Mechanochemical Methods for Single Molecule Biochemistry and Studies of Thiol-Disulfide Exchange in Proteins Pallav Kosuri

Mechanochemical Methods for Single Molecule Biochemistry and Studies of Thiol-Disulfide Exchange in Proteins

Pallav Kosuri

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy under the Executive Committee of the Graduate School of Arts and Sciences

COLUMBIA UNIVERSITY

2012

ABSTRACT

Mechanochemical Methods for Single Molecule Biochemistry and Studies of Thiol-Disulfide Exchange in Proteins

Pallav Kosuri

In this thesis, I develop single molecule methods to detect the formation of covalent bonds in proteins. With the use of mechanical force, I investigated the mechanism of disulfide formation in the protein synthesis pathway, and how enzymes catalyze this process. Disulfide bonds between cysteine residues serve as essential structural and functional components in many proteins. In humans, formation of these covalent bonds takes place during oxidative protein folding, a process that is catalyzed by the enzyme Protein Disulfide Isomerase (PDI). Formation of incorrect disulfides can be pathogenic. It is therefore of utmost importance to understand how

PDI effectively recognizes which atoms to join, even in environments crowded with cysteines.

However, this and other fundamental aspects of oxidative folding remain unexplored due to our current lack of precise experimental methods.

I used a custom built Atomic Force Microscope (AFM) to simultaneously detect protein folding and disulfide formation in single protein molecules. This powerful assay allowed me to determine the sequence of reactions during PDI catalysis of oxidative folding. I discovered that

PDI actually does not recognize which cysteines to join together in its protein substrate. Instead, the enzyme relies on the natural folding pathway of its substrate to guide the pairing of cysteines.

This remarkable finding can explain how one enzyme can catalyze the oxidative folding of a wide variety of proteins.

While disulfide bonds are known to stabilize proteins, it has remained unknown how other cysteine modifications affect the mechanical properties of proteins. I show in a series of

experiments that protein mechanics are strongly modulated by S-glutathionylation, a common post-translational modification involved in cellular redox signaling. These modulatory effects are specific to cryptic cysteines that only become exposed under strain. Relevant to the many proteins that function under mechanical force, this mechanism provides a way to dynamically adjust protein and tissue elasticity. Strain-dependent cysteine modification may represent a new paradigm in mechanochemical regulation.

In a final set of experiments, I show that substrate conformation alters the chemical properties of thioredoxin, an enzyme that catalyzes the reduction of disulfide bonds. The ability of a substrate to influence the p Ka of a catalytic residue may play an important role in determining substrate specificity.

The studies I present in this thesis provide fundamental new insights into the protein synthesis pathway, mechanochemical regulation and enzyme catalysis. Together with the methods provided here, my work sets the stage for the emerging field of mechanical biochemistry.

Table of Contents

1 Background and Significance ...... 2

1.1 The art of making an extracellular protein ...... 2

1.2 Single molecule biology ...... 4

1.3 Force as a probe of biochemistry ...... 5

1.4 Disulfide bonds ...... 7

1.4.1 Disulfide chemistry ...... 7

1.4.2 Disulfides in biology ...... 9

1.4.3 Disulfide formation in cells ...... 10

1.5 Thiol-disulfide oxidoreductases ...... 11

1.5.1 Thioredoxin ...... 12

1.5.2 Protein Disulfide Isomerase ...... 16

1.5.3 Glutaredoxin ...... 20

1.6 Glutathione ...... 21

1.7 Studies of oxidative protein folding in vitro ...... 22

1.7.1 Conventional in vitro assays of oxidative folding ...... 23

1.7.2 Small molecule catalysis of oxidative folding in vitro ...... 24

1.7.3 PDI catalysis of oxidative folding in vitro ...... 27

2 Experimental Techniques ...... 30

2.1 The Atomic Force Microscope ...... 30

2.1.1 Basic principle ...... 31

2.1.2 AFM head ...... 34

2.1.3 Laser ...... 34

2.1.4 Fluid cell ...... 34

2.1.5 Piezoelectric actuator ...... 35

2.1.6 Sample slide ...... 36

2.1.7 Cantilever ...... 36

2.1.8 Data acquisition ...... 37

2.1.9 Supporting structures ...... 37

2.1.10 Force-clamp ...... 38

2.1.11 Calibration ...... 41

2.2 Protein engineering ...... 45

2.2.1 Protein expression and purification ...... 46

2.3 Single Molecule Force Spectroscopy ...... 47

2.3.1 Protein mechanics under force ...... 48

2.3.2 Mechanical protein unfolding ...... 50

2.3.3 Force based assays of protein folding ...... 55

2.3.4 Disulfide detection using force ...... 56

2.3.5 Disulfide cleavage under force ...... 58

2.3.6 Setup of an experiment ...... 60

2.3.7 Pulse protocols ...... 61

2.3.8 Data Analysis ...... 62

3 Single Molecule Study of Oxidative Protein Folding ...... 65

3.1 Physiological significance of PDI activity ...... 66

3.2 Oxidative folding pathways in cells ...... 67

3.3 Spontaneous disulfide oxidation kinetics ...... 69

3.4 PDI-mediated oxidation of folded proteins ...... 70

3.5 Creating a mixed disulfide complex between PDI and a single unfolded protein ...... 71

3.6 Mixed disulfide complexes with PDI enable oxidative folding ...... 73

3.7 Detection of spontaneous enzyme release ...... 74

3.8 Measurement of the PDI spontaneous release rate ...... 78

3.9 Kinetics of oxidative protein folding from an extended state ...... 82

3.10 Interchangeability of oxidoreductase enzymes ...... 83

3.11 Formation of interdomain disulfides triggers misfolding ...... 85

3.12 Discussion and biological implications ...... 89

3.12.1 A kinetic model for the oxidative folding of an extended protein ...... 89

3.12.2 Significance of PDI spontaneous release ...... 92

3.12.3 PDI uses a non-specific mechanism to favor native disulfides ...... 93

3.12.4 Future directions ...... 94

3.13 Materials and methods ...... 95

3.13.1 Single-molecule Atomic Force Microscopy ...... 95

3.13.2 Mechanical fingerprints of I27 32-75 ...... 96

3.13.3 Formation of mixed disulfide complexes ...... 97

3.13.4 Data analysis ...... 98

4 S-Glutathionylation Modulates Protein Mechanics ...... 100

4.1 Physiological significance of S-glutathionylation ...... 100

4.2 Oxidative stress in muscle ...... 101

4.3 Direct detection of S-glutathionylation reactions in a single protein molecule ...... 102

4.4 S-glutathionylation inhibits protein folding ...... 104

iii

4.5 Glutaredoxin rescues folding of S-glutathionylated proteins ...... 106

4.6 Glutathionylation modulates protein mechanics in a native titin domain ...... 107

4.7 A kinetic model for redox regulation of protein stability ...... 112

4.8 Discussion and biological implications ...... 113

4.9 Materials and methods ...... 114

4.9.1 Single-molecule Atomic Force Microscopy ...... 114

4.9.2 Data Analysis ...... 115

5 Conformational Tuning of Cysteine Reactivity ...... 117

5.1 The role of pKa in thioredoxin catalysis ...... 117

5.2 Single molecule assay for pKa determination ...... 119

5.3 Substrate conformation alters the pKa of Cys32 in thioredoxin ...... 122

5.4 Kinetic model for substrate dependent pKa in an enzyme ...... 123

5.5 Discussion and biological implications ...... 124

6 Perspective and Future Directions ...... 127

6.1 Single molecule studies of protein mechanics in muscle ...... 128

6.2 Measuring the fighting power of a protein with Isometric Force Microscopy ...... 131

6.3 Mechanochemical signaling through cysteine modifications ...... 132

References ...... 134

Appendix – Published Articles ...... 156

List of Figures

Figure 1.1 The birth of proteins ...... 2

Figure 1.2 Disulfide structure...... 7

Figure 1.3 Thiol-disulfide exchange reaction ...... 8

Figure 1.4 Co-translational oxidative folding in the ER ...... 11

Figure 1.5 Thioredoxin-mediated disulfide reduction...... 13

Figure 1.6 Human thioredoxin (TRX) ...... 14

Figure 1.7 Protein Disulfide Isomerase (PDI) from yeast ...... 17

Figure 1.8 PDI-mediated disulfide formation ...... 18

Figure 1.9 Glutathione (GSH)...... 21

Figure 2.1 AFM imaging...... 31

Figure 2.2 Basic principle of the AFM ...... 32

Figure 2.3 The custom-built AFM setup...... 33

Figure 2.4 The AFM fluid cell ...... 35

Figure 2.5 Sample slide with drop of solution containing the protein substrate...... 36

Figure 2.6 Cantilevers ...... 37

Figure 2.7 The force feedback system ...... 38

Figure 2.8 Force and length measurements of a single I27 protein molecule ...... 41

Figure 2.9 Frequency spectrum of a cantilever...... 44

Figure 2.10 Molecular structure of the two variants of the I27 protein ...... 45

Figure 2.11 Worm-like chain model ...... 49

Figure 2.12 The free energy of a protein as a function of end-to-end length ...... 51

Figure 2.13 Mechanical unfolding of a wild-type I27 polyprotein ...... 53

Figure 2.14 Force dependence of ubiquitin unfolding ...... 54

Figure 2.15 Protein folding in a wild-type I27 polyprotein...... 56

Figure 2.16 Disulfide detection...... 58

Figure 2.17 Direct detection of single disulfide cleavage reactions ...... 59

Figure 3.1 Summary of reactions catalyzed by PDI...... 68

Figure 3.2 A single molecule approach to study oxidative folding ...... 69

Figure 3.3 Kinetics of catalyzed oxidation of folded I27 32-75 ...... 72

Figure 3.4 Differential accessibility of reduced and oxidized cysteines ...... 73

Figure 3.5 Mixed disulfide complexes with PDI enable oxidative folding...... 75

Figure 3.6 Oxidative folding with PDI A1 shows 25 nm steps in the probe pulse ...... 76

Figure 3.7 Enzyme release determines outcome of oxidative folding ...... 77

Figure 3.8 The Cys35Ser mutation in TRX...... 78

Figure 3.9 Measurement of PDI release kinetics ...... 80

Figure 3.10 The kinetic competition during refolding of I27-PDI ...... 81

Figure 3.11 Refolding of oxidized I27 32-75 in the absence of enzymes ...... 83

Figure 3.12 Measurement of protein folding from an extended state ...... 84

Figure 3.13 Non-native interdomain disulfides prevent native folding...... 87

Figure 3.14 Histogram of atypical steps from all experiments with PDI ...... 88

Figure 3.15 PDI-mediated oxidative folding in the ER...... 91

Figure 3.16 Mechanism of PDI to favor the formation of native disulfide bonds...... 94

Figure 4.1 Architecture of a muscle sarcomere ...... 102

Figure 4.2 Disulfide reduction by GSH, but not TRX, inhibits protein refolding ...... 105

Figure 4.3 Glutaredoxin rescues folding of S-glutathionylated proteins ...... 107

Figure 4.4 Inhibition of protein refolding through glutathionylation of cryptic cysteines ...... 110

Figure 4.5 Strain dependent modulation of I27 stability ...... 111

Figure 5.1 Structure of TRX in a mixed disulfide complex with a substrate polypeptide ...... 118

Figure 5.2 Single molecule detection of TRX-mediated disulfide reduction ...... 120

Figure 5.3 Measurement of reduction rate and pKa ...... 121

Figure 5.4 Force dependent pKa of Cys32 in human TRX...... 123

Figure 5.5 Differential pKa of Cys32 in TRX ...... 124

Figure 6.1 Measuring single molecule protein mechanics in an intact myofibril ...... 130

vii

Acknowledgements

I would like to thank a number of people for their guidance and support throughout my journey here at Columbia University. First and foremost, I feel truly honored to have had Julio Fernández as an adviser and as a mentor. His pioneering spirit, creative mind and love for the art of science have been contagious; these are qualities I hope to possess for the rest of my life.

The members of the Fernandez lab have been like a second family to me. Our heated scientific arguments and our controversial lunch topics, both represent unforgettable components of my PhD experience. In terms of work, they could not have been more supportive: Raul Perez-

Jimenez and Arun Wiita gave me a flying start when I first started, and I have had the pleasure of working closely with Raul ever since. Sergi Garcia-Manyes and Rodolfo Hermans were always available for discussions and advice. Jorge Alegre-Cebollada is a pillar of reason that I could always rely on, and I never once doubted the support I had from Carmelu Badilla. Recently, I have had the fortune of working with David Giganti, whose creativity never ceases to amaze me.

Ronen Berkovich has enlightened my day on many occasions, and so has Ionel Popa, with intelligent criticism and honest advice. Yalda Javadi has been supportive both professionally and as a friend. I also had the honor of working with Jason Feng, Koti Ainavarapu, Tzu-Ling Kuo,

Diego Rojas, Jocelyn Lo, Shayna Busch and Jaime Rivas-Pardo, among many other brilliant people. My external collaborators, who have been of great assistance and support, include Arne

Holmgren, José Manuel Sanchez-Ruiz, Álvaro Ingles-Prieto, Brent Stockwell and Anna Kaplan.

During my rotation projects, I had the fortune of working with three highly distinguished scientists: Arthur Palmer, Ruben Gonzalez and Eric Greene. They all gave me great personal guidance and invaluable training in conducting scientific research. So did Larry Shapiro, Richard

Friesner and Peter Sims, who were truly irreplaceable as members of my committee.

viii

I am grateful for having amazing friends who enriched my life here in New York. I fondly recall aggressive racquet sports with Arun, running and philosophizing with Scott, debating music and craziness with Zack, and innumerable other great memories with Bricklin, Stephanie,

Timur, Daphne, Hector and many other remarkable people.

My family has been a source of never-ending and unconditional support. Venkat and

Victoria have been absolutely incredible, as have my uncles Redy and Prasad. All of my other relatives deserve credit as well, and none of this would of course have been possible were it not for my mom and dad, who since my childhood taught me the virtues of compassion and reason.

My brother deserves a gold medal for standing by my side no matter if we are working overnight in the lab, driving in silence through Fresno or climbing a mountain in eastern Anatolia. Finally,

I would like to thank Bobbi for her unwavering support, for teaching me how to fly and for making me a better person every single day.

Chapter 1 Background and Significance

1 Background and Significance

Figure 1.1 The birth of proteins. Left: Electron micrograph of a pancreatic acinar cell showing endoplasmic reticula (ER, elongated structures) studded with ribosomes (black dots). Secretory proteins are synthesized by ribosomes and fold in the ER before leaving the cell. Center: Molecular model of a ribosome exporting a nascent protein to the ER during ongoing synthesis (adapted from ref [1]). The nascent chain (NC) emerges in an extended state through the protein-conducting channel (PCC) and into the ER, where it acquires disulfide bonds during oxidative folding. Right: In this thesis I present new mechanochemical methods to study the oxidative folding of a single protein molecule (green). My studies explain how enzymes such as PDI (white) catalyze the formation of disulfide bonds.

1.1 The art of making an extracellular protein

All cells make proteins, but few of them have mastered the art like the pancreatic acinar cells

(Figure 1.1, left panel). These cells manufacture massive amounts of enzymes that are secreted to

3 the stomach, where they help us digest our food. The secretory system in acinar cells may be extraordinary, but the pathway for synthesis of extracellular proteins is shared among most cells in our body.

The journey of an extracellular protein begins in the cytosol on a ribosome, where

polypeptide synthesis takes place. As the first residues of the nascent polypeptide peek out of the

ribosomal exit tunnel, they are captured by the signal recognition particle, which halts translation

and mediates docking of the ribosome with the translocase machinery in the ER membrane.

When translation resumes, the nascent polypeptide chain is pushed by the ribosome directly into

the ER lumen (Figure 1.1, center panel). Until now the nascent chain has remained extended, but

as it emerges into the ER it undergoes oxidative folding. During this critical process, the nascent

chain acquires modifications such as disulfide bonds, and eventually adopts a stable native

structural fold. As the matured protein continues on its path to secretion, via the Golgi apparatus

and secretory vesicles, it relies on its native structure to protect it from the harsh environment

that awaits outside the cell.

Thanks to half a century of research in cell biology, molecular biology and biochemistry, we

now know the identity and functions of the main components in the protein synthesis pathway.

The highly conserved enzyme Protein Disulfide Isomerase (PDI) is the main catalyst of oxidative

folding in the ER, where it interacts with a wide range of substrates. Biochemical studies have

revealed the chemical reactions catalyzed by PDI. In addition, the thermodynamic parameters

that govern these reactions under equilibrium conditions have been worked out through years of

studies in bulk. However, oxidative protein folding in the cell may never reach equilibrium. For

instance, a disulfide bond in a folded protein can be shielded from the surrounding medium. The

protein structure can in turn be locked by the disulfide. Like a door that becomes locked from the

4 inside, formation of such disulfides can be irreversible. Consequently, such a protein could be trapped in a metastable state for a time comparable to its lifetime in the ER. In another example, a malformed disulfide can trigger misfolding followed by irreversible aggregation, a process of fundamental importance to the pathogenesis of misfolding diseases. In order to understand these processes we need to know the trajectory of the protein during oxidative folding. A new generation of biophysical techniques enables such investigations, through the study of single molecules.

1.2 Single molecule biology

Studies of single molecules have led to a number of fundamental discoveries in biology, from the dynamic nature of enzymes [2] to catalytic mechanisms of molecular motors [3-6] to the origin of tissue elasticity [7]. These and the many other advances in single molecule biology were made possible by the development of new methods to study individual biomolecules in action.

A variety of approaches have been used to perform measurements on single biomolecules.

To date, most of these rely on fluorescence in one way or another. Single protein or nucleic acid molecules tagged with a fluorescent dye can be visualized with a regular light microscope as long as they can be isolated. Nevertheless, achieving these two conditions turns out to be quite the challenge, because biological systems tend to be both crowded and highly dynamic. In vitro, isolation and localization can be achieved by diluting the concentration of molecules and tethering them to a surface [8]. In addition, a number of different innovative approaches are today available to aid the isolation of single fluorescent molecules. Some rely on limiting the illumination volume, for instance through total internal reflection fluorescence microscopy

(TIRF) [9], zero-mode waveguides (ZMW) [10] or stimulated emission depletion microscopy

(STED) [11]. Other methods such as photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM) use stochastic activation to visualize small numbers of molecules even when the total concentration is high [12, 13].

Fluorescence microscopy can be used to localize molecules and measure individual reactions in real-time based on the fluorescence signal alone [2, 10]. In combination with FRET, where distance fluctuations can be measured, these methods can measure sub-nanometer dynamics of molecules [14-16]. Single molecule methods such as these have given us an unprecedented view of life at the molecular level.

1.3 Force as a probe of biochemistry

Force spectroscopy has emerged as a powerful tool to study the biochemistry of single molecules. In force spectroscopy, mechanical force is used as a perturbation, analogous to the use of light in optical spectroscopy. Mechanical force is a common occurrence in many biological systems. However, the usefulness of force as a scientific tool extends far beyond its physiological relevance. Just as electrophysiological techniques can be used to drive nerve cells to a desired electrical state, force can be used to probe molecular structure and function. By altering the perturbation in a controlled manner while measuring a molecule’s response, we can test hypotheses and begin to construct an accurate quantitative model of that molecule. Single molecule force spectroscopy has contributed greatly to our understanding of protein folding and unfolding [6, 17-20], binding and unbinding [21-23], as well as covalent bond cleavage [24-26].

A variety of single molecule force spectroscopy methods are available, but the basic principle remains the same: A mechanical force is applied to a molecule while measuring its extension. Today there are three predominant methods: optical tweezers, magnetic tweezers and

Atomic Force Microscopy (AFM). The three methods are in theory capable of performing the same experiments. However, optical and magnetic tweezers are usually limited to forces below

100 pN while AFM can operate at higher forces. The limitation of tweezers arises both from the technical difficulty of applying high forces, but also from the mechanical weakness of the most commonly used attachments (such as biotin-avidin or antibody-antigen). Furthermore, while optical and magnetic tweezers typically require extended flexible linkers such as DNA handles, which can obscure the measurement of dynamics of the molecule under study, AFM uses stiff probes that directly apply a force to the molecule.

Due to these limitations, the various methods have been used to study different systems.

Optical and magnetic tweezers have mainly been used to study nucleic acids and the action of

motor proteins, while AFM mainly has been used to study protein folding and unfolding [27].

The access to high forces and the direct extension measurement makes AFM particularly

amenable to structural studies of proteins, as well as kinetic measurements of protein folding,

bond rupture and bond formation.

Of particular significance to this thesis, an AFM based assay was recently developed to

directly detect the cleavage of single disulfide bonds under a constant force [26]. In this study, a

stretching force was applied to a polypeptide chain crosslinked by disulfide bonds. Cleavage of a

disulfide bond under force released the extension of a polypeptide segment with a known length.

This predetermined extension step provided a unique fingerprint for the specific bond cleavage

reaction. While covalent bond cleavage had been measured in earlier reports [28], the use of a

fingerprint in this new constant force assay essentially removed all ambiguity from the analysis.

In this thesis I present developments of this assay that open the door to studies of protein

disulfide formation as well as other thiol modifications in single proteins.

1.4 Disulfide bonds

Disulfides are covalent bonds that are formed between two sulfur atoms. In proteins, they are formed between the γ-sulfurs in cysteine residues. The disulfide creates a crosslink in the otherwise linear polypeptide chain (Figure 1.2).

Figure 1.2 Left: A disulfide creates a strong covalent crosslink in a protein. Right: Chemical structure of two cysteine residues linked by a disulfide bond.

1.4.1 Disulfide chemistry

Disulfide bonds are formed through the oxidation of two thiols:

2 → + 2 ͮ + 2 ͯ ͍͌͂ ͍͍͌͌ ͂ ͙

In biology, however, most disulfides are formed and broken through thiol-disulfide exchange reactions [29, 30]:

+ → + ͍͌ͥ͂ ͍͍͌ͦ͌ͧ ͍͍͌ͥ͌ͦ ͍͌ͧ͂

This S N2 reaction is actually mediated by the deprotonated thiolate anion:

⇌ ͯ + ͮ ͍͌͂ ͍͌ ͂

The mobile proton is in equilibrium with solution. The thiol group’s pK a can therefore be used as

a measure of the proportion of thiols that are found in the deprotonated state at any given time:

ͯ − = ʞ͍͌ ʟ ͤ͂ ͤͅ ͣ͛͠ ʞ͍͌͂ ʟ

A lower pK a signifies a higher reactivity of a thiol group, since this indicates higher prevalence of the reactive thiolate form.

Two consecutive thiol-disulfide exchange reactions can directly transfer a disulfide from one

pair of sulfurs to another. The symmetry of the reaction means that it in principle is reversible,

which adds to the dynamic nature of disulfide bonds in biology (Figure 1.3).

Figure 1.3 Transfer of a disulfide from one protein to another through two sequential thiol-disulfide exchange reactions. Curved arrows indicate nucleophilic attack. Hydrogen atoms on thiols have been omitted for clarity.

Since disulfide formation involves oxidation of the participating sulfurs, the reaction can be characterized in terms of its associated electrochemical redox potential E°. The thermodynamic stability of a disulfide relative to the corresponding thiols can be determined through the Nernst equation:

° = − ͎͌ ̿ ͢͠Ƴͅ , Ʒ ̀͢

Where E° is given as an electrochemical potential in units of volts. Here, R is the gas constant, T is the temperature, n is the number of electrons exchanged in the reaction ( n=2), F is the Faraday constant and Keq is the equilibrium constant for the reaction.

From an experimental point of view, the redox potential of a given disulfide is often measured in relation to a disulfide with a known redox potential such as glutathione. This approach has been used extensively to characterize a wide range of biological disulfides [31].

However, for thiol-disulfide reactions in macromolecules such as proteins, the accessibility of the disulfide can play an important role. Namely, if a sulfur atom is inaccessible inside a protein, the kinetics of any reaction can be so slow as to prevent thermodynamic equilibrium from being reached on a biologically relevant timescale. For instance, none of the 17 disulfide bonds in human serum albumin can be reduced without first denaturing the protein [32]. Furthermore, the disulfide formation pathway in vivo appears to resemble a system in steady state, where

oxidizing equivalents are continually shuttled from molecular oxygen to the constant stream of

newly synthesized proteins traversing the secretory pathway [33].

1.4.2 Disulfides in biology

Disulfide bonds are formed as posttranslational modifications in most secretory and cell-surface

proteins [34]. Formation of disulfides is not only essential for many human proteins, but also

constitutes a critical step in numerous pathologies. Bacteria express a wide range of disulfide

bonded virulence factors, including secreted toxins and adhesion molecules [35]. Virus infection

and replication is similarly dependent on disulfide formation. For instance, docking of HIV-1 to

human immune cells requires the correct folding of the viral gp120 receptor, a protein that

contains nine disulfide bonds [36]. In still other cases, pathogenicity can be caused by

10 dysfunctional oxidative folding, which has been linked to protein misfolding disease [37, 38]. In these cases, aberrant disulfide formation can lead to protein misfolding which can then trigger aggregation. A complete understanding of all of these processes requires a detailed description of how protein disulfides are formed.

1.4.3 Disulfide formation in cells

The general pathway for enzyme-catalyzed disulfide formation in cells has been characterized over the last decades [29, 34, 39]. In both prokaryotes and eukaryotes, the Sec translocase machinery mediates transport of nascent polypeptides from the cytosol to compartments where disulfide formation takes place, such as the periplasmic space in bacteria and the ER in eukaryotes [1, 29, 40, 41]. A nascent polypeptide remains unfolded inside the Sec channel and is, in the case of co-translational export, translocated during ongoing synthesis on the ribosome

[42](Figure 1.4). When emerging on the trans side of the Sec channel, an exposed cysteine in the translocating substrate can become engaged with an oxidase enzyme such as PDI (in eukaryotes) or DsbA (in prokaryotes). This leads to the formation of a mixed disulfide bond between the enzyme and the substrate. Such a mixed disulfide complex is central to oxidative folding, and has been identified in vivo [43-45]. Continued translocation eventually results in the release of a second substrate cysteine. This second cysteine can, upon protein folding, attack the enzyme- substrate mixed disulfide and thereby form an intramolecular disulfide in the substrate.

Figure 1.4 Co-translational oxidative folding in the ER. The polypeptide is extended as it traverses the membrane. PDI catalyzes oxidative folding by engaging in a mixed disulfide complex with its substrate.

1.5 Thiol-disulfide oxidoreductases

Cysteine residues are found in most proteins, and all of these can potentially form disulfide

bonds. Disulfides are therefore heavily regulated in cells, by a class of highly conserved enzymes

known as thiol-disulfide oxidoreductases [29]. These enzymes rely on thiol-disulfide exchange

reactions (Figure 1.3) to catalyze the formation of native disulfide bonds (such as PDI in the ER),

or to reduce erroneous bonds that could be deleterious (such as TRX and GRX in the cytosol).

Thiol-disulfide oxidoreductases are found in all cells from bacteria to human, and they share a number of common features. Almost all belong to the thioredoxin superfamily of proteins, a family characterized by a distinct structural fold and active site sequence [46]. The thioredoxin fold consists of a globular single domain made up of a four-stranded beta-sheet surrounded by three alpha-helices [47]. The active site protrudes from the protein surface and contains a CXXC motif, two reactive cysteines separated by two variable amino acid residues. The identities of the

two variable residues are strongly dependent on the enzyme function, and determines the activity

to some degree [48].

1.5.1 Thioredoxin

The cytosol of cells provides a reducing environment where disulfides are rarely formed [47].

Two enzymatic systems that are present in all organisms interact directly with protein thiols to

maintain them in a reduced state: the thioredoxin and the glutaredoxin system [49].

The 12 kDa enzyme Thioredoxin (TRX) was the first discovered disulfide reductase. It was originally identified in E. coli as the electron donor for ribonucleotide reductase (RNR) [50].

RNR is an essential enzyme for DNA synthesis, which reduces all four ribonucleotides into the corresponding deoxyribonucleotides. RNR itself becomes oxidized as part of the reaction cycle, and has to be reduced before regaining activity [51]. In practice, this occurs as a reduction of a disulfide bond in RNR, a reaction that is mediated by TRX. However, it soon became clear that

TRX functioned as a general disulfide reductase [52], and that its functions in the cell therefore were much more diverse [53].

The catalytic mechanism of TRX is today well understood. An early clue to its reaction cycle was given by the differential reactivity of the two active site cysteines in E. coli Trx, Cys32 and Cys35 [54]. It was found that out of the two active site cysteines, Cys32 showed a higher reactivity, arising from the lower p Ka of this thiol group. The deprotonated form of a cysteine thiol (the thiolate anion form) is much more nucleophilic and thus reactive towards disulfides.

Free cysteine in solution has a p Ka of around 8.7 [55], and is thus mainly found in the protonated state at physiological p H. Meanwhile, the lower p Ka of Cys32 in Trx leads to a much-increased

effective concentration of the reactive deprotonated species, which therefore leads to higher

reactivity at physiological p H.

Holmgren outlined the reaction cycle of TRX with a disulfide in a protein substrate as

follows (Figure 1.5) [53]: The deprotonated thiol in Cys32 performs a nucleophilic attack on one

of the participating sulfur atoms in the disulfide. This leaves TRX covalently attached to its

substrate through a mixed disulfide bond. A second nucleophilic attack by Cys35 then resolves

the mixed disulfide and releases TRX from its substrate. TRX has at this point gained a disulfide

bond while the substrate has been reduced.

Figure 1.5 TRX reduces a disulfide bond in a protein substrate through two consecutive thiol-disulfide exchange reactions.

Oxidized TRX has to be reduced in order to be reactivated. This reaction, which completes the reaction cycle, is mediated by thioredoxin reductase, a TRX-specific reductase enzyme that draws electrons from NADPH in order to reduce TRX [53].

The ability of TRX to act as a non-specific disulfide reductase was first observed with insulin, where it was shown to reduce the disulfide bonds with a kinetic rate approximately

20,000 times faster than an equivalent amount of DTT [52, 56]. This high rate could only in part be due to the lowered p Ka of the active site cysteine, which at most would increase the reaction rate by two orders of magnitude. Other factors must thus play a role in the enzymatic activity of

TRX. Some clues were given by the crystal structure solved for oxidized E. coli Trx at 2.8 Å resolution (PDB: 1srx)[57]. The structure showed that the catalytic cysteine Cys32 was relatively

exposed, and found by a shallow groove that featured a hydrophobic patch presumably involved

in substrate binding (Figure 1.6) [58].

Figure 1.6 Human thioredoxin (TRX). The groove in the active site is highlighted in red. The reactive sulfurs in the two active site cysteine residues Cys32 and Cys35 are shown in yellow, with Cys32 being the most exposed. (PDB:1trs)

Substrate recognition by TRX was further investigated in 1995, when Qin and coworkers reported the structure of human TRX in a mixed disulfide complex with a peptide fragment of its substrate NF-κB, using solution NMR (PDB: 1mdi)[59]. Since the TRX-substrate mixed disulfide is extremely short-lived under normal conditions, with a lifetime of less than a millisecond [60], the authors of the structure had mutated Cys35 in TRX to alanine. This mutation effectively abolished the ability of TRX to release from it substrate. Furthermore, the substrate fragment only contained one cysteine, thus also preventing a cysteine in the substrate from attacking the mixed disulfide bond. The structure revealed that the substrate was coordinated in the hypothetical binding groove of TRX, located adjacent to Cys32. The interaction between the enzyme and substrate consisted of a combination of hydrophobic, electrostatic and hydrogen bonds, both to the substrate backbone and side chain atoms, suggesting a limited amount of specificity. However, the same group later reported a second structure of TRX bound to a substrate peptide fragment from Ref-1 (PDB: 1cqg)[61]. This structure curiously revealed that the orientation of the substrate in the TRX binding groove was

reversed as compared to NF-κB. Aside from this observation, which clearly suggested flexibility

in the substrate recognition of TRX, similar conclusions were drawn regarding the enzyme-

substrate interactions. A more recent crystal structure of barley TRX bound to its substrate barley

α-amylase/subtilisin inhibitor (BASI) showed a binding interface consisting of van der Waals interactions and backbone-backbone hydrogen bonds, potentially indicating the minimum requirement for substrate recognition (PDB: 1iwt)[62]. Still, it remains unknown how the enzyme specifically recognizes the disulfide in its substrate.

Many studies over the years have used site-directed mutagenesis to help clarify the function of TRX. Most of these studies have been focused on the highly conserved active site motif Trp-

Cys-Gly-Pro-Cys. The earliest studies demonstrated the crucial role of the active site cysteines

Cys32 and Cys35 [53]. It was found that mutation of Cys32 abolished reductase activity whereas mutation of Cys35 lead to formation of stable mixed disulfide intermediates (such as those used in the structural studies described above). Next, Trp31 was mutated to a variety of different amino acids [63]. Trp31 forms part of the proposed substrate-binding hydrophobic patch, and as such the hydrophobicity of the substituted residue was expected to play an important role in the retention of activity. Indeed, mutation to other aromatic residues retained or even increased the catalytic activity of the enzyme, whereas mutation to residues with small or charged side chains showed deleterious effects [63]. A Pro34His mutation decreased the reductase activity of TRX and increased the redox potential of the active site disulfide, effectively making the enzyme more oxidase-like from a thermodynamic point of view [64]. Notably, such a WCGHC active site motif is found in oxidases such as PDI where it is highly conserved.

1.5.2 Protein Disulfide Isomerase

The eukaryotic enzyme Protein Disulfide Isomerase (PDI) was discovered by Anfinsen nearly 50 years ago [65, 66]. Originally described as a folding catalyst, it was seen to accelerate the reactivation of denatured and reduced ribonuclease (RNAse) and lysozyme. The enzyme was given the name PDI for the following reason: The refolding of proteins with multiple disulfide bonds such as RNAse and lysozyme in vitro often leads to formation of erroneous disulfide bonds. These have to be isomerized before the native fold can be acquired, and disulfide isomerization is thus the rate limiting step during refolding. Since the newly discovered enzyme catalyzed this reaction, it was given the name Protein Disulfide Isomerase. However, this name is today known to be a slight misnomer, since an essential role of PDI in vivo is to catalyze the formation of disulfide bonds in newly synthesized proteins [39].

The primary sequence of PDI was first published in 1985 and revealed two regions with predicted homology to TRX, each with a WCGHC active site motif [67]. Later studies showed that the enzyme consists of four domains and a C-terminal extension arranged as A1-B1-B2-A2-

C (Figure 1.7), where the catalytically active A1 and A2 domains are highly similar to TRX in sequence and structure [39]. The two B domains share a high sequence similarity with each other and both adopt a TRX fold despite having a low sequence similarity with TRX [68].

For human PDI, the structure of the isolated A1 domain was the first to be determined

(PDB: 1mek)[69, 70], followed by reports on the structure of the B1 domain (PDB:2bjx) [68], the B2 domain (PDB: 3bj5) [71], a B1-B2 construct (PDB: 2kl8) [72], the A2 domain (PDB:

1x5c) and a B1-B2-A2 construct (PDB: 3uem). The full-length structure of human PDI has not yet been solved. However, in 2006 Tian et al. reported the full-length crystal structure of the highly homologous PDI from yeast (PDB: 2b5e) [73]. This structure showed the four domains

arranged in a U-shape, with the two active sites in the A domains facing each other. The B

domains form a surface enriched in hydrophobic residues at the bottom of the U, which the

authors argue may contribute to the substrate binding properties of PDI [73].

Figure 1.7 Protein Disulfide Isomerase (PDI) from yeast. PDI consists of four sequential thioredoxin domains. The redox activity resides in domains A1 and A2, which are functional in isolation. (PDB: 2b5e)

Early biochemical experiments in vitro showed that PDI was capable of catalyzing all three kinds of thiol-disulfide exchange reactions in proteins: reduction, isomerization and oxidation [66, 74-

76]. Darby et al. showed using truncation mutants that, although the full-length enzyme displayed optimal reaction rates, a single isolated A1 or A2 domain was sufficient for catalysis of these reactions [77]. Furthermore, while deletion of PDI is lethal in yeast, a single redox-active domain can rescue oxidative folding and growth in a PDI deletion mutant [78, 79]. Cooperative effects notwithstanding, a single thioredoxin domain can thus be used to understand the basic catalytic mechanism of PDI.

For reduction reactions, a reduced PDI enzyme employs the same reaction pathway as TRX

[39]. For oxidation reactions, where a disulfide is introduced into a substrate protein, this

reaction path is essentially reversed (Figure 1.8): A free cysteine in the substrate attacks the

disulfide in oxidized PDI, forming a mixed disulfide complex. The mixed disulfide is then

resolved as it is attacked by a second cysteine in the substrate. Reduced PDI is thereby released

while the substrate has acquired a disulfide bond [39]. Isomerization reactions, where a disulfide

bond is rearranged within a substrate, can take one of two possible paths: In the first path, a

reduced enzyme cleaves the disulfide in the substrate and an oxidized enzyme then introduces

the new disulfide. In the second path, a reduced enzyme attacks the disulfide in the substrate but

does not release spontaneously. Instead, it remains in the mixed disulfide complex until the

substrate rearranges and a new cysteine in the substrate attacks the mixed bond [80]. Which of

the paths is taken probably depends on the substrate, however the relative contributions of the

two pathways in vivo has yet to be established [34].

Figure 1.8 PDI introduces a disulfide bond into a protein substrate through two consecutive thiol-disulfide exchange reactions.

In the cell, PDI is localized to the ER where it catalyzes the oxidative folding of newly synthesized polypeptides [81]. Many secretory proteins contain disulfides that are essential for function and stability, and thus have to undergo oxidative folding in the ER before being secreted through the golgi system [29]. In higher eukaryotes, secretory proteins are exported cotranslationally via the signal recognition system [40, 42, 82, 83]. These proteins are synthesized as linear polypeptides and emerge still unfolded from the ribosomal channel via the

Sec61 translocon into the endoplasmic reticulum [83-85]. Emerging sequentially into the ER, the

nascent polypeptide encounters PDI, which catalyzes co-translational oxidative folding [86, 87].

This reaction is mediated by the formation of a mixed disulfide bond between PDI and a cysteine

in the nascent polypeptide [29, 88, 89], a disulfide that is then transferred to the folding

polypeptide.

PDI is found mainly in its oxidized form in the ER, where it can reach very high concentrations (up to millimolar) [34, 90, 91]. It is maintained in an oxidized state through direct and specific oxidation by the flavoenzyme ERO1 [92]. Two seminal reports showed that ERO1 accepts electrons from PDI through direct thiol-disulfide exchange, first by isolation of PDI-

ERO1 mixed disulfide complexes in vivo [88], and then by reconstitution of the entire system in vitro [93]. ERO1 then passes electrons to molecular oxygen, the ultimate electron acceptor in the cellular system [94]. Disulfide bonds are thus formed de novo by ERO1, and are then transferred

via PDI to newly synthesized proteins through successive thiol-disulfide exchange reactions.

Nevertheless, many questions remain unanswered at the intersection of folding and disulfide

formation, especially regarding the enzyme-substrate complex. Central questions include, how

does this complex affect protein folding? What are the dynamics of the mixed disulfide? What is

the spontaneous off-rate of PDI? What is the rate-limiting step during oxidative folding catalyzed

by PDI or glutathione? Does disulfide formation guide protein folding or vice versa ? Due to the

complexity of oxidative folding, we are unable to answer these questions without the

development of new experimental techniques. Ideally, these techniques should be able to

independently probe both folding and disulfide formation during oxidative folding.

1.5.3 Glutaredoxin

The cytosolic glutaredoxin (GRX) disulfide reductase system was discovered after characterization of E. coli mutants that were able to grow in the absence of a TRX system [95].

GRX was shown to reduce the disulfide in ribonucleotide reductase only in the presence of reduced glutathione (GSH) [96, 97]. Further studies, where various components of the TRX and

GRX systems were deleted, showed that the two systems had overlapping functions in vivo [98].

GRX contains a Cys-Pro-Tyr-Cys active site motif, and is thus capable of reducing protein disulfide bonds through the same mechanism as TRX. However, the kinetic rate of direct reduction of protein disulfides is at least an order of magnitude lower than that of TRX, leading to the suggestion that this is not the primary reaction path in vivo [49].

The disulfide reductase activity of GRX differs from TRX in that it has a high specificity for protein-glutathione mixed disulfides [99]. This was discovered through a radiolabel assay where

GRX showed a strong affinity towards glutathionylated substrates [100]. GRX was thus found to reduce protein disulfides by catalyzing the reaction of GSH with these substrates [99]. Through this mechanism, protein disulfides are reduced in three steps: A GSH molecule attacks a protein disulfide and forms a protein-glutathione mixed disulfide. The glutathionylated protein is then recognized by GRX, which cleaves the glutathione molecule of the protein, leaving it completely reduced. A second GSH from the surrounding solution then cleaves the GRX-glutathione bond, which effectively regenerates reduced GRX [99].

Structural studies of the reduced (PDB: 1egr)[101] and oxidized (PDB: 1grx) [102] forms of

E. coli GRX, and later of the human enzyme (PDB: 1jhb)[103], revealed a basic form of the

TRX fold and an active site loop that protruded from the otherwise globular structure. An NMR structure of human GRX in a mixed disulfide complex with glutathione displayed the glutathione

21 moiety lodged in a binding cleft that appeared to be structurally specific and featuring multiple specific electrostatic interactions [104]. These observations provide a structural basis for the specificity of GRX towards glutathionylated protein substrates.

Glutathionylation has recently been shown to mediate redox signaling in cells through the modulation of protein function [105]. In addition to its function as a general disulfide reductase,

GRX may therefore also play a regulatory role in cells by catalyzing reversible glutathionylation

[106].

1.6 Glutathione

In addition to enzymatic catalysts, the small molecules GSH and its disulfide bonded dimer form glutathione disulfide (GSSG) are also central to the redox balance in cells (Figure 1.9).

Glutathione is the most abundant thiol compound in vivo and is present at a millimolar concentration in the cytosol of all cells and also in the ER of eukaryotic cells [99, 107]. It is a tripeptide consisting of glutamate, cysteine and glycine, and contains an unusual peptidic γ- linkage that protects it from degradation by aminopeptidases [108]. The cytosolic pool of GSH is maintained in a reduced form by glutathione reductase (GR), which depends on electrons from

NADPH to catalyze the reduction of GSSG [107]. This system normally maintains the cytosolic

GSH:GSSG ratio at around 100:1 [109]. It is widely accepted that one of the main functions of

GSH in cells is to provide a redox buffer that protects the cell from oxidative stress [107].

Figure 1.9 Glutathione (GSH).

The GSH:GSSG ratio of 1:1 to 3:1 in the ER appears to favor the formation of disulfide bonds

[110]. However, despite numerous studies showing that glutathione disulfide (GSSG) can act as

an protein oxidizing agent [111], the role of GSH in oxidative folding in vivo is yet not clear [33,

112].

A GSH molecule can form a mixed disulfide with protein cysteine, resulting in a

posttranslational modification known as S-glutathionylation. This covalent modification is

greatly increased during oxidative stress [105]. In addition, a number of recent studies have

shown that S-glutathionylation plays an important role in redox signaling through modulation of

protein function [113-116].

1.7 Studies of oxidative protein folding in vitro

Oxidative folding has been studied extensively in vitro [111, 117]. In these studies, a protein is

typically denatured and reduced by chemical means, and then refolded after the denaturants have

been removed. It is important to note, however, that while the reaction endpoint in these studies

is identical to the reaction endpoint in vivo (i.e. a natively folded protein), these two situations

start from different states. In other words, the end result may be a correctly folded protein, but

the pathway may be different in vivo . Proteins in the ER fold from an extended state, and many

times vectorially as they emerge from the translocase channel [112]. The methods used to probe

oxidative folding in vitro have so far failed to reproduce these features. Still, these experiments

can lead to general insights regarding the forces and interactions involved in protein folding,

which can then be applied to understand oxidative protein folding in the cell.

1.7.1 Conventional in vitro assays of oxidative folding

A typical in vitro assay for oxidative folding involves starting with a completely denatured and

reduced protein followed by a study of the refolding of the protein in a specific redox

environment [111]. As these experiments have conventionally been carried out in bulk, it is of

high importance to verify the homogeneity and synchrony of the denatured protein sample as the

refolding is initiated. The following section will describe the main assays and their shortcomings.

Refolding assays can be classified into three groups, where each has its own strengths and

weaknesses [34]. The first and most classical is based on gain of function of a substrate such as

RNase A [66]. RNase refolds and regains activity while a coupled assay is used to measure this

activity. This method is relatively simple and, in the case of RNase A, gives a strong signal. The

method is qualitative but cannot accurately measure kinetics of oxidative folding as the readout

is complicated by the activity assay and the possible heterogeneity of the sample. RNase A with

its eight cysteines has over 700 partially folded non-native folding intermediates, some of which

have residual activity [34]. Also, there is no way to separate the folding and disulfide formation

processes.

The second class of oxidative folding assays involves measurement of a biophysical change in a substrate. These assays include fluorescence methods involving tryptophan [118] or extrinsic dyes [119, 120]. These methods are quantitative but are limited to fluorescent substrates, and report on folding rather than disulfide formation. In addition, the fluorescent probe is sensitive to the local chemical environment, which may not reflect the overall folded state of the protein unless the folding is two-state.

In the third class of assays, the oxidation reaction is chemically quenched at different time

points and the substrate populations are subsequently analyzed [74]. This approach has enabled

kinetics measurements on a time scale of several seconds. However there are complications

associated with the accessibility of buried thiols and the use of different quenchers can yield very

different results [111]. In particular, the issue of solvent accessibility presents a significant

complication, as many cysteines are inaccessible to quenching agents in folded proteins [117].

1.7.2 Small molecule catalysis of oxidative folding in vitro

In 1994, Darby, Freedman and Creighton published a study in which they dissected the mechanism of disulfide formation by GSSG [121]. Using a 28-residue unstructured model peptide with two cysteines, they were able to trap and detect all possible redox states of this substrate. They could then measure the rates of individual reactions steps during oxidation, in the absence of influence from formation of native protein structure. Their results showed that GSSG reacted with free cysteines with a rate constant of 0.5-3.0 s -1M-1, while the subsequent

intramolecular attack (forming the substrate disulfide) occurred at a rate of approximately 0.001

s-1. While the first rate could well be representative for exposed cysteines in real proteins, the second rate is probably strongly influenced by the structure of the substrate.

Oxidative folding of structured proteins represents a typical chicken-and-egg problem, since disulfide formation can affect conformational folding and vice versa. Formation of a disulfide can lock in a structural element of the protein and restrict the conformational flexibility.

Meanwhile, conformational folding can potentially either bring cysteines together, allowing them to react, or alternatively shield them from attacking groups. The puzzle of oxidative folding has been approached in vitro in a large number of studies that have mainly been focused on a few model proteins, including ribonuclease (RNase A), lysozyme and bovine pancreatic trypsin inhibitor (BPTI).

Bovine RNase A is a small protein that contains four nonconsecutive native disulfide bonds

(26-84, 40-95, 58-110, 65-72). The various different disulfide bonding patterns between these eight cysteines yields a total number of 764 possible disulfide bonded species. As this heterogeneity is difficult to analyze, states are often bluntly grouped according to how many disulfides are formed (i.e. one, two, three or four disulfides). Scheraga and collaborators have led the effort in characterizing the oxidative folding pathway of RNase A in the presence of small molecule catalysts. In 1997, a detailed report was published that characterized the folding intermediates of RNase A as a function of time and redox conditions, using a mixture of reduced and oxidized DTT [122, 123]. These data were then fit by a kinetic model, allowing the extraction of various rate constants. This and subsequent studies showed that RNase during refolding first populated an ensemble of unstructured states featuring scrambled disulfide bonds.

These bonds were then reshuffled until a state with three native disulfides had been formed. At this point, sufficient native structure was present in order to protect the three disulfides from further cleavage, leaving only the last two cysteines to bond and complete the structure [111,

117, 124]. Under these conditions, oxidation appeared to precede folding initially, necessitating many rounds of disulfide isomerization on average before the correct disulfide bonding pattern was achieved. Once a sufficient number of disulfides were in place, conformational folding helped guide the rest of the process and ensured the directionality of folding [117].

Similar conclusions were drawn from experiments on the oxidative folding of lysozyme, performed by Dobson and coworkers [125]. An initial random search among different disulfide bonded unstructured states eventually led to the formation of two to three native disulfides (out of a total of four native disulfides). At this point, a significant amount of native structure was formed, which in turn funneled the protein into a stable near-native conformation with three

26 native disulfides. This state represented a kinetic trap, as the structure limited the solvent accessibility of the remaining thiols. Consequently, the near-native species was slowly oxidized into the fully native form [125]. In conclusion, a general pathway was proposed, described as a random oxidation and isomerization of disulfides, combined with conformational folding

“locking in” native bonds [117, 125].

Studies on the 59 residue, three-disulfide protein BPTI, on the other hand, indicated a completely different pathway for oxidative folding [126, 127]. Out of the 15 possible single disulfide species, 90% of the sample populated only two of these. In other words, rather than a random search there appeared to be a clear pathway starting as early as the initial oxidation event. A separate study found that the protein at this point already had acquired significant native structure elements [128]. As the folding path progressed, two new species were observed, each with two native disulfide bonds. One of these near-native species, missing only the 30-51 disulfide, was shown to be extremely stable, with a lifetime of weeks. NMR and crystallographic studies revealed that this stability was due to a structure very close to that of native BPTI, and that the two remaining free thiols were buried and therefore inaccessible to oxidizing agents

[111]. When BPTI is refolded in vitro , nearly 50% is trapped in this unproductive state. To summarize, early formation of native structure seemed to drive the protein towards the native state, but could also drive the protein into a kinetic trap. The oxidative folding of BPTI thus appeared to follow a different, and more predetermined, pathway than RNase or lysozyme.

These and other in vitro studies of oxidative folding catalyzed by small molecules reveal a diverse set of possible pathways. A common theme appears to be the ability of folded structures to “lock in” native disulfide bonds. However, this phenomenon can also lead to the creation of trapped intermediate states, where cysteines are locked in non-native configurations. These

features are likely also represented during oxidative folding in the cell. Nevertheless, it remains

unknown which, if any, of the actual pathways are represented in vivo [112].

1.7.3 PDI catalysis of oxidative folding in vitro

The catalytic effect of PDI during oxidative folding has conventionally been studied by adding

PDI to the systems described in the preceding section. The reaction mixture thus contains a combination of oxidized and reduced small molecule catalysts such as GSH and DTT, as well as a mixture of oxidized and reduced PDI and substrate in its varying states of oxidative folding.

While overall effects on the kinetics and folding yield can be easily extracted from these experiments, it has been difficult to draw any strong conclusions regarding the mechanism of catalysis by PDI.

PDI was shown to enhance both the rate and yield of RNase A when refolded in a buffer

containing GSH or DTT [90, 91, 129]. The overall rate of native RNase formation increased by

62-fold at 37°C. A subsequent study showed that the overall pathway of oxidative folding was

not altered by PDI [130]. Instead, PDI appeared to catalyze the reaction by increasing the rate of

interconversion of some of the semi-stable states. These states, that represented kinetic traps in

the oxidative folding pathway, were seen to be less stable in the presence of PDI. In order to

clarify whether this effect was due solely to the redox chemistry of the enzyme, a control

experiment was performed with a PDI mutant lacking the active site cysteines [131]. While this

mutant did not accelerate oxidative folding to the same amount as wild type, it still yielded a

modest 2-fold increase in the overall rate of oxidative folding, indicating that the catalytic

mechanism of PDI extends beyond its redox chemistry [111].

The oxidative folding of lysozyme was accelerated 15-fold after the addition of a catalytic amount of PDI [125]. Although the mechanism of catalysis was not extensively analyzed in this study, it appeared that PDI favored the rearrangement of highly structured intermediates, which were less long-lived in its presence.

The most striking effect was seen in a study of PDI-catalyzed oxidative folding of BPTI

[132]. PDI showed modest rate increases in the initial steps of oxidation, but accelerated the final step of oxidative folding by up to 6,000-fold. As PDI essentially eliminated the kinetic trap observed in previous studies of BPTI oxidative folding, this led the authors to conclude that PDI catalyzed the reaction by unfolding the structured intermediate, an interpretation shared by others in the field [111, 117].

Based on these results and on the wide substrate range of substrates in which PDI catalyzes oxidative folding [133], Gilbert and coworkers proposed a model for PDI activity: PDI does not directly guide the folding protein into a specific structure. Rather, it simply accelerates the formation and isomerization of disulfides in a non-specific manner, thus allowing the substrate protein to find its thermodynamic minimum-energy state in a shorter amount of time [39, 134].

This model can account for the experimental data obtained to date. However, a fundamental validation of the model would require simultaneous and independent measurements of structural folding and disulfide formation in a protein undergoing PDI-catalyzed oxidative folding. Such measurements are now for the first time possible with the use of single molecule force spectroscopy.

Chapter 2 Experimental Techniques

2 Experimental Techniques

2.1 The Atomic Force Microscope

Binnig, Quate and Gerber invented the Atomic Force Microscope in 1986 [135]. A development

on the scanning tunneling microscope, the AFM quickly gained traction due to its simpler

construction and wider range of applications. AFM imaging is done by scanning a sample

relative to a sensitive probe while measuring the interaction forces between sample and probe.

Like electron microscopes, the AFM can acquire images with higher resolution than optical

microscopes. In a recent study, AFM probes ending in a single atom were used to generate the

most detailed images to date of single organic molecules [136](Figure 2.1). As opposed to

electron microscopes, however, the AFM does not require a conductive sample, and can even be

used in liquid media. In addition, the AFM probe can be used to manipulate the sample with a

calibrated force. This unique feature enables force spectroscopy, a method in which the

mechanical properties of a sample are studied through direct interaction.

Developments over the last decade have led to the emergence of single molecule force

spectroscopy [137] [27], in which a sharp probe tip is used to attach a single molecule between

the sample surface and the probe. The tethered molecule can then be studied through its response

to an applied force. This chapter will focus on the development of single molecule force

spectroscopy instrumentation, and applications to study protein biochemistry.

Figure 2.1 AFM imaging of single pentacene molecules. (a) Ball-and-stick model showing the molecular structure of pentacene. (b) STM image. (c and d) AFM images acquired using a tip ending in a single atom. Adapted from ref [136].

2.1.1 Basic principle

The AFM in its simplest form consists of a sample on a movable stage, a probe and a force detector. During a measurement, the sample is moved while the detector measures the resulting force experienced by the probe (Figure 2.2).

The movable stage typically consists of a piezoelectric actuator (piezo, for short), made from a ceramic material that physically expands and contracts in response to applied voltages. The accuracy and reproducibility required for single molecule measurements can be achieved in two different ways. Either the actuator is constructed and calibrated such that its response to a given voltage signal is completely predictable, or alternatively (and more commonly) the actuator is outfitted with an electronic sensor that records its extension in real-time. In the latter case, a strain gauge or a capacitive sensor measures the expansion of the piezo. This signal can then be used as a measurement or as part of a feed-back system for more precise operation. High quality

Piezo-electric actuator systems can today reach a reproducible spatial resolution on the order of

0.1 Å.

Figure 2.2 Basic principle of the AFM. Left: A single protein molecule is held between the sample disc and the cantilever tip. The force on the protein is measured by monitoring the bending angle of the cantilever. Right: Moving the stage down exerts a stronger force on the molecule, which in turn bends the cantilever more. This moves the laser spot downwards on the detector.

The AFM probe is a cantilever with a sharp tip. Usually manufactured from a thin sheet of glass, these cantilevers are microns to millimeters long and flexible. The tip is frequently made of silicon nitride but can be made out of a number of different materials. For single molecule force spectroscopy, the tip has to be sharp enough to ensure the attachment of a single molecule.

The cantilever deforms as it experiences a force from the sample. By detecting this deformation, the force between the tip and the sample can be measured. For minor perturbations, the cantilever behaves like a Hookean spring, where the bending angle is directly proportional to the applied force. This bending angle can be measured by reflecting a laser beam off of the cantilever and onto a photovoltaic detector. A nanometer scale deflection of the cantilever then results in a measurable signal on a standard split photodiode. Since a softer cantilever creates a larger deflection at a given force, the resolution of the force measurement scales with the cantilever softness. For a typical cantilever with a spring constant of 10 pN/nm, forces as low as

10 pN can be measured with relative ease.

The experiments presented in this thesis were performed using a custom built setup (Figure

2.3). The most important components are described in detail in the following sections.

Figure 2.3 The custom-built AFM setup. Left: Close-up view of the AFM head with the fluid cell in place. Center: The AFM head mounted on top of the piezo stage. Right: Rack-mounted control electronics including piezo controller, high-voltage amplifier and force feedback controller. The AFM can be seen in the background on top of the vibrational insulation table.

2.1.2 AFM head

The AFM module was built around a Bruker (formerly Veeco) Multi-Mode head, which provides a mechanical scaffold for a fluid cell, laser aperture and photovoltaic detector (Figure 2.3). The original detector and fluid cell were used while the laser source was replaced in order to improve beam quality. The advantage of using this commercial AFM head is its high mechanical stability and its range of adjustable settings. High quality micrometer screws allow adjustment of the incident laser beam angle and detector position with minimal mechanical drift or creep.

2.1.3 Laser

The original laser was replaced with a 670 nm fiber-coupled diode laser from

Schäfter+Kirchhoff Gmbh, with appropriate collimating and focusing optics attached to the output end of the fiber on the AFM head. This laser has improved beam stability, which decreases time dependent noise in the force measurement. In addition, the high quality of the focusing optics gives a sharper focus on the cantilever, which leads to less light leakage and thereby decreases interference arising from spurious reflections.

2.1.4 Fluid cell

A fused quartz fluid cell from Bruker was used to hold the cantilever and to provide a sealed aqueous environment during the experiments (Figure 2.4). The cell was outfitted with an iridium clip to hold the cantilever and two flow channels with syringe ports to exchange the solution in the reaction chamber. A silicon O-ring provided an air-tight seal from the environment and prevented evaporation.

Figure 2.4 Left: The AFM fluid cell shown roughly to scale. Right: Magnified view of a cantilever chip mounted in the center of the fluid cell.

2.1.5 Piezoelectric actuator

The piezoelectric actuator used was the PicoCube from the manufacturer Physik Instrumente

(PI). This 3D actuator has a range of 6 µm in each dimension and is outfitted with capacitive sensors that record displacements with 0.1 nm precision. The manufacturer provides electronic units that provide the ±250V required to reach the full range of displacements, and also units that read the capacitive sensor in real-time. While the original sensor electronics were used, the high- voltage switching amplifier was replaced due to poor signal-to-noise ratio in the high frequency regime. Since very soft cantilevers are used, even a faint noise signal can be mechanically or acoustically transmitted and influence the force measurement. Therefore, a high performance linear amplifier from Piezo Systems was used instead of the original from Physik Instrumente.

This Piezo Systems amplifier provided better signal to noise ratio and thus less interference with the force measurement. Due to the limited range of the piezo, it was placed on top of a vertical translation stage with manual control, which was used for coarse alignment.

2.1.6 Sample slide

In the single molecule experiments, a molecule was attached on one end to the sample slide and

on the other to the cantilever. The sample slide consisted of a circular glass disc with diameter of

1 cm. On top of the glass were deposited two layers of metal; a 10 nm layer of chrome/nickel

alloy and a 30 nm layer of pure gold that enabled covalent linkage of molecules (Figure 2.5). The

uncoated side of the slide was glued to a ferromagnetic disc, which could then be attached in a

magnetic mount on top of the piezo.

Figure 2.5 Sample slide with drop of solution containing the protein substrate. A pair of tweezers is indicates the scale.

2.1.7 Cantilever

Bruker MLCT triangular cantilevers were used for all experiments (Figure 2.6). These silicon nitride cantilevers are coated with gold to improve reflectivity and have an extremely soft spring constant of ~15pN/nm. The tip radius is typically around 20 nm. These cantilevers can be reused for several experiments and are discarded only after they have sustained mechanical damage or after a buildup of organic material on the tip, which can prevent proper attachment of single molecules.

Figure 2.6 Cantilevers. Left: Bruker MLCT chip showing a series of cantilevers. Cantilever C was used in all experiments presented here. Right: Scanning electron microscope image of a cantilever tip. Scalebars: 100 µµµm (left), 1 µµµm (right).

2.1.8 Data acquisition

Data acquisition and instrument control was done on a PC outfitted with a data acquisition card

(DAQ) from National Instruments. The card enables simultaneous analog and digital input and output, and is used to apply experimental protocols while recording the outputs of the AFM. The

DAQ was controlled through a custom-made control/analysis software package developed for the Igor Pro environment by Wavemetrics.

2.1.9 Supporting structures

To minimize mechanical vibration, the AFM unit is placed on top of a passive vibration isolation platform from Minus K Technologies, which in turn is perched on top of a pneumatic table from

TMC. The setup also included an optical microscope with a 10x objective coupled to a Logitech webcam CCD. This low magnification microscope was used for aligning the laser and focusing on the cantilever. During experiments, an analog oscilloscope functioned as an extra signal monitor.

2.1.10 Force-clamp

A custom-designed active force feedback system was used to apply a constant force to the

sample. This enables operation in force-clamp mode, wherein the AFM applies a constant force

to a single molecule while measuring its extension. The resulting control of the force is optimal

for investigation of force dependent processes such as protein folding and unfolding or bond

cleavage.

The force feedback consists of an electronic circuit with two inputs and one output (Figure

2.7). The inputs are the desired setpoint force from the controller (PC) and the actual force

measured by the AFM. The output, which controls the piezo displacement, is continually

adjusted so as to minimize the difference between the inputs.

Figure 2.7 The force feedback system enables constant force operation of the AFM. The measurement from the AFM and the setpoint given by the user (PC) are processed in real-time by the PID circuit and used to adjust the piezo position. Through automatic adjustment of the piezo position, the sample extension is set to achieve the setpoint force. The force measurement and the piezo position signal are recorded.

As an example, consider a single protein molecule subjected to a stretching force in the AFM.

The protein is attached between the cantilever and a sample slide. Let us assume that initially, the force on the protein is 10 pN while the desired force is set to 100 pN. The protein thus exerts 10 pN on the cantilever, which in turn bends to some small degree. This bending is measured by the photo detector and read by the feedback system as 10 pN. Since this force is too low, the feedback system adjusts the output controlling the piezo, so as to move the sample slide further

away from the cantilever, thereby stretching the protein. This causes the protein to experience a

higher force that in turn is registered by the cantilever and the detector. In this way, the protein is

extended until it experiences a force equal to the desired force. The readout from this experiment

is the output of the feedback, which corresponds to the extension of the protein. Alternatively,

the output of the piezo sensor can be used for higher precision, since this signal circumvents

nonlinear properties of the piezo. The actual measured force can also be recorded, in order to

verify the accuracy of the feedback.

The force feedback system used is known as a negative feedback circuit, and a simple implementation can be built with regular analog electronic components including a few operational amplifiers. The force-clamp electronics rely on PID feedback, which stands for

Proportional-Integral-Derivative. The PID calculates the error signal e(t) as the difference between the two input signals, and continuously generates an output u(t) that depends on the error signal as follows:

/ = + + ͘ ͩʚͨʛ ͅ+͙ʚͨʛ ͅ$ ǹ ͙ʚʛ͘ ͅ ͙ʚͨʛ ͤ ͨ͘

Where Kp, Ki and Kd represent gain factors for the proportional, integral and derivative components of the signal, respectively. These factors can be adjusted manually by altering the resistance of three variable resistors placed in the circuit. For any given physical system, such as a single molecule attached to an AFM probe, the feedback characteristics depend on the value of the gain factors.

A common way to characterize the feedback is in terms of its stability, response time and overshoot. Of these properties, the stability is the single most important feature, as it determines whether the system converges, i.e. whether the desired force can be reached. If not, then the

output signal may oscillate and/or diverge and reach saturation. In either case, the outcome

represents failed control. Oscillation is more likely if the gain is high, and this is often what

limits the gain factors.

The response time is a measure of how fast the feedback recovers after a perturbation.

Assuming it is stable to begin with, the control system may encounter sudden change in one of the inputs. For instance, the setpoint force can be changed or the molecule under study can undergo a structural transition. When this happens, the feedback attempts to recover under the new conditions and reach a new stable state. A faster response time is almost always desirable. In particular during single molecule force spectroscopy, the response time determines what experiments can be performed. In addition, the response time is more important when the experimental system is more dynamic, as it limits the fastest rate that can be measured. The gain factors are therefore set to minimize the response time. Higher gain results in faster response time, but can also compromise the system stability, as explained above. As a consequence, the optimal feedback parameters are those that give a minimal response time while still maintaining stability.

In practice, the values of the optimal gain parameters vary between experimental systems.

They depend on the specific properties of the piezo and the cantilever used, but also on the molecule under study. In some cases, values that are optimal during one part of an experiment could be unstable at a later time point during the same experiment, for example after a change in the setpoint force or even after a structural change in the studied molecule. In the experiments described here, the feedback parameters were set as a compromise, minimizing response time while ensuring stability under all experimental conditions used. The chosen parameters yielded

response time constants that were generally less than 5 ms (Figure 2.8). Rates of up to 200 s -1 could therefore be measured under constant force.

Figure 2.8 Force and length measurements from a single I27 protein molecule held at a setpoint stretching force of 160pN. An abrupt extension of the molecule (due to an unfolding event) at time zero caused the force to drop. This caused the feedback system to extend the protein until the setpoint force was restored. The force was restored with a time constant of 2.5 ms, at which point the protein molecule had extended by 25 nm.

Overshoot can be observed if the feedback overcompensates and passes the setpoint before stabilizing. This phenomenon could potentially introduce artifacts in the measurements, but was never observed at the feedback parameters chosen.

2.1.11 Calibration

The two measurements provided by the AFM, displacement and force, both require accurate

calibration. Displacement is set through the piezo, which has a calibration provided by the

manufacturer. The piezo calibration is given in units of displacement per voltage readout from

the capacitive sensor, in nm/V.

The force on the AFM cantilever is measured by a split photodiode that detects the position

of the reflected laser beam. As the light impinges on the two halves of the detector, each half

reports a voltage signal that is directly proportional to its incident light intensity. These two

signals, A and B, are then processed in an electronic circuit to yield the normalized difference

signal, VD=(A-B)/( A+B). The difference signal is directly proportional to the displacement of the

cantilever and thus to the force.

There is a linear relationship between the difference signal and the actual force under the assumption that the cantilever deflection is small. How small? First of all, the Hookean relationship between the force and the displacement of the cantilever only holds for small angles of deflection, which for Bruker MLCT cantilevers was measured to be up to a few hundred nanometers of tip displacement. Second, the displacements of the laser spot on the detector have to be much smaller than the size of the spot itself. As the spot can be several millimeters in diameter while the spot displacements are on the micrometer scale, this is generally less of an issue.

In order to calculate the scaling factor between the difference signal and the force, two parameters need to be measured. The first one is the spring constant of the cantilever ( kC, given in pN/nm) and the second is the scaling factor between the displacement of the cantilever and the difference signal on the photodiode (kPD , given in nm/V). With knowledge of these two parameters, any difference signal VD can be converted to force through the relationship:

= ∙ ∙ ̀ ͐ ͟ ͟

This relation assumes that the laser is centered on the photodiode (and thus that VD=0 ) when the

force is zero.

Although the manufacturer provides an approximate value of kC for the cantilevers, the real

values were found to deviate by up to 50% from this nominal value. The variation is most likely

due to inconsistencies in the manufacturing process. For these reasons, kC was determined

experimentally for each cantilever used in the experiments, together with a calibration of the kPD parameter.

The thermal tune method was used to calibrate the force measurement of the AFM [138].

Two independent measurements were taken to establish the value of the two unknown parameters, kC and kPD . First, kPD was determined by displacing the cantilever tip by a known amount ∆x (using the calibrated piezo) and measuring the resulting difference signal ∆VD. Then,

the value of kPD is calculated as:

∆ = ͐ ͟ ∆ ͬ

The value of kC can be determined through use of the equipartition theorem, which states that the

kinetic energy of each degree of freedom in a system contains an equal amount of thermal

energy. When applied to the primary vibrational mode of the cantilever, this yields:

ͥ ͦ = ͥ ͦ͟〈ͬ 〉 ͎ͦ͟

2 Where kB is Boltzmann’s constant, T is the temperature and is the mean squared displacement of the cantilever. Recognizing that x can be expressed in terms of the difference signal VD, the spring constant can then be determined as:

∙ ͦ = ͎͟ ʚ͟ ʛ ͟ ͦ 〈͐ 〉

2 These parameters are all known except for , which can easily be measured when the

2 cantilever is placed resting in solution. However, the formula only holds if < VD > is measured

exclusively for the primary vibrational mode, i.e. for the ground harmonic. Meanwhile, a raw

measurement of the cantilever displacement includes all resonant harmonics. The vibrations

pertaining to the ground harmonic need to be isolated somehow. This can be accomplished by

applying a Fourier transform to the data, which generates vibrational amplitudes given as a

function of frequency ( ). Resonant modes of the cantilever are easily separated in the frequency ! 2 domain (Figure 2.9). Moreover, the value of < VD > can be calculated from the transformed data with the use of Parseval’s theorem:

ͦ = ͦ ǹ|ͬʚͨʛ| ͨ͘ ǹ|ͬȬʚ!ʛ| ͘!

2 Where is the Fourier transform of x. The value of for the ground harmonic can therefore ͬȬ be calculated as:

hv ͦ ͦ = 〈͐ 〉 ǹhu |͐ ʚ!ʛ| ͘!

Where and are chosen to encompass the first resonant peak (Figure 2.9). !ͥ !ͦ

Figure 2.9 Frequency spectrum showing thermal fluctuations of a cantilever at rest in solution. The spring constant is measured from the amount of thermal energy in the first resonant peak of the spectrum, i.e. the ground harmonic.

The calibration procedure is an integral part of each experiment. Not only does it yield the specific force scaling parameters for the cantilever used, but it also functions as a diagnostic for the instrumentation. By benchmarking the obtained values against the expected range, any errors or malfunctions can be caught before the actual data collection. Once the calibration is over, the single molecule measurements can begin.

2.2 Protein engineering

Single molecule force spectroscopy benefits from the ability to design specific protein substrates.

Engineered polyproteins were used in the experiments presented in this thesis, because these allow for the unambiguous identification of single molecule attachment, as described in Section

32-75 2.3.8. Two different polyprotein substrates were used: (I27) 8 and (I27 )8 (Figure 2.10).

Figure 2.10 Molecular structure of the two variants of the I27 protein used in this thesis. Left: Wild-type I27. The two native cysteine residues (Cys47 and Cys63) are shown in yellow. The cysteines are too far apart (9.2 Å) to form a disulfide in the native structure. (PDB: 1tit) Right: I27 32-75 . The two engineered cysteines in positions 32 and 75 are close enough in the native state to form a disulfide bond (structure obtained by homology modeling).

The design, expression and purification of these proteins have been described earlier [26, 139].

Both proteins contain an N-terminal His tag used for purification, and two cysteines at the C- terminal for covalent attachment to the gold surface. In addition, the individual protein domains

in the chain are spaced apart by two amino acids (Arg-Ser), which minimizes interactions

between domains. The full polypeptide sequences of the polyproteins are:

N’ -MRGS(H) 6GS(I27-RS) 8(C) 2-C’

and

32-75 N’ -MRGS(H) 6GS(I27 -RS) 8(C) 2-C’

Where I27 represents the wildtype sequence of the 27 th Ig domain (89 amino acids) in human

cardiac titin [140](Figure 2.10, left structure), and I27 32-75 represents the same domain but containing an engineered disulfide between residues 32 and 75. In I27 32-75 , the two native

cysteines Cys47 and Cys63 were mutated to alanine (these residues are distal in the native

structure and therefore do not form a disulfide). Residues Gly32 and Ala75 (which are proximal

in the native structure) were mutated to cysteine and were seen to form a 32-75 disulfide

[26](Figure 2.10, right structure).

The polyproteins (and also most of the enzymes) used in this work were expressed in E. coli and purified as described in the following section.

2.2.1 Protein expression and purification

The polyprotein genes were inserted into pQE80L (Qiagen) expression vectors that enable inducible expression by IPTG. E. coli cultures of strain BLR(DE3) (Novagen) were transformed

and used for large scale expression.

After inducing expression of the polyprotein, cells were harvested and lysed using a

combination of enzyme treatment (lysozyme, DNase and RNase), sonication and high-pressure

treatment in a French pressure cell press. The soluble fraction of the lysate was separated using

centrifugation and then filtered to remove particulate matter.

Purification of the polyprotein was accomplished in two steps. First using a Co 2+ Talon resin affinity column, in which polyprotein was separated based on the affinity of the His tag to the metal ions. Because the eluted fraction still contained some contaminants, typically other His rich proteins, a second purification step was used to obtain a pure protein sample. In this step, an

Äkta FPLC system with a Superdex 200 HR column (GE Biosciences) was used for size exclusion gel filtration chromatography. Protein concentration in the eluate was monitored through the absorbance at 280 nm. The polyproteins were readily discernible as a peak in the chromatogram, and fractions were collected within the expected size range. The fractions were evaluated using SDS PAGE, and the fraction with the highest protein concentration was stored and used in experiments.

2.3 Single Molecule Force Spectroscopy

The workflow of a single molecule force-clamp experiment will be described in the following sections. Before starting the actual experiment, the AFM has to be calibrated and the protein samples have to be prepared. When the setup is ready, the focus shifts to executing the experimental protocol. These protocols consist of a series of sequential force pulses. Through the design of creative protocols, a wide range of different experiments can be executed. The following sections will review the effect of mechanical force on proteins, as kind of information that can be extracted through mechanochemical methods.

2.3.1 Protein mechanics under force

More than a decade of work has established the effect of force on protein molecules [7, 19, 141-

146]. Before considering a folded protein, it is useful to review the basics of polymer physics.

All polymers share certain properties. For instance, while free monomers diffuse rapidly at ambient temperatures and do not aggregate unless there is a strong attractive potential, the simple act of linking them together changes the situation entirely. From an entropic point of view, full extension of a polymer is only represented by a few conformational states. The entropy is instead maximal in a relatively contracted state, where the end points are on average much closer together, which makes the extended state disfavored under equilibrium conditions. However, the aversion of a polymer towards extension can also be understood in a more direct way.

Consider the two endpoints of a completely stretched-out polymer chain. Any thermally induced movement of the segments between these endpoints would tend to bring the endpoints closer together. For instance, in aqueous solution the polymer is continuously bombarded with water molecules. Every such collision creates a tiny contractive force impulse, but when added together this force can become quite considerable. The magnitude of this contractive force abates as the polymer contracts, because not all collisions contract the chain when the segments are not aligned anymore. In practice, the described effect gives rise to a force-extension relationship that accounts well for the entropic behavior of a non-interacting polypeptide molecule in the force range between a few and few hundred piconewtons [146].

The worm-like chain (WLC) is a commonly used physical model that relates the end-to-end length of a polymer (x) to the entropic contractile force (FWLC ):

1 ͯͦ 1 = ͎͟ 1 − ͬ − + ͬ ̀ ʚͬʛ ʨ4 ƴ Ƹ 4 ʩ ͤ ͆ ͆

Where LC is the contour length of the polymer (i.e. the sum of the length of all segments) and p is the persistence length, a polymer-specific property that describes its intrinsic flexibility.

The most important feature is that the force increases to infinity as the end-to-end length of the polymer asymptotically approaches its contour length, or simply put, it would take an infinite force to extend the polymer to its full length (Figure 2.11). The WLC model has been shown to accurately describe the entropic elasticity of both proteins and nucleic acid polymers [146, 147].

Figure 2.11 Worm-like chain model showing the extension of an entropic chain as a function of stretching force. This model is a good description of an unfolded and extended protein. ( p=0.4 nm).

In a folded protein there are other forces at play besides entropy. Hydrophobic and electrostatic interactions, hydrogen bonds, salt bridges and van der Waals interactions all contribute to the formation of the native state during protein folding. Adding all of these factors to the interaction potential creates an enormously much more complex situation, which explains why protein folding is a process that still today cannot be modeled accurately. Nevertheless, whereas the entropic contributions to the force increase in magnitude as the polypeptide is stretched out, all enthalpic interactions fall off with distance. Considering a typical Debye length of less than 1 nm

50 under physiological conditions, only a small fraction of the enthalpic interactions in a folded protein play any role in an extended polypeptide. Even hydrophobic interactions have a range of less than a few nanometers. An unfolded protein therefore behaves as an entropic polymer in the high-force regime.

The proteins considered in the experiments presented here all have well-defined and mechanically stable folded structures [140, 145, 148]. This implies that from a thermodynamic point of view, the folded state represents a global minimum in the free energy of the protein.

This minimum is furthermore separated from any other state by a large energy barrier. In the absence of any perturbation, the notion of the unfolded state is poorly defined. This situation changes in the presence of mechanical strain. When a stretching force is applied to the termini of the folded protein, the protein first tends to align with the direction of that force. This alignment process is referred to as the initial extension [17]. Once the protein is aligned with the force, any subsequent reactions will take place over a defined reaction coordinate: the length of the protein along the axis of the pulling force, which is usually referred to as the end-to-end length of the protein. Since single molecule force spectroscopy enables direct measurement of this variable, it is convenient to express the free energy of the protein along this axis.

2.3.2 Mechanical protein unfolding

A folded protein has an energy minimum at an end-to-end length LF of a few nanometers (Figure

2.12). No other minima exist at higher extensions in the absence of force, due to the associated entropic costs [141]. However, when a pulling force is applied, any extension of the protein along the axis of the force will add a work term to the system. In the case of a constant force, this term is equal to –Fx, where F is the applied stretching force and x is the extension of the protein.

Adding this linear term to the intrinsic protein potential effectively “tilts” the energy landscape

of the protein. When the force is sufficiently large, this tilt creates an additional energy minimum

at an end-to-end length of LU (Figure 2.12). The location of this minimum is at an extension where the stretching force equals the contractile entropic force, F=F WLC . At this minimum, the

protein is unfolded and extended so far that very few, if any, of its native contacts remain intact.

This has been directly verified by showing that the protein behaves as an entropic chain in its

extended state [146]. The end-to-end length of the protein in its unfolded state can therefore be

accurately described by the WLC model (Figure 2.11).

Figure 2.12 The free energy of a protein as a function of end-to-end length. At zero force (yellow trace), the folded state represents an energy minimum at an extension of a few nanometers. No minima are present at high extensions in the absence of force, due to the associated entropic cost. A stretching force tilts the energy landscape (blue trace) and creates a new, unfolded, state at an extended length.

Before proceeding, it is important to note that the mechanically unfolded state described here is radically different from unfolded states sampled in non-mechanical unfolding experiments.

When a protein is denatured using urea, guanidinium chloride, temperature or pressure, the

protein remains in a relatively unextended state due to entropy, even when all of the native

interactions have been severed [149]. On an end-to-end scale such as in Figure 2.12, these states

would therefore all appear at an extension of approximately LF. Meanwhile, the mechanically unfolded state does not exist in the absence of force. What is the relevance of the mechanically unfolded state? I would argue that while it remains questionable to what extent proteins in our body encounter excessive amounts of urea; mechanical forces undoubtedly perturb proteins in vivo . A fundamental understanding of mechanical unfolding is therefore vital to our

understanding of a range of processes from cell adhesion to muscle function.

The mechanically unfolded state is separated from the folded state by an energy barrier, ∆E.

The height of this barrier is roughly equal to the enthalpic interactions that keep the protein folded. Creation of a stable extended state enables the protein to unfold in a two-state manner that is all-or-none because of the barrier. Unfolding occurs with a thermally driven rate kU.

Increasing the force further stabilizes the unfolded state and shrinks the energy barrier by F∆x,

where ∆x is the distance from the folded state to the transition state. The effect of force on the unfolding rate can be accurately described by the Bell model [150]:

ͯ(∆ͯ∆3) Ɵ ∆3 = &û = Ɵ&û ͟ʚ̀ʛ ̻͙ ͙ͤ͟

Here, k0 denotes the unfolding rate extrapolated to zero force (since this rate is only extrapolated,

it does not necessarily have a physical meaning).

In single molecule force spectroscopy, an unfolding event can be detected as a stepwise

increase in length of the protein [145] (Figure 2.8, Figure 2.13). The size of this step ∆LU=L U-LF,

is highly protein specific. The studies in this thesis use polyproteins consisting of a chain of

repeated protein domains. Like the pearls in a pearl necklace, these individual domains do not

interact with each other to the extent that it influences the measurement, and under force they

unfold in independent steps with length ∆LU (Figure 2.13). Assuming ergodicity, we can then extract kinetics by averaging over events within a polyprotein as well as over events in separate molecules.

Figure 2.13 Mechanical unfolding of a wild-type I27 polyprotein. The individual domains unfold independently in 25 nm steps.

The Bell model suggests that the rate of unfolding increases exponentially with the magnitude of the applied force, an observation that has been verified experimentally [144](Figure 2.14).

Figure 2.14 Semi-log plot of ubiquitin unfolding showing that the rate increases exponentially as a function of stretching force.

If the applied force is increased sufficiently it will reach the limit where F∆x =∆E. At this point, the energy barrier is gone and the folded state ceases to be an energy minimum. The protein will consequently unfold in a downhill manner with a rate that is limited only by the diffusion rate.

Taking together all of these observations allows for the interpretation of single molecule force-clamp data (Figure 2.13). At low forces, the protein aligns with the force but remains folded, yielding only an extension equal to the length of the folded structure. At higher forces, there is a chance that the protein unfolds. Unfolding of a protein domain can be detected as an elongation step of ∆LU that scales with force and the contour length of that domain. An unfolding

step is only seen if the protein is in a stable folded state to begin with. If not, then the protein

would unravel to its unfolded length even at low forces, effectively adding this length to the

initial extension. The discrete step of ∆LU can therefore serve as an accurate fingerprint signal of a folded protein. Single molecule force-clamp can thus be used as a powerful tool to control and detect the unfolding of single protein molecules.

2.3.3 Force based assays of protein folding

As described above, single molecule force spectroscopy can be used to detect folded proteins and

also to induce protein unfolding. By combining these two features, the technique can effectively

be used to study protein folding [17, 19, 146]. The basic strategy is as follows: A single protein

molecule is first unfolded mechanically by applying a stretching force high enough to induce

unfolding. During this process, the initially folded state of a domain is verified as an unfolding

step ∆LU. The force is then removed and the protein is left to refold for a set amount of time.

After this refolding interval, a stretching force is applied again. If the protein has successfully

folded during the refolding interval, an unfolding step ∆LU is again detected during this second

force pulse. If, instead, the protein failed to refold during the refolding interval, then the initial extension of this pulse will encompass the fully unfolded length of the protein (Figure 2.15).

The probability of folding PF after a time ∆t can be measured by repeatedly performing this experiment and then calculating the refolding ratio:

∆ = ͊ʚ ͨʛ ͈- !*' ⁄͈/*/'

Where Nrefold and Ntotal are the number of refolded domains and the total number of domains

probed, respectively.

Figure 2.15 Mechanical measurement of protein folding in a wild-type I27 polyprotein. Four domains were initially unfolded under force ( Ntotal =4). After 2 seconds, the force was switched off, which led to rapid collapse of the protein. After an interval of 5 s in the absence of force, the force was switched on again. This time, two unfolding steps were seen, showing that two domains had refolded ( Nrefoldl=2). Recovery of the same unfolded length shows that the same protein was probed in the two force pulses.

2.3.4 Disulfide detection using force

Besides the protein backbone, most of the bonds that contribute to the folded structure of a protein are non-covalent. From a mechanical point of view, the strongest non-covalent bonds in proteins are hydrogen bonds. Depending on their degree of covalence, a single hydrogen bond can withstand up to a few piconewtons before it breaks. However, some proteins feature a

“mechanical clamp” that distributes an applied force between several hydrogen bonds [151].

Distributing the force in parallel increases the resilience of the folded state beyond the strength of any single hydrogen bond, and allows some proteins to remain folded for seconds at forces of over a hundred piconewtons [144, 145].

However, a notable exception among structural protein bonds is the disulfide bond that can

be formed between cysteine residues. While non-covalent bonds break at a force of a few

piconewtons, disulfide bonds are much more resilient and can withstand forces of several

nanonewtons [28]. The stability imparted on proteins by disulfide bonds plays an important role

in biology but also renders them easily detectable through force spectroscopy [148, 152]

In a basic experiment to detect the presence of a disulfide bond, a protein is subjected to a force high enough to unfold it but not high enough to cleave the disulfide. The length of the unfolding step, ∆LU, then depends on whether or not the disulfide is present. If there is no

disulfide, then the unfolding step will be that expected from a normal unfolding event. If there is

a disulfide, however, then the step size upon unfolding will be diminished by a length

corresponding to the polypeptide segment sequestered by the bond. In other words, the protein

will unfold only up to the disulfide, leaving a segment still folded (Figure 2.16). This effect of

disulfide bonds on mechanical protein unfolding has been reported by a number of studies [148,

152-157].

Figure 2.16 Presence of a disulfide bond gives rise to shorter unfolding steps. Top: A reduced I27 32-75 polyprotein unfolds in distinct 25 nm steps. Bottom: Oxidized I27 32-75 domains unfold in 11 nm steps. The disulfide is blocking the extension of the remaining 14 nm of polypeptide chain.

2.3.5 Disulfide cleavage under force

In a study from 2006, Wiita et al demonstrated the detection of single disulfide cleavage reactions under constant force conditions [26]. The assay introduced in this study set the stage for a new field of biophysics, where the effect of force on covalent chemistry could be studied at the single bond level. The authors used an engineered polyprotein consisting of 8 sequential repeats of the I27 domain from human cardiac titin. Each of these I27 domains contained two cysteines introduced at positions 32 and 75, which formed a disulfide. By applying a constant force of 130 pN to the I27 32-75 polyprotein, individual unfolding steps of 11 nm could be

detected. By contrast, wild-type I27 domains that do not contain disulfide bonds unfold in steps

of 25 nm. This difference of 14 nm clearly indicated the presence of disulfide bonds in the folded

protein (Figure 2.16). Cleavage of the disulfide bonds was not observed in the absence of

reducing agents.

When a disulfide reducing agent was added to the reaction solution, 11 nm unfolding steps

were still observed as opposed to 25 nm steps, indicating that the 32-75 disulfide is inaccessible

in the folded state. (If, on the other hand, the disulfides were exposed in the folded proteins, then

some disulfides would have been reduced before the experiment started and 25 nm steps would

have been observed). However, in the presence of the reducing agent, 14 nm steps were observed

after the 11 nm steps (Figure 2.17). The size of these steps corresponded exactly to the expected

extension upon release of the disulfide (25 minus 11). These 14 nm steps thus indicated single

disulfide cleavage reactions. Apparently, partial unfolding as indicated by 11 nm steps exposed

the disulfides to the surrounding media, which then rendered them susceptible to cleavage by the

reducing agent.

Figure 2.17 Direct detection of single disulfide cleavage reactions using force-clamp AFM. A disulfide bonded polyprotein is picked up by an AFM cantilever and mechanically stretched at constant force of 150 pN. The cartoon shows the detection scheme. The polyprotein is first extended by unfolding, right up to the disulfide bond. Each such unfolding event leads to an 11 nm elongation step. Unfolding of a domain exposes the disulfide to nucleophiles (in this case TRX) present in the surrounding solution. Cleavage of a disulfide results in a 14 nm elongation step, which can easily be detected using the AFM. An experimental recording is shown (red trace), with disulfide cleavage events marked by arrowheads.

The ability to control the reactivity of the disulfides meant that the kinetics of the reaction

could be studied in high detail. The results of this study showed that the effect of force on the

kinetics of disulfide reduction was well described by an Arrhenius term, much like protein

unfolding:

∆3 = Ɵ&û ͟ʚ̀ʛ ͙ͤ͟

Where kR0 represents the reduction rate at zero force. In the case of disulfide reduction by DTT,

the value of ∆x was measured to be 0.34 Å. This value was interpreted as the lengthening of the

bond at the transition state of the S N2 reaction [26].

2.3.6 Setup of an experiment

Polyproteins were used in the experiments presented here (see Section 2.2). These proteins consist of sequential repeats of individually folded domains, with short unstructured regions linking the domains. In order to covalently link the molecule to the sample slide, two free cysteine residues are appended to one end of the polypeptide chain. These cysteines are adjacent in the primary sequence and therefore sterically unable to form a disulfide with each other [30].

However, the cysteine thiols can readily form covalent thiol-gold bonds with the surface of the sample slide.

Initially, a freshly evaporated sample slide is mounted on the piezo and cantilever chip is mounted in the fluid cell using a pair of fine tweezers. The fluid cell is mounted in the AFM head, which is positioned on top of the low magnification microscope (Figure 2.3). With help of the microscope, the laser spot can be aligned on the tip of the cantilever. The fluid cell is then removed from the AFM head and a small volume (~20 µL) of solution containing the polyprotein

61 is added, immersing the cantilever in a drop. After making sure that no bubbles have been formed, the fluid cell is then placed back in the AFM head and placed on top of the piezo. Once there is liquid contact between the fluid cell and the sample slide, the calibration procedure begins, as described in Section 2.1.11. When the calibration is finished, the experiment is started.

Each experimental measurement starts with an attempt to pick up a molecule from the sample slide. This is accomplished by pushing the cantilever tip against the surface at a force of

1-2 nN for 0.5-2 s and then retracting. If no molecule was attached then the process is repeated.

When attachment has been achieved, the experimental pulse protocol is initiated. After a completed pulse protocol, the pulling force is increased until the attachment is broken. The data is stored and the process then begins anew. If an additional reagent such as an enzyme needs to be added to the experiment, this is done only after the polyprotein sample has been left to bind to the sample surface for a sufficient amount of time. This time can range anywhere from minutes to hours, but is easily determined by probing the surface with the AFM probe.

Once an experiment is started, it can be set up to run fully automatically for extended amounts of time. With a stable sample and a properly sealed sample cell, an experiment can run for hours or even days without need for intervention. Throughout this time, data is collected and selectively stored based on a few basic requirements. At the end of an experiment, there is usually a large amount of data in need of analysis.

2.3.7 Pulse protocols

Creative design of pulse protocols has enabled the study of a range of different protein processes, including protein folding and unfolding, and disulfide reduction. In addition, a new technique is introduced here, which enables the study of disulfide formation.

The pulse protocols designed to probe disulfide reduction generally consist of two sequential

force pulses: [ unfolding–reduction ] (see for instance Figure 5.2). The unfolding pulse is set at a

force high enough to ensure rapid unfolding of all protein domains under study, yet low enough

to resolve each individual unfolding event. The rapid unfolding of all domains in the substrate

effectively synchronizes the exposure of the intramolecular disulfide bonds, which simplifies the

analysis of reduction kinetics. After a time sufficient to ensure complete unfolding of the

substrate, the force is changed for the reduction pulse. This force can be set to any value, as long

as it is high enough to allow detection of reduction events (at least 30 pN). By repeating this

experiment with different reduction forces, the effect of force on reduction kinetics can be

studied.

The pulse protocols used to study disulfide formation generally consist of three sequential

force pulses: [ denature–∆t–probe ] (Figure 2.15). The denature pulse is here used to completely

extend the substrate protein, while characterizing its initial state. After a refolding interval ∆t, the substrate is probed by again applying a force. If any disulfide bonds have formed during the refolding interval, they are easily detected as they restrict the extension of the substrate during the probe pulse.

2.3.8 Data Analysis

Single molecule experiments tend to generate an abundance of data. In bulk biochemistry, a measured signal usually arises from a combination of different molecules, some of which may not be the ones of interest. By contrast, a single molecule measurement is more binary in the sense that it either captures the molecule of interest or not. If there is an independent way to determine when the correct molecule is being measured, then “bad” measurements can be

excluded from the analysis. Excluding all incorrect measurements is in practice equivalent to

performing the experiment with a 100% pure sample, making data selection a central feature in

single molecule experimentation. However, unambiguous criteria need to be established in order

to ensure that this data selection is unbiased.

One embodiment of such a criterion is the notion of a molecular fingerprint, which has

become common practice in force spectroscopy. A molecular fingerprint is a distinct signal that

can only arise when the molecule of interest is captured. In all experiments presented here, the

usage of polyproteins provided such a fingerprint. When a single protein domain unfolds under

force, the unfolding step ∆LU is highly protein specific [152]. A polyprotein consisting of serially repeated identical domains therefore gives rise to a staircase of identical steps, as each domain unfolds independently after a stochastic amount of time [145]. Since the probability of any other protein generating such a staircase is negligible, this fingerprint verifies that the molecule under study is the one of interest. In practice, the data selection is accomplished by only taking into account measurements that include a minimum number of identical steps (usually 3 or more).

In the refolding experiments, a protein is left in the absence of force during the refolding interval. During this period of time, the cantilever tip may come in contact with the sample slide.

This could potentially lead to the pick-up of a second molecule that would be stretched in the probe pulse. For this reason, a second fingerprint was used in the selection of refolding traces.

This fingerprint was the fully extended length of the polyprotein molecule, as measured at the

end of the denature pulse. Since each molecule is picked up from a different point, this length shows significant variation between individual measurements. Therefore, only data where the same length was recovered at the end of the denature and probe pulses were considered for further analysis.

Chapter 3 Single Molecule Study of Oxidative Protein Folding

3 Single Molecule Study of Oxidative Protein Folding

The following study introduces an entirely novel approach to the study of oxidative folding and presents results that explain how enzymes such as Protein Disulfide Isomerase (PDI) introduce disulfide bonds into folding proteins.

A third of human proteins contain disulfide bonds, including most secretory and cell-surface

proteins. A majority of these proteins acquire their disulfide bonds in the endoplasmic reticulum

during oxidative folding, an essential process catalyzed by PDI [39]. While PDI has recently

been implicated in several protein-misfolding diseases [37, 158, 159], our understanding of how

this enzyme catalyzes oxidative folding has been hampered by the lack of precise analytical

methods. Current assays of oxidative folding, whether genetic or biochemical, reveal very little

detail and are often subject to ambiguity. The technique presented here yields more than an

incremental advance, as we can now for the first time measure protein folding and disulfide

formation separately in the same protein. By probing single proteins caught in the process of

folding, we can count the number of bonds that are formed at any given time. This practically

eliminates interpretational ambiguity and allows us to draw strong conclusions on the

mechanisms of oxidative folding in the cell.

It has been unknown how PDI “knows” which cysteines to join in proteins with more than

two cysteine residues. Formation of incorrect disulfides can be pathogenic by triggering

misfolding and aggregation. In addition, formation of a disulfide can be irreversible if the bond

becomes buried inside the protein. We discovered that PDI actually does not “know” which

cysteines to join together. Instead, the enzyme relies on the natural folding pathway of its

substrate to guide the pairing of cysteines. While the substrate searches for its native

conformation, PDI remains covalently attached to one of its cysteines and enables disulfide

formation once the correct cysteine partner has been brought into place. This remarkable finding

can explain how one enzyme can catalyze the oxidative folding of a wide variety of proteins.

This study presents an enabling technology and brings us closer to a complete description of how proteins are formed.

3.1 Physiological significance of PDI activity

Protein folding is a labile reaction that can be easily affected by a multitude of factors such as mutations, and changes in the physical or chemical environment. Failure to correctly fold is the basis of numerous disorders of central importance to modern medicine [160]. In particular, the third of human proteins that traverse the secretory pathway and that possess disulfide bonds pose unresolved challenges to our understanding of protein folding and disease [81, 161, 162]. Protein

Disulfide Isomerase (PDI) is the main catalyst of oxidative folding in humans [39]. Recent studies have revealed a link between disulfide chemistry and the pathogenesis of misfolding diseases, and specifically implicated PDI as a novel target for treatment of several neurodegenerative disorders including Alzheimer’s disease [37, 158]. These studies stress the importance of understanding how PDI catalyzes oxidative folding.

PDI belongs to a ubiquitous family of enzymes that catalyze thiol-disulfide exchange [39].

These oxidoreductase enzymes share a characteristic structural fold and a highly conserved Cys-

X-X-Cys motif in their active sites [46]. In addition to PDI, the family includes other oxidoreductases such as thioredoxin, glutaredoxin and the bacterial Dsb enzymes [46]. The mechanism of action of these enzymes has been revealed through numerous studies over the past forty years. In all cases, the reaction mechanism involves the formation of a mixed disulfide between a cysteine in the substrate and the N-terminal cysteine in the active site of the enzyme

[53, 163] (Figure 3.1). The C-terminal cysteine can attack and cleave the mixed disulfide, thereby spontaneously releasing the enzyme [39, 80]. Whereas spontaneous release is necessary during reduction of substrate disulfide bonds, it is unknown how this activity affects catalysis of oxidative folding.

3.2 Oxidative folding pathways in cells

Secretory proteins are synthesized as linear polypeptides and emerge from the ribosomal channel via the translocon into the endoplasmic reticulum [83-85]. Emerging sequentially into the ER, the nascent polypeptide encounters PDI, which catalyzes co-translational oxidative folding [86,

87]. This reaction is mediated by the formation of a mixed disulfide bond between the PDI enzyme and a cysteine in the nascent polypeptide (Figure 3.2A) [29, 88, 89]. The mixed disulfide is then transferred to the folding polypeptide. Given the crucial roles of mixed disulfides in oxidoreductase catalysis, many studies have been focused on these ephemeral intermediates. The molecular structure of mixed disulfide complexes have been reported for several enzymes [59, 61, 62, 164, 165]. In addition, mixed disulfide complexes in the process of oxidative folding have been characterized in living cells [43-45, 87]. While all of these studies have provided highly detailed snapshots of mixed disulfide complexes, their dynamics during oxidative protein folding remain unknown. In order to study the intersection of covalent chemistry and protein folding, we need a method that can measure these two concurrent processes independently.

Figure 3.1 Summary of reactions catalyzed by PDI. (A) Thiol-disulfide oxidoreductases in the thioredoxin superfamily, including PDI, catalyze disulfide exchange by forming a mixed disulfide intermediate complex with the substrate (blue). (B) Pathway for disulfide oxidation by PDI. (C) Pathway for disulfide reduction.

Figure 3.2 A single molecule approach to study oxidative folding. (A) As part of the secretory pathway, protein disulfide isomerase (PDI) is thought to form mixed disulfide complexes with nascent polypeptides (blue) undergoing ER translocation. Such complexes can be resolved as the protein completes translocation and oxidative folding. (B) In this study, an Atomic Force Microscope (AFM) was used to establish mixed disulfide complexes between PDI enzymes and a single extended protein. Starting from this state, we investigated how PDI catalyzes oxidative folding. (C) Mixed disulfide complexes were formed by applying a constant stretching force to a folded protein containing a buried disulfide, thus unfolding the protein and exposing the disulfide. A reduced PDI enzyme could then form a mixed disulfide with one of the cysteines in the substrate (see also Figure 3.1). (D) In an experimental recording of the end-to-end length of the I27 32-75 substrate under force, unfolding of the substrate was detected as an 11 nm extension step. Formation of a mixed disulfide was detected as a 14 nm extension step arising from the cleavage of the 32-75 disulfide in the substrate.

3.3 Spontaneous disulfide oxidation kinetics

We used in our experiments the catalytic domain A1 from human PDI, which is sufficient for

catalysis of oxidative folding [77]. The substrate we chose was a polyprotein consisting of

repeats of the 27th Ig domain from human cardiac titin, with each domain containing two

cysteines in positions 32 and 75, which can form a disulfide. We chose this substrate (hereafter

referred to as I27 32-75 ) because it has been characterized extensively in previous investigations.

Several studies have described its folding and unfolding pathways [25, 26, 143, 148], which makes it an ideal substrate for studying the effects of disulfide exchange on protein folding.

We first purified the substrate under reducing conditions. After removing the reducing agent from the buffer, we investigated the structure of the substrate using force-clamp AFM. The top panel in Figure 2.16A shows a representative trace of the refolding properties of this pre-reduced

I27 32-75 . In the denature pulse, each substrate domain unfolded independently in a single step of

25 nm, indicating that none of the domains initially contained disulfide bonds. The recurrence of

25 nm steps in the probe pulse shows that on the timescale of a recording, no disulfides were formed in the absence of catalysts. Thus, in the absence of catalysts, disulfide formation and protein folding take place on entirely different timescales.

On a much longer timescale, disulfide bonds did form spontaneously under our experimental conditions (perhaps due to molecular oxygen leaking into the fluid cell). Our data showed that disulfide bonds were spontaneously formed with a time constant of 12.7 hours under the ambient experimental conditions (Figure 3.3).

3.4 PDI-mediated oxidation of folded proteins

The kinetics of PDI-catalyzed oxidation of folded proteins were measured by adding oxidized enzyme to a coverslide containing pre-reduced I27 32-75 and then unfolding individual substrate

proteins in the AFM, with a single force pulse of 150 pN. Stepsizes ( ∆L) were detected

automatically. Over time, the distribution of step sizes shifted from predominantly 25 nm to

predominantly 11 nm (Figure 3.3). The rate of the shift was proportional to the concentration of

PDI, and a fit to the data yielded a rate constant of 0.13 mM -1s-1. We also measured the rate of

oxidation using a range of other catalysts (Figure 3.3).

Apparently, in its reduced form, the cysteines of I27 32-75 were able to react with oxidized

PDI. Meanwhile, in the oxidized form, the disulfide in I27 32-75 is known to be inaccessible [26,

153]. This intriguing finding can be explained through the constrained flexibility caused by the covalent disulfide. An interesting corollary of this observation is that the 32-75 disulfide never reaches redox equilibrium; once it has been formed it is sterically “locked-in” and cannot be broken (Figure 3.4). This mechanism may be of physiological significance for the retention of native disulfide bonds.

3.5 Creating a mixed disulfide complex between PDI and a single unfolded protein

Using a different strategy, we were able to detect the formation of individual mixed disulfide complexes. Our strategy to establish these complexes is illustrated in Figure 3.2C, and an experimental recording is displayed in Figure 3.2D. We used a custom built Atomic Force

Microscope (AFM) to apply a constant calibrated force to the termini of a single I27 32-75 protein,

while measuring its extension [144]. A force of 130-150 pN enables protein unfolding but does

not break any of its covalent bonds [26, 28]. The polypeptide chain can thus unravel only up to

the disulfide bond, resulting in the 11 nm extension step seen in the recording (Figure 3.2D).

Meanwhile, the 11 nm extension exposes the 32-75 disulfide to the solvent and enables reactions

with reduced enzymes present in the surrounding media [26]. In Figure 3.2D, a reduced PDI

enzyme present in solution reacts with the now solvent exposed disulfide, creating a mixed

disulfide. The enzymatic reaction is captured as a 14 nm step that results from unraveling of the

rest of the substrate as soon as the original disulfide is broken. A 14 nm step in our experiments

thus indicates the formation of a mixed disulfide complex. Once a mixed disulfide was acquired

in a fully extended polypeptide, we could proceed to study its effect on protein folding.

Figure 3.3 Kinetics of catalyzed oxidation of folded I27 32-75 . Pre-reduced I27 32-75 polyproteins were incubated in the AFM fluid cell and randomly picked up and unfolded at 150 pN. Dot graphs: The observed stepsizes

(∆∆∆LU) are displayed as individual datapoints as a function of the time elapsed since the start of the experiment. 25 nm steps indicate reduced domains while 11 nm steps indicate oxidized domains (with formed disulfides, see Figure 2.16). The experiment was repeated with different concentrations of oxidized PDI. The panels on the right show the data binned and fit to single exponentials from which we calculated the rate of PDI catalyzed oxidation of a folded I27 32-75 protein, 0.13 mM -1s-1. Bar graph: The oxidation rates of folded I27 32-75 measured for a number of different catalysts. (PDI FL = full length bovine PDI)

Figure 3.4 Oxidized PDI can introduce a disulfide into a folded I27 32-75 domain (top). However, reduced PDI cannot cleave a disulfide in a folded I27 32-75 domain (bottom). This mechanism may underlie the retention of native disulfides in vivo .

3.6 Mixed disulfide complexes with PDI enable oxidative folding

We used a three part force protocol [ denature – ∆t – probe ] to study the effect of PDI mixed

32-75 disulfide complexes on the folding of an (I27 )8 polyprotein substrate (Figure 3.5A). A

polyprotein was used because it yields multiple events within a single recording, thereby

providing a stronger fingerprint of the reactions [166]. Figure 3.5B shows how each substrate

domain was first completely extended and linked to a PDI enzyme during an initial denature

pulse, as described in the previous section. The force was subsequently switched off and the

substrate was allowed to fold for a set time ∆t. In order to detect the formation of disulfide bonds

and protein folding during ∆t, we once again applied force, thereby halting the folding reaction.

This probe pulse was identical to the initial denature pulse, and in the same manner allowed us

to detect folded domains and disulfide bonds. A 25 nm step appearing during the probe pulse

reports that a domain had folded but not formed a disulfide during ∆t. An 11 nm step, on the

other hand, indicates that a domain had folded during ∆t and also acquired a disulfide. Since PDI

was still present in the surrounding media during this pulse, 14 nm steps would be seen if newly

formed disulfide bonds were again broken. In summary, a 25 nm step provided a signature of a

natively folded domain with no disulfide, whereas an 11 nm step (followed by a 14 nm step)

indicated a natively folded domain with a disulfide (Figure 3.5B) [19, 26].

Figure 3.5C shows an experimental recording of the [denature – ∆t – probe ] unfolding-

32-75 refolding experiment on (I27 )8 in the presence of reduced PDI, and Figure 3.5D shows

histograms of the step sizes detected in several such recordings. During the denature pulse, 11 and 14 nm steps were seen, indicating formation of mixed disulfide complexes between PDI and the substrate. When the force was removed ( ∆t = 5 s), the substrate rapidly collapsed as seen in the experimental trace (Figure 3.5C). The 11 and 14 nm steps seen in the probe pulse unambiguously showed that disulfides had been formed in the substrate during folding. The reaction path for these domains is readily interpreted. When the force was removed, the substrate collapsed and allowed the free cysteine in each domain to attack the mixed disulfide bond, thus releasing the PDI enzyme and establishing the intramolecular 32-75 disulfide bond. These domains also acquired their native fold, as seen from the 11 nm unfolding steps.

3.7 Detection of spontaneous enzyme release

In addition to 11 and 14 nm steps, 25 nm steps were also seen in the probe pulse (Figure 3.5C-D,

Figure 3.6), revealing that some substrate domains had folded without forming disulfide bonds.

Two scenarios can account for these 25 nm steps. Either (i) PDI was still attached to the folded substrate but had failed to introduce the 32-75 disulfide bond; or (ii) PDI had spontaneously released before substrate oxidation could be realized. The effect of the enzyme release rate on oxidative folding would be different in these two scenarios. We could thus distinguish between

the scenarios by using two oxidoreductase enzymes with vastly different release rates; wild-type

and C35S thioredoxin.

Figure 3.5 Mixed disulfide complexes with PDI enable oxidative folding. (A) A mechanical force was applied to a polyprotein consisting of repeated I27 32-75 domains, where each domain contained a disulfide between residues 32 and 75, in a solution containing reduced PDI. (B) The substrate was extended and mixed disulfides were formed between PDI enzymes and a cysteine in each domain, as described in Figure 3.2. The force was then removed, and the resulting folding and oxidation of the substrate was probed after a set time ∆∆∆t. (C) Representative recording showing extension and force measurements for the [denature – ∆∆∆t – probe] force protocol. 11 nm steps indicate unfolding of individual substrate domains. 14 nm steps (arrowheads) indicate formation of mixed disulfide complexes with PDI. After refolding for ∆∆∆t =5 s, four domains had refolded and reformed disulfides, as revealed by the 11 and 14 nm steps. Other traces revealed refolding without disulfide formation (25 nm step, inset, see also Figure 3.6). (D) Step size histograms compiled from several recordings confirm that PDI catalyzed oxidative folding in some domains while other domains refolded in a reduced state.

Figure 3.6 Oxidative folding with PDI A1 shows 25 nm steps in the probe pulse, evidence of incomplete oxidative folding. Arrowheads indicate 14 nm steps each resulting from the formation of a mixed disulfide complex.

Thioredoxin catalyzes the same reactions as PDI [64, 167-170] but releases spontaneously from the mixed disulfide on a much faster (sub-millisecond) timescale, as inferred from reactions with non-protein substrates [60, 171]. Consequently, in a [ denature - ∆t - probe ] experiment with wild-type thioredoxin, all enzymes should have released before the end of the denature pulse.

We would thus expect only 25 nm steps in the probe pulse as the substrate folds in the absence of the enzyme. Experiments with thioredoxin confirmed that only 25 nm steps were detected in the probe pulse (Figure 3.7A).

Spontaneous release from the mixed disulfide is mediated by the C-terminal cysteine in the

active site of oxidoreductases (Cys35 in thioredoxin). We mutated this cysteine to abolish

spontaneous release (TRX C35S, Figure 3.7B; see also Figure 3.8). This resulted in the complete

absence of 25 nm steps and full oxidation of the substrate (as indicated by 11 and 14 nm steps in

the probe pulse). Thus, every substrate domain that folded in the presence of a mixed disulfide

complex successfully completed oxidative folding. These observations indicate that 25 nm steps

are only caused by spontaneous release of the enzyme from the mixed disulfide complex prior to

folding, in agreement with scenario (ii) described above. We could therefore use 25 nm steps as a

proxy for spontaneous enzyme release.

Figure 3.7 Enzyme release determines outcome of oxidative folding. (A) Mixed disulfide complexes with thioredoxin (TRX) do not catalyze disulfide formation in the folding I27 polyprotein. In this trace, seven domains were completely extended during the denature pulse. Four of these refolded subsequently, albeit without disulfides, as seen from the 25 nm steps. The rapid release mechanism of TRX accounts for its inability to catalyze disulfide formation (right panel). (B) The C35S mutation in TRX replaces a reactive sulfur atom with oxygen and thus prevents spontaneous release of the enzyme, see Figure 3.8. TRX C35S catalyzed disulfide formation in the folding polypeptide by remaining in the mixed disulfide complex upon substrate folding. Step size histograms of the probe pulse show that disulfides were formed in all refolded domains.

Figure 3.8 (A) Structure of reduced human thioredoxin (TRX), showing the two redox active sulfur atoms in the active site. The sulfur in Cys32 carries out the initial nucleophilic attack that cleaves the substrate disulfide and forms a mixed disulfide (PDB: 1trs). (B) Single cysteine mutant of human TRX in a mixed disulfide complex with a substrate fragment. The Cys35Ser mutation removes the mechanism for spontaneous release from the mixed disulfide complex (PDB: 1mdi). (C) Experimental trace showing a single I27 32-75 domain being unfolded (11 nm step) and forming a mixed disulfide with TRX C35S (14 nm step). After refolding, the 32-75 disulfide has been reformed, showing that the distal cysteine in the substrate was able to attack the mixed disulfide, despite the steric hindrance imposed by the enzyme.

3.8 Measurement of the PDI spontaneous release rate

In our experiments, we could detect the formation of individual mixed disulfide complexes (14

nm steps in the denature pulse), their presence during protein folding (14 nm steps in the probe pulse) and their spontaneous dissociation prior to folding (25 nm steps in the probe pulse). By varying the time before folding (the duration of the denature pulse ∆tD, see Figure 3.9, Figure

3.10A) and measuring the resulting proportion of 25 nm steps in the probe pulse, we could thus for the first time measure the rate of PDI spontaneous release from a protein substrate. In these

experiments, we used a force protocol optimized for long experiments with split denature and probe pulses (see methods).

For a denature pulse duration ∆ tD = 5 s, the majority of refolded domains contained disulfide bonds (11 nm, 14 nm steps in probe pulse of Figure 3.9A). In contrast, for ∆tD = 30 s, the covalently bound enzymes had more time to release from the mixed disulfide complex, which led to a majority of folded domains without disulfide bonds (25 nm steps in probe pulse of

Figure 3.9B).

The fraction of 25 nm steps in the probe pulse is shown as a function of ∆tD in Figure 3.9C.

The model depicted in Figure 3.9D is the simplest description of the process and contains three rates: koff , kox and kfold . These represent the rate of spontaneous enzyme release (off-rate), the rate

of oxidative folding from the mixed disulfide complex, and the rate of folding of the reduced

substrate, respectively. Notably, we found that the total amount of refolding remained constant

for all values of ∆tD (Figure 3.10C), leading to the conclusion that kox and kfold were approximately equal. We therefore assumed that the ratio displayed in Figure 3.9C was directly proportional to the fraction of remaining mixed disulfide complexes. A least-square fit to the data revealed that PDI spontaneously released at a rate of 0.05±0.02 s-1.

An extrapolation to ∆tD = 0 indicated that even at short durations of the denature pulse, there was a non-zero population of 25 nm steps. This is probably due to the spontaneous dissociation of enzymes after the force was switched off but before the substrate had been oxidized. In addition, the off-rate may be higher at zero force than at 75pN, although this hypothesis would require further experiments to ascertain.

Figure 3.9 Measurement of PDI release kinetics. (A) In experiments with a 5 s denature pulse (∆∆∆tD = 5 s), nearly all PDI enzymes remained in the mixed disulfide complexes during the subsequent protein folding. As a result, oxidative folding mostly completed successfully, as seen from the predominance of 11 and 14 nm

steps in the probe pulse. (B) After a 30 s denature pulse (∆∆∆tD = 30 s), most mixed disulfides had been cleaved through spontaneous release of PDI. Folding was still observed but only few of the folded domains contained disulfides, as seen from the predominance of 25 nm steps in the probe pulse. (C) Fraction of folded domains

that were not oxidized by the start of the probe pulse (25nm step fraction, mean ± SEM), as a function of ∆∆∆tD. The data for PDI fall between the extremes of TRX and TRX C35S. The rate of spontaneous PDI release was calculated from a fit to the model in (D). (D) Kinetic model including the rate of spontaneous enzyme release (koff ), the rate of oxidative folding from the mixed disulfide complex ( kox ), and the rate of folding of the reduced substrate ( kfold ).

Figure 3.10 (A) The kinetic competition during refolding of I27-PDI mixed disulfide complexes. When the

substrate is extended (as during ongoing translocation or during application of mechanical force during ∆∆∆tD), no substrate cysteine is available to cleave the mixed disulfide bond. However, the enzyme can release spontaneously through nucleophilic attack by the C-terminal cysteine in its active site. As soon as the substrate collapses (upon completion of translocation or removal of the mechanical force), its free cysteines become available. The outcome of oxidative folding now depends on the competing reaction kinetics of the cysteines in the enzyme and those in the substrate. (B) Rate of PDI-substrate mixed disulfide formation (black circles) and the 25nm step fraction in the probe pulse (red diamonds, ∆∆∆tD = 5 s), as a function of time from the start of the experiment. Horizontal bars indicate the data range; vertical bars indicate SEM. (C) Refolding of 32-75 I27 from mixed disulfide complexes with PDI, for different durations of the denature pulse ∆∆∆tD (mean ± SEM). The folding time ∆∆∆t was set to 5 s. There is no apparent dependence of the amount of folding on ∆∆∆tD, showing that folding is not noticeably altered by the presence of the mixed disulfide complex. The rates kox and kfold shown in Figure 3.9D are therefore approximately equal. (D) Fraction of 25 nm steps in probe pulse of PDI experiments, as a function of refolding time (mean ± SEM). Length of the denature pulse ∆∆∆tD = 5s. If some mixed disulfides were still present after the substrate had acquired a native fold, then the prevalence of 25 nm steps would be expected to decrease at long refolding times (as the domains became oxidized). The figure shows that the fraction of 25 nm steps does not decrease as a function of refolding time, indicating that the enzyme had released from all of these domains prior to folding.

3.9 Kinetics of oxidative protein folding from an extended state

A mixed disulfide complex with PDI enables the catalysis of oxidative folding, yet it is unknown

how protein folding is affected by the covalent attachment of this enzyme. Our approach allowed

us to directly measure the effect of a covalently bound PDI on protein folding.

We first set out to establish the refolding properties of I27 32-75 in isolation, with and without

the 32-75 disulfide bond. Using the [ denature – ∆t – probe ] protocol, we first measured the

folding kinetics of the reduced substrate. Figure 3.12A shows a representative trace for a

refolding time ∆t = 3s. Steps of 25 nm are seen both in the denature and probe pulses,

confirming that none of the domains contained disulfide bonds. No disulfides were formed

during folding due to the absence of mixed disulfides. (Figure 3.11 shows the equivalent

experiment with oxidized I27 32-75 ). At a longer refolding time ∆t = 10 s, more domains had

refolded (Figure 3.12B). By varying the refolding time ∆t, we could measure the kinetics of refolding. The reduced protein folded at a rate of 0.22 s-1, while presence of the 32-75 disulfide

increased the folding rate by 32 times to 7.1 s-1 (Figure 3.12C).

Strikingly, the rate of PDI-catalyzed oxidative folding was similar to the rate of folding of

the reduced substrate (Figure 3.12C). Despite the steric hindrance caused by covalent

attachment, PDI appeared to interfere minimally with the folding protein. Furthermore, PDI-

catalyzed disulfide formation followed single-exponential kinetics, suggesting a single rate-

limiting step late in the folding pathway. After the disulfide was formed in the substrate, the

native state was acquired very rapidly, as inferred from the fast refolding kinetics of the oxidized

substrate (Figure 3.12C).

Taken together, these data give an idea of how PDI catalyzes disulfide formation in the

folding protein. PDI first becomes covalently attached to the substrate by forming a mixed

disulfide complex. From there on, the enzyme acts as a passive placeholder, allowing the

substrate to collapse and fold in an unhindered manner. At a late stage of folding, close in time to

the acquisition of the native fold, PDI catalyzes the formation of an intramolecular disulfide in

the substrate. This reaction releases PDI and allows the substrate to rapidly complete oxidative

folding.

Figure 3.11 Refolding of oxidized I27 32-75 in the absence of enzymes. No cleavage of disulfide bonds was observed in these experiments.

3.10 Interchangeability of oxidoreductase enzymes

Thioredoxin is a cytosolic reductase whereas PDI is an oxidase localized to the ER. However,

previous studies have revealed an apparent interchangeability in the function of oxidases from

the thioredoxin family [168-170]. These studies suggest that the function of an oxidoreductase in

vivo mainly depends on its equilibrium redox state. However, PDI and thioredoxin have different

kinetic properties [64], which may serve an important role during thiol-disulfide exchange in

folding proteins, a process that is kinetic in nature and may never reach equilibrium in vivo . We showed in this study that thioredoxin catalyzed disulfide formation in folding substrates only when its spontaneous release mechanism was disabled through the C35S mutation (Figure 3.7).

TRX C35S catalyzed disulfide formation in the folding polypeptide by remaining in the mixed disulfide complex upon substrate folding. Notably, TRX C35S exhibited disulfide formation kinetics similar to that of PDI (Figure 3.12C), suggesting that the catalytic mechanisms of

thioredoxin and PDI are highly similar. It thus appears that wild type thioredoxin is intrinsically

capable of catalyzing disulfide formation, but releases rapidly from the mixed disulfide complex

and therefore prevents disulfide formation in folding proteins. This can explain the results of an

earlier study where TRX C35S, but not wild-type thioredoxin, rescued a PDI deletion strain of

yeast [172].

Figure 3.12 Measurement of protein folding from an extended state. (A) Mechanical unfolding of pre-reduced I27 32-75 polyproteins shows only 25 nm steps in the denature pulse, indicating that disulfides were absent in the folded domains. (B) After a longer ∆∆∆t, more domains had folded. Folding kinetics were measured by varying the refolding time ∆∆∆t. Uncatalyzed disulfide formation was not observed in these experiments. (C) Time course of folding (gray squares and black triangles) and oxidative folding (red diamonds and blue triangles) from an extended state. I27 folding is accelerated more than 30-fold by the presence of the 32-75 disulfide (see Figure 3.11 for a representative trace). The rates of disulfide formation catalyzed by PDI and TRX C35S are indistinguishable from the folding rate of the reduced substrate, indicating that disulfide formation occurs late in the oxidative folding process, close to the acquisition of the native state. All data displayed as mean ± SEM.

3.11 Formation of interdomain disulfides triggers misfolding

The probe histograms of PDI and TRX C35S (Figure 3.13B) both revealed a shoulder on the

higher- step size side of the 14 nm peak. The overlaid histograms show that this shoulder was not

present in the denature pulse. This prompted us to perform a more thorough analysis of the step

sizes. Figure 3.13A shows an experimental recording in which the reduction events (14 nm) have

been separated from the unfolding events (11 nm) by means of divided denature and probe

pulses. By binning and averaging the unfiltered data points before and after each step, we

achieved a spatial resolution better than 0.1 nm. The analysis verified that the events in the new

population indeed were different from the 14 nm steps. Whereas the normal reduction events had

a magnitude of 13.8 nm at 75 pN (hereafter referred to as 14 nm steps for simplicity), the new

steps had a magnitude of 15.4 nm (hereafter referred to as 15 nm steps). These steps accounted

for around 20% of the data and were apparently a possible outcome of oxidative folding, yet they

could not have arisen from cleavage of the 32-75 disulfide.

The 15 nm steps were not induced by a high stretching force, as seen in the denature pulse

in Figure 3.13A, and are therefore unlikely to represent unfolding events. Instead, their step size

corresponded exactly to the expected elongation upon cleavage of a disulfide between Cys75 in

one I27 domain and Cys32 in the next domain (15.4±0.1 nm vs. 15.5 nm, see Figure 3.14A), and

we therefore conclude that these steps resulted from cleavage of non-native interdomain

disulfides. Further supporting this conclusion, other non-native disulfides were reproducibly

observed, however much less frequently (Figure 3.14A).

The data also showed that whereas nearly every 14 nm step in the denature and probe pulses

was preceded by a corresponding 11 nm step (97±1% and 91±3%, respectively), none of the 15

nm steps were preceded by a corresponding unfolding step (Figure 3.13A). Furthermore, the

appearance of 15 nm steps was negatively correlated with the fraction of domains that

successfully refolded (Figure 3.14B). This shows that domains affected by the misformed

disulfide were unable to fold into their native conformation. Although the substrate might still

attain some degree of structure beyond our detection limit, the absence of mechanical resistance

of these structures proves that they are not correctly folded. Thus, formation of erroneous non-

native disulfides was sufficient to induce protein misfolding.

Remarkably, the non-native disulfides were formed at a faster rate than the native disulfides

(1.0 s -1, Figure 3.13C). The erroneous interdomain disulfides therefore appear to be formed at an

earlier stage of folding. This is consistent with the substrate being less structured early on during

the folding process. At this stage, two cysteines from adjacent domains would be more likely to

come in contact. The folding pathway thus determines the fidelity of disulfide formation. The

percentage of correctly formed disulfides by PDI was initially low but increased as the protein

progressed along the folding pathway. As the protein approached the native state, the accuracy

approached 100%.

The prevalence of interdomain disulfides did not appear to decrease even at longer waiting

times (Figure 3.13C), indicating that they are relatively stable and not prone to isomerization on

the timescale of our experiments.

Figure 3.13 Non-native interdomain disulfides prevent native folding. (A) Cleavage of a 32-75 intra-domain disulfide results in a 13.8 nm step. If the cleaved disulfide is instead between Cys75 in one domain and Cys32 in an adjacent domain, the resulting step has a magnitude of 15.4 nm. The displayed trace shows unfolding and PDI-mediated disulfide cleavage of four domains in the denature pulse. Upon refolding, two interdomain disulfides were formed but no refolding took place. Cleavage of the interdomain disulfides can be seen in the probe pulse. (B) Superimposed histograms of the denature and probe pulses reveal a population of 15.4 nm steps (black arrows) that appear only after refolding in a mixed disulfide complex with PDI and TRXC35S. None of these steps were preceded by corresponding 11 nm unfolding steps. (C) The kinetics of interdomain disulfide formation by PDI (fraction, mean ± SEM). These disulfides appear to form at a faster rate than the native disulfides, indicating that erroneous bonds form earlier in the folding pathway, before the unfolded substrate has fully segregated into separate domains. The prevalence of interdomain disulfide bonds did not decrease at longer refolding times, which shows that they are relatively stable. See also Figure 3.14.

Figure 3.14 (A) Histogram of atypical steps in the probe pulses from all experiments with PDI. The histogram contains only steps that could not be accounted for by unfolding and reduction of native domains; thus, ordinary 11, 14 and 25 nm steps were excluded. Vertical red lines indicate expected step sizes upon cleavage of possible inter-domain disulfide bonds, labeled as [residue] domain# . For instance, 75 N–32 N+1 represents cleavage of a disulfide between residue 75 in one domain and residue 32 in the next domain. (B) Formation of inter-domain disulfide bonds prevent native folding of I27 32-75 . The appearance of 15 nm steps in the probe pulse correlated strongly with the number of mixed disulfide complexes that were formed in the denature pulse, and therefore with the number of domains in the substrate. The appearance of these steps was negatively correlated both with native disulfide formation and native protein folding. Folding is thus hindered by the formation of inter-domain disulfides. (C) Formation of inter-domain disulfide bonds (mean ± SEM) was linearly dependent on the number of unfolded I27 32-75 domains (as measured from the number of 11 and 14 nm steps in the denature pulse). The x-intercept of the fit was close to 1, thus verifying that at least two domains are required to form an inter-domain disulfide.

3.12 Discussion and biological implications

For a protein, the path to a native fold is lined with potential traps [160]. Protein folding has become a field of intense research due to its fundamental importance as well as its relevance in numerous diseases. The formation of disulfide bonds adds an additional layer of complexity to the folding pathway of many proteins. For these proteins, cells have evolved dedicated pathways to ensure efficient oxidative folding [29, 173].

In eukaryotes, nascent polypeptides targeted to the ER are exported from the cytosol cotranslationally [83]. Emerging in the ER lumen, polypeptides that contain cysteines form mixed disulfide complexes with oxidase enzymes such as PDI [86, 87]. However, it has remained unknown how a mixed disulfide complex affects protein folding, and how oxidation and folding are coupled. To investigate this, we used single molecule AFM to reconstitute mixed disulfide complexes between PDI and an unfolded model protein. Our approach enabled, for the first time, independent kinetic measurements of protein folding and disulfide formation. We showed that

PDI did not interfere with the folding protein, and that the protein folding rate therefore was unaffected by the presence of a mixed disulfide with PDI.

3.12.1 A kinetic model for the oxidative folding of an extended protein

A longstanding question pertains to the causal relationship between disulfide formation and protein folding during oxidative folding. Specifically, does disulfide formation drive protein folding or, conversely, does protein folding provide a driving force for native disulfide formation? Both ideas have been proposed in previous studies [39, 117, 124, 174, 175], without any consensus being reached in the field. In our experiments, disulfide formation was catalyzed by PDI at a late stage of protein folding (Figure 3.15). The enzyme remained in the mixed

90 disulfide complex until the substrate had attained a near-native state. At this point, the substrate disulfide was formed and PDI was released. PDI thus appeared to function more like a passive placeholder than as an active folding catalyst. In conclusion, our results indicate that protein folding provided the driving force during oxidative folding, while PDI simply enabled disulfide formation.

Nascent polypeptides are constrained in an extended state as they traverse the ribosomal exit tunnel and Sec translocase [1]. Before folding can take place, extended polypeptides first have to collapse [17, 19, 142]. In the absence of force, extended full-length proteins collapse in a fraction of a second and then sample an ensemble of compact conformations that over a timescale of several seconds acquire their native contacts. While these compact states have only weak mechanical stability and thus are not completely folded, they have nevertheless been shown to be essential on-pathway intermediates [19]. Such “molten globule” precursor states have been studied extensively, and they are today recognized as a significant deviation from the classical two-state model of protein folding [176-178]. We found in this study that the kinetics of disulfide formation were only marginally faster than the kinetics at which the mechanically stable native state was acquired. Since molten globules are formed on a much faster timescale, we conclude that the molten globule state was not sufficient for disulfide formation. Therefore, PDI catalyzes disulfide formation with high efficiency only after a large fraction of the native contacts have been acquired.

Figure 3.15 PDI-mediated oxidative folding in the ER. A nascent polypeptide is kept physically extended while emerging from the ribosome and is unable to fold or form disulfide bonds before it enters the ER. PDI enables oxidative folding in the ER lumen by forming a mixed disulfide with a cysteine in the nascent polypeptide. The polypeptide collapses before folding takes place, while PDI remains attached throughout this process. Disulfides that are formed at an early stage of folding tend to be non-native. PDI favors the formation of native disulfides by allowing the polypeptide to fold into a near-native state before it catalyzes disulfide formation. By allowing protein folding to guide the pairing of cysteines, PDI can catalyze oxidative folding without the need for substrate specific interactions. Graphs on right side: Color intensity indicates percentage of native contacts (blue bar) and probability of disulfide formation (yellow and orange bars).

3.12.2 Significance of PDI spontaneous release

Co-translational folding is limited by the ribosomal translation rate, which is around 5 amino

acids per second [179] (Figure 3.15). Studies of co-translational folding using NMR have shown

that, whereas proteins can start acquiring some structural elements while still attached to the

ribosome, they acquire their native fold in a domain-wise fashion [180-182]. For a typical 100

amino acid protein domain, this scales to a time of around 20 seconds during which its cysteines

are relatively accessible and therefore able to form mixed disulfides with PDI. Based on the

estimated concentration of PDI in the ER (1 mM) [90, 183] and the rate of mixed disulfide

formation (>0.6 mM -1s-1) [184], there is a high probability of a mixed disulfide being present

when folding occurs.

When an entire protein domain has been translocated into the ER, the mixed disulfide has to remain in place as the protein folds into its native state (Figure 3.15). What happens if PDI dissociates before the substrate is completely folded? An exposed substrate cysteine can form a new mixed disulfide. However, protein folding can cause the burial of cysteine residues, which in turn can prevent thiol-disulfide exchange entirely. For instance, Walker and Gilbert showed that PDI was unable to introduce a disulfide into folded β-lactamase [184]. Disulfide formation

only took place after the substrate spontaneously unfolded, which for β-lactamase could take up to an hour [184]. This timescale is incompatible with many cellular processes, which leads us to conclude that in order for cells to avoid actively unfolding such substrates, PDI needs to introduce disulfides when the nascent protein folds for the first time. This requires that the spontaneous off-rate of PDI is slower than the rate of protein folding. We measured the off-rate of human PDI to be 0.05 s -1. Human PDI is thus an efficient oxidative folding catalyst for substrates that fold in less than 20 s. This time limit is likely sufficient for most secreted proteins,

93 and certainly sufficient for small proteins with simple globular structure. However, protein folding can take place on a wide range of timescales. It is therefore tempting to suggest that the off-rate of PDI is adapted to the folding rate of its substrates. The human PDI family has at least

19 members. Perhaps this diversity is required to accommodate the diversity of protein folding rates.

Spontaneous release by PDI prevented successful oxidative folding in our experiments.

Nevertheless, this mechanism can serve an important purpose as a release timer in situations where the mixed disulfide complex is unproductive [39, 80]. For example, proteins that contain an odd number of cysteines may have one or more unpaired cysteines in their native structures. If

PDI forms a mixed disulfide complex with one of the unpaired cysteines, it would need to release to avoid being trapped. Furthermore, if PDI enzymes formed mixed disulfide complexes with both cysteines in an intended disulfide, oxidative folding would be delayed until one of the

PDI enzymes left. Finally, cysteines in misfolded proteins might never approach each other. Also in this case is there a need for an intramolecular timer that releases the PDI enzyme if the mixed disulfide complex is unproductive.

3.12.3 PDI uses a non-specific mechanism to favor native disulfides

How does PDI ensure correct pairing of cysteines in proteins with more than one disulfide? For a

4-disulfide protein such as RNAse A, there are over 700 possible disulfide-bonded combinations.

In order to catalyze oxidative folding in a wide range of substrates, PDI needs a catalytic mechanism that is general yet still ensures correct pairing for each specific protein. In view of this, Wilkinson and Gilbert proposed that it is the substrate, rather than PDI, that determines the pairing of cysteines [39]. Our data lend strong support to this theory by showing that PDI mixed

disulfide complexes did not affect substrate folding kinetics. PDI could thus allow its substrate to

fold before introducing the disulfide. Despite the vast number of possible disulfides in our multi-

domain substrate, the majority of bonds formed by PDI in our experiments were the correct,

intra-domain disulfides. In other words, PDI favors native disulfides by affording its substrate a

generous amount of time to decide, through folding, which cysteines to join (Figure 3.16). Since

this mechanism does not rely on any substrate-specific interactions besides the mixed disulfide, it

can explain how a single enzyme can catalyze the oxidative folding of a wide variety of proteins.

Figure 3.16 PDI uses a non-specific mechanism to favor the formation of native disulfide bonds.

3.12.4 Future directions

Oxidative folding in the cell is a highly complex process that involves a multitude of

components, including oxidoreductases, chaperones, proline cis-trans isomerases and small

redox molecules such as glutathione. All of these molecules can interact with a protein during its

folding process. This complexity has severely hindered detailed mechanistic studies of oxidative

folding [173]. We have here presented an approach that enables control and measurement of both

folding and disulfide formation in single protein molecules. Through the use of mechanical

force, we can capture fleeting intermediate states and thereby examine in detail the components

of oxidative folding. The methods described here can be used to determine the differences

between oxidase enzymes, as well as the function of other factors involved in oxidative folding

in the cell.

3.13 Materials and methods

3.13.1 Single-molecule Atomic Force Microscopy

The details of our custom-made atomic force microscope have been described previously [17].

We used silicon nitride cantilevers (Bruker) with a typical spring constant of 15 pN nm -1,

calibrated using the equipartition theorem [185]. The force-clamp mode provided a feedback

time constant of 5 ms. The buffer used in all the experiments was 10 mM HEPES, 150 mM

NaCl, 1 mM EDTA, degassed and at pH 7.2. Before the beginning of the experiment, the

enzyme sample was incubated with reduced DTT for 20 minutes at room temperature, to ensure

that the enzyme was fully reduced. 200 µM DTT was used with TRX (10 µM) and TRX C35S

(40 µM). 500 µM DTT was used with PDI A1 (120 µM). The low concentrations of DTT did not directly cause any redox activity on the substrate on the timescale of the experiments. As an additional control to ensure that our results were not influenced by the presence of DTT, we repeated the PDI experiments in the absence of DTT and obtained the same results. This was accomplished by first reducing the enzyme in a solution containing 10 mM reduced DTT for 10 minutes, and then removing the DTT in an FPLC size exclusion column. The purified reduced enzyme was snap-frozen and stored under argon at -80 °C until just before the experiment. The polyprotein substrate was added in a droplet and allowed to bind to a freshly evaporated gold coverslip before the experiments. Every experiment consisted of repeated trials where the tip was pressed against the surface for 0.5 s and subsequently retracted. If attachment was achieved, the

pulse protocol was applied until detachment occurred. The oxidative folding force-clamp

experiments used a triple-pulse [denature – ∆t – probe ] force protocol. The first pulse was

maintained at 130-150 pN for a time long enough to ensure complete unfolding and reduction of

the substrate (at least 5 s). The second pulse was set at 0 pN and maintained for the desired

amount of refolding time. The third pulse was set at a force identical to the first and maintained

until complete unfolding and reduction could be ensured (at least 5 s). For measurement of the

PDI off-rate, split denature and probe pulses were used. This approach has been described earlier for reduction measurements, and allowed for a clearer separation of unfolding and reduction steps [26]. Briefly, a short pulse at a high force (150-170 pN for ~1s) ensured unfolding and exposure of all substrate disulfide bonds. The force was then lowered to 75 pN to detect reduction while minimizing the probability of detachment of the substrate from the AFM tip.

3.13.2 Mechanical fingerprints of I27 32-75

A natively folded I27 32-75 protein can resist piconewton-range stretching forces applied to its

termini for a stochastic amount of time. The rate of thermally driven unfolding increases

exponentially with the applied force [144]. Presence of the 32-75 disulfide did not alter the

mechanical stability of the native state. The 11 nm extension occurs in a single step because once

the native fold is broken, the protein unravels in a downhill manner up until the disulfide (see

Section 2.3 for further details). Importantly, the resulting step size reveals how many residues

have unraveled, and the 11 nm step thus provides a signature of a natively folded and oxidized

I27 32-75 protein [186]. Similarly, once the 32-75 disulfide is broken, the sequestered part of the

protein unravels instantaneously under the applied force.

The expected step size upon cleavage of a disulfide was calculated as:

∆L = kWLC × ( LAA × NAA – LSS )

where kWLC is the worm-like chain fractional extension at a given force, LAA is the contour length

of an amino acid, NAA is the number of amino acids between the two cysteines and LSS is the

contour length of a disulfide bond.

3.13.3 Formation of mixed disulfide complexes

Of the two cysteines in the active site of PDI A1, only the N-terminal cysteine engages in a

mixed disulfide with substrates during its normal activity [163]. The mixed disulfide can be

formed in one of two ways (Figure 3.1, see also section 1.5.2). In the first pathway, an oxidized

PDI enzyme reacts with a free substrate cysteine. This pathway is thought to be predominant

during oxidative folding in vivo . In the second pathway, the mixed disulfide is formed through reaction of reduced PDI with a disulfide in the substrate. This reaction cleaves the substrate disulfide and is therefore necessary for reduction or isomerization activity. Regardless of the pathway, however, the resulting mixed disulfide complex is identical. From an experimental point of view, forming the mixed disulfide complex through the second pathway proved more advantageous, as the complex in this case is formed concomitantly with cleavage of a substrate disulfide. Thus, by detecting disulfide cleavage in a substrate, we could detect the formation of single mixed disulfide complexes in real-time.

3.13.4 Data analysis

We collected and analyzed data using custom-written software in IGOR Pro (Wavemetrics).

Traces were selected based on the fingerprint consisting of at least 3 unfolding events in the

denature pulse. Traces exhibiting step sizes other than 11 nm or 14 nm in the denature pulse were excluded from the analysis to ensure homogeneity (except for experiments with reduced

I27). Only traces showing equal extension at the end of the probe pulse and the end of the denature pulse were included, to ensure that the same protein was stretched in the two pulses.

Step size histograms were generated using all steps >5 nm detected after the initial elastic extension in each force pulse. Histograms in the main text figures were compiled from representative experiments. SEM was estimated through the bootstrap method.

Chapter 4 S-Glutathionylation Modulates Protein Mechanics

100

4 S-Glutathionylation Modulates Protein Mechanics

Mechanochemistry is increasingly recognized as an important component in the regulatory toolkit of cells. Cellular homeostasis depends on protein stability under mechanical loads.

However, it is not known whether the mechanical properties of proteins are subject to regulation.

Here we show that in proteins under strain, cryptic cysteines become targets for glutathionylation, a post-translational modification involved in redox signaling.

Glutathionylation of cryptic cysteines decreases the mechanical stability of a protein. This effect is reversed by the enzymatic activity of glutaredoxin. By reversibly altering protein mechanics, strain-dependent cysteine modification represents a new paradigm in mechanochemical regulation.

4.1 Physiological significance of S-glutathionylation

Oxidative stress alters the cellular redox balance and causes a wide range of physiological

responses [187]. Cysteine thiols are common targets for oxidation [188]. The main thiol-

containing molecule in cells is the tripeptide glutathione, which under normal cellular conditions

is present primarily in its reduced, monomeric form GSH (Figure 1.9). However, during

oxidative stress there is a buildup of oxidized glutathione (GSSG) as well as mixed disulfides

between glutathione and protein cysteines [105]. The latter process, known as protein S-

glutathionylation, is increasingly being viewed as an important posttranslational modification

that can alter protein function [113-116, 188]. For instance, Zweier and coworkers showed in a

recent study that S-glutathionylation decreased the generation of nitric oxide (NO) by endothelial

nitric oxide synthase (eNOS), an enzyme critical to the regulation of vascular function. In a rat

101 model, S-glutathionylation of eNOS led to impaired vasodilation, a pathogenic mechanism in hypertension, atherosclerosis and other cardiovascular disease. Further strengthening their findings, the authors found that S-glutathionylation of eNOS was significantly increased in cases of spontaneous hypertension [114]. S-glutathionylation thus appeared to function as a reversible redox switch that deactivated a regulatory enzyme. However, another study showed that S- glutathionylation caused activation of the calcium ATPase SERCA, providing a mechanism for the regulation of intracellular calcium levels in muscle cells [113]. These and other studies have showed that glutathionylation in vivo plays a multifaceted role in redox signaling and regulation.

4.2 Oxidative stress in muscle

Muscle is a tissue that endures high levels of oxidative stress, both during normal exercise [189,

190] and during pathological conditions such as reperfusion injury and inflammation [191, 192].

Although oxidative stress has been linked to dysfunction and fatigue, we are only beginning to understand the mechanisms by which oxidative stress alters muscle function [189, 192]. In a recent study, exercise was seen to correlate with a marked increase in S-glutathionylation of the muscular protein troponin C in human subjects, revealing a possible pathway for redox regulation [193].

Titin is the main force-bearing component in muscle. Spanning half the sarcomere, it is the only protein that maintains sarcomere integrity in the absence of myosin-actin crosslinks. With over 30,000 residues and a molecular weight of around 3 MDa, it is the largest known protein.

Titin is structured as a polyprotein that contains segments of repeated, independently folded Ig domains (Figure 4.1) [194-196]. Dynamic unfolding and refolding of Ig domains in titin contributes to its elastic properties under strain [7, 197-199]. However, it is not known how these

102 properties are altered by oxidative stress. Specifically, it remains unknown if cysteines in titin are affected by glutathionylation.

Figure 4.1 Architecture of a muscle sarcomere. Four titin polyproteins are shown.

Protein cysteines need to be exposed to be targets for glutathionylation. However, cysteine residues tend to be buried in the hydrophobic core of folded proteins, where they are unreactive and cannot be glutathionylated [200]. On the other hand, even well-structured proteins are known to unfold when they are subject to physiological forces [7, 197, 198, 201, 202]. Through strain-induced unfolding, cysteines that are normally buried can be exposed and are rendered reactive [201]. We speculated that attachment of a glutathione molecule to such a buried residue would interfere with the native structure of the protein. This led us to hypothesize that strain- dependent glutathionylation of cryptic cysteines could alter the mechanical properties of the affected proteins.

4.3 Direct detection of S-glutathionylation reactions in a single protein molecule

There are many biochemical pathways that can lead to S-glutathionylation of a protein [105].

However, our main goal was not to elucidate these pathways but rather to study the effect of

glutathionylation on protein mechanics. We therefore designed an experimental protocol in

which the glutathionylation event gave rise to a signal that we could detect. Briefly, we used

103

GSH to reduce disulfide bonds in a protein placed under force in the AFM. Cleavage of the disulfide then resulted in glutathionylation of a cysteine but also in a detectable extension of the protein. Our strategy is outlined below.

There are two main cytosolic pathways for disulfide reduction, mediated by TRX and GSH

respectively [203]. TRX and GSH both rely on thiol-disulfide exchange reactions that are

– initiated when a thiolate anion in the reducing agent ( R1S ) performs a nucleophilic attack on a

disulfide bond ( R2SSR 3):

ͯ + → + ͯ ͍͌ͥ ͍͍͌ͦ͌ͧ ͍͍͌ͥ͌ͦ ͍͌ͧ

As the original disulfide is cleaved, one of its sulfurs forms a mixed disulfide with the reducing agent ( R1SSR 2). TRX rapidly releases from this mixed disulfide because it has two active cysteines. Cys32 mediates the initial attack and forms the mixed disulfide; Cys35 then rapidly attacks the mixed disulfide, releasing the enzyme. Due to the activity of Cys35, the complex between TRX and its substrate is short-lived, with a measured lifetime of less than a millisecond

[60] (see Section 1.5.1 for a more detailed description of TRX activity).

In contrast to TRX, GSH contains only one cysteine (Figure 1.9). After the initial attack, the substrate is thus left with a covalently linked glutathione moiety. In other words, the substrate has become S-glutathionylated.

We studied the refolding of the model protein I27 32-75 after disulfide reduction by

thioredoxin (TRX) and glutathione (GSH), respectively. Using an Atomic Force Microscope

(AFM), we were able to induce and detect individual disulfide cleavage reactions [25, 26]

(Figure 4.2A-C). We combined this technique with an [unfolding-refolding-probe ] protocol to measure refolding after reduction (see Section 2.3 for experimental details).

104

4.4 S-glutathionylation inhibits protein folding

We applied an initial force of 150pN to a single I27 32-75 polyprotein substrate, which induced stochastic unfolding of individual Ig domains (Figure 4.2A). Unfolding events were detected as stepwise increases in the end-to-end length of the substrate, as single Ig domains unraveled up to the 32-75 disulfide bond (11 nm steps in Figure 4.2D-E). The 32-75 disulfide bonds are inaccessible in the folded substrate. However, once exposed through unfolding, they can be cleaved by reducing agents in the surrounding media [26]. Disulfide bond cleavage reactions by

TRX and GSH were detected as further extension steps due to unraveling of the remainder of the domains (14 nm steps in Figure 4.2D-E, see also Section 2.3.5).

Once the substrate had been completely unfolded and reduced, the force was switched off to enable refolding. The substrate collapsed within a fraction of a second when the force was removed (Figure 4.2D-E), a process that is necessary but not sufficient for folding to take place

[143].

After the refolding interval, we detected the presence of refolded domains by again applying a force of 150 pN. In the case of reduction by TRX, this probe pulse displayed a series of 25 nm steps after 5 s refolding (Figure 4.2D). These steps are distinct signatures of folded and reduced

I27 domains. Since only natively folded domains unfold in a distinct step [19] the 25 nm steps report that a number of domains had successfully refolded in the absence of force. However, the experimental recording in Figure 4.2E shows no 25 nm steps in the probe pulse after reduction

by GSH. Instead, when the probe force was applied the substrate reached its full extension

without evidence for residual structure, indicating that no folding had taken place. Kinetic data

confirmed that, while reduction by TRX did not interfere with refolding, reduction by GSH

significantly inhibited refolding at all refolding durations measured (Figure 4.2F). The high rate

105 of spontaneous release by TRX can explain why this reductase did not interfere with refolding.

However, further data was required to tell whether the GSH-mediated inhibition was related to its reductase activity or caused by non-specific interactions with the substrate.

Figure 4.2 Disulfide reduction by GSH, but not TRX, inhibits protein refolding. (A) A mechanical force was applied to an I27 32-75 polyprotein, where each domain contained a disulfide between residues 32 and 75, in a solution containing reduced TRX or GSH. (B-C) Unfolding of a substrate domain up to the disulfide was detected as an 11 nm extension step. Formation of a mixed disulfide (glutathionylation in C) was detected as a 14 nm extension step arising from disulfide cleavage. While TRX releases from the mixed disulfide on a millisecond timescale, the glutathionylated substrate in C cannot be resolved spontaneously. (D-E) Experimental recordings of the end-to-end length of the I27 32-75 substrate under force, showing 11 nm extension steps (solid arrowheads) indicating partial unfolding up until the disulfide, followed by 14 nm extension steps (diamonds) arising from the cleavage of the 32-75 disulfide in the substrate by TRX (D) or GSH (E). The diamonds in (E) mark single glutathionylation events. Whereas reduction by TRX allowed refolding to take place during an interval of no force (grey segment), as seen from the 25 nm unfolding steps in the probe pulse (hollow arrowheads); no refolding was observed after reduction by GSH, showing that glutathionylation inhibited folding. (F) Kinetics of refolding after reduction by TRX (black circles) or GSH (red squares). Data shown as mean ± s.e.m.

106

To detect any non-specific effects, we repeated the experiments using a pre-reduced I27 32-75 substrate. Because this substrate contained no disulfide bonds, thiol-disulfide exchange reactions could be ruled out under these conditions. With the reduced substrate, we observed no significant difference in refolding between the three experimental conditions (buffer only, TRX, GSH, see

Figure 4.3). The inhibition of folding by GSH is thus dependent on thiol-disulfide exchange.

4.5 Glutaredoxin rescues folding of S-glutathionylated proteins

To further test if glutathionylation had prevented refolding, we added Glutaredoxin (Grx) to the reaction buffer, in an attempt to reverse the inhibitory effect (Figure 4.3B). Grx is a ubiquitous enzyme that specifically catalyzes the de-glutathionylation of substrate proteins [99]:

+ → + ͍͍͊́ ́ͦͬ ͍͊͂ ͍́ͦͬ́

Experimental data collected with GSH and Grx showed 25 nm steps in the probe pulse, indicating rescue of substrate refolding (Figure 4.3A, C). Comparison of all refolding data at 5 s refolding time revealed that a catalytic amount of 45 µM Grx was sufficient to rescue folding back to the control level.

A potential risk in these experiments was that the 14 nm disulfide cleavage events seen in the unfolding pulse perhaps could have been caused by Grx, since this enzyme contains a reactive cysteine and thus in principle is able to reduce any disulfide. We therefore measured the Grx reactivity towards the 32-75 disulfide by unfolding the I27 32-75 substrate in the presence of reduced Grx. The results show that at concentrations up to 180 µM, the rate of Grx-mediated protein disulfide cleavage was negligible as compared to the equivalent rate of GSH (0.07±0.01

107 s-1 vs. 1.2±0.5 s -1). In line with these results, previous studies have reported a strong kinetic

preference of Grx towards protein-glutathione mixed disulfides [100]. We can thus safely

conclude that the 14 nm events observed in our experiments with GSH and Grx were due to

glutathionylation as opposed to direct reduction by Grx.

Figure 4.3 Glutaredoxin rescues folding of S-glutathionylated proteins. (A) Addition of a catalytic amount of Grx to the buffer rescued substrate refolding, as seen from the unfolding steps in the probe pulse (hollow arrowheads). (B) Schematic diagram of a degluathionylation reaction. (C) Quantification of the results, displayed normalized to a control experiment performed in the absence of GSH. The first three experiments from the left are controls performed on pre-reduced substrate without a disulfide. Under these conditions, no mixed disulfide formation could have taken place. Numbers in parentheses indicate GRX concentration in micromolar. ∆∆∆t Refold=5s in all experiments. Data shown as mean ± s.e.m.

4.6 Glutathionylation modulates protein mechanics in a native titin domain

If glutathionylation inhibits protein folding, this effect should not depend on the preceding

reaction path. As mentioned earlier, there are many reaction paths in vivo that lead to

108 glutathionylation of a protein, some of which do not require a protein disulfide [105, 204]. In fact, any solvent accessible thiol can be glutathionylated by exposure to oxidized glutathione

(GSSG):

+ → + ͍͊͂ ͍͍́́ ͍͍͊́ ͍́͂

An increasing number of functional glutathionylation targets have been identified to date [113-

116]. However the number of cysteines that can be classified as potential targets is limited by

their solvent accessibility. Many cysteine residues are buried in the hydrophobic core of folded

proteins, where they are inaccessible and therefore unreactive [200]. If these residues were to

become solvent exposed through mechanical unfolding, they would all be potential targets for

glutathionylation. In fact, a variety different proteins are known to unfold when they are subject

to physiological forces, including elastomeric muscle proteins [7, 197, 198], cytoskeletal

components [201, 205] and vascular factors [202]. In all of these cases, dynamic unfolding and

refolding play critical roles in natural protein function. As a consequence, cysteines that are

normally buried in these proteins can intermittently become exposed through protein unfolding

[201]. Having found that glutathionylation blocked protein refolding, we speculated that strain-

dependent glutathionylation could act as a modulator of force bearing proteins.

We used a polyprotein consisting of repeated wild-type I27 domains to test whether

glutathionylation affected its elastic properties. I27 from human cardiac titin is one of the best-

characterized model proteins for mechanical folding and unfolding [140, 143, 145, 206, 207].

The atomic structure of I27 reveals two solvent inaccessible cysteines in positions 47 and 63,

which at a separation distance of 9.2 Å are too far apart to form a disulfide (Figure 2.10)[140,

207].

109

Through mechanical unfolding-refolding experiments, we investigated whether the cryptic cysteines in I27 were targets for glutathionylation by GSSG upon unfolding (Figure 4.4A-B). If

GSSG could react with the cysteines in the substrate only in the unfolded state, then the effect of glutathionylation should be dependent on the duration of the mechanical strain. We therefore measured the refolding of I27 in the presence of GSSG after varying amounts of time of mechanical strain (hereafter referred to as exposure time). The 25 nm steps seen in these experiments indicate natively folded domains. In the first experiment, the initial unfolding pulse

was maintained for 2 s, and refolding was then probed after a 5 s refolding interval in the

absence of force (Figure 4.4B, top trace). Nearly all domains refolded in this case. In contrast, no

refolding was observed after a 40 s unfolding pulse (Figure 4.4B, bottom trace). This exposure

time dependent effect suggests that GSSG acts only on the unfolded protein. Figure 4.4C shows

a summary of all our experiments. The exposure time did not have any effect on the refolding of

I27 WT in only buffer. Exposure to an equivalent amount of reduced GSH did not affect folding,

which shows that the inhibitory effect was not a result of non-specific interactions between

glutathione and the protein. As a final control, we measured the refolding of an I27 cysteine

deletion mutant (I27 ∆Cys) in the presence of GSSG, but could not detect any effect. Taken

together, these results conclusively prove that GSSG inhibits I27 refolding through

glutathionylation of cryptic cysteines.

A close inspection of the unfolding events in the probe pulse seemed to indicate that some domains unfolded at a higher rate than normal after exposure to GSSG. This would suggest that some refolded domains were mechanically weaker than native I27. To investigate this, we altered the pulse protocol as follows: A low force probe pulse of 110 pN was applied to the

collapsed protein immediately following the refolding interval. After 3 s at this low force, we

110 applied the normal 170 pN probe pulse. The purpose of the low force pulse was to selectively

unfold weaker domains, while the native domains would unfold during the high force pulse.

Figure 4.4 GSSG inhibits protein refolding through glutathionylation of cryptic cysteines in wild-type I27. (A) Glutathionylation of the cryptic cysteines in I27 is enabled by mechanical unfolding. (B) Single molecule refolding recordings of I27 in the presence of 100 mM GSSG. Top trace: After 2 s exposure in the unfolded state, nearly all domains are able to unfold. Bottom trace: After 40 s exposure in the unfolded state, no refolding was observed. (C) Refolded fraction as a function of exposure time. Longer times under mechanical strain led to inhibition of refolding only in the presence of GSSG. A cysteine deletion mutant of I27 did not show any inhibition in the presence of GSSG, showing that the inhibition is due to cysteine-specific glutathionylation. The refolding interval in all experiments was set to 5s ( ∆∆∆t=5 s).

After refolding I27 in buffer, we observed no unfolding during the low force pulse, whereas all domains rapidly unfolded after the normal high force pulse was applied. This result verified that

3s at 110 pN is not sufficient to induce any measurable unfolding of native I27. The top trace in

Figure 4.5A shows similar results after 2s GSSG exposure. However, with increased exposure to

111

GSSG, some unfolding steps began to appear in the low force pulse. After 40 s of exposure to

GSSG in the extended state, only very rarely were unfolding steps seen in the probe pulse,

however all of these took place at the low force. These results show that exposure to GSSG

initially leads to a weaker refolded state, and later prevents refolding altogether.

Figure 4.5 Strain dependent modulation of I27 stability. (A) The stability of refolded wild type I27 domains was probed after varying times of GSSG exposure. A two-part probe pulse was used, where a 3s low force pulse of 110 pN was followed by a high force pulse of 170 pN. Natively folded I27 domains did not unfold during the low force pulse, however a population of weak domains started appearing as the protein was exposed to GSSG. (B) Fraction of domains in the probe pulse with a native fold (red), with a weak structure (blue) and unfolded domains (green, right axis), as a function of GSSG exposure time. All three data sets could be fit to a single three-state consecutive reaction model (shown in C), with a total of three free parameters (solid lines). (C) Consecutive glutathionylation reactions in the extended state. Note that the first reaction is degenerate, which can explain why the first glutathionylation event was measured to occur twice as fast as the second. (D) Mechanical strain enables modification of cryptic cysteines. Under conditions of high stress, cysteine modification in the form of S-glutathionylation modulates protein mechanics in a graded and reversible manner.

112

4.7 A kinetic model for redox regulation of protein stability

Since wild-type I27 contains two cysteines, we hypothesized that these weaker domains had been glutathionylated on one of the two cysteines, while glutathionylation of both cysteines would result in a complete failure to fold. We classified the domains according to the force at which they unfolded. We also counted the domains that had failed to fold, by subtracting the unfolding steps in the probe pulse from the unfolding steps in the denature pulse. Figure 4.5B shows the relative fraction of these three populations (native, weak, unfolded) as a function of GSSG exposure time. We could fit the data with a model consisting of two consecutive and irreversible glutathionylation reactions (Figure 4.5B-C). Strikingly, this model contains only two free parameters, k1 and k2, representing the first and second glutathionylation reaction, respectively.

-1 -1 - We note that the value of k1 was measured to be exactly twice that of k2, at 1.4 s M vs. 0.7 s

1M-1. This is consistent with the two possible reaction sites (two exposed cysteines) available for

the first glutathionylation reaction. Meanwhile, the second reaction can only take place at one

site. The average first-order rate of a single glutathionylation reaction is thus 0.7 s -1M-1. This

value is directly comparable to earlier measurements of glutathionylation rates in an unstructured

peptide [121]. In this study, Darby and coworkers reported first-order rates of 0.5-3.0 s-1M-1 for the glutathionylation of exposed cysteine residues.

We conclude that glutathionylation of cryptic cysteines inhibits the native refolding of an elastomeric protein. Combined with the ability of mechanical strain to expose cryptic sites, this mechanism enables strain-induced modulation of protein mechanics (Figure 4.5D).

113

4.8 Discussion and biological implications

Post-translational modifications play a critical role in cell regulation and signaling. Modifications such as phosphorylation, acetylation, methylation, ubiquitinylation and SUMOylation transmit information by altering protein function and/or structure. We found that glutathionylation of cryptic cysteines inhibits protein folding, representing (to the best of our knowledge) the first reported instance of a mechanochemical post-translational modification.

The strain-dependent effect of glutathione on protein mechanics can be likened to an effect seen in local anesthetics such as lidocaine. Discovered by the Swedish chemist Nils Löfgren, lidocaine inhibits sensory neurons by blocking voltage gated sodium channels. Since the drug preferentially binds to the open state of the channel (a property known as state-dependent binding) the inhibitory effect is stronger in neurons that are more active [208]. Lidocaine therefore selectively inhibits neurons in areas experiencing pain, which leads to the exceptional pharmacological effectiveness of the drug.

Similarly, we observed that the GSSG-mediated folding inhibition was enhanced as a

protein spent more time in the unfolded state (Figure 4.4C). Therefore, this redox regulation of

protein mechanics would selectively affect strained tissue. In this tissue, glutathionylation would

cause an increase in compliance.

Strain-dependent inhibition of protein folding could have many different functions in vivo .

In the most straightforward example, it would allow elastic tissue such as muscle to dynamically adapt to a wide range of conditions without the requirement for protein synthesis or digestion. In this scenario, titin molecules that are exposed to a high level of strain would be glutathionylated in a few domains, which in turn would create a new resting length and relieve the muscle of potentially damaging strain.

114

The inhibition of folding could also trigger downstream effects that could include signaling cascades. For instance, many proteins involved in mechanosensing are known to contain cryptic functional sites that are exposed through mechanical strain [209]. These sites include binding sites [210] and kinase domains [211]. Glutathionylation could in these cases modulate the force- response curve of signaling, providing an additional regulatory layer in mechanosensing. By transmitting information through modulation of protein stability, this mechanism may play a role mechanochemical signal transduction in cells.

In a vivid final example, unfolded proteins are easy targets for degradation by proteases.

Glutathionylation of cryptic cysteines could therefore function as a protein marker (like ubiquitinylation) that facilitates their degradation. It was recently shown, through a series of elegant experiments, that the bacterial proteasome uses mechanical force to unfold proteins prior to proteolysis [6, 20]. In this scenario, as a protein is struggling for survival at the mouth of the proteasome, glutathionylation of a transiently exposed cryptic cysteine could rapidly turn the odds against the protein substrate.

4.9 Materials and methods

4.9.1 Single-molecule Atomic Force Microscopy

The details of our custom-made atomic force microscope have been described previously [17].

We used silicon nitride cantilevers (Bruker) with a typical spring constant of 15 pN nm -1, calibrated using the equipartition theorem [185]. The force-clamp mode provided a feedback time constant of 5 ms. The buffer used in all the experiments was 10 mM HEPES, 150 mM

NaCl, 1 mM EDTA, degassed and at pH 7.2. The polyprotein substrate was added in a droplet

115 and allowed to bind to a freshly evaporated gold coverslip before the experiments. Every experiment consisted of repeated trials where the tip was pressed against the surface for 0.5 s and subsequently retracted. If attachment was achieved, the pulse protocol was applied until detachment occurred. The refolding force-clamp experiments used a triple-pulse [unfolding- refolding-probe ] force protocol (as described in Section 2.3.7). The first pulse was maintained at

150-170 pN for a time long enough to ensure complete unfolding of the substrate (at least 2 s).

The second pulse was set at 0 pN and maintained for the desired amount of refolding time. The third pulse was set at a force identical to the first and maintained until complete unfolding could be ensured (at least 5 s).

4.9.2 Data Analysis

We collected and analyzed data using custom-written software in IGOR Pro (Wavemetrics).

Traces were selected based on the fingerprint consisting of at least 3 unfolding events in the denature pulse. Traces exhibiting other step sizes in the denature pulse were excluded from the analysis to ensure homogeneity. Only traces showing equal extension at the end of the probe pulse and the end of the denature pulse were included, to ensure that the same protein was stretched in the two pulses. SEM was estimated through the bootstrap method.

116

Chapter 5 Conformational Tuning of Cysteine Reactivity

117

5 Conformational Tuning of Cysteine Reactivity

Protein function is often dependent on the protonation state of ionizable groups. Enzymes acquire highly coordinated structures that enable precise modulation of protonation equilibria.

Although much work has been directed at the intramolecular factors in such modulation, the influence of substrates on enzyme protonation remains relatively unexplored. We used single molecule force spectroscopy to show that substrate conformation can have a significant effect on the protonation of a catalytic thiol in thioredoxin (TRX), a universal disulfide reductase. p H titration of the enzyme thiol revealed a shift in pKa from 6.4 to 7.1 in response to strain-induced substrate conformational change. Our experimental results suggest a new mechanism for how substrate conformation can modulate enzyme function.

5.1 The role of pKa in thioredoxin catalysis

Electrostatic interaction between proximal chemical groups is one of the fundamental components that enable the complex biochemistry of life. The local chemical environment can have a profound effect on protonation equilibria, affecting protein stability and function. As a consequence, the p Ka of an ionizable group free in solution can be significantly different from the same group in a folded protein [212]. In enzymes, this effect can be critical for catalysis. One such case is thioredoxin, a universally conserved catalyst of disulfide reduction [53]. Its active site contains a CGPC motif that includes two catalytic cysteines. Thioredoxin cleaves substrate disulfides through a nucleophilic attack by the first catalytic cysteine, Cys32. Thiol deprotonation is required for this reaction to take place [53]. Thioredoxin activity is thus dependent on the protonation state of Cys32. Whereas free cysteine has a thiol p Ka of around 8.7

118

[55], Cys32 in thioredoxin has been shown to exhibit a significantly depressed p Ka. Electrostatic interactions within the enzyme active site are thought to cause this depression, which ensures high activity at physiological p H [213]. Despite much effort, there is still no consensus on the

exact p Ka of this active site thiol. For E. coli thioredoxin, values ranging from 6.3 to 9.9 have

been reported in different investigations using assays that employed different substrates, and in

some cases no substrate [54, 214-219]. For human thioredoxin, a p Ka of 6.3 ±0.1 was measured

using NMR [220]. The lack of consensus has been explained by invoking “microscopic” p Ka values, where the apparent p Ka of Cys32 is affected by the protonation state of neighboring

atoms [213]. However despite these advances, our understanding of p Ka modulation remains incomplete.

We speculated that the substrate might influence enzyme protonation. Thioredoxin, like most enzymes, forms an encounter complex with its substrate prior to catalysis [25]. The substrate is coordinated in a binding groove adjacent to the active site [221] (Figure 5.1). During the lifetime of the encounter complex, the local environment of the catalytic thiol is thus likely to include intermolecular components.

Figure 5.1 Structure of TRX (grey) in a mixed disulfide complex with a substrate polypeptide (blue). The two cysteines in the mixed disulfide are shown in yellow (TRX Cys32 to the left). Although this complex corresponds to a time point after the initial reaction has taken place, it is likely a good approximation for the encounter complex prior to catalysis. We hypothesized that Cys32 in TRX can be influenced by the substrate in the encounter complex. (PDB: 1mdi)

119

5.2 Single molecule assay for pKa determination

By measuring the reactivity of Cys32 in TRX as a function of pH, we were able to determine its pKa. We used an atomic force microscope in force-clamp mode to measure the kinetics of disulfide cleavage by thioredoxin [24, 25]. Our substrate consisted of sequential repeats of human cardiac I27 domains, each with a buried disulfide bond between residues 32 and 75

(I27 32-75 ). In each recording, a single substrate molecule was attached between a gold surface and

the silicon nitride probe tip (Figure 5.2A). We then applied a series of two constant calibrated

force pulses to the substrate while measuring its extension. The initial force was set to 170 pN

for 300 ms to rapidly unfold the substrate, which rendered the disulfides solvent exposed. During

this pulse, each substrate domain unfolded independently up to the disulfide bond, yielding a

series of 11 nm stepwise extension increases. This staircase constituted a fingerprint of single

molecule attachment. No further steps were seen in the absence of thioredoxin, as the applied

force was not sufficient to break the covalent bonds. If reduced thioredoxin was present in

solution, individual disulfide cleavage events were detected during the second, “reduction” pulse.

Each cleavage reaction could be detected through a 14 nm stepwise extension of the substrate as

the polypeptide segment behind the disulfide was released (Figure 5.2B, see also Section 2.3). A

schematic depiction of the experiment is presented in Figure 5.2B and a representative

experimental trace is displayed in Figure 5.2C. By normalizing and summing the reduction pulse

of many such traces, we obtained a precise measure of the disulfide cleavage rate (Figure 5.3A).

120

Figure 5.2 (A) Single substrate proteins consisting of repeated I27 32-75 domains were subject to force in an atomic force microscope, while measuring their extension. (B) Partial unfolding of a domain yields a 11 nm step and exposes the buried disulfide. Nucleophilic attack by deprotonated Cys32 in TRX cleaves the substrate disulfide, yielding a 14 nm step. (C) Representative trace showing single disulfide cleavage events (arrowheads) at 100 pN, pH 7.2.

Thioredoxin activity is known to be dependent on Cys32 deprotonation [53]. The rate scaled linearly with enzyme concentrations up to at least 10 µM, allowing us to use the reaction rate as

an indicator of the overall protonation state of Cys32. The measured reaction rate was indeed

seen to vary significantly at different p H (Figure 5.3A).

The Henderson-Hasselbalch equation was used to relate thiol deprotonation to the solution pH:

ͯ = + log ʞ͍ ʟ ͤ͂ ͤͅ ʦ ʧ ʞ͍͂ ʟ

Where [ S-] and [ SH ] are the concentrations of the deprotonated and protonated form of

thioredoxin Cys32, respectively.

121

The fraction of active, deprotonated enzyme then depends on the solution p H as:

ͯ 1 = ʞ͍ ʟ = ͚ʚͤ͂ ʛ ͯ + 1 + 10 + Ĕͯ+ ʞ͍ ʟ ʞ͍͂ ʟ

The thioredoxin reaction rate was in an earlier work shown to be force dependent, with an

increase in force from 50 to 200 pN leading to a decrease in rate [25]. This effect is due to the

force discouraging the transition state conformation [25, 221]. We postulated that this effect

would be independent of p H. If this was the case, we would be able to describe the force and p H dependency as follows:

, = ∙ = ͌ͤʚ̀ʛ ͌ʚ̀ ͤ͂ ʛ ͌ͤʚ̀ʛ ͚ʚͤ͂ ʛ 1 + 10 + Ĕͯ+

Where R0 is the maximum reaction rate (at complete deprotonation) at a given force F, and f is the fraction of deprotonated (active) enzyme. Although the effects of p H on reaction rate may be more complex, this is the simplest model that is well grounded in theory.

Figure 5.3B shows the data and least-squares fits of this equation at two different forces. The results could be completely accounted for by the model, which lends credibility to our assumptions.

Figure 5.3 (A) Summed and normalized traces at 50 pN and pH 8 ( N=61 traces) and pH 6 ( N=25 traces) reveal single exponential reaction kinetics but different rates. (B) The pH dependency of nucleophilic attack shows different apparent p Ka values at 50 pN (pKa=6.6±0.1) and 150 pN (pKa=7.1±0.1). Errorbars indicate s.e.m.

122

5.3 Substrate conformation alters the pKa of Cys32 in thioredoxin

Strikingly, the fits at different straining forces returned different apparent p Ka values. This is remarkable, because the force was applied to the substrate whereas the p Ka is conventionally

considered a property of the enzyme alone. Any enzyme interaction with the substrate occurs

after the substrate is strained; it therefore seems unlikely that the substrate is exerting any

mechanical force on the enzyme. The most likely explanation is that the conformation of the

substrate influences the protonation state of the enzyme in the encounter complex prior to

nucleophilic attack.

Figure 5.4 shows the measured p Ka values at five different straining forces. At forces up to

100 pN there is a clear trend of increase in p Ka with the force applied to the substrate. At higher forces, however, there is little or no change as the force is increased. This supports the idea that altered substrate conformation is causing the shift in p Ka, as the force effect on the conformation

of an unfolded protein is known to reach an asymptote at high strain (Figure 2.11, see also

Section 2.3.1) [139, 147]. In fact, the conformational change of the substrate at 100 pN has

already exceeded 80% of its high-force asymptote, as given by the wormlike chain model [139].

Further validating our method, the trend in p Ka extrapolated to zero force appears to coincide

with an earlier substrate- and force-free measurement (p Ka=6.3 ±0.1) indicated with a solid line in

Figure 5.4 [220].

123

Figure 5.4 Force dependent p Ka of Cys32 in human TRX. The bulk datapoint was measured using NMR in the absence of substrate (ref [220]).

5.4 Kinetic model for substrate dependent pKa in an enzyme

The finding that the substrate straining force can shift the catalytic cysteine p Ka in thioredoxin implies that proton exchange took place during the timespan of enzyme-substrate interaction prior to nucleophilic attack. In a previous investigation, the encounter complex between thioredoxin and the I27 32-75 substrate had a measured lifetime on the order of milliseconds [25].

By comparison, a reasonable estimate for the timescale of thiol proton exchange is given by

general acid-base dynamics in water, which take place on the sub-nanosecond timescale [222].

Our data is therefore consistent with a scenario in which the Cys32 protonation state reaches a

new equilibrium during the lifetime of the enzyme-substrate encounter complex (Figure 5.5).

The p Ka of an ionizable group can be altered by electrostatic interactions with proximal

groups. For Cys32 in thioredoxin, these are thought to include neighboring residues in the active

site [213, 220]. However, when considering the timescales of relevant processes during catalysis

(vide supra ), it seems that previous investigations have left out half of the picture, namely, the

substrate. Since proton exchange is orders of magnitude faster than most macromolecular

124 interactions, the relevant environment of the catalytic thiol is really that of the enzyme-substrate encounter complex (Figure 5.5).

Figure 5.5 The pKa of Cys32 in TRX is different in the encounter complex from its free form in solution. This situation arises because the vertical reactions that involve proton movements have much faster kinetics than the horizontal reactions that involve macromolecular rearrangements.

The net free energy change around the reaction cycle in Figure 5.5 has to be zero. Furthermore, protonation appeared to be favored in the bound state relative to the unbound state. Therefore, the increased pK a can be explained by the protonated enzyme binding the substrate more tightly

than the unprotonated enzyme. In other words, burying a thiolate anion in the binding interface

appears to decrease the strength of binding. This is hardly surprising as the binding interface

implicated in thioredoxin substrate recognition is considered to be mainly hydrophobic {Krause,

1991 #4568} (see also Section 1.5.1).

5.5 Discussion and biological implications

In human physiology, thioredoxin-mediated disulfide reduction has many essential functions

both intra- and extracellularly. [223, 224]. The low specificity of the enzyme allows it to reduce

disulfides in many different substrates, of which an increasing number are being identified [223,

224]. Although the particular substrate conformations sampled in our experiments are not

necessarily identical to those in vivo , the substrate influence on p Ka is likely a general feature that could have important consequences. This effect may contribute to the variation in catalytic

125 rates measured with different substrates [223], but may also cause a modulation of the catalytic rate for the same substrate if it samples a number of different conformations. Such conformational tuning of enzyme activity would be especially relevant for substrates placed under mechanical stress, such as proteins in muscle and connective tissue. Furthermore, the implications are not limited to thioredoxin; substrate induced p Ka modulation may indeed be

relevant for any enzyme that depends on a certain protonation state for catalysis.

126

Chapter 6 Perspective and Future Directions

127

6 Perspective and Future Directions

In this thesis, I have presented a series of mechanochemical studies that shed light on protein biochemistry. With the use of an AFM and a range of creative protocols, I was able to use mechanical force to dissect the intricacies of thiol chemistry in single protein molecules. My studies provide fundamental insights into processes including enzyme catalysis, oxidative folding and mechanochemical regulation. Most importantly, our new mechanochemical toolkit lays the foundation for future studies in single molecule biochemistry.

In Section 2, I introduced the instrumentation, consisting of a custom-built force-clamp

AFM. The force feedback circuit enables experiments where the force on a single molecule is

held constant while the molecule’s extension is measured. Constant force operation is a unique

feature that allows for the design of pulse protocols to assay aspects of protein biochemistry

inaccessible with any other technique. For instance, native protein structure can be detected

through its mechanical resilience and its fingerprint extension upon unfolding. Disulfide bonds

can be detected in an equivalent manner. I used these features to develop the first single

molecule assay of oxidative protein folding, as described in Section 3.

The mechanochemical assay of oxidative folding I presented in Section 3 is groundbreaking

in two different aspects. It is the first assay of oxidative folding in single molecules. This feature

completely removes the issue of sample heterogeneity, a common obstacle in previous studies.

Furthermore, my assay is also the first one that enables independent kinetic measurements of

protein folding and disulfide formation. This feature for the first time allows us to tease apart

cause and effect during oxidative folding. I used my assay to determine the sequence of events

during PDI-catalyzed oxidative folding, and found that protein folding provides the driving force

behind disulfide formation. I found that PDI, when bound to its substrate, acts as a passive

128 placeholder that allows the substrate to fold in an unhindered fashion. The substrate is therefore afforded the time and opportunity to decide which cysteines to join. Allowing the substrate to call the shots ensures the correct outcome of an otherwise highly labile process where mistakes can have dire consequences.

Another common thiol modification was studied in Section 4, where I demonstrated direct

detection of individual S-glutathionylation events. Although S-glutathionylation is recognized as

a vital component of redox regulation in cells, it has been unknown to what extent cryptic

cysteines are targeted. I found that cryptic cysteines are susceptible to glutathionylation in a

strain-dependent manner. Strikingly, glutathionylation of these cysteines inhibited protein

folding. My results further showed that this modulation of protein mechanics was both graded

and reversible. These findings open doors for new avenues of mechanochemical regulation in

cells and tissue.

In the study presented in Section 5, I investigated the effect of substrate conformation on enzyme activity. I found that the p Ka of a critical cysteine in the active site of thioredoxin could

be modulated by the substrate conformation. My findings show that structural features of the

encounter complex can effectively regulate enzyme activity.

The following section expands on my work and will hopefully inspire future studies.

6.1 Single molecule studies of protein mechanics in muscle

Just like electric charge or electromagnetic radiation, today we know that tissue elasticity at the

molecular level is quantized. Single molecule data have shown that proteins respond to a

stretching force by unfolding in a stepwise fashion. We also know that proteins can recover their

folded state when the force is decreased or removed. Taken together, we now have a strong basis

129 for understanding protein mechanics on a fundamental, single molecule level. These insights allow us to make predictions for protein mechanics in intact tissue. Fundamental responses such as force-induced unfolding are likely to occur in tissue as well. However the complex nature of tissue may harbor emergent features that have yet to be discovered. It would therefore be extremely interesting to study the discrete nature of protein mechanics in situ or even in a living organism.

Myofibrils from striated muscle are perfect candidates for such studies. Their almost crystalline organization of sarcomeres makes them highly isotropic, and all of the main proteins are aligned in parallel (Figure 6.1A). A longitudinally applied force (along the orientation of the filaments) thus becomes homogenously distributed within the fibril. This force would be entirely transmitted through titin in an isolated and relaxed myofibril. In practice, this system is like a massively scaled-up version of the single molecule AFM experiments presented in this thesis.

The difference, however, is that the thermal motion of the large number of titin molecules in parallel and in series would obscure any steps, if we measured the length of the entire myofibril.

A strategy for detecting individual unfolding steps would therefore need to measure the extension of only a tiny fragment of the fibril.

Our strategy is outlined in Figure 6.1. The myofibril is glued between two microneedles, one of which is movable, and observed through a fluorescence microscope. The myofibril is then labeled with primary antibodies against specific epitopes in the proximal region of titin. Since titin is attached symmetrically around the Z-disc, two symmetrically spaced epitopes will be separated by less than a hundred titin Ig domains. Secondary antibodies conjugated to quantum dots are then linked to the primary antibodies, enabling localization with nanometer-scale precision in the fluorescence microscope.

130

At a high antibody concentration, the labeling is seen as two strong bands spaced equidistant

around each Z disc in the myofibril (Figure 6.1B). These bands move apart when a force is

applied through the micropipette, clearly illustrating titin elasticity. If the antibody concentration

is decreased sufficiently, there is a chance to obtain a sample with two single quantum dots

spaced around a single Z-disc (Figure 6.1C). When this is achieved, we would expect to see

signatures of discrete states in the spacing of the quantum dots over time, in response to a

stretching force. This project is still in progress but the results so far appear to be promising.

Figure 6.1 Measuring single molecule protein mechanics in an intact myofibril. (A) Arrangement of elastic titin segment in a muscle sarcomere and fluorescent labelling strategy. Top panel is an electron micrograph of a sarcomere; bottom panel is a schematic of two half I-band regions of titin. I39 and N2A are epitope positions of titin antibodies used in this experiment. The epitope positions of the primary antibody were visualized by secondary antibody conjugated to a fluorescent quantum dot, and the Qdot-separation distance was tracked. Inset: DIC images of a single myofibril glued to the tips of glass microneedles, one of which was attached to a piezoelectric micromotor to facilitate stretch. (B) Striated staining pattern obtained with N2A- antibody. During myofibril stretch the titin epitopes moved apart, i.e. away from the Z-disc (Z). (C) Lower secondary antibody-Qdot concentration allowed detection of individual Qdots. Primary antibody was anti- I39. Shown are images of the same myofibril before and after stretch (left, DIC; middle, fluorescence; right, overlay). Arrows point to Z-disc region and Qdot pair. Scale bars, 5 µm.

131

6.2 Measuring the fighting power of a protein with Isometric Force Microscopy

What force can a protein exert in a given conformation? This question is of fundamental importance for understanding motor proteins as well as elastomeric proteins such as titin. The question has been answered indirectly through such measurements as the stalling force of molecular motors and the force-extension or force-ramp curves of protein folding. However, all of these approaches present significant complications. In the case of motor proteins with a power stroke, the stalling force represents the force at the weakest point during the stroke. When the protein is overpowered at its weakest, it will not be able to move past this position. In the case of protein folding, every study to date has used a flexible probe, meaning that the probe will give way if the protein suddenly exerts a high force, thus obscuring the force exerted by the protein.

I propose to address this question through the development of Isometric Force Microscopy

(or IsoForM, for short). IsoForM will enable position-clamp measurements on single proteins, wherein protein extension is held constant while the exerted force is measured in real-time. An implementation of IsoForM can be built through a simple modification of a force-clamp AFM.

Here follows a brief description.

The extension of the protein ( L) relates to the piezo position ( x) and the detector difference

signal ( VD) as:

= + ͐ʚͨʛ ͆ʚͨʛ ͬ ͟

Where kPD is a scaling constant (see Section 2.1.11). If L is to be held constant at a value LSP , then we need to set the piezo position x as follows:

1 = − ∙ ͬʚͨʛ ͆ ͐ʚͨʛ ͟

132

This can be achieved by feeding the piezo a setpoint signal that is summed with the detector difference signal amplified by an appropriate gain.

One of the central features that can be elucidated using IsoForM is whether a collapsed

protein can undergo folding-unfolding transitions at a constant length, or reach partially folded

states. It will also be interesting to see on what timescale the protein equilibrates at a certain

length. If the substrate is a polyprotein that is constrained at a length that is too long for all of the domains to stay folded, will the domains take turns in unfolding or will one domain stay unfolded all the time? All of this is relevant in vivo in muscle, where titin molecules are

constrained between Z-discs that are too large to budge from the force generated by a single

protein. At a constant sarcomere length, what happens to protein folding?

Furthermore, since the force exerted by a protein is equivalent to the gradient of its potential

energy, IsoForM can be used to map the entire energy landscape of an extended protein, as a

function of end-to-end length. This is not possible with conventional force spectroscopy, because

when the position is not clamped, the protein can skip over energy barriers (as shown in Figure

2.12) and dissipate heat, which is irreversible. With IsoForM, the extension can be changed at an

arbitrarily slow rate, thus probing the actual equilibrium free energy of the molecule.

6.3 Mechanochemical signaling through cysteine modifications

I found that S-glutathionylation, a common component in redox signaling, could alter protein

mechanics when targeted to cryptic cysteines. This raises the question if other post-translational

modifications might have a similar effect on the mechanical properties of proteins. The answer is

almost certainly yes. In addition to glutathionylation, there are many other cysteine modifications

that play important roles in signaling pathways, including nitrosylation, sulfenylation and

133 sulfhydrylation. Furthermore, non-cysteine modifications such as phosphorylation, acetylation, ubiquitinylation, SUMOylation and many others may also modulate protein stability if they are conjugated to cryptic sites. Using the methods I have developed here, the mechanochemical effect of all of these modifications can now be quantified. Due to the sheer number of reactive moieties that are exposed in tissue under strain, the mechanochemical modulation described in our study most definitely will play an important role in vivo .

An interesting synergy arises from the strain-dependence of the stability modulation.

Because glutathionylation alters the response to subsequent strain, the mechanism provides a form of feedback. The effect of this can be two-fold. In proteins with several cryptic cysteines, such as I27, glutathionylation of one site means that subsequent glutathionylation of the second site is more likely (due to the decreased stability of the protein domain). This represents a positive feedback system. However, in a polyprotein with several domains in series, glutathionylation of one domain would lead to that domain unfolding easier in the future, thus taking the load off from the neighboring domains, which in turn would lead to a decreased probability of glutathionylation in these domains. Therefore, glutathionylation of one domain would tend to suppress glutathionylation of adjacent domains in the same protein.

One potential application of the strain-dependence of glutathionylation could be as a research tool to measure strain and unfolding of individual proteins in tissue.

Another potential application of my findings is that cysteine modification agents could be used as pharmaceutical drugs to alter tissue mechanics. By relieving tension in strained tissue, such mechanochemical compounds could perhaps one day be used to treat conditions like arterial hypertension.

134

References

1. Becker, T., S. Bhushan, A. Jarasch, J.P. Armache, S. Funes, F. Jossinet, J. Gumbart, T. Mielke, O. Berninghausen, K. Schulten, E. Westhof, R. Gilmore, E.C. Mandon, and R. Beckmann, Structure of monomeric yeast and mammalian Sec61 complexes interacting with the translating ribosome. Science, 2009. 326 (5958): p. 1369-73.

2. Lu, H.P., L. Xun, and X.S. Xie, Single-molecule enzymatic dynamics. Science, 1998. 282 (5395): p. 1877-82.

3. Funatsu, T., Y. Harada, M. Tokunaga, K. Saito, and T. Yanagida, Imaging of single fluorescent molecules and individual ATP turnovers by single myosin molecules in aqueous solution. Nature, 1995. 374 (6522): p. 555-9.

4. Vale, R.D., T. Funatsu, D.W. Pierce, L. Romberg, Y. Harada, and T. Yanagida, Direct observation of single kinesin molecules moving along microtubules. Nature, 1996. 380 (6573): p. 451-3.

5. Smith, D.E., S.J. Tans, S.B. Smith, S. Grimes, D.L. Anderson, and C. Bustamante, The bacteriophage straight phi29 portal motor can package DNA against a large internal force. Nature, 2001. 413 (6857): p. 748-52.

6. Aubin-Tam, M.E., A.O. Olivares, R.T. Sauer, T.A. Baker, and M.J. Lang, Single- molecule protein unfolding and translocation by an ATP-fueled proteolytic machine. Cell, 2011. 145 (2): p. 257-67.

7. Li, H., W.A. Linke, A.F. Oberhauser, M. Carrion-Vazquez, J.G. Kerkvliet, H. Lu, P.E. Marszalek, and J.M. Fernandez, Reverse engineering of the giant muscle protein titin. Nature, 2002. 418 (6901): p. 998-1002.

8. Ha, T., I. Rasnik, W. Cheng, H.P. Babcock, G.H. Gauss, T.M. Lohman, and S. Chu, Initiation and re-initiation of DNA unwinding by the Escherichia coli Rep helicase. Nature, 2002. 419 (6907): p. 638-41.

9. Zhuang, X., L.E. Bartley, H.P. Babcock, R. Russell, T. Ha, D. Herschlag, and S. Chu, A single-molecule study of RNA catalysis and folding. Science, 2000. 288(5473): p. 2048- 51.

135

10. Uemura, S., C.E. Aitken, J. Korlach, B.A. Flusberg, S.W. Turner, and J.D. Puglisi, Real- time tRNA transit on single translating ribosomes at codon resolution. Nature, 2010. 464 (7291): p. 1012-7.

11. Eggeling, C., C. Ringemann, R. Medda, G. Schwarzmann, K. Sandhoff, S. Polyakova, V.N. Belov, B. Hein, C. von Middendorff, A. Schonle, and S.W. Hell, Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature, 2009. 457 (7233): p. 1159-62.

12. Rust, M.J., M. Bates, and X. Zhuang, Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM). Nature methods, 2006. 3(10): p. 793-5.

13. Betzig, E., G.H. Patterson, R. Sougrat, O.W. Lindwasser, S. Olenych, J.S. Bonifacino, M.W. Davidson, J. Lippincott-Schwartz, and H.F. Hess, Imaging intracellular fluorescent proteins at nanometer resolution. Science, 2006. 313 (5793): p. 1642-5.

14. Blanchard, S.C., H.D. Kim, R.L. Gonzalez, Jr., J.D. Puglisi, and S. Chu, tRNA dynamics on the ribosome during translation. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101 (35): p. 12893-8.

15. Roy, R., S. Hohng, and T. Ha, A practical guide to single-molecule FRET. Nature methods, 2008. 5(6): p. 507-16.

16. Fei, J., P. Kosuri, D.D. MacDougall, and R.L. Gonzalez, Jr., Coupling of ribosomal L1 stalk and tRNA dynamics during translation elongation. Molecular Cell, 2008. 30 (3): p. 348-59.

17. Fernandez, J.M. and H. Li, Force-clamp spectroscopy monitors the folding trajectory of a single protein. Science, 2004. 303 (5664): p. 1674-8.

18. Shank, E.A., C. Cecconi, J.W. Dill, S. Marqusee, and C. Bustamante, The folding cooperativity of a protein is controlled by its chain topology. Nature, 2010. 465 (7298): p. 637-40.

19. Garcia-Manyes, S., L. Dougan, C.L. Badilla, J. Brujic, and J.M. Fernandez, Direct observation of an ensemble of stable collapsed states in the mechanical folding of ubiquitin. Proc Natl Acad Sci U S A, 2009. 106 (26): p. 10534-9.

136

20. Maillard, R.A., G. Chistol, M. Sen, M. Righini, J. Tan, C.M. Kaiser, C. Hodges, A. Martin, and C. Bustamante, ClpX(P) generates mechanical force to unfold and translocate its protein substrates. Cell, 2011. 145 (3): p. 459-69.

21. Kim, J., C.Z. Zhang, X. Zhang, and T.A. Springer, A mechanically stabilized receptor- ligand flex-bond important in the vasculature. Nature, 2010. 466 (7309): p. 992-5.

22. Cao, Y., M.M. Balamurali, D. Sharma, and H. Li, A functional single-molecule binding assay via force spectroscopy. Proc Natl Acad Sci U S A, 2007. 104 (40): p. 15677-81.

23. Le Trong, I., P. Aprikian, B.A. Kidd, M. Forero-Shelton, V. Tchesnokova, P. Rajagopal, V. Rodriguez, G. Interlandi, R. Klevit, V. Vogel, R.E. Stenkamp, E.V. Sokurenko, and W.E. Thomas, Structural basis for mechanical force regulation of the adhesin FimH via finger trap-like beta sheet twisting. Cell, 2010. 141 (4): p. 645-55.

24. Alegre-Cebollada, J., R. Perez-Jimenez, P. Kosuri, and J.M. Fernandez, Single-molecule Force Spectroscopy Approach to Enzyme Catalysis. Journal of Biological Chemistry, 2010. 285 (25): p. 18961-18966.

25. Wiita, A.P., R. Perez-Jimenez, K.A. Walther, F. Grater, B.J. Berne, A. Holmgren, J.M. Sanchez-Ruiz, and J.M. Fernandez, Probing the chemistry of thioredoxin catalysis with force. Nature, 2007. 450 (7166): p. 124-7.

26. Wiita, A.P., S.R. Ainavarapu, H.H. Huang, and J.M. Fernandez, Force-dependent chemical kinetics of disulfide bond reduction observed with single-molecule techniques. Proc Natl Acad Sci U S A, 2006. 103 (19): p. 7222-7.

27. Neuman, K.C. and A. Nagy, Single-molecule force spectroscopy: optical tweezers, magnetic tweezers and atomic force microscopy. Nature methods, 2008. 5(6): p. 491-505.

28. Grandbois, M., M. Beyer, M. Rief, H. Clausen-Schaumann, and H.E. Gaub, How strong is a covalent bond? Science, 1999. 283 (5408): p. 1727-30.

29. Sevier, C.S. and C.A. Kaiser, Formation and transfer of disulphide bonds in living cells. Nat Rev Mol Cell Biol, 2002. 3(11): p. 836-47.

30. Gilbert, H.F., Molecular and cellular aspects of thiol-disulfide exchange. Advances in enzymology and related areas of molecular biology, 1990. 63 : p. 69-172.

137

31. Aslund, F., K.D. Berndt, and A. Holmgren, Redox potentials of glutaredoxins and other thiol-disulfide oxidoreductases of the thioredoxin superfamily determined by direct protein-protein redox equilibria. The Journal of biological chemistry, 1997. 272 (49): p. 30780-6.

32. Sengupta, S., H. Chen, T. Togawa, P.M. DiBello, A.K. Majors, B. Budy, M.E. Ketterer, and D.W. Jacobsen, Albumin thiolate anion is an intermediate in the formation of albumin-S-S-homocysteine. Journal of Biological Chemistry, 2001. 276 (32): p. 30111- 30117.

33. Tu, B.P. and J.S. Weissman, Oxidative protein folding in eukaryotes: mechanisms and consequences. The Journal of cell biology, 2004. 164 (3): p. 341-6.

34. Hatahet, F., L.W. Ruddock, K. Ahn, A. Benham, D. Craik, L. Ellgaard, D. Ferrari, and S. Ventura, Protein disulfide isomerase: a critical evaluation of its function in disulfide bond formation. Antioxid Redox Signal, 2009. 11 (11): p. 2807-50.

35. Heras, B., S.R. Shouldice, M. Totsika, M.J. Scanlon, M.A. Schembri, and J.L. Martin, DSB proteins and bacterial pathogenicity. Nat Rev Microbiol, 2009. 7(3): p. 215-25.

36. Land, A. and I. Braakman, Folding of the human immunodeficiency virus type 1 envelope glycoprotein in the endoplasmic reticulum. Biochimie, 2001. 83 (8): p. 783-90.

37. Uehara, T., T. Nakamura, D. Yao, Z.Q. Shi, Z. Gu, Y. Ma, E. Masliah, Y. Nomura, and S.A. Lipton, S-nitrosylated protein-disulphide isomerase links protein misfolding to neurodegeneration. Nature, 2006. 441 (7092): p. 513-7.

38. Culotta, V.C., M. Yang, and T.V. O'Halloran, Activation of superoxide dismutases: putting the metal to the pedal. Biochim Biophys Acta, 2006. 1763 (7): p. 747-58.

39. Wilkinson, B. and H.F. Gilbert, Protein disulfide isomerase. Biochim Biophys Acta, 2004. 1699 (1-2): p. 35-44.

40. Wickner, W. and R. Schekman, Protein translocation across biological membranes. Science, 2005. 310 (5753): p. 1452-6.

41. Bechtluft, P., R.G. van Leeuwen, M. Tyreman, D. Tomkiewicz, N. Nouwen, H.L. Tepper, A.J. Driessen, and S.J. Tans, Direct observation of chaperone-induced changes in a protein folding pathway. Science, 2007. 318 (5855): p. 1458-61.

138

42. Robson, A. and I. Collinson, The structure of the Sec complex and the problem of protein translocation. EMBO Rep, 2006. 7(11): p. 1099-103.

43. Kadokura, H., H. Tian, T. Zander, J.C. Bardwell, and J. Beckwith, Snapshots of DsbA in action: detection of proteins in the process of oxidative folding. Science, 2004. 303 (5657): p. 534-7.

44. Di Jeso, B., Y.N. Park, L. Ulianich, A.S. Treglia, M.L. Urbanas, S. High, and P. Arvan, Mixed-disulfide folding intermediates between thyroglobulin and endoplasmic reticulum resident oxidoreductases ERp57 and protein disulfide isomerase. Mol Cell Biol, 2005. 25 (22): p. 9793-805.

45. Kadokura, H. and J. Beckwith, Detecting folding intermediates of a protein as it passes through the bacterial translocation channel. Cell, 2009. 138 (6): p. 1164-73.

46. Martin, J.L., Thioredoxin--a fold for all reasons. Structure, 1995. 3(3): p. 245-50.

47. Berndt, C., C.H. Lillig, and A. Holmgren, Thioredoxins and glutaredoxins as facilitators of protein folding. Biochim Biophys Acta, 2008. 1783 (4): p. 641-50.

48. Chivers, P.T., K.E. Prehoda, and R.T. Raines, The CXXC motif: a rheostat in the active site. Biochemistry, 1997. 36 (14): p. 4061-6.

49. Holmgren, A., Thioredoxin and glutaredoxin systems. The Journal of biological chemistry, 1989. 264 (24): p. 13963-6.

50. Laurent, T.C., E.C. Moore, and P. Reichard, Enzymatic Synthesis of Deoxyribonucleotides. Iv. Isolation and Characterization of Thioredoxin, the Hydrogen Donor from Escherichia Coli B. The Journal of biological chemistry, 1964. 239 : p. 3436- 44.

51. Nordlund, P. and P. Reichard, Ribonucleotide reductases. Annual review of biochemistry, 2006. 75 : p. 681-706.

52. Holmgren, A., Reduction of disulfides by thioredoxin. Exceptional reactivity of insulin and suggested functions of thioredoxin in mechanism of hormone action. The Journal of biological chemistry, 1979. 254 (18): p. 9113-9.

139

53. Holmgren, A., Thioredoxin. Annu Rev Biochem, 1985. 54 : p. 237-71.

54. Kallis, G.B. and A. Holmgren, Differential reactivity of the functional sulfhydryl groups of cysteine-32 and cysteine-35 present in the reduced form of thioredoxin from Escherichia coli. J Biol Chem, 1980. 255 (21): p. 10261-5.

55. Danehy, J.P. and C.J. Noel, The Relative Nucleophilic Character of Several Mercaptans toward Ethylene Oxide. Journal of the American Chemical Society, 1960. 82 (10): p. 2511-2515.

56. Holmgren, A., Thioredoxin catalyzes the reduction of insulin disulfides by dithiothreitol and dihydrolipoamide. The Journal of biological chemistry, 1979. 254 (19): p. 9627-32.

57. Holmgren, A., B.O. Soderberg, H. Eklund, and C.I. Branden, Three-dimensional structure of Escherichia coli thioredoxin-S2 to 2.8 A resolution. Proceedings of the National Academy of Sciences of the United States of America, 1975. 72 (6): p. 2305-9.

58. Eklund, H., C. Cambillau, B.M. Sjoberg, A. Holmgren, H. Jornvall, J.O. Hoog, and C.I. Branden, Conformational and functional similarities between glutaredoxin and thioredoxins. The EMBO journal, 1984. 3(7): p. 1443-9.

59. Qin, J., G.M. Clore, W.M. Kennedy, J.R. Huth, and A.M. Gronenborn, Solution structure of human thioredoxin in a mixed disulfide intermediate complex with its target peptide from the transcription factor NF kappa B. Structure, 1995. 3(3): p. 289-97.

60. Chivers, P.T. and R.T. Raines, General acid/base catalysis in the active site of Escherichia coli thioredoxin. Biochemistry, 1997. 36 (50): p. 15810-6.

61. Qin, J., G.M. Clore, W.P. Kennedy, J. Kuszewski, and A.M. Gronenborn, The solution structure of human thioredoxin complexed with its target from Ref-1 reveals peptide chain reversal. Structure, 1996. 4(5): p. 613-20.

62. Maeda, K., P. Hagglund, C. Finnie, B. Svensson, and A. Henriksen, Structural basis for target protein recognition by the protein disulfide reductase thioredoxin. Structure, 2006. 14 (11): p. 1701-10.

63. Krause, G. and A. Holmgren, Substitution of the conserved tryptophan 31 in Escherichia coli thioredoxin by site-directed mutagenesis and structure-function analysis. The Journal of biological chemistry, 1991. 266 (7): p. 4056-66.

140

64. Lundstrom, J., G. Krause, and A. Holmgren, A Pro to His mutation in active site of thioredoxin increases its disulfide-isomerase activity 10-fold. New refolding systems for reduced or randomly oxidized ribonuclease. J Biol Chem, 1992. 267 (13): p. 9047-52.

65. Goldberger, R.F., C.J. Epstein, and C.B. Anfinsen, Acceleration of reactivation of reduced bovine pancreatic ribonuclease by a microsomal system from rat liver. J Biol Chem, 1963. 238 : p. 628-35.

66. Goldberger, R.F., C.J. Epstein, and C.B. Anfinsen, Purification and Properties of a Microsomal Enzyme System Catalyzing the Reactivation of Reduced Ribonuclease and Lysozyme. J Biol Chem, 1964. 239 : p. 1406-10.

67. Edman, J.C., L. Ellis, R.W. Blacher, R.A. Roth, and W.J. Rutter, Sequence of protein disulphide isomerase and implications of its relationship to thioredoxin. Nature, 1985. 317 (6034): p. 267-70.

68. Kemmink, J., N.J. Darby, K. Dijkstra, M. Nilges, and T.E. Creighton, The folding catalyst protein disulfide isomerase is constructed of active and inactive thioredoxin modules. Current biology : CB, 1997. 7(4): p. 239-45.

69. Kemmink, J., N.J. Darby, K. Dijkstra, R.M. Scheek, and T.E. Creighton, Nuclear magnetic resonance characterization of the N-terminal thioredoxin-like domain of protein disulfide isomerase. Protein science : a publication of the Protein Society, 1995. 4(12): p. 2587-93.

70. Kemmink, J., N.J. Darby, K. Dijkstra, M. Nilges, and T.E. Creighton, Structure determination of the N-terminal thioredoxin-like domain of protein disulfide isomerase using multidimensional heteronuclear 13C/15N NMR spectroscopy. Biochemistry, 1996. 35 (24): p. 7684-91.

71. Nguyen, V.D., K. Wallis, M.J. Howard, A.M. Haapalainen, K.E. Salo, M.J. Saaranen, A. Sidhu, R.K. Wierenga, R.B. Freedman, L.W. Ruddock, and R.A. Williamson, Alternative conformations of the x region of human protein disulphide-isomerase modulate exposure of the substrate binding b' domain. Journal of molecular biology, 2008. 383 (5): p. 1144- 55.

72. Denisov, A.Y., P. Maattanen, C. Dabrowski, G. Kozlov, D.Y. Thomas, and K. Gehring, Solution structure of the bb' domains of human protein disulfide isomerase. The FEBS journal, 2009. 276 (5): p. 1440-9.

141

73. Tian, G., S. Xiang, R. Noiva, W.J. Lennarz, and H. Schindelin, The crystal structure of yeast protein disulfide isomerase suggests cooperativity between its active sites. Cell, 2006. 124 (1): p. 61-73.

74. Creighton, T.E., D.A. Hillson, and R.B. Freedman, Catalysis by protein-disulphide isomerase of the unfolding and refolding of proteins with disulphide bonds. J Mol Biol, 1980. 142 (1): p. 43-62.

75. Lundstrom, J. and A. Holmgren, Protein disulfide-isomerase is a substrate for thioredoxin reductase and has thioredoxin-like activity. J Biol Chem, 1990. 265 (16): p. 9114-20.

76. Darby, N.J. and T.E. Creighton, Characterization of the active site cysteine residues of the thioredoxin-like domains of protein disulfide isomerase. Biochemistry, 1995. 34 (51): p. 16770-80.

77. Darby, N.J. and T.E. Creighton, Functional properties of the individual thioredoxin-like domains of protein disulfide isomerase. Biochemistry, 1995. 34 (37): p. 11725-35.

78. Xiao, R., B. Wilkinson, A. Solovyov, J.R. Winther, A. Holmgren, J. Lundstrom-Ljung, and H.F. Gilbert, The contributions of protein disulfide isomerase and its homologues to oxidative protein folding in the yeast endoplasmic reticulum. The Journal of biological chemistry, 2004. 279 (48): p. 49780-6.

79. Xiao, R., A. Solovyov, H.F. Gilbert, A. Holmgren, and J. Lundstrom-Ljung, Combinations of protein-disulfide isomerase domains show that there is little correlation between isomerase activity and wild-type growth. The Journal of biological chemistry, 2001. 276 (30): p. 27975-80.

80. Walker, K.W. and H.F. Gilbert, Scanning and escape during protein-disulfide isomerase- assisted protein folding. J Biol Chem, 1997. 272 (14): p. 8845-8.

81. Gething, M.J. and J. Sambrook, Protein folding in the cell. Nature, 1992. 355 (6355): p. 33-45.

82. Matlack, K.E., W. Mothes, and T.A. Rapoport, Protein translocation: tunnel vision. Cell, 1998. 92 (3): p. 381-90.

142

83. Rapoport, T.A., B. Jungnickel, and U. Kutay, Protein transport across the eukaryotic endoplasmic reticulum and bacterial inner membranes. Annual review of biochemistry, 1996. 65 : p. 271-303.

84. Walter, P., R. Gilmore, and G. Blobel, Protein translocation across the endoplasmic reticulum. Cell, 1984. 38 (1): p. 5-8.

85. Van den Berg, B., W.M. Clemons, Jr., I. Collinson, Y. Modis, E. Hartmann, S.C. Harrison, and T.A. Rapoport, X-ray structure of a protein-conducting channel. Nature, 2004. 427 (6969): p. 36-44.

86. Bulleid, N.J. and R.B. Freedman, Defective co-translational formation of disulphide bonds in protein disulphide-isomerase-deficient microsomes. Nature, 1988. 335 (6191): p. 649-51.

87. Molinari, M. and A. Helenius, Glycoproteins form mixed disulphides with oxidoreductases during folding in living cells. Nature, 1999. 402 (6757): p. 90-3.

88. Frand, A.R. and C.A. Kaiser, Ero1p oxidizes protein disulfide isomerase in a pathway for disulfide bond formation in the endoplasmic reticulum. Molecular Cell, 1999. 4(4): p. 469-77.

89. Gilbert, H.F., Thiol/disulfide exchange equilibria and disulfide bond stability. Methods in enzymology, 1995. 251 : p. 8-28.

90. Lyles, M.M. and H.F. Gilbert, Catalysis of the oxidative folding of ribonuclease A by protein disulfide isomerase: dependence of the rate on the composition of the redox buffer. Biochemistry, 1991. 30 (3): p. 613-9.

91. Lyles, M.M. and H.F. Gilbert, Catalysis of the oxidative folding of ribonuclease A by protein disulfide isomerase: pre-steady-state kinetics and the utilization of the oxidizing equivalents of the isomerase. Biochemistry, 1991. 30 (3): p. 619-25.

92. Sevier, C.S. and C.A. Kaiser, Ero1 and redox homeostasis in the endoplasmic reticulum. Biochim Biophys Acta, 2008. 1783 (4): p. 549-56.

93. Tu, B.P., S.C. Ho-Schleyer, K.J. Travers, and J.S. Weissman, Biochemical basis of oxidative protein folding in the endoplasmic reticulum. Science, 2000. 290 (5496): p. 1571-4.

143

94. Tu, B.P. and J.S. Weissman, The FAD- and O(2)-dependent reaction cycle of Ero1- mediated oxidative protein folding in the endoplasmic reticulum. Molecular Cell, 2002. 10 (5): p. 983-94.

95. Holmgren, A., Hydrogen donor system for Escherichia coli ribonucleoside-diphosphate reductase dependent upon glutathione. Proceedings of the National Academy of Sciences of the United States of America, 1976. 73 (7): p. 2275-9.

96. Holmgren, A., Glutathione-dependent synthesis of deoxyribonucleotides. Purification and characterization of glutaredoxin from Escherichia coli. The Journal of biological chemistry, 1979. 254 (9): p. 3664-71.

97. Holmgren, A., Glutathione-dependent synthesis of deoxyribonucleotides. Characterization of the enzymatic mechanism of Escherichia coli glutaredoxin. The Journal of biological chemistry, 1979. 254 (9): p. 3672-8.

98. Prinz, W.A., F. Aslund, A. Holmgren, and J. Beckwith, The role of the thioredoxin and glutaredoxin pathways in reducing protein disulfide bonds in the Escherichia coli cytoplasm. The Journal of biological chemistry, 1997. 272 (25): p. 15661-7.

99. Lillig, C.H., C. Berndt, and A. Holmgren, Glutaredoxin systems. Biochimica et biophysica acta, 2008. 1780 (11): p. 1304-17.

100. Gravina, S.A. and J.J. Mieyal, Thioltransferase is a specific glutathionyl mixed disulfide oxidoreductase. Biochemistry, 1993. 32 (13): p. 3368-76.

101. Sodano, P., T.H. Xia, J.H. Bushweller, O. Bjornberg, A. Holmgren, M. Billeter, and K. Wuthrich, Sequence-specific 1H n.m.r. assignments and determination of the three- dimensional structure of reduced Escherichia coli glutaredoxin. Journal of molecular biology, 1991. 221 (4): p. 1311-24.

102. Xia, T.H., J.H. Bushweller, P. Sodano, M. Billeter, O. Bjornberg, A. Holmgren, and K. Wuthrich, NMR structure of oxidized Escherichia coli glutaredoxin: comparison with reduced E. coli glutaredoxin and functionally related proteins. Protein science : a publication of the Protein Society, 1992. 1(3): p. 310-21.

103. Sun, C., M.J. Berardi, and J.H. Bushweller, The NMR solution structure of human glutaredoxin in the fully reduced form. Journal of molecular biology, 1998. 280 (4): p. 687-701.

144

104. Yang, Y., S. Jao, S. Nanduri, D.W. Starke, J.J. Mieyal, and J. Qin, Reactivity of the human thioltransferase (glutaredoxin) C7S, C25S, C78S, C82S mutant and NMR solution structure of its glutathionyl mixed disulfide intermediate reflect catalytic specificity. Biochemistry, 1998. 37 (49): p. 17145-56.

105. Gallogly, M.M. and J.J. Mieyal, Mechanisms of reversible protein glutathionylation in redox signaling and oxidative stress. Current opinion in pharmacology, 2007. 7(4): p. 381-91.

106. Shelton, M.D., P.B. Chock, and J.J. Mieyal, Glutaredoxin: role in reversible protein s- glutathionylation and regulation of redox signal transduction and protein translocation. Antioxidants & Redox Signaling, 2005. 7(3-4): p. 348-66.

107. Chakravarthi, S., C.E. Jessop, and N.J. Bulleid, The role of glutathione in disulphide bond formation and endoplasmic-reticulum-generated oxidative stress. EMBO Rep, 2006. 7(3): p. 271-5.

108. Sies, H., Glutathione and its role in cellular functions. Free radical biology & medicine, 1999. 27 (9-10): p. 916-21.

109. Meister, A. and M.E. Anderson, Glutathione. Annual review of biochemistry, 1983. 52 : p. 711-60.

110. Hwang, C., A.J. Sinskey, and H.F. Lodish, Oxidized redox state of glutathione in the endoplasmic reticulum. Science, 1992. 257 (5076): p. 1496-502.

111. Mamathambika, B.S. and J.C. Bardwell, Disulfide-linked protein folding pathways. Annu Rev Cell Dev Biol, 2008. 24 : p. 211-35.

112. Braakman, I. and N.J. Bulleid, Protein folding and modification in the Mammalian endoplasmic reticulum. Annu Rev Biochem, 2011. 80 : p. 71-99.

113. Adachi, T., R.M. Weisbrod, D.R. Pimentel, J. Ying, V.S. Sharov, C. Schoneich, and R.A. Cohen, S-Glutathiolation by peroxynitrite activates SERCA during arterial relaxation by nitric oxide. Nature medicine, 2004. 10 (11): p. 1200-7.

114. Chen, C.A., T.Y. Wang, S. Varadharaj, L.A. Reyes, C. Hemann, M.A. Talukder, Y.R. Chen, L.J. Druhan, and J.L. Zweier, S-glutathionylation uncouples eNOS and regulates its cellular and vascular function. Nature, 2010. 468 (7327): p. 1115-8.

145

115. Biswas, S., A.S. Chida, and I. Rahman, Redox modifications of protein-thiols: emerging roles in cell signaling. Biochemical pharmacology, 2006. 71 (5): p. 551-64.

116. Yang, Y., W. Shi, N. Cui, Z. Wu, and C. Jiang, Oxidative stress inhibits vascular K(ATP) channels by S-glutathionylation. The Journal of biological chemistry, 2010. 285 (49): p. 38641-8.

117. Narayan, M., E. Welker, W.J. Wedemeyer, and H.A. Scheraga, Oxidative folding of proteins. Accounts of chemical research, 2000. 33 (11): p. 805-12.

118. Ruddock, L.W., T.R. Hirst, and R.B. Freedman, pH-dependence of the dithiol-oxidizing activity of DsbA (a periplasmic protein thiol:disulphide oxidoreductase) and protein disulphide-isomerase: studies with a novel simple peptide substrate. Biochem J, 1996. 315 ( Pt 3) : p. 1001-5.

119. Christiansen, C., P.M. St Hilaire, and J.R. Winther, Fluorometric polyethyleneglycol- peptide hybrid substrates for quantitative assay of protein disulfide isomerase. Anal Biochem, 2004. 333 (1): p. 148-55.

120. Kersteen, E.A., S.R. Barrows, and R.T. Raines, Catalysis of protein disulfide bond isomerization in a homogeneous substrate. Biochemistry, 2005. 44 (36): p. 12168-78.

121. Darby, N.J., R.B. Freedman, and T.E. Creighton, Dissecting the mechanism of protein disulfide isomerase: catalysis of disulfide bond formation in a model peptide. Biochemistry, 1994. 33 (25): p. 7937-47.

122. Rothwarf, D.M., Y.J. Li, and H.A. Scheraga, Regeneration of bovine pancreatic ribonuclease A: detailed kinetic analysis of two independent folding pathways. Biochemistry, 1998. 37 (11): p. 3767-76.

123. Rothwarf, D.M., Y.J. Li, and H.A. Scheraga, Regeneration of bovine pancreatic ribonuclease A: identification of two nativelike three-disulfide intermediates involved in separate pathways. Biochemistry, 1998. 37 (11): p. 3760-6.

124. Wedemeyer, W.J., E. Welker, M. Narayan, and H.A. Scheraga, Disulfide bonds and protein folding. Biochemistry, 2000. 39 (15): p. 4207-16.

146

125. van den Berg, B., E.W. Chung, C.V. Robinson, P.L. Mateo, and C.M. Dobson, The oxidative refolding of hen lysozyme and its catalysis by protein disulfide isomerase. EMBO J, 1999. 18 (17): p. 4794-803.

126. Creighton, T.E., Protein folding. Biochem J, 1990. 270 (1): p. 1-16.

127. Weissman, J.S. and P.S. Kim, Reexamination of the folding of BPTI: predominance of native intermediates. Science, 1991. 253 (5026): p. 1386-93.

128. Oas, T.G. and P.S. Kim, A peptide model of a protein folding intermediate. Nature, 1988. 336 (6194): p. 42-8.

129. Shin, H.C. and H.A. Scheraga, Effect of protein disulfide isomerase on the regeneration of bovine ribonuclease A with dithiothreitol. FEBS Lett, 1999. 456 (1): p. 143-5.

130. Shin, H.C. and H.A. Scheraga, Catalysis of the oxidative folding of bovine pancreatic ribonuclease A by protein disulfide isomerase. Journal of molecular biology, 2000. 300 (4): p. 995-1003.

131. Shin, H.C., M.C. Song, and H.A. Scheraga, Effect of protein disulfide isomerase on the rate-determining steps of the folding of bovine pancreatic ribonuclease A. FEBS letters, 2002. 521 (1-3): p. 77-80.

132. Weissman, J.S. and P.S. Kim, Efficient catalysis of disulphide bond rearrangements by protein disulphide isomerase. Nature, 1993. 365 (6442): p. 185-8.

133. Hillson, D.A., N. Lambert, and R.B. Freedman, Formation and isomerization of disulfide bonds in proteins: protein disulfide-isomerase. Methods Enzymol, 1984. 107 : p. 281-94.

134. Gilbert, H.F., Protein disulfide isomerase and assisted protein folding. The Journal of biological chemistry, 1997. 272 (47): p. 29399-402.

135. Binnig, G., C.F. Quate, and C. Gerber, Atomic Force Microscope. Physical Review Letters, 1986. 56 (9): p. 930-933.

136. Gross, L., F. Mohn, N. Moll, P. Liljeroth, and G. Meyer, The chemical structure of a molecule resolved by atomic force microscopy. Science, 2009. 325 (5944): p. 1110-4.

147

137. Butt, H.J., B. Cappella, and M. Kappl, Force measurements with the atomic force microscope: Technique, interpretation and applications. Surface Science Reports, 2005. 59 (1-6): p. 1-152.

138. Hutter, J.L. and J. Bechhoefer, Calibration of Atomic-Force Microscope Tips. Review of Scientific Instruments, 1993. 64 (7): p. 1868-1873.

139. Carrion-Vazquez, M., A.F. Oberhauser, S.B. Fowler, P.E. Marszalek, S.E. Broedel, J. Clarke, and J.M. Fernandez, Mechanical and chemical unfolding of a single protein: a comparison. Proc Natl Acad Sci U S A, 1999. 96 (7): p. 3694-9.

140. Improta, S., A.S. Politou, and A. Pastore, Immunoglobulin-like modules from titin I-band: extensible components of muscle elasticity. Structure, 1996. 4(3): p. 323-37.

141. Berkovich, R., S. Garcia-Manyes, J. Klafter, M. Urbakh, and J.M. Fernandez, Hopping around an entropic barrier created by force. Biochemical and Biophysical Research Communications, 2010. 403 (1): p. 133-7.

142. Berkovich, R., S. Garcia-Manyes, M. Urbakh, J. Klafter, and J.M. Fernandez, Collapse dynamics of single proteins extended by force. Biophysical journal, 2010. 98 (11): p. 2692-701.

143. Garcia-Manyes, S., J. Brujic, C.L. Badilla, and J.M. Fernandez, Force-clamp spectroscopy of single-protein monomers reveals the individual unfolding and folding pathways of I27 and ubiquitin. Biophys J, 2007. 93 (7): p. 2436-46.

144. Schlierf, M., H. Li, and J.M. Fernandez, The unfolding kinetics of ubiquitin captured with single-molecule force-clamp techniques. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101 (19): p. 7299-304.

145. Oberhauser, A.F., P.K. Hansma, M. Carrion-Vazquez, and J.M. Fernandez, Stepwise unfolding of titin under force-clamp atomic force microscopy. Proceedings of the National Academy of Sciences of the United States of America, 2001. 98 (2): p. 468-72.

146. Rief, M., M. Gautel, F. Oesterhelt, J.M. Fernandez, and H.E. Gaub, Reversible unfolding of individual titin immunoglobulin domains by AFM. Science, 1997. 276 (5315): p. 1109- 12.

148

147. Bustamante, C., J.F. Marko, E.D. Siggia, and S. Smith, Entropic elasticity of lambda- phage DNA. Science, 1994. 265 (5178): p. 1599-600.

148. Ainavarapu, S.R., J. Brujic, H.H. Huang, A.P. Wiita, H. Lu, L. Li, K.A. Walther, M. Carrion-Vazquez, H. Li, and J.M. Fernandez, Contour length and refolding rate of a small protein controlled by engineered disulfide bonds. Biophys J, 2007. 92 (1): p. 225- 33.

149. Kohn, J.E., I.S. Millett, J. Jacob, B. Zagrovic, T.M. Dillon, N. Cingel, R.S. Dothager, S. Seifert, P. Thiyagarajan, T.R. Sosnick, M.Z. Hasan, V.S. Pande, I. Ruczinski, S. Doniach, and K.W. Plaxco, Random-coil behavior and the dimensions of chemically unfolded proteins. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101 (34): p. 12491-6.

150. Bell, G.I., Models for the specific adhesion of cells to cells. Science, 1978. 200 (4342): p. 618-27.

151. Lu, H., B. Isralewitz, A. Krammer, V. Vogel, and K. Schulten, Unfolding of titin immunoglobulin domains by steered molecular dynamics simulation. Biophysical journal, 1998. 75 (2): p. 662-71.

152. Li, H. and J.M. Fernandez, Mechanical design of the first proximal Ig domain of human cardiac titin revealed by single molecule force spectroscopy. Journal of molecular biology, 2003. 334 (1): p. 75-86.

153. Ainavarapu, S.R., A.P. Wiita, H.H. Huang, and J.M. Fernandez, A single-molecule assay to directly identify solvent-accessible disulfide bonds and probe their effect on protein folding. J Am Chem Soc, 2008. 130 (2): p. 436-7.

154. Carl, P., C.H. Kwok, G. Manderson, D.W. Speicher, and D.E. Discher, Forced unfolding modulated by disulfide bonds in the Ig domains of a cell adhesion molecule. Proceedings of the National Academy of Sciences of the United States of America, 2001. 98 (4): p. 1565-70.

155. Bhasin, N., P. Carl, S. Harper, G. Feng, H. Lu, D.W. Speicher, and D.E. Discher, Chemistry on a single protein, vascular cell adhesion molecule-1, during forced unfolding. The Journal of biological chemistry, 2004. 279 (44): p. 45865-74.

149

156. Bustanji, Y. and B. Samori, The mechanical properties of human angiostatin can be modulated by means of its disulfide bonds: a single-molecule force-spectroscopy study. Angewandte Chemie, 2002. 41 (9): p. 1546-8.

157. Grandi, F., M. Sandal, G. Guarguaglini, E. Capriotti, R. Casadio, and B. Samori, Hierarchical mechanochemical switches in angiostatin. Chembiochem : a European journal of chemical biology, 2006. 7(11): p. 1774-82.

158. Hoffstrom, B.G., A. Kaplan, R. Letso, R.S. Schmid, G.J. Turmel, D.C. Lo, and B.R. Stockwell, Inhibitors of protein disulfide isomerase suppress apoptosis induced by misfolded proteins. Nat Chem Biol, 2010. 6(12): p. 900-6.

159. Walker, A.K., M.A. Farg, C.R. Bye, C.A. McLean, M.K. Horne, and J.D. Atkin, Protein disulphide isomerase protects against protein aggregation and is S-nitrosylated in amyotrophic lateral sclerosis. Brain, 2010. 133 (Pt 1): p. 105-16.

160. Dobson, C.M., Protein folding and misfolding. Nature, 2003. 426 (6968): p. 884-90.

161. Schroder, M. and R.J. Kaufman, The mammalian unfolded protein response. Annu Rev Biochem, 2005. 74 : p. 739-89.

162. Ron, D. and P. Walter, Signal integration in the endoplasmic reticulum unfolded protein response. Nature reviews. Molecular cell biology, 2007. 8(7): p. 519-29.

163. Walker, K.W., M.M. Lyles, and H.F. Gilbert, Catalysis of oxidative protein folding by mutants of protein disulfide isomerase with a single active-site cysteine. Biochemistry, 1996. 35 (6): p. 1972-80.

164. Paxman, J.J., N.A. Borg, J. Horne, P.E. Thompson, Y. Chin, P. Sharma, J.S. Simpson, J. Wielens, S. Piek, C.M. Kahler, H. Sakellaris, M. Pearce, S.P. Bottomley, J. Rossjohn, and M.J. Scanlon, The structure of the bacterial oxidoreductase enzyme DsbA in complex with a peptide reveals a basis for substrate specificity in the catalytic cycle of DsbA enzymes. The Journal of biological chemistry, 2009. 284 (26): p. 17835-45.

165. Dong, G., P.A. Wearsch, D.R. Peaper, P. Cresswell, and K.M. Reinisch, Insights into MHC class I peptide loading from the structure of the tapasin-ERp57 thiol oxidoreductase heterodimer. Immunity, 2009. 30 (1): p. 21-32.

150

166. Fisher, T.E., P.E. Marszalek, A.F. Oberhauser, M. Carrion-Vazquez, and J.M. Fernandez, The micro-mechanics of single molecules studied with atomic force microscopy. The Journal of physiology, 1999. 520 Pt 1 : p. 5-14.

167. Carvalho, A.P., P.A. Fernandes, and M.J. Ramos, Similarities and differences in the thioredoxin superfamily. Progress in Biophysics & Molecular Biology, 2006. 91 (3): p. 229-248.

168. Debarbieux, L. and J. Beckwith, On the functional interchangeability, oxidant versus reductant, of members of the thioredoxin superfamily. J Bacteriol, 2000. 182 (3): p. 723-7.

169. Stewart, E.J., F. Aslund, and J. Beckwith, Disulfide bond formation in the Escherichia coli cytoplasm: an in vivo role reversal for the thioredoxins. EMBO J, 1998. 17 (19): p. 5543-50.

170. Debarbieux, L. and J. Beckwith, The reductive enzyme thioredoxin 1 acts as an oxidant when it is exported to the Escherichia coli periplasm. Proc Natl Acad Sci U S A, 1998. 95 (18): p. 10751-6.

171. Lappi, A.K. and L.W. Ruddock, Reexamination of the role of interplay between glutathione and protein disulfide isomerase. J Mol Biol, 2011. 409 (2): p. 238-49.

172. Chivers, P.T., M.C. Laboissiere, and R.T. Raines, The CXXC motif: imperatives for the formation of native disulfide bonds in the cell. EMBO J, 1996. 15 (11): p. 2659-67.

173. Chakravarthi, S., C.E. Jessop, and N.J. Bulleid, Chapter 1.5 Oxidative Folding in the Endoplasmic Reticulum , in Oxidative Folding of Peptides and Proteins 2009, The Royal Society of Chemistry. p. 81-104.

174. Welker, E., W.J. Wedemeyer, M. Narayan, and H.A. Scheraga, Coupling of conformational folding and disulfide-bond reactions in oxidative folding of proteins. Biochemistry, 2001. 40 (31): p. 9059-64.

175. Camacho, C.J. and D. Thirumalai, Modeling the role of disulfide bonds in protein folding: entropic barriers and pathways. Proteins, 1995. 22 (1): p. 27-40.

176. Ptitsyn, O.B., Molten globule and protein folding. Advances in protein chemistry, 1995. 47 : p. 83-229.

151

177. Kuwajima, K., The molten globule state as a clue for understanding the folding and cooperativity of globular-protein structure. Proteins, 1989. 6(2): p. 87-103.

178. Baldwin, R.L., C. Frieden, and G.D. Rose, Dry molten globule intermediates and the mechanism of protein unfolding. Proteins, 2010. 78 (13): p. 2725-37.

179. Ingolia, N.T., L.F. Lareau, and J.S. Weissman, Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of Mammalian proteomes. Cell, 2011. 147 (4): p. 789-802.

180. Eichmann, C., S. Preissler, R. Riek, and E. Deuerling, Cotranslational structure acquisition of nascent polypeptides monitored by NMR spectroscopy. Proceedings of the National Academy of Sciences of the United States of America, 2010. 107 (20): p. 9111- 6.

181. Cabrita, L.D., S.T. Hsu, H. Launay, C.M. Dobson, and J. Christodoulou, Probing ribosome-nascent chain complexes produced in vivo by NMR spectroscopy. Proceedings of the National Academy of Sciences of the United States of America, 2009. 106 (52): p. 22239-44.

182. Hsu, S.T., P. Fucini, L.D. Cabrita, H. Launay, C.M. Dobson, and J. Christodoulou, Structure and dynamics of a ribosome-bound nascent chain by NMR spectroscopy. Proceedings of the National Academy of Sciences of the United States of America, 2007. 104 (42): p. 16516-21.

183. Zapun, A., T.E. Creighton, P.J. Rowling, and R.B. Freedman, Folding in vitro of bovine pancreatic trypsin inhibitor in the presence of proteins of the endoplasmic reticulum. Proteins, 1992. 14 (1): p. 10-5.

184. Walker, K.W. and H.F. Gilbert, Oxidation of kinetically trapped thiols by protein disulfide isomerase. Biochemistry, 1995. 34 (41): p. 13642-50.

185. Florin, E.L., M. Rief, H. Lehmann, M. Ludwig, C. Dornmair, V.T. Moy, and H.E. Gaub, Sensing Specific Molecular-Interactions with the Atomic-Force Microscope. Biosensors & Bioelectronics, 1995. 10 (9-10): p. 895-901.

186. Alegre-Cebollada, J., P. Kosuri, J.A. Rivas-Pardo, and J.M. Fernandez, Direct observation of disulfide isomerization in a single protein. Nature chemistry, 2011. 3(11): p. 882-7.

152

187. Finkel, T. and N.J. Holbrook, Oxidants, oxidative stress and the biology of ageing. Nature, 2000. 408 (6809): p. 239-47.

188. Finkel, T., From sulfenylation to sulfhydration: what a thiolate needs to tolerate. Sci Signal, 2012. 5(215): p. pe10.

189. Allen, D.G., G.D. Lamb, and H. Westerblad, Skeletal muscle fatigue: cellular mechanisms. Physiological reviews, 2008. 88 (1): p. 287-332.

190. Powers, S.K. and M.J. Jackson, Exercise-induced oxidative stress: cellular mechanisms and impact on muscle force production. Physiological reviews, 2008. 88 (4): p. 1243-76.

191. Supinski, G.S. and L.A. Callahan, Free radical-mediated skeletal muscle dysfunction in inflammatory conditions. Journal of applied physiology, 2007. 102 (5): p. 2056-63.

192. Valko, M., D. Leibfritz, J. Moncol, M.T. Cronin, M. Mazur, and J. Telser, Free radicals and antioxidants in normal physiological functions and human disease. The international journal of biochemistry & cell biology, 2007. 39 (1): p. 44-84.

193. Mollica, J.P., T.L. Dutka, T. Merry, C. Lamboley, G.K. McConell, M.J. McKenna, R.M. Murphy, and G.D. Lamb, S-glutathionylation of Troponin I (fast) increases contractile apparatusCa2+-sensitivity in fast-twitch muscle fibres of rats and humans. The Journal of physiology, 2012.

194. Schiaffino, S. and C. Reggiani, Molecular diversity of myofibrillar proteins: gene regulation and functional significance. Physiological reviews, 1996. 76 (2): p. 371-423.

195. Gregorio, C.C., H. Granzier, H. Sorimachi, and S. Labeit, Muscle assembly: a titanic achievement? Current opinion in cell biology, 1999. 11 (1): p. 18-25.

196. Granzier, H.L. and S. Labeit, The giant protein titin: a major player in myocardial mechanics, signaling, and disease. Circulation research, 2004. 94 (3): p. 284-95.

197. Minajeva, A., M. Kulke, J.M. Fernandez, and W.A. Linke, Unfolding of titin domains explains the viscoelastic behavior of skeletal myofibrils. Biophysical journal, 2001. 80 (3): p. 1442-51.

153

198. Nedrud, J., S. Labeit, M. Gotthardt, and H. Granzier, Mechanics on Myocardium Deficient in the N2B Region of Titin: The Cardiac-Unique Spring Element Improves Efficiency of the Cardiac Cycle. Biophysical Journal, 2011. 101 (6): p. 1385-1392.

199. Kruger, M. and W.A. Linke, The giant protein titin: a regulatory node that integrates myocyte signaling pathways. The Journal of biological chemistry, 2011. 286 (12): p. 9905-12.

200. Marino, S.M. and V.N. Gladyshev, Cysteine function governs its conservation and degeneration and restricts its utilization on protein surfaces. Journal of molecular biology, 2010. 404 (5): p. 902-16.

201. Johnson, C.P., H.Y. Tang, C. Carag, D.W. Speicher, and D.E. Discher, Forced unfolding of proteins within cells. Science, 2007. 317 (5838): p. 663-6.

202. Zhang, X., K. Halvorsen, C.Z. Zhang, W.P. Wong, and T.A. Springer, Mechanoenzymatic cleavage of the ultralarge vascular protein von Willebrand factor. Science, 2009. 324 (5932): p. 1330-4.

203. Prinz, W.A., F. Aslund, A. Holmgren, and J. Beckwith, The role of the thioredoxin and glutaredoxin pathways in reducing protein disulfide bonds in the Escherichia coli cytoplasm. J Biol Chem, 1997. 272 (25): p. 15661-7.

204. Dalle-Donne, I., R. Rossi, G. Colombo, D. Giustarini, and A. Milzani, Protein S- glutathionylation: a regulatory device from bacteria to humans. Trends in biochemical sciences, 2009. 34 (2): p. 85-96.

205. Gallogly, M.M., D.W. Starke, and J.J. Mieyal, Mechanistic and kinetic details of catalysis of thiol-disulfide exchange by glutaredoxins and potential mechanisms of regulation. Antioxidants & Redox Signaling, 2009. 11 (5): p. 1059-81.

206. Lu, H. and K. Schulten, The key event in force-induced unfolding of Titin's immunoglobulin domains. Biophysical journal, 2000. 79 (1): p. 51-65.

207. Stacklies, W., M.C. Vega, M. Wilmanns, and F. Grater, Mechanical network in titin immunoglobulin from force distribution analysis. PLoS computational biology, 2009. 5(3): p. e1000306.

154

208. Ragsdale, D.S., J.C. McPhee, T. Scheuer, and W.A. Catterall, Molecular determinants of state-dependent block of Na+ channels by local anesthetics. Science, 1994. 265 (5179): p. 1724-8.

209. Vogel, V. and M. Sheetz, Local force and geometry sensing regulate cell functions. Nature reviews. Molecular cell biology, 2006. 7(4): p. 265-75.

210. del Rio, A., R. Perez-Jimenez, R. Liu, P. Roca-Cusachs, J.M. Fernandez, and M.P. Sheetz, Stretching single talin rod molecules activates vinculin binding. Science, 2009. 323 (5914): p. 638-41.

211. Puchner, E.M., A. Alexandrovich, A.L. Kho, U. Hensen, L.V. Schafer, B. Brandmeier, F. Grater, H. Grubmuller, H.E. Gaub, and M. Gautel, Mechanoenzymatics of titin kinase. Proceedings of the National Academy of Sciences of the United States of America, 2008. 105 (36): p. 13385-90.

212. Honig, B. and A. Nicholls, Classical electrostatics in biology and chemistry. Science, 1995. 268 (5214): p. 1144-9.

213. Chivers, P.T., K.E. Prehoda, B.F. Volkman, B.M. Kim, J.L. Markley, and R.T. Raines, Microscopic pKa values of Escherichia coli thioredoxin. Biochemistry, 1997. 36 (48): p. 14985-91.

214. Reutimann, H., B. Straub, P.L. Luisi, and A. Holmgren, A conformational study of thioredoxin and its tryptic fragments. J Biol Chem, 1981. 256 (13): p. 6796-803.

215. Takahashi, N. and T.E. Creighton, On the reactivity and ionization of the active site cysteine residues of Escherichia coli thioredoxin. Biochemistry, 1996. 35 (25): p. 8342- 53.

216. Li, H., C. Hanson, J.A. Fuchs, C. Woodward, and G.J. Thomas, Jr., Determination of the pKa values of active-center cysteines, cysteines-32 and -35, in Escherichia coli thioredoxin by Raman spectroscopy. Biochemistry, 1993. 32 (22): p. 5800-8.

217. Jeng, M.F., A. Holmgren, and H.J. Dyson, Proton sharing between cysteine thiols in Escherichia coli thioredoxin: implications for the mechanism of protein disulfide reduction. Biochemistry, 1995. 34 (32): p. 10101-5.

155

218. Wilson, N.A., E. Barbar, J.A. Fuchs, and C. Woodward, Aspartic acid 26 in reduced Escherichia coli thioredoxin has a pKa > 9. Biochemistry, 1995. 34 (28): p. 8931-9.

219. Nelson, K.J., A.E. Day, B.B. Zeng, S.B. King, and L.B. Poole, Isotope-coded, iodoacetamide-based reagent to determine individual cysteine pK(a) values by matrix- assisted laser desorption/ionization time-of-flight mass spectrometry. Anal Biochem, 2008. 375 (2): p. 187-95.

220. Forman-Kay, J.D., G.M. Clore, and A.M. Gronenborn, Relationship between electrostatics and redox function in human thioredoxin: characterization of pH titration shifts using two-dimensional homo- and heteronuclear NMR. Biochemistry, 1992. 31 (13): p. 3442-52.

221. Perez-Jimenez, R., J. Li, P. Kosuri, I. Sanchez-Romero, A.P. Wiita, D. Rodriguez-Larrea, A. Chueca, A. Holmgren, A. Miranda-Vizuete, K. Becker, S.H. Cho, J. Beckwith, E. Gelhaye, J.P. Jacquot, E.A. Gaucher, J.M. Sanchez-Ruiz, B.J. Berne, and J.M. Fernandez, Diversity of chemical mechanisms in thioredoxin catalysis revealed by single-molecule force spectroscopy. Nat Struct Mol Biol, 2009. 16 (8): p. 890-6.

222. Rini, M., B.Z. Magnes, E. Pines, and E.T.J. Nibbering, Real-time observation of bimodal proton transfer in acid-base pairs in water. Science, 2003. 301 (5631): p. 349-352.

223. Nordberg, J. and E.S. Arner, Reactive oxygen species, antioxidants, and the mammalian thioredoxin system. Free Radic Biol Med, 2001. 31 (11): p. 1287-312.

224. Holmgren, A. and J. Lu, Thioredoxin and thioredoxin reductase: current research with special reference to human disease. Biochem Biophys Res Commun, 2010. 396 (1): p. 120-4.

156

Appendix – Published Articles

Direct observation of disulfide isomerization in a single protein Alegre-Cebollada J, Kosuri P, Rivas-Pardo JA, Fernandez JM Nature Chemistry 3:882-7 (2011) Contributions: I built a mathematical model to interpret the data, performed numerical simulations, made a figure and co-wrote the paper.

Protease power strokes force proteins to unfold Alegre-Cebollada J, Kosuri P, Fernandez JM Cell 145:339-340 (2011) Contributions: I Contributed ideas, designed a model, prepared figures and co-wrote the article.

Single-molecule paleoenzymology probes the chemistry of resurrected enzymes Perez-Jimenez R, Ingles-Prieto A, Zhao Z, Sanchez-Romero I, Alegre-Cebollada J, Kosuri P, Garcia-Manyes S, Kappock TJ, Tanokura M, Holmgren A, Sanchez-Ruiz JM, Gaucher EA, Fernandez JM Nature Structural & Molecular Biology, 18:592-596 (2011) Contributions: I participated in the design of the project, performed experiments and co-wrote the article.

Extended Kalman filter estimates of the contour length of an unfolding protein in single- molecule force spectroscopy experiments Fernandez VI, Kosuri P, Parot P, Fernandez JM Review of Scientific Instruments 80:113104 (2009) Contributions: I performed experiments, wrote an analysis software package and co-wrote the article.

Single-molecule force spectroscopy approach to enzymatic catalysis Alegre-Cebollada J, Perez-Jimenez R, Kosuri P, Fernandez JM, Journal of Biological Chemistry , 285:18961-6 (2009) Contributions: I co-wrote the article.

Diversity of chemical mechanisms in thioredoxin catalysis revealed by single-molecule force spectroscopy Perez-Jimenez R, Li J, Kosuri P, Berne BJ, Fernandez JM Nature Structural & Molecular Biology , 16:890-6 (2009) Contributions: I performed experiments and co-wrote the paper.

Force-clamp spectroscopy detects residue co-evolution in enzyme catalysis Perez-Jimenez R, Wiita AP, Rodriguez-Larrea D, Kosuri P, Gavira JA, Sanchez-Ruiz JM, Fernandez JM Journal of Biological Chemistry , 283:27121-29 (2008) Contributions: I performed experiments and co-wrote the paper.