Sphingosine-1-phosphate-induced release of TIMP-2 from vascular smooth muscle cells inhibits angiogenesis

Angiogenesis is the growth of new blood vessels from the existing vasculature (Carmeliet and Jain, 2011). It is a key process in many (patho)physiological conditions and is tightly regulated. Angiogenic signals induce endothelial cells (ECs) to undergo a phenotypic change resulting in cellular proliferation and migration. This leads to the formation of sprouts from the existing blood vessels (Nicosia, 2009). Following growth of a sprout outward from the vessel, a complex set of cues leads to reformation of specialised endothelial–endothelial cell junctions (adherens junctions) (Carmeliet and Jain, 2011). Ultimately, lumen formation (tubulogenesis) occurs, although this process is still not clearly understood (Xu and Cleaver, 2011). The signals that regulate angiogenesis vary depending on the cellular environment, although typically, the growth factor vascular endothelial growth factor (VEGF) has a major role in most angiogenic scenarios. VEGF expression is induced by hypoxia, which is an important angiogenic stimulus (Fong, 2009).

Sphingolipids, which are produced predominantly from sphingomyelin metabolism, are now known to have important effects on blood vessel function and might have a regulatory role in angiogenesis (Peters and Alewijnse, 2007; Lucke and Levkau, 2010). In particular, sphingosine 1-phosphate (S1P), a bioactive sphingolipid mediator present in plasma and released by activated platelets, can regulate vascular processes including motility and contractility, vascular proliferation and endothelial barrier permeability (Yatomi et al., 1995; Coussin et al., 2002; Kluk and Hla, 2001; Wang and Dudek, 2009). At a cellular level, S1P-induced effects are mediated through seven-transmembrane, G-protein-coupled receptors, S1P1–S1P5 (Meyer zu Heringdorf and Jakobs, 2007). These receptors activate a repertoire of intracellular signalling pathways (Sanchez and Hla, 2004). S1P1, through the heterotrimeric G-protein Gi, leads to activation of several pathways including extracellular signal-regulated protein kinases 1 and 2 (ERK1/2). S1P2 and S1P3 couple to G12 or G13, and Gq, respectively, and predominantly activate intracellular Ca²⁺ release and the monomeric G-protein RhoA (Rosen et al., 2009). Most cell types in the vasculature express at least two isoforms of S1P1, S1P2 or S1P3.

In cardiovascular disease, angiogenesis occurs following myocardial infarction (MI) (van der Laan et al., 2009). This process can lead to an enhanced reperfusion of damaged myocardium and might be therapeutically beneficial. One major stimulus for angiogenesis in this case is probably the hypoxia that results from thrombus formation and subsequent disruption in blood supply to the heart. This hypoxia leads to an upregulation of hypoxia-inducible transcription factors resulting in increased expression of angiogenic growth factors such as VEGF (Zachary and Morgan, 2011). Importantly, in the angiogenic environment produced post MI, there will also be a local increase in the S1P released from activated platelets which constitute the thrombus. Serum concentrations of S1P have been estimated in the high nM to low μM concentration (Yatomi et al., 1995). Such a high local concentration is within the range that would activate S1P receptors present on ECs and exposed smooth muscle cells (SMCs). The effects of S1P on angiogenesis in human arteries is therefore of particular relevance to cardiovascular biology. Owing to the range of intracellular signaling pathways activated by S1P and the relatively promiscuous expression pattern, S1P-induced effects on blood vessel growth in the vasculature are complex (Argraves et al., 2010). Studies have demonstrated that S1P can regulate endothelial cell morphology and migration through a variety of intracellular signals affecting focal adhesion contacts and integrins (Sun et al., 2009; Bayless and Davis, 2003). In general, S1P seems to be angiogenic in vitro, leading to the formation of capillary-like networks on extracellular matrix substrates (Lee et al., 1999; Yonesu et al., 2009). These effects are predominantly through S1P1 activation. In vivo, S1P is also angiogenic, and inhibition of S1P1 and S1P3 prevents vessel formation in Matrigel implant mouse models implicating both these receptor isoforms (Lee et al., 1999; Chae et al., 2004; LaMontagne et al., 2006). However, S1P2 might be angiostatic, because S1pr2^−/− mice demonstrated increased angiogenesis in a mouse model of tumour development (Du et al., 2010). Indeed, in vitro and in vivo S1P2 signalling can inhibit capillary network formation in ECs through activation of RhoA and subsequent inhibition of Rac (Inoki et al., 2006). By contrast, S1P2 has also been shown to induce angiogenesis in a different vascular bed, suggesting tissue specificity (Skoura et al., 2007). Ultimately, therefore, the angiogenic or angiostatic effects of S1P in vivo will be determined by the relative expression of S1P receptor isoforms and tissue environment. Another important aspect of S1P effects on angiogenesis relates directly to the complex interaction of S1P-induced signals from several cell types in the immediate environment where the angiogenesis occurs. Such an environment will have a number of different cell types including vascular SMCs and ECs, all expressing multiple isoforms of the S1P receptor. This will activate a wide range of intracellular mechanisms and the convergence of these signals is likely to be a determining factor in vivo. To date, the effects of this complex interaction on angiogenesis has not been examined.

In the current study, we have used in vitro and ex vivo models of angiogenesis to examine the role of S1P in an angiogenic environment that contains multiple cell types. Our results demonstrate, using a co-culture model that includes both human coronary artery ECs and human coronary artery SMCs, that S1P has an inhibitory effect on angiogenesis. These results were also reflected in ex vivo human arteries. S1P-induced inhibition of angiogenesis is not due to a direct effect on ECs but occurs through S1P-mediated effects on SMCs. The angiostatic effect of S1P in this multi-cell type environment is through activation of S1P2 and the RhoA–Rho-kinase pathway in SMCs, leading to release of tissue inhibitor of metalloproteinases-2 (TIMP-2). TIMP-2 leads directly to a decrease in adherens junction formation. These findings suggest that the effects of S1P on angiogenesis in human arteries are regulated by hitherto unknown interactions with SMCs.

Initially, we determined expression of S1P receptor isoforms in human coronary artery ECs, which has not previously been examined. Immunoblotting revealed that human coronary artery ECs do not express S1P1 above the levels of detection but express both S1P2 and S1P3 (Fig. 1A). HUVECs express S1P1, S1P2 and S1P3.

We developed an in vitro co-culture model of angiogenesis of human coronary artery SMCs, human fibroblasts and human coronary artery ECs. Under specific conditions, the ECs differentiated into an anastomosed multicellular network of lumen bearing tubular structures that reached a maximum size at 14 days. These tubules were visualised by immunostaining for MCAM, which is specifically expressed by ECs (Fig. 1B) (Schmid et al., 2007). Initially, only fibroblasts and human coronary artery ECs (no SMCs) were seeded, which similarly results in the formation of endothelial tubules. Under these conditions, incubation with S1P slightly, but not significantly, reduced tubule formation (Fig. 1C). When ECs were co-cultured with fibroblasts and human coronary artery SMCs, S1P significantly reduced tubule formation by approximately 80%. To determine whether this mechanism involves activation of S1P2 and also the RhoA–Rho-kinase pathway, the effects of the S1P2 antagonist JTE-013 and the Rho-kinase inhibitor Y-27632 were assessed. Incubation with either inhibitor significantly reversed the S1P-induced inhibition of endothelial tubule formation (Fig. 1D).

The anti-angiogenic effect of S1P could be due to an inhibition of adherens junction formation (Vestweber, 2008). This could occur through a change in the expression of VE-cadherin, an important protein in the formation of adherens junctions (Carmeliet et al., 1999), or could be a failure of the adherens junctions to form properly. In coronary artery ECs incubated with S1P for 7 days, there was no difference in VE-cadherin expression (Fig. 2A). The localisation of VE-cadherin was examined in immunofluorescence studies as an indicator of endothelial–endothelial cell junctional integrity. VE-cadherin staining revealed tightly assembled junctions between ECs in co-cultures containing fibroblasts, ECs and SMCs. In co-cultures prepared identically but with S1P added to the medium, VE-cadherin staining showed a distinctly diffuse pattern around EC junctions (Fig. 2B).

To assess whether a diffusible factor is involved in the intercellular inhibitory effect of S1P, SMCs were incubated with S1P for 48 hours. SMC conditioned medium was added to confluent fibroblasts in a separate dish. Human coronary artery ECs were seeded into the dish and tubule formation was assessed after 14 days. Control experiments used conditioned medium from SMCs that had not been incubated with S1P. In parallel, unconditioned medium with and without S1P (incubated in a well with no cells for 48 hours) was added to confluent fibroblasts before seeding with ECs. In unconditioned medium, S1P had a small but non-significant inhibitory effect on tubule formation (Fig. 3A). Co-cultures treated with conditioned medium from unstimulated SMCs demonstrated tubule formation at a similar level to that observed above. However, co-cultures incubated with conditioned medium from S1P-treated coronary artery SMCs had a significant reduction in tubule formation (Fig. 3B). Because activation of S1P2 has been indicated as an anti-angiogenic pathway, we determined whether the conditioned medium from SMCs might increase the expression of S1P2 in ECs. In ECs maintained in S1P-treated conditioned medium there was no difference in the expression of S1P₂ compared with untreated conditioned medium (Fig. 3C). S1P3 was also unchanged.

TIMPs are a family of four proteins that are endogenous inhibitors of angiogenesis and released by vascular cell types (Fabunmi et al., 1996). We determined whether an increase in TIMP-2 activity occurred because this TIMP isoform is reportedly increased following myocardial infarction (Webb et al., 2006). By reverse zymography, co-cultures containing fibroblasts, ECs and SMCs incubated with S1P for 48 hours, showed a significant increase in TIMP-2 release into the medium compared with similar co-cultures without SMCs. Incubation with the Rho-kinase inhibitor Y-27632 significantly reversed this S1P-induced increase in TIMP-2 (Fig. 4A). To establish which cell type(s) were the source of TIMP-2, fibroblasts, ECs and SMCs were incubated separately with S1P. Incubation of human coronary artery SMC monolayers with S1P (1 μM) for 48 hours significantly increased TIMP-2 release in comparison to S1P-treated ECs and fibroblasts (Fig. 4B). Addition of Y-27632 significantly reversed the S1P-induced TIMP-2 release from SMCs. To confirm that TIMP-2 is involved in the inhibition of tubule formation in our co-culture model, human fibroblasts, human coronary artery SMCs and human coronary artery ECs were incubated with a TIMP-2 blocking antibody (2.5 μg/ml) with or without S1P for 14 days. The TIMP-2 blocking antibody significantly reversed the S1P-induced inhibition of endothelial tubule formation (Fig. 4C). After 14 days, the co-cultures containing fibroblasts, ECs and SMCs, were examined for VE-cadherin localisation using immunofluorescence. VE-cadherin staining revealed a diffuse pattern around the EC junctions following S1P treatment (Fig 4D). In co-cultures incubated with S1P and the TIMP-2 blocking antibody, VE-cadherin staining showed tightly assembled junctions between ECs analogous to control samples.

To confirm the results achieved using the in vitro angiogenesis model, we used freshly isolated human arteries. The effects of S1P were determined on rings of freshly isolated human mammary artery embedded in Matrigel and incubated in medium containing serum (to induce angiogenesis) for up to 14 days. In control rings, endothelial sprouts were visible after 5 days and increased throughout the 14 day period (Fig. 5A). These sprouts were confirmed to contain predominantly ECs by uptake of the specific endothelial marker, Dil-tagged acetylated LDL. In rings treated with S1P, the growth of endothelial sprouts was significantly inhibited from day 7 onwards reaching a maximum of approximately 70% reduction at 14 days (Fig. 5B). The inhibitory effect of S1P on EC sprouting was partially but significantly inhibited by pretreatment with the S1P2 receptor antagonist JTE-013. Addition of the Rho-kinase inhibitor, Y-27632, also significantly reversed the S1P-induced inhibitory effect (Fig. 5C). To determine whether the inhibitory effect of S1P on angiogenesis in the ex vivo model was similarly acting through a TIMP-2-mediated mechanism as observed in co-culture experiments, we incubated human mammary artery rings with S1P in the presence of anti-TIMP-2 blocking antibody. The growth of endothelial sprouts was significantly inhibited by blocking TIMP-2 (Fig. 5D).

S1P has been shown to influence several aspects of EC function by a variety of different mechanisms (Lucke and Levkau, 2010). This is typically ascribed to a direct action of S1P on S1P receptor isoforms expressed on the EC membrane. Some in vivo studies have, however, revealed contradictory results. Such variation could be due to interactions with other cell types also expressing S1P receptor isoforms. Our current study using a multi-cell-type co-culture system now clearly demonstrates that the overall effects of S1P on angiogenesis in human arteries are a direct consequence of S1P-mediated signaling mechanisms in vascular SMCs. The observed interaction of SMCs and ECs ultimately dictates the angiogenic response. In our in vitro system, we demonstrate for the first time using human coronary artery ECs in co-culture models of angiogenesis that S1P inhibits tubule formation. This inhibitory effect is not a direct action on ECs but occurs through enhanced TIMP-2 release from SMCs, leading to decreased adherens junction formation. These effects occur through S1P2 signaling and activation of the RhoA–Rho-kinase pathway.

The effects of S1P on ECs are regulated by several mechanisms and, although several studies have reported that S1P is angiogenic (Lee et al., 1999; Chae et al., 2004; LaMontagne et al., 2006), this is likely to be critically dependent on the environment and tissue specificity. In pathological angiogenic environments (including tumour angiogenesis, neovascularisation in the eye and increased blood flow following hind limb ischemia), S1P promotes angiogenesis (Chae et al., 2004; Skoura et al., 2007; Oyama et al., 2008). Studies using S1pr2^−/− mice have demonstrated that induction of tumour angiogenesis is enhanced, suggesting that S1P2 can be angiostatic (Du et al., 2010). Our results using human coronary artery ECs reveal that S1P has an angiostatic effect through an S1P2-mediated mechanism, but only within a multi-cell-type environment containing fibroblasts and SMCs. When SMCs are present, S1P has a dramatic negative effect on tubule formation. Importantly, this is also observed in human arterial rings where, in the presence of SMCs within the vessel wall, endothelial sprouting was inhibited, suggesting relevance to the in vivo environment. This is the first study to demonstrate that SMCs can regulate S1P-mediated effects in angiogenesis.

We now reveal that the inhibitory effect of S1P on angiogenesis occurs through an intercellular signaling mechanism involving TIMP-2 (Fig. 6). Our results indicate that TIMP-2 is released from SMCs and is responsible for the S1P-induced angiostatic effect because block of TIMP-2 prevented tubule formation in vitro and endothelial cell sprouting ex vivo. Previous studies have demonstrated that TIMP-2 can inhibit growth-factor-induced angiogenesis in vitro (Murphy et al., 1993) and prevent angiogenesis in vivo (Valente et al., 1998). More specifically, TIMP-2 can inhibit EC proliferation, EC invasion and tubular morphogenesis (Saunders et al., 2006). The mechanisms of TIMP-2 inhibition of these processes associated with angiogenesis are complex because this TIMP isoform has multiple effects (Stetler-Stevenson and Seo, 2005). The S1P-induced angiostatic effect observed here could potentially be due to inhibition of MMP activity through TIMP-2 release. Recent studies have, however, shown that TIMP-2 forms a complex with pro-MMP-2 and membrane type 1 MMP, which is required for cell surface activation of MMP-2 (Wang et al., 2000). This effect would be pro-angiogenic, the opposite effect to that observed with S1P in our study. At least some of the angiostatic effects of TIMP-2 are now known to be MMP independent (Murphy et al., 1993; Stetler-Stevenson and Seo, 2005). One mechanism involved in this MMP-independent effect is the binding of TIMP-2 to the α3β1 integrin receptor on the endothelial cell membrane (Seo et al., 2003). The intracellular signals involved might include activation of protein tyrosine phosphatase Shp-1 and mitogen-activated protein kinase phosphatase 1 (Seo et al., 2003; Feldman et al., 2004). Such an effect could be involved in the S1P effects observed in the current study. Although we have yet to delineate the mode of action of TIMP-2 in our co-culture model, we have demonstrated at least one downstream TIMP-2-dependent effect. S1P prevented the formation of adherens junctions (as assessed by the localisation of VE-cadherin) through TIMP-2. VE-cadherin localisation at adherens junctions is essential for proper junction integrity and therefore tubule formation (Wang et al., 2010). Further research is required to delineate the exact mechanisms of how TIMP-2 regulates adherens junction formation in ECs.

In human coronary artery SMCs, we have demonstrated that S1P-mediated TIMP-2 release occurs by activation of the S1P2 receptor and subsequent activation of the RhoA–Rho-kinase pathway. The mechanism whereby the RhoA–Rho-kinase pathway leads to increased TIMP-2 release in SMCs is not clear. Because TIMP-2 activity is not increased following treatment of either fibroblasts or ECs with S1P, there is some level of cell-type specificity in this pathway. It has been previously demonstrated in hepatic stellate cells that a Rho-kinase inhibitor suppressed transcription and protein expression of TIMP-1 (Fukushima et al., 2005). In addition, statin treatment in vivo (which inhibits the membrane targeting of Rho by preventing geranylgeranylation) decreases the expression of TIMP-2 in rabbit aorta (Chen et al., 2002). It seems therefore that such mechanisms occur in VSM cells, although the reasons for cell specificity remain unclear.

The principal relevance of the current findings is likely to be in pathological situations where elevated levels of sphingolipids occur as a result of acute arterial thrombus (Yatomi et al., 1995). Our results were observed in coronary artery ECs and therefore will correlate with the angiogenic response that occurs after MI. It has previously been suggested that growth of collateral vessels following MI would be advantageous to allow reperfusion of damaged myocardium and alleviate the ischemic environment (van der Laan et al., 2009). Our data suggest that, in this pathological situation, naturally occurring S1P will actively prevent angiogenesis and limit the development of new collateral vessels. In vivo, it would be expected that an increase in local S1P concentrations could reach the concentration used in this study (Murata et al., 2000). In this case, S1P would be acting detrimentally and suggests that an inhibition of S1P receptor signaling (possibly through S1P2) would be of therapeutic benefit in promoting angiogenesis following MI. It is also of particular interest that the effects of S1P in coronary artery ECs are mediated through TIMP-2. Recent studies have demonstrated that the plasma concentration of TIMP-2 is increased in patients in the later stages of MI (Webb et al., 2006). Potentially, some of this increased TIMP-2 could be caused by S1P-induced release. In a model of MI using Timp2^−/− mice, several parameters indicating cardiac dysfunction were increased, suggesting that TIMP-2 might be beneficial in this case (Kandalam et al., 2010). Therefore, although the S1P-induced release of TIMP-2 could inhibit angiogenesis and decrease reperfusion, in the longer term, TIMP-2 could provide some benefit to cardiac function. This remains to be determined.

In conclusion, the current findings in vitro and ex vivo indicate a pathophysiological role for endogenous S1P as an angiostatic mediator. Uniquely, this inhibition of EC angiogenesis occurs by an interaction with vascular SMCs. S1P-induced stimulation of the vascular SMCs leads to an increase in TIMP-2 release and ultimately results in decreased integrity of EC junctions. These findings will have relevance to the angiogenic healing responses of tissues following an acute thrombotic event.

Purified polyclonal anti-S1P1 receptor antiserum was raised in rabbit against the C-terminal (peptide sequence KDEGDNPETIMSSGNVNSSS) to the human S1P1 receptor protein and has previously been characterised (Coussin et al., 2002). Monoclonal anti-S1P2 was obtained commercially from Oncogene Research Products. Polyclonal antibodies against S1P3 and VE-Cadherin were purchased from Santa Cruz Biotechnology. Monoclonal antibodies against CD146 (anti-MCAM) were from Chemicon. Human polyclonal TIMP-2 blocking antibodies were purchased from R&D Systems. JTE-013 was purchased from Calbiochem. The specific Rho-kinase inhibitor, Y-27632, was from Biomol Research Labs. Dil-tagged acetylated low density lipoprotein (LDL) was purchased from AbD Serotec. Horseradish peroxidase (HRP)-conjugated secondary antibodies were from Dako. All other chemicals were from Sigma.

Normal adult human dermal fibroblasts (Clonetics) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% L-glutamine and 1% penicillin (100 U/ml), streptomycin (100 μg/ml). Human coronary artery SMCs (Clonetics) were cultured in 231 medium containing smooth muscle growth factor supplement. Human coronary artery ECs (Clonetics) were cultured in endothelial basal medium supplemented with 5% FBS and essential growth factors (EGM-2MV) (Clonetics). Human umbilical vein endothelial cells (HUVECs) (Clonetics) were also cultured in EGM-2MV. Cells were cultured in a humidified atmosphere of 5% CO₂ at 37°C.

In vitro tubule formation assay was performed on Matrigel, a substitute extracellular matrix, as previously described (Donovan et al., 2001). Matrigel was distributed into 24-well plates and incubated for 30 minutes at 37°C to allow polymerisation. Human coronary artery ECs or HUVECs were seeded into each well at 3×10⁴ cells/well in EGM-2MV in the presence or absence of S1P (1 μM). Plates were incubated at 37°C for 6 hours. Tubules formed were imaged using a Nikon Camera. Tubular length of each sample was quantified by calculating the mean of four separate fields of view using ImageJ analysis software.

Whole-cell homogenates and membrane preparations were obtained as previously described (Hunter and Nixon, 2006). Cells treated for 1 week with either vehicle or S1P (1 μM) were washed in ice-cold TBS (10 mM Tris-HCl, pH 7.4, 150 mM NaCl) and added to lysis buffer containing protease inhibitors and centrifuged at 18,000 g for 10 minutes at 4°C. The supernatant was collected, added to 6× SDS sample buffer and incubated at 100°C for 5 minutes.

For S1P receptor immunoblots, membrane preparations were used. Untreated cells or cells treated for 1 week with either vehicle or S1P (1 μM) were washed with ice-cold TBS, pH 7.4, suspended in 0.5 ml TBS and centrifuged at 10,000 g for 2 minutes at 4°C. Cells were then suspended in 0.5 ml harvest buffer (1 mM EDTA, 10 mM HEPES, 155 mM NaCl, pH 7.4) containing 5 μM leupeptin, 1 μg/ml pepstatin A and 5 μM PMSF and incubated on ice. Extracts were homogenised (30 strokes) in Jencons Dounce homogeniser. Homogenates were subject to centrifugation at 100,000 g for 1 hour at 4°C. The high-speed supernatant was discarded. Pellets were resuspended in 200 μl harvest buffer containing 1% Triton X-100, sonicated (five short 5 second pulses at 50% amplitude) and incubated on ice for 1 hour. Homogenates were mixed with 6× SDS sample buffer and incubated at 100°C for 5 minutes. Protein was measured using Lowry assay (Bio-Rad) to ensure equal protein loading. Samples were fractionated by SDS-PAGE as previously described (Hunter and Nixon, 2006). Membranes were incubated in primary antibodies followed by detection with HRP-conjugated secondary antibodies. The immunoreactive bands were visualised by enhanced chemiluminescence.

Fibroblasts were seeded onto 24-well plates at 2×10⁴ cells per 0.5 ml DMEM and incubated at 37°C for 5 days until a confluent monolayer had been formed. Human coronary artery SMCs were seeded at 3×10⁴ cells per 0.5 ml in 231 medium on top of the fibroblasts for 24 hours until confluent. Human coronary artery ECs seeded at 3×10⁴ cells per 0.5 ml EGM-2MV were finally added to these co-cultures and incubated for 14 days. Co-cultures of fibroblasts and human coronary artery ECs, but not including the SMC layer, were also prepared in a similar manner. Co-culture consisting of only coronary artery SMC and coronary artery ECs (no fibroblasts) did not develop any tubules.

For preparation of conditioned medium, 24-well plates were seeded with no cells or with human coronary artery SMCs at 3×10⁴ cells per 0.5 ml in 231 medium and incubated at 37°C for 24 hours until confluent. All wells were transferred to EGM-2MV medium containing vehicle or S1P (1 μM) and incubated for 48 hours. The unconditioned medium and conditioned medium (+ SMC wells) was transferred to co-culture monolayers of fibroblasts and human coronary artery ECs and incubated for 2 weeks.

Co-cultures were fixed at room temperature in freshly prepared 4% paraformaldehyde in PBS for 30 minutes. Endothelial tubule formation was assessed by immunostaining with primary antibodies against CD146 (anti-MCAM) or VE-cadherin. Goat anti-mouse IgG and goat anti-rabbit IgG Alexa-Fluor-488- or Alexa-Fluor-546-conjugated secondary antibodies were used (Molecular Probes, Invitrogen). Vessels formed were visualised under fluorescence and/or confocal microscopy. Total tubular formation was assessed from four fields of view per well and quantified using ImageJ analysis software.

TIMP-2 release from sample co-cultures was assessed using reverse zymography. This was conducted with 15% polyacrylamide gels containing 0.1% gelatin and MMP-2 from conditioned medium as previously described (Hawkes et al., 2001). The conditioned medium was prepared by seeding human fibroblasts with incomplete serum-free EBM-2 medium for 24 hours at 37°C. Culture medium aliquots from co-cultures of fibroblasts, human coronary artery SMCs and human coronary artery ECs, with or without S1P (1 μM) treatment for 48 hours, were mixed with 2× SDS sample buffer and subjected to electrophoresis at room temperature. Gels were renatured in 2.5% Triton X-100 for 30 minutes and developed overnight at 37°C. The location of the TIMP activity was visualised as an inhibition of gelatin lysis by MMP-2 by positively stained Coomassie Blue bands.

The angiogenic ring assay was adapted from that previously described (Small et al., 2005). Tissue collection was approved by the North of Scotland Research Ethics Committee and informed written consent was obtained from each patient. The investigation conforms with the principles outlined in the Declaration of Helsinki. A proportion of left internal mammary artery that was surplus to requirements was collected from male patients undergoing elective coronary artery bypass grafting. Mean patient age was 64.9±3.6 years. Patient age ranged from 44 to 87 years. Under sterile conditions, periadventitial tissue was removed from the arterial vessels, washed five times consecutively in ice-cold serum-free DMEM supplemented with 1% penicillin (100 U/ml), streptomycin (100 μg/ml) and sectioned into 1 mm rings. Rings were embedded in 200 μl Matrigel (BD Biosciences) and incubated in at 37°C for 2 weeks in 1 ml EGM-2 (Clonetics) containing 2% FBS and essential growth factors in presence and absence of S1P (1 μM). Media was initially changed on day 3 with subsequent changes every 48 hours. Experiments were performed in triplicate per treatment. Vessel outgrowth from rings was analysed daily over the time course. To confirm the endothelial nature of tubule formation, vessel cultures were incubated for 48 hours for uptake of acetylated LDL tagged with a fluorescent marker (Dil).

Data are means ± s.e.m. Comparisons were made by Student's paired t-test or ANOVA with Tukey post-hoc test where appropriate. P values less than 0.05 were considered statistically significant.

The authors would like to thank Eileen Bishop for invaluable technical assistance.