Bilirubin Oxidation End Products (BOXes) Induce Neuronal Oxidative Stress Involving the Nrf2 Pathway

Delayed ischemic neurological deficit (DIND) is a severe complication after subarachnoid hemorrhage (SAH). Previous studies have suggested that bilirubin oxidation end products (BOXes) are probably associated with the DIND after SAH, but there is a lack of direct evidence yet even on cellular levels. In the present study, we aim to explore the potential role of BOXes and the involved mechanisms in neuronal function. We synthesized high-purity (>97%) BOX A and BOX B isomers. The pharmacokinetics showed they are permeable to the blood-brain barrier. Exposure of a moderate concentration (10 or 30 μM) of BOX A or BOX B to isolated primary cortical neurons increased the production of reactive oxygen species. In the human neuroblastoma SH-SY5Y cells, BOX A and BOX B decreased the mitochondrial membrane potential and enhanced nuclear accumulation of the protein Nrf2 implicated in oxidative injury repair. In addition, both chemicals increased the mRNA and protein expression levels of multiple antioxidant response genes including Hmox1, Gsta3, Blvrb, Gclm, and Srxn1, indicating that the antioxidant response element (ARE) transcriptional cascade driven by Nrf2 is activated. In conclusion, we demonstrated that primary cortical neurons and neuroblastoma cells undergo an adaptive response against BOX A- and BOX B-mediated oxidative stress by activation of multiple antioxidant responses, in part through the Nrf2 pathway, which provides in-depth insights into the pathophysiological mechanism of DIND after SAH or other neurological dysfunctions related to cerebral hemorrhage.


Introduction
Subarachnoid hemorrhage (SAH) is a serious cerebrovascular complication with complex underlying mechanisms inflicting brain perfusion and function. Despite great progress in the understanding of SAH pathophysiology and management of ruptured aneurysms, SAH remains a severe and significant health problem [1]. SAH causes early brain injury, which may be followed after 7 to 14 days by delayed ischemic neurological deficit (DIND). DIND is the leading cause of morbidity and mortality in the patients who survive the initial impact of SAH and have had their aneurysm effectively treated [2]. However, the etiology and pathophysiology of SAH and DNID remain incompletely understood.
Accumulating evidence has shown that oxidative stress, acute and subsequent consisting immunological response, and other factors collectively cause the severe adverse effects of SAH [3,4]. Erythrocyte lysis, a major initiator, is the release of heme and its subsequent degradation cascade, which collectively exert a strong effect on the affected neurons and their electrical excitability [5][6][7]. Various heme and heme degradation products (HHDPs) were identified in the cerebrospinal fluid (CSF) from the patients after SAH or other types of brain hemorrhage [8][9][10][11] and serum from hepatic failure patients [12].
Notably, an array of bilirubin oxidation end products (BOXes) with small molecular weight were identified [11,13,14] and detected in the CSF of a patient with DIND after SAH [8,9,15]. Several studies using crude extraction or synthesized BOXes have shown that BOXes constrict mouse cortical blood vessels and damage the normal contractile of vascular smooth-muscle cells [16,17], suggesting that BOXes are involved in the delayed cerebral vasospasm, a potential cause of DIND. Our previous study showed that a crude BOXes mixture regulates the blood tone by modulation of Ca 2+ -and voltage-gated K + channels (Slo BK Ca ) [18]. However, the direct impact of BOXes on neuronal cells remains to be examined.
In this study, we hypothesized that BOX A or BOX B are important functional regulators of neurons during the progression of DIND after SAH. Therefore, we synthesized BOX A and BOX B following the reported protocol [14,19] and examined their biological effects on cultured primary cortical neurons and neuronal cell lines to elucidate the underlying molecular mechanisms.

BOXes Synthesis and B/P Study. BOX A and BOX B were
synthesized as shown in Figure 1(a). Compound 1 (bromocitraconic anhydride) reacts with triphenylphosphoylide in toluene with reflux. Because the two carbonyl reaction sites are different, two products can be obtained: 3 is the primary product, 2 is the by-product, and two of which can be separated by silica gel column chromatography. Compounds 2 and 3 have the reactivity of palladium-catalyzed coupling reaction and amine acetate substitution reaction (the reaction order is not required). Then, compounds 6 and 7 were hydrolyzed into carboxylic acid intermediates by alkali hydrolysis, and then the target products BOX A and BOX B were obtained by an amino substitution reaction with ammonium chloride. The target products are purified by HPLC and identified by LC/MS/MS. BOX A or BOX B were measured by a validated HPLC-MS/MS method. The assay used an automated system followed by HPLC (Shimadzu LC30AD) using a Waters XSELECT UPLC HSS T3 2.5 μm 2:1 * 50 mm column with gradient elution using mobile phase A containing 0.1% formic acid in water and mobile phase B (0.1% formic acid in acetonitrile) at a flow rate of 500 μL/min. For pharmacological kinetics study, a single dose of 12 mg/kg body weight BOX A or BOX B (powder dissolved in 5% DMSO : 95% saline (pH = 11) [20,21] with final concentration 2.4 mg/mL) was injected in vein (i.v.) with 5 mL/kg body weight in male ICR mice (n = 3) obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). After 30 min, the plasma concentrations (ng/mL) and the brain concentrations (ng/g) were measured. For preparations of mouse plasma and brain samples, 100 μL calibration curve samples in single, QC samples in duplicate, and mouse plasma samples were mixed with 300 μL acetonitrile containing Internal standard (50 ng/mL of propranolol, 200 ng/mL of tolbutamide, and 50 ng/mL of diclofenac) in Eppendorf tubes. After the mixture was vortexed for 1 min, then centrifuged for 10 min, transfer 50 μL supernatant to a 96-well plate with 100 μL pure water, shaking for 10 min, and finally inject 10 μL into LC-MS/MS system. For brain samples, brain samples were added with 5 folds (w/v) phosphate buffer (100 mM, pH 7.4) in terms of the weight of samples, then homogenated. Subsequent procedures were performed as described for plasma. The brain concentration to the plasma concentration was calculated as B/P value. The animal study including the use of embryos for primary neuron cultures was carried out in accordance with regulations for animal experimentation and were approved by the Animal Committee of Tongren Hospital affiliated to Shanghai Jiao Tong University School of Medicine.
For other assays in this study, the ready-for-use BOX A or BOX B solution was prepared as 1,000 times in DMSO as the final concentration in administered cells and stored at -20°C in the dark. Therefore, the final concentration of DMSO in all treatments was 0.1%.

Primary Cortical Neurons and Neuroblastoma Cell Lines.
Mouse primary cortical neurons (PCNs) were cultured as previously described [22] with minor modifications. Briefly, E17.5~E18.5 embryos of either sex were dissected from pregnant C57BL/6J mice for cortex dissection in cold Dulbecco's Modified Eagle Medium (DMEM). The dissected cortices were digested with~0.25% trypsin solution (Gibco) for~12 min and then neutralized by DMEM containing 10% fetal bovine serum (FBS, Gibco) and agitated to make single suspension by pipetting up and down with cutting round tips. The cell suspension was pelleted by centrifugation at 1000 rpm, resuspended in complete medium, and passed 2 Oxidative Medicine and Cellular Longevity through a 70 μm cell strainer to exclude cell debris and aggregates. Cells were seeded at a density of 5 * 10 4 /cm 2 in culture plates or dishes that precoated with 40 μg/mL poly-D-lysine (Sigma-Aldrich) and regarded as 0 days in vitro (DIV). About 2~4 h after plating, the cell culture medium was replaced with neural basal medium containing 1% penicillin/streptomycin, 1% Glutamax™, and 2% B27 supplement (Gibco). Subsequently, the medium was half-changed every 3 days till PCNs developing at 8~11 DIV for the compound treatment or analysis. Human neuroblastoma SH-SY5Y and mouse neuroblastoma Neuro2a cells were obtained from the Shanghai Cell Bank of the Chinese Academy of Science (Shanghai, China) and maintained in DMEM containing 10% FBS and 1% penicillin/streptomycin in appropriate culture plates.

Cell Viability
Assay and TUNEL Assay. Cell viability was determined by using a Cell Counting Kit-8 (CCK-8; DOJINDO) to test the impact of BOX A and BOX B according to the manufacturer's instructions. Briefly, the CCK-8 (10 μL/well) was added to 96-well plate culture dishes and incubated for 2 h for detection by the multifunctional microplate detector (BioTek). Data was interpreted as the percentage of control (vehicle).
For the TUNEL assay, the BOX A-or BOX B-treated cortical neurons were fixed in 4% paraformaldehyde and detected by In Situ Cell Death Detection Kit, Fluorescein (Roche). Before mounting, cells were counterstained with Hoechst33342 (DOJINDO, Japan). The apoptosis rate was calculated as the percentage of TUNEL positive cells to the Hoechst33342 stained cells.
2.5. ROS Detection. DCF-DA (10 μM) was added to the BOX A-or BOX B-or vehicle-treated cells in multiplate incubating for 30 min, subsequently the free probes was washed outandcells were further incubated for 30 min for readouts by using the fluorescence filter (Ex/Em: 488/525). The arbitrary fluorescence intensity was interpreted as the percentage of vehicle-treated group cells. For the ROS detection by imaging, similar incubation with DCF-DA was mentioned above and photographed by fluorescence microscopy (Olympus, Japan).

Mitochondrial Membrane Potential Measurement.
Briefly, SH-SY5Y cells were seeded to grow on 6-well culture plate 24 hr prior to the indicated concentrations of BOX A or BOX B. 48 hr later, mitochondrial membrane potential measurements was carried out with the BD™ MitoScreen (JC-1) Kit (Cat. No. 551302, BD Pharmingen™) according to manufacturer's instruction. The results were analyzed by using BD Accuri C6 Software (TreeStar, San Carlos, CA, USA). 3 Oxidative Medicine and Cellular Longevity the quality control by Agilent 2100 bioanalyzer (RIN > 6:0 and 28S/18S ≥ 0:7) could be qualified to subsequent Agilent SurePrint G3 Mouse Gene Expression 8 × 60 K. The original array figures were subjected to data normalization. After normalization, fold change (multiple of expression difference) and Student's t-test were used for screening and significance statistics (Sinotech Genomics, Shanghai, China).
2.8. Reverse Transcription and qPCR. RNA isolation and RT-qPCR were performed as previously described [23]. Briefly, RNA was extracted by Trizol reagent, 1 μg total RNA was reversed transcribed by ReverTra Ace kit (TOYOBO), and the cDNA was used to quantify by the lightcycle480 system (Roche) with 2 × Power SYBRgreen mastermix (Applied Biosystems, Carlsbad, CA, USA). Quantification of gene  Oxidative Medicine and Cellular Longevity expression was calculated by normalization to GAPDH using the 2 -ΔΔCt method. Primer sequences for qPCR could be available on request.
2.9. Protein Extraction, Western Blotting, and Quantification. As described previously [23], cells were lysed in Glo Lysis Buffer (Promega) supplemented with protease inhibitor and phosphatase inhibitor cocktails (Roche). Next, 20-60 μg of total proteins (as determined by BCA assay kit) was resolved by 10% or 12% SDS-PAGE gel, transferred onto a 0.2-μm PVDF membrane (Millipore, CA, USA), and probed with antibodies as indicated in the figures.
2.11. Data Analysis and Statistics. All figures photographed or cropped were assembled by using the Adobe illustrator CS5 software. All quantitative data was expressed as a percentage of control or vehicle (mock). Statistical analysis was performed on raw data for each group by Student's t-test or ANOVA with multiple comparisons post hoc analysis as indicated in the respective figure legends by GraphPad Prism 8. The resulting P values are provided solely as data descriptors without any inferential intent.

BOX A and BOX B Are
Permeable to the Blood-Brain Barrier. To address the physiological and pathophysiological role of BOXes in brain neurons, we synthetized BOX A and BOX B (Figure 1(a) and Supplemental Figure 1) with a predicted mass~180 Da, which is consistent with the previous reports [14,19]. HPLC separation and subsequent UV-vis spectroscopic measurement verified a purity of more than 97%. Pharmacokinetics studies in ICR male mice showed a moderate distribution in brain and plasma (Figure 1(b)), and the calculated brain/plasma (B/P) ratio showed a more pronounced brain enrichment for BOX B (Figure 1(c)). These results suggested that BOX A or BOX B derived from the ruptured cerebral artery are permeable through the blood-brain barrier and have the potential to affect neuronal function.   Oxidative Medicine and Cellular Longevity to monitor the impact of BOXes on neuroblastoma cells and primary cortical neurons. Regarding cell viability, there was no obvious difference between vehicle and BOXes-treated (1~60 μM) in cortical neuron cells (Figure 2(a)). Similar results were obtained for SH-SY5Y and Neuro2a (Figure 2(b)). In addition, BOX A or BOX B at 30 μM had no apparent impact on cell apoptosis for~48 hr, evaluated with a TUNEL assay (Figures 2(c) and 2(d)) and with a Cleaved Caspase 3 (Figure 2(e)). These results showed no effects on the cell viability and apoptosis in cortical neurons.

BOX A and BOX B Induce Oxidative Stress in Primary
Cortical Neurons. Studies in hepatocytes have shown that BOXes impact on the GSH level [12], indicating its role in redox regulations. To study whether BOX A and BOX B can cause neuronal oxidative stress, 2′,7′-dichlorofluorescein diacetate (DCF-DA), a fluorescence indicator of oxidant generation, was employed to detect reactive oxygen species (ROS) in BOXes-treated primary cortical neurons. The fluorescence showed that both BOX A and BOX B increased DCF-DA fluorescence intensity more than vehicle applications (Figure 3(a)). Similar results were obtained in DCF-DA-labeled primary cortical neurons using a multiplate reader (Figure 3(b)). These results thus suggested a more pronounced ROS production in the presence of BOX A 3.4. BOX A and BOX B Lower the Mitochondrial Potential in SH-SY5Y Cells. The impact of BOXes on the cellular ROS level stimulated the question if mitochondrial functions might be involved. We therefore measured the mitochondrial potential with the indicator JC-1 with and without applica-tion of BOXes. In a flow-cytometry assay to detect the percentage of the aggregated and monomer JC-1 probes, SH-SY5Y cells were treated with BOX A or BOX B. Both BOX A and BOX B increased the fraction of monomer probes (green channel) in a concentration-dependent manner (Figures 4(a) and 4(b)).

BOX A and BOX B Activate the Nrf2 Pathway.
To elucidate the molecular mechanism underlying the BOX A-or BOX Binduced oxidative stress in neurons, a high-throughput gene expression array was used to screen differentially expressed genes (DEGs) between the BOXes-and vehicle-treated cells (Supplemental Table 1). The Gene Ontology (GO) analysis of DEGs showed that expression of those genes related to oxidoreductase activity, oxidation-reduction process, oxidative stress, and cellular response to hydrogen peroxide was greater in the BOX A-or BOX B-treated cells (Supplemental Figure 2). In addition, the pathway analysis used DEGs through GeneAnalytics (http://geneanalytics.genecards.org) showed the Nrf2-related pathway and oxidative stress were highly overpresented in either BOX A or BOX B (30 μM each) treated-primary cortical neurons ( Figure 5(a)). Furthermore, a subset of DEGs involved by Nrf2 and oxidative stress were clustered as shown in the heat map analysis ( Figure 5(b)). Subsequently, RT-qPCR analysis further confirmed the expression changes of several aforementioned genes ( Figure 5(d)).

BOXes Promote Nrf2 Nuclear Translocation and Protein
Expression of Antioxidant Responsive Genes. To further characterize the molecular pathway, Nrf2 immunofluorescence  Oxidative Medicine and Cellular Longevity staining was employed to detect the Nrf2 translocation in human neuroblastoma cell SH-SY5Y after BOXes administration. The immunostaining results showed that Nrf2 accumulated in the nucleus of the BOX A or BOX Btreated cells but not the control cells (Figures 6(a) and  6(b)). In addition, immunoblotting analysis showed that the protein expression of downstream signals of Nrf2 including HO-1, GCLM, BLVRB, GSTA3, and SRXN1 was also markedly increased ( Figure 6(c)).

Discussion
The cellular mechanisms underlying SAH are diverse, including oxidative stress, inflammation, brain edema, and apoptotic neuronal cell death [3]. One major factor, however, is the red blood cell lysis, the release of heme, and the subsequent degradation cascade, which collectively have a strong impact on the affected neurons and their electrical excitability [5][6][7]. A series of HHDPs have been recently identified,  7 Oxidative Medicine and Cellular Longevity and the BOXes were found to be clinically associated with the progression of DIND after SAH [8,9,18], but their direct roles in neurons remain largely unknown. In this study, we synthesized BOX A and BOX B according to the protocols reported previously [14,19] and observed their impacts on primary neurons and neuronal cell lines. The findings include an increase of ROS production and depolarized mitochondrial membrane potentials triggered by BOXes, which activate the Nrf2 signaling pathway to mediate the antioxidant response in cortical neurons and neuroblastoma cells.
To surmount, a total synthesis method with small modifications (Figure 1(a)) was employed to the bottom-up synthesis of BOX A and BOX B from bromocitraconic anhydride and obtained high-purity BOXes (>97%). Furthermore, the pharmacokinetics studies of synthesized BOXes showed a very high concentration distribution between brain and peripheral blood, and BOX B had a much larger B/P ratio than BOX A 30 min after in vein injections (Figures 1(b) and 1(c)), which directly demonstrates that BOXes are transported across the blood-brain barrier. There is accumulated evidence that BOXes are closely related to the pathophysiology of DIND after SAH [8,9]. Seidel et al. used a very high dose of BOXes (>500 μM) and found a change in the morphology and the redox potential (100 μM) of hepatocytes [12]. Here we explored the direct role of BOXes in isolated cortical neurons and found that a moderate concentration of BOXes (10-30 μM) induces ROS production ( Figure 3) without obvious alterations in cell viability and apoptosis of neurons (Figure 2), which is consistent with the previous report that a moderate concentration (~25 μM) of BOX A or BOX B did not affect SH-SY5Y cell viability [24]. Of note, the mitochondrial potential was decreased in a dose-dependent manner after treatment with BOX A or BOX B (Figure 4). The impact of BOXes on ROS generation may advance a concept that BOXes trigger oxidative stress and affect the neurons and eventually lead to DIND after SAH. This is consistent with previous findings that BOXes alter the redox potential in hepatocellular cells [12] and oxidative stress as well as BOXes significantly increased after intracerebral hemorrhage [4]. Yet, the underlying molecular mechanisms are possibly different and require further studies.
Redox homeostasis is tightly regulated by many endogenous redox systems, the cellular antioxidant response mechanisms, or absorption of an antioxidant substance from outside [25]. In neurons and many other cell types, Nrf2 is a master regulator of redox sensors, kept at very low levels via complexing with Keap1 and to be subjected to proteasome-mediated degradation at basal conditions. Acti-vation of Nrf2 leads to translocation to the nucleus to initiate antioxidative response gene transcriptions and phase I/II enzyme expression [26,27]. In addition, the Nrf2 pathway has been shown to be involved in bilirubin-mediated oxidative stress [28]. In the present study, the bilirubin metabolites BOX A and BOX B triggered the oxidative stress and induced several oxidative stress-responsive genes, including Hmox1, Blvrb, Gclm, Srxn1, and Gsta3. This induction is likely due to Nrf2 activation and translocation to nucleus ( Figure 6) to activate the antioxidant response element (ARE) transcriptional cascade ( Figure 5) [29]. The ARE-regulated genes including Hmox1, Blvrb, Gclm, Srxn1, and Gsta3 are extensively investigated and contribute to oxidative stress [29][30][31][32]. Inducible Hmox1 knockout mice have a smaller cerebral hemorrhagic lesion [33], which supports the toxicity of higher levels of BOXes that cause the upregulated Hmox1 in cortical neurons from both mRNA and protein levels ( Figures 5(b), 5(c), and 6(c)). Intriguingly, our findings showed that the Nrf2 accumulation and nuclear translocation and subsequent transcription cascade is triggered by BOXes, which to some extent, is contrary to a previous study that the Nrf2 knockout mice have an enlarged cerebral hemorrhagic lesion [34], indicating a cytoprotective feedback mechanism in BOXes treated neurons. Therefore, BOXes driving the oxidative stress and perturbing the redox homeostasis lowers mitochondrial membrane potential, which may trigger an antioxidative response pathway (Nrf2): Nrf2 is translocated to the nucleus to initiate antioxidative responsive gene transcription, to antagonize this process to maintain the attacked cell survival (Figure 7). Collectively, Nrf2 activation triggered by BOXes provides insight into the pathophysiological mechanism of DIND after SAH or other neurological dysfunctions related to cerebral hemorrhage.

Conclusion
Using synthesized high-purity BOXes and cultured primary cortical neurons, we firstly provide evidence that BOXes directly increase both oxidative and antioxidative signals in neurons. Their mitochondria membrane potential lowering effect, a well-established proapoptosis marker, strongly suggests that BOXes may compromise neuronal functions after SAH. Further research is required to understand the function of BOXes in the development and progression of DIND.

Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
All authors declare no competing interests.