In Vitro Modulation of Redox and Metabolism Interplay at the Brain Vascular Endothelium: Genomic and Proteomic Profiles of Sulforaphane Activity

Sulforaphane (SFN) has been shown to protect the brain vascular system and effectively reduce ischemic injuries and cognitive deficits. Given the robust cerebrovascular protection afforded by SFN, the objective of this study was to profile these effects in vitro using primary mouse brain microvascular endothelial cells and focusing on cellular redox, metabolism and detoxification functions. We used a mouse MitoChip array developed and validated at the FDA National Center for Toxicological Research (NCTR) to profile a host of genes encoded by nuclear and mt-DNA following SFN treatment (0–5 µM). Corresponding protein expression levels were assessed (ad hoc) by qRT-PCR, immunoblots and immunocytochemistry (ICC). Gene ontology clustering revealed that SFN treatment (24 h) significantly up-regulated ~50 key genes (>1.5 fold, adjusted p < 0.0001) and repressed 20 genes (<0.7 fold, adjusted p < 0.0001) belonging to oxidative stress, phase 1 & 2 drug metabolism enzymes (glutathione system), iron transporters, glycolysis, oxidative phosphorylation (OXPHOS), amino acid metabolism, lipid metabolism and mitochondrial biogenesis. Our results show that SFN stimulated the production of ATP by promoting the expression and activity of glucose transporter-1, and glycolysis. In addition, SFN upregulated anti-oxidative stress responses, redox signaling and phase 2 drug metabolism/detoxification functions, thus elucidating further the previously observed neurovascular protective effects of this compound.


Results
SFN affects the expression level of brain vascular endothelial mitochondria-related genes and Nrf2/Nqo1 upregulation. As demonstrated by the 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Fig. 1A), SFN does not impact cell viability at any of the concentrations tested and significantly upregulates Nrf2 expression and its downstream target NAD(P)H: Quinone oxidoreductase 1 (Nqo1; Fig. 1B). The relative impact of SFN on mitochondria-specific gene expression is evident when looking at the Volcano plots ( Fig. 1C1-C3). These Volcano plots represent the effect of SFN on the expression levels of several of mitochondria-specific genes (see also Supplementary Table 2). Fold changes are plotted on X-axis while significance or p-values are plotted on the Y-axis. If we compare the position and density of the red dots, it's clear that the effect is dose-dependent with increases in the expression levels and significance of the changes (see also Supplementary Table 1). A total of 50 genes were upregulated by more than 1.5-fold whereas around 20 genes were repressed down to less than 0.7-fold. Mitochondria-specific genes are highly conservative compared to other nuclear genes 37 and a change in 1.5-fold is considered as highly significant. SFN promotes ATP production through upregulation of key enzymes involved in glucose transport and utilization via glycolytic pathway. To maintain its dynamic and metabolic functions, the brain microvascular endothelium at the BBB is abundantly enriched with glucose transporter 1 (Glut1) protein (~55kD) 38 . Here, we show that SFN promotes Glut1 expression and activity (Fig. 2) in brain vascular endothelium. 24 h exposure to SFN (5 µM) promoted the expression of the glucose transporter Glut1 as demonstrated by western blots (WB) analyses and immunofluorescence in mouse BBB endothelial cells ( Fig. 2A). The effect was not limited to Glut1 expression levels but also extended to its overall transport activity as determined by 2-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)Amino)-2-Deoxyglucose (2-NBDG) uptake (p < 0.05 vs. control; see Fig. 2B).
Following influx into the cytoplasm, glucose undergoes glycolysis to ultimately produce ATP. Hexokinases are the first members of this cycle. Pdk1 plays a key role in regulation of glucose and fatty acid metabolism and homeostasis via phosphorylation of the pyruvate dehydrogenase subunits PDHA1 and PDHA2 while Hk1 is one of four hexokinase isoenzymes that participate in glycolysis and catalyzes the phosphorylation of glucose into Figure 2. SFN modulate the redox-metabolic interplay of the brain vascular endothelium to promote ATP production. SFN promotes upregulation of Glut1 expression (A) and transport activity (B) as determined by 2-NBDG uptake normalized to total protein. (C) Key genes involved in the regulation of glycolysis were also upregulated including hexokinase 1 (Ht1) and pyruvate kinase 2 (Pkm2) as analyzed by MitoChip array (C1) and confirmed by RT-PCR (C2). The overall cellular effect was a substantial increase in ATP production (D). N = 3 independent biological samples per condition assayed in triplicates. "*"p < 0.05, "**"p < 0.01 compared to controls (fold changes). For the time-dependent study (24 vs. 48 h SFN treatment) "+" p < 0.05 vs. 24 h. Blots of Glut1 and β-actin were taken from different gels.
Scientific REPORTS | (2018) 8:12708 | DOI:10.1038/s41598-018-31137-7 glucose-6-phosphate (G6P) 39 to initiate cellular respiration, thus providing enough activation energy to jumpstart the glycolytic process. As shown in Fig. 2C1,C2, SFN significantly upregulated the gene expression level of Hk1 as compared to Hk2 (Fig. 2C1) which was further validated by RT-PCR analysis (Fig. 2C2). Similarly, the gene expression of Pdk1 (a mitochondrial multienzyme complex that catalyzes the oxidative decarboxylation of pyruvate) one of the major enzymes responsible for the regulation of homeostasis of carbohydrate fuels was upregulate by SFN.
Mitochip analysis also showed that the level of glucokinase (Gsk; another hexokinase with a high Km for glucose) was also upregulated by SFN (although results were statistically significant only at the high concentration of 5 µM). The role of GCK is to provide G6P for the synthesis of glycogen.
Pyruvate kinase isozyme type 2 (Pkm2) is a rate-limiting enzyme that catalyzes the last step of the glycolysis cascade and is responsible for net ATP production 40 . Similarly, SFN treatment upregulated Pkm2 gene expression (Fig. 2C1) and results were also verified by RT-PCR analysis (Fig. 2C2). In this respect it is important to note that significant difference in the magnitude of change in gene expression measured by qPCR and MitoChip microarray is attributed to the difference in the dynamic range of detection between the two techniques (like any DNA microarray, the MitoChip is a transcriptomics tool that allows simultaneous measurement of expression of thousand mitochondria-related genes; however, with a smaller dynamic range compared to qPCR).
As shown in Fig. 2D, the overall impact of SFN on cellular bioenergetics was indeed a significant dose-dependent increase in ATP production which was observed at 48 hours exposure (24 hours SFN exposure did not elicit statistically significant changes -data not shown). This finding further corroborates previous data linking SFN with mitochondrial respiration and bioenergetics 41 .

SFN promotes redox signaling and the expression of phase 2 drug metabolism enzymes. Our
MitoChip data clearly shows that SFN significantly upregulated (in a dose-dependent manner) the expression level of catalase (Cat) along with that of several glutathione S-transferases (GSTs) family members including Gstp1 (a phase 2 enzyme involved in xenobiotic metabolism), glutathione-disulfide reductase (GSR), and thioredoxin 1 (Txn1) (Fig. 3A1). Cat is a key antioxidant enzyme responsible for the conversion of hydrogen peroxide to water and oxygen and thereby mitigates the toxic effects of this highly reactive oxidant substance. Complementary to Cat's activity, GST1 is a major detoxifying enzyme that catalyzes the conjugation of endogenous and exogenous toxicants with reduced glutathione. Impact of SFN on the gene expression data of these 2 major detoxification enzymes were further validated ad hoc by RT-PCR analysis (Fig. 3A2).
Results from the mouse MitoChip gene expression arrays also revealed a significant SFN dose-dependent up-regulation of Abc transporters Abcd3 and particularly Abcb6 (Fig. 3B1). The effect in the latter case was both dose and time-dependent and resulted in a nearly 20-fold increase in Abcb6 expression (vs. control) at 48 h exposure (Fig. 3B2). Data were analyzed by one-way ANOVA followed by Tukey's post hoc test.

SFN upregulates the iron transporter Slc40a1 and the cellular oxidative stress response system.
Our mouse MitoChip assay revealed dose-and time-dependent regulation of several Slc transporters by SFN (see also Supplementary Table 3 In addition, we looked at the SFN impact on phase 2 drug metabolism/detoxification genes of the glutathione system. Our results clearly ( Fig. 5A1) indicate that SFN effectively promotes the upregulation of a class of phase 2 enzyme glutathione S-Transferases of the alpha (Gsta1-4) and mu (Gstm1, 2, 5, 7) types which function in the detoxification of electrophilic compounds. In addition, the expression level of the microsomal glutathione S-transferase 1 (Mgst1; also involved in cellular defense against toxic, carcinogenic, and pharmacologically active electrophilic compounds) was upregulated. Mitochip array data were verified (ad hoc) by RT-PCR (Fig. 5A2) demonstrating that the observed effects were both dose-and time-dependent.

Discussion
Our data demonstrates that SFN can positively influence the redox-metabolic interplay at the brain microvascular endothelium (via Nrf2 modulation) and effectively promote glycolysis and ATP production to help sustaining the highly energy demanding functions of these cells.
Our results show that SFN promotes a cascade of bioenergetic events that ultimately leads to increased production of ATP (see also Fig. 2D) and cytoprotection from oxidative stress. This is evident when we look at these effects in the proper contest starting with: (a) upregulation of Glut 1 expression and activity ( Fig. 2A and B) which facilitate glucose entry followed by upregulation of Hk1 (Fig. 2C) which converts glucose into glucose 6 phosphate, thus allowing to maintain a positive gradient for glucose. Hk1 also exhibits a pivotal role in maintaining the homeostasis of mitochondria by restraining ROS generation and conserving mitochondrial membrane potential (Δψ m ) 39 . Glucose entry is then complemented by upregulation of Pkm2, a limiting glycolytic enzyme that catalyzes the final step in glycolysis (Fig. 2C) resulting in an increased production of ATP (Fig. 2D) which was evident following 48 h SFN exposure but not at 24 h (data not shown). This is likely due to the fact that ATP synthesis is the end term of the metabolic process and a significant increase of ATP must be preceded by an amplification of the various bioenergetic steps involved in its production and thus takes longer to become manifest. Concurrently, the metabolic processes involved in the synthesis of ATP also provide extra reductive equivalent (in the form of NADPH) and facilitate the generation of reduced glutathione to counteract cellular ROS and oxidative stress 42,43 .
Scientific REPORTS | (2018) 8:12708 | DOI:10.1038/s41598-018-31137-7 SFN positively impacts redox metabolism. Cytoprotective functions include control of redox metabolism and detoxification systems which encompass efflux transporters and phase 2 drug metabolism enzymes. Our data have clearly shown that SFN promotes the expression levels of several major antioxidative stress and detoxification enzymes including Gstp1, Cat, Gsr, and Txn1 (see Fig. 4). Specifically, Gstp1 is a phase 2 enzyme which catalyzes the conjugation of hydrophobic and electrophilic compounds with reduced glutathione. Loss of Gstp1 has been associated with nuclear oxidative stress damage and potentiation of cell injury induced by ROS 44,45 . Cat promotes the decomposition of hydrogen peroxide to water and oxygen, thus protecting the cell from oxidative damage by ROS 46,47 . Glutathione-disulfide reductase (GSR) instead catalyzes the reduction of glutathione disulfide (GSSG) to its sulfhydryl and active form GSH which is necessary to maintain the reducing environment of the cells and resist oxidative stress 48 . Finally, Txn1 is a small redox protein which has been shown to exert neuroprotection effects against cerebral ischemia/reperfusion injury caused by oxidative stress 49  Among the ABC family of transporters (Fig. 3B1,B2) SFN is clearly able to promote upregulation Abcb6, a porphyrin transporter (importer) which plays a major role in cell survival responses 51,52 and cellular resistance to metal-induced cellular toxicity 53,54 . In addition to Abcb6, SFN also promoted the expression levels of Abcd3 (albeit to a lesser extent) (Fig. 4B1). This is also of interest since not only Abcd3 has a prominent role in cellular oxidative stress 55,56 but also has a tangible presence in brain cells including glial cells. In fact recent studies showed that knockdown of Abcd3 in glial cells markedly elevate oxidative stress responses 57 and its expression was also significantly down-regulated in a cellular model of ischemic/reperfusion injury studied in brain endothelial cells 55 . As a matter of fact SFN has been shown to protect the neurovascular system in stroke 18 as well as neurons exposed to oxygen and glucose deprivation such us during an ischemic injury 16 . In light of these findings, further studies specifically focused on Abcd3 and its transcriptional regulation by SFN would be critical to understanding its role in brain cytoprotection and injuries by oxidative stress.

and is
In addition to Abc transporters, SFN also upregulated Slc40a1 (also known as Ferroportin 1; a well-studied iron importer) which plays a critical role in cerebral iron homeostasis and has been implicated with several neurological disorders including Alzheimer's disease brain pathology 58-60 , abnormal forebrain patterning and neural tube closure and inflammation 61,62 . This is of relevant importance given the inference of iron overload in cellular damage 34 and further outlines the protective potential of SFN. SFN upregulates the detoxification systems. Another facet of SFN activity encompasses phase 2 detoxification systems and specifically GSTs enzymes. These are a group of glutathione S-transferases expressed in different subcellular compartments including cytosol and mitochondria 63,64 and are involved in oxidative stress responses, detoxification functions but also cell signaling and disease prevention/progression. Our results clearly show that SFN induced the upregulation of several of GST enzymes involved in the detoxification and defense against reactive compounds including xenobiotics and endogenous metabolic bioproducts 65 (Fig. 5; see also Supplementary Table 4). This is of importance given their intimate relation to neurological and neurodegenerative disorders including Parkinson disease and epilepsy 66,67 . The specific GST isoforms most affected by SFN include the glutathione S-Transferases of the alpha (Gsta1-4) and mu (Gstm1, 2, 5, 7) types which are involved in the detoxification of active electrophilic compounds by conjugation with glutathione. GST isoenzyme Gstm1 has also been shown to protect cells against TGF beta-and ischemia/reperfusion-induced apoptosis 68 and its antioxidative activity has been recently associated with lower HIV disease progression and severity in HIV infected patients 69 . In addition to GSTs enzymes, noteworthy is the fact that SFN also increased the gene expression level of the microsomal glutathione S-transferase 1 (Mgst1; Fig. 5A1 bottom panel) Mgst1 is a homotrimeric protein with glutathione transferase and peroxidase activities 70 particularly important in the central nervous system in relation to age-related increase of ROS which plays a major role as prodromal factor for the onset of neurodegenerative disorders 71 . In fact, SFN has been shown to reactivate the antioxidant response system during aging 26 . Conclusion and final remarks. Oxidants and electrophiles arising from internal metabolism and xenobiotic sources play an important role in maintaining physiological functions, cell signaling, and cellular defense mechanisms. In conclusion, excessive generation of ROS by both internal and external factors initiates events such as anti-oxidant depletion, lipid peroxidation, and cellular toxicity thereby creating a state of redox imbalance 72 . From a cerebrovascular and CNS perspective, a growing body of evidence indicate that oxidants and electrophiles are among the principal mediators in the initiation and progression of several cerebrovascular and neurodegenerative diseases. The brain microvascular endothelium at the BBB displays abundant expression of various solute carrier family and ATP-dependent transporters, besides tight junction assemblies, for enhanced nutrient/solute uptake, utilization and transport to the brain 38 .
In this contest, activation of the antioxidative response system by SFN can provide an effective counteractive tool against oxidative stress and afford protection against systemic inflammatory challenges, reperfusion/cerebrovascular injuries, and neurodegenerative diseases 14,[16][17][18][19][20][21]73 . On these lines, stimulation of redox activities and cellular metabolism by SFN helps supporting the high demand for bioenergetic fuel (in the form of ATP) and redox cytoprotection from ROS and highly reactive electrophilic substances. Thus, SFN can effectively sustain critical cellular counteractive responses to xenobiotics, oxidative stress stimuli while promoting cell viability (see Sup. Fig. 6). This latter also includes regulation of mitochondrial dynamics (encompassing fusion, fission and biogenesis) and mitochondrial transcription factors (see Sup. Fig. 7) to support the cellular activities.
While Nrf2-dipendent and independent activity of SFN have been previously demonstrated 73 , we did not investigate the specific impact of these modulatory pathways since it was deemed outside the scope of this work. However, we recognize that this is an important aspect of SFN activity that will need to be addressed in future studies. In addition, in vivo studies will be necessary to assess the broader impact of the observed changes in mitochondrial functions and cellular response across the entire neurovascular unit. Immunofluorescence. Following drug treatment, cells were rinsed with PBS and fixed with ice-cold 4% buffered formalin for 10-15 min. Cells were then washed 3X with PBS. Later cells were permeabilized using 0.02% triton 100X and subsequently blocked with 10% goat serum in PBS followed by overnight incubation at 4 °C with primary antibodies in blocking buffer (1:100-150) reactive against human or mouse proteins (Glut1, Nrf2, Nqo1 and Slc40a1). Next day, cells were washed with blocking buffer and incubated for 1 h with Alexa Fluor 488 or 555-conjugated anti-rabbit or anti-mouse antibodies (1:1000) at room temperature. Cells were washed with PBS, dried at room temperature and cover-slipped with DAPI in Prolonged Gold Anti-fade reagent (Invitrogen, Carlsbad, CA). Images were captured using an EVOS XL digital inverted fluorescence microscope at 40 × (Advanced Microscopy Group -AMG, Cat #AME3301).
Western Blotting. Total protein was collected by cell lysis in RIPA buffer with protease inhibitors and quantitated by BCA assay. Equal amounts of protein across samples were subjected to SDS-PAGE separation (4-20% graded gels) as described earlier 76 . Following gel to PVDF membrane electrotransfer of the protein bands, membranes were blocked with blocking buffer (0.1% tween-20 in TBS) containing 5% non-fat dry milk and 1% BSA. Subsequently, membranes were incubated overnight at 4 °C with primary antibodies in blocking buffer (in range of 1:200-1-500), washed repeatedly and probed with HRP-conjugated secondary antibodies (1:5000). Bands were visualized by chemiluminescence reagent using X-ray film-based autoradiography and densities were analyzed by Li-Cor Image Studio software with β-actin as a loading control.

Mouse MitoChip gene expression array.
To assess the transcriptional role of Nrf2 in the regulation of diverse gene networks specific to mitochondrial structure and function in the BBB endothelium (mitochondriome), we employed a custom-designed mouse MitoChip arrays (Agilent Technologies Inc., Santa Clara, CA) developed by National Center for Toxicological Research 33 . It consists of both gDNA and mtDNA encoding 812 genes associated with mitochondrial structure and function, 70 genes involved in both intrinsic and extrinsic apoptotic pathways, 117 genes involved in phase I (cytochrome P450 oxidoreductases), phase II (glutathione S-transferases, UDP-glucuronidases, sulfuryl transferases) of drug metabolism, and 20 housekeeping genes to give a total of 1,019 unique genes on the MitoChip. This is the upgrade of the previous mouse MitoChip of 542 mitochondria-related genes developed for evaluation of mitochondrial role in various drug-induced organ toxicities 35,36 . Following treatment, total RNA was isolated from primary cells by RNAeasy plus mini kit (Qiagen Inc, Germantown, MD) with genomic DNA decontamination 78 . RNA integrity and yield was determined by Nanodrop ™ ND-1000 (Thermo Fisher Scientific, Waltham, MA) and Agilent 2100 Bioanalyzer (Agilent Technologies Inc.). cDNA strand synthesis, labeling, and hybridization were performed at the Nationwide Children's Hospital (Columbus, OH) in a random design of samples on mouse MitoChip. A complete gene list of Mitochip genes is provided in Supplementary Table 4. Following hybridization, arrays were scanned using the Agilent DNA Microarray Scanner (Agilent Technologies Inc., Santa Clara, CA 95051, US). The images were analyzed using Agilent's Feature Extraction software (V10.7.3). Significant changes in the expression of important genes following normalization as observed on MitoChip with an absolute fold change of 1.25 and false discovery rate < 0.05 were further validated by quantitative real-time PCR 35  with gene-specific forward and reverse primer pairs (Integrative DNA technologies, Coralville, IA; Table 1) and SYBR select master mix (Life Technologies). Template-free and RT-negative samples served as internal controls. Amplification was performed in triplicates/sample on Bio-Rad CFX96 Touch Real-Time PCR system (95 °C for 5 min followed by 40 cycles of 95 °C-30 sec, 58 °C-1 min and 72 °C-1 min and a terminal reaction at 72 °C-2 min).
The threshold values (counts, Ct) for target genes and reference genes Rpl21) were determined for each sample and relative expression of each target gene was calculated by ΔΔ Ct method.
Data Analysis. Results are reported as mean ± SEM and analyzed for statistical significance by one-way ANOVA followed by Dunnett's or Tukey's multiple comparison tests using GraphPad Prism Software Inc. (La Jolla, CA, USA). Student's t-test (two-tailed, unpaired) was used to compare control vs. treatment effects. P value < 0.05 was considered for statistical significance.
MitoChip gene expression. Data extraction, pre-processing, normalization, and statistical comparison to obtain differential gene expression were performed using SAS 9.4 (SAS Institute Inc., Cary, NC). The intensity (gPro-cessedSignal) data for all samples were extracted from raw data files (generated by Agilent Feature Extraction software). The intensity data for all 1019 genes from all arrays were transformed to log2 values. The mean intensity of six replicates for each gene was calculated and used as the intensity for a respective gene. Then, an array-wise 50th percentile normalization was performed. Generalized linear model procedure (PROC GLM) in SAS was used to measure statistical significance (p < 0.05) of the contrast between control and different SFN treatment groups to determine differentially expressed genes. A relative fold change in expression level of each gene was calculated as a difference in average log2 intensity values of two comparison groups. Also, raw p-values were adjusted to False Discovery Rate (FDR). An absolute fold change greater than 1.25 and FDR < 5% was used as criteria for differentially expressed genes.
Gene Ontology analysis of MitoChip data. The interpretation of large amounts of complex microarray gene expression data and translating it into biologically meaningful information can be challenging. However, Gene Ontology (GO) has increasingly been used to find patterns within groups of genes 79 . The GO is organized as a hierarchy of annotation terms that facilitate the analysis and interpretation at different levels such as molecular function, cellular compartment, and biological process. It is a powerful tool that enables one to classify genes according to mined functional categories and statistically analyze the obtained classifications, leading to a better understanding and interpretation of microarray data.
Mitochondria-related GO terms based on biological processes or molecular functions were thus obtained from the Mouse Genome Informatics (www.informatics.jax.org) database and NCBI GO annotation (ftp://ftp. ncbi.nlm.nih.gov/gene/DATA/gene2go.gz) database. The overall significance of treatment effect on each GO term was measured by a modified meta-analysis method to combine p-values calculated for each gene within the GO term while considering the inter-gene correlation structure in each GO term 80 .
Data availability. The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.