Impaired mitochondrial function and metabolism in ATG5-deficient endothelial cells
Efficient CRE excision in mice with endothelial-selective Atg5 deletion was demonstrated 15. Endothelial autophagy deficiency was confirmed by accumulation of the autophagy degradation product P62 in aortic endothelial cells of endothelial-selective ATG5-deficient mice (Cdh5.cre-Atg5lox/lox mice) (Supp Fig. 1). RNA sequencing of primary lung endothelial cells from wild-type (WT) and endothelial-selective ATG5-deficient mice was performed. We selected 892 differentially expressed genes (DEG) (fold-change > 1.2, p-value < 0.05, Supp Table S1) and Ingenuity Pathway Analysis (IPA) was performed. Oxidative phosphorylation, mitochondrial dysfunction, and EIF2 signaling were among the most decreased pathways in Atg5 KO endothelial cells. At the same time, sirtuin, p53, Unfolded Protein Response, Ca2+, and phospholipase C signaling cascades were among the most activated pathways (Fig. 1A, Table 1, Supp Fig. 2). IPA identified molecular functions in autophagy-deficient cells related to endothelial homeostasis and maintenance, carbohydrate and lipid metabolism, cell organization, Ca2 + signaling and cellular transport (Supp Table S2). Gene ontology (GO) analysis further confirmed dysregulation of genes implicated in biological functions (BF) such as Ca2 + signaling, cell adhesion, mitochondrial electron transport, tyrosine kinase receptor signaling, fatty acid metabolism, transport, and platelet adhesion, among others (Supp Fig. 2). GO cellular components (CC) confirmed deregulation of genes encoding for mitochondrial proteins in ATG5-deficient endothelial cells (Supp Fig. 2). IPA upstream functional analysis was used to predict the top upstream transcriptional regulators from DEGs and predicted transcriptional regulators to be activated such as RICTOR, KDM5A, BDNF or to be repressed such as STK11, RB1 or IKZF1 (Supp Fig. 2); factors associated with angiogenesis and endothelial function.
Arguing that autophagy deficiency would modulate proteostasis, we analyzed the proteome of primary lung endothelial cells from control and Cdh5.cre-Atg5lox/lox mice. We selected 198 DEGs (FC > 1.2, p < 0.05, Supp Table S3) and performed IPA and GO pathway analyses. Oxidative phosphorylation and mitochondrial dysfunction were among the most dysregulated canonical pathways in Atg5 KO endothelial cells. Canonical pathways regulating cell metabolism, autophagy, and prostanoid signaling were also identified (Fig. 1B, Table 2, Supp Fig. 3). ATG5 deficiency also altered carbohydrate and lipid metabolism functions, cell-to-cell signaling and interaction/organization of organelles (Supp Table S4).
We next confirmed mitochondrial dysfunction in ATG5-deficient murine endothelial cells by using shRNA-mediated ATG5 knockdown (KD) in HUVECs. Significant diminution in ATG5 expression and impaired autophagy were confirmed by immunoblot of ATG5 and LC3B and P62 immunofluorescence (Fig. 1C-D). In ATG5 KD HUVECs, the pool of active mitochondria was reduced, as assessed by a mitotracker red CMXRos staining. In ATG5 KD cells, we observed a rise in the proportion of mitochondria that colocalize within P62, P62 being involved in mitochondrial recycling (Fig. 1. D-E), suggesting that ATG5 deficiency led to the accumulation of non-recycled P62 + mitochondria. Western blot analysis also demonstrated a decrease of the mitochondrial respiratory chains proteins SDHB, COX1, ATP5A and UQCRC2, but not SDHA and COX4 (Fig. 1F). The reduction of active mitochondria in ATG5 KD HUVECs was further confirmed by flow cytometry analysis (Fig. 1G-H). We concluded that ATG5 KD HUVECs had defective oxidative phosphorylation and mitochondrial dysfunction due to a decreased pool of active mitochondria in the absence of a proper recycling process.
Next, we detected metabolites of the tricarboxylic acid (TCA) cycle in ATG5 KD HUVECs. ATG5 KD HUVECs showed an altered abundance of some metabolites: decreased levels of acetoacetate, αketoglutarate, fumarate, and malate and increased levels of hexose, glutamine, and pyruvate (Fig. 1I-J). Taken together, we demonstrate that ATG5 deficiency in endothelial cells leads to a global change in endothelial cell metabolism and mitochondrial dysfunction.
Endothelial Autophagosomal Content
The LC3-positive membrane-bound vesicles of primary lung endothelial cells from GFP-LC3 mice were isolated to analyze autophagosome protein content 25. We selected 391 enriched proteins in autophagosomes (FC > 2, p < 0.01, Supp Table S5). Consistent with the autophagy interplay with RNA transcription and translation, the top enriched canonical pathways were the spliceosomal cycle, EIF2 signaling, transcriptional repression signaling, and cleavage and polyadenylation of pre-mRNA (Fig. 2A, Table 3). Interestingly, autophagosomes contained proteins necessary for the regulation of endothelial function and signaling pathways such as STAT3-, Notch-, PTEN-, NFκB-, VEGF-, TSP1- or tight junction- and epithelial adherens junction-signaling that we identified. GO analysis confirmed the enrichment of pathways related to mRNA processing, but also to angiogenesis, blood vessel morphogenesis, cell adhesion, sprouting angiogenesis, or extracellular matrix organization (Supp Fig. 4). Of note, proteins of the Weibel Palade bodies were detected in autophagosomes (Supp Fig. 4), consistent with previous findings showing that autophagy is involved in Weibel Palade maturation and is related to the blood coagulation defect in endothelial-selective ATG7-deficient mice 26, that we also observed in endothelial-selective ATG5-deficient mice (Supp Fig. 4).
We detected 63 proteins expressed at the plasma membrane and 36 secreted proteins, including members of protein families important to maintain endothelial function such as SDF1-, Endothelin-, TGF-, angiopoietin-, Netrin-, VEGF-, Notch-, Semaphorin-, FGF-pathways (Fig. 2B, Table 3). None of these proteins were dysregulated at the mRNA level in the RNAseq analysis of primary lung endothelial cells isolated from Cdh5.Cre Atg5lox/lox mice as compared to cells from littermate controls. The comparative proteome of these endothelial cells evidenced a significantly decreased abundance of TIE1, ENG, and ECE1 (Fig. 2C).
VEGFR2 colocalizes with autophagosomes, and ATG5 deficiency impact VEGF signaling.
As VEGFR2/KDR was among the most abundant proteins in autophagosomes, we speculate that autophagy could be involved in VEGFR signaling. We confirmed by immunoblot the presence of VEGFR2 in autophagosomes of primary lung endothelial cells (Fig. 2D). Bafilomycin A1 and VEGFA treatments before GFP immunoprecipitation enhanced the detection of VEGFR2 in autophagosomes (Fig. 2E) and promoted partial relocalization of VEGFR2 to GFP-LC3 + dots as shown by immunofluorescence (Fig. 2F). We confirmed partial VEGFR2 colocalization with the autophagosome markers LC3B and ATG16L1 in VEGF-treated primary lung endothelial cells from WT mice (Fig. 2G) and in HUVECs (Fig. 2H).
As no significant change in total endothelial VEGFR2 mRNA or protein content was detected in primary lung endothelial cells from Cdh5.cre-Atg5lox/lox mice, we hypothesized that VEGFR2 subcellular localization or signaling could be impaired by ATG5 deficiency. We found decreased cell surface VEGFR2 expression in endothelial cells on flat mount aortas from Cdh5.cre-Atg5lox/lox mice, which was associated with discrete and local VE-Cadherin junction disruptions (Fig. 3A, B). Such decreased expression of cell surface VEGFR2 was confirmed in vitro in ATG5 KD HUVECs (Fig. 3C, D). Furthermore, VEGF-mediated VEGFR2 activation was impaired in ATG5 KD HUVECs: VEGFR2 phosphorylation was reduced, as well as subsequent eNOS and P38 activation (Fig. 3E). Of note, ATG5 deficiency impaired not only VEGF/VEGFR2 signaling but also HGF/MET signaling in HUVECs (Supp Fig. 5), suggesting that autophagy could be essential for proper signaling of several tyrosine kinase receptors. Similar impaired VEGFR2-eNOS signaling was found in primary lung endothelial cells from Cdh5.cre-Atg5lox/lox mice (Supp Fig. 5). Finally, VEGF-mediated VEGFR2 internalization and recycling were analyzed by quantifying VEGFR2 colocalization to EEA1 + endosomes. After 20min of VEGFA treatment, most of the internalized VEGFR2 was no longer localized in EEA1 + endosomes in control HUVECs, but the majority of remaining EEA1 + endosomes still contained some VEGFR2. Conversely, in ATG5 KD HUVECs, most of internalized VEGFR2 still localized to EEA1 + endosomes, but most of the EEA1 + endosomes did not contain VEGFR2 (Fig. 3F,G). This result suggests impaired VEGFR2 routing and recycling in ATG5-deficient endothelial cells.
Angiogenesis defects in vascular explants of endothelial-selective ATG5-deficient mice.
We explored ex vivo whether endothelial ATG5 deficiency could influence angiogenesis. While extensive vascular outgrowth was observed from aortic rings isolated from Atg5lox/lox mice, the extension of endothelial sprouts was significantly reduced in aortic rings from Cdh5.cre-Atg5lox/lox mice (Fig. 4A-C). Similar observations were found in a choroid sprouting model of microvascular angiogenesis 27 (Fig. 4D-E). An in vivo angiogenesis assay using subcutaneous angioreactor implants showed reduced FGF2- and VEGFA-mediated angiogenesis in Cdh5.cre-Atg5lox/lox mice, confirming impaired neo-angiogenesis in endothelial-selective ATG5-deficient mice (Fig. 4F). Together, our results demonstrate that endothelial autophagy deficiency impairs neo-angiogenesis.
Defective endothelial autophagy results in selective loss of flow-induced vasodilation in mesenteric arteries and kidneys.
Next, we explored the effects of defective endothelial autophagy on endothelium-dependent and endothelium-independent vasodilation. Ex vivo vascular responsiveness to phenylephrine and concentration-response curves to acetylcholine and NO donor sodium nitroprusside (SNP) were performed in isolated mesenteric resistance arteries from Cdh5.cre-Atg5lox/lox mice and WT littermates. Endothelial deletion of Atg5 did not alter the mesenteric arterial compliance. Phenylephrine-induced contraction, endothelium-dependent relaxation in response to acetylcholine and endothelium-independent relaxation in response to SNP (Fig. 5A-C) did not differ between Cdh5.cre-Atg5lox/lox and control mice. Myogenic tone also remained intact (Fig. 5D). Thus, endothelial autophagy did not impact vessel responses to vasoconstrictive and vasodilatory molecules.
As endothelial flow sensing strongly depends on VEGFR2-NO signaling, we explored the role of endothelial autophagy on functional endothelial flow-sensing by measuring flow-mediated vasodilation (FMD) of mouse mesenteric arteries. In pre-constricted vessels isolated from Cdh5.cre-Atg5lox/lox mice, FMD was strongly impaired compared to WT littermates (Fig. 5E). In line with these results, renal flow-mediated pressure was also increased in isolated perfused kidneys from Cdh5.cre-Atg5lox/lox mice when compared to WT littermates (Fig. 5F). Taken together, these results indicate that endothelial ATG5 is required for FMD in resistive vascular beds.
Endothelial autophagy deficiency increases renal and cerebral vascular resistances.
Telemetry, plethysmography, and ultrasonography measurements did not reveal differences in systolic blood pressure, heart rate, or heart structure between Cdh5.cre-Atg5lox/lox and WT littermates (Supp Fig. 6 and Supp Table 7–8). Further, the vascular network in adult retinas was similar between Cdh5.cre-Atg5lox/lox mice and WT (Supp Fig. 7), suggesting no significant developmental effects of endothelial-ATG5 deficiency.
As impaired FMD of resistance arteries may lead to an increase in peripheral vascular resistance, we investigated the role of endothelial autophagy on vascular resistance in kidneys and the brain. Whereas the cardiac output (CO) did not differ between Cdh5.cre-Atg5lox/lox mice and littermate controls (Supp Table 8), the mean blood flow velocity (mBFV) was significantly decreased in the kidneys and basilar artery of Cdh5.cre-Atg5lox/lox mice (Fig. 5G-J). Importantly, micro-computed tomography 3D angio-scanning of the kidneys revealed no difference in the characteristics of renal vascular trees of the two genotypes (Supp Fig. 8), thus suggesting that the decreased renal mBFV in Cdh5.cre-Atg5lox/lox mice was the consequence of increased vascular resistance rather than abnormal renal vascularization. Correlating with such a hypothesis, we observed a decreased phosphorylation of AKT at Ser473 and NOS3 at Ser1177 in isolated renal arteries (Fig. 5K,L) from Cdh5.cre-Atg5lox/lox mice. We also assessed vasodilation of basilar arteries in response to hypercapnia as an in vivo test of endothelial function 28, 29. This was significantly impaired in Cdh5.cre-Atg5lox/lox mice (Fig. 5M,N). These results are consistent with the ex vivo explorations of FMD and demonstrate that endothelial ATG5 is required for endothelial mechanosensing in another important vascular bed.
Endothelial autophagy deficiency exacerbates vascular dysfunction in stressed mice
We next addressed if stress conditions would exacerbate cardiovascular dysfunction triggered by impaired endothelial autophagy.
First, we found that the blood pressure rise in response to angiotensin II was significantly accentuated in endothelium-ATG5 deficient mice relative to WT, while heart rates were similar between the two groups (Fig. 6A-C).
We then tested the flow-mediated outward arterial remodeling. In response to a chronic increase in blood flow, the normal physiological adaptation involves an increase in arterial diameter until shear stress is normalized. Such flow-mediated outward remodeling requires endothelium-mediated dilation. Passive mesenteric arterial diameter in response to increasing intraluminal pressure was measured 1 week after arterial ligation in high flow (HF) and normal flow (NF) arteries isolated from Cdh5.cre-Atg5lox/lox mice and controls. The passive arterial diameter was significantly higher in HF than in NF arteries in control mice, while in Cdh5.cre-Atg5lox/lox mice, the passive arterial diameter was similar in HF and NF arteries (Fig. 6D). Thus, diameter enlargement did not occur in HF arteries from endothelial-selective ATG5 deficient mice, showing that ATG5 is required for flow-mediated outward arterial remodeling.
Next, we analyzed the role of endothelial autophagy in femoral artery responses to wire-induced endothelial denudation and smooth muscle injury. Of note, eNOS signaling was impaired in the femoral arteries of Cdh5.cre-Atg5lox/lox mice at baseline, as shown by decreased phosphorylation of AKT and NOS3 (Supp Fig. 9), indicating a crucial role for ATG5 in the control of NOS3 activity at steady-state in physiological conditions. Arterial wire injury caused the development of concentric neointimal lesions in femoral arteries from Atg5lox/lox control mice. Endothelial ATG5 deficiency increased the neointima formation following wire injury, resulting in increased lesion volume and maximal cross-sectional area (Supp Fig. 9). These results suggest that endothelial autophagy supports the re-endothelialization process following wire injury.
Finally, functional recovery of the heart after myocardial infarction (MI) involves both reperfusion of collateral arteries upon the local increase in blood flow and angiogenesis. Endothelial selective ATG5 deficiency resulted in a significant decrease in the left ventricular shortening fraction 10 days after coronary artery ligation (Supp Table S9). The left ventricular posterior wall width was also significantly smaller in Cdh5.cre-Atg5lox/lox mice, while left ventricular septum size and left ventricle size were similar in Cdh5.cre-Atg5lox/lox mice and Atg5lox/lox littermates (Supp Table S9). Neither the infarct size nor the cardiac fibrosis differed between the groups (Fig. 6E-G), but the number of capillaries per cardiomyocyte was smaller in Cdh5.cre-Atg5lox/lox mice (Fig. 6E,H). Together, these data suggest that while endothelial-selective ATG5 deficiency does not influence normal heart function or tissue injury induced by coronary artery ligation, it impairs functional recovery after MI.