Dimethyl fumarate reduces TNF and Plasmodium falciparum induced brain endothelium activation in vitro

Plasmodium falciparum parasite lines were derived from Malawian children with CM enrolled in a study of the clinicopathologic causes of CM run by the Blantyre Malaria Project (BMP) [1, 4]. Study subjects aged 6 months to 12 years with World Health Organization (WHO) defined CM (coma at least 1 h after termination of a seizure or correction of hypoglycaemia, asexual forms of P. falciparum parasites on peripheral blood smears and no other cause to explain the coma), were enrolled after informed consent was obtained from the parent or guardian. Venous blood was drawn into EDTA coated tubes. Malaria retinopathy status was determined by fundoscopic examination and brain swelling status was determined by MRI imaging, as previously described [13, 40, 41]. The patient’s clinical and demographic information was recorded. From the admission blood draw, parasite lines were obtained by in vitro limiting dilution cloning at 3 different cell concentrations (0.3 IE per well, 3 IEs per well, and 30 IEs per well). For each CM parasite line, parasites were expanded from positive wells at the lowest initial cell seeding concentration for that isolate. Following limited dilution cloning, the parasite lines were then expanded in vitro in Malawi and seed lots were frozen [42]. Cryopreserved parasites were shipped to Albert Einstein College of Medicine, NY. Institutional Review Board (IRB) approvals were obtained from the Albert Einstein College of Medicine, Michigan State University and from the University of Malawi College of Medicine Research and Ethics Committee.

Frozen stocks of CM derived parasite clones were thawed using standard methods and cultured in O⁺ RBCs at 5% hematocrit in supplemented RPMI 1640, including Albumax II (Gibco/BRL, Grand Island, NY) and cultivated under 1% O₂, 5% CO₂, and 94% N₂ gas mixture and at 37 °C [43]. The parasite lines were thawed and expanded for 2–5 cycles prior to freezing and then grown for an additional 10–20 cycles until the co-culture experiments with HBMVEC. P. falciparum laboratory lines and CM isolates were regularly tested and confirmed free of Mycoplasma spp. by the MycoScope PCR detection kit (Genlantis, San Diego, CA). Two long-term cultivated parasite lines with known adhesion traits were employed: ITgICAM1 (DC17 expressing line that binds ICAM1) and IT4var19 (DC8 expressing line that binds EPCR) [22]. These parasites were cultured as above, using RPMI 1640 medium supplemented with 10% Human AB serum (Thermo Fisher Scientific, Pittsburg, PA). The parasite-induced knob-like protrusions were maintained by periodic gelatin flotation, and parasite lines were synchronized by 5% d-sorbitol lysis as needed [44].

Cerebral cortex derived HBMVECs (passage 3) were purchased from Cell Systems (Kirkland, WA) and grown in a standard 75-cm² culture flask (Corning Inc, Corning, NY) coated with 0.2% gelatin (Sigma Aldrich, St. Louis, MO). The cells were maintained at 37 °C in 5% CO₂ and grown in M199 medium (Gibco/BRL, Grand Island, NY) supplemented with 1.6 mM l-glutamine, 50 units/ml penicillin, 50 μg/ml streptomycin, 50 μg/ml ascorbic acid, 25 μg/ml heparin, 5 μg/ml bovine brain extract (Lonza, Wayne, PA), 20% calf serum, 5% human AB serum, and 7.5 μg/ml Sigma growing factor (Sigma Aldrich, St. Louis, MO). Cells were generally split when they reached approximately 90% confluency by lifting with a Trypsin/EDTA solution. HBMVEC cultures were regularly tested and confirmed free of Mycoplasma spp. by the MycoScope PCR detection kit (Genlantis, San Diego, CA). Cells were maintained under 5% CO₂ at 37 °C for all experiments. HBMVEC were used between passages 7–8 for all experiments.

For HBMVEC transcriptional analysis, cells were grown to confluence in collagen coated 6 well plates. On the day of experiment, cells were rinsed with fresh media and incubated with TNF (10 ng/ml), (PEPROTECH Inc., Rocky Hill, NJ), media alone or vehicle control (0.1% DMSO, ATCC, Gaithersburg, MD) for 6 h. To examine the effect of DMF, a set of cells were preincubated with DMF (50 μM) for 1 h prior to the addition of TNF stimulation for 6 h. HBMVEC supernatants were collected and stored for cytokine measurement, cells were rinsed with cold PBS and lysed/homogenized in Trizol (Thermo Fisher Scientific, Pittsburg, PA) for RNA extraction. RNA was purified using the Midi Total RNA Purification kit (Thermo Fisher Scientific, Pittsburg, PA), and DNase I treated. RNA yields were quantified using a Nanodrop Spectrophotometer and quality assessed with the Bioanalyzer RNA 6000 Nano assay (Agilent Technologies Inc., Wilmington, DE).

HBMVEC RNA was processed by standard protocols and hybridized to a human Clariom S microarray (Thermo Fisher Scientific, Pittsburg, PA) which contains probes for 25,000 human gene transcripts at the Genomics Core of Albert Einstein College of Medicine. Gene expression profiles were generated using a robust multi-array average algorithm followed by quantile normalization and batch correction (Transcriptome Analysis Console, Thermo Fisher). Three independent experiments, each done in triplicate, were batch corrected for day of experiment. Differential gene expression between experimental conditions was determined by calculating the p value by Student’s t test and false discovery rate (FDR). Significantly differentially expressed genes that have a FDR ≤ 0.01, and fold change > 2 are reported. p value is calculated by a Fisher’s exact test right tailed using Ingenuity Pathway Analysis (IPA) (Qiagen, Redwood City, CA) to determine if the dysregulated genes are enriched in a canonical pathway. Pathways with at least 8 genes are reported. Z-score statistics indicate the activation state of the pathway.

To characterize the var gene expression in the lab-adapted CM parasite lines, an aliquot of 250–500 μl pellet at 4–5% ring-stage parasitemia from each parasite line was collected, one day prior to the HBMVEC cytoadherence experiments. Total RNA was extracted from samples resuspended in 3–6 ml Trizol LS (Thermo Fisher Scientific, Pittsburg, PA) and stored at − 80 °C. RNA was purified with RNeasy Micro Kit (Qiagen, Germantown, MD) following manufacturer’s instructions and contaminating genomic DNA was eliminated with DNase I treatment. cDNA was synthesized by reverse transcription using Multi-Scribe Reverse Transcriptase Kit and random hexamers (Thermo Fisher Scientific, Pittsburg, PA) at 25 °C for 10 min, 48 °C for 30 min, and 95 °C for 5 min. cDNA samples were considered DNA free when fluorescence remained at baseline after 30 cycles of quantitative PCR (qPCR) with seryl-t-RNA synthetase primers [45]. DBLα tags were PCR-amplified using the previously described varF_dg2 and brlong2 primers [46]. Between 40 and 50 DBLα amplicons were sequenced from each CM parasite line. To classify DBLα tags into predicted CD36 or EPCR binders, BLAST searches were conducted against a custom library of 521 annotated var genes, as described previously [23]. Predictions were considered high confidence when ≥ 4 of the top 5 hits were the same type. This methodology likely underestimates the true level of DC8-EPCR binders (group B/A chimeric var gene) as it can misclassify these DBLα tags as group B/C (CD36 binders). Additionally, the number of cysteine residues in the DBLα amplicons were quantified. Previous work has established that Group A var genes have 2 cysteine residues (C2 group) and group B and C var genes encode 3–5 cysteines with most having 4 cysteines (C4 group) [47].

To determine the number of P. falciparum genotypes in the laboratory adapted CM parasite lines DNA was isolated from IE lysed with 0.15% saponin solution and purified by phenol/chloroform organic solvents [48], and MSP1-PCR was performed as described [49]. Additionally, MSP2-PCR was performed with the M2-OF and M2-OR primers [50] and five amplicons were sequenced from each CM parasite line.

HBMVEC were grown as a monolayer on Biocoat collagen I-coated 8-well chamber slides (Thermo Fisher Scientific, Springfield Township, NJ) until 90% confluence. Trophozoite stage IEs were purified by magnetic columns (Miltenyi Biotec, Gaithersburg, MD), adjusted to 3 or 10% parasitemia and 1% hematocrit in binding media (RPMI 1640, HEPES, BSA, pH 6.4), and added to HBMVEC monolayers in duplicate wells. Co-cultures were incubated for 1 h at 37 °C under semi-static conditions, with constant horizontal agitation at 100 rpm [51]. Unbound IE were gently washed off in pre-warmed binding media, cells were fixed in 2% glutaraldehyde for 2 h at room temperature and stained in 10% Giemsa for 30 min. Number of bound IE to HBMVEC were quantified in 8 fields/well in duplicate wells. IE were manually counted and HBMVEC were automatically counted using a software assisted counter (NIS-Elements Imaging Software, Nikon, Melville, NY). For cytoadherence inhibition assays of the lab lines, monolayers of HBMVEC were incubated with monoclonal antibodies to ICAM1 mAb 15.2 (10 µg/ml) or antibodies to EPCR mAb 252 (50 µg/ml) (Novus Biologicals, Centennial CO), and their corresponding IgG isotype controls, at 37 °C for 30–45 min prior binding assays. All experiments were carried out in three independent experiments.

Data is reported as IE numbers/500 HBMVEC. Laboratory strains ITgICAM1 (ICAM1 binder) and IT4var19 (EPCR binder) served as binding controls, respectively [20, 22]. For cytoadherence under inflammatory conditions, HBMVEC were stimulated with TNF (10 ng/ml) for 20 h prior to binding assays. Binding assays were conducted at 37 °C for 30–45 min. All experiments were carried out in three independent experiments.

HBMVEC activation was measured by the quantification of interleukin-6 (IL-6) levels into cell supernatant under various experimental conditions using standard enzyme-linked immunosorbent assay (ELISA) methods [52]. Parasites were enriched for > 95% trophozoite stage by magnetic columns and added to HBMVEC confluent monolayers in a 96 well plate in M199 complete medium. To determine an optimal IE:HBMVEC stimulation ratio, 2× serial dilutions from 200:1 to 6:1 IE:HBMVEC ratios were tested. HBMVEC were cultured with uninfected erythrocytes as a negative control or TNF (10 ng/ml) (PEPROTECH Inc., Rocky Hill, NJ) as a positive control. The cells were incubated at 37 °C in 5% CO₂ for 6 and 24 h in triplicate. Cell supernatants were collected at each time point, centrifuged at 400×g for 5 min at 4 °C to remove cellular debris, and stored at − 80 °C. To evaluate the effect of DMF on IE-mediated endothelial cells activation, HBMVEC were pre-treated with 50 μM of DMF for 1 h before the co-culture with IE. In all experiments where DMF effect was evaluated the IE:HBMVEC ratio used was 12:1 for the CM derived isolates and 25:1 for the IT4var19 parasite line.

Concentrations of IL-6 in the cell culture supernatants were measured by commercially available human IL-6 ELISA set (BD Biosciences, Palo Alto, CA). The ELISAs were performed according to manufacturer’s instructions. Flat-bottom 96-well microplates were coated with anti-human IL-6 monoclonal antibody at 4 °C overnight. Cell supernatants were assayed as neat or one- to fourfold dilutions. 3,3′,5,5′-tetramethylbenzidine (TMB) substrate reagent set (BD Biosciences) was used to detect the streptavidin–horseradish peroxidase (BD Biosciences) reaction and optical densities were measured at 450 nm with 750 nm wavelength corrections in a microplate reader. The experiments were carried out in triplicate wells in at least three independent experiments.

Expression of endothelial cell surface receptors was determined by flow cytometry using antibodies to ICAM1 (Bio-Rad, Raleigh, NC), VCAM1, EPCR, E-selectin and CD31 (all purchased from R&D systems, Minneapolis, MN).

HBMVEC monolayers were rinsed with PBS and incubated with pre-warmed non-enzymatic cell detachment solution (Sigma, Burlington, MA) for 15–30 min at 37 °C and gently detached with a cell scraper. HBMVEC were collected in polystyrene tubes, rinsed with cold PBS at 300×g for 4 min at 4 °C and transferred to v-bottom 96 well plates. Cells were incubated with live/dead violet dye (Thermo Fisher Scientific, Fair Lawn, NJ) in 100 μl of FACs buffer (PBS, 10% FBS, 0.2% EDTA) for 30 min and rinsed with cold PBS. HBMVECs were incubated with human Fc receptor blocking (eBioscience, Thermo Fisher Scientific, Fair Lawn, NJ) for 20 min, and multiplexed primary antibodies and fluorochrome conjugated antibodies were added to HBMVEC and incubated for 1 h. Excess antibody was washed and HBMVEC incubated with fluorochrome conjugated secondary antibodies for 30 min. Cells were washed twice with cold PBS for 5 min, centrifuged at 300×g for 5 min at 4 °C and fixed with IC fixation buffer (Thermo Fisher Scientific, Fair Lawn, NJ) for 20 min. Excess fixative was washed and cells were resuspended in 1% BSA PBS for analysis. Cell cytometry data was acquired on LSRII flow cytometer and analysis performed with FlowJO software. Assays were performed in 3 to 5 independent experiments.

Modulation of the endothelial NFκB nuclear translocation by DMF and TNF was evaluated with confocal immunofluorescence. HBMVEC were deposited onto lysine and collagen-coated glass bottom dishes (MatTek Corporation, Ashland, MA) and cultured to confluency. HBMVEC monolayers were stimulated with TNF (10 ng/ml) or left unstimulated in M199 medium alone for 30 min or pre-treated with DMF (50 μM) or vehicle DMSO (0.1%) for 1 h prior to TNF stimulation. Cells were rinsed and then fixed with 4% paraformaldehyde (PFA) (Sigma, Burlington, MA) for 30 min, permeabilized and blocked with 0.2% Triton X-100 (Sigma, Burlington, MA), 10% normal goat serum (Thermo Fisher Scientific, Fair Lawn, NJ), 1X PBS and 0.05% sodium azide for 1 h at room temperature. They were incubated overnight with anti-NFκBp65 (Santa Cruz Biotechnology, Dallas, TX) at 4 °C followed by Alexa Fluor-488 (1:300, Invitrogen Carlsbad, CA), and the F-actin probe Alexa Fluor 680 Phalloidin (1:40, Thermofisher Scientific, Fair Lawn, NJ) for 1 h at room temperature. Cells were fixed with 4% PFA and nuclei were stained with 4′-6-diamidine-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA), and mounted with antifade reagent ProLong Gold (Molecular Probes, Thermo Fisher Scientific, Fair Lawn, NJ). Images were collected with a Leica SP2 inverted confocal microscope (Leica Microsystems) under a 63X oil-immersion objective. Experiments were carried out in at least three independent instances.

Statistics were performed using Prism 6 (GraphPad Software). Values reported are the mean ± SEM. All analyses utilized an alpha = 0.05. One-way or two-way ANOVAs with either Sidak’s or Turkey’s post hoc comparisons were used to compare samples where appropriate, as indicated in the figure legends.

HBMVEC transcriptional profiling to define the global effects of DMF treatment on TNF activated cells was conducted. Gene expression profiles of HBMVEC subjected to TNF compared to DMF pre-treatment, and media alone (control) segregated by experimental group in a Principal Component Analysis (PCA) (Fig. 1a). TNF treatment of HBMVEC compared to media alone, altered the expression of 514 genes (fold change > 2, FDR ≤ 0.01, p < 0.005, see Additional file 1 for complete gene list). TNF upregulated cell adhesion molecules, including intercellular adhesion molecule 1 (ICAM1) and vascular cell adhesion molecule 1 (VCAM1). It also induced inflammatory mediators, such as chemokine (C-X-C motif) ligand 6 (CXCL6), Interleukin 1 Beta (IL1β) and NFκB2. TNF treatment upregulated the Neuroinflammation Signaling Pathway, High Mobility Group Protein 1 (HMGB1), NFκB, Triggering Receptor Expressed on Myeloid cells 1 (TREM1), Transforming Growth Factor Beta (TGFβ) and IL-6 signaling pathways, and downregulated the Peroxisome Proliferator-activated Receptor (PPAR) Signaling Pathway when compared to controls (Ingenuity Canonical Pathways, p < 0.01, Z score at least 2 SD above the mean, see Additional file 2 for complete list of pathways).

To determine the effect of DMF treatment on activated cells, HBMVEC were pretreated with DMF (1 h) prior to TNF treatment and compared to TNF treatment alone, which resulted in dysregulation of 348 genes (fold change > 2, FDR ≤ 0.01, p < 0.005, Table 1, see Additional file 3 for complete gene list).

DMF pretreatment prior to TNF stimulation increased transcript levels of several genes found in antioxidant pathways, including heme oxygenase 1 (HMOX1), sulfiredoxin 1 (SRXN1) and oxidative stress induced growth inhibitor 1 (OSGIN1) (Table 1, fold change > 2, FDR ≤ 0.01, p < 0.005, see Additional file 3 for complete list of pathways) as well as the NRF2-mediated oxidase stress response (Fig. 1b, p < 0.01, Z score at least 2 SD above the mean, Additional file 4: Table S4). Extracellular signal-regulated kinase 5 (ERK5) pathway, which is induced in response to oxidative stress is a central mediator of cell survival and apoptotic regulation, was also upregulated [53]. Moreover, DMF pretreatment significantly downregulated VCAM-1 and the parasite cytoadhesion receptor ICAM1, as well as transcripts involved in neuroinflammation signaling (IL-6, TREM1, PPAR, NFκB) (Fig. 1b, p < 0.01, Z score at least 2 SD above the mean, see Additional file 4 for complete gene list). IL-6, an inflammatory cytokine, and a marker of cellular activation levels in the HBMVEC supernatant to confirm the transcriptional data was measured. IL-6 levels increased in a TNF dose dependent manner. IL-6 concentration was significantly less in the supernatant of the DMF pretreated, TNF activated cells (Fig. 2a) compared to TNF activated cells. Moreover, when DMF was added simultaneously with TNF stimulation of HBMVEC, IL-6 levels were significantly reduced (Fig. 2b). To corroborate the DMF inhibitory effect on the NFκB pathway in HBMVEC, NFκB cellular localization by confocal imaging was performed. TNF stimulation resulted in translocation of NFκB from the HBMVEC cytoplasm to the nucleus as expected. In contrast, 1 h pretreatment with DMF treatment inhibited the TNF induced nuclear translocation of NFκB (Fig. 2c).

Previous work from a paediatric CM population has shown that circulating P. falciparum-IE utilize ICAM1 and EPCR to bind to primary HBMVEC [24]. During malaria infection, there is widespread upregulation of vascular endothelial ICAM1 due to cytokines such as TNF [54]. To study whether DMF can modulate the expression levels of parasite cytoadhesion receptors, HBMVEC were treated with TNF plus or minus pre-treatment with DMF. Whereas TNF increased the HBMVEC expression of ICAM1, VCAM1, E-selectin and decreased expression of EPCR surface levels quantified by flow cytometry, there were no changes in the constitutive endothelial marker CD31. DMF pretreatment strongly attenuated the TNF-induced upregulation of ICAM1, and VCAM1 (p < 0.01, one-way ANOVA test) (Fig. 3a), it did not significantly alter EPCR, E-selectin or CD31 surface expression levels.

To determine whether DMF could modify IE binding to TNF-stimulated HBMVEC, the binding of two long-term cultivated parasite lines with known adhesion traits for ICAM1 and EPCR was examined. Confluent monolayers of HBMVEC were stimulated with TNF for 20 h. As expected, there was a major binding increase (~ fivefold) in the ITgICAM1 parasite line (ICAM1 binder) to TNF-activated HBMVEC (Fig. 3b). By comparison, there was a non-significant increase (~ 1.5-fold) in IT4var19 (EPCR binder) parasite line (Fig. 3b). The respective ICAM1 and EPCR-binding dependencies of the ITgICAM1 and IT4var19 parasite lines to HBMVEC was confirmed using blocking antibodies (see Additional file 5). DMF pretreatment abolished the increased binding of the ITgICAM1 parasites (p < 0.0001, one-way ANOVA, Fig. 3b) and had no effect on the binding of IT4var19 parasites to TNF activated HBMVEC (p = 0.99, one-way ANOVA).

The binding of parasite lines derived from Malawian children with CM to HBMVEC (Fig. 4a) was then examined. Parasite lines were derived from 8 comatose children in the research ward of the BMP who were enrolled in 2011–2014. Bacteraemia was excluded in all by blood culture evaluation of the admission blood sample. Five of the parasite lines were from malaria retinopathy positive (Ret+) children (3173, 3180, 3194, 3207 and 3080) and three were from Ret− children (3005, 3013, and 3039). All patients had Blantyre Coma Score of ≤ 2 and anaemia upon admission to the BMP and sustained prolonged coma (Table 2). All patients survived.

By MRI, two patients had mild to moderate severe brain swelling. After limited dilution cloning and in vitro expansion, 7 of 8 clonal parasite lines had a single MSP2 sequence and one parasite line (3005) had two MSP2 genotypes, indicating it was multi-clonal.

Previously, it has been shown that EPCR-binding var transcripts (DC8 var and group A var) are increased in IE from paediatric CM patient cohorts [23]. RNA was isolated from the laboratory adapted CM parasite isolates and sequenced DBLα tags. All eight CM parasite lines expressed a mixture of var transcripts, including predicted EPCR binding (DC8 and group A) and CD36 binding transcripts (group B and C) (Fig. 4, see Additional file 6). The number of different var transcripts ranged from 7 to 21 per CM parasite line, indicating that the parasite lines had likely undergone var gene switching during cultivation following the limited dilution cloning. As observed previously in this patient population [23] there was a high degree of genetic diversity between DBLα tags amplified from the 8 CM parasite lines. Indeed, only one tag was identical between two parasite lines (3194D and 3207s15, predicted CD36 binder).

The binding of CM derived isolates to HBMVEC was measured. At the time of the binding assay, all the parasite lines expressed a high proportion of group A or DC8-EPCR binding var transcripts (range 33–88%, Fig. 4b). Additionally, group A or DC8-EPCR binding var transcripts were the dominant transcripts (> 50% transcripts) in all the parasite lines, except for 3194 (Fig. 4b). The binding levels of the eight CM-derived parasite lines ranged from 191 to 1270 IE per 500 resting HBMVEC (Fig. 5a). TNF activation of HBMVEC induced a significant increase in cytoadherence of 7/8 CM parasite lines (p = 0.0020, Wilcoxon matched-pairs signed rank test) (Fig. 5a, range 237 to 2860 IEs per 500 HBMVEC). There was no significant differences between isolate binding based on donor retinopathy status (baseline binding, p = 0.60 Mann–Whitney test) to either resting or TNF stimulated HBMVEC (p = 0.60 Mann–Whitney test). DMF pre-exposure was tested to determine if it reduced the binding of CM derived isolates to TNF-stimulated HBMVEC. Parasite line 3039 demonstrated reduced binding under DMF treatment (p < 0.001) however, overall DMF did not reduce binding levels of parasite lines to activated HBMVEC (Fig. 5b, p = 0.99, one-way ANOVA).

Previous work has shown that IEs can activate endothelial cells to secrete proinflammatory cytokines, such as IL-6 [28, 52]. To test the effect of DMF to modulate endothelial cell activation by parasites, IE were co-incubated with HBMVEC for 24 h, which allowed the CM parasite lines to cycle through schizogony with associated erythrocyte haemolysis. Consistent with previous findings the IT4var19 parasite line induced minimal IL-6 release from primary HBMVEC (Fig. 6a) [55]. By comparison, the CM parasite lines induced a range of IL-6 release (from 758 to 4158 pg/ml) (Fig. 6a). DMF inhibited the release of IL-6 in all the HBMVEC parasite co-cultures, on average by two-fold with mean IL-6 levels decreased from 3824 to 1629 pg/ml (p < 0.001) (Fig. 6b).