Attenuation of endoplasmic reticulum stress by caffeine ameliorates hyperoxia-induced lung injury

Abstract

Rodent pups exposed to hyperoxia develop lung changes similar to bronchopulmonary dysplasia (BPD) in extremely premature infants. Oxidative stress from hyperoxia can injure developing lungs through endoplasmic reticulum (ER) stress. Early caffeine treatment decreases the rate of BPD, but the mechanisms remain unclear. We hypothesized that caffeine attenuates hyperoxia-induced lung injury through its chemical chaperone property. Sprague-Dawley rat pups were raised either in 90 (hyperoxia) or 21% (normoxia) oxygen from postnatal day 1 (P1) to postnatal day 10 (P10) and then recovered in 21% oxygen until P21. Caffeine (20 mg/kg) or normal saline (control) was administered intraperitoneally daily starting from P2. Lungs were inflation-fixed for histology or snap-frozen for immunoblots. Blood caffeine levels were measured in treated pups at euthanasia and were found to be 18.4 ± 4.9 μg/ml. Hyperoxia impaired alveolar formation and increased ER stress markers and downstream effectors; caffeine treatment attenuated these changes at P10. Caffeine also attenuated the hyperoxia-induced activation of cyclooxygenase-2 and markers of apoptosis. In conclusion, hyperoxia-induced alveolar growth impairment is mediated, in part, by ER stress. Early caffeine treatment protects developing lungs from hyperoxia-induced injury by attenuating ER stress.

bronchopulmonary dysplasia (BPD) is a common but serious lung disorder among extremely premature infants. Currently, BPD primarily affects premature infants born at <30 wk gestation, when the lungs are at saccular stage of development (36). More than 10,000 BPD cases are newly diagnosed each year in the United States (33). The majority of the infants with BPD require prolonged oxygen therapy to maintain their oxygenation (29), and up to one-third will develop pulmonary hypertension as a complication (5). The mortality risk increases to 50% for BPD infants who develop pulmonary hypertension (23). Survivors of BPD suffer from decreased exercise tolerance and impaired lung function into adulthood (51). Early caffeine treatment was shown to decrease BPD incidence in premature infants (42) and is widely used in clinical practice (30). However, the mechanisms involved in caffeine-mediated protection from BPD remain unclear. Delineation of these mechanisms could potentially lead to additional targeted therapies to prevent or ameliorate BPD.

Premature infants are more susceptible to oxidative injury secondary to their inadequate antioxidant defenses as well as their impaired upregulation in response to oxidant stress (26). Endoplasmic reticulum (ER), which mediates protein synthesis, folding, and modification, is sensitive to oxidative stress. ER responds to oxidative stress by unfolded protein response (UPR) to handle the increased need for protein folding or increased protein damage (32). Three ER transmembrane proteins are involved in UPR initiation: inositol-requiring enzyme 1α (IRE1α), PKR-like ER kinase (PERK), and activating transcription factor 6 (ATF6). These three proteins interact with binding immunoglobulin protein (BiP), an ER chaperone, to keep them inactive in the absence of oxidative stress. On ER stress, BiP dissociates from these three proteins to activate UPR by modification of downstream effectors, including the splicing of X-box binding protein 1 (XBP1) mRNA and cleavage of ATF6, and increase CCAAT/enhancer binding protein (C/EBP) homologous protein (CHOP) levels (37). Activated PERK and ATF6 can reciprocally increase the transcription and translation of BiP under oxidative stress.

UPR has three biological functions: adaptive response, feedback control, and cell fate (36). If the response is successful, UPR signaling pathway is turned off by feedback mechanisms. However, protein folding inside the ER also generates reactive oxygen species (ROS) that can potentially aggravate ER stress (41). Prolonged or exaggerated ER stress can overwhelm the initial coping mechanisms involving protein refolding and translation and lead to a vicious cycle of protein degradation and apoptosis (32). There is also a close interaction between ER and mitochondria through calcium sharing (50). During early phase of oxidative stress, the ER-mitochondrial coupling increases to promote mitochondrial bioenergetics and to facilitate the cell adaptation to stress (6). However, chronic ER stress leads to amplification of ROS formation and inflammation (21), which often leads to apoptosis (27) through both mitochondria-dependent and -independent pathways (46). Similar response is also seen in endothelial cells, which leads to impairment in angiogenesis (43) and vasodilation (17). Angiogenesis plays a vital role in postnatal lung development (1), and impaired angiogenesis contributes to the development of BPD (48). Both ER stress (9) and mitochondrial dysfunction (38) contribute to the growth arrest of alveoli in hyperoxia-exposed rodent pups. Increased cyclooxygenase-2 (Cox-2) expression is shown to be involved in ER stress and inflammation-induced lung injury in hyperoxia (9).

Caffeine, a nonselective phosphodiesterase inhibitor, has been shown to prevent the lung inflammation induced by either hyperoxia (52) or intrauterine lipopolysaccharide injection in rat pups (24). Recently, it was reported that caffeine can function as a chemical chaperone to reduce ER stress (20), which may help to protect the lungs from oxidative stress. However, histology studies have so far failed to show improvement in alveolarization by caffeine. It is especially concerning that one recent study showed that caffeine treatment actually aggravated hyperoxia-induced lung injury in mouse pups by increasing apoptosis of alveolar cells (11). The caffeine doses used in previous rodent studies may not be able to provide adequate blood levels using the U.S. Food and Drug Administration (FDA)-recommended conversion table (39). For our current study, we hypothesized that 1) increased ER stress contributes to hyperoxia-induced injury to the developing lungs; 2) early caffeine treatment, given at higher doses, can modulate ER stress in the developing lungs and mitigate hyperoxia-induced impairment of alveolar formation; and 3) continuation of caffeine treatment past hyperoxia exposure will restore lung growth after hyperoxia-induced injury. Our studies tested a novel mechanism by which caffeine treatment attenuates hyperoxia-induced lung injury.

MATERIALS AND METHODS

Mouse antibody for CHOP and rabbit antibodies for PERK, IRE1α, and BiP were from Cell Signaling Technology (cat. no. 9956S; Beverly, MA). Rabbit antibody for serine⁶³⁷ phospho-dynamin-related protein 1 (Drp1; GTX50911) and mouse antibodies for caspase-12 (GTX59923) and 8-hydroxyguanosine (8-OH-dG; GTX41980) were from GeneTex (Irvine, CA). Rabbit antibodies for mitofusin-1 and -2 (Mfn1/2; sc-50330 and sc-50331), cyclooxygenase-1/2 (Cox-1/2; sc-7950 and sc-7951), Bcl-2 (sc-492), Bax (sc-6236), myeloperoxidase (MPO; sc-33596), Drp1 (sc-32898), and phospho-PERK (sc-32577) were from Santa Cruz Biotechnology (Dallas, TX). Rabbit antibodies for phospho-IRE1α (NB100-2323) and XBP1 (NB100-78403) and mouse antibody for ATF6 (NBP1-40256) were from Novus Biologicals (Littleton, CO). Mouse antibodies against rat endothelial cell antigen (RECA-1; ab9774) and NADH:ubiquinone oxidoreductase subunit of complex I (ab14711) were from Abcam (Cambridge, MA). The Caffeine/Pentoxifylline enzyme-linked immunosorbent assay (ELISA; cat. no. 106419) kit was from Neogen (Lexington, KY). In Situ Cell Death Detection POD Kit (cat. no. 11684817910) was from Roche Applied Science (Indianapolis, IN). Caffeine (cat. no. 93784), 3,3′-diaminobenzidine (DAB; D12384), and other chemicals, including nitrotyrosine (3-NT) antibody (N0409), were obtained from Sigma-Aldrich (St. Louis, MO).

Animal care.

Time-dated pregnant Sprague-Dawley rats were obtained from Harlan (Madison, WI) and were acclimated to our animal care facility for 7 days. The use of animals was approved by the Medical College of Wisconsin Institutional Animal Care and Use Committee and conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The animals were kept under a 12-h:12-h light-dark cycle and given free access to chow and water. Four pregnant rats were used for each experiment in the study. The dam and pups were placed with their cages in either >90% oxygen chamber (hyperoxia, H) or room air (normoxia, N) from postnatal day 1 to 10 (P1 to P10). Oxygen concentrations were continuously monitored with an oxygen sensor (Reming Bioinstruments, Redfield, NY).

Pups were allowed to feed ad libitum from nursing mothers. The dams were allowed to recover in room air for 2 h daily. The dose (20 mg·kg⁻¹·day⁻¹) of caffeine used in this study was calculated according to the dose commonly used in premature infants (5–10 mg·kg⁻¹·day⁻¹ caffeine citrate; Ref. 42) corrected to the K_m factor (39). The first dose was given at P2 intraperitoneally via insulin syringe with 29-gauge needles (Becton Dickinson, New York, NY) daily. No loading dose was given. Caffeine was then continued until euthanasia at P10 or P21.

Some pups were euthanized at P10 (alveolar stage P4–P14; Ref. 36), whereas others were recovered in room air until P21 to simulate clinical practice since most premature infants with BPD outgrow their oxygen dependence by the age of 1 yr, before they complete their alveolar stage at age 2–6 yr. Pups were euthanized by carbon dioxide, and lungs and heart were removed en bloc. Blood was collected from the right ventricle at the time of euthanasia for measuring plasma caffeine levels using commercial ELISA kit, and freshly prepared caffeine solution was used as control. A small cut was created on the left atrium, and ice-cold normal saline was gently infused through right ventricle to flush out the blood from lungs before inflation for histology or snap-freezing in liquid nitrogen for protein studies.

Histology study.

For histology studies, the trachea was cannulated by an Instech Solomon (20-gauge) stainless steel feeding tube (Plymouth Meeting, PA), and the lungs were inflated with 10% neutral buffered formalin at 25 cmH₂O (2.4 kPa) for 1 h. Lungs were removed with trachea securely tied with surgical silk under pressure of 25 cmH₂O, fixed additionally in 10% neutral buffered formalin for 24 h, and then embedded in paraffin. Lung paraffin sections (5 μm) were mounted on SuperFrost Plus-coated slides (Denville Scientific, Metuchen, NJ). After deparaffinization, the sections were stained with hematoxylin-eosin (H&E) for histology. Images, devoid of major bronchi and large blood vessels, were captured with a mounted digital camera under ×10 objective of an Olympus IX51 microscope. Mean linear intercept (MLI), or chord length, was used as a method to estimate volume-to-surface ratio of acinar air spaces, whereas radial alveolar count (RAC) and secondary septa were investigated to study the complexity of lung structure (22). Ten equally spaced horizontal lines and eleven equally spaced vertical lines were drawn on each picture, and the number of intercepts through the alveolar wall were counted. MLI was obtained by the number of times the traverses are placed on the lung multiplied by the length of the traverse and then divided by the sum of all of the intercepts. For the RAC, a line from the center of the respiratory tract perpendicular to the nearest connective tissue septum was drawn, and alveoli intercepting with the line were counted. For measurement of secondary septa, elastin was stained with resorcin-fuchsin and Van Gieson solution. Blood vessels < 20 μm in diameter were identified by immunohistochemistry stain with RECA-1 antibody and counted. The in situ TdT-mediated dUTP nick-end labeling (TUNEL) staining was done according to manufacturer’s instructions using DAB as chromophore and counterstained with H&E. Apoptosis was estimated by percentage of brown-stained nuclei under high-power field. Inflammatory cells were stained with MPO antibody. The average of three sections per pup, and five counts per section (fifteen counts per pup), was used for statistical analyses. Five pups per group were obtained for histology.

Transmission electron microscope study.

Rat pups were euthanized at P10. Heart and lungs were obtained en bloc, and 2.5% glutaraldehyde was gently infused through the right ventricle until no visible blood could be seen coming out of the left atrium. The lungs were then immediately put into 2.5% glutaraldehyde for fixation. Sections of 100-nm thickness were obtained by ultramicrotome. Images of the pulmonary vascular endothelial cells were obtained by Hitachi 600 electron microscope.

Measurement of oxidative stress.

8-OH-dG immunofluorescence was used as the biomarker for oxidative stress, and 3-NT immunohistochemistry was used as biomarker for reactive nitrogen stress. Lung sections (5 μm) were stained with 8-OH-dG antibody (1:100) for overnight at 4°C and then treated with Alexa Fluor 488-conjugated secondary antibody for 1 h in room temperature and counterstained with 4′,6′-diamidino-2-phenylindole (DAPI) before imaging under fluorescent microscope. Tissue sections were stained with 3-NT antibody (1:100) for overnight at 4°C and then treated with horseradish peroxidase (HRP)-conjugated secondary antibody for 1 h in room temperature and developed using DAB as the chromophore. Integrated signals were quantified and corrected against the total area of signal by ImageJ.

Western blot analysis.

Whole lung lysates were obtained by homogenizing in MOPS buffer, containing (in mM) 20 3-(N-morpholino)propanesulfonic acid, 2 EGTA, 5 EDTA, 30 NaF, 10 β-glycerophosphate, 10 Na-pyrophosphate, 2 Na-orthovanadate, and 1 PMSF and (in %) 0.5 Nonidet P-40, 1 protease inhibitor cocktail, and 1 phosphatase inhibitor cocktails 2 and 3 (pH 7.0), by Bullet Blender (Next Advance, Averill Park, NY). For Western blot analysis, 30 μg of lysate protein was resolved by SDS-PAGE, transferred to the nitrocellulose membranes (0.22 µm), and then probed with appropriate primary antibodies overnight at 4°C. Signals were generated after incubation with HRP-conjugated goat anti-rabbit (1:10,000) or anti-mouse (1:10,000) IgG using WesternBright ECL Chemiluminescent HRP Substrate (Advansta, Menlo Park, CA) and recorded on CL-Xposure films (Thermo Fisher Scientific, Rockford, IL). Integrated optical density was processed by ImageJ, and β-actin signal was used as the loading control. Samples from six pups were used for each group, and these six values are shown in summary plots, whereas representative blots with three pups per group were shown in figures.

Statistical analysis.

Data were analyzed by MedCalc Statistical Software version 15.2.1 (MedCalc Software, Ostend, Belgium; http://www.medcalc.org). Values are expressed as means ± SD, and a scatterplot was used for the figures. Significant differences between groups were determined by one-way ANOVA with Student-Newman-Keuls post hoc test after Levene test for equality of error variances. A P value < 0.05 was considered statistically significant.

RESULTS

Plasma caffeine level is within therapeutic range.

Plasma caffeine level was below the detection limit in normal saline-treated pups, whereas levels in caffeine-treated pups ranged from 12.9 to 25.7 μg/ml. There was no difference in caffeine levels between normoxic (18.9 ± 5.0 μg/ml) and hyperoxic (18.0 ± 5.2 μg/ml) groups or between P10 (19.3 ± 5.3 μg/ml) and P21 (17.6 ± 4.6 μg/ml) among rat pups treated with caffeine. The levels were within the recommended therapeutic range (5–25 μg/ml; Ref. 35).

Caffeine attenuates hyperoxia-induced alveolar growth impairment.

Hyperoxia (H) significantly increased MLI compared with normoxia group (N). Caffeine-treated hyperoxic pups (HC) had significantly lower MLI compared with saline-treated hyperoxic pups (HC vs. H; Fig. 1A); the MLI for HC group is not different from normoxia + caffeine group at P10. These results indicate that hyperoxia impairs alveolar formation, whereas caffeine treatment protects lungs from hyperoxia-induced injury. Similar changes in MLI were seen at P21 (Fig. 1B) after 10-day recovery in room air. We also saw significant decreases in radial alveolar count and secondary septation in hyperoxia-exposed pups at P10 and P21; caffeine treatment improved both radial alveolar count and secondary septa in the hyperoxia-exposed group.

Fig. 1.

Caffeine treatment reverses the hyperoxia-induced alveolar growth impairment as assessed by mean linear intercept (MLI), radial alveolar count (RAC), and secondary septation. A: at P10, there is a significant increase in MLI with decrease in RAC and secondary septation in hyperoxia (H) group. Early caffeine treatment does not change alveolar growth in normoxia (NC) group but significantly improves alveolar growth in hyperoxia (HC) group, indicating more complex lung structure. B: hyperoxia-induced alveolar simplification persists even after recovery in room air for 10 days (P21); caffeine treatment during normoxia recovery improves alveolar growth in the hyperoxia group. *P < 0.05 compared with the other 3 groups. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine; HPF, high-power field. Scale bar = 100 μm.

Hyperoxia decreases blood vessel formation in lungs.

Blood vessel counts decreased significantly in hyperoxia group at both P10 (Fig. 2A) and P21 (Fig. 2B). Caffeine improved the blood vessel count in the hyperoxia-exposed group at both P10 and P21.

Fig. 2.

Caffeine ameliorates hyperoxia-induced blood vessel formation decrease in the developing lungs. Endothelial cells were identified by mouse anti-rat endothelial cell antigen (RECA-1) antibody in the immunohistochemistry stain. A: the blood vessel counts decrease significantly in hyperoxia group (H) at P10. B: the blood vessel counts remain lower in hyperoxia group at P21 after recovery in room air for 10 days. *P < 0.05 compared with the other 3 groups. #P < 0.05 compared with the H group. Red arrow: blood vessels stained by RECA-1 antibody. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine. Scale bar = 100 μm.

Hyperoxia increases BiP and activates UPR in lungs.

BiP, or 78-kDa glucose-regulated protein, is an ER chaperone protein and a master regulator of ER homeostasis (15). Induction of BiP indicates the presence of acute ER stress (25). We saw increased BiP levels (Fig. 3A) in the hyperoxia group (1.78 ± 0.55-fold); caffeine reduced BiP to normoxic levels (0.98 ± 0.55-fold), suggesting that caffeine ameliorates hyperoxia-induced ER stress at P10.

Fig. 3.

Hyperoxia increases endoplasmic reticulum (ER) stress in developing lungs, which is mitigated by the caffeine treatment. A: at P10, the levels of ER chaperone protein, binding immunoglobulin protein (BiP), PKR-like ER kinase (PERK) and its phosphorylation (P-), and inositol-requiring enzyme 1α (IRE1α) and its phosphorylation are all increased in hyperoxia (H) group. Activated transcription factor-6 (ATF6) levels do not change, but its cleaved form (cATF6) increased in H group. Caffeine treatment in hyperoxia + caffeine (HC) group decreased the levels of BiP, PERK and its phosphorylation, IRE1α and its phosphorylation, and cATF6. B: at P21, the changes are more variable with no significant difference among groups except the increased cATF6 levels in NC and H groups. *P < 0.05 compared with the other 3 groups. #P < 0.05 compared with the N and HC groups. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine.

We also studied all three signaling pathways (PERK, IRE1α, and ATF6) involved in UPR. On sensing the presence of ER stress, both PERK and IRE1α are phosphorylated to activate downstream effectors, whereas ATF6 is cleaved (36). We saw increased phosphorylation in PERK (5.37 ± 3.01-fold) and IRE1α (4.68 ± 0.89-fold) in hyperoxia group. Similar change was also seen in total PERK (6.05 ± 5.25-fold) and total IRE1α (49.62 ± 54.99-fold). No difference in ATF6 was seen, but the cleaved ATF6 was increased in hyperoxia group (4.00 ± 2.45-fold). All three signaling pathways were activated in hyperoxia group at P10, and the responses were all attenuated by caffeine.

Rat pups recovered in room air for 11 days after hyperoxia exposure showed no differences in phospho-PERK, PERK, phospho-IRE1α, and IRE1α between the groups (Fig. 3B), suggesting that ER stress was attenuated during normoxia recovery. However, the levels of cleaved ATF6 (cATF6) remained higher in the hyperoxia group at P21, suggesting incomplete recovery. Caffeine treatment of the hyperoxic pups decreased the cATF6 to normoxic levels, whereas caffeine treatment of normoxic pups increased the cATF6 levels.

Hyperoxia increases UPR downstream effectors in lungs.

Activation of PERK leads to increased CHOP levels (19), whereas activation of IRE1α leads to splicing of XBP1 mRNA (7). CHOP and spliced form of XBP1 (sXBP1) are effectors for the UPR. Hyperoxia increased the CHOP levels (11.53 ± 10.17-fold) and sXBP1 levels (2.13 ± 0.47-fold; Fig. 4A) but not the unspliced XBP1 (uXBP1) levels at P10. Caffeine treatment decreased the levels of these downstream effectors compared with hyperoxia group at P10. No difference was seen in either CHOP or sXBP1 level among the four groups at P21 (Fig. 4B).

Fig. 4.

Caffeine decreases downstream effectors of unfolded protein response (UPR) activated by hyperoxia. A: C/EBP homologous protein (CHOP), downstream effector of PERK, and spliced X-box binding protein 1 (sXBP1), downstream effector of IRE1α, increased in hyperoxia-exposed lungs at P10. Caffeine treatment decreased both CHOP and sXBP1 levels in hyperoxia-exposed lungs. B: there is no difference in CHOP or sXBP1 levels between groups at P21. *P < 0.05 compared with the other 3 groups. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine.

Hyperoxia causes increased ER stress.

Electron microscopy showed dilation of ER in the pulmonary endothelial cells in the hyperoxia group, suggesting increased ER stress and disrupted mitochondrial structure (Fig. 5). The structural changes in the ER and mitochondria were not seen in the caffeine-treated hyperoxia group.

Fig. 5.

Hyperoxia causes ER stress. ER is dilated (between small arrows) in the pulmonary endothelial cells after exposure to hyperoxia. Also noted are the small, disrupted mitochondria (large arrow) in the hyperoxia group. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine. Scale bar = 500 nm.

Hyperoxia impairs mitochondrial function in the lungs.

An interaction between the ER and mitochondria is required for maintaining normal mitochondrial function. The ER-mitochondria interaction requires the participation of Mfn2 (12), and Mfn2 deficiency is associated with ER stress and altered oxidative stress handling (44). Expression of Drp1 counteracts this role of Mfn2 and induces mitochondrial fragmentation, whereas phosphorylation at serine⁶³⁷ of Drp1 promotes ER-mitochondria interaction and prevents mitochondria from fission (28).

At P10 (Fig. 6A), hyperoxia decreased Drp1 serine⁶³⁷ phosphorylation (0.51 ± 0.29-fold) and increased Drp1 levels (1.46 ± 0.23-fold). Levels of Mfn2 (0.50 ± 0.09-fold), but not Mfn1, in the lungs also decreased after hyperoxia at P10. Caffeine prevented the hyperoxia-induced changes in Drp1, phospho⁶³⁷-Drp1, and Mfn2. These results suggest that hyperoxia decreased ER-mitochondrial interaction with more mitochondrial fragmentation, which is prevented by caffeine. Hyperoxia is known to decrease complex I levels, which contributed to the impaired alveolar formation in mouse pups (38). Since we saw impaired mitochondrial dynamics in our rat hyperoxia model, we investigated the changes in mitochondrial complex I protein. Complex I levels in the lungs were decreased by hyperoxia (0.29 ± 0.16-fold). Caffeine prevented this decrease in complex I levels.

Fig. 6.

Mitochondrial biogenesis and bioenergetics in lungs are disrupted by hyperoxia, whereas caffeine ameliorates the effect. A: at P10, mitochondrial fission protein, dynamin-related protein-1 (Drp1), was increased in hyperoxia (H) group. Drp1 serine⁶³⁷ phosphorylation (P⁶³⁷-) and mitochondrial fusion proteins, mitofusin-1 (Mfn1) and Mfn2, were decreased in H group. NADH:ubiquinone oxidoreductase, subunit of complex I of the mitochondrial electron transport chain, was decreased in H group. Caffeine attenuates the effects of hyperoxia on these proteins except for Mfn1. B: at P21, similar changes were seen in H group except for no difference in Drp1 levels. There was also a significant decrease in Mfn1 levels in the H group at P21. *P < 0.05 compared with the other 3 groups. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine.

At P21 (Fig. 6B), the levels of phospho⁶³⁷-Drp1 (0.45 ± 0.22-fold), Mfn1 (0.85 ± 0.12-fold), Mfn2 (0.52 ± 0.08-fold), and complex I (0.25 ± 0.18-fold) decreased in the hyperoxia group with no difference in Drp1 levels between the groups. Caffeine restored these indices of mitochondrial function in the hyperoxia-exposed lungs at P21.

Hyperoxia increases apoptosis in the lungs.

When UPR fails to mitigate the ER stress, the cell will undergo CHOP-mediated apoptosis (19). In rodents, caspase-12 plays a central role in ER-mediated apoptosis and is the upstream regulator for caspases-3 and -9 (34, 45). Hyperoxia increased the caspase-12 levels (2.42 ± 0.22-fold; Fig. 7A) and the cleaved form in lungs at P10. We investigated another index of apoptosis, Bcl-2-to-Bax ratio, which also showed increased apoptosis (decreased Bcl-2-to-Bax ratio) in the hyperoxia group (0.57 ± 0.14-fold; Fig. 7B); these changes were prevented by caffeine treatment. The in situ TUNEL stain further confirmed the findings of the two assays (4.31 ± 0.56-fold; Fig. 7C). Again, caffeine decreased the markers of apoptosis in the hyperoxia group.

Fig. 7.

Caffeine prevents hyperoxia-induced apoptosis in the lungs. A: at P10, caspase-12 and its cleaved form, the upstream protein for caspase-3 and caspase-9, increased in the hyperoxia (H) group. B: the Bcl-2-to-Bax ratio of lung is decreased in H group, consistent with the increase in apoptosis. C: in situ TUNEL stain also shows that hyperoxia increases apoptosis in H group. Caffeine treatment decreases caspase-12 levels and increases Bcl-2-to-Bax ratio in hyperoxia-exposed lungs (HC). D: at P21, no difference in caspase-12 and its cleaved form is noted among N, NC, and H groups, but HC group has decreased caspase-12 level. E: the H group has the lowest Bcl-2-to-Bax ratio, whereas HC group has higher Bcl-2-to-Bax ratio than H group but lower than N and NC groups. F: in situ TUNEL stain shows that only H group has increased apoptosis. →: TUNEL-positive staining. *P < 0.05 compared with the other 3 groups. #P < 0.05 compared with H group. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine. I, intact caspase-12; C, cleaved caspase-12. Scale bar = 50.0 μm.

At P21, there was no change in either caspase-12 or its cleaved form in the hyperoxia group, but caffeine further decreased both levels compared with the normoxia group (0.71 ± 0.19-fold; Fig. 7D). The Bcl-2-to-Bax ratio remained low in both hyperoxia groups, with or without caffeine; however, the ratios were higher in the caffeine-treated group (Fig. 7E). The in situ TUNEL stain showed less apoptosis in caffeine-treated hyperoxia group compared with hyperoxia alone (Fig. 7F).

Hyperoxia increases cyclooxygenase-2 level and inflammation in lungs.

Studies in a mouse model of hyperoxia demonstrated that Cox-2, the inducible isoform of cyclooxygenase, is one of the downstream mediators of UPR that contribute to lung injury and inflammation (9). Consistent with that report, we also observed increased Cox-2 levels in H group (4.74 ± 3.92-fold; Fig. 8A) at P10. Caffeine prevented the increase in Cox-2 during hyperoxia exposure. There was no difference in Cox-1 levels among the four groups. The levels of myeloperoxidase (MPO), biomarker for lung inflammation, followed the same trend and increased in the hyperoxia group (7.42 ± 4.75-fold; Fig. 8B) and were attenuated in the hyperoxia + caffeine group. Parallel to the changes in Cox-2 and MPO levels in the hyperoxia group, we also saw increased MPO-positive inflammatory cell infiltration in the hyperoxia group (Fig. 8C). At P21, similar changes in Cox-2 (8.97 ± 6.43-fold; Fig. 8D), MPO (17.19 ± 14.71-fold; Fig. 8E), and MPO-positive inflammatory cell infiltration (Fig. 8F) were seen in the hyperoxia group, and these changes were also attenuated by caffeine.

Fig. 8.

Hyperoxia (H) increases cyclooxygenase-2 (Cox-2) and myeloperoxidase (MPO) levels in the lung; caffeine ameliorates this response. A: at P10, the levels of Cox-2 increase in H group, but no difference in Cox-1 levels are noted. B: MPO also increases in H group. C: more MPO-positive inflammatory cell infiltrates in H group. At P21, similar changes in Cox-2 (D), MPO (E), and MPO-positive inflammatory cells (F) are seen in H group. →: MPO-positive cell. *P < 0.05 compared with the other 3 groups. N, normoxia; NC, normoxia + caffeine; H, hyperoxia; HC, hyperoxia + caffeine. Scale bar = 100 μm.

Hyperoxia increases oxidative stress in lungs.

Immunofluorescence staining with antibody for 8-OH-dG showed increased oxidative stress in the hyperoxia group (1.80 ± 0.09-fold; Fig. 9A); caffeine attenuated this effect (0.74 ± 0.36-fold). Similarly, protein nitration was increased in the hyperoxia group (1.50 ± 0.08-fold; Fig. 9B), and this nitration was also decreased when caffeine was given (1.31 ± 0.08-fold). Both biomarkers showed that hyperoxia exposure causes oxidative stress to the lungs, consistent with antioxidant effects of caffeine.

Fig. 9.

Hyperoxia exposure increases oxidative stress in the developing lungs. A: hyperoxia increases the colocalization (purple) of DAPI (blue) and 8-hydroxy-deoxyguanosine (8-OH-dG; red) in the immunofluorescence stain, and these levels were decreased by caffeine treatment. B: tissue nitration, by immunohistochemistry stain, is increased in the lungs after exposure to hyperoxia. Nitration is most dramatic in the pulmonary vasculature, and this signal was also decreased by caffeine. *P < 0.05 compared with the other 3 groups. 3-NT, 3-nitrotyrosine. Scale bar = 200 μm for immunofluorescence but 100 μm for immunohistochemistry stain.

DISCUSSION

BPD is a common, long-term complication that occurs primarily in premature infants. Early caffeine treatment in extremely premature infants has been previously shown to decrease the rate of BPD (42). Although not the primary outcome for the study, this groundbreaking observation has dramatically changed our clinical practice (14). The mechanisms involved in the efficacy of caffeine use have not been extensively studied.

We used the hyperoxia rat pup model for our studies since their lungs are at saccular stage at birth, whereas the alveolar stage occurs between P4 and P14 (4, 37). P10 was chosen in our studies to wean the hyperoxia-exposed pups into room air to simulate the clinical scenario, since most BPD infants will outgrow their oxygen dependence at ~1 yr before the completion of alveolar stage by 2–6 yr of life. Our model reproduced the alveolar simplification commonly seen in premature infants with moderate to severe BPD. In the rodent models, an anti-inflammatory effect has been previously shown for caffeine (24, 52) and other nonselective phosphodiesterase inhibitors (2, 47); however, these studies did not report any improvement in lung histology. In contrast, at least one study in a mouse model reported possible harmful effects of caffeine treatment during exposure to hyperoxia (11). This report has led to some concerns about early caffeine use in the absence of studies that support its beneficial effects in the appropriate models (3).

One potential limitation of the previous studies was the dose used in rodent studies. The doses of caffeine citrate used in premature infants are 5–10 mg·kg⁻¹·day⁻¹, which will be equivalent to 15–30 mg·kg⁻¹·day⁻¹ caffeine base for rats and even higher for mice (39). We empirically chose 20 mg·kg⁻¹·day⁻¹ in this study and found that the measured blood caffeine levels were within the acceptable therapeutic range (8–40 μg/ml; Ref. 35). Using this dose, we observed an improved lung structure in treated pups (HC group). Whether lower doses will give different results is not known from our study.

Two mechanisms, ER stress (9, 31) and impaired mitochondrial biogenesis (38), have been reported to contribute to hyperoxia-induced lung growth impairment in rodent pups. Since ER closely interacts with mitochondria (6, 50) and hyperoxia-induced oxidative stress is known to induce ER stress, it is logical to assume that prolonged hyperoxia exposure will exaggerate UPR and affects mitochondrial function. We showed that hyperoxia exposure during the saccular and alveolar stages of lung development in our study has increases of BiP, PERK, IRE1α, sXBP1, cATF6, and CHOP, supporting the notion that oxidative stress increases ER stress and amplifies UPR in the lungs (31, 36). Caffeine has been reported to attenuate protein aggregation and ER stress in neurons (20). In this study, we observed that caffeine doses that gave the recommended therapeutic blood levels are capable of decreasing the global ER stress in the hyperoxia-exposed lungs. The mechanisms involved in this effect of caffeine remain unclear, and the reason for the differential effects of caffeine on cATF6 based on oxygen exposure remains unclear from our study, but increases in cAMP-dependent protein kinase (PKA) function and increased Drp1 phosphorylation may play a role in attenuating the ER stress (53).

The increased mitochondrial fission protein (Drp1) and decreased fusion proteins (Mfn2) in the hyperoxia-exposed lungs in our results suggest disrupted mitochondrial dynamics (49). Hyperoxia also decreased Drp1 phosphorylation, which was restored by caffeine. Drp1 phosphorylation at serine⁶³⁷ inhibits mitochondrial fission and downstream apoptosis (16), which are known to facilitate cell viability in the rapidly growing lungs. The antibody we used for our study detects GTPase effector domain at serine⁶³⁷ in human Drp1, which is equivalent to serine⁶⁵⁶ of rat DRP1 (8). Rat Drp1 serine⁶⁵⁶ is a cAMP-dependent (PKA) phosphorylation site (10). Since caffeine is known to increase cAMP levels, activation of PKA may lead to Drp1 phosphorylation in the hyperoxia-exposed lungs.

In addition to the altered mitochondrial dynamics, we also observed a decrease in complex I levels in the hyperoxia group, suggesting an impairment of mitochondrial bioenergetics. Interaction of ER with the mitochondria requires the presence of Mfn2 (12). The decreased Mfn2 in lungs during hyperoxia may cause disruption of mitochondrial function due to lack of calcium uptake from ER. Both exaggerated UPR and mitochondrial dysfunction can lead to apoptosis. We saw increased caspase-12 and in situ TUNEL staining and decreased Bcl-2-to-Bax ratio in hyperoxia-exposed lungs, indicating increased apoptosis (40); this was reversed by caffeine treatment. Caffeine did not reverse the changes in Mfn1 at P10, suggesting that its efficacy is preferentially through improving the ER-mitochondria interaction.

A previous report indicated that Cox-2 increases UPR in hyperoxia-exposed and IFN-γ-treated mouse lungs and contributes to impaired alveolar formation (9). ER stress also induces inflammation (18, 54). We hypothesized that increase in Cox-2 may be secondary to inflammation induced by hyperoxia (13, 52). Using MPO as a biomarker, we saw evidence of increased inflammation in the hyperoxia group. Caffeine has been shown to exert anti-inflammatory effect in the lungs (24), and our results demonstrated that both MPO and Cox-2 levels decreased in caffeine-treated hyperoxia lungs.

There are some limitations to our study that should be mentioned. Although we observed increased markers of ER stress and impaired lung growth in hyperoxia-exposed lungs at P10, these changes may not be mechanistically linked. Using an ER stress inducer, tunicamycin or thapsigargin, or overexpression of ER stress markers in rodent pups may provide more direct evidence for a role of ER stress in causing decreased lung growth.

In conclusion, early caffeine treatment ameliorates hyperoxia-induced lung injury probably through attenuation of the ER stress and mitochondrial dysfunction (Fig. 10). Caffeine also has anti-inflammatory effect and decreases Cox-2 levels in lungs, which may contribute to the attenuated UPR (9). The attenuation of apoptosis in hyperoxia by caffeine may be secondary to the decreased ER stress and improved mitochondrial function. Our study is among the first to show improved histology with caffeine treatment in the rodent hyperoxia-lung injury model. The dose of caffeine used in our study is higher than the previous reports in literature. Although the blood levels of caffeine were higher than previous reported (24), they were still within the therapeutic range. Our study provides a novel mechanism by which caffeine protects developing lung from oxidative stress induced by hyperoxia exposure.

Fig. 10.

Potential mechanisms by which caffeine can mitigate the hyperoxia-induced lung injury. ┤: Inhibition or suppression.

GRANTS

The study was supported by Internal Support from Department of Pediatrics, Medical College of Wisconsin, Advancing a Healthier Wisconsin Endowment UL1TR001436, National Institute of Child Health and Human Development Grant R03-HD-073274 (R.-J. Teng), Muma Endowed Chair in Neonatology, Children’s Research Institute, and National Heart, Lung, and Blood Institute Grant R01-HL-057268 (G. G. Konduri).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.-J.T. conceived and designed research; R.-J.T., X.J., and T.M. performed experiments; R.-J.T., X.J., T.M., and T.-J.W. analyzed data; R.-J.T., X.J., T.M., A.J.A., and T.-J.W. interpreted results of experiments; R.-J.T. prepared figures; R.-J.T. and T.-J.W. drafted manuscript; R.-J.T., X.J., T.M., A.J.A., T.-J.W., and G.G.K. edited and revised manuscript; R.-J.T. and G.G.K. approved final version of manuscript.