Mechanical ventilation enhances extrapulmonary sepsis-induced lung injury: role of WISP1–αvβ5 integrin pathway in TLR4-mediated inflammation and injury

Mechanical ventilation (MV) is well known to cause an iatrogenic syndrome of ventilator-induced lung injury (VILI). The pathophysiology of VILI includes intrapulmonary inflammatory cell infiltrates, increased vascular permeability and pulmonary edema, and it may occur in ventilation of a healthy lung or worsening of preexisting and coexisting injury [1]. Sensitization of VILI secondary to preexisting acute lung injury (ALI) due to pneumonia [2, 3], intratracheal endotoxin [4‐7] or sterile injury [8‐10] has provided preclinical evidence of a two-hit model.

Although extrapulmonary endotoxemia combined with noninjurious mechanical ventilation leads to VILI [11, 12], evidence of such a two-hit phenomenon in experimental extrapulmonary bacterial sepsis is less clear. Ventilating rodents after polymicrobial sepsis due to cecal ligation and puncture (CLP) [13‐15] has produced equivocal results regarding sensitization to VILI. Mechanical ventilation with injurious high VT (30–40 ml/kg) exacerbated 48 h of CLP-induced lung injury in rats [16]; shorter periods of CLP were not associated with subsequent exacerbation of VILI in mice [17] or rats [9, 18] although such overall injury was accelerated in the latter. Lower VT (15–20 ml/kg) did not exacerbate CLP-induced ALI in intact rats [9] or isolated perfused rat lungs [19]. Since both CLP and VILI have a common TLR4-mediated pathway to inflammation and injury [5, 20‐22] it seems plausible that mechanical ventilation could exacerbate CLP-mediated events within the lung and thus differences are likely to be secondary to variables in experimental protocols (degree of preexisting lung injury, volume and duration of ventilation) as suggested by Yehya et al. [18].

In the current study, we examined the effect of prolonged (6 h), otherwise noninjurious [4] moderate VT ventilation (MTV) in mice with preexisting mild ALI after CLP (12 h). We focused on a novel WNT1 inducible secreted protein (WISP1 or CCN4; also referred to as WNT1 inducible signaling protein-1)–integrin β5 pathway of TLR4-mediated pulmonary inflammation and injury in this two-hit model as we previously noted key roles for WISP1 and beta integrins in VILI [21] and CLP [22, 23] mediated ALI. The mechanism by which WISP1 acts in CLP and/or VILI remains unclear. WISP1 appears, however, to be a modulator of TLR4–CD14 signaling and is known to signal through integrins [24] although the precise integrin is unclear. Although several integrins are important in ALI [25], integrin β5 is a central regulator of increased permeability in VILI [26] and CLP [27].

Thus, the overall hypotheses of this study were that: prolonged ventilation with otherwise noninjurious moderate VT exacerbates ALI in extrapulmonary sepsis; WISP1 and integrin β5 contribute to this two-hit model (i.e., CLP + MTV); and TLR4 is central to the WISP1–TLR4–integrin β5 proinflammatory pathway.

CLP alone led to modest lung injury as demonstrated by histology (Fig. 1a, b) and a significant increase in alveolar-capillary permeability (Fig. 1c, d). MTV alone had no impact on lung injury or permeability, but when applied after CLP it markedly enhanced both the lung injury score and alveolar-capillary permeability. The histopathologic and permeability changes in the two-hit model were completely abrogated in TLR4^−/− mice and partially (but significantly) reduced in cohorts of mice receiving antibodies to either WISP1 or integrin β5. In wildtype mice, we noted that: MTV did not affect intrapulmonary levels of either WISP1 or integrin β5; CLP led to small but significant increases in either WISP1 or integrin β5; and the two-hit model increased either of these molecules 2–3× more than CLP alone (Fig. 2a, b). We previously noted that high VT ventilation increases in WISP1 were abrogated in TLR4^−/− mice [21] and we now note (Fig. 2c) that CLP-induced increases in integrin β5 are abrogated in TLR4^−/− mice.

CLP increased circulating cytokines and chemokines whereas MTV alone did not; the combination of MTV and CLP, however, caused levels of all four mediators to progressively rise above levels measured with CLP alone (Fig. 3). Deletion of TLR4 prevented increases in cytokines and chemokines in the two-hit model and inhibition of WISP1 or integrin β5 blocked further increases induced by MTV (Fig. 3).

CLP significantly increased neutrophil influx in the lung while MTV had no impact; CLP and MTV further significantly increased neutrophil immigration (Fig. 4) that was abolished in TLR4 null mice. Blocking WISP1 or integrin β5 partially prevented the increase in the percentage of PMN induced in the two-hit model compared to combined CLP and MTV in mice receiving control IgG antibody.

We sought further evidence that MTV directly enhanced inflammation in the lungs of septic mice by measuring levels of activated JNK, p38 and ERK MAP kinase. Six hours of MTV had no effect on MAP kinase activation but significantly promoted MAP kinase activation in mice previously subjected to CLP (Additional file 3: Figure S3). TLR4 deletion prevented the increases in MAPK activation in CLP-treated and CLP + MTV-treated mice. Blocking WISP1 or integrin β5 also prevented the increase in MAP kinase phosphorylation induced by MTV in CLP mice.

To explore the mechanism of integrin β5 upregulation by TLR4, we exposed peritoneal macrophages (PM) to LPS and found that ultrapure LPS induced a time and concentration-dependent increase in surface integrin β5 expression (Additional file 4: Figure S4). The increase in integrin β5 expression was TLR4 and MyD88 dependent, but TRIF independent (Additional file 5: Figure S5). We also found that integrin β5 upregulation by LPS was NF-κB dependent (Additional file 6: Figure S6).

LPS-induced increases in IL-6, TNF-α, MIP-2 and MCP-1 in medium of isolated PM were evident within 4–10 h and the addition of rWISP1 induced further increases in all four mediators (Fig. 5). No increase in any of the mediators could be seen in PM from TLR4^−/− mice. Suppression of integrin β5 expression with siRNA (Additional file 7: Figure S7) prevented the WISP1-induced enhancement of LPS-induced IL-6, TNF-α, MIP-2 and MCP-1 released by PM (Fig. 5).

In the current study, MTV did not cause ALI, but exacerbated CLP-mediated increases in alveolar-capillary permeability and indices of pulmonary inflammation (histopathology, cytokines, chemokines, neutrophil influx and activation of MAPK) in wildtype mice: the effects of this two-hit model were completely abrogated in TLR4 null mutants and partially inhibited by neutralizing antibodies to either WISP1 or integrin β5. In PM, activation of TLR4 led to an increase in integrin β5 expression and rWISP1 increased LPS-induced cytokine release in PM that could be inhibited by silencing either TLR4 or integrin β5. Collectively, these data show: that prolonged MTV ventilation exacerbates ALI caused by extrapulmonary sepsis; and an important positive feedback role for the WISP1–integrin β5 pathway in TLR4-mediated exacerbations to this two-hit model.

Mechanical ventilation with high VT is well known to injure healthy lungs and a consensus from experimental and clinical conditions supports the hypothesis that mechanical ventilation can worsen injury in previously damaged lungs [1, 11, 29‐31]. Although systemic [11] endotoxin interacts with mechanical ventilation in producing ALI [16], there is a lack of consensus from studies directed at identifying extrapulmonary bacterial sepsis as a sensitizing condition to VILI. CLP remains the gold standard of experimental polymicrobial sepsis [15] but ventilating rodents after CLP has produced equivocal results regarding sensitization to VILI [9, 18, 19]. Nin et al. [16] reported that mechanical ventilation aggravated CLP-induced multiorgan dysfunction in rats but the investigators used prolonged CLP (24–48 h) and high VT (35 ml/kg). Others have used lower VT ventilation after CLP and did not observe worsening of lung injury [9]. Uematsu et al. [17] noted that high VT ventilation (40 ml/kg) after CLP increased mediator release but did not affect pulmonary function. Yehya et al. [18] showed that high VT (30 ml/kg) accelerated lung injury secondary to previous CLP in rats but the endpoints of injury (lung compliance, pulmonary edema, oxygenation and computed tomography of micro-CT scans) reached the same pathophysiology as from CLP alone. These authors highlighted subtleties in the degree of initial injury with CLP and the magnitude and duration of mechanical ventilation, and speculated that prolonged ventilation with lower VT may indeed enhance lung injury after CLP. In this regard, we unequivocally note that otherwise noninjurious MTV for a prolonged (6 h) time exacerbated underlying CLP-induced ALI. Indeed, whether CLP in itself causes ALI in mice is conjectural. Iskander et al. [32] clearly showed that pulmonary injury after CLP in mice cannot be considered the etiology of death in the acute phase. The authors did note that a severe model of sepsis with significant mortality could show signs of lung injury. We reported previously [22, 23, 28] that the current model of CLP was associated with 80% mortality in 72 h and this presumably accounted for the modest but significant ALI noted at the earlier time period in the current study. Furthermore, we noted that 6 h of MTV (10 ml/kg) without PEEP was void of significant lung injury. In this regard, our model of effects of mechanical ventilation reproduces previous experience of lack of injury with MTV (10 ml/kg, zero PEEP) [4, 28], and was somewhat similar to the results from Hegeman et al. [33] who noted little evidence of ALI in mice after 5 h of mechanical ventilation with 7 ml/kg and 3 cmH₂O of PEEP or 5 h of MV with 15 ml/kg and zero PEEP. These authors did report VILI under both conditions at a longer time period (12 h) [33].

We [21, 34] and others [5, 35] have noted the important role of TLR4 in experimental VILI. We also noted [22, 23, 28] that a considerable component of pathophysiology, inflammation and injury of CLP was due to TLR4 activation (and CD14). The injury, neutrophil sequestration and inflammation due to combined effects of CLP and MTV in the current study were all abrogated in TLR4 null mice.

WISP1 is a secreted matricellular protein involved in cell adhesion, migration, differentiation, proliferation and survival [36]. From an unbiased haplotype association mapping in inbred strains of mice, we identified WISP1 as a candidate gene associated with VILI [21]. We subsequently identified a role for WISP1 in CLP-induced ALI [22, 23]. WISP1 was first noted in the lung to be a component of bleomycin-induced lung injury and fibrosis [37], and subsequently has been reported to be important in epithelial–mesenchymal transition [38], airway remodeling [39] and proliferation of fibroblasts in the context of lung fibrosis [40]. In-vitro mechanical stretch of type II epithelial cells activated innate immunity and increased WISP1 expression, providing fundamental support for its involvement in VILI [41]. We noted that intrapulmonary WISP1 is elevated in VILI [21], CLP [22, 23], combined poly(I:C) and mechanical ventilation [42], and in the current report in combined CLP and MTV; neutralizing antibodies to WISP1 partially reduced lung injury and inflammation in all of these conditions. The potential convergence of WNT/β-catenin signaling and WISP1 adds to its importance in VILI [43] as well as WNT-mediated lung epithelial cell repair [44].

The mechanism by which WISP1 acts in CLP and/or VILI remains unclear. It appears, however, to be a modulator of TLR4–CD14 signaling and is known to signal through integrins [24]. WISP1 coimmunoprecipitated with active, glycosylated TLR4 in lungs of mice subjected to high VT ventilation and rWISP1 augmented LPS-induced TNF-α release in a TLR4–CD14-dependent fashion in PM [21]. The RGD peptide-sensitive response of intact mouse lungs to CLP and the coimmunoprecipitation of WISP1–integrin β6 in lungs of these mice [23] suggested the presence of the WISP1–integrin β6 pathway in mediating TLR4-dependent inflammation and injury. We noted an obligatory role for integrin β3 in WISP1-mediated release of TNF-α in PM [22] and an important role for integrin β3 in polymicrobial sepsis and combined injury from poly I:C instillation and mechanical ventilation [42]. In the current study we noted that siRNA to integrin β5 reduced WISP1-mediated release of multiple cytokines in PM and integrin β5 contributed to lung inflammation and injury with CLP and MTV. Identifying the precise integrin β subunit involved in complex lung injury and signaling in isolated macrophages is complicated by: the multitude of β subunits; the nondiscriminatory inhibition by RGD peptides; and the promiscuity of WISP1 regarding interactions with αVβ integrin subunit receptors [45]. We focused on the interaction of WISP1 and integrin β5 because: others noted that WISP1 induces IL-6 production through integrin β5 receptor in human synovial fibroblasts [46]; and although several integrins are important in ALI [25], integrin β5 is a central regulator of increased permeability in VILI [26] and CLP [27].

We limited our phenotyping of lung injury to inflammation, alveolar-capillary permeability and histopathologic changes, and did not assess lung mechanics or evolution of changes in gas exchange. As such, fundamental issues of alveolar overdistension and physiologic consequences in the current study remain conjectural. It is noteworthy, however, that we recently reported [28] that low VT (6 ml/kg; 6 h) mechanical ventilation was protective after CLP (6 h). We relied on neutralizing antibodies to assess the role of WISP1 and integrin β5 in a two-hit model and partial effects noted in the study may have been secondary to incomplete deletion. For pragmatic reasons, we used PM as a surrogate for the likely cell of interest (alveolar macrophage) and future studies validating these observations with alveolar macrophages, a more challenging cell to isolate in sufficient number, are necessary.

In the current study, we provide evidence that mild lung injury secondary to extrapulmonary sepsis can sensitize intact mouse lung to subsequent prolonged MTV. This two-hit model is TLR4 sensitive and has important proinflammatory contributions from both WISP1 and integrin β5. In isolated PM, we further defined the nature of WISP1 signaling and identified a requisite TLR4-dependent activation of integrin α_Vβ5 and MyD88–NF-κB pathway inflammatory mediator biosynthesis. This two-hit model provides relevant new information regarding unresolved issues of a common risk factor for ARDS (systemic sepsis) and sensitization to often-used MTV.