Transcriptomic Analysis of Alveolar Immune Cells in Acute Respiratory Distress Syndrome: To Lump or to Split?

To the Editor:

We read with interest the article recently published in the Journal by Morrell and colleagues entitled “Alveolar Macrophage Transcriptional Programs Are Associated with Outcomes in Acute Respiratory Distress Syndrome” (1). The authors sought to identify alveolar macrophage (AM) transcriptional signatures associated with poor outcomes in a single-center cohort of patients with acute respiratory distress syndrome. We applaud the authors for their work. Novel methods to improve phenotyping of critically ill patients, including those with acute respiratory distress syndrome, are urgently needed to aid both prognostic and predictive enrichment of clinical trials (2).

The study methodology used by the authors highlights an important question in research that aims to leverage omics technologies to identify clinically informative immune cell signatures in acutely ill patients: should we “lump” or should we “split”? In their study, Morrell and colleagues lump potentially heterogeneous cells together by isolating a broadly defined population of alveolar macrophages through negative selection (1). Marker expression of CD163 and CD71 is shown to support that the isolated cells are truly AMs. Genome-wide transcriptional profiling using microarray was then performed on these cells. Investigating AMs as a single cell type has advantages, including simplifying both sample processing and downstream analysis. In addition, it is plausible that clinically informative transcriptional changes can be identified across a broadly defined cell population irrespective of the complexities of the underlying biology. Indeed, a genomic classifier for lung cancer has been developed using a similar lumping approach with airway epithelial cells (3).

However, in recent years, a growing body of work has demonstrated the power of splitting, or teasing out functionally different subpopulations within broader cell types. We now know that the injured lung contains two distinct populations of AMs: embryonically derived resident AMs and recruited macrophages derived from circulating monocytes (4). Work using RNA sequencing supports that these subsets are transcriptionally and metabolically unique and play distinct roles in lung inflammation, lung fibrosis, and injury resolution (5, 6). With increasing use of single-cell profiling, the breadth and potential clinical importance of AM heterogeneity have become even more apparent. Our group has used single-cell RNA sequencing to identify novel populations of alveolar macrophages expressing profibrotic genes in patients with pulmonary fibrosis (7). In addition, a single-cell approach was recently used to identify five transcriptionally distinct clusters of AMs within the inflamed lung (8).

High-resolution studies aimed at splitting apart relevant populations of AMs in the injured lung can be limited by high cost and complex downstream analysis. In addition, it remains to be seen whether these novel approaches can support discovery of uniquely informative biomarkers or clinical phenotypes. However, it is important to consider how these splitting approaches may complement or improve on studies that analyze an immune cell population in the broadest terms. As an example, in the work by Morrell and colleagues, there were no differences in genome-wide expression profiles between patients with good versus poor clinical outcomes after adjustment for multiple testing (1). This may be because of the noise present in genomic datasets derived from critically ill patients, but it may also be driven in part by the limited resolution of the analysis; an overly broad view can make even complex systems look uniform.

Should we lump or split immune cells when studying the injured lung? If the goal is to advance our understanding of the pathobiologic mechanisms of disease, then splitting using high-resolution approaches offers particular promise. If the aim is to identify clinically informative disease phenotypes, then both approaches may prove useful and synergistic. However, when we choose the 10,000-foot view, we should remember and be informed by the complexity that lies below.