Abstract

   Reactive oxygen species (ROS) are important mediators of both
   physiological and pathophysiological signal transduction in the
   cardiovascular system. The effects of ROS on cellular processes depend
   on the concentration, localization, and duration of exposure. Cellular
   stress response mechanisms have evolved to mitigate the negative
   effects of acute oxidative stress. In this study, we investigate the
   short-term and long-term metabolic and transcriptomic response of human
   umbilical vein endothelial cells (HUVEC) to different types and
   concentrations of ROS. To generate intracellular H[2]O[2], we utilized
   a lentiviral chemogenetic approach for overexpression of human D-amino
   acid oxidase (DAO). DAO converts D-amino acids into their corresponding
   imino acids and H[2]O[2]. HUVEC stably overexpressing DAO (DAO-HUVEC)
   were exposed to D-alanine (3 mM), exogenous H[2]O[2] (10 µM or 300 µM),
   or menadione (5 µM) for various timepoints and subjected to global
   untargeted metabolomics (LC-MS/MS) and RNAseq by MACE (Massive analysis
   of cDNA ends). A total of 300 µM H[2]O[2] led to pronounced changes on
   both the metabolic and transcriptomic level. In particular, metabolites
   linked to redox homeostasis, energy-generating pathways, and nucleotide
   metabolism were significantly altered. Furthermore, 300 µM H[2]O[2]
   affected genes related to the p53 pathway and cell cycle. In
   comparison, the effects of menadione and DAO-derived H[2]O[2] mainly
   occurred at gene expression level. Collectively, all types of ROS led
   to subtle changes in the expression of ribosomal genes. Our results
   show that different types and concentration of ROS lead to a different
   metabolic and transcriptomic response in endothelial cells.

   Keywords: endothelial cells, metabolomics, RNAseq, reactive oxygen
   species, D-amino acid oxidase

1. Introduction

   The vascular system is entirely lined by a single layer of endothelial
   cells (EC), which facilitate the exchange of nutrients and oxygen
   between the blood and surrounding tissues. Angiogenic endothelial cells
   migrate into hypoxic tissue, and endothelial nitric oxide (NO)
   production inhibits complex IV of the respiratory chain. Endothelial
   ATP production therefore highly relies on glycolysis instead of
   oxidative phosphorylation. In turn, EC have a relative small
   mitochondrial volume [[40]1,[41]2]. The endothelial production of ROS
   is triggered by oxidants produced from activated immune cells,
   cytokines, or physical stimuli, such as oscillatory flow [[42]3,[43]4].
   Since ROS can lead to oxidative stress, various anti-oxidant defense
   mechanisms have evolved, such as detoxifying enzymes, redox-sensitive
   gene expression, and a dynamic metabolic response.

   In the context of redox signaling, ROS can be produced and act in a
   confined space in a compartmentalized and controlled manner [[44]5]. It
   is well accepted that not only the concentration, but also the type of
   ROS determines the biological effects as oxidants such as H[2]O[2] and
   superoxide have different chemical properties. H[2]O[2] can selectively
   oxidize low pKa thiols and transition metals in proteins [[45]6,[46]7]
   whereas O[2]^•− reacts preferentially with iron sulfur clusters and
   nitric oxide [[47]8,[48]9,[49]10,[50]11,[51]12]. In response to an
   acute ROS challenge, changes in metabolism can be expected to precede
   changes in gene expression. This is due to the fact that ROS-dependent
   inactivation of metabolic enzymes has immediate consequences on the
   concentration of the upstream and downstream metabolites of the
   affected enzyme. However, the metabolic and transcriptomic response of
   EC to different types and concentrations of ROS has not been studied in
   great detail. Appropriate tools to study the complex cellular response
   to intracellular oxidants, particularly over the course of time, have
   been lacking. Conclusions regarding the biological role of H[2]O[2] in
   signaling are based on findings generated by the addition of H[2]O[2]
   to cultured cells. However, this does not reflect the dynamics of
   intracellular H[2]O[2] flux in the regulation of signaling events
   [[52]13,[53]14,[54]15].

   To compare the metabolic and transcriptomic effects of intracellularly
   generated H[2]O[2] to other types of ROS, we utilized a chemogenetic
   approach based on overexpression of human D-amino acid oxidase (DAO).
   DAO oxidizes D-amino acids to their corresponding imino acids and
   H[2]O[2]. The imine is then non-enzymatically hydrolyzed to its
   corresponding α-keto acid [[55]13,[56]16,[57]17,[58]18]. The enzyme is
   stereospecific for D-amino acids (e.g., D-alanine) which allows for
   precise activation of the enzyme. This makes DAO a useful tool to study
   how the intracellular production of H[2]O[2] modulates cellular
   response. With this tool in hand, we analyzed the different metabolic
   and transcriptomic responses of HUVEC to DAO-derived H[2]O[2] in
   comparison to exposure to exogenous H[2]O[2] (10 µM and 300 µM) or
   menadione (to generate intracellular O[2]^•−) in a time dependent
   manner.

2. Materials and Methods

2.1. Chemicals

   All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA)
   unless otherwise indicated.

2.2. Cell Culture

   Human umbilical vein endothelial cells (HUVEC) were purchased from
   Lonza (CC-2519, Lot No. 371074, 369146, 314457, 192485, 186864, 171772,
   Walkersville, MD, USA). HUVEC were cultured in endothelial growth
   medium supplemented with human recombinant epidermal growth factor
   (EGF), Endo-CGS-Heparin (PeloBiotech, Planegg, Germany), 8% fetal calf
   serum (FCS, #S0113, Biochrom, Berlin, Germany), penicillin (50 U/mL)
   and streptomycin (50 μg/mL) (#15140-122, Gibco/Lifetechnologies,
   Carlsbad, CA, USA). For each experiment, at least three different
   batches of HUVEC at passage 4 were used.

   Human embryonic kidney 293 cells (HEK293) cells were obtained from ATCC
   (Manassas, VA, USA) and Lenti-X 293T cells for virus production were
   purchased from Takara (#632180, Takara, Japan). Both cell lines were
   cultured in Dulbecco’s Modified Eagle Medium High Glucose (DMEM High
   Glucose, Gibco, Carlsbad, CA, USA) supplemented with FCS (8%),
   penicillin (50 U/mL), and streptomycin (50 μg/mL) (#15140-122, Gibco/
   Lifetechnologies, Carlsbad, CA, USA). All cells were cultured in a
   humidified atmosphere (5% CO[2], 37 °C).

2.3. Immunofluorescence

   Cells were seeded on 8-well immunofluorescence plates (Ibidi,
   Gräfeling, Germany). At 80% confluence, cells were washed with PBS,
   fixed with 4% paraformaldehyde and permeabilized with 0.05% Triton
   X-100. After a blocking step with 3% BSA (bovine serum albumin) for 30
   min, cells were incubated at 4 °C overnight with a 1:200 dilution of
   the primary antibody against human D-amino acid oxidase (#ab196563,
   abcam, Cambridge, UK). Cells were washed with 0.3% Tween 20 in PBS and
   incubated with a 1:500 dilution of the secondary antibody (Rabbit IgG
   (Alexa Fluor 647, #[59]A31573, Invitrogen, Carlsbad, CA, USA) for 30
   min. The cells were then washed and counterstained with
   4′,6-diamidino-2-phenylindole (DAPI). Images were captured with a laser
   confocal microscope LSM800 (Zeiss, Jena, Germany) and analyzed with ZEN
   lite software (Zeiss, Jena, Germany).

2.4. Cloning of pLVX2 hDAO for Lentiviral Overexpression

   The pLVX2-CIBN-GFP-CAAX PuroRes [[60]19] vector was a gift from X.
   Trepat (Barcelona, Spain). Human DAO 10xHis-Tag was amplified from
   pCMV6-h-DAO (#HG13372-CH, Sino Biological Inc., Beijing, China) with
   the following primers: 5′- ACA CCT TCG AAA TGC GTG TGG TGG TG -3′ and
   5′- ACA CCG CGG CCG CTT AGT GAT GGT GGT G-3′. Both the vector and the
   PCR product of hDAO, were digested with BstBI/NotI (#FD0124, #FD0593,
   ThermoFisher, Dreieich, Germany), purified by gel extraction, and
   ligated (#15224025, T4 DNA ligase, ThermoFisher, Dreieich, Germany) as
   per the manufacturer’s instructions. The final plasmid (pLVX2 hEF1α
   human DAO 10xHis-Tag) was purified and sequenced.

2.5. Lentiviral Transduction

   Pseudotyped lentiviruses were produced by transfecting LentiX cells
   with pLVX2- hDAO-10xHis-Tag together with lentiviral packaging plasmids
   (#12260 and #12259, Addgene, Watertown, MA, USA) as described
   previously [[61]20] using PEI (polyethyleneimine). Viral supernatants
   were collected, filtered, and snap-frozen on the third day after
   transfection. HUVEC (but also HEK293 cells as control, HEK-DAO) were
   transduced with hDAO-10xHis-Tag viral particles for 24 h and then
   selected with puromycin (2 µg/mL) for 7–10 days.

2.6. Determination of ROS Production

   ROS production was measured as described previously [[62]21] using a
   luminol-based chemiluminescence assay. In brief, luminol (100 μM)
   horseradish peroxidase (HRP, 1 unit/mL) coupled chemiluminescence was
   measured in a Berthold 6-channel luminometer (LB9505, Berthold,
   Wildbad, Germany). All measurements were performed in HEPES-Tyrode
   buffer containing in mmol/L: 137 NaCl, 2.7 KCl, 0.5 MgCl[2], 1.8
   CaCl[2], 5 glucose, 0.36 NaH[2]PO[4], 10 HEPES. PEG-catalase (500 U/mL)
   was directly added to the sample during the measurement.

   H[2]O[2] was also measured with the Amplex Red^®/HRP (50 μm,
   Invitrogen, Carlsbad, USA; HRP, 2 units/mL) fluorimetric assay as
   previously described [[63]7,[64]22]. Fluorescence was determined in a
   microplate reader (excitation 530 nm, emission 590 nm) and normalized
   to the protein amount as determined by the Bradford protein assay. To
   calculate the catalase-sensitive portion of the signal, PEG-catalase
   (500 U/mL) was added to the assay buffer 30 min before starting the
   assay.

2.7. Redox Western Blots

   For peroxiredoxin western blots, cells were exposed to H[2]O[2],
   D-alanine, or L-alanine or menadione for 10 min. Alternatively, cells
   were pre-incubated with the thioredoxin reductase inhibitor auranofin
   (3 µM, 20 min). Free thiols were blocked with N-ethylmaleimide (NEM,
   100 mM). After a washing step with PBS-NEM (100 mM), cells were scraped
   in alkylation buffer (40 mM Hepes, 50 mM NaCl, 1 mM EGTA, Inhibitors,
   100 mM NEM, pH 7.4) and 1% CHAPS (Applichem, Darmstadt, Germany) for
   solubilization. The protein concentration was determined by the
   Bradford assay. Samples were supplemented with non-reducing sample
   buffer (8.5% glycerol, 2% SDS, 6.25% TRIS/HCl pH 6.8, 0.013%
   bromophenol blue) and separated on an SDS-PAGE gel, followed by western
   blotting analysis. After incubation with primary antibodies, membranes
   were analyzed with an infrared-based detection system (LI-COR, Bad
   Homburg, Germany), using fluorescent-dye-conjugated secondary
   antibodies from LI-COR biosciences (Bad Homburg, Germany). The
   following antibodies were used: D-amino acid oxidase (DAO, #ab196563,
   Abcam, Cambridge, UK), peroxiredoxin 1 (Prx1, #MAB3488, R&D systems,
   Minneapolis, MN, USA), peroxiredoxin 2 (Prx2, #AF3489, R&D systems,
   Minneapolis, MN, USA), peroxiredoxin 3 (Prx3, #A304-744, Bethyl
   Laboratories, Montgomery, AL, USA), peroxiredoxin 4 (Prx4, #AF5460, R&D
   systems, Montgomery, AL, USA), and peroxiredoxin-SO[3] (Prx-SO3,
   #ab16830, Abcam, Cambridge, UK).

2.8. Metabolomics

   HUVEC were grown in sister cultures that were treated identically. One
   dish was used for metabolic analysis while the corresponding sister
   culture was used to isolate the total RNA for RNASeq and data
   normalization. HUVEC at 80% confluence, were starved overnight in
   endothelial basal medium (PeloBiotech, Planegg, Germany) supplemented
   with 10 mM L-glutamine and 0.1% FCS. The next day, cells were exposed
   to H[2]O[2] (10 μM or 300 μM), menadione (5 μM), D-alanine (3 mM), or
   basal medium (control sample) for 3, 10, 30, 90, 270, or 900 min. After
   exposure, cells were washed with ice-cold PBS and subsequently
   harvested in 80% LC/MS-grade methanol (Carl Roth, Karlsruhe, Germany)
   containing internal standards. Untargeted global metabolomics was
   performed by Metabolon Inc. (Morrisville, NC, USA) using a Waters
   ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo
   Scientific Q-Exactive high resolution/accurate mass spectrometer
   interfaced with a heated electrospray ionization (HESI-II) source and
   Orbitrap mass analyzer operating at 35,000 mass resolution as
   previously described [[65]23,[66]24,[67]25]. Briefly, cell samples were
   extracted with methanol to remove the protein fraction. The extract was
   divided into five fractions: two for analysis by two separate reverse
   phase (RP)/UPLC-MS/MS methods with positive ion mode electrospray
   ionization (ESI), one for analysis by RP/UPLC-MS/MS with negative ion
   mode ESI, one for analysis by hydrophilic interaction chromatography
   (HILIC)/UPLC-MS/MS with negative ion mode ESI, and one sample as a
   backup. The extract was gradient eluted from a C18 column (Waters UPLC
   BEH C18-2.1 × 100 mm, 1.7 µm) using water and methanol, containing
   0.05% perfluoropentanoic acid (PFPA) and 0.1% formic acid (FA). Another
   aliquot was also analyzed using acidic positive ion conditions,
   however, in this case, the method was chromatographically optimized for
   more hydrophobic compounds. In this method, the extract was gradient
   eluted from the same aforementioned C18 column using methanol,
   acetonitrile, water, 0.05% PFPA and 0.01% FA and was operated at an
   overall higher organic content. Another aliquot was analyzed using
   basic negative ion optimized conditions using a separate dedicated C18
   column. The basic extracts were gradient eluted from the column using
   methanol and water, however with 6.5 mM ammonium bicarbonate at pH 8.
   The fourth aliquot was analyzed via negative ionization following
   elution from an HILIC column (Waters UPLC BEH Amide 2.1 mm × 150 mm,
   1.7 µm) using a gradient consisting of water containing acetonitrile
   with 10 mM ammonium formate, pH 10.8. The MS analysis alternated
   between MS and data-dependent MS^n scans using dynamic exclusion. The
   scan range varied slighted between methods but covered 70–1000 m/z. Raw
   data were extracted, peak-identified, and quality control-processed by
   Metabolon [[68]26]. Compounds were identified by comparison to library
   entries with over 3300 commercially available purified standard
   compounds [[69]24]. A batch correction was performed by Metabolon.
   Following log transformation and imputation of missing values, if any,
   with the minimum observed value for each compound, Mixed Model
   Contrasts were used to identify biochemicals that differed
   significantly between experimental groups. In parallel, metabolomic
   results were normalized to the RNA values of the corresponding sister
   culture, missing values were imputed, and statistically analyzed using
   log transformed data. p values of <0.05 were considered significant.
   Visualizations and plots of metabolomics data were generated using the
   ggplot (3.3.5) package in R (4.1.1). Each treatment was analyzed with
   respect to its control.

2.9. RNA Isolation and RNAseq by Massive Analysis of cDNA Ends (MACE)

   RNA of sister cultures was isolated with the RNA Mini Kit from
   (Bio&Sell, Nuremberg, Germany) combined with on-column DNase digestion
   (DNase-Free DNase Set, Qiagen, Hilden, Germany) to avoid contamination
   by genomic DNA. The libraries were prepared using GenXPro MACE kit
   v.2.0. RNA was sheared to an average size of 350 bps followed by poly-A
   specific cDNA synthesis. The PCR product was purified by SPRI (solid
   phase reversible immobilization) purification and the final product was
   quality controlled on a PerkinElmer LabChip GXII. The fragments were
   ligated to “TrueQuant” (GenXPro property, containing unique molecule
   identifiers). This unique identifier helps to remove PCR duplicates.
   MACE-tags were amplified with 10 PCR cycles and the libraries were
   sequenced on an Illumina NextSeq 500 machine. MACE sequencing reads for
   all samples were quantified against the hg38 transcriptome (obtained
   from Ensembl) [[70]27,[71]28]. Reads not aligned to the transcriptome
   were discarded at this point. Differential gene expression analysis was
   performed using DESeq2 (1.32.0) [[72]29]. Raw transcript counts were
   summed per gene and used in the standard DESeq2 differential gene
   expression analysis workflow, using a negative binomial test over gene
   counts in each of the combinations of conditions. Differences in gene
   expression (Differentially expressed genes, DEG) between conditions
   were considered significant with an adjusted (Benjamini-Hochberg) p
   value < 0.05. Differentially expressed genes per exposure type were
   subjected to pathway enrichment analysis in R using ClusterProfiler
   (4.0.5) and the enrichKEGG function [[73]30]. Significantly enriched
   pathways versus a default random background gene set were those with an
   adjusted (Benjamini-Hochberg) p value < 0.05. Gene heatmaps were
   hierarchically clustered in R using hclust with Euclidean distances and
   the ward.D method.

2.10. Gene Correlation and Transcription Factor Analysis

   Transcription factor (TF) analysis from differentially expressed genes
   (DEGs) was performed using enrichR (3.0) with TFs enriched in the
   “ENCODE and ChEA consensus TFs from ChiP-X” database. Enriched TFs were
   considered significant with an adjusted (Benjamini-Hochberg) p value <
   0.05. Correlations were performed using a Pearson correlation test in
   R. Visualizations and plots were generated using ggplot2 (3.3.5)
   [[74]31].

2.11. Statistics

   Unless otherwise indicated, data are shown as means ± standard error of
   the mean (SEM). Calculations were performed with Prism 9.0 or R
   (4.1.1). For multiple group comparisons, ANOVA followed by Bonferroni
   or Benjamini Hochberg post hoc testing was performed. A p-value of
   <0.05 was considered as significant. n indicates the number of
   individual experiments. No samples were excluded from the analysis.

3. Results

3.1. DAO Is an Efficient Chemogenetic Tool to Produce H[2]O[2]
Intracellularly

   We cloned human DAO into a lentiviral plasmid to overexpress the enzyme
   in HUVEC to generate intracellular H[2]O[2] upon stimulation with
   D-alanine ([75]Figure 1A,B). DAO expression after transduction was
   readily detected in HUVEC by immunofluorescence. To test the activity
   of DAO, H[2]O[2] was measured by luminol/ HRP chemiluminescence.
   Addition of 3 mM D-Ala resulted in a strong increase in
   chemiluminescence ([76]Figure 1C). Likewise, with the Amplex red^®/HRP
   assay, increasing concentrations of D-alanine (1–10 mM), but not
   L-alanine, resulted in an increased H[2]O[2] production. PEG-catalase
   reduced the detected level of H[2]O[2] ([77]Figure 1D) whereas the DAO
   inhibitor 4HF (4H-furo [3,2-b]pyrrole-5-carboxylic acid, 1 µM) almost
   entirely abolished the D-alanine-elicited H[2]O[2] production
   ([78]Supplementary Figure S1). Thus, the HUVEC-DAO system is a valid
   tool to increase cellular H[2]O[2] level in a D-alanine-dependent
   manner.

Figure 1.

   [79]Figure 1
   [80]Open in a new tab

   Generation and characterization of DAO as a chemogenetic approach to
   study the met-abolic and transcriptomic response to intracellular
   H[2]O[2] in HUVEC. (A) Lentiviral overexpression of human D-amino acid
   oxidase (DAO) and its chemical reaction. (B) Immuno-fluorescence for
   DAO in HUVEC-DAO (left) and HUVEC-CTL (empty vector, right). (C)
   H[2]O[2] measurements in HUVEC using Luminol/HRP and (D) Amplex
   red®/HRP assay * p < 0.05 10 mM D-Ala versus 1 mM D-Ala; # p < 0.05 10
   mM D-Ala plus vs 10 mM D-Ala minus PEG-catalase. One-way-ANOVA with
   Bonferroni correction.

3.2. Different Types of ROS Elicit a Differential Metabolic Response in HUVEC

   We designed a large-scale experiment to determine the short-term and
   long-term metabolic and transcriptomic response of HUVEC to different
   types and concentrations of ROS ([81]Figure 2A).

Figure 2.

   [82]Figure 2
   [83]Open in a new tab

   Time course analysis for metabolomics and transcriptomics of HUVEC with
   different oxidative stimuli. (A): Experimental design. (B): Number of
   metabolites significantly altered upon exposure to different ROS in
   HUVEC (n = 3).

   HUVEC-DAO were stimulated with 3 mM D-alanine to generate H[2]O[2]
   intracellularly. In comparison to extracellular H[2]O[2], cells were
   acutely exposed to either low (10 µM) or high (300 µM) concentrations
   of exogenous H[2]O[2]. Additionally, menadione (5 μM), a polycyclic
   aromatic ketone, was included to acutely generate intracellular
   superoxide anions (O[2]^●−) by redox-cycling [[84]32].

   Untargeted metabolomics revealed that the metabolic response to the
   different ROS exposures differs considerably ([85]Supplementary Table
   S1). DAO-derived H[2]O[2] induced only a subtle change in identified
   metabolites (a total of 39 metabolites significantly changed) over the
   course of time. Menadione (5 µM) and low extracellular H[2]O[2] (10 µM)
   showed a similar effect with a total of 93 vs. 73 metabolites
   significantly changed. For both, metabolic changes peaked at 270 min.
   Major metabolic changes were observed in response to the high
   concentration of H[2]O[2] (300 µM), leading to a total of 358
   metabolites changed. Altered metabolites included those involved in
   redox homeostasis, energy, and nucleotide synthesis ([86]Figure 2A,B,
   [87]Supplementary Figure S3, Supplementary Table S2). These changes
   occurred as early as 3 min after the beginning of exposure, and
   reverted to baseline within 270 min.

   The glutathione metabolism is an essential and central antioxidant
   pathway. We therefore analyzed how this pathway is affected by the
   different oxidants ([88]Figure 3A). High H[2]O[2] had the strongest
   effect on glutathione metabolism, leading to a significant increase in
   oxidized glutathione species such as cysteine-glutathione disulfide.
   Furthermore, glutathione exhibited a transient decrease in the
   abundance of its reduced state in response to 300 µM H[2]O[2] that
   recovered within 30 min. Unexpectedly, reduced glutathione was not
   affected by either DAO-derived H[2]O[2] or menadione ([89]Figure 3B).
   Nevertheless, all types of ROS significantly increased
   S-lactoylglutathione after 10 min. This metabolite is an intermediate
   in the detoxification of methylglyoxal, which is generated as a side
   product of upper glycolysis. Methylglyoxal is detoxified by the
   glyoxalase system, a composition of two enzymes, in which the first
   enzyme (glyoxalase I) converts the hemithioacetal that forms
   spontaneously between methylglyoxal and GSH to S-D-lactoylglutathione.
   The second enzyme (glyoxalase II) converts S-D-lactoylglutathione to
   GSH and D-lactate ([90]Supplementary Figure S2C). An increase in
   methyglyoxal can occur under stress situations such as elevated ROS
   formation [[91]33]. Methylglyoxal itself is highly reactive and
   therefore cannot be detected with the analytic methods used here. A
   close look at glycolysis revealed a general increase in metabolites of
   the upper glycolysis pathway (e.g., dihydroxyacetone phosphate, DHAP)
   and a decrease in metabolites of the lower glycolysis pathway (e.g.,
   phosphoenolpyruvate, PEP) in response to exposure to H[2]O[2] (300 µM).
   This effect was, however, not observed after exposure to DAO-derived
   H[2]O[2] or menadione ([92]Supplementary Figure S2A). Recent studies
   demonstrated that a short-term antioxidant response can be mediated by
   redox-sensitive enzymes in the lower glycolysis pathway. High H[2]O[2]
   concentration inhibits glyceraldehyde 3-phosphate dehydrogenase (GAPDH)
   by promoting the formation of an intermolecular disulfide bond
   [[93]34,[94]35]. Therefore, an accumulation of DHAP is noticeable after
   exposure to high H[2]O[2]. The increase in DHAP is consistent with the
   accumulation of S-lactoylglutathione as methyglyoxal is formed during
   the non-enzymatic phosphate elimination of DHAP ([95]Supplementary
   Figure S2C). Additionally, an overall decrease of nucleoside
   triphosphates (ATP and GTP) and an increase of nucleoside
   monophosphates emphasizes the cellular needs during stress defense,
   which includes the generation of nucleotide precursors for repair of
   DNA ([96]Supplementary Figure S2B, Supplementary Table S1). NADPH and
   NADP+ were not detected in the experiment. Notably, only extracellular
   H[2]O[2] at high concentration led to major changes in the metabolism
   of HUVEC. Therefore, we assume that the site and the concentration of
   ROS are central to elicit metabolic changes.

Figure 3.

   [97]Figure 3
   [98]Open in a new tab

   Changes in glutathione and glutathione-related metabolites in HUVEC
   upon exposure to different ROS. (A): Time course metabolic changes in
   glutathione metabolism. (B): Glutathione redox state in HUVEC over the
   course of time treatment with oxidative stimuli. FC = fold change. A&B:
   n = 3. * p-value < 0.05.

3.3. ROS Lead to Different Gene Expression Responses by HUVEC

   Cells invoke specialized gene programs to cope with stress. In order to
   identify whether the type, concentration, and site (intra- vs.
   extracellular) of ROS exposure elicit differential gene expression,
   MACEseq was performed.

   Similar to the metabolic measurements, exposure to 300 µM H[2]O[2]
   induced the most pronounced transcriptomic changes (3540 DEGs,
   differentially expressed genes) in comparison to the other treatments
   (3 mM D-alanine 1575 DEGs; menadione 1236 DEGs) ([99]Figure 4A,
   [100]Supplementary Figure S3, Supplementary Table S3). Ten µM H[2]O[2]
   led to no significant differences in gene expression as compared to
   untreated HUVEC. Thus, we focused our subsequent analysis on 300 µM
   H[2]O[2,] 3 mM D-alanine and 5 µM menadione.

Figure 4.

   [101]Figure 4
   [102]Open in a new tab

   Different ROS lead to a different time-dependent transcriptomic
   response of HUVEC. (A): Differentially expressed genes after
   stimulation with 300 μM H[2]O[2], 3 mM D-alanine or 5 μM menadione over
   the course of 30, 90, 270 and 900 min (significantly changed genes
   (p[adj] < 0.05) are highlighted in blue and red). (B): Pathway
   annotation of significantly altered genes in A. (C): Venn diagram of
   significantly regulated genes in the treatments as indicated. (D):
   Correlation analysis for significantly DEGs comparing exogenous versus
   DAO-derived H[2]O[2].

   With respect to the temporal changes in gene expression, 300 μM
   extracellular H[2]O[2] induced an early gene response 30 min after
   exposure, while most changes in gene expression occurred at 270 min
   ([103]Figure 4A). In contrast, changes caused by 3 mM D-alanine and 5
   µM menadione appeared only after 270 min.

   To gain a comprehensive picture of the molecular pathways affected by
   exposure to different ROS, a pathway enrichment analysis using the
   Kyoto Encyclopedia of Genes and Genome (KEGG) database considering the
   significantly (p[adj] < 0.05) changed genes for all time points was
   performed. High H[2]O[2] concentration (300 µM) significantly induced
   genes of the p53 signaling pathway and cell cycle-related genes.
   Stress-responsive genes such as the cell cycle regulator GADD45A
   (Growth Arrest and DNA Damage Inducible Alpha) and PHLDA3 (Pleckstrin
   Homology Like Domain Family A Member 3) were increased after exposure
   to high concentration of H[2]O[2]. Moreover, induction of CDKN1A/p21
   (Cyclin Dependent Kinase Inhibitor 1A) suggests that H[2]O[2] promotes
   a senescent phenotype in EC ([104]Supplementary Figure S4A,
   Supplementary Table S2). This is consistent with the known effect of
   H[2]O[2] for stress-induced cell cycle arrest [[105]36]. This effect
   did not occur after exposure to menadione or stimulation with
   D-alanine. Nevertheless, all oxidants significantly changed
   ribosome-associated RNAs (rRNA). rRNAs are bound to ribosomal proteins
   to form small and large ribosome subunits. They can be chemically
   modified by ROS, which we assume leads to an upregulation of the rRNA
   transcripts necessary to maintain ribosomal functionality ([106]Figure
   4B, [107]Supplementary Figure S4A). We further described the
   similarities between high concentration of H[2]O[2], DAO-derived
   H[2]O[2,] and exposure to menadione. A Venn diagram shows that 267
   genes were changed by all oxidative stimuli ([108]Figure 4C).

   To gain insights into their regulation, we checked whether these genes
   are under control of similar transcription factors. Transcription
   Factor Enrichment Analysis (TFEA) identified 34 transcription factors
   (TF) such as TAF1, TAF7, MYC, ATF2, and YY1 as putative regulators of
   the DEGs common to all treatments ([109]Supplementary Figure S4B).
   These TF (among others) highly regulate ribosomal-associated genes that
   are, in fact, enriched in all applied oxidative stimuli. Last, we
   investigated whether intracellular H[2]O[2] induces any similar
   transcriptional response to exogenous H[2]O[2] (300 µM) in HUVEC. For
   this, log fold change correlation analysis of the differentially
   expressed genes for each individual time point was performed. As shown
   in [110]Figure 4D, a positive correlation in gene expression exists for
   exposure to H[2]O[2] (300 µM) and DAO-derived H[2]O[2] (3 mM
   D-alanine). This suggests that, despite the fact that H[2]O[2] induces
   the largest changes in gene expression, a positive correlation exists
   for genes at equal time points, independent of the source of H[2]O[2].

   Altogether, the results show that only exogenous H[2]O[2] at high
   concentrations induces the classical stress-induced senescence markers
   (e.g., p21). Both menadione and DAO-derived H[2]O[2] did not elicit
   this effect. Nevertheless, all treatments induce genes linked to
   ribosomal function while a positive correlation occurs in gene
   expression changes between high extracellular H[2]O[2] and DAO-derived
   H[2]O[2].

3.4. Only 300 µM H[2]O[2] Overoxidizes Peroxiredoxins in HUVEC

   As the major changes in metabolomics and transcriptomics of HUVEC were
   almost exclusively induced by exogenous H[2]O[2] at high concentrations
   (300 µM), it appears that the type, the concentration, and the
   localization determine the biological effect of ROS exposure.
   Therefore, we took a close look at redox sensor proteins to detect
   disturbances in redox homeostasis. As thiol-specific antioxidants,
   peroxiredoxins (Prx) act as redox sensors. These enzymes harbor a
   peroxidatic cysteine that can be directly oxidized to protect cellular
   components from oxidative damage. Upon oxidation, Prx can form dimers
   or multimers with other Prx or proteins. Furthermore, the
   redox-sensitive cysteine of Prx can be irreversibly overoxidized to a
   sulfonic acid (RSO[3]^−). Importantly, oxidized Prx are reduced by the
   thioredoxin system, which utilizes NADPH as reducing partner. As the
   oxidative status of Prx is a robust readout of the oxidation level of a
   cell [[111]7,[112]37,[113]38], we investigated how exogenous and
   DAO-derived H[2]O[2] as well as menadione oxidize Prx.

   Addition of D-alanine led to a dose-dependent increase in
   Prx-dimerization in HUVEC-DAO with a pronounced effect on the Prx1 and
   Prx2 isoenzymes but not in Prx3 ([114]Figure 5A,B). The increase in
   Prx-dimerization was accompanied by a decrease in the corresponding Prx
   monomer. This effect on Prx1 and Prx2 dimerization was also detectable
   in HEK293 cells that overexpress DAO (HEK-DAO). Unlike the case with
   HUVEC, Prx3 was oxidized upon addition of 3 mM and 10 mM of D-ala in
   HEK-DAO. Furthermore, DAO-derived H[2]O[2] led to the formation of Prx4
   multimers in HEK-DAO ([115]Supplementary Figure S5).

Figure 5.

   [116]Figure 5
   [117]Open in a new tab

   Oxidation of peroxiredoxins (Prx) in response to exogenous or
   DAO-derived H[2]O[2] or menadione. (A): Representative redox western
   blot (30 µg protein) for Prx1, Prx2, Prx3, and Prx-SO[3] after
   stimulation with different concentration of D- or L-Ala in HUVEC. (B):
   Quantification of redox-western blots by densitometry (n = 3). (C):
   Redox western blot and quantification (D) for Prx1 and Prx2 with
   HUVEC-DAO pre-incubated with auranofin (20 min, 3 µM) prior D- or L-ala
   stimulation, * p < 0.05 as compared to CTL with auranofin.
   One-way-ANOVA with Bonferroni correction.

   Stimulation of HUVEC-DAO with D-alanine increased the formation of the
   thioredoxin-S-S-peroxiredoxin dimer ([118]Figure 5A). Remarkably,
   D-alanine did not lead to overoxidation of peroxiredoxins (Prx-SO[3]),
   which was only induced by 300 µM exogenous H[2]O[2] ([119]Figure 5A,B).
   We therefore speculated that Prx-SO[3] only accumulated in response to
   intracellular H[2]O[2] if the thioredoxin reductase system is
   inhibited. To test this, HUVEC were pre-incubated with auranofin (20
   min, 3 µM) prior to addition of D-alanine. Blocking thioredoxin
   reductase increased the ratio of dimer to monomer of Prx1 in control
   cells by two-fold whereas this ratio for Prx-1-S-S-thioredoxin
   increased by six-fold ([120]Figure 5C,D). Contrary to our expectation,
   pre-incubation with auranofin did not facilitate the accumulation of
   Prx-SO[3] in response to menadione or DAO-derived H[2]O[2] ([121]Figure
   5C). However, the presence of auranofin, menadione and DAO-derived
   H[2]O[2] caused a massive increase in Prx-1 dimer as well as
   Prx-1-S-S-thioredoxin. Thus, the localization and concentration of ROS
   exposure are important factors when considering their effect on the
   redox-status of a cell. While DAO-derived H[2]O[2] and
   menadione-derived superoxide are generated intracellularly, the
   antioxidant system takes effect immediately. Only high amounts of
   H[2]O[2] added to the extracellular space might overcome the cellular
   antioxidant capacity and lead to Prx-overoxidation.

4. Discussion

   In this study we compared how different types and concentrations of ROS
   modulate the metabolic and transcriptomic response of HUVEC over the
   course of time. To generate H[2]O[2] intracellularly, we utilized a
   chemogenetic approach with DAO. DAO contains a peroxisomal targeting
   sequence (PTS), but immunofluorescence showed that the enzyme, in this
   overexpression system, is distributed throughout the whole cell.
   Therefore, we did not genetically manipulate its subcellular
   localization by removing the PTS or addition of a different subcellular
   target sequence. Interestingly, the DAO approach was used previously in
   an endothelial cell line (EA.hy926) to study how the enzyme targeted
   different subcellular compartments modulates endothelial cell
   phosphorylation pathways. DAO-derived H[2]O[2] mediates eNOS
   phosphorylation via AMPK activation when the enzyme is directed to the
   nucleus by an importing sequence. Cytosolic or caveolae-targeted DAO
   had no impact on eNOS phosphorylation [[122]13,[123]39]. Furthermore in
   another study, DAO was overexpressed in the heart of mice and its
   activation, by feeding the mice D-alanine, which led to cardiac
   dysfunction [[124]18].

   Nevertheless, the DAO approach has not yet been explored in an
   untargeted metabolomics and transcriptomics study. In fact, there are
   only few studies that describe the short-term and long-term metabolic
   effects of exposure to H[2]O[2]. In yeast and many human cell types, it
   was shown that inhibition of glycolytic enzymes by H[2]O[2] leads to an
   accumulation of glycolytic intermediates that consequently induce an
   increased flux into the pentose phosphate pathway (PPP), through both
   the oxidative and non-oxidative branches
   [[125]40,[126]41,[127]42,[128]43]. This matches our findings and is
   compatible with the previously shown oxidative inhibition of GAPDH.
   Kuehne et al. (2015) directly showed an upregulation of the PPP in
   human skin fibroblasts as a first line response after exposure to 500
   µM H[2]O[2]. By using ultra-short ^13C labelling experiments, the
   authors provided evidence for multiple cycling of carbon backbones in
   the oxidative PPP, potentially maximizing NADPH reduction. Hence, NADPH
   is required as a reducing equivalent, which maintains the active form
   of catalase and is a cofactor of TRX and GSH reductase
   [[129]35,[130]44]. Additionally, it has been shown that quiescent EC,
   in a manner different from that of proliferating EC, protect themselves
   against oxidative stress by increasing fatty acid oxidation up to
   three-fold to generate NADPH via isocitrate dehydrogenase and malic
   enzyme [[131]45]. Furthermore, H[2]O[2] affected phosphorylated
   nucleotides and produced a reduction in ATP and GTP, which was
   paralleled by an increase in their respective mono-phosphorylated forms
   [[132]46]. It was quite surprising to observe that only H[2]O[2] at
   high concentration induced most metabolic changes in HUVEC. As early as
   3 min after the addition of high concentration of H[2]O[2,] significant
   changes in the antioxidant system became apparent. A profound reduction
   in the GSH/GSSG ratio was observed after 10 min and this ratio was back
   to the basal situation after 270 min, pointing to a transient effect.
   High concentration of H[2]O[2] also increased the concentration of
   methionine sulfoxide and reduced the concentration of isocitrate and
   other TCA cycle metabolites. The later effect was also observed in
   menadione-treated HUVEC and is known to be linked to an inhibition of
   aconitase via oxidation of an Fe−S cluster [[133]9]. The only common
   metabolite that we were able to detect altered in all treatments at the
   10 min time point was S-lactoylglutathione. This metabolite is formed
   upon reaction of glutathione with methylglyoxal [[134]33]. An increase
   in methylglyoxal can occur under stress conditions likely due to
   several events such as the reversible S-glutathionylation and
   inhibition of GAPDH [[135]47]. Methylglyoxal itself was not detected in
   the experiments, likely due to its high reactivity, but considering the
   fact that endothelial cells are highly glycolytic, an efficient removal
   of methylglyoxal in GSH-dependent manner is essential.

   The changes in gene expression observed in HUVEC were mainly caused by
   exposure to 300 µM H[2]O[2] that activated, among others, the p53
   pathway. It is accepted that low to moderate levels of ROS activate
   p53-related genes that increase the time needed for cell repair (e.g.,
   cell cycle arrest and autophagy). With higher levels of ROS, p53
   facilitates cellular stress and induces apoptosis to prevent aberrant
   cell proliferation. Exposure to menadione induces selective increases
   in genes of the p53 pathway, suggesting that the intensity and location
   of the oxidation are important parameters that would determine the
   function of p53 in regulating the signaling outcome [[136]48]. All
   oxidative stimuli had a significant impact on ribosome-associated RNA.
   Both ribosomal RNA and ribosomal proteins can be chemically modified by
   ROS, leading to a loss of their function. Ribosomal RNA is the
   structural and functional core of the ribosome. ROS can chemically
   modify the base and sugar moieties of rRNA, generating a basic site and
   strand breaks. Among proteins of the translational machinery, the
   impact of oxidative stress on the ribosome remains the least studied
   [[137]49,[138]50]. Due to its high cellular abundance, RNA is more
   frequently subject to oxidative damage in comparison to DNA [[139]49].
   Transcription factor analysis revealed ATF2, MYC, and TAF1/7 related
   genes commonly changed among the applied redox stimuli. ATF2 and MYC
   sense cellular stress caused by increased ROS concentrations. They
   regulate DNA damage response and cell cycle progression
   [[140]51,[141]52]. Moreover, TAF-related genes are required for basal
   transcription. Interestingly, all TFs regulate the expression of
   ribosomal genes, e.g., RPL4, which encodes the 60 S subunit of the
   ribosome [[142]53]. Indeed, the findings of this study suggest that EC
   upregulate ribosomal RNAs as a response to different redox stimuli.
   This implies that rRNA is a primary target for oxidative stress in EC
   and this phenomenon has not been extensively studied in EC and can be
   the subject to future research.

   High extracellular concentration of H[2]O[2] had a stronger impact on
   gene expression and metabolism compared to menadione and DAO-derived
   H[2]O[2]. Thus, we analysed the oxidation status of Prx enzymes. Only
   H[2]O[2] at high concentrations induced an overoxidation of Prx
   (Prx-SO[3]). Blocking thioredoxin reductases with auranofin did not
   induce the formation of Prx-SO[3] in D-alanine treated cells. This
   points to a high intracellular reducing capacity of EC and the
   relevance of compartmentalized ROS production for signalling. The
   floodgate model suggests that scavenging enzymes such as Prx must be
   inactivated, so H[2]O[2] can directly oxidize target proteins. The
   irreversible inactivation of Prx by overoxidation allows endogenous
   H[2]O[2] concentrations to build up, which promotes redox signalling by
   direct oxidation of target proteins and other biomolecules. Others have
   shown that reaction of H[2]O[2] with thiols is too slow to outcompete
   peroxiredoxins to exert a direct redox signaling [[143]54]. In fact,
   peroxiredoxins indirectly facilitate H[2]O[2] sensing and oxidize
   target proteins [[144]55,[145]56] in a redox-relay model, transferring
   oxidizing equivalents to the transcription factor STAT3 or to the
   kinases ASK1 and MEKK4 [[146]56,[147]57,[148]58,[149]59].

   As shown here, there is a substantial difference in the metabolic and
   transcriptomic response by EC to extracellular H[2]O[2] vs.
   intracellular H[2]O[2]. This is consistent with increasing recognition
   of subcellular compartmentalization of redox processes. Except for the
   endoplasmic reticulum, intracellular redox potentials are largely more
   reducing in comparison to the extracellular space. In particular, the
   extracellular space contains recognizable amounts of reactive
   intermediates and several available protein targets. Thiol-redox
   regulation of the extracellular space is associated with cellular
   signaling cascades. Indeed, the pathways that respond to ROS are
   essentially the extracellular signal-regulated kinases (ERK1/2), c-Jun
   NH2-terminal kinases (JNKs), and p38 kinase. The ERKs family can be
   activated by growth factors as a response to oxidative stress. Vascular
   smooth muscle cells (VSMC) secrete cyclophilin A, a member of the
   immunophilin family, in response to oxidative stress, which mediates
   ERK1/2 activation. The JNKs and p38 kinases are primarily involved in
   the cellular stress condition and are activated by extracellular
   H[2]O[2] in smooth muscle cells [[150]60,[151]61].

5. Conclusions

   In summary, only high concentrations (300 µM) of extracellular H[2]O[2]
   induced significant changes in metabolic pathways of redox homeostasis,
   energy production, and nucleotide synthesis. Likewise, high
   concentration of H[2]O[2] caused major changes in gene expression and
   oxidation of peroxiredoxin enzymes. Collectively, these findings
   suggest that the source and the concentration of ROS are important to
   elicit changes in metabolism and gene expression. Our data indicate
   that EC have sufficient intracellular reducing capacity to scavenge
   intracellularly produced H[2]O[2]. This is a possible explanation for
   the different effects of intracellular DAO-derived H[2]O[2] and
   menadione-derived superoxide in comparison to high exogenous H[2]O[2].

Acknowledgments