Neuregulin 1-b decrease interleukin -1b – induced RhoA activation, myosin light chain phosphorylation, and endothelial hyper permeability

Abstract
Neuregulin-1 (NRG1) is an endogenous growth factor with multiple functions in the embryonic and postnatal brain. The NRG1 gene is large and complex, transcribing more than twentytransmembrane proteins and generating a large number of isoforms in tissue and cell type-specific patterns. Within the brain, NRG1 functions have been studied most extensively inneurons and glia, as well as in the peripheral vasculature. Recently, NRG1 signaling has been found to be important in the function of brain microvascular endothelial cells, decreasing IL- 1b-induced increases in endothelial permeability. In the current experiments, we have investigated the pathways through which the NRG1-b isoform acts on IL-1b-induced endothelial perme- ability. Our data show that NRG1-b increases barrier function, measured by transendothelial electrical resistance, anddecreases IL-1b-induced hyperpermeability, measured by dex- tran-40 extravasation through a monolayer of brain microvas- cular endothelial cells plated on transwells. An investigation of key signaling proteins suggests that the effect of NRG1-b on endothelial permeability is mediated through RhoA activation and myosin light chain phosphorylation, events which affectfilamentous actin morphology. In addition, AG825, an inhibitor ofthe erbB2-associated tyrosine kinase, reduces the effect of NRG1-b on IL-1b-induced RhoA activation and myosin light chain phosphorylation. These data add to the evidence thatNRG1-b signaling affects changes in the brain microvasculature in the setting of neuroinflammation.

Neuregulin-1 (NRG1) is a growth factor with multiple functions in the embryonic and postnatal brain. Its functions have been studied most extensively in neurons and glia, in a variety of physiological processes, including migration of embryonic interneurons, maturation and survival of oligoden- drocytes (Fernandez et al. 2000; Makinodan et al. 2012) and modulation of synaptic plasticity (Neddens et al. 2009). More recently, NRG1 signaling has been studied in brain microvas- cular endothelial cells (BMEC) (Lok et al. 2009) (Wu et al. 2015). Our previous data show that NRG1-b reduces endothe- lial permeability and neutrophil adhesion during IL-1b incu- bation, and decreases acute blood–brain barrier (BBB) permeability in a mouse model of brain trauma (Lok et al. 2012; Wu et al. 2015). Our current experiments further elucidate the signaling pathways that contribute to NRG1-
b’s effects on brain microvascaular endothelial permeability. BMECs play a prominent role in the integrity of the BBB.Although the BBB is composed of different types of cells, direct exposure of cultured human BMEC to IL-1b induces hyperpermeability even in the absence of other brain parenchymal cell types (Didier et al. 2003). BBB dysfunction, in turn, generates additional inflammatory responses (de Vries et al. 1996; Erickson and Banks 2013) and contributes to the pathogenesis of many neurodegenerative diseases (Zlokovic 2008; Rosenberg 2012). Thus, an understanding of the pathways associated with IL-1b-induced endothelial hyper- permeability will add to the understanding of BBB injury. The model used in our current experiments is one of IL-1b-induced endothelial hyperpermeability. Since IL-1b is a proinflamma- tory cytokine which contributes to neuroinflammation in many central nervous system disorders (Boato et al. 2013; Ghosh et al. 2013; Lyman et al. 2014), the data regarding NRG1 effects in this model may be applicable to a number of brain pathologies in which neuroinflammation plays a role.

Measurement of TEER was performed, using the ECIS Zhelectric cell–substrate impedance system (Applied Biophysics, Troy, NY, USA) configured to a 96W array station. BMEC at 1 9 105/well wereplated on collagen type I coated 96W10Eidf Positron EmissionTomography electrode arrays and were grown to confluence until stable resistance values (900–2000 Ω) were reached. After 7 days, they were exposed to the indicated experimental condition. Readings were acquired continuously at 4000 Hz at 30 min intervals for 24 h. At least three independent experiments containing four to six replicates were performed.Using a transwell system (Corning, Corning, NY, USA), RBE.4 cells were seeded onto the inner surface of collagen-coated inserts (0.4 lm pore size polycarbonate filter), and placed in wells of a 12- well plate. When confluent, the cells were incubated with IL-1b with or without NRG1- b. After 18 h, permeability was measured by adding 0.2 mg/mL of Fluorescein isothiocyanate-labeled dextran (40 kDa; Sigma) to the upper chamber. After 15 min, 100 lL of the sample from the lower compartment was measured for fluorescence.Cells were lysed using lysis buffer (Cell Signaling, Beverly, MA, USA) with phosphatase inhibitor cocktail 2 (Sigma). Cell membrane proteins were separated using plasma membrane protein extraction kit (Abcam, Cambridge, MA, USA). Protein lysates were separated using sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel, transferred to polyvinylidene difluoride membranes, blocked with 5% bovine serum albumin for 1 h, and incubated overnight at 4°C with primary antibodies to: p-MLC, MLC, p-cofilin, cofilin (Cell Signaling), VE-Cadherin (Enzo, Boston, MA, USA), zona occudens-1 (ZO-1) (Invitrogen, Grand Island, NY, USA), occludin (Abcam), claudin 5 (Invitrogen), b actin (Invitrogen), and NaKATPase (Abcam).

After washing and incubation with the secondary antibody, the proteins were visualized with the enhanced chemiluminescence system (GE Healthcare Bio-Science, Pittsburgh, PA, USA) by exposure on film or in the G:box system (Syngene, Frederick, MD, USA). The relative densities of bands were analyzed, using NIH ImageJ, Bethesda, MD, USA.RhoA activation assay (Cytoskeleton, Denver, CO, USA) was performed per manufacturer’s protocols. After lysing cells, equalvolumes of supernatants were incubated with Rhotekin-RBD affinity beads for 1 h at 4°C (10% of total protein was used for measurement of total RhoA). Bound proteins were eluted from the beads with 29sodium dodecyl sulfate buffer and examined by western blot analysis.Endothelial cells treated for 18 h was stained with Alexa Fluor 488- Phalloidin (ThermoFisher Scientific, Grand Island, NY, USA) to examine the structure of filamentous (F)-actin. Washed cells were fixed with 4% paraformaldehyde, washed again and permeabilized for 5 min with 0.1% Triton X-100. The cells were incubated with a 1% solution of bovine serum albumin for 30 min, and stained with Rhodamine-Phalloidin (0.20 mol/L) for 30 min in dark conditions.Three random images of each well (1009 magnification) wereobtained. The intensity of the fluorescent phalloidin signal was measured by obtaining the integrated density of the entire image using Image J, NIH, Bethesda, MD, USA.TEER data were analyzed using one-way ANOVA followed by Dunnett’s post hoc test and presented as the mean percent change from baseline TEER SEM. All other quantitative data wereanalyzed with one-way ANOVA followed by inter-group Tukey Honest Significant Difference post hoc test and presented as mean SD. Statistical significance was set at *p < 0.05, **p < 0.01.

Results
To examine the effect of NRG1 on endothelial barrier function in baseline conditions and in the setting of IL-1b- induced hyperpermeability, the NRG1-b isoform was used. First, the effect of NRG1-b on barrier function at baseline culture conditions was examined in human BMECs (HBMECs) isolated from healthy tissue of the cortex and hippocampus in the resection path during epilepsy surgery. Endothelial barrier tightness was evaluated by measuring the TEER by electric cell–substrate impedance system. After thecells reached a stable monolayer, and baseline measurements (900–2000 Ω 9 cm2) were obtained, HBMECs were exposed to increasing concentrations of NRG1-b ranging from 10 ng/mL (1.25 nM) to 1200 ng/mL (150 nM) (Fig. 1a). Starting at 4 h, significance of at least p < 0.05was observed for comparisons between the individual NRG1- b concentrations and the untreated group, except for NRG1-b 10 ng/mL, which did not reach significance during the time course. NRG1-b gradually increased the average TEER over baseline by as high as 5% (NRG1: 30 ng/mL), 7% (NRG1: 100 ng/mL), 8% (NRG1: 300 ng/mL), 9% (NRG1: 600 ng/mL), and 12% (NRG1: 1200 ng/mL). Maximal effects were reached between 9–12 h post-treatment. Interestingly, the effect on barrier tightness was not only dose responsive but also appeared to be sustained until 24 h, with a slight decline for concentrations above 100 ng/mL. Based on these data, an NRG1-b dose of 100 ng/mL (12.5 nM) was used forsubsequent experiments. This dose is consistent with physiologic levels ranging from 32 to 473 ng/mL, as reported in a study of healthy individuals by Moondra et al. (2009).Next, the possibility that NRG1-b may protect against barrier leakiness when human BMEC were exposed to IL-1b was tested. Addition of IL-1b (100 ng/mL) resulted in a gradual decline in TEER over 24 h, reaching the nadir at 6 h with TEER averaging 84% that of untreated cells, p < 0.05 (Fig. 1b).

In contrast, cells co-incubated with NRG1-b and IL-1b exhibited a smaller decrease in TEER for 12 h, to 91% of steady-state (p < 0.05) compared to incubation with IL-1b alone, then declined over time, although still trending higher than those treated with IL-1b alone without NRG1-b. These data suggest that the addition of NRG1-b to BMEC directly benefits barrier integrity during baseline culture conditions and during a proinflammatory insult. A similar effect was observed in frontal cortex-derived human BMEC, which are commercially available (Fig. 1c). Additionally, NRG1-b’seffect on permeability was tested on rat brain endothelial cells (RBE.4), using transwells. Incubation with IL-1b resulted in an increase in dextran-40 extravasation to 170% that of untreated cells. Co-incubation with NRG1-b decreased the extravasation to 114% (p < 0.05) (Fig. 1d). These data corroborate the findings from TEER measure- ments described for human BMECs.NRG1-b did not change protein levels of occludin, ZO-1, claudin-5, and VE-cadherin in the presence of IL-1b18 h incubation with IL-1b, either in the presence or absence of NRG1-b, did not result in statistically significant changes in protein levels of occludin, ZO-1, claudin-5, and VE- cadherin in whole cell lysates or in the membrane fraction (Fig. 2a–d).The RhoA pathway is known to regulate endothelial perme- ability. The amount of RhoA-GTP in IL-1b-treated cells was1.99 of baseline (p < 0.01), compared to 0.99 of baseline incells treated with IL-1b + NRG1-b (p < 0.01) (Fig. 3a), indicating that NRG1-b prevented the IL-1b-induced increase in RhoA activation. Parallel experiments using RhoA GTP ELISA yielded similar results (data not shown). To evaluate the specificity of the NRG1 signaling pathway in IL-1b- induced RhoA activation, AG825, a selective inhibitor of the erbB2-associated tyrosine kinase, was added during the experiment.

When AG825 was present in the media along with IL-1b and NRG1-b, RhoA-GTP levels increased 2.1- fold, which is a significant difference compared to baseline (p < 0.01) or to IL-1b with NRG1-b (p < 0.01).To determine the relevant signaling pathways, we examined two proteins involved with the membrane barrier functiontreated cells was 1.79 of baseline (p < 0.01), addition of NRG1-b decreased the p-MLC level to 0.29 of baseline, which is significantly different from p-MLC level in cellstreated wth IL-1b alone (p < 0.01) (Fig. 3c). Similar to the case with RhoA, addition of AG825 appeared to negate the NRG1-b mediated decrease in MLC phosphorylation, although statistical significance was not reached (data not shown).Because MLC phosphorylation affects actin and myosin function, with effects on the cell cytoskeleton, filamentous actin (F-actin) staining was performed. 18 h incubation with IL-1b was associated with a decrease in F-actin, which suggests destabilization of the cytoskeleton. This decrease was not seen when cells were incubated with both IL-1b and NRG1-b (Fig. 4).

Discussion
The brain microvascular endothelium, which constitutes a major element of the BBB, is an important determinant of theinflammatory response to brain injury. Dysfunctional endothelial signaling can lead to amplification of the injury. In clinical series, patients with disrupted BBB appear to have an increased risk of developing acute seizures, delayedepilepsy, and cognitive impairment (Rosenberg 2012). Thus, the cellular pathways that affect microvascular endothelial permeability are relevant to the understanding of many central nervous system diseases. IL-1b exposure was used in our in vitro model of cytokine-induced hyperpermeability, because it is a proinflammatory agent involved in many neurological diseases. IL-1b is expressed at a higher level in patients with brain ischemia and trauma, with the potential to exacerbate the underlying pathological processes (Holmin and Hojeberg 2004; Lu et al. 2005; Smith et al. 2014).In our experimental conditions, IL-1b induced a decrease in TEER, in a range consistent with other reports (de Vries et al. 1996; Skinner et al. 2009; Rigor et al. 2012; Yamada et al. 2014). In combination, the TEER and transwell data provide evidence that NRG1-b has a barrier-protective effect in BMECs from different areas of the cortex, and that extends to cells of human and rat origin. The interaction of NRG1-b with RhoA is a novel finding. Since RhoA regulates many cellular processes, including membrane permeability, lamel- lopodia formation, actin cytoskeleton maintenance, and cell migration (Menager et al. 1999; Graupera et al. 2008; Szul- cek et al. 2013; Yuan et al. 2015), we investigated whether some of these pathways are subject to NRG1-b control. Our data suggest that MLC is a downstream target of NRG1-b/ RhoA interactions.

MLC phosphorylation leads to increases in actin and myosin contractility, which can affect cytoskeletal structure and endothelial permeability (Then et al. 2011). The fact that AG825 abolishes the NRG1-b effect on decreasing RhoA activation is a further confirmation that the effect is mediated by erbB receptors and is specific to NRG1.In our model of IL-1b-induced hyperpermeability, no significant changes were observed in the protein expression of occludin, ZO-1, claudin-5 and VE-cadherin. Thesefindings fall within the spectrum of varied responses reported for tight junction protein expression in the presence of IL-1b. In experiments with human corneal epithelial cells, IL-1b did not change ZO-1 and occludin expression, but induced their redistribution and disrupted barrier function (Kimura et al. 2009). In intestinal epithelial cells, IL-1b caused no change in ZO-1, but resulted in down-regulation of occludin, and up- regulation of claudin-1 (Al-Sadi and Ma 2007). Similar to these in vitro data, results from animal models correlating BBB leakage with tight junction protein expression also suggest that the relationship is not always linear. In a model of BBB damage induced by cerebral ischemia, occludin levels were not changed, but ZO-1, claudin-5, and VE- cadherin levels were significantly decreased (Kitashoji et al. 2013). In a rat embolic stroke model, there was no correlation between albumin leakage and levels of tight junction expression (Krueger et al. 2013). In 3-chloropropanediol- induced BBB damage, disruption of the normal paracellular localization of occludin, claudin-5 and ZO-1 was seen, correlating with focal vascular leak of dextran and fibrino- gen. However, partial recovery of barrier function at 6 days, evidenced by decreased dextran and fibrinogen leak, occurred well before normal localization of occludin and claudin-5 (Willis et al. 2004). Additionally, in a human neuropathology study, there was no correlation between tight junction alteration in the cerebral cortex and degree ofAlzheimer’s type pathology, an entity associated with BBB dysfunction (Viggars et al. 2011).The maintenance of a selectively permeable endothelial barrier requires constant adjustments in barrier properties which are regulated by a large array of proteins and signaling pathways (Beckers et al. 2010; Goddard and Iruela-Arispe 2013), as well as by interactions between cell–cell and cell–substrate contacts, soluble mediators, and biomechanical forces (Krishnan et al. 2011). One such factor is the organization of the F-actin cytoskeleton, a dynamic structure with rapid responses to changes in the microenvironment (Shen et al. 2009; Garcia-Ponce et al. 2015).

The role of Rho GTPases, such as RhoA, is being increasingly recog- nized as important determinants of cytoskeletal changes (Spindler et al. 2010; Duluc and Wojciak-Stothard 2014). These systems are complicated, as Rho GTPases may exert barrier-protective or barrier-disruptive effects depending on the specific inducer and environment; with the actions of various GTPase regulators and actin-binding proteins super- imposed on this network (Beckers et al. 2010). It is possible that the tight junction proteins also undergo sequential changes in their expression levels, phosphorylation status, and subcellular localization depending on the stimulus and the microenvironment.In our experimental conditions, IL-1b-induced endothelial hyperpermeability and the effects of NRG1-b in this model are associated with changes in RhoA activation, MLC phosphorylation and F-actin. The finding that NRG1-b is associated with preservation of the cell cortex associated actin during a pro-inflammatory insult is consistent with its barrier-stabilizing properties shown in our experiments, as cell cortex actin is thought to stabilize junctions (Garcia- Ponce et al. 2015). In summary, our data suggest that NRG1- b promotes endothelial permeability through signaling pathways affected by RhoA activation and MLC phospho- rylation.Our study is limited by the fact that endothelial cells in culture cannot completely reproduce the in-vivo situation in which microvascular endothelial cells interact with other types of cells within the neurovascular unit. However, investigating the endothelial cell in isolation is a reasonable initial approach. Our study is also limited by the fact that media conditions in our cell culture system do not accurately reproduce the microenvironment in the blood or interstitial fluid. The effect of NRG1-b on MLC phosphorylation in the presence of IL-1b is seen in different media conditions and in different types of microvascular endothelial cells, suggesting that this is a reliable biological phenomenon. However, many other proteins in addition to MLC are involved in the determination of endothelial permeability, including thosethat constitute the Rac and Cdc42 pathways. NRG1-b’sinteractions with these pathways should be examined in future experiments. Additionally, the active domain of NRG1-b was used for the current set of experiments, and it is unclear whether the effect is isoform specific. Additionally, the effect of RhoA activation, MLC phosphorylation, and changes in F-actin on the spatial distribution of tight and adherens junction proteins should be further studied.

In summary, our data show that NRG1-b decreases IL-1b- induced endothelial hyperpermeability by decreasing RhoA activation and MLC phosphorylation. These data add to the evidence that NRG1-b plays a major role in the barrier function of AG 825 BMEC.