2.1. Mammalian TIMPs

Some of their general properties are summarized in Table 1. The genes for TIMPs-1, -3 and -4, but not TIMP-2, are each nested within an intron in the genes for synapsins; TIMP-1 is associated with synapsin 1, TIMP-3 with synapsin 3 and TIMP-4 with synapsin 2. Synapsins are peripheral neuronal proteins that coat the cytoplasmic surfaces of synaptic vesicles. This inter-relationship appears to be ancient because the gene for the single TIMP in Drosophila melanogaster is nested within an intron of the only Drosophila synapsin (Syn) gene [14]. In the genome of Fugu rubripes, the tiger blowfish, a pair of nested Syn–Timp genes are present together with a Syn gene that does not contain a Timp, but there are two Timp genes that are separate from Syn genes [15]. The nested gene arrangement suggests the possibility of coordinated transcription or alternative splicing between TIMP and synapsin genes, but there appears to be no evidence that this occurs. The gene for TIMP-2 is not nested within a Syn gene, but hosts the gene DDC8 (differential display clone 8), a gene that is expressed in the testis at an enhanced level during spermatogenesis. Recent studies indicate that this gene is also expressed in other tissues in a manner similar to TIMP-2, e.g., in response to an injury to the cerebral cortex. A regulatory relationship between the two genes is suggested by an alternative spliced mRNA variant involving the two genes [16].

All mammalian TIMPs have two distinct domains, an N-terminal domain of about 125 amino acid residues and a C-terminal domain with about 65 residues; the conformation of each domain is stabilized by three disulfide bonds [17]. The four human TIMPs are about 40% identical in sequence to each other; the most similar pair is TIMP-2 and TIMP-4, which are 50% identical in sequence, whereas TIMP-1 is only 37 to 41% identical to the other TIMPs. Although the N- and C-terminal regions of TIMPs have been described as “subdomains” [18], the more definitive term “domain” is appropriate because of their separate evolutionary origin and ability of the N-terminal region to fold and function independently. Recombinant forms of the N-terminal domains of TIMPs, designated N-TIMPs, expressed in heterologous systems, have stable native structures and are fully active as inhibitors of matrix metalloproteinases (MMPs) and some disintegrin-metalloproteinases (ADAMs and ADAMTSs) [19–22]. N-TIMPs have been widely used in characterizing the biochemical and biophysical properties of TIMPs and for investigating structure–function relationships.

2.2. Functional specialization for metalloproteinase inhibition

The four human TIMPs are, in general terms, broad-spectrum inhibitors of the 23 MMPs found in humans, but there are some differences in specificity among them. TIMP-1 is more restricted in its inhibitory range than the other three TIMPs, having a relatively low affinity for the membrane-type MMPs, MMP-14, MMP-16, and MMP-24 as well as for MMP-19 [10]. Also, there are some relatively subtle differences between the affinities of different TIMPs for other MMPs. For example, TIMPs-2 and -3 are weaker inhibitors than TIMP-1 for MMP-3 and MMP-7, contrasting with their affinities for other MMPs [23]. TIMP-3 is unique among the mammalian TIMPs in inhibiting a broader array of metalloproteinases including several members of the ADAM and ADAMTS families [21,22,24–27]. Other TIMPs have limited inhibitory activities for ADAMs: TIMP-1 and TIMP-2 inhibit ADAM10 [24] and ADAM12 [27], respectively. TIMP-3 and N-TIMP-4, but not full-length TIMP-4, inhibit ADAM17 [28]. TIMP-4 was also reported to inhibit ADAM28 [29]. ADAM metalloproteinases differ from the MMPs in domain structures and are highly divergent in catalytic domain sequences: ADAMs are membrane-bound enzymes containing disintegrin, cysteine-rich, EGF-like and transmembrane domains C-terminal to their catalytic domains [3]; and ADAMTS (disintegrin-metalloproteinases with thrombospondin motifs) are secreted proteins with a disintegrin domain and variable numbers of thrombospondin type 1 motifs and other domains in their C-terminal regions [2]. In humans there are 21 members of the ADAM family, of which only 13 are functional proteases, and there are 19 ADAMTSs.

The first non-MMP metalloproteinase reported to be inhibited by TIMP-3 was ADAM17, also known as tumor necrosis factor-α converting enzyme or TACE [21]. ADAM17 catalyzes the release of soluble tumor necrosis factor α (TNF-α) from its membrane-bound precursor as well as the shedding of transforming growth factor (TGF)-α, various receptors, L-selectin and Notch and the non-pathogenic processing of amyloid precursor protein as an α-secretase [30]. It has now been established that TIMP-3 is an effective inhibitor of ADAM10, 12, 17, 28 and 33, and ADAMTS-1, -2, -4 and -5 [10]. ADAMTS-4 and -5 are also called aggrecanases 1 and 2, respectively. The inhibition of ADAM17 and aggrecanases by TIMP-3 appears to be important in regulating inflammatory processes, affecting the progression of disease processes such as cancer, rheumatoid arthritis and osteoarthritis [31,32]. The interaction between TIMP-3 and ADAMTS-4 is enhanced about 4-fold by aggrecan [33]. N-TIMP-3 is a weak inhibitor of ADAMTS-2, procollagen N-proteinase, with a K_i of about 600 nM, but the inhibition is also enhanced by heparin to a K_i of 160 nM [26]. As a potentiator of TIMP-3 inhibitory action, the anti-arthritic agent, calcium pentosan polysulfate, is remarkable; it enhances the binding affinity between TIMP-3 and ADAMTS-4 or -5 more than 100-fold and also blocks the endocytosis of TIMP-3 by chondrocytes [34]. These results suggest that the anti-arthritic activity of pentosan polysulfate is dependent on the endogenous TIMP-3 of cartilage.

MMPs and ADAMs are usually inhibited by the N-terminal domains of TIMPs, but Rapti et al. [35] found that while TIMP-1 and TIMP-3 both inhibit ADAM10, their isolated N-terminal domains do not. This indicates that the mechanism of inhibition of this enzyme by TIMPs differs from those of MMPs and other ADAMs and ADAMTSs. Full-length TIMP-3 and N-TIMP-3 are therefore useful agents for discriminating between the activities of ADAM10 and ADAM17 which often share similar substrates.

2.3. Three-dimensional structures of TIMPs and their complexes

The structures of full-length TIMP-1 and TIMP-2, N-TIMP-1 and N-TIMP-3 have been determined by X-ray crystallography [18,36–41] and the solution structures of N-TIMP-1 and N-TIMP-2 have been elucidated by NMR [42–44]. Most of the structures are of the bound forms of TIMPs in their inhibitory complexes with their target metalloproteinases, exceptions being the crystallographic structure of free TIMP-2 [37], solution structures of free forms of N-TIMP-1 [45] and N-TIMP-2 [42], and the structure of TIMP-2 in its complex with pro-MMP-2 which is stabilized by interactions between the C-terminal domain of the inhibitor and the hemopexin domain of the proenzyme [38] (Table 2).

The structures of full-length TIMP and N-TIMP have a “wedge-shaped” appearance [18,42]. The N-domain has an oligonucleotide and oligosaccharide binding (OB) fold [46], a five-stranded (β strands sA through sF) closed twisted β-barrel with a Greek key topology. This domain also contains three α-helices, one located close to the N-terminus (hI) and two (hII and hIII) near the C-terminus where they form part of the interface with the smaller C-domain. The C-domain has a pair of parallel β strands (sG and sH) connected by a loop, followed by a helix (hIV) and a pair of antiparallel β-strands (sI and sJ) connected by a β-hairpin (Fig. 1A).

Open in a separate window

Fig. 1

Structure of TIMP-1 and its metalloproteinase interaction regions. (A) A ribbon structure of the three-dimensional structure of TIMP-1 showing the locations of the disulfide bonds, the two domains, domain interface and regions (IR I to V) that interact with metzincins and their component residues. This image was generated using Chimera [177]. (B) The structure of the core of the TIMP interaction site indicating interactions with individual subsites in MMP active sites.

The structure of the immediate N-terminal region is highly conserved in all TIMPs, reflecting its crucial function in inhibiting MMPs. The cysteines in the Cys¹-X-Cys³ sequence are attached by disulfide bonds to two other conserved cysteines, Cys⁷⁰ and Cys¹⁰⁰, respectively (numbering based on the sequence of TIMP-1) (Fig. 1B). The third Cys, residue 13, is disulfide bonded with Cys¹²³, a residue that delineates the C-terminus of the inhibitory domain. All structurally characterized inhibitory TIMP-metalloproteinase complexes, including the complex of N-TIMP-3 with ADAM17, are closely similar with respect to the core of the protein–protein interaction interface. The majority (60–75%) of the interactions between a TIMP and a metalloproteinase are made by the continuous ridge formed by the N-terminal five residues, Cys¹-Thr-Cys-Val-Pro⁵ in TIMP-1 (designated here IR I (interaction region I)) and the loop connecting sC and sD (CD loop), residues Met⁶⁶-Glu-Ser-Val-Cys⁷⁰ in TIMP-1 (IR II), that are covalently fastened by the disulfide bond between Cys¹ and Cys⁷⁰ (see Fig. 1). This ridge inserts into the metalloproteinase active site in such a way that the conserved N-terminal Cys¹ of the TIMP lies above the catalytic Zn²⁺ of the MMP. A key to the mechanism of inhibition is the bidentate coordination of the metal ion by the N-terminal α-amino group and carbonyl group of Cys¹, which displaces from the enzyme the water molecule needed for peptide bond hydrolysis. Residue 2, which is Thr or Ser in all TIMPs from vertebrates, interacts with the so-called S1′ specificity pocket of the MMP, a subsite that has a dominant role in determining enzyme specificity. Residues 3–5 interact with primed subsites (that interact with residues of the peptide substrate located to the C-terminal side of the scissile bond) in the protease active site, while Ser⁶⁸ and Val⁶⁹ interact with non-primed subsites S2 and S3 (that interact with residues of the peptide substrate located to the N-terminal side of the scissile bond) of the MMP (Fig. 1B).

Other regions that contribute to the interaction site are the AB loop (IR III) that lies to the right of the CD loop, and the EF loop (IR IV) (Fig. 1A). There is a large variation in the size of the AB loop among different TIMPs and this influences their interactions with MMPs. In the complex of N-TIMP-3 with the catalytic domain of ADAM17, there are also significant interactions involving the C-terminal end of sD (IR V) that are not observed in other structurally characterized TIMP complexes [41]. It is not clear whether these occur in complexes of TIMP-3 with MMPs or are unique to the interaction with ADAM17 and possibly other ADAMs.

In the complexes formed between full-length TIMPs and the catalytic domains of MMPs [18,36,40], there are some contacts between the C-domain of the TIMP and the protease, but these are on the periphery of the interaction site and do not appear to make important contributions to complex formation because the truncated N-TIMPs are high affinity MMP inhibitors. However, studies of the binding kinetics of TIMP-2 with full-length MMP-2 suggest a much higher affinity (6 fM) [47] compared with MMP-2 that lacks the hemopexin domain (275 nM) [48]. This affinity of N-TIMP-2 for full-length MMP-2 was less, although remaining high (40 pM), suggesting that the C-terminal domain of TIMP-2 makes some stabilizing interactions with the C-terminal hemopexin domain of MMP-2. Complexes of TIMP-1 or TIMP-3 with ADAM10 are also expected to have substantial contacts through the C-domain as N-TIMPs fail to inhibit ADAM10 [35]. In contrast, the C-terminal non-catalytic domains in ADAM17 have unfavorable effects on the interaction with N-TIMP-3 [49]. The interaction of TIMP-3 with ADAM17 also differs from those with MMPs in showing cooperative binding, which suggests the presence of multiple TIMP-3 binding sites, and also in the mode of interaction since an N-terminal Ala extension in N-TIMP-3 has a relatively small effect on the inhibition [50].

While close contacts between IR I and IR II and the metalloproteinase are found in all known TIMP complexes, the other regions have variable roles depending on the TIMP and on the target enzyme. Further detailed structural studies are required since such information will be useful for engineering TIMPs with restricted inhibitory spectra.

3.2. Invertebrates

TIMPs from invertebrates are more variable in sequence than those from vertebrates. Apart from the homologues from nematodes, all are two-domain proteins that, with a few exceptions, have arrangements of cysteines that are similar to those of their mammalian relatives (see Fig. 3). However, a large proportion of invertebrate TIMPs have eight cysteines rather than six in their C-terminal domains. Some species, for example, D. melanogaster, have a single TIMP, whereas others have multiple TIMPs (e.g. hydra). Regions of the alignment show striking variations in length. For example, some species variants have large insertions between the third and fourth cysteines of the N-terminal domain (Fig. 3), that appear to reflect variations in length of the AB loop and in the loop between β strands B and C. A single 2-domain TIMP molecule in insects has been suggested to reflect the origin of the four mammalian TIMPs from a single common ancestor during vertebrate evolution. The four TIMPs found in the purple sea urchin and the three TIMPs in Hydra magnipapillata do not correspond to the mammalian subtypes. Nematodes have genes for multiple TIMP homologues that are generally single-domain proteins corresponding to the N-terminal inhibitory domains of mammalian TIMPs but have no region corresponding to the C-domains of mammalian TIMPs [12]. Although this suggests that the single-domain protein is a model for the ancestral protein, H. magnipapillata, which is thought to have diverged from the line leading to the vertebrates at a more ancient time than flies and nematodes [58], has a two-domain TIMP. Therefore, the single-domain nematode TIMP may have developed by loss of a domain from a two-domain ancestor or it could represent a distinct gene line. Based on this consideration and the variable numbers of subtypes in different species, TIMP evolution does not appear to have been a linear process in the invertebrates.

Open in a separate window

Fig. 3

An alignment of the complete amino acid sequences of TIMPs from invertebrates. The location of secondary structure elements projected from the known three-dimensional structures of mammalian TIMPs are marked underneath. “E” denotes residues in extended structures (β-strands) while bars identify α-helices. The species from which the TIMPs are derived are abbreviated as follows: Hydra, Hydra magnipapillata; Nematostella, Nematostella vectensis (sea anemone); Mytilus, Mytilus californianus (mussel); Strong, Strongylocentrus purpuratus (sea urchin); Crassostrea, Crassostrea gigas (oyster); HumanT3, human TIMP-3; Branchiostoma, Branchiostoma floridae (Florida lancelet); Drosophilam, Drosophila melanogaster; Tribolium, Tribolium castaneum (red flour beetle); Culex, Culex quinquefasciatus (Southern house mosquito); Aedes, Aedes aegypti (yellow fever mosquito); Clytia, Clytia hemisphaerica (jellyfish).

The TIMP homologues from nematodes have the characteristic pattern of cysteines and other conserved regions of sequence that are found in the N-domain of TIMPs and some netrin-fold proteins. They also have a secretion signal sequence. However, the amino acid residue present at position 2 in the mature forms of some of these proteins are atypical for a TIMP, for example, Lys, Arg or Gln. Substitution of these amino acids for Thr 2 in human TIMP-1 is known to produce a major loss of affinity for MMPs [59]. This suggests the possibility that these proteins, although homologous with the TIMP N-domains, may not function as MMP inhibitors. Recently, a TIMP homologue from the hookworm, Acylostoma caninum, was cloned and expressed as a His-tagged form cytoplasmically in E. coli [60]. It was soluble and, after purification, was found to have inhibitory activity against some human MMPs. The inhibition studies were conducted at a single relatively high inhibitor concentration of about 500 nM [60]. The implications of these results are unclear, because disulfide bonds do not readily form in proteins expressed in the cytoplasm of E. coli (other TIMPs have been found to be expressed in E. coli as denatured inclusion bodies and require oxidative in vitro folding to become active), and also because such a high concentration was required for MMP inhibition. One possible interpretation is that the TIMP preparation is unfolded and inhibits by acting as an alternative substrate.

3.3. Characterization of invertebrate TIMPs

There have been some studies of expression patterns of TIMPs in non-mammalian species including invertebrates [61–63], but only limited information is available regarding the functional properties of invertebrate TIMPs. Wei et al. cloned the N-terminal inhibitory domain of the single TIMP from a D. melanogaster embryonic cell library (dN-TIMP) and expressed it in E. coli as inclusion bodies. The protein was folded and purified and found to be an effective inhibitor of the two MMPs from Drosophila in vitro [64]. These studies were only semi-quantitative because of the relatively low activities of Drosophila MMPs on synthetic substrates, but more rigorous studies with mammalian MMPs showed that dN-TIMP inhibits MMPs-1, -2, and -14. It resembles human N-TIMP-3 in being a potent inhibitor of ADAM17, but it is a relatively weak inhibitor of ADAM10; this is not surprising because of the weak inhibitory activity of human N-TIMP-1 and N-TIMP-3 for ADAM10 [35]. The relatively low activity against mammalian MMP-1, a collagenase, was proposed to reflect an adaptation to differences in the composition of the extracellular matrix between insects and vertebrates. The Drosophila genome contains no genes for fibrillar collagens. Thus, Drosophila TIMP (dTIMP) would not be under selective pressure to acquire or maintain features required to bind effectively to the active sites of collagenases that cleave fibrillar collagens. Based on modeling, dN-TIMP appears similar in surface charge properties to N-TIMP-3. It is also closer to mammalian TIMP-3 in sequence, functional properties and predicted structure than to the other three TIMPs [64]. Disruption of Timp in Drosophila generates a subviable phenotype. In adult flies, this mutation produces inflated wings, bloated guts and tissue autolysis, impaired phototactic responses, and premature death, indicating that active dTIMP is needed for balanced ECM turnover, appropriate cell–cell or cell–matrix interactions and a functional nervous system in Drosophila [65].

In contrast to the situation in insects, fibrillar collagens have a wide distribution in many other invertebrate species, being present in cnidarians such as hydra and sea anemone as well as the choano-flagellate, Monosiga brevicollis and the demosponge, Amphimedon queenslandica. The most recent common ancestor of cnidarians is thought to be more ancient than that of insects [57], so that it appears that fibrillar collagen genes may have been lost in the line leading to insects. Most or all metazoans appear to have genes that encode TIMPs that might be expected to function as metalloproteinase inhibitors. While there is little information about their functional properties, it is reasonable to speculate that they also regulate turnover of ECM components including fibrillar collagens.

5.1. Cell growth promotion

When the cDNA for TIMP-1 was first cloned [74], it was found to be identical to a factor that has erythroid potentiating activity [75]. Later, TIMP-1 was shown to have cell growth promoting activity on various cell types including keratinocytes and fibroblasts [76,77]. TIMP-2 also has erythroid potentiating activity [78] and cell growth promoting activity on many cell types [79] including metanephric mesenchyme cells during nephron morphogenesis [80]. These effects are independent of MMP inhibition, because TIMPs that lack MMP inhibitory activity either by mutations or reduction and alkylation retained cell growth promoting activity [81] and they were not produced by synthetic MMP inhibitors. In an investigation of the cellular events associated with TIMP-induced cell growth, Wang et al. [82] found that both TIMP-1 and TIMP-2 increased the level of Ras-GTP, but utilize different signaling pathways: TIMP-1 activates the tyrosine kinase/mitogen activated protein kinase (MAPK) pathway, whereas TIMP-2 signaling is mediated by protein kinase A activation which is directly involved in Ras/phosphoinositide 3-kinase (PI3-K) complex formation. This suggests that TIMP-1 and TIMP-2 have distinct receptors. Recent studies have shown that TIMP-1 binds to CD63 [83] and TIMP-2 to α₃β₁ integrin [84], but these interactions have been found to inhibit apoptosis and arrest cell growth, respectively. Thus, the mitogenic activities of the two TIMPs are not explained by the involvement of these receptors, although there seem to be some similarities in the activation of signaling molecules and anti-apoptotic effects may be, in part, related to an effect on cell growth. Recently, D’Alessio et al. [85] have suggested that the effect of TIMP-2 on cell growth is mediated through its binding to MMP-14 on the cell surface and subsequent extracellular regulated kinase (ERK)1/2 activation. Since a TIMP-2 mutant with an extra alanine added to the N-terminus (Ala+TIMP-2), which does not bind to the active site of MMP-14, also activated ERK1/2, it was suggested that the interaction between TIMP-2 and MMP-14 does not involve the active site of the enzyme [85]. This is, however, inconsistent with the observation that a hydroxamate MMP inhibitor that blocks TIMP-2 binding to the active site of MMP-14 inhibited MMP-14-dependent ERK1/2 activation and abrogated cell growth. Further studies are needed to clarify the mechanism by which the TIMP-2–MMP-14 system promotes cell proliferation.

5.2. Cell growth arrest and anti-apoptotic activity of TIMP-1 and TIMP-2 lead to cell differentiation

In addition to their mitogenic activities, TIMP-1 and TIMP-2 suppress the growth of some cells. In vitro studies with human breast epithelial (MCF10A) cells showed that TIMP-1 reduces their growth rate by inducing cell cycle arrest at G₁ through the down-regulation ofcyclin D₁ via the up-regulation of cyclin-dependent kinase inhibitor p27^KIP1 [86]. TIMP-1 also has anti-apoptotic effects on other cell types [87–90]. While this effect of TIMP-1 on hepatic stellate cells was reported to be due to the inhibition of MMPs [90], the effects on other cells are not mediated by MMP inhibition [87–89]. Using the MCF10A cell system, Jung et al. [83] found that the anti-apoptotic activity is mediated through TIMP-1 binding to CD63, a member of the tetraspanin family. CD63 interacts with the β1 subunit of integrins and the TIMP-1–CD63–integrin β1 complex thus constitutively turns on survival signals through the activation of focal adhesion kinase (FAK), PI3-K and ERK pathways [89,91], similar to transformed cell survival signaling. In a 3D matrigel matrix, MCF10A cells form polarized acinar-like structures. The maintenance of these structures requires apoptosis in centrally located cells [83]. The over-expression of TIMP-1 in these cells disrupts the structures by inhibiting caspase-mediated apoptosis and fills the luminal space in the glandular epithelial structure by reducing cell polarization. This is one of the features of ductal cancer development and over-expression of TIMP-1 and its anti-apoptotic effect may be associated with oncogenic cellular transformation.

TIMP-1 levels in serum and TIMP-1 expression in cancerous tissues are associated with poor clinical outcomes in many types of cancer [92–95]. One possible explanation for this is that an elevated level of mesenchymal TIMP-1 increases hepatocyte growth factor (HGF; scatter factor) signaling by blocking a sheddase (possibly ADAM10) of the HGF receptor, cMet, making the liver more susceptible to metastasis [96]. The over-expression of TIMP-1 also elevates HGF and HGF-activating proteases in the liver [96]. Another potential cause is a combination of cell cycle arrest and anti-apoptotic activities of TIMP-1, which leads to cellular transformation. Guedez et al. [87] found that TIMP-1-positive Burkitt’s lymphoma cell lines were resistant to Fas-dependent and Fas-independent (cold-shock, serum deprivation and γ-irradiation) apoptosis, but TIMP-1 negative cells were not. TIMP-1 promotes germinal center B cell survival and differentiation to plasma cells by up-regulating CD40, CD23 and IL-10, and down-regulating CD77 [97]. It also promotes plasmacytic/plasmablastic differentiation in Burkitt lymphoma cells [98], as well as hematopoietic cell differentiation [99].

TIMP-2 suppresses growth factor-stimulated cell proliferation at low to sub-nanomolar concentrations, similar to the concentrations that enhance cell growth [79,100]. This suppressive effect was shown to be mediated by the activation of adenylate cyclase, increased intracellular cAMP levels and the activation of SH2-protein tyrosine phosphatase-1 (SHP-1) which dampens the activity of growth factor receptors [100] and inhibits p42/44 MAPK activation [101]. Stetler-Stevenson et al. [84] identified α₃β₁ integrin as the receptor of TIMP-2, and showed that vascular endothelial cell growth factor (VEGF)-A-stimulated or fibroblast growth factor (FGF)-2-stimulated endothelial cell proliferation is inhibited through the interaction of TIMP-2 with this integrin; this results in the transfer of the β₁ integrin-associated SHP-1 to the VEGF or FGF receptor and its de-phosphorylation. In addition, TIMP-2 induces G₁ growth arrest in endothelial cells through the de novo synthesis of p27^KIP1, which results in the inhibition of cyclin-dependent kinases 4 and 2 and subsequent hypophosphorylation of retinoblastoma protein [102]. These observations may explain how TIMP-2 arrests cell growth and inhibits angiogenesis.

TIMP-2 is expressed in post-mitotic neurons [103] and promotes cell differentiation and neurite outgrowth [104], as the result of cell cycle arrest through decreased expression of cyclins B and D and increased production of cyclin-dependent kinase inhibitor p21^Cip. TIMP-2 expression is found only in α₃ integrin-positive cells [105], suggesting that TIMP-2-α₃β₁ integrin interactions participate in neurogenesis by withdrawing progenitor cells from the cell cycle and permitting their terminal neuronal differentiation.

While cell cycle arrest and the inhibition of apoptosis appear to be closely related to cell differentiation, as in case of TIMP-1, it is not clear whether the anti-apoptotic activity of TIMP-2 is essential for its differentiation-promoting activity because of the limited availability of information about the apoptotic activity of TIMP-2. Nonetheless, the over-expression of TIMP-2 in B16F10 murine melanoma cells rendered these cell more resistant to apoptosis [106] and in a mouse model of atherosclerosis, the over-production of TIMP-2 reduced plaque instability and inhibited the apoptosis of macrophages and macrophage-derived foam cells [107]. Lim et al. [108], on the other hand, reported that TIMP-2 causes the apoptosis of activated T cells and that this is due to inhibition of the cleavage of Fas ligand. Clearly, further investigation of the effects of TIMP-2 on apoptosis in other cell systems would be of considerable interest.

5.3. TIMP-3 and apoptosis

TIMP-3 induces apoptosis in a number of cancer cell lines [109–112] and rat vascular smooth muscle cells [113]. Smith et al. [111] proposed that the pro-apoptotic effects of TIMP-3 result from its inhibition of TACE (ADAM17) with subsequent stabilization of TNF receptors. TIMP-3 inhibits the shedding of a number of cell surface molecules. Ahonen et al. [114] confirmed that TIMP-3 stabilizes three distinct death receptors (TNF receptor 1, FAS, and TNF-related apoptosis inducing ligand receptor 1 (TRAIL-R1)) in melanoma cells in culture and in vivo. TIMP-3 binds to the ECM after secretion [11]. However, recent studies by Troeberg et al. [34] indicate that TIMP-3 secreted from cultured chondrosarcoma cells and chondrocytes does not accumulate in the extracellular space or medium as it is endocytosed by a LDL receptor related protein (LRP). The sequestration of TIMP-3 by binding to the ECM or through endocytosis may help to protect cells from apoptosis while still allowing it to control pericellular proteolysis through its potent inhibition of MMPs, ADAMs and ADAMTSs.

5.4. Angiogenesis and TIMPs

Cartilage, an avascular tissue, contains an anti-angiogenic activity that is closely associated with anti-proteolytic activity, and the latter activity was identified as a collagenase inhibitor [115]. Moses et al. subsequently isolated the anti-angiogenic activity from bovine cartilage and found that its N-terminal sequence is similar to that of human TIMP-1 [116]. It is now well established that all four TIMPs have anti-angiogenic activity (see [117]). Because a number of MMPs, particularly gelatinases and MT1-MMP, are involved in endothelial cell migration and capillary formation, the inhibition of MMPs by TIMPs can prevent angiogenesis [117]. However, TIMP-2 and TIMP-3 also exhibit anti-angiogenic activity that is independent of their MMP inhibitory activities.

As discussed above, the anti-angiogenic activity of TIMP-2 is mediated through its interaction with α₃β₁ integrin. Feldman et al. [118] found that the over-expression of TIMP-2 in murine colon cancer cells inhibited tumor growth and angiogenesis in vivo by up-regulating MAP kinase phosphatase 1 which dephosphorylates p38 MAPK, a molecule involved in endothelial cell proliferation and migration. In addition, TIMP-2 enhances the expression of RECK (the reversion-inducing-cysteine-rich protein with Kazal motif), a membrane-anchored protein with inhibitory activity for MMP-2, -9 and -14 and ADAM10 [119,120]. Enhanced RECK expression is associated with the inhibition of endothelial cell migration [121]. The interaction of TIMP-2 with α₃β₁ integrin reduces Src tyrosine kinase levels which alters the pattern of paxillin phosphorylation. This leads to a signaling switch from Rac 1 to Rap 1, resulting in enhanced RECK expression and loss of a migratory phenotype [122]. The anti-angiogenic activity of TIMP-2 has been localized to a region of sequence in the C-terminal domain [123], which may be the site of interaction with α₃β₁ integrin.

TIMP-3 is also a potent inhibitor of angiogenesis. It binds directly to VEGF receptor 2 and blocks the action of VEGF on endothelial cells [124]. TIMP-3 also binds to angiotensin II type 2 receptor, and the over-expression of both TIMP-3 and angiotensin II type 2 receptor additively inhibits angiogenesis [125].

TIMP-2 and TIMP-3 also play an important role in the stabilization of capillary networks during angiogenesis. This process is important as newly formed capillary networks are predisposed to undergo regression, and their stabilization requires pericyte recruitment. Saunders et al. [126] have shown that endothelial cell–pericyte interactions stimulate the expression of TIMP-3 in pericytes. The TIMP-3 from this source together with TIMP-2 expressed in endothelial cells participates in vascular stabilization by inhibiting a number of MMPs and ADAMs [126] and probably by interacting with VEGFR-2 [124]. These actions of TIMPs inhibit endothelial cell tube formation and tube regression, which leads to the cessation of endothelial cell activation and the development of their quiescence [126]. The recruitment of pericytes and their interaction with endothelial cells also stimulate the assembly of vascular basement membrane matrix assembly which is a crucial step in vessel maturation [127].

6.1. TIMP-1^−/− mice

Nothnick and colleagues found that TIMP-1-null mice exhibit a number of alterations in processes associated with reproduction and steroidogenesis (see Table 3), suggesting that TIMP-1 has a multifaceted role in regulating the murine reproductive cycle involving the uterus, ovary and testis. TIMP-1 is also an important regulator of heart tissue remodeling. TIMP-1^−/− mice display changes in left ventricular structure and cardiac function [133]; they also display exacerbated left ventricular remodeling after myocardial infarction [134,135].

Partial hepatectomy in mice increases TIMP-3 mRNA within 6 h and TIMP-1 and TIMP-4 mRNAs by 48 h. A number of MMPs are also upregulated in regenerating livers in these time periods. Transgenic mice that underexpress TIMP-1 display accelerated hepatocyte cell cycle progression, and in TIMP-1^−/− mice HGF activity release from the ECM is increased in regenerating liver, leading to an increase in the phosphorylation of cMet receptor and activation of p38 MAPK [136]. TIMP-1 over-expressing mice delayed cell cycle progression, suggesting that TIMP-1 negatively regulates HGF activity during liver regeneration by regulating the activities of MMPs that activate and release pro-HGF from the ECM [136].

TIMP-1-null mice have increased resistance to corneal and pulmonary infection by Pseudomonas aeruginosa [137]. This was reversed by a MMP inhibitor, indicating that it results from a loss of MMP inhibitory activity. MMP-9, MMP-7 and MMP-3 are considered to be important for resistance to infections, but the mechanism of their influence is not known. Possibilities include the alteration of innate immunity by MMP-7-catalyzed conversion of pro-defensin to active defensin [138] or the cleavage of syndecan-1 and mobilization of chemokines [139].

Another interesting phenotype of TIMP-1^−/− mice is a decrease in adipose tissue weight when on a high fat diet [140]. This appears to contrast with the increased food intake presented by TIMP-1^−/− female mice despite their hyperleptinemia [141]. Since leptin promotes the expression of TIMP-1 mRNA in the hypothalamus, TIMP-1 contributes to the regulation of feeding and energy balance. It is also notable that TIMP-1^−/− mice have impaired learning and memory as discussed above.

6.2. TIMP-2^−/− mice

TIMP-2-null mice display a number of phenotypes associated with defects in the central nervous system and myogenesis (see Table 3). As discussed earlier, TIMP-2^−/− mice exhibit a deficiency in prepulse inhibitor of the startle reflex (a measure of pre-attentional sensorimotor gating) and in the acquisition of fear-potentiated startle, a model of synaptic plasticity in the amygdala [131]. They also have motor defects, decreased cerebellar neurite outgrowth [132], and delayed neuronal differentiation [104]. These effects are probably due to an abnormal turnover of neuronal ECM and indicate that TIMP-2 is an important promoter of neuronal differentiation.

Altered myotube formation by TIMP-2^−/− myoblasts [142] and weakened muscle and reduced fast-twitch muscle mass in TIMP-2^−/− mice have been reported [143]. In both cases the level of β1 integrin is altered; also, its level is reduced in the fast-twitch muscle of TIMP-2^−/− mice [143], suggesting that ECM-cytoskeletal interactions required for muscle contraction are destabilized.

We thank Professor Gillian Murphy for critical reading of the manuscript. This work is supported by NIH/NIAMS grant AR40994, the Wellcome Trust grant 075473, and a grant from Arthritis Research Campaign in the UK.