1317. Transcriptional Profiling to Discriminate Bacterial vs Viral Lower Respiratory Tract Infections (LRTI) in Hospitalized Adults
Session: Poster Abstract Session: Biomarkers and Correlates of Protection
Saturday, October 5, 2013
Room: The Moscone Center: Poster Hall C
Background: Differentiation between bacterial and viral LRTI remains a significant clinical challenge. The present study was designed to explore the value of transcriptional profiling as a novel diagnostic tool for LRTI.

Methods: We collected whole blood samples for RNA microarray analysis in 118 patients, median [IQR] age: 59 [50-76] yrs hospitalized with LRTI including: 22 bacterial (B), 71 viral (V) and 25 B-V coinfections (Co) and 58 age-matched healthy controls (HC). Comprehensive microbiologic diagnostic testing included: blood and sputum cultures, PCR and DFA in respiratory samples, urine antigen and serology for both B and V agents. We used class comparisons (Mann-Whitney p<0.01; Benjamini MTC) and modular gene analysis to identify distinct biosignatures for B, V and Co. We performed class prediction  (k-NN algorithm) to identify classifier genes that best discriminate B vs V infections.

Results: We analyzed 118 patients with LRTI: 71 were V (32 flu-A, 9 flu-B, 17 RSV, 7 HMPV and 6 V-V coinfection), 22 were B (13 S. pneumoniae, 2 S. aureus, 4 M. catarrhalis and 3 B-B coinfection) and 25 Co. Class comparisons identified 3,376 differentially expressed genes in B vs HC; 2,391 in V vs HC, and 2,628 in Co vs HC. Of those genes, 1,222 were shared among all infections vs HC, and 2,221 genes were specific of each comparison (1,264 B vs HC; 651 V vs HC; 306 Co vs HC). Modular analysis revealed common patterns in the 3 LRTI groups including overexpression of innate immunity, apoptosis and antimicrobial response genes and underexpression of T, B and NK cells, mitochondrial metabolism, lymphoid lineage and protein synthesis genes compared with HC. However, we also identified significant differences among the groups including: Interferon (V>Co>B), inflammation (B>Co>V), neutrophils (B>Co>V); Co had significant underexpression of monocyte and platelets genes compared with B and V. KNN algorithm identified 15 classifier genes that discriminated B vs V LRTI patients with a sensitivity of 95% [95%CI: 77-99] and specificity of 91% [73-98].

Conclusion: Hospitalized LRTI patients with identified B, V, and Co exhibit distinct gene expression patterns. Furthermore, bacterial and viral LRTI can be differentiated at the transcriptional level.

Nicolas M. Suarez, PhD, Center for Vaccines and Immunity, The Research Institute At Nationwide Children's Hospital, Columbus, OH, Eleonora Bunsow, Nationwide Children's Hospital, Columbus, OH, Edward Walsh, MD, Rochester General Hospital, Rochester, NY, Ann Falsey, MD, University of Rochester, Rochester, NY, Asuncion Mejias, MD, PhD, Center for Vaccines & Immunity, The Research Institute at Nationwide Children's Hospital, Columbus, OH and Octavio Ramilo, MD, Pediatrics, Nationwide, Columbus, OH


N. M. Suarez, None

E. Bunsow, None

E. Walsh, None

A. Falsey, None

A. Mejias, None

O. Ramilo, Abbott Molecular: Grant Investigator, Grant recipient
Abbott Labs: Scientific Advisor and Speaker's Bureau, Consulting fee and Speaker honorarium
Merck: Consultant, Consulting fee
Gilead: Scientific Advisor, Consulting fee
Quidel: Scientific Advisor, Consulting fee
Roche: Consultant, Consulting fee

Findings in the abstracts are embargoed until 12:01 a.m. PST, Oct. 2nd with the exception of research findings presented at the IDWeek press conferences.