Nitric oxide maintains endothelial redox homeostasis through PKM2 inhibition

Abstract

Decreased nitric oxide (NO) bioavailability and oxidative stress are hallmarks of endothelial dysfunction and cardiovascular diseases. Although numerous proteins are S‐nitrosated, whether and how changes in protein S‐nitrosation influence endothelial function under pathophysiological conditions remains unknown. We report that active endothelial NO synthase (eNOS) interacts with and S‐nitrosates pyruvate kinase M2 (PKM2), which reduces PKM2 activity. PKM2 inhibition increases substrate flux through the pentose phosphate pathway to generate reducing equivalents (NADPH and GSH) and protect against oxidative stress. In mice, the Tyr656 to Phe mutation renders eNOS insensitive to inactivation by oxidative stress and prevents the decrease in PKM2 S‐nitrosation and reducing equivalents, thereby delaying cardiovascular disease development. These findings highlight a novel mechanism linking NO bioavailability to antioxidant responses in endothelial cells through S‐nitrosation and inhibition of PKM2.

Introduction

Endothelial cells are situated at the interface between the blood and the vessel wall and control vascular tone and homeostasis. Nitric oxide (NO) derived from the endothelial NO synthase (eNOS) plays a crucial role in these processes and in the modulation of endothelial cell activation and vascular inflammation (Siragusa & Fleming, 2016; Jamwal & Sharma, 2018). The continued generation of NO by eNOS has long been associated with a healthy vasculature while decreased NO bioavailability, as a consequence of reduced eNOS activity or the reaction of NO with superoxide anions, has been linked with cardiovascular disease (Forstermann et al, 2017). Certainly, a genetic predisposition toward enhanced NO signaling is clearly linked with a reduced risk of developing coronary artery disease, peripheral artery disease, and stroke in humans (Emdin et al, 2018). Early studies have shown that, in addition to anti‐hypertensive properties, eNOS‐derived NO prevents leukocyte adhesion to the vascular endothelium and leukocyte migration into the vessel wall through inhibition of nuclear factor κB to prevent the upregulation of adhesion molecules (Niu et al, 1994; De et al, 1995; Gauthier et al, 1995; Khan et al, 1996; Tsao et al, 1996). However, it is clear that the impact of NO on endothelial cell function extends beyond nuclear factor κB and that numerous endothelial proteins are targeted by the NO generated by eNOS.

A significant portion of the overall biological response to NO can be attributed to the oxidative modification of cysteine residues to form S‐nitrosothiols in a process referred to as S‐nitros(yl)ation (Stamler et al, 2001; Hess & Stamler, 2012). The co‐localization of eNOS with specific target proteins, via either a direct interaction with eNOS or through scaffolding proteins, is an important mechanism to ensure efficient S‐nitrosation (Stomberski et al, 2018). Indeed, the localization of eNOS to the Golgi apparatus or to the nucleus was shown to increase the S‐nitrosation of specific target proteins in the two subcellular compartments (Iwakiri et al, 2006). Of note, although a large number of proteins were shown to be S‐nitrosated in different cell systems, including endothelial cells, the functional consequences of this post‐translational modification have seldom been investigated in detail. As the changes in NO bioavailability that characterize the development of cardiovascular diseases are bound to affect steady‐state protein S‐nitrosation, the aim of this study was to interrogate the eNOS interactome for novel binding partners and potential S‐nitrosation targets that could contribute to cellular redox regulation and to the NO‐mediated protection of the vascular wall against atherogenesis.

Results

Pyruvate kinase M2 interacts with active eNOS

To identify proteins interacting with the active eNOS that may be novel S‐nitrosation targets with relevance to vascular pathophysiology, eNOS complexes were immunoprecipitated from growth factor‐stimulated endothelial cells expressing FLAG‐tagged wild‐type eNOS and subjected to mass spectrometry (Fig 1A, Appendix Table S1). This approach identified well‐known eNOS‐associated proteins, e.g., calmodulin (CaM) and heat‐shock protein (Hsp) family members, proteins whose relationship with eNOS has not been investigated to date, e.g., nuclear and cytoskeletal proteins, and proteins reported to be S‐nitrosated. One protein belonging to the latter group was a glycolytic enzyme, pyruvate kinase M2 (PKM2).

Figure 1. PKM2 interacts with active eNOS .

• A
  Volcano plot highlighting proteins significantly enriched (open circles, FDR < 0.05) in eNOS‐FLAG immunoprecipitates from growth factor‐stimulated human endothelial cells; n = 3 independent cell batches.
• B
  Nitrite in the cell supernatant of HEK293 cells expressing wild‐type (WT), Y657F (YF), or Y657D (YD) eNOS and treated with solvent (Sol) or ionomycin (Io, 1 μmol/l, 15 min); n = 6 independent experiments (2‐way ANOVA and Bonferroni).
• C
  The FMN/FAD ratio measured in eNOS immunoprecipitates from cells expressing WT, YF, or YD eNOS; n = 6 independent experiments (1‐way ANOVA and Newman–Keuls).
• D
  The consequence of mutating Y657 on the growth factor‐induced formation of eNOS complexes with Hsp90 and calmodulin (CaM); n = 6–19 independent cell batches (1‐way ANOVA and Newman–Keuls).
• E, F
  Effect of mutating Y657 on the formation of complexes between eNOS, Hsp90, and PKM2 in growth factor‐stimulated human endothelial cells. Complex formation was assessed in eNOS‐FLAG immunoprecipitates (E), or PKM2 immunoprecipitates (F); n = 4–8 independent cell batches (1‐way ANOVA and Newman–Keuls).

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.Source data are available online for this figure.

To study the consequences of elevated versus attenuated NO production on PKM2 activity, experiments were performed in the presence of the wild‐type, the Tyr657Phe (Y657F), or Tyr657Asp (Y657D) eNOS mutants. These mutations were selected as the phosphorylation of eNOS on Y657 by the redox‐sensitive kinase proline‐rich tyrosine kinase (PYK2) was previously shown to abrogate eNOS enzymatic activity and may therefore contribute significantly to the impaired endothelial function that characterizes cardiovascular diseases. Consistent with a previous report (Fisslthaler et al, 2008), cells expressing the non‐phosphorylatable Y657F‐eNOS mutant consistently generated more NO than the wild‐type enzyme under basal conditions as well as following Ca²⁺ ionophore stimulation (Fig 1B). The phosphomimetic Y657D‐eNOS mutant, on the other hand, was unable to generate NO, an observation attributable to the lack of FMN binding (Fig 1C). In endothelial cells, the enhanced activity of the Y657F‐eNOS mutant was accompanied by an increased association with Hsp90 and CaM, compared to either the wild‐type or Y657D‐eNOS enzymes (Figs 1D and EV1A). In the same conditions, significantly more PKM2 associated with Y657F‐eNOS than with the wild‐type or Y657D‐eNOS proteins (Figs 1E and EV1A). The association between eNOS and PKM2 appeared to be mediated by Hsp90, as both eNOS and Hsp90 could be detected in PKM2 immunoprecipitates from growth factor‐stimulated endothelial cells (Fig 1F). In line with the fact that PKM2 has been identified as an Hsp90 client protein in other cellular systems (Subbaramaiah et al, 2016; Yang et al, 2016), PKM2 and Hsp90 also formed complexes in HEK293 cells in the absence eNOS (Fig EV1B). To determine whether or not the increased association of PKM2 with eNOS was a consequence of the higher eNOS activity, experiments were repeated using a S633D‐eNOS mutant that also displays a higher NO output (Boo et al, 2003). However, the mutation of eNOS on S633 did not alter Hsp90, CaM, or PKM2 binding compared to the wild‐type enzyme (Fig EV1C). Thus, the increased binding of Hsp90 and not an increase in the activity of eNOS per se appears to be instrumental for the increased binding of PKM2 to the Y657F‐eNOS.

Figure EV1. Consequence of eNOS mutation on the assembly of the eNOS signalosome.

1. The consequence of mutating Y657 on the growth factor‐induced formation of eNOS complexes with Hsp90, CaM, and PKM2 in human endothelial cells as summarized in Fig 1. Human endothelial cells expressing GFP were included as control. Complex formation was assessed in eNOS‐FLAG immunoprecipitates.
2. Formation of complexes between PKM2 and Hsp90 in the absence or presence of eNOS in HEK293 cells. Similar results were obtained in three additional experiments.
3. Effect of mutating Ser633 to Asp (SD) on the formation of complexes between eNOS, Hsp90, and PKM2 in growth factor‐stimulated human endothelial cells. Complex formation was assessed in eNOS‐myc immunoprecipitates; n = 4 independent cell batches. Data are presented as mean ± SEM.

Source data are available online for this figure.

Inhibition of PKM2 by S‐nitrosation contributes to endothelial cell antioxidant responses

In human endothelial cells, the activity of the PKM2 that associated with Y657F‐eNOS or wild‐type eNOS was significantly lower than that associated with the phosphomimetic Y657D mutant (Fig 2A). Also in eNOS‐transfected HEK293 cells, PKM2 exhibited the highest activity when associated with Y657D‐eNOS and its activity decreased progressively when bound to the wild‐type, Y657F, and S633D eNOS enzymes generating increasing amounts of NO (Fig 2B). PKM2 can be S‐nitrosated (Wu et al, 2010; Chung et al, 2015; Zhang et al, 2016), making it tempting to speculate that the inhibition of eNOS‐associated PKM2 could be related to its S‐nitrosation. Therefore, endothelial cells were adenovirally transduced to express wild type or Y657F‐eNOS and treated with H₂O₂, to activate PYK2 and elicit eNOS tyrosine phosphorylation (Loot et al, 2009). Under these conditions, PKM2 S‐nitrosation was significantly lower in cells containing tyrosine phosphorylated wild‐type eNOS than in cells expressing the non‐phosphorylatable Y657F‐eNOS mutant (Fig 2C). Thiol oxidation of PKM2 on Cys358 was previously reported to inhibit its activity (Anastasiou et al, 2011), and while a pronounced S‐nitrosation of wild‐type PKM2 could be detected in Y657F‐eNOS‐expressing cells, no signal was detected in cells expressing a C358S‐PKM2 mutant (Fig 2D). Neither the wild‐type enzyme nor the mutant PKM2 was S‐nitrosated in cells expressing Y657D‐eNOS. Importantly, while the activity of the wild‐type PKM2 that associated with Y657F‐eNOS was low, eNOS‐associated pyruvate kinase activity was significantly higher when the cells were transfected with the S‐nitrosation insensitive C358S‐PKM2 mutant (Fig 2E). That a complete rescue was not achieved can most probably be attributed to the fact that the HEK293 cells studied expressed relatively high levels of endogenous PKM2 that could compete with the C358S‐PKM2 mutant for binding to eNOS (see Fig 2D). To assess the link between eNOS and PKM2 in more detail, PKM2 activity was assessed in eNOS immunoprecipitates from HEK293 cells expressing either Y657F‐ or Y657D‐eNOS. In the presence of increasing concentrations of phosphoenolpyruvate, both the V _max and the K_m of the Y657F‐eNOS‐associated PKM2 were significantly lower than for Y657D‐eNOS‐associated PKM2: V _max: 0.08 ± 0.004 versus 0.06 ± 0.003 U/min and K _m: 0.09 ± 0.03 versus 0.03 ± 0.02 mmol/l for PKM2‐Y657D‐eNOS versus PKM2‐Y657F‐eNOS, respectively (Fig 2F).

Figure 2. Inhibition of PKM2 by S‐nitrosation facilitates the accumulation of reducing equivalents to attenuate oxidative stress.

1. Pyruvate kinase (PK) activity in FLAG immunoprecipitates from human endothelial cells expressing FLAG‐tagged wild‐type (WT), Y657F (YF), or Y657D (YD) eNOS; n = 6 independent cell batches (1‐way ANOVA and Tukey).
2. PK activity in eNOS immunoprecipitates from HEK293 cells expressing WT, YF, YD, or S633D (SD) eNOS; n = 5–9 independent experiments (1‐way ANOVA and Bonferroni).
3. S‐Nitrosation of PKM2 in human endothelial cells expressing FLAG‐tagged WT or YF‐eNOS and treated with H₂O₂ (30 μmol/l) for 15 min; n = 8 independent cell batches (Unpaired Student's t‐test). Dithiothreitol (DTT) was included to demonstrate specificity of the SNO signal.
4. PKM2 S‐nitrosation in HEK293 cells expressing YF or YD eNOS and either FLAG‐tagged WT PKM2 or a C358S (CS) PKM2 mutant. SNO‐FLAG‐PKM2 was detected with an anti‐FLAG antibody after biotin switch technique; n = 4 independent experiments (2‐way ANOVA and Bonferroni). DTT treatment was included to demonstrate specificity of the SNO signal.
5. PK activity in eNOS immunoprecipitates from HEK293 cells co‐expressing YF or YD eNOS and either WT PKM2 or the CS PKM2 mutant; n = 6 independent experiments (2‐way ANOVA and Bonferroni).
6. PKM2 enzyme kinetics measured with increasing concentrations of phosphoenolpyruvate (PEP) in eNOS immunoprecipitates from HEK293 cells expressing YF or YD eNOS. The data were fit with the Michaelis–Menten equation to determine V _max and K _m; n = 4 independent experiments (2‐way ANOVA and Bonferroni).
7. Quantification of pentose phosphate pathway (PPP) intermediates; gluconate‐6‐P (G6P), ribulose‐5‐P (Rl5P), ribose‐5‐P (R5P), sedoheptulose‐7‐P (S7P), fructose‐6‐P (F6P) and erythrose‐4‐P (E4P) in human endothelial cells expressing YF or YD eNOS; n = 6–8 independent cell batches (Unpaired Student's t‐test).
8. Link between the pentose phosphate pathway, generation of NADPH, and reduction of GSSG to GSH.
9. NADPH/NADP⁺ and GSH/GSSG ratios in human endothelial cells expressing YF or YD eNOS; n = 6–8 independent cell batches (Unpaired Student's t‐test).
10. NADPH/NADP⁺ and GSH/GSSG ratios in human endothelial cells treated with solvent (Sol) or the PKM2 inhibitor shikonin (SKN, 1 μmol/l) for 45 min; n = 4 independent cell batches (Unpaired Student's t‐test).

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.Source data are available online for this figure.

The inhibition of PKM2 by thiol oxidation is reported to shift glucose catabolism to the pentose phosphate pathway, thereby generating reducing equivalents required for the generation of reduced glutathione (GSH) and the detoxification of reactive oxygen species (Anastasiou et al, 2011). The eNOS‐dependent S‐nitrosation of PKM2 resulted in the same effect as the majority of the intermediates of the pentose phosphate pathway and cellular NADPH/NADP⁺ and GSH/GSSG ratios were significantly higher in endothelial cells expressing the Y657F‐eNOS than the Y657D‐eNOS mutant (Fig 2G–I). Treating endothelial cells with the PKM2 inhibitor shikonin (Chen et al, 2011) resulted in similar increases in the NADPH/NADP⁺ and GSH/GSSG ratios (Fig 2J).

Enhanced NO bioavailability results in PKM2 S‐nitrosation and inhibition in vivo

To study the consequences of preserved NO bioavailability on PKM2 activity under physiological conditions, Tyr656 (mouse sequence; equivalent to Tyr657 in the human sequence) was mutated to phenylalanine to generate knock‐in mice, hereafter referred to as YF‐eNOS mice (Fig EV2). Under basal conditions, the YF‐eNOS mice demonstrated a tendency to lower systolic blood pressure compared with wild‐type littermates (Fig 3A). In isolated rings of endothelium‐intact aortae, phenylephrine‐induced contractions were consistently attenuated in the YF‐eNOS group, a phenomenon that was not observed following the addition of a NOS inhibitor (Fig 3B). In the same samples, the acetylcholine‐induced NO‐dependent relaxation was slightly but significantly potentiated in aortic rings from YF‐eNOS mice (Fig 3C), while responses to sodium nitroprusside were comparable in both groups (Appendix Fig S1). Acetylcholine‐induced relaxations were also significantly greater in the isolated perfused hindlimb circulation of YF‐eNOS compared with wild‐type mice (Fig 3D), a difference no longer observed following NOS inhibition. These results indicate that the mutation of eNOS Tyr656 to Phe resulted in the generation of a mouse model with slightly improved endothelium‐dependent NO generation compared with wild‐type mice. Consistent with the in vitro findings, PKM2 S‐nitrosation was significantly higher in cultured endothelial cells from YF‐eNOS compared with wild‐type mice (Fig 3E). Also, pyruvate kinase activity in eNOS immunoprecipitates from pulmonary endothelial cells (Fig 3F) and hearts (Fig 3G) was significantly lower in the YF‐eNOS than in the wild‐type group. Proximity ligation assays in murine endothelial cells revealed that the association between eNOS and PKM2 occurred mostly in the cytoplasm and in proximity of the Golgi (Fig 3H).

Figure EV2. Generation of the YF‐eNOS mouse.

1. Schematic representation of the murine eNOS gene and of the targeting vector used to introduce the Y656F mutation.
2. Schematic representation of the Y656F eNOS knock‐in allele. The positive selection cassette (puromycin) flanked by F3 sites was deleted by crossing with a ubiquitously expressing FLP1 recombinase mouse strain.
3. The Y656F mutation generates a BsmI restriction site, allowing identification of the wild‐type, heterozygous, and knock‐in alleles by PCR and subsequent BsmI digestion.
4. Genotyping PCR using the primers described in the Materials and Methods section.

Figure 3. Enhanced NO bioavailability results in PKM2 S‐nitrosation and inhibition in vivo .

• A
  Systolic blood pressure (SBP) in wild‐type (WT) and YF‐eNOS (YF) mice; n = 9 animals per group (Unpaired Student's t‐test).
• B, C
  Phenylephrine (PE)‐induced contraction (B) and acetylcholine (Ach)‐induced relaxation (C) of endothelium‐intact aortic rings from WT and YF mice. Experiments were performed in the absence and presence of L‐NAME (LN, 300 μmol/l); n = 6–10 animals per group (2‐way ANOVA and Bonferroni).
• D
  Acetylcholine (ACh)‐induced vasodilatation of the buffer‐perfused hindlimb in situ. Experiments were performed in the absence and presence of L‐NAME (LN, 300 μmol/l); n = 12–13 animals per group (2‐way ANOVA and Bonferroni).
• E
  PKM2 S‐nitrosation in pulmonary endothelial cells from WT and YF mice; n = 5–6 independent cell batches (Unpaired Student's t‐test). Dithiothreitol (DTT) treatment was included to demonstrate specificity of the SNO signal. The blots show non‐adjacent bands cropped from the same membranes.
• F
  PK activity in eNOS immunoprecipitates from WT and YF pulmonary endothelial cells; n = 7–10 independent cell batches (Unpaired Student's t‐test).
• G
  PK activity in eNOS immunoprecipitates from hearts from WT and YF mice; n = 6 mice per group (Unpaired Student's t‐test).
• H
  Representative images of eNOS‐PKM2 interaction (PLA foci, red) in mouse pulmonary endothelial cells; the Golgi apparatus and endosomes were stained with Golph4 (green), the plasma membrane was stained with CD144, and nuclei were highlighted with DAPI (gray). Only rare PLA foci were found in samples incubated with control mouse and rabbit IgGs, demonstrating the specificity of the reaction; n = 3 independent cell batches. Scale bars: 20 μm.

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns = not significant.Source data are available online for this figure.

Preserved NO bioavailability attenuates angiotensin II‐induced endothelial dysfunction through PKM2 S‐nitrosation and enhanced antioxidant responses in vivo

To study the link between eNOS tyrosine phosphorylation and PKM2 activity in a pathophysiological condition, mice were administered angiotensin II via osmotic minipumps, a procedure that increased the tyrosine phosphorylation of eNOS in wild‐type mice (Fig 4A). After 28 days of angiotensin II, hypertension was manifest in wild‐type mice (Fig 4B) and was accompanied by endothelial dysfunction evidenced by attenuated acetylcholine‐induced, NO‐dependent relaxations (Fig 4C, compare with Fig 3C). Both the increase in blood pressure and the change in vascular reactivity were significantly attenuated in YF‐eNOS mice: −log EC₅₀ was 7.20 ± 0.13 versus 7.58 ± 0.08 mol/l for wild‐type versus YF‐eNOS mice (n = 8, P < 0.05, unpaired Student's t‐test). These results indicated that preventing the phosphorylation of eNOS preserved NO bioavailability. Consistent with preserved NO‐dependent relaxation, the S‐nitrosation of PKM2 was higher in lungs from angiotensin II‐treated YF‐eNOS mice than from wild‐type mice (Fig 4D). Importantly, the ratios of NADPH/NADP⁺ and GSH/GSSG were also significantly higher in the same samples (Fig 4E) and the global angiotensin II‐induced oxidative stress (evidenced by the increased 3‐nitrotyrosine footprint) that was clearly evident in wild‐type mice was prevented in YF‐eNOS mice (Fig 4F). Similarly, in cultured endothelial cells the ratios of NADPH/NADP⁺ and GSH/GSSG were higher in cells from YF‐eNOS than from wild‐type mice in the presence of solvent (Fig 4G). Angiotensin II decreased both NADPH and GSH levels in cells from wild‐type mice, while the reducing equivalents were preserved in cells from YF‐eNOS mice. Consistent with these changes, the angiotensin II‐induced increase in H₂O₂ levels was not observed in pulmonary endothelial cells from YF‐eNOS mice (Fig 4H).

Figure 4. Preserved NO bioavailability attenuates angiotensin II‐induced endothelial dysfunction through preserved PKM2 S‐nitrosation and redox homeostasis in vivo .

Wild–type (WT) and YF‐eNOS (YF) mice were treated with vehicle (Veh) or angiotensin II (AII, 1.44 mg/kg/day) for up to 28 days.

1. Phosphorylation of eNOS on Tyr656 in lung lysates from animals treated for 7 days; n = 6–7 mice per group (Unpaired Student's t‐test).
2. Systolic blood pressure (SBP) in WT and YF mice treated for 28 days; n = 10 mice per group (Unpaired Student's t‐test).
3. Acetylcholine (ACh)‐induced relaxation of aortic rings from the same animals as in panel B; n = 8 mice per group (2‐way ANOVA and Bonferroni).
4. PKM2 S‐nitrosation in lungs from WT and YF mice administered AII for 28 days; n = 7–8 mice per group (Unpaired Student's t‐test). DTT treatment was included to demonstrate specificity of the SNO signal. The blots show non‐adjacent lanes from the same membranes.
5. NADPH/NADP⁺ and GSH/GSSG ratios in lungs from WT and YF mice treated with AII for 28 days; n = 8–9 mice per group (Unpaired Student's t‐test).
6. 3‐Nitrotyrosine (3‐NT) levels in lungs from WT and YF mice treated with vehicle (Veh) or AII (AII) for 28 days; n = 7–9 mice per group (2‐way ANOVA and Bonferroni).
7. NADPH/NADP+ and GSH/GSSG ratios in pulmonary endothelial cells from WT and YF mice treated with AII for 30 min; n = 8 cell batches (2‐way ANOVA and Tukey).
8. H₂O₂ levels in pulmonary endothelial cells from WT and YF mice after treatment with AII (1 μmol/l) for 30 min; n = 8 cell batches (2‐way ANOVA and Tukey).

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.Source data are available online for this figure.

PKM2 S‐nitrosation and inhibition are associated with a reduction in oxidative stress and delayed atherogenesis

To determine whether (i) the well‐characterized anti‐atherosclerotic effects of NO could be linked to the S‐nitrosation and inhibition of PKM2 and (ii) the subsequent increase in reducing equivalents detected in vitro could have a biological consequence, YF‐eNOS mice were crossed with apolipoprotein E‐deficient (ApoE^−/−) mice (ApoE^−/−xYF‐eNOS mice, referred to as ApoE^YF mice). These animals were then subjected to partial ligation of the left carotid artery and fed a cholesterol‐rich diet for up to 21 days.

As early as 2 days after ligation, PKM2 S‐nitrosation was significantly reduced in ligated carotid arteries from ApoE^−/− mice (Fig 5A). This was accompanied by a marked decrease in GSH levels (Fig 5B) and a pronounced increase in 3‐nitrotyrosine in the endothelium as well as the smooth muscle cell layers (Fig 5C). In ligated carotid arteries from ApoE^YF mice, the S‐nitrosation of PKM2 was preserved, as were GSH levels. Also, while the 3‐nitrotyrosine footprint clearly increased in ligated arteries from ApoE^−/− mice, the increase, particularly in endothelial cells, was significantly lower in the ApoE^YF group. All of the latter changes are indicative of attenuated oxidative and nitrosative stress and fit well with the observation that a pronounced endothelial dysfunction was detectable as an attenuated NO‐mediated relaxation in ligated carotid arteries from ApoE^−/− mice 7 days after ligation, while endothelial function was maintained in vessels from ApoE^YF mice (Fig 5D). Similarly, an attenuated NO‐dependent relaxation was detected in aortic rings from ApoE^−/− mice, while samples from ApoE^YF mice responded normally (Fig 5E). To assess the importance of PKM2 inhibition for the antioxidant responses of the vessel wall, ApoE^−/− mice were treated with vehicle or the PKM2 inhibitor and subjected to partial ligation of the left carotid artery in combination with a cholesterol‐rich diet. PKM2 inhibition preserved GSH levels in the endothelium, but not in the smooth muscle cell layer of the carotid arteries (Fig 6A). Moreover, the increased 3‐nitrotyrosine footprint evident in the same arteries was significantly reduced by PKM2 inhibition (Fig 6B).

Figure 5. Preserved NO bioavailability and PKM2 S‐nitrosation protect against vascular oxidative stress in vivo .

ApoE^−/− (−/−) and ApoE^YF (YF) mice were subjected to partial ligation of the left carotid artery (LCA) and given a cholesterol‐rich diet. The right, non‐ligated artery (RCA) was used as a control.

1. Proximity ligation assays showing S‐nitrosated PKM2 (yellow) in the RCA and LCA 2 days after ligation. Phalloidin staining (blue) was included to highlight the vessel wall; n = 4–5 mice per group (2‐way ANOVA and Bonferroni). Scale bars: 50 μm.
2. GSH (green) levels in the RCA and LCA 2 days after ligation. CD31 (red) was included to label endothelial cells. n = 4–5 mice per group (2‐way ANOVA and Bonferroni). Scale bars: 50 μm.
3. 3‐nitrotyrosine (3NT, green) levels in the RCA and LCA 2 days after ligation. CD31 (red) was included to label endothelial cells. n = 5 mice per group (2‐way ANOVA and Bonferroni). Scale bars: 50 μm.
4. Acetylcholine (ACh)‐induced relaxation of the phenylephrine contracted LCA 7 days after ligation; n = 7–8 mice per group (2‐way ANOVA and Bonferroni).
5. Acetylcholine (ACh)‐induced relaxation of the phenylephrine contracted aorta 7 days after ligation; n = 7–8 mice per group (2‐way ANOVA and Bonferroni).

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 6. PKM2 inhibition preserves endothelial cell GSH levels and protects against vascular oxidative stress in vivo .

ApoE^−/− mice were treated with vehicle (Veh) or the PKM2 inhibitor shikonin (SKN, 1.2 mg/kg), subjected to partial ligation of the left carotid artery (LCA) and given a cholesterol‐rich diet. The right, non‐ligated artery (RCA) was used as a control.

1. GSH (green) levels in the RCA and LCA 3 days after ligation. CD31 (red) was included to label endothelial cells. n = 5 mice per group (2‐way ANOVA and Bonferroni). Scale bars: 50 μm.
2. 3‐Nitrotyrosine (3NT, green) levels in the RCA and LCA 3 days after ligation. CD31 (red) was included to label endothelial cells. n = 5 mice per group (2‐way ANOVA and Bonferroni). Scale bars: 50 μm.

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

One early consequence of decreased NO production and increased oxidative stress is the expression of endothelial cell adhesion molecules and the diapedesis of neutrophils and monocytes. As expected, the endothelium of the ligated carotid arteries from ApoE^−/− mice expressed ICAM1, and both endothelial cells and vascular smooth muscle cells expressed VCAM1. The expression of ICAM1 and VCAM1 was, however, barely detectable in ligated carotid arteries from ApoE^YF mice (Fig 7A). No differences were observed in the non‐ligated arteries (Appendix Fig S2A). FACS analyses of carotid arteries 7 days after ligation revealed a significantly attenuated infiltration of immune cells, particularly neutrophils and monocytes, into arteries from ApoE^YF compared with ApoE^−/− mice (Fig 7B–D). The numbers of other infiltrated immune cell subsets and the overall number of circulating immune cells did not differ significantly between the two strains (Appendix Fig S2B and C). Consistent with the published anti‐inflammatory effects of shikonin (Andujar et al, 2010; Yang et al, 2014; Alves‐Filho & Palsson‐McDermott, 2016; Shirai et al, 2016), PKM2 inhibition abrogated the IL‐1β‐induced expression of ICAM1 and VCAM1 in human endothelial cells (Fig 7E). Also, macrophage infiltration was significantly lower in carotid arteries from shikonin‐treated than vehicle‐treated ApoE^−/− mice (Fig 7F), even though neutrophil infiltration was higher (Fig 7G). Other infiltrated immune cell subsets were similar between the two groups (Appendix Fig S3). In keeping with the maintained endothelial function and preserved antioxidant responses observed at early stages, plaque development assessed 21 days after carotid artery ligation was significantly reduced in regions proximal to the aortic arch in arteries from ApoE^YF compared with ApoE^−/− mice (Fig 8A). A protective effect was also evident in carotid arteries from ApoE^−/− mice treated with shikonin (Fig EV3). In a second atherosclerosis model, ApoE^−/− and ApoE^YF mice were fed a Western diet for up to 6 months. Consistent with the findings in the carotid arteries, plaque area was significantly reduced in the aortic arch in ApoE^YF versus ApoE^−/− mice after 4 months (Fig 8B). After 6 months, the total plaque area in the aortic arch and aorta of ApoE^−/− mice was clearly increased, but the ApoE^YF vessels harbored significantly smaller plaques (Fig 8C).

Figure 7. Preserved eNOS activity and PKM2 inhibition exert anti‐inflammatory effects.

ApoE^−/− (−/−) and ApoE^YF (YF) mice were subjected to partial ligation of the left carotid artery (LCA) and given a cholesterol‐rich diet.

• A
  ICAM1 and VCAM1 (green) expression in the LCA 2 days after ligation. CD31 (red) was included to label endothelial cells, and nuclei are highlighted with DAPI (gray). Comparable images were obtained in four additional animals in each group. Scale bars: 50 μm.
• B–D
  Percentage of neutrophils and monocytes infiltrated into the LCA 7 days after ligation and high‐cholesterol diet feeding; n = 9–10 mice per group (Unpaired Student's t‐test).
• E
  Representative images and quantification of ICAM1 (green) and VCAM1 (red) staining in human endothelial cells pre‐treated with the PKM2 inhibitor shikonin (SKN, 1 μmol/l) or solvent (Sol) for 45 min and then stimulated with interleukin‐1β (IL‐1, 20 ng/ml) or solvent (CTL) for 90 min. Nuclei are highlighted with DAPI (gray). n = 6 independent cell batches (2‐way ANOVA and Bonferroni). Scale bars: 20 μm.
• F, G
  Percentage of neutrophils and macrophages infiltrated into the LCA from ApoE^−/− mice treated with Veh or SKN (1.2 mg/kg) 7 days after ligation and high‐cholesterol diet feeding; n = 4–6 mice per group (Unpaired Student's t‐test).

Data information: Data are presented as mean ± SEM. *P < 0.05, ***P < 0.001.

Figure 8. Preserved NO bioavailability delays atherosclerosis development.

• A
  Plaque burden in ligated carotid arteries from ApoE^−/− (−/−) and ApoE^YF (YF) mice fed a high‐cholesterol diet for 21 days; scale bars: 2 mm. Intimal and medial thickness was determined at four specific locations (proximal, middle 1, middle 2, and distal) in cryo‐sectioned arteries stained with Oil Red O; scale bars: 200 μm. n = 7 mice per group (Unpaired Student's t‐test).
• B, C
  Oil Red O staining of plaques in aortae from mice fed a Western diet for four (B) or 6 months (C); scale bars: 3 mm. n = 6 (B) or n = 12 (C) mice per group (Unpaired Student's t‐test).

Data information: Data are presented as mean ± SEM. *P < 0.05, **P < 0.01.

Figure EV3. PKM2 inhibition delays atherosclerosis development.

Plaque burden in ligated carotid arteries from ApoE^−/− mice fed a high‐cholesterol diet and treated with vehicle (Veh) or the PKM2 inhibitor shikonin (SKN, 1.2 mg/kg) for 21 days; scale bars: 2 mm. Intimal and medial thickness was determined at four specific locations (proximal, middle 1, middle 2, and distal) in cryo‐sectioned arteries stained with Oil Red O; scale bars: 200 μm. n = 5 mice per group (Unpaired Student's t‐test). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01.

Discussion

The results of the present investigation identify a novel mechanism that contributes to protective effects of NO in the vasculature, i.e., the S‐nitrosation and inhibition of eNOS‐associated PKM2. PKM2 makes a significant contribution to the homeostasis of the vascular wall by helping to maintain the levels of the reducing equivalents GSH and NADPH, thus protecting against oxidative stress. Indeed, increasing NO output, by preventing the inhibition of eNOS activity that results from its phosphorylation on Tyr657 (Tyr 656 murine sequence), decreased PKM2 activity and increased NADPH/NADP⁺ and GSH/GSSG ratios in vitro and in vivo. The latter effects were also linked with attenuated adhesion molecule expression and immune cell infiltration as well as the development of endothelial dysfunction that characterizes the early stages of vascular disease. All of these changes culminated in a decrease in atherosclerosis in two different models.

Cellular oxidative stress is determined by the balance between the activity of free radical‐generating enzymes and the levels of cellular antioxidants. GSH is a major antioxidant and has multiple roles in protecting cells against oxidative stress by acting as a donor of reducing equivalents in redox reactions to result in the generation of GSSG. The activity of the GSH reductase regenerates GSH from GSSG using NADPH as an electron donor; for reviews, see Ref. (Diaz‐Vivancos et al, 2015; Handy & Loscalzo, 2017), but endothelial cell activation can increase the generation of reactive oxygen species to the extent that the reductase cannot compensate for the oxidation of GSH. The resulting oxidative stress has numerous consequences for cell function and is linked with the scavenging of NO, to generate the more potent oxidant peroxynitrite and promote the development of cardiovascular diseases (Forstermann et al, 2017). Although it is widely appreciated that NO has vascular protective effects, it was only relatively recently that a genetic predisposition toward enhanced NO signaling in humans could be clearly linked with a reduced risk of vascular disease (Emdin et al, 2018). One problem in studying the long‐term protective role of eNOS in in vivo models of disease is that oxidative stress decreases NO bioavailability as well as eNOS activity. In addition to the well‐studied phenomenon of eNOS uncoupling (Forstermann et al, 2017), oxidative stress can elicit the inactivation of eNOS by virtue of the activation of the redox‐sensitive kinase PYK2, which phosphorylates eNOS on Tyr657 (Fisslthaler et al, 2008). Tyr657 is located in the FMN binding domain of eNOS, which is activated by CaM to move flexibly and bridge the distance between the reductase and oxygenase domains to support electron transfer within the eNOS homodimer and NO production (Volkmann et al, 2014). Comparison of the structure of eNOS with that of neuronal NOS indicated that the phosphorylation of a tyrosine residue within the FMN binding domain (Tyr889, rat nNOS sequence) could prevent its movement, thereby locking the FMN domain into its electron‐accepting position and resulting in enzyme inhibition (Garcin et al, 2004). Tyr657 is the equivalent tyrosine residue in the human eNOS sequence, and indeed, it was possible to demonstrate that its phosphorylation was incompatible with FMN binding, which in turn accounts for the previously reported lack of activity of the tyrosine phosphorylated eNOS enzyme (Fisslthaler et al, 2008). Accordingly, the phosphomimetic Y657D‐eNOS mutant was completely inactive, while the non‐phosphorylatable Y657F‐eNOS mutant was characterized by slightly enhanced NO output versus the wild‐type enzyme. Given this information, we speculated that YF‐eNOS mice would be insensitive to PYK2‐dependent inactivation and be an appropriate model to study the consequences of a maintained physiologically relevant level of NO generation in models of cardiovascular disease. Indeed, YF‐eNOS mice demonstrated a slight improvement in NO generation in isolated arteries as well as in the hindlimb circulation and were protected against both the hypertension and the endothelial dysfunction induced by angiotensin II. Moreover, the preserved NO bioavailability in YF‐eNOS mice crossed on to the ApoE^−/− background attenuated adhesion molecule expression and cell infiltration into the vessel wall, as well as the subsequent development of atherosclerosis.

To identify novel mechanisms linking eNOS and NO with vascular homeostasis, eNOS interacting proteins were co‐precipitated with the active enzyme from human endothelial cells. This approach revealed the association of eNOS with two well‐characterized components of its signalosome, i.e., CaM and Hsp90, as well as a number of proteins not previously known to interact with eNOS, among them PKM2. PKM2 is an intriguing target as it catalyzes one of the final steps of glycolysis, converting phosphoenolpyruvate and ADP to pyruvate and ATP. It has been widely studied in cancer, as its expression is responsible for the Warburg effect (Christofk et al, 2008). While in numerous cell types both the PKM1 and PKM2 isoforms are expressed, only PKM2 is abundantly expressed in endothelial cells, where it contributes to vascular integrity and endothelial growth (Kim et al, 2018; Stone et al, 2018). Another point that made PKM2 particularly interesting is that it can be S‐nitrosated and has been identified as a putative target of thioredoxin‐1 S‐transnitrosation (Wu et al, 2010; Chung et al, 2015; Zhang et al, 2016). However, although proteomic approaches have been used to document the S‐nitrosation of PKM2 and numerous additional proteins under various conditions, no study to date has performed a detailed characterization of the actual impact of this post‐translational modification. In this study, we were able to demonstrate the eNOS‐dependent S‐nitrosation of PKM2 in cultured endothelial cells as well as in murine tissues. Importantly, the extent of PKM2 S‐nitrosation was closely linked to the ability of eNOS to generate NO, and the S‐nitrosation of PKM2 was decreased in vitro and in vivo when eNOS activity was impaired, i.e., following the induction of endothelial dysfunction. Cys358, which is situated in a β strand of the PKM2 protein and essential for PKM2 catalytic activity (Anastasiou et al, 2011), was identified as the NO target as its mutation to serine prevented the S‐nitrosation of the kinase. Given that the S‐nitrosation of eNOS‐associated PKM2 resulted in its inhibition, studies were repeated using the PKM2 inhibitor, shikonin. Consistent with its published anti‐inflammatory actions in immune cells and in the activation of nuclear factor κB as well as the expression and secretion of pro‐inflammatory mediators (Andujar et al, 2010; Yang et al, 2014; Alves‐Filho and Palsson‐McDermott, 2016; Shirai et al, 2016), shikonin prevented the IL‐1β‐induced increase in adhesion molecule expression in cultured endothelial cells. Moreover, when given to mice in vivo, shikonin prevented the accumulation of macrophages in the ligated carotid artery and the subsequent development of atherosclerotic plaques. The reason underlying the increase in neutrophil accumulation in the shikonin‐treated group is currently unknown.

How could the association of PKM2 with eNOS and its inhibition by S‐nitrosation have an impact on endothelial cell function and redox stress? When cells are exposed to oxidative stress, cellular reducing equivalents, i.e., reducing peptides and proteins as well as enzymatic antioxidants, are oxidized and eventually depleted. The inhibition of PKM2 is reported to shunt glucose catabolism to the pentose phosphate pathway which plays a key role in protecting against oxidative stress by oxidizing glucose 6‐phosphate to ribose 5‐phosphate and generating NADPH. This NADPH is then used by the GSH reductase to regenerate GSH from GSSG, which is used as a substrate by the glutaredoxin cycle for the reduction of oxidized proteins and by the glutathione peroxidase cycle to reduce hydrogen peroxide to water (Monteiro et al, 2008; Espinosa‐Diez et al, 2015; Forstermann et al, 2017). It therefore follows that increasing flow through the pentose phosphate pathway should protect against the development of endothelial cell dysfunction and atherosclerosis. There is no doubt that NO protects against the development of vascular disease (Emdin et al, 2018), but to date it has not been possible to link maintained eNOS activity with levels of essential reducing equivalents. In this study, it was possible to demonstrate that enhancing/preserving NO production, by mutating Tyr657 in eNOS, maintained the S‐nitrosation of PKM2 and decreased its activity, thus preserving NADPH/NADP⁺ and GSH/GSSG ratios in vitro and in vivo. Overall, this resulted in an attenuation of the endothelial dysfunction elicited by angiotensin II as well as plaque formation in two models of atherosclerosis. In addition to its role in the regeneration of GSH, increased NADPH levels could be expected to have consequences on the activity of NADPH oxidases and superoxide anion generation, and thus contribute to oxidative stress. However, the elevated NADPH/NADP⁺ ratio was coincident with a decrease in oxidative stress and the preservation of endothelial function, indicating that the increased levels of NADPH were not used to fuel reactive oxygen species generation. The NO‐dependent regulation of PKM2 is a mechanism with wide reaching physiological and pathophysiological implications as during the preparation of this manuscript an analogous mechanism was reported to protect against acute kidney injury (Zhou et al, 2019). In the latter study, a mammalian denitrosylase was identified and deleted in mice, resulting in increased global S‐nitrosation, protection against acute kidney injury, and improved survival—all of which was eNOS‐dependent. With the aid of unbiased proteomic and metabolomic approaches, it was possible to demonstrate that the S‐nitrosation and inhibition of PKM2 elicited a metabolic switch toward the pentose phosphate pathway to subsequent enhance redox protection. Given the number of proteins that associate with eNOS in endothelial cells, the S‐nitrosation of PKM2 is unlikely to be the only mechanism responsible for the well‐established protective effect of NO against the development of cardiovascular disease, but both our study and that by Zhou et al indicate that it is a key one.

Taken together, the S‐nitrosation and inhibition of PKM2 is a newly discovered mechanism that contributes to endothelial cell antioxidant responses. As such, it makes a substantial contribution to the protective effects of NO against the occurrence of endothelial dysfunction, vascular inflammation, and atherosclerosis.

Materials and Methods

All data and methods supporting the findings of this study are available within the article, its Expanded View and Appendix. The mass spectrometry data from this publication have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez‐Riverol et al, 2019) partner repository with the dataset identifier PXD013755.

Animals

C57/BL6 mice were purchased from Charles River Laboratories (Sulzfeld, Germany), and eNOS Tyr656Phe knock‐in mice (introduction of an Y656F mutation into eNOS, exon 16, so‐called YF‐eNOS mice) were generated by TaconicArtemis (Cologne, Germany) using C57/BL6 embryonic stem cells for gene targeting (Fig EV2). For atherosclerosis studies, YF‐eNOS mice were crossed back onto an ApoE^−/− background (ApoE^−/− mice were originally purchased from Charles River Laboratories, Sulzfeld, Germany) for nine generations. Male mice homozygous for the eNOS mutation were used for the experiments with age‐matched ApoE^−/− mice. For studies involving shikonin administration, ApoE^−/− mice were purchased from Charles River Laboratories.

Animals were housed in compliance to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health under a 12‐h light–dark cycle with free access to water and a normal chow diet, if not described differently. Both the University Animal Care Committee and the Federal Authority for Animal Research at the Regierungspräsidium Darmstadt (Hessen, Germany) approved the study protocol (F28/42 and FU/1231). Studies were performed using age‐, gender‐, and strain‐matched animals throughout. For the isolation of organs, mice were sacrificed using cardiac puncture in deep inhalation anesthesia or terminal inhalation anesthesia.

Partial carotid artery ligation

Male mice (8–10 weeks of age) were studied, and partial carotid ligation was performed as described (Nam et al, 2009).

Detection of SNO‐PKM2 by modified in situ proximity ligation assay

SNO‐PKM2 was labeled in frozen sections of carotid arteries using the biotin switch technique followed by the PLA probe protocol—Duolink in situ Solutions, with minor modifications.

Pyruvate kinase activity assay

All chemicals were purchased from Sigma‐Aldrich. Pyruvate kinase activity was measured in FLAG/eNOS immunoprecipitates by monitoring pyruvate‐dependent conversion of NADH to NAD⁺ by lactate dehydrogenase (LDH) at room temperature.

S‐Nitrosation

All chemicals were purchased from Sigma‐Aldrich. Iodoacetyl‐PEG2‐biotin and streptavidin were purchased from Thermo Fisher Scientific. S‐Nitrosation was detected using a modified biotin switch technique.

Statistics

Results are presented as mean ± SEM. GraphPad Prism software (versions 5 and 7) was used to assess statistical significance. Differences between two groups were compared by unpaired two‐tailed Student's t‐test, while differences between three groups or more were compared by one‐way ANOVA followed by Newman–Keuls posttest. All experiments in which the effects of two variables were tested were analyzed by two‐way ANOVA followed by Bonferroni posttest. Differences were considered statistically significant when P < 0.05.

Author contributions

MS, JT, S‐IB, BL, AEL, KS, VR, KK, BB, AW, and BF performed experiments and analyzed the data. MS, IW, and JH, analyzed the eNOS interactome. HS, analyzed pentose phosphate pathway intermediates. JT, MS, and IF designed the in vivo research. MS and IF designed the in vitro research and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Appendix Table S1

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

The EMBO Journal (2019) 38: e100938