Research ArticleLung Transplantation

The HMGB1-RAGE axis mediates traumatic brain injury–induced pulmonary dysfunction in lung transplantation

See allHide authors and affiliations

Science Translational Medicine  03 Sep 2014:
Vol. 6, Issue 252, pp. 252ra124
DOI: 10.1126/scitranslmed.3009443

This article has a correction, but has also been retracted. Please see:

Sounding the Alarm for RAGE

Only 20% of lungs are transplantable because traumatic brain injury, a major cause of death in organ doors, may induce acute lung injury. High-mobility group box-1 (HMGB1) release from the injured brain likely contributes to acute lung injury in donors by preferentially interacting with receptor for advanced glycation end products (RAGE) in the lung. Blocking the HMGB1-RAGE axis improves lung function in murine donors with traumatic brain injury and after transplant. In translational studies, lungs sourced from donors with high HMGB1 levels had worse pulmonary function after transplant. Targeting the HMGB1-RAGE axis may increase the number of lungs available for transplantation and improve patient outcomes.


Traumatic brain injury (TBI) results in systemic inflammatory responses that affect the lung. This is especially critical in the setting of lung transplantation, where more than half of donor allografts are obtained postmortem from individuals with TBI. The mechanism by which TBI causes pulmonary dysfunction remains unclear but may involve the interaction of high-mobility group box-1 (HMGB1) protein with the receptor for advanced glycation end products (RAGE). To investigate the role of HMGB1 and RAGE in TBI-induced lung dysfunction, RAGE-sufficient (wild-type) or RAGE-deficient (RAGE−/−) C57BL/6 mice were subjected to TBI through controlled cortical impact and studied for cardiopulmonary injury. Compared to control animals, TBI induced systemic hypoxia, acute lung injury, pulmonary neutrophilia, and decreased compliance (a measure of the lungs’ ability to expand), all of which were attenuated in RAGE−/− mice. Neutralizing systemic HMGB1 induced by TBI reversed hypoxia and improved lung compliance. Compared to wild-type donors, lungs from RAGE−/− TBI donors did not develop acute lung injury after transplantation. In a study of clinical transplantation, elevated systemic HMGB1 in donors correlated with impaired systemic oxygenation of the donor lung before transplantation and predicted impaired oxygenation after transplantation. These data suggest that the HMGB1-RAGE axis plays a role in the mechanism by which TBI induces lung dysfunction and that targeting this pathway before transplant may improve recipient outcomes after lung transplantation.


In addition to the lesions caused at the moment of injury, brain trauma can result in secondary damage, which includes a variety of events that take place in the subsequent hours and days after injury. Included in the possible secondary injury types are indirect effects on the pulmonary system including acute respiratory distress syndrome and acute lung injury (ALI). This is particularly relevant in the context of lung transplantation, where most of the donor lungs are procured from brain-dead donors, of which between 40 and 70% have sustained traumatic brain injury (TBI) (1). Of those evaluated, only about 15% are deemed suitable for transplant (2). The mechanisms by which TBI leads to pulmonary dysfunction are poorly understood. Historically, a combination of catecholamine surge–induced pulmonary vascular permeability and production of inflammatory mediators is thought to compromise lung function (3). There is evidence that systemic inflammatory factors cause pulmonary injury and dysfunction (4).

Recent approaches to identifying the pathophysiological mechanisms of ALI have focused on nontraditional proinflammatory mediators and their receptors. In particular is the class of danger-associated molecular patterns (DAMPs; alarmins), which are often associated with sterile inflammatory responses to events such as ischemia or systemic disease. A DAMP of particular interest is high-mobility group box-1 (HMGB1). Although HMGB1 is typically associated with chromatin, it can be quickly released into the cytoplasm after stress, injury, or disease. Depending on the status of the affected cells, cytoplasmic HMGB1 can be passively released into the extracellular space. Alternatively, HMGB1 can be actively released by cells of the immune and nervous systems after injury, inflammation, or disease (57). Progress to date suggests that HMGB1 receptors include Toll-like receptor 4 (TLR4) and receptor for advanced glycation end products (RAGE) (8). Both TLR4 and RAGE are expressed by many cell types including those in the lung (9) and participate in the onset of innate immune inflammatory processes through activation of nuclear factor κB (NF-κB) (10, 11). HMGB1 binding to RAGE has also been implicated in a number of inflammation-associated diseases including cancer, diabetes, epilepsy, and Alzheimer’s disease (1214). It is well established that RAGE is constitutively expressed at high levels in the lung (9), and RAGE ligation leads to sustained activation of NF-κB and increased RAGE expression, which ensure maintenance and amplification of an inflammatory signal (10, 11).

Here, we examined the involvement of HMGB1 and RAGE in TBI-induced ALI in mice whose lungs were used as donors for transplantation. We also used clinical samples to explore the connection between elevated donor HMGB1 and pulmonary dysfunction before and after lung transplantation. Because HMGB1 is known for its contribution to proinflammatory processes associated with injury and organ damage associated with severe sepsis, TBI disruption of the blood-brain barrier and the release of DAMPs could trigger an inflammatory cascade in tissues rich in RAGE receptors, most prominently the lungs (1518). In an effort to characterize the cause of pulmonary dysfunction after TBI and the role of the HMGB1-RAGE axis, we studied lungs obtained from mice either deficient or sufficient in RAGE before being used as donors for transplantation. Local and systemic inflammatory responses as well as pulmonary function in both donors and recipients were analyzed. Translational studies examined the correlation of systemic HMGB1 concentrations and indices of ALI in human lung transplant donors before donor harvest and in recipients after lung transplantation. Our objective was to determine whether TBI-induced inflammatory changes result in pulmonary dysfunction that is dependent on the HMGB1-RAGE pathway.


TBI induces ALI and increases systemic HMGB1

TBI caused substantial changes in the lung architecture of C57BL/6 mice. These animals had evidence of alveolar hemorrhage at 4 and 8 hours after injury compared with sham-injured mice. By the 12- and 24-hour time points, this alveolar hemorrhage largely had resolved (Fig. 1, A to E). However, these lungs did continue to have evidence of ALI as evidenced by proteinaceous debris and neutrophilic infiltration. As scored by a standardized assessment (19), lungs from animals subjected to TBI had higher ALI scores than control lungs at the same time intervals (Fig. 1F). The mean score for sham injury control lungs was 0.09 ± 0.03, whereas scores at the 4-hour time point after TBI yielded a significantly higher score of 0.62 ± 0.11 (P < 0.01). Scores remained significantly elevated at 8, 12, and 24 hours after TBI, with scores of 0.47 ± 0.11, 0.49 ± 0.16, and 0.56 ± 0.15, respectively (P < 0.01). Additionally, the amount of alveolar hemorrhage at the 4- and 8-hour time points corresponded with ALI scores.

Fig. 1. ALI in C57BL/6 mice 4 to 24 hours after TBI.

Representative hematoxylin and eosin (H&E) histology sections of C57BL/6 mouse lungs after TBI at various time points. (A) Control animals with sham TBI. (B) Four hours after TBI with alveolar hemorrhage. (C) Eight hours after TBI with alveolar hemorrhage. (D) Twelve hours after TBI with interstitial neutrophils. (E) Twenty-four hours after TBI with interstitial neutrophils. (F) ALI scores from a blinded pathologist based on a standardized scoring system from the American Thoracic Society (19). Scores are continuous between 0 and 1, with 0 representing no injury and 1 representing severe ALI. (G) Albumin concentrations in BAL fluid after TBI. (H) Systemic concentrations of HMGB1 24 hours after severe TBI in the sham-injured group, as well as in C57BL/6 and RAGE−/− mice 24 hours after TBI. n = 6 to 8 per condition, Student’s t test comparison with the reference sham-injured group, comparisons versus sham-injured group, *P < 0.01.

Consistent with neutrophilia in the lung were data showing that myeloperoxidase (MPO), a neutrophil-derived enzyme, was greater in mice subjected to TBI compared to controls (327 ± 14.7 ng/ml versus 161 ± 12.7 ng/ml, respectively, P < 0.01; fig. S1). Activated caspase-3 staining, a marker for apoptosis that can occur during tissue injury of many types including the lung, was not increased in either C57BL/6 or RAGE−/− TBI mice compared to sham TBI mice (fig. S2).

Albumin was quantitated in bronchoalveolar lavage (BAL) fluid after TBI to determine whether increased vascular permeability was associated with the ALI scores. In healthy mice subjected to sham injury, the BAL albumin concentration was 52.5 ± 11.2 μg/ml. Albumin concentrations were significantly higher at 4 hours after TBI (303 ± 21.0 μg/ml), then declined progressively at 8 hours (301 ± 34.1 μg/ml), 12 hours (212 ± 17.9 μg/ml), and 24 hours (103 ± 22.1 μg/ml) (Fig. 1G). However, compared to controls, all BAL albumin levels were increased significantly at each time point, indicating TBI-induced vascular permeability (P < 0.01).

To determine whether HMGB1 increased after TBI, serum HMGB1 levels from sham-injured wild-type mice, as well as from TBI wild-type mice, were assayed at 24 hours (Fig. 1H). In the sham injury control animals, the mean serum HMGB1 concentration was 1.8 ± 0.4 ng/ml. However, TBI induced significantly higher HMGB1 levels compared to controls (4.1 ± 0.7 ng/ml, P < 0.01). Nevertheless, even higher levels were seen with RAGE−/− mice (7.9 ± 2.6 ng/ml, P < 0.01).

Expression of RAGE and HMGB1 in different tissues

In an effort to confirm that RAGE, a key receptor for HMGB1, is organ-specific, Western blots were performed with homogenates of lung, spleen, liver, and spinal cord from three wild-type mice. High levels of RAGE were present in the lung samples but not the other tissue types (fig. S3). To localize possible serum sources of HMGB1, we profiled HMGB1 in both the lungs and cortical tissue ipsilateral to the TBI in sham and TBI mice. Although nuclear HMGB1 was observed in the lungs of both sham and TBI mice, there was no evidence of the loss of nuclear HMGB1 immunopositive signal (fig. S4). In contrast, cortical tissue associated with the lesion site exhibited a pronounced decrease in the number of cells that were immunoreactive for HMGB1 when compared with adjacent, uninjured cortical areas 24 hours after TBI. The observed loss of cellular HMGB1 in injured tissue did not appear to be associated with TBI-induced cell loss because numerous cells within the lesion site retained nuclear staining with DAPI (4′,6-diamidino-2-phenylindole) (fig. S5).

TBI induces pulmonary dysfunction

Groups of wild-type C57BL/6 mice were subjected to severe TBI injury followed by assessment of systemic oxygenation as determined by PaO2/FiO2 ratios 4 to 24 hours after TBI (Fig. 2A). This ratio divides the arterial oxygenation concentration by the fraction of inspired oxygen, where increasing values reflect improved ability of the lung to absorb oxygen from air. In the sham injury control group, PaO2/FiO2 ratios were normal at 578 ± 5.23. However, 4 hours after TBI, the PaO2/FiO2 ratio decreased to 351 ± 2.55 and continued to decline at 24 hours (111.5 ± 13.09). Pulmonary function testing at the same intervals demonstrated a similar time-dependent decline in compliance (a measure of the lungs’ ability to expand) (Fig. 2B). For the sham injury control group, compliance was 0.0315 ± 0.0004 ml/cmH2O. By 8 hours after TBI, compliance decreased to 0.0286 ± 0.0008 ml/cmH2O (P = 0.01) and continued to decline for up to 24 hours. Compliance at 24 hours was 0.0205 ± 0.0006 ml/cmH2O, which was significantly lower than that of the control group (P < 0.01). Airway resistance values were also recorded but did not differ between the groups (fig. S6).

Fig. 2. PaO2/FiO2 ratios and compliance values derived from C57BL/6 and RAGE−/− mice 4 to 24 hours after TBI.

(A) PaO2/FiO2 ratios from controls without TBI and wild-type mice 4 to 24 hours after TBI. (B) Static compliance values obtained via pulmonary function testing from controls without TBI and wild-type mice 4 to 24 hours after TBI. (C) PaO2/FiO2 ratios from C57BL/6 and RAGE−/− mice 24 hours after TBI. Animals were administered either moderate or severe TBI. (D) Compliance values from controls and C57BL/6 and RAGE−/− mice 24 hours after TBI. n = 4 to 7 mice per condition, comparisons versus sham-injured group unless otherwise indicated, *P < 0.01. Statistical testing included unpaired t tests and analysis of variance.

Pulmonary dysfunction due to TBI is diminished in the lungs of RAGE−/− mice

In a separate group of experiments, we determined the role of RAGE in TBI-induced lung dysfunction. Wild-type C57BL/6 and RAGE−/− mice were subjected to moderate or severe TBI, and PaO2/FiO2 ratios were assessed at 24 hours. The moderately and severely injured wild-type C57BL/6 mice exhibited low PaO2/FiO2 values at 358 ± 14 and 221 ± 14, respectively. These values were significantly lower than those of the sham injury control mice (554 ± 9.25) (P < 0.01) (Fig. 2C). In contrast, RAGE−/− mice subjected to moderate TBI exhibited PaO2/FiO2 ratios that registered at 541 ± 8.06 and did not differ from sham injury PaO2/FiO2 ratio values (P = 0.32). Similarly, RAGE−/− mice with severe TBI did not appear to be compromised by the injury with a PaO2/FiO2 ratio of 447 ± 3.4, which was significantly higher than for the wild-type C57BL/6 mice subjected to the same severe TBI (P < 0.01). Static compliance was also assessed in wild-type C57BL/6 and RAGE−/− mice subjected to moderate and severe TBI. Static compliance in sham control animals was 0.034 ± 0.0003 ml/cmH2O, whereas compliance declined 24 hours after moderate or severe TBI to 0.0302 ± 0.0006 ml/cmH2O or 0.0226 ± 0.0009 ml/cmH2O, respectively (P < 0.01). Compliance values were improved significantly in RAGE−/− mice subjected to severe injury with a mean value of 0.0317 ± 0.0006 ml/cmH2O compared to the wild-type group subjected to the same injury (P < 0.01) (Fig. 2D). Such results indicate that RAGE plays a major role in the pulmonary dysfunction seen in wild-type mice.

TBI causes lung dysfunction in TLR4−/− mice

In an effort to determine the degree to which TLR4, a key HMGB1 receptor, may contribute to lung dysfunction, we subjected TLR4−/− mice to severe TBI and performed functional analyses after 24 hours. All studies were repeated with groups of mice separate from those reported in Figs. 1 and 2. The values from TLR4−/− mice were compared with a new set of RAGE−/− mice subjected to the same brain injury. The mean PaO2/FiO2 ratio in the TLR4−/− mice was 355 ± 28, which was significantly lower than values present in control wild-type mice and RAGE−/− TBI mice (P < 0.01; Fig. 3A). Additionally, the static compliance value in the TLR4−/− mice was 0.0220 ± 0.0040 ml/cmH2O, again significantly lower than those of the control and RAGE−/− mice groups (P < 0.01; Fig. 3B).

Fig. 3. PaO2/FiO2 ratios and compliance values 24 hours after TBI in RAGE−/−, TLR4−/− mice treated with HMGB1-neutralizing antibody, and mice treated with TAT-4BB decoy peptide.

(A) PaO2/FiO2 ratios in control mice and RAGE−/−, TLR4−/− mice treated with HMGB1 antibody (HMGB1 Ab) or control antibody (CTRL Ab), or treated with 4BB (MyD88) decoy peptide 24 hours after TBI. (B) Compliance values in RAGE−/−, TLR4−/− mice treated with HMGB1 antibody or control antibody, or with 4BB decoy peptide 24 hours after TBI. n = 5 to 7 per condition, comparisons versus sham-injured group unless otherwise indicated, *P ≤ 0.01. Statistical testing included unpaired t tests and analysis of variance.

Effect of neutralizing HMGB1 on lung dysfunction

To determine whether HMGB1 is the primary ligand activating RAGE or TLR4 after TBI, wild-type C57BL/6 mice were treated with an HMGB1-neutralizing antibody or control antibody before TBI. HMGB1 antibody treatment resulted in a PaO2/FiO2 ratio of 408 ± 15.6 at 24 hours after TBI (Fig. 3A), with a mean compliance of 0.0272 ± 0.0015 (Fig. 3B). These values with the neutralizing antibody were significantly improved compared to values observed with the control antibody or in wild-type mice after TBI (P < 0.01). However, these values were not different from those observed in TLR4−/− mice (Fig. 3, A and B).

Preservation of lung function using a Toll/interleukin-1 receptor domain–derived decoy peptide

TLRs use common Toll/interleukin-1R (IL-1R) resistance (TIR) domains, which are pivotal to both interactions of signaling proteins and signal transduction (20). Because TLR TIR dimers serve as recruitment centers for TLR adapter proteins, it is possible to inhibit cellular signaling downstream of TLR4 using cell-permeable decoy peptides as competitive inhibitors of protein-protein interactions (21). TIRAP and MyD88, known adaptor proteins for TLR4, also bind to phosphorylated RAGE (22, 23). Using the cell-permeating decoy peptide, TAT-4BB, to inhibit surface-exposed segments of these adapter proteins, we targeted the ability of HMGB1 to produce signaling in vivo via either TLR4 or RAGE (24). Wild-type C57BL/6 mice were administered a TAT-4BB decoy peptide before exposure to severe TBI. The mean PaO2/FiO2 ratio in these animals was 487.8 ± 31.14 at 24 hours after injury, which was significantly higher than values derived from the TLR4−/− TBI mice (P = 0.01) but did not differ from RAGE−/− mice (P = 0.22; Fig. 3A). Additionally, the static compliance value in the decoy peptide 4BB-treated mice was 0.0286 ± 0.0050 ml/cmH2O, which did not differ from TLR4−/− (P = 0.33) or RAGE−/− mice (P = 0.55; Fig. 3B).

TBI induces pulmonary dysfunction in the setting of preserved cardiac output

In an effort to ensure that TBI-induced pulmonary dysfunction was noncardiogenic in nature, we used gated positron emission tomography (PET) scans to calculate cardiac outputs from four sham injury control mice and four mice 24 hours after severe brain injury (fig. S7A). The data showed that there was no significant difference in heart rates between control animals [440.4 ± 32.3 beats per minute (bpm)] and TBI mice (441.7 ± 18.8 bpm) (fig. S7B). TBI mice did exhibit elevations in average stroke volumes (0.0436 ± 0.0026 ml) when compared with sham injury control volumes (0.034 ± 0.0017 ml) (fig. S7C). Despite the elevated stroke volumes in the sham injury group, there was not a significant difference in cardiac output between the groups.

Murine lung transplant recipients from TBI donors exhibit elevated lung injury scores compared to recipients of lungs from RAGE−/− TBI donors

ALI scores (Fig. 4D) were significantly lower in transplanted lungs from wild-type control donors (Fig. 4A) at 0.12 ± 0.02 when compared to TBI donors, with a mean score of 0.44 ± 0.08 (P < 0.01; Fig. 4B). The mean score in transplanted lungs from TBI RAGE−/− donors was 0.17 ± 0.03 (minimal injury), and these lungs were otherwise histologically normal (Fig. 4, C and D). This score was significantly lower than that for the wild-type C57BL/6 donor group (P < 0.01). Such results demonstrate that eliminating RAGE appears to preserve pulmonary injury after transplantation.

Fig. 4. Histology and ALI scores for transplanted left lungs from control donor mice, wild-type TBI donor mice, or RAGE−/− TBI donor mice.

Representative H&E histology sections from transplanted lungs 5 days after transplant. (A) Transplanted left lung from healthy donor with sham TBI. (B) Transplanted left lung from wild-type TBI donor. (C) Transplanted left lung from RAGE−/− TBI donor. (D) ALI scores from transplant recipients that received lungs from three different donors: wild-type animals with sham TBI, wild-type animals with TBI, and RAGE−/− donors with TBI. n = 6 to 8 per condition, comparison versus sham-injured group, *P < 0.01. Statistical testing included unpaired t tests and analysis of variance.

Elevated IL-10 in TBI RAGE−/− mice and recipients of lungs from TBI RAGE−/− mice

IL-4, IL-6, IL-17a, tumor necrosis factor–α (TNF-α), and interferon-γ (IFN-γ) BAL levels did not differ among donor sham-injured controls, TBI wild-type, and TBI RAGE−/− donor mice (Fig. 5, A to E). However, RAGE−/− donor mice had significantly higher levels of the anti-inflammatory molecule, IL-10, in BAL fluid (MFI 530.0 ± 23) when compared with sham controls (MFI 331.5 ± 16) or TBI wild-type mice (MFI 408.5 ± 13) (P < 0.01) (Fig. 5F). Cytokine profiling was also determined in BAL fluid of transplanted lungs of mice whose donors were either TBI wild-type or RAGE−/− mice. These transplants were performed 24 hours after sham injury or TBI. Cytokine profiling in the BAL fluid of recipients 8 hours after transplant revealed no differences in IL-4, IL-6, IFN-γ, TNF-α, and between wild-type and RAGE−/− donor lungs (Fig. 5, A, B, D, and E). IL-17a was significantly lower in the BAL fluid of recipients when the left lung transplant was derived from a RAGE−/− TBI mouse: 245 ± 63.4 versus 602 ± 33.2 for TBI wild-type C57BL/6 control mice (P < 0.01; Fig. 5C). Similar to observations in mice subjected to TBI alone, IL-10 was significantly elevated in BAL from transplanted lungs from RAGE−/− TBI donors compared to transplanted lungs from wild-type C57BL/6 TBI donors (1846 ± 19.6 versus 1241 ± 21.3, P < 0.01; Fig. 5F).

Fig. 5. BAL cytokine profiles in control and injured donor mice 24 hours after TBI.

Cytokine profiling was also performed in left lung transplant recipients of TBI donor lungs 8 hours after transplant. (A to F) Concentrations of the cytokines IL-4, IL-6, IL-17a, IFN-γ, TNF-α, IL-17A, and IL-10 are shown as mean fluorescence intensity (MFI). n = 5 to 7 per condition, *P < 0.01. Statistical testing included unpaired t tests and analysis of variance.

HMGB1 stimulates NF-κB activation in airway epithelial cells

In an effort to determine the downstream signaling pathways activated by HMGB1 binding to RAGE in the lungs, we stimulated rat alveolar type 2 epithelial cells with increasing concentrations of the non-oxidized (all-thiol) form of HMGB1 that binds only to RAGE. Translocation of NF-κB p65 and p50 from the cytoplasm to the nucleus, an indication of NF-κB activation, was enhanced 24 hours after stimulating AT2 cells with HMGB1 (at least 10 μg/ml) (Fig. 6).

Fig. 6. Representative Western blot of NF-κB activation in rat alveolar type 2 (AT2) cells after exposure to HMGB1.

(A) NF-κB p65 and p50 translocation from the cytoplasm to the nucleus, demonstrated by stimulating AT2 cells with HMGB1 (0, 5, or 10 μg/ml) for 24 hours. (B to E) Densitometric analysis was performed with GAPDH (glyceraldehyde-3-phosphate dehydrogenase) as a loading control for the cytoplasmic samples and lamin B1 as a loading control for the intranuclear samples.

Elevated donor HMGB1 correlates with poor blood oxygenation before and after human lung transplant

Blood was collected from human brain-dead donors before procuring lungs to be used for bilateral lung transplants. Donor-, recipient-, and surgery-specific characteristics are presented in Table 1. All blood samples were obtained from brain-dead organ donors: 11 of the 18 died from TBI, and the rest died of other causes including 4 from anoxia and 3 from stroke. The correlation between donor serum HMGB1 concentrations and the highest PaO2/FiO2 ratio before procurement was determined. HMGB1 concentrations correlated inversely with donor PaO2/FiO2 ratios; high HMGB1 levels were associated with lower PaO2/FiO2 ratios (P < 0.01; Fig. 7A). This correlation was maintained when the data included only the subset of 11 donors that died from TBI (P = 0.05; Fig. 7B). To examine the association between donor HMGB1 and recipient lung function, donor HMGB1 serum concentrations were correlated with recipient PaO2/FiO2 ratios 48 hours after lung transplant. Linear regression revealed that donor HMGB1 concentrations also correlated inversely with recipient PaO2/FiO2 ratios 48 hours after transplant, both in the 18 brain-dead donors (P < 0.01) and in the 11 TBI donors (P < 0.01; Fig. 7, C and D).

Table 1. Donor, recipient, and surgery characteristics (N = 18).

Eighteen bilateral lung transplants were performed, and donor, recipient, and operative variables were recorded. All donor lungs were donated after brain death.

View this table:
Fig. 7. PaO2/FiO2 ratios in lung transplant donors and recipients based on donor HMGB1 serum concentrations at the time of harvest.

(A) PaO2/FiO2 ratios among all brain-dead donors, regardless of mechanism of injury, correlated with donor HMGB1 serum concentrations at harvest. (B) PaO2/FiO2 ratios among brain-dead donors whose mechanism of injury was TBI, and donor HMGB1 serum concentrations at harvest. (C) Recipient PaO2/FiO2 ratios at 48 hours after transplant correlated with donor HMGB1 serum concentrations at harvest. (D) Recipient PaO2/FiO2 ratios at 48 hours after transplant from brain-dead donors whose mechanism of injury was TBI correlated with donor HMGB1 serum concentrations at harvest.


Our results demonstrate that the HMGB1-RAGE ligand-receptor pathway serves as a central signal transduction mechanism for pulmonary dysfunction after TBI. TBI-induced alterations in physiology and histology included reduced systemic arterial oxygen, decreased lung compliance, and pathology consistent with ALI. These findings were RAGE-dependent because systemic oxygenation, lung compliance, and histology were restored almost completely to baseline in RAGE−/− mice subjected to TBI. Considering that most of the lungs used for transplant are obtained from patients who have suffered a TBI, factors such as donor HMGB1 serum concentration may be predictive of lung function before and after transplantation. Our current clinical data suggest that brain injury–induced HMGB1 release may have a role in donor and short-term recipient pulmonary function as well.

It is well established that proinflammatory molecules including HMGB1 are ligands for RAGE (25). Although initial reports suggested that extracellular HMGB1 released by necrotic and inflammatory cells functions as a late-acting cytokine-mediating endotoxin-related lethality in mice (26) and cytokine synthesis in monocytes (27), there has been greater recognition that the effect of HMGB1 activation of RAGE has acute effects on leukocyte chemotaxis and vascular barrier disruption (2830). Moreover, these effects may be important for sustaining inflammatory disease conditions alone (22) or in combination with TLRs (3134). Furthermore, our results with HMGB1 cortical staining demonstrating HMGB1 translocation from the nucleus after TBI agree with reports demonstrating HMGB1 translocation and release after brain insult (35).

A recent report demonstrated that neutralization of HMGB1 may be a new treatment for brain-injured animals because of reductions in brain edema and inflammation (36). Given our results, these beneficial effects can now be extended to the lungs. Given the ability of HMGB1 to bind to both RAGE and TLR4, our experimental outcomes implicate a pronounced role for RAGE because the lung dysfunction was more pronounced in the setting of TLR4 deficiency as compared to RAGE deficiency. One interpretation suggests that HMGB1-RAGE is instrumental for leukocyte infiltration, whereas HMGB1-TLR4 may be responsible for cytokine production (3740). Alternatively, it is known that HMGB1 can exist in either of two isoforms, all-thiol and disulfide, which have varying specificity for RAGE and TLR4. The all-thiol state of the isoform is first released into the extracellular space, where it binds to RAGE and initiates chemoattraction of leukocytes (41). In contrast, the disulfide isoform exhibits a high affinity for TLR4 and induces the release of proinflammatory cytokines and mediators (42). Compared to the half-life of all-thiol (~17 min), the disulfide isoform’s half-life is significantly longer (up to 10 hours) (43). Considering the kinetic differences between the isoforms and the rapid induction of lung injury after TBI, it is apparent that the all-thiol form of HMGB1 binding to RAGE may be the primary pathway of lung injury in the current study. This finding is also supported by our data, showing that the all-thiol form of HMGB1 induced NF-κB activation in airway epithelial cells. However, the exact contribution of both HMGB1 isoforms in TBI-induced lung injury is yet to be established.

The effects of the TAT-4BB decoy peptide and its ability to modulate TBI-induced inflammation are likely dependent on the adapter proteins common to both TLR4 and RAGE (22), although there is a suggestion that the HMGB1-TLR4-RAGE interaction is central to the downstream signaling (44). However, given the apparent commonality of downstream signaling for both TLR4 and RAGE, we hypothesize that decoy peptides would inhibit TBI-associated changes in pulmonary tissue and physiology regardless of the signaling receptor.

Ligand-dependent activation of RAGE may modulate chemotactic actions of leukocytes. Our results reinforce previous reports that HMGB1 may stimulate RAGE activation of NF-κB (45), so it is conceivable that pulmonary dysfunction after TBI includes the production of proinflammatory cytokines such as IL-6, TGF-β (transforming growth factor–β), and IL-17a (4548). However, the only cytokine that appeared to be altered in the BAL after injury was the anti-inflammatory cytokine IL-10. Together with previous observations in a model of hemorrhagic shock (49) and a recent report suggesting that signaling via RAGE is linked to down-regulation of IL-10, it appears that modulation of this important anti-inflammatory cytokine is regulated by RAGE activation. It is possible that increased IL-10 concentrations could dampen the inflammatory process and mitigate injury to the pulmonary system. An example of such an effect may be evidenced by the beneficial effects of soluble RAGE administration in IL-10 null mice subjected to chronic colonic inflammation (50). In addition, IL-10 gene therapy has been used to repair donor human lungs before transplantation (51).

Other work using an ischemia reperfusion injury model demonstrated that the HMGB1-RAGE pathway contributes to IL-17a production from invariant natural killer cells, which are crucial for the initiation of ischemia reperfusion injury (17). Our results corroborated this concept of HMGB1-RAGE–mediated effects on cytokine production in ischemia reperfusion injury. We observed decreased IL-17a protein expression in the BAL of recipients of RAGE−/− TBI donor lungs as compared to recipients of wild-type TBI donor lungs. Additionally, the connection between RAGE and lung dysfunction has been studied outside of the setting of TBI. Circulating soluble RAGE was reported to be a marker of primary graft dysfunction, which is a form of ischemia reperfusion injury that occurs after lung transplantation (52).

Noncytokine pathways have also been linked to TBI-induced lung injury. A recent report implicated high levels of glutamate after brain injury, triggering ALI mediated through interaction of the adenosine A2A receptor and the metabotropic glutamate receptor 5 (mGluR5) on bone marrow–derived cells (53). This pathway may run in parallel to the HMGB1-RAGE axis with overlap of downstream signaling.

In summary, the current study demonstrates a mechanism of TBI-induced ALI that includes a key role for the HMGB1-RAGE pathway. These data may be important for preventing pulmonary complications in patients after TBI. These findings could also have implications in the field of lung transplantation. Most lungs for transplantation are procured from patients after TBI. Remarkably, only 15 to 20% of such lungs are suitable for transplant because of abnormal physiology that includes impaired oxygenation, a critical parameter of lung function. Given that TBI has profound detrimental effects on pulmonary physiology, the current study suggests that targeting the RAGE-HMGB1 axis could serve as a new therapeutic approach to increase the quality of lungs suitable for transplantation.


Study design

This study was designed to identify the role of the HMGB1-RAGE inflammatory pathway in TBI-induced pulmonary dysfunction. To do this, we used murine models of TBI and lung transplantation in wild-type and transgenic mice, as further described below, to better understand the pathways at work. We also administered an HMGB1-neutralizing antibody, as well as a downstream signaling decoy peptide, to further delineate the downstream signaling. Additionally, we used our institution’s prospectively collected lung transplant biobank to correlate donor levels of HMGB1 with donor and recipient lung function.

Human studies

All human studies were approved by the Indiana University Institutional Review Board. Additionally, all lung transplant recipients consented, and all donor families consented for organ donation and research. Using prospectively collected samples from lung transplant recipients and donors, we identified donor cause of death and donor serum levels of HMGB1 using an enzyme-linked immunosorbent assay (ELISA) kit (IBL International) and correlated with donor pulmonary function values and short-term recipient pulmonary values and outcomes.


All mice were housed in the Laboratory Animal Resource Center at Indiana University School of Medicine in accordance with institutional guidelines. All mice were 8 to 12 weeks of age and 24 to 32 g and used as both donors and recipients. Mice were randomly assigned to groups. All studies were approved by the Laboratory Animal Resource Center at the Indiana University School of Medicine. Specific pathogen–free male inbred mice C57BL/6 were purchased from Harlan Sprague Dawley. TLR4−/− mice were purchased from Jackson Laboratory, and RAGE−/− mice were obtained from onsite colonies (54).

Traumatic brain injury

Mice were subjected either to a TBI using a controlled cortical impact model with an electromagnetic impactor (Impactor One, MyNeuroLab) or to sham injury (55). The mice were anesthetized with isoflurane and ketamine and placed in a stereotaxic frame (Kopf Instruments). The skull was exposed in the left frontoparietal cortex using an electric drill. Before the injury, the impacting piston was angled so that the impacting tip (3 mm in diameter) was perpendicular to the exposed cortical surface. This was accomplished by rotating the entire stereotaxic frame on the transverse plane. Sham control animals were exposed to anesthesia, skull exposure, and suturing without the cortical impact. The amount of deformation was set at either 0.5 mm (moderate injury) or 1.0 mm (severe injury). After injury, animals were allowed to recover and were able to eat and drink for 4 to 24 hours before harvest.

Serum and BAL

After euthanization, blood was obtained from the right ventricle and centrifuged (15 min, 1500g, 4°C), and serum was isolated. For BAL, lungs were lavaged with aliquots totaling 3 ml of Ca2+- and Mg2+-free phosphate-buffered saline (PBS) supplemented with 0.1 mM EDTA. Samples were centrifuged (10 min, 1500g, 4°C).

Pulmonary function tests and arterial blood gases

After inducing anesthesia with 1 to 2% isoflurane, all mice were mechanically ventilated with a rodent ventilator using room air, at a rate of 140 breaths per minute, a tidal volume of 0.3 ml, and 2 cmH2O of positive end-expiratory pressure. The animals were placed on a heated (37°C) pad, and pulmonary function tests were then performed with the flexiVent system (Scireq). At the competition of pulmonary function testing, a 25-gauge angiocatheter was placed at the junction of the left ventricle and the ascending aorta, and 2 to 3 ml of arterial blood were obtained. This was run on an i-STAT Point of Care analyzer (Abbott Laboratories). Subsequently, mice were euthanized, and lungs were harvested in 10% neutral buffered formalin.

Histology and lung injury scoring

After mice were euthanized, native and donor lungs were harvested, glutaraldehyde-fixed, and paraffin-embedded. A portion of each lobe of each lung was sectioned and stained with H&E. Lungs were stained with H&E and scored by a blinded pathologist using the Lung Injury Scoring System from the American Thoracic Society Workshop Report (19). These criteria give a continuous score between 0 (no injury) and 1 (severe injury). Additionally, immunohistochemistry was performed on lung sections for caspase-3, a marker of apoptosis.


BAL fluid from animals was collected, and MPO levels were measured using an ELISA kit (Cell Sciences) according to the manufacturer’s instructions.

PET/computed tomography cardiac output and image analysis

Cardiac output was calculated using dynamic high-resolution 18F-NaF PET computed tomography (CT). Subsequently, CT and PET images were co-registered for image analysis, and the Stewart-Hamilton indicator dilution method was used to calculate cardiac output using a left ventricular time-activity curve. The full details are available in the Supplementary Materials.

TAT-4BB decoy peptide

TAT-4BB was obtained from GenScript, containing the essential amino acid sequence of the TAT domain with 4BB (YGRKKRRQRRR-LHYRDRIPGVAIAA). This was administered via a subcutaneous 24-hour micro-osmotic pump from Alzert. Vehicle was used as a negative control.

Serum HMGB1 measurement

HMGB1 serum levels were determined using an ELISA kit (IBL International) according to the manufacturer’s instructions.

HMGB1-neutralizing antibody

Neutralization of HMGB1 was accomplished by administering anti-HMGB1 chicken immunoglobulin Y (IgY)–neutralizing polyclonal antibody (IBL International) dosed at 2 mg/kg both 30 min before TBI. Isotype-matched control antibodies were administered to control animals with the same dosing regimen. Additionally, immunohistochemistry was performed on paraffin-embedded mouse lungs against HMGB1 (ab18256, Abcam) as previously described (56). Brain tissue analysis was also performed with immunocytochemical and immunohistochemical methodologies as previously described (57). Cortical tissue was serially sectioned at 14 μm and was used in immunohistochemical experiments. The primary antiserum used was the rabbit anti-HMGB1 antibody (1:1000; Sigma-Aldrich). Sections were incubated in secondary donkey anti-rabbit conjugated to CY3 (Jackson ImmunoResearch Laboratories Inc.). Nuclear colocalization with DAPI was also performed, and images were imported into Image-Pro Plus (Media Cybernetics) for quantification.

HMGB1-stimulated NF-κB activation in airway epithelial cells

A cell culture of rat alveolar type 2 (AT2) cells was stimulated with increasing concentrations of non-oxidizable lipopolysaccharide (LPS)–free HMGB1 (HMGBiotech) at 0, 5, 10, and 15 μg/ml for 24 hours. Nuclear and cytosolic protein lysates were then collected and analyzed for p50 and p65 NF-κB by Western blot analysis. Further details are supplied in the Supplementary Materials.

Cytokine profiling by cytometric bead array

BAL fluid from animals was collected, and cytokine protein levels of IL-17A, IL-10, TNF-α, IFN-γ, IL-6, IL-4, and IL-2 were measured using the Mouse Th1/Th2/Th17 Cytokine Kit (BD Biosciences) according to the manufacturer’s instructions.

Murine orthotopic left lung transplantation

All surgical procedures were performed with sterile techniques in a previously described method (58), which is detailed in the Supplementary Materials.

Statistical analysis

Data were normally distributed and analyzed statistically with GraphPad Prism (GraphPad Software Inc.). Results are expressed as means ± SEM. Differences between groups were analyzed by unpaired t tests and analysis of variance and Tukey post hoc test. Correlation coefficient between two variables was calculated with the Pearson correlation test. P values not exceeding 0.05 were considered significant.


Materials and Methods

Fig. S1. MPO levels in sham and TBI mice.

Fig. S2. Presence of activated caspase-3 in pulmonary tissue after TBI.

Fig. S3. RAGE is present in the lung but not in other tissues of C57BL/6 mice.

Fig. S4. IHC staining against HMGB1 in lungs of sham and TBI mice.

Fig. S5. HMGB1 staining in normal and injured cortical tissue.

Fig. S6. Airway resistance values after TBI.

Fig. S7. Cardiac output after TBI.

References (5963)


Acknowledgments: We would like to acknowledge the assistance of the Indiana Organ Procurement Organization in assisting with our clinical donor information. Funding: This work was supported by NIH grants R01 HL096845 and National Institute of Allergy and Infectious Diseases P01AI084853 to D.S.W., and R01 DK100905 and Indiana Spinal Cord and Brain Injury Research Fund to F.A.W. Author contributions: Conception and design: D.J.W., F.A.W., and D.S.W.; analysis and interpretation: D.J.W., A.S.A.G., M.L.C., P.R.T., G.D.H., H.E.B., G.E.S., F.A.W., and D.S.W.; animal studies and laboratory experiments: histology and ALI scoring: D.J.W. and G.E.S.; TBI experiments: D.J.W., A.S.A.G., and M.S.R.; murine pulmonary function testing: D.J.W. and A.J.F.; murine lung transplant model: D.J.W. and A.S.A.G.; cardiac output analysis: A.A.R., B.P.M., and P.R.T.; cell culture and immunohistochemistry: P.H.P., R.V., M.L.C., Y.M.A., and Y.K.; protein analysis: B.M.C., Y.K., and M.L.C.; in vivo decoy peptide testing: Y.M.A., M.S.R., and D.J.W.; clinical samples and analysis: D.J.W.; drafting the manuscript for important intellectual content: D.J.W., A.S.A.G., P.R.T., F.A.W., and D.S.W. Competing interests: D.S.W. is a cofounder and chief scientific officer of ImmuneWorks, a biotech company developing therapeutics for immune-mediated lung diseases. The other authors declare that they have no competing interests.
View Abstract

Navigate This Article