Interferon-γ Impairs Human Coronary Artery Endothelial Glucose Metabolism by Tryptophan Catabolism and Activates Fatty Acid Oxidation

The data that support the findings of this study are available from the corresponding author on reasonable request.

No animal experiments were included in this study. Although we used human cells, they were purchased from a vendor, Lonza. They were anonymous with respect to donor, and institutional review board review was not required.

Primary HCAECs (Lonza) were cultured in Endothelial Cell Growth Basal Medium-2 supplemented with EGM-2 Microvascular Endothelial Cell Growth Medium-2 BulletKit without antibiotics unless otherwise specified at 37°C with 5% CO₂. Glucose concentration of this medium was 1.0 g/L (5.5 mM) unless otherwise indicated. Low- and high-glucose media were prepared using the basal medium containing no glucose (Cell Biologics) with the BulletKit supplemented with glucose (Boston Bioproducts) at 2.5 mM and 25 mM, respectively. For studying the effects of tryptophan depletion, confluent HCAECs were incubated with DMEM containing 1.0 g/L glucose and deficient in tryptophan (Thermo Fisher Scientific) with or without exogenous tryptophan supplementation (16.0 mg/L; Sigma) equivalent to the concentration in the standard DMEM formulation. The cells had been prepared commercially from deceased adult male and female donors ranging in age from 46 to 55 years without diabetes or known cardiac diseases whose causes of deaths were noncardiac. All experiments were performed with cells grown to confluence between passages 5 and 7 before being treated with human recombinant IFN-γ (R&D Systems; 1–50 ng/mL) in vitro for the durations specified. For the real-time extracellular flux assays and high-performance liquid chromatography measurements of ATP and ADP (adenosine diphosphate), only passage 6 cells were used. For indoleamine 2,3,-dioxygenase 1 (IDO1) inhibition, cells were cotreated with 1-methyltryptophan (Sigma-Aldrich; 1, 3, and 5 mM) and IFN-γ at 50 ng/mL for 24 h. For poly (ADP-ribose) polymerase (PARP) inhibition, cells were pretreated with 3-aminobenzamide (Sigma-Aldrich; 5 mM) for 1 hour before 24 hours of IFN-γ treatment. For nicotinamide adenine dinucleotide (NAD) precursor supplementation, nicotinamide mononucleotide was added at 0.5 mM to cells pretreated with IFN-γ for 24 hours for an additional 24 hours before measuring total intracellular NAD (NAD[H]). For fatty acid oxidation (FAO) inhibition, 2[6(4-chlorophenoxy)hexyl]oxirane-2-carboxylate (etomoxir; Cayman; 10 μM) was added during the last hour of 24 hours of IFN-γ treatment at 50 ng/mL. For hypoxia studies, cells were treated for 24 hours with 1% O₂, 5% CO₂, with N₂ balance, at 37°C in a hypoxia chamber glove box with or without IFN-γ at 50 ng/mL.

Statistical analyses were performed with GraphPad Prism 8. Data represent mean±SEM from at least 3 independent experiments as specified. Normal distribution was assessed by the Shapiro-Wilk test. Homogeneity of variances was assessed by the F test or Brown-Forsythe test. For normally distributed data, 2-tailed Student t test or Welch’s t test (for unequal variances) was used for 2-group analyses, and 1- or 2-way ANOVA with post hoc Bonferroni, Tukey, or Sidak multiple comparisons test or Welch’s ANOVA test (for unequal variances) with post hoc Dunnett T3 multiple comparisons test was used for multigroup analyses, as specified. For data that deviate from a normal distribution, the Mann-Whitney U test or Wilcoxon matched-pairs signed-rank test was used for 2-group analyses, and the Kruskal-Wallis test with post hoc Dunn multiple comparisons test was used for multigroup analyses. One-sample t test was used to analyze the fold changes in normalized protein densitometric analysis data or intracellular cGMP as specified. P<0.05 was used to define statistical significance.

IFN-γ exposure to confluent, resting HCAECs for 24 hours in vitro in culture medium containing physiological glucose concentration (5.5 mM) resulted in a 20% decrease in basal extracellular acidification rate (P=0.006) measured by the real-time extracellular flux assay, indicating decreased endothelial glycolytic activity (Figure 1A). Consistent with this finding, a 27% decrease in intracellular lactate was observed (P=0.010) (Figure 1B).

HCAECs exposed to IFN-γ had significantly reduced intracellular glucose transport in vitro as demonstrated by a 20% decrease in the fluorescent glucose analogue 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose uptake (Figure 1C and 1D). Similarly, many of the intracellular intermediates in glycolytic pathways measured by liquid chromatography mass spectrometry, including fructose-1,6-bisphosphate, dihydroxyacetone phosphate, glyceraldehyde 3-phosphate, phosphoenolpyruvate, and pyruvate, were significantly decreased after 24 hours of IFN-γ exposure throughout (Figure 1E–1I). Together, these findings indicate that IFN-γ impairs endothelial glycolysis (Figure 1O). Similarly, impaired glycolysis was observed in the absence of serum or growth factor supplementation to the culture medium and, in addition, in low-glucose medium (2.5 mM) (Figure S1).

Given the notable suppression of basal glycolysis, including glucose uptake and pyruvate-to-lactate conversion, we hypothesized that IFN-γ impairs glycolysis by inhibition of multiple glycolytic enzymes at the transcriptional level. Indeed, expression of both SLC2A1 (solute carrier family 2 member 1) and LDHA (lactate dehydrogenase A), the genes encoding the enzymes regulating glucose uptake and pyruvate-to-lactate conversion, respectively, was suppressed after 24 hours of IFN-γ exposure in vitro, each by >40% (P=0.0008 and 0.0286, respectively) (Figure 1J and 1K). LDHA protein expression was similarly decreased (Figure 1L and 1M). In addition, RNA sequencing analysis of HCAECs after 24 hours of IFN-γ exposure was notable for suppression of the expression of the majority of the glycolytic enzymes (Figure 1N), further supporting the conclusion that IFN-γ interferes with endothelial glucose metabolism at the level of transcription (Figure 1O).

Given the significant alterations in endothelial glucose metabolism induced by IFN-γ, we next sought to obtain a more global assessment of the landscape of different cellular metabolic pathways by targeted liquid chromatography mass spectrometry analysis of 144 common intracellular metabolites extracted from HCAECs treated with IFN-γ for 24 hours in vitro (Figure 2A). We observed a decrease in intracellular tryptophan by up to 78% (P<0.0001) (Figure 2B) after 24 hours of IFN-γ exposure with robust antiparallel accumulation of kynurenine (the first major stable metabolite of tryptophan catabolism) by 720% compared with untreated control (P<0.0001) (Figure 2C). We also observed concomitant increases in mRNA and protein expression of IDO1, an IFN-γ–activated enzyme catalyzing tryptophan degradation to kynurenine (Figure 2D and 2E). Increasing 1DO1 activity over time was suggested by a time-dependent increase in the intracellular kynurenine-to-tryptophan ratio (Figure 2F). These findings are consistent with previous reports of tryptophan depletion in noncoronary artery endothelial cells on IFN-γ exposure.^12–14

To understand better the functional consequences of tryptophan depletion on endothelial cells, confluent HCAECs were incubated in DMEM deficient in tryptophan with or without exogenous tryptophan supplementation for 24 hours followed by assessment of changes in endothelial adhesion molecule expression and migratory capacity. Exposure to the tryptophan-deficient medium resulted in a proinflammatory phenotype with robust upregulation of intercellular adhesion molecule (P=0.0005) and vascular cell adhesion molecule (P=0.0083) gene expression (Figure 2G and 2H) and a significant impairment in HCAEC migratory capacity after mechanical (scratch) injury to the endothelial monolayer (Figure 2I and Figure S2).

Kynurenine is an endogenous ligand for the intracellular transcription factor AHR.¹⁵ On ligand- and nonligand-specific activation, AHR translocates to the nucleus, dimerizes with aryl hydrocarbon receptor nuclear translocator (ARNT)/HIF1β (henceforth referred to as ARNT), and binds to the xenobiotic response element (also known as the dioxin response element) to regulate downstream target gene transcription.¹⁶ After IFN-γ treatment and the subsequent intracellular kynurenine surge in HCAECs, we detected an increase in AHR protein expression in endothelial nuclei and nuclear extracts (Figure 3A–3C). This increase in nuclear AHR was accompanied by downstream transcriptional activation of its canonical target genes, such as CYP1B1 (cytochrome P450 family 1 subfamily B member 1) with a >27-fold change in its mRNA expression (Figure 3D). Given that ARNT is a shared binding partner for AHR and HIF1α and necessary for their stabilization and activity as transcription factors,^17,18 we hypothesized that AHR activation and nuclear translocation by the IFN-γ–induced kynurenine increase leads to sequestration of the limited nuclear ARNT and, as a consequence, destabilizes residual HIF1α in normoxia (Figure 3H). Such competition between HIF1α and AHR (activated by an exogenous agonist) has been demonstrated in other endothelial cell types in hypoxia.¹⁹ We further hypothesized that this mechanism likely would explain IFN-γ–induced transcriptional suppression of SLC2A1 and LDHA expression—both well-characterized HIF1 target genes—as well as decreases in the expression of other glycolytic enzymes. We found that IFN-γ treatment of HCAECs for 38 hours in normoxia resulted in significant HIF1α destabilization with a 29% decrease in total nuclear protein expression (P=0.012), whereas total nuclear ARNT remained unchanged (Figure 3C). A similar and significant decrease in nuclear HIF1α was noted with IFN-γ treatment in hypoxia (1% O₂), although the extent of HIF1α destabilization was somewhat less (17%) (data not shown). Coimmunoprecipitation analysis (Figure 3E–3G) comparing the relative quantities of ARNT-bound HIF1α and ARNT-bound AHR in nuclear extracts after the same IFN-γ exposure showed a 30% decrease in ARNT-bound HIF1α/ARNT-bound AHR when compared with untreated controls (P=0.036). Consistent with the decreased nuclear HIF1α availability by IFN-γ, IFN-γ abrogated cobalt chloride–induced hypoxia-responsive element–luciferase reporter activity in HCAECs (Figure 3I) and suppressed another canonical HIF1 target gene expression, VEGFA, by >50% (Figure 3J).

Downstream of IFN-γ–induced HIF1α destabilization in quiescent HCAECs in normoxia, we found global suppression of HIF1 target gene expression (Figure S3 and Table S1). In addition to the significant decreases in the expression of genes encoding almost all of the glycolytic enzymes, other markedly suppressed genes involve broad aspects of endothelial cell survival and function, including cell growth and viability, with a 51% decrease in IGF2 (insulin-like growth factor) and its binding protein (IGFBP) isotypes, as well as a 22% decrease in CDKN1 (cyclin-dependent kinase inhibitor). In addition, there were decreases in the expression of genes involved in vascular remodeling (transforming growth factor β [TGFB]), vasomotor tone (adrenomedullin [ADM], adrenoceptor α 1B [ADRA1B]), thrombosis) serpin family E member 1 [SERPINE1]), and heme (heme oxygenase 1 [HMOX1]) and iron metabolism (transferrin receptor [TFRC]).

To link further IFN-γ–induced activation of IDO1-kyneurenine-AHR pathways and concomitant HIF1α destabilization with IFN-γ–induced suppression of glycolytic enzyme gene expression, we performed a series of targeted gene silencing and pharmacological inhibition studies specific to each pathway. Pharmacological inhibition of IDO1 by 1-methyltryptophan abrogated IFN-γ–induced suppression of SLC2A1 gene expression (Figure 4A) and partially attenuated suppression of LDHA (Figure 4B) in HCAECs. Similarly, gene silencing of AHR attenuated IFN-γ–induced suppression of LDHA gene expression (Figure 4C and 4D and Figure S4). Last, silencing of the von Hippel-Lindau tumor suppressor gene (VHL)²⁰ successfully stabilized nuclear HIF1α protein expression in normoxia (Figure 4E) beyond that of the control siRNA-treated sample and abrogated the IFN-γ–induced suppression of LDHA gene expression (Figure 4F and 4G and Figure S4 and Tables S2 and S3).

Oxidized NAD (NAD⁺) and total NAD(H) are essential elements of cellular bioenergetics and redox homeostasis, serving not only as cofactors for glycolysis but also as important regulators of glycolysis and intracellular signaling (see review²¹). IFN-γ exposure resulted in a significant dose-dependent depletion of intracellular NAD(H) by 20% (P=0.0078) without a significant change in the NAD⁺/NADH ratio (Figure 5A and 5B). These findings were contrary to our initial expectation that the IFN-γ–induced intracellular kynurenine surge would promote de novo NAD⁺ synthesis and increase total intracellular NAD(H). Pretreatment of HCAECs with 3-aminobenzamide, an inhibitor of PARP, completely prevented IFN-γ–induced depletion of intracellular NAD(H) (Figure 5A), suggesting such NAD⁺ depletion is primarily driven by PARP activation. Of note, no significant differences in cell viability were noted between the IFN-γ–treated cells and untreated controls. Supplementation with the NAD⁺ precursor, nicotinamide mononucleotide, also effectively restored the intracellular NAD(H) level after IFN-γ exposure (Figure 5C).

With marked impairment of glycolysis and NAD(H) depletion, we next hypothesized that total intracellular energy balance would be impaired. It is interesting that ATP/ADP and ATP/AMP ratios were unchanged with IFN-γ (Figure 5D and 5E), raising a question about the nature of alternative metabolic pathways and energy source(s) that compensate for impaired glycolysis.

Indeed, despite the inhibition of glycolysis by IFN-γ resulting in a marked depletion of intracellular pyruvate (Figure 1I), a 23% increase in basal mitochondrial oxygen consumption rate (OCR) or oxidative phosphorylation (defined as [basal OCR] – [OCR after antimycin A and rotenone injection]) was noted with IFN-γ (P=0.0032; Figure 6A). This finding was independent of the presence of serum or growth factor supplementation in the basal medium (Figure S1B and S1C). Furthermore, there was a 30% increase in mitochondrial respiration associated with ATP production (ATP-linked OCR, defined as [basal OCR] – [OCR after oligomycin injection]; P=0.0021) (Figure 6B). Overall, there was a significant shift away from glycolysis to oxidative phosphorylation in basal endothelial metabolism after IFN-γ exposure as demonstrated by an increased basal mitochondrial OCR–to–extracellular acidification rate ratio (Figure 6C) and a left and upward shift in the representative extracellular acidification rate–OCR plot (Figure 6D). This finding raised the important issue of the identity of the alternative fuel source(s) that supplies the mitochondrial electron transport chain in the face of impaired glucose metabolism.

In contrast with the global transcriptional suppression of glycolytic enzymes by IFN-γ (Figure 1N), gene expression levels of multiple key enzymes regulating FAO were significantly elevated in HCAECs treated with IFN-γ (Figure 6E). These included up to a 450% increase in acyl-CoA synthetase long-chain family members expression as well as a 57% increase in acyl-CoA dehydrogenase medium-chain expression. For an activated cytosolic long-chain fatty acid to cross the mitochondrial inner membrane for oxidation, transfer of its acyl moiety to carnitine (acyl-carnitine) by CPT1 (carnitine palmitoyl transferase 1) is required. Once in the mitochondrial matrix, the acyl group of the fatty acid is transferred to CoA (fatty acyl-CoA), which is subsequently oxidized to generate ATP. There was a 47% decrease in intracellular butyrylcarnitine, an acyl-carnitine, in IFN-γ–treated HCAECs, as well as a similarly decreasing trend in propionylcarnitine (Figure 6F and 6G). Together with the transcriptional activation of multiple FAO enzymes, such a decrease in precursors to fatty acyl-CoAs was consistent with increased FAO in IFN-γ–treated HCAECs. This metabolic shift from glycolysis to FAO was accompanied by earlier activation of AMP-activated protein kinase by IFN-γ (Figure 6H and 6I).

With pharmacological inhibition of CPT1 (etomoxir), the rate-limiting FAO enzyme, IFN-γ–treated HCAECs could no longer augment basal oxidative phosphorylation (Figure 6J) or ATP-linked mitochondrial OCR (Figure 6K). Similarly, gene silencing of CPT1A attenuated the IFN-γ–induced increase in oxidative phosphorylation (Figure S5A–S5C).

Furthermore, inhibition of FAO disrupted intracellular energy balance in IFN-γ–treated cells as noted by >20% reduction in the ATP/ADP ratio (Figure 6L and 6M), supporting the novel role of FAO in maintaining endothelial intracellular energy balance in the face of IFN-γ–induced impaired glucose metabolism. Of note, no transcriptional activation of the enzymes involved in glutaminolysis or significant changes in intracellular glutamine or glutamate levels were observed with IFN-γ treatment (Figure S5D–S5F), eliminating this amino acid pathway from consideration as an alternate energy source.

Metabolic derangements induced by IFN-γ treatment were accompanied by unfavorable phenotypic changes in HCAECs. Endothelial dysfunction in association with cGMP-NO insufficiency has been well established as an important correlate of heightened cardiovascular risk²² and deemed a precursor to coronary artery atherogenesis.²³ Twenty-five hours of IFN-γ exposure at 50 ng/mL resulted in a 29% reduction in basal intracellular cGMP in HCAECs (P=0.0247; Figure 7A and 7B). This reduction was observed despite increased phosphorylation of the serine 1177 residue of endothelial NO synthase (Figure S6A–S6D), consistent with recent examples of discordance between endothelial NO synthase phosphorylation in human umbilical vein endothelial cells and NO formation.²⁴ Total endothelial NO synthase protein expression did not change (Figure S6E–S6F).

Similar to the effect of tryptophan depletion (Figure 2I and Figure S2), 30 hours of IFN-γ exposure resulted in impaired healing of an endothelial monolayer after mechanical (scratch) injury, a process dependent on endothelial proliferative and migratory function (P=0.0098; Figure 7C–7G) and consistent with previous reports of the antiangiogenic effect of IFN-γ.²⁵

Supplemental Methods

Tables S1–S3

Figures S1–S6

References 59–65