Extracellular matrix


The extracellular matrix (ECM) is a dynamic structure that contains a complex mixture of molecules, including elastin, collagens, laminins, fibronectin, proteoglycans and other insoluble molecules. The ECM not only provides a structural framework to support cells and divide tissues (Klein et al. 2004a;Mott & Werb 2004), but also acts as a reservoir for biologically active molecules, such as cytokines, chemokines, growth factors and apoptotic ligands. The specific structure and composition of the ECM meets the demands of specialized tissues (Klein et al. 2004a). Upon interaction with cells through receptor-mediating signaling induced by growth factors, cytokines and other cell adhesion receptors, the ECM controls cell behavior and creates an influential cellular environment (Brew, Dinakarpandian, & Nagase 2000;Massova et al. 1998). These ECM-cell interactions are crucial for the normal development and function of the organism. To maintain homeostasis, the remodeling of ECM is highly regulated. It is modulated through proteolytic systems that control the hydrolytic degradation of a variety of ECM molecules. By regulating the composition of the ECM structure, these proteolytic systems play a significant role in the control of signals sent by the ECM components. This affects cell migration, proliferation, differentiation and death. Uncontrolled proteolysis is an important pathogenic mechanism observed in a variety of diseases (Massova et al. 1998). The function of these proteolytic systems is very complex, as more than 500 genes encoding proteases or protease-like proteins exist in the human genome (Puente et al. 2003).

The majority of endopeptidases are divided into four different groups: 1) serine proteases, 2) cysteine proteases, 3) aspartic proteases and 4) metalloproteinases. The metalloproteinases are further divided into five subgroups that include the Metzincin superfamily and the gluzincin, inuzincin, carboxypeptidase and DD carboxypeptidase subgroups. The Metzincin superfamily includes five multigene families, the serralysins, the pappalysins, the astacins, the adamalysins (or ADAMs) and the matrix metalloproteinases (MMPs) (Sternlicht & Werb 2001). The MMPs are also known as matrixins and are the main enzymes responsible for the ECM degradation resulting in ECM turnover. The first MMP was described 45 years ago in experiments designed to explain how a metamorphosing tadpole lost its collagen-rich tail, leading to normal maturation to the frog (Gross & Lapiere 1962). This enzyme was shown to be important for the normal degradation of connective tissue in the tadpole tail. Following that discovery, MMPs were found to be present in all living organisms, from the simplest bacteria to the most complex systems in mammals. The physiological role of MMPs has been proven by experimental work using transgenes encoding MMPs, Tissue Inhibitors of Metalloproteinases (TIMPs), mutagenesis for specific MMPs or TIMPs genes (Vu & Werb 2000). These molecules take part in homeostatic mechanisms, such as tissue restoration, remodeling and repair, which are important in normal biological processes, such as fetal tissue development, organ morphogenesis, blastocyst implantation, ovulation, cervical dilation, post-partum uterine involution, endometrial remodeling, hair follicle cycling, bone remodeling and wound healing. Uncontrolled ECM degradation as a result of disruption of the balance between MMPs and their inhibitors is observed in many pathological conditions, such as in osteoarthritis, rheumatoid arthritis, cancer, atherosclerosis, aneurysms, nephritis, tissue ulcers, fibrosis, heart failure, pulmonary emphysema, CNS diseases, fibrotic lung disease, liver fibrosis, otolaryngological disease and eye diseases.

MMPs constitute a family of 24 known Zn+2 and Ca+2 dependent proteases (hence the name metallo-), active at neutral pH. They have overlapping enzymatic activities with regard to ECM (Newby 2006) and non-matrix (Massova et al. 1998) proteins, as well as overlapping chromosomal regions based most likely on gene duplication. Although the MMP nomenclature has reached number 29 (MMP-29), the actual number of enzymes is smaller, since MMP-4, MMP-5, MMP-6, MMP-22 and MMP-29 have been removed from the MMPs list as a result of duplication (Somerville, Oblander, & Apte 2003). Additionally, the human proteinase first published as MMP-18 is now known as MMP-19, and the nomenclature MMP-18 refers to a Xenopus laevis collagenase (Somerville, Oblander, & Apte 2003).

With several exceptions, MMPs share common characteristics, which are:

1) They are extracellular proteins that hydrolyze protein or proteoglycan components of the ECM. Recent studies have identified that MMP-1 (Galt et al. 2002), MMP-2 (Kwan et al. 2004) and MMP-11 (Luo et al. 2002) are also found intracellularly and that they may degrade intracellular proteins (Nagase, Visse, & Murphy 2006).

2) They all have an active catalytic region that contains the zinc-binding active site.

3) They are activated by “cleavage” of the pro-domain.

4) They all have a “pre” domain-signal domain, which directs MMPs to the secretory pathway.

5) They all have a “pro” domain that maintains latency in the MMP molecule by occupying the active zinc site, making the catalytic site inaccessible to substrates.

6) A hemopexin (or else C-terminal) domain is present in all MMPs, except for MMP-7, MMP-23, MMP-26.

7) A hinge domain is found in all MMPs, except for MMP-7, MMP-21, MMP-23 and MMP-26.

8) Matrix metalloproteinases are mainly secreted in a zymogen-latent form, except in MT-MMPs and MMP-11, MMP-21, MMP-28, which contain furin-like recognition domains in their pro-domains and can be activated intracellularly in the trans-Golgi network by serine proteases of the subtilisin family (Velasco et al. 1999).

9) Their enzymatic activity is inhibited by endogenous tissue inhibitors (TIMPs).

10) Specific sequences of aminoacids characterise each family member. The sequence homology with collagenase 1 (MMP-1), the VAAHExGHxxGxxH sequence at the catalytic domain and the PRCGxPD sequence at the prodomain are very important criteria for the categorization of a proteinase within the MMP family.

11) Activation is accompanied by a loss of MW of about 10.000 (Woessner, Jr. 1991). In MMP-12 and MMP-20, the loss is greater (Hernandez-Barrantes et al. 2001).


MMPs are divided into discrete categories according to: 1) their ability to degrade different substrates and 2) their structural domains.

Based on substrate specificity, MMPs are grouped in six categories: the collagenases (MMP-1, MMP-8, MMP-13 and MMP-18 in Xenopus laevis), the gelatinases (MMP-2 and MMP-9), the stromelysins (MMP-3, MMP-10 and MMP-11), the matrilysins (MMP-7 and MMP-26), the membrane-type matrix metalloproteinases (MT-MMPs) (MMP-14, MMP-15, MMP-16, MMP-17, MMP-24, MMP-25) and others that are not grouped in the aforementioned categories (MMP-12, MMP-19, MMP-20, MMP-21, MMP-23, MMP-27, MMP-28 (Nagase, Visse, & Murphy 2006).

The collagenases (MMP-1, MMP-8, MMP-13) form a subgroup that degrades interstitial collagens I (found in the bones, the ligaments, the tendons, the skin, the eye and the internal organs), II (found in the cartilage, the intervertebral disc, the notochord and the vitreous body of the eye) and III (found in the skin, the blood vessels and the internal organs) at their specific site Gly-Ile/Leu bond in the α chains, three fourths of the way from the N terminus, cutting the collagens in two fragments (3/4 and 1/4). The cleavage products denature to form gelatin at physiologic temperature, which is then degraded by a different MMP subgroup: the gelatinases (Nagase & Woessner, Jr. 1999). Native collagen types IV and V cannot be degraded by collagenases (Billinghurst et al. 1997;Sakai & Gross 1967;Welgus et al. 1982). Collagenases are capable of using, apart from collagens, other ECM and non-ECM components as substrates, and recent studies have shown that MMP-1 activates the PAR1 (the protease-activated receptor, by cleaving the same bond between arginine and serine, as thrombins do, which promotes breast cancer and the invasion of the carcinoma cells in other tissues (Boire et al. 2005). MMP-8 is the only interstitial collagenase that can be stored within neutrophil granules, instead of being synthesised and released upon demand (Jeffrey 1998). When compared with MMP-1, MMP-8, which can be highly glycosylated, degrades collagens I and II at a higher rate and collagen III at a lower rate (Jeffrey 1998). MMP-13 is mainly produced from cartilage and bone during development and from chondrocytes in osteoarthritis (Velasco et al. 2000).

Gelatinase A (72 KD gelatinase, MMP-2) and Gelatinase B (92 KD gelatinase, MMP-9) denature interstitial collagen (gelatins) and elastin (Berton et al. 2000; Mecham et al. 1997; Shipley et al. 1996) as well as collagen IV, which constitutes a major constituent of mature basal lamellae (Klein et al. 2004a). MMP-2 is present both in tumours, inflamed and non-inflamed healthy tissue (Polette et al. 2004), and in addition to gelatin, it digests a wide range of substrates. MMP-2, but not MMP-9, is capable of degrading collagens I, II and III (Allan et al. 1995). MMP-9 has the largest molecular weight within the MMP family due to the fibronectin-like domain and the collagen V-like domain. Normally, it is present in trophoblasts, osteoclasts, neutrophils and macrophages (Vu & Werb 1998), and its synthesis is increased in conditions that require tissue remodelling, such as wound healing, tissue development, angiogenesis and tumour invasion (Polette et al. 2004).

Stromelysins are another group of MMPs (MMP-3, MMP-10 and MMP-11). They were named after the first member of the subgroup MMP-3 because of its stromal cell origin, where it was first identified (Nagase 1998). Their main role is to cleave proteins such as fibronectin, proteoglycans and laminin. Stromelysins -1 and -2 have the same structural homology as collagenases (Nagase 1998), and unlike other MMPs that act in neutral pH, MMP-3 has an acidic pH activity range. MMP-10 is like MMP-3 in terms of molecular weight and in proteoglycanase and collagenase activity (Mannello et al. 2006), although its catalytic ability, is lower than MMP-3 (Visse & Nagase 2003).

MMP-7 and MMP-26 form the subgroup of matrilysins. MMP-7 is mostly expressed in glandular epithelial cells in the pancreas, the renal mesangium, the endometrium, the skin and the uterus and is characterised by a wide substrate specificity for various ECM and non-ECM components (Woessner 1998). It is more active than other MMPs against versican, a chondroitin sulphate proteoglycan found in atherosclerotic plaque, and cleaves many stromelysin substrates (Imai, Shikata, & Okada 1995;Quantin, Murphy, & Breathnach 1989). MMP-26 is expressed in the endometrium and other normal cells, as well as in some carcinomas. Its activity includes several ECM molecules and is largely stored intracellularly (Marchenko et al. 2004). MMP-11 has a minor activity in degrading ECM components, but it cleaves serine proteinases inhibitors (serpins) like other MMPs, which have higher affinities for other ECM substrates. MMP-11 is unique among the other stromelysins as it is released as an active enzyme in the extracellular space, because of the intracellular activation by the pro-hormone convertases furin (Pei & Weiss 1995) (Marchenko et al. 2004)

MT-MMPs are anchored to the plasma membrane and can be actively focussed by the cell to degrade their substrates. Six MT-MMPs have been identified, and, in addition to their degrading role, they also activate latent MMPs (Egeblad & Werb, 2002). They are highly expressed in fibroblasts and various tumour cells (Seiki 1998) and play a major role in modifying the pericellular environment and tumour cell behaviour, as well as in inducing cell migration, invasion and angiogenesis. As with MMP-11, MT-MMPs are activated intracellularly by furin-like protein convertases as they are translocated to the plasma membrane (Brinckerhoff & Matrisian 2002;Egeblad & Werb 2002;).

Macrophage metallo-elastase (MMP-12) is expressed in lung macrophages and in tumours. Its main function is thought to be the degradation of elastin, but it is also effective against type IV collagen, fibronectin and laminin. MMP-19 has been found in the liver by c-DNA cloning and in patients with Rheumatoid Arthritis as a T-cell derived autoantigen (Kolb et al. 1997;Pendas et al. 1997), whilst MMP-20 (enamelysin) is primarily located within newly formed tooth enamel (hence its name) and has been shown to cleave amelogenin (Li et al. 2001b). MMP-22 has been found in chicken fibroblasts, and, although a human homolog has been found on the basis of EST sequences, the function of this enzyme remains unknown (Billinghurst et al. 1997).

Based on the structural domains, MMPs are divided into nine different groups which are shown in 1. MMP-7 and MMP-26 are the matrix metalloproteinases with the smallest number of domains. They only consist of the signal, the prodomain and the catalytic domain. The typical structure consisting of signal domain, propeptide, catalytic, hinge and hemopexin domains is present in MMP-1, MMP-3, MMP-8, MMP-10, MMP-12, MMP-13, MMP-19, MMP-20 and MMP-27. MMP-2 additionally has a fibronectin domain of three type II repeats, and MMP-9 has a fibronectin domain like MMP-2 plus an insert within its hinge region similar to collagen V (the collagen V-like domain). MMP-11 and MMP-28, apart from the four main domains, have a furin-susceptible site. This domain is also present in MMP-21, that was initially found in Xenopus (Yang, Murray, & Kurkinen 1997) and recently in mice and humans (Marchenko, Marchenko, & Strongin 2003). The human ortholog lacks the hinge domain, and, in addition to the furin, it also has a 20-residue unconserved proline-rich insert in the pro-domain (Ahokas et al. 2002). With regard to MT-MMPs, they are divided into two groups, according to their domains. The members of the first MT-MMP group, MMP-14, MMP-15, MMP-16 and MMP-24, have a transmembrane region and a short cytoplasmic tail in their molecules (Jiang et al. 2001;Uekita et al. 2001;Valtanen et al. 2000). The members of the second MT-MMP group, MMP-17 and MMP-25, exhibit a glycophosphatidylinositol anchor (GPI-anchor) containing a short hydrophobic signal that anchors to GPI (Itoh et al. 1999;Jiang et al. 2001;Kojima et al. 2000;Sounni & Noel 2005). The catalytic domain of MT-MMPs includes an 8-aminoacid insertion between βII and βIII strands known as the MT-loop. In the MT1-MMP molecule, the role of the MT loop is known, as it creates a pocket in the catalytic domain of this enzyme that interacts with the αβ loop of TIMP-2 (English et al. 2001). MMP-23 lacks a recognizable signal sequence and contains a short prodomain with a single cysteine residue, which is believed to be part of the cysteine-switch mechanism that controls the latency of this enzyme. This cysteine residue is located in the sequence A-L-C-L-L-P-A, which is different from the sequence P-R-C-G-P-D that is present in the other MMPs (Velasco et al. 1999). Moreover, MMP-23 has a unique C-terminal cysteine-rich domain followed by an immunoglobulin domain. It was initially proposed to be categorized as a type II membrane MMP due to the transmembrane domain at the N-terminal of the propeptide. However, since it has a furin-recognition motif in its propeptide, it is cleaved in the Golgi apparatus by a proprotein convertase (Pei, Kang, & Qi 2000;Visse & Nagase 2003), and MMP-23 is, therefore, activated at the extracellular space without the N terminal signal anchor (Pei, Kang, & Qi 2000).


The prodomain is bound with its round side to the active cleft site of the catalytic domain. The propeptide domain of the MMPs (about 80 aminoacids) interacts with features of the active site cleft. It has a unique PRCG (V/N) PD sequence, common to the MMP family. The Cys (C) within the aforementioned sequence interacts with the Zn+2 in the active site, and this interaction maintains the latency of proMMPs (Kleifeld et al. 2000;Van Wart & Birkedal-Hansen 1990). Specifically, the cysteine residue plays a significant role as a fourth inactivating ligand for the catalytic zinc atom, and as a result, water cannot reach the active region of the catalytic domain to activate the enzyme, and so it is kept in latent form (Boire et al. 2005). This mechanism is known as the “cysteine switch”. For activation to take place, the cysteine-to-zinc switch must be disrupted by proteolysis of the propeptide or by ectopic perturbation of the cysteine-zinc interaction (Van Wart & Birkedal-Hansen 1990). Following this process, the thiol group is replaced by a water molecule, which hydrolyses its propeptide to complete its activation and degrade its substrate by attacking the peptide bonds (Somerville, Oblander, & Apte 2003;Vu & Werb 2000).

The catalytic domain (about 170 aminoacids) contains a conserved zinc-binding motif with three histidines linked to a conserved methionine, which forms a unique “Met-turn structure” ([VAIT]-[AG]-[ATV]-H-E-[FLIV]-G-H-[ALMSV]-[LIM]-G-[LM]-X-H-[SITV]-X-X-X-X-X-[LAFIV]-M, where X is any aminoacid (Becker et al. 1995;Bode et al. 1996;Massova et al. 1998;Springman et al. 1990). The three histidines bind the active zinc, and the methionine residue is located underneath the cavity created by the three histidines, resulting in high hydrophobicity in the area, which increases the binding ability of histidines (Bode et al. 1996;Massova et al. 1998). The catalytic domain of MMPs contains an additional structural zinc ion and at least one calcium ion, which are important for the enzymatic activity and structural stability of the MMP molecules. The distance between the zinc in the active site and the structural zinc is 12 A0.

The C-terminal hemopexin-like domain (about 210 aminoacids) contains a four-bladed β-propeller (each composed of four antiparallel β strands connected in a W-like topology, which is strongly twisted) (Jenne & Stanley 1987;Massova et al. 1998), forming a gap in the middle of the structure resembling a propeller, which is filled with a calcium ion. In the fibronectin domain, each fibronectin repeat contains a hydrophobic pocket that binds the gelatin, acting as an exosite for the binding of collagenous substrates. Through their fibronectin II-like domain, proMMP-2 and proMMP-9 are able to bind the insoluble elastin, and, after the formation of that complex, they are not cleaved from other proteolytic systems. When proMMP-2 is attached to insoluble elastin, it is autocleaved to a 62-kDa active form (Shipley et al. 1996).

MMP expression and activation is a step-by-step process that involves gene transcription, synthesis of inactive MMPs, secretion of MMPs onto the extracellular matrix or, in the case of MT-MMPs, display on the plasma membrane and activation of the zymogens, making MMPs capable of degrading their substrates.

With the exception of MMP-8 and MMP-9, which are stored in neutrophil and eosinophil granules (Kahari & Saarialho-Kere 1999), MMPs are produced when ECM degradation and remodelling are necessary. The signal for MMP transcription comes from soluble factors such as cytokines, growth factors and hormones. Factors that have been identified to induce MMP upregulation include both forms of interleukin 1 (IL-1a, IL-1b), tumour necrosis factor (TNFa), transforming growth factor-b (TGF-b) (Johnatty et al. 1997), epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), platelet-derived growth factor (PDGF) (Uzui et al. 2002), Chlamydia pneumoniae heat shock protein 60 for MMP-9 (Bode et al. 1999), CD40 ligand for MMP-1,-3 and -8 (Bode et al. 1999;Mach et al. 1997) and many others. These factors activate MMP transcription via activation of proto-oncogene proteins that bind to cis-acting elements at the promoter site of MMP genes. The most well known regulatory elements are activating protein 1 (AP-1), activating protein 2 (AP-2), polyoma enhancer A binding protein-3 (PEA-3), the NF-kB binding site, the osteoblastic cis-acting element and the TATA box. AP-1 and PEA-3 interact with members of the Jun (c-Jun, Jun B, Jun D), Fos (c-Fos, Fra-1, Fra-2 and FosB) and E26 (Ets) families respectively. Jun and Fos proteins form homodimers and heterodimers, and Ets transcription factors are helix-turn-helix protein molecules that identify the purine rich element A of PEA3 –CGGA[A/T]. They act alone or, as in the case of Ets molecules, they form complexes with other transcription factors. The regulation of matrix metalloproteinases and their inhibitors.Clark IM, Swingler TE, Sampieri CL, Edwards DR.Int J Biochem Cell Biol. 2008;40(6-7):1362-78.

In addition to soluble factors, cell-cell and cell-matrix interactions also induce MMP expression. TNFa and IL-1 as well as ultraviolet B irradiation (Brenneisen et al. 1998) activate the mitogen activated protein kinase (MAPK) path and other kinases, which regulate intracellular signals from membrane receptors as well as from cell-cell and cell-matrix interactions. It is worth mentioning that three parallel MAP kinase pathways have been described; the p44-p42 or Erk ½ path, the c-Jun N terminal kinase/ stress activated protein kinase path and the p38 Map kinase pathway. Other factors, like IL-4 and corticosteroids suppress MMP expression (Damas et al. 2002;Siwik, Chang, & Colucci 2000).


Several agents, such as thiol-modifying agents, organomercurials, reactive oxygen radicals, a variety of denaturing agents as well as conditions of low pH and high temperature can lead to MMP activation in vitro (Nagase 1997).

In vivo, enzymes that are secreted as proenzymes form enzyme cascades, in which each enzyme is activated by the one that precedes it, and upon activation, activates the enzyme that follows it. MMPs and members of the serine proteases like plasmin, urokinase plasminogen activator (u-PA), and tissue plasminogen activator (t-PA) act together in a cascade. The activation of plasminogen to plasmin is mediated by a tissue-type plasminogen activator (t-PA) bound to fibrin or by a urokinase plasminogen activator bound to a high-affinity cellular receptor (u-PAR) (Visse & Nagase 2003). The binding of u-PA with u-PAR leads to rapid generation of plasmin (Ellis et al. 1999;Miles et al. 1991;Ritchie et al. 1999;Vassalli, Sappino, & Belin 1991). Plasmin activates proMMP-3, and both enzymes in combination activate pro-MMP-1 (Nagase et al. 1991). Plasmin has also been found to activate pro-MMP-7, pro-MMP-9, pro-MMP-10 and pro-MMP-13 (Lijnen 2001), whilst in another cascade, the coagulation factors, thrombin and factor Xa, have been reported to activate pro-MMP-2 (Galis et al. 1997;Rauch et al. 2002;Zucker et al. 1995).

Several MMPs are able to initiate the stepwise activation of pro-MMP-2. The most well-known mechanism of pro-MMP-2 activation takes place at the cell surface, where pro-MMP-2 forms a complex with MT1-MMP and TIMP-2, and a neighbouring MT1-MMP cleaves the prodomain of pro-MMP-2 to activate this molecule (Ellerbroek et al. 2001). The activation of proMMP-2 is also mediated by other MT-MMP molecules, except MT4-MMP. In addition to pro-MMP-2, MT1-MMP activates proMMP-13, a process that is facilitated by the presence of active MMP-2 and independent from TIMP-2 (Knauper et al. 2002;Knauper et al. 1996).

MMP-3 is able to activate the latent form of MMP-1, MMP-7, MMP-8, MMP-9 and MMP-13. As part of the cascade, the activated MMP-7 is also capable of activating proMMP-1, proMMP-9 and pro MMP-13. In the same way, MMP-2 activates pro-MMP-9 and MMP-12 activates pro-MMP-2 and MMP-3 (Knauper et al. 2002).

As indicated above, MT-MMPs, MMP-11, MMP-21 (Jones, Sane, & Herrington 2003), MMP-23 and MMP-28- possess a furin-like domain recognition sequence allowing the intracellular activation by furin-like proprotein convertases in the trans-Golgi network during the transfer of MMPs from the endoplasmic reticulum to the plasma membrane (MT-MMPs) and to the extracellular space (MMP-11,-21,-23,-28) (Hotary et al. 2002;Lohi et al. 2001;Pei & Weiss 1995;Zucker et al. 2003). The furin-recognition motif is located between the propeptide and the catalytic domain and it is RXKR in MMP-11 and MT-MMPs, except for the MT-MMP-4, RRRR in MT-MMP-4 (Pei & Weiss 1995;Santavicca et al. 1996;Sato et al. 1996) and MMP-23 (Ohnishi et al. 2001), RSRR in MMP-21 (Ahokas et al. 2002) and RKKR in MMP-28 (Illman et al. 2003;Marchenko & Strongin 2001).


Inhibition of MMPs occurs naturally as a mechanism to control excessive matrix degradation and tissue remodelling. This is exerted by natural inhibitors, known as the tissue inhibitors of metalloproteinases (TIMPs) and by non-specific enzyme inhibitors present in tissues. TIMPs are a family of four known, 20-30 kDa secreted proteins (Massova et al. 1998;Sternlicht & Werb 2001), consisting of an aminoterminal (about 125 aminoacids) and C- terminal (about 65 aminoacids) domain (Murphy et al. 1991;Williamson et al. 1990). TIMP-1,-2 and -4 can be found in soluble form, whilst TIMP-3 predominantly binds to the ECM (Mannello & Gazzanelli 2001). TIMPs inhibit the enzymatic activity of MMPs by forming high-affinity, non-covalent 1:1 complexes with activated MMPs. Although originally it was thought that TIMPs inhibit all the MMPs non-specifically, it has now been shown that MT-MMPs are less effectively inhibited by TIMP-1 Metalloproteinase inhibitors: biological actions and therapeutic opportunities.Baker AH, Edwards DR, Murphy G. J Cell Sci. 2002 Oct 1;115(Pt 19):3719-27.

. The inhibitory activity of TIMP-3 is not restricted to MMPs; it also inhibits ADAM-17,-10,-12 and aggrecanases(Amour et al. 1998) (Amour et al. 2000).

Several other factors have been shown to play a significant role in MMP inhibition. A GPI-anchored reversion-inducing cysteine-rich protein with Kazal motifs (known as RECK) inhibits the post-transcriptional secretion and activation of MMP-2,-9 and MT1-MMP (Oh et al. 2001). It is believed that a2 macroglobulin acts through the low-density lipoprotein receptor-related protein (LDL-RP), inducing the endocytosis of MMPs (Baker, Edwards, & Murphy 2002).

MMPs are activet in several pathological conditions. In the clinical setting, it is important to identify selective inhibitors that can be used therapeutically to control MMP activity. The use of the natural TIMP inhibitors has disadvantages, such as their high molecular weight, their poor oral bioavailability and their potentially harmful effects on cellular function, which prevent their clinical use. In addition, it has been shown that although TIMPs may improve the prognosis of some categories of tumours, there are paradigms, such as in gastric tumors, renal, bladder, colorectal and breast cancer, where TIMPs are involved in more aggressive cancer progress (Glasspool & Twelves 2001;de Mingo et al. 2007). To overcome these difficulties, synthetic inhibitors of MMPs (MMPIs) have been developed. The most well-known MMPIs are batimastat (BB-94), marimastat (BB-2516), Prinomastat (AG3340) and Tanomastat (BAY12-9566) (Glasspool & Twelves 2001), all of which bind reversibly to the active site of MMPs. Most of the potent inhibitors designed to date bind to theright-side of the zinc of the active site cleft, because left-side binding is much weaker, possibly due to its natural ability to prevent the carboxylate product of substrate cleavage from becoming a potent inhibitor of the enzyme (Skiles, Gonnella, & Jeng 2001).


In addition to their role in degrading the ECM to create paths for cell migration during development and tissue repair and regeneration, MMPs take part in enzyme cascades that lead to the activation of the latent forms of several chemoattractants, chemokines, cytokines, growth factors and other enzymes, which act synergistically with MMPs to change the cell behaviour and the extracellular environment. Because of these functions, MMPs are believed to be key enzymes in the mechanisms of tissue morphogenesis. This is supported by observations that MMP activity is present in the embryo from the very first stages. For implantation to take place, trophoblasts need to invade into the maternal residue. Whilst investigating the role of MMP-9 in implantation, anti-MMP-9 antibody has been used in human trophoblast cultures, resulting in inhibition of the invading function of trophoblasts. Additionally, anti-MMP-9 antibody added to cultures of embryonic kidney explants inhibited the uteric bud development (Lelongt et al. 1997). It is now known that MT1-MMP and TIMP-2 are also very important in branching morphogenesis (Kanwar et al. 1999). With regard to mammary development, misregulation of MMP-3 expression in transgenic mice leads to abnormal gland morphology (Witty, Wright, & Matrisian 1995). Additionally, in vitro inhibition of MMP-2 in the embryonic pancreatic islet epithelial cells prevents their normal organization into islets of Langerhans without affecting their differentiation (Miralles et al. 1998). MMP-2 and MMP-9 synthesised in in vitro human skeletal muscle satellite cells may participate in myogenesis and skeletal muscle regeneration (Murphy et al. 1980). MMP-3 is produced in large quantities during chondrocyte-mediated cartilage remodelling, and MMP-13 and MMP-14 are considered to participate in the induction of bone replacement of cartilage during various developmental stages (Jimenez et al. 2001). MMP-9 is also involved in the development of many neural structures, like the hypophysis, the gagglion cell layer of the retina, the uveal tract (Canete et al. 1995) and also in the regulation of the cell microenvironment in the cerebellum (Lijnen 2001).


Wound healing occurs through a series of overlapping processes which include haemostasis, inflammation, proliferation and remodelling. After injury, blood coagulation takes place and fibrin clots are formed to reduce blood loss. Then, the inflammatory phase follows, where neutrophils, macrophages and lymphocytes are attracted to the region. This leads to a proliferative phase where fibroblasts migrate into the site of injury and re-epithelialization and angiogenesis and formation of granulation tissue occur. Finally, in the remodelling phase, remodelling of the tissue takes place and involves the formation of scar tissue. Many different cell types take part in wound healing and these include fibroblasts, keratinocytes, endothelial cells, neutrophils, macrophages, lymphocytes and mast cells (Lobmann et al. 2004).

Formation of fibrin clots involves activation of the clotting cascade and leads to the degradation of fibrinogen from thrombin. Many cytokine mediators such as PDGF, TGF-b and vascular endothelial growth factor (VEGF) are released locally by leucocytes and play an important role in the induction of inflammation and the accumulation of neutrophils and macrophages at the site of injury. TGF-b1 is produced in its latent form by platelets and is mostly responsible in its active form for the rapid attraction of neutrophils and macrophages. Neutrophils are the first migratory cells to arrive at the wound site (peak 24-48 hours post wounding), targeting dead tissues, bacteria and foreign materials with phagocytosis, in order to protect the organism from any possible infection. They also secrete TNF-a and IL-1, which attract and activate keratinocytes and fibroblasts. At a later stage, macrophages are recruited to the wound site after the neutrophils and gradually replace them. In addition to cytokines and growth factors, degraded components of ECM promote macrophage migration to the wound site, where, upon binding through their integrin receptors to ECM components, they acquire the phenotype of tissue macrophages. Apart from their scavenger role at the site of the wound, they also secrete inducers of the proliferative phase, such as TNF-a, TGF-b, IL-1, IL-6, PDGFs, IGF-1.

The proliferative phase of wound healing comprises re-epithelialization and formation of granulation tissue and involves migration of fibroblasts, keratinocytes and vascular endothelial cells to the wound region from neighbouring tissues. Re-epithelialisation in dermal wounds is characterized by migration of keratinocytes from the basement membrane, just under the clot and in contact with the dermal matrix. During re-epithelialisation in both acute and chronic burn wounds, migrating keratinocytes at the migration front that are not in contact with the basement membrane express MMP-1 (Steffensen, Hakkinen, & Larjava 2001). In contrast, proliferative keratinocytes that are in contact with the basement membrane do not seem to express MMP-1 (Steffensen, Hakkinen, & Larjava 2001). MMP-1 is only secreted from keratinocytes in culture that are in contact with collagen I (Pilcher et al. 1998). Activated keratinocytes express transmembrane receptors known as integrins. One of the integrin receptors, the a2b1, recognises collagen type I and is essential for the migration of keratinocytes on collagen I under the influence of MMP-1 activity (Pilcher et al. 1998). The close connection between collagen I, a2b1 and MMP1 helps the migration of keratinocytes with local and controlled proteolysis (Steffensen, Hakkinen, & Larjava 2001). Additionally, at the site of migrating keratinocytes, MMP-10 (stromelysin-2) is expressed in contrast with stromelysin-1 (MMP-3), which is expressed at the site of proliferative keratinocytes. These are located behind the migrating front and are in contact with the basal lamina. MMP-26 is also found in migrating keratinocytes (Ahokas et al. 2005), and it has been reported that MMP-28 is present together with MMP-3 at the site of proliferating keratinocytes (Lohi et al. 2001;Saarialho-Kere et al. 2002). MMP-9, by its natural ability to degrade collagen IV, the main component of the basement membrane, facilitates the detachment of basal keratinocytes and their migration to the wound site during the early stages of wound healing (Hartlapp et al. 2001;Lohi et al. 2001). The role of MMP-9 in wound healing is strengthened by its capability to cleave collagen VII, found in the structure of the dermal-epidermal anchoring fibrils. Migratory keratinocytes are also characterised by the expression of TIMP-1 molecules, instead of the proliferative keratinocytes, in which both TIMP-1 and TIMP-3 are found. Recent experiments in normal and chronic wounds have shown that TIMP-1 and -3 are found only in normally healing wounds and not in chronic ulcers. The formation of granulation tissue is characterised by the replacement of the fibrin clot and the deposition of ECM at the wound region that includes mainly tenascin, fibronectin, collagen III and I. A percentage of the fibroblasts that migrate to the wound differentiate to myofibroblasts (Moulin et al. 1998). PDGF released from degranulated platelets and TGF-b and FGFs released from macrophages recruit fibroblasts to the wound site, which exhibits features of smooth muscle cells, because they produce smooth muscle actin, responsible for their contractile phenotype. Fibroblasts and myofibroblasts play a significant role in the deposition of ECM, a process known as fibroplasia. During the formation of granulation tissue in vivo, fibroblasts express MMP-1 and MMP-2 near the wound front. This facilitates the organization of the newly synthesized ECM. At this stage, the balance between MMPs and TIMPs favours the ECM formation instead of the remodelling phase where the matrix degradation dominates. During the remodelling phase, the stromal cells, including fibroblasts, keratinocytes, endothelial and inflammatory cells, populate the granulation tissue and are believed to continue their expression of MMP-1,-2,-3,-9,-11,-12 and -14 (Steffensen, Hakkinen, & Larjava 2001). As a new basal lamina is formed at the wound site, MMP-1 expression is reduced and the basement membrane-keratinocytes interactions through hemidesmosomes are stabilised. Hemidesmosomes, formed within keratinocytes, bind them to anchoring fibrils of the subtending dermis. In conclusion, when the cells finish the process of wound healing, a stop signal decreases MMP expression and gradually the MMP levels approach the levels of normal tissues (Sudbeck et al. 1997).


Due to their ability to cleave components of the basement membrane, MMPs play a significant role in invasion and metastasis. Through proteolysis of the basement membrane, tumours can spread locally and distantly (Canete et al. 1995;Chang & Werb 2001;Nelson et al. 2000). In the case of metastatic cancers, by degrading the basement membrane, cells escape from the tumour at the primary location and, through the blood or lymphatic circulation, are anchored to distant organs, where by degradation of the basement membrane and proliferation and formation of new blood vessels create secondary tumours. The role of basement membrane degradation in tumour metastasis was initially suggested in 1980 by studies which indicated that MMP secretion from a melanoma cell line was responsible for degrading basement membrane collagen (Canete, Gui, Linask, & Muschel 1995;Nelson, Fingleton, Rothenberg, & Matrisian 2000a. This has been further confirmed by several clinical and research studies in lung, brain, head and neck, colon, breast, thyroid, ovarian, prostate, and gastric carcinomas (Canete et al. 1995;Johansson, Ahonen, & Kahari 2000;Nelson et al. 2000;Stetler-Stevenson, Aznavoorian, & Liotta 1993).

The thought that inhibition of MMPs could be used as a tool for cancer treatment has led to the development of pharmaceutical compounds to block MMP activities. Batimastat (BB94), a pseudopeptide with structural similarity to collagen that inhibits MMP-1,-2,-3,-7, and -9 (Canete et al. 1995;Hoekstra, Eskens, & Verweij 2001), was the first synthetic inhibitor of MMPs tested in clinical trials. Because of its low oral bioavailability due to poor solubility, intraperitoneal administration was used. Side effects including abdominal pain, fever, asymptomatic elevation of liver enzymes, pain at the site of injection when administered intrapleurally, as well as no significant clinical responses led British Biotech to replace it with another, orally administered MMPI: Marimastat. Clinical trials with Marimastat (BB-2516) did not have significant differences compared to chemotherapy (gencitabine) treatment, but further analysis revealed that patients with pancreatic and gastric cancer lacking distant metastases, who have already been treated with chemotherapy, may benefit from Marimastat (Glasspool & Twelves 2001). Additionally, Marimastat seems to increase survival and delay disease progression in patients with advanced gastric cancer (Bramhall et al. 2002) (Lelongt et al. 1997). Musculoskeletal pain is its most important side effect. Today, non-peptido mimetics like BMS-275291, which is assessed for lung, prostate cancer and Kaposi's sarcoma (Lockhart et al. 2003), and non-peptidic chemicals like tetracyclines (Kwan et al. 2004) are being tested in humans.


Glaucoma is a very common cause of irreversible visual loss that affects millions of people worldwide. It is categorized in the following groups; Primary angle closure glaucoma, Primary open angle glaucoma, Paediatric glaucoma and secondary glaucoma (Iridocorneal Endothelial Syndrome (ICE), Inflammatory Glaucoma, Neovascular Glaucoma, Pigmentary Glaucoma, Pseudoexfoliative Glaucoma, Traumatic Glaucoma). In this section we will outline the important role that MMPs play in the pathogenesis of this condition, as well as the potential use of pharmacological MMP inhibition in preventing scar formation following trabeculectomy to promote aqueous humour outflow and reduction of intra-ocular pressure. This is of prime importance, as aberrant wound healing is the leading cause of trabeculectomy failure.


The aqueous humour is produced by the inner non-pigmented epithelial cells of the ciliary processes upon stimulation by circadian rhythms and it is secreted into the posterior chamber. Outflow of the aqueous humour occurs by two different pathways; 85-95% is cleared through the anterior chamber-trabecular meshwork-Schlemm's canal-episcleral veins route and is pressure dependent, whilst 5-15% is cleared through the anterior chamber-ciliary muscle-supraciliary and suprachoroidal space route (uveoscleral outflow) and is pressure independent.


Studies in cultured human trabecular meshwork (TM) cells have shown that these cells synthesize a wide range of ECM molecules including Collagen I, II, IV, V and VI, proteoglycans, fibronectin, laminin and thrombospondin (Hernandez et al. 1987;Polansky et al. 1984;Schachtschabel et al. 1982;Tripathi et al. 1991;Worthen & Cleveland 1982). In Primary Open Angle Glaucoma (POAG), elevated intraocular pressure (IOP) is one of the major risk factors leading to ganglion cell death. Elevated pressure has been ascribed to changes in the trabecular, and more specifically in the juxtacanalicular meshwork, the site with the highest aqueous humour outflow resistance (Rohen 1983). From all the trabecular ECM components, the highly charged glycosaminoglycans (GAGs) are believed to play a significant role in the outflow resistance and subsequently in the elevation of the IOP and in the development of glaucoma. The TM GAGs have been found to be synthesized and degraded more rapidly than those present in cornea and sclera, and to have an average half-life of approximately 1.5 days (Acott et al. 1985;Acott et al. 1988). The strict regulation of the trabecular ECM is very important, as excessive ECM synthesis or reduced degradation can lead to increased resistance of outflow (or even to obstruction of the intratrabecular outflow channels) (Shields 1992). It has been therefore suggested that proteolytic enzymes such as MMPs are important to maintain normal aqueous circulation through the TM channels (Park et al. 1987;Polansky et al. 1984;Shuman et al. 1988). Human and bovine trabecular meshwork cells have been shown to produce four members of the MMP family, the MMP-1,-2,-3 and -9, which through their proteolytic role may control cleavage of trabecular ECM. They also produce TIMP-1, suggesting that release of this inhibitor may balance MMP activity within the TM (Alexander et al. 1991). The basal expression of MMPs is significantly increased after stimulation of the TM cells with TPA, a broad-spectrum phorbol mitogen. Based on these observations it has been proposed that MMPs and their inhibitors may constitute an important part of the homeostatic mechanism that controls the aqueous humour outflow in vivo. In normal bovine aqueous humor, with the methods of zymography and Western blot analysis, MMP-2 and -9 and two TIMP members – the TIMP-1 and -2 - have been detected. Moreover gelatinolytic bands representing probably MMP-1 and -7 and an additional gelatinolytic band at ~ 100 kDa have been observed (Huang et al. 1996). Kee C et al investigated in aqueous humor collected during surgery from patients with POAG, NTG, CACG and cataract the differences in the total protein concentration and in MMP-2 activity (Kee, Son, & Ahn 1999). They observed double total protein concentration in POAG aqueous samples compared to the NTG, CACG and cataract ones. Additionally, the levels of MMP-2 protein and activity were significantly increased in the aqueous humour of patients with POAG when compared with specimens from patients with cataract (control), CACG and NTG. This increased MMP-2 activity could not be ascribed to tissue reaction to elevated pressure, as patients with CACG and POAG had similar IOP, nor to breakdown of the blood-aqueous barrier, because of the use of pilocarpine. The study suggested that although MMP-2 may have a short-term increasing effect in the aqueous humour outflow by cleaving TM ECM molecules, in a long term it may degrade the basal lamina of TM cells causing TM cell apoptosis and glaucoma. This may be of significance as a reduced number of TM cells has been thought to lead to the pathogenesis of glaucoma. (Kimpel & Johnson 1992;Tschumper & Johnson 1990). In contrast to the above findings, two different studies found significantly reduced MMP-2 activity in aqueous samples from patients with POAG. This supports more reasonably the theory that decreased ECM degradation may facilitate the development of POAG (Maatta et al. 2005;Schlotzer-Schrehardt et al. 2003).

Using an organ culture model of human anterior segment, Bradley JMB et al observed that addition of activated MMP -2,-3 and -9 to perfusion media caused a 160% increase in the aqueous outflow (Bradley et al. 1998). In addition, purified stromelysin injected in another experiment of the same study directly three times at 12 hours intervals into the perfusion channel, induced an increase of the outflow by 140%. Confirmation of the involvement of MMPs in this process was achieved by 40% reduction of the outflow when TIMP-2 was added to the perfusion media. Interestingly, extending the anterior segment exposure to MMPs caused a marked decrease in the aqueous outflow, which further demonstrates that TM ECM turnover plays a major role in regulating the aqueous outflow in vivo (Kee, Son, & Ahn 1999).

Balance between MMPs and TIMPs, as that seen in many other biological processes, play an important role in the regulation of the TM ECM turnover. Especially in POAG, this is demonstrated by observations that levels of TIMP-1 in the aqueous humour are significantly higher in aqueous of POAG patients than in similar specimens from CACG, CG and NVG patients and control samples (Gonzalez Avila et al. 1995). Furthermore, when investigating the effect of aqueous in an in vitro human lung fibroblast migration and collagen production model, the samples from POAG patients decreased collagenase activity and increased collagen production. This response was not observed with aqueous from CG, CACG and NVG patients or control specimens (Gonzalez Avila et al. 1995).

A modern method that is widely used to decrease IOP in glaucoma patients by increasing the aqueous humour outflow is laser trabeculoplasty. Normally, there are 200.000-300.000 TM cells per eye and due to the cell apoptosis its number declines with age, which makes the outflow harder. During the process, 50μm burns with an argon-dye laser are applied to the TM (Parshley et al. 1995;Van Buskirk 1989). This induces division of cells in the anterior, non-filtering region of the TM, at the site of insertion into the cornea beneath the Schwalbe's line (Acott et al. 1989;Bylsma et al. 1988). These cells, which are referred as ‘stem cells', divide the first two days after the laser treatment and most of them within two weeks migrate and repopulate the burn sites at the TM (Acott et al. 1989).

Several studies have supported the implication of MMPs in ECM turnover following laser trabeculoplasty. Parshley et al after culturing anterior segment explants for seven days, applied clinical parameters of argon laser trabeculoplasty to the cultures. Examination of the culture media for protein expression revealed a significant increase in latent and active forms of MMP-9, glycosylated or unglycosylated active and latent forms of MMP-3 and TIMP-1 in the laser treated group, when compared with sham treated and untreated explants. However, levels of active and latent forms of MMP-2 remained almost unchanged in all groups (Parshley et al. 1995). In addition, enhanced mRNA levels coding for MMP-3, MMP-9 and TIMP-1, but not for MMP-2 and TIMP-2 were also observed in the laser treated group. Further investigation by the same group using the same in vitro model confirmed the increase in MMP-3 expression after laser treatment and identified that this increase occurs at the juxtacanalicular site and at the site of TM insertion into the cornea and that TM cells were the source of MMP-3 (Parshley et al. 1996). The authors, based a) on the hypothesis that the accumulation of glycosaminoglycans causes outflow resistance and subsequently aqueous outflow reduction that occurs in glaucoma, and b) on the fact that Laser Trabeculoplasty (LTP) increases MMP-3 levels as well as the aqueous outflow, suggest that MMP-3 has a potential therapeutic application in primary open angle glaucoma.

The possibility of using MMPs as a therapeutic tool to increase both ECM degradation in the juxtacanalicular channels and aqueous outflow has led to the investigation of molecular pathways that regulate expression of MMPs. Initially, it was found that induction of MMP expression in TM after LTP is mediated by IL-1β and TNFa (Bradley et al. 2000). Unlike IL-1β and TNFα, whose production and secretion are increased by LTP, IL-1α is increased but not secreted after LTP and remains in its pro-form within the cytoplasm of TM cells (Bradley et al. 2000). The exact role of IL-1α after LTP is not known, but two mechanisms have been proposed; it either triggers the activation and secretion of IL-1β and/or TNFα intracellularly, or it is released at the site of the laser burn but not in sufficient levels to be detected, where it might activate IL-1β and/or TNFα, which in turn induce MMP transcription (Bradley et al. 2000). An important role for IL-1α on MMP stimulation has been shown by observations that following 72 hours culture of TM cells with IL-1α caused an increase in their production of MMP-3, -9 and TIMP-1, with almost no effects on the production of MMP-2 (Samples, Alexander, & Acott 1993). Other studies have also shown that TNFα upregulates MMP-1, -3, -9 and TIMP-1 expression in cultured porcine TM cells, whilst downregulating their production of TIMP-2 without affecting MMP-2 expression (Alexander & Acott 2001). This was shown to involve a cascade of activation of various factors at membrane and intracellular levels: TNFa activates the TNF-R1 55-kDa membrane receptor, which then creates a complex with the TNF receptor-associated death domain (TRADD) and the TNF receptor activation factor (TRAF-2). The steps after the formation of that complex are partially understood, but it is known that c-Raf-1 is attracted to the plasma membrane, where it becomes autophosphorylated and activated. A cascade of events is then followed, where this factor then phosphorylates and activates Mek, which in turn phosphorylates and activates Erk. One of the functions of activated Erk is to phosphorylate and activate transcriptional activator proteins which induce the transcription and expression of various MMPs and TIMPs (Alexander & Acott 2003). Other investigations have revealed that the trabecular protein kinase C μ isoform (PKCμ) triggers MMP/TIMP transcription by activating a different pathway (Alexander & Acott 2001). This was indicated by observations that formation of the TNF-R1/TRADD/TRAF-2 complex caused PKCμ but not c-Raf-1/Mek/Erk cascade activation, as demonstrated by the fact that G06976, a PKCμ inhibitor, inhibited MMP/TIMP expression without blocking Erk phosphorylation (Alexander & Acott 2001).

Morover, other intracellular pathways of MMP activation have been described, which involve the phosphorylation of two sites (S63 and S73) of c-Jun (Hosseini et al. 2006). In this pathway, JNK induces transcription through the AP-1 sites of both the MMP-3 and MMP-9 promoters (Angel & Karin 1991;Auble, Sirum-Connolly, & Brinckerhoff 1992;James et al. 1993). Phosphorylation of JNK usually occurs through the activated MAPK Kinases MKK4 or MKK7 or both of them (Lawler et al. 1998). Treatment of porcine TM cells with TNFα, IL-1α or IL-1β resulted in increased MMP-3 and -9 production as well as enhanced phosphorylation of MKK4, JNK1/2, c-Jun and ATF-2. Involvement of this activation pathway was clearly shown by the demonstration that JNK inhibitor 2 not only prevented the increased expression of both MMP-3 and -9, but also the phosphorylation of c-Jun and ATF-2. Although ATF-2 is phosphorylated and activated through JNK, a connection between ATF-2 and MMP-3 or -9 transcription has not yet been found. The importance of activation of AP-1 sites in the MMP-3 promoter has been shown in a study performed by Pang IH et al 2002 (Pang et al. 2003). The use of an AP-1 activator - the tBHQ - in human TM cultures stimulated the expression of MMP-3, whilst pre-treatment with SR11302, an AP-1 inhibitor, prevented the increase of MMP-3 caused by tBHQ. Moreover, tBHQ raised the aqueous outflow and had a significant effect in IOP reduction both in glaucomatous and non-glaucomatous (normal) eyes when it was tested in an in vitro human eye perfusion organ culture model. This effect was also blocked by inhibition of AP-1 by SR11302 (Pang et al. 2003).

The importance of MMPs in TM ECM regulation has been further supported by studies in a steroid-induced glaucoma (SIG) model (Snyder RW et al 1993). The authors studied in post mortem trabecular meshwork explants and in TM cell cultures the effect of dexamethasone treatment on MMPs. Their experiments indicated that reduction occurs in the levels of secreted (extracellular) MMP-3 and MMP-9 in the TM explants while the levels of MMP-2 remained the same. In previous investigations in human SIG specimens, excessive accumulation of glycosaminoglycans (GAGs) and wall thickening in the juxtacanalicular tissue has been observed (Rohen, Linner, & Witmer 1973;Spaeth, Rodrigues, & Weinreb 1977). This accumulation of GAGs and ECM has been thought to be responsible for the elevated IOP in SIG (Francois 1984;Hernandez et al. 1983;Johnson, Bradley, & Acott 1990;Knepper et al. 1978;Knepper & McLone 1985;Steely et al. 1992;Ticho et al. 1979;Tripathi, Millard, & Tripathi 1990). They also found that the levels of tPA in the TM explants were reduced. tPA is secreted by TM cells (Shuman et al 1988) and is a potent activator of latent forms of MMPs. Addition of dexamethasone to the TM cell cultures caused reduction in stromelysin and tPA, in contrast to MMP-2 that, as in TM organ explants, was found almost unchanged. As in previous studies, MMP-9 was not detected in TM cultures. This is in agreement with previous findings that dexamethasone is capable of inhibiting IL-1a stimulated expression of MMPs in trabecular explant cultures (Samples, Alexander, & Acott 1993) and other tissues (Frisch & Ruley 1987)(Frisch SM et al 1987, Lefebvre V et al 1990). Due to these findings and to the ECM degrading role of MMPs, connection between the accumulation of ECM and the reduction of MMPs has been suggested and has been thought possibly to be responsible for the pathogenesis of SIG.


Exfoliation or pseudoexfoliation syndrome (XFS) is an age-related disorder characterized by deposition of fibrillar ECM in ocular and non-ocular anatomical structures, but especially in the anterior segment. This pathological condition affects approximately 30% of individuals over the age of 60 and it is often complicated by severe secondary COAG and cataract (Schlotzer-Schrehardt & Naumann 2006). The pathogenesis of XFS is not known, but it has been thought to be caused by a disorder in the basement membrane metabolism (Schlotzer-Schrehardt et al. 1992). It is possible that fibrillar ECM, produced in this condition, blocks the juxtacanalicular channels, causing resistance to aqueous outflow, elevation of IOP and subsequently secondary POAG. In some populations, such as the Scandinavians, XFS constitutes the main etiology of more than half of secondary POAG cases. A common sign often found in patients with XFS is the dark pigmentation of the TM. In addition, anterior to the Schwalbe's line, a pigmented streak known as Sampaolesi's line, is usually observed, due to an inferior deposition of ECM. Significant differences exist between exfoliation glaucoma and POAG. Exfoliation glaucoma is characterised by higher IOP, worse prognosis, poor response to LTP treatment and presents many postoperative complications. It often presents monocularly. A comparative study in aqueous humor samples collected from patients with primary open angle glaucoma (POAG), pseudoexfoliation syndrome (PEX), pseudoexfoliation glaucoma (PEXG) and cataract, revealed changes in MMPs and TIMPs levels, which may facilitate a better understanding of the pathological mechanisms involved in these conditions (Schlotzer-Schrehardt et al. 2003). In this study, high levels of the latent form of MMP-2 and latent and active MMP-12 were observed. With the use of ultrasensitive immunoassays, very low quantities of MMP-3,-7 and -9 were also found. It was of interest that both TIMP-1 and TIMP-2 were also increased and the concentration of TIMP-1 was six to eightfold higher than TIMP-2. Additionally, active MMP-2 was detected in very low levels, covering only 0.3-1.5% of the total MMP-2 protein. In a similar comparative study the differences in the expression of various MMPs and TIMPs in the same four patient groups (POAG, PEX, PEXG and controls (cataract) were tested (Maatta et al. 2005). MMP-14 was detected in all groups in both its soluble active and latent form, although this molecule is usually bound to the cellular membrane. Lower levels of MMP-13 and MMP-8 were found too, mostly in their latent forms. MMP-8 was found in three out of four POAG aqueous samples in its 55-KDa less glycosylated pro and active isoforms. In contrast, the 75-KDa highly glycosilated isoform was only found in PEXG samples(Maatta et al. 2005) .Analyses of aqueous samples of patients with PEX and PEXG have revealed increased levels of total protein in individuals with PEX, PEXG and POAG, when compared with cataract patients, only specimens from PEX and PEXG patients had significantly enhanced levels (Maatta et al. 2005). Furthermore, total MMP-2,-3 and TIMP-1 were increased in POAG aqueous compared to specimens from cataract patients. However, these levels were lower than in PEX and PEXG aqueous samples, in which TIMP-2 and MMP-12 were also increased. Increased TIMP-1 in PEX samples was also observed by Ho SL et al {Ho, 2005 1 /id}. Moreover, they detected elevated TIMP-2 levels in POAG, PEX and PEXG compared with cataract; TIMP-2/MMP-2 ratio was >1 in PEX and PEXG and 0.88 in POAG and only 0.34 in control (cataract) samples in that study. In contrast, Gartaganis S.P et al 2001 (Gartaganis et al. 2002) observed lower TIMP-1 levels in PEX aqueous than in patients with cataract and similar TIMP-2 levels in PEX aqueous and cataract aqueous samples. It was also found that although the levels of total MMP-2 and -3 were increased, the activated form of MMP-2 was significantly decreased, especially in the aqueous collected from PEX and PEXG patients (Schlotzer-Schrehardt et al. 2003). More specifically, the ratio of total to activated MMP-2 increased about five times in PEX samples and about four times in POAG samples compared to cataract aqueous samples. With regards to the serum samples that the writers tested as well, only the levels of MMP-9 were significantly decreased in PEX and POAG samples in comparison to the MMP-9 cataract levels and additionally a mild TIMP-1 increase was observed in the PEX serum samples. Since the differences in the levels of MMPs and TIMPs are also found in patients that have not received any glaucoma medication or laser therapy, the authors suggested that the anterior segment tissues and not the blood-aqueous barrier breakdown or any therapeutic approaches are responsible for the concentration changes of the above mentioned molecules. Therefore, the increased levels of TIMP-1 and -2 and the reduced activity levels of MMP-2 may be responsible for the lower degradation and subsequently the accumulation of the PEX material, the resistance on the aqueous outflow, the IOP elevation and the development of secondary glaucoma. This accumulation of PEX material has been suggested to be responsible for the upregulation of the total MMP-2 and -3 as a signal to degrade PEX material. This increase can't be considered as a cause but a result of PEX and PEXG.

Decreased MMP-2 activity in aqueous from PEX is in accordance with the study of Maata M et al, who observed an increase in TIMP-2 (Maatta et al. 2005). On the contrary, Gartaganis S.P. reported that the gelatinolytic activity in PEX samples was increased by about 60% (Gartaganis et al. 2002). Based on previous studies which suggested that neovascularization occurs in PEX (Brooks & Gillies 1983;Helbig et al. 1994;Ringvold & Davanger 1981), the authors justified raised levels of MMP activity in the facilitation of iris neovascularization. Although there are many discrepancies between these studies, the common conclusion of all is that the increased MMPs (active or latent) are not capable of preventing the accumulation of the PEX material. In comparison with normal aqueous humor, elevated concentrations of TGF-β2 (Tripathi RC et al 1994) and –β1 (Schlotzer-Schrechardt U et al 2001, Koliakos GG et al 2001) have been found in aqueous humor samples from patients with POAG and PEX respectively. TGF-β1 and -2 were found to upregulate genes that were related to secreted proteins or extracellular matrix in trabecular meshwork cells (Zhao et al. 2004). This is of significance, as TGF-β2 has been shown to induce the accumulation of ECM, not only by stimulating production of fibronectin from TM cells and inducing irreversible ECM crosslinking, but also by increasing the plasminogen activator inhibitor 1 (PAI1), which subsequently blocks the tPA/uPA→plasmin cascade leading to MMP activation (Fuchshofer R et al 2003). In this study it was shown that TGF-β2 enhancement decreased MMP-2 activity. Furthermore, both TGF-β1 and –β2 induce the expression of tissue transglutaminase in the TM, responsible for cross-linking of the ECM components. This makes the degradation harder (Welge-Lussen, May, & Lutjen-Drecoll 2000) and promotes the accumulation of ECM.


Induction of MMP downregulation is a potential therapeutic target for treatment of most ocular diseases presenting with increased production of these enzymes. However, in uveoscleral outflow the potential target is the upregulation of MMP production for degradation of ECM in the ciliary muscle. Since the extracellular spaces of ciliary muscle form the first part of the uveoscleral outflow, ECM degradation in this region may provide free space for the uveoscleral outflow.

It is known that prostaglandins are effective in IOP reduction and it has been supported that this function is partially succeeded due to the presence of many prostaglandin receptors, mainly FP and EP2 in the ciliary muscle (Matsuo & Cynader 1992;Ocklind et al. 1996). A relationship between prostaglandin (PG) receptor activation and MMP production has been identified in ciliary muscle cells. Studies have shown that a cascade of intracellular events resulting in MMP production can be triggered by PG binding to FP and EP2 receptors present in ciliary muscle(Matsuo & Cynader 1992;Ocklind et al. 1996) This cascade is initiated by G-protein mediated activation of adenyl-cyclase and phospholypases, which then increase the production of cAMP and inositol 1,4,5 triphosphate (IP3). cAMP in turn triggers the expression of c-Jun and c-Fos, which form homo-and heterodimers that bind to the AP-1 transcription promoter sites and induce expression of MMPs (Schachtschabel, Lindsey, & Weinreb 2000). This is supported by in vitro experiments in which human muscle ciliary cells exposed to PGF2a express high intracellular levels of cAMP (Zhan et al. 1998), as well as by the finding of cAMP related increase of c-Fos and c-Jun (Lindsey et al. 1997;Lindsey, To, & Weinreb 1994). Additionally, cAMP analogues increase c-Fos and MMP-1 expression (Schachtschabel, Lindsey, & Weinreb 2000), whilst PGF2a or other PG analogues were also shown to induce upregulation of MMPs in these cells (Weinreb & Zangwill 1997).

In addition to aqueous humor, the uveoscleral extracellular space also contains several ECM components, including collagen I, III, fibronectin and elastin, whilst the basement membrane enfolding the ciliary muscle cells contains laminin and collagen IV. Topical treatment with PGF2a has been shown to induce reduction of the ciliary ECM in monkey eyes in vivo (Lutjen-Drecoll & Tamm 1988) and reduction in collagen I, III and IV production has been observed after addition of PGF2a to human muscle ciliary cells in vitro (Lindsey J et al 1997) (Lindsey et al. 1997) as well as in monkey eyes in vivo (Sagara et al. 1999). These observations are in accordance with the demonstration that enhanced levels of MMP-1, -2, -3 and -9 may be found in culture supernatants of ciliary muscle cells treated with PGs such as PGF2a, 11-deoxy-PGE1, PhXA85 and latanoprost (Akaishi et al. 2004;Lindsey et al. 1996;Ocklind 1998;Weinreb et al. 1997). A further study involving exposure of human ciliary smooth cell cultures to latanoprost acid, the biological active form of latanoprost, revealed increased mRNA levels of MMP-1,-3 and -9 and a dose-dependent increase of MMP-1 mRNA. However, the levels of MMP-2 mRNA were found to be decreased, which was ascribed to the lack of AP-1 sites in the MMP-2 promoter (Weinreb & Lindsey 2002). Other investigations have suggested that lack of transcriptional elevation of MMP-2 may be associated to intracellular storing of MMP-2 into the ciliary muscle cells (Swallow, Murray, & Guillem 1996) or that PG delay MMP-2 transcription compared to the other three MMPs (Gaton et al. 2001;Swallow, Murray, & Guillem 1996;Taraboletti et al. 2000).

In an attempt to explain the molecular mechanisms involved in MMP induction mediated by PGs, it was found that PKC-Erk1/2 dependent pathways are involved in PGF2a induced MMP-2 secretion and activation, as that seen with MMP regulation in the trabecular meshwork (Husain, Jafri, & Crosson 2005). Moreover, it was shown that mRNA and protein levels of TIMP-1 were significantly increased in cultures exposed to latanoprost for 18 and 24 h but not for 6h (Anthony, Lindsey, & Weinreb 2002). In contrast, mRNA and protein levels of TIMP-2 were only slightly elevated after 6h but not after 18h and 24h culture with this compound. Furthermore they observed a time correlation between MMP-1,-2 and TIMP-1 induction. In addition PKC was shown to be involved in the TIMP-1 increase observed after latanoprost exposure of human ciliary muscle cells (Anthony, Lindsey, & Weinreb 2002).

MMP induction by PG has also been observed in vivo (Gaton et al. 2001). Since early studies had shown that PGF2a Isopropyl Ester had an increasing effect in the uveoscleral outflow in cynomolgus monkeys (Gabelt & Kaufman 1989), the immunoreactivity of MMP-1, -2 and -3 in the uveoscleral pathway following PGF2a Isopropyl Ester treatment was examined in order a potential role of MMPs to be investigated (Gaton et al. 2001). They observed a significantly increased immunostaining for these three MMPs. They also observed a link between the ciliary muscle MMP-2 immunoreactivity and IOP reduction. Based on the above results, it has been suggested that by increasing MMP expression, PGs decrease the resistance of the uveoscleral outflow and subsequently reduce the IOP without affecting the trabecular ouflow (Schachtschabel, Lindsey, & Weinreb 2000). It is of interest that previous work from the same group revealed MMP-1 expression in the normal human uveoscleral pathway (ciliary muscle, iris root and sclera exterior to the ciliary muscle) (Schachtschabel, Lindsey, & Weinreb 2000).

Parallel to the effect of prostanoids on the ciliary body, application of latanoprost and PGF2a also increase MMP and TIMP production in the sclera (part of the uveoscleral outflow), as demonstrated by observations that enhanced immunoreactivity for MMP-1, -2 and -3 can be induced by topical application of PGF2a-isopropyl ester in monkey eyes (Sagara et al. 1999). In addition, topical prostaglandin application resulted in increased levels of MMP-1, -2 and -3, enhanced amount of catabolized collagen and subsequently raised scleral permeability in monkeys (Weinreb 2001). Levels of MMP-1, -2 and -3 were also found to be raised in culture supernatants of scleral tissue exposed to PGF2a or PhXA85 -the active form of latanoprost- (Kim et al. 2001). Further studies showed that human scleral tissue cultured with PGF2a produced significantly higher levels of MMP-1 and -9 mRNAs, whilst latanoprost increased MMP-1 mRNA. A dose dependent increase in mRNA levels of MMP-3 and -10 as well as TIMP-1, -2 and -3 was also observed following treatment of human scleral tissue with latanoprost (Weinreb et al. 2004). The expression of TIMPs may occur as a protective mechanism to control excessive MMP scleral ECM degradation. Although TIMP levels are elevated, these are not sufficient to balance MMP activity, which causes increased ECM degradation that facilitates the aqueous outflow through the uveoscleral pathway. Based on these findings, several prostanoids, especially FP agonists, have been used therapeutically to control IOP (Weinreb et al. 2004).


In clinical cases in which lack of IOP control causes progressive optic neuropathy and loss of visual field despite laser trabeculoplasty and use of conventional drug therapies, glaucoma filtration surgery known as well as trabeculectomy is the treatment of choice (American Academy of Ophthalmology 2005). In trabeculectomy, a partial thickness filter formed by removing limbal tissue from beneath the scleral cup creates a new drainage channel for the aqueous humor. This procedure facilitates IOP control as it lowers the IOP without causing hypotony or other complications associated with full thickness procedures. The effects of trabeculectomy will remain as long as the drainage channel remains functional. This has been found to depend on the wound healing response and the formation of scar tissue at the subconjunctival space, which if excessive, can gradually reduce and finally block the aqueous drainage, causing again elevation of the IOP (Wong et al. 2002).

The wound healing process that takes place in the eye after trabeculectomy presents significant similarities with the dermal wound healing (as described above). After the incision, plasma proteins and blood cells are released into the wounded area and a fibrin clot is formed. Neutrophils and macrophages are then attracted to the area, where they degrade the clot by releasing various degrading enzymes, including MMP-8 and -9. Activation, proliferation and migration of fibroblasts is then observed (Wong et al. 2002). Before activation, fibroblasts are found as quiescent undifferentiated mesenchymal cells known as fibrocytes in low numbers in the subconjunctival connective tissue (Tenon's capsule) (Wong et al. 2002). After their activation, fibroblasts, not only produce large amounts of ECM molecules such as collagens, glucosaminoglucans, elastin, but also various MMP molecules, which further facilitate fibroblast migration and organization of the newly formed ECM. MMP-1, MMP-2, TIMP-1 and TIMP-2 staining has been detected in the cytoplasm of fibroblasts from human subconjunctival connective tissue. Comparison between normal and healing conjunctiva specimens showed that MMP-1 and TIMP-1 were located only in the healing subconjunctival tissue and not in normal subconjunctival tissue or conjunctival epithelium (Kawashima Y et al 1998). Because of these findings, it has been suggested that the MMP-TIMP expression and activation alterations may play an important role in the post-operative cell proliferation and subconjunctival scarring (Kawashima et al. 1998). Moreover, expression of other MMPs and TIMPs, including MMP-1, -2, -3, -9, -14 and TIMP-1 and -2 has been detected in cultured human Tenon's fibroblasts (Mietz et al. 2003). During fibroblast migration over the fibronectin interface, traction forces are generated by fibroblasts in the underlying substrate causing wound contraction (Harris, Stopak, & Wild 1981). Fibrovascular granulation tissue is gradually formed at the site of injury, whilst part of the fibroblast population differentiates into myofibroblasts due to mechanical stress and growth factor stimulation (mainly TGF-β and PDGF). After months of continuous remodeling of the granulation tissue and apoptosis of myofibroblasts, dense collagenous subconjunctival scar tissue is formed. Extended fibrosis and tissue contraction finally block the aqueous outflow channel. This causes loss of function of the bleb with subsequent increase of IOP (Harris, Stopak, & Wild 1981).

Antimetabolites such as mitomycin C (MMC) and 5-Fluorouracil (5-FU) have been used as anti-scarring agents to reduce scarring after trabeculectomy ( 1996;Skuta et al. 1992). Many studies have been conducted to prolong the functioning period of the outflow channel. It has been observed that a single five minute application of 5-FU and mitomycin C reduces the acuity of the healing response, mainly due to suppression of fibroblast proliferation, and prolongs the bleb survival (Doyle et al. 1993;Khaw et al. 1992b;Khaw et al. 1992a;Khaw et al. 1993). However, severe complications may occur after these treatments, including bleb leakages, hypotony, endophthalmitis and excessive cell apoptosis, which can cause irreversible vision loss. On this basis, there is the need for new safer and more effective treatments to reduce subconjunctival scarring.

Using an in vitro model of wound healing, it has been shown that inhibition of MMPs induces inhibition of contraction of collagen I lattices populated by HTF (Daniels et al. 2003a;Porter et al. 1998). Comparison between three MMP inhibitors (MMPIs) – Ilomastat, BB-94 and Cell-Tech- revealed inhibition of gel contraction with the application of all the three MMPIs in a dose-dependent manner, although ilomastat was proven to be the most effective (Daniels et al. 2003a). In addition to inhibiting collagen contraction, ilomastat also reduced collagen production from fibroblasts in a dose-dependent manner (Daniels et al. 2003a). This is of special importance, as excessive collagen production and deposition at the trabeculectomy area is responsible for bleb failure (Cordeiro et al. 2000;Daniels et al. 1998). Due to these observations, MMP inhibition has been considered to be an ideal therapeutic target to prevent excessive scarring after glaucoma filtration surgery. Application of ilomastat in an in vivo trabeculectomy model significantly prolonged the bleb survival and lowered the IOP effect (Wong, Mead, & Khaw 2003). Histological assessment showed reduction of scar tissue in the ilomastat treated group when compared to the control group. In the treated group reduction of scar tissue was accompanied by decreased number of myofibroblasts and reduced apoptosis, as well as a bigger bleb area. Comparison of the antiscarring effects of ilomastat with MMC, the main antiscarring post-trabecular therapy, showed that the ilomastat treated group had similar prolonged bleb survival and IOP lowering results to the MMC treated group and significantly better outcome for both factors than the control group. In addition, the subconjunctival tissue morphology was normal in the ilomastat group but hypocellular in the MMC (Wong, Mead, & Khaw 2005). Although ilomastat has potential advantages for clinical application, such as lack of toxicity (based on the observations during the in vivo experiments), target specificity, possible less side effects and better tolerance than antimetabolites, several problems have to be solved before it can be used therapeutically. Determination of optimum doses and formulation of slow release drug delivery systems would facilitate the administration of ilomastat to control intraocular scarring in humans.


The lamina cribrosa is a sieve-like small section of the sclera at the posterior pole that is formed by fibroelastic lamellae, the cribriform plates, which are placed in order to create channels for the exit of the non-myelinated axons of retinal ganglion cells (RGCs). These axons converge at the optic disc and after their passage from the lamina cribrosa become myelinated. In the cribriform plates, astrocytes are organized in a vertical position to the RGC axon and parallel to the optic nerve head. They are separated from the ECM components of the cribriform plates-proteoglycans, collagen and elastic fibers- by basement membranes (Agapova et al. 2001;Agapova et al. 2003;Anderson 1969).

In POAG, the cause of blindness is the progressive and irreversible loss of RGC axons. Parallel to the loss of axons, cupping of the optic disc – backward bowing- and reorganization of the cribriform plates are observed. Several studies suggest that rearrangement of the cribriform plates occurs because of increased IOP (Johnson et al. 1996;Pena et al. 2001) and damages of the nerve axons (Bellezza et al. 2003), which may occur as a consequence of tissue remodeling (Hernandez & Pena 1997). At the ECM level, elastotic degeneration and reduction of collagen take place at the optic disc region (Agapova et al. 2001;Hernandez 1992;Pena, Mello, & Hernandez 2000;Quigley, Brown, & Dorman-Pease 1991;Varela & Hernandez 1997), whilst activated astrocytes migrate from the basement membranes of the cribriform plates into the nerve beams.

MMPs and TIMPs have been shown to be implicated in the pathogenesis of glaucoma, as they play a crucial role in the remodeling of the optic disc structures and in the survival of RGCs. Staining for MMP-1, -2 and -3 has been detected in both normal and glaucomatous eyes (Yan et al. 2000). In normal eyes, MMP-1 was found in the cytoplasm of a small number of laminar and postlaminar glial cells at the ONH and around the axons. In all of these regions and additionally in prelaminar RGCs, staining for MMP-2 and MMP-3 was also detected. It is worth mentioning that MMP-3 staining was observed in fewer glial cells. In glaucomatous eyes, the intensity of staining and the number of stained cells for these three MMPs were increased. In the case of NPG eyes, stronger staining for MMP-2 was revealed around pial blood vessels and axons than in POAG eyes. Staining for MMP-2 was also found at the reorganized laminar plates and at the pial septae that covers the cavernous spaces, as well as in astrocytes in these regions (Yan et al. 2000). Other investigations have reported low MMP-1 reactivity in the optic nerve region of normal eyes, which has been detected in astrocytes in the prelaminar, laminar and in the postlaminar optic nerve. However, in glaucomatous eyes, MMP-1 staining of reactive astrocytes in the cribriform plates and in the nerve bundles as well as in small vessels located in ONH has been observed (Agapova et al. 2001). In addition, MMP-14 was found to be associated in the ONH with astrocytes and blood vessels located at the lamina cribrosa in normal eyes. Enhanced MMP-14 immunoreactivity was found in glaucomatous eyes in the lamina cribrosa and the post laminar optic nerve, laminar region and post laminar optic nerve of normal eyes has also been shown (Agapova et al. 2001). Unlike that reported by Yan X et al, Agapova OA et al did not find quantitative differences in MMP-2 expression between normal and glaucomatous eyes (Agapova et al. 2001). MMP-3 was not found to be associated with astrocytes and MMP-3 staining was detected only in the blood vessels throughout the ONH region. Staining for MMP-7,-9 and -12 in normal and glaucomatous ONH was not found, but TIMP-1 and -2 staining was detected in astrocytes and RGCs axons in the prelaminar and laminar region, as well as in RGCs axons and astrocytes lining the pial septa, with no significant differences among normal and glaucomatous eyes. Based on these observations, it was suggested that increased MMP-1 expression leads to excessive degradation of collagen fibbers, the main pathological characteristic of the ECM in cribriform plates of glaucomatous eyes. This may facilitate the migration of astrocytes to the nerve beams. MMP-14 is thought to play a role in the transformation of quiescent into active astrocytes, TIMP-1 and -2 are believed that after their production at the RGCs soma in the nerve retina layer, they are transferred to the axons and one of their functions is to protect the axons from MMP-proteolysis. This was supported by studies in an experimental model of glaucoma in monkeys, in which increased levels of MMP-1 and -14 were expressed by reactive astrocytes in the glaucoma eyes, when compared with quiescent astrocytes from normal eyes. Interestingly, specific reactive astrocytes expressing these proteolytic enzymes were migrating into the nerve bundles, suggesting that these two MMPs may have an important role in cell movement. Additionally, the astrocytes in the same study obtained reactive phenotype only in the experimental glaucoma eyes and not in eyes that experimental transection of the optic nerve was performed in order the pathological differences among the two pathological situations to be investigated. Moreover, the writers didn't observe migration of the astrocytes into the nerve bundles at the optic nerve transection eyes (Agapova et al. 2003). Furthermore, ECM remodelling has been shown to occur only in glaucoma and not in the case of optic transection, which can be explained by the lack of activated astrocytes and increased MMP levels (Agapova et al. 2003;Pena et al. 2001). These results, together with previous studies, suggest that the stress caused to the cells by high IOP, induces astrocytes to become reactive and subsequently to produce ECM molecules like elastin and tenascin, as well as MMPs, and to commence the remodeling process of the optic nerve head (Pena et al. 1999b;Varela & Hernandez 1997;Wax et al. 2000). It was suggested by Agapova OA et al that increased IOP may lead to enhanced MMP-14 expression, which upon activation of MMP-2 may facilitate the activation and migration of astrocytes. By cleaving collagen, MMP-1 may open paths for the migration of reactive astrocytes, and if not inhibited by TIMP-1, it may degrade other ECM molecules that are important for the axonal survival. It was also proposed by the authors in the same study that the destructive role of MMPs in the RGCs axons may continue even after lowering of the IOP (Agapova et al. 2003).

At cellular level, it has been reported that elevated IOP may affect the RGCs –brain signalling transportation by compressing the axons and affecting neurotrophin transfer. The mechanisms by which RGC apoptosis occurs is unknown, but it has been suggested that ischemia, high levels of extracellular glutamate, nitric oxide production, downregulation of glutamate transporters may play an important role in this process (Chintala 2006). Although it has been generally believed that ischemia causes accumulation of extracellular L-glutamate, which in turn induces overstimulation of glutamate ionotropic receptors (NMDA, APMA, kainate) and NMDA-mediated RGC death, recent observations have shown that glutamate and NMDA citotoxicity do not reduce the survival of purified RGCs, but that of amacrine cells (Ullian et al. 2004).

From all the members of the MMP family, MMP-9 is considered to act negatively on neuronal survival. Previous studies have shown that MMP-9 deficient mice are resistant to neuronal cell apoptosis after brain injury or ischemia (Asahi et al. 2001;Rivera et al. 2002;Wang et al. 2000). Using an experimental model of ischemia induced by transient ligation of the optic nerve, which presents similarities to the glaucoma pathology, increased levels of MMP-9 associated with activation of astrocytes in the retinal gagglion cell layer as well as increased gelatinolytic activity and degradation of laminin in the ILM were observed (Chintala et al. 2002;Zhang, Cheng, & Chintala 2004b). This is of relevance to the glaucoma pathology, as laminin is important for cell-ECM interactions and subsequently for cell survival. The increased levels of MMP-9 observed in this model may be responsible for the degradation of laminin, which triggers the apoptosis of RGCs, as suggested by observations that MMP inhibitors (GM6001) caused a reduction in RGC apoptosis. These findings lead the authors to suggest that MMP-9 also plays a crucial role in this pathogenic process (Zhang, Cheng, & Chintala 2004a). Participation of MMP-9 in pathological mechanisms involved in the development of glaucoma has been further supported by in vivo and in vitro experimental studies. Increased IOP in an experimental rat model of glaucoma was found to raise the levels of MMP-9 and TIMP-1, whilst reducing immunostaining for laminin (Guo L. et al 2005)(Guo et al. 2005). Although MMP-9 activity in this study was related to RGC apoptosis, the mechanisms by which this process occurred are not clear. It was suggested that either high IOP has a direct effect on MMP-9 increase, which then causes damage to the RGCs, or that increased IOP causes mechanical damage to RGC bodies at the ONH. This could possibly lead to MMP-9 upregulation and consequently to ECM changes (Guo et al. 2005). Upregulation of MMP-9 was suggested to be caused by hyper-stimulation of glutamate receptors on RGCs by kainic acid, a non-NMDA receptor agonist (Zhang, Cheng, & Chintala 2004a). The non-NMDA antagonists NBQX and CNQX block this increase and as glutamate accumulates extracellularly in glaucoma ischemia in vivo, it may trigger MMP-9 expression. In addition to inhibiting MMP-9, TIMP-1 has been shown neuroprotective activity (Tan et al. 2003). This suggests that increased levels of TIMP-1, observed in experimental glaucoma (Guo et al. 2005), apart from inhibiting MMP-9 activity, may also protect retinal neurons. Although TGF-β2 accumulation in ONH correlates with IOP exposure, it was found to be decreased in the retina of experimental induced glaucoma. Although this factor is a potent MMP-inhibitor, its reduction can probably explain the ineffective blocking and subsequently the elevated MMP-9 activity (Cordeiro 2002;Guo et al. 2005;Pena et al. 1999a).
Current postoperative treatment of fibrosis after GFS

Antimetabolites such as mitomycin C (MMC) and 5-fluorouracil (5-FU) are used in the ‘scarring war'. They have been shown to be effective in reducing the scarring after trabeculectomy ( 1996;Skuta et al. 1992). Many studies have been published from our lab that describe the increase of the functioning period of the outflow channel in the bleb. Results indicate that a single five minute application of 5-FU or mitomycin C during surgery reduces the acuity of the healing response to decrease fibrotic formation. It is thought this is mainly due to suppression of fibroblast proliferation, prolonging the bleb survival (Doyle et al. 1993;Khaw et al. 1992;Khaw et al. 1994). Unfortunately, severe complications often occur after treatments with these metabolites. The bleb often leaks and there are lots of other effects including hypotony, endophthalmitis and excessive ocular cell apoptosis that can cause irreversible vision loss. Hence safer and more effective agents are needed to reduce scarring and to control healing after GFS.

Synthetic inhibitors of MMPs

Since MMPs take part in several pathological conditions, it is important to identify selective inhibitors that can be used therapeutically to control MMP activity in defined ways. The use of the natural TIMP inhibitors has significant disadvantages such as their high molecular weight and their poor oral bioavailability, which prevent their clinical use. In addition, it has been shown that although TIMPs may improve the prognosis of some categories of tumors, there are examples where TIMPs are involved in more aggressive cancer progress (Glasspool & Twelves 2001b).

To overcome these difficulties, synthetic compounds to block MMP activities (MMPIs) have been designed. Some of the most well-known MMPIs are batimastat (BB-94), marimastat (BB-2516), Prinomastat (AG3340) and Tanomastat (BAY12-9566) (Glasspool & Twelves 2001c). These are hydroxamic acid derivatives that bind reversibly to the zinc in the active site of MMPs. Most of the potent inhibitors designed to date are right-side binders, as left-side binding is much weaker possibly due to its natural ability to prevent the carboxylate product of substrate cleavage from becoming a potent inhibitor of the enzyme (Skiles, Gonnella, & Jeng 2001).

Batimastat is a pseudopeptide with structural similarity to collagen that inhibits MMP-1,-2,-3,-7, and -9 (Canete, Gui, Linask, & Muschel 1995;Hoekstra, Eskens, & Verweij 2001), was the first synthetic inhibitor of MMPs tested in clinical trials for malignances. Because of its low oral bioavailability caused by poor solubility, intraperitoneal administration was used. Side effects included abdominal pain, fever, asymptomatic elevation of liver enzymes and pain at the site of injection. However, there were no significant clinical responses and this led British Biotech to replace it with another, orally administered

MMPI, marimastat.

Although there were not important clinical differences when compared to other chemotherapy (gencitabine), further analysis revealed that patients with pancreatic and gastric cancer lacking distant metastases and already been treated with chemotherapy, may benefit from marimastat (Glasspool & Twelves 2001a). Additionally, marimastat seems to increase survival and delay disease progression in patients with advanced gastric cancer (Lelongt et al. 1997). Musculoskeletal pain is its most important side effect. Today non-peptido mimetics, like BMS-275291 (is being assessed for lung, prostate cancer and Kaposi's sarcoma) (Lockhart et al. 2003) and non-peptidic chemicals like tetracyclines (Kwan, Schulze, Wang, Leon, Sariahmetoglu, Sung, Sawicka, Sims, Sawicki, & Schulz 2004) are being tested in humans.


Ilomastat (4) is a synthetic derived MMP inhibitor. This molecule is referred to as well as Galardin, named after the person who discovered it (Galardy et al. 1994d). Studies in our lab have demonstrated that ilomastat can inhibit MMPs during subconjunctival wound healing without toxic effect. For these reasons our group is focused on Ilomastat for use for scarring inhibition after GFS.

Ilomastat (molecular formula C20H28N4O4, 388.47 g/mol) is a peptide analogue with the formal chemical name of N-[(2R)-2-(hydroxamidocarbonylmethyl)-4-methylpentanoyl]-L tryptophan methylamide. It is a broad spectrum hydroxamate MMP inhibitor (Galardy et al. 1994b). The reported Ki values are as follows: Human MMP-1 (Fibroblast collagenase): 0.4 nM, Human MMP-3 (Stromelysin): 27 nM, Human MMP-2 (72 kDa gelatinase): 0.5 nM, Human MMP-8 (Neutrophil collagenase): 0.1 nM, Human MMP-9 (92 kDa gelatinase): 0.2 nM (Galardy et al. 1994c).

Ilomastat achieves its inhibitory effect through the competitive binding of the catalytic zinc of MMP by its hydroxamic acid (2). Hydroxamic acid inhibitors have two significant disadvantages; (1) short half-life due to the fast metabolism in the liver and (2) low solubility in aqueous solutions (0.03885 mg/ml) (Parker,M.H. Biochemistry 1999, 38, 13592-13601). Ilomastat is soluble in DMSO up to at least 400 mg/ml. In contrast Ilomastat is hardly soluble in water (R.E. Galardy et al. Ann. N.Y. Acad. Sci. 1994 732 315). Relative concentrations of GM6001 in aqueous solutions may be determined by measuring the absorbance of the solution (tyrosine) at 280 nm.

Since it has been found that MMPs play a significant role in wound contraction (Daniels et al. 2003;Porter et al. 1998), their critical involvement was shown in studies where inhibition of MMPs reduced wound contraction in in vitro experiments using collagen I lattices as the wound contraction model (Scott, Wood, & Karran 1998). Both in vitro and in vivo studies in order to test the effect of MMPIs in contraction models have been performed. Daniels et al 2003 (Daniels, Cambrey, Occleston, Garrett, Tarnuzzer, Schultz, & Khaw 2003) tested the effect of three MMPIs – Ilomastat, BB-94 and BMS-275291 (Cell Tech) in HTF populated collagen gels. Observations revealed inhibition of the contraction of the gels with the application of all the three MMPIs in a dose-dependent manner and Ilomastat was observed to be the most effective.

The tested MMPIs were also found to have a non toxic and reversible effect and zymography results indicated significant reduction of the proteolytic activity of the detected MMP bands after the application of the MMPIs. It was also shown that ilomastat inhibited collagen production from fibroblasts in a dose-dependent manner. This was an important finding, as excessive collagen production and deposition at the incision area is mainly responsible for the bleb failure (Cordeiro et al. 2000;Daniels et al. 1998).

Administration of ilomastat in an in vivo contraction model after trabeculectomy was found to significantly prolong the bleb survival in comparison to the control group as well as to have a lowering IOP effect throughout the experiment (Wong, Mead, & Khaw 2003). Histological findings showed that reduction of scar tissue formation in the ilomastat treatment group occurred with decreased cellularity compared to the control group. There was also decreased cell apoptosis (that is known from other studies to be associated to MMC), decreased myofibroblasts in the wound area (possibly because of an inhibitory effect of ilomastat in fibroblast migration) and a large bleb area compared to control group.

The necessity of comparison of the antiscarring effects of ilomastat with MMC led scientists design a new comparative in vivo study (Wong, Mead, & Khaw 2005). The ilomastat treated group had similar prolonged bleb survival and IOP lowering results as the MMC treated group. Importantly, this study showed that the morphology of the subconjunctival tissue was normal in the ilomastat group but hypocellular in the MMC group. It is worth mentioning that in none of our in vivo experiments ilomastat damaged conjunctiva, as it can happen in the case of MMC.


Corticosteroid therapy consists one of the main risk factors for the development of secondary (steroid-induced) glaucoma. As an initial step, Intraocular Pressure (IOP) is increased and if the duration and the level of this increase in significant, can lead to apoptosis of optic neurons and subsequently to the development of glaucoma. Several studies, like (GORDON et al. 1951), have indicated the role of corticosteroids in the enhancement of IOP. Moreover, it is known that cortisol levels are responsible for the diurnal fluctuation of IOP as well as the lack of this normal condition in patients with removed adrenals (Smith 1966) and the raised IOP in adrenal adenomas (excessive cortisol production) (Haas & Nootens 1974). 5% of normal eyes, 92% of POAG patients and 30% of their relatives develop after about 6-8 weeks increased IOP after application of steroid drops which declines several weeks after stopping the steroid treatment (Phillips et al. 1984).

Mifepristone (or RU-486) is a steroid with similar structure to the natural hormone progesterone. Being a competitive inhibitor of the progesterone receptor is categorized as progesterone antagonist. Due to its structural similarity with progesterone and its different function, it is effective for early abortions up to nine months. The maintenance of pregnancy and the nutrition of the embryo make the secretion of progesterone important from the female body. The use of RU-486 sends a stop signal to the production of progesterone, due to total occupation of the progesterone receptors in the uterus and diastoli of the cervix resulting in abortion.

The effect of a similar to Mifepristone compound, the RU486-6, in IOP has been tested in rabbit eyes by (Phillips, Green, Gore, Cullen, & Campbell 1984). The observed during the 13 weeks of 3times/day drops applied a lowering effect of this RU486 analog in IOP, which as it is suggested could have happened either because of steroid blockage leading to increase of the aqueous outflow through trabecular meshwork, or decrease of the aqueous production from the ciliary body, or because of reasons irrelevant to steroid blockage action such as corneo-scleral softening or thinning. Our lab, in a joint project with DANIOLABS, aims to develop further this exciting findings; we decided to test the effectiveness in IOP reduction of three newly created compounds with structure close to the RU-486, the RU-42633, RU-42848 and RU-42698.


The designing of gene targeting techniques have enhanced the understanding over the gene function all over the world (Aigner 2006). It is strongly believed that RNA interference (RNAi), discovered 11 years ago (Fire et al. 1998), through small interfering RNAs (SiRNA) – together with transposons- to have played a vital role in the structuring of the genome of most organisms (Sharp 2001). This high in reliability, specificity as well as efficiency natural mechanism, than other methods used in the past (Fire, Xu, Montgomery, Kostas, Driver, & Mello 1998;Tebes & Kruk 2005), is believed to have been created even before the divergence of animals and plants and to consist the oldest antiviral system. Focusing on the effects of RNAi in the nematode C. Elegans, Fire (1998), applied long double-stranded RNAs (dsRNA) containing the identical sequence to the target gene and detected post-transcriptional silencing of specific genes. This RNAi mechanism (long dsRNAs) was also successful in other organisms including Drosophila. As regards to vertebrates, initial efforts where unsuccessful because dsRNAs trigger the interferon response (Hammond et al. 2000;Tebes & Kruk 2005).

siRNAs are small 19-23 nucleotide duplexes (Bernstein et al. 2001) with 2 nucleotide overhangs at both ends, containing 5΄-phosphates and 3΄-hydroxyls, that cleave mRNA molecules. They are generated in response to introduction of long dsRNAs into the cytoplasm. Modern methods transfect directly siRNAs to the cells.


RNA silencing consists a mechanism for eukariotic cells in order to defend against transposons and pathogens as well as control their own genes expesion. Attempting to detect and isolate short-RNAs with ‘interferance' function from several animal species, short-RNAs that were formed by the host genome and their role was not to produce proteins were founded. The RNA molecules are known as pre-miRNAs, ~70 nucleotide short hairpin molecules, which because of to their self-complementary structure fold on themselves. The enzyme Dicer, which is vital for the function of the siRNA mechanism is also dominant in the case of the miRNA mechanism, as it cleaves the pre-miRNAs into 21-22 nucleotide molecules (miRNAs) (Bernstein, Caudy, Hammond, & Hannon 2001). It has been found that in humans, both in vitro and in vivo, miRNAs induce cleavage of perfectly complementary target RNAs resulting in gene silencing.

After discussing some general similarities between the siRNA and the miRNA mechanisms, in the following pages the intracellular mechanism of siRNA interference is described.


a) In vivo direct application of in vitro synthesised 19-23 bp double stranded small interfering RNAs (siRNAs) containing 2-nt 3΄ overhangs (Aigner 2006).

b)Long dsRNAs, that may cause potent interferon response, which results in total post-transcriptional inhibition of gene expression (Filipowicz et al. 2005;Meister & Tuschl 2004). Long dsRNA, developed by viral replication or viral gene expression, may be detected through the serine-threonine kinase PKR(Katze et al. 1991;Meurs et al. 1992;Williams 2001) or the Toll-like receptor 3 (TLR3) (Alexopoulou et al. 2001). Expression of I interferon (IFN-alpha and IFN-beta) is the result of this process, especially in plasmocytoid dendritic cells (PDC). A multi-step and multi-factor signalling pathway is induced (Leaman et al. 1998) resulting in the upregulation of many interferon-induced genes and this process may cause inhibition of protein synthesis and apoptosis. Further research focusing on RNAi facilitated overcoming this barrier. Specifically, the Dicer enzyme, an endogenous cytoplasmic RNase III-like enzyme, was found to reduce long dsRNAs into small 19-21 nucleotide duplexes (siRNAs) with two nucleotide overhangs in the 3΄ (Aigner 2006). It was shown that dsRNAs greater than 30 bp trigger the interferon response, but not the small siRNAs (Manche et al. 1992). Consequently, siRNAs could be used as a very useful RNAi tool in order to perform gene analysis in mammalian cells.

c) DNA encoding short hairpin RNA (shRNA) expression cassettes that can be delivered to cells, through which intracellular expression of shRNAs is achieved. These shRNAs are then cut and active siRNAs are developed by the host cell. (Plasmids containing polymerase III promoters have been created by several researchers, known to synthesize small RNA, which facilitate synthesis of 50 bp-long single-strand RNA folded in 21-23 bp dsRNAs with a small hairpin at the middle (shRNAs)(Coumoul & Deng 2006). Other vectors allow synthesis of two complementary short RNA duplexes forming then a siRNA (Uprichard 2005).


There are three major strategies for delivery of siRNA into mammalian cells.

1.RNAi effectors

RNAi effectors can be delivered to cells using two different approaches.

a) siRNAs are synthesized by scientists in order to be delivered as a “drug”(Uprichard 2005)

b) through a gene therapy approach, in which viruses are used to deliver DNA encoding the target siRNA into the cell nucleus. Viral mediated delivery offers efficient delivery of DNA with persistent siRNA production by the host cells (Brummelkamp, Bernards, & Agami 2002). Retroviruses, adenoviruses, lentiviruses, herpes viruses, simbis viruses and baciloviruses have all been used in gene therapy studies (Lotze & Kost 2002), but only the first three have been shown to be efficient and to have a long-term effect. In the case of lentivirus mediated gene silencing, Abbas et al. suggested that this mechanism is effective within 72h post-infection and its function was maintained for at least 25 days. Unfortunately, although viral vectors have significant advantages for RNAi (excellent tissue-specific tropism and transduction efficiency), they are linked with risks and safety issues (Uprichard 2005). Despite the significant problems encountered in the past as regards to gene therapy clinical trials, the RNAi benefits outweigh the risk and for that reason adeno- and lenti-virus associated vectors are being evaluated for clinical delivery of shRNAs(Uprichard 2005).

2.Local delivery

Following successful local administration of antisense drugs to the eye, initial clinical trials for RNAi-based treatment of age-related macular degeneration are administering siRNAs with injections directly into the vitreous humor (Check 2005). Intranasal administration for pulmonary delivery and direct delivery into the central nervous system have been considered as additional promissing routes (Uprichard 2005).

3.Systemic delivery

SiRNA stabilization, targeting of the effector to the correct tissue, and inducement of cellular uptake are required for the optimization of systemic delivery (Uprichard 2005).


Dicer and Drosha (Rnase III family enzymes) are important for the formation of siRNAs from dsRNAs. This was initially shown in a Drosophila embryo extract and these RNA molecules were found to contain a 5′ phosphate and a 3′ hydroxyl terminus (Hammond 2005). It is known that these RNA properties are found when RNaseIII enzymes are involved in the cleavage process.

The Droshas, which are 130-160 kDa nuclear proteins have N-terminal proline-rich regions and the human (Hs) enzyme has a domain rich in Arg-Ser (RS) dipeptides, similar to the protein–protein interaction domains found in many splicing factors. Drosha and Dicer work as a complex with proteins that contain dsRBDs and are known as Pasha (in Drosophila) and R2D2 or DGCR8 (in mammals) (Landthaler, Yalcin, & Tuschl 2004) and they are both important for the processing of pre-miRNAs. This complex is as large as of about 650 kDa due to, as it is believed, to the dimerization of the components (Filipowitz et al, 2005). Although WW, which contains 2 conserved tryptophans, is usually expected to interact with proline-rich sequences, the WW module of Pasha is not necessary for the connection with Drosha. The pre-miRNA is released by Drosha and it is transferred from the nucleus to the cytoplasm in an Exportin-5/RAN-GTPase-dependent manner. In the cytoplasm, the pre-miRNA is formed into double stranded miRNA structure by Dicer. The 2 nucleotides 3′ overhang terminus on the pre-miRNA, formed by the Drosha, is captured by the PAZ domain of Dicer, analogous to the recognition of dsRNA termini. Drosha and Dicers present substrate specificity (Drosha is involved in the pre-miRNA processing but not in the processing from a dsRNA terminus). It is believed that the stem–loop structure and especially its size is important for the substrate recognition of Drosha (Zeng, Yi, & Cullen 2005). Unstructured sequences located close to the stem–loop has been suggested to be important for processing, and possibly other proteins-enzymes and not Drosha may be involved in the recognition of these regions. Possibly other factors participate as well, as it seems unlikely Drosha to recognize these sequences as they are located outside its zone of activity (Hammond 2005).


Dicers are not, vert, similar200 kDa proteins. They are shown to have a double role by miRNAs and siRNAs from pre-miRNAs and dsRNAs respectively. Differences in Dicers have been detected between species. Although vertebrates and C. elegans have in their genome single Dicer genes, in Drosophila and in other species two or more Dicers are detected that present specialized functions (Hammond 2005).

Dicers have an N-terminal ATPase/RNA helicase domain, followed by a DUF283 domain with unidentified role and a PAZ domain, also found in Argonaute family proteins. The structure of the two Drosophila (Dm) Dicers is different: In Dcr-1, the helicase domain is smaller. Regarding the PAZ domain, it is the same in humans and in Dcr-1, but the PAZ of Dcr-2 is structurally different.

Dicers are divided into 3 categories (Fig.3). Category I is mainly found in bacteria and yeast. In this category only one RNaseIII domain is detected which is bound to a dsRBD. In contradiction, two Rnase III domains are present in enzymes belonging to categories II and III. Enzymes belonging in category III share a similarity with the Argonaute-family proteins, as they both have in their structure a PAZ domain (Piwi/Argonaute/Zwille) (Tabara et al. 1999).

The two Dicers expressed in Drosophilla have different roles. The role of Dcr-1 seems to be important for the generation of double stranded miRNAs from pre-miRNAs. Dcr-2 on the other hand seems to be important in two steps in order RNA silencing to be achieved; It is involved in the cutting of the dsRNA and the creation of the double stranded siRNA as well as in the structure of the RISC complex. For the first step, Dcr-2 interacts with R2D2, which has two dsRBDs domains (Grishok et al. 2001). In the case of vertebrates as well as in C. Elegans, where, instead of two, only one Dicer protein has been detected, other factors connected with the Dicer may change the activity of the ezyme. Both Dicers (Dcr-1 and Dcr-2) have been found to interact with Argonaute proteins (Fig....) (Bartel 2004). In humans, this interaction is believed to be achieved through regions located at the RNase III and Piwi domains (Hutvagner et al. 2001).

Although it was initially suggested that Dicer works as a dimer performing 4 cleavage reactions, later studies indicate that it is more likely that Dimer may work as a monomer. This is believed because Dicers seem to start their fuction by capturing an already existed terminus and subsequently 2 cleavage reactions are necessary in that case, which occur ~21 nucleotides from the already existed terminus. The existence of terminal position from which the Dicer can start its function is believed to speed up the cleavage process. In contradiction delays happen when Dicers initiate their function by an initial internal binding, as it has been suggested that the binding is less efficient than when an end terminus exists (Hammond 2005).

The most representative model of the structure and the function of the Dicer is shown in Fig.4. An active enzymatic pocket is created by the two Dicer RNaseIII domains, which come one opposite the other and the molecule functions as a pseudo-dimer. Each domain is in charge of cleaving one of the two strands of the double stranded RNA. Is is believed that vital role in the calculation of the 21 nucleotide length new product plays the distance between the PAZ domain the the active pocket generated by the RNase III domains. The pseudodimer alignment seems also to be the reason for the creation of the 2 nucleotide overhang. The existence and fuction of these enzymes is vital; this was shown in experiments in which early embryonic lethality occurred in mice without Dicers. The way and the site that the dsRBD domain binds to the RNA is not yet known (Hammond 2005).


RISC has been suggested to be the last part of the chain in the RNAi process. Argonautes participate in protein complexes like RISCs and miRNPs and they are also known to be linked with Dicer. Argonaute2 (Ago2), the first identified member of the family, has in its structure the PAZ and the PIWI domains. The C-terminal Piwi domain found in the Argonautes presents structural similarities with the RNase H fold. The existence of a domain rich in glutamine is the reason why Drosophila Ago2 is larger than Ago1 and human Ago1–4 (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005).

RISC seems to finction differently in the case of mRNA silencing, as many scientists have suggested and it is working by inducing suppression of translation rather than cleavage of the mRNA.

Scientists have also found additional protein components in the RISC (dFXR, RNA binding protein VIG, the Drosophila homolog of the Fragile X protein, helicase proteins, but their role has not been yet described. Further studies have shown that an oligonucleotide-binding (OB)-like fold exists in the not, vert, similar 130 KDa PAZ domains of Drosophila Ago1 and Ago2 (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005). The model that has been suggested for the siRNA-PAZ interaction is characterised by significant asymetry. The 3' end of the strand domain interacts with PAZ along its full 9-nt length. On the other side the antisense the opposite strand interacts only with the 5'-terminal residue.In humans, four members of the Argonaute family have been detected, the Ago1-4. Although Ago1-4 interact with siRNAs and microRNAs, only complexes that include endogenous or recombinant hAgo2 are capable of cleaving mRNA. Indicative of this is that RNA interference is not achieved in Ago2 negative (knock out) mice (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005).

The Piwi domains has been indicated to have an RNase H role and for that reason it is also known as ‘Slicer' as it is believed that it is in charge of cleaving mRNA in RISCs. RNase H and Ago share a common triplet of aminoacids (DDE), important for the catalysis. Equivalent amino acids appear to be conserved in most eukaryotic Argonaute proteins. The importance of the 2 aspartates of this aminoacid triplet for RNAi has been shown in studies based on mutagenesis of human Argonaute 2 protein (which as it has been previously mentioned is important for RNA inerference) (mRNA cleavage through RISC) (Khvorova, Reynolds, & Jayasena 2003). Apart from this, other studies that have found similarities in the functioning of the RISC and RNA-H (both enzymes generate 5′-phosphate and 3′-OH termini) (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005) are indicative of the function of a role for Argonaute 2 as ‘Slicer'.

The Slicer catalytic model is shown in Fig4. The SiRNA 5′ end of the siRNA guide is linked to the Piwi domain. In this link the 5′ phosphate plays a very important role and it is coordinated by four conserved residues. As it is shown in the graph, the 3′ SiRNA is linked with the PAZ domain and captures the 3′ OH terminal ends of RNA, or duplexes with a 3′ overhang (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005;Hammond 2005). In microRNA/target pairs, the vital role of nucleotides 2–8 in the microRNA for the detection of the target has been indicated. Regarding the first nucleotide, it has been suggested that it is not important for target recognition as well as that when a strong base pair is located at the 5′ SiRNA terminus of the siRNA, the activity of the PIWI domain as a Slicer is declined. As it has been previously mentioned, the cleavage takes place at the PIWI domain as it is shown in figure.... It has also been indicated that the ‘cutting' is measured from the 5΄ SiRNA end, the product then is released without any knowledge on the mechanism that this is achieved and the enzyme at teh end of the process recycles.


Studies designed in order to test how effective are the different forms of double- and single-stranded siRNAs as well as the processing of pre-miRNAs by Dicers and the sequencing of the created miRNAs by RISC pointed out that the strand from the dublex (created after the Dicer processing) with the less thermodynamically stable 5′ N terminus is picked as the miRNA . In Drosophila, it has been proposed that the Dcr-2–R2D2 heterodimer senses the differential stability of the duplex ends and determines which siRNA strand gets selected. Facilitated by the method of 5-iodouracil photocross-linking applied to siRNAs at different positions showed that the less stable end is linked to Dicer an the more stable to R2D2 the siRNA end (Filipowicz, Jaskiewicz, Kolb, & Pillai 2005).


Although RNAi has been suggested to be a gene inhibition tool with high target specificity, off-target effects have been observed post siRNA introduction indicating non-specific inhibition. 8-11 base pairs of homology of the siRNA with a single-stranded RNA were shown to lead to silencing. It is believed that siRNA present the ability to inhibit mRNAs that are characterised by limited sequence similarity with siRNAs. In fact, in some cases regions comprising of only 11–15 contiguous nucleotides of sequence identity were sufficient to induce gene-silencing (Jackson et al. 2003). No significant research has been presented so far regarding the off-target siRNA effects and the predictions of such effects is highly unpredictableThe prediction of these off-target activities is difficult so far, but nowadays scientists, having in their hands the ‘weapon' of searchable genetic information, can in many cases avoid inadvertent off-target silencing.


Three main guidelines which are the optimal length (19–25bp), the existence of a 3′ dinucleotide overhang and a low G/C content ranging between 36% and 52% have been proposed for effective RNAi (Elbashir et al. 2001;Elbashir et al. 2002;Elbashir, Lendeckel, & Tuschl 2001;Holen et al. 2002). More recent studies suggested that the thermodynamic flexibility of the duplex 3′-end (i.e. positions 15–19, sense strand), but not of the 5′-end, correlates with silencing efficacy and that the presence of at least one A/U bp in this region, which decreases the internal stability of the 3′-end, increased silencing efficacy (Reynolds et al. 2004). Furthermore, internal repeats or palindrome sequences are not suggested as facilitate the formation of internal fold-back structures and decrease the possibility of silencing.

Later extensive studies have suggested more specific siRNA structure guidelines for effective interference; the Glutamine and Cytocine percentage in the siRNA structure should be 30–52%; three at least Adenosine or Uracile nucleotides should exist in positions 15–19 of the sense strand; specific nucleotides should be located in specific positions (A at position 3, U at position 10, no G at position 13, no G at position 19, no C at position 19; A is preferred at position 19). By following these guidelines, greater siRNA efficacy is achieved (van Es & Arts 2005).

Serum Amyloid P

The wound healing process that occurs in the eye after glaucoma filtration surgery starts after the initial conjunctival incision. Although many scientific groups have described wound healing extensively, many different theories have been supported as regards to the sequence of the phenomena that occur in wound healing as well as the origin of the cells that participate in wound healing. The healing process in summary, as it has been previously described, seems to start with the release of plasma proteins and blood cells in the wound area, which leads to the formation of a fibrin clot. Neutrophils and macrophages are recruited at the wound area and degrade the clot by expressing several enzymes and MMPs, such as MMP-8 and -9. Activation and migration of fibroblasts to the wound site also takes place. The fibroblasts in normal unwounded tissues are quiescent undifferentiated mesenchymal cells known as fibrocytes. They exist in low numbers in the subconjunctival connective tissue-Tenon's capsule- (1). After their activation, these fibroblasts produce large amounts of ECM molecules such as collagens, glucosaminoglucans and elastin. They also produce MMPs that facilitate cleavage of the ECM.

A decade ago it was postulated for the first time that in the development of fibrotic disorders and in wound healing processes, non activated fibroblasts (fibrocytes) circulating in the blood participate in and lead to scarring (2). Supporting this theory, later studies provided evidence that myofibroblasts do not originate from tissue fibroblasts, but from a bone-marrow-derived precursor (3;4). It has been found that these cells have CD14+ peripheral monocyte-precursor origin and, apart from stromal cell markers (pro-Collagen I, fibronectin, pro-Collagen III), these fibrocytes express hemopoietic markers as CD45, CD13, MHC class II and CD34 (2;5). These cells obtain the characteristic fibroblastic (spindle-shaped) morphology when cultured in vitro (2). Additionally Bucala et al 1994 found that these cells express vimentin (apart from collagen and CD34), indicative of their fibroblastic properties. In tissue damage, it was hypothesised that these cells enter the wound area and, by expressing cytokines and chemokines, cleave the existing ECM and promote angiogenesis, to produce new ECM and to promote contraction (6). Apart from the in vitro studies, these cells were found to be involved in the promotion of fibrosis in an in vivo pulmonary fibrosis model (7). It is believed that they may participate in several fibrotic disorders, such as liver cirrhosis (8;9), and in autoimmune diseases (10). Subsequently, there has been evidence that these circulating fibrocytes are involved in the scarring development (6).

Pentraxins are a protein superfamily that took their name from the presence of a 200 amino acid pentraxin domain in their carboxy-terminal domain. The first described and most well known member of the family is the C reactive protein CRP. The serum amyloid P (SAP), which consists of the protein that we propose to study, is a highly concerved plasma glycoprotein that has been found to share a 51% sequence homology with CRP (11). A difference between the two molecules is that the levels of CRP post infection increase significantly in most mammals and CPR is known as the acute phase protein. On the other hand, small only changes are observed in the levels of SAP as a result of infections (Pepys et al, 1978 apo Pilling 2003). CRP is specialized in binding small nuclear ribonucleoprotein particles; on the other hand SAP is specialized in binding chromatin and free DNA (Du Clos, 1989; Bickerstaff et al 1999). The two molecules, CRP and SAP, present differences to their binding ability to FcgRs; in the case of CRP, a high affinity binding is achieved with FcgRII, lower with FcgRI and no binding with FcgRIII. SAP binds all the three FcgRs, but with higher affinity the FcgRI and the FcgRII. The aforementioned differences, together with different links of the two molecules with apoptotic material, as Pilling et al 2003 suggested, are indicative of the different roles of CRP and SAP in vivo. Especially the differences in the binding of FcgRs have been suggested to be the reason why SAP and not CRP inhibits the differentiation of circulating monocytes to fibrocytes, without the entire mechanism to be yet know.

It has also been suggested that by the presence of SAP in the tissues, which is connected with several ECM proteins, it functions like an inhibitory-regulatory mechanism that blocks the transformation of the circulating monocytes to fibrocytes (43-48 tou Pilling 2003). 7 and 8 of Pilling et al 2003 pointed out that the rapid recruitment of fibrocytes in the wound is indicative of the lack or inactivity of SAP in the wounded area, which may allow the transformation of circulating monocytes to fibrocytes. The lack of SAP or its inactivity in the wounded area can be attributed to the fact that the remodeling and regeneration of the area changes the composition of the ECM proteins in the area, with which, as it was previously mentioned, the SAP is linked. Subsequently, SAP levels in the wounded area are reduced due to the remodeling and cannot inhibit the transformation of monocytes to fibroblasts. In other words, reduction of the SAP levels in a specific tissue may serve as an injury signal in order circulating monocytes, transformed into fibrocytes, are ‘sent' to the specific tissue in order to fix the wound.

Lack of SAP in SAP -/- transgenic mice has been linked to antinuclear autoimmunity and glumeronephritis, which are pathological findings in the systemic lupus erythromatosus (SLE) (12). SAP levels in the human serum are 32±7ug/ml in men and 24±8ug/ml in women (13). SAP levels are reduced in patients with autoimmune diseases such as scleroderma (6), rheumatoid arthritis (14) and mixed connective tissue disease. This finding, together with the observation that circulating fibrocytes participate in the pathological mechanisms of autoimmune diseases, indicates that SAP could be involved in the inhibition of fibrosis and, more specifically, in the inhibition of the differentiation of peripheral mononuclear precursor populations to fibrocytes (6). These research outcomes lead to the conclusion that SAP administration may inhibit scarring in many eye diseases as well as after glaucoma filtration surgery, extending the survival of the bleb created after surgery and subsequently inhibiting the increase of intraocular pressure and blindness. Supportive to this hypothesis is also the study by Naik-Mathuria et al (2008); they created surgically dermal wounds in mice and they treated mice with intradermal or intraperitoneal SAP. Mice treated with intraperitoneal or intradermal SAP were found to have slower wound closure and less wound contraction than controls. Additionally, slower reepithelialization rate was shown in the mice treated with intra-peritoneal SAP. Furthermore this group was found to have significantly lower number of a-SMA expressing fibroblasts compared to controls. Based on this finding, the authors suggested that SAP, by inhibiting the differentiation of circulating monocytes to fibrocytes, may be the reason for the delayed normal wound healing in mice. As the conjunctival wound healing presents significant similarities with the dermal wound healing as well as because part of the fibrocytes population participating in the conjunctival wound healing may originate from the circulating monocytes, the role of the SAP as a scarring inhibitor is investigated.