Blood Gene Expression Predicts Bronchiolitis Obliterans Syndrome

Abstract

Bronchiolitis obliterans syndrome (BOS), the main manifestation of chronic lung allograft dysfunction, leads to poor long-term survival after lung transplantation. Identifying predictors of BOS is essential to prevent the progression of dysfunction before irreversible damage occurs. By using a large set of 107 samples from lung recipients, we performed microarray gene expression profiling of whole blood to identify early biomarkers of BOS, including samples from 49 patients with stable function for at least 3 years, 32 samples collected at least 6 months before BOS diagnosis (prediction group), and 26 samples at or after BOS diagnosis (diagnosis group). An independent set from 25 lung recipients was used for validation by quantitative PCR (13 stables, 11 in the prediction group, and 8 in the diagnosis group). We identified 50 transcripts differentially expressed between stable and BOS recipients. Three genes, namely POU class 2 associating factor 1 (POU2AF1), T-cell leukemia/lymphoma protein 1A (TCL1A), and B cell lymphocyte kinase, were validated as predictive biomarkers of BOS more than 6 months before diagnosis, with areas under the curve of 0.83, 0.77, and 0.78 respectively. These genes allow stratification based on BOS risk (log-rank test p < 0.01) and are not associated with time posttransplantation. This is the first published large-scale gene expression analysis of blood after lung transplantation. The three-gene blood signature could provide clinicians with new tools to improve follow-up and adapt treatment of patients likely to develop BOS.

Keywords: biomarkers; blood; bronchiolitis obliterans syndrome; gene expression; lung transplantation.

Figure 1

Experimental design. (A) Collection modalities and sample classification. For bronchiolitis obliterans syndrome (BOS) patients, two classes of samples were defined depending on the time between blood collection and BOS diagnosis: a prediction class (PRED) combining blood samples collected at least 6 months before BOS diagnosis and a diagnosis class (DIAG) combining blood samples collected at or up to 13 months after BOS diagnosis. For STA patients, samples collected 6 and 12 months after LT were used. (B) Strategy for gene expression analysis. In both identification and validation sets, STA samples collected 6 months posttransplantation were compared with PRED and DIAG class samples. No patient duplicate was included in these groups. For the STA group, comparison between gene expression at 6 and 12 months posttransplantation was performed to exclude irrelevant genes modulated with time.

Figure 2

Differentially expressed genes. (A) Venn diagram of differentially expressed genes. Numbers of genes differentially expressed comparing STA vs. PRED, STA vs. DIAG, and STA at 6 and 12 months posttransplantation are indicated. (B) Principal component analysis and expression heat map (yellow and blue indicate high and low expression, respectively, and red and green indicate bronchiolitis obliterans syndrome and STA, respectively).

Figure 3

Independent validation. Microarray gene expression data (bar histograms) were validated by quantitative PCR in an independent set of patients (dot histograms) comparing STA and PRED (A) and STA and DIAG (B). Mann–Whitney p values are indicated.

Figure 4

Performance of POU2AF1, TCL1A, and B cell lymphocyte kinase (BLK) in prediction of bronchiolitis obliterans syndrome (BOS). (A) ROC curves for POU2AF1, TCL1A, and BLK for the prediction of BOS are displayed. (B) Discriminative characteristics of the three genes or the sum of expression of the three genes. (C) Kaplan–Meier analysis of BOS-free survival categorized by best expression thresholds of discrimination in ROC curves. Log-ranked p values are indicated.

Figure 5

B cell lymphocyte kinase (BLK), POU2AF1, and TCL1A expression is downregulated by terminal respiratory failure. Expression of BLK, POU2AF1, and TCL1A in blood from non-transplanted patients with end-stage chronic respiratory diseases, displaying either cystic fibrosis (CF), pulmonary hypertension (PAH) (A), or idiopathic pulmonary fibrosis (IPF) (B) compared to healthy volunteers (HVs) (29, 30). The p values from Kruskal–Wallis tests with Dunn’s ad hoc pairwise comparisons or Mann–Whitney tests are indicated.