Presented by Slim Fourati

​

REF: Fourati S. et al. (2018). Nat Commun 9(1):4418.

DOI: 10.1038/s41467-018-06735-8

​

A crowdsourced analysis to identify ab initio molecular signatures predictive of susceptibility to viral infection

Acute Upper Respiratory Tract Infection

• ~10 different viruses cause most
• Adults average 2-4 colds/year
• The most common cause of physician visits in the US
• Some numbers:
• Sore throats (pharyngitis) – 7 million visits/year in the US
• Acute sinusitis – 20 million cases/year in the US
• Acute bronchitis – 12 million cases/year in the US
• Influenza – 25 million seek health care annually

Why aren’t we sick all the time?

Host factors

• Prior immunity
• Pathogen load
• Comorbidities
• Intrinsic characteristics
• Genetic
• Epigenetic
• Transcriptomic
• Immunologic

Gene expression signatures predict which patients get sick starting 36 hours post-viral exposure

REF: Zaas AK et al. (2009) Cell Host Microbe 6(3):207-17. Fig. 5

REF: Huang Y et al. (2011) PLoS Genet 7(8):e1002234. Fig. 4

Can we identify early stage predictors of viral infection?

Human Viral Challenge Data

• Data from 7 viral challenge experiments in which healthy controls are exposed to 1 of 4 respiratory viruses (RSV, Rhinovirus, H1N1, H3N2)
• Patients are followed for 10 days with periodic blood draws for gene expression profiling
• Challenge focuses on gene expression timepoints prior to 36 hours post-exposure
• Infection endpoints
• Symptoms severity (study nose, scratchy throats, headache, cough, etc.)
• Viral shedding from nasal lavage samples
• Gene expression profiling on blood samples
• Periodic timepoints before and after viral exposure
• Affy Human Genome U133A 2.0 array and RNA-Seq

🤒

🤒

-24

0 12 24 36 48 60 72 96 120

😊

😊

😊

🤒

Transcriptomics

Respiratory Viral DREAM challenge workflow

*data includes previously published (3/4, accession number not provided to participants) and unpublished data (1/4)

Challenge data and outcomes by viruses

Longitudinal transcriptomic data

Static endpoints

*log symptom score: maximum symptom score over 10 days period after exposure

Log symptom score

Subchallenge 2: symptom onset

Subchallenge 3: log-symptom score

Subchallenge 1: viral shedding

What drive accuracy of models?

Features selected by predictive models

Heme metabolism and Severe COVID-19

Ref: Maestro S et al. (2021) Biomed Phamacother

The Arg134 residue in ORF3a may interact with the iron in heme and then kick off the iron from the heme through a redox reaction, generating dysfunctional hemoglobin

​

The dysfunction of hemoglobin leads to an altered cascade of several biochemical events, which includes low-level oxygenation, elevated free iron level, and down-regulation of heme oxygenase-1 (HO-1)

​

heme catabolite CO also suppresses the NF-κB signaling pathway and thereby impedes replication of virus

​

REF: Maiti BK (2020) ACS Pharmacol. Transl. Sci. 3, 1032–1034

Acknowledgments

Funding:

(Data generation) DARPA, ARO Proposal number 67037-L-DRP

(SF) NIAID U19 AI135972-03; PO: Reed S. Schabman