This study used de-identified pre-existing specimens and clinical records obtained during an outbreak and was determined to be exempt by the Centers for Disease Control and Prevention Human Research Protection Office. CDC Protocol Number: 6920.

Blood samples obtained in EDTA tubes as part of routine clinical care for RVF patients were analyzed using commercial multiplex assays according to the manufacturer’s instructions. The samples were aliquoted, frozen, and gamma irradiated on dry ice with 5x10⁶ rads. In preparation for use in the assays, samples were thawed and centrifuged to remove membrane particles. The freeze thaw process and use of whole blood in these assays were previously demonstrated using 8 human blood samples from healthy controls obtained in the US [22]. Eight healthy human samples were also used to define the normal range for each analyte in this study. These samples were obtained via a normal healthy phlebotomy program in Atlanta, GA. Sex, age and race/ethnicity of the donors were unknown. Fifty-nine analytes were assessed in ten commercially available assays from Invitrogen (Carlsbad, CA, USA), Millipore (Billerica, MD, USA), and Affymetrix (Santa Clara, CA, USA). The largest of these was a 30-plex assay for granulocyte colony stimulating factor (G-CSF), granulocyte macrophage colony stimulating factor (GM-CSF), interferon (IFN)-α, IFN-γ, interleukin (IL)-1β, IL-1RA, IL-2, IL-2R, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p40/p70), IL-13, IL-15, IL-17, tumor necrosis factor (TNF)-α, Eotaxin, interferon gamma induced protein 10 (IP-10), monocyte chemoattractant protein (MCP)-1, monokine induced by gamma interferon (MIG), macrophage inflammatory protein (MIP)-1α, MIP-1β, regulated on activation normal T-cell-expressed and secreted (RANTES), endothelial growth factor (EGF), fibroblast growth factor (FGF)-basic, hepatocyte growth factor (HGF), and vascular endothelial growth factor (VEGF) that was performed using an overnight incubation (Invitrogen). Five additional assays also incubated overnight included three singleplex assays for Ferritin, ADAMTS (a disintegrin and metalloproteinase with thrombospondin motifs)-13, and complement factor H (CFH), a two-plex assay for tissue factor and thrombomodulin, and a four-plex assay for von Willebrand’s factor (vWF), c-reactive protein (CRP), Fibrinogen, and platelet factor (PF)-4 (Millipore). One single-plex assay for IFN-β, an eleven-plex assay for E-selectin, fractalkine, granzyme B, melanoma growth stimulation activity alpha (GRO-α), IL-29, L-selectin, MCP-2, MCP-3, sCD40L, TNF-R2, tissue plasminogen activator (tPA), and a six-plex assay for D-dimer, plasminogen activator inhibitor (PAI-1), platelet endothelial cell adhesion molecule (PECAM), P-selectin, sFas-ligand, TNF receptor (TNF-R1) were performed using a 2 hour incubation (Affymetrix). A two-plex assay for intercellular adhesion molecule (ICAM) and vascular cell adhesion molecule (VCAM) was run with an overnight incubation (Affymetrix). Data were collected on a Luminex 200 (Austin, TX) and all assay results were reported in pg/mL, ng/mL, or mg/dL. For 11 of the biomarkers evaluated (IFN-β, G-CSF, IL-5, IL-13, IL-17, GM-CSF, TNF-α, IL-2, IL-7, IL-4, and tissue factor) levels of the biomarker were near or below the limit of detection in all patient samples, so data are not presented graphically.

RNA was isolated from gamma-irradiated whole blood using MagMax Total RNA Isolation Kit (Life Technologies, Grand Island, NY). qRT-PCR of RNA was conducted with established primer and probe sets for the RVFV L segment [23]. A standard curve, expressed as tissue culture infective dose 50 (TCID₅₀)/mL, was generated from normal human whole blood spiked with a wild-type RVFV stock. The standard curve was used to convert raw Ct values to relative TCID₅₀/mL.

Maxisorp plates (Nalgene-Nunc, Rochester, NY) were coated with RVFV lysate prepared from infected Vero-E6 cells as previously described [24], diluted 1:2000 in PBS, and allowed to adsorb overnight at 4°C. Plates were blocked in 5% milk in PBS with 0.1% Tween-20 (PBST) with 5% FBS for 1 hour at 37°C. Patient samples and controls were serially diluted in blocking buffer and incubated on plates for 2 hours at 37°C. Plates were washed three times in PBST then incubated for 1 hour at 37°C with anti-human IgG HRP or anti-human IgM HRP (Jackson ImmunoResearch Inc, West Grove, PA) diluted 1:5000 in blocking solution. Plates were washed three times in PBST before incubation in TMB substrate (KPL, Milford, MA) for 10 minutes. Reactions were stopped with TMB stop solution, and plates were read at 450nm. Data were analyzed using Excel (Microsoft Corp, Redmond, WA) and Prism (GraphPad Software Inc, La Jolla, CA).

Descriptive analyses were performed using STATA 13.0 (StataCorp, College Station, TX) and Prism. Patient means were compared to those of the controls. P-values for all statistically significant comparisons are included in the supplementary table (S1 Table). Patient biomarkers were analyzed for correlation with RVFV viral load using Spearman’s rank correlation. Control biomarker values were compared to case values during two disease periods: days 7–19 and days 20–33. These time frames were chosen because virus was isolated in all cases until day 20. Wilcoxon rank sum was used to determine the significance of differences in control biomarker levels versus case biomarker levels. The False Discovery Rate technique was used to adjust for multiple comparisons with an overall alpha of 0.05 [25].

Case 1 was a middle aged male butcher who presented to a local hospital in Kabale District with a chief complaint of epistaxis. He had been well until a week prior to admission when he developed a high-grade fever associated with headache, neck pain, joint pain, and worsening fatigue. His admission labs were notable for an elevated white blood cell count of 16,100/μl, a decreased hemoglobin of 7.9 g/dl, and decreased platelets of 29,000/μl. Liver function tests revealed an elevated total bilirubin of 16.8 mg/dl and elevated transaminases. Creatinine and blood urea nitrogen were increased at 12.5 mg/dl and 240.5 mg/dl, respectively. Testing for malaria was negative. He was transferred to another hospital for transfusion and further management and recovered.

Case 2 is a teenage student who presented to a local hospital in Kabale with a five day history of fever, myalgia, joint pain, headache and diarrhea as well as three days of epistaxis. He had cared for goats and sheep at home. On admission he was noted to have epistaxis and scleral icterus. Laboratory evaluation revealed an elevated unconjugated hyperbilirubinemia of 9.72 mg/dl, elevated AST of 470 U/L, decreased total protein of 3.43 g/dl, elevated amylase of 386 U/L, decreased hemoglobin of 7.3 g/dl, and decreased platelets of 45,000/μL. Renal function was normal. He was treated empirically with broad spectrum antibiotics and required blood transfusions. His hospital course was complicated by seizures and altered mental status, but gradually improved.

Case 3 is a young adult builder who presented to a local hospital with fever, vomiting, abdominal pain, arthralgias, myalgias, and photosensitivity. He also had a history of epistaxis and hematemesis. Laboratory data were not available for this case.

The relative viral loads (log TCID₅₀/mL) in each of the three cases decreased over time (Fig 1). All patients still had detectable viral RNA after disease resolution, which likely represents the sensitivity of the assay rather than persistence of infectious virus since viral isolation attempts before day 20 post fever onset were successful and isolation attempts at time points 20 and more days post fever onset were unsuccessful. RVFV IgM was detectable on first date of blood sampling for Cases 1 and 2, at day 13 for Case 1, day 5 for Case 2, but not until the second blood sample for Case 3, on day 11 (Fig 2). RVFV IgG was undetectable in all three cases prior to 10 days post fever onset but was detectable at day 17 for Case 1, day 11 for Case 2, and day 20 for Case 3 (Fig 2).

Correlation of biomarkers with viral load was assessed. Only those that were significant after adjusting for multiple comparisons are presented in Table 1. sICAM-1, CRP, IP-10, eotaxin, tPA, D-dimer, granzyme B, and MCP-2 were all significantly positively correlated with viral load. P-selectin, ADAMTS13, CFH, and Fibrinogen were all significantly negatively correlated with viral load. Of note, three of these factors (CRP, P-selectin and CFH) did not vary significantly outside the range of normal.

Hemostatic derangement has been observed in human RVF cases, and was noted in all three of these patients. Although some markers of coagulopathy have been studied in rhesus monkeys [26], there is limited data regarding human markers of coagulopathy during RVFV infection. For this reason, case biomarkers of coagulopathy were compared to control levels during two disease periods: days 7–19 (virus isolation positive) and days 20–33 (virus isolation negative). Only data in which patient values were significantly outside the normal range are shown in the main figures. Any data obtained but not included in the main figures are presented in the Supplementary material for completeness (S1 Fig). Levels of ADAMTS13, D-dimer, PF4, and vWF were significantly elevated in cases compared to controls during both disease periods (Fig 3). Thrombomodulin was significantly lower in cases compared to controls in both disease periods, but this data appears to be skewed by Case 2 with much lower levels than the others. Fibrinogen levels in cases were significantly lower than control levels early in the disease course, however mean fibrinogen levels increased in cases and were no longer significantly different from control levels after day 20. In contrast, tPA levels in cases were significantly elevated early during the course of illness and decreased such that mean case levels were within the normal range later in the disease course.

Endothelial expression of cell surface molecules can modify vascular stability, coagulation, and leukocyte recruitment [27]. Upregulation of these molecules can be associated with infection and inflammation. Given that microscopic changes in the endothelium have been noted in monkeys infected with RVFV [28], case endothelial marker levels were compared to control endothelial marker levels in the two disease periods. Only data in which patient values were significantly outside the normal range are shown in the main figures. Any data obtained but not included in the main figures is presented in Supplementary material for completeness. Variations in endothelial marker levels were seen during both phases of illness among cases (Fig 4). L-selectin and VEGF levels were significantly higher in cases than controls throughout the illness. However, PECAM-1 was only significantly elevated early in the disease in cases compared to controls. In contrast, case E-selectin levels were normal early in the illness course but increased over time and were significantly elevated in cases late in the course of illness. Several endothelial markers were decreased among cases compared to controls. Mean VCAM levels were significantly lower in cases compared to controls throughout the course of illness. Soluble ICAM-1 levels were within the normal range early in the course of illness but decreased significantly late in the course of illness among cases compared to controls.

Chemokines and cytokines play a vital role in recruitment, activation and maturation of the immune system; for this reason, we compared case levels of a series of cytokines and chemokines to control levels. Only data in which patient values were significantly different from controls are shown in the main figures. Any data obtained but not included in the main figures are presented in the Supplementary material for completeness.

Case levels of IL-8, MCP-2, MCP-3, fractalkine, GRO-α, and IP-10 were significantly elevated when compared to control levels throughout the course of illness (Fig 5). IL-10 and MIP-1β levels in cases were significantly elevated compared to control levels early during the illness course. However MIP-1β levels were within the normal range and did not exhibit marked variation during the disease course. Certain chemokines and cytokines were significantly lower in cases compared to controls. Case IL-1β and eotaxin levels were lower throughout the disease course. IL-12 levels in cases were significantly lower early in the course of illness only.

RVFV infection in vitro and in animal models has been associated with decreased expression of type 1 interferons, secondary to the interferon antagonism activity of the RVFV NSs protein [8]. Consistent with this, IFN-α was lower in cases compared to controls. IFN- γ and IL-29, a type III interferon, were noted to be higher in cases compared to controls but did not fluctuate significantly during the clinical course.

Our assays also examined several different types of growth factors. Mean HGF levels in cases were significantly elevated early during the course of illness only. Mean case levels of FGF-basic were significantly elevated compared to mean control levels throughout the illness course; however, they were still within the range of control values.

Granzyme B, a protease found in cytotoxic lymphocytes and natural killer cells that mediates apoptosis, and three other mediators of apoptosis, sFAS-ligand, sTNF-RI and sTNF-RII were found to be significantly elevated in cases compared to controls (Fig 6).

We also measured non-specific markers of inflammation including C reactive protein (CRP) and ferritin (Supplementary material and Fig 6). Although case CRP levels were within the normal range and were not significantly elevated compared to mean control CRP levels, they did tend to be elevated early in the disease and decreased over time. Ferritin levels were significantly elevated in cases throughout the illness. sCD40L, which is mostly derived from activated platelets, was elevated in all patients throughout the course of infection.