CHAPTER 1 Introduction

CHAPTER 1

Introduction

Osteoporosis is the most common age-related conditions which is characterised by bone mineral density and change in microarchitecture of bone with consequently increasing in bone fragile(1). Osteoporosis has been reported to affect around 10% of population in North America, Europe and Japan(2). Considering the incidence rate of osteoporosis is anticipated to grow with increasing life expectancy. There is an extremely high mortality and morbidity due to bone fracture and loss ability to perform daily life activities. As a report from the World Health Organization (WHO) indicated the mortality and morbidity in Caucasian is as high as approximately 20% in 1-year post fracture. About 30% of survivors will have permanent disability with necessity of intensive care from others(3). For osteoporosis assessment, WHO first announced the tools for screening the patient with osteoporotic condition based on spinal mineral density (BMD) value. A person with osteoporosis has a BMD measurement of 2.5 standard deviations (SD) or more below of a young, normal adult of the same gender, according to WHO criterion(4).

The International Osteoporosis Foundation (IOF) reported the project number of global osteoporotic fractures. As life expectancy continues to rise in many countries, approximately 323 million people aged 65 years or older in 1990 is predicted to surpass 1500 million by the year 2050. Together with the estimated number of hip fractures in 1990 may increase from 1.7 million to 6.3 million by the year 2050. Particularly in Asian population, the estimated number of hip fractures as 0.572 million in 1990 was expected to surpass 3.52 million by the year 2050(5).

In Thailand, osteoporosis is also considered a serious public health concern resulting from long life expectancy. From previous studies, approximately 20% of postmenopausal women aged from 40 to 80 years old had osteoporotic fracture at lumbar spine, and approximately 40% of this population had osteoporotic fracture at femoral neck. Moreover, hip fracture, which is one of the most common fractures, indicated the incidence of 7.45 per 100000 in 1994(6). Phadungkiat et al.(7) reported the incidence of hip fracture in Chiang Mai province

1 as 151.2 per 100000 in 1997 and that of 181.0 per 100000 in 2006 reported by Wongtriratanachai et al(8). The incidence of hip fracture in females is 2.4 times greater than that of males. The most potent factors contributing to high mortality include non-operative treatment, delayed operative treatment, and receiving no anti-resorptive agents (3).

Osteoporosis is a bone disease caused by an imbalance of bone formation and bone resorption. The rapid bone resorption can be affected by general factors that related to aging and sex hormone deficiency, leading to decreased bone formation and reduced bone quality. Especially, the rate of bone resorption resulting from the overactive osteoclasts is greater than that of bone formation. Additionally, the resorption activity of osteoclasts requires several weeks, while the formation activity of osteoblasts requires months to construct a new bone. Therefore, the osteoclastic activity inhibition is interestingly targeted for osteoporosis treatment(9).

In general, the pharmacological treatment of osteoporosis is divided into two approaches, namely inhibition of bone resorption and promotion of bone formation. The medications for bone resorption inhibition include bisphosphonates, calcitonin, denosumab etc., while those to promote bone formation including parathyroid hormones etc(3). A bisphosphonate drug is currently the most popular agent used for prevention of osteoporotic fracture and treatment of osteoporosis. Unfortunately, bisphosphonates were reported to cause many side effects such as digestive problems, flu-like symptoms, bone pain and skin rash.(10) Additionally, the reduced bone resorption by using anti-resorptive agent cooccurs with the reduced bone formation, resulting from their coupling signal. Hence, a novel scheme for development of osteoporotic agents is mainly focused to inhibit bone resorption but no effect to bone formation.

Regards to new findings from previous literatures, bone resorption by osteoclasts is complex and multi-step processes such as the attachment of osteoclasts to bone surface using an integrin receptor prior to bone degradation. During bone resorbing process, osteoclasts release proton ions through membrane into the site of degraded bone via V-ATPase for acidifying the microenvironment, and also secrete the protease enzymes into resorption lacuna. These proteins are considered as targets for development of the osteoporotic agents(9, 11).

Not only the medications used to treat the osteoporotic condition, but medicinal plants also used as osteoporotic prevention such as Traditional Chinese Medicines (TCMs). TCMs is

2 a broad range of medicinal resource in China, which is mostly used to treat various diseases. According to the rapid development of extraction and separation process, a large number of compounds have been identified. Many of them are proven to be active in various biological assessments.(12) The phytochemical compounds possess the osteoprotective property including flavanones, flavones, flavonols etc(13). These herbal compounds exhibited the beneficial activities to represent as the alternative medicines together with calcium and vitamin D for osteoporotic prevention in early stages. Accordingly, natural products have made a significant contribution to drug discovery and nearly half of novel drugs in the market are natural products and their derivatives. The herbal compounds have received increasing attention due to their specific pharmacological activities(14). Hence, the herbal compounds can be used for development the anti-osteoporotic agents toward various targets on osteoclast.

In drug discovery and development, target discovery is an important step in drug discovery process(15) . Interestingly, the agents that can inhibit the function of protein at several targets seem extremely to be a key factor for new drug development. These agents called multi- target agents. Multi-target agents are recently highlighted due to using modern approaches to identify drugs that hit the multiple targets. The main reasons need for the better and safer drugs because the concurrent drugs cannot meet therapeutic target due to their complex diseases such as cancers, autoimmune diseases, and low success rate of drug development that calls for better and safer drug(16). As previous reports, approximately 20 new drugs were launched into the market per year because approximately 90% of new drugs failed in first-in-human testing. Especially in cancer disease, the percentage of drug development failure is highly around 95% in human testing due to their poor safety and less efficacy, which is accounting for 30% of all failures. Therefore, to increase the successful of drug development, the better targets are required. Together with the novel approach, the drug hitting multiple targets might be more successful in drug development(16). A good example of multi-target agents is sunitinib. Sunitinib is an ATP-mimetic kinase inhibitor that binds to the active binding site of several kinase enzymes resulting in enzyme activation and autophosphorylation inhibition(16). Although the multi-target approach is naturally associated with toxicology and off-target side effects, it can be argued that multi-target agents would have a larger therapeutic index than those hitting a single target. Therefore, the multi-target agent could prove to be safer drug. In order to investigate herbal compounds toward several targets on osteoclasts, the combining various approaches including in silico and in vitro methods have been done to find the hit compounds for further developments as multi-target agents against bone resorption process.

1.2 Objectives

This study aims to:

1. To construct an in-house library of natural compounds with anti-osteoresorptive activity 2. To investigate potential multi-target compounds against the selected targets

namely V-ATPase, and cathepsin K

CHAPTER 2

Literature reviews

2.1 Bone

2.1.1 Bone biology and function(17)

Bone is a living, dynamic connective tissue that is continuously remodelled throughout life. Bone is made of approximately 80% of dense cortical bone and 20% of trabecular bone (Figure.1). The most apparent functions of bone are to provide structural support for the body, protect the internal organs, and also act as a reservoir of mineral. Generally, bone composes of two phases including mineral phase and organic phase. In mineral phase, the most component is hydroxyapatite, whereas in organic phase the most component is type I collagen (approximately 90%), non-collagenous protein (NCPs), lipids and water.

Figure 2.1. Bone structure

2.1.2 Bone modelling and remodeling Bone modelling (9, 17)

The skeleton undergoes longitudinal and radial growth, modelling and remodeling throughout life. During embryogenesis, the flat bones are formed by mesenchymal stem cells

5 whilst long bones are established early as cartilages that become gradually replaced by bone. During childhood, bone modelling occurs as a result of the change of bone shaping due to mechanical forces applied on the skeleton. Bone remodeling, including bone resorption and bone formation, also occurs during childhood. Whilst in adulthood bone remodeling is notable process to maintain bone strength. Bone formation and bone resorption are less tightly coupled during modelling than remodeling.

Bone remodeling (9, 11)

Bone remodeling is critical for maintenance of bone homeostasis and healing the fractures by removal the older or damaged bone with healthy new bone. This process occurs throughout lifelong. Briefly, bone remodeling cycle contains of four sequential phases as shown in Figure 2. The recruitment and activation of osteoclasts is the first step. Bone resorption process resulting from the activity of osteoclasts occurs several weeks and then stops functioning. Then, the reversal phase couples bone resorption to bone formation through coupling signals for activating bone formation by osteoblasts. Finally, osteoblasts lay down type I collagen which is produced inside osteoblasts into the resorbed site to replace the damaged bone with healthy new bone.

Figure 2.2. Bone remodeling on the surface of bone

Figure 2.2 showed the bone remodeling process. The recruitment and activation of osteoclasts initially occurs in bone resorption. Osteoclasts are multinucleated cell that derived from hematopoietic stem cells in bone marrow to form preosteoclast precursors. They are activated by various factors such as inflammatory cytokines. Pre-osteoclasts adhere to bone mineral matrix using integrin receptor and then form specialised function, sealing zone, around

6 the location of bone resorption underneath osteoclast. Activated multinucleated osteoclasts secrete the hydrogen ions through the ruffled border for acidifying the compartment underneath osteoclasts to reach pH around 4.7, that is the optimal value for dissolving bone matrix and activating the degraded enzymes. Subsequently, osteoclasts release proteinase enzymes to degrade bone mineral in the bone-resorbing compartment. Generally, bone resorption takes about 2 to 4 weeks before completion when the osteoclast stops functioning and undergo apoptosis.(9)

During reversal phase, bone resorption transforms to bone formation. At the completed site of bone resorption, the resorbed cavities contain various kinds of mononuclear cells such as osteocytes, monocytes, and preosteoblasts recruited to initiate bone formation.(18) The coupling signals connect the completion of bone resorption process to initiation of bone formation at the resorbed area of bone resorption. Osteoclasts also secrete several factors that directly control the activation of preosteoblasts and osteoblasts.

Later, osteoblasts synthesise new collagenous matrix and regulate mineralisation of matrix by launching small matrix vesicles containing high phosphate and calcium concentration, and also inhibitors to degraded enzymes, e.g. pyrophosphate or proteoglycans, resulting from effect of alkaline phosphatase. Normally, bone formation takes about 4 to 6 months before completion.(9)

Bone cells

Bone remodeling is crucial for maintenance a renewal of bone mechanical strength through the replacement of older or damaged bone with new bone tissue. This process is normally occurred throughout lifelong but slow rate of bone formation in elder. To maintain the bone homeostasis, there are four bone cells involving bone remodeling, e. g. osteoclasts, osteoblasts, osteocytes and bone lining cells.(19)

a. Osteoclasts

Osteoclasts are multinucleated giant cells responsible for the degradation of bone. Osteoclasts differentiate from myeloid precursors regulated by cytokines macrophage colony stimulating factors (M-CSFs) and receptor activator of NF-kB ligand (RANKL). RANKL is

7 member of Tumor Necrosis Factor (TNF) superfamily generated by osteoblast and/or osteocytes. M-CSFs play a crucial role for the proliferation, differentiation, and survival of osteoclasts precursors.(17)

b. Osteoblasts

Osteoblasts are large cell responsible for new bone formation. They produce various cell products such as type I collagen, which is the most component of bone matrix protein, enzymes namely alkaline phosphatase and collagenase. This, together with some minor types of collagen, proteoglycans, fibronectin and specific bone proteins, such as osteopontin, bone sialoprotein and osteocalcin, becomes the unmineralised flexible osteoid on which the osteoblast reside. (17, 20)

c. Osteocytes

Osteocytes are the star-shaped type of bone cells embedded within bone matrix. When the osteoblasts are trapped within bone matrix, some of them turn into osteocytes. Osteocytes can form the canalicular network to communicate to bone lining cells, osteoblasts, and other osteocytes to exchange the nutrients and waste via gap junctions. (21)

d. Bone lining cells

Bone lining cell is flat-shaped type of bone cell responsible for covering nonremodeling bone surface. Bone lining cell may control influx and efflux of bone mineral ions in and out of bone extracellular fluid. Additionally, bone lining cells also retain the ability to differentiate into osteoblasts when exposure to parathyroid hormone or mechanical forces. (20)

2.1.3 Potential receptor for osteoclast function

Bone resorption is complex process with various mechanisms inside osteoclastic cells. Osteoclasts are multinucleated cells activated by several factors to resorb bone. The function of osteoclasts begins when the RANK ligands, belongs to TNF superfamily, produced by osteoblasts bind to RANK receptor on osteoclasts. When osteoclasts attach bone mineral matrix, osteoclasts then become polarisation stage.(17)

In polarised osteoclasts, the four types of osteoclastic membranes are involved such as the sealing zone, and ruffled border, basolateral and functional secretory domains. The polarisation of osteoclasts relate to rearrangement of actin cytoskeleton in which F-actin ring that comprises a dense continuous zone of highly dynamic podosome is formed and consequently an area of membrane that develop into the ruffled border (RB) is isolated.(9) This specialised membrane is formed after the osteoclasts are already in contact with extracellular matrix (ECM), in a process which αVβ3 integrin mediates the attachment of the osteoclast podosome to the bone surface.(17)

When osteoclasts attach bone mineral matrix, the actin cytoskeleton of osteoclasts is organised into an actin ring to promote formation of sealing zone around the site of osteoclasts adherence to bone matrix. Sealing zone surrounds and separates the acidified resorption compartment from surrounding bone surface.(22) The maintenance of ruffled border is essential for osteoclastic activity during bone resorption. The ruffled border releases the proton ions via vacuolar-ATPase (V-ATPase) proton pump and chloride channels to acidify the resorption lacuna for mineralising the hydroxyapatite (HAP) crystals.(20)

In resorbed stage, the protons and degraded enzymes such as cathepsin K, matrix metalloproteinase, tartrate-resistant acid phosphatase (TRAP), are transported into the resorption lacuna, where is located underneath the osteoclasts responsible for mineralising the type I collagen from bone matrix. Later, degraded products are moved across the ruffled border and then transferred to the functional secretory domain (FSD) at the plasmamembrane by endocytosis and transcytosis, respectively.(20) As mentioned above, the osteoclastic activity is defined by various morphological and functional characteristics. Briefly, the activation of osteoclast is initiated with matrix recognition, adhesion to the bone surface, followed by cytoskeleton rearrangement.(23)

Briefly, bone resorption is initially activated when RANK ligands, produced by osteoblasts, bind to RANK receptor located on osteoclasts. Then osteoclasts adhere tightly to bone surface via the transmembrane receptor, namely αVβ3 integrin receptor. After that the osteoclast formed specialised organ called ruffled borders (RB) with the presence of various receptors including V-H+-ATPase proton pump receptor, which is responsible for acidifying the compartment underneath the ruffled border, or called resorption lacuna, to reach around pH 4.7 values that is the appropriate condition for dissolution of bone mineral and enzymatic proteases function. Consequently, the lysosomal vesicles containing with cathepsin K enzymes, move across the ruffled border and release the acidic enzymes to the resorption lacuna for degradation of type I collagen, together with matrix metalloproteinases (MMPs) that are also protease enzyme to degrade the collagen at different regions. The resorbing osteoclast showed in Figure 2.3 to express the degraded enzymes and other functional proteins occur in this resorbing phase.

Figure 2.3. Illustration of resorbing osteoclast(24)

2.1.3.1 Vacuolar-ATPase proton pump (25-27)

Vacuolar-ATPase proton pump, or called V-ATPase, is a multi-subunit enzyme presented in eukaryotic cell. V-ATPase is located within endomembrane organelles, including lysosomes, endosomes, vacuoles, golgi apparatus and secretory vesicles, responsible for ATP- dependent proton transportation from cytoplasmic compartment to the opposite site of membrane. These organelles are usually found in acidifying cells such as osteoclasts, renal intercalated cells and epididymis, to play a vital role in such processes as protein degradation,

10 endocytosis, and coupled transport. In addition, V-ATPase was also reported to involve in the regulation of membrane fusion and HIV entry.

The structure of V-ATPase(27)

V- ATPase proton pump enzyme belongs to ATP- dependent proton pump that is responsible for physiological processes. V-ATPase consists of various multiple subunits to organise into two domains such as V1 and V0 domain ( Figure 2.4) . The V1 domain is responsible for the hydrolysis of ATP, which is composed of eight subunits, including A, B, C,

D, E, F, G and H subunit. The V0 domain is responsible for proton translocation, which is composed of six subunits including a, d, c, c’, c” and e subunit. The function of each subunit of V-ATPase is shown in Table 2.1.

Figure 2.4. Structure of V-ATPase (28)

Table 2.1. Summary of subunits and function of V-ATPase (29) Domain Subunits Mass (kDa) Function V1 A 70 Catalytic site, ATP-binding B (B1, B2) 56 Non-catalytic site, ATP-binding C (C1, C2) 42 Peripheral stator, assembly D 34 Central rotor E (E1, E2) 31 Peripheral stator F 14 Central rotor G (G1, G2, G3) 13 Peripheral stator H 50 Peripheral stator V0 a (a1, a2 , a3, a4) 110-116 H+ translocation

c (c, c’, c”) 16-21 H+ translocation D 38 Assembly e (e1, e2) 9 Membrane sector-associated

As shown in Table 2.1, the V1 domain contains of both catalytic site and non-catalytic site which are the site of ATP-binding. Subunit A is responsible for the hydrolysis of ATP whereas subunit B is assumed that is possible to regulate the function of ATP. The A- B conformational changes can stimulate the rotation of central rotor and c and c” subunits. Moreover, the c-c” subunit and a subunit are involved in the pathway of proton translocation through membrane.(30)

Osteoclasts can be able to acidify the microenvironment underneath the ruffled border during bone resorption process, via utilising V-ATPase to pump proton ions from cytoplasm to the resorption lacuna. Toro et al.(31) suggested that during osteoclast differentiation, the levels of most V-ATPase subunits increased and the cell-specific isoforms were expressed (a3, d2). These led to increase levels of subsets of V-ATPase that were stored within intracellular vesicles. The formation of V-ATPase is initially embedded in the vesicles and then directed to the site of ruffled border in coordination with actin filaments, and later fuse with the plasma membrane. Once bone resorption is completed, V-ATPases are re-internalised into the cytoplasm. The osteoclastic cell can move to another site of resorption and transform actin rings and ruffled membranes for another cycles of resorption.

Marshansky et al.(26) indicated that subunit a3 is a specific isoform of osteoclasts which is embedded within ruffled membrane during osteoclast differentiation. Additionally, the mutation of V-ATPase at residue 444, which is located in subunit a3, led to osteopetrosis in human.

Subunit d2 in V0 domain also suggested that the specific subunit for osteoclastic bone resorption. Subunit d2 is a hydrophilic isoform which is not contained with any (32) transmembranes but tightly interaction of V0 and V1 sector of V-ATPase. Nishi et al indicated that d2 isoform played a vital role for couple transport and ATP hydrolysis. In mutagenic study, the mutation at d2 isoform also results in osteopetrosis in mouse.

Crystal structure of V-ATPase

The most recent crystal structure of V-ATPase is crystallised by Mazhab-Jafari et al.(33) with the resolution of 3.9 Å. This structure presented a large cavity between subunit a and the c-ring led to a cytocplasmic half-channel for protons. The c-ring has a distribution of protein- carrying glutamic acids of subunit c” that interacts with Arg735 of subunit a. This structure suggested the protonation of a glutamic acid residue was affected by the protonation and deprotonation of the c-ring with ATP-hydrolysis-driven rotation.

Figure 2.5. Cytoplasmic half-channel and subunit a/c-ring interaction(33)

Figure 2.5 showed the cavity is located between the c-ring and subunit a. This cavity is near the loop consisted of amino acid sequence from 659 to 709 obtained in subunit a. Additionally, Arg735 of subunit a and Glu108 of subunit c” are key residues for binding interaction.(33)

2.1.3.3 Cathepsin K

During osteoclastic bone resorbing process, several degraded enzymes are released from osteoclast to mineralise bone at the resorbed site, such as matrix metalloproteinase (MMP) enzymes and cathepsin K. These enzymes are responsible for cleaving collagen, which is the main component of bone. Historically, MMP was focused as the main degraded enzymes. However, MMP can function at alkaline pH value. Cathepsin K is one of cysteine protease enzymes, that is active in the acidic pH environment, so cathepsin K was then highlighted as main collagenase enzyme during bone resorption. (2) In 1980, Delaisse et al. (34) indicated that leupeptin, the first cysteine protease inhibitors, can inhibit bone resorption in vitro study. The first cysteine protease inhibitors used were peptide aldehydes including leupeptin and antipain which can inhibit the function of tryptic serine proteases and later were followed by the more specific peptide epoxide (E-64)(35) and peptidyl diazomethanes(36).

Additionally, bone biopsies from cathepsin K knockout mice induced osteopetrosis presented reduced bone resorption with higher the number of osteoclasts and increased bone formation.(2) Other in vivo experiments indicated that the removal of cathepsin K in osteoclasts resulted in increased the volume of bone, and also increased the numbers of osteoblasts and osteoclasts. Whilst the removal of cathepsin K in osteoblasts had no effect on the rate of bone formation.(37)

Structure and function of Cathepsin K

Cathepsins are protease enzymes presenting inside cells with acidic compartments such as lysosomes. As yet, there are fifteen human cathepsins which are classified into three groups, e.g. cysteine, aspartyl and serine cathepsins. Cysteine cathepsins include eleven types such as cathepsin B, C, F, H, K, L, O, S, V, W and Z. Aspartyl cathepsins include two types such as cathepsin D, E, and serine cathepsins include two types such as cathepsin A and G. Cathepsin B, K, L, S and V are all found in human skin with collagenase activity.

As previous reports showed that cathepsin K cleave type I collagen at multiple positions. Bromme et al. (2) revealed the recognition sequence of cathepsin K inhibitor for inhibiting the function of cathepsin K. This recognition sequence contained with various residues, Gly-Leu-Lys—Gly-His, that can interact with cathepsin K active site via electrophilic warhead. The site of interaction of cathepsin K is presented in figure 8 and the 3D human cathepsin K crystal structure is showed in Table 2.2.

Figure 2.6. Structure of cathepsin K with nitrile-based inhibitor(38) Table 2.2. 3D structure database of human cathepsin K

PDB entry Method Resolution (Å) Chain Positions

1ATK X-ray 2.2 A 115-329

1AU0 X-ray 2.6 A 115-329

1AU2 X-ray 2.6 A 115-329

1AU3 X-ray 2.5 A 115-329

1AYU X-ray 2.2 A 115-329

1AYV X-ray 2.3 A 115-329

1AYW X-ray 2.4 A 115-329

1BGO X-ray 2.3 A 115-329

1BY8 X-ray 2.6 A 16-329

1MEM X-ray 1.8 A 115-329

1NL6 X-ray 2.8 A/B 115-329

1NLJ X-ray 2.4 A/B 115-329

1Q6K X-ray 2.1 A 115-329

1SNK X-ray 2.4 A 116-329

1TU6 X-ray 1.75 A/B 115-329

1U9V X-ray 2.2 A 113-329

1U9W X-ray 2.3 A 113-329

1U9X X-ray 2.1 A 113-329

1VSN X-ray 2.0 A 115-329

1YK7 X-ray 2.5 A 115-329

1YK8 X-ray 2.6 A 115-329

1YT7 X-ray 2.3 A 115-329

PDB entry Method Resolution (Å) Chain Positions

2ATO X-ray 2.0 A 115-329

2AUX X-ray 2.4 A 115-329

2AUZ X-ray 2.3 A 115-329

2BDL X-ray 2.0 A 115-329

2R6N X-ray 1.95 A 113-329

3C9E X-ray 1.8 A 115-329

3H7D X-ray 2.24 A/E 115-329

3KW9 X-ray 1.8 A 115-329

3KWB X-ray 2.02 X/Y 115-329

3KWZ X-ray 1.49 A 115-329

3KX1 X-ray 1.51 A 115-329

3O0U X-ray 1.8 A 115-329

3O1G X-ray 1.65 A 115-329

3OVZ X-ray 2.02 A 121-329

4DMX X-ray 1.7 A 115-329

4DMY X-ray 1.63 A/B 115-329

4N79 X-ray 2.62 A 115-329

4N8W X-ray 2.02 A 115-329

4X6H X-ray 1.0 A 115-329

4X6I X-ray 1.87 A 115-329

4X6J X-ray 1.59 A 115-329

4YV8 X-ray 2.0 A 115-329

4YVA X-ray 1.8 A 115-329

5J94 X-ray 2.6 A 107-329

7PCK X-ray 3.2 A/B/C/D 16-329

Importantly, potential off- target effect of cathepsin occurs when the cathepsin K inhibitors is possibly interact with other cathepsins. For example, if the inhibitors bind to cathepsin L, it might lead to cardiomyopathy, impaired neovascularization. For cathepsin K, it can lead to neuronal ceroid lipofuscinosis. When it occur with cathepsin S, this is possible to be impaired immune response. (2) Therefore, to improve selectivity to cathepsin K can help avoiding undesirable potential off-target effect.

4.2 Herbal compounds possess osteoprotective activity

According to many efforts to find the natural compounds containing the osteoporosis protective effect, there are several herbal compounds reported osteoprotective activities. Che et al. (13) reviewed the summary of natural products with reported osteoprotective activities derived from Chinese medicinal plants and related properties in cell-based and/or animal models. Their review can be classified natural substances into various group based on the chemical structures such as flavanones, flavones, flavonols, lignans, prenylpropanol derivatives, triterpenoids, and miscellaneous compounds. A summary of natural products from Chinese medicinal herbs showed in Table 2.3.

Other studies showed the medicinal plants such as Cistanche salsa, Anoectochilus formosanus, Acanthopanax senticosus, Herba epidermii and Curcuma longa have showed the potential effects to both osteoclast and osteoblast in vitro, in vivo and therapeutic potency in osteoporosis.(39) Moreover, their effort to search for the potential effect of herbal substances for osteoporosis treatment revealed that they are also related to other bone diseases. In 2010 the collaborative research of Tokyo University and Showa University screened more than 400 active compounds. The extracts that exhibited their anti-osteoclastic function or induced to apoptosis were selected to test in RAW264.7. These extracts were further screened in differentiated osteoblastic cell from MC3T3E1 and chondrocyte from ATDC5 cells. As a result, the extracts of M. azedarach, C. turtschaninovii and C. atratum were selected since these extracts did not induce cell death. (39, 40)

Table 2.3. Source of herbal compounds from Chinese medicinal herbs possess osteoprotective activity (13) Scientific name Part used Active Compounds Osteoprotective property Actaea heracleifolia (Kom.) Rhizome cimicidol-3-O-β-xyloside, -suppression the formation of osteoclast-like J. Compton; cimicidanol-3-O-β-xyloside, cells and their resorbing activity A. dahurica (Turcz. ex Fisch. acetylacteol-3-O-arabinoside -prevention the OVX-induced elevation of & C.A. Mey.) Franch.; serum alkaline phosphatase level A. foetida L. Astragalus mongholicus Root formononetin -up-regulation the gene expression of BMP-2 in Bunge high-throughput assay using MC3TC-E1 cells -increase bone mineral density in female rats Carthamus tinctorius L. Flower matairesinol, tilianine, acacetin and their Inhibition of bone resorption derivatives - inhibition of the Src family kinase - inhibition of osteoclast differentiation - suppression of TRAP-positive multinucleated cells, gene expression of NFATc1, RANKL- mediated intracellular reactive oxygen species (ROS) generation Cistanche deserticola Y.C. Stem 8-hydroxy-2,6-dimethyl-2-octenoic acid, -enhance bone mineral density and bone Ma; echinacoside mineral content C. tubulosa (Forssk.) Beck -regulation of bone metabolic genes; Smad1, Smad5, TGF-β, and TIEG1 -increase of MC3T3E1 cell proliferation, alkaline phosphatase activity, collagen I secreting, osteocalcins.

Cordyceps sinensis (Berk.) cordycepin -inhibition of RANKL-induced osteoclast Sacc. differentiation -down-regulation of mRNA expression of clastogenesis-related genes; TRAP, cathepsin K, MMP-9, NFATc1

- down-regulation of proinflammatory cytokines; IL-1β, TNF-α -prevention of bone loss caused by estrogen deficiency Dioscorea spp. Rhizome diosgenin, diospongins B and C, piperitol, -enhancement the proliferation of MC3TC-E1 sesaminone, syrinaresinol cells with up-regulation of bone marker suppressions; Runx2, osteopontin -inhibition of bone resorption through parathyroid hormone-treated parietal bone of mice Dipsacus asper Wall. ex C.B. Root asperosaponins V and VI, -inducing MC3T3E1 cell maturation and Clarke; hederagenin-3-O-(2-O-acetyl)-α-l- differentiation by BMP-2 formation D. japonicas Miq. arabinopyranoside Drynaria fortunei (Kunze ex Rhizome naringin and other flavos Suppression of osteoclast function Mett.) J. Sm. -down-regulation of osteopontin and osteonectin mRNA expression -interrupting the trafficking of pro-cathepsin K in osteoclasts -suppression the expression of cathepsin K Promotion of osteoblast differentiation and maturation

Eclipta prostrata L. Above- diosmetin, 3′-hydroxybiochanin A, 3′-O- -stimulatory activity on osteoblast proliferation ground methylorobol, echinocystic acid, and alkaline phosphatase activity parts wedelolactone -improved trabecular architecture in OVX rats -inhibition of osteoclast proliferation and differentiation by inhibiting RANKL-induced TRAP activity and reducing number of osteoclast-like cells in RAW264.7 cell line Epimedium brevicornum Leaf icariin, epimedins A, B and C, -osteoblast proliferation and differentiation Maxim.; baohuoside-1, -up-regulation of BMP-2, BMP-4, Runx2, maohuoside A, sagittatoside A, collagen, osterix, Smad4, OPG

E. sagittatum (Siebold & ikarisoside A, icaritin, icariside I and II -inhibition of osteoclast formation induced by Zucc.) Maxim.; RANKL and M-CSF in mouse BM culture E. pubescens Maxim.; E. -suppression the osteoclast differentiation koreanum Nakai marker; TRAP, IL-6, TNF-α, RANKL -inducing osteoclast apoptosis and cell cycle arrest Erythina variegate L. Bark 6-prenylgenistein, 8-prenylgenistein, 6,8- Suppression of osteoclast differentiation and diprenylgenistein maturation -suppression the expression of cathepsin K and down-regulation of OPG mRNA in tibia -decrease TRAP-positive cell numbers in RANKL-treated RAW264.7 cells. Eucommia ulmoides Oliv. Stem bark geniposidic acid, geniposide, aucubin, -osteoblast proliferation 5-hydroxymethyl-2-furaldehyde -suppression the growth of osteoclasts -stimulation the osteogenic differentiation Ferula spp. Resin ferutinin -inhibition of bone resorption -promotion the expression of osteocalcin, osteopontin, type I collagen, Runx2 and osterix in stem cells Ligustrum lucidum W.T. Aiton Fruit oleanolic acid, ursolic acid, tyrosol, -inhibition of osteoclast-like multinucleated cell hydroxytyrosol, oleuropein, and others induced by 1,25-Vit D3 -inhibition of RANKL-induced osteoclast differentiation -promotion of osteoblastic differentiation Lycium babarum L. Fruit polysaccharide -improving the bone mineral density, serum ALP activity, calcium and phosphorous contents in dexamethasone-induced osteoporosis in rats Morinda officinalis F.C. How Root physicion, rubiadin, rubiadin-1-methyl -inhibition of osteoclastic bone resorption ether, 2-hydroxy-1- -enhancement of bone formation methoxyanthraquinone,

1,2-dihydroxy-3-methylanthraquinone, 1,3,8-trihydroxy-2-methoxyanthraquinone, 2-hydroxymethyl-3- hydroxyanthraquinone, 2-methoxyanthraquinone, scopoletin Ormosia henryi Prain Root isoformononetin -inhibition the apoptosis and differentiation of osteoblasts involving the activation of MEK/ERK and Akt pathways Oxytropis falcata Bunge Whole plant isoformononetin -inhibition the apoptosis and differentiation of osteoblasts involving the activation of MEK/ERK and Akt pathways Paeonia lactiflora Pall. Root 6′-O-β-d-glucopyranosylalbiflorin -increase the ALP activity and nodule mineralisation of MC3T3-E1 cells Panax spp. Root ginsenosides -inhibition of bone resorption involving RANKL-induced osteoclast differentiation and TNF-α mRNA expression in RAW264.7 cells -promotion of cell proliferation in MC3T3-E1 cells Podocarpium podocarpum Whole plant podocarnone, luteolin, astragalin, afzelin, -enhancement of osteoblast proliferation (DC.) Yang et Huang kaempferitrin, rutin, involving MEK-ERK and Akt pathway quercetin-7-O-d-glucopyranoside, -increasing alkaline phosphatase activity genistein, laburnetin, wighteone, luteone, -suppression TRAP activity 7-O-methyl-luteone, cajanin Polygonum cuspidatum Root and resveratrol -stimulation the proliferation and differentiation Siebold & Zucc. rhizome of MC3T3-E1 and human bone marrow mesenchymal stem cells -prevention the RANKL-induced osteoclast differentiation Psoralea corylifolia L. Fruit psoralen, isopsoralen, psoralidin, corylin, -stimulation of differentiation of bone bavachin, isobavachin, neobavaisoflavone, mesenchymal stem cells involving Notch bakuchiol, bavachalcone signaling pathway

-up-regulation of expression of type I collagen, osteocalcin and BSP -enhancement osteoblast proliferation and differentiation -inhibition of osteoclast formation from precursor cells -suppression the activation of MEK, ERK and Akt Pueraria lobate (Willd.) Ohwi Root isoformononetin, puerarin and its 6′′- -prevention of bone loss by inhibiting the xyloside production of RANKL, IL-1β, TNFα, MMP-2 and MMP-9 -stimulation of new bone formation in rat model -promotion the osteoblast proliferation and differentiation in OVX animals Rehmannia glutinosa (Gaertn.) Root acteoside -promotion of osteoblast differentiation, ALP DC. activity and mRNA expression of bone-related genes in vitro study -blocking osteoclast activation Salvia miltiorrhiza Bunge Root and salvianolic acids A and B, tanshinones I, -promotion of bone formation markers; ALP, rhizome IIA and VI, cryptotanshinone, OPG in rat osteoblast 15,16-dihydrotanshinone I -suppression the expression of genes such as calcitonin receptor, c-Src kinase, integrin β3, c- Fos and NFATc1-induced RANKL, NF-kB, resulting in the inhibition of osteoclast differentiation -inhibitory effect to cathepsin K -stimulation the gene expression of ALP, OCN, OPG, and RANKL in MC3TC-E1 cells Sambucus williamsii Hance Stem ficusal, ceplignan, dehydrodiconiferyl -suppression of Cbfa1 and cathepsin K mRNA alcohol, dehydrodiconiferyl alcohol-γ′- levels methyl ether, samwinol, erythro-1-(4- -enhancement the OPG/RANKL mRNA hydroxy-3-methoxyphenyl)-2-[4-(3- expression ratios in mice

hydroxypropanyl) -2-methoxyphenoxy]- -reducing the TRAP-positive cell in RANKL- 1,3-propanediol, vanillic acid induced RAW264.7 cells Sophora japonica L. Fruit, seed genistein, 8-prenylkaempferol, -promoting the differentiation and maturation of and root sophoricoside, 2′-methoxykurarinone osteoblasts -inhibition of osteoclast differentiation through the down-regulation of RANKL-induced MAPKs and c-Fos-NFATc1 signaling pathways Sophora flavescens Aiton Root formononetin -up-regulation the gene expression of BMP-2 in high-throughput assay using MC3TC-E1 cells -increase bone mineral density in female rats Trifolium pretense L. Inflorescenc formononetin -up-regulation the gene expression of BMP-2 in e and twig high-throughput assay using MC3TC-E1 cells -increase bone mineral density in female rats Viscum coloratum (Kom.) Twig syringareninol O-β-glucopyranoside, -inhibition the formation of osteoclast-like Nakai 2-homoeriodictyol 7-O-β- multinucleated cells in mouse calvarial glucopyranoside, osteoblasts viscumneosides I, IX and X Adopt from Che et al. (13)

2.4.3 Computational approaches

2.4.3.1 Computational or/in silico approaches in drug discovery and development

Drug discovery and development is a highly complex process which can take about 15 years with expensive costs. Computational or in silico approaches are potential methods in modern drug discovery using computational simulations with less time and highly achievement. These approaches consist of databases, quantitative structure- activity relationships, pharmacophores, homology models and other molecular modelling approaches, machine learning, data mining, network analysis tools and data analysis tools. The applications of computational approaches in drug discovery are cost-effective methods using the experiments alongside the predicted activity from in vitro experiments to improve the compound properties. Hence, computational approaches can help to make a decision and apply to reduce costs drug discovery and development.(41)

2.4.3.2 Multi-target approach

Multi-target strategy is the novel paradigm in drug design and development process. The traditional approach in drug design is one drug one target has been shifted to one drug multiple targets. This strategy requires the combination of the data derived from computational studies, in vitro and in vivo testing, and clinical studies.

To date, the main reasons need for the better and safer drugs because the concurrent drugs cannot meet therapeutic target due to their complex diseases such as cancers, autoimmune diseases, and low success rate of drug development that calls for better and safer drugs. As previous reports, approximately 20 new drugs were launched into the market per year because approximately 90% of new drugs failed in first-in-human testing. Especially in cancer disease, the percentage of drug development failure is highly around 95% in human testing due to their poor safety and less efficacy, which is accounting for 30% of all failures. Therefore, to increase the successful of drug development, the better targets are required. Together with the novel approach, the drug hitting multiple targets might be more successful in drug development. Even a single drug hitting multiple targets may lead to increase the side effects on multiple targets; there are

24 some disagreements that multi-target agents can have a better/safety ration than monotarget agents.(42) Regarding to a review of Sikazwe(43) involved about the notable advantages of utilising a multi-target agent as described below

- monotarget may be not sufficient for therapeutic achievement due to the complex diseases - a multi-target agent may reduce the side effects of individual side drug effect profile from utilising the combination drugs therapy - the pharmacokinetic and metabolism related the toxicity caused by taking many therapeutic drugs, can reduced by utilising a multi-target agent - improving the patient non-compliance - a multi-target agent can be tailored to affect the key disease targets or pathway in order to minimise drug resistance - a multi-target agent can reduce drug-drug interactions particular in drug involving metabolism such as CYP450 induction/inhibition

Monzon et al.(44) suggested that a reasonable alternative to developing combinations of targeted agents is to develop a single agent hitting multiple targets. The multi-target agents are categorised into two subclasses that differences on how they act on the targets and induced cellular effects. One class has been designed to have potent activity on several different targets. The second class has potent activity for single target but has effects on a broad number of additional cellular components. A representative of the first class is the multi-targeted kinase inhibitor, sunitinib. Sunitinib is an ATP- mimetic, which bind to the ATP binding pocket of several protein kinases, inhibiting enzyme autophosphorylation and activation. Examples of second subclass are agents that target protein metabolism such as the proteasome, DNA methylation, or histone deacetylation. Although not designed on purpose, a lot of recent drugs are known for their efficiency to hit multiple targets including aspirin. Aspirin is an anti-inflammatory agent for treating rheumatoid arthritis (RA), pericarditis and Kawasaki diseases. It is also used to prevent transient ischemic attacks, strokes, heart attack and so on.

2.4.3.3 Virtual Screening (45, 46)

In drug discovery process, virtual screening is the essential tools using for investigating the natural substances containing a particularly type of property or biological activity. In order to enhance the quality of screening data generated from highly valuable natural products, the nature of the extracts can be improved. This approach can be able to score and rank molecules in a vast number of chemical libraries according to their likelihood of having affinity for their target. Virtual screening can be categorised into several approaches, one of potential approach is structure-based or target-based virtual screening via molecular docking simulation.

2.4.3.4 Molecular Docking

Molecular docking is extremely useful tool of structure-based drug design which has become an increasingly important tool for drug discovery.(47) Molecular docking is utilised for predicting the pose of bound ligand with binding affinity to its receptor. Docking simulation provides the valuable information about the interaction of small molecule to target receptor and key amino acids. As yet, a variety non-commercial and commercial docking programmes are widely available with various algorithms such as AutoDock(48), FlexX(49), CDOCKER(50), GOLD(51) etc. Several studies have an attempt to compare the effective of various docking approaches. However, different approaches can achieve different success rate due to specific target protein. Typically, molecular docking can be categorised into two steps; namely sampling algorithms and scoring approaches.(47)

Sampling algorithms(47)

In molecular docking, the sampling algorithm is used in prediction of bound ligand conformation within its receptor. More recently, there are several sampling algorithms which have been developed for using in individual molecular docking programmes. The sampling algorithms are shown in Table 2.4.

Table 2.4. Examples of sampling algorithms(47) Algorithms Characteristic Geometry-based, suitable to VS and database Matching algorithms enrichment for high speed Incremental construction Fragment-based and docking incrementally MCSS Fragment-based methods for the de novo design LUDI Fragment-based methods for the de novo design Monte Carlo Stochastic search Genetic algorithms Stochastic search Molecular dynamics For further refinement after docking

Scoring functions(47)

The aim of scoring function is to analyse the reasonable orientations to its receptor. Ideally, the docking score should correlate with binding affinity of ligand to target protein. Scoring functions can be classified in force-field based, knowledge-based, and empirical scoring functions.

Force-field scoring function evaluate the binding affinity by calculating the sum of non-bonded interactions such as electrostatic and van der Waals interaction. The electrostatic terms are calculated by a Coulombic formation. The van der Waals forces are calculated from Lennard-Jones potential function. The docking programme using this method include AutoDock suite. (48) AutoDock force field is essential to calculate the binding energies that the sum of with various molecular mechanic terms such as van der Waals, hydrogen bonding, electrostatics and desolvation energy. Besides, the estimated change in Torsional free energy from unbound ligand to bound ligand is also included. Hence, the complete equation showed below.

Empirical scoring functions have relatively simple energy terms to evaluate. The types of interactions belong to an empirical scoring function include hydrogen bonds, electrostatic interactions, hydrophobic interactions, solvent exclusion volume and electrostatic interactions. Examples of docking programmes using empirical scoring functions include FlexX etc. (49) The formula of this method is shown below.

2.4.3.5 Molecular Dynamics (MD)

Molecular dynamics (MD) is a recent advanced approach to simulate or predict the entire biological system from several hundreds of atoms such as protein in solution or membrane-embedded protein. This approach requires high performance computer together with molecular dynamics algorithm to provide the detailed information on the fluctuations and conformation change of protein.(52)

MD simulation calculates the time-dependent behaviour of molecular system. To investigate potential energy of structure, the force-field algorithm is used to calculate the force acting on every atom.(53) Additionally, MD simulation is based on the second law of Newton, F=ma, where F is the force exerted on the particle, m is its mass and a is its

28 acceleration.(54) For accerelation calculating in every atom in the system, the equation is described below

Fi = miai.

Where Fi is the force exerted on particle i, mi is the mass of particle i, and ai is the acceleration of particle i. For the ith particle; the acceleration at each step is calculated from the negative gradient of the overall potential. The force is the derivative of the potential energy at position r.(54)

r is a vector containing the coordinates for all particles in Cartesian coordinates. The potential energy V function comes from its force field parameters.(54) In general, the biological system is very complex to simulate the virtual environment that mimics the physiological conditions. Hence, MD simulation requires accurate numerical integration of Newton’s second law as described below.

The example of workflow to study in molecular dynamics algorithm is showed in Figure 9.

Figure 2.7. Schematic illustration of molecular dynamics algorithm

2.4.4 In vitro approaches(13) For bone health assessment and the therapeutic response, the bone turnover markers have been available used for clinical studies. Many of these markers are also used in in vitro (cell-based) experiments and in vivo (animal) models. Bone remodeling markers reflect to the metabolic activities of osteoclasts ( resorption) and osteoblasts ( formation) and can be measured in serum or urine. Analytical methods for bone determination of bone remodeling markers include enzyme- linked immunosorbent assay (ELISA), radioimmunoassay (RIA) and electrochemical luminescence.(55) The available biomarkers and analytical methods showed in Table 2.5. Table 2.5. Bone resorption markers(56)

Generally, the main component of bone is type I collagen. During the bone resorbing process, the degraded products are released and pass into the blood circulation and/or urine. Thus, the most bone resorption markers are based on the degraded collagen such as deoxypyridinoline or the cross-linked collagen and telopeptide of type I collagen.

Additionally, osteoclast-specific enzymes such as tartrate-resistant acid phosphatase (TRAP) and cathepsin K also used as biomarkers of bone resorption from osteoclast.

Tartrate-resistant acid phosphatase (TR-ACP or TRAP) is one of specific marker for osteoclast activity. This enzyme expressed and also secreted initially during active bone resorption from osteoclast. TRAP serum level is a useful biomarker for osteoclast activity. More recently, the TRAP5b isoform exhibited as the best indicators for bone resorption and osteoclast number.

Cathepsin K is a cysteine protease enzyme presenting in actively resorbing osteoclast. It cleaves the type I collagen. The serum level of cathepsin K affects the number of osteoclasts and uses as a specific biomarker of osteoclastic activity.

2.4.5 In silico and in vitro approaches to study anti-osteoclastic activity at various proteins

V-ATPase

Hosokawa et al.(57) investigated the signaling between V-ATPases and cytohesin- 2 using in silico docking approaches to study their biological function. Their results indicated that the N-terminal peptides from a-subunit of V-ATPase regulate the enzymatic GEF of cytohesin-2 that is its function as pH-sensing receptor.

Toro et al.(31) presented the enoxacin as a novel V-ATPase-directed osteoclast inhibitor. Virtual screening experiment identified enoxacin interacted to the location of actin binding at subunit B2, resulting in the inhibition of subunit B2 and actin interaction. Over 300000 small molecules in library from National Cancer Institute (NCI) were screened and then there are only 4 compounds possess potently inhibit B2-F-actin interaction. Of these, two compounds were found to inhibit osteoclast formation and activity with an IC50 of 10 µM without affecting cells death. Enoxacin and the other inhibitors identified blocked the formation of multinuclear cells that are positive for tartrate-resistant acid phosphatase activity (TRAP+). Enoxacin demonstrated to inhibit osteoclastic bone resorption in vitro assay, but did not affect osteoblast formation or mineralisation. Papakristos et al.(58) also studied the modelling and mechanics of rotary

V-ATPases using coarse-grain modelling and MD simulations by NAMD 2.8 to investigate the conformational variability and flexibility of their simulation.

Cathepsin K

Cathepsin K is being novel target for osteoporosis treatment. Early development of cathepsin K inhibitors was focused to modify their electrophilic warheads, which interacted to nucleophilic cysteine via reversible or irreversible covalent bond formation. Hence, the main goal of designing cathepsin K inhibitor aims to increase potency and selectivity to their target. However, the irreversible covalent cathepsin K inhiboitor reported to affect potential off-target to various receptors resulting in undesirable side effects such as arthritis, stroke etc. As yet, the development of cathepsin K aims to design their warhead to form reversible covalent bonding linkage to cysteine cathepsin K.(24)

According to the knowledge of mechanism and novel drug development process, the cathepsin K inhibitors have modified on basis of in silico approach. Early cathepsin K inhibitors, e.g. E-64 and related expoxysuccinyl derivatives, were modified for binding cathepsin K enzyme via irreversible covalent bonding. However, there were reported about serious side effects if used chronically such as antigenic and immunologic complications from boosting the immune system. Afterwards, the development of cathepsin K inhibitors intended to modify to be the reversible acting due to the predicted side effects by changing the warhead of cathepsin K inhibitor to be electrophilic warhead such as peptidyl aldehydes, amides, α-keto heterocycles, aliphatic ketones, and nitriles. Additionally, as physiological properties of acidic lysosomes, the inhibitors were designed to contain lipophilic and basic moieties for increasing cell permeability and lysosomotropism but accumulating in acidic lysosomes may result in the off- target inhibition, particularly for cathepsin L, S and V due to their shared about 60% sequence identity. Hence, the strategy to develop novel anti-cathepsin K enzyme is recently shifted to adjust the warhead of inhibitors to be less basic properties or non-basic inhibitors for avoiding the off-target effect.(24)

In order to search for compounds containing with cathepsin K inhibitory effect, the virtual screening is one of essential tools to help in this process. Ravikumar et al.(59) built the 3D-pharmacophore of cathepsin K inhibitors selected from literatures using CATALYST programme. Their 3D-pharmacophore generation exhibited the best hypothesis with correlation coefficient of training set at 0.944 and that of 30 test molecules at 0.909. Then, this modelling was used to screen total 59000 compounds, which were obtained from Maybridge database, to discover new and potent inhibitors toward cathepsin K enzyme. From virtual screening, 2750 molecules were retrieved as hit compounds and consequently 400 compounds were selected to use in docking study due to their fit score>7.00. Their results showed that the pharmacophore modelling agreed very well to the molecular docking.

Qiu et al.(12) investigated the inhibitory effect of Rhizoma Drynariae (RD) extract to cathepsin K, the result showed that the two compounds in RD, Kushennol F (KF) and Sophoraflavanone G (SG), have exerted the anti-osteoclastic activity via inhibiting the function of osteoclast by binding the cathepsin K which is the enzyme for collagen type I degradation. In addition, these compounds also affected to other sites on cathepsin K enzyme. Recent studies found that cathepsin K expressed the other sites for ligand binding which are located near the active site, including the exosite 1 (Tyr87-Gly102) which is responsible for elastin and collagen degradation, and exosite 2 (Gly109-Glu118) which contributes to elastin degradation. Their result also utilised the in silico approach to identify the binding interaction of KF and SG to cathepsin K by using molecular docking and molecular dynamic simulations, which exhibited that both compounds preferred to bind the active site rather than exosite 1 via hydrogen bonding interaction.

Additionally, Lin et al.(60) investigated the nutraceuticals changed microarray mRNA gene expression by using the plant extract from the combination of pomegranates and grape seed whose expressed the antiresorptive activity and bone formation activity. The result suggested that the extract of pomegranates and grape seed containing with ellagic acid and total polyphenols, can be down-regulated the gene expression of cathepsin G (CTSG) and tachychinin receptor 1 (TCR1) which expressed the antiresorptive activity, whereas the combination of all four herbal extracts including promegranate (ellagic acid), grape seed (total polyphenol), quercetin and licorice extract

(glabridin), can up-regulate the gene expression of RUNX2 which demonstrated for bone formation and also down-regulated the expression of that is essential for osteoclastogenesis. This study utilised the microarray for investigation the alteration of mRNA gene expression.

CHAPTER 3

Materials and Methods

3.1 Construction an in-house library of natural compounds

For a library construction, the natural compounds containing the osteoprotective activity are collected and further use to generate a library of herbal compounds for investigating the interaction of ligands to target receptors.

All herbal compounds obtained in this study were reviewed by Che et al.(13) These compounds obtained from various plants, as mentioned before in Table 2.4, that have been reported to possess the osteoprotective activity against osteoporosis or osteonecrosis. These compounds possess the inhibitory effect on bone resorption and/or the stimulatory effect on bone formation, which have been studied in both in vitro and in vivo studies. Herein, 145 compounds are included to construct an in-house library. However, the herbal extracts and medicinal formulas in which the active compounds are unidentified are excluded in this study.

In this study, all selected compounds can be categorised into several groups including flavone and flavonol (Figure 3.1), phenylpropanol (Figure 3.2), triterpenoids (Figure 3.3), flavanones (Figure 3.4), isoflavones (Figure 3.5), monoterpenoid, sesquiterpenoids and diterpenoids (Figure 3.6) and miscellaneous (Figure 3.7).

Figure 3.1. Flavones, flavonols and their glycosides possess osteoprotective property

Figure 3.2. Phenylpropanol derivatives and lignans possess osteoprotective property

Figure 3.3. Triterpenoids possess osteoprotective property

Figure 3.4. Flavanones and their glycosides possess osteoprotective property

Figure 3.5. Isoflavones and their glycosides possess osteoprotective property

Figure 3.6. Monoterpenoid, sesquiterpernoids and diterpenoids possess osteoprotective property

Figure 3.7. Miscellaneous compound types possess osteprotective property

3.2 Virtual Screening (MolecularDocking)

Virtual screening is an essential tool in drug discovery in these days because of spending less time with most effectiveness to define substances which tend to develop as lead-like compounds. To screen numerous compounds in a library, the database with numerous compounds and virtual screening programme are needed.

3.2.1 Target selection

In virtual screening using the target-based concept, the active site of all targets are firstly defined by using the information from the crystal structures derived in protein database. Herein, this study uses protein databank which contains a large number of protein structures, as a database. A criterion for structure selection is overall structures should be the crystal structures from homo sapien with high resolution (Å). All selected targets were obtained in RCSB Protein Databank database. The proposed pdb codes of integrin receptor, V-ATPase and cathepsin K are 5TJ5 and 4X6H, respectively. All 3D structures of proteins showed in Figure 3.8. (A) (B)

Figure 3.8. Selected crystal structures in this study (A) V-ATPase (5TJ5), and (B) Cathepsin K (4X6H)

3.2.2 Preparation of ligands and receptors

Preparation of ligands

The three-dimensional structures of natural compounds containing in a library are downloaded from ZINC12 database. All structures are checked the tautomer, assigned atom/bond types and partial charges. In order to ligand optimisation, overall are optimised by Gaussian 09w programme with same basis set prior to utilising in docking simulation.

Preparation of proteins

Selected target proteins obtained in RSCB protein databank database (www.pdb.com) are prepared by removing all ligands and water molecules inside the complexes. The hydrogen atoms are added, alongside with potential charges. The pretreated structures are minimised using CHARMM-based force field in Discovery Studio 2.5.(50)

3.2.3 Molecular docking approach

In target-based virtual screening, the active site of selected receptor and docking programme are required. To screen a large number of herbal compounds, the rapid docking programme is used in this study such as AutoDock(48), FlexX.(49)

Both ligand and protein are then docked using docking programmes. The main step of docking study can be divided into two steps including searching for the active site by designing grid box or sphere to cover the site of interaction and scoring function for calculating and ranking the docked ligand with highest binding affinity. The active sites of three enzymes are investigated as described previously. The defined binding site of αVβ3 integrin receptor, V-ATPase and cathepsin K, are modified from the study of Choi(61), Mazhab-Jafari(33) and Qiu(12), respectively. The positive controls of V- ATPase and cathepsin K are diphyllin and kushenol F, respectively. The conformation with minimum binding energy in a cluster with most members will be further used in MD simulations. Additionally, all docked compounds are then imposed to identify the three-

45 dimensional pharmacophore generations of each target group using LigandScout(62) programme.

To validate the docking method, the native ligand obtained in the crystal structure is removed. The protein is pretreated followed with docking protocol as described previously, prior to redocking the native ligand into the same binding site. The RMSD value between the pose of redocked ligand compares with that of internal ligand should not more than 1.0 Å. Therefore, this docking protocol can be further use in other ligands. (A) V-ATPase

(B) Cathepsin K

Figure 3.9. Defined binding sites of proteins; (A) cathepsin K, and (B) cathepsin K

3.3 Molecular Dynamics

AMBER(63) is the program for investigating the orientation and movement of ligand while binding in active site of target. Overall docked complexes from previous step were then used in molecular dynamics simulation to determine the stability of ligand during binding toward receptor and conformational changes between ligand and receptor. Binding affinities are calculated by scoring function in AMBER suite program.

Figure 3.10. Schematic diagram of the molecular dynamics simulations. MDs simulate the molecular interactions of ligands and receptors by mimicking the physiological environment in solvated form. There are several steps in MD study, such as energy minimization, equilibrium phase, and molecular dynamic runs. Afterward, the MD results would be analyzed in terms of trajectory and energy contributions

Homology Modelling of Zebrafish (Danio rerio) In the case of cathepsin K of zebrafish (Danio rerio), this crystal structure of zebrafish has not been published on this day. Therefore, a homology modeling of this protein structure should be done by using Modeller version 9.21.

A result of homology modeling predicted a promising three-dimensional structure of cathepsin K of zebrafish. Along with the human cathepsin K, a binding pocket of cathepsin K is well known as a conserved region for ligand binding. Generally, a procathepsin K (amino acids in upstream) would be cleaved before function. Typically, a human cathepsin K expressed this upstream position at 1-114 residues, followed by the active conformation of human cathepsin K structure at 115-329.

Figure 3.11. Binding site of human cathepsin K (64)

3.4 In vitro study Experimental study 1. Zebrafish scales were collected on the left frank of the fish (row A-D). These scales were incubated in lysis buffer (Tris-based buffer, pH 7.5, adding with 1%v/v NP-40). After centrifugation, the supernatant of scale lysate containing with cathepsin K were then collected to analyzed for the cathepsin K activity from the herbal compounds.

Figure 3.12 Left frank of zebrafish illustration for collecting the scales (65)

Figure 3.13 Schematic illustration of cathepsin K extraction from zebrafish scales

2. The herbal compounds in this study were selected from the literature reviews and in silico experiment, including baicalin, catechin, daidzein, hesperidin, EGCG, naringin, quercetin, and rutin. In this experiment, all compounds were dissolved in DMSO and then diluted with the reaction buffer (NaOAc buffer, pH 5.5) to make the final concentrations of 1 µM, 10 µM, and 100 µM respectively. While the Z-Leu-Arg-MCA (a final concentration of 20 µM) was used as the cathepsin K substrate, and Z-Leu-Leu-Leu-H (a final concentration of 10 µM) was used as the specific inhibitor. A decrease of fluorescence intensity was measured by

spectrofluorometer at an excitation wavelength (λex) of 380 nm, and an emission

wavelength (λem) of 460 nm.

Figure 3.14. Flow of anti-cathepsin K screening from zebrafish scales. The experiment investigated the change of substrate color by using the spectrophotoscopy method.

To determine the cathepsin K activity of each assay, the following formula was used: (FL of treated sample –FL of blank) Calculated FL activity (A.U./hr) = incubation time(hr) Where; FL is fluorescence intensity To evaluate the activity of herbal compounds, the fluorescence intensity of a well containing only with substrate was calculated to be 100%, the following formula was used: Cathepsin K activity (%) = 100% x calculated FL of sample / calculated FL of substrate

CHAPTER 4

Results and Discussion

4.1 An in-house library of herbal compounds from literatures As in previous studies, several compounds were used as medicine for osteoporosis treatment, especially Traditional Chinese Medicines (TCM). These compounds obtained the osteoprotective property in terms of anti-resorptive and/or bone formation stimulation in animal models and in vitro study. Che et al. (66) collected various compounds from herbs and categorized into several groups, including flavones, flavonols, flavanones, isoflavones, phenylpropanols, lignans, monoterpenoid, diterpenoids, sesquiterpenoids, triterpenoids, and their derivatives. These compounds were shown in the Table 4.1 below.

Table 4.1. List of herbal compound possesses the osteoprotective property Hernal Compounds 2H-manthraquinone diosgenin luteone salvianolic acid a 3-hydroxyflavone diosmetin maltol salvianolic acid b 5-hydroxyflavone diospongin A maltolglucoside scopoletin 6-hydroxyflavone diospongin B matairesinol scutellarein 6-prenylnaringenin epicatechin morin sesaminone 7-hydroxyflavone ficusal myricetin sophoraflavanone G 8-prenylnaringenin fisetin naringenin sophoricoside acacetin flavanone naringin syringaresinol acteoside flavone neobavaisoflavone tanshinoneI afzelin formononetin nuzhenide tanshinone IIA albiflorin galangin obovatal tanshinone VI amentoflavone geniposide oleosidedies tilianin anthraquinone geniposidic acid oleuropein tyrosol apigenin genistein orobol vanilicacid apigetrin honokiol phloretin wedelolactone apigetrin-6-o-acetyl icariin piperitol wogonin astragalin Icariside ii podophyllotoxin xanthohumol aucubin icaritin psoralen baicalein Ikarisoside A psoralidin

bakuchalcone isoformononetin puerarin bakuchiol isopsoralen quercetin bavachin isorhamnetin quercimeritrin biochaninA isowighteone resveratol cajanin isoxanthohumol rubiadin ceplignan kaempferitrin rubiadin chrysin kaempferol rutin cinnaroside kushenol F sagittatoside A cordycepin kurarinone salidroside corylin laburnetin salvianicacid A cryptotanshinone luteolin salvianolic acid

After that, the three-dimensional structures of these compounds were downloaded from the PubChem and ZINC15 database. These structures were saved and pretreated with HF/6,31G(d,p) method prior to molecular docking. These 3D structures were kept and created as a set of osteoprotective compounds.

4.2 In silico study 4.2.1 Targeted-based virtual screening (Molecular Docking) This study investigated the binding interactions of ligand and receptor. All selected compounds and targets from previous studies were prepared and pretreated for molecular docking simulations.

Cathepsin K The crystal structure of cathepsin k used in this study is crystallized by Borisek et al. Using x-rays crystallography with a resolution of 1.0 angstrom. A binding site of this structure would be similar to that of native ligand, 3XT. The amino acids within the active site, including Gln19, Trp26, Cys25, Asp61, Gy65, Gly66, Tyr67, Ala134, and Leu160 shown in the figure 4.1.

Figure 4.1 Active site of Cathepsin K (PDB: 4X6H)

Method validation (redocking) Prior to utilizing in further simulations, the molecular docking approach should be validated by redocking method. For cathepsin K structure, the original docked compound is N-(Functionalized benzoyl)-homocycloleucyl-glycinonitriles located in the active site of cathepsin K (PDB ID: 4x6h, resolution = 1.0 Å).

Figure 4.2 Human structure of cathepsin K co-crystallized with nitrile-based inhibitors

Molecular docking A total of 108 substances were docked with the pretreated cathepsin K at the active site. The docking result showed in the Table 4.2 and Figure 4.3 in bar plot. The binding modes of all ligands showed in Figure 4.4.

Table 4.2. Docking results of natural substances to cathepsin K (kcal/mol) Substances Dock Score Substances Dock Score Substances Dock Score rutin -12.75 orobol -7.44 2H-manthraquinone -6.22 icariin -11.93 aucubin -7.4 rubiadin -6.21 kaempferitrin -11.35 obovatal -7.39 6-hydroxyflavone -6.16 acteoside -11.23 isowighteone -7.38 apigenin -6.16 salvianolic acid -11.22 myricetin -7.38 flavone -6.12 sagittatoside A -11.17 tanshinone IIA -7.38 genistein -6.12 salvianolic acid b -11.13 matairesinol -7.37 psoralen -6.12 oleuropein -10.85 fisetin -7.33 honokiol -6.10 salvianolic acid a -10.30 cryptotanshinone -7.27 phloretin -6.10 naringin -10.07 baicalein -7.25 flavanone -6.07 cinnaroside -9.86 podophyllotoxin -7.25 anthraquinone -6.06 nuzhenide -9.77 diospongin B -7.23 acacetin -6.05 apigetrin -9.64 luteone -7.22 formononetin -6.00 ikarisoside A -9.58 tanshinone VI -7.22 resveratol -5.92 amentoflavone -9.55 6-prenylnaringenin -7.2 isopsoralen -5.87 apigetrin-6-o-acetyl -9.55 diospongin A -7.17 scopoletin -5.85 quercimeritrin -9.47 epicatechin -7.08 7-hydroxyflavone -5.83 icariside ii -9.02 isoxanthohumol -7.07 cordycepin -5.76 puerarin -8.76 psoralidin -7.02 ceplignan -5.75 sophoricoside -8.66 scutellarein -7.01 salvianicacid A -5.71 afzelin -8.47 isorhamnetin -6.99 bakuchiol -5.7 kurarinone -8.46 oleosidedies -6.96 isoformononetin -5.66 astragalin -8.4 neobavaisoflavone -6.92 piperitol -4.99 geniposide -8.34 bavachin -6.9 maltol -4.62 albiflorin -8.32 morin -6.8 vanilicacid -4.21 kushenol F -8.3 tanshinoneI -6.8 tyrosol -3.96 tilianin -8.17 salidroside -6.78 native -6.51 sophoraflavanone G -8.15 cajanin -6.75 postive -8.3

geniposidic acid -8.1 syringaresinol -6.73

diosgenin -8.07 diosmetin -6.7

ferutinin -7.93 ficusal -6.62

laburnetin -7.93 5-hydroxyflavone -6.59

icaritin -7.84 maltolglucoside -6.51

wedelolactone -7.77 galangin -6.46

bakuchalcone -7.65 xanthohumol -6.42

luteolin -7.65 biochanin A -6.35

wogonin -7.63 kaempferol -6.35

8-prenylnaringenin -7.6 naringenin -6.3

sesaminone -7.58 chrysin -6.27

quercetin -7.54 rubiadin-1-met -6.26

corylin -7.49 3-hydroxyflavone -6.24

-14 -12 -10 -8 -6 -4 -2 0

rutin icariin kaempferitrin acteoside salvianolic acid sagittatoside A salvianolic acid b oleuropein salvianolic acid a naringin cinnaroside nuzhenide apigetrin ikarisoside A amentoflavone apigetrin-6-o-acetyl quercimeritrin icariside ii puerarin sophoricoside afzelin kurarinone astragalin geniposide albiflorin kushenol F tilianin sophoraflavanone G geniposidic acid diosgenin ferutinin laburnetin icaritin wedelolactone bakuchalcone luteolin wogonin 8-prenylnaringenin sesaminone quercetin corylin orobol aucubin obovatal isowighteone myricetin tanshinone IIA matairesinol fisetin cryptotanshinone baicalein podophyllotoxin diospongin B luteone tanshinone VI 6PN diospongin A epicatechin isoxanthohumol psoralidin scutellarein isorhamnetin oleosidedies neobavaisoflavone bavachin morin tanshinoneI salidroside cajanin syringaresinol diosmetin ficusal 5-hydroxyflavone maltolglucoside galangin xanthohumol biochanin A kaempferol naringenin chrysin rubiadin-1-met 3-hydroxyflavone 2H-manthraquinone rubiadin 6-hydroxyflavone apigenin flavone genistein psoralen honokiol phloretin flavanone anthraquinone acacetin formononetin resveratol isopsoralen scopoletin 7-hydroxyflavone cordycepin ceplignan salvianicacid A bakuchiol isoformononetin piperitol maltol vanilic acid tyrosol

Figure 4.3. AutoDock binding energy of herbal compounds with cathepsin K

Binding modes of selected ligands bound cathepsin K

rutin

icariin

kaempferitrin

acteoside

57 sagittatoside A

oleuropein

salvianolic acid a

naringin

cinnaroside

nuzhenide

Figure 4.4. Binding modes of all ligands, rutin, icariin, kaempferitrin, acteoside, sagittatoside A, oleuropein, salvianolic acid a, naringin, cinnaroside and nuzhenide, in cathepsin K binding site

Vacuole-H+-ATPase (V-ATPase) The crystal structure of V-ATPase used in this study was obtained in the RCSB protein databank (www.rcsb.org), PDB ID: 5KNB, with a resolution of 3.251 Å (Figure 4.5). This structure's active site should be the same location of native ligand (adenosine diphosphate; ADP). The binding location of ADP might be between subunit A and subunit B of the V1 domain (Figure 4.6).

Subunit A Subunit B

Subunit A

Subunit B

Figure 4.5. 3D structure of V-ATPase (pdb code: 5KNB) generated by Discovery Visualizer2017R2

Figure 4.6. Binding mode of ADP within its binding site of V-ATPase generated by Discovery Visualizer

Molecular docking A total of 108 substances were docked with the pretreated V-ATPase at the active site (between domain A and B). The docking result showed in the Table 4.3 and Figure 4.7 in bar plot. The binding modes of all ligands showed in Figure 4.8.

Table 4.3 Docking results of natural substances to V-ATPase Substances Dock Substances Dock Substances Dock Score Score Score kaempferitrin -14.00 sesaminone -9.06 oleosidedies -7.93 3- rutin -13.91 xanthohumol -9.05 -7.77 hydroxyflavone sagittatoside A -13.67 cryptotanshinone -9.04 biochanin A -7.76 6- icariin -13.26 isoxanthohumol -8.99 -7.75 hydroxyflavone amentoflavone -12.89 tanshinone VI -8.97 flavanone -7.75 acteoside -12.52 luteone -8.94 honokiol -7.72 cinnaroside -12.45 epicatechin -8.9 flavone -7.62 apigetrin -12.35 ferutinin -8.85 genistein -7.62 ikarisoside A -11.48 matairesinol -8.84 maltolglucoside -7.61 naringin -11.48 baicalein -8.81 anthraquinone -7.58 sophoricoside -11.3 scutellarein -8.75 phloretin -7.42 salvianolic acid a -11.24 kushenol F -8.74 isoformononetin -7.39 7- oleuropein -11.22 bavachin -8.73 -7.21 hydroxyflavone puerarin -11.11 isowighteone -8.72 formononetin -7.17 nuzhenide -10.98 psoralidin -8.72 isopsoralen -7.11 salvianolic acid b -10.95 wogonin -8.67 resveratol -7.05 icarisideii -10.93 diospongin A -8.66 bakuchiol -7.03 afzelin -10.65 diospongin B -8.66 cordycepin -6.72 salvianolic acid -10.62 corylin -8.65 psoralen -6.55 astragalin -10.61 tanshinone I -8.55 ceplignan -6.52 quercimeritrin -10.49 galangin -8.52 scopoletin -6.48 kurarinone -10.4 orobol -8.52 tyrosol -5.79 apigetrin-6-o-acetyl -10.35 salidroside -8.52 salvianic acid A -5.76 sophoraflavanone G -9.99 aucubin -8.49 piperitol -5.71 albiflorin -9.77 morin -8.44 maltol -5.57 icaritin -9.67 fisetin -8.43 vanilic acid -5.12 geniposide -9.6 5-hydroxyflavone -8.42 Native -6.22 tilianin -9.58 diosmetin -8.41 Postive -8.56 bakuchalcone -9.4 ficusal -8.29 luteolin -9.29 kaempferol -8.28 syringaresinol -9.29 acacetin -8.27 diosgenin -9.28 cajanin -8.27 wedelolactone -9.28 rubiadin -8.2 6-prenylnaringenin -9.27 chrysin -8.16 laburnetin -9.2 neobavaisoflavone -8.16 obovatal -9.2 geniposidic acid -8.12 8-prenylnaringenin -9.16 rubiadin -8.12 myricetin -9.15 podophyllotoxin -8.06 tanshinone IIA -9.15 2H-manthraquinone -7.99 isorhamnetin -9.09 apigenin -7.99 quercetin -9.07 naringenin -7.99

-16 -14 -12 -10 -8 -6 -4 -2 0

kaempferitrin rutin sagittatoside A icariin amentoflavone acteoside cinnaroside apigetrin ikarisoside A naringin sophoricoside salvianolic acid a oleuropein puerarin nuzhenide salvianolic acid b icarisideii afzelin salvianolic acid astragalin quercimeritrin kurarinone apigetrin-6-o-acetyl sophoraflavanone G albiflorin icaritin geniposide tilianin bakuchalcone luteolin syringaresinol diosgenin wedelolactone 6-prenylnaringenin laburnetin obovatal 8-prenylnaringenin myricetin tanshinone IIA isorhamnetin quercetin sesaminone xanthohumol cryptotanshinone isoxanthohumol tanshinone VI luteone epicatechin ferutinin matairesinol baicalein scutellarein kushenol F bavachin isowighteone psoralidin wogonin diospongin A diospongin B corylin tanshinone I galangin orobol salidroside aucubin morin fisetin 5-hydroxyflavone diosmetin ficusal kaempferol acacetin cajanin rubiadin-1-met chrysin neobavaisoflavone geniposidic acid rubiadin podophyllotoxin 2H-manthraquinone apigenin naringenin oleosidedies 3-hydroxyflavone biochanin A 6-hydroxyflavone flavanone honokiol flavone genistein maltolglucoside anthraquinone phloretin isoformononetin 7-hydroxyflavone formononetin isopsoralen resveratol bakuchiol cordycepin psoralen ceplignan scopoletin tyrosol salvianic acid A piperitol maltol vanilic acid

Figure 4.7. AutoDock binding contribution of herbal compounds with V-ATPase

Binding modes of selected ligands bound V-ATPase

rutin

icariin

kaempferitrin

acteoside

sagittatoside A

oleuropein

salvianolic acid a

naringin

64 cinnaroside

nuzhenide

Figure 4.8. Binding modes of rutin, icariin, kaempferitrin, acteoside, sagittatoside A, oleuropein, salvianolic acid a, naringin, cinnaroside and nuzhenide within active site of V-ATPase

Table 4.4 Docking results of ligands within V-ATPase and cathepsin K enzymes Rank V-ATPase Dock Score Cathepsin K Dock Score 1 kaempferitrin -14.00* rutin -12.75* 2 rutin -13.91* icariin -11.93* 3 sagittatoside A -13.67* kaempferitrin -11.35* 4 icariin -13.26* acteoside -11.23* 5 amentoflavone -12.89 salvianolic acid -11.22 6 acteoside -12.52* sagittatoside A -11.17* 7 cinnaroside -12.45* salvianolic acid b -11.13 8 apigetrin -12.35 oleuropein -10.85* 9 Ikarisoside A -11.48 salvianolic acid a -10.30* 10 naringin -11.48* naringin -10.07* 11 sophoricoside -11.30 cinnaroside -9.86* 12 salvianolic acid a -11.24* nuzhenide -9.77* 13 oleuropein -11.22* apigetrin -9.64 14 puerarin -11.11 Ikarisoside A -9.58 15 nuzhenide -10.98* amentoflavone -9.55

As in the Table 4.4, the top ten candidates with the lowest binding energy against cathepsin K and V-ATPase were selected to use in the further experiment. The function of cathepsin K exhibited the principal activity on bone degradation, correlated to the number of osteoclasts. The inhibition of cathepsin K decreased bone resorption rate without no effect on the number of osteoclasts, suggesting small disturbances effect for bone remodeling. Hence, the cathepsin K was prioritized over the V-ATPase to select the compounds for later MD study. Additionally, all selected compounds' lowest binding energy should be smaller than that of their positive compounds. These compounds including rutin, icariin, kaempferitrin, acteoside, sagittatoside A, oleuropein, salvianolic acid A, naringin, cinnaroside, and nuzhenide, illustrated the binding modes in Figure 4.4 and Figure 4.8.

4.2.2 Post-MD analysis 4.2.2.1 Trajectory analysis 1) Root-mean-square deviation (RMSD) 2) Root-mean-square fluctuation (RMSF) 3) Dynamic Cross Correlation (DCC) map 4.2.2.2 Energetic analysis Surface Area (PBSA) 1) Poisson–Boltzmann Surface Area (PBSA) 2) Generalized Born Surface Area (GBSA) 3) Pairwise decomposition

Trajectory analysis 1) Root-mean-square deviation (RMSD) This method is calculated for global similarity of protein structure from the initial frame or original structure relative to the reference structure during the whole period of simulation. Basically, RMSD measures the deviation of a target set of coordinates (i.e. a structure) to a reference set of coordinates, with RMSD=0.0 indicating a perfect overlap. RMSD is defined as:

The root-mean-square deviation (RMSD) was calculated from an initial frame of production run which was set as a reference frame, compared to all frames within a whole production simulation. This presented the distribution profile of RMSD values of all complexes by calculating the deviation between atomic positions of reference structure such as carbon (C), nitrogen (N) and oxygen (O) atom, compared to average coordinate set of atoms along the simulation. In this experiment, the RMSD values of ligands and complexes were calculated throughout a 30-ns simulation. This study investigated the binding interaction of the selected ligands against cathepsin K in a time-period of 30 ns. All ligands were able to stabilize within the active site. The RMSD values of 10 complexes were less than 2 angstroms, suggesting the stability of all ligand-bound complexes. These complexes were further used in other analyses.

Figure 4.9. RMSD of all selected ligand in a period of 30 ns Regarding the RMSD values shown in Figure 4.9, some ligands include cinnaroside, icariin, kaempferitrin, kushenol F, nuzhenide, and oleuropein. However, some ligands-bound seemed a bit unstable during this time-period long, especially rutin, sagittatoside A, and salvianolic acid A. Even this result demonstrated the instability of some ligands, the RMSD values of complexes should be considered.

Figure 4.10. RMSD distribution of all complexes in a period of 30 ns After that, the RMSD contribution of the complexes was also calculated for 30 ns to check all the systems' stability. The resultant data demonstrated that a large conformational change throughout these simulations was not observed, suggesting that all complexes were strongly stable during the 30 ns (Figure 4.10). The RMSD value of nuzhenide-bound cathepsin K tended to increase after 30 ns long.

2) Root-mean-square fluctuation (RMSF)

RMSF is used in the study of trajectory in terms of local structural flexibility, thermal stability, and heterogeneity of macromolecules. B-factor is essential factor used in the predicting the protein flexibility, analyzing the binding site and correlated side- chain mobility with conformation, and investigating the protein dynamics (67).

The high RMSF values might indicate the whole structure fluctuates or might reflect only large displacements of a small structural subset within an overall rigid structure. This study investigated the mobility of amino acids within a simulation of 30 ns. The RMSF contribution profiles of all atoms for unbound cathepsin K (black) and all individual ligands (red) are shown in Figure 4.11 and 4.12.

(a) acteoside (b) cinnaroside

(e) kushenol F (f) naringin

(g) nuzhenide (h) rutin

(i) sagittatoside A (j) salvianolic acid a

Figure 4.12. RMSF value of all complexes This analysis explained the movement of all amino acids of the structure. The average movement within a period of 30 ns was calculated from the movement distance change of C-alpha of residues compared to its position at initial frame.

According to a result, the contribution profiles of amino acids in all complexes were not predominantly different. The first 5 residues showed extremely high fluctuation in comparison with other residues. The directly resulted from this 5-residue location in the N-terminal tail, which can freely move along the simulation. Surprisingly, the RMSF value of Met97, Ile113, and Pro114 of most complexes showed a bit lower value than that of the apoprotein, whereas Gly102 showed a slightly higher value than that of apoprotein.

Both RMSD and RMSF are the structural analysis along MD trajectory. However, RMSF is calculated from the averaged over time, indicating a value for each particle, while the average of RMSD is taken over the particles, giving time specific values.

3) Dynamic Cross Correlation (DCC) Map The dynamic cross correlation (DCC) analysis has been extensively applied to quantify the correlation coefficients of motions between atoms. The DCC between the ith and jth atoms is defined by the following equation;

th Where ri(t) denotes the vector of the i atom’s coordinates as a function of time t. While the DCC analysis can provide insight into the correlative motions of atoms, it could overlook some kinds of correlative motions, due to its reliance on displacements from the uniquely determined average coordinate. Generally, DCC has been applied for analysis of backbone fluctuations and domain motions by focusing on Cα atoms (Kasahara, 2014). To elucidate the effect of local interactions on the communications of the entire molecular simulation, this research focused on highly positively correlated pairs.

(a) apoprotein (b) acteoside

(e) kaempferitrin (f) kushenol F

(g) naringin (h) nuzhenide

(i) oleuropein (j) rutin

(k) sagittatoside A (l) salvianolic acid A

Figure 4.13. DCC map of apoprotein (a) and complexes (b-l); (b) acteoside, (c) cinnaroside, (d) icariin, (e) kaempferithrin, (f) kushennol F, (g) naringin, (h) nuzhenide, (i) oleuropein, (j) rutin, (k) sagittatoside A, and (l) salvianolic acid A

This DCC result indicated that the residue pair in range of 80-100 demonstrated a slight change in comparison with those residue pairs of unbound structure. This change indicated the high correlative motion, suggesting these residues had come closer to each other. While the other residue pairs of all systems had no observation change. It emphasizes that the ligand binding with these amino acids showed the configuration change of these amino acids.

The matrix correlation utilized in this work was calculated through CPPTRAJ. The anti-correlated motion was defined as -1 (blue colour) and fully correlated motion was defined as 1 (red colour). As shown in Figure 4.13, the amino acids from 56 to 100

75 in apoprotein represented as anti-correlation which were shown in mostly blue colour. Whilst the DCC map of other complexes in same range of residues showed as a more positive correlation. According to DCC map result, this indicated that the cathepsin K bound with polyphenols resulted in shorter distances between pairs of amino acids in range of 56 to100 than that of apoprotein.

4.2.2.2 Energetic analysis In case of energetic analysis, the binding energy calculations of all complexes were calculated through the usage of MM/PBSA and MM/ GBSA. This calculation was employed by a python module.

1) Molecular Mechanic/Solvation Area (MM/GBSA)

Herein, the binding free energies, ΔG, of all ligands, receptor and complex are calculated through the usage of MM/ GBSA which is a python script available in the AmberTools18 suite program. This analysis is calculated in term of molecular mechanics including van der Waals, electrostatic interaction, hydrogen bond formation and hydrophobic interaction incorporated with Generalized Born Surface Area. GBSA performs an implicit solvent model to calculate the binding free energies within periodic boundary condition. This algorithm simulates the solvent model surrounded the protein which represents as the physiological environment.

This algorithm typically calculates from all extracted snapshots of the production stages in solvated situation. Each snapshot consists of ligand, receptor and complex structures. Finally, ΔGTOT was calculated as shown in Table 4.5.

Table 4.5 MM/GBSA binding free energy component of ligand-protein complex

Compounds Energy components

VDW EGB ESURF ΔGGAS ΔGSOL ΔGTOT Acteoside -40.2989 12.2893 -4.8966 -40.2989 7.3927 -32.9062 Cinnaroside -43.1486 11.7194 -4.8600 -43.1486 6.8594 -36.2892 Icariin -47.1303 14.4525 -5.4292 -47.1303 9.0234 -38.1069 Kaempferitrin -48.4975 14.5882 -5.1839 -48.4975 9.4043 -39.0932 Kushennol F -40.4955 12.8135 -4.9278 -40.4955 7.8857 -32.6098 Naringin -50.3071 15.1527 -5.7032 -50.3071 9.4495 -40.8576 Nuzhenide -54.0503 17.5144 -6.3641 -54.0503 11.1503 -42.8999 Oleuropein -40.7820 12.9162 -4.9835 -40.7820 7.9327 -32.8493 Rutin -41.6943 12.2176 -4.7307 -41.6943 7.4869 -34.2074 Sagittatoside A -47.3248 14.9233 -5.4565 -47.3248 9.4668 -37.8580 Salvianolic acid -35.9442 10.7449 -4.3757 -35.9442 6.3691 -29.5751

Binding energy of each complex was shown in Table. ΔGTOT represented as a

ΔGbinding for complex. In the case of MM/GBSA, nuzhenide exhibited the most binding energy at approximately -42.90 kcal/mol, followed by naringin, kaempferitrin and icariin, respectively.

2) Molecular Mechanic/Poisson-Boltzmann Surface Area (MM/PBSA) MM/PBSA is widely used to calculate the binding free energy for ligand-protein system in an aqueous environment. The solvation contribution of PBSA model is approximated by using a continuum solvent model. This analysis is calculated in term of molecular mechanics including van der Waals, electrostatic interaction, hydrogen bond formation and hydrophobic interaction incorporated with Poisson-Boltzmann Surface

Area. The ΔGTOT of MM/PBSA was calculated as shown in Table 4.6.

Table 4.6 MM/PBSA binding free energy component of ligand-protein complex

Compounds Energy components

VDW EPB ENPOLAR ΔGGAS ΔGSOL ΔGTOT Acteoside -40.2989 15.063 -0.3475 -40.2989 14.7155 -25.5834 Cinnaroside -43.1486 15.6123 -0.1111 -43.1486 15.5012 -27.6474 Icariin -47.1303 17.3103 0.4675 -47.1303 17.7778 -29.3526 Kaempferitrin -48.4975 19.0623 -0.0022 -48.4975 19.0601 -29.4374 Kushennol F -40.4955 17.9139 -0.3783 -40.4955 17.5356 -22.9599 Naringin -50.3071 18.1717 -0.1001 -50.3071 18.0717 -32.2354 Nuzhenide -54.0503 20.5948 0.0827 -54.0503 20.6774 -33.3729 Oleuropein -40.782 15.6497 0.1431 -40.782 15.7927 -24.9893 Rutin -41.6943 15.4244 0.3307 -41.6943 15.7551 -25.9391 Sagittatoside A -47.3248 17.8377 0.4465 -47.3248 18.2842 -29.0406 Salvianolic acid -35.9442 13.6253 0.0209 -35.9442 13.6462 -22.298

Binding energy of each complex was shown in Table. ΔGTOT represented as a

ΔGbinding for complex. In the case of MM/PBSA, nuzhenide exhibited the most binding energy at approximately -33.37 kcal/mol, followed by naringin, kaempferitrin and icariin, respectively.

In comparison of these two methods for energetic analysis, nuzhenide obtained the strong binding contributions in both MM/PBSA and MM/GBSA. The other substances also showed the similar trends. However, the binding free energy calculated from MM/GBSA seems greater than those of MM/PBSA, due to the different algorithm methods.

3) Pairwise decomposition Pairwise decomposition based on MM/GBSA is essential for identifying of residues that have largest effects on binding energy. This experiment investigated the residue decomposition using the python script available in the AmberTools18 suite program. The decompositions of individual residue contribution are shown below.

Acteoside Cinaroside

0 0 -1 -1 -2 -2 -3 -3 -4 -4 -5 -5

Icariin Kaempferitrin

GLY64 GLN21 CYS22 GLY23 CYS25 TRP26 CYS63 GLY65 GLY66 TYR67 ASN161 HIS162 ALA163 LEU209 0 0

-2 -1

-4 -2

-6 -3

-8 -4

Kushenol F Naringin

0 0 -1 -2 -2 -4 -3 -4 -6 -5 -8

Nuzhenide Oleuropein

0 0 -1 -2 -2 -4 -3 -4 -6 -5 -8 -6

Rutin Sagittatoside A

0 0 -2 -2 -4 -4 -6 -6 -8 -8

Salvianolic acid

0 -1 -2 -3 -4 -5

Figure 4.14. Pairwise Decomposition (kcal/mol) of all ligands As in Figure 4.14, the pair decomposition exhibited the binding energy of each amino acid with a cut-off value of -1.0 kcal/mol. According to this criteria, the amino acids, Gly65, Gly66, Tyr67, His162, Leu209, Leu160, Asn161, Ala163, Trp26, Gly64, showed the binding energy with ligand at greater than -1.0 kcal/mol. Tyr67 and Asn161 showed the strong binding affinity in most complexes. However, this analysis depends on the binding pose of ligands and shape of active site residues.

Binding modes of ligands from post-MD simulation Acteoside Acteoside showed the binding interactions through several bonding formations. Regards to decomposition analysis, Tyr67 exhibited the highly binding energy to acteoside, followed by Leu160 and Asn161 with cut-off value at -3. 0 kcal/ mol. Along with the binding mode of acteoside within cathepsin K, acteoside can form hydrogen bonds with Gly23, Cys25 and Gly66. Regards to binding mode of acteoside, this showed that acteoside laid on the pocket of cathepsin K through interactions with hydrophobic amino acids.

Figure 4.15. Binding mode of acteoside from post-MD simulation

Cinnaroside Cinnaroside bound cathepsin K with several hydrophobic amino acids including Asp61, Gly65, Gly66 and Tyr67 that resulted from decomposition analysis. Moreover, the hydrogen bond formations analyzed by visualized program showed that cinnaroside can form hydrogen bond with nitrogen atom of Gly66 and form pi- anion interaction through Asp61.

Figure 4.16 Binding mode of cinnaroside from post-MD simulation

Icariin Icariin uses glycosidic moiety for binding the cathepsin K at S2 subsite. Whilst the core structure itself laid along the groove conformation organized by R- and L- domain. Regards to decomposition result, Asn161 and Tyr67 showed the highly binding energy to icariin with cut-off at -3.0 kcal/mol. To compare with the position of ODN, its position was located far from S2 subsite. Instead, it prefers to bind at the back of cathepsin K through highly hydrophobic interaction with Asn161. Whilst it can form hydrogen bonds with Gly23 and Gly66 and hydrophobic interactions with several surrounding amino acids.

Figure 4.17 Binding mode of icariin from post-MD simulation

Kaempferitrin Kaempferitrin belongs to flavonoid glycoside which is extracted from several plants. Regarding to the energy decomposition, Tyr67 and Asn161 showed the strong binding affinity to ligand with a cut-off value of -3.0 kcal/mol. From binding mode of kaempferitrin within active site of cathepsin K, it exhibited that kaempferitrin can form hydrogen bond with Gly66. Focus on the orientation of ligand, kaempferitrin was fitted on the v-shaped active site of cathepsin K. In addition, due to its large structure, one of phenyl group also bind to the other side of L-domain. The sum of these interactions led to the potent binding affinity to cathepsin K.

Figure 4.18 Binding mode of kaempferitrin from post-MD simulation

Kushenol F Kushenol F was herein used as a positive control for this study due to its potent anti-cathepsin K activity. Focus on the binding interactions of kushenol F, its results indicated that Trp26, Gly66, Asn161 and His162 can form several hydrogen bonds with this ligand. Moreover, ring B of kushenol F interact with Cys25 via π-alkyl interaction, and prenyl moiety of this ligand also bound with amino acids within the active cleft through the hydrophobic interactions. Along with the result from pairwise energy decomposition, it exhibited that Leu160 and Asn161 showed the high- energy decomposition with cut-off value at -3.0 kcal/mol.

Figure 4.19 Binding mode of kushenol F from post-MD simulation

Naringin From decomposition analysis, amino acids namely Gly66, Tyr67 and Leu160, exhibited the strong binding affinity to naringin. Especially, Tyr67 showed the highest affinity at -7.23 kcal/mol. For 2D interactions generated by DS Visualizer 2017R2, the results from these two programs suggested that naringin can form hydrogen bonds with residues including Cys25, Asp61, Gly66 and Asp158. In addition, Tyr67 also form π-π interaction with naringin with strong binding energy.

Figure 4.20 Binding mode of naringin from post-MD simulation

Nuezhenide Nuezhenide was fitted into the v-shaped active site of cathepsin K. Interestingly, in comparison with the binding mode of ODN, the position of nuzhenide within the active site was same location of ODN. The total energy of each residue calculated through pairwise decomposition indicated that Gly66, Tyr67, Leu160, Asn161 and Leu209 showed the higher binding energy to ligand. In addition, nuzhenide can form hydrogen

84 bind formation via residues namely Glu59, Asn60, Tyr67, Asp158 and Asn208, and form hydrophobic interactions through Cys25, Trp26, and Ala163.

Figure 4.21 Binding mode of nuzhenide from post-MD simulation

Oleuropein Oleuropein was located in the active site of cathepsin K similar to the position of nuzhenide. It means that the location of oleuropein was similar as that of ODN which was retrieved from its crystal structure ( PDB ID: 5TDI) . As in binding mode of oleuropein, it suggested that ligand can form hydrogen bond with Gly64, Gly65, Gly66, Leu160 and Asn161 and form π-π stacking interaction with Tyr67. Along with the result from decomposition, Tyr67 was the highest binding energy, followed by Leu160 and Asn161, respectively.

Figure 4.22. Binding mode of oleuropein from post-MD simulation

Rutin Binding mode generated by several programs, rutin was located in the active site of cathepsin K. Rutin possibly interacted with Cys25, Cys63 and Asn161 via hydrogen bonding formations and Tyr67, Ala134 and Ala163 via hydrophobic interactions. Regarding to the energy decomposition, among residues, Asn161 showed the highest energy at -6.32 kcal/mol.

Figure 4.23 Binding mode of rutin from post-MD simulation

Sagittatoside A Sagittatoside A can interact with several residues including Asp158, His162 and Asn208 through hydrogen bond formations. Additionally, this substance also bound to residues Tyr67, Asn161 and Leu209 via hydrophobic interactions. In decomposition analysis, Tyr67, Leu160, Asn161 and Leu209 also possessed the higher binding affinity to the ligand.

Figure 4.24 Binding mode of sagittatoside A from post-MD simulation

Homology modelling of zebrafish

Briefly, Modeller 9.21 is an essential program for predicting and constructing a three-dimensional structure of a protein using a crystal structure obtained in the database, such as PDB, as a template. By the method, an obtained primary sequence of the targeted protein was used to screen for the template, which was shared amino acid identity at a high level. After that, this predicted structure was evaluated for accuracy by using a model evaluation via python script or running on a web-based modeller (ModBase). Finally, the targeted protein structure would be constructed and further used in other studies such as molecular docking or molecular dynamics simulations. This study started with screening for a promising template obtained in the modeller database, pdb_95.pir, and pdb_95.bin. This step provided several predicted templates that would be a promising template for the targeted protein sequence. UniProt's result by computational analysis showed that sequence Q568D6 is represented as a promising template to generate the 3D structure by homology method.

Figure 4.25 sequence of cathepsin K (Danio rerio)

Regards to Figure in more detail, this sequence was separated into two domains; inhibitor and protein domain. Amino acids from 30-90 are inhibitor domain (Inhibitor I29), which belongs to peptidase inhibitor. It is almost found in the N-terminal of cysteine peptidase enzyme. This alpha-helical domain runs through the substrate-binding site. Removal of this domain by proteolytic cleavage leads to the activation of the peptidase activity. While the residue from 118-332 exhibited as a cysteine peptidase enzyme.

Afterward, all predicted proteins were ranked to consider a promising template for the target model. A template structure with a high amino acid identity and less resolution (R-factor) would be selected as the template for this study. In zebrafish cathepsin K, a total of 7 protein templates were selected to use in a further step. All selected templates would be ranked and compared to screen for a promising template, indicating that a human cathepsin K (PDB code: 4x6h) was selected as a template in this study due to its high identity 72.43% and resolution of 1.0 angstrom.

Figure 4.26 Weighted pair-group average clustering based on a distance matrix

A total of 5 predicted structures were generated in the final step. The best proper structure was further evaluated using the DOPE score, suggesting the quality of the generated structure. Lower DOPE score noticeable exhibited the better model. Regards to this experiment, model 3 showed the lowest DOPE score, suggesting the proper model for further analysis.

Figure 4.27 Superimposition of human (green color) and zebrafish cathepsin K (blue color)

Figure 4.28. DOPE profile of predicted model in comparison with template

The homology modelling indicated that the cathepsin K model generated from zebrafish sequence (UniProt) demonstrated the high identity to human sequence. Taken together with the structural analysis, the crucial amino acids within the active site of zebrafish model were quite similar to those of human. This analysis supported the use of zebrafish model in vitro experiment, especially cathepsin K inhibitor screening.

4.3 In vitro study

The candidates in this assay were collected from previous experiment, and from the additional literature reviews. The compounds namely baicalin, catechin, daizein, hesperidin, EGCG, naringin, quercetin, and rutin, were used in order to investigate the anti-cathepsin K activity by using scale lysate from zebrafish. This method extracted the cathepsin K from adult zebrafish scale, and measured the cathepsin K activity by fluorescence spectroscopy method.

The result from the in vitro experiment showed in Table 4.7. 1) The calculated FL of the sample treated with herbal compounds compared to calculated FL of substrate well.

Table 4.7 Cathepsin K activity of candidates at different concentrations

Conc baicalin catechin daidzein hesperidin EGCG naringin quercetin rutin 1 uM 122.59 141.26 211.63 73.21 127.35 118.92 130.66 101.35 10 uM 97.79 116.63 148.75 60.25 106.04 103.47 103.11 83.53 100 uM 79.96 95.99 132.58 58.57 99.74 82.78 67.85 53.92

conc. 1 uM conc. 10 uM conc. 100 uM 100.00 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00

Cathepsin K activity Cathepsin K activity (%) 10.00 0.00

Figure 4.29 Cathepsin K activity of herbal compounds

2) Transform the cathepsin K activity to the inhibitory activity %Inhibition = 100 – (100% x calculated FL of sample / calculated FL of substrate) Conc baicalin catechin daidzein hesperidin EGCG naringin quercetin rutin 1 uM -22.59 -41.26 -111.63 26.79 -27.35 -18.92 -30.66 -1.35 10 uM 2.21 -16.63 -48.75 39.75 -6.04 -3.47 -3.11 16.47 100 uM 20.04 4.01 -32.58 41.43 0.26 17.22 32.15 46.08

Table 4.8 Cathepsin K activity of candidates at different concentrations Conc baicalin catechin daidzein hesperidin EGCG naringin quercetin rutin 1 uM 0.00 0.00 0.00 26.79 0.00 0.00 0.00 0.00 10 uM 2.21 0.00 0.00 39.75 0.00 0.00 0.00 16.47 100 uM 20.04 4.01 0.00 41.43 0.26 17.22 32.15 46.08

50 40 30 20 100 uM 10

0 1 uM % Cathepsin Cathepsin % Inhibition K

1 uM 10 uM 100 uM

Figure 4.30 Cathepsin K inhibition of herbal compounds

According to the Figure 4.30, only hesperidin showed a slight anti-cathepsin K activity, whereas the other compounds did not express cathepsin K inhibition at a concentration of 1 µM level. At 10 µM level, hesperidin rutin and baicalin exhibited the inhibition of cathepsin K activity with a percentage of 39.75, 16.47, and 2.21, respectively.

While at a 100 µM level, almost compounds showed the inhibitory activity to cathepsin K, except daidzein. Rutin showed the strong anti-cathepsin K activity at 46.08%, followed by hesperidin (41.43%), quercetin (32.15%), baicalin (20.04%), naringin (17.22%), catechin (4.01%), and EGCG (0.26) respectively.

CHAPTER 5

Discussion

Osteoporosis is a bone disease that results in fragile bone and increases the risk of a broken bone. According to WHO reports, osteoporosis has become a public health problem with increased global life expectancy since it is a significant leading cause of morbidity and mortality in osteoporotic patients. There are several causes of osteoporosis, such as sex, hormone, heredity. This condition is predominantly caused by the imbalance between bone resorption rate by osteoclasts and the rate of bone formation by osteoblasts. Typically, bone is mostly composed of calcium and follows by other minerals such as phosphate and magnesium. It is also made up of the collagen framework. Incorporation of these substances significantly influences bone strength and flexibility. After the osteoclasts attached bone matrix, lysosomal enzymes and acid phosphatase will be released into the resorbed site underneath the ruffled border. More recent studies investigated osteoclast function to find a novel potential target that would result in a decrease in bone resorption rate. Cathepsin K has become an interesting target for osteoporosis treatment due to its potent collagenase capability. Additionally, there are several reports on cathepsin K and V-ATPase. Cathepsin K is normally functioned in the acidic environment (pH 5.5), resulting in V-ATPase proton release. Thus, we hypothesized that the multi-target inhibitors from natural compounds would be effective and fewer side effects in osteoporotic treatment.

1. Target-based virtual screening of the herbal substances toward cathepsin K and V-ATPase The top ten candidates with the lowest binding energy against cathepsin K and V- ATPase from virtual screening were selected for further experiments. However, cathepsin K possesses a potent collagenase enzyme for bone resorption (68). Additionally, the inhibition of cathepsin K decreased bone resorption rate without affecting the number of osteoclasts (69), suggesting small disturbances effect for bone remodeling. Hence, the

94 cathepsin K was prioritized over the V-ATPase to select the compounds for later MD study. These compounds including rutin, icariin, kaempferitrin, acteoside, sagittatoside A, oleuropein, salvianolic acid A, naringin, cinnaroside, and nuzhenide.

2. Molecular dynamics of candidates All candidates were selected to investigate the dynamic motion on the nanosecond timescale. In trajectory analysis, the RMSD value of complex was less than 2.0 Å, judged to be stable (70). Our experiment showed that the RMSD values of most complexes appeared to be relatively stable. These values indicated the stability of the system for analysis. Even the molecular dynamics of nuzhenide exhibited a gradual increase in the RMSD contributions; the RMSD value was not greater than 2. From the RMSF analysis, the fluctuations of amino acids in all complexes were not predominantly different. The RMSF value of Met97, Ile113, and Pro114 of most complexes showed a bit lower value than that of the apoprotein, whereas Gly102 showed a slightly higher value than that of apoprotein. The DCC map analysis indicated that the cathepsin K bound with the candidates resulted in shorter distances between pairs of amino acids in a range of 56 to100 than that of apoprotein. In energetic analyses, MM/PBSA and MM/GBSA, the pattern of binding free energy of both methods showed a similar trend. To be insight to the substrate binding site of cathepsins, their structures have shown that substrate can bind to active- site cleft through its extended conformation. However, the specificity of cathepsin K substrates or inhibitors indicated that their structures should be long enough to bind residues in S2 pocket. By structure, this S2 subsite consists of several residues with hydrophobic property (71); thereby cathepsin K preferentially accommodates substrates with high hydrophobic property for binding the S2 pocket. In agreement with the decomposition result, residues Gly65, Gly66 and Tyr67 form the S2 pocket. Additionally, cathepsin K has reported for two exosites, namely exosite 1and 2. Exosite 1 exhibits a site for collagen binding, whereas exosite 2 shows a site for elastin degradation (72). To be an insight into the mechanism of cathepsin K, it is responsible for type I collagen degradation. More recently, exosite 1 has been more interested in drug development. Together with numerous clinical reports investigating the serious side

95 effects from post-market drugs, several cathepsin K inhibitors can bind to other targets and lead to serious side effects to other organs, so-called potential off-target.

Figure 5.1. 2D interaction of ODN generated via Discovery Studio Visualizer 2017R2 (A) crystal structure

Odanacatib (ODN) is the novel cathepsin K inhibitor that provides a highlight based on the design and develops the reversible covalent cathepsin K inhibitor by modifying an electrophile warhead of agent for binding the cathepsin. The structure of ODN has been modified to extend to bind amino acids within the S2 site. More recently, this site is investigated to exhibit the selectivity to cathepsin K, so that it would be a light for cathepsin K inhibitor. A 2D interaction pattern of ODN within cathepsin K was generated from Discovery Studio Visualizer 2017R2. Figure 5.1 exhibits the interactions of ODN toward cathepsin K (PDB ID: 5TDI). ODN can form the hydrogen bonds with Gln19, Cys25, Gly66, and Asn161. Significantly, ODN directly forms a covalent bond with the side chain of Cys25 (73). However, ODN was withdrawn in 2017 due to its increasing the risk of stroke.

(A) (B)

Figure 5.2. Exosite 1 (violet) and exosite 2 (purple) of cathepsin K visualized by Discovery Studio Visualizer 2017R2 (A) surface mode, (B) ribbon mode As in the figure above, exosite 1 (violet color) has been recently interested due to its recognition site for type I collagen. Regarding the study of Panwar et al. (74), this site is exhibited as a novel site of development for cathepsin K inhibitor. For natural substances, the tanshinones expressed the cathepsin K inhibitory effect by binding at exosite 1. Dihydrotanshinone-1 (DHT1) is one of the tanshinone groups exhibited the inhibitory to collagenase activity of cathepsin K without any effects on the number of osteoclastic cells. By structure, the amino acids within this site include Val90 to Gly102. While the residues from Tyr110 to Asn117 form exosite 2. Even exosite 2 showed a site for both collagen and elastin degradation, but it less selectivity to inhibit the cathepsin K function. However, these sites provide more valuable information for the development of cathepsin K inhibitor in osteoporotic treatment. In our experiment, the binding mode of cinnaroside showed that it could bind in the S2 and S3 subsite, which was reported to be a specific site for cathepsin K inhibitor. The result from DCC maps also indicated that the complex of cinnaroside bound was altered the correlation profile between amino acids within the L-domain loop. This loop gained more interest since this site represents as a site for collagen binding. Compared with the orientation of odanacatib (ODN), cathepsin K inhibitor, cinnaroside can bind with cathepsin K surface at the same ODN location. These results provided the new findings and enlightened the mechanism of cinnaroside as the potent cathepsin K inhibitor from the herbal substance for osteoporosis treatment.

Qiu et al. (75) investigated the anti-cathepsin K activity of several natural substances using combination approaches, including molecular fishing, molecular docking, and in vitro assay, to find the potential inhibitory activity cathepsin K function. Their findings indicated that kushenol F extracted from Rhizoma drynariae was represented as a potent cathepsin K inhibitor with IC50 value at 8.797 µM. In the case of the binding interaction of kushenol F, its results suggested that kushenol F was fitted within a V-shaped active cleft of enzyme and can form hydrogen bonds with amino acids, including Gly20, Gln21, Cys63, His162, and Trp184. Naringin belongs to the flavonoid family. Structurally, a core structure of aglycone naringenin connected with glycoside moiety at the 7-OH position. Naringin has been reported several beneficial effects on human health, particularly in anti- inflammatory and antioxidative activity. In the case of osteoporosis, naringin exhibited the effect of induced osteoblast differentiation, resulting in reduced bone loss (76). Of particular substrate recognition interests, subsite 2 (S2) shows as the specific site for cathepsin K inhibitor. Molecular modeling studies support that the hydrogen bonds and hydrophobic interaction of nuzhenide are more favorable to preserve an effective affinity towards cathepsin K. Neither nor the interactions of these molecules, the flavonoid with more hydrophobic moieties would be much more potent than that of less hydrophobicity. For the average deviation of nuzhenide-bound protein throughout this simulation compared with unbound protein, the result showed that both simulations' profiles were not significantly different. However, residues in the range of 92 to 101 presented the higher deviation value than that of unbound protein. Surprisingly, these residues were presented to be an exosite 1 in which chondroitin sulfate (CS) was bound for promoting the osteoclast function. Oleuropein, which was extracted from olive oil, possibly obtained the anti- osteoclastic activity by inhibiting osteoclast specific genes such as cathepsin K, MMP-9 (77). In the case of molecular binding interactions, oleuropein indicated that its orientation within the active site of cathepsin K was located in a V-shaped active cleft between L- and R-domain. Compared with other compounds, oleuropein is also bound at the S2 site and beyond with hydrophobic pocket under the S2 site. These interactions supported the favorable position within the cathepsin binding site through the whole simulation.

According to the pairwise decomposition result (Figure 4.14), all selected substances interacted with Gly65, Gly66, Tyr67, His162, Leu209, and almost substances also bound to Leu160 and Asn161. Thereby, these residues were suggested as crucial amino acids for the active-site binding of cathepsin K. Interestingly, among these residues, Tyr67 possessed the highest energy for binding. This finding agrees with some evidence indicating that Tyr67 exhibited as the key residues for ligand-active site binding. Lecaille et al. (71) investigated about the significant role of Tyr67. They indicated that the mutation of Tyr67 predominantly changed the selectivity of subsite 2 (S2).

3. In vitro study of candidates on zebrafish bone research Zebrafish have been interested as a powerful model for bone study. Most bone studies investigated bone metabolism and bone disease in rodent models. The ovariectomized rat (OVX) represents a well-known model for osteoporosis study. Nevertheless, the various limitations of rodent models include expensive and time- consuming for bone induction and repair. The Zebrafish model is exempt from these limitations. They show the incredible benefits over other animal models, including the high similarity of 70% to human genes, rapid development, and easy gene modification. For bone research, zebrafish embryos are a popular model for the study of osteogenesis. The adult zebrafish scales also exhibit essential models for bone metabolism and mineralization, especially in osteoclasts function. As in humans, the early stage of development, the function of scleroblasts (or called osteoblasts in mammal), demonstrates a major bone cell responsible for bone growth. Thus, the adult zebrafish scale is a useful model in this study. The result of in vitro experiment suggests that rutin exhibits strong anti-cathepsin K activity, followed by hesperidin and quercetin. However, more research is needed to clarify the mechanisms underlying the anti- osteoclast activity at the cellular and molecular levels

CHAPTER 6

Conclusions

Osteoporosis is an age-related disease that affects the quality of life, especially the state of health. The anti-resorbing properties of the herbal substances have been investigated for decades. The phytochemical compounds that possess the osteoprotective property exhibited beneficial activities to represent alternative medicines for osteoporotic prevention in early stages. These compounds exhibited the inhibitory effect through several pathways, especially inhibition of osteoclast activity. Of particular interests, cathepsin K became popular in this recent day since the inhibition of enzymes would not influence the number of osteoclasts and apoptosis process of osteoclasts, thus inhibiting the cathepsin K function would not interfered the bone formation by osteoblasts. Additionally, cathepsin K activation requires an acidic environment through the proton release from V-ATPase. However, there are still some aspects involving the mechanism of polyphenols for inhibiting the osteoclast function. This study proposed investigating the inhibitory activity of several the herbal compounds obtained in the literature and providing insight into the molecular interactions of these substances through the usage of in silico experiments, such as molecular docking and molecular dynamics. Our finding suggested that nuzhenide exhibited the strongest binding affinity to the binding site of cathepsin K due to its hydrophobic properties and hydrogen bond- forming with amino acids within the binding site. Whereas the result of in vitro experiment suggests that rutin exhibits strong anti-cathepsin K activity, followed by hesperidin and quercetin. To summarize, our study based on in silico and in vitro approaches provides useful tools for bone research, especially in drug screening.

100

REFERENCES

1. Szulc P, Bouxsein ML. Overview of Osteoporosis: Epidemiology and Clinical Management. Vertebral Fracture Initiative Resource Document. 2011:2.

2. Brömme D, Lecaille F. Cathepsin K Inhibitors for Osteoporosis and Potential Off- target Effects. Expert Opinion on Investigational Drugs. 2009;18(5):585-600.

3. Songpatanasilp T, Sritara C, Kittisomprayoonkul W, Chaiumnuay S, Nimitphong H, Charatcharoenwitthaya N, et al. Thai Osteoporosis Foundation (TOPF) Position Statements on Management of Osteoporosis. Osteoporosis and Sarcopenia. 2016;2(4):191-207.

4. Organization WH. Assessment of Fracture Risk and Its Application to Screening for Postmenopausal Osteoporosis: Report of a WHO Study Group [meeting held in Rome from 22 to 25 June 1992]. 1994.

5. Sato M, Grese TA, Dodge JA, Bryant HU, Turner CH. Emerging Therapies for the Prevention or Treatment of Postmenopausal Osteoporosis. Journal of Medicinal Chemistry. 1999;42(1):1-24.

6. Suriyawongpaisal P, Siriwongpairat P, Loahachareonsombat W, Angsachon T, Kumpoo U, Sujaritputtangkul S, et al. A Multicenter Study on Hip Fractures in Thailand. Journal of the Medical Association of Thailand. 1994;77(9):488-95.

7. Phadungkiat S, Chariyalertsak S, Rajatanavin R, Chiengthong K, Suriyawongpaisal P, Woratanarat P. Incidence of Hip Fracture in Chiang Mai. Journal of the Medical Association of Thailand. 2002;85(5):565-71.

8. Wongtriratanachai P, Luevitoonvechkij S, Songpatanasilp T, Sribunditkul S, Leerapun T, Phadungkiat S, et al. Increasing Incidence of Hip Fracture in Chiang Mai, Thailand. Journal of Clinical Densitometry. 2013;16(3):347-52.

9. Iñiguez-Ariza NM, Clarke BL. Bone Biology, Signaling Pathways, and Therapeutic Targets for Osteoporosis. Maturitas. 2015;82(2):245-55.

10. Abrahamsen B. Adverse Effects of Bisphosphonates. Calcified Tissue International. 2010;86(6):421-35.

101

11. Clarke B. Normal Bone Anatomy and Physiology. Clinical Journal of the American Society of Nephrology. 2008;3(Supplement 3):S131-S9.

12. Qiu ZC, Dong XL, Dai Y, Xiao GK, Wang XL, Wong KC, et al. Discovery of a New Class of Cathepsin K Inhibitors in Rhizoma Drynariae as Potential Candidates for the Treatment of Osteoporosis. International Journal of Molecular Sciences. 2016;17(12):2116.

13. Che CT, Wong M, Lam C. Natural Products from Chinese Medicines with Potential Benefits to Bone Health. Molecules. 2016;21(3):239.

14. Kingston DGI. Modern Natural Products Drug Discovery and its Relevance to Biodiversity Conservation. Journal of Natural Products. 2011;74(3):496-511.

15. Jain KK. Proteomic Technologies for Cancer Target Validation. Drug Discovery Today: Technologies. 2004;1(2):75-8.

16. Hornberg JJ. Chapter 1: Simple Drugs Do Not Cure Complex Diseases: The Need for Multi-targeted Drugs. In: Morphy JR, Harris CJ, editors. Designing Multi- Target Drugs: RSC Publishing; 2012. p. 1-9.

17. Crockett JC, Rogers MJ, Coxon FP, Hocking LJ, Helfrich MH. Bone Remodelling at A Glance. Journal of Cell Science. 2011;124(7):991-8.

18. Kini U, Nandeesh B. Physiology of Bone Formation, Remodeling, and Metabolism. Radionuclide and Hybrid Bone Imaging: Springer; 2012. p. 29-57.

19. Kular J, Tickner J, Chim SM, Xu J. An Overview of the Regulation of Bone Remodelling at the Cellular Level. Clinical Biochemistry. 2012;45(12):863-73.

20. Florencio-Silva R, Sasso GRdS, Sasso-Cerri E, Simões MJ, Cerri PS. Biology of Bone Tissue: Structure, Function, and Factors That Influence Bone Cells. BioMed Research International. 2015;2015:421746.

21. Franz-Odendaal TA, Hall BK, Witten PE. Buried Alive: How Osteoblasts Become Osteocytes. Developmental Dynamics. 2006;235(1):176-90.

22. Rousselle AV, Heymann D. Osteoclastic Acidification Pathways During Bone Resorption. Bone. 2002;30(4):533-40.

102

23. Soysa NS, Alles N. Osteoclast Function and Bone-resorbing Activity: An Overview. Biochemical and Biophysical Research Communications. 2016;476(3):115-20.

24. Rodan SB, Duong LT. Cathepsin K - A New Molecular Target for Osteoporosis. IBMS BoneKEy. 2008;5(1):16-24.

25. Beyenbach KW, Wieczorek H. The V-type H+ ATPase: Molecular Structure and Function, Physiological Roles and Regulation. Journal of Experimental Biology. 2006;209(4):577-89.

26. Marshansky V, Rubinstein JL, Grüber G. Eukaryotic V-ATPase: Novel Structural Findings and Functional Insights. Biochimica et Biophysica Acta (BBA) - Bioenergetics. 2014;1837(6):857-79.

27. Xiao YT, Xiang LX, Shao JZ. Vacuolar H+-ATPase. The International Journal of Biochemistry & Cell Biology. 2008;40(10):2002-6.

28. Cotter K, Stransky L, McGuire C, Forgac M. Recent Insights into the Structure, Regulation and Function of the V-ATPases. Trends in Biochemical Sciences. 2015;40(10):611-22.

29. Qin A, Cheng TS, Pavlos NJ, Lin Z, Dai KR, Zheng MH. V-ATPases in Osteoclasts: Structure, Function and Potential Inhibitors of Bone Resorption. The International Journal of Biochemistry & Cell Biology. 2012;44(9):1422-35.

30. Holliday LS. Vacuolar H+-ATPase: An Essential Multitasking Enzyme in Physiology and Pathophysiology. New Journal of Science. 2014;2014:21.

31. Toro EJ, Ostrov DA, Wronski TJ, Holliday LS. Rational Identification of Enoxacin as a Novel V-ATPase-Directed Osteoclast Inhibitor. Current Protein & Peptide Science. 2012;13(2):180-91.

32. Toei M, Saum R, Forgac M. Regulation and Isoform Function of the V-ATPases. Biochemistry. 2010;49(23):4715-23.

33. Mazhab-Jafari MT, Rohou A, Schmidt C, Bueler SA, Benlekbir S, Robinson CV, et al. Atomic Model for the Membrane-embedded VO Motor of a Eukaryotic V- ATPase. Nature. 2016;539(7627):118-22.

103

34. Delaissé JM, Eeckhout Y, Vaes G. In vivo and In vitro Evidence for the Involvement of Cysteine Proteinases in Bone Resorption. Biochemical and Biophysical Research Communications. 1984;125(2):441-7.

35. Hanada K, Tamai M, Yamagishi M, Ohmura S, Sawada J, Tanaka I. Isolation and Characterization of E–64, a New Thiol Protease Inhibitor. Agricultural and Biological Chemistry. 1978;42(3):523-8.

36. Green GD, Shaw E. Peptidyl Diazomethyl Ketones are Specific Inactivators of Thiol Proteinases. Journal of Biological Chemistry. 1981;256(4):1923-8.

37. Lotinun S, Kiviranta R, Matsubara T, Alzate JA, Neff L, Lüth A, et al. Osteoclast- specific Cathepsin K Deletion Stimulates S1P-dependent Bone Formation. The Journal of Clinical Investigation. 2013;123(2):666-81.

38. Quesne MG, Ward RA, de Visser SP. Cysteine Protease Inhibition by Nitrile- based Inhibitors: A Computational Study. Frontiers in Chemistry. 2013;1.

39. Mukudai Y, Kondo S, Koyama T, Li C, Banka S, Kogure A, et al. Potential Anti- osteoporotic Effects of Herbal Extracts on Osteoclasts, Osteoblasts and Chondrocytes in vitro: BMC Complement Altern Med. 2014;14:29. doi:10.1186/1472-6882-14-29.

40. Li C, Yazawa K, Kondo S, Mukudai Y, Sato D, Kurihara Y, et al. The Root Bark of Paeonia Moutan is a Potential Anticancer Agent in Human Oral Squamous Cell Carcinoma Cells. Anticancer research. 2012;32(7):2625-30.

41. Ekins S, Mestres J, Testa B. In silico Pharmacology for Drug Discovery: Methods for Virtual Ligand Screening and Profiling. British Journal of Pharmacology. 2007;152(1):9-20.

42. Virtual Screening. An Introduction To Chemoinformatics. Dordrecht: Springer Netherlands; 2007. p. 159-81.

43. Sikazwe DMN. The Multi-Target Drug Design Era is Here, Consider it. Drug Design. 2012;1(1):e101.

44. Jose G. Monzon JD. Chapter 11:Combination Agents Versus Multi-Targeted Agents – Pros and Cons. In: Morphy JR, Harris CJ, editors. Designing Multi- Target Drugs: RSC Publishing; 2012. p. 155-80.

104

45. Marianne AG. Protein Structure Prediction in Structure-Based Ligand Design and Virtual Screening. Combinatorial Chemistry & High Throughput Screening. 2009;12(10):940-60.

46. Lionta E, Spyrou G, Vassilatis DK, Cournia Z. Structure-Based Virtual Screening for Drug Discovery: Principles, Applications and Recent Advances. Current Topics in Medicinal Chemistry. 2014;14(16):1923-38.

47. Meng XY, Zhang HX, Mezei M, Cui M. Molecular Docking: A Powerful Approach for Structure-based Drug Discovery. Current Computer-Aided Drug Design. 2011;7(2):146-57.

48. Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, et al. AutoDock4 and AutoDockTools4: Automated Docking with Selective Receptor Flexibility. Journal of Computational Chemistry. 2009;30(16):2785-91.

49. Rarey M, Kramer B, Lengauer T, Klebe G. A Fast Flexible Docking Method using an Incremental Construction Algorithm. Journal of Molecular Biology. 1996;261(3):470-89.

50. Wu G, Robertson DH, Brooks CL, Vieth M. Detailed Analysis of Grid-based Molecular Docking: A Case Study of CDOCKER - A CHARMm-based MD Docking Algorithm. Journal of Computational Chemistry. 2003;24(13):1549-62.

51. Jones G, Willett P, Glen RC, Leach AR, Taylor R. Development and Validation of a Genetic Algorithm for Flexible Docking. Journal of Molecular Biology. 1997;267(3):727-48.

52. Gorb L, Kuz'min V, Muratov E. Application of Computational Techniques in Pharmacy and Medicine: Springer; 2014.

53. Rarey M, Degen J, Reulecke I. Docking and Scoring for Structure-based Drug Design. Bioinformatics-From Genomes to Therapies: Wiley-VCH Verlag GmbH; 2008. p. 541-99.

54. Ciocoiu M. Materials Behavior: Research Methodology and Mathematical Models: Apple Academic Press; 2015.

105

55. Ferreira A, Alho I, Casimiro S, Costa L. Bone Remodeling Markers and Bone Metastases: From Cancer Research to Clinical Implications. BoneKEy reports. 2015;4.

56. Čepelak I, Čvorišćec D. Biochemical Markers of Bone Remodeling-Review. Biochemia Medica. 2009;19(1).

57. Hosokawa H, Dip PV, Merkulova M, Bakulina A, Zhuang Z, Khatri A, et al. The N termini of a-subunit Isoforms are Involved in Signaling between Vacuolar H+- ATPase (V-ATPase) and Cytohesin-2. Journal of Biological Chemistry. 2013;288(8):5896-913.

58. Papachristos K. In silico Investigation of Ion-Pumping Rotary A-and V-type ATPases: Structural and Dynamical Aspects: University of Leeds; 2014.

59. Ravikumar M, Pavan S, Bairy S, Pramod A, Sumakanth M, Kishore M, et al. Virtual Screening of Cathepsin K Inhibitors Using Docking and Pharmacophore Models. Chemical Biology & Drug Design. 2008;72(1):79-90.

60. Lin Y, Kazlova V, Ramakrishnan S, Murray MA, Fast D, Chandra A, et al. Bone Health Nutraceuticals Alter Microarray mRNA Gene Expression: A Randomized, Parallel, Open-label Clinical Study. Phytomedicine. 2016;23(1):18-26.

61. Choi Y, Kim E, Lee Y, Han MH, Kang I. Site-specific inhibition of integrin αvβ3- vitronectin association by a ser-asp-val sequence through an Arg-Gly-Asp binding site of the integrin. PROTEOMICS. 2010;10(1):72-80.

62. Wolber G, Langer T. LigandScout: 3-D Pharmacophores Derived from Protein- Bound Ligands and Their Use as Virtual Screening Filters. Journal of Chemical Information and Modeling. 2005;45(1):160-9.

63. Pearlman DA, Case DA, Caldwell JW, Ross WS, Cheatham TE, DeBolt S, et al. AMBER, A Package of Computer Programs for Applying Molecular Mechanics, Normal Mode Analysis, Molecular Dynamics and Free Energy Calculations to Simulate the Structural and Energetic Properties of Molecules. Computer Physics Communications. 1995;91(1-3):1-41.

106

64. Turk V, Stoka V, Vasiljeva O, Renko M, Sun T, Turk B, et al. Cysteine cathepsins: From structure, function and regulation to new frontiers. Biochimica et Biophysica Acta (BBA) - Proteins and Proteomics. 2012;1824(1):68-88.

65. Sire JY, Girondot M, Babiar O. Marking zebrafish, Danio rerio (cyprinidae), using scale regeneration. The Journal of experimental zoology. 2000;286(3):297- 304.

66. Che C-T, Wong MS, Lam CWK. Natural Products from Chinese Medicines with Potential Benefits to Bone Health. Molecules. 2016;21(3):239-.

67. Kuzmanic A, Zagrovic B. Determination of ensemble-average pairwise root mean-square deviation from experimental B-factors. Biophys J. 2010;98(5):861- 71.

68. Aguda AH, Panwar P, Du X, Nguyen NT, Brayer GD, Brömme D. Structural basis of collagen fiber degradation by cathepsin K. Proceedings of the National Academy of Sciences of the United States of America. 2014;111(49):17474-9.

69. Drake MT, Clarke BL, Oursler MJ, Khosla S. Cathepsin K Inhibitors for Osteoporosis: Biology, Potential Clinical Utility, and Lessons Learned. Endocr Rev. 2017;38(4):325-50.

70. Liu K, Kokubo H. Exploring the Stability of Ligand Binding Modes to Proteins by Molecular Dynamics Simulations: A Cross-docking Study. Journal of Chemical Information and Modeling. 2017;57(10):2514-22.

71. Lecaille F, Chowdhury S, Purisima E, Brömme D, Lalmanach G. The S2 subsites of cathepsins K and L and their contribution to collagen degradation. Protein science : a publication of the Protein Society. 2007;16(4):662-70.

72. Sharma V, Panwar P, O’Donoghue Anthony J, Cui H, Guido Rafael VC, Craik Charles S, et al. Structural requirements for the collagenase and elastase activity of cathepsin K and its selective inhibition by an exosite inhibitor. Biochemical Journal. 2014;465(1):163-73.

73. Law S, Andrault P-M, Aguda AH, Nguyen NT, Kruglyak N, Brayer GD, et al. Identification of mouse cathepsin K structural elements that regulate the potency of odanacatib. Biochemical Journal. 2017;474(5):851-64.

107

74. Panwar P, Søe K, Guido RV, Bueno RV, Delaisse J-M, Brömme D. A novel approach to inhibit bone resorption: exosite inhibitors against cathepsin K. Br J Pharmacol. 2016;173(2):396-410.

75. Qiu Z-C, Dong X-L, Dai Y, Xiao G-K, Wang X-L, Wong K-C, et al. Discovery of a New Class of Cathepsin K Inhibitors in Rhizoma Drynariae as Potential Candidates for the Treatment of Osteoporosis. International journal of molecular sciences. 2016;17(12):2116.

76. Li N, Jiang Y, Wooley PH, Xu Z, Yang S-Y. Naringin promotes osteoblast differentiation and effectively reverses ovariectomy-associated osteoporosis. Journal of Orthopaedic Science. 2013;18(3):478-85.

77. Santiago-Mora R, Casado-Díaz A, De Castro MD, Quesada-Gómez JM. Oleuropein enhances osteoblastogenesis and inhibits adipogenesis: the effect on differentiation in stem cells derived from bone marrow. Osteoporosis International. 2011;22(2):675-84.

108