VEGF-A: A Novel Mechanistic Link Between CYP2C-Derived EETs and Nox4 in Diabetic Kidney Disease

Diabetic kidney disease (DKD) is the leading cause of chronic kidney disease in patients with type 1 diabetes (T1D) and type 2 diabetes (1). The earliest clinical manifestation of DKD is increased urinary albumin excretion, or albuminuria, which eventually progresses to overt proteinuria (1). Podocyte injury is a predominant feature of DKD. It is characterized by significant phenotypic changes, including foot process effacement, alteration in slit diaphragm proteins, and reduction in the number and density of podocytes (2), leading to a breach of the glomerular filtration barrier, which results in proteinuria (3). In addition, DKD is associated with progressive mesangial expansion and accumulation of extracellular matrix (ECM) in the glomerular mesangium and basement membrane, leading to glomerulosclerosis and a reduction in glomerular filtration rate (4). Despite recent advances made in the field, the underlying mechanisms of hyperglycemia-induced glomerular injury remain to be elucidated.

Different CYP450 isoforms can metabolize arachidonic acid to produce 20-hydroxyeicosatetraenoic acid (20-HETE) and epoxyeicosatrienoic acids (EETs). CYP4A and CYP4F subfamilies generate 20-HETE, while CYP2B, CYP2C, and CYP2J subfamilies produce EETs (5). In fact, CYP2C11 and CYP2C23 are the major isoforms responsible for most of the renal epoxygenase activity in the rat kidneys (6,7). Of note, a portion of CYP2C23 is actually present in an enzymatically inactive form (8); thus, our study focuses on investigating the effect of CYP2C11 on renal injury. We have previously provided evidence for a link between CYP2C11 downregulation and the development of DKD (9). CYP2C11 catalyzes the oxidation of 11,12- and 14,15-EETs with almost equal efficiency; 14,15-EET is considered to be the major biologically active metabolite in the kidney (10). EETs are known to be short-lived molecules and are thus hydrolyzed into less active diols, dihydroxyeicosatrienoic acids, via the soluble epoxide hydrolase (sEH) enzyme (11). Given that EETs play a potent renoprotective role, we examined whether an sEH inhibitor, 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA), which inhibits the metabolism of EETs, protects against pathologic changes in DKD.

Increasing evidence highlights that hyperglycemia-induced glomerular injury is mediated by the overproduction of NADPH oxidase–associated reactive oxygen species (ROS) (12,13). We and others have previously shown that Nox-induced ROS production results in podocyte apoptosis/depletion in T1D and type 2 diabetic mice (13–15), as well as in mesangial cell hypertrophy and loss (16,17). Importantly, we have elucidated a sequential activation of CYP4A-induced 20-HETE and Nox4 in T1D-induced kidney injury (9,13). However, little evidence exists addressing the interplay among the epoxygenases, cytochromes, particularly CYP2C11, and Nox4 as well as the cross talk of the CYP2C/Nox4 axis with other key signaling molecules involved in glomerular cell physiology and pathophysiology.

Vascular endothelial growth factor A (VEGF-A) secreted by kidney podocytes is a potent angiogenic factor that plays a major role in regulating podocyte function and glomerular hemodynamics (18). Recently, experimental and clinical studies have reported the association of an early increase of VEGF-A expression with the onset of renal injury (19,20). Diabetes, for instance, has been associated with increased levels of glomerular VEGF-A expression and VEGF receptors in the kidneys (21,22). Therefore, a homeostatic regulated VEGF-A expression is essential to maintain podocyte integrity and normal renal function (23). Yet, the exact underlying signaling pathway responsible for VEGF-A alteration and glomerular injury induction remains to be understood.

This study aims to investigate whether increasing EET bioavailability or inhibiting VEGF-A signaling pathway attenuates diabetes-induced glomerular injury. It also aims to delineate the cross talk between CYP2C-derived EETs and VEGF-A.

All animal procedures were conducted in accordance with the American University of Beirut Animal Care and Use Committee guidelines. Male Sprague-Dawley rats weighing between 200 and 225 g were injected with a single dose of 55 mg/kg streptozotocin (STZ). Control rats were injected with sodium citrate buffer (0.01 mol/L, pH 4.5) alone. Five weeks after diabetes onset, the rats were divided into two groups: 1) untreated T1D rats; and 2) T1D rats treated with 25 mg/L of AUDA in drinking water for 6 weeks (24). In parallel experiments, 5 weeks after diabetes onset, diabetic rats were divided into three different groups: 1) untreated T1D group; 2) T1D group treated with 3 mg/kg SU5416 (Aldrich-Sigma) twice weekly for 8 weeks (25); and 3) T1D group treated with 5 mg/kg humanized monoclonal anti–VEGF-A neutralizing antibody (bevacizumab or Avastin; Genentech) twice weekly for 8 weeks (26). Blood glucose and body weight were measured weekly. Rats were placed individually into metabolic cages for 24 h before sacrifice. Urine and plasma samples were collected for subsequent analysis. Concentration of urinary albumin was measured by using a rat albumin ELISA quantitation kit (Bethyl Laboratories, Montgomery, TX). Urinary and plasma creatinine were determined by a commercially available creatinine assay kit using the Cobas Integra 400 Plus computerized analyzer (Roche Diagnostics). After sacrifice, both kidneys were removed and weighed. A slice of kidney cortex at the pole was fixed with 4% formalin for immunohistochemical analysis or flash-frozen in liquid nitrogen and stored at −80°C for microscopy and image analysis. Cortical tissues from the two kidneys of each mouse were used for isolation of glomeruli by differential sieving with minor modifications as described previously (9,12,13,17).

Immortalized rat podocytes courtesy of Dr. Jeffrey I. Kreisberg and conditionally immortalized human podocytes acquired from Dr. Moïn Saleem were cultured as previously described (27–29) and pretreated with 50 μmol/L of 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA; Cayman Chemical, Ann Arbor, MI) before incubation with 25 mmol/L d-glucose (high glucose [HG]) for 72 h. Results were compared with cells pretreated with 10% DMSO (vehicle) and incubated in the presence of either 5 mmol/L d-glucose (normal glucose [NG]) or 25 mmol/L d-glucose (HG) for 72 h. For siRNA transfection experiments, siGENOME siRNA specific for rat or human VEGF-A were obtained from Thermo Scientific Dharmacon, and the transfection with the siRNA of interest or with the control nontargeting siRNA was performed as previously described (30). Podocytes transfected with nontargeting siRNA were then treated with either NG or HG for 72 h. Podocytes transfected with siRNA (siVEGF-A) were treated with HG for 72 h. In parallel, 30 ng/mL of exogenous recombinant rat VEGF-164 or exogenous recombinant human VEGF-165 were added to cultured rat podocytes in NG for 72 h. Mannitol (25 mmol/L) was used as an osmotic control.

Fixed renal cortical tissues were cut (4-µm–thick sections) and stained with periodic acid Schiff (PAS) reagent to assess glomerulosclerotic index (GSI) and mesangial area and Masson trichrome (MT) stain to evaluate collagen deposition as previously described (31). A quantitative measurement for 25 randomly sampled glomeruli was blindly performed on each group using ImageJ software.

Dual-label immunohistochemistry was used to identify the slit diaphragm protein nephrin, as previously described (13).

mRNA was analyzed by real-time RT-PCR using the ΔΔ threshold cycle method (13,30,32). mRNA expression was quantified using the CFX96 Touch (Bio-Rad Laboratories, Hercules, CA) with SYBR Green dye and predesigned rat RT2-quantitative PCR primers of the corresponding gene of interest.

Treated cells or homogenates from glomeruli isolated from the renal cortex were prepared, and Western blot analysis was performed as previously described (13,30,32) using rabbit polyclonal antibodies against Nox4 (1:250; Santa Cruz Biotechnology), VEGF-A (1:500; Abcam), and collagen IV (1:1,000; Abcam). Densitometric analysis was performed using ImageJ software.

ROS generation was assessed by high-performance liquid chromatography (HPLC) analysis of dihydroethidium-derived oxidation products as previously described (32).

NADPH oxidase activity was measured in cultured podocytes or in glomeruli isolated from kidney cortex, as previously described (9,12,13,17,30,32).

Levels of 14,15-EETs were measured using the 14,15-EET/DHET ELISA kit (Detroit R&D, Inc., Detroit, MI) according to the manufacturer’s protocol.

Apoptosis activity in kidney tissue was examined by an ssDNA Apoptosis ELISA Kit (Chemicon International, Temecula, CA) according to the manufacturer’s instructions.

A caspase-3/CPP-32 colorimetric assay kit (BioVision) was used to assay caspase-3 activity in kidney tissue homogenate according to the manufacturer’s protocol.

The cellular DNA fragmentation ELISA (Roche Diagnostics GmbH, Mannheim, Germany) for detection of BrdU-labeled DNA fragments in culture supernatants and cell lysates was used according to the manufacturer’s protocol (9,12,13,17,30,32).

A Caspase-3 Fluorescence Assay Kit (Cayman Chemical) was used according to the manufacturer’s protocol (9,12,13,17,30,32).

Statistical significance was assessed by one-way ANOVA for multiple comparison of the means with Prism 6 software (GraphPad Software). A P value <0.05 is statistically significant. Results are expressed as mean ± SE.

All data generated or analyzed for this study are included in the published article and its online supplementary files. The complete set of figures generated during the current study is available upon request from the corresponding author. No novel resources were generated/analyzed during the study.

To elucidate the renoprotective mechanisms associated with CYP2C11-derived EET bioavailability, renal injury was assessed in an STZ-induced T1D rat model treated with AUDA. Blood glucose levels and HbA_1c percent were significantly increased in untreated diabetic rats in comparison with their control littermates. There was no significant difference in random blood glucose or HbA_1c percent between diabetic rats treated with AUDA and untreated diabetic rats (Table 1). Urine albumin excretion (UAE), urinary albumin-to-creatinine ratio (UACR), and blood urea nitrogen (BUN) were all significantly increased in untreated diabetic rats in comparison with control rats. Interestingly, treatment with AUDA significantly reduced these kidney markers to near-control levels (Table 1).

Immunohistochemical studies were then performed. T1D was associated with marked increase in GSI and mesangial area expansion. Interestingly, in diabetic rats treated with AUDA, GSI and mesangial area were significantly reduced as compared with untreated diabetic rats (Fig. 1A–C). Collagen deposition percentage (Fig. 1D and E) and mRNA levels of markers of fibrosis, including collagen I (Fig. 1F), collagen IV (Fig. 1G), fibronectin (Fig. 1H), and p21, a marker of cell hypertrophy (Fig. 1I), were all significantly increased in the untreated diabetic group compared with the control group. Treatment with AUDA markedly reversed these observations (Fig. 1D–I). Furthermore, our results show that treatment with AUDA significantly reversed diabetes-induced glomerular cell apoptosis as assessed by the detection of denatured DNA with monoclonal antibody to ssDNA (Fig. 1J) and caspase-3 activity (Fig. 1K). Likewise, our data show T1D was associated with evident podocyte phenotypic changes as measured by the alteration in nephrin, which was restored upon treatment with AUDA (Fig. 1L and M).

Of note, our data show a significant decrease in CYP2C11 mRNA levels in the isolated glomeruli of untreated T1D as compared with their controls. This was paralleled by a decrease in 14,15-EET levels. Treatment with AUDA did not increase CYP2C11 mRNA levels, yet it significantly increased 14,15-EET levels as compared with the untreated diabetic group (Supplementary Fig. 1A and B), suggesting that the dose of AUDA used was able to restore EET bioavailability.

Our results show that superoxide generation is significantly increased in the isolated glomeruli of untreated diabetic rats when compared with their control littermates. Treatment with AUDA significantly decreased ROS production (Fig. 2A), Nox4 mRNA levels and protein expression (Fig. 2B and C), as well as diabetes-induced NADPH oxidase activity (Fig. 2D).

To uncover potential mechanisms underlying the cross talk between CYP2C11/EETs and other signaling pathways in DKD, we examined the role of VEGF-A, which is known to play a role in proteinuria. In this study, VEGF-A mRNA levels and protein expression were both significantly increased in the isolated glomeruli of the untreated T1D group compared with the control group (Fig. 3A and B). Likewise, urinary VEGF excretion was significantly increased in the diabetic group as compared with the control group (Fig. 3C). Intriguingly, treatment with AUDA markedly reversed these observations (Fig. 3). The results imply that AUDA attenuates renal/glomerular injury in T1D by inhibiting the VEGF-A signaling pathway.

To further assess the role of VEGF-A in DKD, two pharmacological inhibitors of VEGF-A were used: bevacizumab, a humanized anti-VEGF monoclonal antibody, and SU5416, a selective tyrosine kinase inhibitor that blocks VEGF receptor autophosphorylation. As expected, blood glucose levels and percent HbA_1c were significantly increased in untreated diabetic rats compared with their control littermates. Blockade of VEGF-A signaling using either anti-VEGF or SU5416 did not show significant alterations in blood glucose levels or HbA_1c percent as compared with the untreated diabetic group. In addition, UAE, UACR, and BUN were all markedly increased in T1D, as anticipated. Treatment with anti-VEGF or SU5416 significantly reversed these changes (Table 2).

To assess the modulation of VEGF-A by VEGF inhibitory drugs, VEGF-A mRNA levels and protein expression were measured in the isolated glomeruli. Our results show that VEGF-A mRNA levels and protein expression were significantly increased in untreated diabetic rats compared with their controls (Supplementary Fig. 2A and B). Besides, urinary VEGF secretion was markedly increased in the untreated diabetic group in comparison with the control group (Supplementary Fig. 2C). While treatment with anti-VEGF significantly decreased VEGF-A mRNA levels and protein expression as well as urinary VEGF, treatment with SU5416 failed to do so (Supplementary Fig. 2).

Our results show that treatment with either anti-VEGF or SU5416 significantly reversed the increase in GSI and mesangial area expansion observed in the untreated T1D rats as compared with the controls (Fig. 4A–C). Similarly, inhibition of VEGF-A using either anti-VEGF or SU5416 markedly decreased collagen deposition (Fig. 4D and E) as well as collagen I (Fig. 4F), collagen IV (Fig. 4G), fibronectin (Fig. 4H), and p21 (Fig. 4I) mRNA levels when compared with untreated T1D rats. Treatment with either anti-VEGF or SU5416 decreased diabetes-induced glomerular cell apoptosis as assessed by the detection of denatured DNA with monoclonal antibody to ssDNA (Fig. 4J) and caspase-3 activity (Fig. 4K) and markedly attenuated diabetes-induced nephrin alteration (Fig. 4L and M).

We next assessed whether VEGF-A regulates NADPH oxidase–induced ROS production and Nox4 expression. Superoxide generation was significantly increased in glomeruli of the untreated diabetic rats as compared with their controls. This observation was markedly inhibited by treatment with either anti-VEGF or SU5416 (Fig. 5A). Furthermore, T1D was associated with a significant increase in Nox4 mRNA (Fig. 5B) and protein expression (Fig. 5C) as well as NADPH oxidase activity (Fig. 5D), which was reversed upon treatment with either anti-VEGF or SU5416 (Fig. 5).

It should be noted that in all of our in vivo experiments, the vehicles of the corresponding drugs were administered to both the control and the untreated diabetic groups. No changes were observed when compared with the nontreated group (no vehicle) (data not shown).

Our in vitro data performed in mice and human podocytes corroborate with our in vivo results and indicate that decreased EET formation mediates the effect of HG on ROS production and exacerbates podocyte injury through a VEGF-A–dependent mechanism, which in turn acts as a mechanistic link between EET formation and Nox4 (Supplementary Figs. 3–5).

In all of our in vitro experiments, equimolar concentrations of mannitol were used as osmotic control and, as expected, did not have any effect on the performed experiments when compared with NG-treated podocytes (data not shown).

In this study, we demonstrate that hyperglycemia-induced glomerular injury is mediated, at least partially, by the alteration in EET production by CYP2C11, the activation of the VEGF-A signaling pathway, and the upregulation of Nox4-associated ROS overproduction. We also demonstrate that blockade of the VEGF-A signaling pathway by either a neutralizing anti-VEGF antibody (Avastin) or a selective tyrosine kinase inhibitor (SU5416) ameliorates glomerular injury by reducing NADPH oxidase activity and Nox4 expression.

Emerging evidence supports a renoprotective role for EETs in kidney injury (33,34). Overexpression of CYP2J2 was shown to attenuate renal damage in STZ-induced diabetic mice (33). In addition, excessive EET production was shown to prevent renal fibrosis by inhibiting renal tubular epithelial–mesenchymal transition in vitro (33). Genetic disruption of sEH in STZ-induced diabetic mice was found to decrease creatinine, BUN, and albumin excretion (34). In line with these findings, we have previously shown that altered expression of CYP450 enzymes is a key factor in the onset and progression of DKD (9). DKD is accompanied by an increased CYP4A expression that leads to an increased formation of 20-HETE, as well as by a decreased CYP2C11 expression that leads to a decrease in 14,15-EETs. On one hand, 20-HETE was found to stimulate ROS production, leading to hypertrophy and increased expression of ECM protein. On the other hand, the decrease in levels of EETs after long-term exposure of cells to HG was also found to stimulate ROS production, leading to hypertrophy and ECM accumulation (9). Interestingly, HET0016, an inhibitor of CYP4A, attenuated HG-induced tubular injury, while the inhibition of EETs reinforced the detrimental effects of HG on the tubules (9). Consistent with these observations, we show that hyperglycemia-induced glomerular injury is mediated by a decrease in CYP2C11 mRNA expression and 14,15-EET formation. Furthermore, we demonstrate that the inhibition of sEH using AUDA attenuates podocyte loss and albuminuria. These findings were further confirmed in vitro, where exposure of rat or human podocytes to HG reduced CYP2C11 protein expression and 14,15-EET formation, leading to podocyte apoptosis. Taken together, these data underline the role of hyperglycemia in inducing glomerular injury by decreasing CYP2C11-derived EETs.

Nox4-derived ROS was shown to increase fibronectin expression and to induce renal hypertrophy (14). In congruence with these observations, previous studies from our laboratory have indicated that HG-induced ROS production was brought about by sequential upregulation of CYP4A/20-HETE, Nox1, and Nox4, resulting in apoptosis of cultured mouse podocytes. Furthermore, inhibition of CYP4A in OVE26 T1D mice attenuated albuminuria and podocyte loss, further validating our results (13). However, limited evidence demonstrates the interplay between EETs and NADPH oxidases. In a recent study, ROS generated by Nox4 and EETs hydrolyzed by sEH contributed to the homocysteine-induced inflammation in vascular smooth muscle cells, leading to vascular remodeling (35). In addition, apocynin, an NADPH oxidase inhibitor, was found to improve cardiac remodeling in chronic renal disease by upregulating EET levels and inhibiting sEH cardiac expression (36). In this study, our data show that hyperglycemia-induced downregulation of CYP2C11-derived EETs upregulates Nox4 expression and increases ROS production. These findings suggest that hyperglycemia-induced ROS production is mediated by the sequential downregulation of CYP2C11 and upregulation of Nox4. In support of our data, growing evidence highlights the role of podocyte sEH in controlling renal function in the setting of hyperglycemia by modulating oxidative stress (37,38). In a recent study, inhibition of sEH with trans-4-[4-(3-adamantan-1-yl-ureido)-cyclohexyloxy]-benzoic acid attenuated renal injury, improved mitochondrial function, reduced ROS production, and decreased endoplasmic reticulum stress in kidneys of db/db mice, suggesting a potential role for sEH inhibition in the treatment of diabetic nephropathy (37). Furthermore, in an experiment on overweight hyperglycemic mice, inhibition of sEH prevented microalbuminuria and renal inflammation. However, in contrast to our findings, sEH inhibition in this study did not alter ROS levels. It did not affect Nox2 expression but instead increased the expression of the antioxidant superoxide dismutase 1 (38).

VEGF has recently emerged as a key contributor to the onset of proteinuria (21,22,39). A previous study has shown that inhibiting VEGF receptor with SU5416 attenuates albuminuria in diabetic mice (22). Many studies, however, have reported a protective role for VEGF-A in maintaining renal vasculature (40,41). In fact, increasing levels of VEGF-A in the kidneys seems to be a possible approach to decrease proteinuria and podocyte apoptosis and to protect renal vasculature (41). Therefore, a robust physiological level of VEGF-A is crucial to maintain normal renal function. In our study, we provide the first evidence of the cross talk between CYP2C-derived EETs and the VEGF-A signaling pathway. We show that treatment with AUDA attenuates renal injury in T1D by decreasing VEGF-A expression. These findings are further confirmed in our in vitro studies, in which we show that a decrease in EET formation mediates the effect of HG on ROS production and worsens podocyte injury via a VEGF-A–dependent mechanism.

To further delineate the VEGF-A signaling pathway and establish its role in diabetic nephropathy, STZ-induced T1D rats were treated using two different pharmacological inhibitors that target VEGF signaling. In this study, we proved that VEGF-A expression was upregulated in the glomeruli of STZ-induced T1D rats. In SU5416-treated rats, no significant difference in VEGF-A expression was noted as compared with diabetic rats. This could be explained by the activation of a positive-feedback loop on VEGF release to compensate for receptor blockade and the inhibition of downstream signaling (39,42). In contrast to the latter observation, treatment with anti-VEGF was able to sequester VEGF-A and to significantly decrease its expression in renal tissues. Likewise, while treatment with anti-VEGF significantly decreased urinary VEGF as expected, treatment with SU5416 failed to do so. In this study, we also show the implication of VEGF-A in DKD as measured by renal structural and functional parameters in STZ-induced T1D animals. Hyperglycemia-induced VEGF-A overexpression exacerbates glomerular injury and ECM expansion. Notably, blocking the activation of VEGF-A signaling was able to reverse these changes. VEGF-A was previously described to hold a sole paracrine function by which it acts locally, diffusing from podocytes to neighboring target cells (43). However, our findings demonstrate that podocytes, aside from being a source of VEGF-A, also act as target cells for VEGF-A. Furthermore, our data show that VEGF inhibition using pharmacological inhibitors in STZ-induced T1D rats restores podocin levels, reduces podocyte depletion/apoptosis, and attenuates albuminuria. These observations endorse a deleterious role of VEGF-A, highlight its plausible autocrine mode of signaling, and corroborate with previously published data showing that the induction of podocyte-specific VEGF164 overexpression in adult transgenic mice led to proteinuria, glomerulomegaly, glomerular basement membrane thickening, mesangial expansion, loss of slit diaphragm proteins, and podocyte effacement (44). Moreover, our findings demonstrate increased NADPH oxidase activity and Nox4 expression in rat podocytes treated either with HG concentrations or with exogenous recombinant VEGF-A. Inhibition of VEGF-A using siRNA in vitro significantly attenuated Nox4 activity and expression as well as ROS generation. Similar results were observed in T1D animals in which both anti-VEGF and SU5416 were able to significantly reduce hyperglycemia-mediated ROS production and Nox4 expression. While our study supports other data from the literature in which VEGF-A acts as an upstream regulator of NADPH oxidase activity and induces endothelial ROS production mainly through Nox2 and Nox4 (45,46), other studies have found that VEGF-A expression is regulated downstream of NADPH oxidases in renal tissues where specific inhibition of Nox4 in podocytes regulated diabetes-induced glomerular VEGF expression and urinary excretion (47). Collectively, these observations suggest a plausible activation loop between VEGF-A and NADPH oxidases, namely Nox4.

In summary, our observations confirm that hyperglycemia-induced glomerular injury is mediated by the downregulation of CYP2C11-derived EET production, followed by the activation of VEGF-A and upregulation of Nox4. Thus, our findings have established a previously unrecognized link between CYP2C11-derived EETs and the VEGF-A signaling pathway. Notably, blocking the activation of VEGF-A was able to curb glomerular injury. As discussed, and comparably to other studies (48,49), we highlight in this study that downregulation of VEGF-A signaling can be considered a potential therapeutic target in DKD. However, the effect of VEGF inhibitors depends on the type of the disease, its stage, the timing of drug administration, and its concentration, as other studies have shown that VEGF inhibition, when applied against tumors in patients without diabetes, can lead to nephrotoxicity and proteinuria (40). Therefore, more rigorous large-scale clinical studies are required to address the role of VEGF inhibition in altering clinical outcomes in patients with DKD.

Acknowledgments. The authors thank the American University of Beirut Animal Care Facility staff for help in taking care of the animals used in this study and the American University of Beirut Research Core Facilities staff.

Funding. This study was funded by a predoctoral scholarship from the American University of Beirut to R.N., a National Priority Research Program regular research grant from Qatar National Research Foundation to A.A.E. and F.N.Z., a regular research grant from the National Council for Scientific Research–Lebanon to A.A.E., and a Medical Practice Plan regular research grant from the American University of Beirut to A.A.E.

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

Author Contributions. R.N. performed the experiment and wrote the manuscript. K.B. performed experiments. H.E.G. contributed to discussion and reviewed and edited the manuscript. N.S.A. and W.S.A. reviewed and edited the manuscript. B.D. helped in performing part of the animal experiments. A.L., F.C., H.K., and A.R.J. provided critical scientific input into the experiments. F.H. helped in statistical analysis of the results. A.A.E. conceived and designed the study. F.N.Z. conceived the study. All authors reviewed the results, provided essential reviews of the manuscript, and approved the final version of the manuscript. A.A.E. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.