Transcriptional Regulation of Endothelial Arginase 2 by HDAC2

HDAC inhibition impairs endothelial-dependent vascular relaxation

Previous findings from Rossig et.al., have shown that Trichostatin A (TSA), the broad-spectrum HDAC inhibitor with promising effects on a variety of human cancer cells, may cause impairment of endothelium-mediated vascular relaxation ¹⁰. This response was attributed to a small decrease in eNOS expression, but the size of this decrement raises the question of whether other genes that are regulated by HDACs may have contributed to the attenuated vascular relaxation responses to acetylcholine (ACh) that Rossig's group observed. Since Arg2 is a competitive inhibitor of eNOS that impairs endothelial function, we used Arg2 KO mice and biochemical inhibitors to examine the possibility that Arg2 may mediate the diminished vascular relaxation caused by TSA. Pre-incubation of aortic rings of wild type (WT) mice with 200nM of TSA for 18 hours led to impairment of acetylcholine-mediated (endothelium-dependent ) vasorelaxation as compared to rings treated with vehicle alone (E_max 31.24±7.8 vs. 81.9±6.2% for controls; p<0.001, n=6) but this TSA-mediated impairment of vasorelaxation was abolished in WT aortic rings that were pretreated with the arginase inhibitor amino-2-borono-6-hexanoic acid (ABH) (E_max 65.7±12.1%; p<0.05; n=6). Pre-incubation of aortic rings from Arg2-/- mice with TSA (200nM) demonstrated no impairment of endothelial cell (EC)-dependent vasorelaxation as compared to vehicle-treated vessels (E_max 66.3±7.54 vs. 79.0±10.7% in controls; ns; n=6) (Figure 1A-1D). There was no significant difference in SNP-mediated vasorelaxation (an endothelium-independent response since SNP is a nitric oxide donor) in aortas that were treated with TSA or MGCD0103 vs. the vehicle-only control. These findings strongly suggest that TSA-mediated impairment of vascular relaxation occurs through the upregulation of Arg2 in the intimal endothelium.

HDACs regulate Arg2 expression and activity

To define and quantitate the nature of TSA modulation of Arg2 expression, HAEC were treated with 400nM TSA for 18 hours as suggested by the manufacturer (Cell Signaling) and arginase activity and protein expression were then measured. TSA increased both total arginase activity and Arg2 expression (Figure 2A and 2B). We further demonstrate that the effect of TSA on Arg2 protein expression was dose-dependent. HAEC were incubated with increasing concentrations of TSA (0, 0.1, 0.3, or 1μM). TSA treatment for 18 hours increased Arg2 in a dose-dependent fashion, with a maximum effect at 1µM. (Figure 2C and online supplemental Figure IA). The effect was also time-dependent with an increase in Arg2 protein expression observed as soon as 8 hrs after TSA (1μM) exposure (Figure 2D and online supplemental Figure IB). TSA dose and incubation time were chosen to be 1μM and 18 hours respectively for subsequent experiments with HAEC. We next utilized adenoviral constructs of Arg1 and Arg2 shRNA to confirm that the increase in arginase activity is due to Arg2 rather than Arg1 upregulation. As shown in Figure 2E, Arg2 shRNA inhibited the TSA-mediated (1μM) increase in Arginase activity, whereas Arg1 shRNA had no effect. To determine whether the TSA effects on increased Arg2 activity and protein expression resulted from changes at the transcriptional level, we examined Arg2 mRNA levels in HAEC and isolated mice aorta that were exposed to TSA using Arg2-specific primers (see methods) for both RTPCR (Figure 2F) and real time PCR reactions. Arg2 mRNA was substantially upregulated in both cultured HAEC and mice aorta that had been pre-incubated with TSA (Figure 2F, 2G and 2H). Incubation of isolated segments of mice aortas in TSA (200nM) also induced a robust increase in arginase activity (Figure 2I). This reduced concentration of TSA was used for experiments with aorta in order to protect their fragile intima from toxicity.

The NAD-dependent Sirtuin family (SIRT or Class III) of HDACs has been implicated in endothelial function ¹¹. To determine whether SIRTs are involved in regulation of Arg2 expression, we incubated HAEC in presence of the SIRT inhibitor nicotinamide (NAM). We observed no changes in the Arg2 mRNA expression in HAEC after incubation with 0.1 mM or 0.2 mM of NAM for 18 hours (Figure 2F). Interestingly, short-term inhibition (30min) of SIRT with NAM (5mM) had an activating effect on HAEC arginase activity whereas TSA (1μM) had no effect in this time frame as shown in Figure 2J. This suggests that SIRTs might be important in the short-term regulation of Arginase 2 activity, and the time course of this effect implies a mechanism involving post-translational modification.

HDAC inhibition by TSA does not affect Arg2 ubiquitination

Post-translational modification by ubiquitin subjects proteins to proteasomal degradation and therefore affects their abundance. Because acetylation and ubiquitination both target lysine residues of proteins there is significant cross talk between these two pathways ¹². To determine whether Arg2 is ubiquitinated and whether increased acetylation resulting from exposure to HDAC inhibitors may compete with ubiquitination sites on Arg2 and thus increase Arg2 abundance by that mechanism, we immunoprecipated FLAG from HAEC co-expressing FLAG epitope-tagged Arg2 and HA-tagged ubiquitin, and immunoblotted for FLAG. As shown in Figure 3, bands migrating at higher molecular weights were observed in cells expressing both Arg2 and ubiquitin. This finding is consistent with polyubiquitination of Arg2. However, pre-incubation with TSA had no effects on the ubiquitination of Arg2, and this excludes ubiquitin as a potential mechanism for TSA-mediated augmentation of Arg2 abundance. This is the first report to show that Arg2 is modified by ubiquitin.

HDAC inhibition-mediated increases in Arg2 expression occur at the transcriptional level via upregulation of Arg2 promoter activity

To determine whether TSA affects transcription of Arg2, HAEC were pre-treated with actinomycin D (ActD) or cycloheximide (CHX) for 1 hour before treatment with TSA (1μM). TSA failed to increase Arg2 expression in cells treated with either agent (Figure 4A). Next we determined whether inhibition of HDAC affects ARG2 promoter activity. HAEC were transiently transfected with different fragments of a -1kb region of the human ARG2 promoter (PI) that were cloned into the pGL3 luciferase (LUC) reporter plasmid. The following day, cells were treated with TSA (1μM) for 8 hours and luciferase activity was measured using a luciferase kit (promega). As shown in Figure 4B, Arg2 promoter activity was increased in cells treated with TSA as compared with control cells that were treated with vehicle alone. The −0.3kb (PII) and −0.65kb (PIII) fragments of the Arg2 promoter exhibited the highest activity after TSA exposure, while the fragment between −1kb and −0.65 kb, PIV, was not responsive to TSA stimulation. The effect of TSA on the Arg2 promoter was also found to be dose-dependent as shown in Figure 4C. These findings strongly suggest that TSA activates the Arg2 promoter.

Inhibition of HDACs by TSA attenuates Nitric oxide Production

Next, we evaluated the functional consequences of HDAC inhibition on downstream endothelial cell signaling. Because eNOS-derived Nitric Oxide plays central role in endothelial function, we assessed eNOS activity by measuring NO production utilizing three different assays. DAF-DA fluorescence (Figure 5A, 5B) and the Siever's NO analyzer (Figure 5C) were used to determine basal NO production from HAEC challenged with TSA (1μM) alone or co-incubated with ABH (1μM). As shown in Figure 5A and 5B, inhibition of HDAC with TSA significantly decreased NO production in HAEC cells, and this failed to occur when cells were co-incubated with TSA and ABH. A DCFH-DiOxyQ-based fluorescence assay (Figure 5C) was utilized to determine NO radical levels in growth medium from 293 cells that were expressing calcium insensitive eNOS (CIeNOS) that were treated with either TSA alone or together with ABH. 293 cells lack native eNOS expression, but do produce abundant Arg2. Cell medium from 293 cells transfected with a control plasmid (LacZ) was used as control background signal that was subtracted from all the groups containing CIeNOS. TSA decreased NO production from 293 cells expressing CIeNOS. In the presence of ABH, TSA failed to attenuate NO production from 293-CIeNOS cells (Figure 5C). These findings support the hypothesis that the increase in Arg2 that is caused by HDAC inhibition is the major underlying mechanism that drives diminished eNOS activity with TSA.

Arg2 expression is specifically regulated by HDAC2

To test whether augmented Arg2 expression was an effect limited to TSA, two other broad-spectrum HDAC inhibitors were tested together with TSA – SAHA and Scriptaid . We found that incubation of HAEC for 18 hours with 1μM of all 3 agents upregulated Arg2 expression to similar levels (Figure 6A). In contrast, these inhibitors had negligible effect on eNOS expression (Figure 6A). To determine the specific HDAC that regulates Arg2 expression, we first utilized class-specific commercially available pharmacological inhibitors (see online supplement Table I for IC₅₀ data). NAM (5mM), a potent Sirtuin (Class III NAD-dependent HDACs) inhibitor had no effect on Arg2 mRNA (Figure 2F) or Arg2 protein levels (Figure 6C, 6D), suggesting that this family of HDACs is not likely to regulate Arg2 expression. The class I-specific inhibitor that is selective for HDAC1 and HDAC3 - entinostat (MS275, 2μM), the class I-specific and HDAC1-selective inhibitor - tacedinaline (CI994, 1μM), the class II-specific MC1568 (0.5μM), and the HDAC6/HDAC8 selective inhibitor-PC34051(100nM) all had no substantial effects on endothelial Arg2 expression. In contrast, the class I specific inhibitor Mocentinostat (MGCD0103,1μM ), which selectively inhibits both HDAC1 and HDAC2, did upregulate Arg2 expression (Figure 6B, 6C, 6D). This screening suggested that HDAC2 was the most likely candidate for a direct role in Arg2 regulation.

To investigate whether MGCD0103 had similar effects on vascular relaxation as those seen with TSA, we incubated isolated mice aortic rings in MGCD0103 (1μM) for 18 hours. Vascular reactivity was then measured in the presence of either ACh or SNP. MGCD0103 significantly attenuated endothelial-dependent vascular relaxation as compared with vessels treated with the vehicle alone (E_max 42.7±8.5 vs. 82.8±7.4%; p<0.001; n=6) as shown in Figure 6D. Endothelium-independent relaxation as assessed by SNP was unaffected by MGCD0103 (Figure 6E). To further determine the role of Arg2 in MGCD0103-mediated vascular dysfunction we measured dose-dependent vascular relaxation in MGCD0103-treated aortas isolated from Arg2−/− mice treated with either ACh or SNP, and found no significant differences in endothelium-dependent relaxation as compared with aortas from Arg2−/− that were treated with vehicle alone (Emax 62.4±16.5 vs. 79.0±10.7%; ns; n=6) (Figures 6F and 6G).

We next transfected HAEC with increasing concentrations of class I HDAC (HDAC1, 2, 3, and 8) cDNAs (0.3 and 1μg). HDAC 1, 3 and 8 overexpression did not have any effect on Arg2 expression (Figure 7A, 7B, 7C), whereas HDAC2 overexpression decreased Arg2 expression (Figure 7D and online supplemental Figure VA). Additionally, we did not observe any substantial change in Arg2 expression with increased expression of other HDACs that have been implicated in the regulation of EC function - HDAC 4, 5, 6 and 7 (online supplemental Figure IIIA and IIIB). To further confirm the involvement of HDAC2 in Arg2 regulation, we silenced HDAC2 in HAEC using siRNA (Qiagen). Arg2 expression was significantly increased with HDAC2 siRNA knockdown (Figure 7E and online supplemental Figure VB). Because we observed a slight increase in Arg2 expression with PC34051, a specific HDAC8 inhibitor, we evaluated whether HDAC8 has an impact on Arg2 expression. We demonstrated that HDAC8 is expressed in HAEC by using RTPCR (online supplemental Figure IIA). HDAC7, which is well known to be expressed by endothelial cells, was used as a positive control. Using siRNA-mediated knockdown and ectopic transfection-mediated overexpression of HDAC8 in HAEC, we showed that HDAC8 is not an important regulator of Arg2 expression (online supplemental Figure IIA, IIB and Figure 7C).

HDAC2 effects on activation of the Arg2 promoter by OxLDL

We next examined whether the atherogenic stimulus OxLDL that is known to increase Arg2 expression ¹³ (also demonstrated in Figure 8A) had any effect on Arg2 promoter activity. Arg2 promoter fragments were cloned into luciferase constructs as described in Figure 4B. Indeed, OxLDL robustly increased the activity of specific Arg2 promoter fragments (Figure 8B). Interestingly, similar to TSA, fragments between -1kb and 0.3kb were not responsive to OxLDL (Figure 8B). This suggested that common Arg2 promoter regions are regulated by atherogenic stimuli and HDACs. To determine whether OxLDL-mediated activation of the Arg2 promoter is a dose-dependent phenomenon, we transfected HAEC with Arg2 (-1kb) promoter in a luciferase construct, and stimulated cells with increasing concentrations of OxLDL (0, 25, 50 and 100μg/mL). As shown in Figure 8C, OxLDL dose-dependently increase Arg2 promoter activity in HAEC. Interestingly, the activating effect of 50 μg/mL of OxLDL on Arg2 promoter activity was completely abolished in HAECs transduced with adenovirus encoding HDAC2. Further we were able to amplify the Arg2 promoter bound to immunoprecipitated HDAC2 from HAEC DNA/protein complexes (Figure 8D) using a ChIP assay. To further determine whether OxLDL modulates HDAC2 abundance, we exposed HAEC to increasing doses of OxLDL (0, 25, 50 and 100μg/mL) and measured HDAC2 expression. HDAC2 expression was progressively decreased by exposure to 50 and 100 μg/mL of OxLDL (Figure 8E and online supplemental Figure VC). These cells simultaneously exhibited a dose-dependent increase in Arg2 expression - suggesting reciprocal regulation of Arg2 by HDAC2 (Figure 8E, second panel). We further demonstrated that OxLDL-mediated increases in HAEC Arg2 expression (Figure 8F) were abolished by HDAC2 overexpression (Figure 8D and online supplemental Figure VD). This finding is further supported by recent observations that link reduced HDAC2 expression with OxLDL exposure ⁹. These results strongly suggest that decreased HDAC2 abundance is a critical downstream mechanism for OxLDL-mediated increase in Arg2 expression during atherogenesis. It is possible that other transcriptional regulators may also modulate Arg2 expression under conditions of cellular stress. The Arg2 promoter sequence, candidate regulators and their potential targets in the Arg2 promoter sequence are summarized in online supplemental Table II and III.

Additionally, we showed that increased expression of a histone acetyl transferase (HAT), CBP, which causes global cell acetylation has no effect on the endothelial Arg2 expression (online supplemental Figure IV). This further supports the notion that the transcriptional effect on Arg2 expression involves HDAC2 interaction with the Arg2 promoter, and this effect is not occurring simply by enhanced transcriptional accessibility due to increased global acetylation.

OxLDL-mediated impairment of vascular relaxation is ameliorated by HDAC2 overexpression

We next determined whether HDAC2 overexpression could be used as a strategy for reversing vascular dysfunction due to atherogenic stimuli such as OxLDL. Isolated mice aortic rings were transduced with either HDAC2 or GFP (control) adenoviruses and incubated with or without OxLDL (50μg/mL) for 48 hours. As shown in Figure 9A, OxLDL significantly impaired endothelial-dependent vascular relaxation (ACh study) as compared with the untreated controls (E_max 52.1 ± 3.1 % vs 93.0 ± 2.8 %; p < 0.0001; n= 6), suggesting that endothelial function is compromised in this group. The sodium nitroprusside dose response effect on OxLDL-incubated aortic rings was not significantly different from that in untreated controls, suggesting that OxLDL does not affect the endothelial-independent component of blood vessel reactivity. Furthermore, mice aortic rings overexpressing HDAC2 (Figure 9C and online supplemental Figure V, panel E) exhibited significantly improved acetylcholine dependent vascular relaxation despite OxLDL exposure as compared with GFP-overexpressing aortic rings exposed to OxLDL (E _max78.4 ± 5.2 % vs 52.1 ± 3.1 %; p < 0.0001; n= 4-6) (Figure 9A). There was no significant difference in sodium nitroprusside-induced vascular relaxation between these groups (Figure 9B). Consistent with studies in cultured HAEC, Arginase 2 expression was significantly attenuated in mice aortas overexpressing HDAC2 (Figure 9C, second panel). Increased expression of HDAC2 in HAEC also decreased total arginase activity, and we have found this activity to be predominantly due to Arg2 (data not shown). We next determined whether increased expression of Arg2 as observed with knockdown of HDAC2 affect nitric oxide production in HAEC. Indeed, HDAC2 shRNA significantly attenuated NO production as determined with DAF-FM DA fluorescence, and this was rescued above the control levels in presence of Arg2 shRNA (Figure 9E). NO production was also significantly augmented in HAEC where only Arg2 was knocked down. The effectiveness of Arg2 and HDAC2 shRNAs in HAEC are demonstrated in online supplemental Figure VI, panels A and B.