Ebola Hemorrhagic Shock Syndrome-on-a-Chip.

Ebola virus, for which we lack effective countermeasures, causes hemorrhagic fever in humans, with significant case fatality rates. Lack of experimental human models for Ebola hemorrhagic fever is a major obstacle that hinders the development of treatment strategies. Here, we model the Ebola hemorrhagic syndrome in a microvessel-on-a-chip system and demonstrate its applicability to drug studies. Luminal infusion of Ebola virus-like particles leads to albumin leakage from the engineered vessels. The process is mediated by the Rho/ROCK pathway and is associated with cytoskeleton remodeling. Infusion of Ebola glycoprotein (GP1,2) generates a similar phenotype, indicating the key role of GP1,2 in this process. Finally, we measured the potency of a recently developed experimental drug FX06 and a novel drug candidate, melatonin, in phenotypic rescue. Our study confirms the effects of FX06 and identifies melatonin as an effective, safe, inexpensive therapeutic option that is worth investigating in animal models and human trials.

Introduction

Ebola hemorrhagic fever is a rapidly progressive and highly fatal condition for which there is no established treatment (Jameson et al., 2018). The Ebola epidemic in West Africa (2014–2016) was a health crisis of unprecedented magnitude and impact, causing more than 11,000 deaths and destabilizing three countries (Nicholas et al., 2016). Currently, World Health Organization and African countries are struggling to contain a new large outbreak in Africa, which has led to hundreds of deaths (Dyer, 2019). Vascular integrity impairment with subsequent blood volume loss (the so-called shock syndrome) is the primary cause of death in patients with Ebola. Despite supportive care, more than 50% of patients die, with significant interindividual differences in disease outcome reported and attributed to viral loads and host factors (de La Vega et al., 2015, Hartley et al., 2017). Despite progress, many challenges remain to be addressed, including improving early diagnosis, predicting disease progression, and developing therapeutic and preventive methods.

The lack of experimental models and sensitive detection tools has historically hindered the early detection of Ebola vasculopathy and the development of Ebola drugs. However, rodent models that can mimic certain aspects of Ebola disease in humans have been developed and are now used along with monkey models as in vivo models of the disease (Bennett et al., 2017, de La Vega et al., 2018, Willyard, 2014). The use of these models has recently led to the development of experimental therapeutic strategies, including small molecules (Warren et al., 2016), antibodies (Olinger et al., 2012, Qiu et al., 2011, Qiu et al., 2012, Wilson et al., 2000), and nanoparticles (Thi et al., 2015), as well as glycofullerenes (Munoz et al., 2016). However, these therapeutics do not directly target hemorrhagic shock syndrome but rather Ebola virus infection. Additionally, animal models are costly and cannot fully recapitulate the physiology and pathology of human organs, making it difficult to predict the efficacy, safety, and toxicity of experimental Ebola drugs (Mestas and Hughes, 2004).

In vitro human models for viral hemorrhagic shock syndrome are currently lacking. However, such models would not only be useful for studying the pathogenesis of Ebola in a human-like setting but would also be critical for diagnostics and drug development. Chip-based disease models are becoming important research tools in biology and medicine (Reardon, 2015, Junaid et al., 2017, Tejavibulya and Sia, 2016). Examples include the modeling of drug-toxicity-induced pulmonary edema in a lung-on-a-chip model (Huh et al., 2012), the modeling of Alzheimer disease in a brain-on-a-chip platform (Park et al., 2015), and the simulation of diabetic nephropathy in a glomerulus-on-a-chip microdevice (Wang et al., 2017). Additionally, there is a growing interest in using in vitro engineered models in vascular medicine (Jeon et al., 2014, Kim et al., 2013, Kim et al., 2017, Qiu et al., 2018, van Duinen et al., 2017, Akbari et al., 2017, Bersini and Moretti, 2015, Chen et al., 2017, Haase and Kamm, 2017, Hovell et al., 2015, Rayner and Zheng, 2016, Sato et al., 2015, Shin et al., 2004, Smith and Gerecht, 2014, Song et al., 2005, Takei et al., 2016, Tien, 2014), yet no chip-based model of viral hemorrhagic shock syndrome has been introduced. Here, we develop, for the first time, a microvessel-on-a-chip based model of Ebola (species Zaire ebolavirus) viral hemorrhagic syndrome and demonstrate its usefulness by exploring the signaling and physical processes that underlie the hemorrhagic syndrome and by targeting those processes using drug candidates.

Results

Here, we describe a simple chip-based model of Ebola-induced vascular integrity loss. To provide the proof-of-principle for this approach and to ensure that the platform can be extended to a low-cost, easy-to-use, high-throughput platform for diagnostics, we included the minimal components needed to model the process. We first generated microvessels within the fabricated OrganoPlates (T-design) using human endothelial cells (primary HUVECs) at the interface of a collagen type 1 network. The chip design allowed us to culture 96 microvessels with heights of 120 μm and widths of 400 μm (see Figures 1A–1D and Video S1). To develop the model and generate all the data for the current study, we have used approximately a total of 550 independent chips. To ensure that the engineered vessel recapitulated the physiological barrier function of a natural vessel, we measured the transport of albumin across the endothelial wall into the collagen network. In a physiological setting, the vessel is expected to be impermeable but to respond dynamically to physiological stimuli. Permeability experiments were carried out after incubating the microvessels with and without histamine (an endogenous biogenic amine known to induce vascular permeability during inflammatory processes) for 40 or 60 min. As shown in Figure 1E, we observed no leakage of albumin from the engineered vessels (control; without stimuli) within a 10-min interval during the permeability assay (see Video S2). Permeability was, however, induced by the administration of histamine, indicating that the endothelial wall is not passive and responds to stimuli as expected (see Figures 1E–1G and Video S3).

Figure 1

Viral Hemorrhagic Syndrome-on-a-Chip

(A) Schematic diagram of the 96 microfluidic devices composing the gradient design (T-design) in the OrganoPlate, based on a 384 wells plate interface on top and 96 microfluidic devices integrated in the bottom.

(B) Each microfluidic tissue chip consists of a microvessel compartment with medium inlet (1) and outlet (4), gel inlet (2), and observation window (3). The dashed highlighted rectangular box indicates the region depicted in (C).

(C) Diagram of the monolayers of human umbilical vein endothelial cells (HUVECs) forming a microvessel next to the ECM in the microfluidic system.

(D) A 3D reconstruction showing the human microvessel-on-a-chip that was formed by cultured HUVECs (red, F-actin) and demonstrated continuous on-chip junctions (green, VE-cadherin). See also Video S1.

(E) Time-lapse fluorescence images of albumin (green) perfusion in the microvessel channel. Scale bar, 200 μm.

(F) 100 μM histamine was pipetted in the medium inlet and outlet of the microfluidic device and incubated for 60 min under perfusion with the Mimetas rocker platform. Subsequently, labeled albumin (green) was added to the medium inlet and outlet and time-lapse fluorescence images of albumin (green) diffusing from the microvessel to the ECM channel were taken. Scale bar, 200 μm. See also Video S2 and S3.

(G) Apparent permeability (Papp) of microvessels in the time response to 100 μM histamine. The control is microvessel without histamine treatment.

Data are represented as mean ± SEM.

Video S1. 3D View of a Microvessel Formed on-chip by Cultured HUVECs, Related to Figure 1

Blue denotes Hoechst-stained nuclei. Green indicates Alexa 488-stained VE-cadherin. Red designates phalloidin-rhodamine-stained F-actin.

Video S2. Video of Albumin (Green) Diffusing from the Microvessel to the ECM Channel, Related to Figure 1

The video length is 10 min

Video S3. Video of Albumin (Green) Diffusing from the Microvessel to the ECM Channel after Treatment with 10 μM Histamine for 60 min, Related to Figure 1

The video length is 10 min

Next, we infused various concentrations of VLPs to determine whether VLPs alone are sufficient to induce permeability and whether the extent of permeability is viral load dependent. Infusion of VLPs led to a dramatic increase in the permeability of the engineered microvessels, as shown in Figures 2A and 2B. The VLPs used in these experiments were non-replicating, indicating that the viral components interact directly with endothelial cells and affect their barrier function, presumably by affecting cellular mechanics and intercellular interactions. Immunostaining of F-actin indicated that Ebola VLPs indeed alter the mechanics and physical interaction of endothelial cells, explaining the induction of permeability (see Figure 2C). However, treatment with VLPs did not result in an apparent rearrangement of VE-cadherin but instead caused a clear increase in actin stress fiber formation, consistent with the findings in previous reports (Wahl-Jensen et al., 2005). Moreover, we observed significant upregulation of E-selectin, a mediator of immune cell recruitment and a biomarker for endothelial dysfunction, clearly indicating the activation of the engineered endothelium (see Figure S1).

Figure 2

Microvascular Dysfunction in the Viral Hemorrhagic Shock Syndrome-on-a-Chip Platform

(A) Apparent permeability (Papp) of microvessels in response to Ebola VLPs at several time points. Microvessels were exposed to 1 μg/mL VLPs, followed by a permeability assay.

(B) Concentration dependence of the VLP effect. Microvessels were treated for 2 h with the indicated concentrations of VLPs, followed by a permeability assay.

(C) Endothelial cells stained for VE-cadherin (green) and F-actin (red) after exposure to 1 μg/mL VLPs for 2 h. A moderate increase in actin filament stress fiber formation was observed (arrowheads). Pearson's correlation coefficient was lower in microvessels exposed to Ebola VLPs than in the control, showing an increase in stress fiber formation and endothelial cell activation.

Data are represented as mean ± SEM.

Next, we assessed whether VLPs affect cellular mechanics by modulating the Rho/ROCK pathway. Over-activation of the Rho/ROCK pathway is an underlying mechanism of several vasculopathies, including endotoxin-induced septic vasculopathy (Tasaka et al., 2005, Li et al., 2010, Suzuki et al., 2012). Given that some of the major pathophysiological mechanisms of Ebola virus disease resemble those of bacterial septic shock (Mahanty and Bray, 2004, Bray and Mahanty, 2003), it is conceivable that the Rho/ROCK pathway may also play a critical role in the pathogenesis of the severe vascular leak observed in Ebola disease (Eisa-Beygi and Wen, 2015). To test this hypothesis, we first determined whether we could stimulate the Rho/ROCK pathway in our engineered system and generate a phenotype similar to that observed after the infusion of Ebola VLPs. We used U46619, a small molecule that activates the Rho/ROCK pathway and measured the time- and concentration-dependent response of the vessels (Kobayashi et al., 2016). Treatment of the microvessels with U46619 (10 μM) increased permeability significantly (see Figure 3A). As the concentration of U46619 increased, the barrier permeability progressively increased (see Figure 3B). Immunostaining of F-actin revealed induced alterations in the cellular cytoskeleton associated with the disruption of the endothelial barrier (see Figure 3C). Subsequently, we investigated whether Rho/ROCK pathway inhibition suppresses the Ebola-VLP-induced vascular phenotype. As RevitaCell Supplement (Hansen et al., 2018) is known to specifically inhibit ROCK, we investigated whether this compound could reverse the Ebola-VLP-induced phenotype. We compared vascular permeability of the microvessels-on-chips exposed to Ebola VLPs only with permeability of microvessels exposed to Ebola VLPs and RevitaCell Supplement simultaneously to inhibit the Rho/ROCK pathway. We observed a full suppression of the VLP-induced permeability upon administration of the inhibitor together with the VLPs (see Figure 4). This result shows that Ebola VLPs critically modulate the Rho/ROCK pathway in hemorrhagic shock syndrome.

Figure 3

U46619 Induces Vascular Permeability in the Microvessel-on-a-Chip Platform

(A) Time dependence of U46619-induced barrier opening. Microvessels were exposed to 10 μM U46619, followed by a permeability assay.

(B) Dose response to U46619 in microvessels. Microvessels were treated with several concentrations of U46619 for 1 h to measure permeability.

(C) Immunofluorescence micrographs of on-chip cultured endothelium. After treatment for 1 h with 10 μM U46619, endothelial cells were stained for VE-cadherin (green) and F-actin (red). An increase in actin stress fiber formation was observed (arrowheads). Pearson's correlation coefficient was lower in microvessels exposed to U46619 than in the control, showing an increase in stress fiber formation and endothelial cell activation.

Data are represented as mean ± SEM.

Figure 4

ROCK-Specific Inhibitor Treatment Abolishes VLP-Induced Vascular Permeability

Microvessels were either left untreated, incubated with 1 μg/mL Ebola VLPs, or exposed to both RevitaCell Supplement (0.1 and 1×) and 1 μg/mL Ebola VLPs simultaneously, with an incubation time of 2 h. Data are represented as mean ± SEM.

Existing evidence suggests that Ebola virus envelope glycoprotein GP1,2 is a key mediator of viral pathogenesis and a determinant of disease severity (Mohan et al., 2015). To test whether GP1,2 alone could simulate Ebola VLPs effect, we infused purified GP1,2 into our engineered vessels and measured the dose and time responses. Figure 5A shows that stimulation with 100 ng/ml GPs1,2 for 120 and 240 min led to a significant increase in vessel permeability. Moreover, we measured the dose-response curve using the chip platform (see Figure 5B). Importantly, the increased permeability of GPs1,2-treated microvessels was associated with the formation of stress fibers (see Figure 5C). These results directly show the ability of Ebola GP1,2 to induce vasculopathy and indicate that our chip-based model can detect both VLP-induced and GPs1,2-induced vascular permeability.

Figure 5

Endothelial Cell Activation Induced by the Ebola Glycoprotein

(A) Permeability assay of microvessels exposed to 100 ng/ml GP1,2 at the indicated time points.

(B) Dose response to GP1,2. Microvessels were incubated with GP1,2 at the indicated concentrations for 2 h to measure permeability.

(C) Immunostaining of endothelial cells for VE-cadherin (green) and F-actin (red) after treatment with 100 ng/ml GP1,2. A moderate increase in actin stress fiber formation was observed (arrowheads). Pearson's correlation coefficient was lower in microvessels exposed to Ebola GPs1,2 than in the control, showing an increase in stress fiber formation and endothelial cell activation.

Data are represented as mean ± SEM.

To demonstrate the applicability of our chip-based assay to pharmacological studies, we used this platform to study the effect of two potential drugs. According to our simple working model, Ebola virus stimulates the Rho/ROCK pathway, thereby inducing actin bundle formation and a tensile force that loosens the intercellular junctions formed by VE-cadherin. We targeted this process at two levels: (1) Rho/ROCK signaling (intracellular), via melatonin and (2) VE-cadherin (extracellular) and the associated actin bundles, via FX06 (see Figure 6A) (Bergt et al., 2016, Uyeki et al., 2016, Petzelbauer et al., 2005). We found that both molecules effectively suppress vasculopathy and further showed that FX06, which binds to VE-cadherin, thus reducing adhesion, also affects actin bundle formation directly (or indirectly via Fyn-mediated signaling (Groger et al., 2009)) (Figure S2). Treatment with FX06 counteracted vascular leakage in VLP-treated vessels, which is consistent with the results of previous animal experiments (Roesner et al., 2009) and the clinical benefit noted in a case report (Wolf et al., 2015) (see Figure 6B). However, vascular integrity was not directly measured in any previous study. Similarly, melatonin reduced vascular permeability in our viral hemorrhagic shock syndrome-on-a-chip model (see Figures 6C and S3). A similar effect of melatonin was also observed when we stimulated the Rho/ROCK pathway with U46619 (Figure S4). This observed effect of melatonin is intriguing, as the effect of melatonin on the permeability of vessels has been previously reported in the context of cancer and septic shock, and the function of melatonin has been attributed to Rho/ROCK pathway modulation and subsequent changes to the cytoskeletal elements including actin stress fibers (Borin et al., 2016, Tang et al., 2016). Melatonin has been proposed as a potential drug for Ebola hemorrhagic shock (Masters et al., 2014, Tan et al., 2014, Wiwanitkit, 2014) but has never been tested experimentally. Given that melatonin is a natural molecule in the body and is safe when administered for at least a year (Chahbouni et al., 2010) and given our observation that melatonin effectively suppresses the Ebola-induced loss of vascular integrity, our study suggests that melatonin is a promising drug for treating Ebola hemorrhagic shock syndrome. However, additional investigations are required to confirm the clinical therapeutic efficacy of melatonin.

Figure 6

FX06 and Melatonin Effectively Ameliorate Vascular Integrity Loss in the Ebola Hemorrhagic Shock Syndrome-on-a-Chip Model

(A) FX06 and melatonin rebalance mechanical forces in endothelial cells to restore vascular integrity in Ebola viral hemorrhagic shock syndrome.

(B) Microvessels exposed to 1 μg/mL Ebola VLPs were treated with FX06 at the indicated concentrations for 2 h. Subsequently, a permeability assay was carried out.

(C) Dose response to melatonin in microvessels exposed to 1 μg/mL Ebola VLPs for 2 h, followed by a permeability assay.

Data are represented as mean ± SEM.

Discussion

Vasculopathy is a critical and fatal consequence of Ebola virus infection (Bah et al., 2015, Escudero-Perez et al., 2014, Lyon et al., 2014). Despite extended in vivo studies, the underlying molecular mechanisms are still elusive, no effective cure is available, and treatment strategies are primarily palliative. To discover and develop new drugs for Ebola in a cost-effective manner with high predictive power, microengineered disease models of human organs are needed. The hemorrhagic shock syndrome-on-a-chip described here is the first of its kind. Despite its simplicity, this model is robust, with significant fidelity in mimicking, at least partly, the structure and functions of human microvessels and Ebola disease-associated vasculopathy. This platform permitted high-throughput simulation of the vascular permeability induced by Ebola VLPs and Ebola GP1,2 as well as an increased surface expression of E-selectin, which mediates disseminated intravascular coagulation and death (Boral et al., 2016, Tan et al., 2014).

Our study provides direct evidence for the usefulness of two candidate drugs, FX06 (which directly targets mechanical elements [Urbschat et al., 2014]) and melatonin (which directly targets biochemical signaling [Yang et al., 2016]), for treating Ebola patients. Although melatonin and FX06 are not antiviral molecules, they can be used to reduce the severity of hemorrhagic shock syndrome. Melatonin has not been used clinically to treat Ebola before; however, it was used routinely in other several clinical settings in humans (Sanchez-Barcelo et al., 2010). Intravenous administration of 60 mg melatonin (equivalent to ∼100 μM in blood) is believed to be safe and with no complications (Kucukakin et al., 2008, Gitto et al., 2004). This indicates that the highest concentration of melatonin (100 μM) used in our study is considered safe and can be translated to future clinical applications. Similarly, the highest concentration of FX06 in our model is clinically relevant and has been used clinically to treat one Ebola patient (Wolf et al., 2015).

Our in vitro model will help in understanding the underlying mechanisms of melatonin and FX06. This will allow the development of new compounds capable of delaying the effect of Ebola-induced vascular leakage. Additionally, the model can be of help in screening for therapeutic monoclonal antibodies (Saphire et al., 2018a, Saphire et al., 2018b), in which the effect of antibodies targeting endothelial cells or VLPs to block the interaction between the two can be assayed. Finally, we stress the importance of the dose-response analysis enabled by the proposed platform. Recently, some Ebola drugs such as favipiravir have been abandoned because of problems with dosing (Dunning and Fischer, 2015). The developed in vitro model can aid in the efforts to develop an effective pharmacokinetic model of drug treatments and therefore, assist in designing the proper dose regimens. This chip-based platform will thus be a valuable tool complementing state-of-the-art technologies for combating current and future Ebola outbreaks.

Limitations of the Study

We note that this study has certain limitations that will be addressed in our future studies. The bidirectionality of the flow in the microvessels is one of the limitations in our system. Currently, we are developing a perfusion pump, which can provide a unidirectional flow for each microvessel channel by covering the whole plate, to solve this problem. We, however, anticipate that the observed vascular permeability remains largely unaffected by the flow directionality, as our preliminary analysis shows (Figure S5). The proposed approach is high-throughput, yet the analysis time can be further reduced using high-content imaging systems.

Our approach allows us to disentangle the direct contribution of VLPs and viral proteins to vascular integrity loss from the contributions of host immunity (indirect mechanism) and the process of infection. The proposed in vitro model can be used in future to address the contributions of immune cells and to study endothelial cell infection. Shed glycoproteins from infected macrophages and dendritic cells can be readily assayed using the proposed approach. The engineered vessels can be further improved by including tissue specific ECM and other vascular cells (e.g., pericytes). The pharmacological analysis can be further extended to investigate the kinetics of recovery (after titration of the inhibitors) in the proposed in vitro model and to translate the results to nonhuman primate and human settings. Certain control and complementary experiments need to be performed (e.g., testing VP40-only VLPs) before exploring the translatability of the results. The platform can also be adapted to investigate therapeutic antibodies and other drug options. Finally, this study will contribute to understanding and detection of other highly dangerous viral infections that cause hemorrhagic shock including Lassa and dengue.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.