Matriptase in Skin: Function and Regulation

Authors Chen, Ya-Wen

Publication Date 2012

Abstract Epidermal differentiation is a carefully orchestrated process that leads to the formation of the critical protective barrier provided by the skin. The process of generating a functional epidermal layer requires progressive remodeling of cell morphol...

Keywords HAI-1; matriptase; protease; Antithrombins; Keratinocytes; Plasminogen

Download date 24/09/2021 03:25:33

Link to Item http://hdl.handle.net/10713/2141 Ya-Wen Chen Greenebaum Cancer Center University of Maryland School of Medicine 655 W. Baltimore Street, RM 10-010 Baltimore, MD 21201 Phone: 410.706.3268 Cellphone: 202.674.0282 Email: [email protected]

Education/Training

Aug. 2007-present Ph.D. candidate, Molecular Medicine, Cancer Biology University of Maryland, Baltimore

Aug. 2006-2007 May M.S., Tumor Biology Georgetown University, Washington, DC

Aug. 1999-2003 May B.S., Biology Tung-Hai University, Taichung, Taiwan

Research Experience

Aug. 2007-present Graduate Research Assistant, Molecular Medicine, Cancer Biology Graduate Program in Life Science University of Maryland School of Medicine, Baltimore, MD Thesis: Matriptase in Skin: Function and Regulation Advisor: Dr. Chen-Yong Lin

Jun. 2008-2008 Aug. Graduate Lab Rotation, Dept of Biochemistry and Molecular Biology Greenbaum Cancer Center, University of Maryland School of Medicine Advisor: Dr. Richard L. Eckert

May 2007-2007 Aug. Research Assistant, Dept of Biochemistry and Molecular Biology Greenbaum Cancer Center, University of Maryland School of Medicine Advisor: Dr. Chen-Yong Lin

Aug. 2006-2007 May Graduate student, Tumor Biology Program Lombardi Cancer Center, Georgetown University, Washington, D.C. Advisor: Dr. Chen-Yong Lin

Jul. 2004-2004 Aug. Intern, Dept of Molecular Medicine & Cellular Oncology M.D. Anderson Cancer Center, Houston, TX Advisor: Dr. Mien-Chie Hung

Jul. 2003-2006 Jul. Research Assistant, Department of Education and Research Taichung Veteran General Hospital, Taichung, Taiwan Advisor: Dr. Dr. Shih-Lan Hsu

Jan. 2003-2003 Mar. Intern, Dept of Molecular Medicine & Cellular Oncology M.D. Anderson Cancer Center, Houston, TX Advisor: Dr. Mong-Hong Lee

Aug. 2000-2003 Jun. Undergraduate Research Assistant, Dept of Education and Research Taichung Veteran General Hospital, Taichung, Taiwan Advisor: Dr. Dr. Shih-Lan Hsu

Sep. 1996-1998 Aug. Intern, Department of Biology National Changhua University of Education, Taiwan Advisor: Dr. Chi-Ying Lee

Honors

2010 Travel Award, University of Maryland, Baltimore 2010 Winner of Biochemical Journal Poster Prize, Plasminogen Activation & Extracellular Proteolysis Gordon Conference, Ventura, CA 2010 2nd place, Poster competition, Student Category of Plasminogen Activation & Extracellular Proteolysis Gordon Conference, Ventura, CA 1999 Masterpiece of Biology of National Science & Experiment Competition, Taiwan 1998 Young Scientist Award of the 1st APEC Youth Science Festival, Seoul, Korea, Representative of Taiwan 1998 The third Award of the 38th National Private and High Schools Science Fair of the Republic of China

Leadership

2007-present President, Taiwanese Student Association of the University of Maryland, Baltimore 2007-2008 President, DC area Taiwanese Student Association

(approx. 800 members across 6 states)

Supervisory/Training experience

2011-present Galen Shi, high school student, Mentor of the Gifted & Talented Intern/Mentor program 2011-present Maegan Wadman, undergraduate student 2011 Manchung Tu, undergraduate student 2011 Osei Xeniel Durand Richards, undergraduate student 2010-present Sean Moore, Masters student 2010 Kimberly Lynn Osborne, PhD student 2007-2008 Moran Choe, PhD student 2008 Yi-Le Hsu, Medical student

Ph.D. Thesis Overall, my thesis work serves to:

 Identify plasminogen as a keratinocyte-selective extracellular stimulus for matriptase activation. The discovery of plasminogen as an initiating signal of the matriptase-driven protease cascade affirms the theory that the matriptase-uPA-plasmin cascade is not unidirectional in the activation of its components, but it is reciprocal.  Reveal that keratinocytes employ antithrombin as a significant endogenous protease inhibitor. The enhanced role of antithrombin in matriptase inhibition in keratinocytes reveals the regulatory adaptation in stratified epithelial cells due to the changes in tissue structure, compared to the polarized epithelial cells.  Reveal that matriptase acts on its molecular targets in two different ways: 1) a rapid activation of prostasin by cell-associated active matriptase under extremely tight control of HAI-1, and 2) the action on several other substrates, including uPA, HGF, and syndecan-1, by secreted active matriptase that is controlled by antithrombin.  Build up the functional linkage between matriptase and its downstream substrate, prostasin.  Demonstrate that the physiological role of matriptase in human skin likely lies in the basal and spinous keratinocytes that are involved in proliferation and early differentiation. The role of matriptase in the early stages of keratinocyte life history was further supported by the increased matriptase zymogen activation in the keratinocytes of the bulge area, likely the stem cells.

Submitted and In preparation

1. Chen YW., Xu ZH., Baksh A., Wang, JK., Swanson R., Olson ST., Johnson M., Lin CY. Keratinocytes adopt antithrombin to control the level of active matriptase shed into extracellular milieu. Submitted to PNAS.

2. Chen YW., Xu ZH., Baksh A., Chou FP., Wang, JK., Chen CY., Shi G., Eckert

R., Johnson M., Lin CY. Matriptase in human epidermis: differentiation-regulated expression and enhanced zymogen activation in the epidermal stem cells. Submitted to Journal of Investigative Dermatology.

3. Chen YW., Xu ZH., Eckert R., Johnson M., Lin CY. Plasminogen-dependent matriptase zymogen activation accelerates pericellular plasmin generation in differentiating primary human keratinocytes. Manuscript under writing.

Peer-Reviewed Publications

1. Chiu YJ., Huang TH., Chiu CS., Lu TC., Chen YW., Peng WH., Chen CY. Analgesic and antiinflammatory activities of the aqueous extract from Plectranthus amboinicus (Lour.) spreng. Both in vitro and in vivo. 2012 Evid Based Complement Alternat Med., In Press.

2. Xu ZH., Chen YW., Battu A., Wilder, P., Weber D., Yu W., MacKerell A., Chen LM., Chai K., Johnson M., Lin CY. Targeting zymogen activation to control the matriptase-prostasin proteolytic cascade. 2011 J Med Chem., 54(21):7567-79.

3. Chou FP., Xu H., Lee MS., Chen YW., Swanson R., Olson ST., Johnson MD., Lin CY. Matriptase is inhibited by extravascular antithrombin in epithelial cells but not in most carcinoma cells. 2011 Am J Physiol Cell Physiol., 301(5):C1093-103.

4. Xu H., Xu ZH., Tseng IC., Chou FP., Chen YW., Wang JK., Johnson MD., Kataoka H., Lin CY. Active matriptase can rapidly be locked in an inactive homodimer and inhibited by other protease inhibitors in the absence of sufficient HAI-1. 2011 Am J Physiol Cell Physiol., 302(2):C453-62.

5. Chen CJ., Wu BY., Tsao PI., Chen CY., Wu MH., Chan YL., Lee HS., Johnson MD., Eckert RL., Chen YW., Chou F., Wang JK., Lin CY. Increased matriptase zymogen activation in inflammatory skin disorders. 2011 Am J Physiol Cell Physiol., 300(3):C406-15.

6. Chen YW., Wang JK., Chou FP., Chen CY., Rorke E., Chen LM., Chai K., Eckert R., Johnson M., Lin CY. Regulation of the matriptase-prostasin cell surface proteolytic cascade by hepatocyte growth factor activator inhibitor-1 (HAI-1) during epidermal differentiation. 2010 J Biol Chem., 285(41):31755-62.

7. Jiang HB., Jans R., Xu W., Rorke E., Lin CY., Chen YW., Fang SY., Zhong YW., Eckert R. Type I transglutaminase accumulation in the endoplasmic reticulum may be an underlying cause of autosomal recessive congenital ichthyosis. 2010 J Biol Chem., 285(41):31634-46.

8. Chen YW., Lee MS., Lucht A., Chou FP., Huang W., Havighurst TC., Kim KM.,

Wang JK., Antalis T., Johnson M., Lin CY. TMPRSS2, a serine protease expressed in the prostate on the apical surface of luminal epithelial cells and released into semen in prostasomes, is misregulated in prostate cancer cells. 2010 Am J Pathol., 176(6):2986-96.

9. Wang JK., Lee MS., Tseng IC., Chou FP., Chen YW., Fulton A., Lee HS., Chen CJ., Johnson MD., Lin CY. Polarized epithelial cells secrete matriptase as a consequence of zymogen activation and HAI-1-mediated inhibition. 2009 Am J Physiol Cell Physiol., 297(2):C459-70.

10. Lin CY., Tseng IC., Chou FP., Su SF., Chen YW., Johnson MD., Dickson RB. Zymogen activation, inhibition, and ectodomain shedding of matriptase. 2008 Front Biosci., 13:621-35.

Review Article

Chen YW., Eckert R., Johnson M., Lin CY. The transmembrane serine protease matriptase: implications for cellular biology and human diseases. 2012 J Med Sci., Invited review.

Book Chapters

Lin CY., Chen YW., Xu ZH., Johnson MD. MS. 681|Matriptase. Handbook of Proteolytic Enzymes (Third edition).

Presentations

Feb. 2012 Plasminogen-dependent matriptase zymogen activation accelerates pericellular plasmin generation in differentiating primary keratinocytes. Protease research interest group, University of Maryland, Baltimore

Oct. 2011 Matriptase in skin: function and regulation. Molecular medicine student seminar, University of Maryland, Baltimore

Aug. 2011 Matriptase protease network in skin. Molecular medicine orientation, Representative of students, University of Maryland, Baltimore

Feb. 2011 Matriptase in skin: function and regulation. Cell signaling research interest group, University of Maryland, Baltimore

Oct. 2009 Matriptase in skin: function and regulation. Cell signaling research interest group, University of Maryland, Baltimore

Abstracts and Poster Presentations

Aug. 2011 Matriptase protease network in skin. Molecular medicine orientation, Representative of students, University of Maryland, Baltimore

May 2011 Matriptase in skin: function and regulation. Cancer Biology Retreat, University of Maryland, Baltimore

Apr. 2011 Matriptase in skin: function and regulation. Graduate Research Conference, University of Maryland, Baltimore

Apr. 2010 The function and regulation of matriptase in skin. Graduate Research Conference, University of Maryland, Baltimore

Feb. 2010 TMPRSS2, a serine protease expressed in the prostate on the apical surface of luminalepithelial cells and released into semen in prostasomes, is misregulated in prostate cancer cells. Gordon Research Conference on Plasminogen Activation and Extracellular Proteolysis, Ventura, CA

Apr. 2009 TMPRSS2, a serine protease expressed in the prostate on the apical surface of luminalepithelial cells and released into semen in prostasomes, is misregulated in prostate cancer cells. Graduate Research Conference, University of Maryland, Baltimore

Selected skills & Techniques

˙Extensive experience in protein purification, identification and characterization ˙Western blot ˙Gel chromatography for protein ˙Enzyme assay ˙Immunoprecipitation ˙Primary cell isolation ˙Monoclonal antibody generation ˙Real-time PCR/PCR ˙Immunohistochemistry and immunocytochemistry ˙Reporter gene assay ˙Lentivirus and Adenovirus generation ˙siRNA and shRNA ˙microRNA technique to generate stable cell line with targeted protein knockdown ˙Mammalian cell expression system ˙Cell culture ˙Gel electrophoresis ˙Organotypic culture ˙Transduction ˙ELISA ˙Molecular cloning ˙Transfection ˙Chromatin immunoprecipitation assay (ChIP) ˙Biotin label ˙Agarose gel electrophoresis ˙DNA/RNA purification ˙Assays for apoptosis, cytotoxicity and cell proliferation ˙Confocal and epi-fluorescence microscopy ˙High-throughput screening

Computer skills MS Office, Graphics programs such as PhotoImpact, Photoshop, Sigma Plot, EndNote, Reference manager

Other languages Native in Mandarin and Taiwanese

Abstract

Title of Dissertation: Matriptase in Skin: Function and Regulation

Name: Ya-Wen Chen, Doctor of Philosophy, 2012

Dissertation Directed by: Chen-Yong Lin, Associate Professor,

Department of Biochemistry and Molecular Biology

Epidermal differentiation is a carefully orchestrated process that leads to the formation of the critical protective barrier provided by the skin. The process of generating a functional epidermal layer requires progressive remodeling of cell morphology and tissue structure, and involves significant pericellular proteolysis that must be regulated in a precisely controlled manner. In particular, the matriptase-driven protease network plays a critical role in epidermal barrier construction as well as in the regenerative processes required for wound healing. In this dissertation, I have identified and characterized novel molecular mechanisms governing the regulation of matriptase, and cellular mechanisms by which matriptase activates its molecular targets and contributes to keratinocyte differentiation and formation of epidermal barrier. First, I identified plasminogen as a keratinocyte-selective extracellular stimulus for matriptase activation. The discovery of plasminogen as an initiating signal of the protease cascade reaffirms the theory that the matriptase-uPA-plasmin cascade is not unidirectional in the activation of its components, but it is reciprocal. In addition to HAI-1, I also revealed keratinocytes employ antithrombin as a significant endogenous protease inhibitor. The enhanced role of antithrombin in matriptase inhibition in keratinocytes reveals the

regulatory adaptation in stratified epithelial cells due to the changes in tissue structure, compared to the polarized epithelial cells. With the dual inhibitory mechanisms, I further revealed that matriptase acts on its molecular targets in two different ways: a rapid activation of prostasin by cell-associated active matriptase under extremely tight control of HAI-1, and the action on several other substrates, including uPA, HGF, and syndecan-1, by secreted active matriptase that is controlled by antithrombin. Last but not least, I also demonstrated that the physiological role of matriptase in human skin likely lies in the basal and spinous keratinocytes that are involved in proliferation and early differentiation. The role of matriptase in the early stages of keratinocyte life history was further supported by the increased matriptase zymogen activation in the keratinocytes of the bulge area. By charactering the prominent regulators, downstream effectors, and the expression and activation states of matriptase in human skin, a clearer picture is emerging regarding the role of matriptase in skin biology.

Matriptase in Skin: Function and Regulation

by Ya-Wen Chen

Dissertation submitted to the faculty of the Graduate School of the University of Maryland, Baltimore in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2012

© Copyright 2012 by Ya-Wen Chen

All rights Reserved

Acknowledgements

Though only my name appears on the cover of this dissertation, a great many people have contributed to its production. I owe my gratitude to all those people who have made this dissertation possible and because of whom my graduate experience has been one that I will cherish forever.

My deepest gratitude is to my advisor, Dr. Chen-Yong Lin. I have been amazingly fortunate to have an advisor who gave me the freedom to explore on my own and at the same time the guidance to recover when my steps faltered. Dr. Lin taught me how to question thoughts and express ideas. His patience and support helped me overcome many crisis situations and finish this dissertation. I hope that one day I would become as good an advisor to my students as Dr. Lin has been to me.

In addition, my thanks go to Dr. Zhenghong Xu, Feng-Pai Chou, Dr. Richard L.

Eckert, Dr. Ellen Rorke, Dr. Haibing Jiang, Wen Xu, Dr. Yap Ching Chew, Dr. Steven

Olson, Dr. Richard Swanson, Dr. Hiroaki Kataoka, Dr. Michael D. Johnson, Dr.

Jehng-Kang Wang, Kathleen Norris, Moran Choe, and Adrienne NH Baksh, for helping me with the work described in this dissertation.

I would like to thank my dissertation committee members: Dr. Amy M. Fulton, Dr.

Toni Antalis, Dr. Pei N Feng, Dr. Antonio Passaniti, and Dr. Richard L. Eckert. Without whose knowledge and assistance, this study would not have been successful. I am deeply grateful to them for the long discussions that helped me sort out the technical details of my work. I am also thankful to them for encouraging the use of correct grammar and consistent notation in my writings and for carefully reading and

iii commenting on revisions of this dissertation.

Many friends have helped me stay sane through these difficult years. Their support and care helped me overcome setbacks and stay focused on my graduate study.

I greatly value their friendship and I deeply appreciate their belief in me, especially my best friend, Chao-Yu Chen and her family who have provided endless support during my

Ph.D. journey.

Most importantly, none of this would have been possible without the love and patience of my family. My family has been a constant source of love, concern, support and strength all these years. They have aided and encouraged me throughout this endeavor. I would like to express my heart-felt gratitude to my family, especially to my parents, sisters, brother, piggy, bibizai, and oakley. I warmly appreciate the generosity and understanding of them. Their supports and encouragements are the reason that I could reach the end of this Ph.D. journey.

iv

Table of Contents

Acknowledgements ...... iii

Table of Contents ...... v

List of Tables ...... xiv

List of Figures ...... xv

Chapter 1 – Introduction ...... 1

1.1 Human skin ...... 2

1.1.1 Functions ...... 2

1.1.2 Skin layers ...... 3

1.1.2.1 The epidermis ...... 3

1.1.2.1.1 Stratum basale ...... 5

1.1.2.1.2 Stratum spinosum...... 7

1.1.2.1.3 Stratum granulosum ...... 8

1.1.2.1.4 Stratum lucidum ...... 9

1.1.2.1.5 Stratum corneum ...... 9

1.1.2.2 The dermis ...... 11

1.1.2.2.1 Papillary region ...... 12

1.1.2.2.2 Reticular region ...... 12

v

1.1.2.2.3 Dermal cell population ...... 12

1.1.2.3 The subcutaneous tissue ...... 13

1.1.3 Difference between human and mouse skin ...... 13

1.1.4 Importance of proteases, protease inhibitors, and proteolytic balance in skin 14

1.2 Proteases and protease inhibitors ...... 16

1.2.1 Introduction ...... 16

1.2.2 Serine proteases ...... 18

1.2.3 Kunitz-type protease inhibitors ...... 18

1.2.4 Serine protease inhibitors (Serpins) ...... 19

1.3 Matriptase-mediated cell surface proteolysis cascade ...... 20

1.3.1 Matriptase cascade in skin ...... 20

1.3.2 Matriptase ...... 21

1.3.2.1 Family, localization, and expression ...... 22

1.3.2.2 Matriptase structure ...... 22

1.3.2.3 Life cycle of matriptase ...... 25

1.3.2.4 Extracellular stimuli for matriptase activation ...... 27

1.3.2.5 Activation machinery ...... 29

1.3.2.6 Matriptase functions...... 30

1.3.3 Matriptase substrates ...... 32

1.3.4 Matriptase inactivation...... 35

vi

1.3.4.1 Kunitz-type serine proteases – HAI-1 and HAI-2 ...... 35

1.3.4.2 Serine protease inhibitors (Serpins) – Antithrombin, alpha-1 antitrypsin

(1AT), and alpha-2-antiplasmin (2AP) ...... 39

1.3.4.2.1 Antithrombin ...... 39

1.3.4.3 Other inhibitors ...... 41

1.4 Hypotheses ...... 41

1.5 Research Objectives ...... 41

Chapter 2 – Regulation of the Matriptase-Prostasin Cell Surface Proteolytic

Cascade by Hepatocyte Growth Factor Activator Inhibitor-1 during Epidermal

Differentiation ...... 43

2.1 Abstract ...... 44

2.2 Introduction ...... 45

2.3 Materials and Methods ...... 47

2.3.1 Antibodies ...... 47

2.3.2 Lentiviral particle preparation and cell transduction for the delivery of small

hairpin RNAs ...... 48

2.3.3 Cell culture and media ...... 48

2.3.4 Organotypic culture...... 49

2.3.5 Acid-induced matriptase activation in HaCaT cells ...... 49

2.3.6 Immunoblot analysis ...... 49

2.3.7 Prostasin-HAI-1 complex purification ...... 50 vii

2.3.8 Mass spectrometry analysis and identification of prostasin ...... 50

2.4 Results ...... 51

2.4.1 Activation of matriptase starts early, and is maintained throughout in vitro

human epidermal differentiation ...... 51

2.4.2 Acid exposure induces simultaneous activation of matriptase and a putative

serine protease in HaCaT cells ...... 54

2.4.3 Purification of the 85-kDa HAI-1 complex and identification of prostasin as a

component of the complex ...... 56

2.4.4 Prostasin activation is dependent on matriptase ...... 58

2.4.5 Matriptase and prostasin activation kinetics in HaCaT cells ...... 62

2.5 Discussion ...... 64

2.6 Acknowledgments ...... 67

Chapter 3 – Antithrombin regulates matriptase-mediated keratinocyte pericellular proteolysis involved in plasminogen activation, syndecan shedding, and activation of hepatocyte growth factor ...... 69

3.1 Abstract ...... 70

3.2 Introduction ...... 71

3.3 Materials and Methods ...... 75

3.3.1 Reagent ...... 75

3.3.2 Antibodies ...... 75

3.3.3 Cell cultures ...... 76

viii

3.3.4 Acid-induced matriptase activation...... 77

3.3.5 Purification of active matriptase ...... 77

3.3.6 Protease activity assay ...... 78

3.3.7 Activation of pro-HGF ...... 79

3.3.8 Immunoblot ...... 79

3.3.9 Antithrombin binding assay ...... 79

3.3.10 Purification and identification of 110-kDa matriptase complexes ...... 80

3.3.11 Syndecan-1 shedding ...... 80

3.4 Results ...... 81

3.4.1 Active matriptase is either rapidly inhibited by HAI-1 or shed from cell surface

of human keratinocytes ...... 81

3.4.2 The pericellular active matriptase can activate pro-HGF and accelerate

plasminogen activation ...... 85

3.4.3 Identification of antithrombin as a significant inhibitor of matriptase in human

keratinocytes ...... 89

3.4.4 Binidng of antithrombin to the surface of keratinocytes depends on its heparin

binding affinity and affects the levels of active protease shed into extracellular

milieu ...... 97

3.4.5 Extracellular active matriptase can shed syndecan-1 ...... 101

3.5 Discussion ...... 102

Chapter 4 – Matriptase in human epidermis: differentiation-regulated expression

ix and enhanced zymogen activation in the epidermal stem cells...... 111

4.1 Abstract ...... 112

4.2 Introduction ...... 113

4.3 Material and Methods ...... 117

4.3.1 Cell lines and Cell cultures ...... 117

4.3.2 Antibodies ...... 118

4.3.3 Immunoblot analysis ...... 119

4.3.4 Organotypic culture...... 120

4.3.5 Immunohistochemistry ...... 120

4.3.6 Crystal violet cell proliferation assay ...... 121

4.3.7 Real-time polymerase chain reaction (PCR)...... 121

4.4 Results ...... 122

4.4.1 Matriptase expression and distribution in skin organotypic culture ...... 122

4.4.2 Elevated matriptase activation during epidermal tissue regeneration and in the

epidermal stem cell ...... 128

4.4.3 Role of matriptase in keratinocyte proliferation and differentiation ...... 131

4.5 Discussion ...... 135

Chapter 5 – Plasminogen-dependent matriptase zymogen activation accelerates pericellular plasmin generation in differentiating primary human keratinocytes . 140

5.1 Abstract ...... 141

x

5.2 Introduction ...... 142

5.3 Material and Methods ...... 146

5.3.1 Reagents ...... 146

5.3.2 Cell culture ...... 147

5.3.3 Purification of serum factor and identification ...... 148

5.3.4 Immunoblot ...... 149

5.3.5 Protease activity assay ...... 149

5.4 Results ...... 150

5.4.1 Interplay of calcium ions and serum stimulates primary human keratinocytes to

activate matriptase...... 150

5.4.2 Plasminogen, the serum factor that induces matriptase activation in primary

human keratinocytes ...... 162

5.4.3 Matriptase activation accelerates production of plasmin ...... 170

5.5 Discussion ...... 176

Chapter 6 – Conclusion and discussion ...... 184

6.1 What does matriptase-driven protease cascade do in human skin and what happens

when it goes wrong? ...... 187

6.2 How does the matriptase-derived protease cascade begin in keratinocytes? ...... 191

6.3 What does matriptase do in keratinocytes? ...... 195

6.4 How is matriptase inhibited in keratinocytes? ...... 198

xi

6.5 Summary ...... 201

Chapter 7 -- Appendix Chapter – TMPRSS2, a Serine Protease Expressed in the

Prostate on the Apical Surface of Luminal Epithelial Cells and Released into Semen in Prostasomes, Is Misregulated in Prostate Cancer Cells ...... 203

7.1 Abstract ...... 204

7.2 Introduction ...... 205

7.3 Materials and Methods ...... 207

7.3.1 Chemicals and reagents...... 207

7.3.2 Cell culture condition ...... 207

7.3.3 Generation of TMPRSS2 monoclonal antibodies ...... 208

7.3.4 Mass spectrometry analysis and identification of proteins ...... 209

7.3.5 Immunohistochemistry ...... 209

7.3.6 IHC staining of prostate tissue microarray for TMPRSS2 and analysis by

AQUA ...... 210

7.3.7 Western blot analysis ...... 212

7.3.8 Diagonal SDS-PAGE ...... 212

7.3.9 PNGase F assay...... 213

7.3.10 Analysis of TMPRSS2 and matriptase shedding ...... 213

7.3.11 Preparation of prostasomes ...... 214

7.4 Results ...... 214

xii

7.4.1 Generation and characterization of TMPRSS2 monoclonal antibodies...... 214

7.4.2 Aberrant subcellular distribution and increased expression of TMPRSS2

protein in prostate cancer cells ...... 217

7.4.3 TMPRSS2 in prostate cancer cell lines ...... 221

7.4.4 TMPRSS2 is present in both noncovalent and disulfide-linked complexes in

LNCaP cells ...... 226

7.4.5 TMPRSS2 is lightly glycosylated ...... 228

7.4.6 The shedding of TMPRSS2 into the extracellular milieu is very limited in

LNCaP cells ...... 231

7.4.7 TMPRSS2 is secreted as both noncovalent and disulfide bond-linked

complexes into human semen ...... 233

7.4.8 TMPRSS2 is secreted as components of the prostasome into human semen 236

7.5 Discussion ...... 239

7.6 Acknowledgements ...... 247

Bibliography ...... 248

xiii

List of Tables

Table 7.1AQUA Scores of TMPRSS2 ...... 222

xiv

List of Figures

Figure 1.1 Schematic illustration of the human skin structure...... 4

Figure 1.2 Schematic representation of the matriptase structure...... 23

Figure 1.3 Schematic representation of the life cycle of matriptase...... 26

Figure 1.4 Schematic representation of the HAI-1 structure...... 37

Figure 2.1 Detection of a novel HAI-1 complex in differentiating human keratinocytes. 52

Figure 2.2 Acid induced activation of matriptase is associated with the formation of a novel 85-kDa HAI-1 complex in immortalized human keratinocytes...... 55

Figure 2.3 Identification of prostasin as a component of the novel 85-kDa HAI-1 complex.

...... 57

Figure 2.4 Acid exposure of HaCaT keratinocytes results in the activation of prostasin associated with matriptase activation...... 59

Figure 2.5 Prostasin activation is dependent on matriptase...... 61

Figure 2.6 Activation and HAI-1 binding kinetics of the matriptase-prostasin proteolytic cascade...... 63

Figure 2.7 The model for the regulation of matriptase-prostasin proteolytic cascade...... 68

Figure 3.1 Shedding of active matriptase by human keratinocyte...... 82

Figure 3.2 Activation of pro-HGF and acceleration of plasminogen activation by active matriptase shed from human keratinocytes...... 87

Figure 3.3 Identification of antithrombin as a component of the novel 110-kDa matriptase complex...... 90

Figure 3.4 The keratinocyte-selected 110-kDa matriptase complex is a protease/serpin complex...... 93

xv

Figure 3.5 Identification of antithrombin as a component of the novel 110-kDa matriptase complex...... 96

Figure 3.6 The shedding of active matriptase is inversely correlated with the levels of membrane-bound antithrombin...... 98

Figure 3.7 Syndecan-1 can be robustly shed by active matriptase...... 103

Figure 3.8 A schematic model for the functional and regulatory divergence of matriptase between the simple/polarized epithelium and the stratified epithelium...... 109

Figure 4.1 Expression of matriptase and HAI-1 in human primary keratinocyte and cellular localization of matriptase and HAI-1 in skin equivalents...... 123

Figure 4.2 Cellular localization of matriptase and HAI-1 in normal human skin...... 126

Figure 4.3 Cellular localization of matriptase and HAI-1 in human normal oral mucosa.

...... 127

Figure 4.4 Elevated matriptase activation in epidermal stem cells of hair follicle...... 130

Figure 4.5 Role of matriptase in keratinocyte differentiation and proliferation...... 132

Figure 5.1 The serum factor-induced activation of matriptase is a primary keratinocyte specific event...... 152

Figure 5.2 The calcium and serum are indispensable to induce evident activation of matriptase...... 154

Figure 5.3 Sequential administration of calcium and serum affects the outcome of matriptase activation...... 155

Figure 5.4 De novo protein synthesis is required for serum/calcium-induced activation of matriptase...... 158

Figure 5.5 Serum/calcium-induced activation of matriptase requires elevating

xvi intracellular calcium and activation of protein kinase C (PKC) and phosphatidylinositol

3-kinases (PI3Ks)...... 160

Figure 5.6 Matriptase activation cannot be stopped by calcium channel blockers or initiated by calcium ionophores...... 163

Figure 5.7 Purification and identification of the serum factor that induces matriptase activation in primary human keratinocyte...... 165

Figure 5.8 Plasminogen is the serum factor that induces matriptase activation in primary human keratinocytes and its effect can be inhibited by 6-AHA...... 168

Figure 5.9 Plasminogen rather than plasmin is the trigger of matriptase activation...... 169

Figure 5.10 Plasmin generation is accelerated by matriptase activation...... 173

Figure 5.11 Acceleration of plasmin generation is done by pericellular active matriptase on cell surface...... 175

Figure 5.12 A schematic model for the reciprocal activation of matriptase-plasminogen activator-plasminogen cascade...... 179

Figure 5.13 Schematic models of the status of matriptase activation in skin under different conditions...... 182

Figure 6.1 Schematic models of the regulation and function of matriptase-driven protease network...... 202

Figure 7.1 TMPRSS2 recombinant protein and mAbs...... 215

Figure 7.2 Verification of TMPRSS2 by proteomic analysis...... 218

Figure 7.3 Cellular localization of TMPRSS2 in normal and carcinomatous prostate tissues...... 219

Figure 7.4 The expression status of endogenous TMPRSS2 protein in human prostate

xvii cancer cells...... 223

Figure 7.5 The expression status of endogenous TMPRSS2 protein in human prostate cancer cells...... 225

Figure 7.6 N-glycosylation of TMPRSS2 is responsible for a modest increase in molecular mass...... 230

Figure 7.7 Shedding of TMPRSS2 to the extracellular milieu by LNCaP cells is very limited...... 232

Figure 7.8 TMPRSS2 is present in multiple complexes in human semen...... 234

Figure 7.9 TMPRSS2 is expressed as components of the prostasome...... 237

Figure 7.10 Verification of the molecular weight of TMPRSS2 by commercially available antibodies...... 241

xviii

Chapter 1 – Introduction

1

1.1 Human skin

1.1.1 Functions

In humans, the largest organ of the integumentary system is the skin, with a total surface of approximately 20 square feet. The skin represents about 16% of body mass and the composition and thickness of it varies depending on anatomical region 1, 2. The skin not only gives us appearance and shape, it also serves other important functions, such as protection, sensation, heat regulation, control of evaporation, storage and synthesis, absorption, and water resistance 3-8. Among these functions, the most important is to form a protective barrier that interfaces with a sometimes hostile environment 2. Skin protects everything that lies beneath it and acts as a cushion against insult to the body. Its continuous, tightly connected surface, lightly coated with oily sebum, allows and limits the inward and outward passage of water, electrolytes and various other substances. For the body to function well, skin is also involved in maintaining the proper body temperature, by regulating blood vessel diameter in the cutaneous vascular system and evaporation of sweat from the surface 8. The three types of nerve endings of the skin, Meissner’s corpuscles, Pacinian corpuscles, and free nerve endings, enable it to gather sensory information from the environment so we can sense touch, pressure, heat and cold, pain, and tissue injury 8. In addition to acting as a protective barrier, skin also plays an active role in the immune system, protecting us from diseases. It normally contains all the elements of the immune system except B cells.

Understanding the structure of the three layers of skin, the epidermis, dermis, and subcutaneous tissues, is the key to understanding how the skin can perform these amazing functions.

2

1.1.2 Skin layers

Human skin is composed of three layers. The first is the epidermis, which provides waterproofing and serves as the physical and chemical barrier between the interior body and exterior environment. The second is the dermis, which serves as the structural support. The third is the subcutaneous tissue, containing larger vessels and nerves 9.

1.1.2.1 The epidermis

The epidermis is the outermost layer of skin and is made up of stratified squamous epithelium, with an underlying basal lamina. The average thickness of epidermis varies from 0.04 mm on the eyelids to 1.5 mm on the palms of the hand and soles of the feet 9, 10.

Keratinocytes are the key cells of the epidermis, which synthesize the 30 different keratins and comprise approximately 80% of the epidermis. The epidermis can be further subdivided into five sublayers based on the differing stages of keratin maturation 2.

From bottom to top, the sublayers of epidermis are named: stratum basale, stratum spinosum, stratum granulosum, stratum lucidum, and stratum corneum 11 (Figure 1.1).

The progressive differentiation of cells of the stratum basale toward the stratum corneum is accompanied by a multitude of changes 1, 7, 9, 12. Keratinocytes divide through mitosis at the stratum basale and push already formed cells into higher layers. As the cells move up the sublayers, they change shape and composition and eventually die due to isolation from the blood supply. The changes of the cell are very diverse, but generally, the cell reorganizes and loses its organelles, increases its size and flattens itself, and the cytoplasm of the cell is released and is replaced by the keratin 13. Cells become more specialized and limited in their structures and functions through this differentiation. The

3

Figure 1.1 Schematic illustration of the human skin structure.

Human skin is composed of the epidermis and the dermis. The epidermis can be subdivided into five sublayers based on the differing stages of keratin maturation, including stratum basale (SB), stratum spinosum (SS), stratum granulosum (SG), stratum lucidum (not shown in this illustration), and stratum corneum (SC).

4 cells eventually reach the stratum corneum and slough off. This process is called keratinization, which takes place within weeks 14, 15. The keratinized layer of skin keeps water in the body and provides a high resistance to environmental insults, making skin a natural barrier to infection. This feature is associated with the numerous cell junctions of keratinocytes in the epidermis 16.

1.1.2.1.1 Stratum basale

The stratum basale (or basal layer) lies adjacent to the dermis and is composed mainly of keratinocytes (Figure 1.1), 10% of which are poorly differentiated 1, 9, 11, 17.

The stratum basale also contains other types of cells, such as the melanocytes, pigment

(melanin) producing cells, and Merkel cells 2. Merkel cells are closely associated with cutaneous nerves and are involved in light touch sensation 2. The basal cells are cube shaped, with a large basophilic nucleus 18. Three subpopulations of keratinocytes in the basal layer have been defined by cell kinetic analysis; they are 1) stem cells, 2) transit amplifying cells, and 3) post-mitotic differentiating cells 11, 17. The stem cells contain large amounts of melanosomes, which are transferred from the adjacent melanocytes, in order to protect their self-renewal ability from ultraviolet (UV) radiation 2, 19, 20. The stem cells express keratin (K) intermediate filaments, mainly K15 and K19, which can be used as markers for this subpopulation 21, 22. The mitosis of stem cells can regenerate other stem cells to maintain the stem cell compartment or to produce daughter cells to replenish the epidermis through differentiation. The transit amplifying cells are the more differentiated daughter cells, compared to their parental cells, which possess the ability to undergo several cycles of proliferation to generate a population of cells that will irreversibly lose their proliferating ability and differentiate.

5

Once differentiated, the keratinocytes move toward the upper layer, stratum spinosum, and keep differentiating from one layer to the other to the final layer, stratum corneum 7, 14. The keratinocytes in the basal layer mainly express K5 and K14, which organize tonofilaments around the nucleus 23-26. The keratinocytes in the basal layer use the keratin filaments connecting to hemidesmosomes to attach to the basement membrane and desmosomes to join the cells together 25-30. Other important molecules that are involved in the dermal-epidermal junction are 21 and 31 integrins 31-34. The integrin 21 binds laminin (1) and collagen (type I and IV) and the integrin 31 focuses on laminin (15) 32, 35-38. The localization of both integrins would explain their roles in association with the dermal-epidermal junction. Besides adhesion, the integrins also regulate the initiation of differentiation. When keratinocytes are maintained in suspension culture, they are forced to lose contact with the extracellular matrix (ECM) and consequently, they differentiate 39, 40. However, loss of contact with the ECM is not the initial signal that triggers cell differentiation since differentiated keratinocytes were observed under regular culture conditions, which does allow ECM-keratinocyte contact

41-43.

Since integrins are involved in several signaling pathways, such as FAK, Ras, ERK, and JNK pathways 44-47, a hypothesis was raised that the proportion occupied by integrin ligands can intervene early in the control of differentiation 48-50. In addition to keratin filaments and integrins, microfilaments, such as actin, myosin, and alpha-actinin, assist in upward migration of cells as they differentiate 9. Stratification is the consequence of keratinocyte differentiation; increased calcium concentration is required for the stratification and formation of desmosomes in cultured keratinocytes 7, 41-43, 51, 52.

6

Normal transit time for keratinocytes from the stratum basale to the stratum granulosum is approximately 14 days, and the cells survive an additional 14 days in the stratum corneum before desquamation.

1.1.2.1.2 Stratum spinosum

As we move up into the stratum spinosum (or spinous layer), which is named after the numerous desmosomal connections of adjacent keratinocytes 7, the morphology of the keratinocytes observed in this layer located closer to the stratum basale are polyhedral with rounder nuclei. The cells located more superficially are more differentiated with a more flattened appearance 9, 18. Functional organelles are still present in the keratinocytes of the stratum spinosum, although the cells are more differentiated than those in the basal layer 9. There is a remarkable activity of phagocytosis in this layer, which degrades foreign molecules in lysosomes or retains foreign molecules in phagosomes. The cells in the spinous layer are active in protein synthesis 53, 54, especially K1 and K10, which are markers of terminal differentiation 23, 24. K5 and K14 are still present, but not made de novo in this layer. These intermediate filaments are more abundant, better organized, and bind to desmosomes to provide mechanical strength

27-30, 55. although the spinous layer contains large amounts of adhesion proteins, there are numerous gap junctions which mediate intercellular communication 56. In the spinous layer, the keratinocytes begin to generate lamellar granules, whose primary activity is in the stratum corneum 56. The lamellar granules are membrane-bound intracellular structures, which are eventually shed into the extracellular milieu to discharge lipid precursors at the junction of the stratum granulosum and stratum corneum.

By doing so, the epidermis establishes a barrier to prevent water loss 56. The contents of

7 lamellar granules include: 1) neutral sugars combined with fats or proteins, 2) various lipids, and 3) acid hydrolases (lysosomal acid lipase) 9, 57, 58. Keratinocytes in the spinous layer then progress to the third layer, stratum granulosum, with the pressure of the underlying cells. It is also worth mentioning that the Langerhans cells, which play an important role in immune reactions as antigen-presenting cells, are mainly located in the middle of the stratum spinosum 2.

1.1.2.1.3 Stratum granulosum

Keratinocytes continue their transition as they travel superficially and form the stratum granulosum (or granular layer). The cells continue to flatten, lose their nuclei and organelles, and their cytoplasm contains large amounts of lamellar bodies and keratohyalin granules which make the cells appear granular, giving rise to the name of this layer 7. The components of lamellar granules and keratohyalin granules are of great importance and are ultimately responsible for the formation of the cornified envelope.

The function of the lamellar granules was described. The keratohyalin granules are aggregates of insoluble proteins including loricrin and profilaggrin 59, 60. Loricrin is a precursor of the cornified envelope and profilaggrin is the precursor of filaggrin, which closely associates with the keratinization of the stratum corneum by promoting hydration and crosslinking of keratin filaments 9, 12, 27, 59, 61, 62. The keratohyalin granules disappear in the cytoplasm during the transition of the stratum granulosum to the stratum corneum. Another important molecule responsible for the formation of the cornified envelope is involucrin, which begins to appear in the spinous layer and crosslinks in the granular layer by the transglutaminases 56 .

8

1.1.2.1.4 Stratum lucidum

The stratum lucidum, also called the clear layer due to the translucent appearance under a microscope, is found in areas of thick skin such as palms and soles 63. The clear layer is composed of three to five layers of dead, flattened keratinocytes, which do not have distinct boundaries 64, 65. The keratinocytes in this layer are filled with eleidin, a transparent substance formed from the keratohyalin and eventually transformed to keratin in the stratum corneum.

1.1.2.1.5 Stratum corneum

The loss of nuclei and organelles causes the keratinocytes in the granular layer to become non-viable corneocytes and form the stratum corneum. A bricks and mortar model has been used to describe the structure of the stratum corneum in which the corneocytes represent the bricks and the intercellular lipids embody the mortar 58. A corneocyte is a dead cell derived from a keratinocyte in the granular layer, which gradually loses its nucleus and organelles by enzymatic degradation 7, 9, 66. The plasma membrane of corneocytes becomes permeable and the nucleus is degraded 40, 66. The corneocytes then become dehydrated and lose approximately 45%-86% of their weight and the cytoplasm is filled with tiny threads of keratin in an organized matrix 7, 9, 66. The rearrangement of keratin cytoskeleton during keratinocyte maturation is done with the help of the filaggrin. At the transition between the stratum granulosum and stratum corneum, the filaggrin is activated by dephosphorylation and enzymatic cleavage from its precursor, profilaggrin 67. The processing of profilaggrin may be facilitated by calcium-dependent enzymes. At the beginning of the stratum corneum, the filaggrin serves as a guide and interacts with keratin filaments causing their aggregation into

9 macrofibrils 61, 68. The tightly packed keratin filaments allow extensive crosslinking of themselves by transglutaminases to form a highly insoluble keratin matrix. Along the journey to the upper stratum corneum, the filaggrin acidifies, loses its affinity to keratin filaments, and undergoes further processing to release free amino acids that bind to water molecules and thus allow the stratum corneum to retain some of its moisture 59, 69-71.

Though the corneocytes are consider as dead cells, they still have some active enzymes, such as acid hydrolases, which are involved in the remodeling of intercellular lipids 9, 58,

71, 72. The intercellular lipids, which fill the intercellular space, are mainly from the lamellar granules 73. When the keratinocytes move into the stratum corneum, the outer envelope of the lamellar granules are degraded by enzymes and release lipids including ceramides and free fatty acids 12, 56, 58, 74. The final intercellular lipids in the stratum corneum are cholesterol, cholesterol esters, long-chain fatty acids, and ceramides 56, 75.

Each corneocyte has a protective protein shell called the cornified envelope which is composed primarily of two proteins, involucrin and loricrin 7, 59, 76-82. The extensive covalent linkage between these proteins are mediated by transglutaminases, activated by increased intracellular calcium concentration 7, 83-90, which could come from the loss of membrane integrity during the transition from the stratum granulosum to the stratum corneum 40, 66, 91 or from the release of calcium while the profilaggrin is dephosphorylated and converted to filaggrin 67. The highly crosslinked proteins make the corneocyte impermeable and resistant to denaturing agents and make the cornified envelope the most insoluble structure of the corneocyte 7, 84, 92, 93. The specialized desmosomes that mediate corneocyte cohesion in the stratum corneum are called corneodesmosomes.

Before desquamation, detachment of cells from the surface of an epithelium, such as skin,

10 the corneocytes and the intercellular lipids work in concert to fulfill the barrier function of skin. The corneocytes are inert with strong cohesion provided by the corneodesmosomes, filled with water-retaining keratin filaments, and surrounded by a lipid matrix. The intercellular lipids and the amino acids from filaggrin allow the stratum corneum to retain its moisture and help to maintain several features, such as elasticity and consequently, the stratum corneum serves as an effective barrier 71.

Degradation of corneodesmosomes is an important step for the desquamation since they are the major structures that hold corneocytes together 75. The closer the corneocytes are to the surface, the more disruption of the keratin filaments, corneodesmosomes, and lipid matrix can be found 29, 71. Disorganization and disappearance of these complexes or structures eventually permit corneocytes to be shed at the surface 29.

1.1.2.2 The dermis

The dermis is the middle layer of skin that connects to epidermis by a basement membrane. It consists of connective tissue and can be divided into two histological regions: the papillary region and the reticular region 1, 2, 94. The dermis is composed of a tough, supportive cell matrix, mainly elastin and collagen 1, 2, 95, and three major types of cells, fibroblasts, macrophages, and adipocytes 2, 64, 65. The major functions of the dermis provide the skin with strength (generated by collagen) and elasticity (generated by elastin) 64, 65. It functions as a cushion to protect organisms from stress and strain. The dermis also contains nerves, such as the mechanoreceptors, which provide the sense of touch and heat, hair follicles, sweat glands, apocrine glands, sebaceous glands, lymphatic vessels and blood vessels 2. Two other important functions of dermis are ensured by the dermal blood vessels. Epidermis is supported by the dermal blood vessels, which

11 provide nutrients to and remove waste from the epidermis. The dermal blood vessels also contribute to regulation of body temperature by control of blood flow in the dermal blood vessels.

1.1.2.2.1 Papillary region

The papillary region is a superficial area adjacent to the epidermis and is composed of thin loosely arranged collagen fibers 2. This region contains numerous dermal papillae that extend toward the epidermis. The dermal papillae are small, fingerlike extensions that increase the surface area between the epidermis and dermis, which further increases the adhesion and interaction between these two layers 6, 64. The large surface area also helps the dermis to support the epidermis by increasing the exchange efficiency of nutrients, oxygen, and wastes 2, 5.

1.1.2.2.2 Reticular region

The reticular region is the deep, thicker area of the dermis. It is mainly composed of collagen and elastin 6, 64. Collagen makes up 70% of the dermis and runs in parallel strands to increase mechanical strength 2, 6, 64. The elasticity and flexibility of the dermis comes from the elastin fibers 2, 6, 64. Also located within the reticular region are the hair follicles, sweat glands, sebaceous glands, and blood vessels 2, 6, 64, 96.

1.1.2.2.3 Dermal cell population

The major cells in the dermis are fibroblasts, which produce elastin, collagen, and structural proteoglycans 2. In addition to fibroblasts, dermal cell populations can be divided into two types. The first are the cells that are associated with structures, such as cells in the hair follicles, blood vessels, glands, nerves, and adipose tissue. These type of cells generally do not move around the dermis 1. The second type are the cells that

12 move freely around the dermis like fibroblasts, lymphocytes, macrophages, mast cells,

Langerhans cells, and others 1, 14.

1.1.2.3 The subcutaneous tissue

The subcutaneous tissue is made of loose connective tissue, elastin and fat.

Fibroblasts, macrophages, and adipocytes are the main cell types found in the subcutaneous tissue. The function of the subcutaneous tissue is to attach the skin to underlying bone and muscle, but it is not part of the skin 2, 6, 64.

1.1.3 Difference between human and mouse skin

The most commonly used animal model for studying skin diseases by far is the mouse model. Mice are economical, easy to house, maintain, and genetically manipulate. The hundreds of mouse models of human diseases are powerful tools to understand skin biology 97-99. Although mouse models are widely used in skin studies, there are differences between mouse and human skin. Mouse skin is composed of three layers, which are epidermis, dermis, and hypodermis, just like human skin, but the anatomy and physiology of each layer within each organism are very different. The epidermis in mice is very thin, having approximately 2-3 layers of living cells (compared to 10-15 layers in human) and there is no stratum granulosum layer 100-102. The hair cycle of mice is approximately 3 weeks, but the hair cycles of humans are highly variable and anatomically-dependent; for example, the human scalp can survive for several years

103. Apocrine sweat glands that are present in the axilla, perianal, and inguinal skin regions of humans are only found in mammary glands in mice 104. Rete ridges, epidermal fingerlike projections that extend downward and interlock with dermal papillae,

13 are not obvious in healthy mouse skin, but can become apparent during wound healing in mice 105. Another difference is the panniculus carnosus, a subcutaneous thin muscle layer, which is only found in the platysma muscle of the neck, palmaris brevis in the hand, and the dartos muscle in the scrotum in humans, but is widely present in mice 73, 104.

The panniculus carnosus also mediates wound healing in mice via rapid wound contraction. On the other hand, angiogenesis, collagen deposition, granulation tissue formation, and re-epithelialization are the characteristics of human wound healing 106.

The substantial differences in wound healing between mice and human should cause us to rethink the use of mouse models in skin studies, when assessing the translational relevance to some physiologic and pathologic processes.

1.1.4 Importance of proteases, protease inhibitors, and proteolytic balance in skin

Epidermal barrier formation requires that keratinocyte differentiation occur in a precisely controlled manner. The processes of keratinocyte proliferation, differentiation and apoptosis must be balanced to lead to the formation of a structured and organized epithelium that provides efficient barrier function 7, 107. Many cellular and biochemical events are initiated at specific times and locations during epidermal differentiation, and the signaling processes involved in the initiation of these events have been extensively studied 108. Controlled proteolysis, which is also required for appropriate differentiation

109, 110, has, however, been understudied. Proteases are key regulatory molecules which activate growth factors, enzymes, membrane receptors, and ion channels, and also process enzyme co-factors, ECM components, structural proteins, and adhesion

14 molecules 109, 110. The importance of proteolysis in epidermal differentiation is emphasized by the many skin disorders that are caused by defects in proteolysis. The proteases implicated in these disorders include lysosomal proteases, cathepsins C, D and

L 111-114, two membrane-associated serine proteases, matriptase and prostasin 115-117, and caspase-14 118, 119. Mutations of the cathepsin C gene lead to Papillon-Lefevre syndrome, an autosomal recessive genetic disorder characterized by palmoplantar keratoderma and periodontitis 111, 113. The stratum corneum morphology is impaired in cathepsin D-deficient mice, which resembles the human skin disorder lamellar ichthyosis, and cathepsin D is involved in the regulation of transglutaminases 1 114. The cathepsin

L-deficient mice develop several skin disorders, including periodic hair loss and epidermal hyperplasia, acanthosis, and hyperkeratosis 112. Dysregulation of matriptase results in epidermal barrier loss leading to severe dehydration and neonatal death in matriptase-deficient mice 115. Matriptase is also required for the maintenance of epidermal homeostasis, since postnatal ablation compromises tight junction formation and barrier stability 115. Furthermore, matriptase mutation in humans is manifested by autosomal recessive ichthyosis and hypotrichosis (ARIH), where it is associated with compromised desquamation and abnormal filaggrin processing 120-122. Deficiency of prostasin, the downstream substrate of matriptase, also presents virtually identical skin disorders in mice and man. Mice present with severe malformation of the stratum corneum due to the disturbance of profilaggrin processing and an impaired skin barrier function, leading to a rapid and fatal dehydration 116.

In addition to proteases, protease inhibitors clearly also function in skin disorders.

These proteins are essential controllers that limit protease activity in both space and time.

15

These include the serine protease inhibitors, lymphoepithelial Kazal-type-related inhibitors (LEKTI) 123, hepatocyte growth factor activator inhibitor-1 (HAI-1) 124, and the lysosomal cysteine protease inhibitor, cystain M/E 125, 126. Premature-termination mutations in the SPINK5, encoding LEKTI, are associated with Netherton syndrome, a severe, autosomal recessive form of ichthyosis, indicating a critical role of SPINK5 in epidermal barrier function 123. Mice lacking the HAI-1/SPINT1 are embryonically lethal due to the impaired placental development 124. Hence, the role of HAI-1/SPINT1 in mice was tested by rescuing the placental function by microinjecting HAI-1-deficient murine embryonic stem cells into wild-type blastocysts to obtain high-chimerism embryos. The HAI-1/SPINT1 deficient mice developed scaly skin reminiscent of ichthyosis, hyperkeratinization, and abnormal hair shafts. The skin of these mice also showed altered proteolytic processing of profilaggrin 124. Deficiency of cystatin M/E leads to juvenile lethality and defects in epidermal cornification due to unrestricted legumain activity 125, 126.

The equally critical roles played by proteases and their inhibitors in skin underscore the importance of regulated proteolysis in epidermal development, differentiation, and homeostasis.

1.2 Proteases and protease inhibitors

1.2.1 Introduction

Proteases are enzymes that hydrolytically cleave peptide bonds and belong to the class of enzymes known as hydrolases. Based on their mechanistic type, proteases are generally grouped into six classes: serine, threonine, cysteine, aspartate, metallo-, and

16 glutamic acid proteases. They, alternatively, may be grouped by their optimal pH in which they are active: acid, neutral, and basic proteases. Most proteases are synthesized as inactive forms called zymogens, which exhibit much weaker proteolytic activity than their active counterparts, and are subsequently activated by proteolysis. The way they act on their substrates can be divided into two different categories: limited proteolysis, which breaks specific peptide bonds of a target protein, leading to the maturation or activation of the target protein, and unlimited proteolysis, which breaks down a complete peptide to individual amino acids. Proteases are involved in a great variety of metabolic processes and regulate one of the fastest “switching on” and “switching off” mechanisms of an organism. Proteases determine the longevity of other proteins involving various vital function of an organism. When the controls of proteolytic activity go wrong, enzymatically active proteases can cause severe harm to cells by activating or destroying proteins in undesirable ways. Proteases, being themselves proteins, are known to be activated by other proteases and inhibited by protease inhibitors, which are also proteins.

Protease inhibitors are molecules that are capable of inhibiting enzymatically active protease, and subsequently, the function of proteases. Most naturally occurring protease inhibitors are proteins and may be classified either by their mechanism of action, or by the type of protease they inhibit 127. The inhibitory mechanisms of protease inhibitors can be divided into two categories: irreversible trapping reaction or reversible tight binding interaction, which is also known as the key and lock mechanism 128, 129.

Protease inhibitors, like the serpin family and the alpha 2-macroglobulin family, use an irreversible trapping reaction to form covalent complexes between protease inhibitors and their target proteases to inhibit the activities of proteases 130. On the other hand,

17 protease inhibitors like Kunitz-type inhibitors and Kazal-type inhibitors utilize high-affinity, but reversible, interactions between protease inhibitors and the active site of their cognate proteases. The activation of proteases zymogens and inactivation by protease inhibitors are generally the two major mechanisms that govern proteolysis.

1.2.2 Serine proteases

Serine proteases perform a large number of biological functions. In mammals, these enzymes are involved in digesting food, fighting infections, helping blood to clot, and assisting sperm entrance into oocytes. The activity of serine proteases depends on nucleophilic attack of the targeted peptidic bond by a serine residue in the active site of the protease. While acting on a substrate, a serine protease first utilizes the serine residue of the serine protease domain to attack the peptide bond of the target substrate, releasing the new amino-terminus and forming an ester bond between the enzyme and the substrate called acyl enzyme intermediate. The ester bond is subsequently hydrolyzed and the new carboxyl-terminus is released.

Substantial progress has recently been made in characterizing members of a subgroup of serine proteases: the type II transmembrane serine proteases (TTSPs) 131, 132.

These trypsin-like (family S1) proteases are linked to cellular membranes via a hydrophobic stretch at the amino terminus; therefore, they represent enzymes whose peptide-bond cleaving activities may be specifically targeted to cellular membranes, in addition to the possibility of being shed into the extra-cellular milieu by the action of other proteases. Currently the TTSPs number 17 members and matriptase is part of this family.

1.2.3 Kunitz-type protease inhibitors

18

Kunitz-type inhibitors are protease inhibitors which utilize the Kunitz domain to inhibit the activity of their target proteases. They are characterized by basic isoelectric points, low molecular weights, and a broad spectrum of activity toward serine proteases

133, 134. Kunitz domains are relatively small, with a peptide chain of around 50 to 60 amino acid residues and have three disulfide bonds to stabilize their structures. The domain was first described in the bovine pancreatic trypsin inhibitor (BPTI)-like protease inhibitors; numerous other members have been described, including the two inhibitors for matriptase, HAI-1 and HAI-2. These inhibitors use high-affinity interactions to form complexes with their target proteases (key and lock mechanism) and thus can be dissociated by low pH or heat.

1.2.4 Serine protease inhibitors (Serpins)

Serpins were first identified as a set of proteins that are able to inhibit proteases. By far, serpins are the largest and most diverse family of protease inhibitors; more than a thousand serpins have been identified, including 36 human serpins 127. Serpins are structurally related, but functionally diverse, and they control essential proteolytic pathways in many branches of life. Unlike most classical protease inhibitors that use a simple key and lock mechanism to block access to the protease active site, serpins form covalent complexes with their target proteases and are therefore irreversible enzyme inhibitors 130. The specificity of a serpin is determined by an exposed region termed the reactive center loop (RCL) and it is this region that forms the initial interaction with the target protease; thus, the RCL of a serpin mimics a substrate for its target protease. The difference between a substrate and a serpin when they meet their cognate protease is that the serpin rapidly undergoes an unusual conformation change, termed the Stressed to

19

Relaxed (S to R) transition 135-138 which causes the hydrolysis of the acyl-enzyme intermediate to be extremely slow 139. The protease hence stays covalently linked to the serpin and is thereby inhibited. Furthermore, since the serpin remains covalently attached to the target protease, inhibition consumes the serpin as well. Since serpins use the RCL to determine the specificity toward their target proteases, serpins are vulnerable to mutations, especially mutations that result in protein misfolding and formation of inactive long-chain polymers called serpinopathies 140-142. Polymerization of serpins reduces the amount of active inhibitors and the accumulation of serpin polymers results in cell death and organ failure 142. The importance of serpins in humans can be demonstrated in approximately 100 human diseases, such as liver cirrhosis caused by polymerization of antitrypsin 142, thrombotic disease caused by antithrombin deficiency, and emphysema caused by antitrypsin deficiency 141, 143.

1.3 Matriptase-mediated cell surface proteolysis cascade

1.3.1 Matriptase cascade in skin

Among these proteases and anti-proteases involved in skin functions, we are particularly interested in a pericellular proteolytic cascade, initiated by matriptase and including other serine proteases such as urokinase-type plasminogen activator (uPA) and prostasin. The matriptase proteolytic cascade plays a critical role in epidermal barrier construction, as well as in the regenerative processes required for wound healing.

Regulation of this protease cascade is well studied in other simple epithelia, involving extracellular stimuli to induce matriptase zymogen activation and four endogenous protease inhibitors, HAI-1, antithrombin, alpha-1 antitrypsin (1AT), and

20 alpha-2-antiplasmin (2AP). However, matriptase regulatory mechanisms in skin remains poorly understood. To date, very little is known about the reciprocal regulations between epidermal differentiation and the control of the protease network activity; however, several lines of evidence indicate an important role for the protease network in normal differentiation and disease processes. First, deletion of key players in this network (matriptase, HAI-1, or prostasin) in mice causes severe barrier defects- leading to postnatal death 115, 116, 124. Second, autosomal recessive matriptase mutations are associated with skin permeability barrier defects in humans 120-122. Third, impaired profilaggrin processing, reduced desquamation, and compromised lipid matrix assembly are evident both in the above-mentioned mouse knockouts, as well as in human matriptase-deficiency disease 116, 117, 120. Fourth, imbalanced proteolysis within this protease network results in the development of epidermal disease 144-146. Given the clear role of the protease network in mediating and/or propagating epidermal disease, the fact that so little is known about proteolysis regulation in skin differentiation and repair is puzzling.

1.3.2 Matriptase

Matriptase is also known as membrane-type serine protease-1 (MT-SP1), tumor-associated differentially expressed gene-15 (TADG-15), and suppressor of tumorigenicity-14 (ST14) 147-149. In 1993, matriptase was first identified as a secreted gelatinase in conditioned medium from human breast cancer cells 150, and later as a transcript found in normal and cancerous colon tissue encoded by a gene located on chromosome 11q24-25 149, 151. Publication of the cDNA sequence of matriptase appeared shortly thereafter 147, 148, 152. Besides playing a role in normal cellular

21 physiology, matriptase may be involved in multiple pathological processes, including carcinogenesis and tumor progression.

1.3.2.1 Family, localization, and expression

Matriptase belongs to the type II transmembrane serine protease (TTSP) family and is expressed by both simple and stratified epithelia in many organ systems, including the skin, breast, lung, small intestine, colon, kidney, ovary, stomach, colon, and prostate 147,

148, 153-158. The widespread epithelial expression of matriptase is consistent with its essential roles in the formation and maintenance of epithelial integrity 145, 159. It is predominately expressed by the epithelial components of human tissues, by epithelial cell lines, and epithelial-derived cancer cell lines 147, 148, 150, 155, 156, 160. The protease is not expressed by fibroblasts present in the stromal component of tissues. For the most part, tissues lacking epithelial components do not show matriptase expression, nor do tumors or tumor cell lines derived from non-epithelial tissues, with the exception of leukemia and lymphoma cell lines 147, 156, 157. In addition to epithelial and carcinoma cells, blood leukocytes 147, 157, monocytes 147, 157, 161-163 and neutrophils 147, 157, chondrocytes 164, mast cells 165, and neural progenitor cells 166 also express matriptase.

1.3.2.2 Matriptase structure

Matriptase is a mosaic protein, containing 855 amino acids and its calculated molecular mass is 94.7-kDa (Figure 1.2). At the amino-terminal end, the protease has an intracellular domain that consists of 54 amino acids, which contains a consensus phosphorylation site for protein kinase C (PKC), followed by a transmembrane domain.

The cytoplasmic tail of the protease can interact with filamins and anchor the protease to the actin cytoskeleton 167. The extracellular stem region of matriptase consists of a

22

Figure 1.2 Schematic representation of the matriptase structure.

Matriptase is a mosaic protein. It has an intracellular domain at the amino-terminal end followed by a transmembrane domain. The extracellular stem region of matriptase consists of a single SEA domain, two CUB domains, four tandem LDLRA domains, and a conserved serine protease domain at the carboxyl-terminal end.

23 single sea urchin sperm protein, enteropeptidase, and agrin (SEA) domain, two tandem

Clr/s, Uegf, and Bone morphogenic protein-1 (CUB) domains, and four tandem low density lipoprotein receptor class A (LDLRA) domains (Figure 1.2). These noncatalytic modules appear to play an essential role in the cellular localization, activation, inhibition, and likely, the substrate specificity of matriptase. The SEA domain consists of about

120 amino acids 168 and has the ability to undergo a posttranslational self-cleavage 169 at

G-149, an event called N-terminal processing. This processing is critical for the maturation and zymogen activation of the protease 170. The two CUB domains and four

LDLRA domains play a critical role in the cellular localization, which targets the protease to the basolateral plasma membrane, and zymogen activation 171. The carboxyl-terminal serine protease domain is structurally similar to that of other TTSPs (Figure 1.2). It includes a conserved histidine/aspartic acid/serine (HDS) catalytic triad essential for proteolytic activity.

Besides those indicated above, four potential N-linked glycosylation sites are also found in the matriptase sequence. Among them, the Asn-722 and Asn-302 are important for matriptase zymogen activation 171. Another residue, Gly-827, in the serine protease domain of matriptase has been shown to have an important role in its activation 172. If mutated Gly-827 causes a rare human genetic skin disorder called autosomal recessive ichthyosis with hypotrichosis (ARIH) 120. A unique insertion loop, 60-loop insertion, was found in the serine protease domain of matriptase. This insertion contains nine residues that form a -hairpin loop which is rotated away from the active site, resulting in a more open cavity than in the case of thrombin, which also contains a similar loop 173.

This unique feature significantly influences the substrate specificity of matriptase.

24

Matriptase has approximate 20 predicted disulfide bonds. One of these disulfide bonds links the serine protease domain and the noncatalytic region which keeps these two parts together after the zymogen activation 174. This disulfide bond is very sensitive to oxidizing and/or reducing environments and is readily cleaved after the cells are lysed.

This cleavage can be prevented by adding 5,5’-dithiobis-(2-nitrobenzoic acid) (DTNB) or other oxidizing reagents to the lysis buffer 175.

1.3.2.3 Life cycle of matriptase

The regulation of matriptase is a complicated process. Matriptase is synthesized as a full-length zymogen with a size of around 95-kDa (Figure 1.3). The full-length matriptase is converted to its mature form via a cleavage at G-149 within the SEA domain yet it is still membrane-associated; this process is also called N-terminal processing and requires no enzyme involvement 169 (Figure 1.3). The activation of matriptase zymogen is by autoactivation 171, a mechanism which is only observed in the initiator proteases of protease cascades and is likely mediated by the intrinsic proteolytic activity of the zymogen. The matriptase zymogen appears to exhibit unusually high inherent proteolytic activity, which can be as much as 3% of that of active matriptase under optimal conditions at pH 6.0 176. The activation of matriptase follows a very classic scheme by cleavage at a canonical activation motif that converts the signal-chain zymogen to a disulfide-linked two-chain active enzyme. The two-chain active matriptase contains a noncatalytic region, 45-kDa, and a serine protease domain, 25-kDa

174. In most matriptase-expressing cells, the active matriptase must immediately act on its substrates since the active matriptase is rapidly inhibited by its inhibitors, such as

HAI-1 and antithrombin, to form complexes and shed into the extracellular lumen 177.

25

Figure 1.3 Schematic representation of the life cycle of matriptase.

Matriptase is synthesized as a full-length zymogen with a size of around 95-kDa. Then it undergoes a conversion of full-length matriptase to its mature form via a cleavage at

G-149 within the SEA domain, also called N-terminal processing. The activation of matriptase zymogen follows a very classic scheme by which a canonical activation motif is cleaved to convert the signal-chain zymogen to a disulfide-linked two-chain active enzyme. The active matriptase is immediately inhibited by its inhibitor, HAI-1, and shed into the extracellular lumen.

26

The shedding of activated matriptase complexes requires a proteolytic cleavage at either

189Lys-Ser190 or 204Lys-Thr205 174. In addition to the activated matriptase complexes, the latent form of matriptase, 70-kDa, is also shed to the extracellular milieu 160, 174.

1.3.2.4 Extracellular stimuli for matriptase activation

As an initiator protease at the top of protease cascade, the most critical step is the initiation of activation of matriptase zymogen. Activation of the initiator protease not only brings matriptase to life, but also switches on the activity of downstream proteases, substrates; those physiological functions are controlled by the matriptase-driven protease cascade. As an initiator protease, matriptase activity is regulated by a number of cell-type specific extracellular stimuli, non-protease factors, on the cell surface.

Sphingosine-1-phosphate (S1P) is the first physiological extracellular stimulus identified from fetal bovine serum which induced matriptase activation on the surface of immortalized breast epithelial cells 174. Though a wide spectrum of biological functions can be induced by S1P in a variety of cell types, e.g., survival, actin cytoskeletal rearrangement, proliferation, cell-shape changes, assembly of adherens junctions, cellular motility, the inhibition of gap junction communication, and contraction 178-180, the functional linkage of S1P and matriptase zymogen activation is presently unknown. The potent cellular effects of S1P are exerted by direct binding of the lysophospholipids to a subfamily of cell surface G-protein coupled receptors termed the endothelial differentiation gene (EDG) receptors, which were renamed S1P receptors in 2002 181.

The relative levels of expression of EDG/S1P receptors and the distinct expression pattern of the EDG/S1P receptors vary in different cell types and are the key factors regulating the power of S1P. How and whether matriptase activation requires EDG/S1P

27 receptors and their downstream signaling remain unknown, but the mechanism apparently does not involve direct binding of S1P to the protease 182. In contrast to the immortalized breast epithelial cells, breast cancer cells have apparently lost

S1P-mediated matriptase activation 183. Cancer cells appear to constitutively activate matriptase and which is shed into conditioned media 183, indicating that malignant epithelial cells have lost the physiological mechanism controlling matriptase activation.

The second identified inducer for matriptase activation is suramin, a sulfide-rich, anionic small molecule 175. The discovery of suramin was unexpected as it was originally used to uncouple the interaction between S1P and S1P3. 175. Suramin alone induces robust matriptase activation, which is more rapid and extensive than induced by S1P 175.

The ability of suramin to induce matriptase activation in several matriptase-expressing epithelial and carcinoma cells distinguishes it from S1P and steroid hormones (see below), which are cell-type specific. The other feature that makes suramin different from S1P and steroid hormones is the subcellular localization of matriptase activation. In

S1P-treated 184 A1N4 mammary epithelial cells, activated matriptase was detected at cell junctions; however, the vesicle-like, intracellular structures in the suramin-treated cells are where activated matriptase is found 175.

To understand the irregular matriptase activation in cancer cells, in contrast to the normal physiological S1P-mediated matriptase activation, our lab endeavored to identify the extracellular stimulators responsible for the dysregulation. Androgens, we identified as a third inducer for matriptase activation in LNCaP prostate cancer cells 184. In hormone-dependent LNCaP cells, androgens play a key role in the activation of matriptase, which is similar to the dependence on S1P for matriptase activation in 184

28

A1N4 cells. Significant suppression of matriptase zymogen activation was seen in hormone-starved LNCaP cells without alteration of the expression of matriptase mRNA

184. This phenomenon was also observed in serum-starved 184 A1N4 cells. Like S1P, the robust matriptase activation induced by androgen is through its receptor, the androgen receptor (AR). However, the time requirement for the two extracellular stimuli to mediate matriptase activation is significantly different. Matriptase activation induced by

S1P through the EDG/S1P receptors is rapid, likely within 10 min; however, androgen-induced matriptase activation is dramatically slower taking at least 6 hours 184.

Though the androgens also regulate several proteases in prostate cancer cells, and include protease/KLK-L1 185, prostate specific antigen (PSA), glandular kallikrein 186, TMPRSS2

187, 188, MMP-2 189, and a protease with high homology with enamel matrix serine proteinase 1 190, the major effects of androgen treatment are attributed to the altered expression of these proteases rather than to the level of zymogen activation. Despite the conspicuous hormone dependence of matriptase activation in prostate cancer cells, hormone-starvation treatment in hormone-dependent breast cancer cells did not alter matriptase activation 184. The striking difference between the prostate and breast cancer cells may result from the high basal level of matriptase activation in breast cancer cells.

1.3.2.5 Activation machinery

The structural and mechanistic diversity of the stimuli of matriptase zymogen activation indicates that there might be an activation machinery that carries out the cleavage of activation in response to these diverse extracellular stimuli. The hypothesis of the presence of activation machinery is initially proposed based on the observation that activation of matriptase induced by S1P and suramin takes place on the cell-cell junctions

29 and intracellular membrane structures, respectively 175, 191. The activation appears where matriptase zymogen accumulates and then propagates along the cell-cell junctions.

The activation machinery can be isolated by homogenizing the cells and collecting the membrane fractions 192. Using these isolated membrane fractions, matriptase activation can be characterized in a cell-free setting. Several interesting features of matriptase activation have been defined using this in vitro, cell-free system. These features include

1) the requirement of lipid bilayer biomembrane 192, 2) the regulation by the chemical environments such as pH and salts 177, 193, 3) the presence of cytosolic suppressor, and 4) the dependence of the interactions of matriptase molecules and other yet to be identified proteins, likely via the noncatalytic domains of matriptase. These biochemical features of matriptase activation can also be seen in intact cells.

1.3.2.6 Matriptase functions

Matriptase is found in the epithelium of all organ systems examined 157, 194, suggesting a role of the protease in common architectural features or functions in all epithelial tissues. Matriptase is indeed required for the maintenance of epithelial integrity in multiple organ systems, as demonstrated by mice with postnatal ablation of matriptase in the epithelial tissues, and include increased transepithelial permeability, loss of tight junction function, and generalized epithelial demise, leading to severe organ dysfunction 159. The requirement of matriptase has also been seen in epidermal barrier function, hair follicle development, and thymic homeostasis in matriptase knockout mice

115, indicating the impact of matriptase dysregulation in specialized epithelial tissues.

Aberrant processing of profilaggrins during epidermal terminal differentiation, which is responsible for the skin barrier function, is the cause of many of the epidermal defects

30 seen in mice with targeted deletion of matriptase 117. The importance of matriptase in epidermal differentiation and barrier function are also found in humans. Four types of matriptase mutations have been found in humans. Instead of postnatal death seen in matriptase knockout mice, patients with matriptase mutations survive yet exhibit ichthyosis, hypotrichosis, and follicular atrophoderma 120-122. The phenotypes seen in matriptase mutation patients are similar to those seen in matriptase hypomorphic mice 195.

Besides loss of barrier function, increased matriptase activation has been seen in several distinct skin diseases of humans, and in keratinocytes close to inflammation areas 196. In addition to epidermal homeostasis, matriptase has been shown to play an important role during neural tube closure 197.

The correlation of matriptase and cancer development and progression has been shown by several groups including ours. When matriptase was overexpressed in gastric cancer cells and grafted to nude mice, a significant increase in metastasis of the cells was observed, resulting in enhanced colonization of the lymph nodes 198. The potent oncogenic activity of matriptase was demonstrated using a transgenic mouse model in which mild overexpression of matriptase is targeted to the skin. All the mice developed spontaneous squamous cell carcinoma and became very susceptible to carcinogen-induced tumor formation 144. Interestingly, the oncogenic activity of matriptase is completely negated when HAI-1, the endogenous inhibitor of matriptase, is simultaneously overexpressed in these transgenic mice 144. The malignant epithelial transformation by dysregulated matriptase expression and the rescue by HAI-1 suggest that it is not the levels of matriptase, but the ratio of matriptase to HAI-1 that is likely to be the driving force that converts this physiological membrane protease into an oncogenic

31 protease. The imbalance between matriptase and HAI-1, and increased matriptase expression are both commonly seen in a wide variety of primary human carcinomas 199,

200 and these dysregulations are associated with poor patient outcomes 201-203.

Surprisingly, reduced matriptase expression has been shown in later stages of breast cancer and is associated with poor patient outcomes as well 204. The reduced level of matriptase in advanced cancer is consistent with the downregulation of the protease during the epithelial to mesenchymal transition (EMT) 205, 206. Moreover, downregulation of HAI-1 is also observed in advanced ovarian cancer 203 providing additional support for the hypotheses that is the ratio of matriptase to HAI-1 that may cause a more significant role in cancer rather than the absolute expression levels of matriptase.

The expression of matriptase in chondrocytes 164, monocytes 161-163, mast cells 165, and neural progenitor cells 166, suggests that this membrane protease may function in a variety of distinct physiological processes beyond the maintenance of epithelial integrity and function. Matriptase activity in condrocytes may contribute to osteoarthritis by initiating the degradation of cartilage matrix 164. Matriptase in monocytes might induce pro-inflammatory cytokines secreted by endothelial cells, therefore; contributing to atherosclerotic lesions 161. Expression of matriptase in mast cells which are involved in many allergy-related diseases, such as asthma, suggests that matriptase has the potential to contribute to these diseases.

1.3.3 Matriptase substrates

The preferred substrate cleavage sequence of matriptase has been identified in detail by use of the purified recombinant matriptase serine protease domain and a P1 lysine

32 positional scanning synthetic combinatorial library and substrate phage display 207. The preferred cleavage sequences are found to be basic P1 residues (Lys or Arg), small residues (Ser or Gly) or Phe at P2, basic residues or Gln at P3, basic residues at P4, and

Ala at the P1’ position. Interestingly, sequences with basic residues at P3 and P4 simultaneously, are less favored, suggesting that the substrate binding pocket contains only one interaction surface for a positively charged side chain and does not accommodate two such residues at P3 and P4. Indeed, the crystal structure of the serine protease domain of matriptase showed a trypsin-like substrate binding pocket (preferring basic residues), a small hydrophobic S2 site, and only one negatively charged surface at

S4 that is predicted to bind the basic P3 or P4 site 173. The revealed structure suggests a substrate cleavage preference of matriptase with a basic P1 residue, a small hydrophobic

P2 residue and a basic P3/P4 residue, which matches the results obtained from the P1 positional scanning synthetic combinatorial library and substrate phage display.

Eight natural substrates of matriptase have been identified to date, including hepatocyte growth factor (HGF) 208, urokinase-type plasminogen activator (uPA) 207, matrix metalloprotease 3 (stromelysin) 209, protease-activated receptor-2 (PAR-2) 207, prostasin 210, macrophage stimulating protein 1 (MSP-1) 211, epithelial sodium channels

(ENaC) 212, and acid-sensing ion channel 1 (ASIC1) 213. Matriptase was also reported to digest fibronectin and laminin 154. No cleavage has been reported in plasminogen,

PAR-1, PAR-3, and PAR-4 which are highly similar structurally to the matriptase substrates HGF and PAR-2 207, 208. The substrates cleaved by matriptase play a role in cellular growth, cell motility, cell survival, extracellular matrix (ECM) remodeling, and skin barrier function. Matriptase has been shown to activate HGF, which initiates cell

33 migration, proliferation, and morphogenesis that is tightly coordinated in the developing embryo and fetus. Matriptase may also play a role in HGF activation that is implicated in the development and cyclic remodeling of adult structures such as the mammary gland.

Besides HGF, matriptase can also act on uPA to initiate the uPA/plasmin cascade which may play a key role in organ development and matrix remodeling. The matrix remodeling could be further encouraged by the activation of HGF, followed by the upregulation of both uPA and its receptor, uPAR 214, 215. As an activator of HGF and uPA, matriptase could contribute to cancer development and progression 216-218.

Matriptase could also contribute to cellular invasion and cell-substratum adhesion by activating matrix metalloprotease 3 (stromelysin), 209 and by degrading components of extracellular matrix 154. Activation of the G protein-coupled receptor, PAR-2 by matriptase may be important for neural tube closure 197 and in atherosclerosis 161. The expression of PAR-2 in epithelial components of human tissues closely overlaps that of matriptase and HAI-1 156, 219, 220, and the functions of PAR-2 activation are as diverse as exocrine secretion, mitogenesis, prostaglandin release, ion channel activation, and phagocytosis 221-229. PAR-2 activation could result in K+ and Ca2+ channel activation and increase prostaglandin release, both of which can lead to increased intestinal transit

230. Several pieces of evidence suggest that the role of matriptase in epidermal barrier function is likely mediated through the activation of prostasin, a glycosylphosphatidylinositol (GPI)-anchored serine protease: 1) The epidermal defects seen in both matriptase and prostasin knockout mice are almost identical 115, 116, 2) Both proteins are co-expressed in the transient layer of the skin 116, and 3) Prostasin activation is undetectable in the skin of matriptase knockout mice 210. Matriptase may also play a

34 role in macrophage activation through the activation of macrophage stimulating protein 1

(MSP-1) 211. In addition, matriptase can regulate ion homeostasis by activating epithelial sodium channels (ENaC) 212 either directly, or by activating prostasin, and acid-sensing ion channel 1 (ASIC1) 213.

1.3.4 Matriptase inactivation

The two major mechanisms that control proteolytic activity of a given protease, in general, are the activation of protease zymogen and the inhibition of active protease.

While the control of matriptase activity also follows this canonical mechanism, one of the most interesting features of matriptase regulation is that both the zymogen activation and the inhibition are closely coupled. To date, five endogenous inhibitors of matriptase have been identified. These include hepatocyte growth factor inhibitor-1 (HAI-1) 155,

HAI-2 231, antithrombin, alpha-1 antitrypsin (1AT), and alpha-2-antiplasmin (2AP) 163.

Among these five inhibitors, HAI-1, the major inhibitor of matriptase, is a membrane-bound, Kunitz-type serine protease inhibitor and is coexpressed and colocalized with matriptase in many epithelial cells 203. HAI-1 is not only an inhibitor of matriptase but also participates in the expression, intracellular trafficking, and zymogen activation of matriptase 171, 175. Besides HAI-1, HAI-2, other Kunitz-type serine protease inhibitors, have been reported to play a role in the regulation of matriptase

231. Antithrombin, 1AT, and 2AP have recently joined HAI-1 in the role of inhibiting matriptase 163. These serine protease inhibitors (serpins) suicidally form irreversible complexes with matriptase through a covalent bond; on the other hand, formation of the matriptase/HAI-1 and matriptase/HAI-2 complexes is reversible.

1.3.4.1 Kunitz-type serine proteases – HAI-1 and HAI-2

35

HAI-1 was originally identified as a strong inhibitor of HGF activator (HGFA) 232.

This inhibitor encoded by the SPINT1 gene locates on chromosome 15q15 and contains

11 exons and 10 introns 233. The messenger RNA (mRNA) of HAI-1 encodes a protein of 513 amino acids. The mature HAI-1, after removal of a 35 amino acid signal peptide, contains 478 amino acids with a calculated molecular mass of 53-kDa 232. The membrane-anchored HAI-1 consists of a cytoplasmic tail with 41 amino acid residues at its carboxyl terminus, followed by a transmembrane domain (Figure 1.4). The extracellular stem region of HAI-1 contains two Kunitz-type serine protease inhibitory domains (Kunitz domain I and II) of approximately 60 amino acids each, both of which are separated by a LDL-receptor-like domain (Figure 1.4). The motif at the N-terminus with eight cysteines (MANEC, formerly called MANSC 234) was recently identified as a new domain for HAI-1. Three putative N-linked glycosylation sites have been found in

HAI-1 232.

Besides being anchored on the plasma membrane, as a 55-kDa form on the cell membrane, HAI-1 can be released as two soluble forms with sizes of 50- and 40-kDa by proteolytic cleavages 235. HAI-1 can be detected either in its unbound, free form or in complexes with activated matriptase and prostasin 155, 235.

HAI-1-mediated inhibition is closely related to matriptase activation induced by several stimuli including S1P, suramin, androgen, and acid 175, 177, 182, 184. The paradoxical requirement of HAI-1 in matriptase activation can be demonstrated by 1) the ablation of matriptase activation when matriptase and a HAI-1 mutant at the LDL receptor class A domain are coexpressed 171, and 2) the translocation and accumulation of

HAI-1 are along with matriptase at cell-cell junctions or in the intracellular vesicle-like

36

Figure 1.4 Schematic representation of the HAI-1 structure.

The membrane-anchored HAI-1 consists of a cytoplasmic tail at its carboxyl terminus, followed by a transmembrane domain. The extracellular stem region of HAI-1 contains two Kunitz-type serine protease inhibitory domains, both of which are separated by a

LDLRA domain. A MANSC (MANEC) domain was recently identified as a new domain for HAI-1 at its amino terminus.

37 structures during matriptase activation 175. Once activated, matriptase must rapidly act on its substrates as the active matriptase will be quickly inhibited by HAI-1 or other matriptase inhibitors. The interaction between matriptase zymogen and HAI-1 is very weak; therefore no stable complex can be detected by immunoblot. In contrast to matriptase zymogen, HAI-1 forms a very stable complex with active matriptase.

Matriptase-HAI-1 complexes are stable in SDS sample buffer in the absence of reducing agents and heating and can be detected by immunoblot 175. The matriptase/HAI-1 complex is initially detected as a 120-kDa complex in cell lysates, which contains a

70-kDa activated matriptase and the 55-kDa mature, membrane-bound HAI-1 175. The ectodomains of the matriptase/HAI-1 complex are shed subsequently from cell surface into the extracellular milieu 174, 184.

While HAI-1 is a well established inhibitor of matriptase, HAI-2, encoded by the

SPINT2 gene, has also been reported as an inhibitor for matriptase and can form

SDS-stable complexes with activated matriptase 231. Although HAI-2 can block matriptase-dependent activation of pro-prostasin and cell surface-bound pro-uPA, and its expression pattern perfectly matches that of matriptase and HAI-1, HAI-2 is usually considered to be redundant or partially redundant with HAI-1 231. However, a recent study has shown an essential function of HAI-2-regulated inhibition of matriptase during tissue morphogenesis in mice with genetic inactivation of the mouse SPINT2 gene 236.

The HAI-2 deficiency-associated defects include abnormal placental labyrinth development associated with loss of epithelial cell polarity, embryonic demise, and neural tube closure. These defects, except the neural tube closure, can be restored by simultaneously genetically inactivating matriptase, suggesting the defects are caused by

38 the dysregulated matriptase activity 236.

1.3.4.2 Serine protease inhibitors (Serpins) – Antithrombin, alpha-1 antitrypsin (1AT), and alpha-2-antiplasmin (2AP)

In addition to the two Kunitz-type inhibitors, HAI-1 and HAI-2, several endogenous serine protease inhibitors (serpins) of matriptase have been identified and include antithrombin, alpha-1 antitrypsin (1AT), and alpha-2-antiplasmin (2AP) 163. These serpins are isolated and identified in their complex forms with activated matriptase from human milk 163.

1.3.4.2.1 Antithrombin

Antithrombin and 1AT are two of the first members of the serpin superfamily to be extensively studied. The major functions of antithrombin and 1AT are to control blood coagulation and inflammation, respectively.

Antithrombin is a small glycoprotein consisting of 432 amino acids. It is produced by the liver and secreted into blood to inactivate several enzymes of the coagulation system. Antithrombin has a half-life of about 3 days and high concentration at approximately 2.3 M in blood plasma 237, 238. Three disulfide bonds, Cys8-Cys128,

Cys21-Cys95, and Cys248-Cys430, and four possible glycosylation sites, N96, N135,

N155, and N192, have been reported in antithrombin. Two types of antithrombin are found in blood plasma, with alpha-antithrombin the dominant form and beta-antithrombin the minor form. The four glycosylation sites of alpha-antithrombin have an oligosaccharide occupying each of the sites by covalent linkage 239. However,

39 beta-antithrombin consistently has one single glycosylation site, N135, which remains unoccupied 240. The activity of antithrombin is increased significantly when it binds to heparin.

The inactivation rate of the physiological target proteases, like thrombin, Factor IXa and Factor Xa, of antithrombin increases 4000, one million and 1000 fold, respectively, in the presence of heparin 241-245. The secret power of heparin to enhance the inactivation rate of antithrombin lies on a specific pentasaccharide sulfation sequence contained within the heparin polymer,

GlcNAc/NS(6S)-GlcA-GlcNS(3S,6S)-IdoA(2S)-GlcNS(6S). Two distinct mechanisms are used by antithrombin upon binding to the pentasaccharide sulfation sequence of heparin to increase inhibition of protease activity: 1) allosteric activation which involves a conformational change within antithrombin 246, and 2) non-allosteric activation which forms a ternary complex between antithrombin and the target protease 247. The conformational change, allosteric activation, within antithrombin in response to pentasaccharide sulfation binding is required to increase inhibition of Factor IXa and

Factor Xa 248-250. When bound to the pentasaccharide sulfation sequence, the amino acid residues P14 and P15, located within the N-terminal of the RCL (also called hinge region), expel themselves from the main body of the protein. Preventing this conformational change abolishes the increased Factor IXa and Factor Xa inhibition

246. Hence, the pentasaccharide-bound form of antithrombin is a more effective inhibitor of Factor IXa and Factor Xa. In addition to the pentasaccharide sulfation sequence, increased thrombin inhibition requires at least an additional 13 monomeric units, due to a requirement that thrombin and antithrombin must be adjacent to each other on the same

40 heparin 251. Therefore, binding of both protease and antithrombin onto heparin dramatically accelerates the interaction between the two parties. In the absence of the pentasaccharide, only 0.25% antithrombin molecules are in the active conformation

246. Therefore, both mechanisms of activation of antithrombin are important to greatly and locally activate antithrombin, in order to perform its inhibitory function.

1.3.4.3 Other inhibitors

Matriptase activity can be also inhibited by naturally occurring protease inhibitors, such as aprotinin, soybean trypsin inhibitor, Bowman-Birk inhibitor, lima bean trypsin inhibitor and sunflower trypsin inhibitor (SFTI-1) 252, 253. Several matriptase catalytic inhibitors have also been developed, including small molecule and peptide-based inhibitors. These inhibitors exhibit great potency against matriptase activity using in vitro assays that, in most cases, have made use of the recombinant matriptase serine protease domain 216, 253-257. Antibody-based inhibitors specifically targeted against matriptase have also been developed 258.

1.4 Hypotheses

We hypothesize that the matriptase-initiated protease network plays essential roles in human skin function via regulation of proliferation and differentiation of keratinocytes.

1.5 Research Objectives

Our research objectives are 1) to define the functional relationship between matriptase and prostasin in human keratinocytes, 2) to elucidate the physiological function of matriptase in human skin, 3) to identify and characterize the keratinocyte-specific stimuli

41 that induce matriptase activation, and 4) to determine how antithrombin regulates pericellular matriptase activity.

42

Chapter 2 – Regulation of the Matriptase-Prostasin Cell Surface

Proteolytic Cascade by Hepatocyte Growth Factor Activator

Inhibitor-1 during Epidermal Differentiation

43

2.1 Abstract

Matriptase, a membrane-tethered serine protease, plays essential roles in epidermal differentiation and barrier function, largely mediated via its activation of prostasin, a glycosylphosphatidylinositol-anchored serine protease. Matriptase activity is tightly regulated by its inhibitor hepatocyte growth factor activator inhibitor-1 (HAI-1) such that free active matriptase is only briefly available to act on its substrates. In the current study we provide evidence for how matriptase activates prostasin under this tight control by HAI-1. When primary human keratinocytes are induced to differentiate in a skin organotypic culture model, both matriptase and prostasin are constitutively activated and then inhibited by HAI-1. These processes also occur in HaCaT human keratinocytes when matriptase activation is induced by exposure of the cells to a pH 6.0 buffer. Using this acid-inducible activation system we demonstrate that prostasin activation is suppressed by matriptase knockdown and by blocking matriptase activation with sodium chloride, suggesting that prostasin activation is dependent on matriptase in this system. Kinetics studies further reveal that the timing of autoactivation of matriptase, prostasin activation, and inhibition of both enzymes by HAI-1 binding are closely correlated. These data suggest that, during epidermal differentiation, the matriptase-prostasin proteolytic cascade is tightly regulated by two mechanisms: 1) prostasin activation temporally coupled to matriptase autoactivation and 2) HAI-1 rapidly inhibiting not only active matriptase but also active prostasin, resulting in an extremely brief window of opportunity for both active matriptase and active prostasin to act on their substrates.

44

2.2 Introduction

Matriptase, a type II transmembrane serine protease, plays essential roles in epidermal differentiation and barrier function 115, 117. Mutations of matriptase are associated with autosomal recessive ichthyosis and hypotrichosis (ARIH) 120. Four mutations have been identified in ST14, the gene that encodes for matriptase 120-122.

Targeted deletion of matriptase in mice also causes severe epidermal defects with impairment of desquamation, lipid matrix formation, and profilaggrin processing.

These defects lead to compromised epidermal barrier function and postnatal death 117.

The severe epidermal defects appear to result from the lack of activation of the

GPI-anchored serine protease prostasin by matriptase, since both matriptase-deficient and prostasin-deficient mice share an almost identical pattern of epidermal defects.

Furthermore, lack of prostasin activation was observed in matriptase-deficient mouse skin 116, 210, suggesting that, much of the function of matriptase in the epidermis is manifested through the activation of prostasin.

Hepatocyte growth factor activator inhibitor-1 (HAI-1) is a transmembrane

Kunitz-type serine protease inhibitor 232. HAI-1 can inhibit matriptase in a competitive and reversible manner, consistent with the characteristics of Kunitz-type serine protease inhibitors 174. What is unusual, however, is its high efficiency of matriptase inhibition by HAI-1. As a consequence of this efficiency, free active matriptase is rapidly inactivated by HAI-1 binding and can only act on its downstream substrates for extremely short time periods after activation 175, 192, 259. This unusually efficient inhibition results not only from the high inhibitory potency of HAI-1 260 and the subcellular co-localization of enzyme and inhibitor, 259 but also from several uncommon

45 biochemical mechanisms. For example, most matriptase-expressing epithelial cells express HAI-1 in significantly higher levels compared to the levels of matriptase, typically in excess of 10-fold molar excess of the former 261. In the absence of HAI-1, newly synthesized matriptase accumulates within the ER/Golgi which results in very low levels of matriptase being expressed 259. HAI-1 may also participate in matriptase zymogen activation, since matriptase activation is significantly reduced when point mutations are introduced into the LDL receptor class A domain of HAI-1, or this domain is deleted 171. This unusual relationship between matriptase and HAI-1 results in two unique consequences: 1) the very short-lived availability of free active matriptase, due to the direct and rapid access of HAI-1 to the newly generated active matriptase, and 2) a paradoxical reduction, rather than increase in matriptase activity in the absence of HAI-1.

The short-lived availability of active matriptase suggests that activation of the downstream matriptase substrate must be coupled with matriptase’s own activation. The paradoxical decrease in matriptase function associated with HAI-1 deficiency appears to be consistent with animal studies in which the epidermal defects seen in HAI-1-deficient mice resemble those in matriptase hypomorphic mice 124, 195.

Although data from genetic manipulation models and studies of tissue distribution clearly demonstrate a close functional linkage between matriptase and prostasin and their essential roles in epidermal function, the details of how and when the activation of prostasin by active matriptase occurs, remain largely unknown. In the current study, we examine the expression and activation state of matriptase and prostasin during epidermal differentiation using an organotypic skin culture model, and dissect the molecular events associated with the activation and inactivation of these enzymes. Our results

46 demonstrate that matriptase and prostasin are rapidly and constitutively activated throughout the course of epidermal differentiation, consistent with their roles in skin function. More importantly, although HAI-1 rapidly inhibits active matriptase, immediately after matriptase zymogen activation, the active matriptase apparently has no difficulty activating prostasin in the brief window of opportunity, prior to its inactivation by HAI-1. Unexpectedly, active prostasin appears to follow the same fate as active matriptase, and is rapidly inactivated by binding with HAI-1. These data suggest that epidermal keratinocytes regulate matriptase-initiated cell surface proteolysis with unprecedented tight control to safeguard themselves from the proteolytic activities of matriptase and prostasin.

2.3 Materials and Methods

2.3.1 Antibodies

Two types of matriptase antibodies were used: the mouse monoclonal antibodies

(mAbs) M32 and M24 and rat mAb 21-9 recognize both latent and activated forms of matriptase 155, 160, 177; mAb M69 recognizes an epitope present only on activated matriptase and is, therefore, able to distinguish activated matriptase from latent matriptase 174, 182. Human HAI-1 protein was detected using the HAI-1 mAb M19 155.

Two prostasin antibodies were used: A prostasin polyclonal antibody, described previously 262, 263, which is able to recognize both activated prostasin in complex with

HAI-1 and uncomplexed prostasin, which is presumably the latent form of the enzyme.

Interestingly this antibody appears to recognize the prostasin-HAI-1 complex better than latent prostasin. The second antibody was a prostasin mAb (Becton, Dickinson and

47

Company, Franklin Lakes, NJ) that only recognizes prostasin after heat-denaturing treatment which results in the dissociation of the prostasin-HAI-1 complex.

2.3.2 Lentiviral particle preparation and cell transduction for the delivery of small hairpin RNAs

Matriptase-targeting and non-targeting control small hairpin RNAs (shRNA) in the vector PLKO.1-puro were purchased from Open Biosystems (Huntsville, AL). The shRNAs constructs were packaged into lentiviral particles using 293T cells using standard methods. Briefly, on day 1, 293T cells were switched to fresh media without antibiotics 3 hours prior to transfection. Ten grams of the shRNA plasmids were mixed with 5 g of VSV-G, 5 g of d89, 50 l CaCl2 and water to a total volume of 500 l.

The mixture was then mixed with 500 l of 2X Hebs buffer and incubated at room temperature for 20 minutes. The resultant mixture was then added dropwise onto a

100-mm dish of 293T cells, and the cells incubated overnight at 37℃, with 5% CO2.

The following day, the transfection media were replaced with fresh media and the cultures again incubated overnight. On the third day, conditioned media were harvested from the plates and filtered through a 22 m PES syringe driven filter unit.

2.3.3 Cell culture and media

The immortalized human keratinocyte line HaCaT was cultured in DMEM (Cellgro,

Manassas, VA) supplemented with 10% FBS (Gemini, West Sacramento, CA), 100 U/ml penicillin, and 100 g/ml streptomycin. Primary human keratinocytes isolated from neonatal foreskin tissue were cultured in keratinocyte serum free media (KSFM, Gibco,

Invitrogen, Carlsbad, CA) supplemented with 25 g/ml bovine pituitary extract and 1.5 ng/ml of recombinant epidermal growth factor. Cultures were maintained in a

48 humidified 5% CO2 atmosphere at 37℃.

2.3.4 Organotypic culture

Organotypic raft cultures were generated based on the method described previously

264, 265. Collagen (type I rat tail; BD Biosciences, Bedford, MA) plugs with 105 J2-3T3 fibroblasts per ml were prepared in 6-well cell culture inserts (0.4 m pore size, PET track-etched membrane; Becton Dickinson, Franklin Lakes, NJ), and incubated for 1 day.

A total of 106 primary human keratinocytes were plated on top of each plug in KSFM supplemented with 25 g/ml bovine pituitary extract and 1.5 ng/ml of recombinant epidermal growth factor. After 2 days, primary human keratinocyte differentiation was initiated by lifting the organotypic cultures to the air-liquid interface, and changing the culture medium to KSFM supplemented with 25 g/ml bovine pituitary extract, 1.5 ng/ml of recombinant epidermal growth factor. The medium was changed every second day for the next 8 days.

2.3.5 Acid-induced matriptase activation in HaCaT cells

Matriptase activation was induced by exposing HaCaT cells to a 0.15 M phosphate buffer (PB), pH 6.0 for 30 minutes at room temperature (RT), as described previously 177.

For the activation kinetics assay, the cells were incubated with PB for various times as indicated in the figure. Cell lysates were harvested and subjected to immunoblot analysis.

2.3.6 Immunoblot analysis

Proteins for immunoblot analysis were prepared from the organotypic cultures and regular cell cultures by washing the cells three times with phosphate-buffered saline (PBS,

Cellgro, Manassas, VA), followed by dissolving cells in 1% Triton X-100 in PBS at 4℃

49 for 5 minutes. Insoluble debris was removed by centrifugation. The protein concentrations in these lysates were determined using the BCA protein assay reagents

(Pierce, Rockford, IL) according to the manufacturer’s instructions. Samples containing the same amount of total protein were mixed with 5X sample buffer and resolved by

7.5% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred to nitrocellulose membrane (Pall Corporation, Pensacola, FL). The membranes were probed with the appropriate antibodies and a HRP conjugated secondary antibody

(Kirkegaard & Perry Laboratories, Gaithersburg, MD), before signal detection with

Western Lightning Chemiluminescence Reagent Plus (Perkin-Elmer, Boston, MA).

2.3.7 Prostasin-HAI-1 complex purification

HaCaT cells were treated with PB for 30 minutes at RT to activate prostasin and induce prostasin-HAI-1 complex formation. After activation, the cells were washed three times with PBS and lysed with 1% Triton X-100 in PBS. The cell lysates were centrifuged to remove insoluble debris and then dialyzed against 20 mM Tris-HCl, pH

8.0, containing 1% Triton X-100 (buffer A). The dialyzed sample was loaded onto a

DEAE Sepharose Fast Flow column (GE Healthcare, Piscataway, NJ) pre-equilibrated with buffer A. The DEAE column was washed with 5 column volume of buffer A, and proteins were eluted with a linear gradient of 0 – 0.5 M NaCl in buffer A. The eluted fractions containing the prostasin-HAI-1 complex were pooled and loaded onto a HAI-1 mAb M19 immunoaffinity column pre-equilibrated with buffer A. The column was washed with buffer A and bound protein was then eluted with 0.1 N glycine-HCl, pH 2.4.

The eluate fractions were neutralized immediately after collection with 2 M Trizma Base.

2.3.8 Mass spectrometry analysis and identification of prostasin

50

The protein bands from SDS-PAGE gels stained with Protoblue (National

Diagnostics, Atlanta, Georgia) were excised, washed, destained, and trypsinized overnight at 37℃ using standard protocols after dithiothreitol (DTT) reduction and iodoacetamide alkylation. Analysis of the tryptic peptides derived from the protein samples was performed by liquid chromatography/mass spectrometry (LS/MS) using services provided by Prottech Inc. (Norristown, PA).

2.4 Results

2.4.1 Activation of matriptase starts early, and is maintained throughout in vitro human epidermal differentiation

Epidermal differentiation involves a host of proteases and protease inhibitors of which matriptase is an essential player in the development of skin barrier function.

Here we used 3-dimensional organotypic skin cultures to examine the functional status of matriptase during epidermal differentiation. When cultured as a monolayer, primary human keratinocytes express matriptase in its mature latent form, yielding a 70-kDa protein band on immunoblots using matriptase mAb M24 (Figure 2.1, Total MTP, lanes

M). The cells are devoid of activated matriptase as shown by the lack of a band recognized by the mAb M69 which can specifically distinguish active matriptase from latent matriptase (Figure 2.1, Activated MTP, lanes M). The cognate inhibitor of matriptase, HAI-1, is also expressed by human keratinocytes in monolayer culture, detectable in its un-complexed form at 55-kDa detected by HAI-1 mAb M19 (Figure 2.1,

HAI-1, lanes M).

51

Figure 2.1 Detection of a novel HAI-1 complex in differentiating human keratinocytes. Primary human keratinocytes were grown in monolayer (M) or in a skin organotypic culture system (R). Cells were harvested on the second or eighth day after they were lifted to the air-liquid interface and analyzed by immunoblot for total matriptase (Total

MTP), activated matriptase (activated MTP), and HAI-1.

52

In order to study how matriptase is regulated during epidermal differentiation, epidermal organotypic cultures were established using primary human keratinocytes grown on collagen lattices prepared with fibroblasts. Two days after being lifted to the air-liquid interface, significant activation of matriptase occurs, as indicated by the appearance of a 120-kDa (activated) matriptase-HAI-1 complex, which can be detected by all three mAbs: M24, M69, and M19 (Figure 2.1, 2 days, lanes R). These data suggest that matriptase activation occurs at an early stage of epidermal differentiation.

Keratinocytes appear to employ HAI-1 to inhibit matriptase in a fashion identical to that of other types of epithelial cells, with the active matriptase being rapidly inactivated by the formation of a 120-kDa matriptase-HAI-1 complex. In addition to this species, however, we noticed that differentiating keratinocytes also form an 85-kDa HAI-1 complex (Figure 2.1, 2 days, HAI-1, lane R). The level of the 85-kDa HAI-1 complex detectable in the lysates appeared to be comparable to that of the 120-kDa matriptase-HAI-1 complex in epidermal keratinocytes, and yet an 85-kDa HAI-1 complex has not been reported in other epithelial systems, such as 184 A1N4 mammary epithelial cells suggesting that this species is not present in other epithelial systems 177.

Since HAI-1 only binds and forms stable complexes with active but not latent serine proteases, the appearance of the 85-kDa HAI-1 complex suggests that a second serine protease may become activated and then bound by HAI-1 in differentiating keratinocytes.

The matriptase-HAI-1 complex and 85-kDa HAI-1 complex were also detected in raft culture lysates prepared on the 8th day of culture (Figure 2.1, 8 Days), suggesting that the activation of matriptase and the putative second serine protease may be maintained during the course of epidermal differentiation.

53

2.4.2 Acid exposure induces simultaneous activation of matriptase and a putative serine protease in HaCaT cells

The appearance of the 85-kDa HAI-1 complex throughout the course of keratinocyte differentiation in the organotypic skin culture model might be related to activation of matriptase in a keratinocyte-selective manner. Previously we discovered using other epithelial cells that robust matriptase activation can rapidly be induced by simply exposing the cells to mildly acidic conditions 177, 192. We, therefore, tested if this would occur in keratinocytes by exposing HaCaT cells to a pH 6.0 buffer, and found that acid exposure of these cells resulted in the rapid conversion of the 70-kDa matriptase zymogen to the 120-kDa matriptase-HAI-1 complex (Figure 2.2, Total MTP and

Activated MTP, comparing lanes 3 to lanes 1), suggesting that keratinocytes indeed share this activation mechanism with other epithelial cells. In addition to the 120-kDa matriptase-HAI-1 complex, the 85-kDa HAI-1 complex was also detected in these cells in response to the mild acid exposure (Figure 2.2, HAI-1, lane 3).

Acid-driven matriptase activation is thought to result from an unusual feature of the matriptase zymogen. The intrinsic activity of the zymogen form of matriptase is greatly enhanced at a mildly acidic pH, and this may contribute to matriptase autoactivation 176.

Furthermore, matriptase zymogen activity can also be inhibited by elevated concentrations of sodium chloride, as can the autoactivation of matriptase in 184 A1N4 mammary epithelial cells 176, 192. The HaCaT cells also appear to exhibit the

NaCl-mediated inhibition of acid-induced matriptase activation since the levels of the

120-kDa matriptase-HAI-1 complex were significantly reduced in the presence of increased NaCl (Figure 2.2, comparing lanes 2 to lanes 3). Interestingly, sodium

54

Figure 2.2 Acid induced activation of matriptase is associated with the formation of a novel 85-kDa HAI-1 complex in immortalized human keratinocytes. HaCaT cells were incubated with basal media (Basal, lane 1) or phosphate buffer pH 6.0 with 0.15 M NaCl (PB+NaCl, lane 2) or phosphate buffer pH 6.0 alone (PB, lane 3) for

30 minutes. Cell lysates were extracted and subjected to immunoblot analyses for total matriptase (total MTP), activated matriptase (activated MTP) and HAI-1.

55 chloride also significantly suppressed formation of the 85-kDa HAI-1 complex (Figure

2.2, HAI-1, comparing lane 2 to lane 3). These data suggest that formation of the

85-kDa HAI-1 complex may be an event common to keratinocytes and that its appearance is associated with matriptase activation.

2.4.3 Purification of the 85-kDa HAI-1 complex and identification of prostasin as a component of the complex

To identify the putative serine protease that binds to HAI-1 to form the 85-kDa complex, the complex was isolated from cell lysates prepared from acid exposed HaCaT cells using a combination of DEAE chromatography, immunodepletion to remove the

120-kDa matriptase-HAI-1 complex using immobilized matriptase mAb 21-9, and immunoaffinity chromatography using immobilized HAI-1 mAb M19. Following these procedures, the 85-kDa HAI-1 complex was visualized by Protoblue protein staining of a

SDS gel under non-boiled and nonreducing conditions (Figure 2.3A, lane 1, a).

Heat-denaturing the sample resulted in dissociation of the 85-kDa HAI-1 complex, and was accompanied by the appearance of bands at 55-kDa and 35-kDa (Figure 2.3, A, lane

2, b and c, respectively). The heat-sensitive nature of the 85-kDa complex is consistent with the noncovalent interactions between serine proteases and Kunitz-type serine protease inhibitors. In-gel trypsin digestion and MS-based protein identification revealed that the 85-kDa protein band contained HAI-1, as expected, and the serine protease prostasin (Figure 2.3B). The constituents of 85-kDa prostasin-HAI-1 complex were further confirmed by MS-based protein identification of the bands yielded by heat-denaturation of the 85-kDa complex with the 35-kDa band being confirmed as prostasin and the 55-kDa band as HAI-1 (Figure 2.3B). Immunoblot analysis using a

56

Figure 2.3 Identification of prostasin as a component of the novel 85-kDa HAI-1 complex. A. Partially purified 85-kDa HAI-1 complex was analyzed by SDS-PAGE under either nonreducing and nonboiled conditions (NRNB) or under nonreducing and boiled conditions (NR, B) and the proteins bands visualized by staining with ProtoBlue. B.

The protein bands indicated by a, b, and c in Figure 3A were subjected to proteomic protein identification by MS/MS. Among the tryptic peptides obtained from protein band a, 12 peptides matched to HAI-1 and 6 peptides matched to prostasin. Thirteen peptides obtained from band b matched to HAI-1, and three from band c matched to prostasin. C. The purified 85-kDa HAI-1 complex was analyzed by immunoblot for prostasin under either nonreducing and nonboiled conditions (NRNB) or nonreducing and boiled conditions (NR, B) using a prostasin polyclonal antibody.

57 prostasin antibody also confirmed that the 85-kDa HAI-1 complex contained prostasin under nonboiled and nonreducing conditions (Figure 2.3C, lane 1); and that the 35-kDa protein band released from heat-denaturing was indeed prostasin (Figure 3C, lane 2).

These data and those described above suggest that keratinocytes simultaneously activate matriptase and prostasin throughout the course of epidermal differentiation, and in response to acid exposure; and that both active matriptase and active prostasin are rapidly inhibited by HAI-1.

2.4.4 Prostasin activation is dependent on matriptase

We next set out to determine if the activation of prostasin in keratinocytes simply occurs at the same time as matriptase activation, or if its activation is coupled to, and dependent on matriptase activation. As shown in Figure 2.2, using the HAI-1 mAb, the formation of the prostasin-HAI-1 complex can be induced by exposing HaCaT cells to a pH 6.0 buffer, and NaCl can suppress complex formation. Using a prostasin polyclonal antibody, we confirmed that the 85-kDa complex was not detected in HaCaT keratinocytes before acid exposure (Figure 2.4, lane 1). In response to the acid exposure, the appearance of the 85-kDa complex can be clearly detected by the prostasin polyclonal antibody (Figure 2.4, lane 3), and its appearance is suppressed by the inclusion of NaCl in the pH 6.0 buffer (Figure 2.4, lane 2), recapitulating the data generated with the HAI-1 mAb shown in Figure 2.2. Since HAI-1 only binds to active serine proteases, formation of the prostasin-HAI-1 complex should be at the cost of latent prostasin, in a situation similar to the disappearance of 70-kDa latent matriptase and the concomitant appearance of 120-kDa matriptase-HAI-1 complex (Figure 2.2). The prostasin polyclonal antibody used in the current study appears to be less sensitive to latent prostasin than to

58

Figure 2.4 Acid exposure of HaCaT keratinocytes results in the activation of prostasin associated with matriptase activation. HaCaT cells were incubated with basal media (Basal, lane 1) or phosphate buffer pH 6.0 with 0.15 M NaCl (PB+NaCl, lane 2) or phosphate buffer pH 6.0 (PB, lane 3) for 30 minutes. Cell lysates were extracted and subjected to immunoblot analyses for prostasin using a prostasin polyclonal antibody. The lower panel (marked with an asterisk) containing proteins near the 35-kDa protein marker shows a longer exposure of the x-film to show the levels of latent prostasin.

59 prostasin-HAI-1 complex. Latent prostasin with a size close to the 35-kDa protein marker was clearly detected in HaCaT cells only after exposing the nitrocellulose membrane to X-ray film for a longer time (Figure 2.4, lane 1 in low panel).

Nevertheless, the conversion of prostasin zymogen to active prostasin and inactivation of active prostasin by the formation of the 85-kDa HAI-1 complex, results in the levels of latent prostasin being significantly decreased in parallel with the formation of the 85-kDa complex (Figure 2.4, low panel, comparing lane 3 with lane 1). The reduction of zymogen activation caused by the NaCl (Figure 2.4,lane 2 in low panel), results in latent prostasin being detected at intermediate levels (Figure 2.4, lower panel) consistent with the partial suppression of the appearance of the 85-kDa complex (Figure 2.4, lower and upper panel). These data clearly demonstrate that prostasin undergoes activation and

HAI-1 complex formation in keratinocytes in response to acid exposure and that matriptase and prostasin activation are tightly coupled.

Matriptase has been shown to become active through an autoactivation process 171, and prostasin activation does not occur in the skin of matriptase-deficient mice 210. This suggests that matriptase may serve as an upstream activator of prostasin. To test this hypothesis we first generated three HaCaT clones in which the level of matriptase expression had been suppressed to various degrees by transduction with lentiviruses carrying matriptase-targeting shRNAs, or non-targeting controls (Figure 2.5A, Total MTP,

49, 52, 53, and NT). The levels of HAI-1 and prostasin were not altered in these clones when compared to the levels in the parental and non-target control cells (Figure 2.5A,

HAI-1 and Prostasin). For better detection of prostasin, a commercially available prostasin monoclonal antibody was used which only detects the protease after

60

Figure 2.5 Prostasin activation is dependent on matriptase.

A. Generation of matriptase knockdown HaCaT cells. HaCaT cells were infected with lentiviruses bearing three different matriptase-targeting shRNAs and a non-targeting shRNA. Stable clones were selected with puromycine and cell lysates were analyzed by immunoblot for total matriptase (Total MTP), HAI-1 and prostasin. Total matriptase and HAI-1 were detected under nonreducing and nonboiled conditions. Prostasin was detected under reduced and boiled conditions using a prostasin monoclonal antibody. B.

Prostasin activation is dependent on matriptase. The matriptase knockdown clones (49 and 52) and the non-target control (NT) cells were either exposed to basal medium (Basal) or phosphate buffer pH 6.0 (PB) for 30 min to induce activation of matriptase and prostasin. Formation of the prostasin-HAI-1 complex, as an indication of prostasin activation, was analyzed by immunoblot using HAI-1 mAb M19.

61 heat-denaturation of the sample prior to SDS-PAGE, which results in dissociation of the prostasin-HAI-1 complex. When matriptase activation was induced by acid exposure of the cells, formation of the 85-kDa prostasin-HAI-1 complex along with 120-kDa matriptase-HAI-1 complex was clearly observed in the non-target control clone (Figure

2.5B, NT, comparing lane 2 to lane 1); but was barely detectible in the matriptase knockdown clones 49 and 52 (Figure 2.5B, 49 and 52). These data clearly demonstrate that prostasin activation depends on matriptase proteolytic action, providing key evidence for a matriptase to prostasin activation cascade.

2.4.5 Matriptase and prostasin activation kinetics in HaCaT cells

We further established the “lockstep” relationship between matriptase activation and prostasin activation by comparing their respective activation kinetics in response to acid exposure. As rapidly as 3 to 4 minutes after starting acid exposure, matriptase began to become activated, with the active enzyme being immediately inhibited by binding to

HAI-1 with the appearance of the 120-kDa matriptase-HAI-1 complex detected by both matriptase and HAI-1 mAbs (Figure 2.6, Total MTP and HAI-1). Matriptase activation reached a plateau after around 10-15 min of acid exposure as there was almost no latent matriptase left at this point (Figure 2.6, Total MTP). The appearance of the 85-kDa prostasin-HAI-1 complex followed similar kinetics to the appearance of the 120-kDa matriptase-HAI-1 complex demonstrated using the HAI-1 mAb M19 (Figure 2.6, HAI-1).

The kinetics of prostasin-HAI-1 complex formation can also be followed using the prostasin polyclonal antibody (Figure 2.6, Prostasin). These kinetic studies demonstrate that matriptase activation, prostasin activation, and inhibition of both proteases by HAI-1 occur at almost the same time in human keratinocytes. These data provide evidence that

62

Figure 2.6 Activation and HAI-1 binding kinetics of the matriptase-prostasin proteolytic cascade. HaCaT cells were treated with either basal medium (C) or phosphate buffer pH 6.0 and harvested at indicated time points (minutes). Cell lysates were prepared by lysis with

1% Triton X-100 in PBS and followed by immunoblot analyses for total matriptase (Total

MTP), HAI-1, and prostasin.

63 in spite of the extremely tight control by HAI-1, the short period in which free active matriptase is present; it is able to activate its downstream substrate, prostasin.

2.5 Discussion

In the present study we investigate how a matriptase-initiated cell surface proteolytic cascade is regulated throughout the course of epidermal differentiation. The proteolytic cascade is constitutively activated from an early stage of epidermal differentiation and the active enzymes generated are immediately inhibited by binding to HAI-1. The constitutive activation of the system followed by the immediate HAI-1-mediated inhibition of both matriptase and prostasin suggests that the proteolytic cascade not only plays an important role in epidermal differentiation but that its activity must also be under tight control. Our data clearly demonstrate that matriptase is able to act on its substrate even in the presence of very high levels of HAI-1. The rapid inhibition of active prostasin by HAI-1 further indicates that the tight regulation of the matriptase-prostasin cascade lies not only at the level of the initiator protease matriptase but also at the secondary protease prostasin. With such a tightly regulated mechanism in which only the first two proteases of the cascade have an extremely limited opportunity to act on their downstream substrates, it is reasonable to suppose that deregulation of the cascade at either one of the two proteases or at their cognate inhibitor HAI-1 would affect the entire protease cascade, and in turn derail epidermal differentiation.

Although several matriptase substrates have been identified based on the cleavage preferences of the enzyme, prostasin is the first matriptase substrate for whose activation can be demonstrated during the course of matriptase’s own activation and in the context

64 of rapid HAI-1-mediated inhibition of active matriptase. The functional linkage between matriptase and prostasin activation is consistent with the observation that prostasin activation does not occur in the skin of matriptase deficient mice 210. The very similar epidermal defects that are observed in matriptase-deficient and prostasin-deficient mice, strongly suggest that prostasin may be the predominant substrate of matriptase in this tissue functional context, and that the lack of prostasin activation may be responsible for the majority of the defects associated with matriptase deletion. In spite of their close relationship in skin biology, matriptase and prostasin may not function in tandem in other epithelial cells, or at least their functional linkage in other epithelial cells is not as straightforward as in keratinocytes. During the development and progression of squamous cell carcinoma induced by initiation with 7, 12-dimethylbenzanthracene

(DMBA) and promotion with phorbol 12-myristate 13-acetate (PMA), matriptase and prostasin are eventually expressed in separate parts of the tumors 194. In polarized epithelial cells grown in moist environments, matriptase is targeted to basolateral plasma membrane and prostasin, a GPI-anchored protein, is likely targeted to the apical plasma membrane 266. While a GPI-anchored protein might be targeted to the apical surface after transiently moving to the basolateral plasma membrane, it is conceivable that the different subcellular localizations of matriptase and prostasin in polarized epithelial cells renders the ability of matriptase to activate prostasin less relevant. Different mechanisms are likely employed by other epithelial cells to activate prostasin, if matriptase is not the activator. Furthermore, the rapid inhibition of prostasin by HAI-1 also may not be a universal mechanism. Prostasin was initially isolated from the seminal fluid as an active protease, but not in a HAI-1 complex 267. The epidermis is a

65 specialized epithelium designed to encounter the terrestrial environment, and while keratinocytes resemble other epithelial types, keratinocytes undergo progressive remodeling of cell morphology and tissue structure during epidermal differentiation.

These cellular remodeling events may alter the subcellular distribution of matriptase and prostasin and allow matriptase and prostasin to team up to form a relatively keratinocyte-selective cell surface protease cascade.

The epithelial sodium channel (ENaC), a membrane-bound ion-channel, is located in the apical membrane of polarized epithelial cells 268. Both prostasin and matriptase have been demonstrated to activate ENaC resulting in their alternative names as mouse channel activating protease (mCAP) 1 and mCAP 3, respectively 212. Since matriptase is targeted to the basolateral membrane in polarized epithelial cells where it is activated and immediately inhibited by HAI-1 binding 269, it is of interest to determine whether or how matriptase could activate the apically localized ENaC. In contrast to matriptase, prostasin is co-localized with ENaC at the apical plasma membrane, and this co-localization makes prostasin a more likely candidate protease activator for ENaC in polarized epithelial cells. In contrast to some non-keratinocyte epithelial cells in which the inhibition of prostasin by HAI-1 seems less effective, active prostasin in keratinocytes, much like active matriptase, is a short-lived species. It remains to be determined whether prostasin is still an effective activator of ENaC in keratinocytes 270.

In conclusion, the matriptase-prostasin proteolytic cascade, a relatively keratinocyte-selective, and tightly regulated cell surface proteolysis system, is essential for epidermal differentiation. Both matriptase and prostasin zymogens and their cognate inhibitor HAI-1 are likely co-localized and anchored at the cell membrane. In response

66 to activation stimuli, such as acid exposure or signals of epidermal tissue differentiation, the matriptase zymogen undergoes autoactivation mediated via the formation of a homodimer that generates active matriptase that in turn rapidly activates prostasin. Both active matriptase and active prostasin likely also rapidly activate downstream substrates before HAI-1 moves in and inhibits both proteases and suppresses the activity of the proteolytic cascade. Although HAI-1 contains two Kunitz domains, Kunitz domain 1 at the N-terminus is likely responsible for the inhibition 271. We summarize these molecular events associated with activation and inactivation of the protease cascade in

Figure 2.7. The unprecedented tight control of this cell surface proteolytic cascade allows keratinocytes to undergo differentiation through a precisely programmed mechanism.

2.6 Acknowledgments

This work was supported by NIH Grants R01-CA-104944 (C.-Y. Lin),

R01-CA-123223 (M.D. Johnson and C.-Y. Lin), the Florida Biomedical Research

Program Grant 06NIR-03 (L.-M. Chen) and the Susan G. Komen for the Cure Grant

BCTR0707538 (K.X. Chai)

67

Figure 2.7 The model for the regulation of matriptase-prostasin proteolytic cascade.

Schematic model of the activation and inactivation of matriptase and prostasin is presented. Both matriptase and prostasin are synthesized as single chain zymogens and anchored on cell membrane. Matriptase undergoes autoactivation via the interaction of two zymogen molecules to generate active matriptase. Active matriptase subsequently activates prostasin and the downstream substrates. Both active matriptase and active prostasin are short-lived due to the rapid HAI-1-meidated inhibition.

68

Chapter 3 – Antithrombin regulates matriptase-mediated keratinocyte pericellular proteolysis involved in plasminogen activation, syndecan shedding, and activation of hepatocyte growth factor

69 3.1 Abstract

Matriptase, a tightly regulated membrane-associated serine protease, plays an essential role in epidermal barrier function through converting the zymogen of the

GPI-anchored serine protease prostasin into an active protease. The matriptase-prostasin proteolytic cascade is tightly regulated by the membrane-bound serine protease inhibitor HAI-1 so that matriptase autoactivation and prostasin activation by active matriptase essentially occur at the same time and are immediately followed by

HAI-1 inhibition. Here we report that a proportion of active matriptase can escape from

HAI-1 inhibition by rapidly shedding from the cell surface. In the pericellular environment, the shed active matriptase is able to activate hepatocyte growth factor

(pro-HGF), accelerate plasminogen activation, and shed syndecan 1. The levels of shed active matriptase are inversely correlated with the levels of antithrombin bound to the surface of keratinocytes. The binding of antithrombin to the surface of keratinocytes requires the serpin inhibitor with functional heparin binding site Lys-125 and

N-glycosylation site Asn-135 unoccupied. This suggests that beta-antithrombin, but not the alpha antithrombin, is responsible for the regulation of pericellular matriptase activity in keratinocytes. While breast and prostate epithelial cells also adapt antithrombin to control matriptase activity, antithrombin might be a negligible matriptase inhibitor in

70 both cells. In contrast, in keratinocytes antithrombin could inhibit matriptase as much as 30% of active matriptase inhibited by HAI-1. Taken together, keratinocytes employ two different serine protease inhibitors to control two different sets of matriptase substrate activation and processing: 1) HAI-1 for activation of prostasin and 2) antithrombin for the pericellular proteolysis involved in HGF activation, accelerating plasminogen activation, and shedding of syndecans.

3.2 Introduction

Epidermal differentiation is a carefully orchestrated process that leads to formation of the critical protective barrier provided by the skin. The process of generating a functional epidermal layer requires progressive remodeling of cell morphology and tissue structure, and involves significant pericellular proteolysis that must be activated and inactivated in a precisely controlled manner. Proteases are key regulatory molecules which activate growth factors, enzymes, membrane receptors and ion channels and also process enzyme co-factors, ECM components, structure proteins, and adhesion molecules

109, 110. The importance of proteolysis in epidermal differentiation is emphasized by the many skin disorders that are caused by defects in proteolysis. The proteases implicated in these disorders include lysosomal proteases, cathepsins C, D and L 111-114, two

71 membrane-associated serine proteases, matriptase and prostasin 115-117, and caspase-14 118,

119. In addition to proteases, protease inhibitors clearly also function in skin disorders.

These proteins are essential controllers that limit protease activity in both space and time.

These include the serine protease inhibitors, lymphoepithelial Kazal-type-related inhibitors (LEKTI) 123 and hepatocyte growth factor activator inhibitor (HAI-1) 124, and the lysosomal cysteine protease inhibitor, cystatin M/E 125, 126.

Among these proteases and protease inhibitors involved in skin functions, matriptase, prostasin, and HAI-1 are functionally linked and form a tightly controlled protease network. Matriptase, a type II transmembrane serine protease (TTSP), functions as an initiator protease that undergoes autoactivation to convert matriptase zymogen to active matriptase. Active matriptase, in turn, activates prostasin, a glycosylphosphatidylinositol (GPI)-anchored serine protease. The functional linkage between the two cell surface serine proteases has been manifested by ready identical skin defects in mice with targeted deletion of both proteases and the lack of active prostasin in the skin of matriptase knockout mice. The functional linkage is also corroborated in cultured human keratinocytes in which loss of prostasin activation is the consequence of the knockdown of matriptase expression 193. A remarkable feature of the regulation of the serine protease cascade is that both matriptase and prostasin are under extremely tight

72 control by HAI-1. HAI-1, an integral membrane, Kunitz-type serine protease, is co-expressed and co-localized with matriptase with an unusual HAI-1:matriptase protein ratio of more than 10:1 in most epithelial and carcinoma cells 261. In addition, HAI-1 is required for proper matriptase synthesis and intracellular trafficking out of the endoplasmic reticulum 259. Furthermore, HAI-1 appears to participate in matriptase autoactivation. As a consequence, active matriptase is inhibited by HAI-1 as rapidly as the active matriptase is generated, as if both matriptase activation and inhibition take place at the same time. Remarkably, with such a short life span, active matriptase is still able to activate prostasin 193. The unusually tight linkage of the three key players of the protease network is consistent with the similar epidermal defects observed in their respective knockout mice models 115, 116, 124.

In addition to prostasin, matriptase is also involved in activation of uPA and hepatocyte growth factor (HGF) 154, 208. HGF activation by matriptase and subsequent induction of the cMET pathway is likely the mechanism involved in the development of spontaneous squamous cell carcinomas in matriptase transgenic mice 217. The activation of uPA by matriptase in monocytes significantly reduces the time required for the initiation of the reciprocal zymogen activation between uPA zymogen and plasminogen

162. In contrast to prostasin that is co-expressed and co-localized with matriptase on the

73 plasma membrane of keratinocytes, both uPA and HGF are secreted, either by keratinocytes or by stromal cells. Despite the growing evidence for the role of matriptase in the activation of these substrates, given that the cellular active matriptase has as extremely short life span, the search for active matriptase with prolonged life would provide cellular evidence for the role of matriptase in the activation of these extracellular substrates. In the current study, we demonstrated that a proportion of active matriptase escapes from HAI-1 inhibition by rapidly shedding into the extracellular milieu in the process of matriptase activation. The secreted active matriptase is able to activate HGF and accelerate plasminogen activation in the pericellular milieu. More importantly, we further discovered that the levels of shed active matriptase are inversely regulated by the levels of membrane-bound antithrombin via heparan sulfate proteoglycans (HSPGs). Interestingly, HSPG syndecan 1 can be shed by active matriptase from the surface of keratinocytes. Our studies reveal that keratinocytes adopt two distinct inhibitory mechanisms to control the two different sets of matriptase functions in keratinocytes: 1) the integral membrane, Kunitz-type inhibitor

HAI-1 for prostasin activation and 2) the HSPG-bound serpin-type inhibitor antithrombin for the activation of uPA, HGF and the shedding of HSPG. Since uPA, HGF, and

HSPGs have been implicated in a variety of keratinocyte regulation and function, the

74 tightly controlled matriptase activities via different protease inhibitors, subcellular localization, and life span of active matriptase could allow the activation and processing of these matriptase substrates in a controllable manner.

3.3 Materials and Methods

3.3.1 Reagent

Horseradish peroxidase (HRP)-conjugated secondary antibodies were purchased from Kirkegaard & Perry Laboratories (Gaithersburg, MD), Western Lightning

Chemiluminescence Reagent Plus was purchased from PerkinElmer Life Sciences

(Waltham, MA). Nitrocellulose membrane was purchased from Pall Corp. (Pensacola,

FL), Boc-Gln-Ala-Arg-AMC was purchased from Enzo Life Sciences (Farmingdale, NY) and Boc-Val-Leu-Lys-AMC was obtained from Bachem Americas, Inc. (Torrance, CA).

Dithiothreitol (DTT) was purchased from Sigma-Aldrich (St. Louis, MD). Protoblue was purchased from National Diagnostics (Atlanta, GE). Dulbecco’s phosphate-buffered saline (DPBS) was obtained from Mediatech Inc. (Manassas, VA).

HGF was prepared as described previously 272.

3.3.2 Antibodies

75 Three types of antibodies were used to detect matriptase. The mouse monoclonal antibody (mAb) M24 and rat mAb 21-9 recognize both latent and activated forms of matriptase 155, 160, 177, 193; rabbit polyclonal antibody matriptase/ST14 (Bethyl,

Montgomery, TX) recognizes an epitope present in the serine protease domain of matriptase and is, therefore, able to detect the matriptase/serpin complex after reducing and boiled the samples. Another mAb M69 recognizes an epitope only present on activated matriptase and is able to distinguish activated matriptase from latent matriptase

174, 182. HAI-1 was detected using the HAI-1 mAb M19 155. Hepatocyte growth factor

(HGF) protein was detected using a goat polyclonal antibody (Santa Cruz, Santa Cruz,

CA) that recognizes the beta subunit of HGF. Antithrombin was detected using a goat polyclonal antibody (R&D, anti-serpin C1, Minneapolis, MN). Syndecan-1 was determined by a goat polyclonal antibody (R&D, Minneapolis, MN).

3.3.3 Cell cultures

HaCaT cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS, heat-inactivated), 100 units/ml penicillin, and 100 g/ml streptomycin. 184 A1N4 cells were maintained in

DMEM/F12 with 0.5% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin.

RWPE1 cells were maintained in RPMI 1640 with 10% FBS, 100 units/ml penicillin, and

76 100 g/ml streptomycin. Human primary keratinocytes were maintained in keratinocyte-SFM (Invitrogen) supplemented with bovine pituitary extract (20-30 g/ml), recombinant epidermal growth factor (rEGF, 0.1-0.2 ng/ml), 100 units/ml penicillin, and

100 g/ml streptomycin.

3.3.4 Acid-induced matriptase activation

HaCaT cells were washed with DPBS and incubated in DPBS, as control, or phosphate buffer (PB), pH 6.0 for 30 minutes at room temperature (RT), as described previously 193. Conditioned buffers were concentrated up to 100 folds for immunoblot or matriptase activity assay. Cell were harvested and lyzed in 1% Triton X-100 in DPBS and subjected to immunoblot analyses. The lysis buffer also contains

5,5'-dithiobis-(2-nitrobenzoic acid) (DTNB) to prevent the cleavage of a disulfide bond that links the serine protease domain and the non-catalytic domain of matriptase.

3.3.5 Purification of active matriptase

Acid-induced matriptase activation was performed using HaCaT cells and the conditioned buffers were collected and concentrated. The 70-kDa active matriptase was purified from the conditioned buffers by immunodepletion to remove the matriptase/HAI-1 complex using immobilized HAI-1 mAb M19 and immunoaffinity

77 chromatography using immobilized matriptase mAb M69. The concentration of active matriptase was determined by active site titration as previously described 273.

3.3.6 Protease activity assay

The activity of matriptase or plasmin was measured by the level of

7-Amino-4-methylcoumarin (AMC) released by the proteases. For matriptase activity, the conditioned buffers were concentrated up to 100 folds using Amicon Ultra-4 centrifugal filter units (Millipore, Billerica, MA). Five microliters of the concentrated samples were used to measure matriptase activity using Boc-Gln-Ala-Arg-AMC as substrate. For plasmin activity, 104/well primary keratinocytes were seeded in 96-well plate. The next day, the cells were incubated with PBS, as control, or pH 6.0 buffer for matriptase activation followed by neutralizing by 2M Trizma base to pH 7.5 and adding

50 nM plasminogen into cells. The activities of plasmin were measured using

Boc-Val-Leu-Lys-AMC as substrate. The released fluorescence resulting from hydrolysis of the peptide substrates was measured using a fluorescent spectrophotometer

(Bechman, DTX 880) with excitation at 360 nm and emission at 480 nm. For gelatin zymography, protein samples were incubated with 5X SDS sample buffer containing no reducing agent and incubated at RT. The proteins were resolved in

SDS-polyacrylamide gel electrophoresis (SDS–PAGE) using a 7.5% polyacrylamide gel

78 containing 1 mg/ml gelatin. The gelatin gels were washed with PBS containing 2.5%

Triton X-100 to remove SDS and incubated in Tris buffer pH 8.0 at 37℃ overnight.

The gels were stained by ProtoBlue.

3.3.7 Activation of pro-HGF

Five micrograms of pro-HGF were mixed with different amounts of active matriptase (0-400 pM) with or without HaCaT cells followed by incubating at 37℃ for

30 min. The samples were subjected to immunoblot analyses.

3.3.8 Immunoblot

The protein extracts were mixed with 5X SDS loading buffer in the absence or presence of reducing agents and incubated either at RT or at 95℃ for 5 min. Protein samples were resolved by 7.5% SDS-PAGE and transferred to nitrocellulose membranes.

The membranes were probed with the desired antibodies and a HRP conjugated secondary antibody, followed by signal detection with Western Lightning

Chemiluminescence Reagent Plus.

3.3.9 Antithrombin binding assay

HaCaT cells were incubated in basal medium for 30 min in CO2 incubator and then followed by addition of the three antithrombin preparations at 230 nM for another 5 min

79 in incubator. The cells were lysed and the levels of antithrombin were assessed by immunoblot.

3.3.10 Purification and identification of 110-kDa matriptase complexes

The 110-kDa matriptase complexes were generated along with the activation of matriptase by HaCaT cells by treating the cells with phosphate buffer, pH 6.0 for 30 min at RT. The conditioned buffers were collected and concentrated. The 110-kDa matriptase complexes were purified by sizing column using HPLC (Beckman) with a

Biosuite 250 sizing column at a flow rate of 2 ml/min, immunodepletion to remove the matriptase/HAI-1 complex using immobilized HAI-1 mAb M19, and immunoaffinity chromatography using immobilized matriptase mAb 21-9 to capture the 110-kDa matriptase complex. Both immunodepletion and immunopurification were previously described 193, 261. The 110-kDa complexes and other bound proteins were eluted from the mAb 21-9 immunoaffinity column by 0.1 M glycine buffer, pH 2.4 and immediately neutralized by 2 M Trizma base. The eluate was resolved by SDS-PAGE and the protein bands were stained with Protoblue. The two protein bands around 110-kDa were sliced out, in-gel digested with trypsin and analyzed by liquid chromatography/mass spectrometry (LS/MS) using service provided by Prottech Inc. (Norristown, PA).

3.3.11 Syndecan-1 shedding

80 HaCaT cells were washed with DPBS and incubated with increasing amount of

active matriptase in basal medium in CO2 incubator for 30 min. After incubation, the conditioned media were collected and analyzed for the amount of syndecan-1 shedding using dot blot assay.

3.4 Results

3.4.1 Active matriptase is either rapidly inhibited by HAI-1 or shed from cell surface of human keratinocytes

In our previous study with a minute scale kinetic analysis right after the induction of matriptase zymogen activation, an unusual tight control of the matriptase-prostasin proteolytic cascade features the simultaneous activation of matriptase and prostasin, followed by the immediate inhibition of both active matriptase and active prostasin by

HAI-1. As shown in Figure 3.1A, the extracellular acidosis converted the 70-kDa matriptase zymogen into the 120-kDa (activated) matriptase-HAI-1 complex (Acid, comparing lane 1 with lane 2). Using HAI-1 mAb, in addition to the matriptase-HAI-1 complex, prostasin-HAI-1 complex with a size close to the 100-kDa marker also appeared along with matriptase activation (Figure 3.1A, Acid, lane 4). The simultaneous activation of matriptase and prostasin can also be induced by treating

81

Figure 3.1 Shedding of active matriptase by human keratinocyte.

A. Human keratinocyte HaCaT cells were either incubated with PBS (C) as a control or a

pH 6.0 buffer (Acid, A) or CoCl2 in basal medium (CoCl2, A) to induce activation of matriptase. Cell lysates were analyzed by immunoblot for matriptase (MTP) or HAI-1

(HAI-1). B. HaCaT cells were incubated with a pH 6.0 buffer to induce matriptase activation. The cells and the conditioned buffer together (Combined), the cells alone

82 (Cellular) and the conditioned buffer alone (Shed) were analyzed for matriptase activity using a matriptase synthetic fluorescent substrate, Boc-Gln-Ala-Arg-AMC. Data are representative of four independent experiments done under similar conditions. C. The conditioned buffer was subjected to immunoprecipitation with an activated matriptase mAb M69. The loading control (Loading), the unbound fraction (Unbound), and the eluent (Elution) were assayed for matriptase activity using the substrate,

Boc-Gln-Ala-Arg-AMC. Data are representative of three independent experiments done under similar conditions. D. HaCaT cells were incubated with DPBS as control or a pH 6.0 buffer to induce matriptase activation. The conditioned buffer was collected and subjected to gelatin zymography.

83 HaCaT human keratinocytes with cobalt chloride (Figure 3.1A, CoCl2) and human primary keratinocytes with plasminogen (Chapter 5). These data suggest that the tightly coupled zymogen activation and the inhibition of active protease is a common phenomenon for the matriptase-prostasin-HAI-1 proteolysis network regardless of the extracellular stimuli for the induction of matriptase activation.

Despite the rapid inhibition by HAI-1, strong proteolytic activity, assessed by the cleavage of Boc-Gln-Ala-Arg-AMC, a commonly used synthetic matriptase substrate, was also generated along with matriptase autoactivation (Figure 3.1B, Combined).

Interestingly, this proteolytic activity was present only in the conditioned buffer but not associated with the cells as almost all the proteolytic activity was detected in the shed fraction (Figure 3.1B, Shed) and almost no activity was detected in the cells (Figure 3.1B,

Cellular). Immunodepletion of the shed fractions using matriptase mAbs removed the proteolytic activity that could then be recovered from the immobilized matriptase mAb, confirming the proteolytic activity being derived from the active matriptase (Figure 3.1C).

The generation of active matriptase was further confirmed by gelatin zymography (Figure

3.1D). Strong gelatinolytic activity with the expected size of 70-kDa, active matriptase, was seen in the conditioned buffer along with the induction of matriptase activation but not in the non-activation control. The activated matriptase with two different fates was

84 also seen in the plasminogen-induced matriptase activation in primary human keratinocytes (Chapter 5). Much weaker matriptase activity was detected in the

conditioned media for the CoCl2-treated cells due to the fact that activation of matriptase

induced by CoCl2 was at much lower levels compared to the pH 6.0 buffer.

3.4.2 The pericellular active matriptase can activate pro-HGF and accelerate plasminogen activation

Since the activation and the inhibition of matriptase-prostasin cascade by HAI-1 has been well documented 193, we have focused on the functional and regulational aspects of the shed active matriptase in the current study. First, we investigated whether the secreted active matriptase could activate and process substrates in the pericellular environments. We are particularly interested in whether HAI-1 on the cell surface would affect the activity of active matriptase. Incubation with increasing amounts at

0-400 pM of active matriptase with pro-HGF in test tube resulted in rapid conversion of single-chain pro-HGF into two-chain active HGF in a dose dependent manner (Figure

3.2A, left panel). Similar effects were also observed with pro-HGF activation by active matriptase when both active matriptase and pro-HGF were incubated with HaCaT keratinocytes (Figure 3.2A, right panel). These data suggest that despite the abundance of HAI-1 on the cell surface, an active matriptase is able to activate pro-HGF with similar

85

86 Figure 3.2 Activation of pro-HGF and acceleration of plasminogen activation by active matriptase shed from human keratinocytes.

A. Pro-HGF was incubated with increasing amounts of active matriptase, as indicated, either with or without HaCaT cells at 37oC for 30 min. Samples were analyzed by immunoblot for HGF cleavage using an antibody directed against the beta subunit of

HGF. B. Primary human keratinocytes were incubated with DPBS (No Activation) or a pH 6.0 buffer (Activation) to induce activation of matriptase, after which the buffer on the cells was adjusted to pH 7.5. PLG (50 nM) was then added to the cells and the generation of plasmin was monitored by the cleavage of a plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC. Data are representative of three independent experiments done under similar conditions. C. Primary human keratinocytes were incubated with a pH 6.0 buffer to induce matriptase activation, followed by buffering to pH 7.5. PLG (50 nM) was then added to the cells in the presence of the shed fractions

(combined) or in the absence of the shed fractions (Cellular) or the shed fraction alone

(Shed). Generation of plasmin was monitored by the cleavage of

Boc-Val-Leu-Lys-AMC. Data are representative of four independent experiments done under similar conditions.

87 potency compared to that seen in the absence of HAI-1.

We have further tested whether active matriptase can enhance plasminogen activation since matriptase has been recognized as a potent activator of pro-urokinase plasminogen activator (pro-uPA). Since HaCaT cells appear to express high levels of active plasminogen activator(s), we used primary human keratinocytes for this experiment. Plasminogen activation commonly features a lag phase followed by a burst of plasmin generation, a phenomenon known as a “reciprocal zymogen activation”.

The trace plasmin in the plasminogen preparation slowly initiates pro-uPA activation and the newly generated active uPA, in turn, activates plasminogen. Through these reciprocal activations between the two proteases, plasmin is explosively generated beyond the lag phase, a period of time required for the accumulation of sufficient uPA and plasmin for the burst of plasmin production. In order to test whether active matriptase in the pericellular environment could affect plasminogen activation, matriptase activation was induced by pH 6.0 buffer followed by neutralizing to pH 7.5 and adding plasminogen to the primary human keratinocytes. The production of plasmin was recorded by the cleavage of the standard synthetic fluorescent substrate of plasmin,

Boc-Val-Leu-Lys-AMC, and the rate of plasmin production was presented as liberated fluoresce per min. As shown in Figure 3.2B, it required almost 40 min for the

88 activation of plasminogen to begin to take off in the absence of active matriptase (Figure

3.2B, No Activation). In contrast, in the presence of active matriptase, the lag phase was shortened by about 25 minutes (Figure 3.2B, Activation).

The matriptase-accelerated plasminogen activation appears to require the pericellular environment. While the combination of the cells and shed active matriptase caused robust plasminogen activation (Figure 3.2C, Combined), neither the shed fractions alone (Figure 3.2C, Shed) nor the cells alone (Figure 3.2C, Cellular) accelerate plasminogen activation. These data suggest that the shed active matriptase can return back to the cell surface to activate plasminogen activator and accelerate the reciprocal zymogen activation. Furthermore, the plasminogen activator was not activated in the process of acid-induced matriptase activation, a situation different from prostasin, in which it can be activated simultaneously during acid incubation.

3.4.3 Identification of antithrombin as a significant inhibitor of matriptase in human keratinocytes

In analogy with the cellular activated matriptase that is under control by HAI-1, the shed active matriptase is also under control by a serine protease inhibitor. The regulation of shed active matriptase was initially noticed by the presence of a 110-kDa matriptase complex in the conditioned buffer (Figure 3.3A, HaCaT, arrow).

89

Figure 3.3 Identification of antithrombin as a component of the novel 110-kDa matriptase complex.

A. Human keratinocytes (HaCaT), mammary (184 A1N4) and prostate (RWPE1) epithelial cells were incubated with a pH 6.0 buffer to induce matriptase activation. The conditioned buffers (S) were concentrated and total cell lysates (T) were analyzed by immunoblot for total matriptase. Arrow indicated 110-kDa matriptase complex. B.

HaCaT cells were incubated with a pH 6.0 buffer and the cell lysates and the conditioned buffers were collected at indicated time points (minutes). Samples were subjected to immunoblot analyses for total matriptase. C. Left panel: Partial purified 110-kDa 90 matriptase complex was analyzed by SDS-PAGE under nonreducing and boiled conditions (NR, B). The protein bands were visualized by staining with ProtoBlue.

The boiling treatment decreased the migration rate of the 110-kDa matriptase complex on

SDS-PAGE to near the 130-kDa marker. The protein bands indicated as a and b were subjected to proteomic protein identification by MS/MS (see Figure 5 for sequence data).

Middle and right panels: HaCaT cells were incubated with human serum followed by a pH 6.0 buffer for matriptase activation. Cell lysates and the concentrated shed fractions were subjected to immunoblot analyses for total matriptase (Total MTP) and antithrombin (AT).

91 Interestingly, keratinocytes shed the 110-kDa matriptase species at relatively much higher levels compared to mammary (184 A1N4) and prostate (RWPE1) epithelial cells. The shedding of the 110-kDa species apparently closely follows the kinetics of activation of matriptase: Matriptase activation begins at the two min post induction of activation, as the matriptase-HAI-1 complex began to appear, and reaches a plateau 10 min later

(Figure 3.3B, Cell lysate); the shedding of the 110-kDa species begins at 6 min and accumulates to high levels for few minutes after the completion of matriptase activation

(Figure 3.3B, Shedding). The 110-kDa matriptase complex is likely to be a matriptase-serpin complex due to its resistance to heat treatment. In addition, their dissociation by reducing agents into a fragment of matriptase serine protease domain covalently bound to the binding protein (Figure 3.4), as we have shown in the previous studies 163 also confirms complex formation.

To identify the putative serpin inhibitor that binds to matriptase to form the 110-kDa complex in keratinocytes, the complex was isolated using a combination of a sizing column using HPLC, immunodepletion to remove the matriptase-HAI-1 complex using immobilized HAI-1 mAb M19 and immunoaffinity chromatography using immobilized matriptase mAb 21-9. Following these procedures, the purified proteins were resolved by SDS-PAGE under non-reducing and boiled conditions. Several protein bands were

92

Figure 3.4 The keratinocyte-selected 110-kDa matriptase complex is a protease/serpin complex.

A. HaCaT cells were treated with pH 6.0 buffer and the conditioned buffers were collected and concentrated. Samples were analyzed by immunoblot under either nonreducing and nonboiled conditions (NRNB), or nonreducing and boiled conditions

(NR, B), or reduced and boiled conditions (R, B). The membranes were probed with total matriptase (Total MTP) or a commercial polyclonal antibody, ST14, which recognizes the serine protease domain of matriptase and can be used to detect matriptase/serpin complexes. B. A schematic model to show the fate of matriptase-serpin complexes after boil or reduce the sample. The shed matriptase-serpin

93 complex remains intact after boiling due to the fact that the protease forms a covalent linkage with the serpin hence resistant to the heat treatment. This covalent linkage cannot be dissociated by incubating the matriptase-serpin complexes with reducing agents. However, the disulfide bond that links the serine protease domain and non-catalytic domain of activated matriptase is dissociated by the incubation with reducing agents.

94 obtained and two protein bands around 130-kDa protein marker were suspected to be the matriptase complex (Figure 3.3C, left panel, a and b). In-gel trypsin digestion and

MS-based protein identification of the two protein bands revealed that these two protein bands indeed make up matriptase, as expected, and the serpin inhibitor antithrombin.

The sequence information can be found in the Figure 3.5. In order to validate this identification, HaCaT cells were incubated with human serum instead of fetal bovine serum followed by inducing matriptase activation with mildly acidic buffer.

Immunoblot analysis of the cell lysate and the conditioned buffer using matriptase mAb

(Figure 3.3C, middle panel, Total MTP) and an antithrombin antibody (Figure 3.3C, right panel, AT) demonstrated that the 110-kDa species was recogized by both antibodies.

These data confirmed that the 110-kDa matriptase complex in the conditioned buffer indeed contained antithrombin (Figure 3.3C, lanes 2). Interestingly, a matriptase-antithrombin complex was also detected at very low level in the cell lysate

(Figure 3.3C, lanes 1), suggesting that the vast majority of matriptase-antithrombin complex have been rapidly shed from the cell surface upon the complex formation.

95

Figure 3.5 Identification of antithrombin as a component of the novel 110-kDa matriptase complex.

The protein bands indicated by a and b in Figure 3C, left panel, were subjected to proteomic protein identification by MS/MS. Among the tryptic peptides obtained from protein band a, 10 peptides matched to matriptase and 18 peptides matched to antithrombin. Ten peptides obtained from b matched to matriptase and 20 matched to antithrombin. These amino acid sequences were presented using a single letter with the sequence numbers at the beginnings and the ends of each peptide.

96 3.4.4 Binidng of antithrombin to the surface of keratinocytes depends on its heparin binding affinity and affects the levels of active protease shed into extracellular milieu

In order to effectively inactivate matriptase, antithrombin must bind to the surface of kerationcytes likely through cell surface heparan sulfate proteoglycans (HSPGs), such as syndecans. The binding of antithrombin to heparin or heparan sulfate is mediated by a two-step mechanism in which the Lys-125 residue of antithrombin provides the initial binding to the charateristic pentasaccharide in heparin. The initial binding is further strengthened by ensuing conformational changes in antithrombin. A recombinant antithrombin bearing a point mutation at the initial heparin binding residue Lys-125 was, therefore, used to investigate the role of heaprin binding site in the binding of antithrombin to the suface of keratinocytes and keratinocytes appear not to retain this antithrombin mutant (Figure 3.6A, lane 1, K125M). In addition, human serum contains two antithrombin species: an α-antithrombin (more than 90%) which is a major form with oligosaccharide side chains occupying all four glycosylation sites and the minor form,

β-antithrombin (5-10%) with an unoccupied Asn-135 residue. Exposing human keratinocytes to the antithrombin pool purified from human serum, resulted in cellular retention of some antithrombin (Figure 3.6A, lane 2, Serum AT). Since β-antithrombin

97

Figure 3.6 The shedding of active matriptase is inversely correlated with the levels of membrane-bound antithrombin.

A. HaCaT cells were incubated with basal medium at 37℃ for 30 min, in order to reduce the level of antithrombin from fetal bovine serum, followed by incubation with different antithrombin variants at 37℃ for 5 min. Cell lysates were extracted and subjected to immunoblot analyses for antithrombin (AT). B. Bovine antithrombin was removed from

HaCaT cells by incubation cells with basal medium at 37℃ for 30 min. Cells were then replenished with three different antithrombin preparations, as indicated, followed by induction of matriptase activation using a pH 6.0 buffer. Cell lysates (Cell lysate) and the conditioned buffers (Shedding) were subjected to immunoblot analyses for total matriptase (Total MTP) and antithrombin (AT). C. Matriptase activity was assayed for the shed fractions associated with matriptase activation from the cells pretreated with the three antithrombin preparations, as indicated, using the matriptase fluorescent substrate,

Boc-Gln-Ala-Arg-AMC. Data was done in triplicate and is a representative example of 98 three independent experiments.

99 exhibits stronger heparin binding than α-antithrombin, the minor antithormbin in serum may represent the species bound to the cells. This hypothesis was supported by the fact that antithrombin Asn-135 mutant exhibits strong retention to keratinocytes (Figure 3.6A, lane 3, N135Q). Taken together, these data suggest that the heparin binding affinity of antithrombin plays an important role in its retention on the cell surface.

In contrast to HAI-1 which is co-expressed and co-localized with matriptase at an excess molar ratio, antithrombin must bind to the cell surface for an effective inhibition of matriptase. In addition to the proximality, the binding to HSPGs also significantly increased the rate constant for antithrombin to inhibit matriptase. This notion suggests that the levels of antithrombin bound to cell surface might be a determinant for the levels of active matriptase shed into the extracellular millieu. The different binding affinity to the keratinocyte cell surface between α- and β-antithrombin enabled us to evaluate the impact of antithrombin on the pericellular matriptase activity. This is particularly relevant for keratinocytes in which antithrombin plays a much more significant role in controlling matriptase as compared to other breast and prostate epithelial cells (Figure

3.3A). To test this hypothesis, keratinocytes were first exposed to the same amounts of

1) antithrombin K125M mutant, 2) antithrombin purified from human serum which contains predominantly α-antithrombin and 3) antithrombin N135Q mutant, the

100 β-antithrombin equivalent. After washing away the excessive unbound antithrombins, the cells were exposed to a pH 6.0 buffer to induce matriptase activation. Both cell lysates and shed fractions were analyzed by immunblot for matriptase and antithrmbin

(Figure 3.6B). Matriptase proteolytic activity in the shed fractions was also analysed

(Figure 3.6C). Matriptase activation and formation of 120-kDa matriptase-HAI-1 complex were not affected by the different levels of antithrombin bound to the cells

(Figure 3.6B, left panel, lanes 1, 2 and 3). The levels of matriptase-antithrombin complex in both cell lysates and the shed fractions were correlated with the levels of antithrombin bound to the cells. More importantly, the levels of shed active matriptase were inversely correlated with the levels of antithrombin bound to the cells (Figure 3.6C).

These data suggest that antithrombin through binding to the cell surface can regulate the levels of shed active matriptase.

3.4.5 Extracellular active matriptase can shed syndecan-1

Given that cell surface HSPGs could serve as the binding sites for antithrombin, the rapid inhibition of active matriptase by antithrombin suggests that matriptase activation might take place in close proximity to HSPGs. While TIMP-3-sensitive metalloproteases, such as MMP-7, play a major role in ectodomain shedding of syndecans 1 and 4, the previous studies also showed that syndecans 1 and 4 can be shed

101 by trypsin-like serine proteases, such as plasmin. Given that matriptase exhibits potent trypsin-like activity and due to close proximity to the cell surface HSPGs, we have hypothesized that matriptase might be able to shed syndecans. To test this, we added purified active matriptase back to human keratinocytes and examined the level of syndecan-1 shed in the conditioned media. As shown in Figure 3.7, increasing amounts of syndecan-1 were shed into the conditioned medium along with the increasing levels of active matriptase added to the cells. It is worthwhile to note that matriptase is a much more potent and robust protease to shed syndecan 1 compared to the other serine proteases and metalloprotease, such as plasmin and MMP-7 based on the concentrations used (1 nM versus 60.24 M for plasmin and 50 nM for MMP-7) and the incubation time

(30 min versus 16h for plasmin and 30 min for MMP-7) 274, 275.

3.5 Discussion

Previous studies have revealed that matriptase activity is inhibited by two different types of serine protease inhibitors: the membrane-bound, Kunitz type serine proteases,

HAI-1 and the secreted, serpin type inhibitor, antithrombin. Inhibition of matriptase by

HAI-1 and antithrombin is a physiological event as both matriptase-HAI-1 and matriptase-antithrombin complexes were initially purified and identified from human

102

Figure 3.7 Syndecan-1 can be robustly shed by active matriptase.

HaCaT cells were incubated with increasing active matriptase, as indicated, at 37℃ for

30 min. The conditioned media were collected and subjected to dot blot analysis for syndecan-1.

103 milk. It has been a mystery how matriptase-expressing cells adapt antithrombin as a physiological inhibitor against matriptase when considering the fact that most of these cells express HAI-1 at excessive ratio to matriptase. In the current studies, our results indicate that the two distinct matriptase inhibitors are apparently involved in the regulation of two different types of substrate processing by matriptase: HAI-1 for prostasin activation and antithrombin for the activation of HGF, acceleration of plasminogen activation, and shedding of syndecan. In addition, despite the rapid inhibition of the nascent active matriptase by both protease inhibitors, several important features and differences in matriptase inhibition have been observed. Although both antithrombin and HAI-1 inhibit matriptase on the cell surface, antithrombin, but not

HAI-1, is responsible for the levels of active matriptase shed into extracellular milieu.

With the relatively prolonged lifespan, the extracellular active matriptase is able to activate and process secreted substrates, such as pro-uPA and pro-HGF. The extracellular active matriptase could also act on substrates anchored on the cell surface, leading to shedding of HSPGs (Figure 3.7) and activation of PAR-2 207. In contrast to the extracellular active matriptase, the cellular active matriptase is a short-lived species, supported by the lack of cellular matriptase proteolytic activity detected in the cells with robust matriptase activation. The short lifespan of cellular active matriptase is the

104 consequence of the excess ratio of HAI-1 to matriptase, the potency of HAI-1 inhibitory activity against matriptase, and the co-localization of HAI-1 with matriptase on the cell surface and in the secretory pathway. Prostasin appears to be the only convincing substrate so far for the cellular active matriptase in keratinocytes. Although cellular active matriptase could also act on other substrates, prostasin activation is likely to be responsible for the most of the epidermal defects associated with matriptase deficiency in rodents. Therefore, the role of HAI-1 in matriptase inhibition is predominantly linked to prostasin activation. In contrast, antithrombin appears to have no role in regulation of prostasin activation. The most interesting feature is that HAI-1 exerts temporal and spatial control to ensure the proteolytic activities of matriptase and prostasin targeting a limited number of substrates without spill-over to other cellular proteins.

While antithrombin is commonly employed by epithelial cells to control matriptase activity 163, keratinocytes are apparently unique among epithelial cells in terms of the extent that antithrombin contributes to matriptase inhibition. As shown in Figure 3.3A, the ratio of matriptase-antithrombin complex to matriptase-HAI-1 complex in both cell lysates and the conditioned buffer are negligible in both mammary and prosate epithelial cells. Although the matriptase-antithrombin:matriptase-HAI-1 ratio in HaCaT keratinocytes varies from experiment to experiment, the ratio is estimated to be around

105 1:2 (Figure 3.3A) or 1:3 (Figure 3.3B). The greater role of antithrombin played in control of matriptase activity in keratinocytes is a likely consequence of the functional and regulational modification and adaption of matriptase in accommodation to the histological and functional divergency between the simple, polarized epithelium and the stratified epithelium. First, both the simple and the stratified epithelial cells express matriptase and employ HAI-1 as an essential and indispensible partner. The partnership between matriptase and HAI-1 is established not only by their widespread co-expression in epithelial tissues and co-localization in the secretory pathway and the cell surface subdomains but also by the functional relationship beyond protease-antiprotease one.

HAI-1 has been shown to participate in matriptase synthesis, intracellular trafficking and zymogen activation. This unusual partnership is manifested by the rescue of the various defects associated with HAI-1 deletion with simultaneous matriptase deletion in several animal models. Secondly, prostasin as a substrate of matriptase could be more physiologically relevant in the stratified epithelium but less important in the simple epithelium. Despite their widespread co-exprssion in many epithelial tissues, matriptase is targeted to the basolateral plasma membrane and prostasin is targeted to the apical plasma membrane in polarized epithelial cells. Although matriptase could activate prostasin during the intracellular trafficking prior to their arrival at different subdomains

106 of the cell surface of polarized epithelial cells, this mechanism would be less important due to the fact that both proteases are in their zymogen forms upon arrival at the cell surface. In contrast to polarized epithelial cells, stratified epithelial cells do not show the typical cell surface polarizartion. This histological modification and adaption in stratified epithelium might bring matriptase and prostasin together and result in the keratinocyte-selective partnetship between the TTSP and the GPI-anchored serine protease. Thirdly, HAI-1 as an inhibitor which must simultaneously suppress both matriptase and prostasin, the availibility of HAI-1 for matriptase in keratinocytes should significantly decrease compared to that in polarized epithelial cells in which the

HAI-1:matriptase ratio could be up to 14. Fourthly, the reduced availability of HAI-1 to matriptase might force keratinocytes to rely on more available cell surface antithrombin to control matriptase than the polarized epithelial cells. While less significant, in polarized epithelial cells, antithrombin is an existing mechanism to control matriptase.

The enhanced role of antithrombin in matriptase inhibition in keratinocytes might be an inevitable consequence of the histological modification for stratified epithelium. Finally, the availibility of antithrombin on the cell surface depends on cell surface HSPGs and the binidng affinity of antithrombin to the HSPGs. This might generate another keratinocyte-selective mechanism in which free active matriptase can be shed into

107 extracellular milieu at relatively high levels compared to the simple epithelial cells and the levels of active matriptase shed can be controlled by the levels of cell surface antithrombin. In Figure 3.8, we have summarized these functional and regulatory modifications and adaptation of matriptase along with the histological modification between simple and stratified epithelium.

In conclusion, we have provided evidence that keratinocytes employ two protease inhibitors, HAI-1 and antithrombin, to regulate distinct substrate processing by matriptase.

108

Figure 3.8 A schematic model for the functional and regulatory divergence of matriptase between the simple/polarized epithelium and the stratified epithelium.

The subcellular distribution of matriptase, prostasin, and HAI-1 and the events following matriptase activation in the simple/polarized and the stratified epithelium is schematically presented. In simple/polarized epithelium, matriptase is targeted to the basolateral surface and prostasin is targeted to the apical surface but HAI-1 interacts with both cell surface subdomains. The polarized expression of matriptase and prostasin is, however, likely not present in stratified epithelium, and as such matriptase gains access to prostasin.

109 When massive matriptase activation is induced, active matriptase is rapidly inhibited by nearby HAI-1 and the shortly-lived active matriptase can act on prostasin only in keratinocytes but not in polarized epithelial cells, due to the differential subcellular distributions of matriptase and prostasin between the two epithelia. Active prostasin is also rapidly inhibited by HAI-1 by forming a complex. A proportion of active matriptase is shed from the cell surface, becoming evident in the stratified epithelial cells, but not in polarized epithelial cells. As does the shedding of active matriptase, inhibition of active matriptase by membrane-bound antithrombin via cell surface heparan sulfate proteoglycans, such as syndecans, becomes more obvious and important in stratified epithelial cells but not in polarized epithelial cells. The escaped active matriptase then can work on its substrates, such as PAR-2, pro-uPA, pro-HGF, and syndecans.

110

Chapter 4 – Matriptase in human epidermis: differentiation-regulated expression and enhanced zymogen activation in the epidermal stem cells

111 4.1 Abstract

Genetic defects in transmembrane serine protease matriptase are the causative mechanism underlying two rare congenital ichthyoses, ARIH and IFAH. Mouse models with deletion or insufficient matriptase phenocopy the histological abnormality and reveal prostasin as the molecular target of matriptase responsible for the epidermal defects and HAI-1 as the indispensible inhibitor required for matriptase skin functions.

Human-relevant models, however, are necessary for the further investigation to gain mechanistic insights into the role of matriptase in skin barrier functions and human skin disease as there are significant differences in the syndrome severity caused by matriptase deletion: uniformly postnatal death in mice but mild skin pathology in humans. In our current studies, the distribution profile of matriptase expression and its zymogen activation state are investigated in human skin, oral epithelium, and during keratinocyte differentiation and epidermal tissue regeneration using an in vitro skin organotypic culture model. Our data reveals that human matriptase is expressed in both basal and spinous keratinocytes and the highest zymogen activation is seen in keratinocytes from the bulge area of hair, suspected to contain skin stem cells. This is in stark contrast to the mouse matriptase which is expressed by suprabasal, but not basal, keratinocytes with highest expression in the outermost living keratinocytes. The data suggests that human

112 matriptase is likely to be involved in the regulation of keratinocyte growth and early differentiation, both of which can be corroborated by the reduced growth and expression of differentiation markers in HaCaT human keratinocytes with reduced matriptase expression. The high matriptase zymogen activation in the bulge skin stem cells and during the epidermal tissue regeneration in the skin organotypic culture suggest that matriptase actively participates in “active” keratinocytes but not in relative quiescent cells in the skin. Dysregulated matriptase activity in skin bulge stem cells could, therefore, be a potential mechanism underlying the ARIH and IFAH.

4.2 Introduction

Genetic defects in type 2 transmembrane serine protease matriptase have been recently identified as a causative abnormality underlying the two rare forms of congenital ichthyosis, autosomal recessive ichthyosis with hypotrichosis (ARIH, OMIM 610765) and ichthyosis, follicular atrophoderma, hypotrichosis and hypohidrosis (IFAH,

OMIM602400) 120-122. The defects in skin barrier function are evident in these patients and they represent the underlying pathogenic mechanisms for the syndrome, similar to the majority of congenital ichthyoses 276. The defects in the skin barrier function for ichthyosis pathogenesis are always associated with impairments of at least one of the

113 three major components of stratum corneum barrier, including intercellular lipid layers, cornified cell envelope and keratin-filaggrin degradation products 276. The disturbed profilaggrin processing, rather than the formation of lamellar granules and cornified cell envelope, appears to be the principal contributor to impaired skin barrier which is manifested in patients with matriptase mutation 122. Histological and ultrastructural analyses further revealed that the impaired degradation of corneodesmosomes and thickened stratum spinosum (acanthosis) and stratum corneum are associated with the skin in ARIH 120. The degradation of both corneodesmosomes and profilaggrin requires intensive proteolysis. Since the potential localization of the matriptase, either the corneodesmosomes or the keratohyalin granules, remains to be elucidated, the proteolytic activity of matriptase could either directly or indirectly be involved in both events. The acanthosis and the thickened stratum corneum could result from an impact on keratinocyte proliferation and differentiation programs by matriptase mutations, propagating through multiple processes rather than just at one specific proteolytic event.

Therefore, matriptase could participate in skin function by regulating multiple processes from the formation of stratum corneum barrier to the maintenance of epidermal homeostasis through the desquamation and/or the replenishment of keratinocytes via highly regulated proliferation and differentiation.

114 An animal model with targeted deletion of matriptase has revealed mechanistic insights for the role of matriptase in skin barrier function. In addition to the replicating the ichthyotic phenotype and the impaired profilaggrin processing seen in the ARIH and

IFAH patients, the impaired tight junction activity has been identified to play a significant role in the loss of epidermal barrier function in the skin of matriptase knockout mice 120-122, 159. The study of matriptase knockout mice also uncovers a prostasin, a glycosylphosphatidylinositol (GPI)-anchored serine protease, as a molecular target of matriptase responsible for the epidermal defects caused by matriptase deficiency in mice 210. Lack of prostasin activation is also seen in the skin of IFAH patients 122.

The functional linkage between the two membrane-bound serine proteases can also be demonstrated in immortalized human keratinocytes in which prostasin zymogen activation depends on matriptase 193. Strikingly, the zymogen activation of both serine proteases is under extremely tight control by hepatocyte growth factor activator inhibitor

(HAI)-1, a Kunitz type serine protease inhibitor, in a way that the zymogen activation of both serine proteases must take place virtually at the same time 193. Interestingly, the targeted deletion of HAI-1 in mice also causes ichthyosis-like skin with aberrant pro-filaggrin processing, acanthosis and enhanced Akt phosphorylation 124. The

115 matriptase-prostasin-HAI-1 cell surface protease network, therefore, is an essential program for skin barrier function.

Mice models appear to be an excellent tool to study the physiological role of matriptase in skin and to understand the development and progression of human skin diseases, given the shared epidermal defects associated with matriptase deficiency, the same downstream molecular target and the same antiprotease mechanism for matriptase between human and rodent. This is particularly true for the matriptase hypomorphic mouse model which phenocopies AIHR patients 120, 195. However, a less severe, non-life threatening clinical phenotype was observed in the patients with deletion of matriptase caused by frame-shift or splice-site mutation in matriptase in IFAH patients as compared to the uniformly postnatal death in matriptase knockout mice 115, 122. There might be differences in the mechanisms underlying the defects in epidermal barrier associated with matriptase deficiency in human versus rodent. Therefore, there is a need to establish human-relevant models to investigate and provide mechanistic insights into the causative role of matriptase in the AIHR and IFAH. To this end, we have compared the distribution profiles of matriptase expression in human skin and oral epithelium with those in an in vitro skin organotypic culture model in which the dynamic processes of robust epidermal differentiation and tissue regeneration can be tracked.

116 The state of matriptase zymogen activation was also investigated in the human skin using a unique matriptase mAb that can distinguish between the activated matriptase from the inactivated matriptase zymogen. Our data revealed that human matriptase is expressed in both basal and spinous keratinocytes in skin. In addition, the process of loss of matriptase in the terminally differentiated keratinocytes was clearly demonstrated in the skin organotypic culture. Interestingly, zymogen activation of matriptase was seen in the keratinocytes only from the bulge area of hair. The distribution profile of human matriptase suggests that human matriptase is likely to be involved in the regulation of keratinocyte growth and early differentiation, both of which can be corroborated by the reduced growth and expression of differentiation markers in HaCaT human keratinocytes along with the reduction of matriptase expression. The distribution profile is in a stark contrast to mouse matriptase which is expressed by suprabasal but not basal keratinocytes with highest expression in the outermost liable keratinocytes. The state of matriptase zymogen activation provides possible mechanistic insights to the pathology ARIH and

IFAH, which could involve the dysregulated matriptase activity in skin stem cells.

4.3 Material and Methods

4.3.1 Cell lines and Cell cultures

117 Human primary keratinocytes were obtained from neonatal foreskin and maintained in Keratinocyte-SFM (Invitrogen) supplemented with bovine pituitary extract (20-30

g/ml), recombinant epidermal growth factor (rEGF, 0.1-0.2 ng/ml), 100 units/ml penicillin, and 100 g/ml streptomycin. Human keratinocyte HaCaT cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Mediatech Inc., Manassas,

VA) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Gemini, West

Sacramento, CA), 100 units/ml penicillin, and 100 g/ml streptomycin. HaCaT cells with matriptase-targeting and non-targeting control small hairpin RNAs (shRNA) were generated as previously described 193 and maintained in DMEM supplemented with 10%

FBS, 2 g/ml puromycin, 100 units/ml penicillin, and 100 g/ml streptomycin. The stably basal-like state HaCaT cells were generated by culturing HaCaT cells in completed

Keratinocyte-SFM, which contains only 0.03mM calcium, for at least one month as previously described 277, 278. This stably basal-like state HaCaT cells were switched from low calcium completed Keratinocyte-SFM to high calcium completed DMEM in order to induce differentiation 277. All cells were incubated in a humidified incubator at

37℃ with 5% CO2.

4.3.2 Antibodies

The total matriptase monoclonal antibody (mAb) M24 and the activated matriptase

118 mAb M69 were generated using matriptase-HAI-1 complex as described previously 155.

The total matriptase mAb M24 recognizes both latent and activated forms of matriptase

155, 160, 177, 193. Monoclonal antibody M69 is a unique matriptase antibody which recognizes an epitope only present on activated matriptase and therefore is able to distinguish activated matriptase from latent matriptase. Human HAI-1 protein was detected using the HAI-1 mAb M19 155, 193.

4.3.3 Immunoblot analysis

Massive matriptase activation in HaCaT cells was induced by applying a 0.15M phosphate buffer, pH 6.0 at room temperature (RT) for 30 minutes, as described previously 177, 193. Proteins for immunoblot analyses were prepared from cell cultures by washing the cells with dulbecco’s phosphate-buffered saline (DPBS) for three times followed by dissolving cells in 1% Triton X-100 in DPBS. Insoluble debris was removed by centrifugation, and the protein concentration was determined using Bradford

(Sigma-Aldrich, St. Louis, MD) according to the manufacturer’s instructions. Samples containing the same amount of total protein were diluted with 5X sample buffer in the absence of reducing agent and resolved by 7.5% SDS-polyacrylamide gel electrophoresis

(SDS-PAGE) followed by transferring to nitrocellulose membrane (Pall Corp., Pensacola,

FL). The membranes were probed with the desired monoclonal antibodies and an

119 HRP-conjugated secondary antibody (Kirkegaard & Perry Laboratories, Gaithersburg,

MD) followed by signal detection with Western Lightning Chemiluminescence Reagent

Plus (PerkinElmer Life Sciences, Waltham, MA).

4.3.4 Organotypic culture

Organotypic rafts were generated based on the protocol described previously with minor revision 193, 264, 265. Briefly, Type I collagen (BD Biosciences, Bedford, MA) plugs with 7.5x104 human primary fibroblasts per plug were prepared in 6-well cell culture inserts (BD Biosciences) and incubated for 2-5 days. A total of 106 primary keratinocytes were applied to each center of collagen plug and incubated for 1-4 days.

Differentiation of the rafts was initiated by lifting them to the air-liquid interface.

4.3.5 Immunohistochemistry

Immunohistochemical staining was performed as previously described with minor modification (Dako, CA) 279. Paraffin-embedded human skin equivalents sections and normal human skin sections were stained using the total matriptase mAb M24, activated matriptase mAb M69, and HAI-1 mAb M19. HRP-labeled anti-mouse IgG and the substrate, diaminobenzidine (DAB), were used for the detection of positive staining.

Cell nuclei were counterstained with hematoxylin. Images were captured using an

Olympus AH2 Vanox Microscope System (Olympus, Melville, NY).

120 4.3.6 Crystal violet cell proliferation assay

HaCaT cells bearing matriptase-targeting and non-targeting were seeded into 6-well plates at 104 cells per well. Cells were incubated overnight for attachment and then fixed for 10 min in a solution of buffered formalin (3.7%, Fisher Protocol*) on desired days. The cells were washed with DPBS and subsequently stained with 0.5% crystal violet solution containing 20% methanol in DPBS for 10 min. After removing excess

stain by washing cells with ddH2O, the crystal violet stained cells were dissolved in 2 ml of a 1% SDS solution and agitated plates on an orbital shaker until color is uniform with no areas of dense coloration in bottom of wells. The optical density of the extracted dye was read with a spectrophotometer (Bechman, DTX 880) at 595 nm. The optical density measurements give a relative number of viable cells present at the time the dye was added and were normalized to the reading on Day 1.

4.3.7 Real-time polymerase chain reaction (PCR)

For each sample, total RNA was extracted from cells using Trizol reagent

(Invitrogen) following the manufacturer’s instructions. Briefly, chloroform was added to homogenized samples followed by centrifugation to isolate the aqueous layer. RNA was precipitated and then washed in 70% ethanol three times. The RNAs were air-dried

and dissolved in ddH2O. One microgram of total RNA was treated with DNase and

121 used for cDNA synthesis. The cDNA synthesis step included a control reaction without reverse transcriptase. Real-time PCR reactions were conducted using SYBR green and gene-specific primers (Supplementary Table S1) and normalized to GAPDH as an internal control. The real-time PCR reactions were as follows: for 2 min at 50℃ and for

10 min at 95℃, followed by 45 cycles of switching between 95℃ for 15 sec and 60℃ for

1 min.

4.4 Results

4.4.1 Matriptase expression and distribution in skin organotypic culture

Primary human keratinocytes maintained in low Ca2+ culture medium express both matriptase and HAI-1 as determined by immunoblot analysis when grown in monolayer

(Figure 4.1A). These data suggest that the proliferating keratinocytes express the cell surface serine protease and its cognate inhibitor. In order to study whether and how epidermal differentiation regulates the expression of matriptase and HAI-1, we examined the distribution profile of the expression of matriptase and HAI-1 using organotypic skin cultures. In this in vitro skin model system, the proliferating keratinocytes are induced to undergo differentiation by growing them in medium containing high Ca2+ and on the interface between fibroblast-primed collagen matrix and air (Figure 4.1B and C). The

122

Figure 4.1 Expression of matriptase and HAI-1 in human primary keratinocyte and cellular localization of matriptase and HAI-1 in skin equivalents.

A. Protein expressions of matriptase and HAI-1 in human primary keratinocyte were analyzed by immunoblot using total matriptase mAb M24 and HAI-1 mAb M19, respectively. B and C. Cellular localization of total matriptase (Total MTP) and HAI-1

(HAI-1) in skin equivalents. The skin equivalents on Day 0, 3, 5, and 9 after lifting them to an air-liquid interface were used to detect the distribution of matriptase and

HAI-1, shown in brown. The skin equivalents were supported on a collagen/fibroblast matrix and nuclei were counterstained blue with hematoxylin. Scale bar: 20 m.

123 skin raft replicates the processes of epidermal differentiation and tissue regeneration.

Both matriptase and HAI-1 are still expressed by primary human keratinocytes after their growth in the epi-medium that contains 1.5 mM Ca2+ and in contact with fibroblast-primed collagen matrix for some of these keratinocytes for 3 days. Both matriptase and HAI-1 were detected on the cell surface in all layers of the cells on day zero when the keratinocytes were lifted to the air-liquid interface (Figure 4.1B and C,

Day 0). These data suggest that the keratinocytes at the earlier differentiation stage induced by Ca2+ and the contact with collagen matrix, apparently still maintain the serine protease system. However, by Day 3, loss of matriptase expression begins to occur at the outermost layer of keratinocytes whose nuclei have become slightly condensed and elongated (Figure 4.1B, Day 3). By Day 5, when there is robust differentiation, the expression of matriptase and HAI-1 is confined to the keratinocyte layers nearest the collagen matrix (Figure 4.1B and C, Day 5). There was no detectable matriptase but

HAI-1 was still seen at reduced levels in the superficial layers which contain the keratinocytes either with condensed nuclei or without nuclei. By Day 9, the skin raft contains basal cells with altered orientation and thicker cornified layers, but expression of matriptase and HAI-1 were maintained at the same level. This analysis reveals that expression of matriptase and HAI-1 is differentiation-dependent. They are highly

124 expressed in proliferating and earlier differentiating keratinocytes which then decrease to their lost expressions in the late stage of differentiation.

The distribution profile of matriptase expression at the Day 9 of the keratinocyte raft culture is remarkably similar to that observed in the stratified epithelium in vivo (Figure

4.2). In human skin epidermis, matriptase is expressed at high levels at cell-cell junctions in the basal layer (SB) and spinous layer (SS). Decreasing expression of matriptase was observed along with the differentiational progression from the SS to the granular layer (SG). The expression becomes undetectable in those keratinocytes with dense cytoplasm (SG), bordering on the cornified layer (SC), in which matriptase was also negative (Figure 4.2B). Despite the abundance of matriptase in epidermis, we did not detect activated matriptase at meaningful levels in all layers of epidermis, while some positive staining was occasionally seen or using much higher concentrations of this mAb

(Figure 4.2D). The loss of matriptase expression in the late stage of keratinocyte differentiation is also seen in the oral epithelium (Figure 4.3A). Oral mucosa contains stratified epithelium that share similar tissue architecture to skin epidermis. Matriptase was detected in the SB and the SS but not in the keratinocytes with the condensed and elongated nuclei from the surface layer (SL). The distribution profile suggests that the physiological function of matriptase in stratified epithelium likely lies in the SB and the

125

Figure 4.2 Cellular localization of matriptase and HAI-1 in normal human skin.

Frozen sections from normal human skin were stained with hematoxylin and eosin (H&E) or immunostained with desired monoclonal antibodies. A. H&E staining of normal human skin. B. Immunoreactivity of total matriptase detected using mAb M32 (Total

MTP). C. Immunoreactivity of HAI-1 detected using mAb M19. D. Immunoreactivity of activated matriptase detected using mAb M69. Nuclei were counterstained blue with hematoxylin. A negative control using Mouse IgG was also performed (not shown).

Scale bar: 25 m. SC: stratum corneum. SG: stratum granulosum. SS: stratum spinosum. SB: stratum basale.

126

Figure 4.3 Cellular localization of matriptase and HAI-1 in human normal oral mucosa.

Immunohistochemical analyses of paraffin-embedded human oral mucosa using total matriptase mAb M24 (A. Total MTP) and HAI-1 mAb M19 (B. HAI-1). A negative control using Mouse IgG was also performed (not shown). Nuclei were counterstained blue with hematoxylin. Scale bar: 50 m. SS-SL: stratum spinosum-surface layer. SS: stratum spinosum. SB: stratum basale. 127 SS rather than the late stage of differentiation.

The distribution profile of HAI-1 expression in the in vitro raft culture and in vivo human epidermis and oral epithelium is almost identical to matriptase, suggesting that the close functional linkage between the protease and the anti-matriptase are well maintained during the differentiation process. An interesting exception was, nevertheless, noticed.

While matriptase expression in human skin is decreasing in the SG, HAI-1 was detected at high levels in this epidermal layer. Furthermore, the subcellular distribution of HAI-1 appears to be altered in these keratinocytes. While HAI-1 was seen at the cell-cell junctions in most keratinocyte layers, the protease inhibitor was detected in the cytoplasm of keratinocytes in the SG. These data suggest that HAI-1 might still play an active role in the late stage of epidermal differentiation by inhibiting other proteases than matriptase, likely in the granules of the SG.

4.4.2 Elevated matriptase activation during epidermal tissue regeneration and in the epidermal stem cell

Like most serine proteases, matriptase is synthesized as a single-chain zymogen and acquires its full proteolytic activity only after converting into a two-chain active form.

The zymogen activation of matriptase is accompanied with a localized conformational change in the serine protease domain and we have generated a set of monoclonal

128 antibodies that apparently recognized these conformational changes and can distinguish the activated matriptase from matriptase zymogen. Furthermore, matriptase zymogen activation is closely followed by the inhibition of the newly generated active matriptase by HAI-1. As a consequence, the end product of matriptase zymogen activation is the

120-kDa matriptase-HAI-1 complex. The specificity of the activated matriptase mAb can be demonstrated by a combination of induction of matriptase activation/inhibition and immunoblot analysis using matriptase mAbs and HAI-1 mAb (Figure 4.4A).

HaCaT keratinocytes express matriptase in its zymogen form as a 70-kDa protein band detected by matriptase mAb (Figure 4.4A, Total MTP, lane 1). Matriptase zymogen in response to extracellular acidosis can be converted into active matriptase that is rapidly inhibited by HAI-1 by forming an inactive complex at 120-kDa. The 120-kDa matriptase-HAI-1 complex can be detected by both matriptase and HAI-1 mAbs (Figure

4.4A, lanes 2). The unique activated matriptase mAb detected only the 120-kDa matriptase-HAI-1 complex, but not matriptase zymogen nor the HAI-1-prostasin complex

(Figure 4.4A, Activated MTP, lane 2). Using this unique activated matriptase monoclonal antibody, we have examined the activation status of matriptase in human skin.

In contrast to the negative staining of activated matriptase in the epidermis, a positive staining of activated matriptase was frequently seen at the bulge area of hair follicles

129

Figure 4.4 Elevated matriptase activation in epidermal stem cells of hair follicle.

A. Acid induced activation of matriptase in immortalized human keratinocytes. HaCaT cells were incubated with DPBS (lanes 1) of phosphate buffer pH 6.0 (lanes 2) for 30 minutes. Cell lysates were extracted and subjected to immunoblot analyses for total matriptase (Total MTP), activated matriptase (Activated MTP), and HAI-1 (HAI-1). B.

Massive matriptase activation detected in the bugle region of human hair follicle.

Frozen sections from normal human skin containing hair follicle were stained with a total matriptase mAb M24, a HAI-1 mAb M19, or a matriptase mAb M69 that specifically recognizes activated matriptase only. A negative control using Mouse IgG was also performed. Nuclei were counterstained blue with hematoxylin. C. A higher magnification for the micrograph showed in black square in Figure 4B, Activated MTP, is presented.

130 (Figure 4.4B) which contains the stem cells of the skin. A higher magnification of the bulge area revealed that the activated matriptase was localized on the cell membrane

(Figure 4.4C). This data suggest that while matriptase is not particularly active in the resting state of the human skin, the membrane-bound serine protease becomes very active and could play important roles in the epidermal stem cells.

4.4.3 Role of matriptase in keratinocyte proliferation and differentiation

A diminishing matriptase expression level in the granular layer of the human skin indicates that the human matriptase may not play an active role during the late stages of epidermal differentiation. Instead, the protease may participate in multiple functions associated with the basal and spinous keratinocytes, such as proliferation and early differentiation in human epidermis. This hypothesis is further supported by the enhanced matriptase activation in the epidermal stem cells (Figure 4.4). To test this hypothesis, we first generated HaCaT clones in which the level of matriptase expression had been suppressed by transduction with lentiviruses carrying matriptase-targeting shRNAs, or non-targeting controls (Figure 4.5A). When the growth rates of both

HaCaT clones were compared to the matriptase knockdown clone, we have observed a slower growth rate than the non-target control clone by 37.5% at day 5 (Figure 4.5B).

These data suggest that the level of matriptase could affect the growth of human

131

Figure 4.5 Role of matriptase in keratinocyte differentiation and proliferation.

A. Matriptase knockdown cell lines. Expression of matriptase and HAI-1 in the cell lines with matriptase-targeting (49 and 52) or non-targeting control (NT) shRNA were examined by immunoblot analyses using total matriptase mAb M24 (Total MTP) and

HAI-1 mAb M19 (HAI-1). B. Decreased proliferation in matriptase knockdown cells.

Growth curves of HaCaT cells with matriptase-targeting (MTP knockdown) and non-targeting control (Non-target) shRNAs estimated by the crystal violet staining assay.

Each point corresponds to the mean and SEM of experiments carried out in quadruplicate.

The results presented are from one experiment representative of the four performed. C.

Matriptase plays a role in early differentiation of human keratinocytes. Real-time PCR analyses of matriptase, HAI-1, K1, K10, and involucrin mRNA expression in the basal-like and differentiated HaCaT cells with matriptase-targeting (MTP KD) and

132 non-targeting control (NT) shRNAs are shown as bar graphs. Bar graphs representing the fold changes of mRNA levels quantified by normalization the controls as indicated on the figure. Mean values ± SD (n=4).

133 keratinocytes.

To test whether matriptase expression might affect the differentiation of human keratinocytes, the expression levels of differentiation marker mRNAs, including keratin 1

(K1), keratin 10 (K10), and involucrin were determined in the matriptase knockdown and the non-target clones. For this testing, both HaCaT clones were maintained in media containing high or low calcium ions to force cells into proliferating or differentiating states, respectively 277. The expression levels of both keratins 1 and 10 were observed in the non-target clones maintained only in media with high calcium, but not in media with low calcium (Figure 4.5C, panels c and d). Likewise, the high calcium media also increased the expression of involucrin (Figure 4.5C, panel e). These data confirmed the differentiation effect of calcium and the differentiation status of these cells. More importantly, the levels of both keratins were significantly lower in the matriptase knockdown clones as compared to the non-target control clone (Figure 4.5 C, panels c and d). Likewise, high calcium failed to increase the expression of involucrin in the matriptase knockdown clone (Figure 4.5C, panel e). These data suggest that matriptase could affect the differentiation of human keratinocytes. The expression of matriptase itself appears to be regulated by differentiation since the higher matriptase expression was observed in keratinocytes grown in media containing high calcium (Figure 4.5C,

134 panel a). In contrast, the expression of HAI-1 was not affected as much as these differentiation markers by matriptase knockdown and calcium-induced differentiation.

4.5 Discussion

The studies of human genetic disorders reveal the importance of matriptase in human skin. In the current studies, we have provided mechanistic insights regarding the roles played by matriptase in human skin. By considering that the highest levels of matriptase zymogen activation were observed in the keratinocytes residing in the bulge area of hair follicles of the human skin, matriptase activity might be actively utilized by skin stem cells. It is interesting to speculate that matriptase dysregulation mutations in patients with ARIH or IFAH might be distantly related to the reduced matriptase activity in these keratinocytes around the bulge area. Matriptase is highly expressed by the keratinocytes in the basal and spinous layers of the epidermis, while much less activated.

Oral epithelium resembles the epidermis in the distribution profile of matriptase expression. The basal layer contains three subpopulations of keratinocytes: 1) stem cells, 2) transit amplifying cells, and 3) post-mitotic differentiating cells 11, 17.

Expression of matriptase among these subpopulations of basal keratinocytes might not show obvious difference as a relatively homogenous expression of matriptase among all

135 keratinocytes in the basal layer. The major function of the basal keratinocytes is through mitosis to maintain the stem cell population and produce daughter cells to replenish the epidermis through differentiation. Once differentiated, the keratinocytes move toward the spinous layer, and continue to differentiate from one layer to the other to the final layer, stratum corneum 7, 14. Given the high expression of matriptase in the first two epidermal layers, it is plausible to hypothesize that matriptase might be involved in growth regulation and early differentiation of keratinocytes. Our functional analysis in which the matriptase expression was reduced by shRNA, supports this hypothesis as the keratinocytes expressing less matriptase grow more slowly and express much less of the differentiation markers, K1, K10 and involucrin in response to the differentiation agent

Ca2+ than the control (Figure 4.5C). The reduced matriptase zymogen activation in the basal keratinocytes, as compared to the bulge cells, could result from the fact that the stem cells in the basal layer of epidermis remain quiescent and have a long cell cycle 280.

We have shown in previous studies that matriptase zymogen activation is turned on rapidly and robustly early in skin organotypic culture. This suggests that matriptase activity might play an important role in the epidermal tissue regeneration and contribute to the early differentiation of keratinocytes 193. Furthermore, increased matriptase zymogen activation in the basal keratinocytes has been observed in several distinct

136 human skin diseases and the increased zymogen activation might be caused by the inflammation associated with these diseases 196. These data suggest that matriptase activity is dynamically regulated in response to the disease development and progression.

Hence, the fact that active matriptase zymogen activation is not seen in the resting epidermis should not be used as evidence to suggest a dispensable role of matriptase in skin. Previously, we have shown that matriptase zymogen activation can be induced by a variety of extracellular stimuli, including reactive oxygen species and extracellular acidosis 177, 196. These known and possible other unknown stimuli of matriptase zymogen activation might be produced over the course of epidermal tissue regeneration and skin disease development. In the resting skin, these extracellular stimuli might be limited and, likewise, matriptase zymogen activation is turned on only at very limited rate.

Taken together, our previous and current data suggest that in the established and resting skin, matriptase might function in the emergency program system in response to acute conditions rather than as a housekeeping activity to maintain epidermal homeostasis.

In contrast to the roles played by matriptase in the epidermal stem cells and the basal and spinous keratinocytes, the TTSP might be less involved in the keratinocyte activity associated with the late stage of keratinocyte differentiation in the granular layer and the desquamation in the cornified layers. Expression analysis of human skin and oral

137 mucosa reveals a decreasing matriptase expression when keratinocytes travel superficially toward the granular layer which is completely lost in the cornified layer of skin and the surface layer in oral epithelium. This conclusion is supported by the results from the in vitro skin organotypic culture in which matriptase expression in keratinocytes is dynamically regulated and begins to disappear when the keratinocytes become differentiated with condensed and elongated nuclei (Figure 4.1B). The loss of matriptase expression at the late stage of differentiation suggests that the protease is less important for the proteolytic events involved in profilaggrin processing and degradation of corneodesmosomes, both of which have been observed in ARIH and IFAH patients.

Despite the unusually tight functional and regulatory linkage between matriptase and

HAI-1 not only in the keratinocytes but also in other type of epithelial cells, HAI-1 might play active roles in the terminal differentiation of keratinocytes in the granular layer via inhibition of other serine proteases than matriptase. The intracellular punctate staining of HAI-1 in the granular keratinocytes suggests that the Kunitz inhibitor might be incorporated into keratohyalin granules and/or the lamellar granules during the processes of the terminal differentiation in the granular layer. In human skin, HAI-1, therefore, might have a better chance than matriptase in directly regulating profilaggrin processing in the keratohyalin granules and lipid matrix formation in the lamellar granules.

138 In summary, we have provided several lines of evidence to support the hypothesis that dysregulated matriptase in ARIH and IFAH could occur in the skin stem cells and/or in the basal and spinous keratinocytes. This is in contrast to the conclusion drawn from the mouse models with matriptase deficiency, in which the suprabasal keratinocytes, particularly the transition cells, are responsible for the defects in the stratum corneum barrier.

139

Chapter 5 – Plasminogen-dependent matriptase zymogen activation accelerates pericellular plasmin generation in differentiating primary human keratinocytes

140 5.1 Abstract

Pericellular plasmin generation through urokinase-type plasminogen activator (uPA) and its membrane receptor (uPAR) in keratinocytes plays important roles in wound healing. Plasminogen activation can be rapidly initiated by the type 2 transmembrane serine protease matriptase that activates pro-uPA and shortens the lag phase which occurs prior to the burst of plasmin generation. Matriptase is synthesized as a zymogen and converted into an active enzyme through autoactivation, which can be induced by extracellular stimuli in a somewhat cell type-specific manner. The functional linkage between matriptase and plasminogen is initially thought to be one-directional: from matriptase through uPA to plasminogen. However, in the current study, we have found that primary human keratinocytes primed with Ca2+ become responsive to serum for rapid matriptase activation through plasminogens. Sub-pM plasminogen is sufficient to induce matriptase activation, and acting on the cell surface, suggesting an outside-in signaling mechanism. Along with matriptase activation, a proportion of active matriptase is shed into extracellular milieu and returns to the cell surface to accelerate plasminogen activation. The novel activity of plasminogen as a matriptase activation inducer and subsequently as an accelerator of plasmin generation reveals that the three serine proteases form a pericellular protease circuit under a positive feedback loop

141 regulation. Given that both uPA and uPAR are upregulated in keratinocytes of the outgrowing epidermal sheet and that plasminogen works in concert with Ca2+ to rapidly and robustly stimulate keratinocytes to activate matriptase, this unusual combination of the prominent keratinocyte differentiation agents and the proteolytic system provides keratinocytes with a timely and effective way to generate plasmin at an early stage of wound healing. In normal intact human skin, the established Ca2+ gradient and the lack of expression of uPA and uPAR would keep the proteolytic system inactive. These well-coordinated regulatory mechanisms allow keratinocytes to precisely generate plasmin when needed.

5.2 Introduction

During epithelial wound healing, one critical step for the correct epithelialization of keratinocytes is to precisely switch on the plasminogen activator (PA)/plasmin system, which starts from activation of PA and leads to the generation of plasmin from its zymogen, plasminogen. Plasmin, a potent serine protease, is not only required for the dissolution of fibrin clots and removal of necrotic tissue along the path of keratinocyte migration during wound healing, but is also involved in migration of keratinocytes and inflammatory cells, activation and release of growth factors 281, 282, activation of matrix

142 metalloproteases 210, angiogenesis 283, and collagen remodeling 284, 285. The regulation of the PA/plasmin system plays multiple roles which are critical for each step of the healing process, and any disruption in this tightly controlled balance will lead to abnormal healing, such as chronic wounds. By restricting the expression of urokinase

PA (uPA, the major PA in keratinocytes) and its membrane receptor, uPAR, on keratinocytes only in the outgrowing epithelial sheet in healing wound 286-290, plasminogen activation is able to temporally and spatially respond to tissue damage and specifically target to the subsequent repair processes. However, the exact mechanism by which what initiates the PA/plasmin system is still unclear. Fulfilling the regulatory mechanism of this delicate protease cascade might allow us to reveal the possible causes and treatments of abnormal healing.

The rapid initiation of the PA/plasmin system is believed to rely on the very low intrinsic enzymatic activity of uPA to generate the initial undetectable plasmin from plasminogen 291. Plasmin itself is a potent activator of pro-uPA (the zymogen of uPA).

The newly generated plasmin, therefore, coverts pro-uPA to active uPA and initiates reciprocal zymogen activation. As a result, a burst of plasmin generation follows a lag phase owing to the need for the accumulation of sufficient plasmin for detection.

Although the generation of plasmin can be achieved by reciprocal zymogen activation,

143 alternative mechanisms for the initiation of the system are not excluded. By demonstrating distinct patterns of plasminogen activation in two monocytic cell lines, a recent study revealed that the type II transmembrane serine protease matriptase can generate active uPA through a plasmin-independent mechanism 162. Matriptase is also synthesized as a single-chain zymogen and is activated by cleavage at the activation motif and the formation of a two-chain active enzyme. Though this canonical activation process is conserved in the vast majority of serine proteases, including uPA and plasminogen, activation of matriptase is characterized by an autoactivation. The autoactivation is a unique feature of the initiator proteases in protease cascades.

Therefore, as the initiator protease, matriptase activation can be triggered via microenvironmental changes and cell type-selective extracellular stimuli. For example, the intrinsic activity of matriptase zymogen can be enhanced under mildly acidic environment 176; hence, matriptase activation can be seen in cells under acidic environments, such as during inflammation 196. Moreover, exposing cells to mildly acidic buffer represents a ubiquitous way to rapidly and robustly stimulate matriptase activation in almost all matriptase-expressing cells 163, 193, 261, 273, 292. In addition to extracellular acidosis, matriptase-expressing cells can also activate matriptase in response to cell type-specific extracellular stimuli, such as exposing lysophospholipids to

144 mammary epithelial cells and androgens to LNCaP prostate cancer cells 182, 184.

Matriptase is also a well-known activator for pro-uPA, though its intrinsic activity to activate pro-uPA is relatively poor compared with plasmin 162. However, matriptase is responsible for the specific activation of uPAR-bound pro-uPA at the cell surface, and can significantly decrease the lag phase required for the burst of plasminogen activation

162. The role of matriptase in plasmin generation through activation of uPAR-bound pro-uPA, the temporal and spatial expression of uPA and uPAR during wound healing, and the high expression of matriptase in the basal and spinous keratinocytes of human skin (Chapter 4) indicate a matriptase-uPA-plasmin pericellular protease cascade might exist in keratinocytes, and this protease cascade might be particularly triggered in the process of wound healing. As the initiator protease, matriptase activation will be the checkpoint for the initiation of the protease cascade. The keratinocyte-selective extracellular stimulus for matriptase activation would, therefore, play the pivotal role in the control of the pericellular plasmin generation. In the current study, we revealed a novel extracellular stimulus of matriptase activation in keratinocytes, which revises the concept of how the cells initiate the PA/plasmin system. The discovery that primary human keratinocytes can rapidly activate matriptase in response to serum treatment following priming of the cells with Ca2+ into a differentiating state, led us to discover the

145 identity of the serum factor. Surprisingly, the serum factor that induces matriptase activation was identified to be a plasminogen. Moreover, plasmin fails to induce matriptase activation which suggests that the enzymatic activity of plasmin is not the reason for activation of matriptase. Given that plasminogen is able to induce rapid matriptase activation at picomolar concentrations, as low as 0.5 pM in as quickly as 30 min, plasminogen appears to transduce outside-in signaling for the rapid matriptase activation. Along with the activation of matriptase, active matriptase is shed into extracellular milieu, and returned back to the cell surface to facilitate the activation of

PA. As a consequence, the lag phase of the PA/plasmin reciprocal zymogen activation is shortened. In summary, our study reveals that the matriptase-uPA-plasmin protease cascade in differentiating keratinocytes can be operated as a pericellular protease circuit with a positive feedback loop, in which plasminogen serves as the initiating signal and plasmin serves as the end product. This tightly regulated protease cascade with the positive feedback mechanism would confer to keratinocytes a timely and efficient way to generate plasmin in response to the influx of blood at the onset of skin wounding.

5.3 Material and Methods

5.3.1 Reagents

146 Human serum and fetal bovine serum (FBS) were obtained from Gemini (West

Sacramenyo, CA). Lysine sepharose, DEAE sepharose fast flow, HiPrep 26/10 desalting column, and HiPrep Q XL 16/10 column were purchased from GE Healthcare

(Piscataway, NJ). Cibacron blue 3GA, human plasmin, Bradford Reagent and calcium chloride were purchased from Sigma-Aldrich (St. Louis, MD). Actinomycin D, cycloheximidine, and Ro-31-8220 were purchased from Calbiochem (Boston, MA).

BAPTA-AM was obtained from Santa cruz biotechnology, Inc. (Santa Cruz, CA).

LY294002 was purchased from Cell signaling technology (Danvers, MA).

Boc-Gln-Ala-Arg-AMC was purchased from Enzo Life Sciences (Farmingdale, NY) and

Boc-Val-Leu-Lys-AMC was obtained from Bachem Americas, Inc. (Torrance, CA).

Protoblue was obtained from National Diagnostics (Atlanta, GE). 6-aminohexanoic acid was purchased from LKT laboratories, Inc. (St. Paul, MN). Human plasminogen was obtained from Technoclone (Vienna, Austria). Matriptase small molecule inhibitors,

F3226-1197 and F3226-1198, were purchased from Ambinter (Paris, France).

5.3.2 Cell culture

Human primary keratinocytes were isolated from foreskin and maintained in the

Keratinocyte SFM (Invitrogen) supplemented with 20-30 g/ml bovine pituitary extract,

0.1-0.2 ng/ml recombinant epidermal growth factor, 100 units/ml penicillin, and 100

147 mg/ml streptomycin. Immortalized basal-like human keratinocytes HaCaT were generated as previously described 277, 278. Briefly, HaCaT cells were switched from a high calcium medium, Dulbecco’s Modified Eagle Medium (Mediatech Inc., Manassas,

VA) supplemented with 10% fetal bovine serum (FBS), to a low calcium medium,

Keratinocyte SFM medium for at least one month. All cells were maintained in a

humidified incubator at 37℃ with 5% CO2.

5.3.3 Purification of serum factor and identification

The factor(s) in the FBS which can induce matriptase activation was purified using conventional chromatographs as shown in Figure 5.7A. FBS was loaded onto Cibacron

Blue 3GA column followed by washing with DPBS ((Mediatech Inc., Manassas, VA ) and eluting with 0.5 N NaCl in DPBS. The eluate was desalted and changed buffer to

20 mM Tris-HCl, pH 7.4 at 25℃ using a HiPrep 26/10 desalting column. The desalted sample was loaded onto a DEAE-sepharose fast flow column pre-equilibrated with 20 mM Tris-HCl, pH 7.4 followed by extensively wash. Proteins bound to DEAE sepharose were eluted with a linear gradient of 0 to 0.1 N NaCl in 20 mM Tris-HCl, pH

7.4. The protein band from SDS-PAGE gels stained with Protoblue was excised, washed, destained, and trypsinized overnight at 37 ℃ using standard protocols.

Analysis of the tryptic peptides derived from the protein sample was performed by liquid

148 chromatography/mass spectrometry (LS/MS) using services provided by Prottech Inc.

(Norristown, PA).

5.3.4 Immunoblot

The concentration of proteins was determined using the Bradford Reagent. Same amount of total proteins were mixed with 5X SDS loading buffer in the absence of reducing agents followed by resolving by 7.5% SDS-polyacrylamide gel and transferred to nitrocellulose membranes (Pall Corp, Pensacola, FL). The membranes were probed with a monoclonal antibody (mAb) M24, which can recognize both the zymogen form and activated form of matriptase. A HRP-conjugated secondary antibody (Kirkegaard &

Perry Laboratories, Gaithersburg, MD) was applied to the membranes followed by signal detection with Western Lightning Chemiluminescence Reagent Plus (PerkinElmer Life

Sciences, Waltham, MA).

5.3.5 Protease activity assay

The activities of plasmin and active matriptase were measured by the liberation of

7-Amino-4-methylcoumarin (AMC) from the fluorescent substrates using a fluorescent spectrophotometer (Bechman, DTX 880) with excitation at 360 nm and emission at 480 nm. For matriptase activity, primary human keratinocytes were incubated with basal medium, as control, or a pH 6.0 buffer to induce matriptase activation for 30 min at room

149 temperature followed by neutralizing the pH back to 8.0. The conditioned buffers were used to measure matriptase activity using the matriptase synthetic fluorescent substrate,

Boc-Gln-Ala-Arg-AMC. For plasmin activity in Figure 5.10B, primary human keratinocytes were incubated with a pH 6.0 buffer for 30 min at room temperature followed by neutralizing the pH back to 8.0 and addition of 50 nM plasminogen. The plasmin activity was monitored by a plasmin synthetic fluorescent substrate,

Boc-Val-Leu-Lys-AMC. The plasmin activities in the rest of the figures were measured under the treatments described in figure legends using a plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC.

5.4 Results

5.4.1 Interplay of calcium ions and serum stimulates primary human

keratinocytes to activate matriptase.

Previously we have demonstrated that the matriptase-prostasin proteolytic cascade is activated at the early stage of epidermal differentiation and tissue regeneration in a

3-dimensional skin organotypic culture using primary human keratinocytes 193. In order to identify the factors that stimulate activation of the cell surface protease cascade, we have systematically analyzed the components of the organotypic culture for the

150 stimulating activity of matriptase activation in a 2-dimentional culture system. We noticed that a combination of the epidermal differentiation agent, calcium ions (Ca2+), and serum are sufficient to stimulate primary keratinocytes to activate matriptase in monolayer culture (Figure 5.1A, lane 3). Like other epithelial cells in which matriptase is expressed primarily in its mature zymogen form that is detected as a 70-kDa protein band by Western blot, primary human keratinocytes grown in low Ca2+ medium also express matriptase in its zymogen form (Figure 5.1 A lane 1). In contrast to skin organotypic culture, activation of matriptase was not induced by the differentiation agent

Ca2+ alone in monolayer culture (Figure 5.1A lane 2). When low levels of fetal bovine serum (2%) was added to the culture medium containing 0.2 mM Ca2+, activation of matriptase becomes evident as shown by the appearance of (activated) matriptase-HAI-1 complex with a size of 120-kDa (Figure 5.1A, lane 3). The serum factor(s) that stimulates matriptase activation is apparently not a lipid-based small molecule and cannot be absorbed by charcoal (Figure 5.1A, lane 4), but is susceptible to heating (Figure 5.1A, lane 5). These data suggest that in the presence of epidermal differentiation agent Ca2+ and a serum factor(s), likely a protein, can activate matriptase in primary human keratinocytes. Interestingly, immortalized basal-like human keratinocytes HaCaT do not respond to the combination of Ca2+ and FBS to activate matriptase (Figure 5.1B).

151

Figure 5.1 The serum factor-induced activation of matriptase is a primary keratinocyte specific event.

A. Human primary keratinocytes were incubated overnight with 0.09 mM calcium (C), or

0.2 mM calcium (Ca), or 0.2 mM calcium plus 2% FBS (FBS), or 0.2 mM calcium plus

2% charcoal-stripped FBS (Charcoal FBS), or 0.2 mM calcium plus 2% heat-treated FBS

(Boiled FBS) followed by immunoblot analyses using a total matriptase mAb M24. B.

The basal-like human immortalized keratinocyte HaCaT cells were incubated overnight with 0.09 mM calcium (C), or 0.2 mM calcium (Ca), or 0.2 mM calcium plus 2% FBS

(Ca+FBS) followed by immunoblot analyses using a total matriptase mAb M24.

152 A combination of Ca2+ and FBS is apparently required for significant matriptase activation. When keratinocytes were grown in medium supplemented with Ca2+ alone up to 1.5 mM or with FBS alone up to 2%, activation of matriptase was not detected or detected at very low levels (Figure 5.2). In contrast, a combination of both stimulates significant matriptase activation (Figure 5.2). These data suggest that Ca2+ and serum work in concert to stimulate keratinocytes to activate matriptase.

The sequence of treatment with Ca2+ and serum is also important for matriptase activation. Activation of matriptase can only be induced by pre-treating the cells with

Ca2+ followed by serum; a reverse treatment with serum first followed by Ca2+ failed to stimulate matriptase activation (Figure 5.3A). These data suggest that Ca2+ and the serum factor(s) play different, but coordinated roles in stimulation of matriptase activation. The differential role of Ca2+ and serum were further revealed by the different time requirements for Ca2+ and serum. After the cells were pretreated with Ca2+ (0.2 mM) overnight, it took as little as 30 min for serum to stimulate matriptase activation and the level of matriptase-HAI-1 complex reached a plateau 1 hr after treatment (Figure

5.3B). In contrast, in order to induce significant matriptase activation by serum, the cells must be pretreated with Ca2+ for at least 8 hrs (Figure 5.3C). These data suggest that primary human keratinocytes seems to be able to rapidly activate matriptase in

153

Figure 5.2 The calcium and serum are indispensable to induce evident activation of matriptase.

Human primary keratinocytes were incubated overnight with variant combinations of calcium and FBS as indicated in the figure. The next day, cells were harvested and lysed followed by immunoblot analyses using a total matriptase mAb M24.

154

Figure 5.3 Sequential administration of calcium and serum affects the outcome of matriptase activation.

A. Human primary keratinocytes were incubated with 0.09 mM calcium (C) or 0.09 mM calcium plus 2% human serum (HS), or 0.2 mM calcium (Ca), or 0.2 mM calcium plus

2% human serum (Ca+HS) over two nights, or incubated with 2% human serum or 0.2 mM calcium overnight and replaced with 0.2 mM calcium (HS->Ca) or 2% human serum

(Ca->HS), respectively, and incubated another overnight. The cells were harvested and lysed followed by immunoblot analyses using a total matriptase mAb M24. B. Human primary keratinocytes were incubated with 0.09 mM calcium (C), or 2% human serum

(HS) overnight, or with 0.2 mM calcium overnight followed by 2% human serum for variant time indicated in the figure. Cell lysates were analyzed by immunoblot using a total matriptase mAb M24. C. Human primary keratinocytes were incubated with 0.09 mM calcium (C), or 2% human serum (HS), or 0.2 mM calcium (Ca) overnight, or with

0.2 mM calcium for variant time periods as indicated in the figure followed by 2% human 155 serum for another 2 h. Cell lysates were analyzed by immunoblot using a total matriptase mAb M24.

156 response to serum only after the cells are primed with Ca2+. Given that Ca2+ induces keratinocyte differentiation, the requisite role of Ca2+ priming in serum-induced matriptase activation suggests that differentiating, but not proliferating, keratinocytes can turn on the matriptase-mediated cell surface proteolytic cascade in response to serum.

The longer time requirement for Ca2+ priming suggests de novo protein synthesis might be involved in Ca2+ priming; in contrast, FBS-induced matriptase activation might be via a rapid signal transduction. The former was confirmed by the fact that both RNA synthesis inhibitor actinomycin D and protein synthesis inhibitor cycloheximidine are able to inhibit matriptase activation induced by Ca2+ and HS in a dose-dependent manner

(Figure 5.4).

Extracellular Ca2+ induces many cellular and molecular events, including elevating intracellular Ca2+and the activation of protein kinase C (PKC) and phosphatidylinositol

3-kinases (PI3Ks). These extracellular Ca2+-induced signaling pathways are apparently involved in the serum-induced matriptase activation, as pretreatment of the cells with intracellular calcium chelator BAPTA-AM (Figure 5.5A), PKC inhibitor Ro-31-8220

(Figure 5.5B), and PI3K inhibitor LY294002 (Figure 5.5C) inhibit the subsequent serum-induced matriptase activation. However, the calcium channel blockers, verapamil and nifedipine, cannot block the activation of matriptase under the stimulation

157

Figure 5.4 De novo protein synthesis is required for serum/calcium-induced activation of matriptase.

A. Primary human keratinocytes were incubated with 0.09 mM calcium alone (C), or 2% human serum alone (HS), or 0.2 mM calcium alone (Ca), or 0.2 mM calcium plus 2% human serum (Ca+HS) for overnight. Variant amounts of actinomycin D were pre-treated to human primary keratinocyte for 30 min followed by adding calcium concentration to 0.2 mM calcium (Ca+HS) for overnight. The 0.2 mM calcium-treated cells were incubated with 2% human serum for another 2h (Ca+HS). Cell lysates were subjected to immunoblot analyses using a total matriptase mAb, M24. B. Primary human keratinocytes were incubated with 0.09 mM calcium alone (C), or 2% human serum alone (HS), or 0.2 mM calcium alone (Ca), or 0.2 mM calcium plus 2% human serum (Ca+HS) for overnight. Variant amounts of cycloheximidine were pre-treated to human primary keratinocyte for 30 min followed by adding calcium concentration to 0.2

158 mM calcium (Ca+HS) for overnight. The 0.2 mM calcium-treated cells were incubated with 2% human serum for another 2h (Ca+HS). Cell lysates were subjected to immunoblot analyses using a total matriptase mAb, M24.

159

Figure 5.5 Serum/calcium-induced activation of matriptase requires elevating intracellular calcium and activation of protein kinase C (PKC) and phosphatidylinositol 3-kinases (PI3Ks).

A. Primary human keratinocytes were incubated with 0.2 mM calcium alone (Ca) or pre-incubated with variant amounts of BAPTA-AM, an intracellular chelator, for 30 min followed by 0.2 mM calcium overnight and 2% human serum for another 2 h. B.

Primary human keratinocytes were incubated with 0.09 mM calcium alone (C), or 2% human serum alone (HS), or 0.2 mM calcium alone (Ca), or pre-treated with variant amount of Ro-31-8220, a PKC inhibitor, for 30 min followed by 0.2 mM calcium overnight and 2% human serum for another 2 h. C. Primary human keratinocytes were incubated overnight with 0.09 mM calcium alone (C), or 2% human serum alone (HS), or

0.2 mM calcium alone (Ca), or 0.2 mM calcium plus 2% human serum (Ca+HS), or pre-treated with variant amounts of LY294002, a PI2K inhibitor, for 30 min followed by

0.2 mM calcium overnight and 2% human serum for another 2 h. All cell lysates were 160 subjected to immunoblot analyses using a total matriptase mAb, M24.

161 of Ca2+ and plasminogen, and calcium ionophores, A23187 and X537A which increase intracellular Ca2+ levels, cannot replace the function of extracellular Ca2+ (Figure 5.6).

5.4.2 Plasminogen, the serum factor that induces matriptase activation in

primary human keratinocytes

The serum factor(s) that induces matriptase activation was purified and identified from fetal bovine serum by a purification scheme, including Cibacron Blue 3GA, DEAE chromatography and HiPrep Q XL using FPLC (Figure 5.7A). In the Q XL elutions, the matriptase activation activity was identified in the column fractions ranging from # 4 through #36 and separated into two groups (Figure 5.7B). The protein profile of these positive fractions revealed that the activity may be derived from a 75-kDa protein that also was detected in multiple fractions with matriptase activation activity. More importantly, the 75-kDa protein was also eluted into two groups with their distribution matching the matriptase activation activity (Figure 5.7C). Since the purity of the

75-kDa protein in the fraction #8 is near homogeneity without clear contaminated proteins as determined by SDS-PAGE using highly concentrated sample, we conclude that the 75-kDa protein is likely the serum factor that induces matriptase activation. The

75-kDa protein is likely a secreted single-chain protein with multiple pairs of disulfide linkages as heating treatment did not cause a change in rate of migration and reducing

162

Figure 5.6 Matriptase activation cannot be stopped by calcium channel blockers or initiated by calcium ionophores.

A and B. Primary keratinocytes were pre-incubated with A23187 or X537A at different concentrations for overnight followed by 2% human serum (HS) for another 2 h. C and

D. Primary keratinocytes were pre-incubated with verapamil or nifedipine for 1 h followed by 0.2 mM calcium overnight and 2% human serum for another 2 h. All cell 163 lysates were subjected to immunoblot analyses using a total matriptase mAb, M24.

164

Figure 5.7 Purification and identification of the serum factor that induces matriptase activation in primary human keratinocyte.

A. Flow chart of purification of the serum factor that induces matriptase activation in primary human keratinocytes is presented. B. Primary human keratinocytes were incubated 0.09 mM calcium (C), or a combination of 0.2 mM calcium and 2% human serum (Ca+HS), or a combination of 0.2 mM calcium and 5% of the column fractions from elutions of the Q XL column. Matriptase activity was examined by immunoblot using a matriptase mAb, M24. C. Protein profiles of the Q XL elutions. Protein bands were stained with Protoblue. D. Protein profiles of the fraction #8 of Q XL column.

The sample was concentrated up to 100 folds using Amicon ultra centrifugal filter unit

165 and protein bands were stained with Protoblue.

166 treatment significantly reduced its migration rate in SDS-PAGE (Figure 5.7D).

The 75-kDa protein was subjected to protein identification via in-gel trypsin digestion, tandem mass spectrometry (MS/MS) and data base searches. Nine obtained tryptic fragments matched to bovine plasminogen, suggesting that the 75-kDa protein is plasminogen (Figure 5.8A). The purification and identification of plasminogen via conventional liquid chromatography was further confirmed by commercial available human glu-plasminogen (Figure 5.8B, PLG+Ca). Plasminogen contains an lysine binding site 293, 294 that is required for its binding to the cell surfaces and the binding can be displaced by 6-aminohexanoic acid (6-AHA) 295. For proper and rapid plasminogen activation, binding to the cell surface is required. Hence, incubating keratinocytes with

6-AHA might prevent the plasminogen from binding to the cell surface and subsequently activating matriptase. Indeed, 6-AHA can suppress the plasminogen-induced matriptase activation (Figure 5.8B, 6-AHA+PLG+Ca). These data suggest that binding of plasminogen to the cell surface is likely required for the induction of matriptase activation.

While matriptase zymogen activation involves a cleavage at the canonical activation motif, plasmin, the active form of plasminogen, plays no role in matriptase activation

(Figure 5.9A). Furthermore, matriptase activation can be induced by plasminogen at a

167

Figure 5.8 Plasminogen is the serum factor that induces matriptase activation in primary human keratinocytes and its effect can be inhibited by 6-AHA.

A. Identified peptide sequences of the serum factor – plasminogen. B. Matriptase activation can be blocked by 6-AHA. Primary human keratinocytes were incubated with 0.09 mM calcium (C), or 0.2 mM calcium alone (Ca), or 0.1 M 6-AHA alone

(6-AHA), or a combination of 0.2 mM calcium and 0.1 M 6-AHA (6-AHA+Ca), or a combination of 0.1 M 6-AHA and 1 g/ml plasminogen (6-AHA+PLG), or a combination of 0.2 mM calcium and 1 g/ml plasminogen (PLG+Ca), or a combination of 0.2 mM calcium, 0.1 M 6-AHA, and 1 g/ml plasminogen (6-AHA+PLG+Ca).

Matriptase activity was examined by immunoblot using a matriptase mAb, M24.

168

Figure 5.9 Plasminogen rather than plasmin is the trigger of matriptase activation.

A. Primary human keratinocytes were incubated with 0.09 mM calcium (C), or 0.2 mM calcium alone (Ca), or 1 g/ml plasmin, or 0.2 mM calcium plus 1 g/ml plasmin for variant times. Cell lysates were subjected to immunoblot analyses using a total matriptase mAb, M24. B. Primary human keratinocytes were incubated overnight with

0.09 mM calcium (C), or 0.2 mM calcium (Ca), or variant amounts of plasminogen as indicated in the figure with 0.09 mM calcium or 0.2 mM calcium. Cell lysates were analyzed by immunoblot using a total matriptase mAb, M24.

169 concentration as low as 0.05 μg/ml (or 0.568 pM) (Figure 5.9B). By considering that plasminogen-induced matriptase activation is a fast event taking less than 30 min (Figure

5.3), requires sub-nM concentration, and does not involves plasmin, plasminogen might function as an extracellular ligand that can engage cell surface receptor(s) to transduce signals, leading to matriptase activation and other cellular and molecular events.

5.4.3 Matriptase activation accelerates production of plasmin

The identification of plasminogen as an extracellular stimulus of matriptase zymogen activation suggests that there might be a proteolytic circuit, starting with plasminogen as extracellular controller to initiate cascading zymogen activation from matriptase to plasminogen activator (PA) to plasminogen and ending with generation of plasmin. Plasminogen activation features a cell associated event and a lag phase prior to a burst of plasmin generation. The former is through the binding of urokinase plasminogen activator to its cellular receptor uPAR. The latter results from a unique process of plasminogen activation by uPA, in which the low intrinsic catalytic activity of pro-uPA generates undetectable plasmin. The initial generated plasmin in turn acts on pro-uPA in a process named reciprocal zymogen activation. After a long lag phase, the rate of plasmin generation becomes exponential. When plasminogen was added to the primary human keratinocytes maintained in culture medium in the absence of Ca2+, the

170 generation of plasmin became very rapid after 40 min incubation. For the keratinocytes maintained in high Ca2+ medium, plasmin generation was even more pronounced with an

86% increase in the rate of plasmin generation at 48 min, suggesting that matriptase activation among other things might accelerate plasmin generation (Figure 5.10A). The role of matriptase in the generation of plasmin was confirmed by the reduced plasmin generation in a dose-dependent manner using matriptase mAb 21-9 (Figure 5.10B), which has been shown to suppress matriptase activation. Likewise, the two small molecule inhibitors, F3226-1197 and F3226-1198, that can suppress matriptase zymogen activation also suppress plasmin generation (Figure 5.10C).

Previously we have discovered that robust matriptase zymogen activation can be induced by transiently exposing matriptase-expressing cells to a mildly acidic buffer.

Within 20 min, the vast majority of matriptase zymogen can be converted into active matriptase at low pH. In human keratinocytes, the generated active matriptase either stays in the cells or is rapidly shed. The cellular active matriptase can activate prostasin and then is rapidly inhibited by HAI-1 to form a 120-kDa matriptase-HAI-1 complex.

This can be demonstrated by the appearance of the 120-kDa matriptase complex at the cost of the 70-kDa matriptase zymogen (Figure 5.11A, comparing lane 2 with lane 1).

The shed active matriptase can be detected by its strong amidolytic activity using a

171

172 Figure 5.10 Plasmin generation is accelerated by matriptase activation.

A. Primary human keratinocytes were incubated with 0.09 mM calcium plus 50 nM plasminogen (C+PLG), or 0.2 mM calcium plus 50 nM plasminogen (Ca+PLG).

Generation of plasminogen was assayed using the plasmin synthetic fluorescent substrate,

Boc-Val-Leu-Lys-AMC. B. Primary human keratinocytes were pre-incubated with rat

IgG, as control, or variant amounts of an anti-rat total matriptase mAb 21-9 for 2 h followed by raising the calcium concentration to 0.2 mM and addition of 50 nM plasminogen. Plasmin activity was assayed using the plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC. C. Primary human keratinocytes were pre-incubated with DMSO, as control, or the two small molecule inhibitor of matriptase, F3226-1197 and F3226-1198 for 2 h followed by raising the calcium concentration to 0.2 mM and addition of 50 nM plasminogen. Plasmin activity was monitored via the plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC. All the data are representative of three independent experiments performed in triplicate.

173

174 Figure 5.11 Acceleration of plasmin generation is done by pericellular active matriptase on cell surface.

A. Primary human keratinocytes were incubated with DPBS (lane 1) or a pH 6.0 buffer

(lane 2) for 30 min at room temperature. Cell lysates were analyzed by immunoblot using a total matriptase mAb, M24. B. Primary human keratinocytes were incubated with DPBS (No Activation) or a pH 6.0 buffer (Activation) for 30 min at room temperature. The conditioned buffers were used to analyze matriptase activity using a matriptase synthetic fluorescent substrate, Boc-Gln-Ala-Arg-AMC. C. Primary human keratinocytes were incubated with a pH 6.0 buffer for 30 min at room temperature. 50 nM plasminogen was added to the cells and conditioned buffer together (Combined), the cells alone (Cellular), and the conditioned buffer alone (Shed) to analyze the plasmin activity using a plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC. D.

Primary human keratinocytes were incubated with 0.2 mM calcium, 50 nM plasminogen, and variant amount of 6-AHA. The generation of plasmin was monitored with a plasmin synthetic fluorescent substrate, Boc-Val-Leu-Lys-AMC. All data are representative of three independent experiments performed in triplicate.

175 standard fluorescent substrate for matriptase (Figure 5.11B). By taking advantage of this unique feature of matriptase activation, we further confirmed the role of matriptase in plasmin generation and the matriptase-accelerated plasminogen activation is a pericellular event. In the presence of the shed active matriptase, the plasmin generation becomes very robust with a shortened lag phase and an increased rate of plasmin generation (Figure 5.11C, Combined). When the shed active matriptase was separated from the cells and incubated with plasminogen, plasmin generation was very slow, confirming that active matriptase is not a good plasminogen activator (Figure 5.11C,

Shed). When plasminogen was added to the post-matriptase activation keratinocytes which contained only very low levels of matriptase zymogen, plasmin generation was also very slow (Figure 5.11C, Cellular). These data suggest that while active matriptase is shed into extracellular milieu, the active matriptase must come to the cell surface for accelerating plasmin generation. The cell-associated process of plasmin generation can be further confirmed by the inhibition of plasmin generation using 6-AHA, which can prevent plasminogen from binding to the cell surface (Figure 5.11D).

5.5 Discussion

Keratinocyte cell morphology and epidermal tissue structure are progressively

176 remodeled in the process of differentiation. This process involves significant pericellular proteolysis that must be precisely controlled. In addition to the expression levels, protease activity is primarily regulated by protease inhibitors and by the process of zymogen activation. The former determines the longevity of active protease and the latter provides the temporal and spatial control. In our previous studies (Chapters 2 and

3), activation of prostasin by active matriptase takes place simultaneously with matriptase zymogen activation and inhibition of active matriptase by HAI-1. Antithrombin, another matriptase endogenous inhibitor, determines the levels of active matriptase shed into extracellular milieu in which active matriptase has a longer life span and can return to the cell surface to activate and process a variety of substrates. With the two protease inhibitors, the action on different substrates by matriptase can be regulated. In the current study, we focus on the control of matriptase zymogen activation. Our study reveals that a combination of Ca2+, the prominent keratinocyte differentiation agent, and plasminogen- the zymogen of plasmin the major protease workhorse at the pericellular environment- confers on keratinocytes critical measures to precisely regulate matriptase activation in the context of epidermal differentiation.

The functional linkage between matriptase and plasminogen is initially thought to be one-directional: from matriptase through uPA to plasminogen. In response to a variety

177 of extracellular stimuli and cellular conditions, matriptase undergoes autoactivation to convert matriptase zymogen to active matriptase. In the current study, plasminogen was identified to be the serum factor that stimulates differentiating keratinocytes to activate matriptase. Our discovery reveals a reciprocal, rather than linear, relation exists between matriptase and plasminogen. This unusual regulatory mechanism makes the matriptase-uPA-plasminogen protease cascade under a control of positive feedback loop, in which the zymogen of downstream effector plasminogen functions as an extracellular message that stimulates keratinocytes to activate the initiator protease matriptase (Figure

5.12). This regulatory mechanism confers on cells a timely and effective control of plasmin generation.

Ca2+ is a potent pro-differentiation agent, particularly for the cultured keratinocytes.

In vivo, a Ca2+ gradient exists in the epidermis and coincides with the differentiation state

296. In the basal and spinous layers, Ca2+ is at low levels and is primarily intracellular.

Large amounts of Ca2+ accumulate in the upper granular layer both inside the cells and in the intercellular matrix. Given that keratinocytes must be primed with Ca2+ in order to become responsive to plasminogen for matriptase activation, this Ca2+-dependency might provide a critical mechanism for keratinocytes to further precisely control the pericellular protease circuit in the context of epidermal differentiation. This dual mechanism for the

178

Figure 5.12 A schematic model for the reciprocal activation of matriptase-plasminogen activator-plasminogen cascade.

The downstream effector plasminogen and calcium-induced cellular effects work in a concert becoming an extracellular stimulus of matriptase activation in primary human keratinocyte. The pericellular active matriptase then comes back to cell surface to activate plasminogen activator and eventually leading to massive generation of plasmin.

179 control of matriptase activation might provide mechanistic insights into the activation status in the normal and diseased human skin revealed in Chapter 4 and in our previous studies 196. While highly expressed by keratinocytes in the basal and spinous layer in normal human skin, matriptase is maintained predominantly in its zymogen form. The lack of matriptase activation might primarily result from the low Ca2+ in basal keratinocytes but not the lack of plasminogen, which was detected in these cells 297. For spinous keratinocytes this could result from both low Ca2+ and lack of plasminogen.

Expression of matriptase is significantly reduced in the granular layer despite its high

Ca2+ levels. Therefore, with the different distribution profile of matriptase, Ca2+, and plasminogen in normal human skin, the activity of the protease circuit can be maintained in very low levels. The inactivity of the protease circuit might rapidly change when the delicate boundary of Ca2+ and plasminogen in normal skin is altered. For example, in skin diseases with chronic inflammation, the skin permeability barrier is altered and results in an increase in Ca2+ efflux. Along with the Ca2+ efflux, matriptase activation might be induced, particularly for the basal keratinocytes which appear have easy access to plasminogen. Indeed, we have observed increased matriptase activation in those basal keratinocytes close to the inflammatory area in several skin diseases 196.

Interestingly, the activation appears to be restricted to the basal layer but not the spinous

180 layer. The lack of plasminogen might be a plausible answer for the lack of matriptase activation in the spinous keratinocytes. Robust matriptase activation and plasmin generation are very likely to occur with the conditions involving the rupture of blood vessels, such as in incision wound, in which keratinocytes from both basal and spinous layers have massive exposure to Ca2+ and plasminogen. A model for the regulation of this protease circuit in normal, inflammation, and wounded skins is proposed in Figure

5.13.

In summary, we discovered that a combination of Ca2+ and plasminogen can induce keratinocytes to activate membrane-bound serine protease matriptase. The role of plasminogen in matriptase zymogen activation reveals that the matriptase-uPA-plasmin proteolytic cascade is under a regulation with positive feedback loop in which the zymogen of the effector protease plasmin stimulates cells to activate the initiator protease matriptase. This unusual mechanism renders the protease cascade as a protease circuit featuring a timely and effective way to generate plasmin. The role of Ca2+ in matriptase activation appears to associate with its pro-differentiating activity and bringing the protease circuit under the control of epidermal differentiation. Since the Ca2+ gradient and the in vivo distribution of matriptase and plasminogen in normal skin do not favor matriptase activation, activity of the protease circuit is very limited in normal skin.

181

Figure 5.13 Schematic models of the status of matriptase activation in skin under different conditions.

A. Under normal condition, the plasminogen only presents in the basal layer of the epidermis. Although matriptase expression can be detected in the basal and spinous layer, the tightly controlled calcium concentration is not enough to work with plasminogen to induce matriptase activation. B. Under inflammatory condition, the tightly regulated calcium level is destroyed due to the altered permeability barriers of the epidermis leading to an increase in calcium in the basal layer and eventually reaching calcium levels that are sufficient to work with plasminogen to induce matriptase activation. C. When skin is wounded, abundant calcium and plasminogen go into the wound site along with blood; hence, a robust matriptase and plasmin generation occurs. 182 Matriptase might, therefore, play a less critical role in maintenance of epidermal homeostasis. The protease circuit could, however, be rapidly turned on if the unfavorable conditions for activation are altered in the disease process and acute insults with influx of Ca2+ and/or plasminogen. In conjunction with its inhibitory mechanism, matriptase activity can be precisely regulated by keratinocytes in a way to accommodate the carefully orchestrated process of epidermal differentiation that leads to formation of the critical protective barrier provided by the skin.

183

Chapter 6 – Conclusion and discussion

184 Epidermal differentiation is a carefully orchestrated process that leads to the formation of the critical protective barrier provided by the skin. The process of generating a functional epidermal layer requires progressive remodeling of cell morphology and tissue structure, and involves significant pericellular proteolysis that must be activated and inactivated in a precisely controlled manner. In particular, the matriptase-driven protease network plays a critical role in epidermal barrier construction as well as in the regenerative processes required for wound healing. In addition to the initiator protease, the network includes extracellular stimuli for matriptase activation, protease inhibitors, and downstream substrates. The importance of this protease network in skin development and homeostasis is evident, as the targeted deletion of its members in mice resulted in epidermal barrier loss and abnormal skin115, 116, 124.

Furthermore, matriptase mutations in humans are characterized by autosomal recessive ichthyosis and hypotrichosis (ARIH), and ichthyosis, follicular atrophoderma, hypotrichosis and hypohidrosis (IFAH)120-122. Though the importance of the matriptase-driven protease cascade in skin is clear, how the protease cascade is initiated and regulated in keratinocytes is still poorly understood.

The two major ways to regulate protease activity are activation of zymogen and inhibition of active proteases by protease inhibitors. In addition, the expression levels

185 of proteases and protease inhibitors can also alter the balance of the protease cascade.

In this dissertation, I have identified and characterized novel molecular mechanisms governing the activation and inhibition of matriptase, and cellular mechanisms by which matriptase activates its molecular targets and contributes to keratinocyte differentiation and formation of epidermal barrier. First, I identified plasminogen as a keratinocyte-selective extracellular stimulus for matriptase activation. The discovery of plasminogen as an initiating signal of the matriptase-driven protease cascade reaffirms the theory that the matriptase-uPA-plasmin cascade is not unidirectional in the activation of its components, but it is reciprocal. In addition to HAI-1, I also revealed keratinocytes employ antithrombin as a significant endogenous protease inhibitor. The enhanced role of antithrombin in matriptase inhibition in keratinocytes reveals the regulatory adaptation in stratified epithelial cells due to the changes in tissue structure, compared to the polarized epithelial cells. With the dual inhibitory mechanisms, I further revealed that matriptase acts on its molecular targets in two different ways: 1) a rapid activation of prostasin by cell-associated active matriptase under extremely tight control of HAI-1, and 2) the action on several other substrates, including uPA, HGF, and syndecan 1, by secreted active matriptase that is controlled by antithrombin. Last but not least, I also demonstrated that the physiological role of matriptase in human skin

186 likely lies in the basal and spinous keratinocytes that are involved in proliferation and early differentiation of keratinocytes. The role of matriptase in the early stages of keratinocyte life history was further supported by the increased matriptase zymogen activation in the keratinocytes of the bulge area. By charactering the prominent regulators, downstream effectors, and the expression and activation states of matriptase in human skin, a clearer picture is emerging regarding the role of matriptase in skin biology.

Rather than sole participation in the maintenance of homeostasis of normal human skin, matriptase might play a more important role in repairing damaged human skin in response to acute insults and pathological states. For the former, as in the process of wound healing, matriptase activation is rapidly induced by the influx of plasminogen and

Ca2+ in blood. For the latter, increased inflammation associated with some skin diseases stimulates basal keratinocytes to activate matriptase, which might participate in recovery from the diseases but could also contribute to the syndromes associated with these skin diseases. We are now likely closer to uncovering the contributors to some skin diseases. This might be particularly helpful in the development of drugs that target the matriptase-driven protease cascade.

6.1 What does matriptase-driven protease cascade do in human skin

187 and what happens when it goes wrong?

The importance of matriptase-driven protease cascade in skin development and homeostasis was first demonstrated by matriptase targeted deletion in mice115. The matriptase-deficient newborn mice died uniformly within 48 h because of compromised epidermal barrier function, which lead to fatal and rapid dehydration. Loss of matriptase also resulted in follicular hypoplasia and wholesale abortion of vibrissal follicles suggesting a role of matriptase in hair follicle development. Targeted deletion of prostasin- another member of the matriptase-derived protease cascade- further supported the important role of the cascade in epidermal barrier function 116. The prostasin-deficient mice also died short after they were born, within 60 h. The cause of death was the rapid, fatal dehydration because of impaired skin barrier function, similar to the matriptase-deficient ones. The fact that both matriptase and prostasin are required for the epidermal barrier function, processing of profilaggrin, and formation of stratum corneum, builds up the functional linkage between matriptase and prostasin as a protease cascade 193, 210. In addition to decreasing protease activity, imbalance of the protease cascade can also disrupt the proper development of epidermal barrier function. Mice lacking HAI-1, the cognate inhibitor for matriptase and prostasin, present growth retardation and die in 16 days 124. They developed ichthyosis-like scaly skin due to

188 hyperkeratinization, abnormal hair shafts without regular cuticularseptation, interfollicular epidermis with acanthosis, and altered processes of profilaggrin. Severe epidermal defects and chronic skin inflammation are also seen in HAI-1-deficient zebrafish 145, 146.

The importance of the matriptase-driven protease cascade in skin was further highlighted by matriptase mutations in human patients. Matriptase mutations were recently identified as responsible for the two rare forms of congenital ichthyosis, the autosomal recessive ichthyosis with hypotrichosis (ARIH, OMIM 610765), and the ichthyosis, follicular atrophoderma, hypotrichosis and hypohidrosis (IFAH, OMIM

602400) 120-122. These patients present congenital ichthyosis, abnormal hair, hyperkeratosis, corneal involvement, and skin barrier defects. The processing of profilaggrin is disturbed in these patients 122 and, hence, is the most likely mechanism that causes the impaired skin barrier. Matriptase might also play a role in epidermal desquamation as the patients with ARIH lost the ability to degrade the stratum corneum corneodesmosomes120. Degradation of both profilaggrin and corneodesmosomes in the late stages of epidermal differentiation require intensive proteolysis; however, the proteolytic activity of matriptase is likely to be indirectly involved in these events in human skin.

189 The fact that humans and mice developed similar phenotypes via disruption of a matriptase-driven protease cascade suggests that this cascade performs similar functions in the whole process of skin development and homeostasis. However, matriptase tissue distributions are distinctly different in the skin of these two species, suggesting that the stages and context that this protease cascade involves might be different. Matriptase is expressed only in the uppermost viable epidermal layer in mouse epidermis 210, suggesting that mouse matriptase actively participates only during the late stages of epidermal differentiation. In contrast, in human epidermis, matriptase is expressed at high levels at cell-cell junctions in the basal layer and spinous layer, but is present only at reduced levels in the granular layer (Chapter 4), which is consistent with our previous in situ hybridization results showing a similar mRNA distribution 157. The diminishing matriptase expression in the granular layer of the human skin indicates that rather than the late stages of epidermal differentiation, human matriptase may participate in multiple functions associated with the basal and spinous keratinocytes, such as proliferation, migration, and early differentiation in human epidermis. This hypothesis was supported by a several finding in this dissertation: 1) the reduced growth and expression of differentiation markers in HaCaT human keratinocytes with a reduction of matriptase expression, 2) the highest zymogen activation in the bulge area of hair, likely the skin

190 stem cells, suggests matriptase actively participates in “active” keratinocytes but not in relative quiescent cells in the skin. It is worth mentioning that thickened stratum spinosum (acanthosis) and stratum corneum seen in the skin of the ARIH patients could result from an impact on keratinocyte proliferation and differentiation by matriptase mutations, propagating through multiple processes rather than just at one specific proteolytic event. Dysregulation of the proliferation and early differentiation will also affect terminal differentiation, which may explain the similar phenotypes seen in matriptase-deficient human and mice.

6.2 How does the matriptase-derived protease cascade begin in keratinocytes?

The severe phenotypes present in matriptase mutant patients highlight how important it is to discover how matriptase influences skin diseases. Since the function of matriptase-driven protease cascade is different in human and mice, investigating the regulation of the protease cascade using human models would be more relevant to human skin diseases.

The matriptase-driven protease cascade starts with matriptase activation when cells receive signals from the extracellular environment. As an initiator protease,

191 matriptase is able to receive signals from the extracellular space and to respond by autoactivation. This feature makes matriptase a unique protease; other proteases require help from already-active proteases to become active themselves. Given this direct signaling, matriptase activation could serve as the checkpoint of initiation of the protease cascade. Matriptase is a very delicate protease. It can be activated by the microenvironment surrounding it. Small changes in pH or redox state are sufficient to induce matriptase activation in simple epithelial cells, leukocytes177, 193, 261, 273, 292, or as we have now shown, keratinocytes 196. The ability of the mildly acidic environment to trigger matriptase activation is attributed to the increased intrinsic activity of matriptase zymogen176. Besides the microenvironment which seems to activate matriptase universally, matriptase can also be activated by several cell type-specific extracellular stimuli. For example, the lysophospholipid sphingosine 1-phosphate (S1P) induces matriptase activation in mammary epithelial cells 182, but not in other epithelial cells including keratinocytes (Chapter 5). Androgens stimulate matriptase activation in prostate cancer cells 184. These activation signals are believed to be transduced to a common activation complex or machinery, which is anchored on a lipid bilayer biomembrane192.

Our quest for keratinocyte-specific matriptase activation began with the

192 observation that matriptase is activated in raft but not in monolayer cultures (Chapter 2).

This observation provided us with a clue that the protease might play an important role during epidermal differentiation and there might be an extracellular stimulus for keratinocytes. The identification of plasminogen as a novel, keratinocyte-selective extracellular stimulus and the requirement of Ca2+-priming prior to plasminogen treatment to induce matriptase activation suggests that a dual regulatory mechanism is utilized by keratinocytes to precisely control protease activation. The requirement of

Ca2+-priming before plasminogen treatment to induce matriptase activation further suggests that the differentiation agent, Ca2+, might significantly increase the levels of an activation program, through which plasminogen transduces the activation signal. The inducible activation program might be lost in immortalized keratinocytes and other cell lines and the cells might eventually become insensitive to plasminogen as they are regularly maintained under high Ca2+ and fetal bovine serum, which contains plasminogen. Binding of plasminogen to the surface of keratinocytes is required for matriptase activation since the effect can be inhibited by 6-AHA. The requirement of

Ca2+ priming and surface binding of plasminogen to induce matriptase activation suggest that the inducible activation program might be a cell surface plasminogen receptor(s); the engagement by plasminogen transduces exogenous signals to the activation machinery,

193 leading to matriptase activation.

These discoveries combined with the dynamic distribution of the three key players (matriptase, plasminogen, and Ca2+) that initiate the matriptase-driven protease cascade explain the way keratinocytes could precisely regulate matriptase activation.

First, matriptase is expressed mainly in the basal and spinous layers (Chapter 4).

Second, plasminogen is synthesized in liver and present both in plasma and interstitial fluids, and can be detected only in the basal layer of human skin297. Last, a Ca2+ gradient exists in the epidermis with the lowest level being detected in the basal layer and gradually increases toward the granular layer. The lack of co-localization of plasminogen and Ca2+ in normal skin ensures only limited matriptase activation in keratinocytes in normal human skin. The delicate balance of matriptase, plasminogen, and Ca2+ gradient renders a way to precisely control the matriptase-driven protease cascade in skin to avoid aberrant activation. As soon as the spatial distribution of these factors is changed, the cells will rapidly initiate the matriptase-driven protease cascade.

For example, when the calcium gradient was disturbed during inflammation, matriptase activation was seen in the basal keratinocytes, which contain plasminogen in a low Ca2+ environment under normal conditions. When keratinocytes are insulted, an influx both of plasminogen and Ca2+ is introduced to the wound bed and matriptase activation is

194 rapidly induced.

6.3 What does matriptase do in keratinocytes?

There are two possible fates of active matriptase: 1) active matriptase stays on the cell surface and acts on its substrates, such as prostasin. 2) active matriptase is shed into the extracellular milieu, and can either return back to the cell surface or stay in the fluids to act on its substrates, such as uPA, HGF, and syndecan-1. Several substrates have been identified for matriptase based on the cleavage preferences of the enzyme, and include prostasin, uPA, stromelysin, PAR-2, MSP-1, ENaC, ASIC1 and HGF. Among these substrates, the two serine proteases, uPA and prostasin, and the HSPG, syndecan-1, are highly relevant to skin function. Prostasin is activated by the cell-associated active matriptase on the cell surface. Pro-uPA is activated on cell surface by the secreted active matriptase which returns back to the cell surface. The secreted active matriptase also activates HGF with similar potency in solution and in the pericellular environment.

Among those matriptase substrates mentioned above, prostasin activation is most tightly coupled with matriptase own activation (Chapter 2) and is activated by the cell-associated active matriptase on the cell surface. Prostasin was first identified as a matriptase substrate in the skin of matriptase knockout mice, which lacks prostasin

195 zymogen activation 210. Moreover, matriptase and prostasin knockout mice display nearly identical epidermal defects 115, 116 suggesting a functional linkage between these two proteases. Despite the stark difference in matriptase distribution in human versus rodent skin (Chapter 4), human keratinocytes still retain prostasin as a matriptase substrate (Chapter 2), as indicated by a loss of prostasin activation in mice with knockout matriptase expression.” The matriptase-prostasin cascade can regulate a variety of targets including ENaC and the G protein-coupled receptor PAR-2. PAR-2 is expressed at high levels in human keratinocytes and is a key player in response to inflammation and epidermal barrier repair. Therefore, matriptase could play a role in wound healing through PAR-2 298. Though the functional linkage between matriptase and prostasin is well connected in keratinocytes, the two proteases may not function in tandem in other epithelial cells, or at least not as directly as in keratinocytes. For example, matriptase and prostasin are eventually expressed in separate parts of squamous cell carcinoma 195 and polarized epithelial cells 266. It is conceivable that the different subcellular localizations of matriptase and prostasin render the ability of matriptase to activate prostasin less relevant. Hence, matriptase is less likely the activator of prostasin outside the keratinocytes.

Besides the active matriptase remaining on the surface which can activate

196 prostasin, active matriptase is also shed into the extracellular milieu in which it has a longer life span and can return to the cell surface to activate and process a variety of substrates, such as uPA, HGF, and syndecan-1. The activation of uPA by matriptase significantly reduces the time required for the initiation of the reciprocal zymogen activation between uPA zymogen and plasminogen 162 and Chapter 5. By activating uPA and other further downstream proteases, such as plasminogen and matrix metalloproteases

(MMPs), matriptase may regulate epidermal wound healing by controlling ECM remodeling and activation of the HGF cell motility factor and other cytokines.

Surprisingly, the conventional one-way direction of the matriptase-uPA-plasmin protease cascade was subverted in keratinocytes as the downstream effector, plasminogen, of the protease cascade, also serves as the initiating signal to form a feed-forward protease circuit. This allows keratinocytes to precisely activate matriptase when needed.

Syndecan-1 is a newly identified substrate of matriptase which also plays a role in wound healing. The impact of syndecan-1 toward wound healing mainly comes from the ectodomain shedding of the HSPG. The release of syndecan-1 extracellular domains may not only convert the membrane-bound receptors into soluble effectors, but also down-regulate signals transduced by syndecan-1. The ectodomain shedding of syndecan-1 is accelerated in response to wound healing and serves to promote cell

197 migration and facilitate wound closure in wound healing (78). Therefore, shedding of the ectodomain of syndecan-1 appears to be an important response in wound healing.

The fact that pericellular active matriptase acts as a novel and potent sheddase for syndecan-1 further links the role of matriptase in epidermal wound healing.

6.4 How is matriptase inhibited in keratinocytes?

Human keratinocytes, like simple epithelium, also adopt HAI-1 and antithrombin as inhibitors to regulate matriptase activity. However, the cells utilize these two inhibitors differently in simple and stratified epithelia. We uncovered several unique features regarding how keratinocytes utilize these two inhibitors to achieve the regulation of matriptase-driven protease network. First, simple epithelium predominantly uses

HAI-1 and to a lesser degree antithrombin as inhibitors of matriptase. Though keratinocytes also employ HAI-1 as a major inhibitor, they rely more on antithrombin.

Second, in addition to matriptase activity, HAI-1 also plays an important role in regulating prostasin activity in keratinocytes. Third, an unexpected role of antithrombin is disclosed in keratinocytes, which regulates the level of pericellular active matriptase that can activate extracellular substrates. Therefore, human keratinocytes appear to utilize these two inhibitors to differentially regulate the function of matriptase: 1) cellular

198 activation of prostasin, a process that is tightly controlled by HAI-1, and 2) activation of extracellular substrates, such as uPA and PAR-2, a process that is controlled by antithrombin.

By studying matriptase in simple epithelial cells, we have learned that the matriptase zymogen undergoes autoactivation for conversion into active matriptase, a process that unexpectedly involves its endogenous protease inhibitor, HAI-1. As a result, free-active matriptase is a very short-lived species, due to the rapid HAI-1 mediated inhibition of the active enzyme through the formation of matriptase-HAI-1 complexes. Human keratinocytes, as expected, also retain this feature of matriptase zymogen activation and rapid HAI-1-mediated inhibition. The major difference between this mechanism in simple and stratified epithelium is that stratified epithelium uses not only HAI-1 to inhibit active matriptase, but also active prostasin, its downstream substrate. This reveals a remarkable feature of the control of the protease network,: matriptase activation, prostasin activation, and inhibition of both proteases by HAI-1 occurs simultaneously. The unusually tight linkage of the three key players of the protease network renders a way to precisely control the effects induced by the protease network.

An unexpected role of antithrombin to regulate the level of pericellular active

199 matriptase in keratinocytes was revealed in this dissertation, which fulfills the long search for cellular evidence of the role of matriptase in activation of the extracellular substrates.

Serine protease inhibitor antithrombin, a major anticoagulant, is efficiently transported into extra-vascular compartments 299-301 where it inhibits matriptase in simple epithelium as shown in our previous studies 163, 273. Though the simple epithelium also utilizes antithrombin as an inhibitor for matriptase, it is less significant compared to HAI-1.

The stratified epithelium, in contrast, apparently relies more heavily on antithrombin to regulate matriptase activation. The enhanced role of antithrombin in keratinocytes might result from the diminished availability of cell surface HAI-1 because prostasin also utilizes HAI-1 as an inhibitor. Moreover, the level of cell surface antithrombin plays a role in the regulation of the amount of pericellular active matriptase. Hence, the binding affinity of antithrombin toward cell surface HSPGs is also a regulator of the matriptase-driven protease network. This unique keratinocyte-selective inhibitory system is able to regulate the level of pericellular active matriptase because 1) the decreased availability of HAI-1 makes keratinocytes rely on antithrombin more, 2) the level of antithrombin of the surface of keratinocyte are controlled by the level of cell surface HSPGs, and 3) the binding affinity of antithrombin on cell surface.

200 6.5 Summary

In summary, keratinocytes express a cell surface protease network comprised of 1) the initiator protease matriptase that is activated in a unique manner in keratinocytes by a combination of Ca2+ and plasminogen; 2) downstream substrates, including proteases, growth factors, membrane receptors, and ion channels; and 3) endogenous protease inhibitors including HAI-1 and antithrombin. The protease network is tightly coupled so that proteolytic activity can be precisely regulated during keratinocyte differentiation and utilized to repair skin function.

Based on our findings, we propose the regulatory scheme outlined in Figure 6.1.

Activation of the matriptase protease network begins with priming of the keratinocytes with a differentiating agent (Ca2+). Significant matriptase activation is then induced by serum-derived plasminogen. Active matriptase triggers activation of prostasin, uPA, and plasmin. These active proteases further activate downstream substrates and effectors that modulate growth factor and receptor function which regulates the indicated biological responses, such as wound healing, barrier function, migration, differentiation.

In addition, activity in this network is suppressed by HAI-1 and antithrombin (AT).

This scheme presents a rational model as to how matriptase modulates keratinocyte barrier formation and response to wounding.

201

Figure 6.1 Schematic models of the regulation and function of matriptase-driven protease network.

See Text for description. CaR stands for calcium receptor; PI3K for phosphatidylinositol 3-kinase; PLC for phospholipase C; PKC for protein kinase C; AP-1 for activator protein 1; PLG for plasminogen; MTP for matriptase; uPA for urokinase-type plasminogen activator; AT for antithrombin; HGF for hepatocyte growth factor; HAI-1 for HGF activator inhibitor 1; MMP for matrix metalloprotease; and PAR-2 for protease activated receptor 2.

202

Chapter 7 -- Appendix Chapter – TMPRSS2, a Serine Protease

Expressed in the Prostate on the Apical Surface of Luminal

Epithelial Cells and Released into Semen in Prostasomes, Is

Misregulated in Prostate Cancer Cells

203 7.1 Abstract

TMPRSS2, a type II transmembrane serine protease, is highly expressed by the epithelium of the human prostate gland. To explore the regulation and function of

TMPRSS2 in the prostate, a panel of monoclonal antibodies with high sensitivity and specificity were generated. Immunodetection showed TMPRSS2 on the apical plasma membrane of the prostate luminal cells, and demonstrated its release into semen as a component of prostasomes, organelle-like vesicles that may facilitate sperm function and enhance male reproduction. In prostate cancer cells, TMPRSS2 expression was increased and the protein mislocalized over the entire tumor cell membrane. In both

LNCaP prostate cancer cells and human semen, TMPRSS2 protein was detected predominantly as inactive zymogen forms as part of an array of multiple noncovalent and disulfide-linked complexes, suggesting that TMPRSS2 activity may be regulated by unconventional mechanisms. Our data suggested that TMPRSS2, an apical surface serine protease, may have a normal role in male reproduction as a component of prostasomes. The aberrant cellular localization, and increased expression of the protease seen in cancer, may contribute to prostate tumorigenesis by providing access of the enzyme to nonphysiological substrates and binding-proteins.

204 7.2 Introduction

TMPRSS2 is an androgen responsive gene that encodes a type II transmembrane serine protease (TTSP) 187, 302, 303. The members of the TTSP family share common protein structures including a transmembrane domain at the N terminus, linker regions with a variety of protein-protein interaction domains, and a canonical serine protease domain at the C terminus 132, 158, 304. TTSPs have been found to play important roles in the development and homeostasis of mammals and the aberrant expression of TTSP genes are reported to contribute to the etiology of several human disorders, including cancer 131. The importance of TMPRSS2 in vivo remains unclear because homozygous

TMPRSS2-null mice are essentially phenotypically normal 305. However, TMPRSS2 was reported to regulate epithelial sodium channel (ENaC) activity in vitro, implying a possible role in epithelial sodium homeostasis 306. TMPRSS2 may play a role in angiogenesis and tubulogenesis in microvesicular endothelial cells, potentially modulating several aspects of prostate tumor biology 307. In addition to its proteolytic activity, TMPRSS2 may also serve as a cell receptor, conducting external signaling or interacting with the extracellular matrix through its extracellular protein binding domains.

Overexpression of TMPRSS2 has been demonstrated in poorly differentiated prostate cancer with significant increase in the mRNA level 187, 188, 308, 309. TMPRSS2 was also

205 reported to be involved in the majority of prostate cancer because of the gene fusion of the 5’-untranslational region of TMPRSS2 with ETS family members, which is implicated in the overexpression of ETS genes in the majority of prostate cancer 187, 310, 311. The coding sequence of TMPRSS2 is not involved in the gene fusion, and as a consequence, there is no resultant recombinant protein for the TMPRSS2-ETS gene fusion and the promoter-less copy of TMPRSS2 is silenced, resulting in reduced expression of

TMPRSS2 mRNA in those prostate cancer patients with the gene fusion 312.

Despite these potentially interesting and important roles for TMPRSS2, characterization of the protein itself remains incomplete and somewhat confusing.

Significant discrepancies in the apparent molecular mass of TMPRSS2 protein, with a calculated size around 54-kDa, have been reported in previous studies, with the most commonly cited being around 70-kDa and 32-kDa 188, 313, 314 as the full-length and serine protease domain of TMPRSS2, respectively. In this study, we have generated a panel of mouse monoclonal antibodies (mAbs) directed against TMPRSS2 that are highly sensitive and specific. Using the newly generated TMPRSS2 mAb, we have investigated the pathophysiological role of TMPRSS2 in prostate tissues by comparing its expression and subcellular distribution in prostate tumors and adjacent normal prostate glands. Functional and regulatory aspects of TMPRSS2 biology were also investigated

206 by examining the activation status of the enzyme and the nature of complexes formed between TMPRSS2 and other proteins.

7.3 Materials and Methods

7.3.1 Chemicals and reagents

Peptide-N-glycosidase F (PNGase F) was purchased from New England Biolabs

(Beverly, MA), and tunicamycin was from Sigma (St. Louis, MO). Commercial

TMPRSS2 antibodies were obtained from Abcam (Cambridge, UK). The matriptase mAb M32 was generated using matriptase-HAI-1 complex as described previously, 155 and the epitope recognized by the mAb was determined to be at the third LDL receptor domain of matriptase 171. Human semen samples were purchased from Lee Biosolutions

(St. Louis, MI).

7.3.2 Cell culture condition

Human prostate cancer cells were obtained from the Tissue Culture Shared Resource of the Lombardi Comprehensive Cancer Center, Georgetown University Medical Center.

LNCaP human prostate cancer cells were maintained in RPMI 1640 supplemented with

10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μg/ml streptomycin, using a

o humidified incubator at 37 C with 5% CO2. PC3 and DU145 human prostate cancer

207 cells were routinely maintained in RPMI 1640 supplemented with 5% FBS, 100 U/ml

o penicillin, and 100 μg/ml streptomycin in a humidified incubator at 37 C with 5% CO2.

7.3.3 Generation of TMPRSS2 monoclonal antibodies

A recombinant protein antigen consisting of a portion of the TMPRSS2 extracellular domains, tagged at the amino terminus with a Flag tag and at the carboxyl terminus with a His tag, was prepared for monoclonal antibody production. DNA encoding the extracellular region of TMPRSS2 was amplified by PCR using the following primers: forward primer: 5’-TAG AAT TCG ATT ACA AGG ATG ACG ACG ATA AGA AGT

TCA TGG GCA GCA AGT GCT CCA AC-3’; reverse primer: 5’-AAC TCG AGC GCC

GTC TGC CCT CAT TTG TCG-3’. A TMPRSS2 cDNA plasmid (a gift from Dr. Peter

Nelson of University of Washington, Seattle, WA) that bears a point mutation R255Q at the canonical activation cleavage site was used as the PCR template. The TMPRSS2 mutant cannot be activated 188. The PCR product was purified, digested with the restriction enzymes EcoRI and XhoI and cloned into a mammalian secretory expression vector, pSecTag2 (Invitrogen, Carlsbad, CA; Fig. 1A). The construct was transfected into HEK293T cells and stable transfectant pools of cells selected with 400μg/ml Zeocin.

Transfectants secreting the TMPRSS2 fusion protein were expanded and conditioned media was then harvested by growing the cells in serum-free DMEM media

208 supplemented with insulin-transferrin-selenium (ITS; Invitrogen, Carlsbad, CA).

Conditioned media were collected every other day, and after clarification by centrifugation proteins were precipitated by the addition of ammonium sulfate to 70% saturation. After desalting, the recombinant TMPRSS2 was purified from the concentrated protein solution by nickel column chromatography (through binding to the

His-tag), followed by Q column chromatography using an AKTA™ prime protein purification system (GE Healthcare Bio-Sciences Corp., Piscataway, NJ). After these procedures, the recombinant protein was purified almost to homogeneity. The purified protein was using to generate monoclonal antibodies using conventional hybridoma technologies 160.

7.3.4 Mass spectrometry analysis and identification of proteins

The protein bands from SDS-gels were excised, washed, destained, and trypsinized overnight at 37°C using standard protocols after dithiothreitol (DTT) reduction and iodoacetamide alkylation. Analysis of tryptic peptides derived from protein samples was performed by liquid chromatography/mass spectrometry (LC/MS) by the Proteomics

Shared Service at Greenebaum Comprehensive Cancer Center, University of Maryland

Baltimore (Baltimore, MD), as describe previously 163.

7.3.5 Immunohistochemistry

209 Immunohistochemistry (IHC) staining was performed using the manufacture’s standard protocol with minor modification (Dako, CA). Paraffin-embedded human prostate tumor tissue sections were obtained from the Histopathology and Tissue Shared

Resource of Georgetown University. The sections were stained using the TMPRSS2 mAb AL20 at a concentration of 2 g/ml for 1 hour. A negative control slide was stained using the mouse IgG antibody at a concentration of 2 g/ml for 1 hour simultaneously to ensure staining quality. HRP-labeled anti-mouse IgG and the substrate, diaminobenzidine (DAB), were used for the detection of positive staining.

Cell nuclei were counterstained with hematoxylin. Images were captured using an

Olympus AH2 Vanox Microscope System (Olympus; Melville, NY).

7.3.6 IHC staining of prostate tissue microarray for TMPRSS2 and analysis by

AQUA

The construction of the prostate tumor tissue microarray (TMA) was described previously 315. Briefly, formalin-fixed, paraffin-embedded prostate specimens from the

University of Wisconsin Pathology archives were used with the approval of the

Institutional Review Board. The TMA consisted of 41 localized prostate cancers

(Pca_local, median Gleason score 7, with a range of 6 to 9), 18 aggressive prostate cancers (Pca_aggr, prostate cancer with lymph node metastasis, median Gleason score 8,

210 with a range of 7 to 9), 18 metastatic prostate cancers (Met, in either lymph nodes or other organs), 24 benign prostate hyperplasia (BPH), 19 high-grade intraepithelial neoplasia (HGPIN), and 48 benign prostate tissues. The median patient age was 62 years (range 37 to 82 years). Cores were 0.6 mm in diameter and spaced 0.8 mm on-center using a Manual Tissue Arrayer (Beecher Instruments, Sun Prairie, WI; Model

MTA-1). Each specimen has duplicated cores. Five micron sections of the TMA were used for staining and analysis. Rabbit E-cadherin antibody (Abnova) and Alexa Flour

555 conjugated goat anti-rabbit (Invitrogen) were used to define and visualize epithelial compartment. TMPRSS2 mouse mAb, biotinylated goat anti-mouse (Biocare Medical), streptavidin-HRP (Biocare Medical), and Alexa Fluor 647 tyramide (Invitrogen) were used to detect and visualize TMPRSS2. 4',6-diamidino-2-phenylindole (DAPI;

Invitrogen) was used to define and visualize nuclear compartment.

After staining, the TMA slide was scanned and images were acquired using the

AQUA platform (HistoRx, CT). Tissue microarray cores without sufficient epithelium

(<5%) or with poor staining quality (i.e., section folding, excess trapping of fluorochrome, etc.) were excluded from analysis. The AQUA_1.5 system (HistoRx, New Haven, CT) was used for TMPRSS2 quantification and analysis. Image acquisition and algorithmic analysis of TMA using automated quantitative analysis have been previously described

211 extensively 315, 316.

7.3.7 Western blot analysis

Proteins for Western blot analysis was prepared from cell cultures by washing the cells three times with PBS followed by lysis with 1% Triton X-100 in PBS. Insoluble debris was removed by centrifugation, and the protein concentration determined using

BCA protein assay reagents (Pierce, Rockford, IL) according to the manufacturer’s instructions. Cell lysates or human semen were diluted with 5X sample buffer in the absence or presence of reducing agent and incubated either at room temperature or 95 oC

(boiled) for five minutes before SDS-polyacrylamide gel (SDS-PAGE). The proteins were resolved by 7.5% SDS-PAGE, transferred to Protran nitrocellulose membranes

(Schleicher and Schuell, Keene, NH), and probed with the desired monoclonal antibodies.

The binding of primary antibody was detected using a goat anti-mouse-HRP-conjugated secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA), and

Western Lightning Chemiluminescence Reagent Plus (Perkin-Elmer, Boston, MA).

7.3.8 Diagonal SDS-PAGE

LNCaP cell lysates or human seminal plasma samples were diluted with 5X sample buffer and incubated either at room temperature or 95 oC for five minutes and then resolved by SDS-PAGE. The gel was sliced into vertical strips, which were then boiled

212 in SDS sample buffer in the absence or presence of DTT. The boiled gel strips were then placed horizontally on the top of fresh SDS-acrylamide gels, and electrophoresed using the same conditions as the first dimension gel, after which the proteins were transferred to nitrocellulose membrane and probed with TMPRSS2 mAb, AL20.

7.3.9 PNGase F assay

Fifty-microgram samples of protein lysate from LNCaP cells was boiled under denaturing conditions (0.5% SDS, 40 mM DTT) for 5 min. The samples were adjusted to

0.05 M sodium phosphate and 1% NP-40. PNGase F (500 U) was added, and the samples were then incubated at 37 oC for 1 hour. The reaction was terminated by boiling the samples with 5X SDS sample buffer containing DTT for 5 min, followed by

SDS-PAGE and immunoblot analysis. Control samples were treated in the same manner except for the addition of the PNGase F.

7.3.10 Analysis of TMPRSS2 and matriptase shedding

LNCaP cells were grown in protein-free media for 24 hours, after which the conditioned media was harvested and concentrated up to 100 fold. The cells were lysated with 1% Triton X-100 in PBS and the protein was quantified using BCA protein assay reagents. Both cell lysates and concentrated media were then subjected to immunoblot using TMPRSS2 mAb, AL20.

213 7.3.11 Preparation of prostasomes

Human semen samples were centrifuged at 1,000g for 15 min to remove spermatozoa and cell debris, after which the supernatants were ultracentrifuged at

105,000g for 2 h. The resulting pellets, which contain prostasomes, were resuspended in PBS containing 1% Triton X-100 and then centrifuged at 21,000g for 15 min to remove the insoluble fraction.

7.4 Results

7.4.1 Generation and characterization of TMPRSS2 monoclonal antibodies

A schematic representation of the domain structure of wild-type TMPRSS2 and the mutant TMPRSS2 expression construct is shown in Figure 7.1A. The recombinant

TMPRSS2 protein was expressed and purified from the conditioned media as described in the Materials and Methods section. The purified protein is seen as a band of approximately 55-kDa on coomassie blue stained gels (Figure 7.1B), consistent with the calculated molecular mass of the TMPRSS2 mutant. A minor protein band of approximately 40-kDa was consistently observed and may represent a proteolytic degradation product of the transmembrane protease. Samples of the purified protein were subjected to mass spectrometry-based proteomic analysis. Thirteen tryptic peptides

214

Figure 7.1 TMPRSS2 recombinant protein and mAbs.

A. Expression of TMPRSS2 in mammalian cells. A schematic representation of the structure of TMPRSS2 is presented at the top. The protease consists of 492 amino acids, has a calculated mass of 53,859-Da, and contains a transmembrane domain, a scavenger receptor cysteine-rich (SRCR) domain, a LDLRA domain, and a serine protease domain.

The two putative N-glycosylation sites are at N-213 and N-249. The serine protease domain is at the carboxyl terminus, and the active site triad was labeled as H-D-S. A truncated and mutated TMPRSS2 expression construct was generated as described in the methods section resulting in the construct shown. B. Purification of TMPRSS2. The recombinant TMPRSS2 protein was produced and purified from HEK 293T cells and

215 resolved by SDS-PAGE and stained by colloidal Coomassie blue to show the protein patterns and to asses the purity of the TMPRSS2 preparation. C. Characterization of a

TMPRSS2 monoclonal antibody. In order to evaluate the sensitivity of the TMPRSS2 monoclonal antibody, AL20, different amounts of purified TMPRSS2 protein as indicated were resolved by SDS-PAGE under reducing and boiled conditions. Immunoblot analysis was conducted with the AL20 monoclonal antibody at 2 μg/ml.

216 were identified, all of which matched TMPRSS2 (Figure 7.2) confirming the identity of the purified recombinant protein. Monoclonal antibodies directed against TMPRSS2 were generated by conventional hybridoma technologies using the purified TMPRSS2 mutant protein. More than 20 positive clones were selected based on immunodetection of recombinant TMPRSS2 protein (data not shown), of which mAb AL20 was selected having the highest sensitivity and specificity. As shown in Figure 7.1C, when used at 2

µg/ml, the mAb was able to detect as little as 2.5 ng of recombinant TMPRSS2. These data demonstrate that the TMPRSS2 mAb, AL20, is a sensitive immunological reagent.

This antibody was used in the studies described below.

7.4.2 Aberrant subcellular distribution and increased expression of TMPRSS2

protein in prostate cancer cells

Using the highly sensitive TMPRSS2 mAb, we compared the subcellular localization and expression of TMPRSS2 in prostate cancer cells versus their normal counterparts. Preliminary immunohistochemical staining of prostate tumor tissue sections from seven different patients showed plasma membrane localization of

TMPRSS2 in all specimens (Figure 7.3, A-F), consistent with its integral transmembrane protein nature. In the nonaffected areas of the prostate tumor tissue, which are representative of the normal prostate gland, TMPRSS2 was predominantly detected at the

217

Figure 7.2 Verification of TMPRSS2 by proteomic analysis.

The purified TMPRSS2 recombinant protein was resolved by SDS-PAGE and stained by

Coomassie blue. The protein band was subjected to proteomic identification by MS/MS and the peptides identified are indicated by lines under the sequence.

218

Figure 7.3 Cellular localization of TMPRSS2 in normal and carcinomatous prostate tissues.

Paraffin-embedded human prostate tumor tissue sections were stained by immunohistochemistry using TMPRSS2 mAb AL20 (panels B, C, E, and F) and mouse

IgG (panels A and D). Positive staining for TMPRSS2 was observed as brown precipitates (diaminobenzidine), and the nuclei were counterstained with hematoxylin.

The micrographs in panels A, B and C were taken from non-cancerous acini portions of the prostate tumor (non-affected) and panels D, E, and F from the cancer portions. The sizes of bars are indicated in each panel. Open arrows show TMPRSS2 on the apical surface of luminal cells. Closed arrows show the difference of TMPRSS2 protein expression in non-affected prostate tissue and carcinomatous prostate tissue.

219 apical plasma membrane of the luminal epithelial cells of the prostate acini (Figure. 3, B and C, open arrows). The staining was essentially limited to the uppermost portion of plasma membrane either directly exposed to the lumen or in regions of cell-cell contacts in these histologically normal prostate tissues. In the adjacent prostate tumor, however, although the TMPRSS2 was still predominantly localized at the cell surface, the staining was observed to be around the entire cell surface (Figure 7.3, F, close arrows). In addition, the expression levels of TMPRSS2 protein in the prostate cancer cells were much higher than in the neighboring prostate luminal epithelial cells (Figure 7.3, compare

E with B).

To further examine TMPRSS2 protein levels in a more quantitative way and to examine the role of TMPRSS2 in prostate cancer and in tumor progression, a prostate cancer tissue microarray (TMA) was stained using the TMPRSS2 antibody and protein levels quantitatively determined using the AQUA system. The TMA was constructed with duplicate cores of 41 localized prostate cancers (Pca_local) 18 aggressive prostate cancers (Pca_aggr), 18 metastatic prostate cancers (Met), 24 benign prostate hyperplasia

(BPH), 19 high-grade intraepithelial neoplasia (HGPIN), and 48 benign prostate tissues.

The reliability of the TMA and AQUA system was tested using a known prostate cancer

-methylacyl coenzyme A racemase (AMACR) (data not shown) 315. We found

220 that the TMPRSS2 protein levels were higher in proliferative prostate diseases compared to benign prostatic tissue (Table 7.1). TMPRSS2 levels were significantly higher in BPH and localized and aggressive prostate cancers (mean AQUA scores, 316, 302, and 366, respectively) than benign prostatic tissue (mean AQUA score, 185). However, the elevation of TMPRSS2 protein expression in metastatic cancer and HGPIN (mean AQUA score, 242 and 249, respectively) did not reach statistical significance when compared with that in benign prostatic tissue (Table 1). Taken together, these data suggest that both the expression level and subcellular localization of TMPRSS2 may be deregulated in prostate cancer cells.

7.4.3 TMPRSS2 in prostate cancer cell lines

The expression of TMPRSS2 by prostate cancer cell lines, including LNCaP, PC3, and DU145, has been well characterized at the mRNA level. TMPRSS2 mRNA was consistently detected in LNCaP and not in DU145; PC3 may not express the protease or express at very low level 187, 188. TMPRSS2-ETS gene fusion was not found in LNCaP and PC3 cells 311. So as an additional control for the specificity of our antibody, the expression status of TMPRSS2 at the protein level was investigated in these three cancer cell lines. Of these three prostate cancer cell lines, TMPRSS2 protein was only detected in LNCaP cells (Figure 7.4A) and not in PC3 or DU145 cells (Figure 7.5), consistent with

221 Table 7.1AQUA Scores of TMPRSS2

Dx Age (y) GS N AQUA Score P-value

Benign 61.3 80 185±18

BPH 67.8 39 316±42 0.005*

HGPIN 62.9 36 249±25 0.097

Pca_local 58.2 6-9 (7) 72 302±25 0.003*

Pca_aggr 62.2 7-9 (8) 29 366±60 0.016*

Met 67.6 24 242±40 0.301

GS, Gleason score; N, core number analyzed; P, significance level using

ANCOVA

*P<0.05.

222

Figure 7.4 The expression status of endogenous TMPRSS2 protein in human prostate cancer cells.

A. TMPRSS2 protein expression in LNCaP prostate cancer cells. LNCaP cells were lysed using PBS containing 1% Triton X-100. Equal amounts of total cellular proteins were mixed with SDS sample buffer in the absence (NR) or presence (R) of reducing agents and incubated either at room temperature (NB) or 95oC (B) for 5 minutes, followed by SDS-PAGE and western blot analysis using the monoclonal antibody AL20.

NRNB stands for nonreducing and non-boiled conditions. NR,B stands for nonreducing and boiled conditions. R,B stands for reducing and boiled conditions. B. TMPRSS2 is present in non-covalent-linked complexes in LNCaP cells. LNCaP cell lysates were resolved by diagonal gel electrophoresis under non-reducing and non-boiled conditions in the first dimension and under non-reducing and boiled condition in the second dimension.

Immunoblot analysis of the second dimensional gel was carried out with the TMPRSS2 mAb AL20. C. TMPRSS2 is also present in disulfide bond-linked complexes in LNCaP 223 cells. LNCaP cell lysates were resolved by diagonal gel electrophoresis under non-reducing and boiled conditions in the first dimension and under reducing and boiled condition in the second dimension. Immunoblot analysis of the second dimensional gel was performed with the TMPRSS2 mAb AL20.

224

Figure 7.5 The expression status of endogenous TMPRSS2 protein in human prostate cancer cells.

PC3 and DU145 cells were lysed using PBS containing 1% Triton X-100. Equal amounts of total cellular proteins were mixed with SDS sample buffer in the absence (NR) or presence (R) of reducing agents and incubated either at room temperature (NB) or

95oC (B) for 5 minutes, followed by SDS-PAGE and Western blot analysis using the monoclonal antibody AL20. NRNB stands for non-reducing and non-boiled condition.

NR, B stands for non-reducing and boiled condition. R, B stands for reducing and boiled condition.

225 the mRNA expression status.

In LNCaP cells, endogenous TMPRSS2 was detected as a major protein band near to the 55-kDa marker under nonreducing and nonboiled conditions (Figure 7.4A, lane 1,

NRNB). Some additional minor bands were also detected, particularly after longer exposure to X-ray film. Under boiled and nonreducing conditions (Figure 7.4A, lane 2,

NR, B), a second protein band with a size of approximately 40-kDa was seen. The ratio of the 40-kDa species relative to the 56-kDa form varied among different experiments.

Under boiled and reducing conditions, TMPRSS2 was detected as two protein bands with apparent sizes of 58- and 42-kDa (Figure 7.4A, lane 3, R,B). The decreased rate of migration of TMPRSS2 on SDS-PAGE caused by chemical reduction is consistent with the fact that TMPRSS2 contains many cysteine residues, which likely participate in 9 disulfide bonds.

7.4.4 TMPRSS2 is present in both noncovalent and disulfide-linked complexes

in LNCaP cells

The appearance of the 40-kDa TMPRSS2 fragment after boiling treatment (Figure

7.4A, comparing lane 2 with lane 1) suggests that the 40-kDa TMPRSS2 species may be present in some form of noncovalent complexes, which can be dissociated by boiling treatment. Furthermore, the epitope recognized by the TMPRSS2 mAb on the 40-kDa

226 TMPRSS2 species seems to be largely masked in the putative TMPRSS2 complexes, given that the TMPRSS2 mAb only detects these putative complexes very weakly under nonreducing and nonboiled conditions (Figure 7.4A, lane 1). We therefore set out to test this hypothesis using a two-dimensional, diagonal gel electrophoresis (Figure 7.4B). In the first dimension under nonboiled and nonreducing conditions, TMPRSS2 was detected as a major 56-kDa protein band with few minor bands with greater or smaller sizes than the 56-kDa TMPRSS2 monomer (Figure 7.4B, 1D NRNB). In the second dimension under boiled and nonreducing conditions, the major 56-kDa TMPRSS2 protein band was detected at the same size as 56-kDa, suggesting that the 56-kDa species is likely to be a monomer and not in a complex with other proteins. In addition to the major 56-kDa band, the 56-kDa species was also detected at positions corresponding to 80-90-kDa and

120-kDa in the first dimensional gel, despite the fact that there was no TMPRSS2 detected at these sizes in first dimensional gel. These data suggest that the 56-kDa

TMPRSS2 species can form multiple noncovalent complexes in which the epitope recognized by mAb AL 20 is largely masked. Furthermore, in the second dimensional gel, the 40-kDa TMPRSS2 species was detected as a continuous protein band with corresponding molecular weight from 120-kDa through 40-kDa in the first dimension

(Figure 7.4B). These data suggest that the 40-kDa TMPRSS2 species undergoes similar

227 complex formation to the 56-kDa species. These data indicate that TMPRSS2 is able to form noncovalent complexes.

In addition to the noncovalent complexes, TMPRSS2 also forms disulfide-linked complexes, which can be clearly detected by diagonal gel electrophoresis (Figure 7.4C).

In the first dimension, under boiled and nonreducing conditions, TMPRSS2 was detected as a major 56-kDa band and a minor 40-kDa band, as expected (Figure 7.4C, 1D NR,B).

In the second dimension under boiled and reducing conditions, both the 56- and 40-kDa

TMPRSS2 species were detected along the diagonal line (Figure 7.4C, 2D R,B), suggesting that both are single-chain TMPRSS2 zymogen species. Consistent with the previous result, exposure to the reducing agents resulted in minor changes to their migration rates between the two gel systems. The 56-kDa TMPRSS2 zymogen was also detected in the second dimension in three positions corresponding to the molecular weights of around 130-kDa, 250-kDa and a very high molecular weight species in the first dimensional gel. These data suggest that there are at least three disulfide-linked

TMPRSS2 zymogen complexes present in LNCaP cells. Similar disulfide-linked complexes involving the 40-kDa TMPRSS2 species were also detected in this experiment.

7.4.5 TMPRSS2 is lightly glycosylated

228 Because the calculated mass of TMPRSS2 is 53,859-Da, the 58-kDa TMPRSS2 species detected in LNCaP cells is likely the mature full-length form. Human

TMPRSS2 contains two putative N-glycosylation sites, N-213 and N-249 302 and the mouse protein contains three, N-111, N-212, and N-474 303. The difference between the calculated mass of approximately 54-kDa and 58-kDa is most likely attributable to posttranslational modifications, specifically glycosylation. We used two different approaches to investigate the extent to which glycosylation contributes to the apparent mass of endogenous TMPRSS2. Tunicamycin blocks assembly of the GlcNAc-dolichol complex and has been broadly used to study the glycosylation of proteins, and so we cultured LNCaP cells in the presence of tunicamycin for various time periods up to eight hours (Figure 7.6A). After two hours of tunicamycin treatment a 53-kDa TMPRSS2 species began to appear, and after between four to eight hours of treatment, the 58-kDa

TMPRSS2 form had disappeared (Figure 7.6A). The 53-kDa form is essentially the same as the calculated mass of the protease (Figure 7.6A). Exposure to tunicamycin also resulted in the disappearance of the 40-kDa TMPRSS2 species and the appearance of a smaller TMPRSS2 species that could be detected using more protein and longer exposure of the blot to X-ray film (Figure 7.6A, lane 8*). In an alternate approach we made use of the enzyme PNGase F, which cleaves between the innermost

229

Figure 7.6 N-glycosylation of TMPRSS2 is responsible for a modest increase in molecular mass.

A. LNCaP cells were treated with tunicamycin (5 g/ml) for the indicated times. Cell lysates were analyzed for TMPRSS2 by immunoblot using the mAb AL20. To facilitate detection of the deglycosylated 40-kDa species, a longer exposure of the immunoblot for the 8 hour time point was taken (8*). B. LNCaP cell lysates were incubated in the absence (-) or presence (+) of PNGase F and TMPRSS2 was probed with the mAb AL20.

230 N-acetylglucosamine (GlcNAc) and asparagine residues of high mannose glycosylation.

Treatment with this enzyme should remove all the N-glycosylation from proteins such as

TMPRSS2. As shown in Figure 7.6B, digestion with PNGase F resulted in the disappearance of both the 56-kDa and 40-kDa TMPRSS2 and the concomitant appearance of two smaller TMPRSS2 species. Taken together, these data suggest that glycosylation contributes about 4 to 5-kDa to the apparent molecular mass of TMPRSS2, and removal of the N-glycan results in an apparent molecular mass almost identical to the mass calculated based on the amino acid sequence.

7.4.6 The shedding of TMPRSS2 into the extracellular milieu is very limited in

LNCaP cells

Shedding to the extracellular milieu after zymogen activation has been observed for other TTSP family members, such as matriptase 184. Taking advantage of the specificity and sensitivity of the TMPRSS2 antibody, we examined whether LNCaP cells release

TMPRSS2 into media. Matriptase was used as a positive control for the quality of the conditioned media because this enzyme is known to be activated and shed by LNCaP cells 184. As expected, TMPRSS2 was readily detected in the LNCaP cell lysates but was completely absent from the conditioned media. In contrast to TMPRSS2, matriptase was detected both in the lysates and the conditioned media (Figure 7.7, A and

231

Figure 7.7 Shedding of TMPRSS2 to the extracellular milieu by LNCaP cells is very limited.

LNCaP cells were grown in serum-free media for 24 hours, and then the conditioned medium and cell lysates were harvested. The conditioned medium was concentrated up to 100 fold. The cell lysates (C) and the conditioned medium (M) were analyzed for total matriptase (Total MTP) using the matriptase mAb M32 and TMPRSS2 using the mAb AL20.

232 B). Furthermore, both the latent matriptase at 70-kDa and activated matriptase in HAI-1 complex at 95-kDa were detected in the conditioned media (Figure 7.7), suggesting that matriptase underwent activation and shedding. These data suggest that shedding of

TMPRSS2 appears to be very limited in LNCaP cells.

7.4.7 TMPRSS2 is secreted as both noncovalent and disulfide bond-linked

complexes into human semen

Although TMPRSS2 is apparently not secreted by LNCaP cells (Figure 7.7), the expression of TMPRSS2 on the apical plasma membrane of the prostatic luminal epithelial cells suggests that TMPRSS2 may be secreted into semen. We therefore examined the expression status of TMPRSS2 in human semen from twenty three healthy men of different ages and races. When the samples were not treated with reducing agents or high temperature (nonreducing and nonboiled conditions), TMPRSS2 from human semen was detected as multiple protein bands of 40-kDa, 56-kDa, 68-kDa,

100-kDa, 120-kDa, and 140-kDa (Figure 7.8A, lane 1, NRNB). These species may be

TMPRSS2 complexes because their size was greater than that of full-length TMPRSS2.

Under boiled and nonreducing conditions, the protease was also detected as multiple protein bands, though the bands were of significantly increased intensity (Figure 7.8A, lane 2, NR,B) suggesting that the heat treatment may either expose the epitope

233

Figure 7.8 TMPRSS2 is present in multiple complexes in human semen.

A. Equal amount of human semen were mixed with SDS sample buffer in the absence

(NR) or presence (R) of reducing agents and incubated either at room temperature (NB) or 95℃ (B) for 5 minutes, followed by SDS-PAGE and western blot analysis using

TMPRSS2 mAb AL20. B. Human semen was resolved by diagonal gel electrophoresis under non-reducing and boiled conditions in the first dimension and under reducing and boiled condition in the second dimension. Immunoblot analysis of the second dimensional gel was carried out with the TMPRSS2 mAb AL20.

234 recognized by TMPRSS2 mAb or cause the dissociation of TMPRSS2 complexes, or both. After treating with reducing agents, TMPRSS2 was detected as one major protein band of 42-kDa and a minor band of 27-kDa (Figure 7.8A, lane 3, R,B). The disappearance of most of the higher molecular weight TMPRSS2 species and the appearance of the 42-kDa species after chemical reduction suggests that TMPRSS2 is present in multiple disulfide-linked complexes in semen that are converted to the 42-kDa species after chemical reduction (Figure 7.8A, lane 3, R,B). This hypothesis was further supported by a diagonal gel electrophoresis analysis with the second dimension under reducing conditions (Figure 7.8B). Immunoblot analysis revealed that all of these

TMPRSS2 complexes species were converted to the 42-kDa species. These data suggest that TMPRSS2 is secreted into human semen as the 42-kDa species in the form of multiple noncovalent and disulfide-linked complexes. Although a 56-kDa TMPRSS2 species was detected in semen under boiled and nonreducing conditions (Figure 7.8A, lane 2), this 56-kDa species is not the same as the full-length TMPRSS2 detected in

LNCaP cells under boiled and nonreducing conditions (Figure 7.4A, lane 2), because of the fact that the 56-kDa species is a disulfide-linked complex that contains the 42-kDa species. A minor 27-kDa species detected in semen may be a degradation product because its levels vary among different batches of human semen and increases along with

235 the disappearance of the high molecular weight species during the processing of the semen.

7.4.8 TMPRSS2 is secreted as components of the prostasome into human

semen

In addition to the formation of multiple protein complexes, we also noticed that semen-derived TMPRSS2 complexes appear to be present in gigantic complex(es) with apparent sizes of up to several million Daltons, because they eluted together in the void volume of a Superose 6 sizing column (Figure 7.9A, upper panel). Human semen is known to contain organelle-like membrane vesicles, named prostasomes, which have previously been isolated in the void volume of various sizing columns and by ultracentrifugation 317, 318, and so we inferred that TMPRSS2 and its complexes may reside in these membrane vesicles. Two approaches were taken to demonstrate that

TMPRSS2 is a component of prostasomes. When seminal plasma was loaded onto

Superose 6 sizing column, in the absence of detergent, all the TMPRSS2 species were found to elute in fractions 17 through 22, which represents the void volume (Figure 7.9A, upper panel). In the presence of Triton X-100, TMPRSS2 species eluted in fractions 31 through 35, which contain proteins with molecular mass between 200- and 29-kDa

(Figure 7.9A, lower panel). The nonionic detergent apparently solubilized the

236

Figure 7.9 TMPRSS2 is expressed as components of the prostasome.

A. Seminal plasma was fractionated in the absence (upper panel) or presence (lower panel) of 1% Triton X-100 by the Superose 6 sizing column. The Superose 6 fractions were examined by immunoblot analysis using TMPRSS2 mAb AL20. In the absence of

Triton X-100, TMPRSS2 was detected in the void volume as indicated by co-elution with blue dextran (upper panel). In the presence of non-ionic detergent, TMPRSS2 was eluted in the fractions whose apparent molecular masses were distributed from 200-kDa to 29-kDa as determined by comparison to standard proteins which included blue dextran

(> 2,000-kDa), thyroglobulin (669-kDa), apoferritin (443-kDa), b-amylase (200-kDa), carbonic anhydrase (29-kDa) as indicated (lower panel). B. Upper panel: Flow chart of 237 human semen fractionation. Human semen was fractionated by centrifugation at

1,000xg for 15 min into sperm fractions and seminal plasma. The latter was further fractionated by centrifugation at 105,000xg for 2 hours into the insoluble fractions, representing the prostasomes, and the soluble fractions, the seminal fluids.

Prostasomes were then lysed with 1% Triton X-100 in PBS and the insoluble fractions were further extracted by 1% SDS in PBS. Lower panel: The four semen fractions were subjected to SDS-PAGE and followed by immunoblot analyses for TMPRSS2. The

TTSP, detected in seminal plasma (lane 1), was fractionated into the prostasome Triton

X-100-soluble fractions (lane 3), but not in the seminal fluid (lane 2) and the Triton

X-100-insoluble fractions (lane 4).

238 membrane vesicles and released the TMPRSS2 from prostasomes, resulting in the change in the sizing column elution profile. Because prostasomes can be precipitated by ultracentrifugation, we fractionated human semen into 4 different fractions by differential centrifugation as described in Figure 7.9B. These fractions include seminal plasma

(after removal of sperm and cell debris), seminal fluid (after removal of prostasomes), prostasome-Triton X-100 soluble fraction, and prostasome-Triton X-100 insoluble fraction. All of the TMPRSS2 complexes detected in the seminal plasma (Figure 7.9B, lane 1) were all spun down with the prostasomes and detected in the prostasome

Triton-soluble fractions (Figure 7.9B, lane 3) and not in the seminal fluid fraction (Figure

7.9B, lane 2). These data suggest that TMPRSS2 is secreted by prostate luminal epithelial cells into semen as components of prostasomes.

7.5 Discussion

TMPRSS2, an androgen responsive gene, is highly expressed in prostate epithelial cells and in human prostate cancers. Here we report new insights into the regulation and function of this TTPS, including 1) its presence in high-molecular-weight complexes which, are SDS-resistant and disulfide-linked, 2) its localization to the apical plasma membrane of prostate luminal epithelial cells, 3) its presence in prostasomes,

239 membranous vesicles in semen, and 4) its increased expression and misregulated subcellular localization in prostate cancer cells.

TMPRSS2 was previously reported to have a mass of either 70-kDa or -65 kDa 188,

313, 314, which is different from that determined in this study and is greater than the calculated mass. These discrepancies may be attributable to the use of different

TMPRSS2 antibodies, which were raised against synthetic peptides or recombinant proteins produced in bacteria. As shown in the Figure 7.10, the three commercial

TMPRSS2 antibodies we tested, produced against synthetic peptides, appear to be suboptimal for the detection of endogenous TMPRSS2 (Figure 7.10). In contrast, our mAbs were generated by using TMPRSS2 produced in HEK293T cells and have proven to be highly specific and very sensitive (Figures 1C and 10). In our study, the vast majority of TMPRSS2 was detected either at an apparent mass of 58-kDa in LNCaP cells or 42-kDa in semen. The 58-kDa species appears to be the full-length form of the protein and the size is close to the 53,859-Da calculated size of the protease. The small difference between the calculated and observed size is due to glycosylation (Figure 7.6).

In LNCaP prostate cancer cells, the 58-kDa species was consistently detected as the major form of TMPRSS2, though the 42-kDa species was also detected and its levels varied from experiment to experiment. In contrast, the 42-kDa TMPRSS2 appears to be

240

Figure 7.10 Verification of the molecular weight of TMPRSS2 by commercially available antibodies.

Endogenous TMPRSS2 protein from LNCaP cells was used to test the specificity of three commercially available antibodies. LNCaP cells were lysed using PBS containing 1%

Triton X-100. Equal amounts of total proteins were mixed with SDS sample buffer in the absence (NR) or presence (R) of reducing agents and incubated either at room temperature (NB) or 95 oC (B) for 5 minutes. Proteins were resolved by SDS-PAGE followed by immunoblottings with the three TMPRSS2 antibodies (ab56110, ab56111, and ab56112) and the TMPRSS2 mAb AL20. NRNB stands for non-reducing and non-boiled conditions. NR,B stands for non-reducing and boiled conditions. R,B stands

241 for reducing and boiled conditions.

242 the major form detected in human semen and 58-kDa full-length TMPRSS2 was not detected (Figure 7.8A, lane 3, R,B). The conversion of full-length TMPRSS2 to 42-kDa form may occur prior to its release. In spite of their differential presentation in LNCaP cells and semen, both the 58-kDa and 42-kDa TMPRSS2 forms seem to resemble one another with respect to the formation of complexes. This is particularly evident when comparing the pattern of disulfide-linked complexes containing the 58-kDa vs. the

42-kDa form (Figure 7.4C).

Detection of TMPRSS2 under reducing and boiled conditions as a 58-kDa protein band close to the calculated size of the full-length TMPRSS2 suggests that the TTSP is synthesized as a latent protease, in common with most serine proteases. TMPRSS2 is, however, unusual among TTSPs regarding its formation of multiple complexes.

TMPRSS2 can form non-covalent and disulfide-linked complexes both in vitro in LNCaP cells and in vivo in prostatic secretions. The vast majority of the TMPRSS2 released from complexes is in the latent form which would not be expected to form stable complexes with endogenous protease inhibitors. The TMPRSS2 complexes are, therefore, apparently not a product of conventional mechanisms governing the zymogen activation and inhibition of active serine proteases. Furthermore, TMPRSS2 has 22 cysteine residues. Although the serine protease domain in TMPRSS2 contains eight

243 Cys residues in conserved positions, the pairing counterpart cysteine for Cys140 is absent.

This unpaired Cys in the serine protease domain may be the structural element which facilitates the formation of disulfide-linked complexes. It will be of great interest to determine whether the extensive formation of TMPRSS2 complexes represents a novel mechanism by which the protease is regulated and to explore the functional significance of complex formation.

Human TMPRSS2 contains 492 amino acid residues. Its activation theoretically involves a proteolytic cleavage between amino acids 255 Arg-Ilu 256, resulting in two fragments with similar sizes of around 27-kDa, held together through a disulfide linkage.

While the exact epitope recognized by our TMPRSS2 mAb has not been determined, the detection of a 27-kDa protein band by the TMPRSS2 mAb under reducing conditions is likely an indication of TMPRSS2 activation. The level of TMPRSS2 activation in

LNCaP prostate cancer cells and human semen, however, appears to be very limited since there was no detectable 27-kDa TMPRSS2 protein present under reducing conditions

(Figures 4A and 8A).

In addition to the unusual complex formation, the targeting of the protease to the apical plasma membrane and its shedding as a component of prostasomes into semen provides insights into the physiological roles of TMPRSS2. Like some other

244 prostasome proteins, which were also detected on the apical plasma membrane of the prostate luminal cells 319, the presence of TMPRSS2 both on the apical plasma membrane of luminal cells in the prostate gland and in the prostasomes suggests that TMPRSS2 may be a marker for prostasomes and its physiological functions may be associated with this organelle-like structure. ENaC, a putative TMPRSS2 substrate, is also localized at the apical plasma membrane 306. If TMPRSS2 can be converted into an active enzyme at the apical membrane, the protease may activate ENaC to regulate sodium homeostasis within the prostate gland. The presence of TMPRSS2 in prostasomes suggests that it may also regulate prostasome functions. Prostasomes have been proposed to be involved in protection of the spermatozoa in the acidic milieu of the vagina 320, 321, stabilization of sperm plasma membrane and delay of acrosome reaction 322-325, enhancement of sperm motility 321, 326-330, protection of spermatozoa from phagocytosis and complement attack, and inhibition of viral activity. Many of the proposed biological functions of the prostasomes facilitate sperm function and enhance male reproduction. However, the loss of TMPRSS2 in TMPRSS2-deficient mice did not adversely influence fertility, reduce survival, result in prostate hyperplasia or carcinoma, or alter prostatic luminal epithelial cell regrowth following castration and androgen replacement 305. It remains to be determined if the lack of phenotype in the

245 TMPRSS2-deficient mouse is due to the fact that TMPRSS2 is not involved in these processes, or redundancy with other proteases fulfilling the functions of TMPRSS2, or reflects difference between humans and mice.

TMPRSS2 recently received a great deal of attention as a participant in gene fusions of the promoter regions of the protease with the ETS family members found in the majority of human prostate cancers 187, 310. This gene fusion is a frequent and early event in prostate cancer pathogenesis and may promote cancer progression into a more aggressive phenotype through the engagement of the downstream targets of ETS, including plasminogen activation pathway that mediates cancer invasion and metastasis

331. While the gene fusion should not affect the function of the TMPRSS2 gene product as a transmembrane serine protease, the gene fusion appears to reduce the expression of

TMPRSS2 at mRNA levels in those prostate tumors harboring TMPRSS2/ERG 312.

Although the expression of TMPRSS2 transcript is decreased in prostate cancer cells with

TMPRSS2/ETS fusions, several studies have shown that TMPRSS2 expression is higher in prostate cancer tissue than in normal prostate, and increases with Gleason score, tumor grade and stage 187, 188, 308, 309, 312, 332. Furthermore, our current study further confirms the increased expression of TMPRSS2 in prostate cancer at the protein level using the AQUA system to quantitate IHC staining in a prostate TMA. Therefore, TMPRSS2 may

246 contribute to prostate cancer not only by the increased expression but also through aberrant subcellular localization due to the loss of epithelial polarity in the transformed cells. This latter abnormality may allow the protease to inappropriately gain access to and/or activate some cancer-promoting substrates which its normal subcellular localization would preclude under normal physiological conditions.

7.6 Acknowledgements

This study was supported by National Cancer Institute (NCI) Grants RO1 CA

096851 and RO1 CA 104944 (to C.-Y. Lin) and Taiwan National Science Council Grant

NSC 97-2320-B-002-052-MY3.

247 BibliographyReference List

1. Breathnach,A. & Bannister,L.H. Integumental system: skin and breasts in Gray's anatomy (ed. Williams,P.L.) 376-412 (Churchill Livingston, New York, 1995).

2. David Gawkrodger Dermatology, An Illustrated Colour Text2002).

3. Proksch,E., Brandner,J.M., & Jensen,J.M. The skin: an indispensable barrier. Exp. Dermatol. 17, 1063-1072 (2008).

4. Madison,K.C. Barrier function of the skin: "la raison d'etre" of the epidermis. J. Invest Dermatol. 121, 231-241 (2003).

5. Stucker,M. et al. The cutaneous uptake of atmospheric oxygen contributes significantly to the oxygen supply of human dermis and epidermis. J. Physiol 538, 985-994 (2002).

6. Gerard J Tortora The Integumentary System in Principles of Anatomy and Physiology 139-145 (John Wiley & Sons Inc., 2008).

7. Eckert,R.L. Structure, function, and differentiation of the keratinocyte. Physiol Rev. 69, 1316-1346 (1989).

8. Arthur C.Guyton & John E.Hall Body temperature, temperature regulation, and fever in Textbook of medical physiology 911-922 (W B Saunders Co, 1996).

9. Holbrook,K.A. & Wolff,K. The structure and development of skin in Dermatology in General Medicine (eds. Fitzpatrick,T.B., Eisen,A.Z., Wolff,K., Freedberg,I.M. & Austen,K.F.) 93-131 (McGraw-Hill, New York, 1987).

10. Odland,G.F. Histology and fine structure of the epidermis in The Skin by 30 authors (eds. Helwig,E.B. & Mostofi,F.K.) 28-46 (Williams & Wilkins, Baltimore, 1971).

11. Wheater,P.R., Burkitt,H.G., & Daniels,V.G. Skin in Wheater's functional histology, a text and colour atlas (eds. Burkitt,H.G., Young,B. &

248 Heath,J.W.) 153-169 (Churchill Livingston, New York, 1993).

12. Watt,F.M. Terminal differentiation of epidermal keratinocytes. Curr. Opin. Cell Biol. 1, 1107-1115 (1989).

13. Matoltsy,A.G. & Matoltsy,M. A study of morphological and chemical properties of keratohyalin granules. J. Invest Dermatol. 38, 237-247 (1962).

14. Lindsay,D.T. The integument in Functional human anatomy (ed. Simth,J.M.) 345-375 (Mosby Year Book, St. Louis, 1996).

15. Stenn,K.S. & Malhotra,R. Epithelialization in Wound Healing, Biochemical & Clinical aspects (eds. Cohen,K.I., Diegelmann,R.F. & Lindblad,W.J.) 115-127 (W. B. Saunders, Philadelphia, 1992).

16. Matoltsy,A.G. Desmosomes, filaments, and keratohyaline granules: their role in the stabilization and keratinization of the epidermis. J. Invest Dermatol. 65, 127-142 (1975).

17. Ross,M.H., Romrell,L.J., & Kaye,G.I. Integumentary system in Histology, a text and atlas (ed. Coryell,P.A.) 370-403 (Williams & Wilkins, Baltimore, 1995).

18. Klein-Szanto,A.J. Stereologic baseline data of normal human epidermis. J. Invest Dermatol. 68, 73-78 (1977).

19. Lavker,R.M. & Sun,T.T. Epidermal stem cells. J. Invest Dermatol. 81, 121s-127s (1983).

20. Lajtha,L.G. Stem cell concepts. Differentiation 14, 23-34 (1979).

21. Lyle,S. et al. The C8/144B monoclonal antibody recognizes cytokeratin 15 and defines the location of human hair follicle stem cells. J. Cell Sci. 111 ( Pt 21), 3179-3188 (1998).

22. Michel,M. et al. Keratin 19 as a biochemical marker of skin stem cells in vivo and in vitro: keratin 19 expressing cells are differentially localized in function of anatomic sites, and their number varies with donor

249 age and culture stage. J. Cell Sci. 109 ( Pt 5), 1017-1028 (1996).

23. Fuchs,E. & Green,H. Changes in keratin gene expression during terminal differentiation of the keratinocyte. Cell 19, 1033-1042 (1980).

24. Woodcock-Mitchell,J., Eichner,R., Nelson,W.G., & Sun,T.T. Immunolocalization of keratin polypeptides in human epidermis using monoclonal antibodies. J. Cell Biol. 95, 580-588 (1982).

25. Selby,C.C. An electron microscope study of the epidermis of mammalian skin in thin sections. I. Dermo-epidermal junction and basal cell layer. J. Biophys. Biochem. Cytol. 1, 429-444 (1955).

26. Odland,G.F. The fine structure of the interrelationship of cells in the human epidermis. J. Biophys. Biochem. Cytol. 4, 529-538 (1958).

27. Coulombe,P.A., Bousquet,O., Ma,L., Yamada,S., & Wirtz,D. The 'ins' and 'outs' of intermediate filament organization. Trends Cell Biol. 10, 420-428 (2000).

28. Kouklis,P.D., Hutton,E., & Fuchs,E. Making a connection: direct binding between keratin intermediate filaments and desmosomal proteins. J. Cell Biol. 127, 1049-1060 (1994).

29. Caputo,R. & Peluchetti,D. The junctions of normal human epidermis. A freeze-fracture study. J. Ultrastruct. Res. 61, 44-61 (1977).

30. Garrod,D.R. Desmosomes and hemidesmosomes. Curr. Opin. Cell Biol. 5, 30-40 (1993).

31. Peltonen,J. et al. Localization of integrin receptors for fibronectin, collagen, and laminin in human skin. Variable expression in basal and squamous cell carcinomas. J. Clin. Invest 84, 1916-1923 (1989).

32. Carter,W.G., Wayner,E.A., Bouchard,T.S., & Kaur,P. The role of integrins alpha 2 beta 1 and alpha 3 beta 1 in cell-cell and cell-substrate adhesion of human epidermal cells. J. Cell Biol. 110, 1387-1404 (1990).

33. Hertle,M.D., Adams,J.C., & Watt,F.M. Integrin expression during

250 human epidermal development in vivo and in vitro. Development 112, 193-206 (1991).

34. Larjava,H. et al. Novel function for beta 1 integrins in keratinocyte cell-cell interactions. J. Cell Biol. 110, 803-815 (1990).

35. Adams,J.C. & Watt,F.M. Expression of beta 1, beta 3, beta 4, and beta 5 integrins by human epidermal keratinocytes and non-differentiating keratinocytes. J. Cell Biol. 115, 829-841 (1991).

36. Carter,W.G., Kaur,P., Gil,S.G., Gahr,P.J., & Wayner,E.A. Distinct functions for integrins alpha 3 beta 1 in focal adhesions and alpha 6 beta 4/bullous pemphigoid antigen in a new stable anchoring contact (SAC) of keratinocytes: relation to hemidesmosomes. J. Cell Biol. 111, 3141-3154 (1990).

37. Carter,W.G., Ryan,M.C., & Gahr,P.J. Epiligrin, a new cell adhesion ligand for integrin alpha 3 beta 1 in epithelial basement membranes. Cell 65, 599-610 (1991).

38. Rousselle,P. & Aumailley,M. Kalinin is more efficient than laminin in promoting adhesion of primary keratinocytes and some other epithelial cells and has a different requirement for integrin receptors. J. Cell Biol. 125, 205-214 (1994).

39. Adams,J.C. & Watt,F.M. Changes in keratinocyte adhesion during terminal differentiation: reduction in fibronectin binding precedes alpha 5 beta 1 integrin loss from the cell surface. Cell 63, 425-435 (1990).

40. Green,H. Terminal differentiation of cultured human epidermal cells. Cell 11, 405-416 (1977).

41. Watt,F.M. & Green,H. Stratification and terminal differentiation of cultured epidermal cells. Nature 295, 434-436 (1982).

42. Watt,F.M. Selective migration of terminally differentiating cells from the basal layer of cultured human epidermis. J Cell Biol. 98, 16-21 (1984).

251 43. Boyce,S.T. & Ham,R.G. Calcium-regulated differentiation of normal human epidermal keratinocytes in chemically defined clonal culture and serum-free serial culture. J. Invest Dermatol. 81, 33s-40s (1983).

44. Juliano,R.L. & Haskill,S. Signal transduction from the extracellular matrix. J. Cell Biol. 120, 577-585 (1993).

45. Giancotti,F.G. Integrin signaling: specificity and control of cell survival and cell cycle progression. Curr. Opin. Cell Biol. 9, 691-700 (1997).

46. Zachary,I. & Rozengurt,E. Focal adhesion kinase (p125FAK): a point of convergence in the action of neuropeptides, integrins, and oncogenes. Cell 71, 891-894 (1992).

47. Wary,K.K., Mainiero,F., Isakoff,S.J., Marcantonio,E.E., & Giancotti,F.G. The adaptor protein Shc couples a class of integrins to the control of cell cycle progression. Cell 87, 733-743 (1996).

48. Adams,J.C. & Watt,F.M. Fibronectin inhibits the terminal differentiation of human keratinocytes. Nature 340, 307-309 (1989).

49. Watt,F.M., Kubler,M.D., Hotchin,N.A., Nicholson,L.J., & Adams,J.C. Regulation of keratinocyte terminal differentiation by integrin-extracellular matrix interactions. J. Cell Sci. 106 ( Pt 1), 175-182 (1993).

50. Watt,F.M. & Jones,P.H. Expression and function of the keratinocyte integrins. Dev. Suppl185-192 (1993).

51. Hennings,H. et al. Calcium regulation of growth and differentiation of mouse epidermal cells in culture. Cell 19, 245-254 (1980).

52. Watt,F.M., Mattey,D.L., & Garrod,D.R. Calcium-induced reorganization of desmosomal components in cultured human keratinocytes. J. Cell Biol. 99, 2211-2215 (1984).

53. Stern,I.B. & Sekeri-Pataryas,K.H. The uptake of 14C-leucine and 14C-histidine by cell suspensions of isolated strata of neonatal rat epidermis. J. Invest Dermatol. 59, 251-259 (1972).

252 54. Fukuyama,K. & Epstein,W.L. Epidermal keratinization: localization of isotopically labeled amino acids. J. Invest Dermatol. 47, 551-560 (1966).

55. Coulombe,P.A. The cellular and molecular biology of keratins: beginning a new era. Curr. Opin. Cell Biol. 5, 17-29 (1993).

56. Adam Friedman A review of skin structure. Skin and aging 18, (2010).

57. Ishida-Yamamoto,A. et al. Epidermal lamellar granules transport different cargoes as distinct aggregates. J. Invest Dermatol. 122, 1137-1144 (2004).

58. Elias,P.M. Epidermal lipids, barrier function, and desquamation. J. Invest Dermatol. 80 Suppl, 44s-49s (1983).

59. Steven,A.C., Bisher,M.E., Roop,D.R., & Steinert,P.M. Biosynthetic pathways of filaggrin and loricrin--two major proteins expressed by terminally differentiated epidermal keratinocytes. J. Struct. Biol. 104, 150-162 (1990).

60. Matoltsy,A.G. & Matoltsy,M. A study of morphological and chemical properties of keratohyalin granules. J. Invest Dermatol. 38, 237-247 (1962).

61. Dale,B.A., Resing,K.A., & Presland,R.B. Keratohyalin granule proteins in The Keratinocyte Handbook (eds. Leigh,I., Lane,B. & Watt,F.) 323-350 (Cambridge University Press, Cambridge, 1994).

62. Fleckman,P., Dale,B.A., & Holbrook,K.A. Profilaggrin, a high-molecular-weight precursor of filaggrin in human epidermis and cultured keratinocytes. J. Invest Dermatol. 85, 507-512 (1985).

63. Lauren E.Flynn & Kimberly A.Woodhouse Burn Dressing Biomaterials and Tissue Engineering in Biomedical Materials (ed. Roger Narayan) 371-405 (Springer, 2009).

64. McGrath,J.A., Eady,R.A., & Pope,F.M. Rook's Textbook of

253 Dermatology(Blackwell Publishing,2004).

65. Tortora,G. & Derrickson,B. Principles of Anatomy and Physiology(John Wiley & Sons Inc., Hoboken, NJ, 2009).

66. Lavker,R.M. & Matoltsy,A.G. Formation of horny cells: the fate of cell organelles and differentiation products in ruminal epithelium. J. Cell Biol. 44, 501-512 (1970).

67. Dale,B.A., Scofield,J.A., Hennings,H., Stanley,J.R., & Yuspa,S.H. Identification of filaggrin in cultured mouse keratinocytes and its regulation by calcium. J. Invest Dermatol. 81, 90s-95s (1983).

68. Dale,B.A., Presland,R.B., Lewis,S.P., Underwood,R.A., & Fleckman,P. Transient expression of epidermal filaggrin in cultured cells causes collapse of intermediate filament networks with alteration of cell shape and nuclear integrity. J. Invest Dermatol. 108, 179-187 (1997).

69. Scott,I.R., Harding,C.R., & Barrett,J.G. Histidine-rich protein of the keratohyalin granules. Source of the free amino acids, urocanic acid and pyrrolidone carboxylic acid in the stratum corneum. Biochim. Biophys. Acta 719, 110-117 (1982).

70. Harding,C.R. & Scott,I.R. Histidine-rich proteins (filaggrins): structural and functional heterogeneity during epidermal differentiation. J. Mol. Biol. 170, 651-673 (1983).

71. Rawlings,A.V., Scott,I.R., Harding,C.R., & Bowser,P.A. Stratum corneum moisturization at the molecular level. J. Invest Dermatol. 103, 731-741 (1994).

72. Wolff-Schreiner,E.C. Ultrastructural cytochemistry of the epidermis. Int. J. Dermatol. 16, 77-102 (1977).

73. Matoltsy,A.G. & Parakkal,P.F. Membrane-coating granules of keratinizing epithelia. J. Cell Biol. 24, 297-307 (1965).

74. Lampe,M.A., Williams,M.L., & Elias,P.M. Human epidermal lipids: characterization and modulations during differentiation. J. Lipid Res. 24,

254 131-140 (1983).

75. Heather Brannon. Stratum Corneum Anatomy - The key to healthy, attractive skin. 2009. Ref Type: Online Source

76. Rice,R.H. & Green,H. Presence in human epidermal cells of a soluble protein precursor of the cross-linked envelope: activation of the cross-linking by calcium ions. Cell 18, 681-694 (1979).

77. Banks-Schlegel,S. & Green,H. Involucrin synthesis and tissue assembly by keratinocytes in natural and cultured human epithelia. J. Cell Biol. 90, 732-737 (1981).

78. Mehrel,T. et al. Identification of a major keratinocyte cell envelope protein, loricrin. Cell 61, 1103-1112 (1990).

79. Hohl,D. et al. Characterization of human loricrin. Structure and function of a new class of epidermal cell envelope proteins. J. Biol. Chem. 266, 6626-6636 (1991).

80. Hanigan,H. & Goldsmith,L.A. Endogenous substrates for epidermal transglutaminase. Biochim. Biophys. Acta 522, 589-601 (1978).

81. Simon,M. & Green,H. Enzymatic cross-linking of involucrin and other proteins by keratinocyte particulates in vitro. Cell 40, 677-683 (1985).

82. Buxman,M.M., Buehner,G.E., & Weupper,K.D. Isolation of substrates of epidermal transglutaminase from bovine epidermis. Biochem. Biophys. Res. Commun. 73, 470-478 (1976).

83. Thacher,S.M. & Rice,R.H. Keratinocyte-specific transglutaminase of cultured human epidermal cells: relation to cross-linked envelope formation and terminal differentiation. Cell 40, 685-695 (1985).

84. Peterson,L.L., Zettergren,J.G., & Wuepper,K.D. Biochemistry of transglutaminases and cross-linking in the skin. J. Invest Dermatol. 81, 95s-100s (1983).

255 85. Rothnagel,J.A. & Rogers,G.E. Transglutaminase-mediated cross-linking in mammalian epidermis. Mol. Cell Biochem. 58, 113-119 (1984).

86. Peterson,L.L. & Buxman,M.M. Rat hair follicle and epidermal transglutaminases. Biochemical and immunochemical isoenzymes. Biochim. Biophys. Acta 657, 268-276 (1981).

87. Goldsmith,L.A. et al. Vertebral epidermal transamidases. Biochim. Biophys. Acta 351, 113-125 (1974).

88. Rice,R.H. & Green,H. The cornified envelope of terminally differentiated human epidermal keratinocytes consists of cross-linked protein. Cell 11, 417-422 (1977).

89. Goldsmith,L.A. & Martin,C.M. Human epidermal transamidase. J. Invest Dermatol. 64, 316-321 (1975).

90. Ogawa,H. & Goldsmith,L.A. Human epidermal transglutaminase. II. Immunologic properties. J. Invest Dermatol. 68, 32-35 (1977).

91. Fuchs,E. Epidermal differentiation and keratin gene expression. J. Cell Sci. Suppl 17, 197-208 (1993).

92. Matoltsy,A.G. & Matoltsy,M. A study of morphological and chemical properties of keratohyalin granules. J. Invest Dermatol. 38, 237-247 (1962).

93. Lagermalm,G., Philip,B., & LindbergERG,J. Occurrence of thin membranes in the surface layers of human skin and in finger nails. Nature 168, 1080-1081 (1951).

94. William D.James, Timothy Berger, & Elston,D.in Andrews' Diseases of the skin: Clinical Dermatology 1-12 (Saunders, 2005).

95. James G.Marks Jr. & Jeffery J.Millerin Lookingbill and Marks' Principles of Dermatology 8-9 (Elsevier Inc, 2006).

96. Schon,M., Benwood,J., O'Connell-Willstaedt,T., & Rheinwald,J.G.

256 Human sweat gland myoepithelial cells express a unique set of cytokeratins and reveal the potential for alternative epithelial and mesenchymal differentiation states in culture. J. Cell Sci. 112 ( Pt 12), 1925-1936 (1999).

97. Bedell,M.A., Jenkins,N.A., & Copeland,N.G. Mouse models of human disease. Part I: techniques and resources for genetic analysis in mice. Genes Dev. 11, 1-10 (1997).

98. Bedell,M.A., Largaespada,D.A., Jenkins,N.A., & Copeland,N.G. Mouse models of human disease. Part II: recent progress and future directions. Genes Dev. 11, 11-43 (1997).

99. Scheid,A., Meuli,M., Gassmann,M., & Wenger,R.H. Genetically modified mouse models in studies on cutaneous wound healing. Exp. Physiol 85, 687-704 (2000).

100. Sundberg,J.P. & King,L.E., Jr. Mouse mutations as animal models and biomedical tools for dermatological research. J. Invest Dermatol. 106, 368-376 (1996).

101. Krawczyk,W.S. A pattern of epidermal cell migration during wound healing. J. Cell Biol. 49, 247-263 (1971).

102. Croft,C.B. & Tarin,D. Ultrastructural studies of wound healing in mouse skin. I. Epithelial behaviour. J. Anat. 106, 63-77 (1970).

103. Porter,R.M. Mouse models for human hair loss disorders. J. Anat. 202, 125-131 (2003).

104. Wong,V.W., Sorkin,M., Glotzbach,J.P., Longaker,M.T., & Gurtner,G.C. Surgical approaches to create murine models of human wound healing. J. Biomed. Biotechnol. 2011, 969618 (2011).

105. P.Sundberg Skin and adnexa of the laboratory mouse in The Laboratory Mouse (eds. Hans J Hedrich & Gillian Bullock) 195-206 (Elsevier Academic Press, San Diego, 2004).

106. Midwood,K.S., Williams,L.V., & Schwarzbauer,J.E. Tissue repair

257 and the dynamics of the extracellular matrix. Int. J. Biochem. Cell Biol. 36, 1031-1037 (2004).

107. Elias,P.M. Skin barrier function. Curr. Allergy Asthma Rep. 8, 299-305 (2008).

108. Elias,P.M. Defensive Functions of the Stratum Corneum: Integrative Aspects in Skin Barrier (eds. Elias,P.M. & Feingold,K.R.) 5-14 (Taylor & Francis Group, New York, 2006).

109. Meyer-Hoffert,U. Reddish, scaly, and itchy: how proteases and their inhibitors contribute to inflammatory skin diseases. Arch. Immunol. Ther. Exp. (Warsz. ) 57, 345-354 (2009).

110. Zeeuwen,P.L. Epidermal differentiation: the role of proteases and their inhibitors. Eur. J. Cell Biol. 83, 761-773 (2004).

111. Hart,T.C. et al. Mutations of the cathepsin C gene are responsible for Papillon-Lefevre syndrome. J. Med. Genet. 36, 881-887 (1999).

112. Toomes,C. et al. Loss-of-function mutations in the cathepsin C gene result in periodontal disease and palmoplantar keratosis. Nat. Genet. 23, 421-424 (1999).

113. Egberts,F. et al. Cathepsin D is involved in the regulation of transglutaminase 1 and epidermal differentiation. J. Cell Sci. 117, 2295-2307 (2004).

114. Roth,W. et al. Cathepsin L deficiency as molecular defect of furless: hyperproliferation of keratinocytes and pertubation of hair follicle cycling. FASEB J. 14, 2075-2086 (2000).

115. List,K. et al. Matriptase/MT-SP1 is required for postnatal survival, epidermal barrier function, hair follicle development, and thymic homeostasis. Oncogene 21, 3765-3779 (2002).

116. Leyvraz,C. et al. The epidermal barrier function is dependent on the serine protease CAP1/Prss8. J Cell Biol. 170, 487-496 (2005).

258 117. List,K. et al. Loss of proteolytically processed filaggrin caused by epidermal deletion of Matriptase/MT-SP1. J Cell Biol 163, 901-910 (2003).

118. Demerjian,M. et al. Acute modulations in permeability barrier function regulate epidermal cornification: role of caspase-14 and the protease-activated receptor type 2. Am. J. Pathol. 172, 86-97 (2008).

119. Denecker,G. et al. Caspase-14 protects against epidermal UVB photodamage and water loss. Nat. Cell Biol. 9, 666-674 (2007).

120. Basel-Vanagaite,L. et al. Autosomal recessive ichthyosis with hypotrichosis caused by a mutation in ST14, encoding type II transmembrane serine protease matriptase. Am. J. Hum. Genet. 80, 467-477 (2007).

121. Avrahami,L. et al. Autosomal recessive ichthyosis with hypotrichosis syndrome: further delineation of the phenotype. Clin. Genet. 74, 47-53 (2008).

122. Alef,T. et al. Ichthyosis, follicular atrophoderma, and hypotrichosis caused by mutations in ST14 is associated with impaired profilaggrin processing. J. Invest Dermatol. 129, 862-869 (2009).

123. Chavanas,S. et al. Mutations in SPINK5, encoding a serine protease inhibitor, cause Netherton syndrome. Nat. Genet. 25, 141-142 (2000).

124. Nagaike,K. et al. Defect of Hepatocyte Growth Factor Activator Inhibitor Type 1/Serine Protease Inhibitor, Kunitz Type 1 (Hai-1/Spint1) Leads to Ichthyosis-Like Condition and Abnormal Hair Development in Mice. Am. J. Pathol. 173, 1464-1475 (2008).

125. Zeeuwen,P.L., van Vlijmen-Willems,I.M., Hendriks,W., Merkx,G.F., & Schalkwijk,J. A null mutation in the cystatin M/E gene of ichq mice causes juvenile lethality and defects in epidermal cornification. Hum. Mol. Genet. 11, 2867-2875 (2002).

126. Zeeuwen,P.L. et al. Evidence that unrestricted legumain activity is 259 involved in disturbed epidermal cornification in cystatin M/E deficient mice. Hum. Mol. Genet. 13, 1069-1079 (2004).

127. Rawlings,N.D., Tolle,D.P., & Barrett,A.J. Evolutionary families of peptidase inhibitors. Biochem. J. 378, 705-716 (2004).

128. Fischer E. Einfluss der Configuration auf die Wirkung der Enzyme. Ber. Dt. Chem. Ges. 27, 2985-2993 (1894).

129. Koshland,D.E. Application of a Theory of Enzyme Specificity to Protein Synthesis. Proc. Natl. Acad. Sci. U. S. A 44, 98-104 (1958).

130. Egelund,R. et al. An ester bond linking a fragment of a serine proteinase to its serpin inhibitor. Biochemistry 37, 6375-6379 (1998).

131. Bugge,T.H., Antalis,T.M., & Wu,Q. Type II transmembrane serine proteases. J. Biol. Chem. 284, 23177-23181 (2009).

132. Hooper,J.D., Clements,J.A., Quigley,J.P., & Antalis,T.M. Type II transmembrane serine proteases. Insights into an emerging class of cell surface proteolytic enzymes. J Biol. Chem. 276, 857-860 (2001).

133. Laskowski,M., Jr. & Kato,I. Protein inhibitors of proteinases. Annu. Rev. Biochem. 49, 593-626 (1980).

134. Fritz,H. & Wunderer,G. Biochemistry and applications of aprotinin, the kallikrein inhibitor from bovine organs. Arzneimittelforschung. 33, 479-494 (1983).

135. Loebermann,H., Tokuoka,R., Deisenhofer,J., & Huber,R. Human alpha 1-proteinase inhibitor. Crystal structure analysis of two crystal modifications, molecular model and preliminary analysis of the implications for function. J. Mol. Biol. 177, 531-557 (1984).

136. Whisstock,J.C. & Bottomley,S.P. Molecular gymnastics: serpin structure, folding and misfolding. Curr. Opin. Struct. Biol. 16, 761-768 (2006).

137. Gettins,P.G. Serpin structure, mechanism, and function. Chem. Rev.

260 102, 4751-4804 (2002).

138. Whisstock,J.C., Skinner,R., Carrell,R.W., & Lesk,A.M. Conformational changes in serpins: I. The native and cleaved conformations of alpha(1)-antitrypsin. J. Mol. Biol. 296, 685-699 (2000).

139. Huntington,J.A., Read,R.J., & Carrell,R.W. Structure of a serpin-protease complex shows inhibition by deformation. Nature 407, 923-926 (2000).

140. Yamasaki,M., Li,W., Johnson,D.J., & Huntington,J.A. Crystal structure of a stable dimer reveals the molecular basis of serpin polymerization. Nature 455, 1255-1258 (2008).

141. Carrell,R.W. & Lomas,D.A. Conformational disease. Lancet 350, 134-138 (1997).

142. Lomas,D.A., Evans,D.L., Finch,J.T., & Carrell,R.W. The mechanism of Z alpha 1-antitrypsin accumulation in the liver. Nature 357, 605-607 (1992).

143. Bruce,D., Perry,D.J., Borg,J.Y., Carrell,R.W., & Wardell,M.R. Thromboembolic disease due to thermolabile conformational changes of antithrombin Rouen-VI (187 Asn-->Asp). J. Clin. Invest 94, 2265-2274 (1994).

144. List,K. et al. Deregulated matriptase causes ras-independent multistage carcinogenesis and promotes ras-mediated malignant transformation. Genes Dev. 19, 1934-1950 (2005).

145. Carney,T.J. et al. Inactivation of serine protease Matriptase1a by its inhibitor Hai1 is required for epithelial integrity of the zebrafish epidermis. Development 134, 3461-3471 (2007).

146. Mathias,J.R. et al. Live imaging of chronic inflammation caused by mutation of zebrafish Hai1. J. Cell Sci. 120, 3372-3383 (2007).

147. Takeuchi,T., Shuman,M.A., & Craik,C.S. Reverse biochemistry: use of macromolecular protease inhibitors to dissect complex biological

261 processes and identify a membrane-type serine protease in epithelial cancer and normal tissue. Proc. Natl. Acad. Sci. U. S. A 96, 11054-11061 (1999).

148. Tanimoto,H. et al. Ovarian tumor cells express a transmembrane serine protease: a potential candidate for early diagnosis and therapeutic intervention. Tumour. Biol. 22, 104-114 (2001).

149. Cao,J. et al. Characterization of colorectal-cancer-related cDNA clones obtained by subtractive hybridization screening. J Cancer Res. Clin Oncol 123, 447-451 (1997).

150. Shi,Y.E. et al. Identification and characterization of a novel matrix-degrading protease from hormone-dependent human breast cancer cells. Cancer Res. 53, 1409-1415 (1993).

151. Zhang,Y., Cai,X., Schlegelberger,B., & Zheng,S. Assignment1 of human putative tumor suppressor genes ST13 (alias SNC6) and ST14 (alias SNC19) to human chromosome bands 22q13 and 11q24-->q25 by in situ hybridization. Cytogenet. Cell Genet. 83, 56-57 (1998).

152. Lin,C.Y., Anders,J., Johnson,M., Sang,Q.A., & Dickson,R.B. Molecular cloning of cDNA for matriptase, a matrix-degrading serine protease with trypsin-like activity. J Biol. Chem. 274, 18231-18236 (1999).

153. Kim,M.G. et al. Cloning and chromosomal mapping of a gene isolated from thymic stromal cells encoding a new mouse type II membrane serine protease, epithin, containing four LDL receptor modules and two CUB domains. Immunogenetics 49, 420-428 (1999).

154. Satomi,S. et al. A role for membrane-type serine protease (MT-SP1) in intestinal epithelial turnover. Biochem. Biophys. Res. Commun. 287, 995-1002 (2001).

155. Lin,C.Y., Anders,J., Johnson,M., & Dickson,R.B. Purification and characterization of a complex containing matriptase and a Kunitz-type serine protease inhibitor from human milk. J Biol. Chem. 274,

262 18237-18242 (1999).

156. Oberst,M. et al. Matriptase and HAI-1 are expressed by normal and malignant epithelial cells in vitro and in vivo. Am J Pathol. 158, 1301-1311 (2001).

157. Oberst,M.D. et al. Characterization of matriptase expression in normal human tissues. J Histochem. Cytochem. 51, 1017-1025 (2003).

158. Szabo,R. et al. Type II transmembrane serine proteases. Thromb. Haemost. 90, 185-193 (2003).

159. List,K. et al. Epithelial Integrity Is Maintained by a Matriptase-Dependent Proteolytic Pathway. Am. J. Pathol. 175, 1453-1463 (2009).

160. Lin,C.Y. et al. Characterization of a novel, membrane-bound, 80-kDa matrix-degrading protease from human breast cancer cells. Monoclonal antibody production, isolation, and localization. J Biol. Chem. 272, 9147-9152 (1997).

161. Seitz,I. et al. Membrane-type serine protease-1/matriptase induces interleukin-6 and -8 in endothelial cells by activation of protease-activated receptor-2: potential implications in atherosclerosis. Arterioscler. Thromb. Vasc. Biol. 27, 769-775 (2007).

162. Kilpatrick,L.M. et al. Initiation of plasminogen activation on the surface of monocytes expressing the type II transmembrane serine protease matriptase. Blood 108, 2616-2623 (2006).

163. Tseng,I.C. et al. Purification from human milk of matriptase complexes with secreted serpins: mechanism for inhibition of matriptase other than HAI-1. Am. J. Physiol Cell Physiol 295, C423-C431 (2008).

164. Milner,J.M. et al. Matriptase is a novel initiator of cartilage matrix degradation in osteoarthritis. Arthritis Rheum. 62, 1955-1966 (2010).

165. Cheng,M.F. et al. Matriptase expression in the normal and neoplastic mast cells. Eur. J Dermatol. 17, 375-380 (2007).

263 166. Fang,J.D., Chou,H.C., Tung,H.H., Huang,P.Y., & Lee,S.L. Endogenous expression of matriptase in neural progenitor cells promotes cell migration and neuron differentiation. J Biol. Chem. 286, 5667-5679 (2010).

167. Kim,C. et al. Filamin is essential for shedding of the transmembrane serine protease, epithin. EMBO Rep. 6, 1045-1051 (2005).

168. Bork,P. & Patthy,L. The SEA module: a new extracellular domain associated with O-glycosylation. Protein Sci. 4, 1421-1425 (1995).

169. Levitin,F. et al. The MUC1 SEA module is a self-cleaving domain. J Biol Chem. 280, 33374-33386 (2005).

170. Cho,E.G. et al. N-terminal processing is essential for release of epithin, a mouse type II membrane serine protease. J Biol. Chem. 276, 44581-44589 (2001).

171. Oberst,M.D., Williams,C.A., Dickson,R.B., Johnson,M.D., & Lin,C.Y. The activation of matriptase requires its noncatalytic domains, serine protease domain, and its cognate inhibitor. J Biol. Chem. 278, 26773-26779 (2003).

172. Desilets,A. et al. Mutation G827R in matriptase causing autosomal recessive ichthyosis with hypotrichosis yields an inactive protease. J. Biol. Chem. 283, 10535-10542 (2008).

173. Friedrich,R. et al. Catalytic domain structures of MT-SP1/matriptase, a matrix-degrading transmembrane serine proteinase. J Biol. Chem. 277, 2160-2168 (2002).

174. Benaud,C., Dickson,R.B., & Lin,C.Y. Regulation of the activity of matriptase on epithelial cell surfaces by a blood-derived factor. Eur. J Biochem. 268, 1439-1447 (2001).

175. Lee,M.-S., Kiyomiya,K., Benaud,C., Dickson,R.B., & Lin,C.Y. Simultaneous activation and HAI-1-mediated inhibition of matriptase induced at activation foci in immortal human mammary epithelial cells. Am J Physiol Cell Physiol 288, C932-C941 (2005).

264 176. Inouye,K., Yasumoto,M., Tsuzuki,S., Mochida,S., & Fushiki,T. The Optimal Activity of a Pseudozymogen Form of Recombinant Matriptase under the Mildly Acidic pH and Low Ionic Strength Conditions. J. Biochem. 147, 485-492 (2009).

177. Tseng,I.C. et al. Matriptase activation, an early cellular response to acidosis. J. Biol. Chem. 285, 3261-3270 (2010).

178. Hla,T. Physiological and pathological actions of sphingosine 1-phosphate. Semin. Cell Dev. Biol. 15, 513-520 (2004).

179. Pyne,S. & Pyne,N.J. Sphingosine 1-phosphate signalling in mammalian cells. Biochem. J. 349, 385-402 (2000).

180. Spiegel,S. & Milstien,S. Exogenous and intracellularly generated sphingosine 1-phosphate can regulate cellular processes by divergent pathways. Biochem. Soc Trans. 31, 1216-1219 (2003).

181. Chun,J. et al. International Union of Pharmacology. XXXIV. Lysophospholipid receptor nomenclature. Pharmacol. Rev. 54, 265-269 (2002).

182. Benaud,C. et al. Sphingosine 1-phosphate, present in serum-derived lipoproteins, activates matriptase. J Biol. Chem. 277, 10539-10546 (2002).

183. Benaud,C.M., Oberst,M., Dickson,R.B., & Lin,C.Y. Deregulated activation of matriptase in breast cancer cells. Clin Exp. Metastasis 19, 639-649 (2002).

184. Kiyomiya,K.I. et al. Matriptase activation and subsequent shedding with HAI-1 is induced by steroid sex hormones in human prostate cancer cells, but not in breast cancer cells. Am J Physiol Cell Physiol. 291, C40-C49 (2006).

185. Yousef,G.M., Obiezu,C.V., Luo,L.Y., Black,M.H., & Diamandis,E.P. Prostase/KLK-L1 is a new member of the human kallikrein gene family, is expressed in prostate and breast tissues, and is hormonally regulated. Cancer Res 59, 4252-4256 (1999).

265 186. Wolf,D.A., Schulz,P., & Fittler,F. Transcriptional regulation of prostate kallikrein-like genes by androgen. Mol. Endocrinol. 6, 753-762 (1992).

187. Lin,B. et al. Prostate-localized and androgen-regulated expression of the membrane-bound serine protease TMPRSS2. Cancer Res 59, 4180-4184 (1999).

188. Afar,D.E. et al. Catalytic cleavage of the androgen-regulated TMPRSS2 protease results in its secretion by prostate and prostate cancer epithelia. Cancer Res 61, 1686-1692 (2001).

189. Liao,X. et al. Androgen stimulates matrix metalloproteinase-2 expression in human prostate cancer. Endocrinology 144, 1656-1663 (2003).

190. Nelson,P.S. et al. Molecular cloning and characterization of prostase, an androgen-regulated serine protease with prostate-restricted expression. Proc. Natl Acad. Sci U. S. A 96, 3114-3119 (1999).

191. Hung,R.J. et al. Assembly of adherens junctions is required for sphingosine 1-phosphate-induced matriptase accumulation and activation at mammary epithelial cell-cell contacts. Am J Physiol Cell Physiol 286, C1159-C1169 (2004).

192. Lee,M.S. et al. Autoactivation of matriptase in vitro: requirement for biomembrane and LDL receptor domain. Am. J. Physiol Cell Physiol 293, C95-C105 (2007).

193. Chen,Y.W. et al. Regulation of the matriptase-prostasin cell surface proteolytic cascade by hepatocyte growth factor activator inhibitor-1 (HAI-1) during epidermal differentiation. J Biol. Chem. 285, 31755-31762 (2010).

194. List,K., Hobson,J.P., Molinolo,A., & Bugge,T.H. Co-localization of the channel activating protease prostasin/(CAP1/PRSS8) with its candidate activator, matriptase. J. Cell Physiol 213, 237-245 (2007).

195. List,K. et al. Autosomal ichthyosis with hypotrichosis syndrome 266 displays low matriptase proteolytic activity and is phenocopied in ST14 hypomorphic mice. J. Biol. Chem. 282, 36714-36723 (2007).

196. Chen,C.J. et al. Increased matriptase zymogen activation in inflammatory skin disorders. Am. J Physiol Cell Physiol 300, C406-415 (2011).

197. Camerer,E. et al. Local protease signaling contributes to neural tube closure in the mouse embryo. Dev. Cell 18, 25-38 (2010).

198. Ihara,S. et al. Prometastatic effect of N-acetylglucosaminyltransferase V is due to modification and stabilization of active matriptase by adding beta 1-6 GlcNAc branching. J Biol. Chem. 277, 16960-16967 (2002).

199. Uhland,K. Matriptase and its putative role in cancer. Cell Mol. Life Sci. 63, 2968-2978 (2006).

200. List,K. Matriptase: a culprit in cancer? Future. Oncol. 5, 97-104 (2009).

201. Kang,J.Y. et al. Tissue microarray analysis of hepatocyte growth factor/Met pathway components reveals a role for Met, matriptase, and hepatocyte growth factor activator inhibitor 1 in the progression of node-negative breast cancer. Cancer Res. 63, 1101-1105 (2003).

202. Saleem,M. et al. A novel biomarker for staging human prostate adenocarcinoma: overexpression of matriptase with concomitant loss of its inhibitor, hepatocyte growth factor activator inhibitor-1. Cancer Epidemiol Biomarkers Prev. 15, 217-227 (2006).

203. Oberst,M.D. et al. Expression of the serine protease matriptase and its inhibitor HAI-1 in epithelial ovarian cancer: correlation with clinical outcome and tumor clinicopathological parameters. Clin Cancer Res. 8, 1101-1107 (2002).

204. Kauppinen,J.M. et al. ST14 gene variant and decreased matriptase protein expression predict poor breast cancer survival. Cancer Epidemiol. Biomarkers Prev. 19, 2133-2142 (2010).

267 205. Lyons,J.G. et al. Snail up-regulates proinflammatory mediators and inhibits differentiation in oral keratinocytes. Cancer Res. 68, 4525-4530 (2008).

206. Gemmill,R.M. et al. ZEB1-responsive genes in non-small cell lung cancer. Cancer Lett. 300, 66-78 (2011).

207. Takeuchi,T. et al. Cellular localization of membrane-type serine protease 1 and identification of protease-activated receptor-2 and single-chain urokinase-type plasminogen activator as substrates. J Biol. Chem. 275, 26333-26342 (2000).

208. Lee,S.L., Dickson,R.B., & Lin,C.Y. Activation of hepatocyte growth factor and urokinase/plasminogen activator by matriptase, an epithelial membrane serine protease. J Biol. Chem. 275, 36720-36725 (2000).

209. Jin,X. et al. Matriptase activates stromelysin (MMP-3) and promotes tumor growth and angiogenesis. Cancer Sci. 97, 1327-1334 (2006).

210. Netzel-Arnett,S. et al. Evidence for a matriptase-prostasin proteolytic cascade regulating terminal epidermal differentiation. J Biol Chem. 281, 32941-32945 (2006).

211. Bhatt,A.S. et al. Coordinate expression and functional profiling identify an extracellular proteolytic signaling pathway. Proc. Natl. Acad. Sci. U. S. A 104, 5771-5776 (2007).

212. Andreasen,D., Vuagniaux,G., Fowler-Jaeger,N., Hummler,E., & Rossier,B.C. Activation of epithelial sodium channels by mouse channel activating proteases (mCAP) expressed in Xenopus oocytes requires catalytic activity of mCAP3 and mCAP2 but not mCAP1. J. Am. Soc. Nephrol. 17, 968-976 (2006).

213. Clark,E.B., Jovov,B., Rooj,A.K., Fuller,C.M., & Benos,D.J. Proteolytic cleavage of human acid-sensing ion channel 1 by the serine protease matriptase. J Biol. Chem. 285, 27130-27143 (2010).

214. Ried,S. et al. Activation mechanisms of the urokinase-type plasminogen activator promoter by hepatocyte growth factor/scatter

268 factor. J. Biol. Chem. 274, 16377-16386 (1999).

215. Moriyama,T. et al. Simultaneous up-regulation of urokinase-type plasminogen activator (uPA) and uPA receptor by hepatocyte growth factor/scatter factor in human glioma cells. Clin. Exp. Metastasis 17, 873-879 (1999).

216. Forbs,D. et al. In vitro inhibition of matriptase prevents invasive growth of cell lines of prostate and colon carcinoma. Int J Oncol 27, 1061-1070 (2005).

217. Szabo,R. et al. c-Met-induced epithelial carcinogenesis is initiated by the serine protease matriptase. Oncogene(2011).

218. Suzuki,M. et al. Inhibition of tumor invasion by genomic down-regulation of matriptase through suppression of activation of receptor-bound pro-urokinase. J Biol Chem 279, 14899-14908 (2004).

219. Kataoka,H. et al. Distribution of hepatocyte growth factor activator inhibitor type 1 (HAI-1) in human tissues. Cellular surface localization of HAI-1 in simple columnar epithelium and its modulated expression in injured and regenerative tissues. J Histochem. Cytochem. 47, 673-682 (1999).

220. D'Andrea,M.R. et al. Characterization of protease-activated receptor-2 immunoreactivity in normal human tissues. J. Histochem. Cytochem. 46, 157-164 (1998).

221. Darmoul,D., Marie,J.C., Devaud,H., Gratio,V., & Laburthe,M. Initiation of human colon cancer cell proliferation by trypsin acting at protease-activated receptor-2. Br. J Cancer 85, 772-779 (2001).

222. Kawabata,A. et al. Activation of protease-activated receptor-2 (PAR-2) triggers mucin secretion in the rat sublingual gland. Biochem. Biophys. Res. Commun. 270, 298-302 (2000).

223. Kawabata,A., Nishikawa,H., Kuroda,R., Kawai,K., & Hollenberg,M.D. Proteinase-activated receptor-2 (PAR-2): regulation of salivary and pancreatic exocrine secretion in vivo in rats and mice. Br. J. Pharmacol.

269 129, 1808-1814 (2000).

224. Kong,W. et al. Luminal trypsin may regulate enterocytes through proteinase-activated receptor 2. Proc. Natl. Acad. Sci. U. S. A 94, 8884-8889 (1997).

225. Nguyen,T.D. et al. Trypsin activates pancreatic duct epithelial cell ion channels through proteinase-activated receptor-2. J Clin Invest 103, 261-269 (1999).

226. Bertog,M. et al. Basolateral proteinase-activated receptor (PAR-2) induces chloride secretion in M-1 mouse renal cortical collecting duct cells. J Physiol 521 Pt 1, 3-17 (1999).

227. Sharlow,E.R. et al. The protease-activated receptor-2 upregulates keratinocyte phagocytosis. J. Cell Sci. 113 ( Pt 17), 3093-3101 (2000).

228. Miyata,S., Koshikawa,N., Yasumitsu,H., & Miyazaki,K. Trypsin stimulates integrin alpha(5)beta(1)-dependent adhesion to fibronectin and proliferation of human gastric carcinoma cells through activation of proteinase-activated receptor-2. J Biol. Chem. 275, 4592-4598 (2000).

229. Kawabata,A. et al. The protease-activated receptor-2 agonist induces gastric mucus secretion and mucosal cytoprotection. J. Clin. Invest 107, 1443-1450 (2001).

230. Kawabata,A. et al. In vivo evidence that protease-activated receptors 1 and 2 modulate gastrointestinal transit in the mouse. Br. J. Pharmacol. 133, 1213-1218 (2001).

231. Szabo,R. et al. Potent inhibition and global co-localization implicate the transmembrane kunitz-type serine protease inhibitor hai-2 in the regulation of epithelial matriptase activity. J. Biol. Chem. 283, 29495-504 (2008).

232. Shimomura,T. et al. Hepatocyte growth factor activator inhibitor, a novel Kunitz-type serine protease inhibitor. J Biol. Chem. 272, 6370-6376 (1997).

270 233. Itoh,H. et al. Genomic structure and chromosomal localization of the human hepatocyte growth factor activator inhibitor type 1 and 2 genes. Eur. J. Biochem. 267, 3351-3359 (2000).

234. Guo,J. et al. MANSC: a seven-cysteine-containing domain present in animal membrane and extracellular proteins. Trends Biochem. Sci. 29, 172-174 (2004).

235. Shimomura,T. et al. Multiple sites of proteolytic cleavage to release soluble forms of hepatocyte growth factor activator inhibitor type 1 from a transmembrane form. J Biochem. (Tokyo) 126, 821-828 (1999).

236. Szabo,R. et al. Regulation of cell surface protease matriptase by HAI2 is essential for placental development, neural tube closure and embryonic survival in mice. Development 136, 2653-2663 (2009).

237. Collen,D., Schetz,J., de,C.F., Holmer,E., & Verstraete,M. Metabolism of antithrombin III (heparin cofactor) in man: effects of venous thrombosis and of heparin administration. Eur. J. Clin. Invest 7, 27-35 (1977).

238. Conard,J., Brosstad,F., Lie,L.M., Samama,M., & Abildgaard,U. Molar antithrombin concentration in normal human plasma. Haemostasis 13, 363-368 (1983).

239. Bjork,I. & Olson,S.T. Antithrombin. A bloody important serpin. Adv. Exp. Med. Biol. 425, 17-33 (1997).

240. Brennan,S.O., George,P.M., & Jordan,R.E. Physiological variant of antithrombin-III lacks carbohydrate sidechain at Asn 135. FEBS Lett. 219, 431-436 (1987).

241. Olson,S.T. et al. Role of the antithrombin-binding pentasaccharide in heparin acceleration of antithrombin-proteinase reactions. Resolution of the antithrombin conformational change contribution to heparin rate enhancement. J. Biol. Chem. 267, 12528-12538 (1992).

242. Olson,S.T. & Bjork,I. Predominant contribution of surface approximation to the mechanism of heparin acceleration of the

271 antithrombin-thrombin reaction. Elucidation from salt concentration effects. J. Biol. Chem. 266, 6353-6364 (1991).

243. Jordan,R.E., Oosta,G.M., Gardner,W.T., & Rosenberg,R.D. The kinetics of hemostatic enzyme-antithrombin interactions in the presence of low molecular weight heparin. J. Biol. Chem. 255, 10081-10090 (1980).

244. Griffith,M.J. Kinetics of the heparin-enhanced antithrombin III/thrombin reaction. Evidence for a template model for the mechanism of action of heparin. J. Biol. Chem. 257, 7360-7365 (1982).

245. Bedsted,T. et al. Heparin and calcium ions dramatically enhance antithrombin reactivity with factor IXa by generating new interaction exosites. Biochemistry 42, 8143-8152 (2003).

246. Langdown,J., Johnson,D.J., Baglin,T.P., & Huntington,J.A. Allosteric activation of antithrombin critically depends upon hinge region extension. J. Biol. Chem. 279, 47288-47297 (2004).

247. Johnson,D.J. et al. Crystal structure of monomeric native antithrombin reveals a novel reactive center loop conformation. J. Biol. Chem. 281, 35478-35486 (2006).

248. Carrell,R.W., Stein,P.E., Fermi,G., & Wardell,M.R. Biological implications of a 3 A structure of dimeric antithrombin. Structure. 2, 257-270 (1994).

249. Schreuder,H.A. et al. The intact and cleaved human antithrombin III complex as a model for serpin-proteinase interactions. Nat. Struct. Biol. 1, 48-54 (1994).

250. Whisstock,J.C. et al. Conformational changes in serpins: II. The mechanism of activation of antithrombin by heparindagger. J. Mol. Biol. 301, 1287-1305 (2000).

251. Petitou,M. et al. Synthesis of thrombin-inhibiting heparin mimetics without side effects. Nature 398, 417-422 (1999).

272 252. Yamasaki,Y., Satomi,S., Murai,N., Tsuzuki,S., & Fushiki,T. Inhibition of membrane-type serine protease 1/matriptase by natural and synthetic protease inhibitors. J Nutr. Sci. Vitaminol. (Tokyo) 49, 27-32 (2003).

253. Long,Y.Q. et al. Synthesis and evaluation of the sunflower derived trypsin inhibitor as a potent inhibitor of the type II transmembrane serine protease, matriptase. Bioorg. Med Chem. Lett. 11, 2515-2519 (2001).

254. Galkin,A.V. et al. CVS-3983, a selective matriptase inhibitor, suppresses the growth of androgen independent prostate tumor xenografts. Prostate 61, 228-235 (2004).

255. Steinmetzer,T. et al. Secondary amides of sulfonylated 3-amidinophenylalanine. New potent and selective inhibitors of matriptase. J Med. Chem. 49, 4116-4126 (2006).

256. Desilets,A., Longpre,J.M., Beaulieu,M.E., & Leduc,R. Inhibition of human matriptase by eglin c variants. FEBS Lett. 580, 2227-2232 (2006).

257. Enyedy,I.J. et al. Structure-based approach for the discovery of bis-benzamidines as novel inhibitors of matriptase. J Med Chem. 44, 1349-1355 (2001).

258. Farady,C.J., Sun,J., Darragh,M.R., Miller,S.M., & Craik,C.S. The mechanism of inhibition of antibody-based inhibitors of membrane-type serine protease 1 (MT-SP1). J Mol. Biol. 369, 1041-1051 (2007).

259. Oberst,M.D. et al. Hepatocyte growth factor activator inhibitor 1 (HAI-1) regulates activation and expression of matriptase, a membrane-bound serine protease. Am J Physiol Cell Physiol 289, C462-C470 (2005).

260. Kirchhofer,D. et al. Tissue expression, protease specificity, and Kunitz domain functions of hepatocyte growth factor activator inhibitor-1B (HAI-1B), a new splice variant of HAI-1. J Biol Chem 278, 36341-36349 (2003).

261. Xu,H. et al. Mechanisms for the control of matriptase activity in the absence of sufficient HAI-1. Am. J. Physiol Cell Physiol 302, C453-C462

273 (2012).

262. Chen,M., Fu,Y.Y., Lin,C.Y., Chen,L.M., & Chai,K.X. Prostasin induces protease-dependent and independent molecular changes in the human prostate carcinoma cell line PC-3. Biochim. Biophys. Acta 1773, 1133-1140 (2007).

263. Chen,M., Chen,L.M., Lin,C.Y., & Chai,K.X. The epidermal growth factor receptor (EGFR) is proteolytically modified by the Matriptase-Prostasin serine protease cascade in cultured epithelial cells. Biochim. Biophys. Acta(2007).

264. Stark,H.J., Baur,M., Breitkreutz,D., Mirancea,N., & Fusenig,N.E. Organotypic keratinocyte cocultures in defined medium with regular epidermal morphogenesis and differentiation. J. Invest Dermatol. 112, 681-691 (1999).

265. Stark,H.J., Szabowski,A., Fusenig,N.E., & Maas-Szabowski,N. Organotypic cocultures as skin equivalents: A complex and sophisticated in vitro system. Biol. Proced. Online. 6, 55-60 (2004).

266. Paladino,S., Pocard,T., Catino,M.A., & Zurzolo,C. GPI-anchored proteins are directly targeted to the apical surface in fully polarized MDCK cells. J. Cell Biol. 172, 1023-1034 (2006).

267. Yu,J.X., Chao,L., & Chao,J. Prostasin is a novel human serine proteinase from seminal fluid. Purification, tissue distribution, and localization in prostate gland. J. Biol. Chem. 269, 18843-18848 (1994).

268. Myerburg,M.M., Harvey,P.R., Heidrich,E.M., Pilewski,J.M., & Butterworth,M. Acute Regulation of ENaC in Airway Epithelia by Proteases and Trafficking. Am. J. Respir. Cell Mol. Biol.(2010).

269. Wang,J.K. et al. Polarized epithelial cells secrete matriptase as a consequence of zymogen activation and HAI-1-mediated inhibition. Am. J. Physiol Cell Physiol 297, C459-C470 (2009).

270. Ovaere,P., Lippens,S., Vandenabeele,P., & Declercq,W. The emerging roles of serine protease cascades in the epidermis. Trends

274 Biochem. Sci.(2009).

271. Fan,B., Wu,T.D., Li,W., & Kirchhofer,D. Identification of hepatocyte growth factor activator inhibitor-1B as a potential physiological inhibitor of prostasin. J Biol Chem 280, 34513-34520 (2005).

272. Kataoka,H. et al. Hepatocyte growth factor activator inhibitor type 1 is a specific cell surface binding protein of hepatocyte growth factor activator (HGFA) and regulates HGFA activity in the pericellular microenvironment. J Biol. Chem. 275, 40453-40462 (2000).

273. Chou,F.P. et al. Matriptase is inhibited by extravascular antithrombin in epithelial cells but not in most carcinoma cells. Am. J Physiol Cell Physiol(2011).

274. Ding,K., Lopez-Burks,M., Sanchez-Duran,J.A., Korc,M., & Lander,A.D. Growth factor-induced shedding of syndecan-1 confers glypican-1 dependence on mitogenic responses of cancer cells. J. Cell Biol. 171, 729-738 (2005).

275. Subramanian,S.V., Fitzgerald,M.L., & Bernfield,M. Regulated shedding of syndecan-1 and -4 ectodomains by thrombin and growth factor receptor activation. J. Biol. Chem. 272, 14713-14720 (1997).

276. Akiyama,M. & Shimizu,H. An update on molecular aspects of the non-syndromic ichthyoses. Exp. Dermatol. 17, 373-382 (2008).

277. Deyrieux,A.F. & Wilson,V.G. In vitro culture conditions to study keratinocyte differentiation using the HaCaT cell line. Cytotechnology 54, 77-83 (2007).

278. Micallef,L. et al. Effects of extracellular calcium on the growth-differentiation switch in immortalized keratinocyte HaCaT cells compared with normal human keratinocytes. Exp. Dermatol. 18, 143-151 (2009).

279. Chen,Y.W. et al. TMPRSS2, a Serine Protease Expressed in the Prostate on the Apical Surface of Luminal Epithelial Cells and Released into Semen in Prostasomes, Is Misregulated in Prostate Cancer Cells.

275 Am. J. Pathol.(2010).

280. Staniszewska,M., Sluczanowska-Glabowska,S., & Drukala,J. Stem cells and skin regeneration. Folia Histochem. Cytobiol. 49, 375-380 (2011).

281. Rifkin,D.B. et al. Plasminogen/plasminogen activator and growth factor activation. Ciba Found. Symp. 212, 105-115 (1997).

282. Koolwijk,P. et al. Cooperative effect of TNFalpha, bFGF, and VEGF on the formation of tubular structures of human microvascular endothelial cells in a fibrin matrix. Role of urokinase activity. J. Cell Biol. 132, 1177-1188 (1996).

283. Chapman,H.A. Plasminogen activators, integrins, and the coordinated regulation of cell adhesion and migration. Curr. Opin. Cell Biol. 9, 714-724 (1997).

284. Singer,A.J. & Clark,R.A. Cutaneous wound healing. N. Engl. J. Med. 341, 738-746 (1999).

285. Li,F. et al. Targeted inhibition of wound-induced PAI-1 expression alters migration and differentiation in human epidermal keratinocytes. Exp. Cell Res. 258, 245-253 (2000).

286. Romer,J. et al. The receptor for urokinase-type plasminogen activator is expressed by keratinocytes at the leading edge during re-epithelialization of mouse skin wounds. J. Invest Dermatol. 102, 519-522 (1994).

287. Kramer,M.D., Schaefer,B., & Reinartz,J. Plasminogen activation by human keratinocytes: molecular pathways and cell-biological consequences. Biol. Chem. Hoppe Seyler 376, 131-141 (1995).

288. Romer,J. et al. Differential expression of urokinase-type plasminogen activator and its type-1 inhibitor during healing of mouse skin wounds. J. Invest Dermatol. 97, 803-811 (1991).

289. Gissler,H.M., Frank,R., & Kramer,M.D. Immunohistochemical

276 characterization of the plasminogen activator system in psoriatic epidermis. Br. J. Dermatol. 128, 612-618 (1993).

290. Justus,C., Muller,S., & Kramer,M.D. A monoclonal antibody recognizing plasminogen and plasmin--altered reactivity in psoriatic lesions. Br. J. Dermatol. 117, 687-694 (1987).

291. Ellis,V. & Dano,K. Potentiation of plasminogen activation by an anti-urokinase monoclonal antibody due to ternary complex formation. A mechanistic model for receptor-mediated plasminogen activation. J. Biol. Chem. 268, 4806-4813 (1993).

292. Xu,Z. et al. Targeting zymogen activation to control the matriptase-prostasin proteolytic cascade. Submitted(2011).

293. Wiman,B., Lijnen,H.R., & Collen,D. On the specific interaction between the lysine-binding sites in plasmin and complementary sites in alpha2-antiplasmin and in fibrinogen. Biochim. Biophys. Acta 579, 142-154 (1979).

294. Varadi,A. & Patthy,L. Kringle 5 of human plasminogen carries a benzamidine-binding site. Biochem. Biophys. Res. Commun. 103, 97-102 (1981).

295. Burge,S.M., Marshall,J.M., & Cederholm-Williams,S.A. Plasminogen binding sites in normal human skin. Br. J Dermatol. 126, 35-41 (1992).

296. Menon,G.K., Grayson,S., & Elias,P.M. Ionic calcium reservoirs in mammalian epidermis: ultrastructural localization by ion-capture cytochemistry. J. Invest Dermatol. 84, 508-512 (1985).

297. Isseroff,R.R. & Rifkin,D.B. Plasminogen is present in the basal layer of the epidermis. J Invest Dermatol. 80, 297-299 (1983).

298. Hachem,J.P. et al. Serine protease signaling of epidermal permeability barrier homeostasis. J. Invest Dermatol. 126, 2074-2086 (2006).

277 299. Carlson,T.H., Atencio,A.C., & Simon,T.L. In vivo behavior of radioiodinated rabbit antithrombin III. Demonstration of a noncirculating vascular compartment. J Clin. Invest 74, 191-199 (1984).

300. Carlson,T.H., Simon,T.L., & Atencio,A.C. In vivo behavior of human radioiodinated antithrombin III: distribution among three physiologic pools. Blood 66, 13-19 (1985).

301. Xu,Y. & Slayter,H.S. Immunocytochemical localization of endogenous anti-thrombin III in the vasculature of rat tissues reveals locations of anticoagulantly active heparan sulfate proteoglycans. J. Histochem. Cytochem. 42, 1365-1376 (1994).

302. Paoloni-Giacobino,A., Chen,H., Peitsch,M.C., Rossier,C., & Antonarakis,S.E. Cloning of the TMPRSS2 gene, which encodes a novel serine protease with transmembrane, LDLRA, and SRCR domains and maps to 21q22.3. Genomics 44, 309-320 (1997).

303. Jacquinet,E. et al. Cloning and characterization of the cDNA and gene for human epitheliasin. Eur. J. Biochem. 268, 2687-2699 (2001).

304. Netzel-Arnett,S. et al. Membrane anchored serine proteases: a rapidly expanding group of cell surface proteolytic enzymes with potential roles in cancer. Cancer Metastasis Rev. 22, 237-258 (2003).

305. Kim,T.S., Heinlein,C., Hackman,R.C., & Nelson,P.S. Phenotypic analysis of mice lacking the Tmprss2-encoded protease. Mol. Cell Biol. 26, 965-975 (2006).

306. Donaldson,S.H. et al. Regulation of the epithelial sodium channel by serine proteases in human airways. J. Biol. Chem. 277, 8338-8345 (2002).

307. Aimes,R.T. et al. Endothelial cell serine proteases expressed during vascular morphogenesis and angiogenesis. Thromb. Haemost. 89, 561-572 (2003).

308. Vaarala,M.H., Porvari,K., Kyllonen,A., Lukkarinen,O., & Vihko,P. The TMPRSS2 gene encoding transmembrane serine protease is

278 overexpressed in a majority of prostate cancer patients: detection of mutated TMPRSS2 form in a case of aggressive disease. Int. J. Cancer 94, 705-710 (2001).

309. Vaarala,M.H., Porvari,K.S., Kellokumpu,S., Kyllonen,A.P., & Vihko,P.T. Expression of transmembrane serine protease TMPRSS2 in mouse and human tissues. J. Pathol. 193, 134-140 (2001).

310. Hermans,K.G. et al. TMPRSS2:ERG fusion by translocation or interstitial deletion is highly relevant in androgen-dependent prostate cancer, but is bypassed in late-stage androgen receptor-negative prostate cancer. Cancer Res. 66, 10658-10663 (2006).

311. Tomlins,S.A. et al. Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310, 644-648 (2005).

312. Mwamukonda,K. et al. Quantitative expression of TMPRSS2 transcript in prostate tumor cells reflects TMPRSS2-ERG fusion status. Prostate Cancer Prostatic. Dis.(2009).

313. Lucas,J.M. et al. The androgen-regulated type II serine protease TMPRSS2 is differentially expressed and mislocalized in prostate adenocarcinoma. J. Pathol. 215, 118-125 (2008).

314. Wilson,S. et al. The membrane-anchored serine protease, TMPRSS2, activates PAR-2 in prostate cancer cells. Biochem. J. 388, 967-972 (2005).

315. Warren,M. et al. Protein Expression of Matriptase and its Cognate Inhibitor HAI-1 in Human Prostate Cancer: A Tissue Microarray and Automated Quantitative Analysis. Appl. Immunohistochem. Mol. Morphol. 17, 23-30 (2008).

316. Rahrmann,E.P. et al. Identification of PDE4D as a proliferation promoting factor in prostate cancer using a Sleeping Beauty transposon-based somatic mutagenesis screen. Cancer Res. 69, 4388-4397 (2009).

279 317. Arienti,G., Carlini,E., De Cosmo,A.M., Di,P.P., & Palmerini,C.A. Prostasome-like particles in stallion semen. Biol. Reprod. 59, 309-313 (1998).

318. Carlsson,L. et al. Characteristics of human prostasomes isolated from three different sources. Prostate 54, 322-330 (2003).

319. Nilsson,B.O., Jin,M., Einarsson,B., Persson,B.E., & Ronquist,G. Monoclonal antibodies against human prostasomes. Prostate 35, 178-184 (1998).

320. Arienti,G., Carlini,E., & Palmerini,C.A. Fusion of human sperm to prostasomes at acidic pH. J. Membr. Biol. 155, 89-94 (1997).

321. Arienti,G. et al. The motility of human spermatozoa as influenced by prostasomes at various pH levels. Biol. Cell 91, 51-54 (1999).

322. Arienti,G., Carlini,E., Polci,A., Cosmi,E.V., & Palmerini,C.A. Fatty acid pattern of human prostasome lipid. Arch. Biochem. Biophys. 358, 391-395 (1998).

323. Carlini,E., Palmerini,C.A., Cosmi,E.V., & Arienti,G. Fusion of sperm with prostasomes: effects on membrane fluidity. Arch. Biochem. Biophys. 343, 6-12 (1997).

324. Cross,N.L. Human seminal plasma prevents sperm from becoming acrosomally responsive to the agonist, progesterone: cholesterol is the major inhibitor. Biol. Reprod. 54, 138-145 (1996).

325. Cross,N.L. & Mahasreshti,P. Prostasome fraction of human seminal plasma prevents sperm from becoming acrosomally responsive to the agonist progesterone. Arch. Androl 39, 39-44 (1997).

326. Carlsson,L., Ronquist,G., Stridsberg,M., & Johansson,L. Motility stimulant effects of prostasome inclusion in swim-up medium on cryopreserved human spermatozoa. Arch. Androl 38, 215-221 (1997).

327. Fabiani,R., Johansson,L., Lundkvist,O., Ulmsten,U., & Ronquist,G. Promotive effect by prostasomes on normal human spermatozoa

280 exhibiting no forward motility due to buffer washings. Eur. J. Obstet. Gynecol. Reprod. Biol. 57, 181-188 (1994).

328. Fabiani,R., Johansson,L., Lundkvist,O., & Ronquist,G. Enhanced recruitment of motile spermatozoa by prostasome inclusion in swim-up medium. Hum. Reprod. 9, 1485-1489 (1994).

329. Fabiani,R., Johansson,L., Lundkvist,O., & Ronquist,G. Prolongation and improvement of prostasome promotive effect on sperm forward motility. Eur. J. Obstet. Gynecol. Reprod. Biol. 58, 191-198 (1995).

330. Stegmayr,B. & Ronquist,G. Promotive effect on human sperm progressive motility by prostasomes. Urol. Res. 10, 253-257 (1982).

331. Tomlins,S.A. et al. Role of the TMPRSS2-ERG gene fusion in prostate cancer. Neoplasia. 10, 177-188 (2008).

332. Bi,X. et al. Association of TMPRSS2 and KLK11 gene expression levels with clinical progression of human prostate cancer. Med. Oncol.(2009).

281