In this study, we compared gene expression profiles of MSCs derived from lung, bone marrow and adipose tissue of COPD patients and non-COPD controls. MSCs from each source had a specific gene signature. Comparing COPD to control, the only genome-wide significantly different expressed gene in LMSCs was CSGALNACT1, while higher number of differentially expressed genes were observed in BM-MSCs and AD-MSCs.
The differences between COPD and control in MSCs from extrapulmonary tissues may reflect systemic effects of either smoking or the disease. Of note, we did not correct for presence of tumors in the lung tissue, which may have had effects systemic effects as well. The different signatures of MSCs from lung tissue versus other sources and between COPD and control should be taken into account when considering MSCs for therapeutic strategies in lung disease, the latter especially when using autologous MSCs. As for the tissue specificity, a specific profile of regenerative factors may be needed to realize tissue regeneration in each specific organ. For instance, HGF and FGF10 are known to mediate alveolar repair and stimulate proliferation of alveolar epithelial progenitors[14–16, 19]. Notably, we previously observed that both factors are lower expressed in COPD-derived LMSCs compared to those from controls[20]. Here, we show that genes encoding both growth factors are predominantly expressed in MSCs from the lung. Rolandsson and co-workers also observed important differences between LMSCs and BM-MSCs using a microarray[12]. Their study was the first to confirm that lung and bone marrow resident MSCs possess tissue specific properties. Although LMSCs had a higher colony forming capacity and lower osteogenic differentiation potential, the authors observed an overall more similar gene expression pattern in LMSCs and BM-MSCs compared to our study, with 89 genes differently expressed. Similar to our study, MSCs from lung and bone marrow were from different donors using the same culture protocol, whereas the isolation and culture protocols differ between our studies. Further, we isolated LMSCs from explanted peripheral lung tissue, while Rolandsson and co-workers used transbronchial biopsies in live patients, although it is unclear how this would explain the higher similarity between MSCs from different sources. In line with our findings, Rolandsson and co-workers showed that FOXF1 as well as HOXB5 and SFRP1 were amongst the lung-specific genes. All these genes have been demonstrated crucial for human lung development and branching[21, 22]. We observed that signature genes of LMSCs include FOXF1, TBX2, TBX4, SCN7A and ETS2, and that a stromal cell subset exists in lung tissue in vivo with a similar expression profile. Of interest, forkhead box F1 (FOXF1) is a lung embryonic mesenchyme-specific transcription factor with persistent expression into adulthood in mesenchymal stromal cells[23]. In murine studies, Foxf1 + cells were shown to encompass a stem cell subset of collagen 1-expressing mesenchymal cells with clonogenic potential and capacity to generate lung epithelial organoids[24]. Interactions between FOXF1 and sonic hedgehog (SHH), T-box transcription factor (TBX4), TBX2 and FGF10 pathways have been described, proposing an essential transcriptional network during early lung organogenesis[25]. SCN7A encodes an atypical sodium channel. It has been identified as signature gene of the stromal tumor micro-environment associated with survival of lung cancer[26] and is expressed by alveolar fibroblasts[27]. Ets2 a ubiquitous transcription factor that is induced by HGF-MET signaling and is activated after phosphorylation at threonine-72[28]. Previous studies highlighted the importance of phosphorylated Ets2 in lung inflammation and extracellular matrix remodeling, pathways involved in pulmonary fibrosis[29]. It will be of interest to further study the role of these LMSC signatures genes in lung tissue regenerative processes.
Strikingly, the differences between COPD and control were most pronounced in AD-MSCs, followed by BM-MSCs, and the lowest number of differentially expressed genes was found in LMSCs. So far, clinical studies in COPD using cell-based strategies have focused on autologous BM-MSCs. We observed that the top-hit gene upregulated in COPD-derived BM-MSCs was HLA-DRB, encoding the major histocompatibility complex (MHC) region DRB5. Genetic variation in this gene has been associated with interstitial lung disease[30] and with circulating levels of IL-6[31], a pro-inflammatory cytokine with higher levels in COPD. As for AD-MSCs, the pathways differently expressed between cells from COPD patients and controls suggest abnormalities in extracellular matrix-growth factor binding, and may thus reflect impaired adhesion/migration responses. The extent of differentially expressed genes in AD-MSCs may be a consequence of metabolic alterations that have been associated with COPD, although caution needs to be taken given the small sample number of AD-MSC donors in our study. Despite this, it is tempting to speculate on the implications of observed abnormalities in native AD-MSCs in COPD. To the best of our knowledge, it is unknown whether AD-MSCs from subcutaneous adipose tissue in the thoracic cavity have the potential to migrate into the lung tissue upon injury. The ability to differentiate towards adipocytes/adipocyte-like cells could be of relevance, as adipocytes highly resemble lipofibroblasts, which are well known to support regenerative processes[32]. The highest upregulated gene in AD-MSCs from COPD patients was HAPLN1, encoding hyaluronan and proteoglycan link protein 1. HAPLN1 is known to be expressed in lung fibroblasts, stabilizing aggregates of proteoglycan monomers with hyaluronic acid in the ECM, which can lead to fibrotic remodeling[33]. Collectively, differences between COPD and control-derived BM-MSCs and AD-MSCs may be of relevance when considering autologous MSCs for the treatment of COPD. Notably, we should also take into account that MSCs may change their phenotype upon administration.
It was somewhat surprising to find only 2 genes with genome-wide significance to be differently expressed between the lung-derived cells from COPD and control donors. This may reflect absence of major differences between COPD and control-derived LMSCs, at least in these specific subjects, but may also be due to the loss of a COPD-specific phenotype upon in vitro expansion, although this was apparently not the case for AD-MSCs and BM-MSCs. As mentioned earlier, we previously observed differences in several read-outs between LMSCs from COPD patients and controls[7]. The difference between our two studies is that LMSCs were previously grown in high-glucose media (25 mM), while here we used low-glucose media (5.5 mM) in order to be able to compare to the MSCs from the other sources. Future studies will have to reveal whether a low-glucose (normal) environment can normalize defects observed in LMSCs from COPD.
The most strongly upregulated gene in LMSCs from COPD patients was CSGALNACT1, which encodes the key enzyme that initiates the biosynthesis of chondroitin sulfates and dermatan. Although the functional consequences of high CSGALNACT1 expression of need further investigation, our data suggest that LMSCs can modulate the ECM in their micro-environment, resulting in higher chondroitin and/or dermatan sulfate ratios and as consequence potentially lower heparan sulfate ratios. Of interest, lower levels of heparan sulfate proteoglycans have been observed in COPD lung tissue[34]. Proteoglycans bind growth factors and thus instruct cellular attachment, proliferation and differentiation. Specifically, heparan sulfates act as co-factors to enhance FGF10 signaling[35], thereby potentially supporting alveolar epithelial activation as well as mobilization and recruitment of lung-resident MSCs[36].
The most strongly downregulated gene in COPD was a pseudogene, CSPG4P13. Pseudogenes can act as decoy for microRNAs, potentially enhancing the expression of their respective genes. The protein encoded by CSPG4, chondroitin sulfate proteoglycan 4, is a well-known marker for pericytes, but further investigation has to show the potential consequences of lower CSP4P13 expression in LMSCs.
A limitation of our studies is that the translation to protein data needs to be largely confirmed, as previously done for lower HGF and decorin levels in LMSCs from COPD patients versus controls[20]. We did perform staining for chrondroitin sulfate in decellularized scaffolds reseeded with LMSCs, confirming their ability to modify the ECM. However, no differences were readily apparent visually between scaffolds re-seeded with COPD and control-derived LMSCs and quantification was challenging given the presence of CS on empty scaffolds. Therefore, further functional studies will be required in order to confirm the differences between COPD and control derived MSCs.
Together, our data suggest that for cell-based strategies using MSCs, the differences in gene expression profiles between MSCs from different sources should be taken into consideration. LMSCs may be optimally equipped for lung tissue repair because of the expression of specific growth factor genes. Autologous MSCs from COPD patients may show abnormal regenerative responses, even or especially when cells from extrapulmonary sources are considered.