Growth differentiation factor 15 (GDF15) is an independent predictor of cardiovascular disease, and increased GDF15 levels have been associated with endothelial dysfunction in selected patients. We therefore investigated whether GDF15 modulates endothelial function in aortas of wild-type (WT) and GDF15 knockout (KO) mice. Vascular contractions to phenylephrine and relaxation to ACh were assessed in aortas obtained from healthy WT and GDF15 KO mice. The effects of GDF15 pretreatment and the involvement of ROS or caveolae were determined. Phenylephrine-induced contractions and ACh-mediated relaxations were similar in WT and GDF15 KO mice. Pretreatment with GDF15 inhibited contraction and relaxation in both groups. Inhibition of contraction by GDF15 was absent in denuded vessels or after blockade of nitric oxide (NO) synthase. Relaxation in WT mice was mediated mainly through NO and an unidentified endothelium-derived hyperpolarizin factor (EDHF), whereas GDF15 KO mice mainly used prostaglandins and EDHF. Pretreatment with GDF15 impaired relaxation in WT mice by decreasing NO; in GDF15 KO mice, this was mediated by decreased action of prostaglandins. Disruption of caveolae resulted in a similar inhibition of vascular responses as GDF15. ROS inhibition did not affect vascular function. In cultured endothelial cells, GDF15 pretreatment caused a dissociation between caveolin-1 and endothelial NO synthase. In conclusion, GDF15 impairs aortic contractile and relaxing function through an endothelium-dependent mechanism involving altered caveolar endothelial NO synthase signaling.
- growth differentiation factor 15
growth differentiation factor 15 (GDF15), also called macrophage inhibitory cytokine (MIC)-1, is a distant member of the transforming growth factor (TGF)-β family and the result of posttranslational modification of the product of an early response gene (4). Upregulation of GDF15 has been demonstrated after chemical and hyperoxic injury of the lung, by surgical, chemical, and heat-induced injury of the liver (19), and after ischemia-reperfusion injury of the heart (15), indicating that GDF15 is a general mediator of the organ injury response. Furthermore, GDF15 is emerging as an independent predictor of cardiovascular disease (CVD) in the elderly and a predictor of prognosis in patients with established CVD (17, 20, 21). An association between GDF15 levels and endothelial dysfunction was found in an elderly population (21). Hence, endothelial dysfunction is considered the first step in the chain of events leading to atherosclerosis and CVD. It comprises an imbalanced production of vasodilating and vasoconstricting substances by the endothelium and is associated with increased adhesion molecule expression and reduced anticoagulant properties. Results from a recent study (16) have demonstrated that GDF15 reduces leukocyte recruitment and inflammation by inhibiting endothelial integrin activation in an experimental model of myocardial infarction using GDF15-deficient mice and recombinant GDF15. Thus, current evidence suggests that GDF15 may play an important role in the development of endothelial dysfunction in CVD.
To further explore this hypothesis, we investigated vascular responsiveness in the aorta of wild-type (WT) and GDF15 knockout (KO) mice and the acute modulatory effect of recombinant GDF15 in these animals. Contractile responses to vasoconstrictor agents were determined in both intact and endothelium-denuded aortic rings to assess the presence of basally released endothelium-derived relaxing factors (EDRFs). Furthermore, endothelium-dependent relaxation to ACh was assessed in preconstricted rings. The contribution of the different EDRFs, i.e., nitric oxide (NO), prostaglandins (PGs), and endothelium-derived hyperpolarizing factor (EDHF), to vasomotor responses was examined using inhibitors of NO and PG synthesis. Since endothelial dysfunction in CVD is closely related to increased oxidative stress and inactivation of endothelial NO in these conditions (1, 2, 24), the effect of scavengers of ROS was also studied.
Furthermore, endothelial NO synthase (eNOS) associates with caveolin (CAV)-1, the principal structural component of caveolae, which function as signal-transducing organelles (for a recent review, see Ref. 22). When the endothelium is not stimulated, this association of CAV-1 with eNOS limits the activity of eNOS, resulting in a low basal release of NO. In contrast, CAV-1/eNOS association facilitates the stimulated release of NO (by ACh) because of the favorable spatial organization of signaling molecules (e.g., the ACh receptor, Ca2+ channels/stores, and eNOS) in close proximity. Recent findings from experimental studies have suggested that a reduction in the number of caveolae and altered caveolar signaling may underlie the impairment of endothelial NO release in CVD (for a review, see Ref. 33). Therefore, we also studied the effect of GDF15 on eNOS-CAV-1 interactions in cultured human umbilical vein endothelial cells (HUVECs). Finally, as GDF15 levels are low in healthy mice, we explored the effects of genetic deletion of GDF15 on vascular function in a model of type 2 diabetes mellitus (T2DM), a condition associated with increased GDF15 levels (12).
The protocols for animal care and use were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Committee for Animal Experiments [Dierexperimentencommissie (DEC)] of the University of Groningen (permit nos. DEC 5391 and DEC5757). Basic experiments were performed using seven male mice (12–14 wk old, 25.1–29.4 g, C57BL/6, Harlan) and seven male GDF15 KO mice (12–14 wk old, 22.6–25.3 g, from internal breeding; 5 breeding pairs of GDF15 KO mice on a C57BL/6 background were obtained from Johns Hopkins University, Baltimore, MD) housed under standard conditions with free access to food and drinking water throughout the study. The effects of 4-hydroxy-2,2,6,6-tetramethylpiperidin-1-oxyl (tempol), catalase, and the caveolae-disrupting agent methyl-β-cyclodextrin (MCD) were studied in aortic rings from an additional group of 19 male mice (10–14 wk old, 24–31 g, C57BL/6, Harlan). Mice were anesthetized with isoflurane (2% in oxygen), and the thoracic aorta was removed and temporarily stored in cold physiological saline solution.
Organ bath experiments with the isolated aorta.
Freshly isolated thoracic aortic rings (1.5–2 mm in length) were mounted on 200-μm stainless wires in individual myograph baths (Danish Myo Technology, Aarhus, Denmark) as previously described (26). Briefly, baths containing 6 ml Krebs solution were warmed to 37°C and preequilibrated and continuously aerated with 95% O2-5% CO2 to maintain pH at 7.4. The length of the aortic strips was assessed by microscopy. Aortic rings were equilibrated for 40 min until they were at a steady baseline. Rings were then primed and checked for viability by two consecutive stimulations with KCl (60 mM) followed by washings and renewed stabilization to obtain reproducible contractile responses.
Vascular experiments are shown in Table 1. Contraction responses were measured as cumulative concentration-response curves to phenylephrine (PE; 10 nM–100 μM) followed by a single concentration of KCl (90 mM). Endothelium-dependent relaxation was assessed by obtaining concentration-response curves to ACh (10 nM–300 μM) in rings precontracted with PE (1 μM) followed by stimulation with a high concentration of the NO donor sodium nitroprusside (SNP; 0.1 mM) to assess maximal endothelium-independent dilation. To study the role of the endothelium in vasoconstrictor effects, rings were denuded by removing the endothelial cell layer by rubbing the luminal side of the vessel with a moistened wire. To examine the contribution of different EDRFs in modulating vasoconstrictor responses and mediating endothelium-dependent relaxation, inhibitors of PG and NO synthesis were used. To this end, PG components were assessed by preincubating rings (20 min) with the nonspecific cyclooxygenase inhibitor indomethacin (10 μM) (33). The NO component was examined by subsequent incubation with both NO synthesis inhibitor NG-monomethyl-l-arginine (l-NMMA; 1 μM) and indomethacin. The remaining ACh-mediated relaxation was attributed to an unidentified EDHF (18).
Acute treatment effects of GDF15.
To study the acute treatment effects of GDF15 on vasoconstrictor and endothelium-dependent relaxation responses, rings were preincubated for 20 min either with vehicle (control) or GDF15 (50 ng/ml) before the start of the vascular protocols. To investigate whether GDF15 exerts its effects by acting on receptors of the TGF-β superfamily, additional comparisons were made with rings preincubated with TGF-β1 (3 ng/ml). To study whether the effect of GDF15 on endothelium-dependent relaxation was via modulation of ROS and/or caveolae function, aortic rings were preincubated for 40 min with the membrane-permeable radical scavengers tempol (100 μmol/l) and catalase (500 U/ml) (29) or with the caveolae-disrupting agent MCD (1 mM) (33). ACh concentration-response curves were obtained as described above.
Coimmunoprecipitation of eNOS and CAV-1 in HUVECs.
HUVECs were obtained from the Endothelial Cell Facility of the University Medical Center Groningen/University of Groningen and were grown to confluence in six-well plates and pretreated with GDF15 (50 ng/ml) or vehicle for 30 min before being washed once with 0.5 ml PBS. After trypsinization, 2 ml full Claycomb medium (Invitrogen) was added, and cells were transferred to a 14-ml Falcon tube. Cells were then centrifuged for 5 min at 1,000 rpm, and the pellet was resuspended in 1 ml ice-cold PBS. Cells were then centrifuged (3 min at 6,000 rpm at 4°C) and resuspended [in cold PBS containing 1 mM dithiobis(succinimidyl propionate)] a second time before being incubated for 30 min on ice. Subsequently, 10 μl of 100 mM glycine were added to the cells and incubated for 15 min on ice followed by a wash with ice-cold PBS (3 min, 6,000 rpm, 4°C). Cells were lysed in immunoprecipitation buffer [20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 10 mM NaF, 2 mM Na3VO4, 2 mM PMSF, and 1% Triton X-100] supplemented with a mixture of protease inhibitors (Complete, Roche Diagnostics, Almere, Netherlands). Precleared lysates were subjected to immunoprecipitation with a polyclonal rabbit CAV-1 antibody (Cell Signaling). Precipitated proteins were eluted from the beads by boiling in SDS sample buffer and then separated by SDS-PAGE. Immunoblot assays were performed using a monoclonal eNOS antibody (Transduction Laboratories, Breda, Netherlands). In a reversed experiment, immunoprecipitation was performed with the monoclonal eNOS antibody with subsequent detection with the CAV-1 antibody.
T2DM in mice.
As GDF15 levels are rather low in healthy mice, we also explored the effects of genetic deletion of GDF15 on vascular function in a model of T2DM, a condition associated with increased GDF15 levels (12). Five female db/+ mice (B6.Cg-Dock7m+/+ Lepr db/J, 7 wk of age), heterozygous for a leptin receptor mutation, were purchased from The Jackson Laboratories (Charles River, L'Arbresle, France). To evaluate the effect of genetic deletion of GDF15 in the db/db T2DM mouse model, homozygous GDF15 KO mice were interbred with heterozygous db/+ mice. The desired male genotypes, db/db GDF15+/+ (n = 10) and db/db GDF15−/− (n = 8) mice, were subsequently obtained by successive rounds of inbreeding. Nondiabetic db/+ GDF15−/− mice (n = 6) were used a controls. Genotyping for mouse leptin receptor mutation was performed as previously described (13).
Mice were housed under standard conditions with free access to food and drinking water throughout the study and followed for the development of diabetes and associated renovascular damage. To this end, mice were placed in metabolic cages for 24 h at ages of 6, 8, 10, 12, 14, 16, and 18 wk, and food and water intake were measured. Glucose was measured in urine and blood samples obtained from the tail vein or eye socket. Plasma glucose concentrations were determined by the Accu-chek glucose meter (Roche Diagnostics). At the age of 18 wk, mice were anesthetized with isoflurane (2%), and the aorta and kidney were collected. The aorta was studied as vascular ring preparations during organ bath experiments as described above, and kidney sections were stained for periodic acid-Schiff (PAS) for the quantification of tubular injury as previously described in detail elsewhere (27). In brief, tubular morphology was evaluated by the assessment of four markers of damage: tubular necrosis, loss of the brush border, denudation of the basement membrane, and intraluminal casts. Each parameter was graded on a scale from 0 to 3, according to the extent of the injury (where 0: <5%, 1: 5–25%, 2: 25–75%, and 3: >75%). In total, 30 tubules/kidney were analyzed, and the histological score was calculated. Thus, the total histological score ranged from 0 to 90.
Krebs solution was prepared freshly and was of the following composition (in mM): 120.4 NaCl, 5.9 KCl, 2.5 CaCl2, 1.2 MgSO4, 25.0 NaHCO3, 1.2 NaH2PO4, and 11.5 glucose; all components were obtained from Merck (Darmstadt, Germany). The stock solution (10 mM) for indomethacin was prepared in 96% ethanol. Tempol, freshly prepared catalase, and MCD were dissolved in saline solution. All other drugs were dissolved in deionized water. l-NMMA acetate salt was purchased from MP Biomedicals (Illkirch, France). Recombinant human TGF-β1 and GDF15/MIC-1 were obtained from Peprotech (London, UK) and dissolved in PBS. All other compounds were purchased from Sigma. The concentrations presented in the concentration-curve responses are expressed as final molar concentrations in the organ baths.
Calculations and statistical analysis.
Contractions to KCl and PE are given in milliNewtons. Relaxation responses to ACh and SNP are expressed as percentages of the preconstriction with PE. In addition to that, the area under each individual curve (AUC; in arbitrary units) was determined for ACh-induced relaxation (SigmaPlot version 10.0, Systat Software, San Jose, CA). The AUC was used to present total endothelium-dependent relaxation and for the subsequent analysis of differences in ACh-mediated relaxation with and without inhibitors present to estimate the contribution of the different EDRFs, i.e., PGs for the part sensitive to cycloxygenase inhibition with indomethacin, NO for the part sensitive to NOS inhibition with l-NMMA, and EDHF by means of exclusion of PG and NO (34). Data are presented as means ± SE, and n refers to the number of animals in each group. Statistical analysis was done with SPSS 16.0.2 for Windows (SPSS, Chicago, IL). Differences between full concentration-response curves were tested with repetitive ANOVA; differences between points were tested with one-way ANOVA. P values of <0.05 (two tailed) were considered as statistically significant.
Vascular function in WT and GDF15 KO mice.
To investigate the effects of GDF15 on vascular function, we first compared aortic contraction and relaxation in healthy WT and GDF15 KO mice. Animal characteristics are shown in Table 2. GDF15 KO mice had lower body weights, increased left and right kidney weights, and decreased liver weight compared with WT control mice.
The addition of PE and KCl induced concentration-dependent contraction in isolated aorta rings, which did not differ between both groups, i.e., pEC50 and maximal elastance for PE and maximal elastance for KCl did not differ between aortas of WT and GDF15 KO mice (Table 2). Similarly, total endothelium-dependent relaxation to ACh did not differ in aorta rings of WT and GDF15 KO mice, although the major contribution of individual EDRFs to ACh-mediated relaxation demonstrated a trend from NO and an unidentified EDHF in WT mice and from mainly PG in GDF15 KO mice.
Acute pretreatment with GDF15 inhibits vascular contraction via a mechanism that is endothelium dependent and involves the NO pathway.
Acute pretreatment with GDF15 significantly reduced PE-induced contraction in both WT and GDF15 KO animals (Fig. 1). This effect appeared to not be mediated by some general action of GDF15 on TGF-β receptors since the inhibitory effect was not observed after acute pretreatment with a high concentration of TGF-β1 in either in WT (Fig. 1A) or GDF15 KO (Fig. 1B) animals. Furthermore, the inhibitory effect of GDF15 seemed to involve a general depression of vascular contractile since (receptor independent) contraction to KCl in the WT aorta was also significantly reduced by GDF-15: contraction to 30, 60, and 90 mM KCl was 1.7 ± 0.2 versus 0.9 ± 0.1, 3.8 ± 0.3 versus 2.9 ± 0.3, and 4.6 ± 0.4 versus 3.9 ± 0.3 mN for control mice versus GDF15 KO mice, respectively (P < 0.05 for all).
To investigate the mechanism through which GDF15 inhibited vascular contractility, PE-induced contraction was also assessed in aortas of WT mice in endothelium-denuded rings or during pharmacological inhibition of different EDRFs. While in endothelium-intact rings the acute pretreatment with GDF15 again significantly reduced PE-mediated contraction (Fig. 2A), GDF15 was without effect on PE-mediated contraction in endothelium-denuded rings (Fig. 2A) or when l-NMMA was present to inhibit NO synthesis (Fig. 2B). In contrast, pretreatment with indomethacin to inhibit PG synthesis did not affect the inhibitory effect of GDF15 on PE-mediated contraction (data not shown). These findings together suggest that GDF15 opposed vascular contraction via stimulation of endothelium-derived NO production.
Acute pretreatment with GDF15 inhibits endothelium-dependent relaxation in WT and GDF15 KO mice.
Interestingly, in addition to inhibition of contractile responses, acute GDF15 pretreatment also decreased endothelium-dependent relaxation, as shown by the profound reduction in ACh-induced relaxation in both WT and GDF15 KO mice (Fig. 3). Here too the effect appeared independent of a general action of GDF15 on TGF-β receptors since the inhibitory effect was not observed after acute pretreatment with a high concentration of TGF-β1 in either WT (Fig. 3A) or GDF15 KO (Fig. 3B) animals. AUC analysis of the different EDRFs contributing to total ACh-induced relaxation further showed that GDF15 most notably reduced the NO component (Fig. 4). In addition, GDF15 appeared to limit PG contribution in GDF15 KO mice (although significance was not reached in this group due to the large variation), whereas the EDHF component remained relatively unaffected by GDF15 in both groups (Fig. 4). Moreover, the maximal endothelium-independent relaxation to SNP was similar in control and GDF15-pretreated aorta rings (98.1 ± 0.9 and 89.8 ± 4.5, respectively), suggesting that NO sensitivity of the vascular smooth muscle was unaffected by GDF15. Taken together, these data suggest that acute GDF15 pretreatment limits the endothelial synthesis of EDRFs, notably NO.
Effects of ROS inhibition and caveolae disruption on aortic responses.
It emerged from the above findings that GDF15 modulates vascular responses in an endothelium-dependent way, involving the NO pathway in particular. We therefore investigated the potential involvement of ROS and caveolae, two systems substantially affecting NO production. As in previous experiments, acute pretreatment with GDF15 inhibited contraction to PE in aorta rings of WT mice, resulting in a significantly reduced AUC (Fig. 5A, solid bars). Pretreatment with the ROS scavengers tempol and catalase did not affect PE-mediated contraction, nor did it attenuate GDF15-mediated inhibition of PE-induced contraction (Fig. 5A, open bars). In contrast, pretreatment with the caveolae-disrupting agent MCD resulted in significantly smaller contractions to PE and, importantly, precluded the GDF15-mediated reduction in PE-mediated contraction (Fig. 5A, shaded bars).
ROS inhibition and caveolae disruption similarly affected the action of GDF15 on ACh-mediated relaxation, i.e., GDF15 inhibited ACh-induced relaxation, resulting in a significantly decreased AUC (Fig. 5B, solid bars). Pretreatment with tempol and catalase did not affect ACh-mediated relaxation, nor the inhibitory effect of GDF15 on ACh-mediated relaxation (Fig. 5B, open bars). MCD pretreatment lowered ACh-mediated relaxation by 61%, although the variation was too large to obtain statistical significance (Fig. 5B, shaded bar). Importantly, however, GDF15 did not inhibit ACh-mediated relaxation in MCD-pretreated rings.
Thus, these data suggest that GDF15 modulates vascular function by inhibition of caveolar function in endothelial cells.
Effects of GDF15 pretreatment on the interaction between eNOS and CAV-1 in cultured HUVECs.
To further explore the effects of GDF15 on endothelial caveolar signaling, we investigated the interaction between eNOS and CAV-1. As the aortic segments contained too few endothelial cells for immunoprecipitation experiments, we conducted these experiment in cultured HUVECs. Total eNOS and CAV-1 levels in lysates of HUVECS were not affected by pretreatment with GDF15 (Fig. 6). However, pretreatment with GDF15 substantially lowered the association of eNOS and CAV-1, as demonstrated by the significant decrease in coimmunoprecipitation of eNOS and CAV-1 (Fig. 6).
Effect of genetic deletion of GDF15 on the development of endothelial dysfunction in T2DM.
As genetic deletion of GDF15 in healthy animals did not severely affect vascular function in healthy animals, we additionally studied GDF15 KO effects on vascular function in db/db mice, an established model of T2DM with endothelial dysfunction (7, 35). Although measurements of GDF15 levels are technically not feasible in murine blood samples, several human studies (6, 9, 12, 21) have shown increased GDF15 plasma levels in patients with T2DM. Therefore, to confirm the upregulation of GDF15 in the mouse model of T2DM, we measured GDF15 expression in white adipose tissue of db/+ and db/db mice by real-time PCR. GDF15 expression increased from 0.35 ± 0.35 arbitrary units in nondiabetic db/+ control mice to 1.3 ± 0.5 arbitrary units in diabetic db/db mice (P < 0.05).
Endothelial function was studied at the age of 18 wk. db/+ GDF15 KO mice had normal ACh-induced relaxation curves, and the AUC was similar to GDF15 KO mice (Fig. 7). In contrast, db/db mice displayed gross endothelial dysfunction, as demonstrated by the absence of relaxation to ACh (Fig. 7). Interestingly, KO of GDF15 partly rescued the endothelial dysfunction in db/db mice, by significantly increasing the AUC to ACh compared with db/db mice (Fig. 7).
In addition to the differences in ACh-induced relaxation, db/db and db/db GDF15 KO mice differed in the severity of diabetic complications. db/db GDF15 KO mice had increased renal damage compared with db/db mice, as demonstrated by increased urinary glucose loss (2.1 ± 0.5 vs 0.8 ± 0.1 g/day), increased urine production (19 ± 4 vs. 13 ± 3 g), and higher tubular damage scores (29 ± 5 vs. 12.8 ± 4). Therefore, we cannot exclude a contribution of these renal factors on the observed changes in ACh-induced relaxation in db/db and db/db GDF15 KO mice.
GDF15, a distant TGF-β family member, is emerging as an independent predictor of CVD, and increased GDF15 levels have been associated with endothelial dysfunction. However, an active role of GDF15 on vascular function to explain this has not been previously explored. Here, we show, for the first time, that GDF15 directly modulates vascular contraction and relaxation responses in an endothelium-dependent fashion that involves the NO pathway and may be explained by an interference of GDF15 in the eNOS-CAV-1 interaction.
The intact vascular endothelium is an established regulator of vasomotor tone and other processes in the vessel wall, by the production of vasoactive substances (among others). This includes EDRFs such as NO, PGs, and an unidentified EDHF (18), which may be released both basally or after stimulation of the endothelium (e.g., with ACh) to oppose vasoconstriction (for a recent review, see Ref. 11). Our present findings with isolated murine aortic preparations suggest that GDF15 enhances the basal release of NO (or its bioactivity) from the endothelium, thus suppressing vascular contraction. In support of this, we found that GDF15 inhibits contraction to PE and KCl, which was abrogated 1) in endothelium-denuded aortas and 2) by blockade of eNOS in endothelium-intact aortas. In addition to the stimulation of basal NO release, our results show that GDF15 counteracts the endothelial NO release stimulated by ACh, as evidenced by the marked reduction in the NO component of ACh-stimulated relaxation. The above findings together suggest that GDF15 acts on the endothelial NO system to enhance basal NO release yet impair the action of receptor-stimulated NO release.
One obvious mechanism by which GDF15 may exert its effects is by acting on receptors of the TGF-β superfamily (23). Indeed, TGF-β has been shown to stimulate ROS production by NADPH oxidase (5), which, in turn, may decrease the bioavailability of NO, thus accounting for the impairment of ACh-induced NO-mediated relaxation observed in the present study. However, GDF15 still inhibited ACh-mediated relaxation in the presence of the ROS scavengers tempol and catalase, indicating that GDF15 did not modulate vascular function through a route involving the NADPH oxidase/ROS pathway. Moreover, TGF-β did not affect ACh-mediated relaxation or PE-mediated contraction, indicating that the effects of GDF15 are distinct from those of TGF-β. Together, these results indicate that the GDF15 mechanism does not involve the TGF-β/ROS route.
As an alternative explanation of the GDF15 action, we investigated its action on caveolae. A large body of literature has demonstrated that the function of eNOS is modulated in caveolae, which are specialized microdomains that compartmentalize signal transduction in endothelial cells (for a recent review, see Ref. 22). CAV-1, a structural protein within the caveolus, has been identified as a key regulator of eNOS activity (34). In a recent study (3), it was demonstrated that the scaffolding domain of CAV-1 acts as an endogenous negative regulator of eNOS function, thereby limiting basal NO release. In the present study, we found both GDF15 pretreatment and disruption of caveolae using the compound MCD to result in a similar inhibition of vascular contraction. Moreover, MCD treatment abrogated the inhibitory action of GDF15 on vascular contraction.
Thus, GDF15 effects on both vasoconstriction and relaxation are consistent with a disruption of the interaction between CAV-1 and eNOS. To further address this, we studied this interaction in cultured endothelial cells using immunoprecipitation. In line with our assumption, pretreatment with GDF15 resulted in a significant decrease in the association between eNOS and CAV-1. Taken together, the results of the present study suggest that under basal conditions, GDF15 causes a dissociation between eNOS and its inhibitor CAV-1, resulting in enhanced basal NO release opposing vasoconstriction (Fig. 8). As an additional consequence, loss of eNOS from caveolae will attenuate its interaction with signaling molecules in caveolae, resulting in an impairment of ACh-induced NO release.
Despite the pronounced effects of GDF15 on endothelial function, genetic deletion of GDF15 was not associated with major hemodynamic effects in GDF15 KO mice. Furthermore, vascular contraction and relaxation curves were similar for WT and GDF15 KO mice. The absence of a clear vascular phenotype of GDF15 KO mice despite the strong vascular effects of acute GDF15 pretreatment may be explained by the observation that GDF15 expression is generally low in healthy animals and mainly induced after tissue damage (15, 25, 36, 37). We therefore also studied the involvement of GDF15 on vascular function in a model of T2DM, a condition featuring elevated GDF15 levels. To this end, we used db/db mice that were WT or KO for the GDF15 gene. First, upregulation of GDF15 was confirmed by increased GDF15 expression in white adipose tissue of db/db mice with an intact GDF15 gene. Furthermore, we found these mice to display severe endothelial dysfunction, as previously demonstrated by others (14). Importantly, genetic deletion of GDF15 in the db/db model partly rescued the mice from the impairment of endothelial function, suggesting a role for GDF15 in mediating the development of endothelial dysfunction in diabetes. It should be noted, however, that genetic deletion of GDF15 in db/db mice also had additional effects outside the vasculature, including an increase in diabetic kidney damage. Although its seems unlikely that enhancement of renal damage rescues systemic vascular function, we cannot fully exclude that genetic deletion of GDF15 in db/db mice has indirectly caused this effect. The final confirmation awaits development of tissue-specific GDF15 KO mice.
GDF15 has been attributed great potential as a biomarker in CVD, especially for prognosis, but so far the exact mechanism of GDF-15 in CVD remains unknown (30, 32). Our present results now demonstrate an active role of GDF15 in the development of endothelial dysfunction, a key feature in the onset and progression of CVD. GDF15 reduced agonist-induced NO release, resulting in endothelial dysfunction, yet increased basal NO-release. Similar findings may also be observed under CVD conditions, in patients with chronic heart failure (among others). Impaired endothelium-dependent dilation of resistance vessels in these patients is thought to be involved in the impaired vasodilator capacity in the peripheral circulation (e.g., during exercise), whereas enhanced basal release of NO from the endothelium of resistance vessels was suggested to play an important compensatory role in chronic heart failure (10). Furthermore, increased circulating levels of GDF15 have been associated with an increased risk of developing heart failure in apparently healthy individuals, whereas the GDF15 level and its increase over time have been associated with adverse outcomes in patients with established heart failure (31). Our data also suggest that GDF15 exerts its action by altering caveolar signaling of the endothelium. Caveolae have acquired increasing attention as cellular organelles contributing to the pathogenesis of several cardiovascular conditions, including cardiac hypertrophy, atherosclerosis, and heart failure (8). Consequently, they have been implicated as a potential target for the treatment of renocardiovascular disease (8, 28, 33); our present data suggest that this may include the preclusion of the GDF15 action on caveolar eNOS signaling.
In conclusion, GDF15 impairs vascular contraction and agonist-induced vascular relaxation of the mouse aorta through a mechanism involving altered caveolar eNOS signaling. We propose a model in which uncoupling of eNOS to CAV-1 by GDF15 results in both effects. Our findings may provide a functional explanation for the close associations found between GDF15 and CVD.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: M.M., R.H.H., and L.E.D. conception and design of research; M.M. and A.v.B. performed experiments; M.M., H.B., S.L., P.V., A.v.B., and L.E.D. analyzed data; M.M., S.L., and P.V. interpreted results of experiments; M.M. prepared figures; M.M. and L.E.D. drafted manuscript; H.B., R.H.H., and L.E.D. edited and revised manuscript; R.H.H. and L.E.D. approved final version of manuscript.
The authors thank Marcel van Es for assisting with the coimmunoprecipitation procedure.
- Copyright © 2013 the American Physiological Society