Gene expression profiling is rapidly becoming an essential tool for research and drug discovery and may soon play a central role in clinical diagnostics. This study profiles a large cohort of genes in human lung allografts by applying a high-sensitivity PCR approach to bronchoscopically acquired biopsy specimens. We identify several genes whose transcripts correlate strongly with histopathologic airway rejection/lymphocytic bronchitis. The data also identify genes overexpressed in EBB or TBB specimens, and others correlated between EBB and TBB samples from a given patient. Our work applies multiplex, real-time PCR to bronchoscopic biopsy specimens from lung transplant recipients. This approach has potential advantages of sensitivity and wide dynamic range, and therefore, can assess high-abundance and low-abundance transcripts in the same assay; reproducibility, as validated by the sub-set of genes, subject to little variation; and speed as PCR steps can be completed in 1 day.
The success in harvesting mRNA suitable for analysis is also notable. Once the extraction technique was refined, no sample was insufficient for analysis because of degradation or small size. This implies that we have not stretched the sensitivity limits of this assay. The most recent applications of this technique suggest that it can profile similar numbers of genes in a few cells obtained from tissue sections by laser capture microscopy (unpublished data).
Although the expression of about 33% of the transcripts was consistently low, this relates to low relative abundance rather than small sample size. In our samples, enough mRNA remained to assay more than 1,000 additional unique transcripts. These archived transcripts, indexed for pathologic grade in parallel EBB and TBB specimens, are a resource for future profiling experiments. The current study identifies a sub-set of genes of potential diagnostic utility. Screening of further cohorts of genes may identify additional informative transcripts. The ultimate goal is to winnow the list of candidate genes to a maximally informative sub-set. Our experience with these experiments leads us to predict that a panel of transcripts, rather than the product of a single gene, will be more useful.
Adaptation of this technique to a 384-well format raises the possibility that such an assay can be used in high-throughput fashion in a clinical lab setting with samples from many patients. One potential drawback is the high initial cost of TaqMan-type probes and primers for each gene included in the analysis; although once synthesized, these probe/primer sets can be used for many assays. Another limitation, compared with microarray approaches, is the number of genes that can be analyzed simultaneously (practically, around 200). However, this is offset by much higher sensitivity and dynamic range compared with current, hybridization-based microarrays, which require larger amounts (1–10 μg) of starting RNA and therefore are not applied easily to small biopsy samples.
Although our results identify 6 transcripts that correlate strongly with pathologic grade in EBB specimens, they also reveal variability in transcript levels relative to histopathology. This is reflected in the modest sensitivity and specificity of individual transcript levels compared with the gold standard of histopathologic grading of biopsy specimens obtained in parallel at the same sitting.
Even if levels of these particular transcripts perfectly reflected the degree of lymphocytic bronchitis in a given sample, there would, however, remain limits to sensitivity and specificity in relation to the histopathology of parallel biopsy specimens because of patchy distribution of disease. We know that patchiness exists from the variation in pathologic grade in separate biopsy samples obtained from the same patient during the same bronchoscopy. This is also implied by the low yield of bronchoscopic biopsy specimens for diagnosing OB, which can produce severe functional deficits despite overt involvement of a small fraction of airways.18
Additionally, OB may show scarring without inflammation in its inactive phase.
Thus, sensitivity and specificity likely would improve if we had a better gold standard, though presently no suitable alternative exists.31
A more practical strategy for improving sensitivity and specificity of transcriptional profiling is to increase the number of samples—perhaps pooling them to obtain a broader representation of a potentially patchy process.
A comparison of transcription profiles of EBB and TBB specimens in the same patient, as shown in and , identifies transcripts strongly overexpressed in one type of biopsy specimen compared with the other. This suggests the possibility of deriving a sub-set of genes characteristic of airway mucosa and alveolar parenchyma, respectively. Such a sub-set would be useful in estimating airway and alveolar contributions in potentially mixed samples, as in TBB, in which a sample can range from mostly airway to mostly lung parenchyma.
Of the genes surveyed, the most consistently (p < 0.001) overexpressed in EBB relative to TBB specimens are mucin genes MUC4 and MUC5AC (but not MUC1). Interestingly, neither MUC4 nor MUC5AC correlates with rejection, suggesting that they are markers of airways but not necessarily of inflammation or disease. Expression of these mucin genes in EBB specimens is so disproportionate that in the few paired biopsy samples in which EBB and TBB expression is approximately equal (see ) it is likely that the TBB contained mainly airway rather than alveolar material.
Among the larger set of genes overexpressed in TBB relative to EBB specimens, the most extreme examples are ALOX5 and ICAM1. Because neither of these genes correlates with pathologic grade and both are expressed 12-fold higher in TBB than EBB samples, they may reflect alveolar content of a given sample independent of pathology. Other genes overexpressed in TBB specimens are MMP9 and TIMP2 (). In the case of MMP9, disproportionate expression in TBB samples may be due to alveolar macrophages, which are a source of MMP9 and also of ALOX5.32
On the other hand, the most likely sources of differential alveolar expression of ICAM1 are vascular endothelium and type I epithelium.33
A comparison of paired EBB and TBB samples also reveals several transcripts with very high linear correlations in the same patient ( and ). Interestingly, none of the transcripts that correlate well (r
> 0.8, p
< 0.001) in EBB/TBB pairs correlate with pathologic grade. CTGF is a prominent example featuring a wide range of expression, the basis of which is intriguing because of the correlation of CTGF with lung fibrosis34
and, potentially, with OB. The variation in CTGF transcription could be inherited or acquired and be due to differences unrelated to the accumulation of mononuclear cells in bronchial and alveolar tissues.
In this regard, the lack of correspondence between high-grade bronchial inflammation and acute alveolar rejection in the same patient and bronchoscopy () is also interesting. This suggests that the processes are at least partly independent and may reflect a different pathogenesis. It is also consistent with the recent appreciation of sub-types of acute allograft rejection featuring differences in immune activation and cellular proliferation indistinguishable by light microscopy,35
recognition of a chronic onset form of BOS not strongly linked to acute rejection,36
and involvement of innate as well as adaptive immune mechanisms in rejection.37
These data not only shed light on the feasibility and diagnostic implications of measuring multiple transcripts in bronchoscopic biopsies but also provide clues about the mechanism. The finding that type I collagen, MMP, and fibronectin transcripts increase with increasing grades of lymphocytic bronchitis is especially meaningful because these gene products are associated with tissue remodeling and fibrosis, as in OB. It is perhaps significant that while airway MMP transcripts rise, those encoding natural MMP-inhibiting TIMPs do not. This may produce a protease–anti-protease imbalance in the airway wall.
On the other hand, TIMP levels are high in absolute terms—higher than many housekeeping genes. If mRNA levels mirror TIMP and MMP levels in these samples, there may be an excess of TIMP even after induction of MMP expression. Nonetheless, some MMPs can be activated even while bound to TIMP38
so that the TIMP/MMP ratio may be less important than the amount of MMP produced and the presence of an activating mechanism.
The nature and importance of protease/anti-protease imbalance in the evolution of remodeling and fibrosis in the airways of transplanted lungs is not well understood and is likely to be complex. Although the inhibition of some MMPs may favor the accumulation of collagen and other matrix proteins, it is also likely that active proteases are required for fibrosis to develop. This is because when one cell type replaces another (such as fibroblasts for epithelium) in remodeling tissues, or when new cells migrate into areas previously occupied by extracellular matrix, existing matrix proteins must be broken down. Thus, the action of MMPs and other proteases likely are essential for the evolution of fibrotic airway lesions, but the process may be altered by imbalances or dysregulation at critical phases.
Another transcript that increases greatly with higher grades of bronchitis is DEFB2. Defensins are a family of cationic peptides with broad-spectrum anti-microbial activity. DEFB2 is expressed in epithelia of many organs, including airway, where it is found in surface epithelia and serous cells of sub-mucosa glands.29
Moreover, β-defensins may be chemotactic for dendritic and T cells.39
Levels of SCYA18, a CC-class lymphocyte-selective chemokine preferentially expressed in lung,30
also increase in high-grade bronchitis, although transcript levels are lower overall than those of other positively correlated genes. Because of their chemotactic effects, DEFB2 and SCYA18 induced in the allograft airway could augment rejection responses.
In summary, this work demonstrates a method for measuring large cohorts of gene transcripts in small, bronchoscopically acquired EBB and TBB specimens. The results provide a view of molecular events that accompany allograft inflammation and identify specific bronchitis-associated transcripts, which may be useful in diagnosing rejection events in the airways. These findings help explain mechanisms of acute rejection and OB, the major obstacles to long-term survival in lung transplant recipients.