Background: Acute lung rejection is a risk factor for chronic rejection, which jeopardizes long-term recipient survival. Presently, acute rejection is diagnosed with the use of transbronchial lung biopsies, which are invasive, expensive, and subject to sampling error. We seek to improve acute rejection diagnostics by identifying genes whose expression in bronchoalveolar lavage (BAL) cells best classifies acute rejection versus no rejection.
Methods: BAL samples were analyzed from 32 subjects whose concurrent histology showed acute rejection (n=14) or no rejection (n=18). Gene expression was measured with Affymetrix microarrays. Quantitative real-time polymerase chain reaction confirmed the microarray results for selected genes. The nearest shrunken centroid method with 10-fold cross validation defined the classification model. A total of 250 iterations of the algorithm were performed to determine the misclassification error rate and the most influential genes in determining classifiers.
Results: The estimated overall misclassification rate was <20%. Seven transcripts were present in every classifier, and 52 transcripts were present in >70% of classifiers; these transcripts are related to T-cell function, cytotoxic CD8 activity, and granulocyte degranulation. Eleven of the 52 genes were analyzed with quantitative real-time polymerase chain reaction; all were found to significantly different between the groups, with 10 of 11 increased in acute rejection samples. The proportions of lymphocytes and neutrophils in BAL samples increased in acute rejection but did not outperform the gene-based classifier.
Conclusions: There is a prominent acute rejection-associated signature in BAL cells characterized by increased T-cell, CD8 cytotoxic cell, and neutrophil gene expression. These findings lay the foundation for development of rapid PCR-based assays of gene expression for clinical acute rejection diagnosis.