Clinical and Immunologic Correlates of Vasodilatory Shock Among Ebola Virus–Infected Nonhuman Primates in a Critical Care Model

Abstract

Background

Existing models of Ebola virus infection have not fully characterized the pathophysiology of shock in connection with daily virologic, clinical, and immunologic parameters. We implemented a nonhuman primate critical care model to investigate these associations.

Methods

Two rhesus macaques received a target dose of 1000 plaque-forming units of Ebola virus intramuscularly with supportive care initiated on day 3. High-dimensional spectral cytometry was used to phenotype neutrophils and peripheral blood mononuclear cells daily.

Results

We observed progressive vasodilatory shock with preserved cardiac function following viremia onset on day 5. Multiorgan dysfunction began on day 6 coincident with the nadir of circulating neutrophils. Consumptive coagulopathy and anemia occurred on days 7 to 8 along with irreversible shock, followed by death. The monocyte repertoire began shifting on day 4 with a decline in classical and expansion of double-negative monocytes. A selective loss of CXCR3-positive B and T cells, expansion of naive B cells, and activation of natural killer cells followed viremia onset.

Conclusions

Our model allows for high-fidelity characterization of the pathophysiology of acute Ebola virus infection with host innate and adaptive immune responses, which may advance host-targeted therapy design and evaluation for use after the onset of multiorgan failure.

Ebola virus (EBOV; Zaire ebolavirus), a nonsegmented negative-strand filovirus and etiologic agent of EBOV disease (EVD), has led to increasingly frequent outbreaks over the past decade [1]. EVD is characterized by profuse vomiting and diarrhea, followed by shock, multiorgan failure, and coagulopathy [1]. Despite available licensed monoclonal antibody (mAb) therapies, mortality remains high, particularly among individuals with high viral loads and those who seek care late during illness [2, 3]. In a randomized trial evaluating the efficacy of mAb114 or REGN-EB3, mortality was 69.9% and 63.6%, respectively, among individuals with an EBOV reverse transcription–polymerase chain reaction (RT-PCR) cycle threshold value ≤22.0, despite mAb administration. For each day delay in treatment relative to illness onset, an 11% increase in mortality was observed. Consequently, there is a need to develop therapies that reduce mortality from EVD among individuals who present later during illness, typically with high viral loads and evidence of organ failure.

EBOV infection in nonhuman primates (NHPs) recapitulates nearly all aspects of the pathophysiology of EVD observed in humans, and so these are used to evaluate vaccines and therapies [1, 4]. Notable differences in the NHP model, particularly when 1000 plaque-forming units (PFU) of EBOV is administered intramuscularly, include a more rapidly lethal disease course and an absence of gastrointestinal fluid loss that contributes to shock in humans [1, 4]. To assess the benefit of therapies administered after the onset of critical illness due to EBOV infection, NHP models of supportive critical care are needed. To date, 2 groups of investigators have implemented supportive care models for EBOV infection in NHPs. The first group provided fluid and vasopressor administration for the management of shock [5, 6]. The second group provided respiratory support with invasive mechanical ventilation in addition to fluid and vasopressor support [7, 8]. In each model, supportive care alone did not provide a survival benefit, in part due to the high dose (1000 PFU) of EBOV administered intramuscularly. We implemented an NHP intensive care unit (ICU) model that builds on this work. We inserted a Swan-Ganz pulmonary artery catheter to gain further insight into the daily progression of EBOV pathophysiology by providing measurements of heart function (cardiac index), vascular tone (systemic vascular resistance index), as well as the filling pressures of both sides of the heart (right, central venous pressure; left, pulmonary arterial wedge pressure), which informed our fluid resuscitation and vasopressor administration algorithms. We also incorporated broad and unbiased daily immune cell phenotyping of neutrophils and peripheral blood mononuclear cells (PBMCs). Such a characterization of immune cell subsets has not been defined in EVD. We believe that our model can serve as a platform to evaluate therapies initiated after the onset of shock and multiorgan failure in this high-lethality model.

METHODS

Animal Use

This study was performed with the approval and oversight of the National Institutes of Health’s (NIH’s) Division of Clinical Research Animal Care and Use Committee, Scientific Advisory Board, and Institutional Biosafety Committee. Two age-matched adult Indian-origin rhesus macaques were used in this study weighing 7.46 kg (NHP1, female; 6 years) and 8.54 kg, (NHP2, male; 7 years). They were cared for and treated humanely according to the following policies: the US Public Health Service Policy on Humane Care and Use of Animals (2000), the Guide for the Care and Use of Laboratory Animals (1996), and the US Government Principles for Utilization and Care of Vertebrate Animals Used in Testing, Research, and Training (1985). This study was performed in the animal biosafety level 4 containment unit at the Integrated Research Facility at Fort Detrick (National Institute of Allergy and Infectious Diseases [NIAID], NIH). All Integrated Research Facility animal facilities and programs are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International and are in full compliance with the Federal Select Agent Program's requirement for work with a tier 1 select agent.

EBOV Variant and Study Design

The study design is shown in Figure 1. Briefly, on study day 0 (D0), the animals were inoculated intramuscularly with a target dose of 1000 PFU of EBOV/H sapiens-tc/GIN/2014/Makona-C05 (EBOV/Mak; GenBank accession KX000400). A back titer performed on the inoculum revealed that 1300 PFU was administered to both animals. The isolate was received from Dr Gary P. Kobinger, Public Health Agency of Canada (BioSample SAMN03611815), and passaged twice at the Integrated Research Facility on Vero E6 cells. Following inoculation, the animals were monitored in their cages until D3 and then sedated, with 24-hour critical care support algorithms initiated (supplementary methods). The study endpoint was either D14 or when mean arterial blood pressure (MAP) could not be maintained >40 mm Hg despite high-dose dual vasopressor therapy. Complete postmortem examinations were performed with tissues collected, fixed, and processed for histopathology and immunohistochemistry (IHC) as previously described [9].

Figure 1.

Critical care in acute Ebola virus infection model: study design and blood sampling. Created with BioRender.com. BSL-4, biosafety level 4; EBOV, Ebola virus (Zaire ebolavirus); IM, intramuscular; PBMC, peripheral blood mononuclear cell; PFU, plaque forming unit.

Sedation, Procedures, Supportive Care Algorithms, and Clinical Sampling

Detailed information regarding the sedation algorithms, invasive procedures, supportive care algorithms, and clinical sampling performed in this study is presented in the supplementary methods. A list of consumables is available in Supplementary Table 1.

Physiology Data Cleaning, Smoothing, and Visualization

Continuously captured physiologic data were cleaned to remove artificially elevated pressure readings due to the manipulation of lines for blood draws and measurements of pulmonary arterial wedge pressure and cardiac output. The cleaned data were smoothed in Prism version 8.3.0 (GraphPad) via Savitzky-Golay to 100 neighbors for heart rate, mean noninvasive blood pressure, MAP, and pulse oximetry; zero-order smoothing to 200 neighbors for pulmonary arterial pressure and central venous pressure); and zero-order smoothing to 2 neighbors for core body temperature, which was recorded only once per hour. All physiologic data were visualized on graphs made with Prism.

EBOV Viral Load by Quantitative RT-PCR

A total of 200 µL of plasma and 10% clarified tissue homogenates inactivated by TRIzol LS (Invitrogen) were prepared in a KingFisher 96-well V-bottom deep-well plate (Thermo Fisher Scientific) with 10 µL of proteinase K from the MagMAX Viral Pathogen II Nucleic Acid Isolation Kit (Applied Biosystems). Overall 550 µL of Binding Solution and 20 µL of MVP II Binding Beads were added to the sample wells and mixed. The KingFisher Flex Purification System (Thermo Fisher Scientific) was used to automatically extract total nucleic acids with the customized program MVP_Flex_200 µL. Total nucleic acids were eluted in 70 µL of the Elution Buffer and stored in TempAssure PCR Flex-Free 8-Tube Strips (USA Scientific) until use. EBOV RNA levels were determined with the ABI Fast Dx Real-time PCR System with EBOV Zaire master mix stock (310719-01M; Critical Reagent Program) and SuperScript II Reverse Transcriptase with Platinum Taq DNA Polymerase (Invitrogen). Results were reported as EBOV RNA gene copies per milliliter of plasma or milligram of tissue.

Plaque Assay

The method of determining the EBOV viral titer in plasma and tissues has been described [10], with samples for this study serially diluted 1:10 for a total of 6 dilutions.

Immunohistochemistry

IHC for EBOV-VP40 and EBOV-GP was performed as previously described [11]. IHC for myeloperoxidase (MPO) to identify neutrophils within tissues from EBOV-infected and uninfected macaques for comparison was performed according to the following procedure. Formalin-fixed paraffin-embedded 4-μm-thick tissue sections were deparaffinized and rehydrated through a series of graded ethanol. Antigen retrieval was conducted with Diva Decloaker citrate buffer (DV2004; Biocare Medical) for 30 minutes at 95°C. After rinses with Tris-buffered saline with Tween 20 (TBST), an Avidin and Biotin block (SP-2001; Vector Laboratories) was applied for 15 minutes each with a TBST rinse after each application. Background Sniper (BS966; Biocare Medical) was applied for 30 minutes. Sections were then incubated with rabbit polyclonal anti-MPO antibody (PA5-16672; Invitrogen) or rabbit polyclonal IgG isotype control antibody (ab27478; Abcam) or plain diluent for 60 minutes. After rinses with TBST, the sections were incubated with Peroxidazed 1 (PX968; Biocare Medical) for 10 minutes. The slides were rinsed with TBST and incubated with a secondary biotinylated donkey anti-rabbit antibody (711-065-152; Jackson Immunoresearch) for 30 minutes. The sections were rinsed with TBST, and the VECTASTAIN Elite ABC-HRP Kit was applied for 30 minutes (PK-6100; Vector Laboratories). After rinses with TBST, the sections were visualized by applying the Betazoid DAB Kit (BDB2004; Biocare Medical) for 5 minutes.

Staining and Flow Cytometry Acquisition for Neutrophils and Peripheral Blood Mononuclear Cells

Detailed protocols for the staining of neutrophils in whole blood and for the isolation and staining of PBMCs for flow cytometry acquisition are presented in the supplementary methods with a list of antibodies and reagents detailed in Supplementary Tables 2 to 5.

Whole Blood Neutrophil Flow Cytometry Data Analysis

To analyze changes in neutrophil subsets over time in response to EBOV infection, we employed a high-dimensional flow cytometry approach using OMIQ (Dotmatics). FCS files (flow cytometry standard) were compensated and transformed with the hyperbolic arcsine transformation. CD66abce⁺ neutrophils were selected by removal of doublets with side scatter area vs height; exclusion of dead cells; and removal of CD20⁺, CD3⁺, HLA-DR⁺, CD123⁺, and CD16⁺ cells to remove B cells, T cells, monocytes, dendritic cells (DCs), and natural killer (NK) cells. Next, to eliminate bias from the sample collection, we down sampled each FCS file to 20 000 events. Dimensional reduction was performed with the uniform manifold approximation and projection (UMAP) algorithm [12] and default settings with all remaining fluorescent signals to characterize neutrophils (CD66abce-APC, CD11b-BV510, CD64-BV711, CD62L-BV786, lactoferrin-PE, CD87-PerCp-eFluoor 710, MPO-eFluor 450). Metaclustering was performed with the FlowSOM algorithm (flow cytometry–specific self-organizing map) [13] into a 7 × 7–node grid based on the similarity of surface expression [14]. The elbow criterion method for the within-cluster sum of squares demonstrated an inflection at k = 12, which was used for all subsequent analyses (Supplementary Figure 1). Additional analyses were performed on exported data in Prism version 9.3.1 (GraphPad).

PBMC Flow Cytometry Data Analysis

We employed a high-dimensional flow cytometry approach using OMIQ to analyze daily changes to all PBMC populations during acute EBOV infection. Initial gating was performed in FlowJo (BD Biosciences). Lymphocytes were selected by forward and side scatter, singlet events by forward scatter height vs area and side scatter height vs area, and live cells by exclusion of LIVE/DEAD Blue dye. FCS files of gated populations were exported and uploaded to OMIQ (Dotmatics) with associated metadata. Data were scaled, with cells gated on CD45 positivity and subsampled to 60 000 events per sample. Dimensional reduction was performed with the UMAP algorithm [12] and default settings with all fluorescent signals except CD45-BUV-395, LIVE/DEAD Blue, and KZ52-PE (KZ52 was not used for analysis because of low sensitivity). Metaclustering was performed in 17 × 17 clusters with the FlowSOM algorithm [13] by using all fluorescent signals except those aforementioned. The elbow method for the within-cluster sum of squares demonstrated an inflection at k = 47, which was used for all subsequent analyses. Expression of each marker overlaid on the UMAP projections of the PBMCs is displayed in Supplementary Figure 2. We annotated broad PBMC lineages in the data set based on the following marker proteins: CD8⁺ T cells (CD3⁺, CD8⁺), CD4⁺ T cells (CD3⁺, CD4⁺), NK cells (CD16⁺, NKG2a⁺; lineage negative), B cells (CD20⁺), monocytes (CD11b⁺, CD11c⁻; in UMAP monocyte cluster), classical DCs (CD11c⁺), and plasmacytoid DCs (HLA-DR⁺, CD4⁺, CD123⁺). Metaclusters (MCs) 3 and 32 had subpopulations in >1 lineage and so were excluded from lineage-specific analyses. Additional analyses were performed on exported data in Prism version 9.4.0 (GraphPad).

RESULTS

Viral Load Exponentially Rises After D5

EBOV was undetectable in plasma until D5 postinfection when high levels of viral RNA (approximately 10⁷ genome equivalents/mL) and viable virus (approximately 10⁴ PFU/mL) became detectable (Supplementary Figure 3). Between D5 and D6, an approximately 3-log rise in plasma RNA levels and viable virus was observed, with a less steep rise thereafter.

Decompensated Vasodilatory Shock, Organ Failure, and Coagulopathy Occur After D6

On D4, animals began to display evidence of compensated shock, an early phase during which the body responds to decreased organ perfusion by increasing heart rate and MAP (Figure 2A). By D5, animals remained in compensated shock with MAP >65 mm Hg and stable to increased cardiac index (Figure 2A and 2B). On D6, animals progressed to decompensated vasodilatory shock, a late phase when vasopressor therapy is required to maintain an adequate perfusion pressure (MAP >65 mm Hg), which in our animals was due to declining vascular tone (systemic vascular resistance index) despite adequate cardiac filling pressures and stable cardiac index. By D7 to D8, shock became irreversible with progressive vasodilation and inability to maintain a MAP >65 mm Hg despite rapid escalation in dual vasopressor therapy.

Figure 2.

Clinical data for NHP1 (left; female) and NHP2 (right; male) overlaid with the quantification of EBOV viral load in plasma by quantitative reverse transcription–polymerase chain reaction and stage of shock: A, hemodynamic measures with vasopressor dosing administration; B, cardiovascular function; C, acid-base status; D, liver injury and dysfunction; E, renal dysfunction; F, coagulopathy. Blood pressure data depict mean arterial pressure measurements. ALP, alkaline phosphatase; ALT, alanine transaminase; aPTT, activated partial thromboplastin time; AST, aspartate transaminase; BUN, blood urea nitrogen; CI, cardiac index; CVP, central venous pressure; EBOV, Ebola virus (Zaire ebolavirus); MVe, exhaled minute volume; NHP, nonhuman primate; PAWP, pulmonary arterial wedge pressure; PT, prothrombin time; SVRI, systemic vascular resistance index.

Serum lactate levels began to rise by D4 to D5 contributing to progressive metabolic acidosis with serum pH declining <7.3 by D6, despite compensatory respiratory responses with increased minute ventilation (Figure 2C). Hepatic and renal injury or dysfunction became evident after D6 with serum aspartate aminotransferase, alanine aminotransferase, and serum creatinine levels rising above preinfection baseline levels and steadily increasing thereafter (Figure 2D and 2E). Serum fibrinogen levels increased by D5 and D6 and subsequently declined by D7. Hemoglobin and hematocrit levels precipitously dropped on D7 to D8 (Figure 3A), concordant with declining platelet counts (Supplementary Table 6A and 6B) and fibrinogen levels and rising prothrombin time, activated partial thromboplastin clotting time, and D-dimer levels (Figure 2F, Supplementary Table 7A and 7B), indicative of consumptive coagulopathy.

Figure 3.

Highlighted results from complete blood counts for NHP1 (left; female) and NHP2 (right; male) by study day: A, red blood cell count, hemoglobin concentration, and hematocrit; B, white blood cell count and neutrophil count and percentage; C, lymphocyte count and percentage; D, monocyte count and percentage. HCT, hematocrit; Hgb, hemoglobin; NHP, nonhuman primate; RBC, red blood cell; WBC, white blood cell.

Neutrophil Count Nadir and Population Shift by D6

To gain insight into neutrophil biology in our model, we assessed absolute neutrophil counts (ANCs) and the phenotypic profiles of neutrophil populations based on 7 surface and intracellular markers. We identified 12 circulating CD66⁺ neutrophil MCs, which varied by day postinfection (Figures 4A and 4B and 5). On D0, both animals had an ANC of approximately 5000 cells/μL (Figure 3B) composed primarily of MCs 1 and 4 (Figures 4B and 5, Supplementary Figures 4 and 5). MC1 had markers of maturity [15] (CD11b^hiCD87^hiCD64^hi) and high granule expression (MPO^hilactoferrin^hi; Figures 4C and 5). MC4 had markers of intermediate maturity (CD11b^intermediateCD87^intermediateCD64^intermediate) and high granule expression. Following ICU procedures on D3 (Figure 3), a rise in the ANC was observed in NHP2, and MCs 5 and 9 predominated in both animals. MC5 had markers of immaturity (CD11b^loCD87^loCD64^lo) and intermediate granule expression (MPO^intermediatelactoferrin^intermediate). MC9 had markers of maturity and intermediate granule expression. On D4, the ANC dipped and then increased above baseline on D5, with MC1 accounting for >90% of circulating neutrophils in both animals on these days. On D6, ANC nadir was observed in both animals. MCs 2 and 3 predominated among the few circulating neutrophils in NHP1 on D6. MCs 2 and 3 had markers of maturity and high granule expression but relatively lower CD62L expression than MC1. MCs 1 and 9 predominated in NHP2 on D6. ANC rebounded in both animals after D6. In NHP1, MCs 2 and 3 predominated on D7. In NHP2, MCs 1, 2, and 3 predominated on D7, and MCs 1, 3, and 4 predominated on D8.

Figure 4.

Cell clustering showed that distinct populations of neutrophils emerge following an Ebola virus infection. A, UMAP of flow cytometry data colored by the FlowSOM MC assignment from pooled neutrophil samples and samples differentiated by day. In some plots, parts of MC2 were grouped with MC1 after UMAP dimensionality reduction. In those cases, the label for MC2 is located in the middle of the MC2 populations on the UMAP plot. B, The proportion of each MC by day as determined by FlowSOM. C, The corresponding heat map characterizing each MC by median surface marker expression. FlowSOM, flow cytometry–specific self-organizing map; MC, metacluster; MPO, myeloperoxidase; UMAP, uniform manifold approximation and projection.

Figure 5.

Schematic of change in proportion of neutrophil MCs with the highest proportions (>10%) over time by day. Proportions are averages from both animals and are relative to total neutrophils measured in the animals. Proportions are best fit to represent the stacked bar graph in Figure 4B and to prevent overlapping of the MCs for clarity. Cell markers: hi = high expression; int = intermediate expression: lo = low expression. ICU, intensive care unit; MC, metacluster.

Progressive Lymphopenia With Few Changes in T-, B-, and NK-Cell Populations Prior to D5

Absolute lymphocyte counts in both animals declined below baseline levels by D3, nadired by D6, and subsequently rebounded (Figure 3C). We identified 15 T-cell MCs, 12 B-cell MCs, and 2 NK-cell MCs (Figure 5A; Supplementary Figures 6A–C, 7, and 8). T cells made up 37% to 72% of PBMCs during the study (Figure 6B, Supplementary Table 8) and were comprised of both antigen-experienced (CD95⁺) [16] T cells (CD8⁺ MCs 36, 37, 40, and 41; CD4⁺ MCs 28, 34, 35, 38, 39, 46, and 47; CD4⁺CD8⁺ double-positive MC42; and CD4⁻CD8⁻ double-negative [DN] MC45) and antigen-naive T cells (MCs 43 and 44) (Supplementary Figure 7). The proportion of most individual T-cell MCs was stable from D0 to D5 (Supplementary Figure 6A, Supplementary Table 9). A notable exception was the CXCR3^hi MC34, which decreased in proportion by D5 and became nearly undetectable thereafter (Figure 7, Supplementary Figure 9A and 9B). Additionally, the DN MC45, which made up 3.25% circulating T cells at D0, accounted for 11% by D7.

Figure 6.

Flow cytometry results from 29-color spectral cytometry. A, UMAP projection of all MCs from all animals and time points. Each MC is labeled at its centroid. B, Proportion of total PBMCs by each major lineage per day. C, Proportion of total monocytes by each monocyte MC per day. D, Expression of CD163 in monocytes by day. DC, dendritic cell; MC, metacluster; NK, natural killer; PBMC, peripheral blood mononuclear cell; UMAP, uniform manifold approximation and projection.

Figure 7.

Notable changes in PBMC MC proportions after EBOV infection. Notable changes in MC proportions after EBOV infection: in the first 5 days, left of the dotted line; from days 4 to 8, right of the dotted line. Cell MCs that increased in proportion, black arrows; cell MCs that decreased in proportion, white arrows. Proportions are relative to the total number of cells per type (eg, MC34 decreased as a proportion of total T cells). DC, dendritic cell; B, B cells; EBOV, Ebola virus; MC, metacluster; Mono, monocytes; NK, natural killer cells; PBMC, peripheral blood mononuclear cell; T, T cells.

B cells made up 18% to 34% of PBMCs throughout the study (Figure 6B, Supplementary Table 8) and included MCs with expression markers consistent with the following [17]: transitional B cells (CD27⁻IgM/D⁺CD38^hi) in MC9; naive B cells (CD27⁻IgM/D⁺CD38^lo/−) such as MCs 1, 2, 8, 14, and 27; non–class-switched memory B cells (CD27⁺IgM/D⁺CD38^lo/−) such as MCs 13 and 23; and class-switched memory B cells (CD27⁺IgM/D⁻CD38^lo/−) such as MCs 18, 21, and 22 (Supplementary Figure 7). We also detected an antigen-experienced (CD95⁺) IgD/M⁻CD27⁻ DN B-cell population in MC29. The proportion of most individual B-cell MCs was stable from D0 to D5 (Supplementary Figure 6B, Supplementary Table 10). Notable exceptions were CXCR3^hi MCs 1 and 21, which decreased in proportion by D5 and became nearly undetectable thereafter (Figure 7, Supplementary Figure 9A and 9B). Decreased proportions of MCs 2, 13, and 22 and increased proportions of MCs 9, 14, and 29 were observed after D5 (Figure 7). Notably, DN MC29 composed only 3.85% of circulating B cells at D0, increasing to 17.5% by D7.

NK cells made up <4% of PBMCs across the study (Figure 6B, Supplementary Table 8). We identified 2 populations: MC30 (CD16⁺CD8⁺NKG2a⁺) and MC31 (CD16⁻CD8⁺NKG2a⁺; Supplementary Figure 6C). Both animals had an increase in the proportion of MC31 by the terminal time point, representing 73% of NK cells (Figure 7, Supplementary Table 11). Expression of the activation marker CD69 increased in both MCs starting at D5 and peaking at the terminal time points (Figure 7, Supplementary Figure 10).

Decreased Classical and Increased DN Monocytes Circulating by D4

Absolute monocyte counts nadired in both animals on D6 and subsequently rebounded above baseline in NHP1 only (Figure 3D). Monocytes made up 2% to 30% of PBMCs during the study (Figure 6B, Supplementary Table 8). We identified 13 monocyte MCs: MCs 7, 10, 12, and 17, consistent with [18] classical monocytes (CD14⁺CD16⁻); MC4, a mixed population of intermediate and nonclassical monocytes (CD14^dimCD16⁺); MCs 6, 11, 15, 16, 19, 20, and 26, as DN monocytes (CD14⁻CD16⁻); and MC5, a mixed population of CD206^hi cells consistent with alternative M2 activation [19] (Figure 5C, Supplementary Figure 7). A gradient of increasing expression of CD38, a marker of inflammation [20], was observed in classical MCs (17 > 10 and 12 > 7) and DN MCs (20 > 11 > 19, 6, 26, and 1). The monocyte repertoire began shifting at D4 with a decline of all classical MCs (7, 10, 12, and 17) and expansion of DN MCs 6, 11, and 20 (Figure 7, Supplementary Table 12). Of all monocytes, MCs 6 and 11 had the highest expression of CD163, a hemoglobin-haptoglobin scavenger, and expansion of these MCs corresponded with an overall increased detection of CD163 (Figure 5D, Supplementary Figure 10C and 10D). By D6, there was near complete loss of classical MCs, and the proportion of M26, a DN CD163^lo population, significantly increased (Figure 7). Strikingly, DN monocytes accounted for just 15% of monocytes at D0 but increased to >97% of the monocyte repertoire at D7 and D8.

DCs at baseline made up 0.31% and 2.91% of PBMCs in NHP1 and NHP2, respectively, but the repertoire diminished by D6 and did not recover (Figure 6B, Supplementary Table 8). We identified 3 populations of DCs: 2 classical (CD11c⁺), MC24 (CD11b⁻) and MC25 (CD11b⁺), and 1 plasmacytoid, MC33 (CD11c⁻CD123⁺CD206⁺; Supplementary Figures 7D and 8). At baseline, NHP1 had nearly equal proportions of all 3 MCs, while in NHP2, MC25 was dominant (67.5%) and MC33 was low (3.9%; Supplementary Table 13). Despite this initial difference, both animals saw a decrease in MC25 during the study with a rise in MC24 until the terminal time point when MC33 became dominant (Figure 7).

Hemorrhage, Necrosis, and EBOV-Infected Monocytes and Macrophages at Necropsy

Gross and histopathologic findings of tissues collected at necropsy in both animals were typical of lesions in rhesus macaques with acute EBOV infection, including widespread hemorrhage, necrosis, and inflammation with positive IHC staining for EBOV-VP40 and EBOV-GP proteins within circulating monocytes and tissue macrophages (supplementary results, Supplementary Figure 11). The brain and spinal cords of both animals were devoid of any significant findings associated with EBOV infection, such as an absence of EBOV-VP40 and EBOV-GP antigen staining within parenchyma, although some circulating monocytes within these tissues stained positive. Despite this, dorsal root ganglia from along the spinal cord, trigeminal ganglia, and peripheral autonomic ganglia had multifocal mild to moderate neuronopathy, characterized by neuronal degeneration and necrosis with interstitial macrophage activations, as well as positive IHC staining VP40 antigen within interstitial macrophages and stromal cells (Supplementary Figure 12). Additionally, nearly all tissues collected from both animals at necropsy had replication-competent EBOV isolated, including the eye and optic nerve, reproductive organs, brain cortex, and lumbar spinal cord (Supplementary Table 14). Of note, the bone marrow showed positive IHC staining in macrophages with mild hypocellularity but was otherwise intact.

To investigate the disappearance of neutrophils from circulation on D6, we performed additional IHC staining for MPO, which was contained within primary neutrophil granules. EBOV-infected macaques showed markedly increased MPO⁺ cells in the white pulp of the spleen, liver, kidney, lung, and heart as compared with uninfected controls (Supplementary Figure 13, Supplementary Table 15). However, there were no significant changes in MPO⁺ cells in the gonad, duodenum, and colon in infected vs uninfected macaques.

DISCUSSION

Understanding the pathophysiology of EVD in humans has increased in the past decade with large-scale outbreaks providing the opportunity for enhanced observation [1]. While mAb-based therapies have improved EVD survival, mortality remains high among individuals seeking care after the onset of critical illness. Our NHP ICU model is intended to provide new insights into EVD pathogenesis and allow for evaluation of therapies initiated after critical illness onset. In this study, we observed few physiologic or immunologic changes from D0 to D3, except for a D3 shift in circulating neutrophil populations, likely in response to invasive procedures.

By D4, animals exhibited an increased heart rate and a subtle increase in serum fibrinogen levels consistent with early inflammatory responses prior to viremia onset. D4 also revealed a shift in circulating monocyte populations with a decreased proportion of circulating classical monocytes, which perhaps traffic directly into infected lymph nodes during this previremic stage. Trafficking of classical monocytes into lymph nodes via high endothelial venules, rather than draining from infected sites via afferent lymphatics, has been described in other infections where lymph nodes become inflamed [21, 22]. We observed an increased proportion of circulating DN monocytes at D4. DN monocytes have been recently described in NHPs and humans infected with EBOV [18, 23]. In NHPs, this population was phenotypically and transcriptionally consistent with proliferative monocyte progenitors [18]. Therefore, an increased proportion of circulating DN monocytes at D4 is consistent with early emergency myelopoiesis.

By D5, animals transitioned from compensated to decompensated vasodilatory shock coincident with the onset of high-level viremia. While metabolic acidosis was detectable at D5, clinical laboratory evidence of renal or hepatic injury was not. In blood, absolute lymphocyte counts declined with few shifts in circulating T-, B-, and NK-cell populations. Given the early stage of infection, prior to onset of adaptive immunity, T- and B-cell death rather than migration into tissues likely accounted for the decreased absolute lymphocyte count. This is consistent with prior reliable observations of indirect lymphocyte death in NHPs and humans infected with EBOV [24, 25]. A potential exception might be CXCR3^hi T- and B-cell populations, which proportionally decline in circulation by D5 and become nearly undetectable thereafter. CXCR3 is a chemokine receptor highly expressed on Th1-type CD4⁺ T cells, effector CD8⁺ T cells, and naive and memory B cells, which coordinates trafficking of cells into inflamed peripheral and lymphoid tissue following activation by interferon-inducible ligands CXCL9, CXCL10, and CXCL11 [26–29]. Also notable on D5 was the predominance of DN CD163^hi monocytes in circulation. CD163 is a member of the scavenger receptor cysteine-rich family and is upregulated on immunomodulatory M2 monocytes associated with resolving inflammation and wound repair [30]. Its mRNA and protein expression are upregulated by macrophage colony-stimulating factor during phagocytic differentiation [31]. An increased proportion of these cells on D5 further indicated emergency myelopoiesis. From D6 on, the proportion of circulating DN CD163^lo monocytes significantly increased. Activated monocytes are known to shed CD163, and high levels of soluble CD163 have been associated with EVD severity [32].

During D6 to D8, animals developed progressive and ultimately irreversible shock, acidosis, organ failure, and consumptive coagulopathy. ANCs nadired on D6 with decreased circulation of CD62L^hi mature neutrophils. CD62L, or L-selectin, is a leukocyte-cell surface adhesion molecule that facilitates leukocyte adhesion to endothelium and migration into tissues. D6 neutrophil recruitment into tissues was associated with rapidly progressive shock and organ failure, perhaps in part attributable to neutrophil degranulation and reactive oxygen species generation. An abrupt decline in hemoglobin concentration on D7 to D8 suggests hemorrhage into tissues at this stage, supported by necropsy findings. A precipitous drop in platelet counts on D6 and declining serum fibrinogen levels with rising prothrombin time, activated partial thromboplastin clotting time, and D-dimer levels by D7 indicated that peak consumptive coagulopathy was coincident with putative timing of tissue hemorrhage. Given that neutrophils are known to mediate tissue injury and immunothrombosis during severe infection, harmful neutrophil effector functions are a potential therapeutic target during the late stages of EBOV infection. On D7 to D8, an increased proportion of circulating CD4⁻CD8⁻ DN T cells and IgD/M⁻CD27⁻ DN B cells, as well as transitional and naive B cells, was observed. The contribution to EVD pathogenesis of these cell phenotypes is not known. Heterogenous DN T-cell subpopulations have been implicated as proinflammatory or immunoregulatory in malignant, autoimmune, and infectious diseases [33] and may represent γδ T cells [34]. DN B cells, thought to be unique memory B cells, contribute to the pathogenesis of infectious and noninfectious inflammatory conditions [35].

In this study, NHPs received care comparable to that of critically ill humans in modern ICUs, including ventilatory support, vasopressors, empiric antibiotics, and stress-dose steroids, as well as fluid, electrolyte, and acid-base management. However, the timing of death in our study was similar to that observed in other NHP studies where EBOV 1000 PFU was administered intramuscularly with or without supportive care [5–8, 36–38]. While this model, with near uniform lethality, does not reflect the typical dose or route of infection in humans [4] or the improved survival observed in humans who receive ICU care [39, 40], it successfully predicted the efficacy of now licensed vaccines [41–43] and mAb-based therapies initiated prior to critical illness onset [44, 45]. Our study incrementally adds to the understanding of EBOV pathophysiology and pathogenesis with our characterization of the progression of vasodilatory shock, despite adequate cardiac filling pressures and function with a Swan-Ganz pulmonary artery catheter, the timing of which was closely associated with organ failure and consumptive coagulopathy after D6. We also characterized the daily kinetics of circulating neutrophil and PBMC populations, providing new insights into early shifts in monocyte populations, timing of neutrophil migration into tissues, and near homogenous distribution of lymphocyte loss. Additionally, we emphasize the need to better define the phenotypic and functional profiles of specific T- and B-cell populations during EBOV infection.

We acknowledge limitations of this resource-intensive model, including that it can be accomplished in only a few maximum-containment laboratories worldwide and requires dozens of staff approved to work in this environment. Indeed, while our observations are limited to 2 animals with some variability in the disease course between them, this model undoubtedly provides an unrivaled opportunity to define daily kinetics of physiologic, immunologic, and molecular responses longitudinally across strain, dose, and route of EBOV infection in future studies. The model also provides a promising platform to evaluate the efficacy of therapies initiated after the onset of critical illness and to closely define the in vivo cellular and molecular mechanisms of these therapies alongside clinically relevant outcomes, such as severity of shock, organ dysfunction, and coagulopathy, as well as mortality. Finally, utility of this model extends beyond EBOV and can be applied to inform the pathogenesis and treatment of other highly lethal emerging and reemerging pathogens that threaten public health.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank the dedicated research facility staff at the Integrated Research Facility, NIAID, for their expertise and care, which made this study possible.

Author contributions. D. S. C. and S. R. S. conceived of the study, wrote the animal study plan, and designed the critical care algorithms with assistance from M. R. H., R. J. R., R. H., H. K., K. C., R. B., M. S. C., L. M. H., and J. R. S. D. S. C., M. S. C., R. J. R., K. C., and R. B. performed animal procedures and oversaw animal care for the study. K. H. performed daily hematologic and coagulation panels and preserved and inventoried study samples. T. L. B. performed RT-PCR. D. J. D., F. R.-H., C. J., P. S., and J. H. performed plaque assays. S. M. A. stained fresh neutrophils, preserved and stained PBMCs, designed the PBMC flow cytometry panel, and performed all flow cytometry with assistance from B. B. S. W., J. R. S., A. P. P., K. M. V., and D. S. C. designed the neutrophil flow panel and gave critical input to the design of the PBMC flow panel. A. P. P. and H. L. T. performed flow cytometry analysis with upstream assistance from S. M. A. and B. B. D. X. L. performed the necropsies, histopathologic analysis, and IHC. S. R. S. performed physiologic data analysis with assistance from S. W. S. R. S. drafted the manuscript with critical input from D. S. C., A. P. P., H. L. T., J. R. S., K. M. V., and S. M. A. All authors approved the submitted version of the manuscript.

Data availability. Raw data files to support the conclusions in this article are available upon request of the corresponding author.

Disclaimer. The opinions, interpretations, and conclusions contained herein are those of the authors and are not necessarily endorsed by the NIH.

Financial support. This work was supported in part by the Intramural Research Program, NIAID, NIH. This project has also been funded in part by the contract with Laulima Government Solutions, LLC (HHSN272201800013C) with federal funds from the NIAID, NIH, Department of Health and Human Services. S. M. A, R. J. R., K. C., R. B., D. J. D., D. X. L., T. L. B., K. H., B. B., F. R.-H., C. J., P. S., J. H., L. M. H., and M. R. H. performed this work as employees under this contract.

Supplement sponsorship. This article appears as part of the supplement “10^th International Symposium on Filoviruses.”