Pulmonary Epithelial TLR4 Activation Leads to Lung Injury in Neonatal Necrotizing Enterocolitis

This article is featured in In This Issue, p.679

Necrotizing enterocolitis (NEC) is the leading cause of death from gastrointestinal disease in premature infants, and it is characterized by the sudden onset of intestinal necrosis leading to death in nearly a third of cases (1–3). The mortality from NEC has increased during the past decade due to an increase in the overall survival of extremely premature infants (4), illustrating the importance of understanding the underlying mechanisms and causes of morbidity in patients with this disease. One of the most important long-term health sequela associated with NEC is the development of severe inflammatory lung disease (5), which is more severe than the lung disease that develops in premature infants in the absence of NEC (6, 7). Importantly, however, the mechanistic steps that link the development of NEC with the development of lung injury remain largely unexplained.

In seeking to understand the factors leading to the development of NEC-associated lung disease, we and others have sought to understand the events that lead to the development of NEC in the first place (8–13). In this regard, it has been established that the development of NEC requires activation of the LPS receptor, namely TLR4, on the lining of the intestinal epithelium (8, 14, 15). Genetic or pharmacologic inhibition of TLR4 prevents NEC in mice (13, 16–19), and the expression of TLR4 in the intestinal epithelium is higher in the premature as compared with the full-term mouse and human infant (11, 15), explaining in part the reasons for which the premature infant is at risk for NEC development. Furthermore, we recently described that TLR4 plays a critical role in the regulation of epithelial differentiation via effects on Notch signaling in the intestinal stem cells (8), providing insights as to why TLR4 is higher in the premature developing gut as compared with the full-term gut. Thus, in the postnatal period, persistently elevated TLR4 expression on the still premature intestinal epithelium interacts with colonizing microbes and causes a proinflammatory response leading to mucosal injury (14, 15), whereas TLR4 signaling on the endothelium leads to impaired gut perfusion and mucosal death (20). TLR4 has also been shown to be expressed in the lung (21–23), where it may either contribute to or protect from the development of lung disease, a seeming contradiction that has not been fully resolved, in part due to the lack of mice deficient in TLR4 on the pulmonary epithelium.

We now hypothesize that TLR4 signaling on the lung leads to the development of NEC-associated lung disease, potentially through interaction with gut-derived TLR4 ligands. In support of this hypothesis, we reveal that the development of NEC-associated lung injury in human tissue and mouse models is associated with increased TLR4 expression and signaling in the lung. By using mice that specifically lack TLR4 on the pulmonary epithelium, we further show that NEC-associated lung disease requires the activation of TLR4 on the pulmonary epithelium by the gut-derived TLR4 ligand high-mobility group box 1 (HMGB1), which leads to the recruitment of neutrophils through upregulation of the chemoattractant CXCL5. Strikingly, the aerosolized delivery of a novel TLR4 small molecule inhibitor reverses these effects and prevents NEC-associated lung disease in mice. Taken together, these findings raise insights into the development of NEC-induced lung injury, suggesting the possibility that novel TLR4-targeted strategies may provide therapeutic approaches for this devastating complication of NEC.

The human bronchiole epithelial cell line HBE 135-E6E7 was obtained from the American Type Culture Collection and modified to be deficient in the Tlr4 gene by transduction of lentiviral particles containing The RNAi Consortium TLR4 short hairpin RNAs (shRNAs). Lentiviral particles were generated using the four-plasmid lentiviral packaging system (Invitrogen) and the The RNAi Consortium TLR4 shRNA clone (catalog no. RHS4533-EG7099, GE Healthcare Dharmacon) using permissive HEK293 cells as we have previously described (8). In control experiments, HBE 135-E6E7 cells were transduced with lentivirus particles containing scrambled shRNA. Stable integration of lentiviruses in HBE cells was obtained by selection using puromycin-containing media (5 μg/ml), and knockdown of the gene of interest was verified by RT-PCR. Where indicated, cells were treated with ultrapure LPS (Escherichia coli 0111:B4 purified by gel filtration chromatography, >99% pure; Sigma-Aldrich, 6 h at 25–50 μg/ml as indicated) or purified HMGB1 (rHMGB1, the gift of Dr. Kevin Tracey [The Feinstein Institute for Medical Research], 6 h at 2.5 μg/ml).

Sources of Abs and other reagents were as follows: cleaved caspase-3 (Cell Signaling Technology), DAPI (Invitrogen), inducible NO synthase (iNOS; BD Biosciences), myeloperoxidase (MPO; Thermo Scientific), human CXCL5 (Abcam), mouse CXCL5 (Cedarlane Laboratories), inhibitory anti-CXCL5 (R&D Systems), Ly6G 1A8 (BioLegend), CD11B-PE (BioLegend), and Ly6G-FITC (BioLegend). The novel TLR4 inhibitor compound 34 (C34; 2-acetamidopyranoside, C₁₇H₂₇NO₉, molecular mass 389 Da) was described by our group recently and was synthesized as in Hackam and colleagues (24, 25).

The animal experiments described in these studies were approved by the University of Pittsburgh Animal Care and Use Committee (protocol 12040382) and by Johns Hopkins University Animal Care Committee (protocol M014M362) and were performed according the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.

All human intestinal tissue was obtained and processed as discarded tissue via waiver of consent with approval from the University of Pittsburgh Institutional Review Board (protocols 0606072 and PRO11110007) and in accordance with the University of Pittsburgh anatomical tissue procurement guidelines. Intestinal samples were obtained from human premature neonates undergoing resection for NEC, at time of stoma closure, as we have performed previously (14, 15). In the present study, the mean age of patients with NEC that underwent intestinal resections was found to be lower than that of the controls, but because the generally accepted practice is to perform stoma closure quite soon after the original operation, most control bowel was still obtained from infants prior to their due date (mean age NEC, 26 ± 4 versus 38 ± 2 wk, p < 0.05). Human infant lung samples were obtained and processed at autopsy from either patients with NEC or age-matched infants that did not develop NEC or sepsis and died of unrelated conditions that did not affect the lungs, with approval from the University of Pittsburgh Institutional Review Board (Committee on Oversight of Research Involving the Dead no. 491) and in accordance with the University of Pittsburgh anatomical tissue procurement guidelines. All samples were de-identified via an independent honest broker assurance mechanism (approval no. HB 043) and transferred to Johns Hopkins University under the guidance of material transfer agreement approval (no. A26558) for analysis.

C57BL/6, Scgb1a1^Cre-ERT2 (B6N.129S6(Cg)-Scgb1a1^{tm1(cre/ERT)Blh}/J), and Villin-cre (B6.Cg-Tg(Vil-cre)997Gum/J) were purchased from The Jackson Laboratory. Hmgb1^loxP mice were the gift of Eugene Chang (University of Chicago) and were then bred with Villin-cre mice to create mice lacking HMGB1 from the intestinal epithelium (Hmgb1^ΔIEC) as we have described (26). Tlr4^−/− mice and mice in which TLR4 was selectively deleted from the intestinal epithelium (Tlr4^ΔIEC) were generated in our laboratory as recently described (8). Mice in which TLR4 was specifically deleted from the airway epithelium (TLR4^ΔBAEC) were generated by breeding Tlr4^loxP mice (8) with Scgb1a1^cre-ERT2 mice (The Jackson Laboratory). The progeny were found to lack TLR4 in the airway epithelium as determined by PCR, and to lack an inflammatory response to the intratracheal instillation of LPS (Supplemental Fig. 1). Mice in which TLR4 was expressed only in the intestinal epithelium were generated as described (11, 15).

NEC was induced in 8-d-old mice as we have described and validated in previous reports using formula gavage (Similac Advance infant formula [Abbott Nutrition]/Esbilac (PetAg) canine milk replacer, 2:1) five times per day and hypoxia (5% O₂, 95% N₂) for 10 min in a hypoxic chamber (Billups-Rothenberg) twice daily for 4 d, supplemented with enteric bacteria obtained from an infant with NEC requiring surgery (14, 15, 27, 28). This protocol results in the development of patchy necrosis and cytokine induction that mimics that seen in human NEC (14).

For bronchoalveolar lavage, mice were euthanized using CO₂ and the chest was opened by midline incision, and the lungs were lavaged using PE-90 tubing inserted into the exposed trachea using 0.5 ml sterile saline per lavage (total lavage vol of 2 ml/mouse). The lavage fluid was then centrifuged for 10 min at 200 × g and the cell pellet was resuspended in 1 ml HBSS. The cells were then counted by flow cytometer (Accuri C6, BD Biosciences), and the percentage of neutrophils was determined by Ly6G and CD11b double-positive cell counts.

For tracheal administration of reagents, mice were first anesthetized by inhalation of isoflurane and then administered (by nasal instillation) 50 μg LPS (E. coli 055:B5; Sigma-Aldrich), rHMGB1 (4 μg/g), isotope control rat IgG (25 mg/kg), rat anti-Ly6G mAb (clone 1A8, catalog no. 127620, BioLegend), and rat anti-CXCL5 Ab (catalog no. MAB 433, R&D Systems). Each Ab was dissolved in 50 μl saline and was administered either on the day of the experiment (endotoxin inhalation) or once a day during the NEC model starting the day prior to NEC induction. C34 (10 μg/kg) was administered via aerosol 1 d before the NEC model started and then one dose every day for the duration of the model.

Quantitative real-time PCR (qRT-PCR) was performed using the Bio-Rad CFX96 real-time system as described previously (15) using the primers listed in Table I. Total RNA was isolated from samples of either lung or the terminal ileum of mice, or from specimens resected from human infants during surgery. The expression levels of the proinflammatory cytokines were measured relative to the housekeeping gene RPLO. Immunofluorescence staining was performed on 4% paraformaldehyde-fixed 5-μm-thick paraffin sections following Ag retrieval, as described (15, 29), and was assessed on a Zeiss LSM710 confocal microscope. SDS-PAGE was performed as described by Hackam and colleagues (15, 30), in which intestinal samples were collected in RIPA buffer (BP-115, Boston BioProducts) containing phosphatase and protease inhibitor mixture (BP-480, Boston BioProducts; PIC02, Cytoskeleton) and homogenized with homogenizing beads on a BeadBlaster (BenchMark) and centrifuged at 4°C at 16,000 × g for 5 min. Supernatants were collected and equal amount proteins were loaded on SDS-PAGE gel before transferring to cellulous membrane for Ab detection.

Single-cell suspensions from bronchoalveolar lavage fluid or from mouse lung isolates were subjected to flow cytometry. To isolate single-cell suspensions from mouse lung, the lungs were minced and incubated in 50 μg/ml Liberase solution (Roche) for 30 min at 37°C and agitated at 750 rpm. The cells were then disassociated with an 18-gauge needle. The tissue digest was passed through a 40-μm cell strainer into a tube with wash buffer and centrifuged at 400 × g at 4°C for 5 min. The pellet was then resuspended in 50 ml 1% BSA (VWR International) in PBS and centrifuged at 400 × g at 4°C for 5 min and the supernatant was discarded. The cell pellet was resuspended at 2.5 × 10⁷ cells/ml in FACS buffer. Single-cell suspensions were then incubated with anti-CD16/CD32 (BD Biosciences) to block Fc receptor binding (20 min, 4°C). Cells were pelleted by centrifugation and resuspended in primary fluorochrome-conjugated Abs (see above) in ice-cold FACS buffer. After washing with 1% BSA in PBS, at least 100,000 live cells per sample were collected for analysis on a BD Accuri C6 flow cytometer, in which neutrophils were counted as Ly6G and CD11b double-positive after excluding dead cells, which were 7-aminoactinomycin D⁺. Data analysis was performed using FlowJo software as in Egan et al. (31).

Sandwich ELISA analysis was performed according to manufacturers’ instructions (human or mouse CXCL5 DuoSet, R&D Systems; human and mouse HMGB1, IBL International). Briefly, capture Ab was incubated on 96-well flat-bottom plates overnight. Plates were washed and blocked with 5% BSA (1 h, room temperature) and samples were added to the plate, incubated overnight (4°C), washed extensively, and then incubated with biotinylated detection Ab (2 h, room temperature). Following washes, streptavidin–alkaline phosphatase was added to the wells and the enzymatic reaction was stopped after 30 min by the addition of an equal volume of 0.2 N sulfuric acid, and the color change was read on a spectrophotometer (450 nm, Molecular Dynamics). Data were normalized to the standards according to manufacturers’ instructions and quantified using GraphPad Prism software.

Where indicated, data were analyzed for statistical significance by a two-tailed Student t test or ANOVA using GraphPad Prism 6 software. Statistical significance was determined as having a p value <0.05, and data are represented as mean ± SEM as indicated. All experiments were repeated at least in triplicate, with at least five pups per group for experimental NEC assessed.

We first sought to characterize the development of lung injury in a mouse model of NEC and to compare the pathologic features and biochemical findings with the human disease in premature infants. As shown in Fig. 1A, the appearance of the small intestine in NEC in premature infants is characterized by a loss of villous architecture and sloughing of the intestinal villi, consistent with the original descriptions of this disease (32). The findings in the intestinal mucosa of humans with NEC were very similar to those observed in TLR4-expressing wild-type mice, in which NEC was induced by the exposure to hypoxia and the administration of formula (Fig. 1B), as we have described and validated (14, 15, 20, 27, 28). In both the human disease and the mouse model, we also detected significantly increased expression of proinflammatory factors in the intestinal mucosa, including IL-8 and iNOS (Fig. 1C, 1D). Importantly, in humans and mice with NEC, we also observed the presence of significant pulmonary injury, characterized by airspace destruction, the influx of neutrophils (Fig. 1A, 1B), and the expression of the proinflammatory molecules iNOS and IL-8 in the lung as determined by qRT-PCR (Fig. 1C, 1D). Of note, the expression of TLR4 expression was significantly higher in the intestinal and pulmonary epithelium of mice and humans with NEC as compared with controls (Fig. 1E), and it was also significantly greater in the mouse lung at the age in which mice are susceptible to NEC development (days 5–10 postnatally). Taken together, these findings suggest that signaling between TLR4 in the intestine and the lung could influence the development of lung injury in the setting of NEC. In support of this possibility, we now show that mice lacking TLR4 in the intestinal epithelium (TLR4^ΔIEC), whose generation we have previously reported (8), are protected from the development of intestinal injury and do not display lung injury when exposed to the NEC model (Fig. 2A). Furthermore, mice expressing TLR4 only on the intestinal epithelium (TLR4^IEC-over), whose generation we have also previously reported (29), were found to develop intestinal injury after exposure to the NEC model, yet lung disease did not occur (Fig. 2). These findings imply that TLR4 signaling on the intestine and lung are required for NEC-associated lung injury, as explored in greater detail below.

To determine specifically whether TLR4 signaling in the lung is required for the development of pulmonary injury in mice with NEC, we next generated mice that selectively lack TLR4 on the bronchoalveolar epithelium as described in 2Materials and Methods (Supplemental Fig. 1), herein called TLR4^ΔBAEC for clarity, and subjected these mice to experimental NEC. The TLR4^ΔBAEC mice were healthy and fertile, displayed no obvious lung phenotype, reproduced at expected Mendelian ratios, and did not induce proinflammatory cytokines in response to intratracheal LPS as compared with wild-type mice, confirming the success of the deletion strategy (Supplemental Fig. 1). Importantly, when subjected to the model of experimental NEC, TLR4^ΔBAEC mice were found to develop significant intestinal injury, yet they were largely protected from the development of lung injury when compared with wild-type mice subjected to the experimental model, as manifested by reduced histological evidence of injury, reduced infiltration of MPO⁺ neutrophils (Fig. 2A), and decreased expression of the proinflammatory molecules KC and iNOS (Fig. 2B). Along with the findings in Fig. 1, these results illustrate that the development of NEC-associated lung injury in mice requires TLR4 signaling in the lung epithelium. We therefore next sought to evaluate potential TLR4 ligands involved.

At sites of intestinal inflammation, dying cells release proinflammatory molecules such as HMGB1, which is an endogenous ligand for TLR4 (33–36) and which has been linked to the development of lung injury in other settings but not in the setting of neonatal inflammation (26, 37, 38). As shown in Fig. 3A, NEC in human infants and in wild-type mice was associated with significant HMGB1 release into the systemic circulation, consistent with prior reports linking HMGB1 expression with NEC in mice (39, 40). We now show that circulating HMGB1 levels were not increased in mice lacking TLR4 in the intestinal epithelium (TLR4^ΔIEC mice) upon exposure to the NEC model, indicating that the release of HMGB1 into the circulation in NEC is dependent on TLR4 signaling in the intestinal epithelium (Fig. 3B). To evaluate whether HMGB1 release from the intestine could play a role in NEC-associated lung injury, we first administered rHMGB1 directly into the lungs of wild-type mice and observed the development of significant lung inflammation as manifested by the destruction and inflammation of the airways, the induction of expression of proinflammatory KC and iNOS mRNA, and an influx of neutrophils in the lung (Fig. 3C). Importantly, the effects of rHMGB1 on the induction of lung injury required the presence of TLR4 on the lung epithelium, as the administration of rHMGB1 to TLR4^ΔBAEC mice, which lack TLR4 on the lung epithelium, did not develop significant lung inflammation (Fig. 3C). To determine whether gut-derived HMGB1 could play a role in inducing lung injury in the setting of NEC, we next generated mice that lack HMGB1 in the intestinal epithelium (HMGB1^ΔIEC), which were healthy and fertile, as we recently described (26). As shown in Fig. 3B, when subjected to experimental NEC, HMGB1^ΔIEC mice displayed significantly reduced HMGB1 in the circulation compared with wild-type mice, consistent with the notion that the release of HMGB1 into the circulation in NEC is largely gut derived. Strikingly, when subjected to experimental NEC, although HMGB1^ΔIEC mice still developed significant inflammation in the intestine (Fig. 4A), examination of the lungs from HMGB1^ΔIEC mice with NEC revealed significantly less lung inflammation as compared with wild-type mice subjected to the NEC model (Fig. 4B). Importantly, note that the administration of neutralizing anti-HMGB1 Ab to wild-type mice that had been exposed to experimental NEC also resulted in significantly reduced lung injury as compared with mice with NEC that were administered an equimolar concentration of nonspecific IgG (Fig. 3D), illustrating the importance of circulating HMGB1 in the development of NEC-associated lung injury. Taken together, these studies indicate that gut-derived HMGB1 is required for the induction of lung injury in NEC. We next sought to evaluate the potential mechanisms by which this could occur.

In the next series of studies, we sought to determine how HMGB1 release from the intestine could lead to inflammation in the lung in the setting of NEC in newborn mice. To do so, we focused on our earlier observation that NEC in mice and humans was associated with an influx of neutrophils into the lung (Fig. 1A, 1B). To investigate whether the neutrophil influx could play a role in the induction of lung injury in NEC, we next assessed the extent of NEC-associated lung injury after neutrophil depletion using anti-Ly6G Ab. As shown in background experiments described in Supplemental Fig. 2, anti-Ly6G Ab significantly reduced the influx of pulmonary neutrophils in LPS-treated mice compared with control mice that were administered IgG, confirming the efficacy of the neutrophil inhibitory strategy. Importantly, neutrophil depletion with anti-Ly6G Ab significantly reduced the degree of lung injury in neonatal mice that were subjected to experimental NEC, as manifested by reduced histological inflammation of the lung (Fig. 5A), reduced MPO expression within the airspace, and significantly reduced lung expression of the proinflammatory genes KC and iNOS as compared with mice administered nonspecific IgG and induced to develop NEC (Supplemental Fig. 3). It is noteworthy that the neutrophil depletion strategy did not reduce the degree of gut inflammation or severity of NEC (Fig. 5B, Supplemental Fig. 3), a finding that is consistent with the lack of effect of neutrophils in inducing intestinal injury in NEC that we have previously observed (15) and that excludes the possibility that the protection in the lung was merely a result of attenuation of NEC severity in the intestine.

To determine whether intestinal-derived HMGB1 could induce lung injury through the recruitment of neutrophils into the lung, we next focused our attention on CXCL5, which is a major neutrophil chemoattractant on the pulmonary epithelium (41). The expression of CXCL5 was significantly increased in the lungs of human premature infants with NEC as compared with control infants, and it was also significantly greater in the lungs of wild-type neonatal mice with experimental NEC compared with age-matched control mice, as assessed both by qRT-PCR (Fig. 5C) as well as by immunostaining (Fig. 5D). Importantly, the induction of CXCL5 in the lungs of mice and humans with NEC was largely due to the effects of gut-derived HMGB1, which activated TLR4 on the pulmonary epithelium, as demonstrated by the following lines of evidence: 1) the induction of NEC resulted in increased expression of CXCL5 in the lungs of wild-type mice, but not in HMGB1^ΔIEC mice (Fig. 5C); 2) treatment of wild-type human airway epithelial cells in culture with HMGB1 led to a significant release of CXCL5 measured by ELISA as shown in Fig. 5C, which was not seen in TLR4-deficient cultured airway epithelial cells (genetic and phenotypic evidence for TLR4 knockdown in the human airway epithelial cells is shown in Supplemental Fig. 4); and 3) the instillation of rHMGB1 into the airways resulted in the induction of CXCL5 expression in the airway epithelium of wild-type mice but not TLR4^ΔBAEC mice (Fig. 5E), confirming the importance of pulmonary TLR4 in mediating the induction of CXCL5 expression in response to HMGB1. As shown in Fig. 6A, rHMGB1 administration caused an increase in the percentage of neutrophils in the bronchoalveolar lavage fluid of saline-treated mice, which was reduced by preadministration of anti-CXCL5 Abs, whereas anti-CXCL5 Ab significantly reduced the degree of NEC-associated lung injury in newborn mice, without affecting the degree of intestinal injury (Fig. 6B, 6C). Taken together, these findings reveal that HMGB1 release from the injured intestinal epithelium in NEC leads to lung inflammation via the TLR4-mediated induction of CXCL5 and subsequent recruitment of neutrophils. We therefore sought to determine whether this pathway could be manipulated for potential therapeutic benefit.

Having shown in Fig. 1 that TLR4 signaling in the lung is required for the induction of NEC-associated lung injury, we next sought to determine whether we could inhibit TLR4 pharmacologically in the lung to prevent lung inflammation in the setting of NEC. In approaching this possibility, we have recently identified a novel family of TLR4 inhibitors that are highly effective and nontoxic at low concentrations (24, 25). Our lead compound, C34, was recently identified to be a 2-acetamidopyranoside (molecular mass of 389 Da) with the formula C₁₇H₂₇NO₉ (24, 25). We now show that the administration of aerosolized C34 significantly inhibits TLR4 in the lung epithelium, as manifested by the protection from LPS-induced proinflammatory cytokine induction (KC and iNOS) in the lungs of newborn mice (Fig. 7A). Importantly, the daily administration of aerosolized C34 to mice that were subjected to experimental NEC significantly improved the histological appearance and reduced the degree of proinflammatory cytokine expression in the lungs as compared with mice that were subjected to experimental NEC and received aerosolized saline alone (Fig. 7B–D). Furthermore, the administration of aerosolized C34 significantly reduced the induction of CXCL5 expression in the mouse lung (Fig. 7C) and abrogated the recruitment of neutrophils into the mouse lung (Fig. 7E), consistent with the interruption of the TLR4-mediated neutrophil recruitment cascade that we revealed to be important in NEC-associated lung injury. It is noteworthy that the aerosolized administration of C34 did not block the degree of intestinal injury in mice with NEC (Fig. 7B, 7D), reflecting the fact that the aerosolized route does not deliver effective dosage of the TLR4 inhibitor to the intestine. Taken together, these findings illustrate a link between TLR4 signaling in the gut in the pathogenesis of NEC leading to lung damage that can be reversed through TLR4 inhibition and interrupting the HMGB1–CXCL5 signaling cascade. This pathway is described in schematic form in Fig. 8.

The development of lung injury in premature infants represents a major cause of long-term morbidity and develops in nearly half of all infants born before 36 wk gestation (42). Importantly, in the setting of inflammatory processes such as NEC, the development of lung injury is more severe and lasts for a greater duration than does the lung injury that is seen in premature infants in the absence of NEC, although the underlying reasons for this difference have remain largely unexplained (5, 6, 43). To address this directly, we have now developed a mouse model of NEC in which lung injury develops that shares important features with the human condition. Using this model, we have identified a critical role for TLR4 signaling on the pulmonary epithelium in the development of NEC-associated lung disease, and have demonstrated a pathway by which TLR4 signaling on the intestinal epithelium leads to gut-derived HMGB1 release, which activates TLR4 on the lung epithelium, leading to neutrophil recruitment via the induction of CXCR5. We can be confident that the experimental model of NEC that involves hypoxia/formula gavage did not directly injure the lungs, as mice lacking TLR4 in the intestinal epithelium subjected to experimental NEC did not sustain lung injury, an observation that also confirms the importance of TLR4 on the intestinal epithelium in mediating the lung damage (Figs. 1, 2). It is also noteworthy that TLR4 expression is significantly higher in the lungs of mice and humans with NEC as compared with controls that do not have this disease, and that TLR4 expression was significantly higher at the time that NEC develops in the mouse, perhaps providing insights into the particular susceptibility of NEC-associated lung disease in this population. Collectively, these findings extend our understanding of the pathways that lead to the development of lung inflammation in the presence of NEC, and they raise the possibility that airway-targeted TLR4 inhibition could offer benefit to patients to prevent the long-term sequelae of NEC.

The question arises as to how TLR4 activation on the intestinal epithelium leads to HMGB1 release. Data from our prior publications (11, 26, 44) as well as those from other investigators (45, 46) strongly indicate that the release of HMGB1 from the intestinal epithelium after TLR4 activation requires enterocyte apoptosis, that enterocyte apoptosis is induced by the activation of TLR4, and that the mechanisms leading to enterocyte apoptosis require the induction of endoplasmic reticulum stress on the enterocytes. Specifically, we have shown that TLR4 activation on the intestinal epithelium leads to the induction of endoplasmic reticulum stress within the intestinal epithelium, which leads to enterocyte apoptosis (11). Previous authors have shown that cell death by apoptosis is a potent inducer of HMGB1 release (45, 46). In data not including the present study, we have shown that, when cultured enterocytes or newborn mice were treated with reagents to reduce endoplasmic reticulum stress or to block apoptosis pathways prior to the treatment with LPS in the induction of NEC, the release of HMGB1 was blocked and lung injury did not occur. Taken together, these findings suggest that TLR4-induced enterocyte apoptosis is required for the release of HMGB1 and the development of lung injury.

The finding that TLR4 signaling on the pulmonary epithelium is required for the development of lung inflammation in the setting of NEC may appear at first glance to be somewhat counterintuitive, yet in fact it supports an emerging story that TLR4 can exert both protective and injurious roles in the host depending on the cell of expression and the environmental cues present. Specifically, although TLR4 plays a critically important role in host defense and coordination of the adaptive and innate immune responses when expressed on leukocytes (47), exaggerated TLR4 signaling on mucosal surfaces can lead to tissue injury in the liver (48), pancreas (49, 50), and small intestine (8, 51, 52). TLR4 signaling in the lung can also induce proinflammatory signaling in chronic obstructive pulmonary disease (21) and asthma (22), and it has been shown to induce inflammatory changes in response to the endogenous ligand hyaluronan in the setting of acute lung injury (23). Despite these important studies, it has not been possible to reliably determine whether the role for TLR4 signaling in these disease processes acts on the lung epithelium or on other cell types owing to the lack of available mice in which TLR4 is deleted from the pulmonary epithelium. We have now addressed this directly by generating mice lacking TLR4 on the pulmonary epithelium, and have shown that the development of NEC-associated lung injury stems from exaggerated TLR4 signaling on the lung epithelium. We have also identified one potential ligand, namely HMGB1, that can interact with pulmonary TLR4 in mediating the lung injury in the setting of NEC. Further studies will be required to identify other potential TLR4 ligands (either endogenous or exogenous) that could play a role in mediating lung injury in the setting of the inflamed bowel seen in NEC.

Although helpful in providing a mechanistic explanation for the development of NEC-associated lung injury, the present findings may also shed light as to the role of TLR4 on the pulmonary epithelium in the first place. Given that these results may be somewhat specific for the early postnatal lung in which TLR4 expression is elevated at baseline, we now posit that the present findings raise the possibility that TLR4 signaling may play a physiologic role in the regulation of normal lung function in the early newborn period. In comparison, we have shown that TLR4 expression on another epithelial surface, namely the intestinal epithelium, plays a critical role in normal intestinal stem cell differentiation and cell fate through a canonical TLR4–Notch signaling pathway (8). It is conceivable therefore that the pulmonary epithelium is similarly dependent on TLR4 for normal function, perhaps via effects on pulmonary stem cells, and that these effects may in part explain some of the underlying differences between the premature as compared with the full-term lung. Further studies involving the role of TLR4 on cultured lung epithelium may be required to determine the nature and extent of TLR4 on the regulation of lung stem cell growth, signaling, and differentiation.

Perhaps one of the most potentially translational findings of the present study is the observation that the administration of a TLR4 inhibitor directly into the airway had a protective role in NEC-induced lung injury by interrupting the pathways leading to neutrophil recruitment. Given that current therapies for lung injury in the premature population are largely nonspecific (53, 54), this molecule-targeted approach may offer the potential for improvements over current care models. It is also possible that patients who exhibit elevated TLR4 signaling or expression within the lung at baseline may be most likely to benefit from intra-airway TLR4 inhibition. Such strategies will need to be balanced by potentially negative effects on neutrophil function so as not to impair host defense. The inhibition of TLR4 in the airway under conditions in which TLR4 signaling is exaggerated in the premature infant is in some respects akin to the administration of surfactant that restores pulmonary compliance and therefore allows for adaptation, thereby reducing hyaline membrane disease (55). By reversing the deleterious effects of exaggerated TLR4 signaling in the premature lung through TLR4 inhibition, molecular adaption may be achieved, and NEC-induced lung injury may be attenuated or even prevented.

We readily acknowledge that there are several limitations of the present study. For instance, the findings may be applicable largely to the newborn period, before a full complement of immune cells and microbial species has been established. Moreover, although the principal findings were verified in human samples, it is possible that the findings that pulmonary TLR4 activation is required for the induction of neonatal lung disease may be specific to the mouse model of NEC that we used, and may not be broadly applicable to other models, as have been described in rats and piglets (56). Further studies will be required to determine whether TLR4 plays a developmental role in the lung, and whether mutations in TLR4 in the neonatal lung modulate the risk for pulmonary disease development in the premature infant.

In summary, as described in schematic form in Fig. 8, using a novel series of mouse reagents in which TLR4 is deleted from the pulmonary epithelium, we now have identified a novel signaling pathway that exists between the lung and the intestine in mice, which might shed light on the development of a more severe lung injury in premature infants with NEC. The increased TLR4 expression in the lungs of infants who develop NEC correlates with the onset of lung inflammation, and it may help explain the increased susceptibility to lung injury in this population via CXCL5 induction. Collectively, these findings illustrate the role for pulmonary TLR4 in the development of lung injury that occurs in the presence of NEC, highlight a novel link between the gut and the lung in the pathogenesis of this disease, and provide novel therapeutic approaches for this devastating complication of NEC.

The authors have no financial conflicts of interest.