The Natural Protective Mechanism Against Hyperglycemia in Vascular Endothelial Cells: Roles of the Lipid Peroxidation Product 4-Hydroxydodecadienal and Peroxisome Proliferator–Activated Receptor δ

Glucose-free Dulbecco's modified Eagle's medium was from Life Technologies (Grand Island, NY). Biological Industries (Beth-Haemek, Israel) supplied antibiotics, bovine fibronectin, FCS, and soybean trypsin inhibitor. Calbiochem (Darmstadt, Germany) supplied GW501516 and 4-HNE. 4-HDDE was synthesized as described (16). Sigma-Aldrich (Rehovot, Israel) supplied 12-HETE, WY14643, baicalein, cytochalasin B, GSK0660, l-glucose, troglitazone, and the anti–α-tubulin antibody. The following polyclonal antibodies were used: rabbit anti-calreticulin (Stressgen Biotechnologies, Victoria, BC, Canada); rabbit anti–GLUT-1 (courtesy of Dr. H.-G. Joost, the Institute of Human Nutrition, Bergholz-Rehbrücke, Germany); anti-(C terminus) GLUT-3 and -GLUT-4 (Millipore, Billerica, MA); anti-PPARα, -PPARγ, and -PPARδ (Cayman Chemicals, Ann Arbor, MI); anti-PPARδ (H-74; Santa Cruz Biotechnology, Santa Cruz, CA); and horseradish peroxidase–conjugated anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA). Abcam (Cambridge, U.K.) supplied the chromatin immunoprecipitation (ChIP) assay kit. American Radiolabeled Chemicals (St. Louis, MO) supplied 2-[1,2-³H(N)]deoxy-d-glucose (1.48 TBq/mmol). Mirus Bio-Corporation (Madison, WI) supplied the TransIT-LT1reagent and Biontex Laboratories (Munich, Germany) the Metafectene. Promega (Madison, WI) supplied the dual-luciferase reporter assay system, ligase, Pfu DNA polymerase, pGEM T-easy plasmid, random primers, restriction enzymes, and RNA polymerase. RNasin was from TaKaRa (Tokyo, Japan). Real-time PCR reagents were purchased from Applied Biosystems (Carlsbad, CA). ReddyMix PCR Master was from Thermo Scientific (Epsom, Surrey, U.K.). The PPARδ small-interfering RNA (siRNA) was from Dharmacon (Lafayette, CO). The pcDNA3 and pEGFP-N1 plasmids and the MCF10 cDNA library were kindly provided by Dr. R. Hertz and Dr. R. Reich, respectively (The Hebrew University, Jerusalem, Israel). The pSVPORT1-hRXR vector and the 3×PPAR response element (PPRE)-TK-luciferase plasmid were donated by Dr. B.M. Spiegelman (Dana Farber Cancer Institute, Boston, MA). The plasmids pCMX-hPPARγ1 and pCMX-hPPARγ2 and the respective empty plasmids were kindly provided by Dr. R. Evans (Howard Hughes Medical Institute, La Jolla, CA). The pSG5 and pSG5-hPPARα vectors were prepared in Dr. B. Staels' laboratory. All primers were synthesized by Sigma-Aldrich (Rehovot, Israel). Organic solvents were from Frutarom (Haifa, Israel) and Mallinckrodt Baker (Deventer, Holland).

Primary cultures of bovine aortic endothelial cells were prepared and characterized as described previously (20). EA.hy926 cells were cultured and maintained as described (21).

The [H³]dGlc uptake assay in VEC cultures was conducted as previously described (22). The nonspecific [H³]dGlc uptake that was determined in the presence of 10 μmol/l cytochalasin B in the uptake mixture was <4% of the total uptake. Cell numbers were determined microscopically in a hemocytometer following cell detachment with a trypsin-EDTA solution for endothelial cells (Sigma-Aldrich, Rehovot, Israel). Soybean trypsin inhibitor (50 μg/ml) was used to stop the reaction. Trypan blue exclusion tests showed <1–3% dead cells following trypsinization.

Cell lysates were prepared, and Western blot analyses of GLUT-1 calreticulin, PPARδ, and α-tubulin were performed as previously described (22) or according to the antibody suppliers' protocols. Cell surface biotinylation of VECs and determination of plasma membrane–associated GLUT-1 were carried as described (23).

RNA isolation, cDNA synthesis, and real-time PCR analyses of calreticulin and GLUT-1 mRNA in VECs were performed as described by Totary-Jain et al. (13), using the same primers. Both mRNA levels were normalized to 18S rRNA.

VEC cultures, at 60% confluency, were cotransfected with DNA complexed to TransIT-LT1reagent in 2 ml growth medium, according to the manufacturer's instructions. The DNA consisted of 165 ng pSG5-hPPARα, pCMX-hPPARγ1, pCMX-hPPARγ2, pcDNA-hPPARδ, or the corresponding empty plasmids. It also contained the following plasmids: pSVPORT1-hRXR (82.5 ng), 3×PPRE-TK-luciferase reporter (500 ng), the renilla luciferase (100 ng), and pEGFP-N1 (82.5 ng). The latter served to assess of the yield of transfection by fluorescence microscopy (>80%; excitation, 490 nm; emission, 515 nm). After 24 h, the cultures were washed, received fresh medium, and incubated for additional 24 h. Firefly luciferase–induced luminescence was determined with the dual-luciferase reporter assay in a Mithras LB-940 luminometer and normalized to Renilla luciferase activity as an internal control, according to the kit's instructions (Berthold Technologies, Bad Wilbad, Germany).

Polar lipid extraction from culture media or plasma and high-performance liquid chromatography (HPLC) analysis were performed according to Zanardi et al. (24), with some modifications. Briefly, VEC cultures in 15-cm plates were incubated for 24 h with serum-free media and the indicated additions. The media were then collected, cleared by centrifugation, and loaded on prewashed Supelclean LC-18 SPE tubes (6 ml/1gr; Supelco, Bellefonte, PA) and washed with 15 ml cold water and petroleum ether (boiling range 40–60°C). The polar lipid fraction was then eluted with 3 ml cold methanol, dried under N₂, dissolved in 300 μl cold methanol, filtered through a Teflon syringe filter (0.45 μmol/l; National Scientific, Rockwood, TN), sealed under N₂, and kept at −70°C. HPLC analysis was performed in a Merck Hitachi machine with a Supelcosil LC-18-DB column (5 μmol/l particle size, 25 cm × 10 mm; Supelco, Bellefonte, PA) connected to an ultraviolet detector (295 nm). For 4-HDDE, the elution (1.0 ml/min) started with a gradient of acetonitrile:water (42:58), flowed by a linear gradient that progressed over 25 min to 100% acetonitrile. For 4-HNE detection, the initial ratio in the mixture was 30:70. Pure 4-HDDE and 4-HNE were used to resolve their respective peaks at 10.0 and 4.2 min. The recovery of the standards was 85–90%. Clarity-Lite software was used to analyze and quantify HPLC data.

VECs at 5.5 mmol/l glucose were transfected with pcDNA-hPPARδ, pSVPORT1-hRXR, and pEGFP-N1 plasmids, as described above, and incubated for 24 h. The cells were then treated with 100 nmol/l of GW501516 or with DMSO and incubated for additional 24 h. Cell lysates were then prepared, sonicated, fractionated, and immunoprecipitated with anti-PPARδ (H-74) and anti–histone H3 (positive control, ab-1791; Abcam, Cambridge, U.K.) and taken for PCR with specific primers, according to the protocol of the ChIP assay kit as described in the online appendix (available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1207/DC1).

Subconfluent bovine aortic endothelial cultures were transfected with 100 nmol/l of the siRNA 5′-AAAATGATGTCACTGAAGGGC-3′ targeted to bovine PPARδ, using metafectene according to the manufacturer's instructions. Control cells were treated with metafectene only. Cells were harvested 72 h after transfection, and lysates were prepared and used for Western blot analyses.

Results are given as means ± SE. Statistical analyses were performed using the nonparametric Mann-Whitney test.

The potential role of PPARα, -γ, or -δ in the regulation of glucose uptake in VECs was elucidated by studying the capacity of their respective selective agonists, WY14643, troglitazone, and GW501516 (25), to downregulate the rate of glucose uptake in VECs under normal glucose levels (Fig. 1 A). The control treatment (25 mmol/l glucose) shows the characteristic high-glucose–induced ∼50% reduction in the rate of hexose uptake in comparison to the 5.5 mmol/l glucose incubation (10). GW501516, but not troglitazone or WY14643, mimicked high glucose and comparably downregulated the uptake rate in cells under normal glucose, indicating that only PPARδ participates in the downregulatory response. Maximal effect of GW501516 was obtained between 1 and 100 nmol/l and within 36–48 h (supplemental Fig. S1A and B, which can be found in the online appendix. The vehicle (DMSO) had no significant effect on the rate of glucose uptake.

Figure 1 B shows that the PPARδ antagonist GSK0660 (26) effectively prevented the downregulation of glucose uptake induced by 25 mmol/l glucose in VECs, whereas the basal high rate of glucose uptake in cells at 5.5 mmol/l glucose was not altered. Supplemental Fig. S1C confirms the specificity and the competitive nature of GSK0660-induced inhibition of PPARδ.

Baicalein, a specific 12-LO inhibitor, was used to test the hypothesis that 12-LO metabolites and PPARδ cooperate in the downregulatory response. Fig. 1,C shows that the inhibition of 12-LO with baicalein prevented high-glucose–induced downregulation of glucose uptake. Yet, GW501516 downregulated the uptake in the presence of baicalein. No synergistic interactions between GSK0660 and baicalein in high-glucose VEC cultures were found, suggesting that 12-LO and PPARδ act in sequence (Fig. 1 D). The cellular content of PPARδ was not altered by the ambient glucose, GW501516, or baicalein (supplemental Fig. S2). Supplemental Fig. S3 indicates that the slight increase in the osmotic pressure in the high-glucose–containing culture medium had no effect on the glucose transport system.

The data in Fig. 2 A–C show that GW501516 reduced GLUT-1 mRNA and protein levels and plasma membrane abundance in cells exposed to 5.5 mmol/l glucose to the downregulated levels induced by the 25 mmol/l glucose incubation. Moreover, GW501516 also prevented baicalein-mediated upregulation of GLUT-1 expression and its subcellular distribution in cells exposed to high glucose, thus linking 12-LO metabolites to PPARδ in the mechanism of the downregulation of GLUT-1 expression. Supplemental Fig. S4 shows that the cell content of GLUT-3 and GLUT-4 in VECs was not altered by high glucose or GW501516.

The hypothesis that high glucose levels activate PPARδ was further investigated in VECs that were transfected with a PPRE-luciferase reporter vector (3×PPRE-TK-luciferase). Figure 3,A shows a 1.5-fold increase in luciferase activity in cells exposed to 25 mmol/l glucose in comparison with the low-glucose incubation. GSK0660 abolished this stimulatory effect, suggesting a PPARδ-dependent transcription of the luciferase gene. The control experiment in supplemental Fig. S5 shows that GSK0660 fully inhibited GW501516-induced transcription of luciferase in the same assay. VECs overexpressing the various hPPAR isotypes (α, γ1, γ2, or δ), hRXR, and the PPRE-luciferase reporter expression plasmids were used. Control cells received the empty plasmids. All four hPPAR isotypes were successfully overexpressed (supplemental Fig. S6). Figure 3,B shows that high glucose increased luciferase activity 1.3- to 1.5-fold in VECs that were transfected with all empty plasmids. This effect, which was also found in VECs transfected with the reporter construct only (Fig. 3 A), is attributed to the activation of an endogenous PPAR by high glucose. The observation that high glucose markedly increased luciferase activity (2.6-fold) only in cells overexpressing hPPARδ suggests that PPARδ is this endogenous receptor that is activated under hyperglycemic conditions.

High glucose increases 12-LO expression in VECs and consequently augments the synthesis and secretion of 12-HETE in VECs (10). When tested in the present luciferase reporter assay, 12-HETE (3 μmol/l, 24 h) increased luciferase activity 1.70 ± 0.03-fold (n = 3) in hPPARδ-expressing VECs, in comparison with control cells. This is most likely an underestimated value because 12-HETE is prone to a rapid oxidation in the open air (t_1/2 = 10–12 min [10]). This degradation is most likely the reason for the failure of exogenously added 12-HETE to downregulate the rate of hexose transport in VEC cultures (10).

Radical-induced peroxidation of the 12-LO and 15-LO products, 12-HpETE and 15-HpETE, respectively, generate the corresponding biologically active aldehydes 4-HDDE and 4-HNE (16,18). It has been reported that 15-HETE and 4-HNE activate PPARδ (19). We asked whether 4-HDDE could activate PPARδ in VECs. Figure 4,A shows a remarkable 4.3 ± 0.1-fold increase in hPPARδ-dependent, but not hPPARα-, hPPARγ1-, or hPPARγ2-dependent, stimulation of the reporter luciferase activity in cells treated with 100 nmol/l 4-HDDE. Like high glucose and GW501516 (Figs. 1 and 2), 4-HDDE (50 nmol/l) significantly reduced the rate of hexose uptake (Fig. 4,B) and the cell content of GLUT-1 mRNA and protein (Fig. 4,C and D) in VECs under 5.5 mmol/l glucose. The upregulatory effect induced by baicalein in high-glucose cultures was also reversed by 4-HDDE (Fig. 4,B–D). The inhibitor GSK0660 abolished 4-HDDE–induced downregulation of the rate of hexose transport in VECs exposed to 5.5 mmol/l glucose (Fig. 4 E). The concentrations of 4-HDDE used in these experiments (50–100 nmol/l) did not compromise VEC viability (supplemental Fig. S7). The similar effects of 4-HDDE and GW501516 suggest that the former is an endogenous ligand for PPARδ.

Extraction of polar lipid and HPLC analyses were performed to measure 4-HDDE production in VECs exposed to 2 or 25 mmol/l glucose. High glucose increased the secretion of 4-HDDE 5.9 ± 3.1-fold higher than at the low-glucose incubation (Fig. 5,A), while baicalein blocked this effect, confirming that 4-HDDE is indeed derived from a 12-LO metabolite. The initial step in lipid peroxidation is a hydrogen atom abstraction by ROS (16); therefore, antioxidants can attenuate this process (27). Figure 5,A and B shows that the antioxidant N-acetylcysteine completely prevented high-glucose–dependent generation of 4-HDDE and the downregulation of hexose transport. Fig. 5 C depicts a comparable increase in 4-HDDE generation in the plasma of hyperglycemic Zucker diabetic rats (details on the animals are given in the online appendix). Representative HPLC tracings and 4-HDDE peaks in culture medium and plasma extracts are shown in supplemental Fig. S8. Noteworthy, similar HPLC analyses of the same extracts did not reveal significant differences in 4-HNE levels between the low- and high-glucose incubations (data not shown).

Calreticulin, whose expression is significantly increased in VECs exposed to high glucose, destabilizes GLUT-1 mRNA (13). Figures 6,A–D confirm previous findings (13) on high-glucose–induced increased expression of calreticulin mRNA and protein levels in VECs under high glucose. This effect was eliminated in the presence of baicalein, suggesting that 12-LO metabolites participate in the regulation of calreticulin expression. Both GW501516 (Figs. 6,A and B) and 4-HDDE (Figs. 6,C and D) increased the expression levels of calreticulin mRNA and protein in VECs exposed to 5.5 mmol/l glucose. Both compounds also reversed the effect of baicalein and restored calreticulin mRNA and protein expression to the levels measured under the high-glucose incubation. Inversely, the inhibition of PPARδ with GSK0660 prevented high-glucose–induced increased expression of calreticulin in VECs (Fig. 6 E). These data suggest that PPARδ participates in the regulation of calreticulin expression in VECs.

The potential of PPARδ to interact with PPREs in the calreticulin gene was determined. We identified four PPRE sequences in the promoter region of the bovine calreticulin gene (3,000 bp upstream of the coding sequence) located at 5′-1,710–1,732, 1964–1986, 2,139–2,161, and 2,255–2,277 bp. The ChIP assay was used to determine whether PPARδ binds to these elements. DNA samples from VECs overexpressing hPPARδ and hRXR and treated with 100 nmol/l GW501516 or the vehicle were immunoprecipitated with an anti-PPARδ antibody and taken for the respective PCR analyses. Figure 6 F depicts a specific binding interaction of PPARδ with the PPRE located at 5′-2139–2,161 bp in this promoter region.

PPARδ expression in VECs was silenced with a specific siRNA, and the expression levels of calreticulin and GLUT-1 were determined (Fig. 7). A 50–60% reduction in PPARδ expression in VECs exposed to 25 mmol/l led to a 30–40% decrease in the content of calreticulin and a 30–40% increase in GLUT-1 expression in comparison with the respective controls. Similar effects were observed in VECs exposed to 5.5 mmol/l glucose (data not shown).

Finally, we used the EA.hy926 cell line to determine whether human-derived endothelial cells regulate the glucose transport system like bovine VECs. This cell line, which was derived from primary cultures of human umbilical cord vascular endothelial cells (21), has been shown in many studies to preserve markers and functions of the primary cells. High glucose downregulated the rate of glucose transport and total GLUT-1 level and increased calreticulin expression in a PPARδ-dependent manner in these cells in (supplemental Fig. S9A–C). Exposure of EA.hy926 cells to normal glucose conditions and 4-HDDE mimicked the effect of high glucose and reduced the rate of hexose transport, while GSK0660 prevented this effect (Fig. S9D). These data suggest that both bovine and human VECs use a similar mechanism to downregulate the glucose transport system.

Two key factors that mediate high-glucose–induced downregulation of the glucose transport system in bovine aortic endothelial cells and in the human-derived EA.hy926 cells have been identified: the lipid peroxidation product 4-HDDE and its cognate nuclear receptor PPARδ. The latter increases the expression of the protein calreticulin that was shown before to destabilize GLUT-1 mRNA (13).

The augmented production of 4-HDDE results from high-glucose–induced 12-LO expression and activity and glucose-derived ROS. The mechanism responsible for the increased expression of 12-LO is not yet known. Numerous studies have proven that hyperglycemia promotes the generation of ROS (3,7). We showed before an augmented production of ROS in bovine aortic endothelial cell primary cultures under high-glucose conditions (28). Two observations confirm the role of ROS in the generation of 4-HDDE from 12-LO metabolites. First, the inhibition of 12-LO activity with baicalein significantly reduced 4-HDDE secretion from VECs under high-glucose conditions. Second, the antioxidant N-acetylcysteine blocked 4-HDDE production in VECs under similar conditions. The capacity of baicalein and N-acetylcysteine to prevent high-glucose–induced downregulation of glucose uptake is attributed to the impeded production of 4-HDDE. Collectively, these data indicate that high-glucose–induced overexpression of 12-LO and overproduction of ROS underlie the augmented generation of 4-HDDE.

The physiological, pathophysiological, or cytotoxic effects of 4-hydroxyalkenals depend on their absolute concentrations. Esterbauer et al. (18) concluded that at concentration >20 μmol/l, 4-HNE exhibited cytotoxic effects due to its chemical reactivity and the formation of stable adducts with macromolecules. Bacot et al. (16,29) linked 4-hydroxyalkenals to membrane disorders due to the formation of stable adducts with ethanolamine phospholipids. It is important to emphasize that the effective concentrations of 4-HDDE (1–100 nmol/l) used in the present study were not cytotoxic to VECs. The present study shows that a moderate production of ROS, sufficient to initiate the generation of nontoxic levels of 4-HDDE in VECs, is required for an effective cellular defense against harmful effects of hyperglycemia. Notwithstanding, an excessive and uncontrolled oxidative stress in VECs may lead to the production of cytotoxic levels of 4-hydroxyalkenals. Many studies have indeed suggested that an excessive 4-hydroxyalkenal production underlies various pathological conditions (30,31).

Tissue-specific production of 4-hydroxyalkenals correlates with the particular pattern of expression of the various lipoxygenases and synthesis of their corresponding HpETEs. For example, the peroxidation products 4-HHE and 4-HNE, but not 4-HDDE, were found in the diabetic rat retina, which predominantly expresses 5-LO and 15-LO (29,32). Our present findings on 4-HDDE– and PPARδ-dependent downregulation of the glucose transport system in human VECs correlate well with previous reports on high-glucose–induced expression of 12-LO in human VECs (4).

We found before that high glucose levels increased the expression of 12-LO but not of 15-LO in VECs (10). Subsequently, high glucose promotes the generation of 4-HDDE from 12-HpETE but not of 4-HNE, which is derived from 15-HpETE. Nevertheless, it has been shown before that 4-HNE also activates PPARδ (19). Using the luciferase reporter assay we found that the potency of 4-HDDE to activate the system was 200-fold higher than that of 4-HNE (50 nmol/l vs. 10 μmol/l). These disparate potencies suggest that the binding affinity of 4-HDDE to the ligand binding domain (LBD) in PPARδ is markedly higher than that of 4-HNE. This may reflect the increased hydrophobicity of the 4-HDDE in comparison with 4-HNE (logP values 3.48 and 2.45, respectively, calculated with the Molinspiration Chemoinformatics software). X-ray analyses of crystal structures of the PPARδ LBD identified a network of hydrogen bonds with His⁴¹³, Tyr⁴²⁷, His²⁸⁷, and Thr²⁵³ that are involved in the binding interaction of eicosapentanoic acid (33). His⁴¹³ has also been implicated in a hydrogen-bonding interaction with the 4-hydroxy group of 4-HNE (19). The specific molecular binding kinetics and affinity of 4-HDDE to the PPARδ LBD remain to be analyzed.

Many studies over the last decade have assigned PPARδ key metabolic regulatory functions (34). For instance, it augments lipogenesis and glycolysis in the liver (35), increases fatty acid oxidation in adipocytes (36), and increases oxidative metabolism in skeletal muscles (37,38). Of interest are recent studies on protective effects of PPARδ ligands against the development of atherosclerosis by regulating lipid homeostasis, decreasing the expression of inflammatory genes, and attenuating macrophage migration. We speculate that 4-HDDE–activated PPARδ may mediate some of these effects.

This study links PPARδ activation to the transcription of the calreticulin gene. We have previously found that the final step in the downregulation of GLUT-1 expression in VECs is calreticulin-mediated destabilization of the transporter mRNA (13). Because calreticulin is a multifunctional protein (39), it is feasible that PPARδ-regulated transcription of calreticulin may affect some of these functions (e.g., calcium storage, cell adhesion, chaperoning of malfolded proteins). The present study has identified a PPRE in the promoter of the calreticulin gene that seems to interact with PPARδ. The precise nature of this interaction and its contribution to the assembly of an active transcription complex of calreticulin need further investigations.

Recently, Gross and Staels (40) have stressed the need for the development of novel PPARα and PPARγ agonists for the treatment of type 2 diabetes. We suggest that novel PPARδ agonists may reduce the risk of dysfunctional endothelial cells in blood vessels by downregulating the rate of glucose uptake and protecting them from the deleterious effects of an increased influx of glucose. Clearly, 4-HDDE is not considered a potential pharmacological agonist due to its severe side effects when present at concentrations that allow irreversible covalent binding and cross-linking of macromolecules. Nonetheless, molecular and chemical analyses of the interaction of 4-HDDE with the PPARδ LBD may provide a platform for a rational design of potent and safe PPARδ agonists.

Finally, our previous and current findings on the autoregulation of the glucose transport system in VECs are summarized in the model shown in Fig. 8. High glucose levels increase the expression of the arachidonic acid, metabolizing enzyme 12-LO, which leads to an augmented production of 12-HpETE and its immediate metabolite 12-HETE. Concomitantly, high-glucose–derived ROS initiates the peroxidation of 12-HpETE and the generation of 4-HDDE. This molecule, and possibly 12-HETE, interact specifically with and activate PPARδ, which in turn binds to a PPRE in the promoter region of the calreticulin gene and augments its transcription. Calreticulin binds to a specific 10-nucleotide sequence located in the 3′-untranslated region of GLUT-1 mRNA, destabilizing and rendering it susceptible to degradation. Consequently, the cell content of GLUT-1 protein and its plasma membrane abundance are significantly reduced and downregulation of glucose uptake ensues.

This work was supported by grants from the German-Israel Foundation for Scientific Research and Development (I-750-165.2/2002), the Israel Ministry of Health, the Yedidut Foundation Mexico, the David R. Bloom Center for Pharmacy at the Hebrew University, The Dr. Adolf and Klara Brettler Center for Research in Molecular Pharmacology and Therapeutics at The Hebrew University and COST (European Cooperation in Science and Technology) Actions BM0602 and B35.

Y.R. Y.S.-M., G.C., E.A., and A.G. received fellowships from The Hebrew University Center for Diabetes Research. S.S. is affiliated with the David R. Bloom Center for Pharmacy and The Dr. Adolf and Klara Brettler Center for Research in Molecular Pharmacology and Therapeutics at The Hebrew University.

No potential conflicts of interest relevant to this article were reported.

We thank Drs. R. Hertz, M. Armoni, and T. Kahan for helpful comments and discussions.