Computational Prediction and Experimental Validation of Cytochrome C Oxidase Main-Chain Flexibility and Allosteric Regulation of the K-Pathway

COMPUTATIONAL PREDICTION AND EXPERIMENTAL VALIDATION OF CYTOCHROME C OXIDASE MAIN-CHAIN FLEXIBILITY AND ALLOSTERIC REGULATION OF THE K-PATHWAY

Leann Marie Buhrow

A DISSERTATION

Submitted to Michigan State University in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Cell and Molecular Biology

2012

ABSTRACT

COMPUTATIONAL PREDICTION AND EXPERIMENTAL VALIDATION OF CYTOCHROME C OXIDASE MAIN-CHAIN FLEXIBILITY AND ALLOSTERIC REGULATION OF THE K-PATHWAY

Leann Marie Buhrow

Comparison of crystallographic structures and deuterium accessibility of different redox states of cytochrome c oxidase (CcO) have suggested conformational changes of mechanistic significance. To predict the intrinsic flexibility and low energy motions in CcO, this work has analyzed available high-resolution crystallographic structures with ProFlex and elNémo computational methods. CcO is predicted to undergo rotational motions on the interior and exterior of the membrane, driven by transmembrane helical tilting and bending, coupled with rocking of the

β-sheet domain. Consequently, the proton K-pathway becomes sufficiently flexible for internal water molecules to alternately occupy upper and lower parts of the pathway.

At the entrance of the K-pathway, a conserved crystallographically-defined steroid binding site had been previously identified. Binding of diverse amphipathic molecules including detergents, fatty acids, steroids, and porphyrins affect the activity of the Rhodobacter sphaeroides

CcO variant E101A, as well as the wild type and bovine enzymes. Detergent inhibition is observed for the E101A variant but may be overcome in the presence of micromolar concentrations of steroids and porphyrin analogs. Computational modeling of lauryl maltoside, bilirubin, and protoporphyrin IX into the conserved membrane site shows energetically favorable binding modes for these ligands and suggests that a groove at the interface of subunits I and II, including the entrance to the K-pathway, mediates competitive ligand interactions involving two overlapping sites. The high affinity and specificity of a number of compounds for this region, and its conservation and impact on CcO activity, support its physiological significance.

Physiological ligands, specific for the steroid binding site, were identified by combining three computational approaches: ROCS comparison of ligand shape and electrostatics,

SimSite3D analysis of similarity to ligand binding sites in the Protein Data Bank, and SLIDE screening of small molecules by docking. Together, the results suggest several steroids, adenine and guanine nucleotides, NAD+, FAD, and phosphorylated isoprenes as top candidates for interacting at this site, along with bile acids and porphyrins. In vitro oxygen consumption assays support some of these predicted interactions. In the wild type R. sphaeroides CcO, ATP and GDP are mildly inhibitory while the steroidal deoxycholate and fusidic acid ligands are highly inhibitory.

Cytochrome c titration assays indicate nucleotides inhibit CcO activity in low cytochrome c conditions, similar to the observed ATP inhibition of mammalian CcO. These finding suggest that nucleotides regulate CcO on the conserved subunit I-III core, potentially at the steroid binding site.

Overall this work predicts CcO conformational changes required for catalysis, including the conformational change of the K-pathway, and describes the first report of allosteric regulation of bacterial CcO by nucleotides. These results have been used to understand allosteric regulation by restricting conformational changes, generate a two-site model for lipid and ligand- specific regulation, and propose CcO regulation by arresting the enzyme in a state which cannot produce oxygen radical byproducts.

ACKNOWLEDGEMENTS

I would like to thank:

• My co-advisors, Shelagh Ferguson-Miller and Leslie Kuhn, for their guidance,

support, and direction in both research and life.

• Past and present members of the Ferguson-Miller and Kuhn labs including Drs.

Carrie Hiser, Jian Liu, Denise Mills, Ling Qin, Jon Hosler, Shujuan Xu, Namjoon

Kim, Xi Zhang, and Jeff Van Voost and Fei Li, Matt Tonero, and Nan Liu. I would

especially like to thank Carrie Hiser for her technical support, brainstorming

sessions, and being a wonderful collaborator on our lipids project. I would also like

to thank Ling Qin and Jeff Van Voost for their inspirational drive. Without their

crystal structures and SimSite3D methods, this work would not be possible.

• My committee members: Drs. Benning, Hausinger, Hegg, and Yan for their

suggestions, critiques, and diverse research interests.

• The Cell and Molecular Biology and Quantitative Biology Graduate Programs at

MSU for allowing me to focus on protein structure and function and expand my

interests into computational structural biology. I would especially like to thank Dr.

Susan Conrad for her support and guidance. I would also like to thank Helen Geiger

and Becky Mansel for their friendship and endless technical assistance.

• My advisors at the University of Wisconsin-Parkside: Drs. MacWilliams, Lewis, and

Wood for their mentoring and encouragement to attend graduate school.

• My fellow students at MSU and UW-Parkside for their friendship throughout the

years. Without your laughter and shenanigans, my life would be a more serious and

boring experience.

• My family, especially John, Terri, and Emily Buhrow. I love you!

TABLE OF CONTENTS

LIST OF TABLES vii

LIST OF FIGURES viii

CHAPTER 1: Rhodobacter sphaeroides cytochrome c oxidase as a model system for understanding energy coupling and allosteric regulation in membrane proteins 1 References 27

CHAPTER 2: From static structure to living protein: computational analysis of cytochrome c oxidase main-chain flexibility 40 Introduction 41 Materials and Methods 44 Results and Discussion 52 Conclusions 77 References 78

CHAPTER 3: Structural predictions and functional consequences of porphyrin, steroid, and detergent ligands binding to the cytochrome c oxidase steroid binding site 84 Introduction 85 Materials and Methods 93 Results and Discussion 96 Conclusions 109 References 110

CHAPTER 4: Three-pronged computational approach to predict regulatory ligands of cytochrome c oxidase 115 Introduction 116 Materials and Methods 119 Results and Discussion 124 Conclusions 178 References 179

CHAPTER 5: Perspectives on CcO K-pathway conformational change and allosteric regulation 186 References 193

LIST OF TABLES

TABLE 2.1: CcO structures analyzed by ProFlex and elNémo 46

TABLE 2.2: Comparison of oxidized and reduced two-subunit RsCcO structures 47

TABLE 2.3: Diverse membrane protein folds analyzed by ProFlex and elNémo 60

TABLE 2.4: elNémo percentage of atoms significantly displaced and relative frequencies of normal modes 60

TABLE 2.5: CcO helical conformational changes in elNémo low-energy modes 64

TABLE 3.1: RsCcO and BtCcO steroid binding site characterization 98

TABLE 3.2: Effects of bile acids on E101A mutant and WT RsCcO 101

TABLE 4.1: ROCS predicted ligands 125

TABLE 4.2: SimSite3D predicted binding sites 134

TABLE 4.3: Percentage of protein binding site matches between the CcO steroid binding site and diverse Binding MOAD proteins 143

TABLE 4.4: SLIDE docked ligands and their protein interactions 145

vii

LIST OF FIGURES

FIGURE 1.1: The electron transport chain 8

FIGURE 1.2: Structure of bacterial and mammalian cytochrome c oxidases 10

FIGURE 1.3: Cytochrome c oxidase oxygen reduction mechanism 15

FIGURE 2.1: Architecture of the two-subunit structure essential for RsCcO activity 42

FIGURE 2.2: Dependence of CcO flexibility on thermal energy increase, analyzed by ProFlex hydrogen bond dilution profiles 45

FIGURE 2.3: Mean squared displacement values within RsCcO transmembrane helices 49

FIGURE 2.4: ProFlex flexibility comparison the of two and four subunit crystal structures of RsCcO 52

FIGURE 2.5: ProFlex prediction of main-chain flexibility and stability in RsCcO 54

FIGURE 2.6: Thermal denaturation of the RsCcO structure by ProFlex 55

FIGURE 2.7: elNémo simulation of low frequency motions in RsCcO 58

FIGURE 2.8: Relative flexibility of membrane proteins assessed using ProFlex 61

FIGURE 2.9: RsCcO, β2 adrenergic receptor, KcsA potassium channel, and VDAC residue displacement 61

FIGURE 2.10: Planes of relatively stationary residues in the CcO transmembrane helices 67

FIGURE 2.11: Comparison of RsCcO normal modes and crystallographic temperature factors 69

FIGURE 2.12: Water molecule movement in the K-pathway 71

FIGURE 2.13: Conformational gating in the RsCcO oxygen channels 74

viii FIGURE 3.1: Bile acid structures 86

FIGURE 3.2: Deoxycholate resolved near the K-pathway entrance of RsCcO 89

FIGURE 3.3: RsCcO and BtCcO steroid binding site with conserved residues and bound water molecules 98

FIGURE 3.4: The steady-state activities of the detergent-solubilized RsCcO E101A 99

FIGURE 3.5: Potential binding orientations of known lipidic ligands in the RsCcO steroid binding site 103

FIGURE 3.6: RsCcO K-pathway rigidification upon ligand binding 107

FIGURE 4.1: Structure of bacterial and mammalian cytochrome c oxidases 117

FIGURE 4.2: ROCS predicted 2D ligand structures 130

FIGURE 4.3: ROCS aligned crystallographic deoxycholate and predicted ligands 132

FIGURE 4.4: SimSite3D predicted analogous chemical points between the RsCcO steroid binding site and sites found in Binding MOAD 142

FIGURE 4.5: Similar binding modes were predicted by SimSite3D aligned binding site and SLIDE docked ligands 147

FIGURE 4.6: Ethanol-soluble ligand additives do not affect WT CcO in standard or low detergent conditions 151

FIGURE 4.7: Ethanol-soluble ligand additives do not affect the E101A mutant in standard or low detergent conditions 153

FIGURE 4.8: GDP and an ATP analog mildly inhibit while fusidic acid and deoxycholate strongly inhibit WT CcO in both standard and low detergents 155

FIGURE 4.9: Water-soluble ligands are able to stimulate the E101A mutant in standard detergent 158

FIGURE 4.10: GDP, FAD, and fusidic acid stimulate while an ATP analog inhibits the E101A mutant in low detergent 159

FIGURE 4.11: An ATP analog stimulates then inhibits WT CcO in both standard and low detergent 161

FIGURE 4.12: An ATP analog mildly stimulates the E101A mutant in both

ix standard and low detergent 162

FIGURE 4.13: GDP inhibits WT CcO in both standard and low detergent 163

FIGURE 4.14: GDP mildly stimulates the E101A mutant in both standard and low detergent 164

FIGURE 4.15: Fusidic acid inhibits WT CcO in both standard and low detergent 165

FIGURE 4.16: Fusidic acid stimulates then inhibits the E101A mutant in both standard and low detergent 166

FIGURE 4.17: FAD stimulates WT CcO in both standard and low detergent 167

FIGURE 4.18: FAD stimulates the E101A mutant in both standard and low detergent 168

FIGURE 4.19: The nucleotides ADP, ATP, and GDP mildly inhibit WT CcO in low substrate conditions 171

FIGURE 4.20: The steroidal inhibitors deoxycholate and fusidic acid mildly inhibit WT CcO in low substrate conditions 172

FIGURE 4.21: The steroidal inhibitors deoxycholate and fusidic acid mildly inhibit WT CcO in low substrate conditions 173

FIGURE 4.22: Deoxycholate inhibition of E101A supports noncompetitive inhibition with cytochrome c 174

CHAPTER 1: Rhodobacter sphaeroides cytochrome c oxidase as a model system for understanding energy coupling and allosteric regulation in membrane proteins

Discovery of cytochromes and cytochrome c oxidase

Cytochromes, meaning cellular pigments, were originally identified and characterized from diverse tissues by C. S. McMunn beginning in 1884 [1-2]. Cytochromes were observed, in tissues without blood contamination, based on the presence of their unique four banded reduced absorbance spectra [1]. However, it was originally believed that these spectra were a result of contaminating hemoglobin degradative products and not integral components of all analyzed cells [1]. It was not until the mid 1920s when David Keilin detected this spectrum in insect abdominal tissue and was able to eliminate the possibility of hemoglobin contamination that cytochromes were again studied. Keilin's work supported cytochromes as distinct from hemoglobin, widely distributed in cellular tissues, and composed of three distinct chromophores

[1-3]. These chromophores were differentiated and named based on the order of their absorbance peaks in the four banded spectra, where the cytochrome peak closest to the infrared end of the spectrum (800 nm) was referred to as cytochrome a, the intermediate absorbing cytochrome was referred to as cytochrome b, and the cytochrome peak closest to the ultraviolet end of the visible spectrum was referred to as cytochrome c [1-3]. It is the absorbance peak near

606 nm that gave cytochrome oxidase the designation of cytochrome t, called the α-band. This absorbance peak along with a nearer-to-ultraviolet absorbance peak at 444 nm, named the γ- or

Soret-band, is currently used to characterize the cytochrome c oxidase cofactor environment and redox state [4].

Otto Warburg expanded the functional knowledge of cytochromes by developing the

Warburg oxygen sensitive respirometer. This instrument allowed respiratory oxygen consumption rates to be quantified and respiratory inhibitors including cyanide, carbon

2 monoxide, and carbon dioxide to be identified [2-3]. In 1929-1938, work by Dixon, Keilin, and

Hartree functionally characterized a cytochrome a containing protein that could oxidize cytochrome c, consume molecular oxygen, and readily undergo changes in redox state [1-2].

These founding works described and named cytochrome c oxidase (CcO; E.C. 1.9.3.1 [1]).

Additional strides were made in the mid 1950s, when Maley, Lardy, and Lehninger directly connected CcO activity to oxidative phosphorylation [1-2], and the independent groups of

Okunki, Hatefi, Green, Griffiths, and Wharton developed purification protocols for the detergent solubilization of CcO [2]. The most significant modern advance has been the atomic structure determination of mammalian CcO, which elucidated the protein's active site, location of heme and copper redox metal centers, and the number of accessory subunits [5]. At the same time, the structure of a bacterial cytochrome c oxidase was solved [6], which showed an essentially identical active site structure, validating both structures and opening up the field to investigation of structure/function relationships by the powerful tools of molecular biology.

The study of Rhodobacter sphaeroides cytochrome c oxidase as a model system for integral membrane protein mechanism, structure, and assembly

Heme/copper oxidases include a diverse group of enzymes that participate in aerobic respiration by reducing molecular oxygen to water and pumping protons across an associated membrane, thereby generating an electrochemical gradient [6-7]. This superfamily includes the

A-type aa3 oxidases traditionally studied from mammalian and bacterial sources, the B-type oxidases including ba3 from Thermus thermophilus, and C-type oxidases including cbb3 oxidase

[7]. Despite the low sequence identity between members of the heme/copper oxidase superfamily, these enzymes are highly similar in overall protein fold, use of at least one

3 conserved proton uptake pathway, and their heme/copper binuclear active site [6]. The study of

A-type CcO is most translatable to human health and disease as mammals use this enzyme as their sole terminal oxidase [8-9].

Bos taurus, R. sphaeroides, and Paracoccus denitrificans have all been used as model organisms to investigate the structure and function of CcO [5, 10-15]. For this work, CcO from

R. sphaeroides 2.4.1 has been selected for investigation due to its simplified structure, genetic manipulability, and high (39-52%) sequence identity to the mitochondrially encoded subunits of mammalian CcO [8]. The genetically tractable R. sphaeroides CcO is also thought to be one of the most similar bacterial oxidases as compared to human CcO based on phylogenetic studies

[16].

Cell metabolism of Rhodobacter sphaeroides

R. sphaeroides is a purple nonsulfur bacterium with a versatile metabolism including fermentation, aerobic and anaerobic respiration, and photosynthesis [17-18]. These bacteria are classified as purple based on their bacteriochlorophyll and carotenoid derived cellular color and classified as nonsulfur based on their inability to utilize high concentrations of hydrogen sulfide during photosynthesis [17-18]. Photosynthesis is the process of energetic molecule production

(i.e., ATP and NADPH) by capturing and harvesting light energy. In higher plants and cyanobacteria, light capturing involves a complex of antenna proteins, cholorophyll, and carotenoids [19]. During energetic molecule production, electrons are acquired by the splitting of water and passed through the photosynthetic machinery starting with photosystem II, plastoquinone, cytochromes b6f, plastocyanin, and photosystem I, which reduces ferredoxin and

NADP to NADPH [19]. This electron transport is also coupled to proton pumping across a membrane, production of an electrochemical gradient, and ATP synthesis.

Purple photosynthetic bacteria differ in their photosynthetic apparatus and mechanism as compared to higher plants. These differences include the use of non-water electron donors and long wavelength absorbing bacteriochlorophyll pigments [20]. Additionally, R. sphaeroides contains a single photosystem that is most similar to photosystem II in higher plants [17, 20].

This system uses an excited bacteriochlorophyll P870 in the photosynthetic reaction center to sequentially reduce bacteriopheophytin and bound quinone, a quinone pool, cytochrome bc1, cytochrome c2, and back to the photosystem [17, 20]. The reduction of cytochrome bc1 may produce reducing equivalents (i.e., NADH) by reversing the electron flow through the electron transport chain complexes III and I [17, 20].

As an alternative to photosynthesis, R. sphaeroides may catabolize foodstuffs using either fermentation or respiration. The advantage of respiration, involving oxygen as the electron sink, over fermentation is the amount of released free energy as the redox potential difference between respiration substrates and products is much greater than that of fermentation. During catabolism, sugars are broken down by glycolysis, the pentose phosphate pathway, or the Entner-Doudoroff pathway, while fatty acids are broken down by β-oxidation [21]. The product, acetyl-CoA, may be further oxidized to carbon dioxide by the tricarboxylic acid cycle. The molecules reduced by these processes, NADH and FADH2, may then be utilized directly or indirectly used to synthesize ATP via the electron transport chain [22]. The advantage of the electron transport chain is that the reduction of molecular oxygen to water captures approximately half of the

5 reaction energy of this highly exergonic process, while limiting cellular damage [22].

Oxidative phosphorylation

Peter Mitchell published his chemiosmotic hypothesis in 1961 that described the coupling of membrane transport processes to the production of energetic molecules such as ATP [23].

This mechanism is now understood to couple electron transport chain redox reactions to proton pumping, thereby generating an electrochemical gradient used to drive energetically demanding cellular processes. The electron transport chain is composed of four multisubunit integral membrane proteins. The detailed study of these proteins has been enabled by the 1960s application of chemically defined ionic and non-ionic detergents, including cholate, to purify mitochondrial membrane proteins and the reconstitution of these proteins into proteoliposomes

[24].

NADH dehydrogenase (complex I) couples proton pumping to the two-electron transfer from an NADH substrate to quinone [22, 25]. This protein is composed of 45 subunits and has an L-shaped structure with a hydrophobic arm embedded in the membrane and a hydrophilic domain projecting from the membrane into the mitochondrial matrix (Figure 1.1) [25]. NADH dehydrogenase sequentially passes electrons from NADH to the bound flavin mononucleotide, then through seven iron-sulfur centers, and finally to a quinone bound at the transmembrane interface [22, 25]. It is thought that the quinone cofactor must extend at least 10 Å out of the membrane environment toward the final iron-sulfur center in order to participate in electron transfer [25]. The hydrophobic domain of NADH dehydrogenase is composed of three homologous subunits that are proposed to be involved in proton pumping. One of these

6 hydrophobic subunits includes an amphipathic helix that runs parallel to the surface of the membrane and appears to connect the entire hydrophobic domain [25]. The current model for proton pumping involves quinone reduction driving conformational changes of this transverse helix, which in turn drives conformational changes required for proton pumping in each of the membrane subunits [25].

In addition to NADH dehydrogenase, succinate dehydrogenase or complex II is capable of reducing quinones. Succinate dehydrogenase functions in both the electron transport chain and the tricarboxylic acid cycle by coupling the oxidation of succinate to fumarate in the mitochondrial matrix and the reduction of quinone [22, 26]. This protein exists as a trimer, with catalysis independently occurring in each of its monomers. Each monomer consists of a flavoprotein, a hydrophilic iron-sulfur protein, and two integral membrane proteins (Figure 1.1).

In the redox reaction, electrons are sequentially passed from succinate to a covalently bound flavin adenine nucleotide, three iron-sulfur centers, and finally to the quinone [22, 26].

Succinate dehydrogenase also includes a heme b cofactor. It is hypothesized that this heme molecule, rather than being involved in electron transport, serves as an electron sink that shortens the lifetime of the semiquinone intermediate and thus limits the formation of reactive oxygen species [26].

The reduced quinols shuttle electrons through the membrane environment to cytochrome bc1, also known as complex III of the electron transport chain. Cytochrome bc1 is an integral component of aerobic respiration and bacterial photosynthesis [22, 27]. Each monomer of the functional dimer contains 11 polypeptide chains with two heme b molecules, one heme c, and

7 one iron-sulfur cluster. These proteins catalyze electron transfer from the two-electron carrier quinol to the single electron carrier cytochrome c through a process described as the proton motive Q cycle [28]. This cycle involves a bifurcated electron transfer mechanism with one electron donated to the iron-sulfur center, the bound cytochrome c1, and finally to the cytochrome c protein, while a second electron is sequentially transferred from the quinol to the low-spin heme b, high-spin heme b, and finally to a second quinone molecule [27]. Proton pumping is also coupled to the Q cycle as quinols are protonated on the internal side of the membrane and deprotonated on the external side of the membrane [22, 27].

The Q cycle must occur twice to transfer the two electrons carried by quinol to the single electron carrier, cytochrome c. Additionally, a single cycle revolution produces an unstable semiquinone radical [27]. Highly reactive and damaging radicals are produced as byproducts of electron transport by cytochrome bc1 and NADH dehydrogenase [27]. Cells have evolved mechanisms to mitigate damage by reactive oxygen species including dissipation by superoxide dismutase, catalase, and glutathione peroxidase [29] or oxidative phosphorylation regulation by cytochrome c oxidase.

Figure 1.1: The electron transport chain (following page). The electron transport chain is composed of mitochondrial complexes I-IV and ATP synthase. Complex I (PDB: 2FUG [133] and 3RKD [134]) couples the electron transfer from NADH to quinone (purple sticks) and proton pumping, while complex II (PDB: 1ZOY [135]) transfers electrons from succinate to quinone. Complex III (PDB: 3H1L [136]) oxidizes these quinols, reducing cytochrome c and pumping protons. Cytochrome c (green ribbon with magenta heme stick cofactor; PDB: 3CYT [137]) transfers a single electron to complex IV (PDB: 2DYR [138]) where molecular oxygen is reduced to water and additional protons are pumped. The complex I, III, and IV proton pumping from the mitochondrial matrix (left side), across the inner membrane (gray plane), to the intermembrane space (right side) generates an electrochemical gradient that is used by ATP synthase (PDB: 2XND [139]) to generate ATP from ADP and inorganic phosphate by rotary catalysis. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation.

Figure 1.1 (cont’d)

Cytochrome c accepts electrons from cytochrome bc1 and reduces cytochrome c oxidase.

Cytochrome c oxidase, complex IV (CcO), is an integral membrane protein with three mitochondrial encoded subunits and ten nuclear encoded subunits (Figure 1.2) [5]. CcO sequentially passes electrons from cytochrome c through copper and heme redox centers to reduce molecular oxygen at its active site [8, 30]. Oxygen reduction is coupled to pumping up to four protons from the matrix to the mitochondrial inner membrane space and the resultant generation of an electrochemical gradient [8, 30].

Figure 1.2: Structure of bacterial and mammalian cytochrome c oxidases. (A) Rhodobacter sphaeroides (PDB: 1M56 [37]) and (B) Bos taurus (PDB: 2DYR [138]) CcO structures are depicted as ribbons with homologous subunits I-III colored in green, cyan, and magenta, respectively. The bacterial subunit IV and mammalian, nuclear encoded accessory subunits are represented in uniquely colored ribbons. Heme (magenta sticks) and copper (copper colored spheres) cofactors are also shown. The conserved D- and K-pathways for proton uptake are represented as dashed lines, leading from the internal side of the membrane to the active site. The approximate location of the membrane is shown with gray spheres and curved lines. The bacterial interfacial lipids are represented as atomic detailed lipids (gray sticks in panel A only).

The electrochemical gradient produced by the electron transport chain generates a proton motive force that may be used by ATP synthase to synthesize ATP from ADP and inorganic phosphate by rotary catalysis [22, 31]. ATP synthase is a highly conserved multisubunit protein found in mitochondrial, chloroplast, and bacterial membranes [22, 31]. It is composed of an integral membrane domain known as F0 and a hydrophilic domain known as F1 linked together by central and peripheral stalks [31]. F0 contains channels through which protons may diffuse down their gradient. The energy that is released drives the rotation of F0 which in turn rotates the closely associated central stalk, leading to conformational changes in the F1 domain [31]. The F1 domain is composed of six alternating α and β-subunits that create three interfacial active sites.

As the central stalk rotates, it sequentially impacts the F1 subunits causing them to undergo a series of conformational changes in each of the catalytic sites, ultimately converting ADP and inorganic phosphate to ATP [22, 31]. In addition to generating ATP, ATP-synthase forms dimer and oligomer states that influence inner membrane curvature, local membrane potential, and mitochondrial morphology [32-33].

Rhodobacter sphaeroides cytochrome c oxidase

Bacterial electron transport chains vary from their mitochondrial counterparts in their use of additional and alternative cytochromes and extensive branching, incorporating several terminal oxidases [17]. These terminal oxidases are thought to exist as an evolved mechanism to survive under varying oxygen conditions with high affinity oxidases, including the R. sphaeroides cytochrome cbb3, functioning maximally in microaerobic conditions and lower affinity oxidases, including the R. sphaeroides cytochrome aa3, functioning in aerobic conditions

[17]. The native terminal oxidases in R. sphaeroides are thought to have 70% redundant activity

[17]. The deletion of the cytochrome cbb3 gene for the study of CcO alone stimulates photosynthesis and represses respiration [17]. Therefore both oxidases are typically natively expressed for CcO studies.

R. sphaeroides CcO is a four subunit enzyme with its first three subunits being homologous to the mitochondrially encoded core of mammalian CcO (Figure 1.2) [34-36].

Subunit I is a 62.6 kDa, twelve α-helical transmembrane protein that includes the sites of heme coordination, the binuclear active site, proton uptake pathways, and the oxygen uptake channel

[4, 37]. Subunit II is a 33 kDa protein with two transmembrane α-helices and an extracellular β- sheet domain [4, 37]. This subunit also contains the initial site of electron acceptance from cytochrome c, the copper B site [37]. Subunit III is a 30 kDa protein [4], with seven transmembrane α-helices which are grouped into two bundles, allowing a V-shaped cleft to be formed in which lipids are bound [37-38]. Subunit III contains no redox metal centers nor proton pathways but plays a role in stabilizing the binuclear center through a currently unknown mechanism [4, 38-39]. Subunit IV (6.3 kDa) is composed of a single transmembrane α-helix with no identified function to date [37].

CcO accepts electrons from cytochrome c which are consumed upon reduction of molecular oxygen at its active site. Electrons are sequentially passed through a bimetallic copper

A, heme a, and finally to the binuclear active site composed of heme a3 and copper B [8, 30].

Electron transport is coupled to the uptake of protons through two conserved pathways: the K- and D-pathways (Figure 1.2). The K-pathway, named after the conserved and critical Lys362 residue (R. sphaeroides numbering is used unless otherwise noted), begins with a conserved

12 carboxylate residue on the interior of the membrane and leads to the active site via Lys362 and

Thr359. This uptake pathway takes up two protons during active site reduction [40]. Ultimately, these protons are consumed in the reduction of molecular oxygen. The D-pathway, named after the critical entry residue Asp132, takes up all other protons [6, 41-42]. This pathway is delineated by a chain of water molecules leading between the conserved Asn132 to Asp286.

Protons are transported to Asp286, where they are shuttled to the active site to be consumed or pumped to the opposing side of the membrane [43-44]. A clear definition of the proton exit pathway has yet to be identified but may involve the heme propionate groups, associated waters, or a pair of conserved arginines: Arg481 and Arg482 [45-46]. An additional pathway for proton exit, the H-pathway, has been proposed based on the mammalian structure. However, due to the lack of conservation of the H-pathway residues across species, it is not likely that this channel is universally used for proton pumping [47].

Through the advance of membrane protein crystallography, a bifurcated hydrophobic oxygen pathway has been defined in mammalian and bacterial CcO as well as in T. thermophilus ba3 oxidase [48-49]. This bifurcated channel in subunit I has two entries, one between helices I,

II and III and another between helices IV and V with both branches leading to the binuclear active site. Functional relevance of the helices IV-V entry has been supported by molecular dynamics simulations of oxygen diffusion [50] and mutational analysis [51]. The second entryway, near helices I-III, has higher oxygen affinity and has been suggested to function as an oxygen vestibule [49].

CcO oxygen reduction chemistry has been studied largely using Raman and resonance

Raman spectroscopy. Raman spectroscopy is a technique used to characterize low frequency motions (e.g., bond vibrations or rotations) by monitoring the ability to shift the frequency of monochromatic light [52-55]. Specific chromophore excitation and monitoring of Raman shifts, called resonance Raman spectroscopy, has been extensively applied to understand the CcO reaction cycle intermediates [52-55]. Based on these studies, a model of CcO oxygen reduction has been generally accepted (Figure 1.3) [8, 52-55]. First, a single electron is passed from copper A, through heme a, to reduce copper B and form the E intermediate. Upon the addition of a second electron, both copper B and heme a3 become reduced and form the R intermediate.

Molecular oxygen can then bind to the doubly reduced binuclear center to form the A intermediate. This intermediate is immediately reduced in a four electron reaction, breaking the oxygen double bond using electrons from CuB, heme a3 and a neighboring Tyr288, to form a ferryl heme a3 and cuprous hydroxide copper B, called the P intermediate. Finally, the addition of a third electron from cytochrome c restores Tyr288 to its native state, and the addition of a fourth electron returns the active site to its original state with a hydroxyl group bound to each of the heme a3 and copper B metals. The protonation of these hydroxyl groups forms the water molecule product.

The cytochrome c oxidase oxygen reduction chemistry is coupled to proton uptake and pumping. Specifically, pumping is observed in the P to F, F to O, and O to R intermediate transitions [8, 56]. The stochiometry of this pumping, or CcO efficiency, may be controlled directly or by intrinsic uncoupling where protons are taken up from the external side of the membrane [30]. This control of proton pumping efficiency has been suggested not only to regulate CcO activity but to contribute to the control of oxidative phosphorylation [30].

Figure 1.3: Cytochrome c oxidase oxygen reduction mechanism. Each observed reaction intermediate is represented as a box with the binuclear active site heme a3 iron and copper B metals, redox state, and ligands depicted. A covalently linked His284/Tyr288 is found near the CcO active site and is represented as 'YOH'. Proton, electron, and oxygen uptake are depicted by arrows leading into the reaction cycle, while water molecule release is depicted by arrows diverging from the cycle. Proton pumping is represented by dashed lines during oxidized to E- intermediate, peroxy to ferryl, and ferryl to oxidized state transitions. This figure was created by combining figures provided by Dr. Denise Mills and [56].

Regulation of cytochrome c oxidase

CcO may be a key regulator of cell metabolism as this protein controls oxidative phosphorylation efficiency, has highly regulated expression and assembly, and is the only protein in the electron transfer chain that has a high level of tissue specific variation [57-60]. The

15 number of accessory CcO subunits, subunits in addition to the subunit I-III core, varies between evolutionary clades, where bacteria typically contain one subunit, Saccharomyces cerevisiae contain seven, plants and fish vary between seven and ten, and mammals contain ten [60-62]. It is hypothesized that these subunits and their isoforms evolved in the eukaryotic host genome to regulate the activity of mitochondrially encoded CcO [60]. However, in no case has the functional or regulatory role of these subunits been unequivocally established.

The function of accessory subunits has been reported to include promoting enzyme assembly as seen in the case of subunit VIb (with accessory subunits described in mammalian numbering [63]), contributing to structural stability as seen with subunits Vb and VIa/b [5], and providing allosteric regulatory sites specific for oxidative phosphorylation or hormonal control

[60, 64]. These functions appear to be adapted to both tissue specific need and environmental resource availability via isoforms. CcO has both liver and heart isoforms where liver isoforms are expressed in tissue poor in mitochondria and with high CcO activity and heart isoforms are expressed in tissue rich in mitochondria and low CcO activity [60]. Liver and heart isoforms have been observed for subunits VI, VIa/b, and VIII [60-61, 65-67].

In addition to alternative accessory subunits, CcO activity appears to be regulated by phosphorylation and feedback inhibition by adenine nucleotides. At least fourteen phosphorylation sites have been characterized on mammalian CcO, including three sites on the subunit I-III core [61, 68-69]. These sites include Tyr304/Tyr327 (B. taurus / R. sphaeroides numbering) and Ser115/Ser156 in subunit I and Ser126/Ser173 in subunit II, all of which are highly conserved from bacteria to mammals [61, 70-71]. The phosphorylation of

Tyr304/Tyr327 inhibits CcO activity and affects small molecule allosteric regulation of CcO

[70]. Other phosphorylation sites are located on the accessory subunits IV, Va/b, and VIa with subunit VIa phosphorylation being dependent on the expressed tissue specific isoform [61, 72].

The relative concentrations of ATP and ADP have also been observed to regulate CcO activity by feedback inhibition. Under very high [ATP]/ [ADP] ratios, in vitro activity of CcO is inhibited and the affinity for cytochrome c is reduced [73], possibly due to allosteric binding of

ATP on the matrix domains of subunits IV and VIa [68]. However, the observation of this inhibition is not clearly understood and is highly variable in differing detergent and phosphorylation conditions [74]. Due to CcO inhibition only being observed under high [ATP]/

[ADP] and low reduced/oxidized cytochrome c conditions, it is postulated that adenine nucleotide regulation contributes to maintaining the homeostatic 100-140 mV membrane potential, a level that will lower radical oxygen species production and proton backflow [74].

Role of membrane lipids in cytochrome c oxidase stabilization and activity

The investigation of CcO function was greatly advanced with the development of purification and proteoliposome reconstitution protocols beginning in the 1950s [2]. These purification protocols relied on solubilization of the enzyme in anionic detergents and subsequent reincorporation of the protein into lipid environments. The removal of native phospholipids [75-

76] and addition of cholate [77] or Triton-X100 [78] for solubilization drastically decreased enzyme turnover but activity could be restored upon addition of exogenous phospholipids [79].

The role of lipids as functional components of CcO, rather than just as an annular layer of hydrophobic solvent, remains incompletely understood [80]; however, crystallographic structure

17 comparisons support the conservation and hence functional significance of CcO associated lipids.

Comparisons of mammalian and bacterial CcO crystal structures have identified conserved lipid binding grooves on the membrane surface of CcO [12]. These grooves contain highly conserved residues and are bound by lipids or detergents in similar binding orientations.

Cardiolipin and conserved detergent binding sites are present at the subunit/subunit interfaces as well as monomer/monomer interfaces in the mammalian enzyme, suggesting that lipids may play a key role in protein-protein interactions either by stabilizing subunit associations [37-38], oligomeric states [5, 12], or its association with other integral membrane proteins [81-84]. In R. sphaeroides CcO subunits I and IV are associated almost exclusively through lipid mediated contacts with four lipids bound at both sides of the membrane [37]. The subunit I/III interface also contains two conserved lipid binding sites which are occupied by either phosphotidylglycerol or phosphotidylethanolamine [5, 37]. Removal of these subunit I/III interfacial lipids destabilizes the catalytic core and increases catalytic turnover-induced enzyme inactivation, known as suicide inactivation [38]. Based on the location of crystallographically observed lipids and detergents and on inhibition observed upon delipidation, CcO may structurally and functionally require specific lipid components that cannot be substituted by a non-specific amphipathic bilayer environment.

Need for computational structural biology to elucidate cytochrome c oxidase function

Cytochrome c oxidase has been studied for nearly a century due having a central role in bioenergetics as well as being a model system for studying oxygen chemistry, proton pumping,

18 integral membrane protein assembly, and cofactor insertion. This body of work has enabled research groups to understand the mechanism of reducing molecular oxygen without radical generation [52-55] and of proton uptake through a series of protonatable residues and conserved water molecule networks [41, 85-88]. Eukaryotic and prokaryotic CcO assembly variants [89-

90] and subsequent oxidase characterization have elucidated integral membrane protein assembly as well as cofactor insertion mechanisms as reviewed in [91]. Finally, methodological advances in spectroscopy [52-55] and crystallography are now available to the scientific community based on the efforts to study CcO, including new effective detergents [92], native membrane lipid retention [12], lipid cubic phase crystallization [7], and in-beam spectral monitoring [88]. In spite of all the experimental advances, there is a clear need to utilize rapidly developing and increasingly powerful computational methods to address the still outstanding questions regarding

CcO function.

First, the mechanism of coupling oxygen reduction to proton pumping remains a topic of debate. Currently available experimental techniques and previous modeling efforts are limited to understanding local protein changes on short timescales or averaged over long timescales, with no global view of coupling during the entire reaction. The current view suggests CcO conformational changes may be involved in alternate use of the proton pathways in a redox dependent manner. However, inconsistencies are observed when comparing crystallographic

[13, 47] and solvent accessibility [56] studies of CcO. Additionally, there is no clear mechanism for proton uptake through the discontinuous K-pathway nor oxygen uptake through the proposed hydrophobic channel, especially in light of these conformational changes. The fundamental question of coupling oxygen reduction, proton pumping, and oxygen uptake from a global

19 perspective and considering the millisecond reaction time frame, may be addressed using a combination of intrinsic protein backbone flexibility and low energy mobility calculations (See

Chapter 2). Additionally, comparisons of these models to experimentally obtained data may provide a more unified CcO mechanism and testable hypotheses for future work.

Secondly, it is unclear as to what role native lipids play in the structure and function of

CcO. The removal of certain lipids or replacement by specific detergents reduces enzyme stability and activity and, in some cases, results in permanent inhibition. Are the studies of purified integral membrane proteins then misleading or invalid in the absence of at least a subset of native lipids? For example, the reported activity of the purified CcO E101A variant varies between 5-50% of the activity of wild type (WT) [86, 93]. This variation may be explained entirely based on detergent conditions (Hiser, C., et al., in review) (See Chapter 3). The identification of key lipid components and development of methods to retain these lipids is of critical importance. Insights into the question of detergent inhibition, for the E101A variant and potentially other systems, may be gained by identifying conserved lipid binding grooves and by docking known inhibitory detergents into characterized sites to assess their ability to interact competitively.

Finally, whole cell and isolated mitochondrial studies suggest that a number of small molecules regulate metabolism and cell fate at the level of CcO. These molecules include hormones such as testosterone and thyroid horomones [68], porphyrins including bilirubin and protoporphyrin IX [94-95], palmitic acid [96], and adenine nucleotides [11, 68, 73]. However, it is unclear as to the location and mechanism of CcO regulation by most small molecules. Native

20 regulatory ligands of CcO may be identified, and their protein interactions characterized, using a combination of ligand comparisons, protein binding site comparisons, and small molecule docking (See Chapter 4). The mechanism by which these native ligands affect activity may again be understood in the context of CcO functional coupling through conformational change.

Computational predictions of protein flexibility and ligand interactions

Crystallographic characterization of the mammalian and bacterial CcO structures has elucidated the protein's fold and active site structure, and validated the use of bacterial model systems [5-6]. However, crystal structures have only provided a static view of CcO and have prompted many functional questions. In order to study a protein's physiological functions, which occur from femtosecond to second timescales [97], a wide range of biophysical and computational approaches must be applied. One of the most widely applied techniques for computationally predicting protein dynamics is molecular dynamic (MD) simulations. MD utilizes protein structures at atomic detail and complex potential functions to predict the behavior of macromolecules [97]. However, MD has several disadvantages for the study of CcO. First, this method's accuracy relies on calculating many short timescale steps that accumulate to physiological timescales, typically nanoseconds. The reaction mechanism of CcO is substantially longer than nanosecond timescales, and therefore, only local protein behavior may be predicted [98]. Secondly, MD simulations of large membrane proteins have been shown to be inaccurate due to the size of the macromolecule system and unsuitable hydrophobic potential functions [99].

The behavior of CcO, like other large membrane proteins, may be studied using

21 alternative methods. One such method is normal modes analysis (NMA). NMA uses a harmonic approximation to predict vibrational dynamics of macromolecules [98]. One form of NMA, called the elastic network model, represents a protein not on an atomic scale, but by secondary structure topology about α-carbons or individual residues [100]. This elastic network variation is especially useful in predicting cooperative motions [101] which may be critical in understanding

CcO coupling of electron acceptance, proton pumping, and oxygen uptake. Despite the simplicity of the structural information and potential functions, NMA has predicted behavior consistent with experimental data including directional motion, domain movements, hinge locations, and allosteric regulation [98]. This method may then be applied to understand global

CcO low energy motions and allosteric conformational changes driven by the binding of small molecules such as nucleotides.

Two molecular models of protein recognition for their cognate ligands have been proposed. Fisher's lock and key model describes the rigid body complementarity between a protein binding site and ligand, generating specific high affinity complex formation [102] with a population of the protein in the right equilibrium conformation [103]. Koshland's induced fit model describes the process of altering protein conformation by binding of a ligand, resulting in a better fit of the protein-ligand complex than in the initial bound state [104]. Experimental characterization of protein flexibility supports this induced fit model with proteins varying from rigid globular to intrinsically disordered structures, but kinetic analysis suggests that lock and key conformational selection is more prevalent than previously thought [103, 105]. A combination of these models may be closest to the truth, especially where the purpose of the ligand binding event is to alter protein conformation to regulate activity.

Docking is the computational prediction of the orientation and affinity of a ligand in a protein binding site, considering chemical and geometric complementarity [106]. Docking is composed of two steps: a semi-exhaustive search to identify ligand binding poses and scoring to rank these poses and predict binding affinity. The search for chemical and geometric complementarity is exponentially more difficult as the number of rotational degrees of freedom

(i.e., number of rotatable bonds) increases. The greater the ligand flexibility, the greater the difficulty in accurately modeling the protein-ligand complex [106]. However, the accuracy of docking is also dependent on the incorporation of flexibility in both the protein binding site and the ligand, as most ligand-protein interactions involve some degree of induced fit [105]. Ligand flexibility may be incorporated in docking using Monte Carlo simulations of ligand conformations (e.g., International Coordinate Mechanics [107]), docking of rigid ligand fragments and subsequent incorporation of flexible ligand connecting regions (e.g., FlexX [108],

DOCK [109], etc.), docking of ligand conformer libraries (e.g., FRED [OpenEye Scientific

Software, Santa Fe, NM], Glide [Schrödinger, Inc.], SLIDE [110], etc.), genetic algorithms that combine dockings of ligand fragments with high affinities into a single docking with high affinity throughout the ligand (e.g., GOLD [111], Autodock [112], etc.), and simulated annealing where ligand temperatures are cycled to overcome local energy barriers (e.g., Autodock [112])

[106]. Protein flexibility remains a significant challenge. However, some docking methods, like

SLIDE, incorporate side chain rotations [113] or protein backbone hinging [106, 113-116]. After complementary protein-ligand complexes have been identified, a scoring function is used to rank the complexes and predict binding affinities. Scoring functions may be based on least-squares fitting or machine learning-based optimized functions using empirical data or statistical analysis

23 of known holoenzyme complexes [106]. Ultimately, docking may be used to understand the plausibility of ligand binding to a given site or the detailed molecular interactions within a protein-ligand complex.

Molecular docking accuracy is improved when multiple protein structures are included, either from nuclear magnetic resonance ensembles, accumulated crystal structures with slight variations, or computationally predicted structures. MD [117-118], NMA, and simulated annealing techniques have all been utilized to generate computational ensembles of physiologically relevant protein conformations either for understanding intrinsic protein flexibility or for improving small molecule docking [114, 119]. Alternative methods involving the incorporation of crystallographically observed backbone rotamer libraries (e.g., FlexE [120]) or intrinsic flexibility predictions (e.g., ProFlex [121], SOFTSPOTS [122], etc.) have also been applied to characterize protein flexibility [119]. Ultimately, predicting protein flexibility may be used to understand mechanistically critical conformational changes and interactions of proteins with small molecules.

In the absence of a known protein binding site or protein-ligand complex, ligand based alignment methods are often applied. These methods function by superimposing entire rigid ligands (e.g., OEChem RMSD [OpenEye Scientific Software, Santa Fe, NM]), several rigid ligand fragments separated by a flexible substructure (e.g., LigAlign [123]), or by chemical or geometric complementarity (e.g., ROCS [OpenEye Scientific Software, Santa Fe, NM]). By combining chemical and geometric complementarity, ligand alignments increasingly identify truly similar ligands by reducing the false positive selection rates [124]. These methods are

24 computationally advantageous as they require less computing time and resources than molecular docking or protein based methods [124]. However, the combination of ligand comparison methods with small molecule docking has been observed to increase the success of high throughput screening [125]. While the combination of ligand comparisons and protein binding site complementarity methods has identified more diverse ligand classes [124]. Importantly, combining ligand comparison methods with a second protein structure-based method can provide information about steric and chemical complementarity of the ligand with the protein.

Protein binding site comparison methods have been a popular topic of research with the vast increase of protein structures determined to atomic resolution. One group of binding site comparison methods uses knowledge of previously characterized amino acids sequences to propose a function for uncharacterized proteins [126]. Evolutionary conservation in a given site may also support functionality or specificity between homologs in a single protein family [127].

Alternatively, chemical and geometric information may be used to characterize diverse protein binding sites. These techniques often represent and compare sites based on three dimensional templates [128-130], functional group geometries [131], or surface complementarity [132].

Additionally, protein binding site comparisons may also be supported or validated using molecular docking [127].

Application of computational methods to predicting CcO flexibility and ligand binding may elucidate fundamental, unanswered questions regarding the mechanism of this protein’s ability to couple chemical reaction energy (oxygen reduction) to proton transfer and how this coupling is controlled by ligand binding. First, the coupling of oxygen reduction, proton

25 pumping, and oxygen uptake will be addressed using a combination of intrinsic flexibility and low energy mobility calculations to generate novel insight into the CcO reaction mechanism.

Secondly, observations of detergent inhibition will be rationalized based on conserved lipid binding grooves and the docking of known inhibitory detergents into these sites. Finally, potential native regulatory ligands of CcO will be identified and their protein interactions characterized using ligand and protein comparison and small molecule docking approaches. The mechanisms of detergent inhibition and small molecule regulation will be explained in the context of the functional implications of conformational change.

REFERENCES

1. Snyder, R. 2000. Cytochrome P450, the oxygen-activating enzyme in xenobiotic metabolism. Tox. Sciences. 58: 3-4.

2. Wikström, M., K. Krab, and M. Saraste. 1981. Cytochrome oxidase: A synthase. Academic Press. Toronto.

3. Nichols, P. 1984. Cytochromes and cell respiration: Carolina Biology Readers Series. Scientific Publications Dept., Carolina Biological Supply Co. Burlington, NC.

4. Hosler, J. P., J. Fetter, M. M. J. Tecklenburg, M. Espe, C. Lerma, and S. Ferguson-Miller. 1992. Cytochrome aa3 of Rhodobacter sphaeroides as a model for mitochondrial cytochrome c oxidase. Purification, kinetics, proton pumping, and spectral analysis. J. Biol. Chem. 267(34): 24264-24212.

5. Tsukihara, T., H. Aoyama, E. Yamashita, T. Tomizaki, H. Yamaguchi, K. ShinzawaItoh, R. Nakashima, R. Yaono, and S. Yoshikawa. 1996. The whole structure of the 13-subunit oxidized cytochrome c oxidase at 2.8 angstrom. Science. 272: 1136-1144.

6. Ferguson-Miller, S., and G. T. Babcock. 1996. Heme/copper terminal oxidases. Chem. Rev. 96(7): 2889-2908.

7. Tiefenbrunn, T., W. Liu, Y. Chen, V. Katritch, C. D. Stout, J. A. Fee, and V. Cherezov. 2011. High resolution structure of the ba3 cytochrome c oxidase from Thermus thermophilus in lipid environment. Plos ONE. 6(7): e22348-e22348.

8. Brzezinski, P., and R. B. Gennis. 2008. Cytochrome c oxidase: Exciting progress and remaining mysteries. J. Bioenerg. Biomembr. 40: 521-531.

9. Richter, O. M., and B. Ludwig. 2009. Electron transfer and energy transduction in the terminal part of the respiratory chain - lessons from bacterial model systems. Biochim. Biophys. Acta. 1787(6): 626-634.

10. Frank, V., and B. Kadenbach. 1996. Regulation of the H+/e- stoichiometry of cytochrome c oxidase from boving heart by intramitochondrial ATP/ADP ratios. FEBS Lett. 382: 121- 124.

11. Kadenbach, B., J. Napiwotzki, V. Frank, S. Arnold, S. Exner, and M. Hüttemann. 1998. Regulation of energy transduction and electron transfer in cytochrome c oxidase by

adenine nucleotides. J. Bioenerg. Biomembr. 30: 25-33.

12. Qin, L., C. Hiser, A. Mulichak, R. M. Garavito, and S. Ferguson-Miller. 2006. Identification of conserved lipid/detergent-binding sites in a high-resolution structure of the membrane protein cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 16117-16122.

13. Qin, L., J. Liu, D. A. Mills, D. A. Proshlyakov, C. Hiser, and S. Ferguson-Miller. 2009. Redox-dependent conformational changes in cytochrome c oxidase suggest a gating mechanism for proton uptake. Biochemistry. 48: 5121-5130.

14. Ostermeier, C., A. Harrenga, U. Ermler, and H. Michel. 1997. Structure at 2.7 Å resolution of the Paracoccus denitrificans two-subunit cytochrome c oxidase complexed with an antibody FV fragment. Proc. Natl. Acad. Sci. USA. 94: 10547-10553.

15. Forte, E., F. M. Scandurra, O. M. Richter, E. D'Itri, P. Sarti, M. Brunori, B. Ludwig, and A. Giuffré. 2004. Proton uptake upon anaerobic reduction of the Paracoccus denitrificans cytochrome c oxidase: a kinetic investigation of the K354M and D124N mutants. Biochemistry. 43(10): 2957-2963.

16. Gray, M. W., G. Burger, and B. F. Lang. 1999. Mitochondrial evolution. Science. 283: 1476- 1481.

17. Mackenzie, C., J. M. Eraso, M. Choudhary, J. H. Roh, X. Zeng, P. Bruscella, Á. Puskás, and S. Kaplan. 2007. Postgenomic adventures with Rhodobacter sphaeroides. Annu. Rev. Microbiol. 61: 283-307.

18. McEwan, A. G. 1994. Photosynthetic electron transport and anaerobic metabolism in purple non-sulfur phototrophic bacteria. Anton. Leeuw. 66: 151-164.

19. Nelson, N., and A. Ben-Shem. 2005. The complex architecture of oxygenic photosynthesis. Nat. Rev. Mol. Cell. Biol. 5: 971-982.

20. Hoffa, A. J., and J. Deisenhofer. 1997. Photophysics of photosynthesis. Structure and spectroscopy of reaction centers of purple bacteria. Phys. Rep. 287(1-2): 1-247.

21. Bar-Even., A., A. Flamholz, E. Noor, and R. Milo. 2012. Rethinking glycolysis: On the biochemical logic of metabolic pathways. Nat. Chem. Biol. 8: 509-517.

22. Smeitink, J., L. van den Heuve, and S. DiMauro. 2001. The genetics and pathology of oxidative phosphorylation. Nat. Rev. Genet. 2: 342-352.

23. Mitchell, P. 1961. Coupling of phosphorylation to electron and hydrogen transfer by a chemiosmotic type of mechanism. Nature. 191: 144-148.

24. Alberts, B., A. Johnson, J. Lewis, M. Raff, K. Roberts, and P. Walter. 2002. Molecular biology of the cell, 4th edition. Garland Science. New York.

25. Efremov, R. G., R. Baradaran, and L. A. Sazanov. 2010. The architecture of respiratory complex I. Nature. 465(7297): 441-445.

26. Hao, G.-F., F. Wang, H. Li, X.-L. Zhu, W.-C. Yang, L.-S. Huang, J.-W. Wu, E. A. Berry, and G.-F. Yang. 2012. Compuational discovery of picomolar QO site inhibitors of cytochrome bc1 complex. J. Am. Chem. Soc. 134: 11168-11176.

27. Zhang, Z., L. Huang, V. M. Shulmeister, Y.-I. Chi, K. K. Kim, L.-W. Hung, A. R. Crofts, E. A. Berry, and S.-H. Kim. 1998. Electron transfer by domain movement in cytochrome bc1. Nature. 392: 677-684.

28. Mitchell, P. 1976. Possible molecular mechanisms of the proton motive function of cytochrome systems. J. Theor. Biol. 62: 327-367.

29. Henchcliffe, C., and M. F. Beal. 2008. Mitochondrial biology and oxidative stress in Parkinson disease pathogenesis. Nat. Clin. Pract. Neuro. 4: 600-609.

30. Hosler, J. P., S. Ferguson-Miller, and D. A. Mills. 2006. Energy transduction: Proton transfer through the respiratory complexes. Annu. Rev. Biochem. 75:165-187.

31. Walker, J. E., and V. K. Dickson. 2006. The peripheral stalk of the mitochondrial ATP synthase. Biochim. Biophys. Acta. 1757: 286-296.

32. Davies, K. M., C Anselmi, I. Wittig, J. D. Faraldo-Gómez, and W. Kühlbrandt. 2012. Structure of the yeast F1Fo-ATP synthase dimer and its role in shaping the mitochondrial cristae. Proc. Natl. Acad. Sci. USA. 9109(34): 13602-13607.

33. Paumard, P., J. Vaillier, B. Coulary, J. Schaeffer, V. Soubannier, D. M. Mueller, D. Brèthes, J. P. di Rago, and J. Velours. 2002. The ATP synthase is involved in generating mitochondrial cristae morphology. EMBO J. 21(3): 221-230.

34. Shapleigh, J. P., and R. B. Gennis. 1992. Cloning, sequencing and deletion from the chromosome of the gene encoding subunit I of the aa3-type cytochrome c oxidase of Rhodobacter sphaeroides. Mol Microbiol. 6(5): 635-642.

35. Cao, J., J. Shapleigh, R. Gennis, A. Revzin, and S. Ferguson-Miller. 1991. The gene encoding cytochrome c oxidase subunit II from Rhodobacter sphaeroides; comparison of the deduced amino acid sequence with sequences of corresponding peptides from other species. Gene. 101(1):133-7.

36. Cao, J., J. Hosler, J. Shapleigh, A. Revzin, and S. Ferguson-Miller. 1992. Cytochrome aa3 of Rhodobacter sphaeroides as a model for mitochondrial cytochrome c oxidase. The coxII/coxIII operon codes for structural and assembly proteins homologous to those in yeast. J. Biol. Chem. 267(34): 24273-24278.

37. Svensson-Ek, M., J. Abramson, G. Larsson, S. Törnroth, P. Brzezinski, and S. Iwata. 2002. The X-ray crystal structures of wild-type and EQ(I-286) mutant cytochrome c oxidases from Rhodobacter sphaeroides. J. Mol. Biol. 321: 329-339.

38. Varanasi, L., D. Mills, A. Murphree, J. Gray, C. Purser, R. Baker, and J. Hosler. 2006. Altering conserved lipid binding sites in cytochrome c oxidase of Rhodobacter sphaeroides perturbs the interaction between subunits I and III and promotes suicide inactivation of the enzyme. Biochemistry. 45(50): 14896- 14907.

39. Bratton, M. R., M. A. Pressler, and J. P. Hosler. 1999. Suicide inactivation of cytochrome c oxidase: catalytic turnover in the absence of subunit III alters the active site. Biochemistry. 38(49): 16236-16245.

40. Brändén, M., F. Tomson, R. B. Gennis, and P. Brzezinski. 2002. The entry point of the K- proton-transfer pathway in cytochrome c oxidase. Biochemistry. 41: 10794-10798.

41. Fetter, J. R., J. Qian, J. Shapleigh, J. W. Thomas, A. García-Horsman, E. Schmidt, J. Hosler, G. T. Babcock, R. B. Gennis, and S. Ferguson-Miller. 1995. Possible proton relay pathways in cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 92: 1604-1608.

42. Lepp, H., L. Salomonsson, J.-P. Zhu, R. B. Gennis, and P. Brzezinski. 2008. Impaired proton pumping in cytochrome c oxidase upon structural alteration of the D pathway. Biochim. Biophys. Acta. 1777: 897-903.

43. Egawa, T., K. Gaesan, M. T. Lin, M. A. Yu, J. P. Hosler, S. R. Yeh, D. L. Rousseau, and R. B. Gennis. 2011. Differential effects of glutamate-286 mutations in the aa(3)-type cytochrome c oxidase from Rhodobacter sphaeroides and the cytochrome bo(3) ubiquinol oxidase from Escherichia coli. Biochim. Biophys. Acta. 1807(10): 1342-1348.

44. Vakkasoglu, A. S., J. E. Morgan, D. Han, A. S. Pawate, and R. B. Gennis. 2006. Mutations which decouple the proton pump of the cytochrome c oxidase from Rhodobacter sphaeroides perturb the environment of glutamate 286. FEBS Lett. 580(19): 4613-4617.

45. Lee, H. J., L. Ojemyr, A. Vakkasoglu, P. Brzezinski, and R. B. Gennis. 2009. Properties of Arg481 mutants of the aa3-type cytochrome c oxidase from Rhodobacter sphaeroides suggest that neither R481 nor the nearby D-propionate of heme a3 is likely to be the proton loading site of the proton pump. Biochemistry. 48(30): 7123-7131.

46. Mills, D. A., L. Geren, C. Hiser, B. Schmidt, B. Durham, F. Millet, and S. Ferguson-Miller. An arginine to lysine mutation in the vicinity of the heme propionates affects the redox potentials of the hemes and associated electron and proton transfer in cytochrome c oxidase. Biochemistry. 44(31): 10457- 10465.

47. Yoshikawa, S., K. Muramoto, K. Shinzawa-Itoh, H. Aoyama, T. Tsukiara, K. Shimokata, Y. Katayama, and H. Shimada. 2006. Proton pumping mechanism of bovine heart cytochrome c oxidase. Biochim. Biophys. Acta.1757: 1110-1116.

48. Luna, V. M., Y. Chen, J. A. Fee, and C. D. Stout. 2008. Crystallographic studies of Xe and Kr binding within the large internal cavity of cytochrome ba3 from Thermus thermophilus. Biochemistry. 47: 4657-4665.

49. Luna, V. M., J. A. Fee, A. A. Deniz, and C. D. Stout. 2012. Mobility of Xe atoms within the oxygen diffusion channel of cytochrome ba3 oxidase. Biochemistry. 51: 4669-4676.

50. Hofacker, I., and K. Schulten. 1998. Oxygen and proton pathways in cytochrome c oxidase. Proteins. 30: 100-107.

51. Riistama, S., A. Puustinen, M. I. Verkhovsky, J. E. Morgan, and M. Wikström. 2000. Binding of O2 and its reduction are both retarded by replacement of valine 279 by isoleucine in cytochrome c oxidase from Paracoccus denitrificans. Biochemistry. 39: 6365-6372.

52. Han, S., S. Takahashi, and D. L. Rousseau. 2000. Time dependence of the catalytic intermediates in cytochrome c oxidase. J. Biol. Chem. 275(3): 1910-1919.

53. Proshlyakov, D. A., T. Ogura, K. Shinzawa-Itoh, S. Yoshikawa, and T. Kitagawa. 1996. Resonance raman/absorption characterization of the oxo intermediates of cytochrome c oxidase generated in its reaction with hydrogen peroxide: pH and H2O2 concentration dependence. Biochemistry. 35(26): 8580-8586.

54. Proshlyakov, D. A., M. A. Pressler, and G. T. Babcock. 1998. Dioxygen activation and bound cleavage by mixed-valence cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 95(14): 8020-8025.

55. Ogura, T., S. Yoshikawa, and T. Kitagawa. 1985. Resonance raman spectra for catalytic intermediates of cytochrome c oxidase detected with a mixed flow transient apparatus. Biochim. Biophys. Acta. 832(2): 220-223.

56. Busenlehner, L. S., L. Salomonsson, P. Brzezinski, and R. N. Armstrong. 2006. Mapping protein dynamics in catalytic intermediates of the redox-driven proton pump cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 15398-15403.

57. Pacelli, C., D. Latorre, T. Cocco, F. Capuano, C. Kukat, P. Seibel, and G. Villani. 2011. Tight control of mitochondrial membrane potential by cytochrome c oxidase. Mitochondrion. 11: 334-341.

58. Villani, G., and G. Attardi. 1997. In vivo control of respiration by cytochrome c oxidase in wild-type and mitochondrial DNA mutation-carrying human cells. Proc. Natl. Acad. Sci. USA. 94: 1166-1171.

59. Hüttemann, M., S. Helling, T. H. Sanderson, C. Sinkler, L. Samaviti, G. Mahapatra, A. Varughuese, G. Lu, J. Liu, L. I. Grossman, J. W. Doan, K. Marcus, and I. Lee. 2012. Regulation of mitochondrial respiration and apoptosis through cell signaling:

Cytochrome c oxidase and cytochrome c in ischemia/reprefusion injury and inflamation. Biochim. Biophys. Acta. 1817: 598-609.

60. Pierron, D., D. E. Wildman, M. Hüttemann, G. C. Markondapatnaikuni, S. Aras, and L. I. Grossman. 2012. Cytochrome c oxidase: Evolution of control via nuclear subunit addition. Biochim. Biophys. Acta. 1817: 590-597.

61. Hüttemann, M., I. Lee, L. I. Grossman, J. W. Doan, and T. H. Sanderson. 2012. Phosphorylation of mammalian cytochrome c and cytochrome c oxidase in the regulation of cell densiny: Respiration, apoptosis, and human disease. Adv. Exp. Med. Biol. 748: 237-264.

62. Das, J., S. T. Miller, and D. L. Stern. 2004. Comparison of diverse protein sequences of the nuclear-encoded subunits of cytochrome c oxidase suggest conservation of structure underlies evolving functional sites. Mol. Biol. Evol. 21: 1572-1582.

63. LaMarche, A. E. P., M. I. Abate, S. H. P. Chan, and B. L. Trumpower. 1992. Isolation and characterization of COX12, the nuclear gene for a previously unrecognized subunit of Saccharomyces cerevisiae cytochrome c oxidase. J. Biol. Chem. 276: 22473-22480.

64. Arnold, S., F. Goglia, and B. Kadenbach. 1998. 3,5-Diiodothyronine binds to subunit Va of cytochrome-c oxidase and abolishes the allosteric inhibition of respiration by ATP. Eur. J. Biochem. 252: 325-330.

65. Hüttemann, M., B. Kadenbach, and L. I. Grossman. 2001. Mammalian subunit IV isoforms of cytochrome c oxidase. Gene. 267: 111-123.

66. Grossman, L. I., and M. I. Lomax. 1997. Nuclear encoded genes for cytochrome c oxidase. Biochim. Biophys. Acta. 1352:174-192.

67. Hüttemann, M., T. R. Schmidt, and L. I. Grossman. 2003. A third form of cytochrome c oxidase subunit VII is present in mammals. Gene. 312: 95-102.

68. Bender, E., and B. Kadenbach. 2000. The allosteric ATP-inhibition of cytochrome c oxidase activity is reversibly switched on by cAMP-dependent phosphorylation. FEBS Lett. 466: 130-134.

69. Lee, I., E. Bender, and B. Kadenbach. 2002. Control of mitochondrial membrane potential and ROS formation by reversible phosphorylation of cytochrome c oxidase. Mol. Cell. Biochem. 234-235 (1-2): 63-70.

70. Lee, I., A. R. Salomon, S. Ficarro, I. Mathes, F. Lottspeich, L. I. Grossman, and M. Hüttemann. 2004. cAMP-dependent tyrosine phosphorylation of subunit I inhibits cytochrome c oxidase activity. J. Biol. Chem. 280(7): 6094-6100.

71. Fang, J. K., S. K. Prabu, N. B. Sepuri, H. Raza, H. K. Anandatheerthavarada, D. Galati, J.

Spear, and N. G. Avadhani. 2007. Site specific phosphorylation of cytochrome c oxidase subunits I, IVi1, and Vb in rabbit hearts subjected to ischemia/reprefusion. FEBS Lett. 581: 1302-1310.

72. Tsukihara, T., K. Shimokata, Y. Takayama, H. Shimada, K. Muramoto, H. Aoyama, M. Mochizuki, K. Shinzawa-Itoh, E. Yamashita, M. Yao, Y. Ishimura, and S. Yoshikawa. 2003. The low spin heme of cytochrome c oxidase as the driving element of the proton- pumping process. Proc. Natl. Acad. Sci. USA. 100: 15304-15309.

73. Arnold, S., and B. Kadenbach. 1997. Cell respiration is controlled by ATP, an allosteric inhibitor of cytochrome-c oxidase. Eur. J. Biochem. 249: 350-354.

74. Ramzan, R., K. Staniek, B. Kadenbach, and S. Vogt. 2010. Mitochondrial respiration and membrane potential are regulated by the allosteric ATP-inhibition of cytochrome c oxidase. Biochim. Biophys. Acta. 1797: 1672-1680.

75. Yu, C. A., L. Yu, and T. E. King. 1975. Studies on cytochrome-oxidase - interactions of cytochrome-oxidase protein with phospholipids and cytochrome-c. J. Biol. Chem. 250: 1383-1392.

76. Vik, S. B., and R. A. Capaldi. 1977. Lipid requirements for cytochrome-c oxidase activity. Biochemistry. 16: 5755-5759.

77. Van Buuren, K. J. H., and B. F. Van Gelder. 1974. Biochemical and biophysical studies on cytochrome-c oxidase: Effect of cholate on enzymic activity. Biochim. Biophys. Acta. 333: 209-217.

78. Antalik, M., D. Jancura, G. Palmer, and M. Fabian. 2005. A role for the protein in internal electron transfer to the catalytic center of cytochrome c oxidase. Biochemistry. 44: 14881-14889.

79. Brierley, G. P., and A. J. Merola. 1962. Studies of electron-transfer system: Phospholipid requirements in cytochrome oxidase. Biochim. Biophys. Acta. 64: 205-217.

80. Ferguson-Miller, S., C. Hiser, and J. Liu. 2012. Gating and regulation of the cytochrome c oxidase proton pump, Biochim. Biophys. Acta 1817: 489-494.

81. Stroh, A., O. Anderka, K. Pfeiffer, T. Yagi, M. Finel, B. Ludwig, and B. Schägger. 2004. Assembly of respiratory complexes I, III, and IV into NADH oxidase supercomplex stabilizes complex I in Paracoccus denitrificans. J. Biol. Chem. 279(6): 5000-5007.

82. Dudkina, N. V., R. Kouril, K. Peters, H. P. Braun, and E. J. Boekema. 2010. Structure and function of mitochondrial supercomplexes. Biochim. Biophys. Acta. 1797: 664-670.

83. Althoff, T., D. J. Mills, J. L. Popot, and W. Kuhlbrandt. 2011. Arrangement of electron transport chain components in bovine mitochondrial supercomplex I1III2IV1. EMBO J.

30: 4652-4664.

84. Bottinger, L., S. E. Horvath, T. Kleinschroth, C. Hunte, G. Daum, N. Pfanner, and T. Becker. 2012. Phosphatidylethanolamine and cardiolipin differentially affect the stability of mitochondrial respiratory chain supercomplexes. J. Mol. Biol. 423: 677-686.

85. Jünemann, S., B. Meunier, R. B. Gennis, and P. R. Rich. 1997. Effects of mutation of the conserved lysine-362 in cytochrome c oxidase from Rhodobacter sphaeroides. Biochemistry. 36: 14456-14464.

86. Tomson, F. L., J. E. Morgan, G. P. Gu, B. Barquera, T. V. Vygodina, and R. B. Gennis. 2003. Substitutions for glutamate 101 in subunit II of cytochrome c oxidase from Rhodobacter sphaeroides result in blocking the proton-conducting K-channel. Biochemistry. 42: 1711- 1717.

87. Seibold, S. A., D. A. Mills, S. Ferguson-Miller, and R. I. Cukier. 2005. Water chain formation and possible proton pumping routes in Rhodobacter sphaeroides cytochrome c oxidase. Biochemistry. 44: 10475-10485.

88. Liu, J., L. Qin, and S. Ferguson-Miller. 2011. Crystallographic and online spectral evidence for role of conformational change and conserved water in cytochrome oxidase proton pump. Proc. Natl. Acad. Sci. USA. 108: 1284-1289.

89. Varanasi, L. and J. P. Hosler. 2012. Subunit III-depleted cytochrome c oxidase provides insight into the process of proton uptake by proteins. Biochim. Biophys. Acta. 1817(4): 545-51.

90. Neehaul, Y., Y. Chen, C. Werner, J. A. Fee, B. Ludwig, and P. Hellwig. 2012. Electrochemical and infrared spectroscopic analysis of the interaction of the Cu(A) domain and cytochrome c(552) from Thermus thermophilus. Biochim. Biophys. Acta. 1817(10): 1950-1954.

91. Carr, H. S., and D. R. Winge. 2003. Assembly of cytochrome c oxidase withing the mitochondrion. Acc. Chem. Res. 36(5): 309-316.

92. Rosevear, P., T. Van Aken, J. Baxter, and S. Ferguson-Miller. 1980. Alkyl glucoside detergents. A simpler synthesis and their effects on the kinetic and physical properties of cytochrome oxidase. Biochemistry. 19: 4108-4115.

93. Qin, L., M. A. Sharpe, R. M. Garavito, and S. Ferguson-Miller. 2007. Conserved lipid- binding sites in membrane proteins: a focus on cytochrome c oxidase. Curr. Opin. Struc. Biol. 17: 444-450.

94. Malik, S. G., K. A. Irwanto, J. D. Ostrow, and C. Tiribelli. 2010. Effect of bilirubin on cytochrome c oxidase activity of mitochondria from mouse brain and liver. BMC Res. Notes. 3: 162-167.

95. Vaz, A. R., M. Delgado-Esteban, M. A. Brito, J. P. Bolanos, D. Brites, and A. Almeida. 2010. Bilirubin selectively inhibits cytochrome c oxidase activity and induces apoptosis in immature cortical neurons: Assessment of the protective effects of glycoursodeoxycholic acid. J. Neurochem. 112: 56-65.

96. Lee, I., and B. Kadenbach. 2001. Palmitate decreases proton pumping of liver-type cytochrome c oxidase. Eur. J. Biochem. 268: 6329-6334.

97. Dror, R.O., M. O. Jensen, D. W. Borhani, and D. E. Shaw. 2010. Exploring atomic resolution physiology on a femtosecond to millisecond timescale using molecular dynamics simulations. J. Gen. Physiol. 135: 555–562.

98. Isin, B., K. C. Tirupula, Z. N. Oltvai, J. Klein-Seetharaman, and I. Bahar. 2012. Identification of motions in membrane proteins by elastic network models and their experimental validation. Methods Mol. Biol. 914: 285-317.

99. Grossfield, A., S. E. Feller, and M. C. Pitman. 2007. Convergence of molecular dynamics simulations of membrane proteins. Proteins. 67: 31–40.

100. Bahar, I., T. R. Lezon, A. Bakan, and I. H. Shrivastava. 2010. Normal mode analysis of biomolecular structures: Functional mechanisms of membrane proteins. Chem. Rev. 110: 1463–1497.

101. Bahar, I., and A. J. Rader. 2005. Coarse-grained normal mode analysis in structural biology. Curr. Opin. Struct. Biol. 15: 586–592.

102. Fischer, E. 1894. Einfluss der configuration auf die wirkung der enzyme. Ber Dtsch Chem Ges 27: 2985–2993.

103. Vogt, A. D., and E. Di Cera. 2012. Conformational selection or induced fit? A critical appraisal of the kinetic mechanism. Biochemistry. 51(30): 5894-5902.

104. Koshland, D. E. 1958. Application of a theory of enzyme specificity to protein synthesis. Proc. Natl. Acad. Sci. USA. 44: 98–104.

105. Meireles, L., M. Gur, A. Bakan, and I. Bahar. 2011. Pre-existing soft modes of motion uniquely defined by native contact topology facilitate ligand binding to proteins. Protein Sci. 20(10): 1645-1658.

106. Dhanik, A., and L. E. Kavraki. 2012. Protein-ligand interactions: Computational docking. eLS. John Wiley & Sons, Ltd: Chichester.

107. Abagyan, R., M. Totrov, and D. Kuznetsov. 1994. ICM- a new method for protein modeling and design: Applications to docking and structural prediction form the disordered native conformation. J. Comp. Chem. 15: 488-506.

108. Rarey, M., B. Kramer, T. Lengauer, and G. Klebe. 1996. A fast flexible docking method using an incremental construction algorithm. J. Mol. Biol. 261(3): 470-489.

109. Moustakas, D. T., P. T. Lang, S. Pegg, E. Pettersen, I. D. Kuntz, N. Brooijmans, and R. C. Rizzo. 2006. Development and validation of a modular, extensible docking program: DOCK 5. J. Comput. Aided Mol. Des. 20(10-11): 601-619.

110. Schnecke, V., Swanson, C. A., Getzoff, E. D., Tainer, J. A., and Kuhn, L. A. 1998. Screening a peptidyl database for potential ligands to proteins with side-chain flexibility, Proteins: Struct., Funct., Genet. 33, 74-87.

111. Jones, G., P. Willett, R. Glen, A. Leach, and R. Taylor. 1997. Development and validation of a genetic algorithm for flexible docking. J. Mol. Biol. 267(3): 727-748.

112. Morris, G., D. Goodsell, R. Halliday, R. Huey, W. E. Hart, R. K. Belew, and A. J. Olson. 1998. Automated docking using a Lamarckian genetic algorithm and an emperical binding free energy function. J. Comp. Chem. 19(14): 1639-1662.

113. Zavodszky, M. I., and L. A. Kuhn. 2005. Side-chain flexibility in protein-ligand binding: The minimal rotation hypotheses. Proteins: Struct. Funct. Bioinfo. 57: 243-261.

114. Schnecke, V., and L. A. Kuhn. 1999. Flexibility screening for molecules interacting with proteins. In Rigidity Theory and Applications. Plenum Publishing. New York. pp. 385- 400.

115. Gohlke, H., L. A. Kuhn, and D. A. Case. 2004. Change in protein flexibility upon complex formation: Analysis of Ras-Raf using molecular dynamics and a molecular framework approach. Proteins: Struct. Funct. Bioinf. 56: 322-337.

116. Zavodszky, M. I., M. Lei, A. R. Day, M. F. Thorpe, and L. A. Kuhn. 2004. Modeling correlated main-chain motions in proteins for flexible molecular recognition. Proteins: Struct. Funct. Bioinf. 57: 243-261.

117. Carlson, H. A., K. M. Masukawa, and J. A. McCammon. 1999. Method for including the dynamic fluctuations of a protein in computer-aided drug design. J. Phys. Chem. A. 103: 10213-10219.

118. Broughton, H. B. 2000. A method for including protein flexibility in protein-ligand docking: Improving tools for database mining and virtual screening. J. Mol. Graphics. Model. 18: 247-257.

119. Carlson, H. A. 2002. Protein flexibility and drug design: How to hit a moving target. 6: 447- 452.

120. Claussen, H., C. Buning, M. Rarey, and T. Lengauer. 2001. FlexE: Efficient molecular

docking considering protein structure variations. J. Mol. Biol. 308(2): 377-395.

121. Jacobs, D. J., A. J. Rader, L. A. Kuhn, and M. F. Thorpe. 2001. Protein flexibility predictions using graph theory. Proteins. 44: 150-165.

122. Anderson, A. C., R. H. O'Neil, T. S. Surti, and R. M. Stroud. 2001. Approaches to solving the rigid receptor problem by identifying a minimal set of flexible residues during ligand docking. Chem. Biol. 8: 445-457.

123. Heifetsa, A., and R. H. Liliena. LigAlign: Flexible ligand-based active site alignment and analysis. J. Mol. Graph. Model. Graphics. 29(1): 93-101.

124. Ebalunode, J. O., Z. Ouyang, J. Liang, and W. Zheng. 2008. Novel approach to structure- based pharacophore search using computational geometry and shape matching techniques. J. Chem. Inf. Model. 48: 889-901.

125. Steindl, T. M., D. Schuster, C. Laggner, and T. Langer. 2006. Parallel screening: A novel concept in pharmacophore modeling and virtual screening. J. Chem. Inf. Model. 46: 2146-2157.

126. Lichtarge, O., and M. E. Sowa. 2002. Evolutionary predictions of binding surfaces and interactions. 12: 21-27.

127. Campbell, S. J., N. D. Gold, R. M. Jackson, and D. R. Westhead. 2003. Ligand docking: Functional site location, similarity and docking. Curr. Opin. Struc. Biol. 13: 389-395.

128. Di Gennaro, J. A., N. Siew, B. T. Hoffman, L. Zhang, J. Skolnick, L. I. Neilson, and J. S. Fetrow. 2001. Enhanced functional annotation of protein sequences via the use of structural descriptors. J. Struct. Biol. 134: 232-245.

129. Rantanen, W., K. A. Denessiouk, M. Gyllenberg, T. Koski, M. S. Johnson. 2001. A fragment library based on Gaussian mixtures predicting favorable molecular interactions. J. Mol. Biol. 313: 197-214.

130. Zhao, S., G. M. Morris, A. J. Olson, and D. S. Goodsell. 2001. Recognition templates for prediction adenylate-binding sites in proteins. J. Mol. Biol. 314: 1245-1255.

131. Schmitt, S., D. Kuhn, and G. Klebe. 2002. A new method to detect related function among proteins independent of sequence and fold homology. J. Mol. Biol. 323: 387-406.

132. Kinoshita, K., J. Furui, and H. Nakamura. 2001. Identification of protein functions from a molecular surface database, eF-site. J. Struct. Funct. Genomics. 2: 9-22.

133. Sazanov, L. A. And P. Hinchliffe. 2006. Structure of the hydrophilic domain of respiratory complex I from Thermus thermophilus. Science 311: 1430-1436.

134. Tang, X., C. Yang, Y. Gu., C. Song, X. Zhang, Y. Wang, J. Zhang, C. L. Hew, S. Li, N. Xia, and J. Sivaraman. 2011. Structural basis for the neutralization and genotype specificity of hepatitis E virus. Proc. Natl. Acad. Sci. USA. 108(25): 10266-10271.

135. Sun, F., X. Huo, Y. Zhai, A. Wang, J. Xu, D. Su, M. Bartlam, and Z. Rao. 2005. Crystal structure of mitochondrial respiratory membrane protein complex II. Cell. 121: 1043- 1057.

136. Berry, E. A., L. S. Huang, and N. Minagawa. 2010. Ascochlorin is a novel, specific inhibitor of the mitochondrial cytochrome bc(1) complex. Biochim. Biophys. Acta. 1797: 360-370.

137. Takano, T. And R. E. Dickerson. 1980. Redox conformational changes in refined tuna cytochrome c. Proc. Natl. Acad. Sci. USA. 77(11): 6371-6375.

138. Shinzawa-Itoh, K., H. Aoyama, K. Muramoto, H. Terada, T. Kurauchi, Y. Tadehara, A. Yamasaki, T. Sugimura, S. Kurono, K. Tsujimoto, T. Mizushima, E. Yamashita, T. Tsukihara, and S. Yoshikawa. 2007. Structures and physiological roles of 13 integral lipids of bovine heart cytochrome c oxidase. EMBO J. 26. 1713-1725.

139. Watt, I. N., M. G. Montgomery, M. J. Runswick, A. G. W. Leslie, and J. E. Walker. 2010. Bioenergetic cost of making an adenosine triphosphate molecule in animal mitochondria. Proc. Natl. Acad. Sci. USA. 107: 16823-16827.

CHAPTER 2: From static structure to living protein: Computational analysis of cytochrome c oxidase main-chain flexibility

Reprint (adapted) with permission from Buhrow, L., S. Ferguson-Miller, and L.A. Kuhn. 2012. From static structure to living protein: Computational analysis of cytochrome c oxidase main- chain flexibility. Biophys. J. 102: 2158-2166. Copyright 2012 Elsevier B. V.

Introduction

Cytochrome c oxidase is responsible for reducing molecular oxygen and coupling the released energy to pumping protons across its associated membrane [1-3]. Structural and mutational studies of Rhodobacter sphaeroides (Rs) CcO have defined two conserved proton pathways, the K- and D-pathways [4-9]; while a bifurcated hydrophobic channel leading from the oxygen-rich membrane environment to the active site in bacterial and mammalian CcO has been defined based on the absence of water molecules in the interior channel and co- crystallization with high-pressure xenon gas (Figure 2.1) [5, 10-11]. There is little agreement on the detailed mechanism of coupling of these oxygen reduction and proton pumping functions of

CcO.

Hydrogen/deuterium exchange and x-ray crystal structures of CcO in distinct redox states have identified regions that undergo conformational change during the reaction cycle, including rotation of the heme a3 porphyrin ring and farnesyl tail and movement of helix VIII, which contains critical K-pathway residues [12]. The redox state of RsCcO proteins during x-ray beam exposure has been microspectrophotometrically monitored. Results suggest that the initial protein conformation (whether oxidized or reduced) is apparently stabilized by crystal contacts and maintained despite reduction of the CcO metal centers during data collection [13].

Structural and solvent accessibility changes support alternating proton pathway accessibility to the active site, with the K-pathway open during metal reduction only and the D-pathway open in all the other reaction intermediates [12-15].

Figure 2.1: Architecture of the two-subunit structure essential for RsCcO activity. Subunit I, in green ribbons, includes 12 transmembrane α-helices, heme and metal cofactors, and the proton and oxygen pathways. Subunit II, in blue ribbons, is composed of two transmembrane α- helices, the β-sheet domain, and the CuA site (PDB ID 2GSM [7]). (A) Electrons are sequentially passed from cytochrome c, positioned based on [54], through the bimetallic CuA and heme a to the active site heme a3 and CuB. The D-proton pathway contains a chain of well resolved water molecules shown in red spheres, while the K-pathway contains two bound water molecules. (B) A hydrophobic channel, shown as dashed lines, leads from the oxygen-rich membrane environment to the heme a3/CuB center.

Previous studies of the coupling of oxygen reduction and proton pumping are limited in availablility and scope. Crystallography provides structural snapshots constrained by intermolecular crystal contacts, rather than a continuum of conformations. Deuterium exchange mass spectrometry provides data for a limited subset of the protein fragments. Computational methods including molecular dynamics, quantum mechanics/molecular mechanics, and electrostatic analyses have predicted continuous hydrogen-bonding chains within the K-pathway, identified potential proton exit pathways, and evaluated the effects of hydration and redox state 42 on the protonation of critical residues [16-23]. However, these studies are limited to local motions that occur on the nanosecond timescale.

In this study, main-chain conformational change, coupling between mobile regions, and the functional implications of such motions are studied by computational analysis of CcO flexibility and low energy vibrational motions using ProFlex and elNémo. Therefore, this effort focuses on aspects of CcO conformation change shared by related structures, for instance, CcO structures in different redox states, resolved by different research groups, and derived from related organisms. ProFlex has successfully predicted flexibility and protein coupling for ligand- bound and ligand-free protein structures [24-25], while elNémo normal modes analysis has described the main-chain motions of a number of integral membrane proteins [26-29]. The conformational changes predicted in CcO, and their influence on function, are considered in the context of available data.

Materials and Methods

Identification of flexible regions using ProFlex

In ProFlex [24], the input protein structure is represented as a three-dimensional molecular framework with atoms represented as vertices and bonds as edges. Graph-theoretic analysis, based on constraint counting [30], decomposes a protein structure into rigid, isostatic and flexible regions based on the number of internal degrees of freedom, offset by the number of covalent and non-covalent bonds and bond-coordination angle constraints [24]. Protein regions linked by more bonds than needed to form a stable structure are considered rigid. Regions with the same number of bond constraints and bond-rotational degrees of freedom are defined as isostatic (barely rigid), whereas regions with fewer constraints than bond-rotational degrees of freedom are identified as flexible. This level of flexibility is represented as an energy-dependent hydrogen dilution profile (Figure 2.2).

For ProFlex, the three-dimensional atomic coordinates corresponding to one biological unit were selected from the highest-resolution bacterial structures (Table 2.1): 2- and 4-subunit

RsCcO, 2-subunit Paracoccus denitrificans (Pd) CcO, and the 3-subunit Thermus thermophilus

(Tt) CcO structure, which is structurally equivalent to the 2-subunit RsCcO. Oxidized and reduced 2-subunit RsCcO structures were produced by the same research group (Table 2.2).

ProFlex analysis predicts intrinsic protein flexibility and has shown little sensitivity to small changes in analyzed structures [24], including redox state. This is desirable for the current analysis, which focuses on global conformational change and coupling. Analyzed structures of

CcO include two heme groups, three copper ions and calcium and magnesium, where appropriate. Bound water molecules that were buried were included in the ProFlex analysis,

44 based on identification by PRO_ACT [34] and Consolv [35]. Protons and partial charges were assigned for salt bridge identification using WhatIf (version 6 [36]) for protein and water atoms and InsightII with CVFF parameters for heme groups (Accelrys, Inc., San Diego, CA).

Figure 2.2: Dependence of CcO flexibility on thermal energy increase, analyzed by ProFlex hydrogen bond dilution profiles. These profiles indicate the rigid and flexible regions of the protein backbone by residue number along the x-axis, with change in energy along the y-axis. At the top of the y-axis, the energy is 0 kcal/mol, descending to increasingly negative values (corresponding to thermal energy), at which only the strongest hydrogen bonds and salt bridges remain. Flexible regions appear as thin gray lines along the x-axis, while each mutually rigid region of main chain appears as a colored bar; different colors represent independently rigid regions. For each energy at which the protein becomes more flexible, a new line (representing the new rigid decomposition) is drawn in the profile. Here, the ProFlex hydrogen bond dilution profile of the reduced RsCcO crystal structure (PDB entry 3FYE [12]) is depicted. ProFlex comparisons between different bacterial structures in the Results are based upon selecting the thermal energy for each at which the second largest rigid cluster (subunit II helix II shown in yellow bars) is maximal in size (number of atoms). Other independently rigid clusters, depicted in green and purple, are located on internal or exterial protein loops.

Table 2.1: CcO structures analyzed by ProFlex and elNémo. High resolution two and four subunit structures from R. sphaeroides, P. denitrificans, and T. thermophilus were selected for flexibility and low energy motion analysis using ProFlex and elNémo. For the R. sphaeroides structures, one biological unit was selected from each unit cell, with chain identifiers A and B for two-subunit structures or A-D for four-subunit structures. The P. denitrificans structure was studied in the absence of the co-crystallized antibody Fv fragment, using subunits I and II (chain identifiers A and B). The three-subunit T. thermophilus structure (chains A, B, and C), analogous to the RsCcO and PdCcO two-subunit structures, was also studied. All analyzed structures are in the oxidized state except for the two-subunit reduced RsCcO structure.

PDB Entry Description Source Resolution (Å) R-Factor (R-Free) Two-subunit oxidized 2GSM [7] R. sphaeroides 2.00 0.214 (0.232) aa3 oxidase 3FYE [12] Two-subunit reduced R. sphaeroides 2.15 0.196 (0.221) aa3 oxidase Four-subunit 1M56 [5] R. sphaeroides 2.30 / 2.69 0.236 (0.275) aa3 oxidase Two-subunit with 1AR1 [46] P. denitrificans 2.70 0.207 (0.261) antibody Fv fragment Recombinant 1XME [47] T. thermophilus 2.30 0.217 (0.236) Cytochrome ba3 oxidase

Table 2.2: Comparison of oxidized and reduced two-subunit RsCcO structures. The oxidized and reduced structures of RsCcO are substantially similar, as they were solved using the same method by the same research group [7, 12]. The generation of reduced crystals was accomplished by soaking oxidized crystals in a stabilizing solution, which included 10 mM sodium dithionite, prior to flash cooling and data collection. The reduced structure shows significant displacement of the heme a3 porphyrin ring and farnesyl tail, movement of subunit I helix VIII, which contains critical K-pathway residues, and the resolution of additional K- pathway water molecules [12]. Analyses of these structures using ProFlex and elNémo methods are considerably similar despite the slight loss of resolution for the reduced structure.

Oxidized [7] Dithionite Reduced [12] Space Group P 212121 P 212121 Cell Dimensions a=125.0 b=131.6 c=176.8 a=124.6 b=131.5 c=176.2 Molecules per Assymetric Unit 2 2 Resolution Range (Å) 20-2.0 50-2.15 Completeness (%) 96.1 99.1 Number of Unique Reflections 184,839 151,414 Redundancy 4.5 4.6 6.0 6.5 Rmerge (%) I/σ 14.8 20.8 Number of Refined Atoms 13,648 13,615 R-Factor/R-Free (%) 21.4/23.2 19.6/22.1 Average B-Factor 36.5 45.8 RMSD Bond Length (Å) 0.013 0.012 RMSD Bond Angle (°) 1.253 1.227 RMSD (Å) Between Structures 0.219 (1429 atoms aligned to 1429 atoms)

Identification of low energy motions by normal mode analysis using elNémo

Collective protein motions have been studied since the early 1980s using normal mode analysis [37]. This approach decouples protein motions about an initial, low energy conformation into oscillating orthogonal modes that range from low-frequency, global motions

(low numbered modes) to high-frequency, localized motions (high numbered modes). elNémo normal mode analysis [38] was applied to sample the low-frequency global motions in CcO.

47 elNémo places a harmonic potential, or spring, between each pair of non-hydrogen atoms within

8 Å, then solves for the frequency and direction of maximal motion in each mode. CcO, with over 6,300 protein and cofactor atoms, requires a rotational-translational-block representation.

Two consecutive amino acid residues form a block, and blocks are then coupled by springs. This representation has been shown to have little effect on the resulting low frequency motions [39].

High-resolution bacterial CcO structures were analyzed by elNémo (Table 2.1). For proper interpretation by elNémo, the heme, copper, and magnesium Protein Data Bank (PDB) coordinates were renamed as ATOM records, and heme groups were assigned unique chain identifiers. Bound water molecules were not included, as they do not affect the accuracy of results [40-41].

The modes output by elNémo include six trivial zero frequency modes, 1-6, corresponding to rigid-body protein rotation and translation. This work is focused on the internal motions, starting with mode 7. The displacement in the direction of maximal motion was identified by elNémo, and 11 structural snapshots (centered on the input crystal structure) were output to represent the dynamics trajectory for each mode. The normalized mean-squared

2 displacement value for each Cα , , was graphed as a function of residue number (Figure 2.3).

Significantly mobile and immobile regions were defined as three or more consecutive Cα’s with

values at least 2 standard deviations above or below the mean for all residues, respectively.

Figure 2.3: Mean squared displacement values within RsCcO transmembrane helices (following page). Cα motions in elNémo (A) mode 7, (B) mode 8, and (C) mode 9 are graphed as a function of helix residue numbers (PDB entry 2GSM [7]), with helices individually colored and labeled. The stationary point in each helix, defined as the residue(s) undergoing least displacement, appears as a minimum in the values. Large displacements of helix termini reflect tilting, bending and compression motions.

Figure 2.3 (cont’d)

Evolutionary conservation

RsCcO amino acid conservation was measured and mapped onto the structure using

ConSurf [42]. Multiple sequence alignments were constructed by searching the SwissProt database [43] for all protein sequences with 30-95% sequence identity to RsCcO subunits I and

II, using five iterations of PSI-BLAST with an E-value cutoff of 0.001. The resulting sequences

(254 subunit I and 148 subunit II sequences) were aligned using CLUSTALW [44]. The ConSurf neighbor joining method was used to construct a phylogenetic tree of the sequences such that only one representative of each clade was included, to keep clusters of similar sequences from dominating the conservation scores.

Cavity identification

RsCcO internal protein cavities were identified by Hollow [45], which places probe atoms in the protein interior on a 0.5 Å grid at points not occupied by protein atoms. A molecular surface tangent to the probe atom van der Waals spheres was used to visualize the cavities. PyMOL [Schrödinger, Inc.] was used to render all structural figures.

Results and Discussion

Global motions in CcO

Unlike computational methods previously applied to CcO, ProFlex and elNémo probe regions of global flexibility, intra-protein coupling, and main-chain motions. The elNémo analysis simulates low-energy main-chain protein motions, while ProFlex identifies underconstrained (coupled flexible) regions and rigid regions connected by hinges that allow motion relative to other parts of the protein. Two-subunit RsCcO structures were selected for analysis, as these structures are of higher, isotropic resolution. The loss of two accessory subunits had little effect on the flexibility of CcO's catalytic core (Figure 2.4).

Figure 2.4: ProFlex flexibility comparison the of two and four subunit crystal structures of RsCcO. (A) The four-subunit structure of RsCcO (PDB entry 1M56 [5]) was analyzed with ProFlex to evaluate the influence of subunits III-IV on the catalytic core flexibility (subunits I- II). (B) When subunits III-IV are removed from the four-subunit structure to analyze subunits I- II alone, the catalytic core is observed to have very similar flexible regions to those seen in the presence of subunits III-IV (panel A). An exception is the short N-terminal helix of subunit I appears rigid in the four-subunit structure (colored blue as part of the core rigid region in panel A), whereas in the structure with subunits III and IV removed (B), it becomes flexible.

The degree of flexibility and coupling within CcO is highlighted by ProFlex analysis

(Figure 2.5) of oxidized and reduced RsCcO (PDB entries 2GSM [7] and 3FYE [12]) and the closest homolog PdCcO (PDB entry 1AR1 [46]). At the initial hydrogen-bond energy (Figure

2.2), all three structures have a stable core consisting of helices I-VI, X-XII, and the subunit II β- sheet domain. The middle of helices VII-IX and helix II of subunit II, which surround the K- pathway, are flexible (Figure 2.5), as are the exterior surface where cytochrome c interacts and the C-terminus of subunit I. At a higher but still moderate energy with a less-dense hydrogen bond network (hydrogen bonds at least -4.2 kcal/mol in strength for RsCcO), all helices except those packed against the heme groups (II, VI, and X) attained significant flexibility, while remaining coupled (Figure 2.6). At significantly higher energy (H-bonds stronger than -7.6 kcal/mol in RsCcO; Figure 2.6), RsCcO became a coupled flexible structure surrounding the heme groups. These results indicate that most regions of CcO, except for the helices surrounding the K-pathway (VII and VIII) and residues coordinating heme groups, behave as a coupled structure that is either mutually flexible or rigid, depending on the energy.

Figure 2.5: ProFlex prediction of main-chain flexibility and stability in RsCcO. (A) Oxidized RsCcO [7], (B) reduced RsCcO [12], and (C) PdCcO [46] were analyzed. Deep blue indicates regions of greatest stability, with a gradient to light blue (somewhat rigid) to gray (isostatic) to orange (flexible) to red (most flexible). The stable proteins core consists of transmembrane helices I-VI, including the D-pathway, and X-XII. Most of the subunit II β-sheet domain is also internally rigid. The greatest intramembrane flexibility occurs within helices VII-VIII and subunit II helix II surrounding the K-pathway, while the cytochrome c interaction site and subunit I C-terminus exhibit the highest overall flexibility.

Figure 2.6: Thermal denaturation of the RsCcO structure by ProFlex (following page). As the energy increases from top to bottom in these panels (with the right view in each pair rotated by 180º about the vertical axis), ProFlex recalculates the internal flexibility. (A and B) At the initial analyzed energy (-2.938 kcal/mol), RsCcO (PDB entry 2GSM [7]) is composed of a rigid core of helices I-VI, X-XII and the subunit II β-sheet domain. Helices VII-VIII surrounding the K-pathway show the greatest transmembrane flexibility, with central hinges connecting stable helical ends which can undergo rigid-body motion. The cytochrome c interaction site and subunit I and II C-termini are the most flexible regions outside the membrane. (C and D) At higher but still moderate energy (-4.248 kcal/mol), all helices except for those nearest the heme groups (II, VI, and X) attain significant flexibility, while the β-sheet domain remains barely stable. (E and F) At the highest sampled energy (-7.645 kcal/mol), RsCcO becomes a flexibly coupled structure surrounding the heme groups.

Figure 2.6 (cont’d)

elNémo was used to predict low-energy motions about the equilibrium conformation of

CcO (Table 2.1). The three lowest frequency protein motions, modes 7-9, were selected as representing global motions as at least 30% of the residues experienced significant displacement

(PDB entry 2GSM [7]). Mode 7 exhibited counter-rotational twisting on the interior and exterior

1 of the membrane (Figure 2.7A; Animation 1 ) involving helical tilting and bending (Figure 2.3).

The protein located near the inside of the membrane rotated about an axis perpendicular to the membrane plane, whereas on the outside of the membrane the protein rotated about an axis tilted by ~30º, parallel to the subunit II β-strands. Mode 8 involved counter-rotational motions of CcO on each side of the membrane about an axis perpendicular to the membrane plane (Figure 2.7B;

Animation 2). Mode 9 showed a C-clamp-like motion of subunits I and II, with the pontoon helix approaching the interior C-terminus of subunit I near helices I, III, and XII, then receding

(Figure 2.7C; Animation 3). Mechanical twisting motions of integral membrane proteins on either side of the membrane previously have been described in potassium, mechanosensitive, and nicotinic acetylcholine receptor proteins [26-29, 48].

1 Animations 1-3 are available on the Biophysical Journal website. Animations: elNémo low- energy motions about the equilibrium conformation of CcO (PDB entry 2GSM [7F]). Subunits I and II are depicted as green and cyan ribbons, respectively, while heme cofactors are represented as magenta tubes. Regions of the protein are highlighted. The highly flexible subunit I C-terminus is colored red, the pontoon helix is colored orange, the interfacially coupled regions of subunits I and II (helix VII-VIII loop, the ends of helix IX and subunit II helices I-II, and the subunit II β-strand) are colored yellow, the subunit I helix III-IV loop is in dark cyan, the K-pathway is colored blue with critical residues shown in sticks, and the D- pathway is represented in dark purple with critical residues shown in stick diagram. 57

Figure 2.7: elNémo simulation of low frequency motions in RsCcO. Low-energy modes of internal RsCcO motion (PDB entry 2GSM [7]), are shown by superposition of the two extreme conformations in each mode, with subunit I in blue and subunit II in orange. Arrows indicate motion of Cα’s undergoing a displacement of at least 4Å. (A) Counter-rotational motions on the membrane interior relative to the exterior were seen in mode 7. On the inside of the membrane, RsCcO helices rotated about an axis perpendicular to the membrane, while on the outside, the rotation axis was 30º from the membrane plane, parallel to the β-strands. (B) Mode 8 involved counter-rotational motion about an axis perpendicular to the membrane. (C) Mode 9 showed clamp-like compression and expansion, drawing the protein on the N- and P-sides of the membrane to the membrane space nearest helices I, III, and XII. All three modes showed major rocking of the β-sheet domain while maintaining the subunit I-II interface. Results on PdCcO (PDB entry 1AR1 [46]) and TtCcO (PDB entry 1XME [47]) were substantially similar.

Together the ProFlex and elNémo analyses suggest that CcO is composed of a stable, coupled core with greater flexibility and independence observed in the K-pathway helices. This stable, coupled core may undergo low energy motions, including counter-rotational twisting and compression driven by transmembrane helical tilting and bending motions. With additional energy or active site redox changes, the K-pathway helices may undergo independent motions.

This is supported by their flexible, centrally located hinge residues that allow the helical termini to under greater amplitude motions. These motions, similar to those seen in the crystal structure, are predicted to allow water molecule movement along the K-pathway for proton uptake.

The internal flexibility and low energy motions of CcO were compared to other integral membrane proteins. The β2 adrenergic receptor, KcsA potassium channel, and voltage- dependent anion channel (VDAC) were selected as representing diverse membrane proteins with similar structural resolution to the analyzed CcO proteins and experimental characterization of conformational change (Table 2.3). elNémo results indicate that low energy motions of CcO have a similar immobile protein core, mobile secondary structural ends, and a similar percentage of displaced residues relative to the receptor and channel proteins, though β2 adrenergic receptor shows greater mobility at the lowest energy (Table 2.4; Figure 2.8). ProFlex results suggest that

CcO and the β2 adrenergic receptor are more rigid than the transporter/channel proteins (Figure

2.9), perhaps due to tighter packing within their helical bundles relative to the pore-forming membrane proteins.

Table 2.3: Diverse membrane protein folds analyzed by ProFlex and elNémo. To compare low energy motions of membrane proteins, directly we analyzed a G-protein coupled receptor: the β2 adenergic receptor, the α-helical KcsA potassium channel, and a β-barrel transporter protein: the voltage-dependent anion channel (VDAC). These proteins were selected as they represent diverse integral membrane proteins, have available crystal structures of similar resolution to that of the analyzed CcO structures, and have been studied in terms of conformational change or dynamics required for function.

R-Factor PDB Entry Description Source Resolution (Å) (R-Free) 2RH1 [54] β2 adrenergic receptor Homo sapiens 2.4 0.20 (0.23) 1K4C [55] Potassium channel Streptomyces lividans 2.0 0.22 (0.23) KcsA 3EMN [56] Voltage-dependent Mus musculus 2.3 0.24 (0.28) anion channel

Table 2.4: elNémo percentage of atoms significantly displaced and relative frequencies of normal modes. The low energy motions generated by elNémo are given only as relative amplitudes in the direction of maximum displacement for a given frequency or normal mode. It is possible to compare the percentage of atoms with significant motion in a given protein at a given frequency, called the “collectivity” [38]. Values approaching 1.0 signify global motion of the entire protein, while values approaching 0.0 signify localized motion. In the two lowest energy modes, modes 7 and 8, the percentage of atoms displaced in CcO is consistent with those of the KscA potassium channel and VDAC with about 50% of atoms being significantly displaced. The additional mobility of β2 adrenergic receptor (71% of atoms displaced in the mode 7) may be due to the protein sampling of active protein conformations in the absence of agonists [11].

Protein Mode 7 Mode 8 Mode 9 RsCcO Displaced atoms (%) 47 47 38 Relative frequency 1.00 1.04 1.08 β2 adrenergic Displaced atoms (%) 71 58 68 receptor Relative frequency 1.00 1.25 1.46 Potassium channel Displaced atoms (%) 50 68 68 KcsA Relative frequency 1.00 1.44 1.44 Voltage-dependent Displaced atoms (%) 46 53 67 anion channel Relative frequency 1.00 1.09 1.28

Figure 2.8: Relative flexibility of membrane proteins assessed using ProFlex. (A) RsCcO,

(B) β2 adrenergic receptor, (C) KcsA potassium channel, and (D) voltage-dependent anion transporter were analyzed using ProFlex (Structural details can be found in Table 2.3). The proteins are represented as ribbon diagrams and colored by degree of rigidity, from blue for highly rigid regions, to gray for barely rigid regions, to red for highly flexible regions.

Figure 2.9: RsCcO, β2 adrenergic receptor, KcsA potassium channel, and VDAC residue displacement (following page). Displacements of residue in modes (A, E, I, M) 7, (B, F, J, N) 8, and (C, G, K, O) 9 are colored spectrally where blue is the most stationary and red is the most displaced. To compare these results to CcO displacements, the most mobile and least mobile residues in the three lowest energy modes were defined (D, H, L, P). Three or more consecutive residues undergoing displacement of one standard deviation above or below the average displacement in the three lowest energy modes are colored orange and blue, respectively, while three or more consecutive residues undergoing displacements greater than two standard deviations above the average are colored red.

Figure 2.9 (cont’d)

Coupled local motions

Local motions underlying the global conformational changes include an inter-domain hinge between subunits I and II and coupled helical deformation. ProFlex results for the oxidized and reduced RsCcO structures (Figure 2.5) indicate the greatest transmembrane flexibility is located in the central regions of helices VII-VIII. Additionally, the secondary structures at the interface of subunit I and the subunit II β-sheet domain are flexibly coupled.

This region includes the helix VII-VIII loop, the ends of helix IX and subunit II helices I-II, and the interfacial subunit II β-strand (residues Gly225-Val231) (Figure 2.1). This flexibly coupled region acts like an inter-domain hinge which allows the coordinated counter rotational movements of subunit I together with subunit II β-sheet domain rocking as predicted by elNémo

(Figure 2.7 and Animations 1-3). In elNémo mode 7, helices I, IX, and X maintain an interface with the β-sheet domain, while conformational changing (Table 2.5 and Animation 1). In modes

8 and 9, subunit II helices I-II bend and straighten significantly in concert with the β-sheet domain rocking (Table 2.5 and Animations 2-3). Since the cytochrome c binding site involves one face of this domain (Figure 2.1), these motions could couple with cytochrome c binding or affect its affinity for the site. Inter-domain hinging while maintaining an interface has been observed previously in normal mode analysis of nicotinic receptors [48].

Table 2.5: CcO helical conformational changes in elNémo low-energy modes. Relative magnitudes of helix tilting, elongation, and internal conformational change were measured by comparing the two extreme structural snapshots from the trajectory for each of the three lowest- frequency elNémo modes. In order to measure helical tilt angles, each helix was represented by two line segments connecting the Cα at each helix end with the central most stationary residue in the helix. Stationary residues and mean displacement are defined as in the elNémo section of the Methods. Helical tilting was measured by the minimum angle change of the helix on either side of the membrane, to offset the additional angular effect of bending in some helices. Internal conformational changes were measured by calculating the root mean squared displacement (RMSD) of the helix Cα’s between the extreme conformations in each mode. Helical length change was determined as the difference in distance between the helix terminal Cα’s between the extreme conformations. Measurements deviating by more than one standard deviation above or below the mean value for all helices appear in boldface are indicated by “^” or “v”, respectively, whereas “^^” or “vv” indicates two standard deviations above/below the mean

Table 2.5A Stationary Helical Conformational Length Mode Helix 2 Residue(s) Tilt (º) Change (RMSD, Å) Change (Å) -3 II S44-V45 3.4x10 2.45 1.042^ 0.740 -3 III H127I 2.2x10 5.51 0.406 0.170 -3 IIII L152I 2.3x10 5.06 0.316V 0.520 -3 IVI I191I 1.3x10 5.44 0.274V 0.250 -3 VI I250I 1.0x10 4.87 0.436 0.610 -4 VII A294I 6.0x10 2.78 0.676 0.14V -3 VIII V329I 2.4x10 1.73 0.357 1.000 7 -3 VIIII T352I-A356I 3.6x10 4.58 0.468 0.00V -4 IXI F391I 7.0x10 1.13V 0.454 0.200 -3 XI A427I-V428I 1.2x10 5.41 0.421 0.840 -3 XII G463I 1.6x10 3.21 0.999 2.64^ -3 XIII L503I 2.1x10 5.46 1.37^^ 2.98^ -3 III I68II 1.4x10 6.09^ 0.25V 2.7^ -3 IIII I111II 3.4x10 3.97 0.25V 2.44^

Table 2.5 (cont’d)

Table 2.5B Stationary Helical Conformational Length Mode Helix 2 Residue(s) Tilt (º) Change (RMSD, Å) Change (Å) -3 II S44I- V45I 3.4x10 4.44 0.822 2.47^ -3 III G101I & H127I 2.2x10 8.31^^ 0.85 0.970 -3 IIII L152I 2.3x10 5.94^ 0.794 0.15V -3 IVI I191I 1.3x10 3.28 1.055^ 2.22^ -3 VI I250I 1.0x10 0.26V 0.996 1.500 -4 VII Y275I & A294I 6.0x10 3.16 0.779 0.760 -3 VIII V329I 2.4x10 1.24V 0.682 1.640 8 -3 VIIII A351I - A356I 3.6x10 9.56^^ 0.878 2.95^ -4 IXI F391I 7.0x10 7.28^ 0.317V 0.910 -3 XI A427I-V428I 1.2x10 1.12V 0.321V 0.16V -3 XII G463I 1.6x10 3.68 0.325V 0.280 -3 XIII L503I 2.1x10 5.35 0.484 0.03V -3 III I68II 1.4x10 4.69 1.333^ 1.010 -3 IIII I111II 3.4x10 5.26 1.229^ 1.630

Table 2.5 (cont’d)

Table 2.5C Stationary Helical Conformational Length Change Mode Helix 2 Residue(s) Tilt (º) Change (RMSD, Å) (Å) -3 II G37I 7.5x10 1.67 0.687 1.130 -3 III L11I- F116I 2.3x10 3.49 0.777 0.970 -3 IIII S142I 7.0x10 2.77 0.703 4.26^^^ -3 IVI I209I 1.8x10 3.33 0.540 2.22^ -3 VI I232I 1.4x10 2.09 0.468 1.500 -4 VII P293I 6.0x10 2.07 0.503 0.760 -3 VIII M320I-V321I 4.5x10 1.27V 0.504 1.660 9 -3 VIIII T359I & K362I 5.6x10 2.26 0.786 1.460 -3 IXI W383I 1.4x10 1.01V 0.523 0.06V -3 XI A434I 1.3x10 3.10 0.418 0.16V -4 XII K454I 8.0x10 2.40 0.549 1.050 -3 XIII E488I 5.0x10 0.00V 0.874 0.03V -3 III L62II 6.1x10 2.99 1.352^^ 0.5 -3 IIII Q127II 6.2x10 0.00V 1.42^^ 1.63

Within the lipid bilayer, a relatively stationary set of RsCcO residues is found near the mid-plane in elNémo modes 7 and 8 (Figure 2.10), whereas in mode 9, a second set of stationary residues is found near the inside face of the membrane. This is similar to mechanosensitive channels, in which a pivot plane runs through the plane of the membrane and key channel gating residues [26]. One potential consequence of the tilting and internal conformational change of the

RsCcO helices about these pivots is a change in distance across the membrane, as observed for mechanosensitive channels hypothesized to use a “twist-to-open” function [26].

Figure 2.10: Planes of relatively stationary residues in the CcO transmembrane helices. RsCcO (PDB code 2GSM [7]) subunits I and II are colored green and cyan, respectively. Heme cofactors are represented by magenta tubes, and copper A, magnesium, and copper B appear as copper or lime colored spheres. The most stationary helical residues, based on elNémo normalized mean squared displacement () values (Table 2.5), appear in (A) orange for modes 7 and 8 (in which overlapping sets of residues were stationary) and (B) red for mode 9.

The subunit II β-sheet domain is identified as a single, internally rigid domain by the

ProFlex method, while low energy motions of this region support that its rocking behavior is essentially rigid-body, based on the ~1 Å main-chain root mean standard deviation (RMSD) following least-squares superposition of this domain between the two extreme conformations in each mode. Experimental analysis of movement for the subunit II β-sheet domain is limited.

Crystal structure comparisons provide little information, as this domain is involved in strong crystal contacts in most RsCcO structures (Figure 2.11C) [7, 12]. However, deuterium exchange analysis supports the flexibility of the subunit I/β-sheet domain interface, with the subunit II

Gly225-Val231 β-strand found to be differentially solvent accessible [14]. Two-dimensional surface-enhanced infrared absorption spectroscopy experiments on membrane bound RsCcO indicate a possible change in association of α-helices and β-strands upon protein activation [47].

In order to better understand the mechanistic implications of the movements that Proflex and elNémo predict, these global motions are considered in the context of the functional regions of

CcO.

Figure 2.11: Comparison of RsCcO normal modes and crystallographic temperature factors. (A) Segments undergoing the most 2 (red) or least (blue) displacement in modes 7-9, based on having three consecutive Cα’s with a displacement, , two standard deviations above or below the mean value from all three modes. (B) Crystallographic temperature factors (B-values) of reduced RsCcO (PDB entry 3FYE [12]), measuring fluctuation in atomic position. (C) Crystal contacts of subunit II with neighboring molecules. One molecule of RsCcO is represented in dark gray (subunit I) and medium gray (subunit II). Neighboring molecules appear in light gray.

Protein flexibility along the K-pathway suggests a proton conducting mechanism involving water molecule movement

The CcO K-pathway is responsible for supplying protons to the active site during metal center reduction, where they are consumed to make water. Mutagenesis and x-ray crystallographic studies have identified functionally important residues and two stable water molecule binding sites along the K-pathway [5, 7, 12-13, 50]. These water molecule sites

(HOH6516 and HOH6533 in PDB entry 2GSM [7]), are located between conserved Lys362 and

Ser299, and Tyr288 and Thr359, respectively (Figure 2.1). A continuous hydrogen-bond network for proton transfer from the K-pathway entry to the active site is not apparent in most crystal structures, though it is predicted by molecular dynamics calculations, which show concurrent motion of Thr359 [18].

According to ProFlex, the most flexible intra-membrane region in oxidized and reduced

RsCcO and in PdCcO involves the helices composing the K-pathway, VII-VIII and subunit II helix II, which could move independently of the rigid core of subunit I (Figure 2.5).

Interestingly, in the same region, the low energy motions in elNémo mode 8 suggest the repositioning of water molecules along the K-pathway. These motions alternately produce and occlude two small cavities (Figure 2.12), which correspond to the well-resolved K-pathway water molecule positions, even though this method does not consider water molecules in the input structure.

ProFlex and elNémo analyses of intrinsic protein flexibility and low-energy protein motions are consistent with the crystallographic and deuterium accessibility data involving the

K-pathway. Specifically, crystallographic analysis shows movement of heme a3 in the reduced

70 form, affecting the conformation of the central region of helix VIII and allowing new water molecules to be resolved in the active-site entrance [12]. Deuterium accessibility studies also support conformational change in helix VIII upon reduction, as residues Val354 -Trp366 rapidly incorporate deuterium during metal reduction [14].

Figure 2.12: Water molecule movement in the K-pathway. K-pathway residues, in green tubes, and associated water molecules, in red spheres, lead to the active site heme a3, in magenta tubes and CuB depicted as a gold sphere (PDB entry 2GSM [7]). Two molecular surfaces are shown in gold and blue mesh for the two extreme conformations of elNémo mode 8. The blue cavity (coincident with crystallographic Wat6515) appears in one extreme conformation from elNémo, while the gold cavity (coincident with Wat6533) appears in the other, suggesting that conformational change is associated with water molecule repositioning.

The D-pathway remains rigid but may be regulated by the flexible subunit I C-terminus

The D-pathway of CcO conducts protons from the inside of the membrane to both the active site for oxygen reduction chemistry and the exterior for pumping. In contrast to the K- pathway, the D-pathway contains a conserved hydrogen-bonded chain of water molecules [5, 7,

12-13]. Both ProFlex and elNémo analyses indicate relative rigidity along helices II-V composing the D-pathway (Figures 2.5 and 2.7). Only at energies at which the enzyme would partly unfold did these regions become flexible or flexibly linked to the rest of subunit I in

ProFlex (Figure 2.6). Regions surrounding the D-pathway were substantially immobile in elNémo normal modes, based on at least three consecutive residues having displacements at least two standard deviations below the average for all residues (Figure 2.11A). Rigidity of the D- pathway residues and associated water molecules is supported by low crystallographic temperature factor (B) values (Figure 2.11B) [7, 12], absence of structural changes between the oxidized and reduced structures [12-13], and lack of deuterium exchange in D-pathway peptide fragments [14].

In contrast to the intra-membrane portion of the D-pathway, the solvent-exposed subunit I

C-terminus was determined to be highly flexible, based on bond-rotational degrees of freedom identified by ProFlex (Figure 2) and displacement of C-terminal residues Thr520 - Thr550 by at least two standard deviations above the average residue in elNémo normal modes (Figure 2.11A).

Crystallographic B-values also identify this as one of the most mobile regions of RsCcO (Figure

2.11B), with residues beyond Thr550 not usually resolved [7, 12-13]. While C-terminal mobility is anticipated due to its irregular structure, this region could be essential for regulating the D- pathway. In reaction intermediates with the D-pathway hypothesized to be fully functional

(oxidized, peroxy, and ferryl), the peptide fragment from this region incorporates ~80%

72 deuterium, whereas incorporation is reduced to ~50% in intermediates with an inactive D- pathway [14]. This suggests that C-terminus flexibility is related to D-pathway activity, possibly by recruiting protons via its acidic residues (Glu552, Asp558 , Asp559 , and Glu561).

Oxygen supply to the CcO active site may be gated

In RsCcO, a branching hydrophobic channel leading from the membrane environment to the protein's active site has been identified based on solvent accessible surfaces within crystal structures and co-crystallization with xenon gas, which has similar hydrophobicity and size as oxygen [5, 10-11]. This bifurcated channel has two entries [5], one between helices I, II, and III and another between helices IV and V, with the ceiling formed by the helix III–IV loop (residues

Pro160 -Tyr175) (Figure 2.1B). The channel has been hypothesized to be responsible for oxygen supply to the active site but remains a topic of debate. Molecular dynamics of oxygen diffusion from the helix IV-V entry to the hydrophobic channel and active site suggest that this branch of channel may supply oxygen for enzyme turnover [10]. Furthermore, mutagenesis of hydrophobic residues lining the channel indicated channel occlusion results in reduced oxygen binding to the active site [51].

Motions of RsCcO sampled by elNémo in low-energy modes 7-9 resulted in alternation between open and constricted states in the oxygen pathway (PDB IDs 2GSM [7]) (Figure 2.13).

Gating occurs in both branches of the channel between the entrance and their junction (Figures

2.13B and 2.13C). The helix I-II-III entrance is also periodically closed by the motions of helix

III coupled to large amplitude motions of the pontoon helix over the entrance (Figure 2.7C). The importance of this oxygen pathway and its constriction points can be inferred from the conserved

residues lining and constricting the channel (Figures 2.13B and 2.13C). Additionally, the

hydrophobic channel of TtCcO, which naturally functions under significantly reduced oxygen

tension, was analyzed with elNémo. TtCcO contains substantially longer hydrophobic channels

with more uniform diameter. These channels did not experience constriction, but instead

remained continuously dilated.

Figure 2.13: Conformational gating in the RsCcO oxygen channels (following page). (A) Converging, hydrophobic channels for oxygen, defined by thick dashed lines, are formed by residues in helices I-VI, in gray ribbons, and the external loop joining helices III and IV, in cyan. Motions sampled in all three normal modes result in alternating constriction, shown as discontinuous blue cavities, and dilation, shown as a virtually continuous Y-shaped channel in pink. Constriction occurs at two points. (B) In the channel starting near helices IV-V, ConSurf indicates moderate conservation, colored in green surface and side chain tubes, of channel forming residues and increasingly high conservation, colored in blue, between the constriction (Val194, Leu246, and Ile250) and the active sites. (C) The channel starting at helices I-III is also increasingly conserved between its constriction (Phe108, Trp172, and Leu174) and the active sites.

Figure 2.13 (cont’d)

Previous comparison of Bos taurus (Bt) CcO, RsCcO, PdCcO, and TtCcO structures defined hydrophobic channels in the same region but with differing lengths and diameters [11].

BtCcO, RsCcO, and PdCcO channels were shorter than the long, continuous channel of TtCcO.

On the other hand, RsCcO channels are of considerably larger volume while being lined with large hydrophobic residues that produce constriction points along the channel [11]. Based on crystallographic comparisons in the presence and absence of xenon gas [11] and the low-energy motion simulations of CcO presented here, we propose that variation in the pathway diameter of

RsCcO may regulate oxygen occupancy next to the active site, whereas the larger, unconstricted channel in TtCcO enhances oxygen flow to the active site.

In addition to providing an oxygen channel ceiling in RsCcO, the helix III-IV loop has also been suggested to be involved in proton pumping. Substitutions in this loop

(Gly171Asp, Trp172Phe, Trp172Gln, and Tyr175Ala) show reduced proton pumping and increased proton backflow [52-53]. Molecular dynamics simulations suggest this loop can move deeper into the membrane, allowing transient water molecule chains to form between the top of the D-pathway and the hypothesized proton exit regions including the Mg2+ site [53]. ProFlex analysis supports the flexibility of this loop (RsCcO Gly161-Trp172; Figure 2.5A and 2.5B) and is in agreement with the changes of solvent accessibility between reaction intermediates

(Gly169-Tyr175) [14]. Flexibility, conservation, mutational analysis, and differential solvent accessibility in reaction intermediates all support a role of the helix III-IV loop in CcO pumping.

Conclusions

Normal modes and flexibility analysis indicate that the core of CcO undergoes global counter-rotational and compressional motions on the interior and exterior of the membrane, coupled with rocking of the β-sheet domain. These motions are a consequence of helix tilting, bending, and compression, potentially regulating oxygen flux to the active site while the D- pathway integrity is maintained. The subunit I C-terminus is predicted to be the most mobile region in the protein, suggesting conformational regulation of proton uptake at the D-pathway entrance. The K-pathway helices are the most flexible intra-membrane region, supporting water molecule movement in the pathway and are consistent with conformational changes in crystal structures. This work predicts global motions and underlying internal flexibility of CcO that is consistent with experimental observations, while providing a new perspective on K-pathway proton conductance, regulation of proton uptake in the D-pathway, and gating of oxygen flux to the active site.

The CcO main-chain flexibility predictions have been used to design further experiments including substitutions of flexible residues in the conformationally-changing K-pathway.

Additionally, these results have been used to rationalize the mechanisms of K- and D-pathway alteration by mutagenesis and small molecule binding. A conserved hydrophobic binding region may inhibit CcO by altering K-pathway conformational changes and proton uptake (See

Chapters 3 and 4). A D-pathway entry double variant may increase helical flexibility and allow water molecules to supply protons to the middle of the pathway (Buhrow, unpublished data).

Overall, these findings suggest that main-chain flexibility and conformational changes play a significant role in integral membrane protein function.

REFERENCES

1. Hosler, J. P., S. Ferguson-Miller, and D. A. Mills. 2006. Energy transduction: Proton transfer through the respiratory complexes. Annu. Rev. Biochem. 75: 165-187.

2. Brzezinski, P., and R. B. Gennis. 2008. Cytochrome c oxidase: Exciting progress and remaining mysteries. J. Bioenerg. Biomembr. 40: 521-531.

3. Diaz, F., 2010. Cytochrome c oxidase deficiency: Patients and animal models. Biochim. Biophys. Acta. 1802: 100-110.

4. Fetter, J. R., J. Qian, J. Shapleigh, J. W. Thomas, A. García-Horsman, E. Schmidt, J. Hosler, G. T. Babcock, R. B. Gennis, and S. Ferguson-Miller. 1995. Possible proton relay pathways in cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 92: 1604-1608.

5. Svensson-Ek, M., J. Abramson, G. Larsson, S. Törnroth, P. Brzezinski, and S. Iwata. 2002. The x-ray crystal structures of wild-type and EQ(I-286) mutant cytochrome c oxidases from Rhodobacter sphaeroides. J. Mol. Biol. 321: 329-339.

6. Brändén, M., F. Tomson, R. B. Gennis, and P. Brzezinski. 2002. The entry point of the K- proton- transfer pathway in cytochrome c oxidase. Biochemistry. 41: 10794-10798.

7. Qin, L., C. Hiser, A. Mulichak, R. M. Garavito, and S. Ferguson-Miller. 2006. Identification of conserved lipid/detergent-binding sites in a high-resolution structure of the membrane protein cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 16117-16122.

8. Lepp, H., L. Salomonsson, J.-P. Zhu, R. B. Gennis, and P. Brzezinski. 2008. Impaired proton pumping in cytochrome c oxidase upon structural alteration of the D pathway. Biochim. Biophys. Acta. 1777: 897-903.

9. Zhu, J., H. Han, A. Pawate, and R. B. Gennis. 2010. Decoupling mutants in the D-channel of the aa3-type cytochrome c oxidase from Rhodobacter sphaeriodes suggest that a continuous hydrogen-bonded chain of waters is essential for proton pumping. Biochemistry. 49: 4476-4482.

10. Hofacker, I., and K. Schulten. 1998. Oxygen and proton pathways in cytochrome c oxidase. Proteins. 30: 100-107.

11. Luna, V. M., Y. Chen, J. A. Fee, and C. D. Stout. 2008. Crystallographic studies of Xe and Kr binding within the large internal cavity of cytochrome ba3 from Thermus thermophilus.

Biochemistry. 47: 4657-4665.

12. Qin, L., J. Liu, D. A. Mills, D. A. Proshlyakov, C. Hiser, and S. Ferguson-Miller. 2009. Redox- dependent conformational changes in cytochrome c oxidase suggest a gating mechanism for proton uptake. Biochemistry. 48: 5121-5130.

13. Liu, J., L. Qin, and S. Ferguson-Miller. 2011. Crystallographic and online spectral evidence for role of conformational change and conserved water in cytochrome oxidase proton pump. Proc. Natl. Acad. Sci. USA. 108: 1284-1289.

14. Busenlehner, L. S., L. Salomonsson, P. Brzezinski, and R. N. Armstrong. 2006. Mapping protein dynamics in catalytic intermediates of the redox-driven proton pump cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 15398-15403.

15. Sharpe, M. A., and S. Ferguson-Miller. 2008. A chemically explicit model for the mechanism of proton pumping in heme-copper oxidases. J. Bioenerg. Biomembr. 40: 541-549.

16. Pomès, R., G. Hummer, and M. Wikström. 1998. Structure and dynamics of a proton shuttle in cytochrome c oxidase. Biochim. Biophys. Acta. 1365: 255-260.

17. Wikström, M., M. I. Verkhovsky, and G. Hummer. 2003. Water-gated mechanism of proton translocation by cytochrome c oxidase. Biochim. Biophys. Acta. 1604: 61-65.

18. Cukier, R. I. 2005. A molecular dynamics study of water chain formation in the proton- conducting K channel of cytochrome c oxidase. Biochim. Biophys. Acta. 1706: 134-146.

19. Song, Y., E. Michonova-Alexova, and M. R. Gunner. 2006. Calculated proton uptake onanaerobic reduction of cytochrome c oxidase: Is the reaction electroneutral? Biochemistry. 45: 7959-7975.

20. Henry, R. M, C.-H. Yu, T. Rodinger, and R. Pomès. 2009. Functional hydration and conformational gating of proton uptake in cytochrome c oxidase. J. Mol. Biol. 387: 1165- 1185.

21. Sugitani, R., and A. A. Stuchebrukhov. 2009. Molecular dynamics simulation of water in cytochrome c oxidase reveals two water exit pathways and the mechanism of transport. Biochim. Biophys. Acta. 1787: 1140-1150.

22. Ghosh, N., X. Prat-Resina, M. R. Gunner, and Q. Cui. 2009. Microscopic pKa analysis of Glu286 in cytochrome c oxidase (Rhodobacter sphaeroides). Biochemistry. 48: 2468- 2485.

23. Zhang, J., and M. R. Gunner. 2010. Multiconformation continuum electrostatics analysis of the effects of a buried Asp introduced near heme a in Rhodobacter sphaeroides cytochrome c oxidase. Biochemistry. 49: 8043-8052.

24. Jacobs, D. J., A. J. Rader, L. A. Kuhn, and M. F. Thorpe. 2001. Protein flexibility predictions using graph theory. Proteins. 44: 150-165.

25. Hespenheide, B. M., A. J. Rader, M. F. Thorpe, and L. A. Kuhn. 2002. Identifying protein folding cores from the evolution of flexible regions during unfolding. J. Mol. Graph. Model. Graphics. 21: 195-207.

26. Valadié, H., J. J. Lacapčre, Y.-H. Sanejouand, and C. Etchebest. 2003. Dynamical properties of the MscL of Escherichia coli. J. Mol. Biol. 332: 657-674.

27. Shrivastava, I. H., and I. Bahar. 2006. Common mechanism of pore opening shared by five different potassium channels. Biophys. J. 90: 3929-3940.

28. Miloshevsky, G. V., and P. C. Jordan. 2006. The open state gating mechanism of gramicidin a requires relative opposed monomer rotation and simultaneous lateral displacement. Structure. 14: 1241-1249.

29. Szarecka, A., Y. Xu, and P. Tang. 2007. Dynamics of heteropentameric nicotinic acetylcholine receptor. Proteins. 68: 948-960.

30. Maxwell, J. 1864. On the calculation of the equilibrium and stiffness of frames. Philos. Mag. 27: 294-299.

31. Rader, A. J., B. M. Hespenheide, L. A. Kuhn, M. F. Thorpe. 2002. Protein unfolding: Rigidity lost. Proc. Natl. Acad. Sci. USA. 99: 3540-3545.

32. Tanford, C. 1980. The hydrophobic effect. Wiley, New York.

33. Keating, K. S., S. C. Flores, M. B. Gerstein, and L. A. Kuhn. 2009. StoneHinge: Hinge prediction by network analysis of individual protein structures. Protein Sci. 18: 359-371.

34. Williams, M. A., J. M. Goodfellow, and J. M. Thornton. 1994. Buried waters and internal cavities in monomeric proteins. Protein Sci. 3: 1224-1235.

35. Raymer, M. L., P. C. Sanschagrin, W. F. Punch, S. Venkataraman, E. D. Goodman, and L. A. Kuhn. 1997. Predicting conserved water-mediated and polar ligand interactions in proteins. J. Mol. Biol. 265: 445-464.

36. Vriend, G. 1990. WHAT IF: A molecular modeling and drug design program. J. Mol. Graph. 8: 52-56.

37. Brooks, B., and M. Karplus. 1983. Harmonic dynamics of proteins. Proc. Natl. Acad. Sci. USA 80: 6571-6575.

38. Suhre, K., and Y.-H. Sanejouand. 2004. elNémo: A normal mode web server for protein movement analysis and the generation of templates for molecular replacement. Nucleic

Acids Res. 32: W610-W614.

39. Tama, F., F. X. Gadea, O. Marques, and Y.-H. Sanejouand. 2000. Building-block approach for determining low-frequency normal modes of macromolecules. Proteins. 41: 1-7.

40. Roux, B., and M. Karplus. 1988. The normal modes of the gramicidin-a dimer channel. Biophys. J. 53: 297-309.

41. Essiz, S. G., and R. D. Coalson. 2007. Langevin dynamics of molecules with internal rigid fragments in the harmonic regime. J. Chem. Phys. 127: 104-109.

42. Landau, M., I. Mayrose, Y. Rosenberg, F. Glaser, E. Martz, T. Pupko, and N. Ben-Tal. 2005. ConSurf 2005: The projection of evolutionary conservation scores of residues on protein structures. Nucleic Acids Res. 33: W299-W302.

43. Bairoch, A., and B. Boeckmann. 1991. The SWISS-PROT protein sequence data bank. Nucleic Acids Res. 19: 2247-2249.

44. Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignments. Nucleic Acids Res. 22: 4673- 4680.

45. Ho, B. K., and F. Gruswitz. 2008. HOLLOW: Generating accurate representations of channel and interior surfaces in molecular structures. BMC Struct. Biol. 8: 49.

46. Ostermeier, C., A. Harrenga, U. Ermler, and H. Michel. 1997. Structure at 2.7 Å resolution of the Paracoccus denitrificans two-subunit cytochrome c oxidase complexed with an antibody FV fragment. Proc. Natl. Acad. Sci. USA. 94: 10547-10553.

47. Hunsicker-Wang, L. M., R. L. Pacoma, Y. Chen, J. A. Fee, and C. D. Stout. 2005. A novel cryoprotection scheme for enhancing the diffraction of crystals of recombinant cytochrome ba3 oxidase from Thermus thermophilus. Acta Cryst. D61: 340-343.

48. Taly, A., M. Delarue, T. Grutter, M. Nilges, N. Le Novère, P.-J. Corringer, and J.-P. Changeux. 2005. Normal mode analysis suggests a quaternary twist model for the nicotinic receptor gating mechanism. Biophys. J. 88: 3954-3965.

49. Nowak, C., T. Laredo, J. Gebert, J. Lipkowski, R. B. Gennis, S. Ferguson-Miller, W. Knoll, and R. L. C. Naumann. 2010. 2D-SEIRA spectroscopy to highlight conformational changes of the cytochrome c oxidase induced by direct electron transfer. Soft Matter. 6: 5523-5532.

50. Koepke, J., E. Olkhova, H. Angerer, H. Müller, G. Peng, and H. Michel. 2009. High resolution crystal structure of Paracoccus denitrificans cytochrome c oxidase. Biochim. Biophys. Acta. 1787: 635-645.

52. Namslauer, I., and P. Brzezinski. 2009. A mitochondrial DNA mutation linked to colon cancer results in proton leaks in cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 106: 3402-3407.

53. Seibold, S. A., D. A. Mills, S. Ferguson-Miller, and R. I. Cukier. 2005. Water chain formation and possible proton pumping routes in Rhodobacter sphaeroides cytochrome c oxidase. Biochemistry. 44: 10475-10485.

54. Rasmussen, S. G. F., H.-J. Choi, J. J. Fung, E. Pardon, P. Casarosa, P. S. Chae, B. T. DeVree, D. M. Rosenbaun, F. S. Thian, T. S. Kobilka, A. Schnapp, I. Konetzki, R. K. Sunahara, S. H. Gellman, A. Pautsch, J. Steyaert, W. I. Weis, and B. K. Kobilka. 2011. Structure of a nanobody-stabilized active site of the β2 adrenoceptor. Nature. 469: 175-181.

55. Zhou, Y., J. H. Morais-Cabral, A. Kaufman, and R. MacKinnon. 2001. Chemistry of ion coordination and hydration revealed by a K+ channel-Fab complex at 2.0 Å resolution. Nature. 414: 43-48.

56. Ujwal, R., D. Cascio, J. P. Colletier, S. Faham, J. Zhang, L. Toro, P. Ping, J. Abramson. 2008. The crystal structure of mouse VDAC1 at 2.3 Å resolution reveals mechanistic insights into metabolic gating. Proc. Natl. Acad. Sci. USA. 105: 17742-17747.

57. Roberts, V. A., and M. E. Pique. 1999. Definition of the interaction domain for cytochrome c on cytochrome c oxidase. J. Biol. Chem. 274: 38051-38060.

CHAPTER 3: Structural predictions and functional consequences of porphyrin, steroid, and detergent ligands binding to the cytochrome c oxidase steroid binding site

This work was a major collaboration with Drs. Carrie Hiser, Ling Qin, and Denise Mills. I contributed cholate and deoxycholate titrations to our initial manuscript. Reprint (adapted) with permission from Qin., L., D.A. Mills, L. Buhrow, C. Hiser, and S. Ferguson-Miller. 2008. A conserved steroid binding site in cytochrome c oxidase. Biochemistry. 47(38): 9931-9933. Copyright 2008 American Chemical Society.

In our subsequent studies, I contributed assays on a variety of bile acid effects, computational studies of ligand binding, intrinsic flexibility calculations, and I had substantial input in data interpretation and writing of this paper. The second manuscript is in review as Hiser, C.*, L. Buhrow*, J. Liu, Leslie A. Kuhn, and S. Ferguson-Miller. A conserved amphipathic ligand binding region influences K-path dependent activity of cytochrome c oxidase. Biochemistry. *These authors contributed equally to this work. In this Chapter, my portion of these works are presented and discussed.

Introduction

Specific, conserved lipid binding sites have been identified in high-resolution X-ray crystal structures of cytochrome c oxidase (CcO) [1-4]. The presence of these sites, as well as the interchangeability between native lipids and crystallographic detergents, clarify earlier reports of lipids and detergents modulating CcO activity [5-9]. An interesting case of CcO modulation is observed with the bile acid cholate. Cholate (Figure 3.1) is a very effective detergent often used to purify Bos taurus CcO (BtCcO) and has been reported to inhibit the enzyme strongly [7]. Cholate molecules are found co-crystallized in all high resolution structures of the bovine enzyme including at a site proximal to one of the proton uptake pathways, the K-pathway [10]. This bound cholate interacts with amino acid residues whose bacterial homologs have been shown to be critical for K proton pathway function [10-11].

The conservation of this cholate-occupied site was observed upon the crystallization of

Rhodobacter sphaeriodes CcO (RsCcO) in the presence of a related bile acid. The structure of wild-type (WT) RsCcO with deoxycholate (DXC; Figure 3.1), a bile acid detergent similar to cholate and only varying by one hydroxyl group position, was resolved in the same binding site as cholate in BtCcO (Figure 3.2) [12]. The deoxycholate steroid ring system is bound at the subunit I/II interface where it is associated with the protein surface that is normally within the membrane bilayer. While the deoxycholate carboxylate tail further strengthens this ligand’s binding by interacting with a crystallographic cadmium ion. Cadmium is a critical component of the crystallization conditions for RsCcO, as well as a known proton pathway inhibitor [12].

Figure 3.1. Bile acid structures. RsCcO WT and the E101A variant activities were monitored in the presence of (A) cholate, (B) chenodeoxycholate, (C) ursodeoxycholate, (D) lithocholate, (E) deoxycholate, (F) glycocheno-deoxycholate, (G) glycoursodeoxycholate, and (F) taurocholate. Ligand structures were depicted using eMolecules (Solano Beach, CA) molecular structure search tool.

Figure 3.1 (cont’d)

Figure 3.2. Deoxycholate resolved near the K-pathway entrance of RsCcO. The two-subunit structure of RsCcO has been solved in the presence of deoxycholate (PDB: 3DTU [12]). The catalytic core subunits I and II are depicted in green and cyan ribbons, respectively with heme (magenta sticks) and copper (orange spheres) cofactors. The two proton uptake channels, D- and K-pathways, are indicated by black dashed lines. The water molecules within the pathways and at the subunit interfaces are shown as red spheres and key proton pathway residues are shown as sticks. The resolved deoxycholate molecule (gray sticks) is located near the K-pathway entrance.

Mutagenic and kinetic investigations of the bacterial CcO have established the presence of two proton uptake pathways: the D-pathway and the K-pathway. The D-pathway functions as

89 an uptake pathway for both substrate protons (consumed in the reduction of molecular oxygen) and pumped protons; while the K-pathway takes up substrate protons only. The K-pathway leads from the subunit I/II interface on the negative side of the membrane through the conserved residues E101, Lys362, Ser299, Thr359, and Tyr288 (R. sphaeriodes numbering) into the active site composed of heme a3 and copper B [11].

The entrance of the K-pathway, defined by E101, is the location of cholate and deoxycholate binding and will be termed the steroid binding site. This binding location suggests that the observed stimulatory and inhibitory effects of these ligands may be due to blockage or facilitation of K-pathway dependent proton uptake. Spectral analysis of heme cofactors supports this hypothesis as K-pathway variants or the addition of inhibitory detergents result in greater heme a reduction, presumably due to their blockage of proton uptake required to stabilize the reduced state of heme a3 [11, 13-15, Hiser, C., et al. in review].

Substitutions of K-pathway residues, including E101A, leads to severe inhibition of overall activity but maintain proton pumping functionality [11, 15-16]. In the standard lauryl maltoside (LM) concentration (i.e., 0.06%), the activity of RsCcO E101A is approximately 5-8% of wild type (WT) CcO [15, 17]. In lower LM conditions (i.e., 0.01%), but still sufficient to give full WT activity, the E101A variant is more active with approximately 50% WT activity (Hiser,

C., et al., in review). LM is routinely used to solublize and assay CcO, since the original finding that it was uniquely effective in dispersing, stabilizing and activating the enzyme [18]. The inhibitory effect of LM on the E101A enzyme suggests that LM strongly interacts with this variant and may impact the K-pathway function.

Activity recovery of another CcO proton pathway variant has been observed upon the addition of lipidic carboxylates. The D-pathway variant D132A is extremely inactive with less than 5% of WT turnover [19]. It can be stimulated three to seven fold upon the addition of micromolar concentrations of arachidonic acid but not the arachidonate alcohol derivative [19].

Arachidonic acid is thought to bind at the entry of the D-pathway such that its carboxyl group may chemically rescue D132A. Based on the crystallographic positioning of deoxycholate in the

RsCcO structure, it is also reasonable that this bile acid may stimulate the E101A enzyme by replacement of the substituted carboxyl group.

In addition to lipidic carboxylates, heterocyclic amphiphiles have also been shown to affect CcO activity. Porphyrin ligand titrations including the heme precursor protoporphyrin IX

(PPIX) and, to a lesser degree, the heme breakdown product bilirubin (BR) have been shown to stimulate and then inhibit the activity of E101A enzyme in a manner dependent on LM concentration (Hiser, C., et al., in review). This detergent dependence supports a LM interaction at, or close to, the steroid binding site and its competition with other amphipathic ligands.

Therefore, the effects of exogenous ligands would be expected to reflect both the ligand’s direct effect and its ability to compete with LM for binding. The initial stimulation observed in the presence of PPIX and bilirubin has been suggested to be due to the displacement of highly inhibitory LM, while the later inhibition is thought to be a result of ligand binding at the steroid binding site (Hiser, C., et al., in review).

The differential effects of steroids, porphyrins, and LM on E101A compared to WT

RsCcO suggest that their binding behavior is affected by the change in chemistry at the mouth of the K-pathway, namely the loss of the E101 carboxyl group (Hiser, C., et al., in review). To date, ligand interactions near the K-pathway have only been defined crystallographically for deoxycholate in the WT RsCcO [12]. The interaction sites of PPIX, bilirubin, and LM remain unclear.

In this study, the functional effects of diverse bile acids were investigated. The crystallographically-observed ligands, cholate and deoxycholate, were studied to determine the possibility and magnitude of chemical rescue of the K-pathway E101A variant. Additionally, the structural specificity for the RsCcO steroid binding site was evaluated using biological variants of these bile acids. The feasibility and potential binding modes of known inhibitory porphyrin ligands and LM detergent molecules were predicted using small molecule docking and knowledge of molecular mimicry from diverse proteins. Finally, solvent accessibility and crystal structure comparisons have identified redox-dependent conformational changes in the K- pathway that impact steroid binding [30, 35]. To investigate whether ligand binding to the steroid binding site could affect such conformational changes, the ProFlex main-chain flexibility method was applied to RsCcO in the presence and absence of bound deoxycholate. This work identifies diverse ligand regulators of RsCcO as well as testable hypotheses for the molecule binding interactions between the steroid binding site and porphyrin or detergent molecules.

Materials and Methods

Consurf amino acid conservation mapping

The amino acid conservation of CcO was evaluated using the Consurf method [20] as described in the Chapter 2 Materials and Methods section and in [21] with over 250 CcO amino acid sequences analyzed.

Purification of CcO

R. sphaeroides strains overexpressing the 37-2 WT CcO [22] or the E101A variant CcO

[17] were grown and cell membranes were prepared as described in [22]. The RsCcO was

2 isolated by metal affinity chromatography as described for crystallography in [17].

Activity assays

Oxygen uptake rates were measured with a Clark-type electrode and turnover rates

(Turnover Number: electrons per second per CcO molecule) were calculated as described in [23].

Assay mixtures contained 100 mM HEPES pH 7.4, 24 mM KCl, 2.8 mM ascorbate, 1 mM

N,N,N',N'-tetramethyl-p-phenylenediamine (TMPD), 0.06% LM with 30 μM bovine heart cytochrome c as the substrate and bile acids as noted. All bile acids (Figure 3.1) were from

Sigma-Aldrich (St. Louis, MO).

Protein and ligand structures

The crystal structures of BtCcO (PDB: 2DYR [4]), RsCcO (PDB: 2GSM [1]), RsCcO with deoxycholate (PDB: 3DTU [12]), and Paracoccus denitrificans CcO (PdCcO) (PDB: 3HB3

2 RsCcO WT and E101A proteins were provided courtesy of Drs. Carrie Hiser, Denise Mills, and Ling Qin. 93

[3]) were used for ligand binding and flexibility analysis. Crystal structures of ferrochelatase in complex with cholate (PDB: 2PO7 [24]) or heme (PDB: 2QD2 [25]) and serum albumin in complex with biliverdin (PDB: 2VUE [26]) or fusidic acid (PDB: 2VUF [26]) were used for ligand docking. LM, biliverdin, and protoporphryin IX ligand structures were obtained from

PdCcO (PDB: 3HB3 [3]), serum albumin (PDB: 2VUE [26]), and ferrochelatase (PDB: 2QD2

[25]) structures, respectively.

SLIDE small molecule docking

The groove between subunits I and II in the crystal structure of RsCcO with bound deoxycholate removed (PDB: 3DTU [12]) was used as the target for SLIDE (Screening Ligands by Induced-fit Docking, Efficiency) prediction of the binding modes of experimentally identified ligands [27-28]. SLIDE characterizes a binding pocket by a template of chemistry-labeled points which have favorable positions for protein-ligand hydrophobic interactions or hydrogen bonds.

In this work, the known interactions of LM were used to create the template. To sample ligand flexibility fully, low energy conformations of all docked ligands were generated using Omega v2

(OpenEye Scientific Software, Santa Fe, NM). Using distance and geometric constraints, SLIDE predicts the orientation of ligand binding by sampling all orientations that yield good shape and chemical complementarity between the ligand and protein, then chooses the orientation of the conformer with the most favorable ∆Gbinding according to SLIDE’s OrientScore [Tonero, in preparation]. SLIDE AffiScore, a weighted sum of favorable polar and hydrophobic interactions and unfavorable (unsatisfied/repulsive) interactions between the protein and ligand, was used to assess the relative ∆Gbinding of these diverse ligands in the binding groove.

Ligand modeling based on crystallographic interactions in other proteins

Serum albumin and ferrochelatase have been crystallized in separate complexes with steroid and heme-like ligands in a single binding site (see Protein and ligand structures, above).

To test whether heme-like ligands could mimic the deoxycholate steroid ring interactions in

RsCcO in the same orientation as they exhibit in serum albumin and ferrochelatase, the overlaid ligands were transposed into the RsCcO steroid binding site. To do so, the ferrochelatase-bound cholate (PDB: 2PO7 [24]) and serum albumin-bound fusidic acid (PDB: 2VUF [26]) ligands were aligned to the RsCcO deoxycholate (PDB: 3DTU [12]) based on least-squares fitting of their steroid ring atoms using OEChem RMSD (OpenEye Scientific Software, Santa Fe, NM).

The heme- and steroid-bound structures were then superimposed for serum albumin (PDB:

2VUE and 2VUF [26]), and separately for ferrochelatase (PDB: 2PO7 [24] and 2QD2 [25]) based on least-squares fitting of their protein backbone atoms using PyMOL molecular graphics software (Schrödinger, Inc., New York, NY). The resulting overlaid heme and biliverdin ligands were located in the same reference frame as the RsCcO steroid binding site. Whether the heme ligands were sterically permitted and energetically favorable in this orientation was assessed using PyMOL and SLIDE prediction of ∆Gbinding, respectively.

ProFlex flexibility analysis

The main chain flexibility of CcO in the presence (PDB: 3DTU [12]) and absence of deoxycholate (PDB: 2GSM [1]) was evaluated using the ProFlex method [29] as described in the

Chapter 2 Materials and Methods section and in [21].

Results and Discussion

Characterization of the steroid binding sites in mammalian and bacterial CcO

Comparing the cholate binding site in BtCcO with the deoxycholate site in RsCcO indicates that they have important structural similarities beyond their location at the entrance region of the K-pathway (Table 3.1). The consensus between cholate and deoxycholate binding includes interactions with the conserved K-pathway residue E62/E101 (Bos taurus/ R. sphaeroides numbering) in subunit II, a conserved threonine T66/T105 one helical turn above this entrance point of the K-pathway, and positive charges interacting with the ligand carboxyl group consisting of a cadmium ion (RsCcO) or a pair of arginine residues (BtCcO) (Figure 3.3).

Cholate in BtCcO contacts subunit VIa of the other monomer in the crystallographic dimer, as well as the highly conserved subunit II residue W65, while deoxycholate in RsCcO contacts subunit I via conserved residues A319 and P358. In both cases several water molecules are stabilized by the presence of these ligands, participating in their interactions with subunits I and

II (Figure 3.3).

Table 3.1: RsCcO and BtCcO steroid binding site characterization. The RsCcO with bound deoxycholate (PDB: 3DTU [12]) and bovine CcO with bound cholate (PDB: 2DYR [4]) were aligned based on least-squares fitting of their protein backbone atoms using PyMOL molecular graphics software (Schrödinger, Inc., New York, NY). The ligand binding sites were defined as any residue with a non-hydrogen atom within 5 Å of the protein's bound ligand. Common interactions were defined if identical functional group(s) or ion made the same interactions with a respective ligand atom. Subunits are represented as subscripts.

RsCcO BovCcO Ligand Deoxycholate Cholate Common Protein- E101II E62II Ligand Interactions I102II T63II

T105II T66II

2+ R14IV Cd R17IV Unique Contacts P315I A58II

Y318I Q59II

A319I E60II

P358I V61II

I361I I64II

W65II

I67II

Figure 3.3. RsCcO and BtCcO steroid binding site with conserved residues and bound water molecules. The steroid binding site side chains from (A) R. sphaeroides and (B) B. taurus are colored by degree of conservation in CcO sequences as calculated in [21] where >75%, 50- 75%, or less than 50% conservation are colored dark blue, medium blue, and gray, respectively. Bound deoxycholate or cholate (green sticks), K-pathway water molecules (red spheres), and a cadmium ion (orange sphere) are shown.

Stimulation of the RsCcO E101A variant by co-crystallized bile acids

To test whether the co-crystallized bile acids cholate and deoxycholate were positioned in such a way in the bile acid binding site to rescue the RsCcO E101A variant chemically, these ligands were titrated and the CcO activity monitored using an oxygen reduction activity assay.

Micromolar concentrations of both cholate and deoxycholate stimulated the E101A protein approximately 10-12 fold (Figure 3.4), reaching activities comparable to the activity of RsCcO

E101A in minimal detergent (0.01% LM) (Hiser, C., et al., in review). Deoxycholate stimulates

E101A enzyme to this maximal activity at much lower concentrations than cholate, suggesting that deoxycholate has a significantly higher affinity for the steroid binding site. As the

98 stimulated activities are similar to the minimal detergent activity, it is clear that cholate and deoxycholate do displace the inhibitory LM, but without further activation, it is unclear whether these bile acids are actually chemically rescuing the mutated the K-pathway carboxyl group.

However, further studies with these and other carboxylate ligands (Hiser, C., et al., in review) suggest a biphasic effect of detergent displacement combined with true activation, followed by inhibition.

Figure 3.4. The steady-state activities of the detergent-solubilized RsCcO E101A [12]. The oxygen reduction of RsCcO E101A was monitored with increasing concentrations of the bile - acids cholate and deoxycholate. A maximum velocity of ~750 e /sec/aa3 was reached upon the addition of micromolar concentrations of either ligand. Deoxycholate had a greater ability to stimulate this mutant as compared to cholate, potentially due to its higher affinity for the RsCcO steroid binding site.

Bile acid structural specificity for stimulating E101A RsCcO activity

To elucidate the structural specificity for the observed stimulation of the E101A variant, activity was measured with a variety of bile acids at 250 μM final concentration in moderate

(0.06% LM) detergent conditions (Table 3.2 and Figure 3.1). Cholate and deoxycholate remained the strongest effectors at this concentration. Only chenodeoxycholate and lithocholate were close to their effectiveness exhibiting ~10 fold stimulation. Taurocholate stimulated only three-fold and glycoursodeoxycholate (GUDCA) stimulated by 40%. Ursodeoxycholate and glycochenodeoxycholate had no observable effect. These results suggest a non-conjugated tail may be important for the optimal carboxyl group placement and hydroxyl groups projecting toward the protein and away from the lipid environment are critical for stimulating the E101A enzyme (Table 3.2). All the bile acids tested had no significant effect on WT RsCcO activity at

250 μM concentration under these conditions. Their differential effect on the variant presumably was due to their ability to displace the E101A-specific inhibitory LM ligand.

100

Table 3.2. Effects of bile acids on E101A mutant and WT RsCcO. Purified RsCcO was assayed in 0.06% LM as described in Materials and Methods in the absence (Buffer) or presence of 250 μM of each bile acid. All numbers are average values of at least three independent trials with the standard deviation. Due to its poor solubility in water, lithocholate was solubilized in 100mM CHES pH 8.8 with 20% ethanol and was added to a final concentration of 75μM.

RsCcO E101A turnover RsCcO WT turnover - - Bile acid additive number (e /sec/aa3) number (e /sec/aa3) Buffer 60 ± 10 1270 ± 20 Cholate 800 ± 10 1260 ± 10 Deoxycholate (DXC) 780 ± 20 1210 ± 20 Chenodeoxycholate 620 ± 60 1240 ± 30 Lithocholate 530 ± 60 1150 ± 80 Taurocholate 150 ± 10 1360 ± 80 Glycoursodeoxycholate 90 ± 40 1160 ± 40 (GUDCA) Glycochenodeoxycholate 60 ± 30 1360 ± 60 Ursodeoxycholate 50 ± 10 1260 ± 60

Modeling bilirubin and protoporphyrin interactions based on other crystal structures

Given the variety of compounds that appear to interact with relatively high affinity at what was originally observed to be a bile acid/steroid binding site, it is important to note that other structurally characterized proteins show interchangeability of ligands, in particular steroids, porphyrins and other amphipathic molecules [24-26]. Knowledge of how heme and steroid molecules bind relative to one another in the same site in another protein – mimicking each others’ key protein interactions – provides a testable hypothesis for how they might bind relative to one another in the RsCcO deoxycholate site. Ferrochelatase was crystallized with its substrate

PPIX and separately with the bile acid cholate [24-25], showing how a steroid and porphyrin could bind to the same site (Figure 3.5A). Similarly, human serum albumin was crystallized with

101 biliverdin, a natively transported heme catabolic product, or fusidic acid, an antibiotic that is structurally similar to bile acids, in the same binding site [26]. To test whether these porphyrin ligands could mimic the deoxycholate steroid ring interactions in RsCcO in the same orientation they exhibit in serum albumin and ferrochelatase, their PPIX and biliverdin ligands were transposed into the RsCcO steroid binding site as described in the Materials and Methods. The transposed ferrochelatase heme group interpenetrated the RsCcO protein surface, and thus would be sterically disallowed in this binding orientation, while the serum albumin biliverdin was transposed into the RsCcO site with no steric hindrance and interacted favorably with RsCcO according to SLIDE evaluation of ∆Gbinding (Figure 3.5B). This binding mode suggests that bilirubin may bind to the RsCcO steroid binding site in an energetically favorable manner using a crystallographically-defined ligand conformation.

102

Figure 3.5. Potential binding orientations of known lipidic ligands in the RsCcO steroid binding site. (A) Heme (magenta sticks) and cholate (black sticks) bound in the same site in two different crystal structures of ferrochelatase (PDB: 2PO7 [24] and 2QD2 [25]). (B) An energetically favorable orientation of the bilirubin analog biliverdin (red sticks) in RsCcO based on ligand transposition from serum albumin. Energetically favorable SLIDE dockings of (C) PPIX (magenta sticks), (D) bilirubin (red sticks), and (E) two LM flexible conformers (yellow and magenta sticks) into RsCcO. (B-E) The crystallographically bound deoxycholate (black sticks; PDB: 3DTU [12]) and/or LM (blue sticks; PDB 3HB3 [3]) are shown as reference molecules.

Docking of protoporphyrin IX and bilirubin into the steroid binding site

Alternative binding interactions for these ligands in the steroid binding site were identified by SLIDE small molecule docking of low energy PPIX and bilirubin conformers. The top four PPIX dockings (scoring two standard deviations greater than the average docked PPIX score) depicted two possible binding modes in which half the PPIX is located in the steroid binding site and one propionate group makes protein interactions similar to the deoxycholate's

103 carboxyl tail (Figure 3.5C). The top eight docked bilirubin conformers (based on SLIDE scores at least two standard deviations above the mean) shared a single binding mode (Figure 3.5D).

This binding mode, like human serum albumin binding of bilirubin, creates an “L-shaped” ligand conformation [26] with the bilirubin dipyrrinone group located in the same site as the steroid ring. The SLIDE docking of PPIX and bilirubin support the interaction of these ligands with the steroid binding site and provide testable hypotheses for ligand-protein side chain interactions.

Modeling detergent binding to the RsCcO steroid binding region

Predicting possible binding orientations of detergents in the steroid binding site is currently a significant challenge; however, an alternative interaction mode is suggested by a

PdCcO crystal structure containing a LM bound in a unique orientation just above the steroid binding site [3]. In this structure, the maltose head group is represented as flexible and located in the middle of the membrane region, while the tail is embedded in a hydrophobic groove that is occupied by an alkyl tail in all oxidase structures (Figure 3.5E). To test whether ligands binding to this unusual LM binding site could also interact with the steroid site, low energy conformations of LM were docked into the PdCcO LM site using SLIDE. Among the many energetically favorable dockings identified that had acyl tails occupying the PdCcO LM tail binding site, the maltose head group positions varied between being solvent-exposed or protein associated, occluding the RsCcO steroid binding site (Figure 3.5E). This overlap explains how

LM and other detergents could compete with bile acid binding. It may also explain the inhibitory effects of certain detergents on the E101A variant, since a detergent head group binding in the steroid binding site could interfere directly with K-pathway proton uptake.

104

This work addresses the effects of various amphipathic and steroid-like molecules that appear to interact at a crystallographically-defined steroid binding site located at the entrance of the K-pathway of bacterial and mammalian CcO. It is interesting that none of the more than 50

CcO crystal structures in the Protein Data Bank has any ligand resolved in this site, except additives like cholate or deoxycholate. This finding suggests that in the native structure this site is not a phospholipid binding site, since detergents typically occupy phospholipid sites in the bacterial CcO that contain native lipids in the BtCcO [2]. Additionally, in RsCcO and BtCcO crystal structures, an alkyl tail is always resolved in the deep groove where the tail of the unusual

PdCcO LM position is located. This region mediates the interaction between subunit I and the second transmembrane helix of subunit II and is highly conserved. In BtCcO, the alkyl tail position is occupied by the fatty acid chain of a cardiolipin molecule whose other chains and head group are part of the contact region between two monomers that forms the dimer interface

[4].

Mechanisms of ligand activation/ inhibition of CcO

There are several potential mechanisms by which amphipathic compounds could activate or inhibit CcO by favoring or disfavoring proton uptake in the K-pathway. One hypothesis relates to the disruption or stabilization by ligand binding of a proton conducting network in the

K-pathway, which involves hydrogen bonded amino acids and water molecules that are likely required for proton transfer. Waters in the lower K-pathway have been crystallographically resolved in PdCcO [3], RsCcO [30-31], and BtCcO (PDB: 2DYR [4]), and a role for E101 in organizing water near the K-pathway entrance has been invoked for bacterial CcO [3, 32-33].

Deoxycholate binding results in the resolution of three additional water molecules located

105 between the ligand’s planar ring system and the lower K-pathway (Figure 3.3). The resulting changes in hydrogen bonding at the K-pathway entry may affect the continuity of the pathway or reduce mobility of water molecules involved in conducting protons.

Another possibility, which is not mutually exclusive with the first, is that the presence of amphipathic ligands in the region associated with the K-pathway may either promote or restrict conformational change observed to occur in the reduced state [30-31]. These conformational changes are postulated to be necessary to open the K-pathway for rapid proton access to the heme a3 region [21, 30, 34]. The central region of the K-pathway (subunit I, helix VIII, residues

355-364) has been shown by hydrogen/dueterium exchange to experience altered solvent accessibility upon enzyme reduction [35]. Intrinsic protein flexibility analysis also suggests this region is highly flexible, influencing water molecule positions along the K-pathway [21].

Binding of deoxycholate appears to lower the flexibility of CcO, based upon lower average B-values of WT RsCcO crystals with the ligand bound as compared to unbound (Figure

3.6A-B) [12]. ProFlex main-chain flexibility prediction also suggests that deoxycholate reduces conformational flexibility of the K-pathway region relative to the unbound state (Figure 3.6C-D).

In addition, upon reduction of RsCcO crystals there is loss of bound deoxycholate [32], suggesting that ligands like deoxycholate may stabilize the oxidized state and prevent conformational changes required for efficient functioning of the K-pathway.

106

Figure 3.6. RsCcO K-pathway rigidification upon ligand binding. (A) RsCcO (PDB: 2GSM [1]) and (B) deoxycholate-bound RsCcO (PDB: 3DTU; [12]) are colored by crystallographic B- value. (C) RsCcO and (D) deoxycholate-bound RsCcO are colored by internal protein flexibility as determined by the ProFlex. Deep blue represents greatest protein main-chain rigidity while red represents greatest flexibility.

Potential physiological implications of the CcO steroid binding site

The physiological significance of the steroid binding site remains elusive. However, this

107 site is conserved from bacteria to mammals in both amino acid residues and ability to bind physiologically relevant steroidal and porphyrin molecules with reasonably high affinity. Studies of purified CcO support bile acids [36] and glucocorticords [37] as activity regulators. Studies of mitochondrial function have also implicated steroids as strong effectors [38-39]. Whole cell and tissue studies suggest that porphyrins and bile acids modulate cell fate, specifically at the level of

CcO activity [40-41]. Together these results suggest that the CcO steroid binding site is sensitive to a variety of physiological ligands with previously uncharacterized molecular effects.

Originally, the cholate bound sites in BtCcO were suggested to be adenine nucleotide regulatory sites due to the similar size, shape, and hydrophobic protein binding characteristics of these ligands [10]. There is considerable evidence that adenine nucleotides strongly affect mammalian CcO activity [42-43]. However, activity effects of nucleotide binding to mammalian or bacterial CcO have yet to be connected to the steroid binding site.

It is also plausible that the steroid binding site, given its location at the dimer interface, is involved in the oligomeric state of CcO. The currently available crystallographic structures of

BtCcO show a homodimer with two cholate molecules resolved that the interface. In fact, the cholate bound to the steroid binding site bridges the subunit I/II catalytic core of one monomer to the nuclear-encoded subunit VIa of the second monomer. There is evidence that subunit VIa has three adenine nucleotide binding sites, a phosphorylation site, regulates the cytochrome c affinity, and affects overall activity [44]. Therefore, regulating the CcO monomer/dimer equilibrium via the steroid binding site may play a key role in respiratory activity and feedback activation/inhibition.

108

Conclusions

The current studies investigate a variety of ligands that can impact the crystallographically-defined steroid binding site at the subunit I/II interface in mammalian and bacterial CcO. A number of subunit II residues in this site are highly conserved, including the K- pathway entry residue E101. The E101A variant can be stimulated 10-12 fold with micromolar concentrations of bile acids which bind at this steroid binding site. Of these, deoxycholate has the strongest affinity for this site. PPIX and bilirubin ligands which had previously been shown to stimulate CcO may reasonably interact at the same steroid binding site based on docking and ligand overlays from diverse proteins. Additionally, the observed LM competition with steroids and porphyrins may be due to a second, overlapping site at the subunit I/II interface. Steroid, porphyrin, and detergent ligands may inhibit CcO by restricting K-pathway conformational changes or by preventing proper water molecule organization. To define detailed protein-ligand interactions further, key residues involved in ligand binding, and regulatory ligands that are physiologically relevant, further molecular modeling and in vitro ligand screen approaches will be required.

109

REFERENCES

110

REFERENCES

1. Qin, L., C. Hiser, A. Mulichak, R. M. Garavito, and S. Ferguson-Miller. 2006. Identification of conserved lipid/detergent-binding sites in a high-resolution structure of the membrane protein cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 16117-16122.

2. Qin, L., M. A. Sharpe, R. M. Garavito, and S. Ferguson-Miller. 2007. Conserved lipid-binding sites in membrane proteins: a focus on cytochrome c oxidase. Curr. Opin. Struct. Biol. 17: 444-450.

3. Koepke, J., E. Olkhova, H. Angerer, H. Muller, G. Peng, and H. Michel. 2009. High resolution crystal structure of Paracoccus denitrificans cytochrome c oxidase: New insights into the active site and the proton transfer pathways. Biochim. Biophys. Acta. 1787: 635-645.

4. Shinzawa-Itoh, K., H. Aoyama, K. Muramoto, H. Terada, T. Kurauchi, Y. Tadehara, A. Yamasaki, T. Sugimura, S. Kurono, K. Tsujimoto, T. Mizushima, E. Yamashita, T. Tsukihara, and S. Yoshikawa. 2007. Structures and physiological roles of 13 integral lipids of bovine heart cytochrome c oxidase. EMBO J. 26: 1713-1725.

5. Brierley, G. P., and A. J. Merola. 1962. Studies of electron-transfer system. 48. Phospholipid requirements in cytochrome oxidase. Biochimica. Biophysica. Acta. 64(2): 218-228.

6. Vanneste, W. H., M. Ysebaert, and H. S. Mason. 1974. Decline of molecular activity of cytochrome-oxidase during purification. J. Biol. Chem. 249: 7390-7401.

7. Yu, C. A., L. Yu, and T. E. King. 1975. Studies on cytochrome-oxidase - Interactions of cytochrome-oxidase protein with phospholipids and cytochrome-c. J. Biol. Chem. 250: 1383-1392.

8. Vik, S. B., and R. A. Capaldi. 1977. Lipid requirements for cytochrome-c oxidase activity. Biochemistry. 16: 5755-5759.

9. Vik, S. B., and R. A. Capaldi. 1980. Conditions for optimal electron-transfer activity of cytochrome-c oxidase isolated from beef-heart mitochondria. Biochem. Bioph. Res. Commun. 94: 348-354.

10. Tsukihara, T., H. Aoyama, E. Yamashita, T. Tomizaki, H. Yamaguchi, K. Shinzawa-Itoh, R. Nakashima, R., Yaono, and S. Yoshikawa. 1996. The whole structure of the 13-subunit oxidized cytochrome c oxidase at 2.8 angstrom. Science. 272: 1136-1144.

111

11. Brändén, M., F. Tomson, R. B. Gennis, and P. Brzezinski. 2002. The entry point of the K- proton-transfer pathway in cytochrome c oxidase. Biochemistry. 41: 10794-10798.

12. Qin, L., D. A. Mills, L. Buhrow, C. Hiser, and S. Ferguson-Miller, 2008. A conserved steroid binding site in cytochrome oxidase. Biochemistry. 47: 9931-9933.

13. Konstantinov, A. A., S. Siletsky, D. Mitchell, A. Kaulen, and R. B. Gennis. 1997. The roles of the two proton input channels in cytochrome c oxidase from Rhodobacter sphaeroides probed by the effects of site-directed mutations on time-resolved electrogenic intraprotein proton transfer. Proc. Natl. Acad. Sci. USA. 94: 9085-9090.

14. Vygodina, T. V., C. Pecoraro, D. Mitchell, R. Gennis, and A. A. Konstantinov. 1998. Mechanism of inhibition of electron transfer by amino acid replacement K362M in a proton channel of Rhodobacter sphaeroides cytochrome c oxidase. Biochemistry. 37: 3053-3061.

15. Tomson, F. L., J. E. Morgan, G. P. Gu, B. Barquera, T. V. Vygodina, and R. B. Gennis. 2003. Substitutions for glutamate 101 in subunit II of cytochrome c oxidase from Rhodobacter sphaeroides result in blocking the proton-conducting K-channel. Biochemistry. 42: 1711- 1717.

16. Ma, J. X., P. H. Tsatsos, D. Zaslavsky, B. Barquera, J. W. Thomas, A. Katsonouri, A. Puustinen, M. Wikstrom, P. Brzezinski, J. O. Alben, and R. B. Gennis. 1999. Glutamate- 89 in subunit II of cytochrome bo(3) from Escherichia coli is required for the function of the heme-copper oxidase. Biochemistry. 38: 15150-15156.

17. Qin, L., D. A. Mills, C. Hiser, A. Murphree, R. M. Garavito, S. Ferguson-Miller, and J. Hosler. 2007. Crystallographic location and mutational analysis of Zn and Cd inhibitory sites and role of lipidic carboxylates in rescuing proton path mutants in cytochrome c oxidase. Biochemistry. 46: 6239-6248.

18. Rosevear, P., T. VanAken, J. Baxter, and S. Ferguson-Miller 1980. Alkyl glycoside detergents: A simpler synthesis and their effects on kinetic and physical properties of cytochrome c oxidase. Biochemistry. 19: 4108-4115.

19. Fetter, J., M. Sharpe, J. Qian, D. Mills, S. Ferguson-Miller, and P. Nicholls. 1996. Fatty acids stimulate activity and restore respiratory control in a proton channel mutant of cytochrome c oxidase. FEBS Lett. 393: 155-160.

20. Landau, M., I. Mayrose, Y. Rosenberg, F. Glaser, E. Martz, T. Pupko, and N. Ben-Tal. 2005. ConSurf 2005: The projection of evolutionary conservation scores of residues on protein structures. Nucleic Acids Res. 33: W299-W302.

21. Buhrow, L., S. Ferguson-Miller, and L. A. Kuhn. 2012. From static structure to living protein: Computational analysis of cytochrome c oxidase main-chain flexibility. Biophys. J. 102: 2158-2166.

112

22. Hiser, C., D. A. Mills, M. Schall, and S. Ferguson-Miller. 2001. C-terminal truncation and histidine-tagging of cytochrome c oxidase subunit II reveals the native processing site, shows involvement of the C-terminus in cytochrome c binding, and improves the assay for proton pumping. Biochemistry. 40: 1606-1615.

23. Hosler, J. P., J. Fetter, M. M. J. Tecklenburg, M. Espe, C. Lerma, and S. Ferguson-Miller. 1992. Cytochrome-aa3 of Rhodobacter sphaeroides as a model for mitochondrial cytochrome c oxidase - Purification, kinetics, proton pumping, and spectral analysis. J. Biol. Chem. 267: 24264-24272.

24. Dailey, H. A., C. K. Wu, P. Horanyi, A. E. Medlock, W. Najahi-Missaoui, A. E. Burden, T. A. Dailey, and J. Rose. 2007. Altered orientation of active site residues in variants of human ferrochelatase. Evidence for a hydrogen bond network involved in catalysis. Biochemistry. 46: 7973-7979.

25. Medlock, A. E., T. A. Dailey, T. A. Ross, H. A. Dailey, and W. N. Lanzilotta. 2007. A pi-helix switch selective for porphyrin deprotonation and product release in human ferrochelatase. J. Mol. Biol. 373: 1006-1016.

26. Zunszain, P. A., J. Ghuman, A. F. McDonagh, and S. Curry. 2008. Crystallographic analysis of human serum albumin complexed with 4Z,15E-bilirubin-IX alpha. J. Mol. Biol. 381: 394-406.

27. Schnecke, V., C. A. Swanson, E. D. Getzoff, J. A. Tainer, and L. A. Kuhn. 1998. Screening a peptidyl database for potential ligands to proteins with side-chain flexibility. Proteins. 33: 74-87.

28. Zavodszky, M. I., P. C. Sanschagrin, R. S. Korde, and L. A. Kuhn. 2002. Distilling the essential features of a protein surface for improving protein-ligand docking, scoring, and virtual screening. J. Comput. Aid Mol. Des. 16: 883-902.

29. Jacobs, D. J., A. J. Rader, L. A. Kuhn, and M. F. Thorpe. 2001. Protein flexibility predictions using graph theory. Proteins. 44: 150-165.

30. Qin, L., J. Liu, D. A. Mills, D. A. Proshlyakov, C. Hiser, and S. Ferguson-Miller. 2009. Redox-dependent conformational changes in cytochrome suggest a gating mechanism for proton uptake. Biochemistry. 48: 5121-5130.

31. Liu, J. A., L. Qin, and S. Ferguson-Miller. 2011. Crystallographic and online spectral evidence for role of conformational change and conserved water in cytochrome oxidase proton pump. Proc. Natl. Acad. Sci. USA. 108: 1284-1289.

32. Ferguson-Miller, S., C. Hiser, and J. Liu. 2012. Gating and regulation of the cytochrome c oxidase proton pump. Biochim. Biophys. Acta. 1817: 489-494.

113

33. Sharpe, M. A., L. Qin, S. Ferguson-Miller. 2005. Proton entry, exit and pathways in cytochrome oxidase: Insight from conserved water, in biophysical and structural aspects of bioenergetics. The Royal Society of Chemistry. Cambridge, UK, Cambridge. pp26-54.

34. Sharpe, M. A., and S. Ferguson-Miller. 2008. A chemically explicit model for the mechanism of proton pumping in heme-copper oxidases. J. Bioenerg. Biomembr. 40: 541-549.

35. Busenlehner, L. S., L. Salomonsson, P. Brzezinski, and R. N. Armstrong. 2006. Mapping protein dynamics in catalytic intermediates of the redox-driven proton pump cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 15398-15403.

36. Van Buuren, K. J, H., and B. F. Van Gelder.1974. Biochemical and biophysical studies on cytochrome c oxidase. XIII. Effect of cholate on enzymic activity. Biochim. Biophys. Acta. 333: 209–217.

37. Simon, N., P. Jolliet, C. Morin, R. Zin, S. Urien, J.-P. Tillement. 1998. Glucocorticoids decrease cytochrome c oxidase activity of isolated rat kidney mitochondria. FEBS Lett. 435: 25–28.

38. Molano, F., A. Saborido, J. Delgado, M. Moran, and A. Megias. 1999. Rat liver lysosomal and mitochondrial activities are modified by anabolic-androgenic steroids. Med. Sci. Sport Exer. 31: 243-250.

39. Starkov, A. A., R. A. Simonyan, V. I. Dedukhova, S. E. Mansurova, L. A. Palamarchuk, and V. P. Skulachev. 1997. Regulation of the energy coupling in mitochondria by some steroid and thyroid hormones. Biochim. Biophys. Acta. 1318: 173-183.

40. Vaz, A. R., M. Delgado-Esteban, M. A. Brito, J. P. Bolanos, D. Brites, and A. Almeida. 2010. Bilirubin selectively inhibits cytochrome c oxidase activity and induces apoptosis in immature cortical neurons: assessment of the protective effects of glycoursodeoxycholic acid. J. Neurochem. 112: 56-65.

41. Malik, S. G., K. A. Irwanto, J. D. Ostrow, and C. Tiribelli. 2010. Effect of bilirubin on cytochrome c oxidase activity of mitochondria from mouse brain and liver. BMC Res. Notes. 3: 162-165.

42. Arnold, S., and B. Kadenbach. 1997. Cell respiration is controlled by ATP, an allosteric inhibitor of cytochrome-c oxidase. Eur. J. Biochem. 249: 350-354.

43. Frank, V., and B. Kadenbach. 1996. Regulation of the H+/e(-) stoichiometry of cytochrome c oxidase from bovine heart by intramitochondrial ATP/ADP ratios. FEBS Lett. 382: 121- 124.

44. Kadenbach, B., J. Napiwotzki, V. Frank, S. Arnold, S. Exner, and M. Huttemann. 1998. Regulation of energy transduction and electron transfer in cytochrome c oxidase by adenine nucleotides. J. Bioenerg. Biomembr. 30: 25-33.

114

CHAPTER 4: Three-pronged computational approach to predict regulatory ligands of cytochrome c oxidase

This work was completed in collaboration with Drs. Carrie Hiser and Jeffery Van Voorst and will be submitted as Buhrow, L., C. Hiser, J. R. Van Voorst, S. Ferguson-Miller, and L. A. Kuhn. Three-pronged computational approach to predict regulatory ligands of cytochrome c oxidase.

115

Introduction

Cytochrome c oxidase (CcO) is thought to be a key regulator of oxidative phosphorylation rate and efficiency [1-4] through the use of iso-forms of nuclear-encoded accessory subunits [4-9], phosphorylation [10-11], and allosteric regulation by adenine nucleotides and adrenal or thyroid hormones [12-16]. Eukaryotic CcO contains a variable number of nuclear-encoded subunits that are hypothesized to be involved in sensing metabolic demand and oxygen tension. Additionally, fourteen phosphorylation sites on mammalian CcO have been well characterized [10-11], including Tyr304/Tyr347 (Bos taurus / Rhodobacter sphaeroides numbering) and Ser115/Ser156 in subunit I and Ser126/Ser173 in subunit II (Figure

4.1) [10, 17]. The phosphorylation of Tyr304/Tyr347 not only inhibits CcO but abolishes allosteric regulation by nucleotides [17]. Adenine nucleotides have been shown to both activate and inhibit mammalian CcO [12-16]. However, the location(s) and mechanism of this inhibition remain controversial. Mammalian subunits IV and VI have been suggested to include adenine nucleotide allosteric binding sites on their matrix domains [18-19], while subunit V has been suggested to include an allosteric site specific for thyroid hormone regulation (Figure 4.1) [12].

116

Figure 4.1: Structure of bacterial and mammalian cytochrome c oxidases. Rhodobacter sphaeroides (PDB: 3DTU [25]) and Bos taurus (PDB: 2DYR [64]) CcO structures are depicted as ribbons with homologous subunits I and II colored in green and cyan, respectively. Heme (magenta sticks) and copper (copper colored spheres) cofactors and crystallographic bile acid ligands (gray spheres) are also present. The conserved K-pathway for proton uptake is represented as a dashed line, leading from the internal side of the membrane to the active site. The approximate location of the membrane is shown with gray spheres and curved lines. (A) Three phosphorylation sites observed on the mammalian enzyme are conserved in the bacterial CcO core and include Tyr347 and Ser156 in subunit I and Ser173 in subunit II (red spheres). (B) Three nuclear encoded mammalian subunits [IV (dark blue), Va (dark pink), and VIa (orange)] have been implicated in nucleotide and hormone allosteric regulation.

The steroid/bile acid binding site in mammalian CcO has also been suggested to be an

ADP regulatory site, based on shape, chemistry, and binding preference similarites between ADP and the crystallographically bound cholate molecule in the bovine CcO (bovCcO; Figure 4.1)

[20]. Occupation of the CcO steroid binding site at the entrance of the critical K proton uptake pathway has been shown to inhibit electron transfer between the a and a3 heme cofactors in the mammalian enzyme [21-23]. In the R. sphaeroides CcO (RsCcO), this site has been shown to bind a diverse group of ligands including detergents, fatty acids, steroids, and porphyrins [Chapt

117

3; Hiser, C., et al., in review]. However, the native regulatory ligand(s) for either mammalian or bacterial CcO binding sites remain(s) elusive.

To understand CcO regulation at the steroid binding site, a three-pronged computational approach was applied to predict natively abundant ligands that are specific for this site. Ligand comparisons were performed using the ROCS method to understand which native ligands are structurally and chemically similar to the crystallographically bound bile acid, allowing these ligands to mimic binding. Protein binding site comparisons were performed using the SimSite3D method to understand which native protein binding sites are most similar to the RsCcO steroid binding site. The native ligands bound to similar sites may then be investigated as candidate molecules for interaction with the steroid site. Finally, the ligands predicted to bind to the steroid binding site were docked using the SLIDE method to evaluate potential interactions with the protein. All commercially available, soluble, and high scoring candidate ligands were tested for their ability to affect the RsCcO wild type (WT) and the E101A variant in varying detergent and substrate conditions. Oxygen consumption rate measurements support a number of nucleotide, flavin, and steroid ligands as regulators of CcO by binding to the steroid site.

118

Materials and Methods

Ligand database selection and processing

The Binding Mother of All Databases (Binding MOAD [24]) is a collection of 9836 crystallographic protein-ligand, protein-cofactor, and protein-ligand-cofactor complexes at or below 2.5 Å resolution. These complexes are filtered to include only physiologically relevant small molecules, thereby eliminating catalytically required small molecules (e.g., active site metals, hemes, porphyrins, etc.), covalently bound ligands, crystallographic additives (e.g., buffers, detergents, ions, etc.), and other macromolecules, defined as peptides longer than 10 amino acids and DNA fragments larger than 4 nucleic acid residues. The redundant Binding

MOAD database that contains multiple copies of proteins with >90% sequence identity was used to maintain ligand diversity. Identical protein-ligand complexes were manually removed after analysis to reduce a bias for highly represented structures in the Protein Data Bank (PDB). This database was selected to compare against the RsCcO steroid binding site as it is limited to physiological metabolites and addresses the question of which ligand(s) may regulate CcO in vivo.

Geometric and chemical comparison of the RsCcO bound deoxycholate to diverse metabolites

Rapid Overlay of Chemical Structures (ROCS; OpenEye Scientific Software, Sante Fe,

NM) aligns a database of molecules or molecule conformations to a reference ligand structure based on maximal shape and chemical similarity [65]. Ligands are represented by a Gaussian function about each atom center [65]. ROCS aligns these distributions such that they have maximal overlap as assessed by a Tanimoto scoring metric. Tanimoto scores, a variation of the

Jaccard index, describe the correlation between vectors with continuous or binary attributes [65].

119

This metric, calculated by the quotient of the union and intersection of two sets of vectors, may independently evaluate the maximal shape and chemical similarities between ligands. The geometric (ShapeTanimoto) and chemical (ColorTanimoto) similarities are each independently scored and scaled between zero and one, resulting in a maximum combined score

(TanimotoCombo) of two for exact ligand matches.

The ROCS method was applied to identify physiological ligands similar to deoxycholate

(DXC) whose interactions with CcO are potentially capable of being mimicked. The RsCcO's bound DXC (PDB: 3DTU [25]) was compared to the crystallographic ligands in Binding MOAD and all low energy conformers of these ligands, as determined by Omega v2 (OpenEye Scientific

Software, Sante Fe, NM). The TanimotoCombo score was used to rank the results as the entire molecule, both in shape and chemical complementarity, is of importance. All ligands scoring two to three standard deviations or greater than the average ROCS TanimotoCombo score were grouped into dominant chemical classes (e.g., bile acids, steroids, etc.) with redundant ligands removed. Ligands were selected for testing based on TanimotoCombo score ranking, commercial availability, and ability to be solubilized at a concentration sufficient for assaying

(100 μM was the minimal final concentration tested).

Chemical and shape comparisons of the RsCcO steroid binding site and diverse crystallographic binding sites

SimSite3D is a protein binding site screening tool that returns a ranked list of binding sites most similar to the site of interest. Each binding site is represented using a set of pharmacophore points with associated bonding geometry and a solvent accessible molecular

120 surface [Van Voorst, in preparation]. The pharmacophore points denote positions where non- hydrogen atoms, of the appropriate chemistry, could be placed to form favorable interactions with atoms in the protein. Polar pharmacophore points have an associated direction to estimate the strength of potential hydrogen bond interactions, while the molecular surface ensures that highly ranked binding sites will have a shape as compared to that of the site of interest [Van

Voorst, in preparation].

The RsCcO steroid binding site with DXC removed (PDB: 3DTU [25]) was used as the query binding site. The SimSite3D standard protein template was used and included all pharmacophore points within 3.0 Å of at least one non-hydrogen atom of DXC. The molecular surface of the subunit I/II interface within 4.0 Å of DXC was computed using the MSMS

3 method [26].

All ligand binding sites found in Binding MOAD were ranked based on SimSite3D score.

Sites scoring two standard deviations greater than the average SimSite3D score across Binding

MOAD were analyzed as potentially physiologically relevant. These sites were grouped into dominant chemical classes (e.g., steroid, nucleotide, or lipid binding proteins) with redundant protein-ligand crystallographic complexes removed. The top scoring ligand classes' pharmacophore points that matched the pharmacophore points of the RsCcO steroid binding site were mapped onto the RsCcO structure and binding site residues.

3 Dr. Jeff van Voorst executed the SimSite3D method comparing the RsCcO steroid binding site to Binding MOAD. All subsequent data analysis, representations, and interpretation were completed by Leann Buhrow. 121

Small molecule docking of predicted ligands into the RsCcO steroid binding site

The RsCcO steroid binding site with bound DXC removed (PDB: 3DTU [25]) was used as the target for SLIDE (Screening Ligands by Induced-fit Docking, Efficiency) [27-28] prediction of the binding modes of ROCS and SimSite3D candidate ligands. SLIDE characterizes favorable positions for protein-ligand hydrophobic interactions or hydrogen bonds in a binding pocket by generating a template of chemistry-labeled points. To sample candidate ligand flexibility, Omega v2 was used to generate all low-energy conformations of the docked ligands.

SLIDE predicts the ligand binding orientation by sampling all conformations that exhibit shape and chemical complementarity between the protein and ligand. The best orientation is then chosen according to the most favorable ΔGbinding according to SLIDE's OrientScore [Tonero, in preparation]. SLIDE's AffiScore uses a weighted sum of favorable hydrophobic and hydrogen- bonding interactions and unfavorable (unsatisfied/ repulsive) interactions between the protein and docked ligand to compare relative ΔGbinding of diverse ligands. Additionally, the interactions between the diverse docked ligands and the RsCcO steroid binding site were calculated and tabulated.

Purification of CcO

R. sphaeroides strains overexpressing the 37-2 WT CcO [29] or the E101A variant CcO

[30] were grown and cell membranes were prepared as described in [31], and the RsCcO was

4 isolated by metal affinity chromatography as described for crystallography in [30].

4 Proteins were provided by Dr. Carrie Hiser. 122

RsCcO activity determined by oxygen consumption assays

Oxygen uptake rates were measured with a Clark-type electrode and turnover rates

(Turnover number: electrons per second per CcO molecule) were calculated as described in [31].

Assay mixtures contained 100 mM HEPES pH 7.4, 24 mM KCl, 2.8 mM ascorbate, 1 mM

N,N,N',N'-tetramethyl-p-phenylenediamine (TMPD) with 0-30 μM bovine heart cytochrome c as the substrate, and with varying levels of lauryl maltoside (LM) and other additives as noted. All assays included 5.6 mM EDTA. Farnesyl diphosphate (FPP) was purchased from Echelon

Biosciences Incorporated (Logan, UT). All other additives were from Sigma-Aldrich (St. Louis,

MO).

123

Results and Discussion

Chemical and geometric comparisons between the CcO crystallographic deoxycholate and diverse, natively abundant ligands

Mammalian and bacterial CcO have been crystallized in the presence of bile acid detergents [20, 25]. These detergents, cholate in the case of bovCcO and DXC in the case of

RsCcO, are bound at a conserved site, near the entrance of the K-pathway (Figure 4.1). The ability of these steroid-derived detergents to interact at a conserved binding site and affect activity supports the possibility that they mimic structurally similar, native regulatory molecules.

In order to understand which natively abundant ligands are significantly similar to the crystallographic conformation of DXC, the ROCS method was used to evaluate chemical and geometric similarities between DXC and Binding MOAD ligands. Binding MOAD was selected because this database contains nearly 10,000 native ligands with available structural information.

The ROCS method identified six chemical classes of ligands as significantly similar based on scoring three standard deviations or better than the average TanimotoCombo score across Binding MOAD. Ligand classes include bile acids, steroids, retinoic acid analogs, amino acid analogs, nucleotide and flavin analogs, and thyroid hormone analogs (Tables 4.1A-4.1F).

These candidate ligands are structurally similar in that they contain a cyclic head group, in some cases a four-membered steroid ring system or nucleotide nitrogenous base, and a flexible tail region that often contains a phosphate or carboxylate (Figure 4.2). The alignment of these ligands superimposes the DXC steroid ring system with the diverse ligands' cyclic head groups and superimposes the DXC carboxylated tail with the diverse ligands flexible tail regions (Figure

4.3).

124

Table 4.1. ROCS predicted ligands. ROCS-analyzed ligands scoring at least three standard deviations above the average TanimotoCombo score across Binding MOAD were selected as similar to deoxycholate. These ligands were grouped into dominant chemical classes with redundant ligands removed. (A) Bile acids are cholic acid derivates. (B) Steroids contain a four- membered steroid ring system and exclude bile acids. (C) Retinoic acid ligands are retinol derivatives. (D) Amino acid analogs contain a common amino acid. (E) Nucleotide and flavin analogs contain a nucleic acid base or a flavin moiety. (F) Thyroid hormone ligands are derivatives of one of the thyroid hormones. Nucleotides, flavins, and nicotinamide ligands have been suggested to interact at the CcO steroid binding site by protein binding site comparisons. (G) Representatives of these ligands were selected as highly similar to the crystallographically bound deoxycholate as they scored at least two standard deviations above the average ROCS score across Binding MOAD.

Table 4.1A. ROCS predicted bile acids Rank Ligand ID Conformer Ligand name Tanimoto Shape Color Combo Tanimoto Tanimoto 1 DXC 1 Deoxycholic acid 1.51 0.84 0.66 2 CHD 1 Cholic acid 1.49 0.92 0.57 3 CHC 8 Chenodeoxycholic acid 1.25 0.86 0.4 4 GCH 111 Glycocholic acid 1.24 0.82 0.42 6 TCH 308 Taurocholic acid 1.19 0.79 0.4 7 CHO 104 Glycochenodeoxycholic 1.14 0.81 0.34 acid 8 TUD 302 Taurochenodeoxycholic 1.09 0.77 0.32 acid 25 12H 1296 12-Hydroxydodecanoic 0.95 0.54 0.41 acid 27 NPC 1478 4-Hydroxy-3- 0.95 0.67 0.28 nitrophenylacetyl- ε-aminocaproic acid anion 52 T24 1015 (9z,11e,13s,15z)-13- 0.89 0.54 0.35 Hydroxyoctadeca-9,11,15- trienoic acid

125

Table 4.1 (cont’d)

Table 4.1B. ROCS predicted steroids Rank Ligand ID Conformer Ligand name Tanimoto Shape Color Combo Tanimoto Tanimoto 5 TH2 115 Testosterone 1.21 0.83 0.38 hemisuccinate 9 CI2 1 5β-Dihydroprogesterone 1.09 0.8 0.29 10 ANO 1 5β-Androstane-3,17-dione 1.07 0.78 0.29 11 HE7 67 Estradiol-17β- 1.02 0.71 0.31 hemisuccinate 12 HCY 3 Hydrocortisone 1.01 0.81 0.2 14 C0R 4 Corticosterone 1 0.78 0.22 15 ANB 1 Androsta-1,4-diene- 0.99 0.72 0.27 3,17-dione 16 ASD 1 4-Androstene-3-17-dione 0.99 0.72 0.27 17 STR 1 Progesterone 0.99 0.72 0.27 18 CLR 393 Cholesterol 0.98 0.71 0.27 19 ERG 18 Ergosterol 0.98 0.72 0.26 20 AS4 4 Aldosterone 0.97 0.76 0.21 21 HC3 12 25-Hydroxycholesterol 0.96 0.71 0.26 22 1CA 3 Desoxycorticosterone 0.96 0.73 0.23 23 HCR 16 7-Hydroxycholesterol 0.96 0.71 0.24 24 TES 1 Testosterone 0.96 0.71 0.25 28 FFA 1 Epitestosterone 0.94 0.61 0.33 29 SNL 3 Spironolactone 0.94 0.68 0.26 30 LAN 14 Lanosterol 0.94 0.69 0.25 32 R18 1 Metribolone 0.93 0.67 0.26 35 FUA 7 Fuscidic acid 0.93 0.61 0.32 36 HC2 4 20-Hydroxycholesterol 0.92 0.74 0.19 37 DHT 1 5-α-dihydrotestosterone 0.92 0.68 0.24 38 DEX 2 Dexamethasone 0.92 0.73 0.19 40 PLO 1 Pregnenolone 0.91 0.69 0.23 43 PDN 1 Cortisone 0.9 0.7 0.21 44 PO1 93 (9β,13ρα,14β,17α)-2- 0.9 0.66 0.24 methoxyestra-1,3,5(10)- triene-3,17-diyl disulfamate 45 EST 1 Estradiol 0.9 0.7 0.21 50 EQU 1 Equilenin 0.9 0.69 0.21 60 17H 1 Tetrahydrogestrinone 0.89 0.66 0.23 51 667 46 6-Oxo-8,9,10,11- 0.89 0.68 0.22 tetrahydro- 7hcyclohepta[c][1]benzop yran-3-Osulfamate

126

Table 4.1 (cont’d)

Table 4.1C. ROCS predicted retinoic acid analogs Rank Ligand Conformer Ligand name Tanimoto Shape Color ID Combo Tanimoto Tanimoto 13 RE9 44 Synthetic retinoic acid 1.01 0.65 0.36 26 564 107 6-(5,5,8,8-tetramethyl-5,6, 7,8 0.95 0.64 0.31 tetrahydro-naphtalene-2- carbonyl)-naphtalene-2 carboxylic acid 34 184 11 6-[hydroxy-(5,5,8,8-tetra-methyl- 0.93 0.61 0.32 5,6,7,8-tetrahydro-naphtalen-2- Yl)-methyl]-naphtalene-2- carboxylic acid 4-[3-oxo-3-(5,5,8,8-tetra-methyl- 39 156 36 5,6,7,8-tetrahydro-naphthalen-2- 0.92 0.61 0.31 Yl)-propenyl] -benzoic acid 3-Methyl-7-(5,5,8,8-tetramethyl- 5,6,7,8-tetra hydro-naphthalen-2- 58 R13 31 Yl) -octa-2,4,6-trienoic acid 0.89 0.63 0.26

Table 4.1D. ROCS predicted amino acid analogs Rank Ligand Conformer Ligand name Tanimoto Shape Color ID Combo Tanimoto Tanimoto 31 BP5 15 3-(2,2'-Bipyridin-5-Yl)-L-alanine 0.93 0.65 0.28 33 CCL 785 N-6-[(Cyclopentyloxy) 0.93 0.68 0.25 carbonyl]-D-lysine 42 NHL 166 (4s)-4-(2-Naphthyl methyl)-D- 0.91 0.63 0.28 glutamic acid 46 PDE 795 Para-Nitrophenyl phospho- 0.9 0.7 0.2 nobutanoyl D-alanine 48 MP2 702 N-[(benzyloxy)carbonyl]-L- 0.9 0.65 0.25 cysteinylglycine 53 DNS 554 N-6-{[5-(Dimethylamino)-1- 0.89 0.64 0.25 naphthyl]sulfonyl}-L-lysine 55 F6F 272 2-{[4-(Trifluoromethoxy) 0.89 0.68 0.21 benzoyl]amino} ethyl dihydrogen phosphate 56 HEP 510 Phenyl[1-(Nsuccinyl- 0.89 0.62 0.27 amino)pentyl]phosphonate

127

Table 4.1 (cont’d)

Table 4.1E. ROCS predicted nucleotide and flavin analogs Rank Ligand ID Conformer Ligand name Tanimoto Shape Color Combo Tanimoto Tanimoto 41 P1S 4 (6ar,12ar)-3-(Hydroxy- 0.91 0.7 0.21 methyl)-6h [1,3]dioxolo [5,6][1]benzofuro[3,2- C]chromen-6a(12ah)-Ol 47 EMA 1215 (Adenin-9-Yl-ethoxy 0.9 0.66 0.24 methyl)-hydroxyphos phinyl-diphosphate 49 RMB 116 N1-(5'-Phospho- α- 0.9 0.7 0.2 ribosyl)-5- methylbenzimidazole 54 GCQ 548 Gemcitabine diphosphate 0.89 0.67 0.22 57 RPD 128 (C8-R)-hydantocidin 5'- 0.89 0.66 0.22 phosphate 59 IMU 144 Phosphoric acid mono-[5- 0.89 0.63 0.25 (2-amino-4-oxo-4,5- dihydro-3h-pyrrolo[3,2- D]pyrimidin-7-Yl)-3,4- dihydroxypyrrolidin-2- Ylmethyl] ester

Table 4.1F. ROCS predicted thyroid hormone analogs Rank Ligand ID Conformer Ligand name Tanimoto Shape Color Combo Tanimoto Tanimoto 61 IH5 1 {3,5-Dichloro-4-[4- 0.89 0.68 0.2 hydroxy-3- (propan-2-Yl) phenoxy]phenyl}acetic acid 63 T3 21 T3 thyroid hormone 0.85 0.6 0.25

128

Table 4.1 (cont’d)

Table 4.1G. ROCS predicted flavins, nicatinamides, and nucleotides Ligand ID Conformer Ligand name Tanimoto Shape Color Combo Tanimoto Tanimoto TMP 131 Thymidine-5'-phosphate 0.85 0.65 0.2 FAD 175 Flavin-adenine dinucleotide 0.85 0.68 0.17 NAI 53 1,4-Dihydronicotinamide 0.83 0.64 0.2 adenine dinucleotide IDP 892 Inosine-5'-diphosphate 0.82 0.66 0.17 GDP 359 Guanosine-5'-diphosphate 0.82 0.57 0.25 TTP 1 Thymidine-5'-triphosphate 0.82 0.61 0.21 UMP 133 2'-Deoxyuridine 5'- 0.82 0.61 0.2 monophosphate ATP 2943 Adenosine-5'-triphosphate 0.81 0.59 0.22 CDP 162 Cytidine-5'-diphosphate 0.81 0.62 0.19 CMP 11 Adenosine-3',5'- 0.81 0.67 0.13 cyclicmonophosphate UDP 112 Uridine-5'-diphosphate 0.79 0.61 0.18 GMP 10 Guanosine 0.79 0.61 0.18 UTP 1494 Uridine 5'-triphosphate 0.79 0.62 0.17 ADP 130 Adenosine-5'-diphosphate 0.78 0.58 0.21 GTP 109 Guanosine-5'-triphosphate 0.78 0.51 0.27

129

Figure 4.2. 2D ROCS predicted ligand structures. The comparison of the RsCcO bound deoxycholate and Binding MOAD ligands by the ROCS method suggested (A) testosterone hemisuccinate, (B) progesterone, (C) hydrocortisone, (D) cholesterol, (E) aldosterone, (F) retinoic acid, and (G) thyroid hormone were highly similar. Ligand structures were depicted using eMolecules (Solano Beach, CA) molecular structure search tool.

130

Figure 4.2 (cont’d)

131

Figure 4.3. ROCS aligned crystallographic deoxycholate and predicted ligands. The RsCcO bound deoxycholate and Binding MOAD steroids, retinoic acid, thyroid hormone, nucleotides, and flavins were significantly similar in shape and chemistry as determined by ROCS. These top ranked ligands: (A) testosterone hemisuccinate, (B) progesterone, (C) hydrocortisone, (D) cholesterol, (E) aldosterone, (F) fusidic acid, (G) thyroid hormone, (H) retinoic acid, (I) FAD, and (K) GDP were aligned to the crystallographic deoxycholate (gray sticks).

The ROCS predicted retinoic acid and thyroid hormone analogs have been previously shown to affect RsCcO activity in vitro (Hiser, C., et al., in review and Hiser, C., unpublished data). Retinoic acid, a known regulator of lipid and steroid metabolism [32], specifically stimulated the K-pathway variant E101A at micromolar concentrations by interacting at the steroid binding site (Hiser, C., et al., in review), while micromolar concentrations of the T3 thyroid hormone inhibit both the WT and E101A RsCcO enzymes (Hiser, C., unpublished data).

The identification of known regulatory ligands that interact near the K-pathway supports the ability of ROCS to identify ligands capable of interacting at the steroid binding site.

132

It has been previously suggested that the steroid site was in fact an ADP allosteric regulatory site [20]. The similarity between DXC and abundant adenine-containing ligands and nucleotides was evaluated by the ROCS method. Thirteen different nucleotides were determined to be structurally similar enough to DXC to score two standard deviations or better than the average TanimotoCombo score across Binding MOAD (Table 4.1G). FAD and an NAD+ analog were also identified using this significance threshold (Table 4.1G).

In each of the ligand chemical classes identified by the ROCS method, the highest scoring ligands based on TanimotoCombo that were commercially available and soluble (100 μM minimal final concentration) were tested for their ability to affect CcO activity. ROCS candidate ligands that were experimentally evaluated include testosterone, estradiol, hydrocortisone, aldosterone, fusidic acid (FA), FAD, NAD+, GDP, ATP, and ADP. These results are presented and discussed in the later section: In vitro regulation of CcO by the natively abundant candidate ligands.

Chemical and geometric comparisons between the CcO steroid binding site and diverse, native protein binding sites

To predict more varied candidate ligands than those similar to DXC, the CcO steroid site was compared to diverse, native ligand binding sites in Binding MOAD using the SimSite3D method. SimSite3D compares protein binding sites in crystallographically defined proteins based on chemistry and geometry with no knowledge of the existing bound ligand or amino acid sequences. The ligands bound to sites similar to the CcO steroid binding sites may then serve as candidate ligands for CcO.

133

The SimSite3D method identified five classes of proteins, characterized by their bound ligands, based on scoring two standard deviations or better than the average SimSite3D score across Binding MOAD. These protein classes bound flavin, lipid, nicotinamide, nucleotide, and steroid ligands (Tables 4.2A-4.2E). Lipidic ligands were classified as triacylglycerol, phospholipids, sphingolipids, fatty acids, isoprenes, or any derivative of these molecules (Table

4.2 legend lists all identified lipidic ligands).

Table 4.2. SimSite3D predicted binding sites (following pages). SimSite3D analyzed protein binding sites scoring at least two standard deviations above the average SimSite3D score across Binding MOAD were selected as similar binding sites. These binding sites were grouped into dominant chemical classes and redundant protein binding sites were removed. Flavin binding sites must bind a ligand containing a flavin moiety. Lipid binding proteins bound triacylglycerol, phospholipid, sphingolipid, fatty acid, isoprenes, or any derivative of these molecules. Lipids included tristearoyl glycerol (PDB Ligand: TGL), a farnesyl diphosphate and its analog (PDB Ligands: FPP and FII), geranylgeranyl diphosphate and its analog (PDB Ligands: GRG and MGM), phosphatidylethanolamine and its analogs (PDB Ligands: PEV, PEE, and EPH), the cytochrome bc1 Q-site inhibitor pentachlorophenol (PDB Ligand: AGA), myristic acid (PDB Ligand: MYR), phosphtidylglycerol (PDB Ligand: P6L), ricinoleic acid (PDB Ligand: RCL), α- hydroxyfarnesyl phosphonic acid (PDB Ligand: HFP), palmitic acid (PDB Ligand: PLM), stearic acid (PDB Ligand: STE), and C18-ceramide (PDB Ligand: 18C). Nicotinamide binding sites must bind a ligand containing a nicotinamide moiety. Nucleotide binding sites must bind a ligand containing a nucleic acid base. Steroid binding sites must bind a ligand containing a 4- membered steroid ring system.

134

Table 4.2 (cont’d)

Table 4.2A. SimSite3D predicted flavin binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

12 1FFU Carbon monoxide dehydrogenase FAD -2.81 18 2R9Z Glutathione amide reductase FAD -2.72 22 1QO8 Flavocytochrome c3 fumarate reductase FAD -2.63 26 1E1K Adrenodoxin reductase FAD -2.61 33 1FEA Crithidia fasciculata trypanothione reductase FAD -2.56 36 1PBD P-Hydroxybenzoate hydroxylase FAD -2.54 37 1VDV Bovine milk xanthine dehydrogenase y-700 FAD -2.54 49 1XAN Human glutathione reductase FAD -2.46 54 1QX4 S127P mutant of cytochrome b5 reductase FAD -2.41 82 1JEH Yeast e3 lipoamide dehydrogenase FAD -2.31 92 1DOE 4-OH Benzoate hydroxylase FAD -2.29 94 2NVK Thioredoxin reductase FAD -2.28 100 1XKK Epidermal growth factor receptor FMM -2.27 107 2FJA Adenosine 5'-phosphosulfate reductase FAD -2.24 125 1CC2 Cholesterol oxidase FAD -2.21 138 3CGD Bacillus anthracis coenzyme a-disulfide FAD -2.18 reductase 179 1DS7 FMN-dependent nitroreductase FMN -2.08 190 3MDD Acyl-coA dehydrogenase FAD -2.06 195 1AN9 D-amino acid oxidase FAD -2.05 202 2FJD Adenosine-5-phosphosulfate reductase SFD -2.04 207 2J07 Thermus DNA photolyase FAD -2.03 214 2Q0K Oxidized thioredoxin reductase FAD -2.02 226 2VNK Ferredoxin-NADP(H) reductase FAD -2.00 231 2A8X Lipoamide dehydrogenase FAD -2.00

135

Table 4.2 (cont’d)

Table 4.2B. SimSite3D predicted lipidic binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

5 2DYS Bovine heart cytochrome c oxidase TGL -2.99 27 1JCS Rat protein farnesyltransferase FII -2.61 43 1S63 Human protein farnesyltransferase FPP -2.50 51 3DST Rab GGTase GRG -2.45 55 2E8V S. cerevisiae geranylgeranyl pyrophosphate GRG -2.41 synthase 63 2E2X Sec14 homology module of neurofibromin PEV -2.37 75 1Q16 Nitrate reductase a AGA -2.34 77 1ND2 Rhinovirus 16 MYR -2.33 103 1N4Q Protein geranylgeranyltransferase type-I MGM -2.26 106 1YOK NR5 orphan receptors sf-1 and lrh-1 P6L -2.25 112 2R40 Ecdysone receptor EPH -2.24 135 1O5M Tricyclic farnesyl protein transferase FPP -2.18 136 1FK7 Maize lipid-transfer protein complexes RCL -2.18 148 1QBQ Rat farnesyl protein transferase HFP -2.16 151 3BDQ Sterol carrier protein-2 like-2 PLM -2.15 168 1PPJ Bovine cytochrome bc1 complex PEE -2.11 178 1FK4 Maize lipid-transfer protein complexes STE -2.09 181 2E3R Cert start domain 18C -2.08 185 1FK2 Maize lipid-transfer protein complexes MYR -2.07 189 1LN1 Human phosphatidylcholine transfer protein DLP -2.06 215 3DSU Rab geranylgeranyl transferase FPP -2.02

136

Table 4.2 (cont’d)

Table 4.2C. SimSite3D predicted nicotinamide binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

1 1NNU Enoyl-acyl-carrier-protein reductase NAD -3.14 2 1KY8 Non-phosphorylating glyceraldehyde-3- NAP -3.13 phosphate dehydrogenase 4 1IY8 Levodione reductase NAD -3.02 6 2JL1 Citrobacter sp. triphenylmethane reductase NAP -2.99 7 1E6W 3-Hydroxyacyl-coA dehydrogenase NAD -2.98 8 1AHI 7 α-Hydroxysteroid dehydrogenase NAD -2.93 10 1VHD Iron containing alcohol dehydrogenase NAP -2.85 13 1D8A E. coli enoyl reductase NAD -2.79 14 2ID2 CoA-independent aldhs NAP -2.76 16 1H5Q Mannitol dehydrogenase NAP -2.74 19 1M8F M. thermoautotrophicum nicotinamide NAD -2.72 mononucleotide adenylyltransferase R11A 20 1V35 Eoyl-acp reductase NAI -2.71 21 1PS9 E. coli 2,4-dienoyl CoA reductase NAP -2.69 24 2IMP Lactaldehyde dehydrogenase NAI -2.62 25 1G0O Trihydroxynaphthalene reductase NDP -2.61 29 1X1T d-3-Hydroxybutyrate dehydrogenase NAD -2.59 38 2BD0 Chlorobium tepidum sepiapterin reductase NAP -2.53 42 1QRR SQD1 protein NAD -2.50 45 1T90 Methylmalonate semialdehyde dehydrogenase NAD -2.48 47 1GZ6 (3R)-Hydroxyacyl-coA dehydrogenase fragment NAI -2.48 61 1IB0 NADH-dependent cytochrome b5 reductase NAD -2.38 65 1XHL Tropinone reductase-II NDP -2.36 66 1YVE Acetohydroxy acid isomeroreductase NDP -2.36 67 1CD2 Pneumocystis carinii dihydrofolate reductase NAP -2.36 69 1J3I Plasmodium falciparum dihydrofolate NDP -2.36 reductase-thymidylate synthase 70 1OJZ C3stau2 NAD -2.36

137

Table 4.2 (cont’d)

Table 4.2C Continued. SimSite3D predicted nicotinamide binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

73 1A71 Horse liver alcohol dehydrogenase NAD -2.35 88 1EK6 Human UDP-galactose 4-epimerase NAI -2.30 95 1KVQ UDP-galactose 4-epimerase NAD -2.28 97 1HE2 Human biliverdin IX β reductase NAP -2.28 98 1GJR Ferredoxin-NADP+ reductase NAP -2.28 111 2GZ1 Aspartate semialdehyde dehydrogenase NAP -2.24 127 1B15 Alcohol dehydrogenase NAE -2.21 128 1B2L Alcohol dehydrogenase NDC -2.21 130 2JD1 1-Deoxy-D-xylulose 5-phosphate NDP -2.20 reductoisomerase 133 1Y7T NAD(H)-depenent malate dehydrogenase NDP -2.19 140 1WNB E. coli YDCW aldehyde dehydrogenase NAD -2.17 143 1JAY Coenzyme F420H2:NADP+ oxidoreductase NAP -2.16 144 1RKX CDP-D-Glucose 4,6-dehydratase NAD -2.16 157 2DTE T. acidophilum aldohexose dehydrogenase NAI -2.13 159 1UXT Non-phosphorylating glyceraldehyde-3- NAD -2.13 phosphate dehydrogenase (gapn) 161 1IE3 R153C E. coli malate dehydrogenase NAD -2.12 169 2FZI DHFR complexes NAP -2.11 171 1RPN GDP-D-Mannose 4,6-dehydratase NDP -2.10 174 3CE6 M. tuberculosis S-adenosyl-L-homocysteine NAD -2.10 hydrolase 192 1E7W Leishmania pteridine reductase NDP -2.06 197 1DIG Human methylenetetrahydrofolate NAP -2.05 dehydrogenase / cyclohydrolase 200 1DYR P. carinii dihydrofolate reductase NDP -2.05 206 2DFV Hyperthermophilic threonine dehydrogenase NAD -2.03 212 1B16 Alcohol dehydrogenase NAQ -2.02 217 2DT5 TTHA1657 (at-rich dna-binding protein) NAD -2.02

138

Table 4.2 (cont’d)

Table 4.2D. SimSite3D predicted top nucleotide binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

9 2PTR E. coli adenylosuccinate lyase mutant H171A 2SA -2.88 11 2GQR E. coli phosphoribosylaminoimidazole ADP -2.84 succinocarboxamide synthetase 28 2A14 Human indolethylamine N-methyltransferase SAH -2.60 31 1ATP CAMP-Dependent protein kinase ATP -2.57 39 1S7N Ribosomal L7/L12 α-N-protein COA -2.52 acetyltransferase 44 3ELW Wesselsbron virus methyltransferase SAM -2.49 53 2HK9 Shikimate dehydrogenase ATR -2.41 76 2H7C Human carboxylesterase COA -2.34 86 2PX5 Murray Valley Encephalitis virus NS5 2'-O SAH -2.30 methyltransferase domain 90 3TMK Yeast thymidylate kinase T5A -2.29 91 1QZ5 Rabbit actin ATP -2.29 99 2Q4V Thialysine n-acetyltransferase (SSAT2) ACO -2.27 101 1Q99 S. cerevisiae SR protein kinase ANP -2.27 102 2FF3 WH2/β-Thymosin motif-containing proteins ATP -2.26 105 1L3R Transition state mimic of the catalytic subunit ADP -2.26 of cAMP-dependent protein kinase 109 1O6K PKB kinase domain S474D ANP -2.24 129 1V0P P. falciparum protein kinase PVB -2.21 137 3MCT Vaccinia methyltransferase VP39 SAH -2.18 147 2JI8 Oxalyl-coA decarboxylase ADP -2.16 150 2BZG Thiopurine S-methyltransferase SAH -2.15 176 1AV5 Protein kinase C interacting protein (PKCI) AP2 -2.09 183 1DR2 Chicken liver dihydrofolate reductase TAP -2.08 185 1DTP Catalytic domain of diphtheria toxin APU -2.08

139

Table 4.2 (cont’d)

Table 4.2D Continued. SimSite3D predicted top nucleotide binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

185 1DTP Catalytic domain of diphtheria toxin APU -2.08 187 2VBQ Aminoglycoside N-acetyltransferases BSJ -2.07 192 2CGJ L-Rhamnulose kinase ADP -2.06 197 2A94 P. falciparum lactate dehydrogenase AP0 -2.05 204 1NW5 β Class N6-adenine DNA methyltransferase SAM -2.04 210 1E8X Phoshoinositide 3-kinase ATP -2.02 224 1M2G Sir2 homologue APR -2.01 78 1S4O Yeast α 1,2-mannosyltransferase Kre2p/Mnt1p GDP -2.33 96 1R27 Catalytic and electron-transfer subunits MGD -2.28 (NarGH) of the integral membrane protein, respiratory nitrate reductase (Nar) 104 1RDS Ribonuclease MS GPC -2.26 122 2G9X Thr 160 phosphorylated CDK2/cyclin A NU5 -2.22 153 1GUA Human RAP1A GNP -2.14 209 2OM2 Human G[α]i1 GDP -2.03 214 3DAG [Fe]-Hydrogenase holoenzyme (HMD) FEG -2.02 217 1H1S Human THR160-phospho CDK2/cyclin A 4SP -2.02 57 2BZN Guanosine monophosphate reductase 2 IMP -2.40 62 1G93 α-1,3-Galactosyltransferase UPG -2.38 72 1UQT Trehalose-6-phosphate U2F -2.35 143 1I2B Mutant T145A SQD1 protein complex UPG -2.16 189 2NOM Terminal uridylyl transferase 4 DUT -2.06 219 2IV7 WAAG Glycosyltransferase UDP -2.01

140

Table 4.2 (cont’d)

Table 4.2E. SimSite3D predicted steroid binding protein matches Rank PDB Protein Ligand Code SimSite3D Score

202 2AA2 Mineralocorticoid Receptor AS4 -2.12 210 1ZHY Yeast oxysterol binding protein Osh4 CLR -2.11 223 2A3I Mineralocorticoid Receptor C0R -2.09 268 2Q1H Corticoid Receptor AS4 -2.02 286 1F3B Murine class alpha glutathione S-transferase GBX -2.01

The chemically and geometrically similar diverse protein binding sites were aligned by

SimSite3D to the CcO steroid site based on five conserved CcO residues: P315 in subunit I and

H96, S98, E101, and T105 in subunit II (Figure 4.4). The protein binding site classes that bound flavins, nicotinamide, or nucleotide ligands contained chemical groups analogous to the entire steroid site as at least 50% of these binding sites matched all five CcO residues (Table 4.3 and

Figure 4.4). Protein binding sites that bound lipidic or steroidal ligands were less similar to the steroid site, as these sites matched the H96 polar region and the P315 hydrophobic crevice better than the other CcO residues (Table 4.3 and Figure 4.4).

141

Figure 4.4. SimSite3D predicted analogous chemical points between the RsCcO steroid binding site and sites found in Binding MOAD. Analogous chemical groups are used to evaluate binding site similarities. (A) RsCcO was determined to be highly similar to nicotiamide, nucleotide, and flavin binding sites as the proteins matched RsCcO subunit I P315 and subunit II H96, S98, E101, and T105. (B) RsCcO was also determined to be similar to steroid and lipid binding sites as these sites match the steroid binding site H96 and the P315 residues.

142

Table 4.3. Percentage of protein binding site matches between the CcO steroid binding site and diverse Binding MOAD proteins. SimSite3D aligns ligand binding sites based on shape and chemical similarity by aligning chemically labeled template points. The aligned points between the CcO steroid binding site and diverse proteins in Binding MOAD may then be mapped back to the steroid binding site residues. Five CcO residues were identified as analogous to diverse sites in all high scoring alignments. The percentage of matches between these residues and diverse protein binding sites were tabulated and divided based on ligand chemical class.

CcO residue Nucleotide sites Nicotinamide sites Flavin sites Steroid sites Lipidic sites P315 52 63 87 60 36 H96 70 91 100 60 100 S98 54 67 65 60 29 E101 66 100 50 0 29 T105 96 100 65 20 19

The SimSite3D predicted sites include those that bind phospholipids and isoprenes which have been previously shown to affect CcO activity (Hiser, C., et al., in review). It has been observed that delipidation of CcO during purification inhibits the enzyme [33-34]. This inhibition may be overcome by the addition of exogenous lipids, including asolectin [35].

Although the location(s) of lipid binding and mechanism of the inhibition by delipidation are not completely understood, the addition of a lyso-lipid specifically affects the K-pathway E101A variant (Hiser, C., et al., in review). Additionally, micromolar concentrations of the isoprenoid phytanic acid greatly stimulate the E101A variant and appear to bind at the entrance of the K- pathway, to displace inhibitory detergent, and to rescue the E101A enzyme (Hiser, C., et al., in review).

In each of the protein binding site classes identified by the SimSite3D method, the highest scoring bound ligands that were commercially available and soluble (100 μM minimal final

143 concentration) were tested for their ability to affect CcO activity. SimSite3D candidate ligands that were experimentally evaluated include hydrocortisone, aldosterone, FAD, NAD+, GDP, ATP,

ADP, and farnesyl diphosphate (FPP). Striking overlap with the ROCS top ligands is observed as both methods identified testosterone, estradiol, hydrocortisone, aldosterone, FA, FAD, NAD+,

GDP, ATP, and ADP.

Predicted interactions between candidate ligands and the CcO steroid binding site

The ROCS method of ligand comparison and SimSite3D method of protein binding site comparison identified natively abundant candidate ligands that may affect CcO by specifically interacting at the steroid binding site. However, neither of these methods provide any information about the interaction between the candidate ligands and CcO steroid binding site.

SimSite3D attempts to assess these interactions by aligning the protein binding sites alone but also including the ligand in the same frame of reference. In this way, the ligand binding orientation can be predicted but is optimized for the native binding site and not the CcO steroid binding site. To understand the binding orientation and relative energetic favorability of the natively abundant candidate molecules, these ligands were docked into the RsCcO steroid binding site.

The SLIDE method was used to dock ADP, ATP, GDP, FAD, and NAD+ into the CcO steroid binding site. The top scoring ligand dockings were selected based on scoring two standard deviations or better than the average SLIDE OrientScore across all low energy conformations of a single ligand. The top scoring docked ADP, ATP, GDP, FAD, and NAD+ ligands were analyzed to determine their binding orientations, relative binding affinities, and

144 interaction with the CcO binding site residues.

Single ligand binding orientations were identified for ADP, ATP, and NAD+, while two possible binding orientations were identified for GDP and FAD ligands (Table 4.4). GDP was docked such that the phosphate groups interact at the entrance of the K-pathway and the nitrogenous base is either located within the binding pocket or solvent exposed. FAD was docked such that either its nitrogenous base or flavin moiety was located in the steroid binding site. The two binding orientations for GDP and FAD ligands are both energetically favorable as determined by SLIDE OrientScore (Table 4.4).

Table 4.4. SLIDE docked ligands and their protein interactions. The SLIDE method was used to dock riboside candidate ligands into the CcO steroid binding site. The top scoring binding orientations were characterized as scoring two standard deviations or better than the average SLIDE OrientScore for an individual molecule. These top scoring dockings were divided based on the bound ligand fragment where either the nucleic acid base, flavin moiety, or neither was bound. The top scoring ligands interacted with eight residues in the steroid binding site. The percentage of the top scoring binding mode interactions with these residues was calculated.

Percentage of docking ligand inactions Ligand Binding mode Orient Affi P315 Y318 H96 S98 E101 I102 W104 T105 Score Score ADP Adenine Bound -8.0 -7.0 100 33 100 100 100 100 33 100 ATP Adenine Bound -8.7 -7.9 88 77 100 88 100 85 62 77 GDP Guanine Bound -8.1 -7.1 90 90 90 40 80 50 80 60 Guanine Soluble -8.1 -7.5 100 14 100 71 79 0 0 7 FAD Adenine Bound -9.7 -10 83 50 83 83 100 67 0 67 Flavin Bound -9.2 -9.9 100 82 91 64 100 64 41 55 + NAD Adenine Bound -9.3 -8.6 100 54 100 87 97 79 36 90

145

The relative affinities of the docked ligands were also ranked based on SLIDE AffiScore.

FAD in either orientation has the greatest affinity, NAD+ has intermediate affinity, and ADP, ATP, and GDP have the lowest affinity (Table 4.4). However, binding of all these ligands to the steroid site were sterically allowed and energetically favorable.

Finally, in all cases where the steroid binding site was fully occupied, the docked ligands interacted with eight steroid binding site residues, including the five residues identified by

SimSite3D. The eight interacting residues included P315 and Y318 in subunit I and H96, S98,

E101, I102, W104, and T105 in subunit II (Table 4.4). The majority of the top ligand orientations interacted with P315, H96, S98, and E101. P315, E101, I012, and T105 are all conserved and have greater than 50% sequence identity between more than 250 diverse organisms (Chapter 3).

Interestingly, both the SimSite3D protein binding site comparison and SLIDE docking methods identified a very similar binding orientation for all the analyzed riboside ligands (Figure

4.5). These ligands bound to the CcO steroid binding site via interactions of the ligand phosphate groups with H96 and E101, the ligand ribose with T105, and the nitrogenous base or other cyclic group with the P315 hydrophobic pocket. The identification of a single binding orientation using two independent methods and between structurally similar ligands supports the validity of the interaction of ADP, ATP, GDP, FAD, and NAD+ in the CcO steroid binding site using this orientation. Additionally, the identification of the same critical residues by two independent methods suggests that these interactions are important for ligand binding and allosteric regulation and may guide mutagenesis in subsequent studies.

146

Figure 4.5 (following page). Similar binding modes were predicted by SimSite3D aligned binding site and SLIDE docked ligands. SimSite3D aligned binding sites and SLIDE small molecule docking predicted similar binding modes for nucleotides, NAD+, and FAD. In all cases, the di- or triphosphate groups are bound at near the entrance of the K-pathway, specifically near H96 and E101. The ribose group was bound in the center of the steroid binding site and interacted with T105. Finally, the nucleic acid base, nicotinamide, or flavin moiety was located furthest into the membrane domain, often interacting near P315. The subunit I (green ribbon) /subunit II (cyan ribbon) interface is depicted with interacting amino acid residues in stick diagram. For the SimSite3D aligned sites, these residues are P315, H96, S98, E101, and T105; while the SLIDE docked sites, these residues are P315, Y318, H96, S98, E101, I102, W104, and T105. Binding modes determined by SimSite3D: (A) ADP (magenta sticks), (B) ATP (salmon sticks), (C) GDP (orange sticks), (G) NAD+ (blue sticks) with its adenine base bound, (H) NAD+ (blue sticks) with its nicotinamide moiety bound, (I) FAD (purple sticks) with its adenine base bound, and (J) FAD (purple sticks) with its flavin moiety bound. Binding modes determined by SLIDE: (D) ADP (magenta sticks), (E) ATP (salmon sticks), (F) GDP (orange sticks), (K) NAD+ (blue sticks) with its adenine base bound, (L) NAD+ (blue sticks) with its nicotinamide moiety bound, (M) FAD (purple sticks) with its adenine base bound, and (N) FAD (purple sticks) with its flavin moiety bound.

147

Figure 4.5 (cont’d)

148

The CcO steroid binding site is an amphipathic binding crevice in the membrane domain and is unlike other observed nucleotide, flavin, and nicotinamide binding sites. The binding of adenine-containing nucleotides either as coenzymes, cofactors, substrates, or allosteric regulators has been observed to be dependent on a three residue loop where the distal ends of the loop make backbone hydrogen bonds to the adenine ring while the center of the loop is highly hydrophobic

[36]. In the steroid binding site, the P315 backbone carbonyl group may interact with the adenine base using its N6 amine group, similar to other adenine-specific binding motifs [36].

Additionally, T318 may stack with the nitrogenous base as seen in purine binding proteins [36].

However, no other similarities are observed. FAD is thought to bind to two consensus motifs:

GXGGXXG with a glutamate or aspartate residue interacting with the FAD ribose in a βαβ structure [37] or a TXXXXhφhhGD [38-39] motif where X is any residue, h is hydrophobic, and

φ is aromatic. The only similarity observed in the CcO steroid binding site is the presence of two conserved glycines, G324 and G327, that create a GXXG motif. NAD+ and NADP ligands bind to dehydrogenases and reductases using a GXXXGXG motif for phosphate interactions and

YXXK for ribose interactions [40-42]. Neither of these motifs is observed in the CcO steroid binding site. Based on sequence motifs from soluble proteins, the CcO steroid site may interact with nucleotides, flavins, and nicotinamides in a unique manner required for the hydrophobic environment, where part of the binding interaction must be mediated by lipid from the bilayer.

In vitro regulation of CcO by the natively abundant candidate ligands

ROCS ligand comparisons and SimSite3D protein binding site comparisons identified an overlapping but diverse set of natively abundant, steroid binding site candidate ligands. The top scoring, commercially available ligands with high solubility were selected for experimental

149 confirmation studies. These ligands included testosterone, estradiol, hydrocortisone, aldosterone,

FA, FAD, NAD+, GDP, ATP, ADP and farnesyl diphosphate (FPP).

The oxygen consumption rate, or turnover number, of CcO was determined polarographically using an oxygen-sensitive electrode. The variation in CcO WT and the E101A variant turnover was initially assessed in the presence and absence of the maximal concentration of each candidate ligand. The E101A protein was studied along with the WT as this enzyme has been shown to be highly sensitive to ligands binding at the steroid site either due to chemical rescue or displacement of inhibitory detergent [25, 30, Chapter 3, and Hiser, C., et al., in review].

Standard (0.06% LM) and low (0.01% LM) detergent conditions were also investigated as the steroid binding site has been suggested to be partially occluded by an inhibitory LM detergent molecule in the case of the E101A variant (Chapter 3 and Hiser, C., et al., in review).

The hydrophobic steroids testosterone, estradiol, hydrocortisone, and aldosterone were solubilized in ethanol. The activity of WT and E101A CcO were compared in the presence of these compounds or the same volume of ethanol. These ligands failed to affect either the WT or

E101A enzyme in either standard or low detergent conditions (Figures 4.6 and 4.7). The lack of effect may be explained as a result of none of these steroids binding to the steroid binding site at the concentrations tested (100-700 μM). In order to test these ligands at greater concentrations, an alternative solvent system with greater steroid solubility and with minimal affects on CcO must be identified. However, binding at greater concentrations is unlikely to have physiological significance.

150

Figure 4.6. Ethanol-soluble ligand additives do not affect WT CcO in standard or low detergent conditions. The five best scoring steroids from ROCS ligand comparisons and SimSite3D protein binding site comparisons were tested at maximal concentration for their ability to affect WT CcO activity in both (A) standard (0.06% LM) and (B) low (0.01% LM) detergent conditions. In all cases, steroid ligands failed to affect WT CcO as compared to their solvent, ethanol, alone.

151

Figure 4.6 (cont’d)

152

Figure 4.7. Ethanol-soluble ligand additives do not affect the E101A mutant in standard or low detergent conditions. The five best scoring steroids from ROCS ligand comparisons and SimSite3D protein binding site comparisons were tested at maximal concentration for their ability to affect the E101A variant activity in both (A) standard (0.06% LM) and (B) low (0.01% LM) detergent conditions. In all cases, steroid ligands failed to affect the E101A variant as compared to their solvent, ethanol, alone.

153

Figure 4.7 (cont’d)

154

The remaining hydrophilic candidate ligands were solublized in water to millimolar concentrations, which are physiological, and compared to the activity of WT and E101A CcO alone. In the WT enzyme, GDP and an ATP analog [adenosine 5′-(β,γ-imido) triphosphate] mildly inhibited while the steroids FA and DXC strongly inhibited activity in both standard and low detergent conditions (Figure 4.8). In the E101A variant, all of the hydrophilic candidate ligands were able to mildly stimulate activity in standard detergent conditions (Figure 4.9), suggesting that these ligands weakly interact with the CcO steroid binding site and are capable of displacing a proportion of the inhibitory LM. GDP, FAD, and FA stimulate while an ATP analog inhibits the E101A variant in low detergent (Figure 4.10), suggesting that these ligands are eliciting a specific effect unrelated to detergent inhibition. Since GDP, an ATP analog, FAD, and

FA were capable of eliciting an effect on the CcO WT and/or E101A variant, these ligands were further studied by ligand titrations.

Figure 4.8. GDP and an ATP analog mildly inhibit while fusidic acid and deoxycholate strongly inhibit WT CcO in both standard and low detergents (following page). In addition to steroids, candidate ligands from ROCS ligand comparisons and SimSite3D protein binding site comparisons include nucleotides. High scoring SimSite protein binding site classes also included flavin, nicotinamide, and phosphorylated isoprene binding sites. The best scoring ligands from these classes were tested at maximal concentration for their ability to affect WT CcO activity in both (A) standard (0.06% LM) and (B) low (0.01% LM) detergent conditions. Water-soluble predicted steroidal ligands, deoxycholate and fusidic acid, were also tested. GDP and an ATP analog were able to mildly inhibit WT CcO, while the water-soluble steroidal ligands deoxycholate and fusidic acid were able to inhibit activity strongly.

155

F igure 4.8 (cont’d)

156

Figure 4.8 (cont’d)

157

Figure 4.9. Water-soluble ligands are able to stimulate the E101A mutant in standard detergent. In addition to steroids, candidate ligands from ROCS ligand comparisons and SimSite3D protein binding site comparisons include nucleotides. High scoring SimSite protein binding site classes also included flavin, nicotinamide, and phosphorylated isoprene binding sites. The best scoring ligands from these classes were tested at maximal concentration for their ability to affect the E101A mutant in standard (0.06% LM) detergent conditions. Water-soluble predicted steroidal ligands, deoxycholate and fusidic acid, were also tested. In all cases, the predicted ligands were able to slightly stimulate E101A, potentially due to the displacement of inhibitory LM.

158

Figure 4.10. GDP, FAD, and fusidic acid stimulate while an ATP analog inhibits the E101A mutant in low detergent. In addition to steroids, candidate ligands from ROCS ligand comparisons and SimSite3D protein binding site comparisons include nucleotides. High scoring SimSite protein binding site classes also included flavin, nicotinamide, and phosphorylated isoprene binding sites. The best scoring ligands from these classes were tested at maximal concentration for their ability to affect the E101A mutant in low (0.01% LM) detergent conditions. Water-soluble predicted steroidal ligands, deoxycholate and fusidic acid, were also tested. GDP, FAD, and fusidic acid are all able to stimulate E101, while an ATP analog inhibits activity.

The ATP analog mildly inhibited WT and mildly stimulated the E101A variant in both standard and low detergents conditions at a single, 5 mM concentration. This analog was titrated from 0-10 mM, conditions similar to the concentrations previously observed to regulate mammalian CcO in vitro [12, 14-16, 18-19] and found in isolated mitochondria [43]. The ligand

159 stimulated WT at low concentrations and inhibited the enzyme at increasing concentrations

(Figure 4.11). This pattern of stimulation followed by inhibition has been observed previously upon addition of amphipathic ligands to the E101A variant in standard detergent conditions

(Hiser, C., et al., in review). The behavior had been hypothesized to be a result of ligands binding to the CcO steroid binding site, displacing inhibitory detergent, followed by ligand- specific inhibition. However, this behavior had been previously suggested to be specific for the

E101A variant and was not observed in the WT enzyme. ATP bound to the steroid binding site may extend further into the membrane environment than the previously studied ligands, allowing it to interact with and displace detergents. Displacement of the inhibitory detergent and subsequent allosteric inhibition of the K-pathway is consistent with the ATP analog titration.

This same ATP analog was only able to stimulate the E101A variant mildly to an activity level similar to that of E101A in minimal detergent (Figure 4.12). This behavior may be explained by the inability of the ATP analog to bind to the modified steroid site in the variant form, or the inability of the bound ligand to displace the inhibitory detergent.

160

Figure 4.11. An ATP analog stimulates then inhibits WT CcO in both standard and low detergent (previous page). Millimolar concentrations of an ATP analog were tested for their ability to affect WT CcO activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of the analog stimulated and then inhibited the enzyme, suggesting that ATP may compete with LM in the WT enzyme and then may allosterically inhibit by interacting at the steroid binding site.

161

Figure 4.12. An ATP analog mildly stimulates the E101A mutant in both standard and low detergent. Millimolar concentrations of an ATP analog were tested for their ability to affect the E101A mutant activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of the analog slightly stimulated the mutant, suggesting that ATP may compete with LM but without the E101A carboxyl group cannot optimally bind to the steroid binding site.

GDP mildly inhibited the WT enzyme in both standard and low detergent concentrations and greatly stimulated the E101A variant in low detergent conditions at a single, 5 mM concentration. Again, this ligand was titrated from 0-10 mM. In both detergent conditions, GDP only inhibited WT CcO but to nearly the same activity as the ATP analog, approximately 800 electrons per second per CcO (Figure 4.13). This suggests that the smaller GDP ligand is unable to extend further into the membrane environment to displace inhibitory detergent but may itself allosterically inhibit the enzyme to a similar level as ATP.

162

Figure 4.13. GDP inhibits WT CcO in both standard and low detergent. Millimolar concentrations of GTP were tested for their ability to affect WT CcO activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of GDP inhibited the enzyme, suggesting that GDP does not interact with LM as strongly as ATP but still allosterically inhibits by interacting at the steroid binding site.

GDP was able to stimulate the E101A variant, specifically in low detergent concentration conditions (Figure 4.14). This detergent-dependent activation is again consistent with the inability of the smaller GDP to extend into the membrane environment to displace inhibitory detergent. However, the activation of the E101A variant is less clear. Potential mechanisms for activation by previously observed lipidic carboxylates have been suggested to include chemical rescue by correctly placed negatively charged functional groups or the reorganization of K- pathway entry water molecules (Hiser, C., et al., in review). Either of these mechanisms are also reasonable for GDP activation of E101A. 163

Figure 4.14. GDP mildly stimulates the E101A mutant in both standard and low detergent. Millimolar concentrations of GDP were tested for their ability to affect the E101A mutant activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of the analog stimulated the mutant, suggesting that GDP may allosterically activate by interacting at the steroid binding site but cannot displace LM as well as ATP.

FA is a potent activator and inhibitor of WT and the E101A enzyme in both standard and low detergent conditions (Figures 4.15 and 4.16). For both enzymes, this biphasic behavior is highly similar to DXC titrations [25, Hiser, C., et al., in review]. These steroids are able to compete with inhibitory detergent, in the case of the E101A protein, and at higher concentrations are able to inhibit the WT and E101A enzymes. Interestingly, FA binds to the same protein binding site as the porphyrin, biliverdin, in serum albumin [44] and causes jaundice [45].

Porphyrins such as protoporphyrin IX and bilirubin have been shown to activate and inhibit CcO by binding to the steroid site, showing competition with bile acids (Hiser, C., unpublished data).

164

Additionally, the use of FA as an antibiotic is not recommended in combination with antagonistic quinol analog antibiotics [46]. Therefore, it may be reasonable for the steroid binding site to be regulated by native porphyrin or quinol molecules. Future studies investigating these possibilities may be aided with the use of the highly soluble FA. The lack of identification of porphyrin and quinol ligands using ROCS and SimSite3D may be understood as these methods compare the RsCcO steroid binding site to Binding MOAD, which contains no porphyrin or quinol ligands.

Figure 4.15. Fusidic acid inhibits WT CcO in both standard and low detergent. Millimolar concentrations of fusidic acid were tested for their ability to affect WT CcO activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of fusidic acid inhibited the enzyme, suggesting that this steroidal antibiotic may have CcO inhibitory side effects.

165

Figure 4.16. Fusidic acid stimulates then inhibits the E101A mutant in both standard and low detergent. Below one millimolar concentrations of fusidic acid stimulated, while increasing concentrations of the steroid analog inhibited the mutant, when tested for its affect on the E101A mutant activity in standard (0.06% LM) and low detergent (0.01% LM).This steroidal antibiotic with a carboxylic group may be effective in displacing LM from the E101A mutant and rescuing activity before exerting inhibitory effects at the steroid binding site.

FAD appears to be a mild inhibitor of WT CcO and a potent stimulator of the E101A CcO in both standard and low concentrations of detergent conditions (Figures 4.17 and 4.18). The inhibition of WT CcO in the presence of FAD is similar to that of the nucleotide titrations, where the activity is of the protein is reduced by approximately 20-30% in the presence of milimolar concentrations of ligand. Based on the SLIDE docking results, FAD may bind to the steroid binding site in either a flavin bound orientation or an adenine bound orientation. Therefore, the

166

FAD inhibition could be mimicking the interactions natively used for adenine nucleotide inhibion. Alternatively, flavinoid ligands may affect activity at the steroid binding site. To distinguish between these hypotheses, the flavanoid ligand FMN that does not contain a nitrogenous base may be tested for its ability to affect WT CcO activity in vitro. Additionally, the importance of the flavoinoid moiety is supported based on FAD’s affects on the CcO variant.

The E101A variant was stimulated to nearly WT protein activity upon additional of FAD (Figure

4.18), suggesting that FAD effectively displaces detergent and nearly restores the native entrance of the K-pathway. This stimulation was not observed upon the addition of nucleotides, suggesting that flavanoids may have a specific affect on this site.

Figure 4.17. FAD mildly inhibits WT CcO in both standard and low detergent. Millimolar concentrations of FAD were tested for their ability to affect WT CcO activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of FAD slightly inhibited the activity of the enzyme in a similar manner as observed with nucleotide additives.

167

Figure 4.18. FAD stimulates the E101A mutant in both standard and low detergent. Millimolar concentrations of FAD were tested for their ability to affect the E101A mutant activity in standard (0.06% LM) and low detergent (0.01% LM). Increasing concentrations of FAD stimulated the enzyme to nearly WT activity levels and were effective in displacing LM from the steroid binding site.

ROCS ligand comparisons, SimSite3D protein binding site comparisons, and SLIDE dockings identified candidate ligands specific for the CcO steroid binding site with potential ligand binding orientations. In vitro confirmation studies supported the regulation of CcO by the natively abundant ATP, GDP, and FAD molecules. Also implicated are porphyrins and quinols as potential regulators, as FA is found to strongly interact not only with CcO but also with binding sites where porphyrin and quinol bind. The feedback inhibition of CcO by ATP has been previously reported but was thought to be specific only for eukaryotic CcO [47]. Allosteric

168 regulation by GDP has not been previously observed. However, GDP had been reported to regulate oxidative phosphorylation and thermoregulation in mitochondria [48-50]. This regulation may be in part due to the allosteric inhibition of brown fat uncoupler proteins [51-52] but has also been suggested to abolish thyroid and corticosterone regulation of the mitochondria, both of which have been observed to regulate mammalian CcO [12, 18-19]. Allosteric regulation of CcO by FAD has also not been previously reported and remains unclear. However, flavins play a key role in the β-oxidation of fatty acids and, when reduced, are highly reactive with molecular oxygen, especially at low mitochondrial pH [53]. Therefore, it is possible that allosteric activation of CcO by flavin molecules may be a mechanism of coordinating oxidative phosphorylation and lipid metabolism or preventing the formation of FAD-catalyzed reactive oxygen species. Overall, these potential CcO effecter molecules may be specific for the steroid binding site and optimize the protein's activity based on the metabolic demands of the cell.

Allosteric regulation of CcO may be observed by varying cytochrome c substrate concentrations

The inhibition of mammalian CcO by ATP has been shown by studying the effect of the ligand as a function of substrate concentration, cytochrome c [12, 14-16, 19]. It has been suggested that the observed sigmoidal inhibition is a result of altering the affinity for cytochrome c by ATP binding on the matrix domains of nuclear encoded subunits IV or VI [18-19]. The allosteric regulation of bacterial CcO has not been observed previously, and this has been suggested to be due to the lack of nuclear encoded subunits with ATP regulatory sites [47]. As this work is the first to illustrate nucleotide effects on bacterial CcO, the candidate ligands specific for the steroid binding site were tested for their ability to affect the activity of RsCcO

169 under varying cytochrome c concentrations.

Cytochrome c titrations were preformed in the presence of the candidate ligands including testosterone, estradiol, hydrocortisone, aldosterone, FA, FAD, NAD+, GDP, ATP, and

ADP. The hydrophobic steroids testosterone, estradiol, hydrocortisone, and aldosterone failed to affect CcO under varying concentrations of cytochrome c (data not shown). Millimolar concentrations of ADP, ATP, and GDP were all able to inhibit CcO, especially at cytochrome c concentrations below 15 µM (Figure 4.19). This inhibitory behavior is represented by a sigmoidal curve, similar to the behavior observed in mammalian CcO in the presence of ATP [12,

18-19]. The steroidal ligands DXC and FA were observed to be strong inhibitors in all tested levels of substrate (Figure 4.20). However, these ligands were more effective at inhibiting CcO in a more reduced state as supported by an increased inhibition at higher cytochrome c concentrations. The predicted NAD+ and FAD cofactors were also tested, with only FAD able to inhibit CcO (Figure 4.21). The same sigmoidal behavior was observed, with approximately the same level of inhibition, as seen in the presence of nucleotides (Figures 4.19 and 4.21).

170

Figure 4.19. The nucleotides ADP, ATP, and GDP mildly inhibit WT CcO in low substrate conditions. Millimolar concentrations of nucleotides were tested for their ability to affect WT CcO under increasing concentrations of the cytochrome c substrate. In low cytochrome c conditions, ADP (♦), an ATP analog (▼), and GDP (►) inhibited the enzyme with sigmoidal behavior. This observation is consistent with the sigmoidal inhibition of bovine CcO in the presence of high concentrations of ATP, suggesting that these nucleotide effects are an effect of allosteric regulation on the CcO core subunits.

171

Figure 4.20. The steroidal inhibitors deoxycholate and fusidic acid mildly inhibit WT CcO in low substrate conditions. Millimolar concentrations of steroidal inhibitors were tested for their ability to affect WT CcO under increasing concentrations of the cytochrome c substrate. Both deoxycholate and fusidic acid inhibited the enzyme in minimal to saturating concentrations of cytochrome c with slightly stronger inhibition observed with saturating concentrations of cytochrome c.

172

+ Figure 4.21. The FAD cofactor inhibits while NAD has no affect on WT CcO in low substrate conditions. Millimolar concentrations of predicted adenine-containing cofactors were tested for their ability to affect WT CcO under varying concentrations of the cytochrome c substrate. In low cytochrome c conditions and the absence of metals (EDTA added), FAD inhibited the enzyme with sigmoidal behavior similar to the inhibition of bovine CcO in the presence of high concentrations of ATP.

The cytochrome c substrate titrations suggest that mammalian and bacterial CcO proteins are inhibited by nucleotides on their common, subunit I-III enzyme core. As the steroid binding site was specifically probed to identify these ligands and the neighboring E101A variant is highly affected by these molecules, this site may be a reasonable binding site for allosteric regulation.

However, further studies are required to determine the location and mechanism of inhibition.

173

The nucleotide inhibition of mammalian CcO was argued to be a result of negative cooperativity between cytochrome c substrate molecules on the mammalian dimer [12-13]. The bacterial CcO is believed to exist principally as a monomer, especially in detergent solubilized conditions [31]. Therefore, the sigmoidal inhibition of CcO in the presence of nucleotides at low cytochrome c concentrations cannot be explained by negative cooperativity. Additionally, cytochrome c titrations support noncompetitive inhibition by DXC at the steroid binding site

(Figure 4.22). This noncompetitive inhibition by DXC indicates regulation at the steroid binding site does not alter the affinity of CcO for its cytochrome c substrate. However, the exact mechanism of inhibition remains unclear.

Figure 4.22. Deoxycholate inhibition of E101A supports noncompetitive inhibition with cytochrome c (following page). Millimolar concentrations of deoxycholate were tested to determine if the observed inhibition affected the CcO affinity for its cytochrome c substrate. The double reciprocal plot is consistent with noncompetitive inhibition where the Km of cytochrome c is unchanged by the deoxycholate allosteric inhibition.

174

Figure 4.22 (cont’d)

Nucleotides may allosterically inhibit CcO by binding to the steroid binding site and arresting the protein in an oxidized state

Conserved, K-pathway residue variants and a number of amphipathic molecules including cholate, Triton X100, LM, and CHOBIMALT are proposed to inhibit bacterial and mammalian CcO by blocking K-pathway dependent reduction of heme aoxidized [21-23, Hiser, C., et al., in review]. The K-pathway is responsible for taking up a proton from the internal side of the membrane and supplying it to the active site during the catalytic transition from oxidized (O intermediate) to one-electron reduced (E intermediate) and then to two-electron reduced (R intermediate) (Figure 1.3) [54-55]. K-pathway inhibition affects the equilibrium between heme a

175 and a3 so as to favor a being reduced, a3 oxidized. This effect may be observed spectrally as an increased steady-state ratio of reduced heme a to reduced heme a3 [22-23, Hiser, C., et al., in review].

Occupation of the steroid binding site also has been shown to produce this effect. In all cases, it is presumed to be due to blocking the K-pathway proton uptake (Chapter 3; Hiser, C., et al., in review). Two hypotheses, not mutually exclusive, have been suggested to explain the mechanism of K-pathway blockage. Substitutions, or the binding of certain ligands to this site may disrupt or constrain the dynamic water molecules, likely required for proton conductance

(Chapter 3; Hiser, C., et al., in review). Indeed the crystal structure with DXC bound shows increased number of well-resolved, stable water molecules in the K-pathway entry.

Alternatively, side chain substitutions or ligand binding to the steroid site may restrict conformational changes of subunit I helix VIII (Chapter 3; Hiser, C., et al., in review). This region has been shown by crystallography [56], solvent accessibility [57], and intrinsic flexibility calculations [58] to be highly dynamic, especially during active site reduction. It is unclear as to the advantage of blocking the K-pathway, thereby arresting the enzyme entirely, rather than regulating activity by an alternate means such as intrinsic uncoupling of proton pumping [59].

The rationale for controlling CcO by blocking the K-pathway may be understood by considering the conservation of this pathway and its role in proton uptake. The K-pathway is the only fully conserved proton uptake pathway in heme/copper oxidases as it has been identified in all known oxidases from thermal and soil bacteria to mammals [20, 54, 60-62], whereas some of these oxidases (e.g., B-type Thermus thermophilus ba3) appear not to have conserved the D

176 proton pathway. Therefore, the regulation of the K-pathway represents both an ancient and current target. By blocking the K-pathway, the enzyme is arrested is a semi-oxidized state

(copper A, heme a reduced, copper B, heme a3 oxidized), a state in which oxygen cannot bind to the active site [55]. Without the oxygen bound, CcO remains in an innocuous state where no reactive oxygen species may be produced as a byproduct. This state has been experimentally observed and has been termed the enzyme's resting state [63].

Functional implications of arresting CcO

CcO is considered to be a key regulator of oxidative phosphorylation, contributing to maintaining the membrane potential at a level that may limit the formation of deleterious byproducts. However, the role played by slowing overall electron transfer rates is not clear.

However, the steroid binding site has been shown to be polyspecific for nucleotides, bile acids and other steroids, porphyrins, isoprenes, and diverse lipidic carboxylates (Chapter 3; Hiser, C., et al., in review), suggesting that multiple levels of regulation may be accomplished at this site.

Inhibition of bacterial CcO activity could contribute to reduction of the cytochrome c pool, potentially reducing alternative forms of cytochrome c that function in nitrogen reduction or photosynthesis. Alternatively, arresting CcO may be critical for a bacterium to switch its metabolic state, such as the transition from a low affinity terminal oxidase to a high affinity terminal oxidase or induction of photosynthesis. Additionally, eukaryotic cells may have evolved additional modes of regulation as compared to their ancestral counterparts. Inhibiting the activity of mitochondrial CcO may play a role in supplying the important signaling intermediate, cytochrome c, for initiating programmed cell death, or apoptosis.

177

Conclusions

Three diverse yet complementary computational methods have been applied in this study to predicted natively abundant ligands specific for the conserved CcO steroid binding site. The top candidates, supported by ROCS and SimSite3D methods, include steroids, adenine and guanine nucleotides, NAD+, FAD, and phosphorylated isoprenes, along with the previously known bile acids, retinoic acid, and thyroid hormone regulators. Porphyrin ligands were shown to be high affinity ligands as well (Chapter 3), but were excluded from this search due to their role as prosthetic groups. Both protein-based methods identified P315, H96, S98, E101, and

T105 as key protein contacts that could serve as targets for substitutions in further studies.

Additionally, all ribosides were predicted to bind in a single binding orientation. Consistent findings from diverse methods support the validity of candidate ligands, their interactions with the protein, and binding orientation.

In vitro oxygen consumption assays validated a number of candidate ligands based on their affects on CcO. ATP and GDP are mildly inhibitory to RsCcO while the steroidal DXC and

FA ligands are highly inhibitory. Cytochrome c titration assays indicate nucleotides inhibit CcO activity in low cytochrome c conditions, similar to the observed ATP inhibition of mammalian

CcO. This suggests that nucleotides specifically regulate by allosterically inhibiting the CcO enzyme core, potentially at the steroid binding site. Regulation of the K-pathway may provide a means of arresting CcO in a state where the active site is oxidized, has low affinity for oxygen, and thus has little potential for generating oxygen radical byproducts.

178

REFERENCES

179

REFERENCES

1. Pacelli, C., D. Latorre, T. Cocco, F. Capuano, C. Kukat, P. Seibel, and G. Villani. 2011. Tight control of mitochondrial membrane potential by cytochrome c oxidase. Mitochondrion. 11: 334-341.

2. Villani, G., and G. Attardi. 1997. In vivo control of respiration by cytochrome c oxidase in wild-type and mitochondrial DNA mutation-carrying human cells. Proc. Natl. Acad. Sci. USA. 94: 1166-1171.

3. Hüttemann, M., S. Helling, T. H. Sanderson, C. Sinkler, L. Samaviti, G. Mahapatra, A. Varughuese, G. Lu, J. Liu, L. I. Grossman, J. W. Doan, K. Marcus, and I. Lee. 2012. Regulation of mitochondrial respiration and apoptosis through cell signaling: Cytochrome c oxidase and cytochrome c in ischemia/reprefusion injury and inflamation. Biochim. Biophys. Acta.1817: 598-609.

4. Pierron, D., D. E. Wildman, M. Hüttemann, G. C. Markondapatnaikuni, S. Aras, and L. I. Grossman. 2012. Cytochrome c oxidase: Evolution of control via nuclear subunit addition. Biochim. Biophys. Acta. 1817: 590-597.

5. Das, J., S. T. Miller, and D. L. Stern. 2004. Comparison of diverse protein sequences of the nuclear-encoded subunits of cytochrome c oxidase suggest conservation of structure underlies evolving functional sites. Mol. Biol. Evol. 21: 1572-1582.

6. LaMarche, A. E. P., M. I. Abate, S. H. P. Chan, and B. L. Trumpower. 1992. Isolation and characterization of COX12, the nuclear gene for a previously unrecognized subunit of Saccharomyces cerevisiae cytochrome c oxidase. J. Biol. Chem. 276: 22473-22480.

7. Hüttemann, M., B. Kadenbach, and L. I. Grossman. 2001. Mammalian subunit IV isoforms of cytochrome c oxidase. Gene. 267: 111-123.

8. Hüttemann, M., T. R. Schmidt, and L. I. Grossman. 2003. A third form of cytochrome c oxidase subunit VII is present in mammals. Gene. 312: 95-102.

9. Grossman, L. I. and M. I. Lomax. 1997. Nuclear encoded genes for cytochrome c oxidase. Biochim. Biophys. Acta. 1352: 174-192.

10. Hüttemann, M., I. Lee, L. I. Grossman, J. W. Doan, and T. H. Sanderson. 2012. Phosphorylation of mammalian cytochrome c and cytochrome c oxidase in the regulation of cell densiny: Respiration, apoptosis, and human disease. Adv. Exp. Med. Biol. 748:

180

237-264.

11. Fang, J. K., S. K. Prabu, N. B. Sepuri, H. Raza, H. K. Anandatheerthavarada, D. Galati, J. Spear, and N. G. Avadhani. 2007. Site specific phosphorylation of cytochrome c oxidase subunits I, IVi1, and Vb in rabbit hearts subjected to ischemia/reprefusion. FEBS Lett. 581: 1302-1310.

12. Arnold, S., F. Goglia, and B. Kadenbach. 1998. 3,5-Diiodothyronine binds to subunit Va of cytochrome-c oxidase and abolishes the allosteric inhibition of respiration by ATP. Eur. J. Biochem. 252: 325-330.

13. Arnold, S., and B. Kadenbach. 1997. Cell respiration is controlled by ATP, an allosteric inhibitor of cytochrome-c oxidase. Eur. J. Biochem. 249: 350-354.

14. Bender, E., and B. Kadenbach. 2000. The allosteric ATP-inhibition of cytochrome c oxidase activity is reversibly switched on by cAMP-dependent phosphorylation. FEBS Lett. 466: 130-134.

15. Lee, I., E. Bender, B Kadenbach. 2002. Control of mitochondrial membrane potential and ROS formation by reversible phosphorylation of cytochrome c oxidase. Mol. Cell. Biochem. 234-235 (1-2): 63-70.

16. Lee, I., A. R. Salomon, S. Ficarro, I. Mathes, F. Lottspeich, L. I. Grossman, and M. Hüttemann. 2005. cAMP-dependent tyrosine phosphorylation of subunit I inhibits cytochrome c oxidase activity. J. Biol. Chem. 280(7): 6094-6100.

17. Hopper, R. K., S. Carroll, A. M. Aponte, D. T. Johnson, S. French, R. F. Shen, F. A. Witzmann, R. A. Harris, R. S. Balabon. 2006. Mitochondrial matrix phosphoproteome: Effect of extra mitochondrial calcium. Biochemistry. 45: 2524-2536.

18. Napiwotzki, J., K. Shinzawa-Itoh, S. Yoshikawa, and B. Kadenbach. 1997. ATP and ADP bind to cytochrome c oxidase and regulatte its activity. Biol. Chem. 378(9): 1013-1021.

19. Frank, V., and B. Kadenbach. 1996. Regulation of the H+/e(-) stoichiometry of cytochrome c oxidase from bovine heart by intramitochondrial ATP/ADP ratios. FEBS Lett. 382: 121- 124.

20. Tsukihara, T., H. Aoyama, E. Yamashita, T. Tomizaki, H. Yamaguchi, K. ShinzawaItoh, R. Nakashima, R. Yaono, and S. Yoshikawa. 1996. The whole structure of the 13-subunit oxidized cytochrome c oxidase at 2.8 angstrom. Science. 272: 1136-1144.

21. Vik, S. B., and R. A. Capaldi. 1977. Lipid requirements for cytochrome-c oxidase activity. Biochemistry. 16: 5755-5759.

22. Konstantinov, A. A., S. Siletsky, D. Mitchell, A. Kaulen, and R. B. Gennis. 1997. The roles of the two proton input channels in cytochrome c oxidase from Rhodobacter sphaeroides

181

probed by the effects of site-directed mutations on time-resolved electrogenic intraprotein proton transfer. Proc. Natl. Acad. Sci. USA 94: 9085-9090.

23. Antalik, M., D. Jancura, G. Palmer, and M. Fabian. 2005. A role for the protein in internal electron transfer to the catalytic center of cytochrome c oxidase. Biochemistry. 44: 14881- 14889.

24. Benson, M. L., R. D. Smith, N. A. Khazanov, B. Dimcheff, J. Beaver, P. Dresslar, J. Nerothin, and H. A. Carlson. 2007. Binding MOAD, a high-quality protein-ligand database. Nucleic Acids Res. 36: D674-D678.

25. Qin, L., D. A. Mills, C. Hiser, A. Murphree, R. M. Garavito, S. Ferguson-Miller, and J. Hosler. 2007. Crystallographic location and mutational analysis of Zn and Cd inhibitory sites and role of lipidic carboxylates in rescuing proton path mutants in cytochrome c oxidase. Biochemistry. 46: 6239-6248.

26. Sanner, M., A. J. Olsen, J. C. Spehner. 1996. Reduced surface: an efficient way to compute molecular surfaces. Biopolymers. 38(3): 305-320.

27. Schnecke, V., C. A. Swanson, E. D. Getzoff, J. A. Tainer, and L. A. Kuhn. 1998. Screening a peptidyl database for potential ligands to proteins with side-chain flexibility. Proteins. 33: 74-87.

29. Hiser, C., D. A. Mills, M. Schall, and S. Ferguson-Miller. 2001. C-terminal truncation and histidine-tagging of cytochrome c oxidase subunit II reveals the native processing site, shows involvement of the C-terminus in cytochrome c binding, and improves the assay for proton pumping. Biochemistry. 40: 1606-1615.

30. Qin, L., D. A. Mills, C. Hiser, A. Murphree, R. M. Garavito, S. Ferguson-Miller, and J. Hosler. 2007. Crystallographic location and mutational analysis of Zn and Cd inhibitory sites and role of lipidic carboxylates in rescuing proton path mutants in cytochrome c oxidase. Biochemistry. 46: 6239-6248.

31. Hosler, J. P., J. Fetter, M. M. J. Tecklenburg, M. Espe, C. Lerma, and S. Ferguson-Miller. 1992. Cytochrome-aa3 of Rhodobacter sphaeroides as a model for mitochondrial cytochrome c oxidase - Purification, kinetics, proton pumping, and spectral analysis. J. Biol. Chem. 267: 24264-24272.

32. Amengual, J., J. Ribot, M. L. Bonet, and A. Palou. 2008. Retinoic acid treatment increases lipid oxidation capacity in skeletal muscle of mice. Obesity. 16: 585-591.

33. Yu, C. A., L. Yu, and T. E. King. 1975. Studies on cytochrome-oxidase - interactions of

182

cytochrome-oxidase protein with phospholipids and cytochrome-c. J. Biol. Chem. 250: 1383-1392.

34. Vik, S. B., and R. A. Capaldi. 1977. Lipid requirements for cytochrome c oxidase activity. Biochemistry. 16: 5755-5759.

35. Brierley, G. P., and A. J. Merola. 1962. Studies of electron-transfer system: Phospholipid requirements in cytochrome oxidase. Biochim. Biophys. Acta 64: 205-217.

36. Denessiouk, K. A., V.-V. Rantanen, and M. S. Johnson. 2001. Adenine recognition: A motif present in ATP-, CoA-, NAD-, NADP-, and FAD-dependent proteins. Proteins. 44: 282- 291.

37. Wierenga, R. K., P. Terpstra, and W. G. J. Hol. 1986. Predicting of the occurence of the ADP- binding βαβ-fold in proteins, using amino acid sequence fingerprint. J. Mol. Biol. 187: 101-107.

38. Eggink, G., H. Engel, G. Vriend, P. Terpstra, and B. Witholt. Rubredoxin reductase of Pseudomonas oleovorans. Structural relationship to other flavoprotein oxidoreductases based on one NAD and two FAD fingerprints. J. Mol. Biol. 212:135-142.

39. Eppink, M. H., H. A. Schreuder, and W. J. Van Berkel. 1997. Identification of a novel conserved sequence motif in flavoprotein hydroxylases with a putative dual function in FAD/NAD(P)H binding. Protein Sci. 6: 2454-2458.

40. Jörnvall, H., M. Persson, and J. Jeffrey. 1981. Alcohol and polyol dehydrogenases are both divived into two protein types, and structural properties cross-relate the different enzyme activities within each type. Proc. Natl. Acad. Sci. USA. 78: 4226-4230.

41. Jörnvall, H., B. Persson, M. Krook, S. Atrian, R. Gonzàlez-Duarte, J. Jeffrey, and D. Ghosh. 1995. Short-chain dehydrogenase/reductase (SDR). Biochemistry. 34: 6003-6013.

42. Ghosh, D., V. Z. Pletnev, D. W. Zhu, Z. Wawrzak, W. L. Duax, W. Pangborn, F. Labrie, and S.X. Lin. 1995. Structure of human estrogenic 17β-hydroxysteroid hydrogenase at 2.20 Å resolution. Structure. 3: 503-513.

43. Beis I., and E. A. Newsholme. 1975. The contents of adenine nucleotides, phosphagens and some glycolytic intermediates in resting muscles from vertebrates and invertebrates. Biochemistry. 152(1): 23–32.

44. Zunszain, P. A., J. Ghuman, A. F. McDonagh, and S. Curry. 2008. Crystallographic analysis of human serum albumin complexed with 4Z,15E-bilirubin-IX alpha. J. Mol. Biol. 381: 394-406.

45. Kutty, K. P., I. V. S. Nath, K. R. Kothandaraman, J. A. Barrowman, P. G. Perkins, M.-U. P Ra, and S.-N. Huang. 1987. Fusidic acid-induced hyperbilirubinemia. Digest. Dis. Sci.

183

32(8): 933-938.

46. Turnidge J, and P. Collignon. 1999. Resistance to fusidic acid. Int. J. Antimicrob. Agents. 12(2): S35–S44.

47. Follmann, K., S. Arnold, S. Ferguson-Miller, and B. Kadenbach. 1998. Cytochrome c oxidase from eukaryotes but not from procaryotes is allosterically inhibited by ATP. Biochem. Mol. Biol. Int. 45(5): 1047-1055.

48. Ricquier, D., C. Lin, and M. Klingenberg. Isolation of the GDP binding protein from brown adipose tissue mitochondria of several animals and amino acid composition study in rat. Biochem Biophys Res Commun. 106(2): 582–589.

49. Lin, C. S., and M. Klingenberg. 1982. Characteristics of the isolated purine nucleotide binding protein from brown fat mitochondria. Biochemistry. 21(12): 2950–2956.

50. Strieleman, P. J., K. L. Schalinske, and E. Shrago. Partial purification and functional reconstitution of GDP sensitive brown adipose tissue mitochondria uncoupling protein using octyl glucoside. Biochem Biophys Res Commun. 127(2): 509–516.

51. Cannon, B. and J. Nedergaard. 2004. Brown adipose tissue: function and physiological significance. Physiol. Rev. 84: 277-359

52. Nicholls, D. G. and R. M. Locke. 1984. Thermogenic mechanism in brown fat. Physiol. Rev. 64: 1-64.

53. Müller, F. 1987. Flavin radicals: Chemistry and biochemistry. Free Radical Bio. Med. 3: 215- 230.

54. Brändén, M., F. Tomson, R. B. Gennis, and P. Brzezinski. 2002. The entry point of the K- proton-transfer pathway in cytochrome c oxidase. Biochemistry. 41: 10794-10798.

55. Sharpe, M. A., and S. Ferguson-Miller. 2008. A chemically explicit model for the mechanism of proton pumping in heme–copper oxidases. J. Bioenerg. Biomembr. 40(5): 541–549.

56. Qin, L., J. Liu, D. A. Mills, D. A. Proshlyakov, C. Hiser, and S. Ferguson-Miller. 2009. Redox-dependent conformational changes in cytochrome suggest a gating mechanism for proton uptake. Biochemistry. 48: 5121-5130.

57. Busenlehner, L. S., L. Salomonsson, P. Brzezinski, and R. N. Armstrong. 2006. Mapping protein dynamics in catalytic intermediates of the redox-driven proton pump cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 15398-15403.

58. Buhrow, L., S. Ferguson-Miller, and L. A. Kuhn. 2012. From static structure to living protein: Computational analysis of cytochrome c oxidase main-chain flexibility. Biophys. J. 102: 2158-2166.

184

59. Hosler, J. P., S. Ferguson-Miller, and D. A. Mills. 2006. Energy transduction: Proton transfer through the respiratory complexes. Annu. Rev. Biochem. 75:165-187.

60. Svensson-Ek, M., J. Abramson, G. Larsson, S. Törnroth, P. Brzezinski, and S. Iwata. 2002. The x-ray crystal structures of wild-type and EQ(I-286) mutant cytochrome c oxidases from Rhodobacter sphaeroides. J. Mol. Biol. 321: 329-339.

61. Tomson, F. L., J. E. Morgan, G. P. Gu, B. Barquera, T. V. Vygodina, and R. B. Gennis. 2003. Substitutions for glutamate 101 in subunit II of cytochrome c oxidase from Rhodobacter sphaeroides result in blocking the proton-conducting K-channel. Biochemistry. 42: 1711- 1717.

62. Tiefenbrunn, T., W. Liu, Y. Chen, V. Katritch, C. D. Stout, J. A. Fee, and V. Cherezov. 2011. High resolution structure of the ba3 cytochrome c oxidase from Thermus thermophilus in lipid environment. Plos ONE. 6(7): e22348-e22348.

63. Antonini, E., M. Brunori, A. Colosimo, C. Greenwood, and M. T. Wilson. 1977. Oxygen “pulse” cytochrome c oxidase: Functional properties and catalytic relevance. Proc. Natl. Acad. Sci. USA. 74(8): 3128-3132.

64. Shinzawa-Itoh, K., H. Aoyama, K. Muramoto, H. Terada, T. Kurauchi, Y. Tadehara, A. Yamasaki, T. Sugimura, S. Kurono, K. Tsujimoto, T. Mizushima, E. Yamashita, T. Tsukihara, and S. Yoshikawa. 2007. Structures and physiological roles of 13 integral lipids of bovine heart cytochrome c oxidase. EMBO J. 26. 1713-1725.

65. Swann, S. L., S.P. Brown, S.W. Muchmore, H. Patel, P. Merta, J. Locklear and P.J. Hajduk. 2011. A unified, probabilistic framework for structure- and ligand-based virtual screening. J. Med. Chem. 54(5): 1223-1232.

185

CHAPTER 5: Perspectives on CcO K-pathway conformational change and allosteric regulation

186

Cytochrome c oxidase (CcO) has been a attractive system for investigating catalytic mechanism of controlled oxygen reduction, proton pumping, and the structure and assembly of membrane proteins. Additionally, studies of CcO are appealing for translational research.

Specifically, improperly assembled or dysfunctional CcO has been implicated as a cause of encephalophathies, heart diseases, and motor neuron diseases [1]. Due to the critical role CcO plays in aerobic metabolism and human health, this protein has been studied for nearly a century, producing a wealth of knowledge and technological advancements. However, due to the size, hydrophobicity, and complexity of CcO, a number of key questions remain.

For the last five years, we have combined biochemical and computational approaches to investigate allosteric regulation of the CcO K-pathway by amphipathic molecules. These ligands affect conformational change and proton conducting networks critical for K-pathway function.

Our results have been used to identify conformational change, rationalize confounding experimental findings, and design hypothesis-driven studies of CcO structure and function.

Three key studies have generated a unified understanding of K-pathway conformational change and the mechanism by which diverse amphipathic ligands regulate these changes.

From static structure to living protein: computational analysis of cytochrome c oxidase main- chain flexibility

CcO has been observed to undergo conformational changes during reaction intermediate transitions by solvent accessibility and crystallographic studies [2-3]. By applying the elastic network model and intrinsic flexibility calculations, a model of global counter-rotational and compressional motions on the interior and exterior of the membrane, coupled with rocking of the

187

β-sheet domain of CcO has been described. These motions are generated by helix tilting, bending, and compression, potentially regulating oxygen flux to the active site and proton uptake through the K-pathway. The K-pathway helices are the most flexible intra-membrane region and may function by controlling water molecule movement from the internal side of the membrane to the CcO active site.

Due to the large size and hydrophobicity of CcO, traditional methods for experimentally monitoring conformational changes are not applicable. How may we then confirm our computational model and elaborate our findings on the independent behavior of the K-pathway?

By comparing our flexibility analysis with amino acid conservation of CcO, three conserved glycines in the K-pathway region have been identified. Mutations of these residues with decreased backbone flexibility while maintaining the relative size and hydrophobicity of the K- pathway (glycine to alanine targeted mutagenesis) may be used to artificially constrain this region and relate rigidification to functionality.

It is also interesting that diverse membrane proteins, specifically α-helical pumps, have been predicted to undergo counter rotational motions in what has been called a “twist-to-open” motion [4]. However, functionally-critical lipids associated with these proteins have yet to be incorporated into the flexibility model. What role would lipids have on this twisting? How would this increased packing affect channel opening and closing?

Structural predictions and functional consequences of porphyrin, steroid, and detergent ligands binding to the cytochrome c oxidase steroid binding site

188

In a related study of CcO, the counter-rotational and compressional model of CcO flexibility was applied to understand the affects of ligand binding, specifically near the K- pathway. A conserved, crystallographically-defined ligand binding site has been identified in mammalian and bacterial CcO and has been called the steroid site [5-6]. Ligand binding by a diverse group of small molecules and detergents to this steroid site drastically affects the activity of the K-pathway E101A variant. Therefore, the E101A protein has been identified as a sensitive assay system for ligand binding, competition, and chemical rescue [7].

Investigation of the CcO steroid site using the E101A enzyme assay system has been complicated by the increased affinity for and inhibition by lauryl maltoside (LM) detergent

(Hiser, C., et al., in review). Based on comparisons of bacterial CcO crystal structures and small molecule docking, we have proposed an overlapping, two-site model. In this model, LM binds to an upper, conserved hydrophobic crevice, occupied by cardiolipin in mammalian structures.

Diverse, amphipathic molecules such as steroids and porphyrins bind to the lower, steroid binding site to both displace the overlapping LM and elicit ligand-specific affects. By applying intrinsic flexibility calculations to both the apo- and deoxycholate-bound structures, detergent inhibition and ligand-specific effects have been hypothesized to be a consequence of affecting the mechanistic conformational changes of the K-pathway. Overall, these results have been used to explain numerous experimental observations and design experiments to confirm the two-site model of K-pathway inhibition.

In our assay system, we are specifically investigating the E101A variant's activity changes in the presence of ligands. By using the E101A protein, the activating and inhibitory

189 ligand binding may be tied to this site. Further competition studies with the crystallographically observed deoxycholate, strengthen the finding that these diverse ligands do in fact interact at or near the steroid binding site. However, by investigating the E101A variant, we have taken our focus off of the WT enzyme, though it is always studied in parallel. Therefore, the native ligands that affect the WT enzyme in vivo remain a topic of debate.

Detergent inhibition and competition have both been hypothesized to be a result of ligand binding to a conserved upper, lipid binding site that is occupied by cardiolipin in the mammal

CcO structures [5]. Cardiolipin is a unique lipid that has been shown to be involved in CcO oligomerization and supramolecular complex assembly [5, 8]. One potential rationale for this upper cardiolipin site and lower regulatory site may then to be to differentially regulate monomeric and multimeric CcO.

This differential regulatory hypothesis may be better understood when comparing two isoforms of mammalian CcO. One tissue specific subunit IVa isoform, called the heart isoform due to its location in heart and other contractile muscle, is located at the dimeric interface and contains a unique phosphorylation site that promotes oligomerization [9]. In this same set of tissues, CcO is more tightly controlled by small molecules including nucleotides and hormones than in non-contractile tissues. Therefore, in metabolically demanding tissue, CcO may be phosphorylated and bind cardiolipin to promote dimerization or supracomplex formation. These complexes may then expose the steroid binding site for specific allosteric regulation based on cellular demand.

190

In some ways, this steroid binding site is analogous to the NADH dehydrogenase quinone binding site [10]. Both sites have been shown to be occupied by a diverse group of amphipathic ligands, including the non-steroidal ligands which affect CcO. It is also not unreasonable that the regulation of oxidative phosphorylation by small molecules may affect one or more of the electron transport chain complexes.

Steroid and detergent molecules inhibit mammalian and bacterial CcO [11, Hiser, C., et al., in review]. However, most of these ligands have little physiological consequence. What ligands are therefore capable of regulating CcO? What evolutionary advantage would blocking the K-pathway provide, as it is critical for function but is not involved in proton pumping efficiency? Computational studies of natively abundant ligands that may be specific for the steroid site have been undertaken in an attempt to understand which ligands are capable of regulating proton uptake through the K-pathway.

Three-pronged computational approach to predict regulatory ligands of cytochrome c oxidase

The application of three diverse, yet complementary computational methods have been used to predict natively abundant ligands specific for the conserved CcO steroid binding site.

These top ligand candidates were tested for their ability to affect RsCcO in vitro. ATP and GDP have been shown to be mildly inhibitory in a manner similar to the ATP inhibition observed in mammalian CcO regulation. These results suggest that nucleotides specifically regulate by allosterically inhibiting the CcO enzyme core, potentially at the steroid binding site. The consistency between diverse computational methods support the accuracy of predicted candidate ligands, protein-ligand interactions, and ligand binding orientations. Finally, our results have led

191 to a model of arresting CcO by blocking the K-pathway in a state that cannot generate oxygen radical byproducts.

We have combined cutting-edge computational approaches with experimental data to understand allosteric regulation of the K-pathway. Our results suggest that, based on the metabolic need of a cell, oxidative phospholyation may be controlled by arresting CcO in a state where no oxygen radical byproducts can be formed. Potential allosteric regulators include ATP,

GDP, and FAD, all of which may play a role in co-regulating metabolic components. The regulatory ability of these ligands may be optimized upon formation of CcO dimers or supramolecular complexes. This work serves as a basis for future crystallography, molecular biology, and cell biology studies to examine allosterstic regulation of CcO, by limiting K- pathway conformationally change in response to energetic demand.

192

REFERENCES

193

REFERENCES

1. Diaz, F., 2010. Cytochrome c oxidase deficiency: Patients and animal models. Biochim. Biophys. Acta. 1802: 100-110.

2. Qin, L., J. Liu, D. A. Mills, D. A. Proshlyakov, C. Hiser, and S. Ferguson-Miller. 2009. Redox- dependent conformational changes in cytochrome suggest a gating mechanism for proton uptake. Biochemistry. 48: 5121-5130.

3. Busenlehner, L. S., L. Salomonsson, P. Brzezinski, and R. N. Armstrong. 2006. Mapping protein dynamics in catalytic intermediates of the redox-driven proton pump cytochrome c oxidase. Proc. Natl. Acad. Sci. USA. 103: 15398-15403.

4. Valadié, H., J. J. Lacapčre, Y.-H. Sanejouand, and C. Etchebest. 2003. Dynamical properties of the MscL of Escherichia coli: A normal mode analysis. J. Mol. Biol. 332: 657-674.

6. Qin., L., D.A. Mills, L. Buhrow, C. Hiser, and S. Ferguson-Miller. 2008. A conserved steroid binding site in cytochrome c oxidase. Biochemistry. 47(38): 9931-9933.

7. Qin, L., D. A. Mills, C. Hiser, A. Murphree, R. M. Garavito, S. Ferguson-Miller, and J. Hosler. 2007. Crystallographic location and mutational analysis of Zn and Cd inhibitory sites and role of lipidic carboxylates in rescuing proton path mutants in cytochrome c oxidase. Biochemistry. 46: 6239-6248.

8. Pfeiffer, K., V. Gohil, R. A. Stuart, C. Hunte, U. Brandt, M. L. Greenberg, and H. Schägger. 2003. Cardiolipin stabilizes respiratory chain supercomplexes. J. Biol. Chem. 278(52): 52873-52880.

9. Fang, J. K., S. K. Prabu, N. B. Sepuri, H. Raza, H. K. Anandatheerthavarada, D. Galati, J. Spear, and N. G. Avadhani. 2007. Site specific phosphorylation of cytochrome c oxidase subunits I, IVi1, and Vb in rabbit hearts subjected to ischemia/reprefusion. FEBS Lett. 581: 1302-1310.

10. Okun, J. G., P. Lummen, and U. Brandt. 1999. Three classes of inhibitors share a common binding domain in mitochondrial complex I (NADH:ubiquinone oxidoreductase). J. Biol. Chem. 274, 2625-2630.

194

11. Van Buuren, K. J. H., and B. F. Van Gelder. 1974. Biochemical and biophysical studies on cytochrome-c oxidase. Effect of cholate on enzymic activity. Biochim. Biophys. Acta. 333:209-217.

195