OUP user menu

Statins inhibit cyclooxygenase-2 and matrix metalloproteinase-9 in human endothelial cells: anti-angiogenic actions possibly contributing to plaque stability

Marika Massaro, Antonella Zampolli, Egeria Scoditti, Maria Annunziata Carluccio, Carlo Storelli, Alessandro Distante, Raffaele De Caterina
DOI: http://dx.doi.org/10.1093/cvr/cvp375 311-320 First published online: 27 November 2009


Aims Cyclooxygenase (COX)-2 expression is increased in inflammation and angiogenesis and also in atherosclerotic plaques, where it co-localizes with metalloproteinases (MMPs) involved in the fibrous cap weakening. Insight into the regulation of COX-2 and MMP-9 expression suggests the involvement of a Rho-dependent pathway. Because statins interfere with Rho activation, we investigated the statin effect on COX-2 and MMP expressions in the human endothelium.

Methods and results Simvastatin and atorvastatin were incubated with endothelial cells for 12 h before stimulation with phorbol myristate acetate or tumour necrosis factor-α, for times suitable to assess the endothelial tube differentiation on Matrigel and COX-2 and MMPs activities, proteins, and mRNA expressions. At 0.1–10 µmol/L, both statins reduced COX-2 expression and activity, without affecting COX-1. The statin effect was reversed by mevalonate and geranylgeranyl-pyrophosphate and mimicked by the Rho inhibitor C3 transferase, indicating the involvement of Rho in the signal transduction pathway leading to COX-2 expression. In parallel, statins, as well as COX-2 inhibitors, reduced the MMP-9 stimulated release and the endothelial tubular differentiation.

Conclusion In the human vascular endothelium, statins reduce COX-2 and MMP-9 expression and activity. Through this mechanism, statins exert an anti-angiogenic effect possibly contributing to the cholesterol-lowering-unrelated protective effects of statins against plaque inflammatory angiogenesis and rupture.

  • Statins
  • Cyclooxygenase-2
  • Metalloproteinases
  • Angiogenesis
  • Atherosclerosis

1. Introduction

Acting as effective competitive inhibitors of HMG-CoA reductase, the rate-limiting enzyme in the synthesis of cholesterol,1 statins have now become the most widely used drugs in the treatment of hypercholesterolaemia.2 However, opinions differ on whether the effects of statins on cardiovascular risk are fully accounted for by their lipid-lowering effects, and many believe that their efficacy in clinical trials is related to cholesterol-lowering independent properties.3,4 Statins have indeed been suggested to affect alternative targets involved in inflammatory and/or proliferative phenomena occurring in atherosclerosis,5 preventing the synthesis of important isoprenoid intermediates such as farnesylpyrophosphate (Fpp) and geranylgeranyl-pyrophosphate (GGpp). Through such mechanisms, statins would inhibit post-translational lipid modification and activation of some small G-proteins including Ras and Rho, critically involved in eukaryotic cell signalling.6

Angiogenesis is now emerging as a key regulator of plaque growth and instability.7 Statins can both promote collateral vessel formation in ischaemic tissues and inhibit angiogenesis. More recent reports have, however, reconciled these conflicting data showing biphasic dose-dependent effects, with low doses inducing and higher doses inhibiting angiogenesis (for a review on these issues, see Ii and Losordo8), as well as with effects tightly dependent on micro-environmental (hypoxic vs. inflammatory) conditions.9

Cyclooxygenase (COX)-2 is an inducible pro-inflammatory enzyme with important pro-angiogenic effects.10 Molecular mechanisms of COX-2 involvement in angiogenesis are complex and not fully explained, but the occurrence of a functional relationship between COX-2 activity and the release of metalloproteinases (MMPs) has been demonstrated in several cellular models,11,12 although not in endothelial cells.

Since the activation of COX-2 promoter involves Ras- and Rho-dependent pathways, which are potential targets for statins,1315 we sought to characterize the anti-angiogenic properties of statins investigating their effects on the expressions of COX-2 and related MMPs in human cultured endothelial cells. Using in vitro angiogenic assays, we here confirm a pro-angiogenic role for COX-2 activity and its involvement in the endothelial MMP-9 release. Here, statins effectively reduce the tubular neovascular endothelial cell organization by modulating COX-2 expression and reducing the related MMP-9 release.

2. Methods

A more detailed description of methods and a full description of chemicals used in this study are provided in the Supplementary material online.

2.1 Cell cultures and treatments

Human umbilical vein endothelial cells (HUVECs), human saphenous vein endothelial cells (HSVECs) and bovine aortic endothelial cells (BAECs) were harvested and maintained as described previously.16 They were treated conforming with the principles outlined in the Declaration of Helsinki (see Cardiovascular Research 1997;35:2–3). Human cells were obtained from discarded umbilical and saphenous veins and treated anonymously. As such, approval from the University Ethics Review Board was not necessary. When not otherwise specified, experiments were performed in HUVECs, but control experiments were also performed in HSVECs and BAECs, with identical results. Simvastatin and atorvastatin (0.001–10 µmol/L) were added to the medium of confluent endothelial monolayers in the presence of 2.5% fetal bovine serum 12 h before stimulation with phorbol myristate acetate (PMA, 10 nmol/L) or tumour necrosis factor-α (TNF-α, 10 ng/mL). In additional experiments, mevalonate, Fpp, GGpp, and cholesterol were added to the medium containing simvastatin or atorvastatin to address the statin site of action along the mevalonate pathway. To confirm the involvement of specific isoprenoids in the mevalonate pathway, FPT inhibitor I GGTI-286 and the Rho inhibitor C3 transferase were added to the medium instead of statins.

Cellular toxicity by statin treatments was monitored through a variety of techniques including cell counting and Trypan Blue exclusion,17 3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide (MTT) assay, and the assessment of total cellular protein content (amido-black assay).17 In preliminary experiments aimed at evaluating drug toxicity, treatment of HUVECs up to 10 µmol/L simvastatin or atorvastatin for 12 h did not produce any sign of toxicity (data not shown). Statins that were not chemically activated had no effect on PMA-mediated COX-2 expression (data not shown).

2.2 Western blotting

After drug treatment and stimulation, endothelial monolayers were washed twice with cold phosphate-buffered saline, lysed, and processed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, as described previously.18 In some experiments, blots were stripped and re-probed with monoclonal antibodies against COX-1 and/or against endothelial nitric oxide synthase (eNOS). Blots were scanned with an Agfa Arcus II scanner (Mortsel, Belgium), and band densities quantified with the aid of Scion Image 1.4 software (Scion Corporation, Frederick, MD, USA).

2.3 Measurement of 6-keto-PGF

After drug treatment and stimulation, aliquots of culture medium were collected, and the concentration of 6-keto-PGF (the stable product of the non-enzymatic catabolism of PGI2) determined by radioimmunoassay.19

2.4 Cellular expression of COX-2

To measure the cellular expression of COX-2, cells were grown in 96-well plates until they reached about 90% confluence, exposed to simvastatin (with or without isoprenes) for 12 h or to l-NAME for 1 h and stimulated with PMA before processing by a modified enzyme immunoassay (EIA) procedure.18

2.5 Immunocytochemistry

Cells were grown directly on gelatin-covered cover slips (Thermanox, ProSciTech, Thuringowa, Queensland, Australia). After drug treatment and stimulation, monolayers were fixed and processed as described previously.18

2.6 Isolation of RNA and northern analysis

Northern blots were performed as published previously.20

2.7 Gelatin zymography

Gelatin zymography was carried out as described by Bellosta et al.21

2.8 Measurement of MMP-9 release

MMP-9 released in the supernatant was quantified using an EIA, using a commercial kit by Oncogene Research Products (now a Calbiochem brand), and using a sheep polyclonal antibody raised against the human zymogen form of MMP-9 protein. The assay was performed according to the manufacturer's instructions.

2.9 Rho kinase assay

Rho kinase activity was measured using a commercial kit by Cyclex MBL International Corporation (Woburn, MA, USA), following the manufacturer's instructions.

2.10 In vitro angiogenesis assay

The formation of vascular-like structures by HUVECs was assessed on the growth-factor-reduced basement membrane matrix ‘Matrigel’ (11.1 mg/mL, Becton Dickinson Biosciences, Bedford, MA, USA). HUVECs were grown in culture flasks until they reached about 80% confluence and were then treated with simvastatin, with or without isoprene intermediates, at the concentrations indicated. After 12 h, HUVECs were gently detached with 0.25% trypsin/EDTA, and the trypsinized cells collected, counted, and resuspended in low serum (2.5%) M199 medium. Meanwhile, the bottoms of 24-well culture plates were coated with Matrigel (50 µL per well), diluted 1:2 with M199 medium. After gelation at 37°C for 30 min, gels were overlaid with 500 µL of low-serum M199 medium containing 4 × 104 cells per well. Media were supplemented with PMA in the presence or absence of simvastatin, isoprenoids, or NS-398 and incubated at 37°C for 18 h. Tube formation was monitored by an inverted phase contrast microscope (Leica, Wetzlar, Germany) and picture taken by an attached digital output Canon Powershot S50 camera. Several tube formation indices were measured after image conversion into pseudocolours to best point out the confluent areas (shown in white in the derivatized picture), the nodal structure (shown in blue), the branching point (shown in green), and the tubes (shown in yellow, orange, or red depending upon maturation degree) by using the web-based Image analysis Core software.22

2.11 Statistical analysis

Values were expressed as mean ± SD. Differences between two groups were determined by the unpaired Student's t-test. Multiple comparisons were performed by one-way analysis of variance (ANOVA), and individual differences were then tested by the Fisher's protected least significant difference test after the demonstration of significant differences between groups by ANOVA.

3. Results

3.1 Statins inhibit COX-2 expression induced by mitogenic and inflammatory stimuli

As shown in Figure 1A and C, HUVEC treatment with simvastatin reduced the PMA-stimulated expression of COX-2 in a concentration-dependent manner, with a maximum effect at 10 µmol/L (−70 ± 10%; P < 0.01), without significantly affecting the expression of COX-1 (Figure 1B). Simvastatin effect was time-dependent, since a much lower effect was obtained when the drug was co-administered with the stimulus (see Supplementary material online, Figure S1). The 50% inhibitory concentration (IC50) for simvastatin was around 1 µmol/L (Figure 1D). Since 1 µmol/L simvastatin closely approaches concentrations achievable in vivo,23 we used this concentration in most subsequent experiments. Similar results were obtained with atorvastatin (see Supplementary material online, Figure S2, upper panel) and using TNF-α as inducer (Figure 2A, upper panel). Furthermore, under the same experimental conditions, comparable results were observed using HSVECs as alternative cellular model (data not shown).

Figure 1

Simvastatin reduces PMA-stimulated COX-2 expression in a concentration-dependent fashion. Simvastatin (SIM) was incubated with confluent HUVECs for 12 h, after which PMA was added for an additional 12 h. Parallel monolayers were used to perform Western analysis for COX-2 (A) and COX-1 (B) and immunostaining for COX-2 (C). (D) Values of COX-2 and COX-1 expressions at Western analysis were obtained by band scanning and are shown as fold induction over control. Three independent experiments were performed. *P < 0.05 and **P < 0.01 between groups joined by the horizontal lines (ANOVA with post hoc testing).

Figure 2

Statins reduce COX-2 protein expression and activity induced by TNF-α. (A) Simvastatin (SIM) and atorvastatin (ATV) were incubated with confluent HUVECs for 12 h, after which TNF-α was added for an additional 12 h. Monolayers were used to perform Western analysis, and supernatants were collected to assess the production of 6-keto-PGF in the medium. A significant reduction in COX-2 protein expression (upper panel) and enzyme activity (lower panel) is evident upon simvastatin treatment. Three independent experiments were performed. *P < 0.05 and **P < 0.01 between groups joined by the horizontal lines (ANOVA with post hoc testing). (B) In some experiments, blots, as in (A), were stripped and re-probed with an anti-eNOS antibody. Three independent experiments were performed. *P < 0.05 between groups joined by the horizontal lines (ANOVA with post hoc testing).

The inhibition of PMA- and TNF-α-mediated COX-2 expression by simvastatin and atorvastatin was accompanied by a parallel reduction in COX enzyme activity, as determined by measuring the accumulation of 6-keto-PGF in the culture medium (see Supplementary material online, Figure S2 and Figure 2A, lower panels). In all cases, and contrary to clinical situations, simvastatin was more potent than atorvastatin in the presence of both PMA and TNF-α (see Supplementary material online, Figure S2 and Figure 2) and was therefore used preferentially in the subsequent experiments. Similar results were obtained also using Northern analysis, thus indicating a pre-translational effect by statins in the regulation of COX-2 expression (see Supplementary material online, Figure S3).

As an independent proof of statin efficacy in this system, under the same experimental conditions associated with COX-2 inhibition, we observed that both simvastatin and atorvastatin determined a clear increase in basal eNOS expression and reverted the decrease of eNOS induced by TNF-α, in agreement with data previously reported by others.24,25 (Figure 2B). On this background, to exclude that simvastatin effect on COX-2 expression was mediated by the increased NO production (through the upregulation of eNOS), we exposed simvastatin-treated cells to the eNOS inhibitor Nω-nitro-l-arginine methyl ester (l-NAME). As shown in Figure 3B (column 10 vs. 3), cell exposure to l-NAME did not prevent the simvastatin effect on COX-2 expression, thus excluding NO as a mediator of the statin effect in this system.

Figure 3

Effect of isoprenoid intermediates of the mevalonate pathway on COX-2 expression. (A) Confluent HUVECs were treated with simvastatin (SIM) in the presence or absence of mevalonate, GGpp, and Fpp for 12 h. Monolayers were then stimulated with PMA for a further 12 h, after which cell lysates were processed by Western analysis, with immunoblots probed first for COX-2 and then, after stripping, for COX-1 and/or β-actin. Images reported are representative of three separate experiments. Three independent experiments were performed. *P < 0.05 and **P < 0.01 between groups joined by the horizontal lines (ANOVA with post hoc testing). (B) Confluent HUVECs were treated as in (A), but also in the presence or absence of cholesterol, l-NAME, and C3 transferase. COX-2 expression was assessed by EIA. Original readings were derived from three independent experiments, each consisting of six replicates for each condition. **P < 0.05 between groups joined by the horizontal lines (ANOVA with post hoc testing).

3.2 Mevalonate and GGpp, but not Fpp, overcome the inhibition of COX-2 protein expression by simvastatin

As shown in Figure 3A and B, the reduction of the stimulated COX-2 expression by simvastatin was reverted by mevalonate, but not by cholesterol, thus preliminarily suggesting that isoprenoid intermediates, and not cholesterol, are involved in the regulation of COX-2 expression by statins. Consistently, we observed that GGpp, but not Fpp, was able to overcome the effect of simvastatin, suggesting that the inhibition of geranylgeranylation, but not that of farnesylation, is involved in this effect of simvastatin. To confirm this, we tested whether GGTI-286, an inhibitor of geranylgeranyltransferase I or the FPT inhibitor I, an inhibitor of farnesyltransferase, mimicked simvastatin effect. We observed that although the addition of FPT inhibitor I had negligible effects, the addition of GGTI-286 reduced, in a concentration-dependent manner, the stimulated expression of COX-2 with potency comparable with simvastatin for GGTI-286 concentrations approaching to 10 µmol/L (Figure 4).

Figure 4

Effect of prenyltransferase inhibitors FTI-277 or GGTI-286 on PMA-dependent induction of COX-2. HUVECs were treated with simvastatin (SIM) or only with inhibitors of prenyltransferases and then stimulated with PMA for further 12 h. The levels of COX-2 were then assessed by Western analysis. Three independent experiments were performed. *P < 0.05 and **P < 0.01 between groups joined by the horizontal lines (ANOVA with post hoc testing).

3.3 Ras-like proteins as possible targets of simvastatin effect

Isoprenoid intermediates serve as important lipid attachments for the post-translational modification and plasma membrane translocation of a variety of proteins, including the small GTP-binding protein Ras and the Ras-like proteins Rho, Rab, Rac, Ral, and Rap.25 Since the small GTP-binding protein Rho is involved in the mitogenic signalling pathway leading to COX-2 expression,13 we hypothesized that Rho might be the specific target of simvastatin activity. To address this hypothesis, we treated HUVECs with the cell-permeable Rho inhibitor C3 transferase, a drug that selectively ADP-ribosylates—thus inactivating—Rho A/B/C proteins. As expected, the Rho inhibitor C3 transferase reproduced simvastatin effect, strongly reducing PMA-stimulated COX-2 expression (Figure 3B, bar 8 vs. bar 2). In agreement with the known mechanism by which C3 transferase inactivates Rho, the addition of GGpp to C3 transferase did not revert the inhibition of COX-2 expression (Figure 3B, bar 9 vs. bar 8). We then directly assessed the effect of simvastatin on Rho kinase activity. As shown in the Supplementary material online, Figure S4, the same simvastatin concentrations able to reduce PMA-stimulated COX-2 expression significantly reduced Rho activity (by 39 ± 15%, P < 0.01).

3.4 Simvastatin reduces endothelial angiogenic activity

We next checked the functional consequence of the COX-2 downregulation by simvastatin in terms of inhibition of endothelial angiogenic activity, performing in vitro Matrigel assays. As shown in Figure 5 and in the related Table 1, cell treatment with 1 µmol/L simvastatin negatively affected several tube formation indices, among which the total length skeleton and the number of branching points. Here, again, mevalonate and GGpp, but not Fpp, overcame simvastatin effect. Under the same experimental conditions, NS-398, a potent and selective inhibitor of COX-2 activity,26 also reduced, although to a lesser extent, both the total length skeleton and the number of branching points, confirming the role of COX-2 activity in angiogenesis.

Figure 5

Effect of simvastatin and NS-398 on angiogenic activity of endothelial cells. HUVECs were treated with simvastatin in the presence or absence of mevalonate, GGpp, and Fpp for 12 h or with NS-398 for 30 min and, after enzymatic detachment, plated on Matrigel-coated 24-well plates and stimulated with PMA. After incubation at 37°C for 24 h, the images were taken, transformed, and analysed as indicated in section 2. Legend: 1, PMA 10 nmol/L; 2, SIM 1 µmol/L + PMA 10 nmol/L; 3, SIM 1 µmol/L + mevalonate 200 µmol/L + PMA 10 nmol/L; 4, SIM 1 µmol/L + Fpp 5 µmol/L + PMA 10 nmol/L; 5, SIM 1 µmol/L + GGpp 5 µmol/L + PMA 10 nmol/L; 6, NS-398 5 ng/mL + PMA 10 nmol/L.

View this table:
Table 1

Effect of simvastatin treatment on several indices of tube-like structures formation in Matrigel

PMA 10 nmol/LSIM 1 µmol/L + PMA 10 nmol/LSIM 1 µmol/L + mevalonate 200 µmol/L + PMA 10 nmol/LSIM 1 µmol/L +Fpp 5 µmol/L + PMA 10 nmol/LSIM 1 µmol/L +GGpp 5 µmol/L + PMA 10 nmol/LNS-398 5 ng/mL + PMA 10 nmol/L
Mean single tube index0.080 ± 0.0050.020 ± 0.004*0.075 ± 0.006#0.030 ± 0.002*0.05 ± 0.03#0.035 ± 0.003*
Tube formation index4.80 ± 0.123.03 ± 0.25*4.31 ± 0.23#3.37 ± 0.13*5.12 ± 1.13#3.20 ± 0.65*
Total length skeleton (green) (px)39 444 ± 27830 752 ± 850*41 404 ± 375#31 438 ± 567*43 930 ± 474#32 906 ± 453*
No. of branching points (green)134 ± 880 ± 9**155 ± 7##83 ± 6*149 ± 12##102 ± 8
No. of tubes (yellow, orange, red)95 ± 5102 ± 1098 ± 7113 ± 10106 ± 2398 ± 9
No. of nodal structures (blue)35 ± 338 ± 534 ± 534 ± 434 ± 536 ± 5
Mean tube length (px)78 ± 669 ± 780 ± 667 ± 770 ± 969 ± 8
Total length satisfactory tubes (orange) (px)4444 ± 3001439 ± 33**3821 ± 80##2296 ± 97*2847 ± 80#2087 ± 127**
Total length poorly developed tubes (red) (px)2942 ± 1785265 ± 365**3968 ± 150#5121 ± 121*4229 ± 177#4687 ± 233**
  • To best interpret the data, see pseudocolour panels in Figure 5. The data were obtained from three separate experiments and were reported as mean ± SEM.

  • px, pixel; SIM, simvastatin; PMA, phorbol myristate acetate; Fpp, farnesylpyrophosphate; GGpp, geranylgeranyl-pyrophosphate.

  • Mean single tube index = sum of all single tube indices/no. of tubes.

  • Single tube index = linearity factor × tube length–width ratio.

  • Tube formation index = (mean single tube index)2 × (1−confluent area) × (no. of branching points/total length skeleton).

  • *P < 0.05 SIM vs. PMA.

  • **P < 0.01 SIM vs. PMA.

  • #P < 0.05 isoprenoid vs. SIM.

  • ##P < 0.01 isoprenoid vs. SIM.

Since the degradation of basement membrane by secreted proteolytic enzymes, such as MMPs, is a crucial step in angiogenesis,27 and since COX-2 activity has been involved in the regulation of MMP expression,11,12 we investigated the effects of simvastatin and the contribution of COX-2 activity in the regulation of MMP expression and secretion by HUVECs and BAECs. As assessed by gelatine zymography, simvastatin treatment of HUVECs before PMA stimulation selectively reduced the gelatinolytic activity corresponding to MMP-9 (by 40 ± 10% with 1 µmol/L simvastatin; P < 0.01) without significantly affecting the other induced gelatinolytic bands (Figure 6A). Under the same experimental conditions, there was a corresponding reduction also in the release of MMP-9 zymogen protein as assessed by EIA (Figure 6B, by 39 ± 5% with 10 µmol/L simvastatin; P < 0.01). As for COX-2, the addition of mevalonate and GGpp, but not of Fpp, overcame the inhibitory effect of simvastatin, strongly suggesting a similar mechanism in the regulation of gene expression of COX-2 and MMP-9. Similar results were obtained with BAECs (data not shown).

Figure 6

Simvastatin reduces the PMA-stimulated release of MMP-9 at gelatin zymography and EIA. Confluent HUVECs were treated with simvastatin (SIM) in the presence or absence of mevalonate, GGpp, or Fpp (12 h before PMA stimulation) or only with NS-398 (30 min before PMA stimulation). After PMA stimulation, media were collected and analysed by gelatin zymography (A) or by enzyme-linked immunoassay (B). The image in (A) is representative of three separate experiments yielding similar results. In (B), the original readings were derived from three independent experiments each consisting of four replicates for each condition. *P < 0.05 and **P < 0.01 between groups joined by the horizontal lines (ANOVA with post hoc testing). Legend for (B): 1, basal; 2, PMA 10 nmol/L; 3, SIM 1 µmol/L + PMA 10 nmol/L; 4, SIM 1 µmol/L + PMA 10 nmol/L + mevalonate 200 µmol/L; 5, SIM 1 µmol/L + PMA 10 nmol/L + GGpp 5 µmol/L; 6, SIM 1 µmol/L + PMA 10 nmol/L + Fpp 5 µmol/L; 7, PMA 10 nmol/L + NS-398 1 mg/mL.

We next examined the effect of NS-398 on the PMA-stimulated release of MMPs at both zymography and EIA. We observed that the addition of NS-398 to the culture medium 30 min before PMA stimulation selectively reduced, in a concentration-dependent fashion, the stimulated production of MMP-9 (Figure 6B and Supplementary material online, Figure S5, lanes 5, 6, 7, 8 vs. 3). NS-398 exerted its maximum effect at 1 and 5 µg/mL, concentrations that selectively inhibit COX-2 activity.

To rule out the possibility that the observed inhibitory effect by simvastatin and NS-398 on MMP-9 release was due to an interference with the activation—rather than with the extracellular release—of MMP-9, HUVECs were stimulated with PMA for 24 h, after which time incubation media were collected and divided into aliquots, to which increasing concentrations of simvastatin (0.1–10 µmol/L) or NS-398 (0.1–10 µg/mL) were added. Media were then incubated for 24 h at 37°C in the absence of cells and analysed by zymography. Neither simvastatin nor NS-398 had any appreciable effect on MMP activity (data not shown), thus excluding the possibility that simvastatin and NS-398 interfere with the extracellular activation of these proteases.

Finally, we failed to show an overcoming of the effect of NS-398 by the addition of PGE2 and carbaprostacyclin (a stable analogue of PGI2), suggesting the involvement of other COX metabolites in the recognized functional link between COX-2 activity and MMP-9 production in endothelial cells.

4. Discussion

A discrete body of evidence supports the clinical relevance of statin benefits in cardiovascular disease not mediated by the well-known cholesterol-lowering properties of these drugs. Such benefits are related to the ability of statins to inhibit the isoprenylation of several signalling proteins, in turn influencing the expression of a variety of pro-inflammatory, pro-atherogenic genes.4

We here report that treatment of human endothelial cells with two lipophilic but structurally different statins, simvastatin and atorvastatin, under the same experimental conditions associated with a clear upregulation of eNOS expression, inhibits the endothelial pro-angiogenic potential and the stimulated expression of COX-2 and MMP-9. To induce the expression of COX-2 and MMP-9 as well as the angiogenic activity, we used two well-known pro-angiogenic and pro-inflammatory stimuli: the phorbol ester PMA, the relevance of which has been recently revalidated by Xu et al.28 in angiogenesis, and the cytokine TNF-α, the pro-inflammatory and pro-angiogenic activity of which is also well appreciated in cardiovascular disease, as recently reviewed by Kofler et al.29 We here observed that statin treatment significantly reduces the expression of COX-2 induced by both stimuli, as well as the angiogenic activity induced by PMA. Importantly, however, we also demonstrate that the specific inhibitor of COX-2 activity NS-398, similar to statins, reduces both the angiogenic activity and the release of MMP-9, thus allowing us to propose for the first time the existence of a causal link between COX-2 activity and MMP-9 release in the human vascular endothelium.

We demonstrate the importance of HMG-CoA reductase inhibition in the observed effects by showing the overcoming of the effects of statins on COX-2 and MMP-9 by the immediate product of HMG-CoA reductase, mevalonate, but not by cholesterol, the final product of the mevalonate pathway. This proves that such effects can be attributed to the ability of statins to induce a depletion in cell isoprenoids rather than of cell cholesterol. To better understand the mechanisms involved, we further investigated the metabolites downstream of the mevalonate pathway co-treating endothelial cells with statins in the presence of the intermediate isoprenoids GGpp and Fpp. Although the addition of GGpp completely reverted statin effects, Fpp was devoid of any activity. This indicates that statins interfere with COX-2 and MMP-9 by negatively interfering with the geranylgeranylation of some cellular proteins, with candidates being the Rho subfamily of monomeric GTPases (Rho A, Rho B, Rho C, Rac1, Rac2, and Cdc42), recently shown to be involved in the regulation of both COX-215 and MMP-930 gene expression. Consistently, the inhibitor of geranylgeranyltransferases, but not that of farnesyltransferases, mimicked the effect of statins on COX-2 expression. We further investigated the specific contribution of activated Rho proteins in the signalling events leading to COX-2 using a selective inhibitor of geranylgeranylated Rho A/B/C. As expected, the addition of the Rho inhibitor C3 transferase closely mimicked statin effect on COX-2, thus proving the involvement of Rho in the endothelial induction of COX-2 by the pro-mitogen PMA, as previously recognized upon pro-inflammatory stimulation with LPS and TNF-α,13 thus allowing us to point out the exact site of statin interference. Although COX-2 expression was reported also to depend on farnesylated proteins such as Ha-Ras or Ki-Ras,31 inhibition of farnesylation does not seem to be relevant to explain statin effects in our experimental conditions.

By providing new insights into the molecular mechanisms by which statins downregulate mitogen-mediated COX-2 expression, our results expand those by Inoue et al.,32 also obtained in endothelial cells, as well as those by Wu et al.33 and Habib et al.34 in different cellular models (epithelial and monocytoid cells, respectively), in which atorvastatin and simvastatin similarly reduced LPS-stimulated expression of COX-2. Our results are apparently in contrast with the data reported by Smith et al.,35 showing an increase in COX-2 mRNA upon long-term (48 h) treatment with lovastatin. We believe that such discrepancy is only apparent, since we did not investigate the effect of long-term statin treatment under basal condition, but instead focused on the effect of statin treatment on the (TNF- and PMA-) stimulated expression of COX-2.

The role of endothelial COX-2 activity in inflammation and atherogenesis has been long discussed and still remains a matter of debate.36 In apolipoprotein E- and in low density lipoprotein receptor-deficient mice, as well as in C57BL/6 mice, it has been clearly shown that COX-2 activity promotes atherosclerosis despite the selective suppression of prostacyclin production.37 Such experimental results are supported by those obtained in high-risk patients in several studies,38,39 proposing the benefit of some curtailing of COX-2 activity when patients are studied on top of treatment with low-dose aspirin. In contrast, several independent studies have demonstrated the early pathogenetic role of angiogenesis in the development and instabilization of atherosclerotic plaques (reviewed in Doyle and Caplice7), as well as the role of COX-2 activity in orchestrating angiogenesis.10 Against this background, the observation that COX-2 co-localizes with MMP-9 and membrane type-1 MMP in the endothelial lining of vasa vasorum of human atherosclerotic aortas suggests the possibility of a pathogenetic role for the endothelial COX-2-MMP axis in atherosclerosis.40 In such a context, statin treatment, reducing (but not totally suppressing) endothelial COX-2 activity and decreasing the related MMP-9 secretion, might curb early plaque neoangiogenesis, while not completely blocking the production of prostacyclin. Consistent with our findings, statin treatment has been found to interfere with angiogenesis in vitro8 and, more importantly, in in vivo models of atherosclerosis.41,42

Alternative hypotheses have been proposed to explain the anti-angiogenic properties of statins, including the increased concentration of the cell cycle inhibitors p19, p21, and Wnt5a, as well as the reduction of angiogenesis-related genes such as PAI-1, vitronectin, HoxD3, and Notch4 (reviewed in Ii and Losordo8). Although our findings are not excluding the coexistence of other mechanisms, the causal link provided by us between statin effects on COX-2 inhibition and the formation of tube-like structures in Matrigel provides a coherent explanation of a variety of previous observations.

Although both simvastatin and atorvastatin significantly reduced the stimulated expression of COX-2 (thus suggesting a class rather than drug-specific effect), the inhibitory efficacy was here greater with simvastatin than with atorvastatin. This finding might appear surprising based on the higher IC50 of simvastatin than of atorvastatin for the purified human HMG-CoA reductase (11.2 vs. 8.2 nmol/L, respectively).43 Here, however, solubility issues and ancillary properties may play a role in this hierarchy of potency. Simvastatin has indeed shown stronger antioxidant properties than atorvastatin,44 which might turn out to be relevant in the light of redox sensitivity of COX-2.18

In conclusion, we demonstrate that statins reduce COX-2 protein expression and activity, as well as MMP-9 release in human endothelial cells. We also demonstrate the existence of a functional link between COX-2 activity and MMP-9 release and point out to the inhibition of Rho geranylgeranylation as the crucial target for statin action for such phenomena. Our results therefore offer a plausible explanation for a beneficial role for statins as anti-inflammatory/anti-angiogenic agents potentially preventing crucial steps in plaque instabilization.


This study was partially funded through grants from the Italian Ministry of the University and of the Istituto Italiano Ricerche Cardiovascolari (to R.D.C.).


The authors express gratitude to the personnel in the Division of Obstetrics and Gynecology at the Vito Fazzi Hospital in Lecce for providing umbilical cords.

Conflict of interest: none declared.


View Abstract