Celecoxib Impairs Heart Development via  Inhibiting Cyclooxygenase-2 Activity in Zebrafish Embryos

• The selective cyclooxygenase-2 inhibitor celecoxib has been widely used in pregnant patients, yet its potential effects on fetal heart development remain unknown.

• Celecoxib exposure impairs heart development in zebrafish embryos by inhibiting cyclooxygenase-2 activity.

PREGNANT patients who suffer from systemic lupus erythematosus and other autoimmune diseases require pharmacologic control of pain and other inflammatory complications during pregnancy.1Cyclooxygenase inhibitors are among the most commonly prescribed drugs for controlling these complications.2Because the conventional nonselective cyclooxygenase inhibitors usually have severe adverse gastrointestinal effects, the selective cyclooxygenase-2 (COX-2) inhibitor celecoxib has been widely used in pregnant patients because it causes minimum adverse gastrointestinal effects.1,3However, recent studies have reported that celecoxib has adverse effects on the cardiovascular system in adults.4,5The goal of this study is to examine whether celecoxib also affects heart development in embryos.

The zebrafish is a powerful model organism for assessing drug-induced cardiotoxicity in vivo .6Transgenic zebrafish with vascular endothelial cells expressing fluorescent proteins allow in vivo  imaging of the endocardium and heart valve in live embryos.7Drugs diluted in medium can be absorbed by zebrafish larvae through the skin and gills. In addition, accumulating evidence has shown that drug-induced cardiotoxicity in zebrafish well resembles its effects in humans.6,8–10In the current study, we used zebrafish as an animal model to examine the effect of celecoxib on heart development. We found that celecoxib treatment caused serious defects in heart development, including abnormality in heart looping and the absence of heart valve formation. These defects were mimicked by treatment with the COX-2 inhibitor NS398 but significantly relieved by COX-2 product prostaglandin E²(PGE²). Thus, celecoxib exposure impairs embryonic heart development in zebrafish by inhibiting COX-2 activity.

Wild-type AB and transgenic Tg(flk1:GFP ) zebrafish, in which vascular endothelial cells express green fluorescent protein (GFP), were obtained (Zebrafish International Resource Center, Eugene, OR) and maintained at 28°C with an automatic fish-housing system (ESEN, Beijing, China) in the National Zebrafish Resources of China (Shanghai, China). Embryos were staged as described previously.11Zebrafish handling procedures were approved by the Institute of Neuroscience, Shanghai Institutes for Biologic Sciences, Chinese Academy of Sciences. Wild-type AB embryos were used for in situ  hybridization experiments; Tg(flk1:GFP)  embryos were used for assessing atrioventricular valve leaflet and looping development.

Zebrafish embryos were treated with celecoxib (Pfizer Pharmaceuticals LLC, Caguas, Puerto Rico), NS398, SC560 (Cayman Chemical, Ann Arbor, MI), or acetaminophen (Sigma-Aldrich, St. Louis, MO) from the tailbud stage (10.3 h postfertilization [hpf]). Celecoxib, NS398, SC560, and acetaminophen were dissolved in dimethyl sulfoxide (DMSO) (Sigma-Aldrich) to prepare stock solutions, which were then diluted to experimental concentration in embryo medium. Controls for each experiment included a negative control with embryos incubated in embryo medium and a solvent control with embryos incubated in 0.3% DMSO. Experiments were performed in triplicate. Optimal concentrations of celecoxib, NS398, SC560, and acetaminophen were 25, 60, 20 μM, and 5 mM, respectively, at which drug treatments produced obvious defects in heart development with minimal mortality before 96 hpf. At 72 hpf, drug-containing embryo medium was replaced with drug-free embryo medium.

Prostaglandins (PGs)—including PGE², prostaglandin I²(PGI²), and prostaglandin F^2α(PGF^2α) —and thromboxane A²receptor agonist U46619 (Cayman Chemical) were dissolved in DMSO to generate stock solutions before dilution in embryo medium at experimental concentrations. In rescue experiments, we supplemented celecoxib-exposed embryos with 1–25 μM PGE², PGF^2α, or PGI², and 1–10 μM U46619. Embryos were coincubated with celecoxib and PGE², PGF^2α, PGI², or U46619 from the tailbud stage to 72 hpf. Embryos were also treated with PGE², PGF^2α, PGI², or U46619 alone as negative controls.

To assess blood regurgitation, live embryos were mounted in 1% low melting point agarose on ventral view and imaged using a microscope equipped with a high-speed motion scope camera (Neurolucida E600FN; Nikon Corporation, Tokyo, Japan). For analysis of celecoxib-induced heart looping defects, embryos were positioned and imaged as described above. Images were captured to measure the angle between the midline and atrioventricular canal with Image-Pro Plus software (Media Cybernetics, Inc., Bethesda, MD). The position of the atrioventricular canal shifted slightly during heartbeat, introducing errors in the measurement of the angle. To reduce this error, three images—which correspond to the early phase, metaphase, and telophase of atrium contraction, respectively—were used to obtain a mean angle for each embryo.

Digoxigenin-labeled probes (Roche Applied Science, Mannheim, Germany) for cardiac myosin light chain 2, ventricular myosin heavy chain, atrial myosin heavy chain, bone morphogenetic protein 4 (BMP4), and versican were synthesized using linearized plasmids. Whole mount in situ  hybridization experiments were performed as described previously.12 

To examine heart valve development, Tg(flk1:GFP)  zebrafish embryos were mounted in 1.0% low melting point agarose with a ventral view for imaging.12Confocal imaging was performed at 25°C (FV1000-MPE; Olympus Corporation, Tokyo, Japan). Bidirectional confocal scans were captured at 6–10 frames/s. Time-series imaging was initiated at a random time in the cardiac cycle with 300–600 frames per embryo.

For normally distributed data (the angle of heart looping), statistical analysis was performed (Prism 4; GraphPad Software, San Diego, CA; SPSS version 13.0; SPSS Science Software, Chicago, IL) using an unpaired two-tailed Student t  test between two groups and one-way ANOVA plus Newman–Keuls multiple comparison test between more than two groups. For data that were not normally distributed (e.g. , heart rate), Kruskal-Wallis test plus Dunn's posttest was performed. For categorical variables (e.g. , incidence of pericardial edema, valve formation), chi-square test was performed. Statistical methods are specified in each figure legend. Summary data are represented as mean ± SD. Differences between groups were considered to be significant with a P  value of less than 0.05.

To examine the effect of celecoxib on heart development, we exposed zebrafish embryos to the drug from 10.3 to 72 hpf at concentrations of 10, 25, and 40 μM. Celecoxib-induced mortality was exhibited in a dose-dependent manner. Supplemental Digital Content 1 presents mortality rates of 48- and 72-hpf zebrafish embryos treated with celecoxib at different concentrations, https://links.lww.com/ALN/A670. All embryos treated with 40 μM celecoxib were dead by 72 hpf. In contrast, embryos treated with 10 μM celecoxib showed no obvious drug-induced effects in gross morphology or mortality.

We then exposed embryos to celecoxib at 10, 17.5, 25, and 32.5 μM, examining the drug's effects on heart development. Celecoxib-treated embryos exhibited serious pericardial edema and low heart rate compared with control embryos at 48 (Supplemental Digital Content 1, https://links.lww.com/ALN/A670) and 72 hpf (fig. 1). At 72 hpf, the incidence of pericardial edema was 13.4, 23.9, 61.9, 89.2, and 98.4% in the control, 10, 17.5, 25, and 32.5 μM celecoxib groups, respectively (fig. 1C). In addition, heart rates for these study groups were, respectively, 168.5 ± 11.9, 136.4 ± 21.2, 99.4 ± 17.4, 72.4 ± 14.8, and 38.5 ± 10.4 per minute (fig. 1D).

A similar dosage of celecoxib was used in a previous zebrafish study,13in which 20 μM celecoxib reduced the number of hematopoietic stem cells. These dosages may be clinically relevant because serum concentration of celecoxib and other COX-2 inhibitors in patients ranges from 20 to 100 μM.14,15 

During zebrafish development, the initially midline-straight embryonic heart tube transforms into a helically wound loop that subsequently loops to the right. This process is called heart looping and is conserved in vertebrates.16Therefore, we next examined whether celecoxib treatment affects heart looping in zebrafish larvae.

Heart configuration was clearly observed by high-resolution microscopy. Supplemental Digital Content 1 demonstrates heart configuration in DMSO and celecoxib-treated zebrafish embryos, https://links.lww.com/ALN/A670. We found that celecoxib-treated embryos failed to undergo heart looping (fig. 2). By 72 hpf, control embryos had completed the process of heart looping, with the atrium and ventricle transferring to the left and right of the midline, respectively (fig. 2A). In celecoxib-treated embryos, the atrium and ventricle remained central and linear in a tubular structure (fig. 2B).

To determine the extent to which heart looping was affected by celecoxib treatment, we measured the angle between the midline and atrioventricular canal in ventral views (figs. 2C and D). This angle provides an index of heart looping. As shown (fig. 2E), celecoxib treatment with 25 and 32.5 μM significantly decreased the angle (DMSO, 85.3 ± 11.8°; 25 μM celecoxib, 38.9 ± 15.6°; 32.5 μM celecoxib, 37.2 ± 15.4°). These results indicate that celecoxib exposure disrupts heart looping during development.

The primary function of the heart valve is to prevent retrograde blood flow. To determine whether celecoxib exposure causes defects in heart valve development, we performed in vivo  time-series confocal imaging of transgenic Tg(flk1:GFP)  zebrafish larvae, which permitted visualization of the endocardium and valve structure in intact zebrafish embryos.7 

At 85 hpf, the heart valve in control embryos was clearly observed at the atrioventricular region, whereas there was only a single sheet of endocardial cells at the atrioventricular junction in celecoxib-treated embryos. Supplemental Digital Content 1 presents heart valve morphology in control and 25-μM celecoxib–treated embryos at 85 hpf, https://links.lww.com/ALN/A670. To determine whether the celecoxib-induced valve absence was the result of developmental delay, we examined heart valve formation at a later stage. At 96 hpf, the heart valve leaflet in control embryos was well developed (fig. 3A), whereas the heart valve in 25-μM celecoxib–treated embryos was absent at the atrioventricular junction (fig. 3B). This defect was more readily observed by high-speed confocal scanning during the heartbeat. Although the valve leaflet in control embryos was open or closed during heart contraction or relaxation, respectively, no valve leaflet was observed at the atrioventricular boundary in celecoxib-treated embryos (figs. 3D and E).

The percentages of heart valve formation were 90.6, 82.5, 49.7, and 16.3% in the control, 10-, 17.5-, and 25-μM celecoxib–treated groups, respectively (fig. 3C). These results suggest that celecoxib treatment prevents heart valve formation.

To further examine whether celecoxib exposure causes abnormal blood flow, we used video microscopy to monitor hemodynamics. At 96 hpf, the valve leaflet in control embryos effectively prevented blood regurgitation at the atrioventricular junction. Supplemental Digital Content 2 demonstrates normal blood flow in control embryos, https://links.lww.com/ALN/A671. Only 7% (2 of 28) control embryos exhibited blood regurgitation. However, there was serious blood regurgitation in 90% (35 of 39) celecoxib-treated embryos, consistent with the observed absence of heart valve. Supplemental Digital Content 3 depicts blood regurgitation in celecoxib-treated embryos, https://links.lww.com/ALN/A672.

To determine how celecoxib affects heart valve formation, we performed whole mount in situ  hybridization to examine the expression pattern of the early valve differentiation markers BMP4 and versican. Both genes are initially expressed throughout the heart.17After 37 hpf, their expression is restricted to the valve-forming region. This restricted expression pattern is required for heart valve development.17 

At 48 hpf, although the expression of BMP4 and versican was restricted to the atrioventricular junction in control embryos (figs. 4A and C), as reported previously,17we found that, in celecoxib-treated embryos, both genes were expressed at the ventricle or atrium in addition to the atrioventricular junction (figs. 4B and D). However, it is noteworthy that the expression of versican in the otoliths was not changed by celecoxib exposure (figs. 4E and F), indicating that celecoxib-induced changes in versican expression are heart-specific. Therefore, these findings imply that celecoxib-induced malformation of the heart valve may be caused by altered expression of early differentiation markers of the heart valve.

In addition, we examined the effect of celecoxib treatment on the expression of cardiac markers. At 48 hpf, cardiac myosin light chain 2 is expressed strongly in the ventricle and weakly in the atrium. In contrast, atrial- and ventricular-specific myosin heavy chain markers are expressed in the atrium and ventricle, respectively.18We found that the expression patterns of these genes in celecoxib-treated embryos were indistinguishable from those in control embryos (fig. 5).

To examine whether celecoxib-induced heart defects are the result of inhibition of COX-2 activity, we treated zebrafish embryos with a mixture of celecoxib and a COX-2 product (i.e. , PGE², PGF^2α, or PGI²) or U46619. PG concentrations ranged from 1 to 25 μM; U46619 concentration ranged from 1 to 10 μM.

We found that application of 25 μM PGE²significantly alleviated 25 μM celecoxib-induced defects in heart looping (figs. 6A, B, C, and D; 35.0 ± 24.2° in celecoxib-treated embryos vs.  65.5 ± 22.6° in cotreated embryos; P < 0.001) and heart valve malformation (figs. 6E, F, G, and H; 19.9% in celecoxib-treated embryos vs . 62.3% in cotreated embryos; P < 0.001). However, coincubation with PGI², PGF^2α, and U46619 did not relieve celecoxib-induced defects. Supplemental Digital Content 1 depicts the gross morphology of zebrafish embryos in each study group, https://links.lww.com/ALN/A670. These results indicate that celecoxib-induced heart defects are largely the result of decreased PGE²production, consistent with previous reports that PGE²is the predominant prostanoid effector.33,34 

We next examined whether exposure to another COX-2 inhibitor, NS398, could mimic the heart defects induced by celecoxib treatment. We incubated zebrafish embryos with NS398 at concentrations ranging from 30 to 100 μM. Similar to celecoxib treatment, we found that the mortality of NS398-treated embryos increased as drug concentration increased (data not shown). Exposure to 60 μM NS398 led to defects similar to those observed in celecoxib-treated embryos. NS398-treated embryos exhibited serious pericardial edema (figs. 7A, B, and C; 74.7% in NS398-treated embryos vs . 19% in control embryos; P < 0.001), abnormal heart looping (figs. 7D, E, and F; 49.1 ± 28.9° in NS398-treated embryos vs.  83.9 ± 13.7° in control embryos; P < 0.001), and a low rate of heart valve formation (figs. 7G, H, and I; 36.5% in NS398-treated embryos vs . 87.2% in control embryos; P < 0.001). Consistent with heart valve malformation, we observed serious blood regurgitation in NS398-treated embryos (data not shown).

Acetaminophen, a commonly prescribed analgesic and antipyretic drug, is regularly consumed by pregnant women suffering from pain and fever.19Recent studies20,21have reported that acetaminophen suppresses COX-2 activity and PGE²release, and that it may be a relatively selective COX-2 inhibitor. In addition, acetaminophen may cause cardiovascular risk in adults.22,23In zebrafish embryos, acetaminophen can decrease COX-2 expression.24 

In addition, we found that acetaminophen exposure caused heart defects similar to those induced by celecoxib treatment, such as pericardial edema (data not shown), abnormal heart looping (figs. 8A, B, and C; 54.7 ± 25.3° in 5-mM acetaminophen–treated embryos vs.  86.3 ± 18° in control embryos; P < 0.01), and a low rate of heart valve formation (figs. 8D, E, and F; 38.1% in 5-mM acetaminophen–treated embryos vs . 88.9% in control embryos; P < 0.001). Taken together, these findings suggest that celecoxib-induced defects were caused by the inhibition of COX-2 activity.

Celecoxib is a selective inhibitor of COX-2. It is one of the most commonly prescribed drugs and is used by some pregnant patients for controlling pain and other inflammatory complications. However, the adverse cardiovascular effects of celecoxib complicate its use in adults.4,5Recently, there have been some concerns regarding the effects of celecoxib on fetal development. For example, clinical trials showed that taking celecoxib or other cyclooxygenase inhibitors during early pregnancy increased the risks associated with specific fetal cardiac anomalies.25Animal studies on fetal lambs showed that celecoxib has adverse effects on constriction of the fetal ductus arteriosus.26However, it remains unknown whether celecoxib exposure causes defects in heart development. In the current study, we found that celecoxib exposure seriously impaired heart development in zebrafish embryos. Celecoxib-treated embryos failed to undergo heart looping and displayed defects in heart valve formation. The restricted expression pattern of the heart valve markers BMP4 and versican was expanded in celecoxib-treated embryos. These defects were markedly corrected by PGE²treatment and mimicked by NS398 exposure, suggesting that celecoxib-induced abnormalities in heart development are the result of inhibition of COX-2 activity.

COX-2 has roles in a variety of physiologic and developmental processes.27,28For example, global deletion of COX-2 in mice caused reproductive and renal defects.29,30Because of extra-embryonic development, zebrafish present an ideal model for investigating the functional role of cyclooxygenases and their products.13,31,32Cyclooxygenases and their downstream PGs have been identified in zebrafish.33,34Scherz et al.  7showed that zebrafish heart expresses COX-2. They found that exposure to the COX-2 inhibitors NS398 and CAY10404 from 56 to 72 hpf changed the shape and reduced the size of ventricle cells, causing the valve leaflet to shift ventricularly. In the current study, we observed that celecoxib or NS398 exposure affected ventricle development in zebrafish, leading to a shrunken ventricle. However, we did not observe the valve leaflet at the atrioventricular junction, the ventricle, or the atrium in celecoxib- or NS398-treated embryos. Because the zebrafish atrioventricular valve begins to form before 43 hpf,17one explanation for this discrepancy is that we treated embryos with COX-2 inhibitors at earlier stages than previous investigators (i.e. , at 10.3 hpf rather than 56 hpf). It is possible that COX-2 plays a critical role in atrioventricular valve formation at an early stage.

Heart valve formation requires precise cell-signaling and gene-expression patterning in the myocardium and endocardium at the atrioventricular boundary.35For example, both BMP4 and versican are involved in heart valve formation. They initially express throughout the anteroposterior extent of the heart in zebrafish. After 37 hpf, their expression is restricted to the atrioventricular valve-forming region, coincident with the time of prevalvular cushion formation.36Misexpression of these genes induces defects in heart valve formation.17,37It is believed that the restricted expression pattern of BMP4 and versican is required for the cells at the border between the atrium and ventricle to adopt a tertiary cell fate during valve development. In the current study, we found that celecoxib exposure resulted in a loss of restriction in BMP4 and versican expression in the heart. In celecoxib-treated embryos, both genes were diffusely expressed in the ventricle or atrium—similar to jekyll and NXT2 mutants, which have exhibited similar defects in heart valve development,17,36—suggesting that celecoxib-induced heart valve defects may be the result of altered expression patterns in these genes.

In mammals, PGs play critical roles in physiologic and pathologic processes.27,38For example, deletion of the PGE²receptor subtype 2 in mice causes arterial dilatation, salt-sensitive hypertension, and reduced fertility.39In zebrafish, PGE²is the primary effector of prostanoid and is regulated by cyclooxygenase-1 (COX-1) and COX-2.34PGE²binds to cell surface receptors and plays important roles in physiologic and developmental processes.13,32Our results showed that celecoxib-induced heart defects could be attenuated by exogenous PGE². In contrast, application of PGF^2α, PGI², or a manipulation of thromboxane A²signaling, failed to rescue celecoxib-induced defects in heart development. This result suggests that PGE²plays a critical role in heart development.

In the current study, we found that celecoxib-induced defects in heart development were largely corrected by PGE²and mimicked by the COX-2 inhibitor NS398, indicating that COX-2 plays a major role in celecoxib-induced effects. Celecoxib is a selective COX-2 inhibitor and shows selectivity for inhibition of COX-2 375 times more than for COX-1.40In addition, we found that the COX-1 selective inhibitor SC560 could also cause heart developmental defects. Supplemental Digital Content 1 demonstrates SC560-induced defects in heart development, https://links.lww.com/ALN/A670. These results suggest that both cyclooxygenases are involved in heart development.7,34Consistent with the fact that COX-1, but not COX-2, is expressed in trunk vasculature and required for vascular development in zebrafish embryos,31we found that only SC560 (20 μM), but not celecoxib (25 μM), treatment impaired vascular development and resulted in shorten or absent intersomitic vessels. Supplemental Digital Content 1 depicts vascular development in control, celecoxib, and SC560-treated embryos, https://links.lww.com/ALN/A670. This finding implies that celecoxib, at the dosages used in our study, mainly affects COX-2—but not COX-1—activity.

During the past decade, numerous studies have focused on understanding inherited causes of congenital heart diseases, but the adverse effect of drug treatment on fetal heart development has received less attention. Herein, we demonstrate for the first time that celecoxib exposure affects heart development, including the heart looping and valve formation. We hope to direct attention to fetal safety in pregnant women using celecoxib.