Neuregulin-1 attenuates experimental cerebral malaria (ECM) pathogenesis by regulating ErbB4/AKT/STAT3 signaling

We previously reported that mortality associated with HCM correlates with increased serum CXCL10 levels and that treatment of both endothelial hCMEC/D3 and neuroglial M059K cells with CXCL10 at a physiologically relevant concentration of 0.02 μg/ml in vitro induced apoptosis [6]. To test the protective effects of NRG-1, we treated hCMEC/D3 and M059K cells, respectively, with CXCL10 and NRG-1 ranging from 50 to 250 ng/ml for 24 h and assayed for apoptosis by TUNEL. NRG-1 attenuated apoptosis of hCMEC/D3 and M059K cells induced by CXCL10 at concentrations as low as 50 ng/ml and up to a maximum effect at 250 ng/ml (Fig. 1a). As a follow-up to our first report on the induction of apoptosis by heme in human brain microvascular endothelial cells [7], we next tested whether NRG-1 had anti-apoptotic effect against hCMEC/D3 and M059K cells treated with heme. When M059K cells were treated with heme at the concentrations used for hCMEC/D3 endothelial cells [7], M059K cell apoptosis occurred at around 20%. Although apoptosis was modest compared with that in endothelial hCMEC/D3 cells (52%) induced by heme [7], protection of apoptosis by NRG-1 was significant (Fig. 1b). Furthermore, NRG-1 attenuated apoptosis of both hCMEC/D3 and M059K cells either induced by CXCL10 or heme in a dose-dependent manner. In agreement with Lok et al. [27], we used 100 ng/ml of NRG-1 in the following experiments because at this concentration more than 50% of apoptotic cells were rescued (Fig. 1a, b). We next conducted a dose response assay with heme ranging from 10 to 60 μM to assess protection of NRG-1 against heme-induced apoptosis. Both hCMEC/D3 and M059K cells were treated with 10 to 60 μM of heme along with 100 ng/ml NRG-1 for 24 h followed by measurement of cell viability using MTT assay. NRG-1 attenuated 20–30% of heme-induced hCMEC/D3 cell death at each heme concentration (Fig. 1c). Similar observations were made when neuroglial M059K cells were treated with different doses of heme, reducing 8 to 20.6% cell death in the presence of 100 ng/ml of NRG-1 for 24 h (Fig. 1d). Figure 1e indicated that hCMEC/D3 used in the experiments express the characteristic endothelial cell marker von Willebrand factor VIII (VWF).

The BBB functions as a barrier to restrict movement of potentially toxic substances from the circulation to the central nervous system [28]. It has been demonstrated that a polymerized cell-free Hb (HbG) (extracellular Hb) triggers BBB disruption and oxidative stress in a guinea pig exchange transfusion model [29]. However, it is unclear whether hemoglobin-derived heme, which induces cytotoxic oxidative stress, alters the permeability/integrity of BBB and/or disrupts intercellular tight junctions (TJs) of endothelia in addition to promoting endothelial cell and astrocyte apoptosis. We treated hCMEC/D3 cells grown in a monolayer with 30 μM heme and NRG-1 (100 ng/ml) followed by staining with antibodies against tight junction proteins claudin5, ZO-1, and occludin. Claudin5 was located on cell membranes (arrows) while the membrane-bound staining disappeared after the cells were treated with 30 μM heme for 24 h (Fig. 2a, b). NRG-1 treatment prevented reduction of membrane-bound claudin5 by heme (Fig. 2c, d). The fluorescence intensity of claudin5 signals in hCMEC/D3 cells treated with heme, NRG-1, heme/NRG-1, and corresponding control were quantitatively analyzed in 500 cells per treatment using an ImageJ software program (Version 1.51j8, National Institutes of Health (NIH)). The average relative intensity of cells was 177.44 ± 16.45 in controls. The average immunofluorescence intensity associated with claudin5 antibody of cells treated with heme was 129.33 ± 13.16 (p < 0.05). In contrast, the average immunofluorescence intensity with claudin5 antibody increased to 165.23 ± 17.11 when NRG-1 was added to heme-treated cells (p < 0.05, Fig. 2e). Similar observations were obtained with ZO-1 (Fig. 2f–i) as those obtained with claudin5. The average relative intensity of ZO-1 of heme-treated cells decreased to 96.99 ± 10.35 as compared to controls 136.16 ± 13.00 (p < 0.05, Fig. 2j). The average immunofluorescence intensity with ZO-1 antibody increased to 125.45 ± 11.78 when NRG-1 was added to heme-treated cells (p < 0.05, Fig. 2j). The expression of occludin (Fig. 2k–n) showed the same pattern as that of claudin5 and ZO-1. The average relative intensity of occludin in heme-treated cells decreased to 88.04 ± 9.24 compared to controls 137.34 ± 14.23 (p < 0.05). The average immunofluorescence intensity of occludin antibody increased to 144.10 ± 15.00 when NRG-1 was added to heme-treated cells (p < 0.05, Fig. 2o). These results suggest that NRG-1 protects against heme-induced damage of tight junction integrity in the BBB.

To further examine the role of NRG-1 on the in vivo BBB conditions, we employed a two-dimensional (2D) in vitro BBB model to promote a phenotype that more closely mimicks in vivo physiological or pathological conditions as previously described [28, 30‐33]. To enable better attachment of endothelial cells to support surfaces for immunofluorescent staining, the protocol was slightly modified. In our 2D transwell co-culture system, the monolayer of hCMEC/D3 was placed in the lower surface of the filter (brain side), with the upper chamber (blood side) containing neuroglial cell cultures from human A172 glioma (Fig. 3a–d). Treatment was initiated after 7 days of co-culture between the two cell types. The endothelial cells were allowed to acquire the appropriate BBB phenotype needed for the assay and to produce cytokines or key soluble substances [34]. This syngeneic co-culture model of endothelial and neuroglia/astrocyte cells increases tight junction resistance between the endothelial cells compared to endothelial cells grown alone (one-dimensional). Expression of tight junction protein markers were assayed by staining with antibodies against tight junction proteins such as claudin5, ZO-1, and occludin. Claudin 5 was expressed on cell membranes (arrows) in co-cultured hCMEC/D3 with human neuroglia as observed with hCMEC/D3 monolayer culture. The membrane-bound expression was disrupted and/or disappeared, while cytoplasm stained more densely after cells were treated with 30 μM heme for 24 h (Fig. 3a, b, arrows). NRG-1 attenuated the disruption of claudin5 expression induced by heme (Fig. 3c, d). Quantitative analysis of immunoreactive signals of claudin5 antibody in hCMEC/D3 cells indicate decreased staining from pronounced (401.80 ± 39.00) in control to moderate (303.21 ± 32.67) after treatment with heme (n = 500, p < 0.05), while staining for claudin5 proteins in cells treated with NRG-1 (385.00 ± 39.97, p < 0.05, Fig. 3e) increased significantly. Similar results were observed for ZO-1(Fig. 3f–j). The average immunofluorescence intensity for ZO-1 decreased to moderate levels (114.69 ± 11.60) in heme-treated cells when compared to control (192.17 ± 20.01, n = 500, p < 0.05), while the intensity of staining returned to pronounced levels (187.46 ± 16.78) when cells were treated with NRG-1 (p < 0.05, Fig. 3j). The expression pattern of occludin was similar to that of claudin5 (Fig. 3k–o). The average relative intensity of occludin in cells treated with heme decreased to 170.11 ± 15.89 as compared to that in control 303.34 ± 28.98 (n = 500, p < 0.05). The average immunofluorescence intensity with occludin antibody increased to 489.24 ± 47.92 when NRG-1 was added to the cells treated with heme (p < 0.05, Fig. 3o).

BBB function can be examined by assessing permeability of confluent endothelial cells to sodium fluorescein (SF; MW = 376 kDa) [28]. Heme-induced endothelial cell permeability was assessed by growing hCMEC/D3 cells to confluence on the inner surface of gelatin-coated transwell inserts followed by incubation of cells with 30 μM heme, 10 ng/ml IL-1β, 1 μg/ml lipopolysaccharide (LPS), and 100 ng/ml NRG-1 for 24 h as indicated, respectively, while LPS and IL-1β served as positive controls. Permeability was measured by adding 10 μM sodium fluorescein to the upper chamber as previously described [35] (Fig. 4). Incubation with heme, LPS, and IL-1β resulted in increased permeability of the endothelial monolayer by 1.93-fold, 1.92-fold, and 2.33-fold, respectively. Addition of NRG-1 during incubation of the cells reduced endothelial permeability by 1.32-fold, 1.50-fold, and 1.8-fold, respectively (Fig. 4).

Neuregulin-1 gene encodes various isoforms which are generated by transcription from six promoters and alternative splicing of primary transcripts [19, 36]. NRG-1 variants vary with four distinct domains: (1) the N-terminal domain, (2) the EGF-like domain, (3) the juxtamembrane domain, and (4) the C-terminal domain [37]. All functional variants contain EGF-like domain essential for interaction with ErbB receptors [37]. The EGF domain differs in the inclusion of exon A or B, which is referred as to α isoforms if it is derived from exon A or β isoforms derived from exon B [36, 37]. β isoforms have higher affinity for the receptor and tend to produce a more robust, long-term signal as compared to α isoforms [36, 37]. NRG-1 is mainly expressed in neurons and Schwann cells of the nervous system in human [38] and mouse [19], in human/mouse endothelial cells [27, 39] and cardiomyocytes [40], and in human cornea epithelial and stroma cells [37]. They play complicated roles in the development of the nervous and the cardiovascular systems. NRG-1 is initially synthesized as a transmembrane protein; however, it does not reach the cell surface and is expressed as soluble or cytosolic proteins [18] or nuclear protein [41] without a transmembrane domain. We assessed total NRG-1 protein expression level in serum by enzyme-linked immunosorbent assay (ELISA) of ECM (n = 5/group). Circulating blood NRG-1 level significantly increased in mice 6 days post infection and declined to a level below that of uninfected mice (controls) 8 days post infection when fatal brain damage occurred (Fig. 5a). This decline in NRG-1 suggests inadequate protection of endogenous serum NRG-1 against severe late-stage factors of ECM indicating that low plasma NRG-1 level correlates with clinical severity of ECM. To determine the relationship between tissue NRG-1 expression especially in brain tissue and malaria infection, total RNA, and whole-cell protein lysate were extracted from ECM mice at different time points. Infection of C57BL/6 mice with Plasmodium berghei ANKA (PbA) upregulated NRG-1 mRNA in the brain cortex (Fig. 5b) suggesting that NRG-1 expression may be protective against PbA-induced damage in the brain. Consistent with the levels of mRNA detected by real-time reverse transcription polymerase chain reaction (qRT-PCR) assay (Fig. 5b), total NRG-1 protein levels in the cortex tested by Western blot were significantly elevated in PbA-infected C57 BL/6 mice (Fig. 5c, d).

Next, we investigated whether exogenously infused recombinant human NRG1-beta1 (rhNRG-1) containing EGF-like domain into ECM mice activates the NRG-1/ErbB4 signaling pathway, which is functionally active in ischemia/reperfusion-induced injuries [17]. Real-time quantitative PCR results showed that PbA infection did not change ErbB4 expression in the cortex (Fig. 6a). Whole extracts of mouse cortex tissues were analyzed by Western blots with antibody to the phosphotyrosine residue Tyr1284 within ErbB4 and total ErbB4 antibody. Extracts of cortex showed gradually decreased ErbB4 phosphorylation at Tyr 1284 after 2 days of PbA infection and was hardly detectable by day 12 (upper panel of Fig. 6b). Protein extracts from pooled hippocampus of mice were also probed with antibodies to phosphorylated ErbB4 (Tyr1284) and total ErbB4. The pattern of pErbB4 protein expression from pooled hippocampus tissues was similar to those in the cortex (lower panel of Fig. 6b). Phosphorylated ErbB4 decreased when mice were infected with P. berghei ANKA. Consistent with total mRNA expression, levels of total ErbB4 protein did not change in either the cortex or hippocampus. Immunofluorescence (IFC) results confirmed the findings of Western blot analysis whereby total ErbB4 protein remained unchanged 12 days after infection (Fig. 6c, panels a, b). In addition, IFC data showed that ErbB4 co-localized with the neuronal marker-NeuN (Fig. 6d, panels a‐c), but not with microglial cell marker Iba1 (Fig. 6d, panels d‐f), or astrocyte marker glial fibrillary acidic protein (GFAP) (Fig. 6d, panels g‐i). Remarkably, exogenous NRG-1 (at both low (5 μg/kg) and high (25 μg/kg) doses) in combination with/without artemether (ARM) treatment partially restored expression of pErbB4 in both cortex (Fig. 7a) and pooled hippocampus tissues (Fig. 7c). pErbB4 levels increased in PbA-infected mice treated with NRG-1 and ARM individually, while there was robust increase in animals treated with both NRG-1 and ARM (NRG-1+ARM or NRG-1/ARM represents combination of NRG-1 and ARM), particularly in pooled samples prepared from hippocampal tissues (Fig. 7c). These data suggest that the ErbB4 receptor is activated through phosphorylation after the treatment with NRG-1 infusion. We further examined the putative downstream pathways activated by NRG-1/ErbB4. Our interest in signal transducer and activation of transcription 3 (STAT3) and AKT signaling pathways derives from the reported importance of STAT3 in the pathogenesis of cerebral malaria [7, 42] while AKT activation is a typical downstream signal molecule for NRG-1/ErbB4 [17]. Activated STAT3 induced by Plasmodium infection was inhibited by treatment with NRG-1 and/or ARM as expected (Fig. 7b, d). In addition, pAKT is upregulated both in brain cortex (Fig. 7b) and hippocampus (Fig. 7d) of PbA-infected mice when treated with NRG-1 and/or ARM compared to untreated control. These results strongly indicate that NRG-1 plays an important role in attenuation of mortality associated with ECM through ErbB4/STAT3/AKT signaling. We also examined whether this mechanism applied to in vitro cell culture systems. To this end, hCMEC/D3 cells were grown and treated with 30 μM of heme with or without 100 ng/ml of NRG-1. Whole cell lysates were isolated under each condition as indicated in Fig. 8, transferred to membranes and probed with pSTAT3/tSTAT3 and pAKT/tAKT antibodies. Consistent with the observation made in our in vivo study the results show that NRG-1 reduced heme-induced STAT3 activation and increases heme-reduced ErbB4 and AKT activation (Fig. 8).

Infection of mice with P. berghei ANKA was performed as described previously [26]. Mice were selected and randomized into treatment groups after diagnosis with ECM [26] on days 5 to 6 post infection. To determine the therapeutic benefit of NRG-1 on ECM-associated brain damage and mortality and to compare NRG-1 with artemether (ARM) treatment, PbA-infected mice were treated daily via i.p. injection with 50 μl doses of NRG-1 (5 or 25 μg/kg/day, EGF-like domain, R & D Systems, Minneapolis, MN, USA) [NP_039250] or artemether prepared in coconut oil (25 mg/kg/day, Sigma-Aldrich, St Louis, MO, USA) from days 6 to day 11 post infection. The mice treated daily with 50 μl saline solution (i.p.) from days 6 to 11 post infection were used as the controls. Mice were checked several times daily for mortality and signs of ECM neurological symptoms such as ataxia, loss of reflex, and hemiplegia. All murine ECM experiments were terminated 19 days after PbA infection with animals being euthanized accordingly [26].

Rabbit antibody against STAT3, phospho-STAT3, rabbit antibody against AKT, phospho-AKT, anti-NeuN, anti-Iba1, and anti-glial fibrillary acidic protein (GFAP) were purchased from Cell Signaling Technology Inc. (Danvers, MA, USA). Polyclonal antibody to β-actin, lipopolysaccharide (LPS), and sodium fluorescein were purchased from Sigma-Aldrich (St. Louis, MO, USA). Rabbit polyclonal antibodies against claudin5, ZO-1, and phosphorylated ErbB4 (monoclonal) antibody were purchased from Abcam (Cambridge, MA, USA). Rabbit polyclonal antibody against occludin and attachment factor (AF) was purchased from Invitrogen (Waltham, MA, USA). Rabbit antibody against VWF was purchased from Dako (Santa Clara, CA, USA). Anti-ErbB4 (polyclonal) and NRG-1 (polyclonal) antibodies were purchased from Santa Cruz (Dallas, TX, USA). All secondary antibodies used for Western blot were purchased from Calbiochem (La Jolla, CA, USA). Hemin (heme) was purchased from Frontier Scientific (Logan, UT, USA). Recombinant human NRG1β1/HRG1β EGF domain protein (γhNRG-1) and recombinant human CXCL10/IP-10 protein were purchased from R&D systems (Minneapolis, MN). Recombinant human interleukin-1β (IL-1β) was purchased from Millipore (Billerica, MA, USA).

The hCMEC/D3 a cell line (Cellutions Biosystems Inc., Ontario, Canada) was derived from human temporal lobe microvessels and were immortalized with hTERT and SV40 large T antigen. The cells have been extensively characterized for brain endothelial phenotype and are a model of human blood brain barrier (BBB) function. hCMEC/D3 cells were cultured at 37 °C with 5% CO2 in endothelial basal medium-2 (Lonza) supplemented with 5% fetal bovine serum (FBS; American Type Culture Collection (ATCC), Manassas, VA, USA), growth factors and other supplements including human recombinant epidermal growth factor (hEGF), hydrocortisone, GA-100 (Gentamicin, Amphotericin-B), human recombinant vascular endothelial growth factor (VEGF), recombinant human fibroblast growth factor-b (hFGF-b), recombinant long R insulin-like growth factor (R3-IGF-1), ascorbic acid, heparin, 100 U/ml of streptomycin, and 100 U/ml of penicillin (Lonza). The cells were harvested and passaged at about 70–90% confluence as described previously [6]. The hCMEC/D3 cells (2 × 10⁵ cells/ml) were seeded in 35-mm tissue culture dish and incubated at 37 °C in 5% CO₂ for 24–48 h for future use. M059K cells purchased from ATCC were cultured at 37 °C with 5% CO2 in DMEM: K-12 medium.

The hCMEC/D3 or M059K cells were seeded at 1 × 10⁴ cells in 100 μl of medium per well into 96-well plates and serum-starved for 24 h, incubating with 100 ng/ml γhNRG-1 for 30 min, followed by exposing to heme at 30 μM or 0.02 μg/ml CXCL10 for 24 h. MTT assay was performed in accordance to the manufacturer’s instructions. Ten microliters of MTT reagent (the ratio of MTT reagent to medium is 1:10) was added into each well and incubated in the dark at room temperature for 2 to 4 h. Absorbance at 570 nm was measured using 650 nm as reference filter by a CytoFluorTM 2300 plate reader and the software CytoFluorTM 2300 v. 3A1 (Millipore Co, Bedford, MA, USA).

Peripheral blood NRG-1 quantification was assayed by ELISA (MyBioSource, San Diego, CA, USA). At indicated days after PbA infection, a subset of mice (n = 5/group) was anesthetized, then 500 μL of peripheral blood was acquired through cardiac puncture. Blood was immediately placed in a biopsy tube with EDTA to prevent clotting. Plasma was obtained by centrifuging the whole blood at 1200 g for 15 min at room temperature. ELISA was performed on all samples to quantify plasma levels of mouse NRG-1. The samples were read on a microplate reader (TECAN) at 450 nm.

Endothelial cell permeability assay was performed as previously described [28, 43, 55]. In brief, the hCMEC/D3 cells were cultured in complete growth media EGM-2 MV (Lonza, Walkersville, MD, USA). The hCMEC/D3 cells were seeded into the inner surface of attachment factor (AF)/ collagen-coated transwell inserts (24 mm diameter, 0.4-μm pore size polyester filter; Corning, Corning, NY, USA), which were placed in wells of a 6-well plate with human neuroglial/astrocytes (A172 glioma cells). About 7 days after seeding, when hCMEC/D3 cells grown on the inner surface of the insert were confluent (confirmed by testing permeability to media), the cells were serum starved overnight with EBM media without growth supplement, then incubated with IL-1β (10 ng/ml), LPS (1 μg/ml), and heme (30 μM) for 24 h. Thirty minutes prior to addition of IL-1β, LPS and heme, NRG1 (100 ng/ml = 12.5 nM), or vehicle (PBS) was added. After 24 h, media in both upper and lower chambers were removed and replaced with fresh media. Permeability was measured by adding 0.1 mg/ml of Sodium Fluorescein (MW, 376 kDa, Sigma, St. Louis, MO, USA) to the upper chamber. After incubation for 0.5 h, the sample from the lower compartment was collected and 100 μl of sample was measured for fluorescence at excitation 485 nm and emission 535 nm. All three independent experiments were performed in triplicate.

Cell pellets were stored in Trizol reagent and homogenized in fresh Trizol. Total RNA was isolated from cells using an RNeasy Mini Kit (Qiagen, Valencia, CA, USA). Total RNA was quantified using the Nanodrop N-1000 by Agilent Biosystems (Santa Clara, CA, USA). cDNAs were synthesized from the isolated RNA using iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Inc.). Reverse transcription was performed by using random hexamers at 25 °C for 5 min, 42 °C for 30 min, and 85 °C for 5 min. Quantitative PCR was performed using iQ SYBR Green Supermix (Bio-Rad Laboratories, Inc.) in a CFX96 Real-Time PCR System machine (Bio-Rad Laboratories, Inc.). The data was analyzed using CFX96 Real-Time PCR System (Bio-Rad Laboratories, Inc.). Primer sequences for the genes were described in Table 1.

Cells were lysed with lysis buffer (50 mM HEPES, 150 mM NaCl, 1.5 mM MgCl2, 1 mM EGTA, 10% glycerol, 1% Nonidet P-40, 100 mM NaF, 10 mM sodium pyrophosphate, 0.2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, and 10 μg/ml leupeptin). Samples were separated by SDS/PAGE, and separated proteins were transferred to nitrocellulose membranes and identified by immunoblotting. Primary antibodies were obtained from commercial sources; these antibodies were diluted at the ratio of 1:1000 according to the manufacturer’s instruction. Blots were developed with Supersignal Pico or Femto substrate (Pierce). A densitometric analysis of the bands was performed with the ImageQuant program (Bio-Rad).

Statistically significant differences were determined using Prism statistical software (Graph Prism 4.03, San Diego, CA, USA). All data were presented as mean ± SEM of at least three independent experiments. For data analysis, unpaired t test or one-way ANOVA with Dunnett’s or Bonferroni’s post test was applied. All P values resulted from two-sided statistical tests and statistical significance was set at *p < 0.05, **p < 0.01, and ***p < 0.001.