Observations in previous studies indicate that DLFs are morphologically and functionally distinct from AFs (1
). The present study expands on these earlier observations by identifying global differences in gene expression, and by evaluating their relationship to regulatory networks and their potential functional importance. In total, 1,085 genes were differentially regulated between AFs and DLFs, confirming their genomically distinct nature. Genes up-regulated in AFs were significantly enriched with functional groups involved in ECM and ECM organization, whereas genes up-regulated in DLFs were significantly enriched with functional groups participating in actin binding and cytoskeletal organization. All these GO categories are consistent with the myofibroblast-like characteristics of DLFs, which may normally play an important role in parenchymal changes with respiration. However, this function is only one of many that DLF may play in the lung parenchyma, and unfortunately, gene expression studies are unable to define these specific roles. Moreover, specific molecular functions and pathobiologic disorders associated with the gene expression profile in DLFs were also strongly associated with SMAD3 and MAPK8. The up-regulation of SMAD3 and MAPK8 expression and activation in DLFs was subsequently confirmed by quantitative RT-PCR and protein analysis. The TGF-β canonical signaling molecules, SMAD3 and MAPK8, were significantly expressed and activated in DLFs. Thus, genomic and phenotypic analyses, along with activation pathways, all support the critical importance of TGF-β signaling pathways, SMAD3, and MAPK8 in differentiating DLFs from AFs.
Our work is related to two previous microarray studies that, rather than characterizing fibroblasts from the same anatomic location (lung) as described here, compared gene expression from skin fibroblasts of various anatomic sites (including the foot, arm, and scalp) to lung, liver, prostate, and aortic fibroblasts (5
). The differential expression of genes involved in the synthesis of ECM, cell migration, and growth/differentiation, including TGF-β signaling, distinguished fibroblasts from different anatomic sites. These findings suggest that fibroblasts encode positional signals to guide cells to differentiate or migrate. Thus, it is tempting to postulate that despite their close anatomic proximity, genomic differences between AFs and DLFs encode distinct positional signals that facilitate their differential roles in the lungs. AFs may be more involved in the organization of ECM and immune responses, whereas DLFs may be more involved in lung regeneration and repair. TGF-β1 signaling, through differences in the expression and activation of SMAD2/3 and MAPK8 (JNK1), appears to be central to this differentiation.
Although considerable interest has arisen regarding the differences in fibroblast phenotypes from various diseases, including interstitial pulmonary fibrosis (23
), pulmonary hypertension (26
), scleroderma (27
), systemic sclerosis (29
), and asthma (30
), as well as the migration of fibrocytes into the lung (31
), little discussion has centered on the possibility that different fibroblast phenotypes normally exist in the lung. To our knowledge, the finding of two genomically different human lung fibroblast phenotypes (i.e., AFs, innately more fibroblast-like, and DLFs, more “myofibroblast-like”) in structurally different lung regions is unique, with implications for fibrotic lung disease. Moreover, at least in asthma, the differences in gene expression have more to do with location than disease state in human lung fibroblasts. A clear precedent for “normal” phenotypic differences exists in the liver, where four different liver fibroblast–myofibroblast phenotypes were described (32
). These different fibroblasts in liver are thought to contribute differentially to the fibrogenic process in hepatic disease (34
). Because these different fibroblast phenotypes have not been clearly delineated in the lung, studies that purport to show a different “diseased” fibroblast phenotype may in actuality show a fibroblast–myofibroblast infiltration from a different region of the lung. In fact, despite the global differences in gene expression profiles according to lung region, no disease-specific differences were evident in gene expression between asthmatic and normal control fibroblast regions. This evaluation is the most extensive to date of differences in gene expression profiles between asthmatic and normal lungs, and supports the concept that in asthma, only small (or no) basal differences in gene expression are present compared with normal control samples. Finally, studies that use purchased “lung fibroblasts” should also be interpreted with caution, because these likely represent either a mix of fibroblast phenotypes, or a predominance of parenchymal fibroblasts. Further, regenerative therapies for lung disease should also consider these differences in fibroblasts, replacing diseased areas of the lung with appropriately “matched” fibroblasts.
The DLF network is significantly enriched with genes linked to myofibroblast-like characteristics, suggesting that the distal lung may be preprogrammed to repair injury more rapidly than the airways, and may contain the more important fibroblast cell type for fibrotic processes. This may also partly explain the relative differences in levels of fibrosis between traditional parenchymal disease (e.g., idiopathic pulmonary fibrosis [IPF]) and airway disease (e.g., asthma).
TGF-β1 is a myogenic factor that induces the differentiation of fibroblasts to myofibroblasts (3
). However, the relative importance of TGF-β–related pathways to regional lung fibroblast phenotypic differences was not previously reported. The literature supports SMAD3 as a key canonical signaling molecule for the effects of TGF-β1 in the differentiation of fibroblasts (35
). Network analysis predicted that SMAD3 plays an important role in the expression of α-SMA, differentiating DLFs from AFs. In addition to significant differences in the expression of SMAD3 in gene arrays, total and phospho-SMAD3 proteins were also significantly increased in DLFs, supporting a critical role of SMAD3 in driving their myofibroblast-like characteristics. Interestingly, our network analysis suggested distinct roles for SMAD2 and SMAD3 in TGF-β1 signaling. Although SMAD2 was involved in both AF and DLF networks, only SMAD3 was involved in DLF networks. Finally, the overexpression and activation of SMAD3, but not of SMAD2, increased the expression of α-SMA and cytoskeletal organization, perhaps explaining differences in the impact of TGF-β on the differentiation of myofibroblasts in DLFs, compared with collagen and ECM in AFs (39
). The molecular mechanisms driving the active SMAD3 pathway in DLFs require further study.
In addition to the canonical SMAD pathways, microarray and quantitative RT-PCR data support a role for MAPK8 (JNK1) in the distal lung myofibroblast phenotype. Interestingly, regional differences in total MAPK8 protein were not evident. However, the phosphorylation of MAPK8 in Western blot analyses was greater in DLFs than in AFs. Numerous factors, including the lower sensitivity of Western blots and of translational and other protein-directed processes, could dampen the differences observed in MAPK8 at the mRNA level. However, the concurrent enhanced expression and activation of threonine and tyrosine kinase could promote the activation of MAPK8 (40
). In particular, cdc42 signaling (enhanced in the DLF network) can cause the activation of JNK through a mixed-lineage kinase–3 pathway (Figure E1C). Activated JNK could then also phosphorylate SMADs (in particular, SMAD3), and further contribute to the differentiation of myofibroblasts through TGF-β1–activated pathways (35
). Interestingly, the elevated activation of JNK was a feature of fibrotic lung fibroblasts isolated from patients with pulmonary fibrosis associated with systemic sclerosis. Whether this activation is attributable to a disease-specific effect, as previously thought, or to the overgrowth of a DLF phenotype in patients with systemic sclerosis, remains unclear (42
In conclusion, to the best of our knowledge, this is the first study to compare global gene expression profiles between fibroblasts from distinctly different lung regions (airway and parenchyma). As microarray and network analyses suggest, and as follow-up studies confirm, DLFs exhibit a higher basal activation of SMAD3 and MAPK8, which promotes the more myofibroblast-like phenotype present in distal lung and parenchymal tissue. The differences in AFs and DLFs suggest that they will respond differently to injury, activate alternative regenerative pathways, and control different lung activities, thus confirming profound differences between these cell types. Further studies are needed to determine the genomic reasons for the increased SMAD3 and MAPK8 signaling and the mechanisms governing the myofibroblast-like phenotype in DLFs, as well as the structural, immunological, and even disease-related implications of these phenotypic differences. The existence of these two distinct lung phenotypes should be considered in all future studies of injury, regeneration, and repair in the human lung.