A Four-Biomarker Blood Signature Discriminates Systemic Inflammation Due to Viral Infection Versus Other Etiologies

The innate immune system of humans and other mammals responds to pathogen-associated molecular patterns (PAMPs) that are conserved across broad classes of infectious agents such as bacteria and viruses. We hypothesized that a blood-based transcriptional signature could be discovered indicating a host systemic response to viral infection. Previous work identified host transcriptional signatures to individual viruses including influenza, respiratory syncytial virus and dengue, but the generality of these signatures across all viral infection types has not been established. Based on 44 publicly available datasets and two clinical studies of our own design, we discovered and validated a four-gene expression signature in whole blood, indicative of a general host systemic response to many types of viral infection. The signature’s genes are: Interferon Stimulated Gene 15 (ISG15), Interleukin 16 (IL16), 2′,5′-Oligoadenylate Synthetase Like (OASL), and Adhesion G Protein Coupled Receptor E5 (ADGRE5). In each of 13 validation datasets encompassing human, macaque, chimpanzee, pig, mouse, rat and all seven Baltimore virus classification groups, the signature provides statistically significant (p < 0.05) discrimination between viral and non-viral conditions. The signature may have clinical utility for differentiating host systemic inflammation (SI) due to viral versus bacterial or non-infectious causes.


Item Title
Supplementary Figure   Resolution of patients with acute respiratory illness (ARI) into three clusters corresponding to bacterial infection, viral infection, and non-infectious illness (GSE63990).

Supplementary Figure S1 | Pan-viral signature scores for children with and without detectable human rhinovirus (Baltimore Group IV)
Human rhinoviruses (HRV, Baltimore Group IV, positive-sense single-stranded RNA) are the most frequent etiological agents of respiratory tract infections in both adults and children (Heinonen et al., 2016). Children infected with HRV display a broad spectrum of clinical signs, and questions have been raised over the etiological role of HRV in respiratory disease.
Dataset GSE67059 was generated in a study (Heinonen et al., 2016) that identified a whole blood gene expression signature for differentiating pediatric subjects with or without detectable human rhinovirus (HRV). The study included respiratory symptomatic and asymptomatic subjects as either inpatients or outpatients. Asymptomatic subjects included healthy children, and also those presenting for minor surgery that did not involve the respiratory system. GSE67059 comprised one of our validation datasets, and was not used in discovery of the pan-viral signature.
We observed differences in the pan-viral signature when comparing subjects that were HRV(-) to subjects that were HRV(+), irrespective of whether the infection-positive subjects had symptoms or were inpatients or outpatients (Supplementary Figure S1).  Respir. Crit. Care Med. 193, 772-782 (2016).

infection with H3N2 influenza A (Baltimore Group V)
Dataset GSE30550 (Huang et al., 2011) was used in the discovery of the pan-viral signature. The following samples from GSE30550 were used in the discovery process: infection-negative samples taken pre-inoculation (0 hours); and infection-positive samples taken at times corresponding to peak symptom presentation (21-69 hours). Supplementary Figure S2 presents a box-and-whisker plot of the the pan-viral signature score for symptomatic human subjects, sampled over a 108 hour period following intra-nasal inoculation of influenza H3N2. A significant increase in pan-viral signature score relative to its pre-inoculation value can be seen as early as 36 hours, and reaches a maximum at 69 hours.
infection with H3N2 influenza A (Baltimore Group V). Box-and-whisker plot of the pan-viral signature score for symptomatic human subjects, based on sampling over a 108-hour period following intra-nasal inoculation of influenza H3N2. The time-course was visualized by fitting the observed points to a third-order polynomial.
A parallel time-course for earliest detection and peak score in mice (Supplementary Figure S3) in response to experimental inoculation with influenza H3N2 suggests that this virus produces a similar pan-viral signature response in the two species.

infection with H3N2 influenza A (Baltimore Group V)
GSE57384 is a validation dataset that was not used in discovery of the pan-viral signature. In the study represented by GSE57384 (Qiu et al., 2015), mice were intra-nasally inoculated with influenza A virus (H3N2) and followed over time. Three mice were sacrificed at 9 days prior to inoculation as a control, and then daily from day 0 to day 10 post-inoculation. As shown in Supplementary Figure S3, the pan-viral signature first showed a consistent increase on day 2, and peaked on day 3. It then decreased to pre-inoculation levels by day 9. A similar response curve was observed in experimental human inoculation with the same virus strain (Supplementary Figure S2). Note: OASL1 is the mouse ortholog of the human OASL gene (Elkhateeb et al., 2016) and was used here.

Supplementary Figure S4 | Pan-viral signature score over time, for liver biopsy tissue from chimpanzees inoculated with Hepatitis C Virus or Hepatitis E Virus (Baltimore Group IV)
To date, the chimpanzee is the only animal model for Hepatitis C (HCV) and Hepatitis E (HEV) infection in humans. Dataset GSE22160, used in validation of the pan-viral signature, was generated in a study of the temporal gene expression responses in liver biopsies from chimpanzees inoculated with either HCV or HEV (Yu et al., 2014). Supplementary Figure S4 presents the pan-viral signature in the chimpanzee liver biopsy samples, as a function of time post-exposure, which correlated with the first detection of viral antigen in regional lymph nodes, but preceded the first detectable viremia (Day 4) or elevated body temperature (Day 5) as detailed in Table 3

Supplementary Text S1 | Description of the FEVER study
Study Identifiers: • Title: Expression profiling in patients presenting to hospital with fever.
• A consort diagram for the FEVER study is given in Supplementary Figure S6.

Supplementary Text S2 | Description of the GAPPSS study
Study Identifiers: • Title: Genotypes and Phenotypes in Pediatric SIRS and Sepsis (GAPPSS).
• Signed consent from English-speaking parents.

Control Group
• Inclusion criteria: Admitted to intensive care, 1-18 years of age, at least 2 SIRS criteria, in-dwelling vascular catheter, not previously enrolled in the study, > 10kg, open heart surgery requiring bypass.
• Exclusion criteria: Not expected to survive, pre-or post-operative positive microbiology, malignancy or immune disorder, corticosteroid use (pre-, post-or intra-operative).

Sepsis Group
• Inclusion criteria: Admitted to intensive care, 38 weeks Estimated Gestational Age -18 years of age, at least 2 SIRS criteria and one organ dysfunction, in-dwelling vascular catheter, not previously enrolled in the study, strongly suspected or documented bacterial infection, > 4kg, severe respiratory dysfunction requiring invasive or non-invasive positive pressure mechanical ventilation.
• Exclusion criteria: Not expected to survive, PICU nosocomial infection, ward of the state, corticosteroid use (pre-, post-or intra-operative).

Viral Group
• Inclusion criteria: Admitted to intensive care, 38 weeks Estimated Gestational Age -18 years of age, in-dwelling vascular catheter, not previously enrolled in the study, Positive PCR verifying a viral infection, > 4kg, severe respiratory dysfunction requiring invasive or non-invasive positive pressure mechanical ventilation.
Clinical Workup: • Specimens collected: blood and nasal swab for clinical microbiology and virology.