The 2013–2016 Ebola virus (EBOV) crisis in West Africa devastated the health care and wider infrastructures of many communities, and subsequent sporadic cases associated with sexual transmission have been reported (1
). The outbreak was unprecedented in scale, providing an opportunity for an in-depth analysis of infected humans and observations of previously unknown aspects of EBOV biology. The epidemic occurred at a time when high-resolution genomic analysis could be used to analyze samples obtained from patients during the outbreak (2–4
). Using these approaches, viral genome evolution has been investigated (2
), and the host immune response has been shown to play an important role in the outcome of infection (5
). Additionally, differential cellular transcriptional profiles of blood samples from patients who presented at an Ebola virus treatment center have been shown to be associated with an outcome of either survival or death. Differences in activation of natural killer cells suggested a large acute-phase response and activation of the innate immune response (4
Many of the patients from West Africa who had Ebola virus disease (EVD) would have had an underlying burden of infectious disease. Additionally, during an EVD breakout, other infections may be present and/or translocate from one organ to another, complicating disease and patient management and influencing infection outcome. Guinea, where the 2013–2016 outbreak originated, experiences high levels of malaria transmission (>1 case per 1,000 population), with Plasmodium falciparum
being responsible for 100% of cases in 2013 (6
as a coinfection has been implicated in the outcome of EVD (7
), and a cohort study suggested that the presence of Plasmodium
was associated with survival (8
). The presence of GB virus C (GBV-C) (formally known as hepatitis G virus) was examined in 49 patients positive for EBOV from Sierra Leone (9
), where both survival and GBV-C status were found to be associated with age (9
). There have been several case reports of Gram-negative septicemia in patients with EVD (10
), although very little data exist regarding the frequency of this septicemia. However, the incidence of bacteremia appeared to be low upon clinic admittance (11
). Assessing the presence of bacterial infection of the blood from large numbers of patients infected with high-consequence infections is complicated, particularly under field conditions of high containment, and the primary need is to provide an accurate diagnosis of EBOV.
During the outbreak in Guinea, the European Mobile Laboratory provided a frontline diagnostic service by testing blood samples for EBOV via quantitative reverse transcription-PCR (qRT-PCR) and for the presence of Plasmodium
spp. by using an immunochromatographic assay (12
). Patients who tested positive for EBOV were admitted into the Ebola Virus Treatment Centre and treated according to guidelines from Médecins Sans Frontières (13
). These patients subsequently either went on to succumb to EVD (hospitalized fatalities) or survived infection (hospitalized survivors). The European Mobile Laboratory archived large numbers of these leftover diagnostic samples, primarily consisting of RNA. Additionally, oral throat swabs were taken from people in the community who died (community deaths) to test for EBOV. Blood samples were also taken from patients who were convalescent for EBOV to prove that these patients were EBOV negative and thus could be discharged from the Ebola Virus Treatment Centre.
A genomics approach was used to sequence nucleic acids present in the samples and to map reads to microorganisms (not mapping to the human genome). Identification of transcripts to pathogens would be indicative of other microorganisms potentially present in the blood of patients with EVD, indicating translocation from the gut during the acute phase and reflecting underlying burdens of disease. The data indicated that the genomics pipeline successfully identified and quantified the amount of EBOV. The data suggested that the presence of Plasmodium had been underreported based on the antigen capture test and highlighted how a genomics approach could provide additional evidence of this parasite and connect it to an identified patient sample—particularly for retrospective analyses. The study found that most patients with acute illness on admittance to the Ebola Virus Treatment Centre already had extensive evidence in their blood of nucleic acid mapping to bacteria (the blood should have been effectively sterile), as well as nucleic acid mapping to viruses and parasites. The types of microbes identified argued strongly for bacterial translocation across the gut wall. These results have implications for the design of supportive care platforms, including empirical antimicrobial use.
This study investigated the presence of transcripts from viral, bacterial, fungal, and parasitic pathogens in patients infected with EBOV. The study made use of samples taken for routine diagnostic testing for EVD. This is the first systematic analysis of pathogens in patients infected with EBOV based on an unbiased deep sequencing approach. This approach did not rely on specific amplification of pathogens in selected growth media or enrichment of genomic sequences (e.g., by RT-PCR or PCR). Samples included those taken from patients acutely ill with laboratory-confirmed EVD infection prior to being admitted to the Ebola Virus Treatment Centre. Most of these patients were admitted around 6 days into the illness (based on when the patient thought they started feeling unwell) (Fig. 1
). This was still early in the disease course, but based on clinical studies the majority of patients would be predicted to have entered the gastrointestinal stage of illness (which occurs after 3 days [27
]), and therefore may also be predicted to have bacterial translocation from the gut to the blood. Blood samples taken from convalescent patients were used as a control to establish background levels of transcripts mapping to pathogens. As noted, the taking of blood from this group of patients would have been different from that from patients with acute EVD. This may have had an impact on which microbes were present, although we note that there was no one uniform microbial signature in the blood from patients with acute EVD. In the main, blood samples were used for the analysis presented herein, although some plasma sample results were also incorporated. There may be subtle differences in the transcriptome mapping to nucleic acids from pathogens in these samples, as sequence reads from blood would map to both extra- and intracellular pathogens, whereas with plasma, sequence reads would be predicted to map to extracellular pathogens.
The presence of other microorganisms can have a significant impact on the morbidity and mortality associated with primary infection. In a recent study, parasitemia (not clinical malaria) was positively associated with survival in patients with EVD (8
). This study used PCR-based approaches to identify the presence of P. falciparum
. Data from the analysis of samples gathered by the European Mobile Laboratory suggested that high parasite load (as determined by RDT, which likely indicates clinical malaria) corresponded with a poor outcome in children between 5 and 14 years old (12
). The deep sequencing analysis suggested a larger proportion of patients with EVD had sequence reads mapping to P. falciparum
than was recorded by use of RDT. P. falciparum
RDTs have a limited sensitivity and therefore mainly detect patients with higher levels of parasites in the blood, as found with acute malaria disease. Molecular techniques have shown a high background prevalence of Plasmodium
The effect of Plasmodium
on patient outcome with this data set varied according to the method by which Plasmodium
status was determined (Fig. 6
). Via the RDT, Plasmodium
appeared to impact patient outcome (P
= 0.053; analysis of variance [ANOVA] performed using software from SPSS, Inc.). However, this effect was eliminated once P. falciparum
status was confirmed using RNA sequencing data. This could be because with RNA-seq a larger proportion of individuals in the sample population tested positive for P. falciparum
: 84% (156 out of 186) compared to 33% (40 out of 121) by RDT. The discrepancy between RDT and RNA-seq data was reduced as the number of reads to Plasmodium
spp. increased. The RDT detects patients with high parasitemia and misses patients with low parasite levels. Thus, it is not surprising that a high Plasmodium
spp. load has an impact on patient outcome, whereas a low Plasmodium
load has a less pronounced effect or no impact. Individuals with a high Plasmodium
load are likely to have been sampled at the time they were experiencing the acute phase of the infection. Acute-phase infections are associated with strong systemic proinflammatory responses (30
), which are known to have antiviral effects. One hypothesis is that these responses may also act synergistically with virus-induced immunopathogenic responses. People with a low parasite load load are likely to harbor a chronic asymptomatic infection, which is characterized by mild or no inflammation and thus might be predicted not to interfere with Ebola virus infection. This was investigated by comparing the abundance of gene transcripts associated with the innate response stimulated by P. falciparum
in EBOV-positive patients who had the highest sequence read depth mapping to P. falciparum
and were also RDT positive, versus EBOV-positive patients who had no reads mapping to P. falciparum
and were also RDT negative. The analysis made use of matched transcriptomic data to investigate the host response to EBOV infection (4
). Overall, there was no significant difference in the abundance of gene transcripts associated with the innate response (Fig. 7
), but there were potential differences in the abundance of transcripts associated with blood coagulation (Fig. 8
). We note, however, that this analysis was based on a limited number of patients with matched host transcriptome data.
Nucleic acids from several bacterial species were identified in the blood and plasma samples from patients with EVD (Fig. 3
). Sequence reads mapping to multiple organisms were identified in the patient samples, for example, Prevotella
a genus of Gram-negative bacteria. Several studies have found that Prevotella
dominates the African gut microbiome (31
), giving credence to the hypothesis of the translocation of bacteria from the gut to the blood. However, across both hospitalized survivors and hospitalized fatalities, in a minority of patients there was no evidence of nucleic acids from bacteria, at least at the sequence read depths used in this study. Additionally, the approach did not provide antimicrobial sensitivity or clear identification of dominant causative organisms in a polymicrobial picture. The use of data generated by mapped reads on bacterial identification should be treated with some caution, due to the close phylogenetic relationships between nonpathogenic and pathogenic representatives of a species. However, many of the identified bacteria are associated with sepsis, and consideration should be given to appropriate antibiotic therapy. Clinical guidelines suggest empirical antibiotic treatment for patients with EVD who often present with nonspecific symptoms and signs that are difficult to distinguish from malaria, typhoid fever, or bacterial sepsis (33
). The rationale for empirical antibiotic therapy in EVD is 2-fold. First, it is used to provide broad-spectrum antibiotic cover for suspected EVD patient who have a range of infectious diseases, recognizing the lack of diagnostic capacity and resources in Ebola Treatment Centres. Second, it is to provide empirical Gram-negative/anaerobic antibiotic cover for patients with confirmed EVD who are at risk of secondary bacteremia, which is thought to originate from translocation across the bowel wall that is associated with the gastrointestinal clinical stage of EVD. Most patients showed evidence of nucleic acids from Pseudomonas
in their blood, and supportive antibiotic therapy should take into account the possible presence of this bacterial species.
Unsurprisingly, oral swab samples from community deaths indicated bacterial diversity and bacterial load and were included for completeness (Fig. 4
). This technique became routinely utilized in standard postmortem pathological sampling during this outbreak, with recognized limitations. The bacteria present in this sample set, and their growth, may be reflective of agonal processes or may represent growth postmortem. These samples included Wohlfahrtiimonas chitiniclastica
, which was likely deposited by flies postmortem, and Neisseria meningitidis
, a causative agent of meningitis, which was identified in both blood samples and swabs. Mass spectrometry also confirmed the identification of some of the species in throat swabs, thus validating the sequencing and bioinformatics approaches. Note that it was not possible to directly compare a throat swab with a blood sample from the same patient. Throat swabs were only used in the community deaths category, and only blood samples were taken when the patient presented at the treatment center. The species identified by mass spectrometry included bacteria and fungi. These microorganisms can be found as transitional flora in the upper respiratory tract. However, in case of a weakened immune system, these opportunistic pathogens may cause bacterial sepsis, pneumonia, or bacteremia. The successfully cultured bacteria are likely to constitute only a small part of the real bacterial flora, but many of the bacteria and fungi cannot be cultured due to the long storage time at −80°C and repeated freeze-thaw cycles. This illustrates the versatility of the sequencing pipeline that only requires RNA, returning large overlapping data sets and not requiring multiple steps at high containment.
Nucleic acids from several viruses were identified that correlated with the age of the patient, such as HHV4 (Epstein-Barr virus) (Fig. 9
). No evidence was found to suggest that, of the viruses potentially identified, there was a correlation with outcome or that any virus had a synergistic effect. One transcript corresponding to the envelope glycoprotein of human immunodeficiency virus 1 (HIV-1) was found (Table S1
), but no other HIV-1 transcripts were identified. The prevalence of adult HIV-1 in Guinea is 1.7% (2012 estimate). Therefore, in the patient cohort examined in this study, we would have predicted approximately 3 cases. However, if the virus was present in low copy number, then possibly higher sequence read depths would have been required to identify HIV-1. In addition, more specialist approaches are required to reliably identify HIV-1 and assembly genomes and rely on preamplification (34
). The evidence suggested potential activation of HHV4, and this could contribute to T cell dysfunction in these patients and, as noted, immune cell subsets may have different frequencies in different patients with EVD.
The deep sequencing and bioinformatics approaches showed that nucleic acids from viral, bacterial, fungal, and parasitic organisms were present in the blood taken from many patients. Many of the potentially identified bacteria are associated with sepsis and consideration should be given to treating patients with antibiotic therapy. While Plasmodium spp. can be detected by the applied rapid test under field conditions, the deep sequencing data suggest that this approach did not detect normal parasite carriage in the EBOV-infected patient population, and this may have affect on lethality calculations.
This is the first study to demonstrate that a potential range of pathogens are present in the blood of patients with EVD and provide evidence for existing treatment guidelines (antimicrobial therapy). The study provides a comprehensive open access data set for researchers to investigate a range of microbial factors linked to EBOV pathogenesis and outcome. Although we have learned much about the clinical syndrome of EVD and natural history of the disease, the pathogenesis of disease is still not understood. Much of the research focus focused on the 2013–2016 outbreak has been with regard to developing clinical trials for EBOV-specific therapeutics, but little evidence exists for different supportive care platforms, including empirical antimicrobial use. Deep sequencing of historical samples from patients can help fill this vital knowledge gap and illustrates research requirements for future EBOV outbreaks, mainly to build in prospective, preapproved protocols for EVD with serial sampling, matched clinical data, and correlation with other standardized microbiological techniques.
We acknowledge the help of Sophie Duraffour in the implementation of this study.
This work was carried out in the context of the project EVIDENT (Ebola Virus Disease: Correlates of Protection, Determinants of Outcome, and Clinical Management), which received funding from the European Union’s Horizon 2020 Research and Innovation Program under grant agreement 666100, and in the context of service contract IFS/2011/272-372 funded by the Directorate-General for International Cooperation and Development. The European Mobile Laboratory is a technical partner in the WHO Emerging and Dangerous Pathogens Laboratory Network (EDPLN) and the Global Outbreak Alert and Response Network (GOARN) and is headed by S.G.
The work was funded by the Centre for Defence Enterprise (to D.A.M. and J.A.H.) and by the U.S. Food and Drug Administration (to J.A.H. and M.W.C.) program for Ebola Virus Disease: Correlates of Protection, Determinants of Outcome and Clinical Management, grant HHSF223201510104C. The research was also funded by the National Institute for Health Research Health Protection Research Unit (NIHR HPRU) in Emerging and Zoonotic Infections at the University of Liverpool in partnership with Public Health England (PHE) and the Liverpool School of Tropical Medicine (LSTM), which directly supported N.Y.R. and X.L.
The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, the Department of Health or Public Health England.
S.H., N.Y.R, B.P., and I.G.-D. made equal contributions to this work.
M.W.C., A.D.C., S.G., R.W., K.S. D.A.M., and J.A.H conceived and designed the experiments. D.A.M., J.A.H. S.H., X.I., N.Y.R., I.G.-D., E.-G.Z., and G.B. performed the experiments. S.H., T.F., N.Y.R., G.B., K.S., A.D.C., L.F.P.N., L.R., D.A.M., and J.A.H. analyzed the data. All other authors were involved in sample gathering under the auspices of the European Mobile Laboratory outbreak response or data handling/processing. M.W.C., A.D.C., K.S., T.F., N.Y.R., D.A.M., and J.A.H. cowrote the paper, and all authors read and commented on the paper.