Ventilator-induced Lung Injury Promotes Inflammation within the Pleural Cavity

Abstract

Mechanical ventilation contributes to the morbidity and mortality of patients in intensive care, likely through the exacerbation and dissemination of inflammation. Despite the proximity of the pleural cavity to the lungs and exposure to physical forces, little attention has been paid to its potential as an inflammatory source during ventilation. Here, we investigate the pleural cavity as a novel site of inflammation during ventilator-induced lung injury. Mice were subjected to low or high tidal volume ventilation strategies for up to 3 hours. Ventilation with a high tidal volume significantly increased cytokine and total protein levels in BAL and pleural lavage fluid. In contrast, acid aspiration, explored as an alternative model of injury, only promoted intraalveolar inflammation, with no effect on the pleural space. Resident pleural macrophages demonstrated enhanced activation after injurious ventilation, including upregulated ICAM-1 and IL-1β expression, and the release of extracellular vesicles. In vivo ventilation and in vitro stretch of pleural mesothelial cells promoted ATP secretion, whereas purinergic receptor inhibition substantially attenuated extracellular vesicles and cytokine levels in the pleural space. Finally, labeled protein rapidly translocated from the pleural cavity into the circulation during high tidal volume ventilation, to a significantly greater extent than that of protein translocation from the alveolar space. Overall, we conclude that injurious ventilation induces pleural cavity inflammation mediated through purinergic pathway signaling and likely enhances the dissemination of mediators into the vasculature. This previously unidentified consequence of mechanical ventilation potentially implicates the pleural space as a focus of research and novel avenue for intervention in critical care.

Mechanical ventilation is a crucial component of supportive treatment in the ICU, both for patients suffering from respiratory failure and for those undergoing major surgical procedures. It is, however, widely accepted that mechanical ventilation invokes nonphysiological forces within the lung that can worsen outcome, a phenomenon termed ventilator-induced lung injury (VILI). The “biotrauma” hypothesis of VILI proposes that ventilation exacerbates inflammation and that this is causally linked to permeability and physiological dysfunction of the lungs. After systemic dissemination, this inflammation is further suggested to contribute to the multiple-system organ failure that is commonly observed in patients with acute respiratory distress syndrome (ARDS) (1, 2).

Although the inflammatory consequences of mechanical ventilation within the lungs themselves have been widely studied, there has been almost no consideration given to the pleural cavity, despite its intimate relationship with the lung parenchyma. Pleural effusions are a very common finding in the ICU, occurring in up to 80% of ventilated patients with ARDS, and are associated with longer duration of ventilation and ICU stay, although the reasons are uncertain (3). Pleural effusion drainage has been shown to improve oxygenation and, possibly, respiratory mechanics, although whether physical constraints of effusions (e.g., compression of respiratory structures) significantly impact clinical outcomes is questionable (4). It has been shown in patients with sepsis that pleural fluid samples can contain high levels of inflammatory mediators (5), and although the source of such mediators is unknown, the pleural cavity contains populations of cavity-specific macrophages (6) and other leukocytes that could potentially be a source of locally produced cytokines. Moreover, mesothelial cells lining the pleural cavity are exposed to substantial mechanical forces during breathing. Pleural mesothelia undergo significant cyclic stretching during lung expansion (7, 8), whereas movement of the visceral and parietal pleura relative to each other imposes substantial shear stress (9). Cultured pleural mesothelial cells are capable of releasing a wide range of mediators in response to inflammatory stimuli (10). Furthermore, both mechanical stretching and shear stress induce the release of endothelin-1 from mesothelial cells in vitro (11).

Therefore, within this study, we hypothesized that high tidal volume (Vt) mechanical ventilation induces inflammation within the pleural cavity as a result of pleural cell activation. Our findings indicate that VILI (perhaps uniquely) both upregulates pleural cavity cytokines and enhances dissemination of proteins into the circulation, potentially implicating the pleural space as an important locus of inflammation and therapeutic target within critically ill patients.

Some of the results of these studies have been previously reported in the form of an abstract (12).

Methods

(For detailed methods, see Supplemental Methods in the data supplement.)

Injury Models

Animal experiments were performed under the Animals (Scientific Procedures) Act 1986 of the United Kingdom, using male C57BL/6 mice aged 8–12 weeks. The in vivo VILI model has been described previously (13, 14) (see the data supplement for details). Anesthetized mice were ventilated using either an injurious high-Vt strategy (Vt = 32–36 ml/kg) or a protective low Vt strategy (7–8 ml/kg, with recruitment maneuvers every 30 min). Mice were ventilated for predetermined periods up to 3 hours or until a mortality surrogate was reached. In inhibition studies, 50 μl saline or PPADS (purinergic receptor inhibitor) were delivered intrapleurally before ventilation protocols.

As a comparison to the VILI model, experiments using an acute acid-aspiration model were performed (15). Mice were anesthetized, intratracheally instilled with saline or 0.15 M HCl (50 μl), and ventilated with low Vt (including recruitment maneuvers) for 3 hours.

After termination, plasma and BAL fluid (BALF) samples were collected. Pleural lavage fluid (PLF) was recovered by piercing the parietal pleura, instilling 400 μl saline, and withdrawing as much as possible from the cavity. Cytokine concentrations were determined by ELISA, and ATP levels were determined using an ATPlite kit (PerkinElmer). Flow cytometry was used to assess the activation of pleural cavity cells through surface and intracellular markers, as well as the production of extracellular vesicles (EVs) (16).

Protein Movement Studies

To determine blood-to–pleural cavity and blood-to–alveolar space permeability during VILI, 50 μl Alexa Fluor 594–conjugated albumin (AF594-albumin) was injected intravenously. After 3 hours of ventilation with low or high Vt, we collected plasma, BALF, and PLF, and fluorescence was evaluated. Plasma fluorescence was determined after a 100-fold dilution, whereas BALF and PLF samples were measured undiluted. Dye movement was evaluated as ratio of fluorescence units in BALF or PLF versus plasma.

To determine pleural cavity–to-blood or alveolar space–to-blood movement of protein, 50 μl AF594-albumin was instilled into the pleural cavity or alveolar space, followed by ventilation for 3 hours with low or high Vt. High-Vt settings were matched on the basis of plateau pressure to invoke similar degrees of lung strain (34.3 ± 2.27 vs. 33.3 ± 1.47 cm H₂O for intrapleural vs. intratracheal dye, respectively). Serial plasma samples from intrapleurally and intratracheally dosed mice were collected at intervals throughout ventilation, and fluorescence was determined. Dye movement was evaluated as the ratio of fluorescence units in undiluted plasma versus initial instillate (after 100-fold dilution).

In Vitro Studies

Rat pleural mesothelial cells (4/4 RM-4) were seeded onto collagen-coated flexible-bottom plates and exposed to 20% cyclic stretch (or held static) for 15 minutes or 4 hours, using a Flexcell FX-6000 Tension system. After 15 minutes, supernatants were collected, ARL67156 (ecto-ATPase inhibitor) was added, and ATP levels were determined. After 4 hours, supernatants were recovered, and EV numbers were determined by flow cytometry (16).

Statistics

Data were evaluated for distribution using the Shapiro-Wilk test of residuals. Normally distributed data were analyzed with student’s t test, a one-way ANOVA with Sidak’s test, or a two-way ANOVA with Tukey’s multiple comparisons test. Nonnormally distributed data were analyzed with the Mann-Whitney test or Kruskal-Wallis test followed by Dunn’s test.

Results

The physiological characteristics of the VILI model are shown elsewhere (see Figure E1 in the data supplement). Consistent with previous studies (14, 17), in this “one-hit” model of high-Vt ventilation, respiratory mechanics were initially stable until ∼90 minutes, after which, there was a relatively rapid deterioration. In contrast, low-Vt ventilation with recruitment maneuvers induced no impairment in respiratory mechanics, and all animals survived until the predetermined endpoint. As expected, 3 hours of ventilation with high Vt induced substantial increases in respiratory system elastance (Figure 1A), BALF total protein (Figure 1B), and fluorescent marker translocation from blood to BALF (reflecting enhanced barrier permeability) (Figure 1C). In addition, ventilation with high Vt promoted significant increases in BALF cytokines (Figures 1D–1F).

Figure 1.

Permeability and inflammation in alveolar and pleural compartments during ventilator-induced lung injury (VILI). Samples were taken at either 3 hours of ventilation or when injury had reached a point to attain the mortality surrogate if this occurred first (only with mice ventilated at a high tidal volume [Vt]). (A) Elastance change was calculated as percent increase compared with the value at the start of ventilation with either high or low Vt. (B and C) We determined (B) the concentration of total protein in BAL fluid (BALF) and (C) the ratio of Alexa Fluor 594-albumin (AF 594-albumin) present in BALF to that in plasma. (D–F) Markers of inflammation (D) IL-6, (E) CXCL1, and (F) CCL2 were measured in BALF using ELISA. (G–K) Similarly, (G) total protein in pleural lavage fluid (PLF), (H) the ratio of AF 594-albumin present in PLF to that in plasma and pleural fluid (I) IL-6, (J) CXCL1, and (K) CCL2 were evaluated. Data in (A) through (D) and (H) through (K) were nonnormally distributed and analyzed by Mann-Whitney test. Data in (E) through (G) were normally distributed and analyzed by Student’s t test. Data are displayed as individual data points with solid lines indicating either the mean or the median value for normal or nonnormal data, respectively. n = 6–7 for each dataset. *P < 0.05. **P < 0.01. ***P < 0.001.

On evaluation of PLF samples, we similarly observed that total protein levels were raised after 3 hours of ventilation with high Vt, associated with increased permeability of the pleural barriers (Figures 1G and 1H). Moreover, we also observed substantial increases in inflammatory cytokines IL-6, CXCL1, and CCL2 in the pleural space (Figures 1I–1K). Of note, the levels of PLF cytokines were similar in magnitude to those found in plasma (IL-6: PLF, 318 ± 264 pg/ml vs. plasma, 604 ± 165 pg/ml; CXCL1: PLF, 261 ± 206 pg/ml vs. plasma, 481 ± 272 pg/ml; CCL2: PLF, 137 ± 56 pg/ml vs. plasma, 125 ± 48 pg/ml).

To explore whether this phenomenon was specific to VILI, we performed similar measurements in a model of acute acid-aspiration lung injury. As with the VILI model, acid aspiration induced significant increases in respiratory elastance and resistance over 3 hours (Figure 2A; see Figure E2), along with increases in BALF total protein, IL-6, CXCL1, and CCL2 (Figures 2B–2E). Notably, elastance increase and BALF cytokine concentrations were higher after acid aspiration compared with VILI. However, in complete contrast to the VILI model, acid aspiration did not lead to increases in total protein or any cytokines measured within the pleural space (Figures 2F–2I).

Figure 2.

Permeability and inflammation in alveolar and pleural compartments in acid-aspiration model. Samples were taken 3 hours after either saline or HCl instillation. (A) Elastance change was calculated as percent increase compared with the value immediately after instillation of either saline or HCl. (B–I) Concentration of BALF (B) total protein, (C) IL-6, (D) CXCL1, (E) CCL2, and pleural lavage fluid (PLF) (F) total protein, (G) IL-6, (H) CXCL1, and (I) CCL2 were determined. Data in (A), (B), and (I) were nonnormally distributed and analyzed by Mann-Whitney test. Data in (C) through (H) were normally distributed and analyzed by Student’s t test. Data are displayed as individual data points, with solid lines indicating either the mean or the median value for normal or nonnormal data, respectively. n = 6 for each dataset. *P < 0.05. **P < 0.01. ****P < 0.0001. ns = not significant.

As resident macrophages have previously been implicated in inflammatory responses within the pleural cavity (18), we focused our investigations on whether these may be responsible for inflammation during VILI, with B-cell responses explored as a comparator. PLF cells were first identified on the basis of the expression of CD11b and CD19, the majority of PLF cells falling into either the CD11b⁺CD19⁻ or the CD11b⁻CD19⁺ population (Figure 3A). CD19⁺ cells were assumed to be populations of B cells (19), whereas CD11b⁺CD19⁻ cells were confirmed as macrophages on the basis of the absence of circulating monocyte (Ly6C) and neutrophil (Ly6G) markers (Figure 3B). Consistent with previous findings, macrophages existed as two distinct populations, with a major population of F480^hi cells showing higher granularity (Figure 3C); and higher expression of ICAM2 (Figure 3D), TIM-4 (Figure 3E), and CD86 (Figure 3F); but lower expression of MHCII (Figure 3G), compared with F480^lo macrophages (19–21). After 3 hours of ventilation with high Vt, the numerically dominant cell populations in the pleural cavity (F480^hi macrophages and CD19⁺ cells) were unchanged in number, although there was a small but significant increase in the number of F480^lo macrophages (Figures 4A–4C). Numbers of neutrophils and monocytes remained negligible (data not shown). To clarify whether leukocytes in the pleural cavity were activated during ventilation with high Vt and, thus, potentially responsible for inflammatory mediator production, we evaluated surface ICAM-1 and whole-cell IL-1β expression by flow cytometry. ICAM-1 expression was significantly increased in F480^hi macrophages and showed a close-to-significant increase in F480^lo macrophages (P = 0.056) but not on CD19⁺ B cells (Figures 4D–4F). Cellular IL-1β levels were significantly increased in F480^lo macrophages (Figures 4G–4I), indicating that macrophages specifically showed signs of inflammatory activation during VILI.

Figure 3.

Identification of cells recovered from pleural cavity lavage. (A) Cells were first distinguished on the basis of the expression of CD11b and CD19. (B) CD11b⁺ cells were then confirmed as being negative for the circulating monocyte marker Ly6C and the neutrophil marker Ly6G. (C–G) CD11b⁺ cells could be identified as two distinct populations on the basis of the expression of F480 as well as (C) side scatter (SSC) characteristics and expression of (D) ICAM-2, (E) TIM-4, (F) CD86, and (G) MHCII.

Figure 4.

Cell quantification and activation in pleural lavage fluid (PLF) samples by flow cytometry. (A–C) Numbers of (A) F480^hi macrophages, (B) F480^lo macrophages, and (C) CD19⁺ cells were determined after 3 hours of ventilation with high or low Vt. (D–I) Additionally, surface ICAM-1 expression on (D) F480^hi macrophages, (E) F480^lo macrophages, and (F) CD19⁺ cells, and whole-cell IL-1β expression in (G) F480^hi macrophages, (H) F480^lo macrophages, and (I) CD19⁺ cells were evaluated. Data were nonnormally distributed and analyzed by Mann-Whitney test. Data are displayed as individual data points, with solid lines indicating the median value. n = 5–7 for each dataset. *P < 0.05. MFI = mean fluorescence intensity; Vt = tidal volume.

To further explore the nature of inflammatory responses, we determined the numbers of EVs released from the different resident cell types, identified previously as markers of cellular activation. These measurements were performed at an earlier time point of ventilation (1 h) to minimize possible confounding effects of EV ingress from outside the pleural space or uptake by cells. The data demonstrate a significant increase in the number of CD11b⁺ EVs in the PLF after just 1 hour of ventilation with high Vt (Figure 5A). MHCII⁺ and ICAM2⁺ EVs showed a numerical, but not statistically significant, increase (data not shown), whereas CD19⁺ EVs were unchanged (Figure 5B). It is interesting that there was a close-to-significant (P = 0.076) increase in T1α⁺ EVs (Figure 5C), suggesting the possibility of mesothelial cell activation. To confirm that EV data represent local inflammatory cell activation rather than infiltration from elsewhere (e.g., the alveolar space, as T1α is also present on alveolar epithelial cells), EVs expressing either CD11b or CD11c (alveolar macrophage marker) were determined in PLF and BALF after 1 hour of ventilation with high or low Vt. Consistent with the concept of a local EV source, CD11c⁺ EV numbers only increased in BALF after VILI, whereas CD11b⁺ EVs only increased in PLF (see Figure E3). Thus, the data indicate little, if any, direct movement of mediators between the alveolar space and the pleural space.

Figure 5.

Extracellular vesicle (EV) identification and quantification in pleural lavage fluid (PLF) samples by flow cytometry. (A–C) Extracellular vesicles expressing surface markers for (A) CD11b, (B) CD19, and (C) T1α were quantified after 1 hour of ventilation with high or low Vt. Data were nonnormally distributed and analyzed by the Mann-Whitney test. Data are displayed as individual data points, with solid lines indicating median value. n = 6–7 for each dataset. *P < 0.05.

Given the aforementioned findings, we postulated a mechanistic paradigm whereby exposure of mesothelial cells to stretch induces the release of factors that subsequently promote local macrophage responses. Initially, in vitro studies were performed that showed that mesothelial cells exposed to stretch for 4 hours released T1α⁺ EVs, with no significant increase in cell death on the basis of lactate dehydrogenase assay (Figures 6A and 6B). We determined the secretion of ATP, a known damage-associated molecular pattern (DAMP), from mesothelial cells and found a significant increase after just 15 minutes of stretch (Figure 6C). Subsequently, we explored whether ATP released from stretched mesothelial cells may be responsible for pleural macrophage activation in vivo. First, we determined ATP concentration within the pleural cavity, which was significantly increased after 1 hour of ventilation with high Vt (Figure 6D). We then performed experiments utilizing the nonspecific purinergic receptor blocker PPADS, which demonstrated that both CD11b⁺ EV production and proinflammatory cytokine upregulation were attenuated by purinergic receptor inhibition (Figures 6E–6H).

Figure 6.

Pleural cavity purinergic receptor involvement in ventilator-induced lung injury (VILI). (A) Pleural mesothelial cells (4/4 RM-4) were exposed to 20% cyclic stretch or held static for 4 hours for measurement of T1α⁺ extracellular vesicles. (B) Presence of lactate dehydrogenase was evaluated in terms of percentage of dead cells (% signal after cell lysis) to clarify whether cell stretching was associated with cell death. (C) Pleural mesothelial cells were exposed to 20% cyclic stretch or held static for 15 minutes for measurement of ATP secretion. (D) Subsequently, ATP concentration was evaluated in pleural lavage fluid (PLF) samples from mice ventilated with high or low Vt for 1 hour. In separate experiments, mice were dosed intrapleurally with 50 μl saline or 50 mM PPADS before ventilation with high or low Vt for 1 hour. (E–H) PLF samples were evaluated for (E) CD11b⁺ EVs by flow cytometry, and (F) soluble IL-1β, (G) IL-6, and (H) CXCL1. Note that data in (E) were obtained with a more sensitive flow cytometer than that used to obtain the data in Figure 5; hence, the differences in absolute CD11b⁺ EV counts between figures. Data in (A), (B), (C), and (E) were normally distributed and analyzed by Student’s t test or one-way ANOVA with Sidak’s multiple comparisons test as appropriate. Data in (D), (F), (G), and (H) were nonnormally distributed and analyzed by Mann-Whitney test or Kruskal-Wallis test with Dunn’s multiple comparisons test. Data are displayed as individual data points, with solid lines indicating mean or median value. n = 5–6 for each dataset. *P < 0.05. ***P < 0.001. EV = extracellular vesicle.

Finally, having demonstrated the initiation of an inflammatory response within the pleural cavity during ventilation with high Vt, we considered whether this remained a localized response or whether there was the potential for inflammatory mediators to disseminate. Therefore, we determined the movement of fluorescence-labeled albumin out of the pleural and alveolar spaces during 3 hours of ventilation. The immediate impact of bolus administration on respiratory mechanics meant that, despite matched plateau pressure, Vt was slightly (but not significantly) lower in mice ventilated with high Vt after intratracheal administration of protein compared with intrapleural administration (28.3 ± 3.04 ml/kg vs. 31.2 ± 2.64 ml/kg). However, the increase in elastance over the time course was similar between intratracheally and intrapleurally dosed mice that were ventilated with high Vt (percent increase: 10.3 ± 14.0 vs. 5.59 ± 6.19, respectively), indicating that similar degrees of VILI were induced. After intrapleural administration, there was a gradual appearance of dye within plasma over the course of 3 hours of ventilation with low Vt, which was dramatically enhanced at all time points by ventilation with high Vt (Figure 7). After intratracheal administration, high-Vt ventilation also increased dye appearance in plasma, compared with low-Vt ventilation. However, amounts of dye translocating to plasma were significantly lower after intratracheal administration compared with intrapleural administration, regardless of ventilation strategy (Figure 7; see Figures E4 and E5). Overall, these data demonstrate that ventilation with high Vt promotes movement of protein from the pleural space to the circulation and does so to a greater extent or more rapidly than occurs from the alveolar space.

Figure 7.

Translocation of labeled protein from pleural cavity or alveolar space to circulation. Time course of appearance of AF 594-albumin within plasma, expressed as a ratio of plasma fluorescence to that of the original instillate during 3 hours of ventilation with high or low Vt. Data were normally distributed and were analyzed by two-way ANOVA followed by Tukey’s multiple comparisons test. (A) Time course of changes within different groups, with data displayed as mean ± SD. A Time × Group interaction showed significance at P < 0.0001. (B) Comparisons at each time point, displayed as individual data points with differences between each group at a given time. *P < 0.05. **P < 0.01. N = 6–7 for each dataset at each time point. High IPL = high Vt, intrapleural dye; High IT = high Vt, intratracheal dye; Low IPL = low Vt, intrapleural dye; Low IT = low Vt, intratracheal dye.

Discussion

Although mechanical ventilation is a life-saving component of ICU treatment, the potential for negative consequences is well known, including the suggestion that ventilation exacerbates the development of multiple-system organ failure among patients with ARDS. A number of possible explanations for this have been proposed (22), but the focus has mostly been attached to the inflammatory response. Both local intrapulmonary and circulating cytokines have been linked to VILI in preclinical models and associated with increased mortality in patients with ARDS, leading to the concept that mechanical ventilation promotes organ injury through the systemic propagation of inflammation (2). Explanations for this dissemination of mediators involve either passive leak from alveolus to blood or ventilation directly impacting on endothelial cells and/or leukocytes within the pulmonary vasculature (23, 24). Perhaps surprisingly, there has been no consideration of a possible involvement of the pleural cavity, even though it is exposed to substantial physical forces during lung expansion. In this study, we show—for the first time, to our knowledge—that injurious mechanical ventilation provokes inflammation within the pleural space, which may contribute to the inflammatory load of patients.

In our initial experiments, we determined the levels of inflammatory cytokines in BAL and pleural fluid of ventilated animals. As expected, inflammatory mediators within the alveolar space were increased with high Vt, but we also demonstrated heightened inflammation within the pleural cavity. It is important to note that the levels of cytokines in PLF after VILI were comparable in magnitude with the concentration in plasma. The mouse pleural cavity comprises a single compartment surrounding both lungs, containing a volume of ∼20 μl fluid (25). Here, we utilized a lavage recovery procedure involving the instillation of 400 μl saline, allowing washout of the cavity surrounding both lungs but diluting pleural lining fluid ∼20-fold. Thus, the cytokine levels in pleural fluid before dilution would likely be substantially higher than observed in plasma, consistent with local production rather than movement from the circulation (26). Total protein levels in PLF were significantly increased after ventilation with high Vt, which, given that plasma protein levels were unchanged between low and high Vt (27.7 ± 1.88 mg/ml vs. 28.2 ± 3.57 mg/ml), is consistent with the development of an exudative effusion (27) during VILI. It should be pointed out that, in these initial experiments, only 1/6 mice survived the full 3-hour high-Vt protocol, with others meeting the mortality surrogate earlier. Therefore, animals that were ventilated with high Vt were, on average, ventilated for less time than other groups. This is, however, very unlikely to have any material impact on our conclusions, as analysis (data not shown) found that all mediators and markers of permeability tended to increase with longer ventilation time with high Vt. Nevertheless, for subsequent experiments, ventilation settings were adjusted slightly to ensure that all animals who were ventilated with high Vt survived the entire period.

Crucially, within an acid-aspiration model of lung injury that induced similar alveolar-epithelial permeability, and even greater increases in BALF cytokines than the VILI model, we found no increases in PLF protein or cytokine concentrations. Thus, the development of pleural cavity inflammation seems to be a unique response to VILI compared with other etiologies of acute lung injury, at least within this early phase.

The major leukocyte types identified within the pleural cavity—which, we hypothesized, may be responsible for the aforementioned cytokine secretion—included CD19⁺ B cells and CD11b⁺ macrophages, the latter of which existed in two distinct populations on the basis of the relative expression of F480, MHCII, ICAM2, TIM-4, and CD86 (19–21). The numerically dominant F480^hi MHCII^lo population is considered to be a self-proliferating population derived from embryonic precursors, whereas the F480^lo MHCII^hi population is derived from circulating monocytes (20). Although there was no change in the numbers of F480^hi macrophages, the number of F480^lo macrophages did increase significantly with high-Vt ventilation. It is unclear whether this reflects recruitment of cells or altered efficiency of recovery.

As confirmation that resident pleural cavity cells were, indeed, directly activated by VILI, we demonstrated increased expression of ICAM-1 and IL-1β on or in macrophages, but not B cells, by flow cytometry. ICAM-1 was used, as we have previously demonstrated this to be a robust marker of alveolar macrophage activation during VILI (14). Although ICAM-1 expression was significantly increased only on F480^hi cells, there was a numerical increase on F480^lo macrophages. These had substantially higher basal expression levels than F480^hi cells, which may have made detection of a signal more difficult. ICAM-1 has traditionally been explored for a role in leukocyte migration, but recently, it has been shown to also play important roles in the activation of tissue-resident leukocytes (28).

To gain further insight into the inflammatory responses during ventilation, we investigated EV production from resident cells. These measurements were performed at an earlier time point of ventilation (1 h) both to minimize the possibility that these EV may be translocating in from elsewhere and to minimize the loss of EVs as a result of uptake by cells (29). Of note, our protocols are designed to recover a population consisting primarily of microvesicles (16), but we cannot rule out the presence of other EVs, including large exosomes (although this distinction is not particularly relevant here, given that we are using these measurements purely to gain information regarding the cell types activated, rather than invoking them as playing a pathophysiological role). Consistent with our previous observation of macrophage activation, we found significant upregulation of CD11b⁺ EVs, but not CD19⁺ EVs. Intriguingly, there were indications in our study (although not significant) of upregulated T1α⁺ EVs in the pleural space. T1α (podoplanin) is known to be abundantly expressed on the surface of pleural mesothelium (30), suggesting a direct response of the mesothelial cells to VILI. We subsequently demonstrated that pleural mesothelial cells that were exposed to stretch in culture released increased numbers of T1α⁺ EVs, confirming that these cells respond to cyclic stretch. It is theoretically possible that the T1α⁺ EVs observed in vivo could have moved from the alveolar space, as T1α is also present on the surface of alveolar epithelial cells. However, we believe that this is very unlikely, as alveolar macrophage–derived (CD11c⁺) EVs were increased in BALF, but not PLF, after VILI. These findings further support the conclusion that local macrophages were activated and that our observations were not due to spillover of mediators from the alveolar to the pleural space.

With regard to the precise mechanisms involved, we postulated a paradigm whereby the deformation of visceral and/or parietal pleural mesothelial cells during ventilation with high Vt leads to the release of signals that promotes macrophage activation and the consequent secretion of inflammatory cytokines into the pleural space. ATP is known to be rapidly released from a wide variety of cells exposed to mechanical stimulation (31, 32), and extracellular ATP is an important DAMP promoting inflammasome activation in macrophages. Therefore, we considered this to be a strong candidate as a signal that may be released by pleural mesothelium during VILI. Here, we show—for the first time, to our knowledge—that pleural mesothelial cells that are exposed to cyclic stretch in vitro release ATP and that in vivo VILI led to upregulated extracellular ATP in the pleural cavity. Moreover, use of the nonspecific purinergic receptor inhibitor PPADS promoted the almost-complete abrogation of CD11b⁺ (pleural macrophage) EV release and soluble IL-1β, IL-6, and CXCL1. Although the precise purinergic receptor remains uncertain, these data strongly support the hypothesis that extracellular ATP (or potentially other nucleotide DAMPs) released from stretched pleural mesothelial cells promotes local macrophage-driven inflammation. We cannot exclude that mesothelial cells might also directly contribute to cytokine upregulation; indeed, they have been reported to be responsible for mediator production during pleural disease (33) and recently during COVID-19–related fibrotic damage (34). However our findings would suggest that, in the early phase of VILI at least, this is likely to be a minor contribution compared with that of macrophages.

The pathophysiological consequences of our finding of ventilator-induced pleural inflammation remain unclear. A thorough exploration of this lies beyond the scope of this study, although we would speculate that it may have more impact on the extrapulmonary consequences of ventilation rather than a compulsory role in lung dysfunction per se (at least in the very acute phase explored here). We show that ventilation with high Vt, as well as inducing pleural cavity inflammation, also dramatically and rapidly enhances the egress of protein from the cavity into the circulation. It is interesting that movement from the pleural cavity seems to be of greater magnitude and/or rapidity than movement of protein from the alveolar space. The kinetics of dye appearance in plasma indicate that intrapleural protein seemed to translocate in a consistent manner over time during ventilation with high Vt, whereas intraalveolar protein translocation remained relatively limited until displaying a more rapid increase toward the end of the experiments. We would speculate that this may reflect different mechanisms of movement from the two compartments. More rapid translocation from the pleural cavity may reflect either greater permeability of the mesothelium-to-circulation barriers or alterations in lymphatic transport. This latter is a major mechanism of protein movement out of the pleural space (35), and crucially, lung lymph flow has been shown to be increased with ventilation with greater Vt (36). To note, increased respiratory rate has also been reported to promote epithelial cell permeability (37) and lung lymph flow (36), although in the present study, our VILI model involved a reduction of respiratory rate, so increased movement of mediators would be purely related to increased Vt. In contrast, the “late” increase in dye translocation from the alveolar space likely reflects the development of epithelial barrier breakdown. Thus, the data would indicate that injurious mechanical ventilation increases inflammatory cytokine levels within both the pleural cavity and the alveolar space but that egress from the pleural cavity into the circulation is likely to occur earlier, in the absence of overt lung injury.

The biotrauma hypothesis of ventilator-induced lung injury essentially proposes that ventilation enhances extrapulmonary inflammation, organ injury, and mortality through the “spillover” of cytokines from within the lungs or pulmonary vasculature (23). Here, we have demonstrated—for the first time, to our knowledge—that a possible alternative mechanism exists whereby VILI uniquely promotes both pleural cavity inflammation and the likely rapid dissemination of mediators from the pleural space into the vasculature. Although the precise relevance of this still remains to be proven (as, indeed, does any paradigm linking ventilation, inflammation, and organ failure), this novel mechanism potentially opens new therapeutic avenues, not least because the delivery of agents into the pleural space may be both more direct and less hampered by physical considerations (e.g., edema, altered regional perfusion) than other approaches.