Introduction

Due to their strategic localization, vascular endothelial cells (EC) constantly face oscillating blood glucose concentrations in relation to the pre- and post- prandial cycles1,2,3. However, uncontrolled hyperglycaemia promotes endothelial dysfunction4, a primary event that foreruns atherosclerosis and, therefore, cardiovascular diseases5,6. Physiologically, EC uptake glucose from the blood mainly through the Glucose Transporter 1 (GLUT1) and then use part of it for their own metabolism while delivering the rest to the surrounding tissues7. EC are highly glycolytic since most of the energy they produce derives from glycolysis8. Focusing on Human Umbilical Vein EC (HUVEC), the amount of glucose oxidized in the glycolytic pathway is about 200-fold higher than the glucose oxidized in the tricarboxylic acid (TCA) cycle8. Accordingly, mitochondrial content in EC is lower than in other cells9, thereby indicating that endothelial mitochondria play a role in sensing cell stress and integrating signals from the microenvironment rather than in energy production9,10. Mitochondria are also the headquarters of fatty acid catabolism. EC metabolize fatty acids to form Acetyl-CoA11, used as a fuel to sustain TCA cycle in conjunction with other anaplerotic substrates derived from glucose and/or amino acids, mainly to sustain anabolic pathways and to maintain redox homeostasis through the generation of nicotinamide adenine dinucleotide phosphate hydrogen (NADPH), necessary to reduce glutathione12,13. Moreover, Fatty Acid β-Oxidation (FAO) contributes to the maintenance of endothelial differentiationd-glucose alters mitochondrial shape and promotes the accumulation of lipid droplets. The next step was to investigate mitochondrial function as well as some aspects of endothelial metabolism. In agreement with previous studies utilizing different imaging techniques12, 37, 38, we observed a higher number of round and a lower count of elongated mitochondria in cells cultured in high d-glucose, which indicates an imbalance between fission and fusion39. Accordingly, DRP1, the predominant regulator of mitochondrial fission40, was upregulated and OPA1, a GTPase playing an important role in fusion, was downregulated in HUVEC in high d-glucose. Interestingly, similar results were achieved in EC isolated from diabetic patients that show a lower mitochondrial network than healthy controls41. Increased mitochondrial fission has also been reported in retinal EC exposed to high glucose and linked to the reduction of OPA142. Moreover, in high glucose treated human retinal EC and in the retinal microvasculature of human donors with documented diabetic retinopathy, mitochondrial fusion is impaired because of the hypermethylation of Mfn2 promoter43. Imbalances in fission and fusion seem to be a common response to high glucose as found both in vivo and in vitro. Indeed, augmented mitochondrial fission was detected in skeletal muscle cells from diabetic patients, and associated with low amounts of OPA144. The downregulation of OPA1 was described also in myoblasts and pancreatic β-cells and correlated with insulin resistance45. In murine microvascular cells and podocytes, mitochondrial fission by high glucose was due to ROCK-mediated activation of DRP124. Since an imbalance of fusion and fission not only alters mitochondrial shape but also disrupts their function42, it is noteworthy that HUVEC cultured in high d-glucose accumulate mtROS, which play a role in promoting mitochondrial fission46 together with ROS derived from cytosolic sources47. In fact, there are experimental pieces of evidence showing that NAD(P)H oxidases (NOX)48, cyclooxygenase and nitric oxide synthase47, 49 induce oxidative stress EC in high d-glucose, thereby contributing to endothelial dysfunction. In particular, the study by Gray et al. showed in human aortic endothelial cells exposed to high glucose an increased expression of Nox1, located in the plasma membrane50, accompanied by an augmentation of oxidative stress. Furthermore, within the same study, the deletion of Nox1, but not Nox4, correlated with reduced ROS formation48. These ROS might trigger mtROS production through a "crosstalk" between the plasma membrane and mitochondria, potentially contributing to the amplification of ROS in subcellular compartments essential for the activation of redox signalling50. Although it is widely accepted that mtROS production correlates positively with ΔΨm, the relationships between ΔΨm and mtROS production are not fully understood yet. In fact, opposite correlations between ΔΨm and mtROS production have also been observed in some pathological conditions and mitochondrial disorders51. An increase in mtROS levels may be due to the overproduction and/or the decrease in enzymatic or non-enzymatic antioxidants capable of catalyzing the breakdown or scavenging of these species. Mitochondria of EC contain several potential sources of ROS associated with nutrient oxidation, such as Complex I, Complex II, Monoamine Oxidases, cytochrome c and pyruvate dehydrogenase52. Since we observed that the high level of d-glucose decreased the electrochemical gradient, oxygen consumption and ATP production in HUVEC, in our experimental model the main source of mtROS might be the increased activity of pyruvate dehydrogenase. Accordingly, Nishikawa et al. observed that the inhibition of glycolysis-derived pyruvate transport into mitochondria by 4-hydroxycyanocinnamic acid completely inhibited high glucose-induced ROS production in cultured bovine aortic EC22. In the mitochondria, the two main ROS degrading pathways are the thioredoxin-2 and glutathione systems, both requiring the reducing power of NADPH to carry out their antioxidant activities53. The proton gradient formed by the flow of electrons through the respiratory chain plays a fundamental role in regulating ROS levels because the return of the proton through nicotinamide nucleotide transhydrogenase is necessary for the supply of NADPH to these antioxidant systems53. In addition, high d-glucose induces the upregulation of the thioredoxin interacting protein (TXNIP) that inhibits the antioxidant function of thioredoxin and promotes mtROS accumulation in HUVEC54.

Interestingly, metformin, a mainstay of therapy in diabetic patients and particularly beneficial for the vascular system, mitigates ROS production by inhibiting DRP1-mediated mitochondrial fission in HUVEC55 and retards atherosclerosis in diabetic mice by ameliorating endothelial dysfunction through the reduction of mitochondrial fragmentation and the attenuation of oxidative stress55. We also detected a decline in mitochondrial potential, which is considered a signal of bioenergetic impairment eventually resulting in apoptosis56. However, in our experimental model no signs of apoptosis were observed, in agreement with previous evidence indicating that apoptosis requires longer times to be detected in high glucose-treated HUVEC57. As above mentioned, we also found lower amounts of ATP and decreased O2 consumption in high d-glucose-treated cells, in agreement with recent data in human aortic EC whose mitochondrial function was investigated by respirometry58. In the immortalized endothelial EA.hy926 cell line at least 6 days of culture in high glucose are necessary to detect a statistically significant decrease in basal OCR59. This different kinetics can be ascribed to the marked differences occurring between primary HUVEC and EA.hy926 cells60.

The decrease in O2 consumption could stem from a decrease in the O2 reduction reaction in water catalysed by Complex IV. This could be a consequence of the alteration in the electron transport chain due to the decreased activity of Complex II. It is noteworthy that O2 is also consumed by reactions catalysed by cytosolic enzymes mentioned earlier as well as by processes such as the redox cycle and lipid peroxidation. The reduction of mitochondrial oxidative processes mainly involves glutamine and fatty acids, and represents a trigger to enforce glycolysis52. Consequently, we investigated some aspects of lipid and glucose metabolism in HUVEC exposed to high d-glucose for 24h. We found that GLUT1 is rapidly but transiently upregulated in HUVEC in high glucose, in agreement with previous studies showing that HUVEC chronically exposed to high glucose do not modulate GLUT147. To dispose the overload of glucose, the glycolytic pathway is potentiated, as evidenced by the increase of hexokinase and lactate dehydrogenase59. This results in the overproduction of pyruvate and, consequently, lactate along with the accumulation of glycolytic intermediates that are shunted to the polyol, pentose phosphate and hexosamine pathways, all implicated in the insurgence of endothelial dysfunction1,47. Among the glycolytic by-product, methylglyoxal is implicated in the vascular complications of diabetes, because it plays a role in the formation of advanced glycation end-products and the production of ROS61. Methylglyoxal production during glycolysis has been well documented in different kinds of cells and tissues62. Incubation of vascular smooth muscle cells with 25 mM glucose for 3h increases methylglyoxal production 3.5-fold and enhances oxidative stress. Moreover, 25 mM glucose and methylglyoxal induce endothelial dysfunction in rat aortic rings as well as in cultured rat aortic EC and HUVEC63. Therefore, it is reasonable to assume that even in our experimental model the increase in methylglyoxal may contributed to alter mitochondrial function through the inhibition of Complex II activity and decrease in Δψm64. In the mitochondria, the conversion of excessive pyruvate into acetyl-CoA increases the production of citrate and its exportation in the cytosol as a substrate for ATP citrate lyase, which cleaves citrate to regenerate acetyl-CoA and oxaloacetate. Under conditions of glucose excess, the function of this pathway is to direct acetyl-CoA away from the mitochondria and back to the cytosol for the synthesis of fatty acids and sterols65. As a result, the cells can dispose glucose in excess as triglycerides stored in lipid droplets. However, an excessive increase in triglyceride synthesis can determine a high consumption of NADPH that might alter cellular redox homeostasis and, as a consequence, contribute to a decrease in the activity of the cytosolic glutathione antioxidant system resulting in endothelial dysfunction. It is relevant that lipid droplets were detected in EC lining atheromas66,67 and in arteries from patients bearing a loss of function mutation of ATGL68, a critical enzyme of triglyceride lipolysis. The significance of triglyceride-rich lipid droplets is not clear at the moment. Beyond serving as an energy resource, endothelial lipid droplets also function as a defense mechanism against lipotoxicity67. This mechanism might be applicable also to HUVEC in high d-glucose where ROS are overproduced. Lipid droplets also reduce mitochondrial fragmentation and ROS production69, as upon stress lipid droplets and mitochondria physically interact so that noxious proteins present on the outer mitochondrial membrane can be cleared70. In general, the accumulation of lipid droplets might be interpreted as a compensatory mechanism to dump high glucose-driven storage of triglycerides. The upregulation of ATGL in HUVEC cultured in high glucose sounds intriguing in the light of experiments showing that primary EC lacking ATGL accumulate lipid droplets. Since (i) FFAs are ligands for the lipid sensing nuclear receptor PPAR-γ, (ii) PPAR-γ upregulates ATGL71 and (iii) culture in high d-glucose increases the amounts of PPAR-γ54, we propose that ATGL upregulation is mediated through the FFA-PPAR-γ pathway. ATGL also promotes lipophagy72, and, in HUVEC exposed to high glucose, this might represent an initial step to prevent excessive accumulation of lipid droplets. However, in our experimental setting we did not observe any difference in autophagy, the physiological process which eliminates damaged or senescent organelles. It is possible that this process might require longer exposure times to become activated in HUVEC exposed to high d-glucose. In line with this issue, we propose that mitophagy may not play a significant role in the early remodeling of mitochondrial network in HUVEC cultured in high glucose. In this context, the upregulation of BNIP3 can be envisioned as a mechanism to regulate mitochondrial dysfunction73. Initially, it has a role in reducing respiration, decreasing ΔΨm and ATP synthesis, before potentially serving as a driver of mitophagy. Further studies in kinetics should be performed to test this hypothesis.

Another result that caught our attention is the lack of dose dependence in some responses to high glucose. In particular, Δψm and O2 consumption are significantly and reproducibly lower in cells exposed to 11.1 mM than 30 mM glucose. Moreover, mtROS are higher in HUVEC treated with 11.1 mM than 30 mM glucose. We hypothesize that very high concentrations of glucose activate rapid and robust adaptive mechanisms that limit some detrimental effects of glucose overload, while lower, albeit pathological, concentrations do not grant to reach the threshold of anti-stress defense. In Fig. 6, a schematic representation of results is reported.

Figure 6
figure 6

Schematic representation of the results, created in Biorender.com. GLUT1 Glucose Transporter 1, GLU-6P Glucose-6-phosphate, Glycerol-3P Glycerol-3-posphate, ROS Reactive Oxygen Species, PLIN2 Perilipin-2, LD Lipid droplets, CPT1A Carnitine Palmitoyl Transferase 1A, mtROS mitochondrial ROS, Δψm mitochondrial membrane potential, ATP adenosine triphosphate, OCR oxygen consumption rate, TCA tricarboxylic acid cycle, FAO Fatty acid β-oxidation.

Full size image

In conclusion, we found a relation between the ultrastructural changes of HUVEC exposed to high glucose and their metabolic derangement. More experiments are necessary for a clear definition of the time sequence of these events.

Materials and methods

Cell culture

HUVEC were purchased from the American Type Culture Collection (ATCC Manassas, Virginia, USA) and cultured in medium M199 supplemented with 10% fetal bovine serum, 1 mM l-Glutamine, 1 mM Penicillin–Streptomycin (Euroclone, Milano, Italy), 1 mM Sodium Pyruvate, 5 U/mL Heparin and 150 µg/mL Endothelial Cell Growth Factor on collagen-coated dishes (50 µg/mL) (Sigma-Aldrich, St. Louis, MO, USA). d-glucose was used at concentrations of 11.1 mM and 30 mM and l-glucose (Sigma-Aldrich) was utilized as control of osmolarity (30 mM). 30 mM glucose corresponds to severe hyperglycaemia in diabetic individuals and is used in many studies on EC, whereas 11.1 mM glucose is a pathological concentration rarely used in in vitro experiments.

Sample preparation for Cryo Soft X-ray tomography (Cryo-SXT)

HUVEC were seeded onto gold Quantifoil R 2/2 holey carbon-film microscopy grids at a concentration of 1 × 104 cell/cm2. Cells were incubated at 37 °C in 5% CO2 for 24h with medium containing physiological (5.5 mM, CTR) or high concentrations of d-glucose (11.1 mM and 30 mM). l-glucose was used as control of osmolarity at a concentration of 30 mM. The samples were then gently rinsed twice with phosphate buffered saline (PBS) (Euroclone) and soon thereafter HUVEC were frozen-hydrated by a rapid plunge freezing in liquid ethane bath cooled with liquid nitrogen using a Leica EM GP robot. Excess water was removed before plunge freezing via blotting to obtain a total ice thickness well below 5 µm. Frozen specimens were transferred into the full field soft X-ray transmission microscope of the beamline of the ALBA-Light Source74, where Cryo-SXT tomographic measurements of whole frozen hydrated cells were performed. The cryogenic conditions were maintained during all the experiment.

Cryo soft X-ray tomography

Cryo-SXT images were recorded at the MISTRAL beamline of the ALBA light source, where photons extracted from a bending magnet source are directed on the sample by a capillary condenser facing the monochromator exit slit. Behind the sample, a zone plate with an outermost zone width of 40 nm acts as the objective lens of the microscope, generating a magnified image of the sample on a direct illumination CCD detector74. Cryo-SXT was carried out at 520 eV to optimize the contrast between the carbon-rich organelles membranes and the water-rich cytoplasmic solutions. For each cell, a tilt series was acquired using an angular step of 1° on a 140° angular range. The effective pixel size in the images was 13 nm. No radiation damage was detected at our spatial resolution. Each transmission projection image of the tilt series was normalized using flat-field (incident intensity) images of 1-s acquisition time. The tilt series were manually aligned using eTomo in the IMOD tomography software suite75. With the aim to decrease as much as possible the deviations from an ideal rotation that creates artefacts in the reconstructed tomograms, the rotation of Au fiducials of 150 nm diameter (BBI Solutions—Freiburg, Germany) was followed. According with the Beer–Lambert law, the transmission T(x,y) is given by:

$$T\left(x,y\right)=\frac{I\left(x,y\right)}{{I}_{0}\left(x,y\right)}={e}^{(-\int \mu (x,y,{E}_{0})dt)}={e}^{(-\int {\mu }_{m}(x,y,{E}_{0})\rho d{t}_{m})}$$

where I is the transmitted intensity by the sample; I0 is the incident beam intensity; µ is the X-ray linear absorption coefficient (LAC) at incident energy E0, µm(E0) = µ/ρ is the mass absorption coefficient at the same energy; ρ is the matrix density; x and y are the coordinates in the transversal plane at the sample position and the integral is extended through all the sample thickness. All the transmission tilt series have been converted in absorbance A using ImageJ by applying the following expression:

$$A=\mu \left({E}_{0}\right)t=-ln\left(T\right)$$

The absorbance tilt series were finally reconstructed with TomoJ76, a plugin of ImageJ (National Institute of Health, Bethesda, MD, USA)77, using the ART iterative-algorithms with 15 iterations and a relaxation coefficient of 0.01. Finally, the images were segmented by Amira (Thermo Fisher Scientific, Waltham, MA, USA) and the “Volren” module enables to render the segmented regions at the same time with different colormaps.

Fractional anisotropy

The Fractional Anisotropy (FA) was calculated for all segmented mitochondria and lipid droplets in every cell culture condition. The eigenvalues λ1, λ2 and λ3 were automatically extracted by Amira software and the FA has been calculated by implementing the following formula78:

$$FA=\sqrt{\frac{3}{2}} \frac{\sqrt{\left({\lambda }_{1}-\lambda \right)+\left({\lambda }_{2}-\lambda \right)+\left({\lambda }_{3}-\lambda \right)}}{\sqrt{{\lambda }_{1}^{2}+{\lambda }_{2}^{2}+{\lambda }_{3}^{2}}}$$

Western blot

HUVEC were lysed in 50 mM Tris–HCl (pH 7.4) containing 150 mM NaCl (Sigma-Aldrich), 1% NP40 (Sigma-Aldrich), 0.25% sodium deoxycholate (Sigma-Aldrich), protease inhibitors (10 µg/mL Leupeptin, 10 µg/mL Aprotinin and 1 mM Phenylmethyl-Sulfonyl Fluoride, PMSF) (Sigma-Aldrich), and phosphatase inhibitors (1 mM sodium fluoride, 1 mM sodium vanadate, 5 mM sodium phosphate) (Sigma-Aldrich). Lysates (40 µg/lane) were separated by SDS-PAGE and transferred to nitrocellulose sheets. Western Blot analysis was performed using antibodies against OPA1, DRP1, LC3 B-I/-II, BECLIN (Cell Signalling, Euroclone, Pero, Italy), CYPD, CPT1A, GLUT1 (Thermo Fisher Scientific), BNIP3 (Sigma-Aldrich), p62 (Invitrogen, Carlsbad, CA, USA), ATGL, PLIN2 and mitochondrial oxidative phosphorylation complexes (OXPHOS) (Abcam, Cambridge, UK). Actin (Santa Cruz, Dallas, Texas, USA) was the control of equal loading. After washing, secondary antibodies labelled with horseradish peroxidase (GE Healthcare, Waukesha, WI, USA) were used. Immunoreactive proteins were detected with Clarity™ Western ECL substrate by ChemiDoc MP Imaging System (Bio-Rad). Densitometry of the bands was performed with ImageJ. The Western blots shown are representative and the densitometric analysis was performed calculating the ratio between the protein of interest and actin on three independent experiments ± Standard Deviation (SD).

Real-time PCR

Real-time PCR (RT-PCR) was performed three times in triplicate using the CFX96 Touch Real-Time PCR Detection System (Bio-Rad) exploiting the TaqMan Gene Expression Assay (Life Technologies, Monza, Italy). The following primers were used: Hs02596873_s1 (MT-ND1) and Hs03654441_s1 (RNA45S5) as internal reference gene. Relative changes in gene expression were analysed by the 2−ΔΔCt method.

Mitochondrial membrane potential (ΔΨm) and mtROS production

ΔΨm was quantified by measuring fluorescence intensities of red-shifted aggregates (in functional mitochondria) and green-shifted JC-1 (in damaged mitochondria) monomers to evaluate mitochondrial viability using JC-1 probe (Thermo Fisher Scientific). The cells were incubated with the probe at 37 °C for 10 min, and fluorescence (λex/em red = 575/590 nm, λex/em green = 460/510 nm) was measured using the Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific). The red/green ratio was calculated for each sample79.

mtROS production was measured by MitoSOX™ Red mitochondrial superoxide indicator (Invitrogen). After the treatments in a 96-well plate, the cells were incubated for 10 min at 37 °C with the reagent, protected from light. Fluorescence was measured at λex/em = 510/580 nm using the Varioskan LUX Multimode Microplate Reader.

Extracellular O2 consumption

The Oxygen Consumption Rate (OCR) was measured by the Extracellular O2 Consumption Reagent (Abcam), according to the manufacturer’s instructions. In particular, the assay is based on the ability of oxygen to quench the excited state of the reagent. During the cell respiration, the oxygen is depleted in the surrounding environment increasing the fluorescent signal. After the treatments, the cells were incubated with the Extracellular O2 Consumption Reagent and each well was sealed by adding pre-warmed High Sensitivity mineral oil80. H2O2 was used as positive control. Then the plate was introduced into the Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific), pre-set to 37 °C. The fluorescent signal was measured every 2 min for 180 min at λex/em = 380/650 nm and normalized on the cell number. The results are the mean of three independent experiments ± SD.

ATP quantification

The CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, Wisconsin, USA) was used to determine the quantification of the mitochondrial ATP, according to the manufacturer’s instructions. This assay relies on the properties of a thermostable luciferase that, in the presence of Mg2+, catalyses an oxidative reaction thus producing bioluminescence. Starting from luciferin and the co-factors molecular oxygen and ATP, it is produced oxyluciferin and emitted light. After the treatments, ATP content was measured on permeabilized mitochondria. Thus, the cells were trypsinized, permeabilized, resuspended in isolation buffer (100 mM KCl, 50 mM TRIS, 5 mM MgCl2, 1.8 mM ATP, 1 mM EDTA), and centrifuged at 600g for 10 min at 4 °C. The supernatant was centrifuged at 10,000g for 15 min at 4 °C to allow the sedimentation of mitochondria, which were incubated in CellTiter-Glo Reagent, diluted in culture medium with a 1:1 ratio, for 10 min at room temperature. The luciferase activity was monitored using the Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific). The fluorescent results were normalized on the cell number. The results are the mean of three independent experiments performed in triplicate ± SD.

FAO and triglycerides quantification

FAO was monitored by Fatty Acid Oxidation assay (Abcam) in living cells seeded in a 96-well black plate (Greiner Bio-One). Oleate was the substrate utilized to measure fatty acid-driven oxygen consumption54,81. Cells were treated with high glucose for 24h and, at the end of the experiment, were rinsed twice with 90 μL of pre-warmed FA-Free Measurement Media (containing 150 μM oleate-BSA conjugate) and then added 10 μL of extracellular O2 consumption reagent. Extracellular O2 Consumption Reagent (Abcam) was added into all the wells except for the blank control well. Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 0.625 µM), used as positive control, induces maximal electron transport chain activity by dissipating the mitochondrial membrane potential. Etomoxir (40 µM), an inhibitor of the carnitine transporter CPT1, was used as negative control. At the end of the experiment, the wells were sealed with pre-warmed high sensitivity mineral oil. The Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific) was used (λex/em = 380/650 nm) was measured every 2 min for 180 min54,81.

Triglycerides were quantified using Triglyceride Quantification Kit (Sigma-Aldrich), according to the manufacturer’s recommendations. Triglycerides are hydrolyzed by lipoprotein lipase to glycerol and free fatty acids. Glycerol is then measured by coupled enzyme reactions resulting in the final production of a quinoneimine dye that shows an absorbance at 540 nm. The increase in absorbance at 540 nm is directly proportional to triglyceride concentration of the sample. Fluorescence (λex/em = 535–587 nm) was monitored using the Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific). The fluorescent results were normalized on the cell number54.

The results are the mean of three independent experiments performed in triplicate ± SD.

Lactate quantification

L-Lactate was quantified using the luminescence-based Lactate-Glo™ Assay (Promega), according to the manufacturer’s recommendations. In the presence of NADH, a pro-luciferin reductase substrate is converted to luciferin by reductase, emitting light proportionally to the amount of lactate in the sample. Luminescence was monitored using the Varioskan LUX Multimode Microplate Reader. The results were normalized on the cell number. The results are the mean of three independent experiments performed in triplicate ± SD.

Statistical analysis

All the results are the mean of three independent experiments performed in triplicate ± SD. The data were analysed using one-way ANOVA. The p-values deriving from multiple pairwise comparisons were corrected using the Bonferroni method. The statistical analysis was performed with the software GraphPad Prism. Statistical significance was defined as p-value < 0.05. Regarding the figures, *p < 0.05; **p < 0.01; ***p < 0.001.