Welcome to the Neighborhood: Tissue-Resident Lung Natural Killer Cells in Chronic Obstructive Pulmonary Disease and Viral Infections

A growing literature demonstrates a key role for natural killer (NK) cells in chronic obstructive pulmonary disease (COPD) pathogenesis. NK cells are essential innate immune effectors, activated to release cytokines or to kill target cells on the basis of tightly regulated signals perceived via their activating and inhibitory receptors. Aberrant activation of NK cells can result in tissue damage and chronic inflammation. Peripheral blood NK cells from patients with COPD have lower cytotoxicity compared with those from smokers without COPD or never-smokers (1, 2). By contrast, NK cells from respiratory compartments, including BAL, sputum, and lung tissue, demonstrate increased relative cytotoxicity in COPD (3–6). Thus, anatomic location must be considered when evaluating the potential involvement of NK cells, particularly regarding their cytotoxic responses. During viral infections, a leading cause of COPD exacerbations, NK cells are crucial to limit initial disease severity but can also exacerbate pathology (7), suggesting that there are many questions still to be explored.

The lung contains a complex network of nonrecirculating immune cells, known as tissue-resident cells. Focus has been primarily on tissue-resident T cells, while less is known about tissue-resident NK (trNK) cells. However, data from other organs indicate that there are important functional differences between how trNK and circulating NK (cNK) cells that are recruited to tissues respond to infection (8). This disparity raises the question of whether lung trNK and cNK cells will also behave differently during a COPD exacerbation. By focusing primarily on studies of NK cells as a single entity, have we been missing important nuances?

In a study reported in this issue of the Journal, Cooper and colleagues (pp. 553–565) used an experimental murine model combining cigarette smoke exposure and influenza A virus (IAV) infection, plus human pathological lung tissues infected ex vivo with IAV, to compare the phenotype and functions of trNK versus cNK cells (9). The authors provide novel information about murine lung tissue–resident (CD49a⁺ [cluster of differentiation 49a]) NK cells, showing that they constitute two subsets on the basis of expression of CD49b and the transcription factor EOMES (eomesodermin) (Figure 1). CD49a⁺ trNK cells were similarly identified in human lung tissues, with the greatest expression of CD49a (an adhesion receptor) found on immature CD56^bright NK cells, as they previously reported (10). The percentage of CD56^bright NK cells that were positive for CD49a was increased in subjects with spirometrically more severe disease, resulting in a statistically significant inverse correlation.

Figure 1.

Lung natural killer (NK) subsets and viral lung infection. Markers in black font denote increased expression compared with markers in light gray font. Tissue-resident NK (trNK) cells can be distinguished from circulating NK (cNK) cells by the presence of CD49a (cluster of differentiation 49a); in mice they can be further separated on the basis of CD49b expression. cNK and trNK cells display unique phenotypes: trNK subsets express more CD69 and CD103 (indicative of tissue retention and localization) and greater basal NKG2D (natural killer group 2/member D; an activating receptor). The transcription factor EOMES (essential for NK development) is expressed on all three NK subsets, with highest expression on CD49a⁺ CD49b⁺ trNK cells and lowest expression on CD49a⁺ CD49b⁻ trNK cells. On the basis of the results of Cooper and colleagues (9), we postulate that influenza A virus (IAV) infection in individuals without chronic obstructive pulmonary disease (COPD) results in a rapid NK response, including prompt population expansion, increased expression of NKG2D (necessary to recognize IAV-infected cells), a cytotoxic response limited to the CD49a⁺ CD49b⁻ trNK subset, and little to no IFN-γ production. The outcome is a protective response with viral clearance and subsequent resolution. By contrast, the NK response in COPD is both delayed and aberrant. Cytotoxic responses shift from the purview of the smaller CD49a⁺ CD49b⁻ trNK subset to the more numerous cNK cells, and IFN-γ production is significantly increased. These alterations result in a dysfunctional response, leading to excessive inflammation and tissue damage. EOMES = eomesodermin.

In response to IAV infection of control air–exposed mice, cNK and trNK cells rapidly expanded, with significantly increased numbers by Day 3 that returned to baseline by Day 7. The CD49a⁺ CD49b⁻ trNK subset also increased expression of CD107a, a marker of degranulation, implying that this population contributes to elimination of virally infected cells. This finding is congruent with this investigative group’s previous results using resected lung tissue from individuals without COPD, in which ex vivo IAV infection also led to increased proportion of CD49a⁺ trNK expressing CD107a (10). In the present study, however, the response of cigarette smoke–exposed mice to IAV was delayed, with greater cNK and trNK numbers observed only at Day 7 after infection. Furthermore, degranulation (CD107a expression) no longer increased on the relatively small CD49a⁺ CD49b⁻ trNK subset (∼10% of total lung NK cells) but instead was significantly increased on the much larger cNK (∼90%) subset.

Given the importance of NK cells for early control of viral infections, these results suggest that smoking impairs the early and tightly controlled local NK cytotoxic response while allowing a delayed and excessive response that leads to greater lung damage (Figure 1). This interpretation mirrors the “Goldilocks hypothesis,” which originally hypothesized that the severity of exacerbations of COPD could result either from immune responses being “too little” or “too much.” The present study refines that concept by suggesting that for antiviral NK cells, it is a combination of too little followed by too much (11). On the basis of the finding that IAV infection can increase NK cytotoxicity toward a normally resistant epithelial cell line (12), this excessive response might involve killing of noninfected bystander epithelial cells, driving COPD progression. However, even though CD107a is an accepted surrogate for cytotoxicity, a limitation of the present study is the absence of cytotoxicity data, an important future direction.

A strength of this study is its combination of murine and human data, some but not all of which are congruent between species. The differences in NK cell production of IFN-γ are worth highlighting. In mice, IAV infection led to increased IFN-γ production by both cNK and CD49a⁺ CD49b⁺ trNK subsets, regardless of cigarette smoke exposure. By contrast, in the human experiments, only NK cells from COPD lungs produced IFN-γ in response to IAV, which was attributable predominantly to trNK cells. In this group’s previous study, human lung NK cells also produced significantly more IFN-γ in response to ex vivo IAV infection, compared with uninfected or ultraviolet-killed virus (10). This study extends those previous data, which did not distinguish those without from those with COPD. Other limitations are minor and are thoroughly discussed (e.g., the use of only female mice in a single large experiment).

Importantly, using state-of-the-art flow cytometry, this study helps set a new standard for NK research by demonstrating the necessity of including both cNK and trNK populations in future analyses. However, one could ask if even that approach will be too limited. Human lung trNK subsets (CD49a⁺ CD103⁺ vs. CD49a⁺ CD103⁻) differ in degranulation and cytokine production; lung CD103⁺ trNK cells are postulated to have an intraepithelial location, as has been shown for CD8⁺ T cells (13). Using high-dimensional flow cytometry, another study identified 1,000 unique NK subpopulations, 13 of which were altered in patients who experienced prior exacerbations, and 3 of those clusters were significant contributors in a linear regression model of previous exacerbation risk (14). That analysis did not include CD49a expression, so the tissue residency of those clusters is uncertain, but one cluster that was CD69⁺ (suggesting tissue retention) demonstrated increased cytotoxicity compared with bulk NK cells.

Taken together, it is clear that future studies must take measures to analyze the many effector functions of multiple NK cell subsets comprehensively, including at a minimum trNK and cNK cells. In summary, it is clear that important new insights have been gained from this study, but additional work is needed to understand fully the role of lung trNK cells in health and disease.