A Means to an End(othelium): Using Peripheral Blood Mononuclear Cell 'Omics to Peer into the Pulmonary Vasculature

Although most often considered a disease of the airways and airspaces, chronic obstructive pulmonary disease (COPD) has long been associated with pulmonary vascular loss and dysfunction (1). Cigarette smoke is well recognized as the primary risk factor for COPD in the developed world (2) and has effects on the systemic and pulmonary vasculature (3). Some evidence from animal models suggests that pulmonary vascular remodeling associated with cigarette smoke exposure precedes the development of emphysema (4). Loss of pulmonary microvasculature has been shown to accompany parenchymal loss in COPD and appears to contribute to measures of COPD severity above and beyond those that can be explained by emphysema (5). This has led some to proffer the “vascular hypothesis” of emphysema that contends that its most proximate cause is damage to the pulmonary endothelium caused by cigarette smoke exposure (6).

Taken together, this makes the pulmonary endothelium a tantalizing target for study, but, by the nature of its location, it remains difficult to safely sample in a living organism. In the current issue of AnnalsATS, Buschur and colleagues (pp. 884–894) have undertaken a novel strategy to study the pulmonary endothelium by analyzing gene expression in peripheral blood mononuclear cells (PBMCs) (7), which are known to have direct interaction with the pulmonary vascular endothelium (8). The authors used a discovery-and-replication population selected from participants in the Multi-Ethnic Study of Atherosclerosis (MESA) COPD cohort, who had microarray data on PBMC gene expression as well as magnetic resonance imaging measures of pulmonary microvascular perfusion. They then employed the MESA Lung cohort as a second, larger replication population with less smoke exposure (40% never-smokers) and a lower percentage of COPD cases (14%). In MESA Lung, pulmonary vasculature was instead assessed via dual-energy computed tomography (DECT), which cannot be directly compared to magnetic resonance imaging measurements. Gene expression in PBMCs in the MESA Lung cohort was analyzed via RNA sequencing, but, of note, this was performed in 2010 and 2011, whereas DECT was performed in 2017 and 2018.

The authors’ discovery analysis in the MESA COPD cohort found 642 associations between the three different microvascular perfusion traits and levels of gene expression after adjusting for common covariates. A significant subset of associations between pulmonary microvascular blood flow and pulmonary microvascular blood volume were duplicated, likely because of the high correlation of these two traits (Spearman’s ρ = 0.92). The TNF-α (tumor necrosis factor-α) signaling pathway via the NF-κB (nuclear factor-κB) gene set was significantly enriched for genes with expression that were negatively associated with PMBF and PMBV but was positively associated with mean transit time (MTT), which is consistent with its negative correlation with PMBF and PMBV. This is also consistent with the NF-κB pathway’s known upregulation in COPD (9), as PMBF and PMBV are reduced in COPD (10). The only single gene to be significantly associated with both PMBV and PMBF in both discovery and replication MESA COPD analyses was NFKB2. There was little overlap between genes associated with MTT and those associated with PMBV and PMBF, and NPC1 and FOXN3 were the only two gene expression associations with MTT in the discovery set that remained significant in the MESA COPD replication set. Replication of the results from the initial MESA COPD study in the MESA Lung cohort was low (18 of 642 associations, or 2.8%), possibly due to the time between the gene expression and microvasculature measurements in the MESA Lung cohort, and no genes replicated across both replication populations. A sensitivity analysis testing for the association of gene expression with emphysema had no significant results, which may have been due to the low incidence of emphysema present in this population.

It is interesting to consider these results in the context of a recently published study of the blood-based transcriptomic and proteomic correlates of emphysema by Suryadevara and colleagues (11). Even though three genes from Table 2 in the study of Buschur and colleagues were initially associated with emphysema in the discovery cohort (ZNF74, PDXK, and DHX30), none of these associations were replicated in the internal and external replication populations used by Suryadevara and colleagues. This suggests that these two separate studies detected distinct molecular signatures with some degree of overlap, mirroring observations from the original analysis of pulmonary microvascular dysfunction in MESA COPD that noted changes in PMBV and PMBF in regions both with and without emphysema (10).

The primary strength of the study by Buschur and colleagues is precisely that it investigates the molecular signals associated with pulmonary microvascular perfusion and that it does so with a robust methodology. The authors have employed ingenuity to use existing data to query a relevant hypothesis, but this use of existing data has also given rise to many of the limitations of their work. The relatively small population studied here naturally limits statistical power, as does the temporal distance between the measurement of gene expression and the DECT measures in the MESA Lung replication cohort. The secondary sampling of the endothelium via changes in PBMCs is pragmatic but ultimately places these results at some remove from the endothelial pathophysiology. Finally, the use of bulk microarray data to measure gene expression means a loss of sensitivity compared with RNA sequencing and a loss of resolution compared with single-cell RNA sequencing, both methods that were not widely available when the microarray data were collected.

Single-cell RNA sequencing could offer some unique insights on the endothelium through similar PBMC sampling with the use of causal reasoning models to generate inferences about upstream influences on the transcriptome (12). The advent of spatial transcriptomics adds another promising method for investigation, although its requirement for whole tissue would likely require an animal model or decedent tissues. There is much more to be learned about the pulmonary microvasculature’s effects on COPD pathophysiology, and this study by Buschur and colleagues is a reminder of the importance of future investigation in this space.