 sumoylation and of the involvement of this possible modification by PIAS3 in regulating FOXC2 activity would be very interesting. In summary, PIAS3 was identified as a protein partner of FOXC2 throughY2H screening of a HTM cDNA library. The physical interaction between FOXC2 and PIAS3 was confirmed by co-IP and immunofluorescence analyses. PIAS3 can up-regulate the FOXC2-mediated transactivation. This up-regulation of FOXC2 transactivity might be the result of the recruitment of transcription co- activator complexes by PIAS3, the opened chromatin structure introduced by PIAS3 or the altered the posttranslational modification of FOXC2 trigger by PIAS3. However additional studies need to be performed in order to further elucidate the mechanisms underlying the positive regulation of FOXC2 transactivation by PIAS3.

127 Figure 3-1: Interaction phenotype displayed by colony 2F2021 in the yeast two-hybrid library screen using GAL4DBD-FOXC2 fusion construct as a bait vector

Colony 2F2021 in the black box grew on the selective medium without histidine (-HIS+3AT) and showed positive X-Gal staining (X-GAL ASSAY), while the growth on medium containing 5FOA was inhibited, indicating a protein-protein interaction in the yeast. (The prey construct in 2F2021 was sequenced later and found to encode human PIAS3). A-E are control yeast strains supplied by the ProQuest Two-Hybrid System (Invitrogen) displaying a spectrum of interaction strength. A: No interaction; B: weak interaction; C: moderately strong interaction; D: strong interaction; E: very strong interaction. F and G are patches of experimental negative control yeast cells. F is the yeast cells co-transformed with the FOXC2-bait construct and the empty prey vector. G is the yeast cells co- transformed with both empty prey vector and bait vector.

128 -HIS+3AT X-GAL ASSAY

V-. .. .to' 0 1

C «2F2021 F 6 i

-URA 5-FOA ^ V 1 o B I—I 2F2021 i 2F202ii ii i 1 i

129 Figure 3-2: Retransformation assays.

The 2F2021 prey construct was re-cotransformed into yeast with the original FOXC2 bait construct (solid-lined box) and the interaction phenotypes were confirmed. The 2F2021 prey construct was also co-transformed with an empty bait construct to test for prey self-activation (dash-lined box). A-E are control yeast strains supplied by the ProQuest Two-Hybrid System (Invitrogen) displaying a spectrum of interaction strength. A: No interaction; B: weak interaction; C: moderately strong interaction; D: strong interaction; E: very strong interaction. F is the yeast cells co-transformed with the FOXC2-bait construct and the empty prey vector.

130 HIS+3AT X-GAL ASSAY

Jig Tg» *-<^" • g S A" A 'f g B 8 !•*•••••• G!4 • C

>URA 5-FOA • "•0I*

*#.••••** ifJ en ^^^^'

131 Figure 3-3: Confirmation of the interaction between FOXC2 and PIAS3.

Co-immunoprecipitation of recombinant FOXC2 and PIAS3 was performed. HeLa cell lysates co-transfect with V5-FOXC2 and Xpress-PIAS3 (+) or cell lysates co-transfected with empty V5 plasmid and Xpress-PIAS3 (-) were immunoprecipitated with an anti-V5 antibody. The captured protein complexes were subjected to immunoblot analysis and the presence of Xpress-PIAS3 was detected with an anti-PIAS3 antibody. Input fraction represented 5% of the protein extracts used in the immunoprecipitation. *, IgG bands.

132 Input IP: a-V5 _^_ + - + 83KD— *«,*w» «taM> —Xp-PIAS3 ^* * 62KD-—

IB: a-PIAS3

133 Figure 3-4: Subcellular localization of FOXC2 and PIAS3.

HeLa cells were transfected with Xpress-PIAS3 expression vectors or FOXC2EGFP constructs respectively or co-transfected with both constructs. Xpress-PIAS3 were stained with mouse monoclonal anti-Xpress antibody followed by Cy3 donkey anti-mouse IgG secondary antibody. The red staining corresponds to Xpress-PIAS3 protein while the green corresponds to FOXC2EGFP. The merge shows an overlay of red and green staining.

134 F0XC2-EGFP r ^\ F0XC2-EGFP DAPI

Xpress-PIAS3 r ^ g-Xpress DAPI

FOXC2-EGFP+Xpress-PIAS3 r "^ FOXC2-EGFP a-Xpress Merge

135 Figure 3-5: The effect of PIAS3 on FOXC2 transactivation.

Upper panel: HeLa cells were transfected with FGF19RE luciferase reporter along with empty vector; V5-FOXC2 expression vector, Xpress-PIAS3 expression vector or V5-FOXC2 and Xpress-PIAS3 together, respectively. To control for transfection efficiencies, cells were co-transfected with pRL-CMV vectors, which expresses renilla luciferase. The total amount of transfected DNA was equalized with empty vectors. Two days after the transfection, dual-luciferase assays were performed. The activity of the firefly luciferase was normalized by that of the Renilla luciferase, and the value of empty vector was scaled to 1. Lower panel: Immunoblots showing the expression levels of V5-FOXC2 and Xpress-PIAS3 were presented.

136 FGF19RE

18 16 S 14 - u X . >12j I |- 10 - ffi 2 8 - 9

2 0 empty vector FOXC2 PIAS3 FOXC2&PIAS3

t. '. ,»-!• V5-FOXC2

Xp-PIAS3

137 Figure 3-6: FOXC binding site is required for the transactivation of PIAS3.

HeLa cells were transfected with increasing amounts of Xpress-PIAS3 expression vector along with FGF19RE luciferase reporter construct or FGF19RE- luciferase reporter construct. To control for transfection efficiencies, cells were co- transfected with pRL-CMV vector, which expresses renilla luciferase. The total amount of transfected DNA was equalized with empty vector. Two days after transfection, dual-luciferase assays were performed. The activity of firefly luciferase was normalized by that of the renilla luciferase, and the value on FGF19RE luciferase reporter with no PIAS3 was defined as 100%.

138 Luc if erase activity (%) o-J. toO OoJ .Oo .

O CO CD ] -n O

70 m Figure 3-7: PIAS3 does not bind to FOXC consensus binding sequence and does not affect FOXC2 DNA binding.

Upper panel: COS-7 cells transfected with V5-FOXC2 expression vector or Xpress-PIAS3 or co-transfected with V5-FOXC2 and increasing amounts of Xpress-PIAS3. Two days after transfection, cell lysates were subjected to EMS As using 32P-labeled FOXC binding site probes. EMSA binding buffer alone (the first line) and COS-7 cell lysate (the second line) were also used in EMSA reactions as negative controls. Lower panel: Immunoblots showed the amounts of V5-FOXC2 and /or Xpress-PIAS3 used in each corresponding EMSA reaction.

140 FOXC2 * * *

•^ rr • i^* "'v^e*' '"""""'I

J **>>'•

complce^

* *«"? / %a < sit iL„ '# 's\vV

FOXC2

=FIAS3

141 Figure 3-8: PIAS3 can up-regulate the transcriptional activity of endogenous FOXC2.

(A) Upper panel: HeLa cells transfected with control siRNA or FOXC2 siRNA first and then transfected with FGF19RE luciferase reporter construct along with empty pcDNA4 or Xpress-PIAS3, respectively. To control for transfection efficiencies, cells were co-transfected with pRL-CMV vector, which expresses renilla luciferase. Two days later, dual-luciferase assays were performed. The activity of firefly luciferase was normalized by that of renilla luciferase and the value from cells with control siRNA and empty vector transfection was scaled to 1. *, significant difference (p < 0.01) between PIAS3-induced expression of luciferase reporter with siRNA control transfection and PIAS3-induced expression of luciferase reporter with siFOXC2 transfection. Lower panel: Immunoblots showed the amounts of Xpress-PIAS3, endogenous FOXC2 and TFIID which was used as loading control in each corresponding transactivation assay. (B) HeLa cells were transfected with empty pcDNA4 plasmid or with Xpress-PIAS3 construct. Two days after transfection, nuclear extracts were subject to immunoblot analysis with an anti-FOXC2 antibody or an anti-Xpress anti-body to determine the amount of endogenous FOXC2 (middle) or Xpress- PIAS3 (top). TFIID was used as a loading control (bottom).

142 * e 1 1 5 • 4.S1 a u 3.67 S 4

i 3 • o

b 1.00 0.92 1 •

0 J I 1 slctr/empty sictr/P!AS3 siFOXC2/empfy siFOXC2/PIAS3

TFIID B

Xp-PIAS3

FOXC2 &Mm

TFIID

143 Chapter 4. Analyses of the regulation of target genes by FOXC1 and FOXC2

144 Introduction Forkhead Box (FOX) proteins are a family of transcription factors that contain an evolutionarily conserved DNA-binding domain known as the forkhead domain (FHD). The name comes from the spiked-head phenotype in embryos of the Drosophila fork head mutant (1). Subsequent identification of the 110-amino acid forkhead DNA binding domain that is highly conserved between fork head and the mammalian hepatocyte nuclear factor 3 (FTNF3) transcription factors indicates this domain defines a novel transcription factor family (129, 176). The FHD is a subtype of the helix-turn-helix motif with two wing-like structures flanking the three a helices giving rise to a butterfly-like appearance and therefore it is also known as the winged-helix motif (5). Since the discovery of the first forkhead gene, many others have been identified in a variety of eukaryotic organisms, from yeast to human. The number of forkhead genes has increased during evolution with larger numbers in more complex organisms indicating that increased complexity in body plan may be a driving force behind the expansion of the forkhead gene family (3). The nomenclature of the chordate forkhead genes was unified in 2000 and these genes are divided into 18 subclasses (A to Q), according to the amino acid sequence of their FHDs (177). Forkhead transcription factors regulate a variety of cellular and developmental processes, such as cell fate specification, cell growth and differentiation, tumorigenesis and metabolic homeostasis (3, 128). FOXC1 and FOXC2 are the two members in the FOXC subfamily. The chromosomal localization and genomic organization of these two genes suggest that they may derive from a common ancestral gene through inter- and intra- chromosomal duplications (3, 128). Human FOXC1 and FOXF2 are located in proximity on chromosome 6p25, while FOXC2 and FOXF1 are closely located on chromosome 16q24 (3). This arrangement is possibly the results from two chromosome duplication events during evolution. The first duplication of a ancestral gene gave rise to ancestral FOXC and FOXF genes. The second more recent duplication of the entire locus, followed by transferring of one copy to a different chromosome, then gave rise to the present four genes. Additional

145 sequence and functional divergence arose from these processes as well (3). In agreement with this inference that FOXC1 and FOXC2 arise from a common ancestor, the two proteins share a strong amino acid sequence homology. Their FHDs are essentially identical with only two amino acid difference. FOXC1 contains glutamic acid residues at positions 90 and 110, whereas FOXC2 carries aspartic acid residues at the corresponding positions. However, the N- and C- terminal sequence flanking FHD are more variable between FOXC1 and FOXC2 (45). The protein sequence comparison of FOXC1 and FOXC2 suggests that they may bind to the same target genes, nevertheless these two proteins may differ in some functions. During mouse embryonic development, Foxcl and Foxc2 have largely overlapping expression patterns. In addition, Foxcl and Foxcl deficient mouse models often demonstrate similar developmental defects in the tissues that they are co-expressed, indicating functional interaction between the two genes. Co- expression of Foxcl and Foxc2 have been found in paraxial, cephalic and nephrogenic mesoderm, somites, endothelial cells of the heart and blood vessels and mesenchyme of the aortic arches, valves, and outflow tract (48, 49, 51, 53-55, 96). As a result, it is not surprising to find that homozygous null mutants for each gene have many developmental abnormalities in skeletal, cardiovascular and genitourinary systems (48, 52, 53, 55, 95, 96). Interestingly, Foxcl+"; Foxc2+' compound heterozygotes show non-allelic non-complementation in cardiovascular, urogential and skeletal development. The compound heterozygotes die prenatally with similar abnormal phenotypes in cardiovascular and urogential systems and in somitogenesis to those of either single homozygous null mutants, suggesting that Foxcl and Foxc2 have similar, dose-dependent functions, and can partially compensate for each other (51, 53, 55). In agreement with this statement, some developmental defects are more severe or only seen in compound Foxcl+; Foxc2'A homozygotes. For example, compound Foxcl'1'\ Foxc2~' homozygotes lack of somites formation and segmentation of the presomitic mesoderm; whereas the somites do form in either single homozygotes or the compound heterozygotes (51). In the absence of Foxcl

146 and Foxc2, transcription of paraxis, Mespl, Mesp2, Hes5 and Notchl in the anterior presomitic mesoderm is lost and as is the formation of sharp boundaries of Dill, Lfng and ephrinB2 expression, which suggest that Foxcl and Foxc2 interact with the Notch signaling pathway in prepatterning of anterior and posterior domains of the presumptive somites (51). In compound homozygotes, paraxial mesoderm acquires intermediate mesoderm fate as indicated by the medially expansion of intermediate mesoderm markers, Osrl, Lhxl and Fg/8, into the paraxial domain, indicating that Foxcl and Foxc2 together regulate the specification of mesoderm to paraxial versus intermediate fates (178). In addition to somitogenesis and trunk mesoderm specification, the development of cardiovascular system seems to be more severely affected in compound homozygous mice. Foxcl'"; Foxc2~' homozygotes die earlier than either single homozygotes with reduction in heart size, reduction in the size of the first branchial arch, absence of the second branchial arch, the outflow tract, the inflow tract and right ventricle (51, 124). Foxcl and Foxc2 are both expressed in the developing second heart field, which contributes to the formation of the cardiac outflow tract. The expression of markers for the second heart field, including Fg/8, FgflO, Tbxl are significantly decreased in compound homozygotes. Moreover, Foxcl and Foxc2 directly regulate the expression of Tbxl in the second heart field through two Foxc binding sites on the promoter of Tbxl (179), while Tbxl controls the expression of Fg/8 and FgflO in this region (180). Taken together, Foxc genes play key roles in the morphogenesis of the cardiac outflow tract by acting upstream of the Tbxl-Fgf cascade (124). Foxcl'''; Foxc2'/' homozygotes display arteriovenous malformation (fusion of arteries and veins without an intervening capillary) and lack of expression of arterial markers including Notchl, Notch4, D114, Jagged 1, Hey2, and ephrinB2 (103). The expression of DU4 and Hey2 are directly regulated by both Foxcl and Foxc2. VEGF have a synergistic effect with Foxcl and Foxc2 on activation of D114 and Hey2 in endothelial cells, suggesting that Foxc proteins interact with VEGF and Notch signaling to regulate arterial gene expression (126).

147 Disease-causing mutations have been identified in human FOXC1 and FOXC2 genes. Mutations in FOXC1 result in Axenfeld-Rieger (AR) malformations characterized by the severe developmental defects in the anterior segment of the eye and higher predisposition to glaucoma (32, 33). Mutations in FOXC2 lead to lymphedema with distichiasis (LD) characterized by the early onset lymphedema of lower limbs and an accessory row of eyelashes (79, 85). Although LD patients with FOXC2 mutations do not typically present with severe ocular phenotypes, a closely eye examination of patients with FOXC2 R121 H and FOXC2 S125L reveals mild anterior segment abnormalities (92). Interestingly, similar disease-causing mutations to FOXC2 R121H and FOXC2 S125L occur at parologous positions in FOXC1. Moreover, these two FOXC2 mutants and their parologous FOXC1 mutants exhibit similar molecular defects in their functions, shown as the incapability to bind DNA and transactivation (45). These findings suggest that both FOXC1 and FOXC2 are required for the development of the anterior segment of the eye. Consistant with this suggestion, Foxcl and Foxc2 have largely overlapping expression patterns in the periocular mesenchyme and in the anterior segment structures derived from periocular mesenchyme during mouse ocular development (49, 58, 95). Foxc2+/~ mice have similar ocular abnormalities as those of Foxcl+/'. Foxcl+/' ; Foxc2+/~ compound heterozygotes also have ocular findings that are comparable to either single heterozygotes (61). However, defects in the ciliary body, congenital corneal vascularization and open eyelids at birth are specific to the Foxcl+' ; Foxc2+' compound heterozygotes (61). The fact that ocular defects not found in either single heterozygote are discovered in double heterozygotes demonstrates that Foxcl and Foxc2 have overlapping functions in the developing eye. Taking together that Foxcl and Foxc2 have overlapping expression patterns and functions in the developing eye and that their DNA binding domains are almost identical, I hypothesize that similar to the situation in cardiovascular system, FOXC1 and FOXC2 share at least some direct target genes during ocular development. Previous work from our laboratory has identified FGF19 and FOXOIA as direct targets of FOXC1 in ocular cell lines and in zebrafish eyes (65, 71). In this

148 chapter, my preliminary results demonstrate that FOXC2 is also able to regulate the expression of FGF19 and FOXOIA in the ocular cell lines. Interestingly, in the transactivation assays using luciferase reporter genes under the control of either the FGF19 or FOXOIA promoter, I did not observe a synergistic or additive effect from co-transfection of FOXC1 and FOXC2 on inducing the expression of the luciferase gene. Instead, an intermediate expression level was obtained on the FGF19RE reporter and an expression level close to that of FOXC1 alone was obtained on the FOXOIA reporter. The lack of synergy or addition in the presence of both FOXC1 and FOXC2 does not result from competition for binding-sites between the two FOXC transcription factors, since the transactivation of WT FOXC2 together with a DNA-binding deficient FOXC1 mutant or the transactivation of WT FOXC1 together with a DNA-binding deficient FOXC2 mutant was still lower than that of WT FOXC2 or WT FOXC1 alone. Finally I show that FOXC1 and FOXC2 interact with each other in an in vitro pull-down assay, suggesting that FOXC1 and FOXC2 may regulate the activity of each other by forming a protein complex.

149 Methods

Plasmids N-terminally tagged 6xHis::FOXCl bacterial expression vector, pET28- FOXC1, was built by Fred Berry and has been previously described (41). N- terminally tagged 6xHisXpress::FOXCl was expressed in pcDNA4 His-Max B. The construct was assembled by Ramsey Saleem and has been previously described (42). pcDNA4-FOXCl R127H was constructed by Ramsey Saleem and has been previously described (45). FOXC1 full length cDNA was amplified by PCR from pcDNA4-FOXCl and subcloned into the pcDNA3.1/nV5-DEST vector (Invitrogen) in-frame to the V5 epitope by gateway technology. N-terminally tagged 6xHisXpress::FOXC2 and mutant FOXC2 R121H fusion proteins were expressed in pcDNA4 His-Max C. The constructs were assembled by Fred Berry and have been previously described (45). FOXC2 cDNA was amplified by PCR from pcDNA4-FOXC2 and subcloned into the pcDNA3.1/nV5-DEST vector (Invitrogen) in-frame to the V5 epitope by gateway technology or into the bacterially expressed pET28b vector in frame with the 6xHis epitope. All newly built vectors were sequenced to confirm that no mutations were introduced into the cDNAs and the cDNAs were in-frame to the epitopes. The pGL3-TK- luciferase reporter vector under the control of a 354 bp fragment from FGF19 promoter (FGF19RE), which contains a FOXC binding site, was constructed by Yahya Tamimi and has been previously described (65). The pGL3-basic luciferase reporter vector under the control of a DNA fragment, corresponding to 580 bp upstream and 250 bp downstream of the putative transcription start site of FOXOIA gene (FOXOIARE), was built by Fred Berry and has been previously described (71).

Mammalian cell culture and transfection HeLa cells and non-pigmented ciliary epithelium cells (NPCE) cells were maintained in Dulbeco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum in a 37°C, humidified incubator under an atmosphere containing a constant 5% CO2. The day prior to transfections, plates

150 were seeded (106 cells in 10ml media per 100mm Petri dish and 4xl04 cells in 1ml media per 15mm well ) so as to reach approximately 30-50% confluence the following day. Transfections were performed using FuGene 6 (Roche) or TranlT- LT1 transfection reagent (Minis bio) according to the manufacturer's protocol. Transfected cells were subjected to protein extraction or Dual-luciferase assays (Promega) 48 hours after transfection.

Immunoblot analysis The detailed description can be found in Chapter 2. Primary antibodies and the amount of each antibody used in immunoblot are anti-Xpress (1:5000), anti- V5 (1:5000). HRP-conjugated secondary antibodies used are goat anti-mouse (1:5000).

Transactivation assays Transactivation assays were performed using Promega Dual-luciferase reporter assay system according to the manufacturer's instructions. Subcultured HeLa cells on 24-well tissue culture plates were transfected respectively with pcDNA4-FOXCl expression vectors, pcDNA4-FOXCl R127H expression vector, pcDNA4-FOXC2 expression vectors, pcDNA4-FOXC2 R121H, both WT FOXC1 and FOXC2 expression vectors, both WT FOXC1 and FOXC2 R121H expression vectors or both WT FOXC2 and FOXC1 R127H expression vectors along with either pGL3-TK-FGF19RE luciferase reporter or pGL3-basic- FOXOIA luciferase reporter, and pRL-CMV (Promega). The total amount of transfected DNA was equalized with empty vectors. The transactivation assays were performed two days after transfection. The detailed procedure can be found in Chapter 2

Chromatin Immunoprecipitation (ChIP) Approximately 108 NPCE cells were used for ChIP analyses. Cells were cross-linked with 1% (final concentration) of formaldehyde added directly into the culture media 10 minutes at room temperature with gentle shaking. The crosslink reaction was stopped by incubating with 0.125M (final concentration)

151 glycine for 5 minutes at room temperature with gentle shaking. The culture media was then removed and cells were washed with PBS twice. Cells were pooled into 50ml conical tubes by scraping with PBS (5ml PBS for each 100mm plate) and spun down at 650 x g for 5 minutes at 4°C. The cell pellets were resuspended in 5 ml of ChIP cell lysis buffer [5mM PIPES pH8.0, 85mM KC1, 0.5% IPEGAL CA- 630, ImM PMSF, 1/200 dilution of mammalian protease inhibitor cocktail (Sigma)] and left on ice for 10 minutes. The cell lysates were subject to homogenizing by 6 strokes of a Dounce homogenizer (B pestle) and centrifuged at 2500 x g for 5 minutes at 4°C. The cell pellet was resuspended in 1 ml of ChIP nuclear lysis buffer (50mM Tris-Hcl, pH8.0, lOmM EDTA, 1% SDS, ImM PMSF, 1/200 dilution of mammalian protease inhibitor cocktail) and incubated on ice for 10 minutes. The lysates was sonicated by six-15 seconds bursts at setting 7 on a Sonic dismembrator 60 (Fisher Scientific) so as to shear chromatin. The lysate was centrifuged at 15,000 x g for 10 minutes at 4°C and subsequently pre- cleared with 50 ul of protein G beads (Sigma), which were blocked beforehand with BSA and sheared salmon sperm DNA for 1 hour at 4°C with rotating. Following the preclear, cell lysate was centrifuged at 6000 x g for 1 minute at 4°C and the supernatant was split into 4 eppendorf tubes with 250 ul aliquot each. Each aliquot was diluted with 1 ml of ChIP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-Hcl, pH8.0, 167 mM NaCl) and incubated overnight at 4°C with 2 ug of anti-FOXCl (Abeam), anti-FOXC2 (Abeam), anti-GFP (Abeam) and anti- acetylated lysine 9 of Histone 3 (ACK9 Histone H3, Cell Signaling), respectively. In the next day, 25 ul of protein G beads (blocked with BAS and sheared salmon sperm DNA) were added into each tube and incubated with rotating for another 2 hours at 4°C. The protein G beads were collected by centrifugation at 6000 x g for 1 minute at 4°C, followed by three 5-minutes washes with ChIP wash buffer (0.1% SDS, 1% Triton X-100, 2mM EDTA, 150mM NaCl, 20mM Tris-Hcl, pH8.0) and one 5- minutes final wash with final wash buffer (0.1% SDS, 1% Triton X-100, 2mM EDTA, 500 NaCl, 20mM Tris-Hcl, pH8.0). The protein-DNA complexes were eluted from the beads by adding 300 ul of elution buffer (1% SDS, lOOmM NaHCQ3) and rotated

152 15 minutes at room temperature. The beads were removed by centrifugation at 6000 x g for 1 minute at 4°C and the supernatants were transferred into fresh tubes. Each tube of supernatant was treated with 0.1 mg of protease K (Invitrogen) overnight in a 65°C water bath. The DNA was purified by PCR cleanup protocol (QIAGEN). In brief, the supernatant treated with protease K was mixed with 1.2 ml of PB buffer and transfer onto a Qiaquick PCR cleanup column, followed by 1 minute spin at maximum speed. The column was then washed with 750 ul of PE buffer and spun 1 minute at maximum speed. Finally, the purified DNA was eluted in 80 ul of MilliQ water. Collected DNA was diluted 1 in 1000 and 5 ul of the dilution was used in each PCR reaction. A 100 bp fragment in the FGF19 promoter containing a FOXC binding site was amplified with the following primer set: 5'-GCA GCC AAG CCA GTT AGC-3' and 5'-AGA TCC TCC AGC CGG AAC-3'. A 140bp fragment in the FOXOIA promoter containing a FOXC binding site was amplified with the following primer set: 5'-AGT ACT CGG CTC TGC TGC TC-3' and 5'- GGG GTA GTG GGG TGT TTT TC-3'.

153 Results

Both FOXC1 and FOXC2 bind to the FGF19 and FOXOIA promoter in vivo I performed chromatin immunoprecipitation (ChIP) in non-pigmented ciliary epithelial (NPCE) cells to determine whether endogenous FOXC2, like its paralog FOXC1 (65), could bind to the FGF19 and FOXOIA promoters in vivo. Before ChIP analysis, the expression of endogenous FOXC1 and FOXC2 in NPCE cells was determined. NPCE nuclear extract was subjected to immunoblot analysis with antibodies against FOXC1 and FOXC2, respectively. As illustrated in Figure 4-1 A, the FOXC2 antibody detected a band at an apparent molecular weight of 65 kDa, consistent with the predicted molecular weight of FOXC2. FOXC1 migrated at a higher molecular weight of approximately 70 kDa, again in line with the higher predicted molecular weight of FOXC1 compared with FOXC2. Because FOXC1 and FOXC2 share a high degree of amino acid sequence identity (98% in the FHD, 36% in sequence other than the FHD), the specificities of the antibody against FOXC1 and antibody against FOXC2, which would be used in the ChIP analysis, were examined. Cell lysates containing recombinant Xpress tagged FOXC1 or Xpress tagged FOXC2 were immunoblotted with anti-Xpress antibody, anti-FOXC 1 antibody and anti-FOXC2 antibody, respectively. As shown in Figure 4-1 B, the FOXC1 antibody only recognized the recombinant Xpress tagged FOXC1 but not the recombinant FOXC2 protein. Similarly, the anti- FOXC2 antibody only recognized the recombinant Xpress tagged FOXC2 and did not show cross reaction to FOXC1 protein. The results of ChIP analysis were illustrated in Figure 4-1 C. The upstream promoters of the FGF19 and FOXOIA genes containing FOXC-binding sites were recovered by PCR amplification from FOXC1, FOXC2 and histone-H3-acetylated lysine-9 immunoprecipitates. These data indicate that both FOXC1 and FOXC2 can occupy the promoter regions of FGF19 and FOXOIA genes in vivo.

154 Both FOXC1 and FOXC2 activate transcription from the FGF19 and FOXOIA promoters Dual-Luciferase assays were performed to test if FOXC2 can activate transcription from FGF19 regulatory element and FOXOIA regulatory element in the cells. The luciferase construct under the control of a 354 bp amplicon of the FGF19 promoter (FGF19RE), which containing a FOXC-binding site, was built previously in our laboratory (71). This luciferase construct was co-transfected into HeLa cells with pcDNA4-FOXCl or pcDNA4-FOXC2 or empty pcDNA4 plasmids. The luciferase activities were monitored afterward. As indicated in Figure 4-2 A, promoter activity was enhanced in the cells when either FOXC1 or FOXC2 was transfected. Similarly, FOXC1 and FOXC2 also significantly enhanced the expression from the luciferase construct (FOXOIARE) with a DNA fragment, 580 bp upstream and 250 bp downstream of the predicted transcription start site of FOXOIA gene, placed in front of a luciferase reporter (45) (Figure 4-2 B).

FOXC1 and FOXC2 do not function synergistically or additively in transactivation assays In order to determine how FOXC1 and FOXC2 activate the same promoter, dual-luciferase assays were performed. HeLa cells were co-transfected with pcDNA4-FOXCl, pcDNA4-FOXC2 and FGF19RE luciferase reporter construct or FOXOIARE luciferase reporter construct. The luciferase activities were then monitored. Surprisingly, in the presence of both FOXC1 and FOXC2, an intermediate expression level was obtained on the FGF19RE reporter, which is significantly higher than that from FOXC1 alone but significantly lower than that from FOXC2 alone (Figure 4-3 A). An expression level close to that from FOXC1 alone but significantly lower than that from FOXC2 alone was obtained on the FOXOIA reporter, when FOXC1 and FOXC2 were co-transfected. (Figure 4-3 B). These results indicated that FOXC1 and FOXC2 do not function synergistically or individually in the context of FGF19 promoter or FOXOIA promoter. To determine if FOXC 1 and FOXC2 compete with each other for the

155 same FOXC-binding site, the expression constructs encoding mutant FOXC1 R127H and mutant FOXC2 R121H, respectively, were used in the transactivation assays. FOXC1 R127H and FOXC2 R121H both are disease-causing mutations that occur at parologous positions in the FHDs of FOXC1 and FOXC2 (42, 45). Previous studies have showed that these two mutant proteins lose their DNA- binding ability and transactivity abilities (65, 71). As showed in Figure 4-3 A and B, neither FOXC1 R127H nor FOXC2 R121H can enhance the luciferase activity when FGF19RE reporter or FOXOIARE reporter was used in the transactivation assays, which is consistent with previous studies. Although these two mutant proteins do not bind to DNA, when FOXC1 R127H and wild type FOXC2 or FOXC2 R121H and wild type FOXC1 were co-transfected, the expression levels of the luciferase gene were still lower than the levels when either wild type protein was expressed alone. These results indicated that competition for FOXC- binding sites is not the underlying mechanism for the non-synergistic and non- additive effect of the two FOXC proteins on their common target genes. Instead, other mechanisms, such as the two FOXC proteins forming a protein complex so as to affect their transactivity, may exist. Very interestingly, in the transactivation assays on either FGF19 promoter or FOXOIA promoter, the luciferase activity from the co-transfection of FOXC2 R121H and WT FOXC1 is significantly lower than that from the co-transfection of both WT FOXC1 and FOXC2 (Figure, 4-3 A and B), indicating that the mutant protein FOXC2 R121H has a dominant negative effect on WT FOXC 1.

FOXC1 and FOXC2 are able to interact with each other I conducted Ni2+ pull-down assays to determine if FOXC 1 and FOXC2 can interact with each other and regulate their transactivity. Full-length wild type FOXC1 and FOXC2 were subcloned separately into the bacterial expression vector (pET28), allowing expression of FOXC 1 and FOXC2 as 6xHis tagged fusion proteins. The cell lysate prepared form HeLa cells transfected with V5- tagged FOXC 1 was incubated with 6xHis-FOXC2 bound to Ni -agarose beads and the cell lysate prepared from HeLa cells transfected with V5-tagged FOXC2

156 was incubated with 6xHis-F0XCl bound to Ni -agarose beads. As a negative control, the same amount of cell lysate was incubated with empty Ni2+-agarose beads. Immunoblot analysis using an anti-V5 antibody showed that FOXC1 and FOXC2 can interact with each other (Figure 4-4 A and B). These results indicate that FOXC1 and FOXC2 can form a protein complex and that the interaction between these two FOXC proteins may therefore regulate their function.

157 Discussion The FGF19 and FOXOIA genes have been previously identified as direct targets of FOXC1 in ocular cell lines and the developing zebrafish eye (65, 71). Fgfl9 regulates the development of cornea through FGFR4/MAK pathway. Knocking down of Fgfl9 in the developing zebrafish embryos leads to anterior segment dysgenesis in the eye (65). FOXOIA is a mediator of FOXC1-dependent cell viability and resistance to oxidative stress in the eye (71). FOXC1 and FOXC2 are the two highly close related members of FOXC transcription factor family. The DNA binding domains of FOXC 1 and FOXC2 are 98% identical, resulting in the prediction that their DNA binding specificities also are identical. In this study, I found that both FOXC1 and FOXC2, in vivo, bind to the FOXC- binding sites within the enhancer/promoter regions of FGF19 and FOXOIA genes in ocular cell line, where both of the FOXC proteins are expressed. Moreover, both FOXC proteins are able to activate the promoters of FGF19 and FOXOIA in the transactivation assays. These results are consistent with our hypothesis that FOXC1 and FOXC2 share common target genes in the ocular tissues. These results are also consistent with several recent studies, which have shown that FOXC1 and FOXC2 can both directly regulate the expression of a number of common down-stream target genes in other, non-ocular tissues (69, 126, 179). In epithelial cells, these common target genes include D114 and Hey2, which are key players of VEGF/Notch signaling pathway during arterial fate specification (126). Tbxl, a major contributor to the normal cardiac and cranial-facial development, is also a common FOXC1 and FOXC2 target gene (69, 179). Taken together, all these results strongly suggest that the direct downstream target genes of FOXC 1 and FOXC2 largely overlapped. My studies, however, indicate that FOXC1 and FOXC2 are not simply redundant, but rather these two proteins have a complex relationship with each other in how they co-regulate their target genes. In my study, comparing the induced expression of luciferase gene from co-transfection of FOXC 1 and FOXC2 with those from transfection of FOXC 1 or FOXC2 alone on either the FGF19RE reporter construct or on the FOXOIARE reporter construct, co-

158 transfection of FOXC1 and FOXC2 did not exhibit a synergistic or additive effect on inducing the expression of the luciferase gene. An intermediate transactivity from FOXC co-transfection, which is lower than that from FOXC2 alone, but higher than that from FOXC1 transfection alone, was observed on the FGF19RE reporter construct. On the FOXOl ARE reporter construct, the level of transactivity from FOXC co-transfection is close to that from FOXC1 alone, but is lower than that from FOXC2. These results suggest that FOXC1 and FOXC2 do not transactivate the reporter genes in a synergistic or an independent manner, but instead that they may compete for the FOXC-binding sites. However, co- transfection of mutant FOXC1 R127H, which does not bind DNA, with wild type FOXC2 did not restore FOXC2 transactivity to the level of wild type FOXC2 alone. A similar result was obtained when wild type FOXC 1 was transfected with mutant FOXC2 R121H, a mutant that also has lost its DNA binding ability. The failure of the DNA-binding deficient mutant FOXC proteins to rescue the transactivity of their wild type parolog proteins indicates that there may not be a simple competition between FOXC1 and FOXC2 for the DNA binding sites and therefore promoted me to examine the possibility that FOXC 1 and FOXC2 can interact with each other and subsequently regulate their transcriptional activities. The results from my Ni2+ pull-down assays (Figure 4-4) support this hypothesis and demonstrate that FOXC1 and FOXC2 physically interact. This is the first time that the two FOXC proteins have been shown to be able to interact with each other. Previously, members from the FOXP subfamily have been shown to be able to form homo- or heterodimers through their N-terminal leucine zipper domains; a dimerization required for the FOXP proteins to bind DNA (181). However, how forming heterodimer between FOXC1 and FOXC2 regulates their transactivation is not clear. Since co-transfection of FOXC 1 and FOXC2 can still transactivate the luciferase reporter gene, though not in a synergistic or an additive manner, one possibility is that the DNA binding ability of FOXC 1 or FOXC2 is not significantly affected by their interaction. Instead, either FOXC1 or FOXC2 in the heterodimer appears to have an equal opportunity to bind the FOXC-binding site. Heterodimer formation may disturb the efficiency of the recruitment of other

159 transcription co-factors or lead to the formation of a less potent transcriptional activator complex, resulting in an intermediate transactivity in the luciferase assays when both FOXC proteins are transfected. This hypothesis is supported by a previous study showing that the interaction between FOXG1 and FOXH2 interferes with the recruitment of Smad2 onto FOXH2 and as a result interferes with the antiproliferative activity of TGF|3 signaling mediated by FOXH2 (182). Alternatively, only FOXC1 and FOXC2 monomers bind the FOXC-binding site, whereas the heterodimer formed by FOXC1 and FOXC2 does not. The formation of dimer between FOXC1 and FOXC2 may function as a buffer system to a increased FOXC1 and FOXC2 concentration in the cells. This scenario could as well lead to an intermediate transactivity when both FOXC1 and FOXC2 are present in the cells. Alterations in FOXC1 causes eye malformations, not only when the gene dosage is decreased, but also when it is increased by chromosomal duplications (35-37). Other known examples in which duplication or deletion of a single gene cause similar conditions include hereditary neuropathy with liability to pressure palsy and Charcot-Marie tooth disease type 1A, and emphasize the importance of precise control of gene dosage in the normal development (183). My results showing that FOXC1 and FOXC2 regulate each other's transcriptional activity through forming a protein complex may point to a mechanism used to fine-tune the overall FOXC protein activity in the cells that highly express FOXC 1 and FOXC2 together during the embryonic development. Studies of Foxc knock-out mice have found that Foxcl or Foxc2 null mice or Foxcl and Foxc2 compound heterozygous mice have similar defects in skeleton, cardiovascular and kidney development (51, 53, 55). Moreover, compound Foxcl and Foxc2 null mice display more severe abnormalities than either of the single null mice (51, 103, 124, 178). From these studies it was proposed that Foxcl and Foxc2 play similar, dose-dependent roles during normal embryonic development (51, 53, 55, 103, 124, 178). However, my results showing that FOXC1 and FOXC2 can physically interact clearly indicate that the relationship between FOXC1 and FOXC2 is more complex. Due to their interaction, co-transfection of

160 F0XC1 and FOXC2 resulted in an expression level of the reporter gene in between the expression levels induced by either FOXC1 or FOXC2 alone (Figure 4-3); whereas the dose-dependent model would predict an expression level that is higher than those induced by either FOXC1 or FOXC2 alone. However, my results and the dose-dependent model are not mutually exclusive. The studies of Foxcl and (or) Foxc2 deficient mouse models reveal the overall phenotypic manifestations and indicate the requirement of the intact overall activity of Foxc transcription factors for the normal development; whereas, my studies focus on only two of the genes that are transcriptional regulated by FOXC1 and FOXC2 and provide evidence for an additional level of regulation of the FOXC activities In the transactivation assays on either the FGF19 promoter or FOXOIA promoter, the luciferase activity from the co-transfection of FOXC2 R121H and WT FOXC1 was lower than that from the co-transfection of both WT FOXC1 and FOXC2, indicating that the mutant protein FOXC2 R121H has a dominant negative effect on WT FOXC1 (Figure 4-3). Interestingly, a detailed ocular examination of LD patients with FOXC2 point mutations within FHD, including FOXC2 R121H, found that these patients displayed mild anterior segment dysgenesis, such as iris hypoplasia (92). This impaired function of WT FOXC1 by FOXC2 R121H may provide an explanation for the ocular phenotypes that are found in LD patients with FOXC2 R121H point mutation. As discussed above, WT FOXC1 and FOXC2 interact with each other and may help with keeping the total transcriptional activity of FOXC proteins in an appropriate level and prevent it from being extremely high in the cells with FOXC1 and FOXC2 co-expressed. FOXC2 R121H mutant proteins can neither bind to DNA nor transactivate the expression of a reporter gene (45). Formation of a heterodimer or protein complex between WT FOXC1 and FOXC2 R121H may reduce the FOXC activity below a critical level and thus result in the ocular defects. In summary, I found that the two previously identified FOXC1 direct target genes, FGF19 and FOXOIA, are also directly regulated by FOXC2. FOXC1 and FOXC2 interact with each other and regulate the total transcriptional activity of the FOXC proteins in the cells co-expressing both FOXC1 and FOXC2. These

161 results provide evidence of additional regulatory mechanism for the FOXC transcriptional factors, the dose of which are very critical for the normal embryonic development and therefore are tightly controlled. Furthermore, the previously suggested haploinsufficient and LD-causing mutant FOXC2 R121H protein demonstrated a dominant-negative effect on wild type FOXC1, which may lead to a decrease in FOXC activity below a critical level and subsequently the anterior segment abnormalities in the eye.

162 Figure 4-1: Both FOXC1 and FOXC2 bind to the FGF19 and FOXOIA promoter in vivo

(A). Immunoblot analyses of FOXC1 and FOXC2 protein expression in NPCE nuclear extract. (B). Left: FOXC1 antibody does not cross react to FOXC2 protein. Cell lysates transfected with Xpress tagged FOXC1 (XpFOXCl) and Xpress tagged FOXC2 (XpFOXC2) were analyzed by immunoblot with anti-Xpress antibody or anti-FOXCl antibody, respectively. The anti-Xpress antibody recognized both Xpress tagged FOXC1 and Xpress tagged FOXC2. The anti- FOXCl antibody only recognized Xpress tagged VOXCX. Right: FOXC2 antibody does not cross react with FOXC1 protein. Cell lysates transfected with Xpress tagged FOXC1 and Xpress tagged FOXC2 were analyzed by immunoblot with anti-Xpress antibody or anti-FOXC2 antibody. The anti-Xpress antibody recognized both Xpress tagged FOXC1 and Xpress tagged FOXC2. The anti- FOXC2 antibody only recognized Xpress tagged FOXC2. (C). Binding of FGF19 and FOXOIA promoter regions by FOXC1 and FOXC2 in vivo. Chromatin from NPCE cells were cross-linked and were immunoprecipitated with antibodies against GFP, FOXC1, FOXC2, or acetylated lysine 9 of Histone H3 (ACK9-H3). Sonicated ChIP input is a 1:10,000 dilution of cross-linked chromatin used for immunoprecipitations. Immunoprecipitated DNA was PCR amplified using primers designed to amplify regions flanking putative FOXC binding sites on the promoters of the FGF19 and FOXOIA genes (65, 71).

163 o CD >

g * a § FGF19^ | O O D O z I I I I •n •o XpFOXd O o X m X o 3 water control XpF0XC2 C

mvi-oxo d o_ ChIP input n p i fl> |j[ XpFOXCI X aGFPChIP -n »•* O X XpFOXC2 o aFOXCIChIP o aFOXC2ChlP ^H a ACK9-H3 ChIP (J £ w 00 O o» 09 O Up; • Genomic DNA

E[ > DNA Ladder s | XpF0XC2 I X XpFOXCI

-n XpFOXC2 O X XpFOXCI o Figure 4-2: Both FOXC1 and FOXC2 activate transcription from the FGF19 and FOXOIA promoters

(A). Transactivation assays of the FGF19 promoter region. The FGF19RE luciferase reporter construct was co-transfected with empty pcDNA4 or pcDNA4- FOXC1 expression vector or pcDNA4-FOXC2 expression vector into HeLa cells. To control for transfection efficiencies, cells were also transfected with pRL- CMV, which expresses renilla luciferase. Activity of firefly luciferase was normalized by that of renilla luciferase and the value of empty vector was scaled to 1. (B). Transactivation assays of the FOXOIA promoter region. The FOXOl ARE luciferase reporter construct was co-transfected with empty pcDNA4 or pcDNA4-FOXCl expression vector or pcDNA4-FOXC2 expression vector into HeLa cells. To control for transfection efficiencies, cells were also transfected with pRL-CMV, which expresses renilla luciferase. Activity of firefly luciferase was normalized by that of renilla luciferase and the value of empty vector was scaled to 1.

165 FGF19 promoter

14 n

12

I 8 5 6

\* on 2 - \ -'r 1 pcDNA4 FOXC1 WT FOXC2WT

B FOX01A promoter

3D -

30 - b. S 8 25 • > * ,n o. 20 • E a S 15 - 0) •o> £ 10 - at ft 5 -

0 - • 1 1 pcDNA4 FOXC1 WT FOXC2WT

166 Figure 4-3: FOXC1 and FOXC2 do not function synergistically or additively in transactivation assays

(A). Upper panel: Dual luciferase assays of the FGF19 promoter region. HeLa cells were co-transfected with the FGF19RE reporter construct and expression vector(s), as indicated. To control for transfection efficiencies, cells were also transfected with pRL-CMV, which expresses renilla luciferase. The total amount of transfected DNA was equalized with empty vector. Activity of firefly luciferase was normalized by that of renilla luciferase and the value of empty vector was scaled to 1. *, significant difference (p < 0.05). Lower panel: Immunoblots showed wild type and/or mutant FOXC protein(s) expression in each corresponding luciferase assay. (B). Upper panel: Dual luciferase assays of the FOXOIA promoter region. HeLa cells were co-transfected with the FOXOl ARE reporter construct and expression vector(s), as indicated. To control for transfection efficiencies, cells were also transfected with pRL-CMV, which expresses renilla luciferase. The total amount of transfected DNA was equalized with empty vector. Activity of firefly luciferase was normalized by that of renilla luciferase and the value of empty vector was scaled to 1. *, significant difference (p < 0.05). Lower panel: Immunoblots showed wild type and/or mutant FOXC protein(s) expression in each corresponding luciferase assay.

167 FGF19 promoter

B FOX01A promoter

168 Figure 4-4: FOXC1 and FOXC2 interact with each other

(A). HeLa cell lysates transfected with V5 tagged FOXC1 (V5-FOXC1) were subjected to Ni pull-down assays with the 6xHis tagged FOXC2 bound to the Ni2+-NTA beads (Ni-FOXC2) or the empty beads (Ni). Bound proteins were analyzed by immunoblot using an anti-V5 antibody. (B). HeLa cell lysates transfected with V5 tagged FOXC2 (V5-FOXC2) were subjected to Ni2+ pull­ down assays with the 6xHis tagged FOXC1 bound to the Ni2+-NTA beads (Ni- FOXC1) or the empty beads (Ni). Bound proteins were analyzed by immunoblot using an anti-V5 antibody.

169 / .0 A,

J, 4?

80KD

58KD—1

«••**«*.

oc-V5

B

/

J1 * 4"

80KD

58KD —

oc-V5

170 Chapter 5. General Discussion and Further Directions

171 FOXC1 and FOXC2 interacting proteins can be identified by the HTM yeast two-hybrid system Several different experimental methods, such as protein affinity chromatography and co-immunoprecipitation, are available to identify or confirm protein-protein interactions. Filamin A was isolated as FOXC1 interacting partner in tissues that are involved in IOP homeostasis from a protein affinity chromatography experiment, in which purified 6xHis tagged FOXC1 were coupled to the Ni2+ agarose columns (77). Co-immunoprecipitation was used to confirm the hypothesis that both transcription factors, FOXC1 and PITX2 can physically interact with each other, providing an explanation why mutations in both genes cause the same condition (78). Screening a cDNA library using a yeast two-hybrid system is another method to identify novel protein-protein interactions. The yeast two-hybrid system is an in vivo genetic system that uses transcriptional activity as a measure of protein- protein interactions occurring in living yeast cells. It has several advantages that make it very useful for analysis of protein-protein interactions. It is highly sensitive since the genetic reporter gene strategy results in a significant amplification and thus it can do a better job in detecting weak or transient interactions. Another feature of this system is that the interactions are detected within yeast cells, an in vivo and eukaryotic environment. The yeast two-hybrid cDNA library screening is a high throughput method for identifying interactions between many proteins with relative minimal requirements to initiate a screening comparing with other biochemical approaches where high quantities of purified proteins are needed (184, 185). However, some concerns apply to Y2H screening for protein-protein interactions. One of them is the use of overexpressed fusion proteins. The fusion may alter the conformation of the prey and/or bait protein(s); while the overexpression may lead to unspecific interactions. Some protein posttranslational modifications such as phosphorylation that might critic for the interactions do not or inappropriately occur in yeast, which will lead to false results. Finally, although some proteins are never present in the same cell at the same time, they may interact when co-expressed in the yeast. The possibility of

172 giving false results means that all interactions identified from Y2H screen should be confirmed by other assays, such as co-immunoprecipitation and pull-down assays (184). The ProQuest yeast two-hybrid system I used was purchased from Invitrogen, which takes advantage of the yeast transcription factor, GAL4. GAL4 has a DNA binding domain (DBD) and a transcriptional activation domain (AD). The coding sequence of protein of interest such as FOXC1 and FOXC2 in my research is fused with the coding sequence of GAL4DBD. The resulting fusion protein will bind to the GAL4 DNA binding sites which are upstream of the reporter genes in the yeast nucleus and functions as bait to trap interacting proteins. The prey proteins to be screened are encoded by a cDNA library fused with GAL4AD sequence. The function of GAL4 is reconstituted, if the bait and prey proteins interact and bring the DBD and AD together in close proximity. The functional GAL4 then activates the expression of the reporter genes that can be detected by selective assays. In the ProQuest Y2H system, vectors used to express DBD fusion proteins and AD fusion proteins contain ARS/CEN element which can maintain protein expression level at a relatively low level so as to reduce the nonspecific interaction and false positive during Y2H screening. Therefore, the positive results in the ProQuest system are more likely true interactors, while some real interactions may be missed. Since both FOXC transcription factors are expressed in the trabecular meshwork (TM) in the eye, a cDNA library built from TM cells is a good resource for isolating biologically relevant FOXC-interacting proteins. mRNA was extracted from a primary human trabecular meshwork cell lines and the GAL4 AD fused cDNA library was custom built also by Invitrogen. Using full length FOXC1 or FOXC2 as bait to screen the HTM Y2H cDNA library, p32 and PIAS3 were repeatedly isolated as FOXC1- and FOXC2- interacting protein, respectively, suggesting they are true interacting partners for FOXC1 and FOXC2. However there is still a possibility of generating false positive results from the screening due to the effects such as overexpression of fusion proteins, incorrect protein folding and posttranslational modification in the

173 yeast, even though the ProQuest system gives a lot of effort to reduce false positives. Due to this reason, the FOXCl/p32 and FOXC2/PIAS3 interactions were further confirmed to be true interactions by other assays such as co- immunoprecipitation and Ni + agarose pull-downs. In addition to being used to identifying protein partners of FOXC proteins, the same Y2H system and Y2H cDNA library were also successfully used to detecting proteins that interact with PITX2 (Acharya M and Sharp M unpublished data) and optineurin (Rezaie T unpublished data), all of which have important functions in the eye. Taken together, the HTM Y2H cDNA library is a valuable tool for identification of interacting partners for proteins with ocular functions.

Protein-protein interactions can regulate the transactivity of FOXC transcription factors The FOXC transcription factors, FOXC1 and FOXC2, regulates gene expressions that give rise to and characterize a wide variety of biological processes including cell differentiation and proliferation, cell fate specification and organogenesis. It is well established that interactions of transcription factors with other proteins can affect the transcription factors' stability, cellular localization as well as post-translational modifications of transcription factors so as to control transcriptional activity to an appropriate level. In addition, protein- protein interactions are also involved in recruitment of transcriptional co- regulators, which can promote or repress transcriptional activation. Transcriptional co-regulators are proteins that complex with transcriptional factors and regulate transcriptional activation through various mechanisms, such as helping with the assembly of transcriptional complex, communicating with the general transcription machinery and altering chromatin structure so as to change chromatin accessibility. Therefore, the interactions of transcription factors with other proteins are recognized as important components involved in conversion of cellular signaling inputs into the transcriptional regulation of target genes. Therefore, looking for interacting proteins of a transcription factor by various biochemical or genetic techniques has become a very important part of the studies

174 that try to understand the function of the transcription factor and how its activities are being regulated in certain biological processes. In my experiment, I have discovered p32 and PIAS3 as FOXC1 and FOXC2 interacting proteins in ocular cells, respectively and also found that the two FOXC proteins directly interact to form a protein complex. p32 was isolated as FOXC1 interacting protein from Y2H HTM cDNA library screenings. This interaction requires the forkhead DNA binding domain of FOXC 1. In transactivation assays, p32 has an inhibiting effect on FOXC1 transactivity. However, the incapability of the mutant FOXC1 Fl 12S, which is deficient in transactivation, to interact with p32 indicates that p32 may not be just a transcriptional co-repressor for FOXC1. Instead, p32 may be involved in both positive and negative regulation of FOXC 1 activity. It is possible that p32 functions as an anchor protein for the recruitment of either transcription co-repressors or co-activators onto FOXC1. Using FOXC2 as bait to screen the same cDNA library identified PIAS3 as FOXC2 interacting partner. Surprisingly, PIAS3 is able to induce FOXC2-mediated luciferase gene expression in transactivation assays and this induction depends on the FOXC- binding site on the promoter of the luciferase gene. Therefore, these data suggest that PIAS3 can up-regulate the activity of endogenous transcription factors, which may include FOXC2. Knocking down FOXC2 results in a decreased PIAS3- induced expression of the reporter gene, supporting the idea that PIAS3 can activate endogenous FOXC2. PIAS3 may help with the recruitment of transcription co-activators such as CBP/p300 onto FOXC2 or be involved in posttranslational modification of FOXC2 as so to up-regulate the transactivity of FOXC2. Interestingly, from Ni2+ pull-down assays, I found that the two FOXC proteins can interact with each other, which likely is involved in regulating the total FOXC transcription activity. FOXC1 and FOXC2 have almost identical forkhead domains and both play important roles in the development of several organs such as eye, heart and vessels, bone and kidney. FOXC1 and FOXC2 have been suggested to be the converging points of different signaling pathways on the same gene (53). My discovery of a direct interaction between FOXC1 and FOXC2 places FOXC1 and FOXC2 in a common molecular setting which is

175 consistent with a role for both proteins in the activation of common target genes. This finding further connects the two transcription factors at the level of protein- protein interaction and provides a novel mechanism by which the two transcription factors regulate common target genes. Several proteins interacting with FOXC transcription factors in ocular cells or in vascular epithelial cells have been identified and their effect on regulating FOXC1 or FOXC2 activity has also been determined. The actin-binding protein filamin A (FLNA) interacts with FOXC1 in the cell nucleus and inhibits FOXC1 activity through retargeting FOXC1 to an HP la, heterochromatin-rich region of nucleus and promoting an inhibiting interaction with PBXla. These findings provide a novel mechanism by which FOXC1 activity responds to the cellular mechanical forces through the influence of structural proteins such as FLNA (77). Interestingly, the two AR malformation associated transcription factors, FOXC 1 and PITX2, interact with each other, tying the two transcription factors in a common pathway. Moreover, PITX2 can negatively regulate the transcriptional activity of FOXC 1 and therefore provides a mechanism of how PITX2 deletions and mutations can lead to increased FOXC1 activity (78). SMAD proteins have been suggested as common binding partners of FOX proteins. Foxcl and Foxc2 have been showed to be able to interact with Smad3 and Smad4. Particularly, the interaction between adjacent Foxc2 and Smad3/4 on the promoter of Pai-1 synergistically induces the gene expression in the arterial epithelial cells, revealing the mechanism by which Foxc2 mediates TGF(3 signaling in regulating Pai-1 expression in epithelial cells (121). Foxc proteins mediate VEGF-Notch signaling-induced expression of DU4 and Hey2, two elements involved in Notch signaling pathway and playing a role in arterial fate specification. Notably, by forming a transcription activation complex containing Foxc2, Su(H) and the Notch intracellular domain (NICD), Hey2 expression is further induced in the epithelial cells (126). A summary of the interacting proteins of FOXC 1 or FOXC2 discovered from my studies and previous studies is show in Figure 5-1. The interactions between FOXC transcription factors and their interacting proteins, as discussed above, were individually confirmed and their effects on the

176 transactivity of F0XC1 or FOXC2 were also individually examined. However, it is possible that large protein complexes containing some, all of these proteins exist in the cells. Therefore as a result of forming protein complexes, these proteins might together regulate the FOXCl/FOXC2-mediated gene expression. Furthermore, biological events might enhance or weaken these interactions and as a result regulate the activities of FOXC1/ FOXC2. For example, mechanical stress may repress FOXC1 activity through translocation of FLNA into the nucleus, where it interacts with FOXC1. Therefore, knowing the dynamics of the interactions between FOXC1/FOXC2 and their interacting proteins will help us understand the function of FOXC1/FOXC2. Further investigations are necessary to investigate this possibility.

Signaling pathways in the eye associated with FOXC transcription factors and their interacting proteins Foxcl and Foxc2 are co-expressed in the periocular mesenchyme giving rise to the anterior segment. Mouse models deficient for either gene display similar phenotypes of anterior segment dysgenesis. The identity of interacting partners of FOXC transcription factors is important to understand their normal function. As FOXC transcription factors and their interacting proteins have been shown to respond to distinct types of extracellular signals, interactions between them may coordinate these functionally important stimuli. Signals from the TGF(3 superfamily are such important stimuli, playing key roles in cell proliferation, differentiation, migration, extracellular matrix remodeling and apoptosis. The TGFp superfamily includes TGFp and BMP/GDF (bone morphogenetic protein/growth and differentiation factor) sub-families. On activation, the secreted TGFp/BMP molecules bind to their membrane bound ligands, which will subsequently activate cytoplasmic SMAD proteins. Activated SMAD proteins enter the nucleus where they associate with transcription factors and regulates target gene expression (105, 186). Members of the TGFp superfamily are important in ocular development. Bmp4 ' mice have severe anterior segment dysgenesis with hypoplastic and compressed trabecular

177 meshwork resulting from reduced number of trabecular beam and less extracellular matrix (187). Tgf|3 molecules are produced in the forming lens and their receptor is expressed in the surrounding periocular mesenchyme and lens as well (75, 188). Overexpression of Tgf/31 in the lens of transgenic mice severely disrupts corneal, anterior chamber and lens development (189). In addition, Tgf/32 ' mice have defects in cornea development (190). Notably, neural crest cell- derived periocular mesenchyme that lacks responsiveness to Tgfp fails to express Foxcl in the forming chamber angle and corneal endothelium (75). In agreement with this finding, Foxcl expression is up-regulated by TGFp in fibroblasts, in ex vivo eye cultures and some cancer cells (75, 191). Interestingly, FOXC1 is a mediator for TGFp 1-induced G0/G1 growth arrest in HeLa cells and TGFpi- induced differentiation of mesenchymal cells in the primodial sternum culture, which may be the consequence of increased expression of FOXC1 induced by TGFpi (48, 191). Moreover, the capability of FOXC 1 to form complexes with SMAD proteins on the promoter of target gene may also contribute to or regulate the FOXC1 ability as mediator for TGFp signaling (121). The above data indicate that the lens-derived TGFp signaling direct the development of surrounding periocular mesenchymal cells into the structures constituting the anterior segment through FOXCL Although it has remained unclear that whether the expression of FOXC2 is induced by TGFp signals, given the similarity of ocular phenotypes of Foxc2 and Tgf/3 mutant mice and the ability of FOXC2 to interact with SMAD proteins, it is likely that FOXC2 and TGFp family members also act in the same pathway. In addition to the TGFp signaling, epidermal growth factor (EGF)-like growth factors, such as TGFa and EGF, direct the development of anterior segment formation, which may be through a mechanism involving FOXCL It is known that TGFa and EGF bind to their cell membrane-bound receptors, EGFRs, which contain tyrosine kinase activity to exert their biological function. Upon ligand binding, the tyrosine kinase activity of EGFR is activated, which thereafter leads to a cascade of phosphorylation events including the phosphorylation and

178 activation of the MEKs (MAPK/ERK Kinases) and ERKs. FOXC1 is a downstream effecter of the MAPK Kinases. Activation of the ERK1/2 mitogen- activated protein kinase through EGF treatment can phosphorylate the serine residue of FOXC1 at amino acid 272, which is required for the stability and maximal FOXC1 transcriptional activation (46). Either the overexpression of TGFa or EGF in the lens of the transgenic mice (192) or the loss of TGFa expression in Tgfd' mice can result in anterior segment dysgenesis (193), which is a reminiscent of the ocular phenotype in Foxcl''' mice. Interestingly, p32 is also a substrate of activated ERK1/2 and ERK activation is required for the translocation of p32 into cell nucleus (194). Since p32 interacts with FOXC1, its translocation to the nucleus upon the activation of the MAP kinase might contribute to the MAP kinase-induced alteration of FOXC1 stability and activity. Therefore, p32, together with FOXC1, are parts of the MAP kinase cascade and their interaction may contribute to the EGF-like growth factors-mediated ocular development. Additionally, since FOXC1 is also a mediator of TGFp signal as discussed above, the interaction between FOXC1 and p32 may also be a convergent point for EGF signals and TGFp signals during ocular development. PIAS3 has been identified as an interacting protein and regulator of several transcription factors. Its role in repressing STAT3 transcription activity has been well established. The STAT signaling pathway is activated in response to cytokines, such as Interleukin-6 (IL-6) and it may play a role in HTM cells in response to mechanical stress or oxidative stress, since mechanical stress and oxidative stress are able to induce the expression of several cytokines including IL-6 (195, 196). Study of the interplays mediated by PIAS3 between STAT and other transcription factors has been of great interest. For example, MITF and STAT3 are two transcription factors that play a key role in the regulation of proliferation and apoptosis in mast cells and melanocytes. PIAS3 is an inhibitor for both proteins. Upon IL-6 or stem cell factor (SCF) activation, MITF is phosphorylated leading to the release of PIAS3, which then potentially can bind to STAT3. Such shuttling of PIAS3 from MITF to STAT3 provides a mechanism for the regulation of the activity of MITF and STAT3 in response to IL-6 or SCF

179 signals (197). Growth factor independent 1 (Gfi-1) is a dominant oncogene in the process of lymphomagenesis. The association of Gfi-1 and PIAS3 relieves the inhibition of STAT3 by PIAS3 and subsequently enhances STAT3-mediated transcription activity in T cells, which provides an explanation of the oncogenic potential of Gfi-1 (198). Therefore, examination of the molecular interactions between STAT3, PIAS3 and FOXC2 in HTM cells under circumstance of mechanical or oxidative stress, which contributes to the development of glaucoma, would be of great interest and would be helpful to understand the role of FOXC2 in response to stress in the ocular tissues. My results clearly support the hypothesis that FOXC proteins require protein-protein interactions to function. In addition, the interactions between the FOXC proteins and their interacting partners could represent crosstalk points that allow coordinated transcriptional responses to extracellular signals. Findings from my work contribute to the long term effort to understanding the molecular function and regulation of FOXC transcription factors. Together with findings from previous studies, my work demonstrates that FOXC transcription factors do interact with other proteins and these protein-protein interactions are involved in the up-regulation or down-regulation of the activity of FOXC transcription factors via various mechanisms. These studies of FOXC interacting proteins are revealing the complex protein networks associated with FOXC transcription factors, which will definitely lead us to a better understanding of FOXC-mediated gene expression and the role of FOXC transcription factors in disease.

Future Direction

Further characterization of the FOXC2/PIAS3 and FOXC1/FOXC2 interaction As discussed previously, more work need to be done to confirm the idea that PIAS3 can enhance the transcription activity of both ectopically expressed and the endogenously expressed FOXC2. Using GAL4 driven luciferase reporter gene

180 and GAL4DBD fused F0XC2 in the transactivation assays would avoid using the luciferase reporter containing FOXC-binding site and could provide evidence for the positive modulation of FOXC2 by PIAS3. Optimization of the conditions for siRNA transfection and repetition of the experiment shown in Figure 3-8 will give more confidence that PIAS3 can activate endogenous FOXC2. In addition, the idea that PIAS3 activates endogenous FOXC2 can be further supported by detection the interaction between endogenous PIAS3 and FOXC2 and visualization of the colocalization in the same cells, which would require high- quality antibodies specific to the proteins. PIAS3 contains several evolutionary conserved domains or motifs. More than 20 transcription factors have been suggested to interact with PIAS3. The regions of PIAS3 that are involved in protein-protein interactions have been identified in many studies, but different regions of PIAS3 seem to be involved in different protein-protein interactions. An N-terminal region between the SAP domain and SP-RTNG domain of PIAS3 is responsible for the interaction with STAT3 and MITF (199). A broader N-terminal region of PIAS3 covering the SAP, PINIT and SP-RING domains interacts with the p65 subunit of nuclear factor-KB

(NF-KB), while the LXXLL motif within the SAP domain is required for the repression effect of PIAS3 on p65 (171). The SP-RING domain can interact with both SMAD3 and CBP (174, 197), whereas the C-terminal region of PIAS3 can bind to Gfi-1 and Interferon regulatory factor-1 (IRF-1) (121). Like PIAS3, FOXC2 also comprise a number of functional domains, including the DNA- binding forkhead domain, the transcription activation and transcription inhibitory domains (200). My Y2H cDNA library screenings for FOXC2 interacting protein have isolated two positive cDNA clones encoding the full-length PIAS3 protein, suggesting that the N-terminal region or the intact 3-dimensional structure of PIAS3 is necessary for its interaction with FOXC2. Further Y2H experiments or in vitro pull-down assays using deletion mutants of the two proteins are needed to map the protein-protein interaction domains for both FOXC2 and PIAS3. These findings will further characterize the FOXC2/PIAS3 interaction and may give a clue to the functional consequence of this interaction.

181 Some evidence has shown that PIAS3 exhibit differential activities towards its interacting partner in a cell type, promoter context dependent manner (158, 200). PIAS3 inhibits androgen receptor-mediated transactivation of a luciferase reporter gene driven by two copies of an androgen response element in prostate cancer cells LNCaP, whereas it induces androgen receptor-medicated transactivation of a luciferase reporter gene under control of the androgen responsive promoter of prostate specific antigen in the same type of cells (158, 200). Although PIAS3 showed a significant degree of colocalization with POU class 5 homeobox 1 gene product (Oct4) across all three tested cell types, it is only capable of relocating Oct4 to the nuclear periphery in C2C12 cell context (148). All the experiments I did so far to characterize the FOXC2/PIAS3 interaction were performed in HeLa cells. It would be interesting to do these experiments in other cell lines, especially in HTM cells, to determine whether there is a cell type difference in regard to the effect of PIAS3 towards FOXC2. I have subcloned PIAS3 cDNA into 3 different mammalian expression vectors, pcDNA4 (Invitrogen), pcDNA3.1nV5/DEST (Invitrogen) and pCMVFLAG (Sigma). However, none of them was successfully transfected into our ocular cell lines in a sufficient amount, indicating that ocular cell lines may be sensitive to the overexpression of PIAS3. Utilization of other type of transfection reagents, optimization of the amount of DNA and reagents used, or creation of an ocular cell line stably expressing PIAS3 may solve this problem. In Ni2+ pull-down assays, I have shown that FOXC1 and FOXC2 can interact with each other. Confirmation of FOXC1/FOXC2 interaction in the cells by co- immunoprecipitation analyses could also be conducted to further confirm this interaction. Both FOXC transcription factors have several functional domains including FHD DNA binding, transcriptional activation and transcriptional inhibition domains (38, 121), all of which can mediate protein-protein interactions. For example, p32 binds to FOXC1 through the FOXC1 FHD while the FOXC1 C- terminal activation domain is required for FOXCl's interaction with PITX2 (78). Further experiments using deletion mutants of the two proteins are needed to map the protein-protein interaction domains for both FOXC1 and FOXC2.

182 I have showed that FOXC1 and FOXC2 can bind to the promoter of FGF19 and FOXOIA in vivo and induce expression of luciferase reporter gene under the control of these two promoters in the luciferase assays. Interaction between FOXC1 and FOXC2 can regulate their transcription activity in luciferase assays. Detection of the expression level of FGF'19 and FOXOIA after siRNA knocking down of FOXC1 or/and FOXC2 would provide additional evidence at the endogenous protein level for my discovery that FOXC1 and FOXC2 can induce the expression of FGF19 and FOXOIA.

Identification of additional FOXC1 or FOXC2 interacting proteins Additional candidate proteins that may interact with FOXC2 are other PIAS proteins, because functional redundancy among mammalian PIAS proteins has been suggested from previous studies. In the mouse, deletion of only one Pias gene at a time only results in mild phenotypes (201-204). In addition, many molecular analyses have shown that different PIAS proteins can interact with and regulate the same transcription factor (156, 166). Moreover, there is evidence that PIAS proteins may associate with each other and thus influence each other's function (205). Experiments that examine interactions between FOXC2 and other PIAS proteins in addition to PIAS3 can be conducted. It would also be interesting to determine their overall effect on FOXC2-mediated transcription in a cell line that co-expresses these proteins. In my Y2H experiments, I only screened 2x10e5 of clones using FOXC2 as the bait. A screen of lxl0e6 of co-transformants would be needed to reach the level of screening saturation (184). Further Y2H screening of the HTM library using FOXC2 may lead to identification of additional FOXC2 interacting proteins. Although a screening saturation has been reached with FOXC1 as the bait, no other proteins have been identified as FOXC-interacting partner except p32. In order to identify new FOXC 1-interacting proteins, a mutant FOXC1 without the FHD, which is required for the interaction with p32, could be used in the Y2H screens.

183 Figure 5-1: Summary of the known interacting proteins of FOXC1 or FOXC2

This figure is a schematic illustration of the FOXC1 or FOXC2-interacting proteins described in the accompanying discussion. Lines with arrowheads represent activation relationships, while inhibitory relationships are represented as lines with blunt arrowheads. Interacting proteins that were identified from my studies are in black, while interacting proteins that were discovered from previous studies are in grey color.

184 SMAD3/4 (Fujita, H. et al. 2006) PIAS3

PITX2 Su(H) (Hayashi, H. et al. 2008) (Berry, FB. et al. 2006)

FLNA/PBX1a (Berry, FB. et al. 2005)

Target genes (FGF19IFOX01A)

185 Bibliography

1. Weigel, D., Jurgens, G., Kuttner, F., Seifert, E. and Jackie, H. (1989) The homeotic gene fork head encodes a nuclear protein and is expressed in the terminal regions of the Drosophila embryo. Cell, 57, 645-58. 2. Myatt, S.S. and Lam, E.W. (2007) The emerging roles of forkhead box (Fox) proteins in cancer. Nat Rev Cancer, 7, 847-59. 3. Carlsson, P. and Mahlapuu, M. (2002) Forkhead transcription factors: key players in development and metabolism. Developmental biology, 250, 1-23. 4. Kaufmann, E. and Knochel, W. (1996) Five years on the wings of fork head. Mechanisms of development, 57, 3-20. 5. Clark, K.L., Halay, E.D., Lai, E. and Burley, S.K. (1993) Co-crystal structure of the FTNF-3/fork head DNA-recognition motif resembles histone H5. Nature, 364, 412-20. 6. Jin, C, Marsden, I., Chen, X. and Liao, X. (1999) Dynamic DNA contacts observed in the NMR structure of winged helix protein-DNA complex. Journal of molecular biology, 289, 683-90. 7. van Dongen, M.J., Cederberg, A., Carlsson, P., Enerback, S. and Wikstrom, M. (2000) Solution structure and dynamics of the DNA-binding domain of the adipocyte-transcription factor FREAC-11. Journal of molecular biology, 296, 351-9. 8. Weigelt, J., Climent, I., Dahlman-Wright, K. and Wikstrom, M. (2001) Solution structure of the DNA binding domain of the human forkhead transcription factor AFX (FOX04). Biochemistry, 40, 5861-9. 9. Pierrou, S., Hellqvist, M., Samuelsson, L., Enerback, S. and Carlsson, P. (1994) Cloning and characterization of seven human forkhead proteins: binding site specificity and DNA bending. The EMBO journal, 13, 5002-12. 10. Overdier, D.G., Porcella, A. and Costa, R.H. (1994) The DNA-binding specificity of the hepatocyte nuclear factor 3/forkhead domain is influenced by amino-acid residues adjacent to the recognition helix. Molecular and cellular biology, 14, 2755-66.

186 11. Marsden, I., Jin, C. and Liao, X. (1998) Structural changes in the region directly adjacent to the DNA-binding helix highlight a possible mechanism to explain the observed changes in the sequence-specific binding of winged helix proteins. Journal of molecular biology, 278, 293-9. 12. Hjalt, T.A. and Semina, E.V. (2005) Current molecular understanding of Axenfeld-Rieger syndrome. Expert reviews in molecular medicine, 7, 1-17. 13. Shields, M.B. (1983) Axenfeld-Rieger syndrome: a theory of mechanism and distinctions from the iridocorneal endothelial syndrome. Transactions of the American Ophthalmological Society, 81, 736-84. 14. Fitch, N. and Kaback, M. (1978) The Axenfeld syndrome and the Rieger syndrome. Journal of medical genetics, 15, 30-4. 15. Lines, M.A., Kozlowski, K. and Walter, M.A. (2002) Molecular genetics of Axenfeld-Rieger malformations. Human molecular genetics, 11,1177-84. 16. Sadeghi-Nejad, A. and Senior, B. (1974) Autosomal dominant transmission of isolated growth hormone deficiency in iris-dental dysplasia (Rieger's syndrome). The Journal of pediatrics, 85, 644-8. 17. Cunningham, E.T., Jr., Eliott, D., Miller, N.R., Maumenee, I.H. and Green, W.R. (1998) Familial Axenfeld-Rieger anomaly, , and sensorineural hearing loss: a possible new genetic syndrome. Archives of ophthalmology, 116,78-82. 18. Moog, U., Bleeker-Wagemakers, E.M., Crobach, P., Vies, J.S. and Schrander-Stumpel, C.T. (1998) Sibs with Axenfeld-Rieger anomaly, hydrocephalus, and leptomeningeal calcifications: a new autosomal recessive syndrome? American journal of medical genetics, 78, 263-6. 19. Bekir, N.A. and Gungor, K. (2000) Atrial septal defect with interatrial aneurysm and Axenfeld-Rieger syndrome. Acta ophthalmologica Scandinavica, 78, 101-3. 20. Vaux, C, Sheffield, L., Keith, C.G. and Voullaire, L. (1992) Evidence that Rieger syndrome maps to 4q25 or 4q27. Journal of medical genetics, 29, 256-8. 21. Heon, E., Sheth, B.P., Kalenak, J.W., Sunden, S.L., Streb, L.M., Taylor, CM., Alward, W.L., Sheffield, V.C. and Stone, E.M. (1995) Linkage of

187 autosomal dominant iris hypoplasia to the region of the Rieger syndrome locus (4q25). Human molecular genetics, 4, 1435-9. 22. Datson, N.A., Semina, E., van Staalduinen, A.A., Dauwerse, H.G., Meershoek, E.J., Heus, J.J., Frants, R.R., den Dunnen, J.T., Murray, J.C. and van Ommen, G.J. (1996) Closing in on the Rieger syndrome gene on 4q25: mapping translocation breakpoints within a 50-kb region. American journal of human genetics, 59, 1297-305. 23. Gould, D.B., Mears, A.J., Pearce, W.G. and Walter, M.A. (1997) Autosomal dominant Axenfeld-Rieger anomaly maps to 6p25. American journal of human genetics, 61, 765-8. 24. Mirzayans, F., Mears, A.J., Guo, S.W., Pearce, W.G. and Walter, M.A. (1997) Identification of the human chromosomal region containing the iridogoniodysgenesis anomaly locus by genomic-mismatch scanning. American journal of human genetics, 61, 111-9. 25. Phillips, J.C, del Bono, E.A., Haines, J.L., Pralea, A.M., Cohen, J.S., Greff, L.J. and Wiggs, J.L. (1996) A second locus for Rieger syndrome maps to chromosome 13ql4. American journal of human genetics, 59, 613-9. 26. Semina, E.V., Reiter, R., Leysens, N.J., Alward, W.L., Small, K.W., Datson, N.A., Siegel-Bartelt, J., Bierke-Nelson, D., Bitoun, P., Zabel, B.U. et al. (1996) Cloning and characterization of a novel bicoid-related homeobox transcription factor gene, RIEG, involved in Rieger syndrome. Nature genetics, 14, 392-9. 27. Palmer, C.G., Bader, P., Slovak, M.L., Comings, D.E. and Pettenati, M.J. (1991) Partial deletion of chromosome 6p: delineation of the syndrome. American journal of medical genetics, 39, 155-60. 28. Plaja, A., Vidal, R., Soriano, D., Bou, X., Vendrell, T., Mediano, C, Pueyo, J.M., Labrana, X. and Sarret, E. (1994) Terminal deletion of 6p: report of a new case. Annales de genetique, 37, 196-9. 29. Zurcher, V.L., Golden, W.L. and Zinn, A.B. (1990) Distal deletion of the short arm of chromosome 6. American journal of medical genetics, 35, 261-5.

188 30. Larsson, C, Hellqvist, M., Pierrou, S., White, I., Enerback, S. and Carlsson, P. (1995) Chromosomal localization of six human forkhead genes, freac-1 (FKHL5), -3 (FKHL7), -4 (FKHL8), -5 (FKHL9), -6 (FKHL10), and -8 (FKHL12). Genomics, 30, 464-9. 31. Mears, A.J., Mirzayans, F., Gould, D.B., Pearce, W.G. and Walter, M.A. (1996) Autosomal dominant iridogoniodysgenesis anomaly maps to 6p25. American journal of human genetics, 59, 1321-7. 32. Mears, A.J., Jordan, T., Mirzayans, F., Dubois, S., Kume, T., Parlee, M., Ritch, R., Koop, B., Kuo, W.L., Collins, C. et al. (1998) Mutations of the forkhead/winged-helix gene, FKHL7, in patients with Axenfeld-Rieger anomaly. American journal of human genetics, 63, 1316-28. 33. Nishimura, D.Y., Swiderski, R.E., Alward, W.L., Searby, C.C., Patil, S.R., Bennet, S.R., Kanis, A.B., Gastier, J.M., Stone, E.M. and Sheffield, V.C. (1998) The forkhead transcription factor gene FKHL7 is responsible for glaucoma phenotypes which map to 6p25. Nature genetics, 19, 140-7. 34. Chakrabarti, S., Kaur, K., Rao, K.N., Mandal, A.K., Kaur, I., Parikh, R.S. and Thomas, R. (2009) The transcription factor gene FOXC1 exhibits a limited role in primary congenital glaucoma. Investigative ophthalmology & visual science, 50, 75-83. 35. Lehmann, O.J., Ebenezer, N.D., Jordan, T., Fox, M., Ocaka, L., Payne, A., Leroy, B.P., Clark, B.J., Hitchings, R.A., Povey, S. et al. (2000) Chromosomal duplication involving the forkhead transcription factor gene FOXC1 causes iris hypoplasia and glaucoma. American journal of human genetics, 67, 1129-35. 36. Chanda, B., Asai-Coakwell, M., Ye, M., Mungall, A.J., Barrow, M., Dobyns, W.B., Behesti, H., Sowden, J.C., Carter, N.P., Walter, M.A. et al. (2008) A novel mechanistic spectrum underlies glaucoma-associated chromosome 6p25 copy number variation. Human molecular genetics, 17, 3446-58. 37. Nishimura, D.Y., Searby, C.C., Alward, W.L., Walton, D., Craig, J.E., Mackey, D.A., Kawase, K., Kanis, A.B., Patil, S.R., Stone, E.M. et al. (2001) A spectrum of FOXC1 mutations suggests gene dosage as a mechanism for

189 developmental defects of the anterior chamber of the eye. American journal of human genetics, 68, 364-72. 38. Berry, F.B., Saleem, R.A. and Walter, M.A. (2002) FOXC1 transcriptional regulation is mediated by N- and C-terminal activation domains and contains a phosphorylated transcriptional inhibitory domain. The Journal of biological chemistry, 111, 10292-7. 39. Saleem, R.A., Murphy, T.C., Liebmann, J.M. and Walter, M.A. (2003) Identification and analysis of a novel mutation in the FOXC1 forkhead domain. Investigative ophthalmology & visual science, 44, 4608-12. 40. Murphy, T.C., Saleem, R.A., Footz, T., Ritch, R., McGillivray, B. and Walter, M.A. (2004) The wing 2 region of the FOXC1 forkhead domain is necessary for normal DNA-binding and transactivation functions. Investigative ophthalmology & visual science, 45, 2531-8. 41. Saleem, R.A., Banerjee-Basu, S., Berry, F.B., Baxevanis, A.D. and Walter, M.A. (2001) Analyses of the effects that disease-causing missense mutations have on the structure and function of the winged-helix protein FOXC1. American journal of human genetics, 68, 627-41. 42. Saleem, R.A., Banerjee-Basu, S., Berry, F.B., Baxevanis, A.D. and Walter, M.A. (2003) Structural and functional analyses of disease-causing missense mutations in the forkhead domain of FOXC1. Human molecular genetics, 12, 2993-3005. 43. Ito, Y.A., Footz, T.K., Murphy, T.C., Courtens, W. and Walter, M.A. (2007) Analyses of a novel L130F missense mutation in FOXC1. Archives of ophthalmology, 125, 128-35. 44. Saleem, R.A., Banerjee-Basu, S., Murphy, T.C., Baxevanis, A. and Walter, M.A. (2004) Essential structural and functional determinants within the forkhead domain of FOXC1. Nucleic acids research, 32, 4182-93. 45. Berry, F.B., Tamimi, Y., Carle, M.V., Lehmann, O.J. and Walter, M.A. (2005) The establishment of a predictive mutational model of the forkhead domain through the analyses of FOXC2 missense mutations identified in patients

190 with hereditary lymphedema with distichiasis. Human molecular genetics, 14, 2619-27. 46. Berry, F.B., Mirzayans, F. and Walter, M.A. (2006) Regulation of FOXC1 stability and transcriptional activity by an epidermal growth factor-activated mitogen-activated protein kinase signaling cascade. The Journal of biological chemistry, 281, 10098-104. 47. Wang, W.H., McNatt, L.G., Shepard, A.R., Jacobson, N., Nishimura, D.Y., Stone, E.M., Sheffield, V.C. and Clark, A.F. (2001) Optimal procedure for extracting RNA from human ocular tissues and expression profiling of the congenital glaucoma gene FOXC1 using quantitative RT-PCR. Molecular vision, 7, 89-94. 48. Kume, T., Deng, K.Y., Winfrey, V., Gould, D.B., Walter, M.A. and Hogan, B.L. (1998) The forkhead/winged helix gene Mfl is disrupted in the pleiotropic mouse mutation congenital hydrocephalus. Cell, 93, 985-96. 49. Hiemisch, H., Monaghan, A.P., Schutz, G. and Kaestner, K.H. (1998) Expression of the mouse Fkhl/Mfl and Mfhl genes in late gestation embryos is restricted to mesoderm derivatives. Mechanisms of development, 73, 129-32. 50. Sasaki, H. and Hogan, B.L. (1993) Differential expression of multiple fork head related genes during gastrulation and axial pattern formation in the mouse embryo. Development (Cambridge, England), 118, 47-59. 51. Kume, T., Jiang, H., Topczewska, J.M. and Hogan, B.L. (2001) The murine winged helix transcription factors, Foxcl and Foxc2, are both required for cardiovascular development and somitogenesis. Genes & development, 15, 2470- 82. 52. Hong, H.K., Lass, J.H. and Chakravarti, A. (1999) Pleiotropic skeletal and ocular phenotypes of the mouse mutation congenital hydrocephalus (ch/Mfl) arise from a winged helix/forkhead transcriptionfactor gene. Human molecular genetics, 8, 625-37. 53. Winnier, G.E., Kume, T., Deng, K., Rogers, R., Bundy, J., Raines, C., Walter, M.A., Hogan, B.L. and Conway, S.J. (1999) Roles for the winged helix transcription factors MF1 and MFH1 in cardiovascular development revealed by

191 nonallelic noncomplementation of null alleles. Developmental biology, 213, 418- 31. 54. Swiderski, R.E., Reiter, R.S., Nishimura, D.Y., Alward, W.L., Kalenak, J.W., Searby, C.S., Stone, E.M., Sheffield, V.C. and Lin, J.J. (1999) Expression of the Mfl gene in developing mouse hearts: implication in the development of human congenital heart defects. Dev Dyn, 216, 16-27. 55. Kume, T., Deng, K. and Hogan, B.L. (2000) Murine forkhead/winged helix genes Foxcl (Mfl) and Foxc2 (Mfhl) are required for the early organogenesis of the kidney and urinary tract. Development (Cambridge, England), 127, 1387-95. 56. Mattiske, D., Kume, T. and Hogan, B.L. (2006) The mouse forkhead gene Foxcl is required for primordial germ cell migration and antral follicle development. Developmental biology, 290, 447-58. 57. Green, M.C. (1970) The developmental effects of congenital hydrocephalus (ch) in the mouse. Developmental biology, 23, 585-608. 58. Gage, P.J., Rhoades, W., Prucka, S.K. and Hjalt, T. (2005) Fate maps of neural crest and mesoderm in the mammalian eye. Investigative ophthalmology & visual science, 46, 4200-8. 59. Kidson, S.H., Kume, T., Deng, K., Winfrey, V. and Hogan, B.L. (1999) The forkhead/winged-helix gene, Mfl, is necessary for the normal development of the cornea and formation of the anterior chamber in the mouse eye. Developmental biology, 211, 306-22. 60. Mattiske, D., Sommer, P., Kidson, S.H. and Hogan, B.L. (2006) The role of the forkhead transcription factor, Foxcl, in the development of the mouse lacrimal gland. Dev Dyn, 235,1074-80. 61. Smith, R.S., Zabaleta, A., Kume, T., Savinova, O.V., Kidson, S.H., Martin, J.E., Nishimura, D.Y., Alward, W.L., Hogan, B.L. and John, S.W. (2000) Haploinsufficiency of the transcription factors FOXC1 and FOXC2 results in aberrant ocular development. Human molecular genetics, 9, 1021-32. 62. Topczewska, J.M., Topczewski, J., Solnica-Krezel, L. and Hogan, B.L. (2001) Sequence and expression of zebrafish foxcl a and foxclb, encoding

192 conserved forkhead/winged helix transcription factors. Mechanisms of development, 100, 343-7. 63. Topczewska, J.M., Topczewski, J., Shostak, A., Kume, T., Solnica-Krezel, L. and Hogan, B.L. (2001) The winged helix transcription factor Foxcla is essential for somitogenesis in zebrafish. Genes & development, 15, 2483-93. 64. McMahon, C, Semina, E.V. and Link, B.A. (2004) Using zebrafish to study the complex genetics of glaucoma. Comp Biochem Physiol C Toxicol Pharmacol, 138,343-50. 65. Tamimi, Y., Skarie, J.M., Footz, T., Berry, F.B., Link, B.A. and Walter, M.A. (2006) FGF19 is a target for FOXC1 regulation in ciliary body-derived cells. Human molecular genetics, 15, 3229-40. 66. Jerome, L.A. and Papaioannou, V.E. (2001) DiGeorge syndrome phenotype in mice mutant for the T-box gene, Tbxl. Nature genetics, 27, 286-91. 67. Lindsay, E.A., Vitelli, F., Su, H., Morishima, M., Huynh, T., Pramparo, T., Jurecic, V., Ogunrinu, G., Sutherland, H.F., Scambler, P.J. et al. (2001) Tbxl haploinsufficieny in the DiGeorge syndrome region causes aortic arch defects in mice. Nature, 410, 97-101. 68. Merscher, S., Funke, B., Epstein, J.A., Heyer, J., Puech, A., Lu, M.M., Xavier, R.J., Demay, M.B., Russell, R.G., Factor, S. et al. (2001) TBX1 is responsible for cardiovascular defects in velo-cardio-facial/DiGeorge syndrome. Cell, 104, 619-29. 69. Yamagishi, H., Maeda, J., Hu, T., McAnally, J., Conway, S.J., Kume, T., Meyers, E.N., Yamagishi, C. and Srivastava, D. (2003) Tbxl is regulated by tissue-specific forkhead proteins through a common Sonic hedgehog-responsive enhancer. Genes & development, 17, 269-81. 70. Tamimi, Y., Lines, M., Coca-Prados, M. and Walter, M.A. (2004) Identification of target genes regulated by FOXC1 using nickel agarose-based chromatin enrichment. Investigative ophthalmology & visual science, 45, 3904-13. 71. Berry, F.B., Skarie, J.M., Mirzayans, F., Fortin, Y., Hudson, T.J., Raymond, V., Link, B.A. and Walter, M.A. (2008) FOXC1 is required for cell

193 viability and resistance to oxidative stress in the eye through the transcriptional regulation of FOXOIA. Human molecular genetics, 17, 490-505. 72. Rice, R., Rice, D.P., Olsen, B.R. and Thesleff, I. (2003) Progression of calvarial bone development requires Foxcl regulation of Msx2 and Alx4. Developmental biology, 262, 75-87. 73. Rice, R., Rice, D.P. and Thesleff, I. (2005) Foxcl integrates Fgf and Bmp signalling independently of twist or noggin during calvarial bone development. DevDyn, 233,847-52. 74. Matt, N., Dupe, V., Gamier, J.M., Dennefeld, C, Chambon, P., Mark, M. and Ghyselinck, N.B. (2005) Retinoic acid-dependent eye morphogenesis is orchestrated by neural crest cells. Development (Cambridge, England), 132, 4789- 800. 75. Ittner, L.M., Wurdak, H., Schwerdtfeger, K., Kunz, T., Ille, F., Leveen, P., Hjalt, T.A., Suter, U., Karlsson, S., Hafezi, F. et al. (2005) Compound developmental eye disorders following inactivation of TGFbeta signaling in neural-crest stem cells. Journal of biology, 4, 11. 76. Libby, R.T., Smith, R.S., Savinova, O.V., Zabaleta, A., Martin, J.E., Gonzalez, F.J. and John, S.W. (2003) Modification of ocular defects in mouse developmental glaucoma models by tyrosinase. Science (New York, N. Y, 299, 1578-81. 77. Berry, F.B., O'Neill, M.A., Coca-Prados, M. and Walter, M.A. (2005) FOXC1 transcriptional regulatory activity is impaired by PBX1 in a filamin A- mediated manner. Molecular and cellular biology, 25, 1415-24. 78. Berry, F.B., Lines, M.A., Oas, J.M., Footz, T., Underhill, D.A., Gage, P.J. and Walter, M.A. (2006) Functional interactions between FOXC1 and PITX2 underlie the sensitivity to FOXC1 gene dose in Axenfeld-Rieger syndrome and anterior segment dysgenesis. Human molecular genetics, 15, 905-19. 79. Fang, J., Dagenais, S.L., Erickson, R.P., Arlt, M.F., Glynn, M.W., Gorski, J.L., Seaver, L.H. and Glover, T.W. (2000) Mutations in FOXC2 (MFH-1), a forkhead family transcription factor, are responsible for the hereditary

194 lymphedema-distichiasis syndrome. American journal of human genetics, 67, 1382-8. 80. Erickson, R.P., Hudgins, L., Stone, J.F., Schmidt, S., Wilke, C. and Glover, T.W. (1995) A "balanced" Y;16 translocation associated with Turner-like neonatal lymphedema suggests the location of a potential anti-Turner gene on the Y chromosome. Cytogenetics and cell genetics, 71, 163-7. 81. Mangion, J., Rahman, N., Mansour, S., Brice, G., Rosbotham, J., Child, A.H., Murday, V.A., Mortimer, P.S., Barfoot, R., Sigurdsson, A. et al. (1999) A gene for lymphedema-distichiasis maps to 16q24.3. American journal of human genetics, 65, 427-32. 82. Erickson, R.P., Dagenais, S.L., Caulder, M.S., Downs, C.A., Herman, G., Jones, M.C., Kerstjens-Frederikse, W.S., Lidral, A.C., McDonald, M., Nelson, C.C. et al. (2001) Clinical heterogeneity in lyrnphoedema-distichiasis with FOXC2 truncating mutations. Journal of medical genetics, 38, 761-6. 83. Brice, G., Mansour, S., Bell, R., Collin, J.R., Child, A.H., Brady, A.F., Sarfarazi, M., Burnand, K.G., Jeffery, S., Mortimer, P. et al. (2002) Analysis of the phenotypic abnormalities in lyrnphoedema-distichiasis syndrome in 74 patients with FOXC2 mutations or linkage to 16q24. Journal of medical genetics, 39,478-83. 84. Bahuau, M., Houdayer, C, Tredano, M., Soupre, V., Couderc, R. and Vazquez, M.P. (2002) FOXC2 truncating mutation in distichiasis, lymphedema, and cleft palate. Clinical genetics, 62, 470-3. 85. Finegold, D.N., Kimak, M.A., Lawrence, E.C., Levinson, K.L., Cherniske, E.M., Pober, B.R., Dunlap, J.W. and Ferrell, R.E. (2001) Truncating mutations in FOXC2 cause multiple lymphedema syndromes. Human molecular genetics, 10, 1185-9. 86. Mellor, R.H., Brice, G., Stanton, A.W., French, J., Smith, A., Jeffery, S., Levick, J.R., Burnand, K.G. and Mortimer, P.S. (2007) Mutations in FOXC2 are strongly associated with primary valve failure in veins of the lower limb. Circulation, 115, 1912-20.

195 87. Yildirim-Toruner, C, Subramanian, K., El Manjra, L., Chen, E., Goldstein, S. and Vitale, E. (2004) A novel frameshift mutation of FOXC2 gene in a family with hereditary lymphedema-distichiasis syndrome associated with renal disease and diabetes mellitus. Am J Med Genet A, 131,281-6. 88. Bell, R., Brice, G., Child, A.H., Murday, V.A., Mansour, S., Sandy, C.J., Collin, J.R., Brady, A.F., Callen, D.F., Burnand, K. et al. (2001) Analysis of lymphoedema-distichiasis families for FOXC2 mutations reveals small insertions and deletions throughout the gene. Human genetics, 108, 546-51. 89. Traboulsi, E.I., Al-Khayer, K., Matsumoto, M., Kimak, M.A., Crowe, S., Wilson, S.E., Finegold, D.N., Ferrell, R.E. and Meisler, D.M. (2002) Lymphedema-distichiasis syndrome and FOXC2 gene mutation. American journal of ophthalmology, 134, 592-6. 90. Brooks, B.P., Dagenais, S.L., Nelson, C.C., Glynn, M.W., Caulder, M.S., Downs, C.A. and Glover, T.W. (2003) Mutation of the FOXC2 gene in familial distichiasis. JAapos, 7, 354-7. 91. Sholto-Douglas-Vernon, C, Bell, R., Brice, G., Mansour, S., Sarfarazi, M., Child, A.H., Smith, A., Mellor, R., Burnand, K., Mortimer, P. et al. (2005) Lymphoedema-distichiasis and FOXC2: unreported mutations, de novo mutation estimate, families without coding mutations. Human genetics, 117, 238-42. 92. Lehmann, O.J., Tuft, S., Brice, G., Smith, R., Blixt, A., Bell, R., Johansson, B., Jordan, T., Hitchings, R.A., Khaw, P.T. et al. (2003) Novel anterior segment phenotypes resulting from forkhead gene alterations: evidence for cross-species conservation of function. Investigative ophthalmology & visual science, 44, 2627- 33. 93. Kaestner, K.H., Bleckmann, S.C., Monaghan, A.P., Schlondorff, J., Mincheva, A., Lichter, P. and Schutz, G. (1996) Clustered arrangement of winged helix genes fkh-6 and MFH-1: possible implications for mesoderm development. Development (Cambridge, England), 122, 1751-8. 94. Miura, N., Wanaka, A., Tohyama, M. and Tanaka, K. (1993) MFH-1, a new member of the fork head domain family, is expressed in developing mesenchyme. FEBS letters, 326, 171-6.

196 95. Winnier, G.E., Hargett, L. and Hogan, B.L. (1997) The winged helix transcription factor MFH1 is required for proliferation and patterning of paraxial mesoderm in the mouse embryo. Genes & development, 11, 926-40. 96. Iida, K., Koseki, H., Kakinuma, H., Kato, N., Mizutani-Koseki, Y., Ohuchi, H., Yoshioka, H., Noji, S., Kawamura, K., Kataoka, Y. et al. (1997) Essential roles of the winged helix transcription factor MFH-1 in aortic arch patterning and skeletogenesis. Development (Cambridge, England), 124, 4627-38. 97. Kanzaki-Kato, N., Tamakoshi, T., Fu, Y., Chandra, A., Itakura, T., Uezato, T., Tanaka, T., Clouthier, D.E., Sugiyama, T., Yanagisawa, M. et al. (2005) Roles of forkhead transcription factor Foxc2 (MFH-1) and endothelin receptor A in cardiovascular morphogenesis. Cardiovascular research, 65, 711-8. 98. Furumoto, T.A., Miura, N., Akasaka, T., Mizutani-Koseki, Y., Sudo, H., Fukuda, K., Maekawa, M., Yuasa, S., Fu, Y., Moriya, H. et al (1999) Notochord- dependent expression of MFH1 and PAX1 cooperates to maintain the proliferation of sclerotome cells during the vertebral column development. Developmental biology, 210, 15-29. 99. Dagenais, S.L., Hartsough, R.L., Erickson, R.P., Witte, M.H., Butler, M.G. and Glover, T.W. (2004) Foxc2 is expressed in developing lymphatic vessels and other tissues associated with lymphedema-distichiasis syndrome. Gene Expr Patterns, 4,611-9. 100. Dale, R.F. (1987) Primary lymphoedema when found with distichiasis is of the type defined as bilateral hyperplasia by lymphography. Journal of medical genetics, 24, 170-1. 101. Kriederman, B.M., Myloyde, T.L., Witte, M.H., Dagenais, S.L., Witte, C.L., Rennels, M., Bernas, M.J., Lynch, M.T., Erickson, R.P., Caulder, M.S. et al. (2003) FOXC2 haploinsufficient mice are a model for human autosomal dominant lymphedema-distichiasis syndrome. Human molecular genetics, 12, 1179-85. 102. Petrova, T.V., Karpanen, T., Norrmen, C., Mellor, R., Tamakoshi, T., Finegold, D., Ferrell, R., Kerjaschki, D., Mortimer, P., Yla-Herttuala, S. et al. (2004) Defective valves and abnormal mural cell recruitment underlie lymphatic vascular failure in lymphedema distichiasis. Nature medicine, 10, 974-81.

197 103. Seo, S., Fujita, H., Nakano, A., Kang, M., Duarte, A. and Kume, T. (2006) The forkhead transcription factors, Foxcl and Foxc2, are required for arterial specification and lymphatic sprouting during vascular development. Developmental biology, 294, 458-70. 104. Takemoto, M., He, L., Norlin, J., Patrakka, J., Xiao, Z., Petrova, T., Bondjers, C, Asp, J., Wallgard, E., Sun, Y. et al. (2006) Large-scale identification of genes implicated in kidney glomerulus development and function. The EMBO journal, 25,1160-74. 105. Gould, D.B., Smith, R.S. and John, S.W. (2004) Anterior segment development relevant to glaucoma. The International journal of developmental biology, 48, 1015-29. 106. Pressman, C.L., Chen, H. and Johnson, R.L. (2000) LMX1B, a LIM homeodomain class transcription factor, is necessary for normal development of multiple tissues in the anterior segment of the murine eye. Genesis, 26, 15-25. 107. Cederberg, A., Gronning, L.M., Ahren, B., Tasken, K., Carlsson, P. and Enerback, S. (2001) FOXC2 is a winged helix gene that counteracts obesity, hypertriglyceridemia, and diet-induced insulin resistance. Cell, 106, 563-73. 108. Gronning, L.M., Baillie, G.S., Cederberg, A., Lynch, M.J., Houslay, M.D., Enerback, S. and Tasken, K. (2006) Reduced PDE4 expression and activity contributes to enhanced catecholamine-induced cAMP accumulation in adipocytes from FOXC2 transgenic mice. FEBS letters, 580, 4126-30. 109. Dahle, M.K., Gronning, L.M., Cederberg, A., Blomhoff, H.K., Miura, N., Enerback, S., Tasken, K.A. and Tasken, K. (2002) Mechanisms of FOXC2- and FOXD1-mediated regulation of the RI alpha subunit of cAMP-dependent protein kinase include release of transcriptional repression and activation by protein kinase B alpha and cAMP. The Journal of biological chemistry, 277, 22902-8. 110. Kim, J.K., Kim, H.J., Park, S.Y., Cederberg, A., Westergren, R., Nilsson, D., Higashimori, T., Cho, Y.R., Liu, Z.X., Dong, J. et al. (2005) Adipocyte- specific overexpression of FOXC2 prevents diet-induced increases in intramuscular fatty acyl CoA and insulin resistance. Diabetes, 54, 1657-63.

198 111. Gronning, L.M., Cederberg, A., Miura, N., Enerback, S. and Tasken, K. (2002) Insulin and TNF alpha induce expression of the forkhead transcription factor gene Foxc2 in 3T3-L1 adipocytes via PI3K and ERK 1/2-dependent pathways. Molecular endocrinology (Baltimore, Md, 16, 873-83. 112. Ridderstrale, M., Carlsson, E., Klannemark, M., Cederberg, A., Kosters, C, Tornqvist, H., Storgaard, H., Vaag, A., Enerback, S. and Groop, L. (2002) FOXC2 mRNA Expression and a 5' untranslated region polymorphism of the gene are associated with insulin resistance. Diabetes, 51, 3554-60. 113. Hansen, J.B., Jorgensen, C., Petersen, R.K., Hallenborg, P., De Matteis, R., Boye, H.A., Petrovic, N., Enerback, S., Nedergaard, J., Cinti, S. et al. (2004) Retinoblastoma protein functions as a molecular switch determining white versus brown adipocyte differentiation. Proceedings of the National Academy of Sciences of the United States of America, 101, 4112-7. 114. Toh, S.Y., Gong, J., Du, G., Li, J.Z., Yang, S., Ye, J., Yao, H., Zhang, Y., Xue, B., Li, Q. et al. (2008) Up-regulation of mitochondrial activity and acquirement of brown adipose tissue-like property in the white adipose tissue of fsp27 deficient mice. PLoS ONE, 3, e2890. 115. Yang, X., Enerback, S. and Smith, U. (2003) Reduced expression of FOXC2 and brown adipogenic genes in human subjects with insulin resistance. Obesity research, 11,1182-91. 116. Carlsson, E., Almgren, P., Hoffstedt, J., Groop, L. and Ridderstrale, M. (2004) The FOXC2 C-512T polymorphism is associated with obesity and dyslipidemia. Obesity research, 12, 1738-43. 117. Carlsson, E., Groop, L. and Ridderstrale, M. (2005) Role of the FOXC2 - 512C>T polymorphism in type 2 diabetes: possible association with the dysmetabolic syndrome. International journal of obesity (2005), 29, 268-74. 118. Kovacs, P., Lehn-Stefan, A., Stumvoll, M., Bogardus, C. and Baier, L.J. (2003) Genetic variation in the human winged helix/forkhead transcription factor gene FOXC2 in Pima Indians. Diabetes, 52, 1292-5. 119. Osawa, H., Onuma, H., Murakami, A., Ochi, M., Nishimiya, T., Kato, K., Shimizu, I., Fujii, Y., Ohashi, J. and Makino, H. (2003) Systematic search for

199 single nucleotide polymorphisms in the FOXC2 gene: the absence of evidence for the association of three frequent single nucleotide polymorphisms and four common haplotypes with Japanese type 2 diabetes. Diabetes, 52, 562-7. 120. Mani, S.A., Yang, J., Brooks, M., Schwaninger, G., Zhou, A., Miura, N., Kutok, J.L., Hartwell, K., Richardson, A.L. and Weinberg, R.A. (2007) Mesenchyme Forkhead 1 (FOXC2) plays a key role in metastasis and is associated with aggressive basal-like breast cancers. Proceedings of the National Academy of Sciences of the United States of America, 104, 10069-74. 121. Fujita, H., Kang, M., Eren, M., Gleaves, L.A., Vaughan, D.E. and Kume, T. (2006) Foxc2 is a common mediator of insulin and transforming growth factor beta signaling to regulate plasminogen activator inhibitor type I gene expression. Circulation research, 98, 626-34. 122. Hayashi, H., Sano, H., Seo, S. and Kume, T. (2008) The Foxc2 transcription factor regulates angiogenesis via induction of integrin beta3 expression. The Journal of biological chemistry, 283, 23791-800. 123. Xue, Y., Cao, R., Nilsson, D., Chen, S., Westergren, R., Hedlund, E.M., Martijn, C, Rondahl, L., Krauli, P., Walum, E. et al. (2008) FOXC2 controls Ang-2 expression and modulates angiogenesis, vascular patterning, remodeling, and functions in adipose tissue. Proceedings of the National Academy of Sciences of the United States of America, 105, 10167-72. 124. Seo, S. and Kume, T. (2006) Forkhead transcription factors, Foxcl and Foxc2, are required for the morphogenesis of the cardiac outflow tract. Developmental biology, 296, 421-36. 125. Gitler, A.D., Lu, M.M., Jiang, Y.Q., Epstein, J.A. and Gruber, P.J. (2003) Molecular markers of cardiac endocardial cushion development. Dev Dyn, 228, 643-50. 126. Hayashi, H. and Kume, T. (2008) Foxc transcription factors directly regulate D114 and Hey2 expression by interacting with the VEGF-Notch signaling pathways in endothelial cells. PLoS ONE, 3, e2401.

200 127. Connell, F., Brice, G. and Mortimer, P. (2008) Phenotypic characterization of primary lymphedema. Annals of the New York Academy of Sciences, 1131, 140-6. 128. Lehmann, O.J., Sowden, J.C., Carlsson, P., Jordan, T. and Bhattacharya, S.S. (2003) Fox's in development and disease. Trends Genet, 19, 339-44. 129. Lai, E., Prezioso, V.R., Tao, W.F., Chen, W.S. and Darnell, J.E., Jr. (1991) Hepatocyte nuclear factor 3 alpha belongs to a gene family in mammals that is homologous to the Drosophila homeotic gene fork head. Genes & development, 5, 416-27. 130. Sasai, N., Mizuseki, K. and Sasai, Y. (2001) Requirement of FoxD3-class signaling for neural crest determination in Xenopus. Development (Cambridge, England), 128, 2525-36. 131. Grosso, S., Farnetani, M.A., Berardi, R., Vivarelli, R., Vanni, M., Morgese, G. and Balestri, P. (2002) Familial Axenfeld-Rieger anomaly, cardiac malformations, and sensorineural hearing loss: a provisionally unique genetic syndrome? American journal of medical genetics, 111,182-6. 132. Sommer, P., Napier, H.R., Hogan, B.L. and Kidson, S.H. (2006) Identification of Tgf betali4 as a downstream target of Foxcl. Development, growth & differentiation, 48, 297-308. 133. Petersen-Mahrt, S.K., Estmer, C., Ohrmalm, C., Matthews, D.A., Russell, W.C. and Akusjarvi, G. (1999) The splicing factor-associated protein, p32, regulates RNA splicing by inhibiting ASF/SF2 RNA binding and phosphorylation. The EMBO journal, 18, 1014-24. 134. Honore, B., Madsen, P., Rasmussen, H.H., Vandekerckhove, J., Celis, J.E. and Leffers, H. (1993) Cloning and expression of a cDNA covering the complete coding region of the P32 subunit of human pre-mRNA splicing factor SF2. Gene, 134,283-7. 135. Wang, Y., Finan, J.E., Middeldorp, J.M. and Hayward, S.D. (1997) P32/TAP, a cellular protein that interacts with EBNA-1 of Epstein-Barr virus. Virology, 236, 18-29.

201 136. Dedio, J., Jahnen-Dechent, W., Bachmann, M. and Muller-Esterl, W. (1998) The multiligand-binding protein gClqR, putative Clq receptor, is a mitochondrial protein. J Immunol, 160, 3534-42. 137. Soltys, B.J., Kang, D. and Gupta, R.S. (2000) Localization of P32 protein (gClq-R) in mitochondria and at specific extramitochondrial locations in normal tissues. Histochemistry and cell biology, 114, 245-55. 138. Jiang, J., Zhang, Y., Krainer, A.R. and Xu, R.M. (1999) Crystal structure of human p32, a doughnut-shaped acidic mitochondrial matrix protein. Proceedings of the National Academy of Sciences of the United States of America, 96, 3572-7. 139. Even, Y., Durieux, S., Escande, M.L., Lozano, J.C., Peaucellier, G., Weil, D. and Geneviere, A.M. (2006) CDC2L5, a Cdk-like kinase with RS domain, interacts with the ASF/SF2-associated protein p32 and affects splicing in vivo. Journal of cellular biochemistry, 99, 890-904. 140. Chattopadhyay, C, Hawke, D., Kobayashi, R. and Maity, S.N. (2004) Human p32, interacts with B subunit of the CCAAT-binding factor, CBF/NF-Y, and inhibits CBF-mediated transcription activation in vitro. Nucleic acids research, 32, 3632-41. 141. Zhang, B., Chambers, K.J., Faller, D.V. and Wang, S. (2007) Reprogramming of the S WI/SNF complex for co-activation or co-repression in prohibitin-mediated estrogen receptor regulation. Oncogene, 26, 7153-7. 142. Fang, M., Li, J., Blauwkamp, T., Bhambhani, C, Campbell, N. and Cadigan, K.M. (2006) C-terminal-binding protein directly activates and represses Wnt transcriptional targets in Drosophila. The EMBO journal, 25, 2735-45. 143. Gu, W., Malik, S., Ito, M., Yuan, C.X., Fondell, J.D., Zhang, X., Martinez, E., Qin, J. and Roeder, R.G. (1999) A novel human SRB/MED-containing cofactor complex, SMCC, involved in transcription regulation. Molecular cell, 3, 97-108. 144. Balciunas, D., Galman, C, Ronne, H. and Bjorklund, S. (1999) The Medl subunit of the yeast mediator complex is involved in both transcriptional

202 activation and repression. Proceedings of the National Academy of Sciences of the United States of America, 96, 376-81. 145. Zhou, S., Fujimuro, M., Hsieh, J.J., Chen, L., Miyamoto, A., Weinmaster, G. and Hayward, S.D. (2000) SKIP, a CBF1-associated protein, interacts with the ankyrin repeat domain of NotchIC To facilitate NotchIC function. Molecular and cellular biology, 20, 2400-10. 146. Honkanen, R.A., Nishimura, D.Y., Swiderski, R.E., Bennett, S.R., Hong, S., Kwon, Y.H., Stone, E.M., Sheffield, V.C. and Alward, W.L. (2003) A family with Axenfeld-Rieger syndrome and Peters Anomaly caused by a point mutation (Phel 12Ser) in the FOXC1 gene. American journal of ophthalmology, 135, 368- 75. 147. Sachdev, S., Bruhn, L., Sieber, H., Pichler, A., Melchior, F. and Grosschedl, R. (2001) PIASy, a nuclear matrix-associated SUMO E3 ligase, represses LEF1 activity by sequestration into nuclear bodies. Genes & development, 15,3088-103. 148. Tolkunova, E., Malashicheva, A., Parfenov, V.N., Sustmann, C, Grosschedl, R. and Tomilin, A. (2007) PIAS proteins as repressors of Oct4 function. Journal of molecular biology, 374, 1200-12. 149. Liu, B., Gross, M., ten Hoeve, J. and Shuai, K. (2001) A transcriptional corepressor of Statl with an essential LXXLL signature motif. Proceedings of the National Academy of Sciences of the United States of America, 98, 3203-7. 150. Liu, B., Liao, J., Rao, X., Kushner, S.A., Chung, CD., Chang, D.D. and Shuai, K. (1998) Inhibition of Statl-mediated gene activation by PIAS1. Proceedings of the National Academy of Sciences of the United States of America, 95, 10626-31. 151. Chung, CD., Liao, J., Liu, B., Rao, X., Jay, P., Berta, P. and Shuai, K. (1997) Specific inhibition of Stat3 signal transduction by PIAS3. Science (New 7or£, TV.7 278, 1803-5. 152. Rycyzyn, M.A. and Clevenger, C.V. (2002) The intranuclear prolactin/cyclophilin B complex as a transcriptional inducer. Proceedings of the National Academy of Sciences of the United States of America, 99, 6790-5.

203 153. Betz, A., Lampen, N., Martinek, S., Young, M.W. and Darnell, J.E., Jr. (2001) A Drosophila PIAS homologue negatively regulates stat92E. Proceedings of the National Academy of Sciences of the United States of America, 98, 9563-8. 154. Palvimo, J.J. (2007) PIAS proteins as regulators of small ubiquitin-related modifier (SUMO) modifications and transcription. Biochemical Society transactions, 35, 1405-8. 155. Schmidt, D. and Muller, S. (2003) PIAS/SUMO: new partners in transcriptional regulation. Cell Mo I Life Sci, 60, 2561-74. 156. Shuai, K. and Liu, B. (2005) Regulation of gene-activation pathways by PIAS proteins in the immune system. Nature reviews, 5, 593-605. 157. Okubo, S., Hara, F., Tsuchida, Y., Shimotakahara, S., Suzuki, S., Hatanaka, H., Yokoyama, S., Tanaka, H., Yasuda, H. and Shindo, H. (2004) NMR structure of the N-terminal domain of SUMO ligase PIAS1 and its interaction with tumor suppressor p53 and A/T-rich DNA oligomers. The Journal of biological chemistry, 279, 31455-61. 158. Gross, M., Liu, B., Tan, J., French, F.S., Carey, M. and Shuai, K. (2001) Distinct effects of PIAS proteins on androgen-mediated gene activation in prostate cancer cells. Oncogene, 20, 3880-7. 159. Duval, D., Duval, G., Kedinger, C, Poch, O. and Boeuf, H. (2003) The 'PINIT' motif, of a newly identified conserved domain of the PIAS protein family, is essential for nuclear retention of PIAS3L. FEBS letters, 554, 111-8. 160. Hochstrasser, M. (2001) SP-RTNG for SUMO: new functions bloom for a ubiquitin-like protein. Cell, 107, 5-8. 161. Jackson, P.K. (2001) A new RING for SUMO: wrestling transcriptional responses into nuclear bodies with PIAS family E3 SUMO ligases. Genes & development, 15,3053-8. 162. Joazeiro, C.A. and Weissman, A.M. (2000) RING finger proteins: mediators of ubiquitin ligase activity. Cell, 102, 549-52. 163. Weissman, A.M. (2001) Themes and variations on ubiquitylation. Nat Rev Mol Cell Biol, 2, 169-78.

204 164. Kotaja, N., Karvonen, U., Janne, O.A. and Palvimo, J.J. (2002) PIAS proteins modulate transcription factors by functioning as SUMO-1 ligases. Molecular and cellular biology, 22, 5222-34. 165. Minty, A., Dumont, X., Kaghad, M. and Caput, D. (2000) Covalent modification of p73alpha by SUMO-1. Two-hybrid screening with p73 identifies novel SUMO-1-interacting proteins and a SUMO-1 interaction motif. The Journal of biological chemistry, 275, 36316-23. 166. Shuai, K. (2006) Regulation of cytokine signaling pathways by PIAS proteins. Cell research, 16, 196-202. 167. Megidish, T., Xu, J.H. and Xu, C.W. (2002) Activation of p53 by protein inhibitor of activated Statl (PIAS1). The Journal of biological chemistry, 277, 8255-9. 168. Nojiri, S., Joh, T., Miura, Y., Sakata, N., Nomura, T., Nakao, H., Sobue, S., Ohara, H., Asai, K. and Ito, M. (2004) ATBF1 enhances the suppression of STAT3 signaling by interaction with PIAS3. Biochemical and biophysical research communications, 314, 97-103. 169. Nishida, T. and Yasuda, H. (2002) PIAS 1 and PIASxalpha function as SUMO-E3 ligases toward androgen receptor and repress androgen receptor- dependent transcription. The Journal of biological chemistry, 211, 41311-7. 170. Miyauchi, Y., Yogosawa, S., Honda, R., Nishida, T. and Yasuda, H. (2002) Sumoylation of Mdm2 by protein inhibitor of activated STAT (PIAS) and RanBP2 enzymes. The Journal of biological chemistry, 277, 50131-6. 171. Long, J., Wang, G., Matsuura, I., He, D. and Liu, F. (2004) Activation of Smad transcriptional activity by protein inhibitor of activated STAT3 (PIAS3). Proceedings of the National Academy of Sciences of the United States of America, 101,99-104. 172. Levy, C, Nechushtan, H. and Razin, E. (2002) A new role for the STAT3 inhibitor, PIAS3: a repressor of microphthalmia transcription factor. The Journal of biological chemistry, 277, 1962-6. 173. Wu, L., Wu, H., Ma, L., Sangiorgi, F., Wu, N., Bell, J.R., Lyons, G.E. and Maxson, R. (1997) Mizl, a novel zinc finger transcription factor that interacts

205 with Msx2 and enhances its affinity for DNA. Mechanisms of development, 65, 3- 17. 174. Nakagawa, K. and Yokosawa, H. (2002) PIAS3 induces SUMO-1 modification and transcriptional repression of IKF-l.FEBS letters, 530, 204-8. 175. Lyst, M.J., Nan, X. and Stancheva, I. (2006) Regulation of MBD1- mediated transcriptional repression by SUMO and PIAS proteins. The EMBO journal, 25, 5317-28. 176. Weigel, D. and Jackie, H. (1990) The fork head domain: a novel DNA binding motif of eukaryotic transcription factors? Cell, 63, 455-6. 177. Kaestner, K.H., Knochel, W. and Martinez, D.E. (2000) Unified nomenclature for the winged helix/forkhead transcription factors. Genes & development, 14, 142-6. 178. Wilm, B., James, R.G., Schultheiss, T.M. and Hogan, B.L. (2004) The forkhead genes, Foxcl and Foxc2, regulate paraxial versus intermediate mesoderm cell fate. Developmental biology, 271, 176-89. 179. Maeda, J., Yamagishi, H., McAnally, J., Yamagishi, C. and Srivastava, D. (2006) Tbxl is regulated by forkhead proteins in the secondary heart field. Dev Dyn, 235, 701-10. 180. Xu, H., Morishima, M., Wylie, J.N., Schwartz, R.J., Bruneau, B.G., Lindsay, E.A. and Baldini, A. (2004) Tbxl has a dual role in the morphogenesis of the cardiac outflow tract. Development (Cambridge, England), 131, 3217-27. 181. Li, S., Weidenfeld, J. and Morrisey, E.E. (2004) Transcriptional and DNA binding activity of the Foxp 1/2/4 family is modulated by heterotypic and homotypic protein interactions. Molecular and cellular biology, 24, 809-22. 182. Dou, C, Lee, J., Liu, B., Liu, F., Massague, J., Xuan, S. and Lai, E. (2000) BF-1 interferes with transforming growth factor beta signaling by associating with Smad partners. Molecular and cellular biology, 20, 6201-11. 183. Osborne, L.R. (2001) 'Forkhead' gene expression balanced on a knife-edge. Trends in molecular medicine, 7, 51. 184. Van Criekinge, W. and Beyaert, R. (1999) Yeast Two-Hybrid: State of the Art. Biological procedures online, 2, 1-38.

206 185. Phizicky, E.M. and Fields, S. (1995) Protein-protein interactions: methods for detection and analysis. Microbiological reviews, 59, 94-123. 186. Guo, X. and Wang, X.F. (2009) Signaling cross-talk between TGF- beta/BMP and other pathways. Cell research, 19, 71-88. 187. Chang, B., Smith, R.S., Peters, M., Savinova, O.V., Hawes, N.L., Zabaleta, A., Nusinowitz, S., Martin, J.E., Davisson, M.L., Cepko, C.L. et al. (2001) Haploinsufficient Bmp4 ocular phenotypes include anterior segment dysgenesis with elevated intraocular pressure. BMC genetics, 2, 18. 188. Pelton, R.W., Saxena, B., Jones, M., Moses, H.L. and Gold, L.I. (1991) Immunohistochemical localization of TGF beta 1, TGF beta 2, and TGF beta 3 in the mouse embryo: expression patterns suggest multiple roles during embryonic development. The Journal of cell biology, 115, 1091-105. 189. Flugel-Koch, C, Ohlmann, A., Piatigorsky, J. and Tamm, E.R. (2002) Disruption of anterior segment development by TGF-betal overexpression in the eyes of transgenic mice. Dev Dyn, 225, 111-25. 190. Saika, S., Saika, S., Liu, C.Y., Azhar, M., Sanford, L.P., Doetschman, T., Gendron, R.L., Kao, C.W. and Kao, W.W. (2001) TGFbeta2 in corneal morphogenesis during mouse embryonic development. Developmental biology, 240,419-32. 191. Zhou, Y., Kato, H., Asanoma, K., Kondo, H., Arima, T., Kato, K., Matsuda, T. and Wake, N. (2002) Identification of FOXC1 as a TGF-betal responsive gene and its involvement in negative regulation of cell growth. Genomics, 80, 465-72. 192. Reneker, L.W., Silversides, D.W., Xu, L. and Overbeek, P.A. (2000) Formation of corneal endothelium is essential for anterior segment development - a transgenic mouse model of anterior segment dysgenesis. Development (Cambridge, England), 127, 533-42. 193. Luetteke, N.C., Qiu, T.H., Peiffer, R.L., Oliver, P., Smithies, O. and Lee, D.C. (1993) TGF alpha deficiency results in hair follicle and eye abnormalities in targeted and waved-1 mice. Cell, 73, 263-78.

207 194. Majumdar, M., Meenakshi, J., Goswami, S.K. and Datta, K. (2002) Hyaluronan binding protein 1 (HABPl)/ClQBP/p32 is an endogenous substrate for MAP kinase and is translocated to the nucleus upon mitogenic stimulation. Biochemical and biophysical research communications, 291, 829-37. 195. Li, G., Luna, C, Liton, P.B., Navarro, I., Epstein, D.L. and Gonzalez, P. (2007) Sustained stress response after oxidative stress in trabecular meshwork cells. Molecular vision, 13, 2282-8. 196. Liton, P.B., Luna, C, Bodman, M., Hong, A., Epstein, D.L. and Gonzalez, P. (2005) Induction of IL-6 expression by mechanical stress in the trabecular meshwork. Biochemical and biophysical research communications, 337, 1229-36. 197. Rodel, B., Tavassoli, K., Karsunky, H., Schmidt, T., Bachmann, M., Schaper, F., Heinrich, P., Shuai, K., Elsasser, H.P. and Moroy, T. (2000) The zinc finger protein Gfi-1 can enhance STAT3 signaling by interacting with the STAT3 inhibitor PIAS3. The EMBO journal, 19, 5845-55. 198. Yamamoto, T., Sato, N., Sekine, Y., Yumioka, T., Imoto, S., Junicho, A., Fuse, H. and Matsuda, T. (2003) Molecular interactions between STAT3 and protein inhibitor of activated STAT3, and androgen receptor. Biochemical and biophysical research communications, 306, 610-5. 199. Jang, H.D., Yoon, K., Shin, Y.J., Kim, J. and Lee, S.Y. (2004) PIAS3 suppresses NF-kappaB-mediated transcription by interacting with the p65/RelA subunit. The Journal of biological chemistry, 279, 24873-80. 200. Junicho, A., Matsuda, T., Yamamoto, T., Kishi, H., Korkmaz, K., Saatcioglu, F., Fuse, H. and Muraguchi, A. (2000) Protein inhibitor of activated STAT3 regulates androgen receptor signaling in prostate carcinoma cells. Biochemical and biophysical research communications, 278, 9-13. 201. Liu, B., Mink, S., Wong, K.A., Stein, N., Getman, C, Dempsey, P.W., Wu, H. and Shuai, K. (2004) PIAS1 selectively inhibits interferon-inducible genes and is important in innate immunity. Nature immunology, 5, 891-8. 202. Wong, K.A., Kim, R., Christofk, H., Gao, J., Lawson, G. and Wu, H. (2004) Protein inhibitor of activated STAT Y (PIASy) and a splice variant lacking

208 exon 6 enhance sumoylation but are not essential for embryogenesis and adult life. Molecular and cellular biology, 24, 5577-86. 203. Roth, W., Sustmann, C, Kieslinger, M., Gilmozzi, A., Irmer, D., Kremmer, E., Turck, C. and Grosschedl, R. (2004) PIASy-deficient mice display modest defects in IFN and Wnt signaling. J Immunol, 173, 6189-99. 204. Liu, B., Yang, R., Wong, K.A., Getman, C, Stein, N., Teitell, M.A., Cheng, G., Wu, H. and Shuai, K. (2005) Negative regulation of NF-kappaB signaling by PIAS1. Molecular and cellular biology, 25, 1113-23. 205. Tan, J.A., Hall, S.H., Hamil, K.G., Grossman, G., Petrusz, P. and French, F.S. (2002) Protein inhibitors of activated STAT resemble scaffold attachment factors and function as interacting nuclear receptor coregulators. The Journal of biological chemistry, 277, 16993-7001.

209 Appendix: Antibody information and suppliers

Primary antibody Synthetic peptide Supplier and product recognized number Mouse IgGi DLTDDDDDK Invitrogen; R910-25 monoclonal a-Xpress Mouse IgG2a GKPIPNPLLGLDST Invitrogen; R960-25 monoclonal a-V5 Rabbit polyclonal a- HHHHHH Santa Cruz; Sc-803 His Goat polyclonal a- RTSGAFVYDCSKF Abeam; ab5079 FOXC1 Goat polyclonal a- RHAAPYSYDCTKY Abeam; ab5060 FOXC2 Goat polyclonal a-p32 N/A Dr. Tom Hobman Rabbit polyclonal a- Human PIAS3 amino acid Santa Cruz; Sc-14017 PIAS3 451-619 Rabbit polyclonal a- Acetylated Lys9 of Histone Cell signaling; #9671 Acetyl-Histone H3 H3 Lys9 Rabbit polyclonal a- Full length GFP Abeam; ab6556 GFP Rabbit polyclonal a- N/A Santa Cruz; Sc-204 TFIID

Secondary antibody Supplier and product number HRP-conjugated goat a-mouse IgG Jackson ImmunoResearch; 115-035-003 HRP-conjugated goat a-rabbit IgG Jackson ImmunoResearch; 111-035-003 HRP-conjugated donkey a-goat IgG Jackson ImmunoResearch; 705-035-003 Alexa Fluor 594 donkey a-goat IgG Molecular Probes; Al 1058 Alexa Fluor 488 donkey a-mouse IgG Molecular Probes; A21202 Cy2 donkey a-mouse IgG Jackson ImmunoResearch; 715-225-150 Cy3 goat a-mouse IgG

210