Published ahead of print on September 4, 2003, doi:10.1164/rccm.200305-644OC
© 2003 American Thoracic Society Gene Expression Profiling of Bronchoalveolar Lavage Cells in Acute Lung RejectionDepartment of Medicine and Institute of Human Genetics, University of Minnesota, Minneapolis, Minnesota Correspondence and requests for reprints should be addressed to Marshall I. Hertz, M.D., University of Minnesota, 420 Delaware St. SE, MMC 276, Minneapolis, MN 55405. E-mail: hertz001{at}umn.edu
Lung transplantation is effective for many diseases that are unresponsive to other therapy. However, long-term survival of recipients is limited by the development of bronchiolitis obliterans syndrome. Acute rejection is a major risk factor for bronchiolitis obliterans syndrome, but noninvasive biomarkers have not been identified. To address this deficiency, gene expression microarrays were performed using bronchoalveolar lavage cells of lung transplant recipients with acute rejection (n = 7) and with no rejection (n = 27). The cell and differential counts were similar. Signal values for genes between groups were compared using t tests. One hundred thirty-five genes were upregulated in the acute-rejection group, including genes involved in acute rejection, immune response genes with an unknown role in rejection, genes not known to have a role in rejection, and genes of unknown function. Two-dimensional hierarchical clustering grouped all acute rejection samples into one cluster and the majority of the no-rejection samples into a second cluster. The acute-rejection samples showed significant changes in gene expression for seven biological pathways. Bronchoalveolar lavage cells are a reliable RNA source for microarray analysis, which is powerful in identifying acute-rejection genes. The individual genes, patterns of gene expression, or biologic pathways identified may represent novel biomarkers for acute rejection.
Key Words: lung transplantation • allograft rejection • microarray Lung and heart–lung transplantations are effective treatment modalities for many diseases that are not responsive to other medical and surgical therapies (1, 2). However, long-term survival of recipients is limited by the development of bronchiolitis obliterans syndrome. Acute rejection of the lung allograft occurs in 30–50% of lung transplant recipients and is the leading risk factor for the development of bronchiolitis obliterans syndrome (3–7). Although many lung transplant recipients with acute rejection suffer from dyspnea and demonstrate a decline in pulmonary function, these findings are nonspecific. In addition, the diagnosis requires invasive lung biopsies, and the treatment of acute rejection demands escalation of immunosuppression, both of which can cause additional morbidity and mortality. Therefore, identification of reliable, relatively noninvasive biomarkers of acute rejections would constitute an important advance in the management of this group of patients. Acute rejection is characterized histologically by perivascular and peribronchial lymphocytic inflammation. The cellular infiltrate is primarily comprised of lymphocytes, and the inflammation extends to the alveolar wall. As in other transplanted organs, acute lung rejection is mediated primarily by a CD8 T-lymphocyte cytotoxic response initiated via recognition of graft alloantigens by CD4 and CD8 T lymphocytes. The response involves a complex cascade of cytokines and markers of activated lymphocytes, some of whose genes and gene products have been identified in bronchoalveolar lavage (BAL) fluid and cells during acute rejection. Although these prior investigations have been extremely useful in identifying markers of acute rejection, they have limitations: (1) The number of genes studied is limited. (2) The genes studied are participants in a limited number of cellular processes. (3) They exclude analysis of genes not already known or expected to be associated with graft rejection. The application of large-scale microarray-based technology can overcome these limitations. Although there have been several studies of microarray analysis of allograft rejection, none specifically address lung transplantation (8–23). Therefore, in this report, we present the results of gene expression profiling of BAL cells of lung transplant recipients with acute rejection with the aim of identifying biomarkers, which will enable improved prevention, prediction, and treatment of lung allograft rejection.
Patients and Samples Lung transplant recipients surviving 30 days after transplantation at the University of Minnesota were eligible for enrollment in the study, which was approved by the university's institutional review board. All patients signed informed consent. Patients underwent scheduled surveillance bronchoscopy (consisting of BAL and transbronchial biopsies) to identify rejection or infection at 1, 2, 4, 6, 8, and 12 months after transplantation and when clinically indicated. Each bronchoscopy consisted of BAL (100–140 cc of normal saline instilled in one subsegmental location) and transbronchial biopsies (8–12 samples total from two subsegmental locations within the same lung). After sending BAL fluid for clinical tests, approximately 50 ml of BAL fluid was immediately placed on ice. Cells were separated by centrifugation and washed twice with phosphate-buffered saline. BAL cell counts and differentials were determined with a hemocytometer. Patients with bronchopulmonary infections (as determined by history, exam, chest radiograph, and the results of routine laboratory tests and cultures) were excluded from analysis.
Microarray Preparation
Data Analysis Signal values (expression levels) for transcript abundance were calculated using Microarray Suite 5.0 (Affymetrix, Santa Clara, CA). Chip quality control and data normalization were done using the Refiner Module of Gene Expressionist 4.0 (GeneData, Basel, Switzerland); t tests were used to compare signal values for the two groups using the data-mining tool SAM (Significance Analysis of Microarrays) (25). Genes that differed significantly between the rejection and no-rejection groups and the corresponding statistical values were reported. The relationships of significant transcripts with the samples groups were determined with two-dimensional hierarchical clustering and were visualized using Treeview (26). Further analyses of significant genes across biologic pathways were investigated with the use of GenMAPP (27), which identifies pathways as significant based on a Z score of more than 2.
Analysis of Variability in Acquisition and Preparation of BAL Specimens To establish the feasibility and reliability of using gene expression microarrays to analyze BAL cells, we investigated how results are affected by potential sources of variability in microarray preparation. To evaluate variability in microarray preparation, a single BAL sample from one patient was equally divided at the time of acquisition, and the two aliquots were processed identically as described in METHODS. The microarray signal value data from the two aliquots showed a very high correlation with each other (see Figure E1 in the online supplement). BAL sampling location might also affect microarray results: During a typical bronchoscopy, the BAL cells are obtained from one subsegment of one lobe of one lung. It is assumed that this results in a cell sample that is representative of the pathophysiologic processes occurring in the entire transplanted lung (or both lungs in the case of bilateral single lung transplant recipients). To evaluate this, we obtained separate BAL samples from the right middle lobe and lingula of a bilateral single-lung transplant recipient in good health (no evidence of infection or rejection). The samples were collected and processed separately and in parallel. The microarray signal value data from the paired samples showed a very high correlation (see Figure E2 in the online supplement), demonstrating that BAL cell samples taken from different sites within the lungs of a transplant recipient without focal disease have nearly identical gene expression patterns.
Patients and Samples
Analysis of Gene Expression Based on Rejection Status SAM was used to calculate the false discovery rate (specificity) for varying degrees of statistical stringency determined by a tuning parameter termed the  value (25). Each gene was assigned a q value, which is like the familiar p value adapted to the analysis of a large number of genes; genes with a q value of less than 0.05 were considered significant (25).
We compared the signal values of each transcript represented in the microarray between the acute-rejection and no-rejection groups. To do so, we first examined the relationship between the
The gene expression data corresponding to the value of 1.50 were used for further analysis, as it was correlated with a high specificity. As shown in Table 2, this value corresponded to a list of 135 genes at a false discovery rate of 0.94%; this suggested that at most, approximately one of the included genes might represent a false-positive result. Using a higher value (increased statistical stringency) did not reduce the false discovery rate appreciably. Alternatively, decreasing statistical stringency by using a lower value added more candidate genes but with the associated specificity was lower.
A sampling of the immune response genes that demonstrated significant changes in expression in the acute rejection group as compared with the no-rejection group corresponding to the
Cluster Analysis of Significant Genes and Relationship to Rejection Status Cluster analysis was performed for all samples based on the 135 significant genes (Figure 2 and Table E1 in the online supplement). The clustering algorithm separated the samples into two major groups: One group included all of the acute-rejection samples, and the other group included the majority (21 of 27) of the no-rejection samples; six no-rejection samples clustered with the acute-rejection samples.
The five subjects represented by more than one BAL sample afforded the opportunity to compare the importance of acute-rejection status in sorting samples obtained from specific individuals. Patients A and B each contributed one acute-rejection sample and one no-rejection sample; each of these samples clustered with their appropriate groups according to rejection status. Patient C contributed two acute-rejection samples, each of which clustered with other acute-rejection samples. Conversely, patient D had three no-rejection samples that all clustered within the no-rejection group. Patient E contributed two no-rejection samples; one grouped with the majority of no-rejection samples, but one clustered with the acute-rejection samples.
Relationship of Significant Genes in Process Pathways: GenMAPP
We have shown that microarray analysis of BAL cells of lung transplant recipients is a reliable, reproducible, and powerful tool for the identification of genes and gene expression patterns indicative of acute rejection of the lung allograft. Adequate amounts of high-quality RNA can be extracted from BAL cells obtained as part of a routine bronchoscopy examination, and highly reproducible microarray results can be obtained with this material. The comparison of acute-rejection and no-rejection samples showed no significant difference in the BAL cell counts or cell differentials. As expected, the preponderance of cells was macrophages. BAL cell counts and differential counts are generally accepted to have little diagnostic value for acute lung rejection (28), and these findings are consistent with this. Despite this, the difference in gene expression between the rejection and no-rejection groups suggests that the rejection samples contain increased numbers of functionally activated lymphocytes. Therefore, although lymphocytes comprise a relatively small fraction of the total BAL cells, their pronounced changes in gene expression did not go undetected using these methods. Therefore, fractionation of the cells before RNA extraction may not be necessary to detect significant gene expression changes indicative of acute rejection.
Using statistical methods to analyze microarray-based gene expression patterns is an evolving science. In particular, choosing the appropriate threshold for determining statistical significance is, in part, dependent on the purpose of the analysis. Using the SAM method of analysis, we chose the gene expression profiles associated with a
Among the genes identified in this study known to be mediators in the acute-rejection response were T-cell receptors ( Cluster analysis reliably grouped all acute-rejection samples and the majority of the no-rejection samples. Therefore, the patterns of gene expression displayed in Figure 2 may be relatively unique for rejection status. Clustering based on these gene expression patterns was reproducible whether the samples were from different patients or from the same patient at different times. The reproducibility of these gene expression patterns will have to be confirmed prospectively with a validation set of future samples. There were six no-rejection samples that clustered with the acute rejection group. To eliminate the possibility that the biopsy scores were in error, the biopsy grades of these six specimens (along with six other no rejection samples to eliminate bias) were verified after blind reassessment by our lung pathologist. Aside from biopsy scores, these samples shared no significant differences or similarities with the other samples with respect to bronchoscopy results (culture, virology, cell counts, and differential counts) or other clinical factors (immunosuppressive medications, time since transplantation, and reason for transplantation). The potential explanations for these misclassifications are several. Although the algorithm clustered these no-rejection samples with the acute-rejection group, the gene expression profiles of the no-rejection samples in question are not entirely similar to those of the acute-rejection samples; there are similarities, but across a limited number of genes (Figure 2). Also, the gene expression profiles of the no-rejection samples in question are not entirely dissimilar to the other group of no-rejection samples (Figure 2). Therefore, although these six no-rejection samples clustered with the acute-rejection samples, their gene expression patterns share some similarity with the other no-rejection samples. Also, there is sampling error inherent to the technique of transbronchial biopsies for the detection of lung rejection. For example, one patient had a decline in her pulmonary function tests at the time of her no-rejection biopsy; the decline in the pulmonary function tests persisted, and acute rejection was diagnosed on a subsequent biopsy. It is also possible that this pattern of gene expression persists in patients who have had prior episodes of acute rejection. One patient had acute rejection (A0B2) 1 month before, and another had four episodes of acute rejection over the previous 11 months. This pattern of gene expression may represent some immune dysregulation that represents increased future risk of either rejection or infection. One patient acquired pneumonia because of respiratory syncytial virus within 1 month; another developed cytomegalovirus pneumonitis and an episode of acute rejection within 4 months. Two more patients developed persistent, asymptomatic shedding of cytomegalovirus. Finally, these patients with misclassified samples and the patients with the acute-rejection samples may have clinical resemblances that are not yet manifest. For example, these six patients and the seven with acute rejection may be at increased risk for bronchiolitis obliterans syndrome; therefore, these gene expression patterns could actually indicate future risk of this outcome. These patients will be followed prospectively to determine whether this is indeed true. The finding of the inflammatory response pathway (Figure E3B) and the multiple chemokine receptors identified in the peptide G-protein–coupled receptor pathway (Figure E3E) is not unexpected, as this is consistent with the pathogenesis of acute rejection. Of particular importance is the finding of upregulation of the TGF-ß signaling pathway (Figure E3A), as TGF-ß expression is increased in patients with bronchiolitis obliterans syndrome even before the onset of airflow obstruction (37, 38). TGF-ß itself was not noted to be upregulated in the sample set of genes chosen for this analysis, but many downstream mediators, including the TGF-ß receptor, were upregulated (Table E2 and Figure E3A). The association of apoptosis with allograft rejection has been demonstrated with several organs, including the lung (Figure E3C) (30, 31, 36, 39, 40). Specifically granzyme B and perforin have been identified as important mediators, and both were upregulated in the acute-rejection group (Figure E3C) (30, 31). The Wnt signaling pathway is of potentially great interest for several reasons. During development, Wnts have diverse roles in governing cell fate, proliferation, migration, polarity, and death; in adults, Wnts function in homeostasis, and inappropriate activation of the Wnt pathway is implicated in a variety of cancer (41). Interestingly, Wnt5a, which was notably upregulated in our acute rejection specimens, has specifically been found to play a part in lung development (41). The significance of this complex pathway in allograft rejection has yet to be determined. All of these findings, taken together, suggest that gene expression microarrays of BAL cells yield information for the both the clinical and research progress of acute lung rejection. We have shown that there exists very little variation inherent in the techniques required for producing microarray results. There is a very high correlation between gene expression patterns in BAL cells obtained from both lungs of a lung transplant recipient at a given time. These quality control studies support the fact that gene expression microarray analysis can be used to produce reliable and reproducible results from BAL cells. Microarray analysis is a powerful tool to identify candidate genes involved in acute rejection of the lung allograft. At a high level of statistical stringency, we have found 135 genes (of known and unknown functional importance) significantly upregulated in those patients with acute rejection. These genes comprised a distinct pattern of gene expression for acute rejection of the lung allograft. Components of biologic pathways for the inflammatory response, TGF-ß signaling, G-protein–coupled receptors, nucleotide G-protein–coupled receptors, apoptosis, cytokine–CXC chemokines, and Wnt signaling were significantly upregulated in patients with acute rejection, and represent potential areas of additional investigation. These patterns of gene expression, gene pathways, or the genes themselves may represent novel biomarkers that may lead to advances in the prediction, noninvasive diagnosis, and treatment of lung allograft rejection.
Supported by National Institutes of Health PPG 5PO1-AI50162-02 and by a grant from the Minnesota Medical Foundation. This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Conflict of Interest Statement: V.J.G. has no declared conflict of interest; J.D.L. has a managed stock portfolio that includes health care and pharmaceutical stocks; T.R.B. has no declared conflict of interest; R.A.K. has no declared conflict of interest; M.I.H. has no declared conflict of interest. Received in original form May 13, 2003; accepted in final form September 1, 2003
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
and ß), granzymes (
(
